xref: /netbsd-src/sys/netinet6/in6.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: in6.c,v 1.269 2018/07/04 00:35:33 kamil Exp $	*/
2 /*	$KAME: in6.c,v 1.198 2001/07/18 09:12:38 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)in.c	8.2 (Berkeley) 11/15/93
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: in6.c,v 1.269 2018/07/04 00:35:33 kamil Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_compat_netbsd.h"
70 #include "opt_net_mpsafe.h"
71 #endif
72 
73 #include <sys/param.h>
74 #include <sys/ioctl.h>
75 #include <sys/errno.h>
76 #include <sys/malloc.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/sockio.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/syslog.h>
85 #include <sys/kauth.h>
86 #include <sys/cprng.h>
87 #include <sys/kmem.h>
88 
89 #include <net/if.h>
90 #include <net/if_types.h>
91 #include <net/if_llatbl.h>
92 #include <net/if_ether.h>
93 #include <net/if_dl.h>
94 #include <net/pfil.h>
95 #include <net/route.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 
100 #include <netinet/ip6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/nd6.h>
103 #include <netinet6/mld6_var.h>
104 #include <netinet6/ip6_mroute.h>
105 #include <netinet6/in6_ifattach.h>
106 #include <netinet6/scope6_var.h>
107 
108 #ifdef COMPAT_50
109 #include <compat/netinet6/in6_var.h>
110 #endif
111 
112 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
113 
114 /* enable backward compatibility code for obsoleted ioctls */
115 #define COMPAT_IN6IFIOCTL
116 
117 #ifdef	IN6_DEBUG
118 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
119 #else
120 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
121 #endif /* IN6_DEBUG */
122 
123 /*
124  * Definitions of some constant IP6 addresses.
125  */
126 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
127 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
128 const struct in6_addr in6addr_nodelocal_allnodes =
129 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
130 const struct in6_addr in6addr_linklocal_allnodes =
131 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
132 const struct in6_addr in6addr_linklocal_allrouters =
133 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
134 
135 const struct in6_addr in6mask0 = IN6MASK0;
136 const struct in6_addr in6mask32 = IN6MASK32;
137 const struct in6_addr in6mask64 = IN6MASK64;
138 const struct in6_addr in6mask96 = IN6MASK96;
139 const struct in6_addr in6mask128 = IN6MASK128;
140 
141 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
142 				     0, 0, IN6ADDR_ANY_INIT, 0};
143 
144 struct pslist_head	in6_ifaddr_list;
145 kmutex_t		in6_ifaddr_lock;
146 
147 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
148 	struct ifnet *);
149 static int in6_ifaddprefix(struct in6_ifaddr *);
150 static int in6_ifremprefix(struct in6_ifaddr *);
151 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
152 	const struct sockaddr_in6 *, int);
153 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
154 static int in6_update_ifa1(struct ifnet *, struct in6_aliasreq *,
155     struct in6_ifaddr **, struct psref *, int);
156 
157 void
158 in6_init(void)
159 {
160 
161 	PSLIST_INIT(&in6_ifaddr_list);
162 	mutex_init(&in6_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
163 
164 	in6_sysctl_multicast_setup(NULL);
165 }
166 
167 /*
168  * Add ownaddr as loopback rtentry.  We previously add the route only if
169  * necessary (ex. on a p2p link).  However, since we now manage addresses
170  * separately from prefixes, we should always add the route.  We can't
171  * rely on the cloning mechanism from the corresponding interface route
172  * any more.
173  */
174 void
175 in6_ifaddlocal(struct ifaddr *ifa)
176 {
177 
178 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
179 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
180 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
181 	{
182 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
183 		return;
184 	}
185 
186 	rt_ifa_addlocal(ifa);
187 }
188 
189 /*
190  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
191  * if it exists.
192  */
193 void
194 in6_ifremlocal(struct ifaddr *ifa)
195 {
196 	struct in6_ifaddr *ia;
197 	struct ifaddr *alt_ifa = NULL;
198 	int ia_count = 0;
199 	struct psref psref;
200 	int s;
201 
202 	/*
203 	 * Some of BSD variants do not remove cloned routes
204 	 * from an interface direct route, when removing the direct route
205 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
206 	 * cloned routes, they could fail to remove the cloned routes when
207 	 * we handle multple addresses that share a common prefix.
208 	 * So, we should remove the route corresponding to the deleted address.
209 	 */
210 
211 	/*
212 	 * Delete the entry only if exactly one ifaddr matches the
213 	 * address, ifa->ifa_addr.
214 	 *
215 	 * If more than one ifaddr matches, replace the ifaddr in
216 	 * the routing table, rt_ifa, with a different ifaddr than
217 	 * the one we are purging, ifa.  It is important to do
218 	 * this, or else the routing table can accumulate dangling
219 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
220 	 * which will lead to crashes, later.  (More than one ifaddr
221 	 * can match if we assign the same address to multiple---probably
222 	 * p2p---interfaces.)
223 	 *
224 	 * XXX An old comment at this place said, "we should avoid
225 	 * XXX such a configuration [i.e., interfaces with the same
226 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
227 	 * XXX agree, especially now that I have fixed the dangling
228 	 * XXX ifp-pointers bug.
229 	 */
230 	s = pserialize_read_enter();
231 	IN6_ADDRLIST_READER_FOREACH(ia) {
232 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
233 			continue;
234 		if (ia->ia_ifp != ifa->ifa_ifp)
235 			alt_ifa = &ia->ia_ifa;
236 		if (++ia_count > 1 && alt_ifa != NULL)
237 			break;
238 	}
239 	if (ia_count > 1 && alt_ifa != NULL)
240 		ifa_acquire(alt_ifa, &psref);
241 	pserialize_read_exit(s);
242 
243 	if (ia_count == 0)
244 		return;
245 
246 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
247 
248 	if (ia_count > 1 && alt_ifa != NULL)
249 		ifa_release(alt_ifa, &psref);
250 }
251 
252 /* Add prefix route for the network. */
253 static int
254 in6_ifaddprefix(struct in6_ifaddr *ia)
255 {
256 	int error, flags = 0;
257 
258 	if (in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) == 128) {
259 		if (ia->ia_dstaddr.sin6_family != AF_INET6)
260 			/* We don't need to install a host route. */
261 			return 0;
262 		flags |= RTF_HOST;
263 	}
264 
265 	/* Is this a connected route for neighbour discovery? */
266 	if (nd6_need_cache(ia->ia_ifp))
267 		flags |= RTF_CONNECTED;
268 
269 	if ((error = rtinit(&ia->ia_ifa, RTM_ADD, RTF_UP | flags)) == 0)
270 		ia->ia_flags |= IFA_ROUTE;
271 	else if (error == EEXIST)
272 		/* Existance of the route is not an error. */
273 		error = 0;
274 
275 	return error;
276 }
277 
278 /* Delete network prefix route if present.
279  * Re-add it to another address if the prefix matches. */
280 static int
281 in6_ifremprefix(struct in6_ifaddr *target)
282 {
283 	int error, s;
284 	struct in6_ifaddr *ia;
285 
286 	if ((target->ia_flags & IFA_ROUTE) == 0)
287 		return 0;
288 
289 	s = pserialize_read_enter();
290 	IN6_ADDRLIST_READER_FOREACH(ia) {
291 		if (target->ia_dstaddr.sin6_len) {
292 			if (ia->ia_dstaddr.sin6_len == 0 ||
293 			    !IN6_ARE_ADDR_EQUAL(&ia->ia_dstaddr.sin6_addr,
294 			    &target->ia_dstaddr.sin6_addr))
295 				continue;
296 		} else {
297 			if (!IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
298 			    &target->ia_addr.sin6_addr,
299 			    &target->ia_prefixmask.sin6_addr))
300 				continue;
301 		}
302 
303 		/*
304 		 * if we got a matching prefix route, move IFA_ROUTE to him
305 		 */
306 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
307 			struct psref psref;
308 			int bound = curlwp_bind();
309 
310 			ia6_acquire(ia, &psref);
311 			pserialize_read_exit(s);
312 
313 			rtinit(&target->ia_ifa, RTM_DELETE, 0);
314 			target->ia_flags &= ~IFA_ROUTE;
315 
316 			error = in6_ifaddprefix(ia);
317 
318 			ia6_release(ia, &psref);
319 			curlwp_bindx(bound);
320 
321 			return error;
322 		}
323 	}
324 	pserialize_read_exit(s);
325 
326 	/*
327 	 * noone seem to have prefix route.  remove it.
328 	 */
329 	rtinit(&target->ia_ifa, RTM_DELETE, 0);
330 	target->ia_flags &= ~IFA_ROUTE;
331 	return 0;
332 }
333 
334 int
335 in6_mask2len(struct in6_addr *mask, u_char *lim0)
336 {
337 	int x = 0, y;
338 	u_char *lim = lim0, *p;
339 
340 	/* ignore the scope_id part */
341 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
342 		lim = (u_char *)mask + sizeof(*mask);
343 	for (p = (u_char *)mask; p < lim; x++, p++) {
344 		if (*p != 0xff)
345 			break;
346 	}
347 	y = 0;
348 	if (p < lim) {
349 		for (y = 0; y < NBBY; y++) {
350 			if ((*p & (0x80 >> y)) == 0)
351 				break;
352 		}
353 	}
354 
355 	/*
356 	 * when the limit pointer is given, do a stricter check on the
357 	 * remaining bits.
358 	 */
359 	if (p < lim) {
360 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
361 			return -1;
362 		for (p = p + 1; p < lim; p++)
363 			if (*p != 0)
364 				return -1;
365 	}
366 
367 	return x * NBBY + y;
368 }
369 
370 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
371 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
372 
373 static int
374 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
375 {
376 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
377 	struct	in6_ifaddr *ia = NULL;
378 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
379 	struct sockaddr_in6 *sa6;
380 	int error, bound;
381 	struct psref psref;
382 
383 	switch (cmd) {
384 	case SIOCAADDRCTL_POLICY:
385 	case SIOCDADDRCTL_POLICY:
386 		/* Privileged. */
387 		return in6_src_ioctl(cmd, data);
388 	/*
389 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
390 	 */
391 	case SIOCSIFADDR:
392 	case SIOCSIFDSTADDR:
393 	case SIOCSIFBRDADDR:
394 	case SIOCSIFNETMASK:
395 		return EOPNOTSUPP;
396 	case SIOCGETSGCNT_IN6:
397 	case SIOCGETMIFCNT_IN6:
398 		return mrt6_ioctl(cmd, data);
399 	case SIOCGIFADDRPREF:
400 	case SIOCSIFADDRPREF:
401 		if (ifp == NULL)
402 			return EINVAL;
403 		return ifaddrpref_ioctl(so, cmd, data, ifp);
404 	}
405 
406 	if (ifp == NULL)
407 		return EOPNOTSUPP;
408 
409 	switch (cmd) {
410 	case SIOCSNDFLUSH_IN6:
411 	case SIOCSPFXFLUSH_IN6:
412 	case SIOCSRTRFLUSH_IN6:
413 	case SIOCSDEFIFACE_IN6:
414 	case SIOCSIFINFO_FLAGS:
415 	case SIOCSIFINFO_IN6:
416 		/* Privileged. */
417 		/* FALLTHROUGH */
418 	case OSIOCGIFINFO_IN6:
419 	case SIOCGIFINFO_IN6:
420 	case SIOCGDRLST_IN6:
421 	case SIOCGPRLST_IN6:
422 	case SIOCGNBRINFO_IN6:
423 	case SIOCGDEFIFACE_IN6:
424 		return nd6_ioctl(cmd, data, ifp);
425 	}
426 
427 	switch (cmd) {
428 	case SIOCSIFPREFIX_IN6:
429 	case SIOCDIFPREFIX_IN6:
430 	case SIOCAIFPREFIX_IN6:
431 	case SIOCCIFPREFIX_IN6:
432 	case SIOCSGIFPREFIX_IN6:
433 	case SIOCGIFPREFIX_IN6:
434 		log(LOG_NOTICE,
435 		    "prefix ioctls are now invalidated. "
436 		    "please use ifconfig.\n");
437 		return EOPNOTSUPP;
438 	}
439 
440 	switch (cmd) {
441 	case SIOCALIFADDR:
442 	case SIOCDLIFADDR:
443 		/* Privileged. */
444 		/* FALLTHROUGH */
445 	case SIOCGLIFADDR:
446 		return in6_lifaddr_ioctl(so, cmd, data, ifp);
447 	}
448 
449 	/*
450 	 * Find address for this interface, if it exists.
451 	 *
452 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
453 	 * only, and used the first interface address as the target of other
454 	 * operations (without checking ifra_addr).  This was because netinet
455 	 * code/API assumed at most 1 interface address per interface.
456 	 * Since IPv6 allows a node to assign multiple addresses
457 	 * on a single interface, we almost always look and check the
458 	 * presence of ifra_addr, and reject invalid ones here.
459 	 * It also decreases duplicated code among SIOC*_IN6 operations.
460 	 */
461 	switch (cmd) {
462 	case SIOCAIFADDR_IN6:
463 #ifdef OSIOCAIFADDR_IN6
464 	case OSIOCAIFADDR_IN6:
465 #endif
466 #ifdef OSIOCSIFPHYADDR_IN6
467 	case OSIOCSIFPHYADDR_IN6:
468 #endif
469 	case SIOCSIFPHYADDR_IN6:
470 		sa6 = &ifra->ifra_addr;
471 		break;
472 	case SIOCSIFADDR_IN6:
473 	case SIOCGIFADDR_IN6:
474 	case SIOCSIFDSTADDR_IN6:
475 	case SIOCSIFNETMASK_IN6:
476 	case SIOCGIFDSTADDR_IN6:
477 	case SIOCGIFNETMASK_IN6:
478 	case SIOCDIFADDR_IN6:
479 	case SIOCGIFPSRCADDR_IN6:
480 	case SIOCGIFPDSTADDR_IN6:
481 	case SIOCGIFAFLAG_IN6:
482 	case SIOCSNDFLUSH_IN6:
483 	case SIOCSPFXFLUSH_IN6:
484 	case SIOCSRTRFLUSH_IN6:
485 	case SIOCGIFALIFETIME_IN6:
486 #ifdef OSIOCGIFALIFETIME_IN6
487 	case OSIOCGIFALIFETIME_IN6:
488 #endif
489 	case SIOCGIFSTAT_IN6:
490 	case SIOCGIFSTAT_ICMP6:
491 		sa6 = &ifr->ifr_addr;
492 		break;
493 	default:
494 		sa6 = NULL;
495 		break;
496 	}
497 
498 	error = 0;
499 	bound = curlwp_bind();
500 	if (sa6 && sa6->sin6_family == AF_INET6) {
501 		if (sa6->sin6_scope_id != 0)
502 			error = sa6_embedscope(sa6, 0);
503 		else
504 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
505 		if (error != 0)
506 			goto out;
507 		ia = in6ifa_ifpwithaddr_psref(ifp, &sa6->sin6_addr, &psref);
508 	} else
509 		ia = NULL;
510 
511 	switch (cmd) {
512 	case SIOCSIFADDR_IN6:
513 	case SIOCSIFDSTADDR_IN6:
514 	case SIOCSIFNETMASK_IN6:
515 		/*
516 		 * Since IPv6 allows a node to assign multiple addresses
517 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
518 		 */
519 		error = EINVAL;
520 		goto release;
521 
522 	case SIOCDIFADDR_IN6:
523 		/*
524 		 * for IPv4, we look for existing in_ifaddr here to allow
525 		 * "ifconfig if0 delete" to remove the first IPv4 address on
526 		 * the interface.  For IPv6, as the spec allows multiple
527 		 * interface address from the day one, we consider "remove the
528 		 * first one" semantics to be not preferable.
529 		 */
530 		if (ia == NULL) {
531 			error = EADDRNOTAVAIL;
532 			goto out;
533 		}
534 		/* FALLTHROUGH */
535 #ifdef OSIOCAIFADDR_IN6
536 	case OSIOCAIFADDR_IN6:
537 #endif
538 	case SIOCAIFADDR_IN6:
539 		/*
540 		 * We always require users to specify a valid IPv6 address for
541 		 * the corresponding operation.
542 		 */
543 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
544 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
545 			error = EAFNOSUPPORT;
546 			goto release;
547 		}
548 		/* Privileged. */
549 
550 		break;
551 
552 	case SIOCGIFADDR_IN6:
553 		/* This interface is basically deprecated. use SIOCGIFCONF. */
554 		/* FALLTHROUGH */
555 	case SIOCGIFAFLAG_IN6:
556 	case SIOCGIFNETMASK_IN6:
557 	case SIOCGIFDSTADDR_IN6:
558 	case SIOCGIFALIFETIME_IN6:
559 #ifdef OSIOCGIFALIFETIME_IN6
560 	case OSIOCGIFALIFETIME_IN6:
561 #endif
562 		/* must think again about its semantics */
563 		if (ia == NULL) {
564 			error = EADDRNOTAVAIL;
565 			goto out;
566 		}
567 		break;
568 	}
569 
570 	switch (cmd) {
571 
572 	case SIOCGIFADDR_IN6:
573 		ifr->ifr_addr = ia->ia_addr;
574 		error = sa6_recoverscope(&ifr->ifr_addr);
575 		break;
576 
577 	case SIOCGIFDSTADDR_IN6:
578 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
579 			error = EINVAL;
580 			break;
581 		}
582 		/*
583 		 * XXX: should we check if ifa_dstaddr is NULL and return
584 		 * an error?
585 		 */
586 		ifr->ifr_dstaddr = ia->ia_dstaddr;
587 		error = sa6_recoverscope(&ifr->ifr_dstaddr);
588 		break;
589 
590 	case SIOCGIFNETMASK_IN6:
591 		ifr->ifr_addr = ia->ia_prefixmask;
592 		break;
593 
594 	case SIOCGIFAFLAG_IN6:
595 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
596 		break;
597 
598 	case SIOCGIFSTAT_IN6:
599 		if (ifp == NULL) {
600 			error = EINVAL;
601 			break;
602 		}
603 		memset(&ifr->ifr_ifru.ifru_stat, 0,
604 		    sizeof(ifr->ifr_ifru.ifru_stat));
605 		ifr->ifr_ifru.ifru_stat =
606 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
607 		break;
608 
609 	case SIOCGIFSTAT_ICMP6:
610 		if (ifp == NULL) {
611 			error = EINVAL;
612 			break;
613 		}
614 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
615 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
616 		ifr->ifr_ifru.ifru_icmp6stat =
617 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
618 		break;
619 
620 #ifdef OSIOCGIFALIFETIME_IN6
621 	case OSIOCGIFALIFETIME_IN6:
622 #endif
623 	case SIOCGIFALIFETIME_IN6:
624 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
625 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
626 			time_t maxexpire;
627 			struct in6_addrlifetime *retlt =
628 			    &ifr->ifr_ifru.ifru_lifetime;
629 
630 			/*
631 			 * XXX: adjust expiration time assuming time_t is
632 			 * signed.
633 			 */
634 			maxexpire = ((time_t)~0) &
635 			    (time_t)~(1ULL << ((sizeof(maxexpire) * NBBY) - 1));
636 			if (ia->ia6_lifetime.ia6t_vltime <
637 			    maxexpire - ia->ia6_updatetime) {
638 				retlt->ia6t_expire = ia->ia6_updatetime +
639 				    ia->ia6_lifetime.ia6t_vltime;
640 				retlt->ia6t_expire = retlt->ia6t_expire ?
641 				    time_mono_to_wall(retlt->ia6t_expire) :
642 				    0;
643 			} else
644 				retlt->ia6t_expire = maxexpire;
645 		}
646 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
647 			time_t maxexpire;
648 			struct in6_addrlifetime *retlt =
649 			    &ifr->ifr_ifru.ifru_lifetime;
650 
651 			/*
652 			 * XXX: adjust expiration time assuming time_t is
653 			 * signed.
654 			 */
655 			maxexpire = ((time_t)~0) &
656 			    (time_t)~(1ULL << ((sizeof(maxexpire) * NBBY) - 1));
657 			if (ia->ia6_lifetime.ia6t_pltime <
658 			    maxexpire - ia->ia6_updatetime) {
659 				retlt->ia6t_preferred = ia->ia6_updatetime +
660 				    ia->ia6_lifetime.ia6t_pltime;
661 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
662 				    time_mono_to_wall(retlt->ia6t_preferred) :
663 				    0;
664 			} else
665 				retlt->ia6t_preferred = maxexpire;
666 		}
667 #ifdef OSIOCFIFALIFETIME_IN6
668 		if (cmd == OSIOCFIFALIFETIME_IN6)
669 			in6_addrlifetime_to_in6_addrlifetime50(
670 			    &ifr->ifru.ifru_lifetime);
671 #endif
672 		break;
673 
674 #ifdef OSIOCAIFADDR_IN6
675 	case OSIOCAIFADDR_IN6:
676 		in6_aliasreq50_to_in6_aliasreq(ifra);
677 		/*FALLTHROUGH*/
678 #endif
679 	case SIOCAIFADDR_IN6:
680 	{
681 		struct in6_addrlifetime *lt;
682 
683 		/* reject read-only flags */
684 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
685 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
686 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
687 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
688 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
689 			error = EINVAL;
690 			break;
691 		}
692 		/*
693 		 * ia6t_expire and ia6t_preferred won't be used for now,
694 		 * so just in case.
695 		 */
696 		lt = &ifra->ifra_lifetime;
697 		if (lt->ia6t_expire != 0)
698 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
699 		if (lt->ia6t_preferred != 0)
700 			lt->ia6t_preferred =
701 			    time_wall_to_mono(lt->ia6t_preferred);
702 		/*
703 		 * make (ia == NULL) or update (ia != NULL) the interface
704 		 * address structure, and link it to the list.
705 		 */
706 		int s = splsoftnet();
707 		error = in6_update_ifa1(ifp, ifra, &ia, &psref, 0);
708 		splx(s);
709 		if (error)
710 			break;
711 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
712 		break;
713 	}
714 
715 	case SIOCDIFADDR_IN6:
716 		ia6_release(ia, &psref);
717 		ifaref(&ia->ia_ifa);
718 		in6_purgeaddr(&ia->ia_ifa);
719 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
720 		ifafree(&ia->ia_ifa);
721 		ia = NULL;
722 		break;
723 
724 	default:
725 		error = ENOTTY;
726 	}
727 release:
728 	ia6_release(ia, &psref);
729 out:
730 	curlwp_bindx(bound);
731 	return error;
732 }
733 
734 int
735 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
736 {
737 	int error, s;
738 
739 	switch (cmd) {
740 	case SIOCSNDFLUSH_IN6:
741 	case SIOCSPFXFLUSH_IN6:
742 	case SIOCSRTRFLUSH_IN6:
743 	case SIOCSDEFIFACE_IN6:
744 	case SIOCSIFINFO_FLAGS:
745 	case SIOCSIFINFO_IN6:
746 
747 	case SIOCALIFADDR:
748 	case SIOCDLIFADDR:
749 
750 	case SIOCDIFADDR_IN6:
751 #ifdef OSIOCAIFADDR_IN6
752 	case OSIOCAIFADDR_IN6:
753 #endif
754 	case SIOCAIFADDR_IN6:
755 
756 	case SIOCAADDRCTL_POLICY:
757 	case SIOCDADDRCTL_POLICY:
758 
759 		if (kauth_authorize_network(curlwp->l_cred,
760 		    KAUTH_NETWORK_SOCKET,
761 		    KAUTH_REQ_NETWORK_SOCKET_SETPRIV,
762 		    so, NULL, NULL))
763 			return EPERM;
764 		break;
765 	}
766 
767 	s = splsoftnet();
768 #ifndef NET_MPSAFE
769 	KASSERT(KERNEL_LOCKED_P());
770 #endif
771 	error = in6_control1(so , cmd, data, ifp);
772 	splx(s);
773 	return error;
774 }
775 
776 static int
777 in6_get_llsol_addr(struct in6_addr *llsol, struct ifnet *ifp,
778     struct in6_addr *ip6)
779 {
780 	int error;
781 
782 	memset(llsol, 0, sizeof(struct in6_addr));
783 	llsol->s6_addr16[0] = htons(0xff02);
784 	llsol->s6_addr32[1] = 0;
785 	llsol->s6_addr32[2] = htonl(1);
786 	llsol->s6_addr32[3] = ip6->s6_addr32[3];
787 	llsol->s6_addr8[12] = 0xff;
788 
789 	error = in6_setscope(llsol, ifp, NULL);
790 	if (error != 0) {
791 		/* XXX: should not happen */
792 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
793 	}
794 
795 	return error;
796 }
797 
798 static int
799 in6_join_mcastgroups(struct in6_aliasreq *ifra, struct in6_ifaddr *ia,
800     struct ifnet *ifp, int flags)
801 {
802 	int error;
803 	struct sockaddr_in6 mltaddr, mltmask;
804 	struct in6_multi_mship *imm;
805 	struct in6_addr llsol;
806 	struct rtentry *rt;
807 	int dad_delay;
808 	char ip6buf[INET6_ADDRSTRLEN];
809 
810 	/* join solicited multicast addr for new host id */
811 	error = in6_get_llsol_addr(&llsol, ifp, &ifra->ifra_addr.sin6_addr);
812 	if (error != 0)
813 		goto out;
814 	dad_delay = 0;
815 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
816 		/*
817 		 * We need a random delay for DAD on the address
818 		 * being configured.  It also means delaying
819 		 * transmission of the corresponding MLD report to
820 		 * avoid report collision.
821 		 * [draft-ietf-ipv6-rfc2462bis-02.txt]
822 		 */
823 		dad_delay = cprng_fast32() % (MAX_RTR_SOLICITATION_DELAY * hz);
824 	}
825 
826 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
827 	/* join solicited multicast addr for new host id */
828 	imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
829 	if (!imm) {
830 		nd6log(LOG_ERR,
831 		    "addmulti failed for %s on %s (errno=%d)\n",
832 		    IN6_PRINT(ip6buf, &llsol), if_name(ifp), error);
833 		goto out;
834 	}
835 	mutex_enter(&in6_ifaddr_lock);
836 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
837 	mutex_exit(&in6_ifaddr_lock);
838 
839 	sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
840 
841 	/*
842 	 * join link-local all-nodes address
843 	 */
844 	sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
845 	    0, 0, 0);
846 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
847 		goto out; /* XXX: should not fail */
848 
849 	/*
850 	 * XXX: do we really need this automatic routes?
851 	 * We should probably reconsider this stuff.  Most applications
852 	 * actually do not need the routes, since they usually specify
853 	 * the outgoing interface.
854 	 */
855 	rt = rtalloc1(sin6tosa(&mltaddr), 0);
856 	if (rt) {
857 		if (memcmp(&mltaddr.sin6_addr,
858 		    &satocsin6(rt_getkey(rt))->sin6_addr,
859 		    MLTMASK_LEN)) {
860 			rt_unref(rt);
861 			rt = NULL;
862 		} else if (rt->rt_ifp != ifp) {
863 			IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
864 			    "network %04x:%04x::/32 = %04x:%04x::/32\n",
865 			    __func__, rt->rt_ifp, ifp, ifp->if_xname,
866 			    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
867 			    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
868 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
869 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
870 			rt_replace_ifa(rt, &ia->ia_ifa);
871 			rt->rt_ifp = ifp;
872 		}
873 	}
874 	if (!rt) {
875 		struct rt_addrinfo info;
876 
877 		memset(&info, 0, sizeof(info));
878 		info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
879 		info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
880 		info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
881 		info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
882 		/* XXX: we need RTF_CONNECTED to fake nd6_rtrequest */
883 		info.rti_flags = RTF_UP | RTF_CONNECTED;
884 		error = rtrequest1(RTM_ADD, &info, NULL);
885 		if (error)
886 			goto out;
887 	} else {
888 		rt_unref(rt);
889 	}
890 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
891 	if (!imm) {
892 		nd6log(LOG_WARNING,
893 		    "addmulti failed for %s on %s (errno=%d)\n",
894 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
895 		    if_name(ifp), error);
896 		goto out;
897 	}
898 	mutex_enter(&in6_ifaddr_lock);
899 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
900 	mutex_exit(&in6_ifaddr_lock);
901 
902 	/*
903 	 * join node information group address
904 	 */
905 	dad_delay = 0;
906 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
907 		/*
908 		 * The spec doesn't say anything about delay for this
909 		 * group, but the same logic should apply.
910 		 */
911 		dad_delay = cprng_fast32() % (MAX_RTR_SOLICITATION_DELAY * hz);
912 	}
913 	if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
914 		;
915 	else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
916 		  dad_delay)) == NULL) { /* XXX jinmei */
917 		nd6log(LOG_WARNING,
918 		    "addmulti failed for %s on %s (errno=%d)\n",
919 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
920 		    if_name(ifp), error);
921 		/* XXX not very fatal, go on... */
922 	} else {
923 		mutex_enter(&in6_ifaddr_lock);
924 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
925 		mutex_exit(&in6_ifaddr_lock);
926 	}
927 
928 
929 	/*
930 	 * join interface-local all-nodes address.
931 	 * (ff01::1%ifN, and ff01::%ifN/32)
932 	 */
933 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
934 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
935 		goto out; /* XXX: should not fail */
936 
937 	/* XXX: again, do we really need the route? */
938 	rt = rtalloc1(sin6tosa(&mltaddr), 0);
939 	if (rt) {
940 		/* 32bit came from "mltmask" */
941 		if (memcmp(&mltaddr.sin6_addr,
942 		    &satocsin6(rt_getkey(rt))->sin6_addr,
943 		    32 / NBBY)) {
944 			rt_unref(rt);
945 			rt = NULL;
946 		} else if (rt->rt_ifp != ifp) {
947 			IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
948 			    "network %04x:%04x::/32 = %04x:%04x::/32\n",
949 			    __func__, rt->rt_ifp, ifp, ifp->if_xname,
950 			    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
951 			    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
952 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
953 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
954 			rt_replace_ifa(rt, &ia->ia_ifa);
955 			rt->rt_ifp = ifp;
956 		}
957 	}
958 	if (!rt) {
959 		struct rt_addrinfo info;
960 
961 		memset(&info, 0, sizeof(info));
962 		info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
963 		info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
964 		info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
965 		info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
966 		info.rti_flags = RTF_UP | RTF_CONNECTED;
967 		error = rtrequest1(RTM_ADD, &info, NULL);
968 		if (error)
969 			goto out;
970 #undef	MLTMASK_LEN
971 	} else {
972 		rt_unref(rt);
973 	}
974 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
975 	if (!imm) {
976 		nd6log(LOG_WARNING,
977 		    "addmulti failed for %s on %s (errno=%d)\n",
978 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
979 		    if_name(ifp), error);
980 		goto out;
981 	} else {
982 		mutex_enter(&in6_ifaddr_lock);
983 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
984 		mutex_exit(&in6_ifaddr_lock);
985 	}
986 	return 0;
987 
988 out:
989 	KASSERT(error != 0);
990 	return error;
991 }
992 
993 /*
994  * Update parameters of an IPv6 interface address.
995  * If necessary, a new entry is created and linked into address chains.
996  * This function is separated from in6_control().
997  * XXX: should this be performed under splsoftnet()?
998  */
999 static int
1000 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
1001     struct in6_ifaddr **iap, struct psref *psref, int flags)
1002 {
1003 	int error = 0, hostIsNew = 0, plen = -1;
1004 	struct sockaddr_in6 dst6;
1005 	struct in6_addrlifetime *lt;
1006 	int dad_delay, was_tentative;
1007 	struct in6_ifaddr *ia = iap ? *iap : NULL;
1008 	char ip6buf[INET6_ADDRSTRLEN];
1009 
1010 	KASSERT((iap == NULL && psref == NULL) ||
1011 	    (iap != NULL && psref != NULL));
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 #ifdef FORCE_P2PPLEN
1084 		int i;
1085 #endif
1086 
1087 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1088 			/* XXX: noisy message */
1089 			nd6log(LOG_INFO, "a destination can "
1090 			    "be specified for a p2p or a loopback IF only\n");
1091 			return EINVAL;
1092 		}
1093 		if (plen != 128) {
1094 			nd6log(LOG_INFO, "prefixlen should "
1095 			    "be 128 when dstaddr is specified\n");
1096 #ifdef FORCE_P2PPLEN
1097 			/*
1098 			 * To be compatible with old configurations,
1099 			 * such as ifconfig gif0 inet6 2001::1 2001::2
1100 			 * prefixlen 126, we override the specified
1101 			 * prefixmask as if the prefix length was 128.
1102 			 */
1103 			ifra->ifra_prefixmask.sin6_len =
1104 			    sizeof(struct sockaddr_in6);
1105 			for (i = 0; i < 4; i++)
1106 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
1107 				    0xffffffff;
1108 			plen = 128;
1109 #else
1110 			return EINVAL;
1111 #endif
1112 		}
1113 	}
1114 	/* lifetime consistency check */
1115 	lt = &ifra->ifra_lifetime;
1116 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1117 		return EINVAL;
1118 	if (lt->ia6t_vltime == 0) {
1119 		/*
1120 		 * the following log might be noisy, but this is a typical
1121 		 * configuration mistake or a tool's bug.
1122 		 */
1123 		nd6log(LOG_INFO, "valid lifetime is 0 for %s\n",
1124 		    IN6_PRINT(ip6buf, &ifra->ifra_addr.sin6_addr));
1125 
1126 		if (ia == NULL)
1127 			return 0; /* there's nothing to do */
1128 	}
1129 
1130 	/*
1131 	 * If this is a new address, allocate a new ifaddr and link it
1132 	 * into chains.
1133 	 */
1134 	if (ia == NULL) {
1135 		hostIsNew = 1;
1136 		/*
1137 		 * When in6_update_ifa() is called in a process of a received
1138 		 * RA, it is called under an interrupt context.  So, we should
1139 		 * call malloc with M_NOWAIT.
1140 		 */
1141 		ia = malloc(sizeof(*ia), M_IFADDR, M_NOWAIT|M_ZERO);
1142 		if (ia == NULL)
1143 			return ENOBUFS;
1144 		LIST_INIT(&ia->ia6_memberships);
1145 		/* Initialize the address and masks, and put time stamp */
1146 		ia->ia_ifa.ifa_addr = sin6tosa(&ia->ia_addr);
1147 		ia->ia_addr.sin6_family = AF_INET6;
1148 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1149 		ia->ia6_createtime = time_uptime;
1150 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1151 			/*
1152 			 * XXX: some functions expect that ifa_dstaddr is not
1153 			 * NULL for p2p interfaces.
1154 			 */
1155 			ia->ia_ifa.ifa_dstaddr = sin6tosa(&ia->ia_dstaddr);
1156 		} else {
1157 			ia->ia_ifa.ifa_dstaddr = NULL;
1158 		}
1159 		ia->ia_ifa.ifa_netmask = sin6tosa(&ia->ia_prefixmask);
1160 
1161 		ia->ia_ifp = ifp;
1162 		IN6_ADDRLIST_ENTRY_INIT(ia);
1163 		ifa_psref_init(&ia->ia_ifa);
1164 	}
1165 
1166 	/* update timestamp */
1167 	ia->ia6_updatetime = time_uptime;
1168 
1169 	/* set prefix mask */
1170 	if (ifra->ifra_prefixmask.sin6_len) {
1171 		if (ia->ia_prefixmask.sin6_len) {
1172 			/*
1173 			 * We prohibit changing the prefix length of an
1174 			 * existing autoconf address, because the operation
1175 			 * would confuse prefix management.
1176 			 */
1177 			if (ia->ia6_ndpr != NULL &&
1178 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) !=
1179 			    plen)
1180 			{
1181 				nd6log(LOG_INFO, "the prefix length of an"
1182 				    " existing (%s) autoconf address should"
1183 				    " not be changed\n",
1184 				    IN6_PRINT(ip6buf,
1185 				    &ia->ia_addr.sin6_addr));
1186 				error = EINVAL;
1187 				if (hostIsNew)
1188 					free(ia, M_IFADDR);
1189 				return error;
1190 			}
1191 
1192 			if (!IN6_ARE_ADDR_EQUAL(&ia->ia_prefixmask.sin6_addr,
1193 			    &ifra->ifra_prefixmask.sin6_addr))
1194 				in6_ifremprefix(ia);
1195 		}
1196 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1197 	}
1198 
1199 	/* Set destination address. */
1200 	if (dst6.sin6_family == AF_INET6) {
1201 		if (!IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
1202 		    &ia->ia_dstaddr.sin6_addr))
1203 			in6_ifremprefix(ia);
1204 		ia->ia_dstaddr = dst6;
1205 	}
1206 
1207 	/*
1208 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1209 	 * to see if the address is deprecated or invalidated, but initialize
1210 	 * these members for applications.
1211 	 */
1212 	ia->ia6_lifetime = ifra->ifra_lifetime;
1213 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1214 		ia->ia6_lifetime.ia6t_expire =
1215 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1216 	} else
1217 		ia->ia6_lifetime.ia6t_expire = 0;
1218 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1219 		ia->ia6_lifetime.ia6t_preferred =
1220 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1221 	} else
1222 		ia->ia6_lifetime.ia6t_preferred = 0;
1223 
1224 	/*
1225 	 * configure address flags.
1226 	 * We need to preserve tentative state so DAD works if
1227 	 * something adds the same address before DAD finishes.
1228 	 */
1229 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1230 	ia->ia6_flags = ifra->ifra_flags;
1231 
1232 	/*
1233 	 * Make the address tentative before joining multicast addresses,
1234 	 * so that corresponding MLD responses would not have a tentative
1235 	 * source address.
1236 	 */
1237 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1238 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1239 		ia->ia6_flags |= IN6_IFF_DETACHED;
1240 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1241 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp) &&
1242 	           ip6_dad_count > 0) {
1243 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1244 	}
1245 
1246 	/*
1247 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1248 	 * userland, make it deprecated.
1249 	 */
1250 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1251 		ia->ia6_lifetime.ia6t_pltime = 0;
1252 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1253 	}
1254 
1255 	if (hostIsNew) {
1256 		/*
1257 		 * We need a reference to ia before calling in6_ifinit.
1258 		 * Otherwise ia can be freed in in6_ifinit accidentally.
1259 		 */
1260 		ifaref(&ia->ia_ifa);
1261 	}
1262 
1263 	/* Must execute in6_ifinit and ifa_insert atomically */
1264 	mutex_enter(&in6_ifaddr_lock);
1265 
1266 	/* reset the interface and routing table appropriately. */
1267 	error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew);
1268 	if (error != 0) {
1269 		if (hostIsNew)
1270 			free(ia, M_IFADDR);
1271 		mutex_exit(&in6_ifaddr_lock);
1272 		return error;
1273 	}
1274 
1275 	/*
1276 	 * We are done if we have simply modified an existing address.
1277 	 */
1278 	if (!hostIsNew) {
1279 		mutex_exit(&in6_ifaddr_lock);
1280 		return error;
1281 	}
1282 
1283 	/*
1284 	 * Insert ia to the global list and ifa to the interface's list.
1285 	 * A reference to it is already gained above.
1286 	 */
1287 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1288 	ifa_insert(ifp, &ia->ia_ifa);
1289 
1290 	mutex_exit(&in6_ifaddr_lock);
1291 
1292 	/*
1293 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1294 	 * not just go to unlink.
1295 	 */
1296 
1297 	/* join necessary multicast groups */
1298 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1299 		error = in6_join_mcastgroups(ifra, ia, ifp, flags);
1300 		if (error != 0)
1301 			goto cleanup;
1302 	}
1303 
1304 	if (nd6_need_cache(ifp)) {
1305 		/* XXX maybe unnecessary */
1306 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1307 		ia->ia_ifa.ifa_flags |= RTF_CONNECTED;
1308 	}
1309 
1310 	/*
1311 	 * Perform DAD, if needed.
1312 	 * XXX It may be of use, if we can administratively
1313 	 * disable DAD.
1314 	 */
1315 	if (hostIsNew && if_do_dad(ifp) &&
1316 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1317 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1318 	{
1319 		int mindelay, maxdelay;
1320 
1321 		dad_delay = 0;
1322 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1323 			struct in6_addr llsol;
1324 			struct in6_multi *in6m_sol = NULL;
1325 			/*
1326 			 * We need to impose a delay before sending an NS
1327 			 * for DAD.  Check if we also needed a delay for the
1328 			 * corresponding MLD message.  If we did, the delay
1329 			 * should be larger than the MLD delay (this could be
1330 			 * relaxed a bit, but this simple logic is at least
1331 			 * safe).
1332 			 */
1333 			mindelay = 0;
1334 			error = in6_get_llsol_addr(&llsol, ifp,
1335 			    &ifra->ifra_addr.sin6_addr);
1336 			in6_multi_lock(RW_READER);
1337 			if (error == 0)
1338 				in6m_sol = in6_lookup_multi(&llsol, ifp);
1339 			if (in6m_sol != NULL &&
1340 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1341 				mindelay = in6m_sol->in6m_timer;
1342 			}
1343 			in6_multi_unlock();
1344 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1345 			if (maxdelay - mindelay == 0)
1346 				dad_delay = 0;
1347 			else {
1348 				dad_delay =
1349 				    (cprng_fast32() % (maxdelay - mindelay)) +
1350 				    mindelay;
1351 			}
1352 		}
1353 		/* +1 ensures callout is always used */
1354 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1355 	}
1356 
1357 	if (iap != NULL) {
1358 		*iap = ia;
1359 		if (hostIsNew)
1360 			ia6_acquire(ia, psref);
1361 	}
1362 
1363 	return 0;
1364 
1365   cleanup:
1366 	in6_purgeaddr(&ia->ia_ifa);
1367 	return error;
1368 }
1369 
1370 int
1371 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1372 {
1373 	int rc, s;
1374 
1375 	s = splsoftnet();
1376 	rc = in6_update_ifa1(ifp, ifra, NULL, NULL, flags);
1377 	splx(s);
1378 	return rc;
1379 }
1380 
1381 void
1382 in6_purgeaddr(struct ifaddr *ifa)
1383 {
1384 	struct ifnet *ifp = ifa->ifa_ifp;
1385 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1386 	struct in6_multi_mship *imm;
1387 
1388 	/* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */
1389 	KASSERT(IFNET_LOCKED(ifp));
1390 
1391 	ifa->ifa_flags |= IFA_DESTROYING;
1392 
1393 	/* stop DAD processing */
1394 	nd6_dad_stop(ifa);
1395 
1396 	/* Delete any network route. */
1397 	in6_ifremprefix(ia);
1398 
1399 	/* Remove ownaddr's loopback rtentry, if it exists. */
1400 	in6_ifremlocal(&(ia->ia_ifa));
1401 
1402 	/*
1403 	 * leave from multicast groups we have joined for the interface
1404 	 */
1405     again:
1406 	mutex_enter(&in6_ifaddr_lock);
1407 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1408 		struct in6_multi *in6m __diagused = imm->i6mm_maddr;
1409 		KASSERT(in6m == NULL || in6m->in6m_ifp == ifp);
1410 		LIST_REMOVE(imm, i6mm_chain);
1411 		mutex_exit(&in6_ifaddr_lock);
1412 
1413 		in6_leavegroup(imm);
1414 		goto again;
1415 	}
1416 	mutex_exit(&in6_ifaddr_lock);
1417 
1418 	in6_unlink_ifa(ia, ifp);
1419 }
1420 
1421 static void
1422 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1423 {
1424 	int	s = splsoftnet();
1425 
1426 	mutex_enter(&in6_ifaddr_lock);
1427 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1428 	ifa_remove(ifp, &ia->ia_ifa);
1429 	/* Assume ifa_remove called pserialize_perform and psref_destroy */
1430 	mutex_exit(&in6_ifaddr_lock);
1431 
1432 	/*
1433 	 * Release the reference to the ND prefix.
1434 	 */
1435 	if (ia->ia6_ndpr != NULL) {
1436 		nd6_prefix_unref(ia->ia6_ndpr);
1437 		ia->ia6_ndpr = NULL;
1438 	}
1439 
1440 	/*
1441 	 * Also, if the address being removed is autoconf'ed, call
1442 	 * nd6_pfxlist_onlink_check() since the release might affect the status of
1443 	 * other (detached) addresses.
1444 	 */
1445 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0) {
1446 		ND6_WLOCK();
1447 		nd6_pfxlist_onlink_check();
1448 		ND6_UNLOCK();
1449 	}
1450 
1451 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1452 
1453 	/*
1454 	 * release another refcnt for the link from in6_ifaddr.
1455 	 * Note that we should decrement the refcnt at least once for all *BSD.
1456 	 */
1457 	ifafree(&ia->ia_ifa);
1458 
1459 	splx(s);
1460 }
1461 
1462 void
1463 in6_purgeif(struct ifnet *ifp)
1464 {
1465 
1466 	IFNET_LOCK(ifp);
1467 	in6_ifdetach(ifp);
1468 	IFNET_UNLOCK(ifp);
1469 }
1470 
1471 void
1472 in6_purge_mcast_references(struct in6_multi *in6m)
1473 {
1474 	struct	in6_ifaddr *ia;
1475 
1476 	KASSERT(in6_multi_locked(RW_WRITER));
1477 
1478 	mutex_enter(&in6_ifaddr_lock);
1479 	IN6_ADDRLIST_WRITER_FOREACH(ia) {
1480 		struct in6_multi_mship *imm;
1481 		LIST_FOREACH(imm, &ia->ia6_memberships, i6mm_chain) {
1482 			if (imm->i6mm_maddr == in6m)
1483 				imm->i6mm_maddr = NULL;
1484 		}
1485 	}
1486 	mutex_exit(&in6_ifaddr_lock);
1487 }
1488 
1489 /*
1490  * SIOC[GAD]LIFADDR.
1491  *	SIOCGLIFADDR: get first address. (?)
1492  *	SIOCGLIFADDR with IFLR_PREFIX:
1493  *		get first address that matches the specified prefix.
1494  *	SIOCALIFADDR: add the specified address.
1495  *	SIOCALIFADDR with IFLR_PREFIX:
1496  *		add the specified prefix, filling hostid part from
1497  *		the first link-local address.  prefixlen must be <= 64.
1498  *	SIOCDLIFADDR: delete the specified address.
1499  *	SIOCDLIFADDR with IFLR_PREFIX:
1500  *		delete the first address that matches the specified prefix.
1501  * return values:
1502  *	EINVAL on invalid parameters
1503  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1504  *	other values may be returned from in6_ioctl()
1505  *
1506  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1507  * this is to accommodate address naming scheme other than RFC2374,
1508  * in the future.
1509  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1510  * address encoding scheme. (see figure on page 8)
1511  */
1512 static int
1513 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1514 	struct ifnet *ifp)
1515 {
1516 	struct in6_ifaddr *ia = NULL; /* XXX gcc 4.8 maybe-uninitialized */
1517 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1518 	struct ifaddr *ifa;
1519 	struct sockaddr *sa;
1520 
1521 	/* sanity checks */
1522 	if (!data || !ifp) {
1523 		panic("invalid argument to in6_lifaddr_ioctl");
1524 		/* NOTREACHED */
1525 	}
1526 
1527 	switch (cmd) {
1528 	case SIOCGLIFADDR:
1529 		/* address must be specified on GET with IFLR_PREFIX */
1530 		if ((iflr->flags & IFLR_PREFIX) == 0)
1531 			break;
1532 		/* FALLTHROUGH */
1533 	case SIOCALIFADDR:
1534 	case SIOCDLIFADDR:
1535 		/* address must be specified on ADD and DELETE */
1536 		sa = (struct sockaddr *)&iflr->addr;
1537 		if (sa->sa_family != AF_INET6)
1538 			return EINVAL;
1539 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1540 			return EINVAL;
1541 		/* XXX need improvement */
1542 		sa = (struct sockaddr *)&iflr->dstaddr;
1543 		if (sa->sa_family && sa->sa_family != AF_INET6)
1544 			return EINVAL;
1545 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1546 			return EINVAL;
1547 		break;
1548 	default: /* shouldn't happen */
1549 #if 0
1550 		panic("invalid cmd to in6_lifaddr_ioctl");
1551 		/* NOTREACHED */
1552 #else
1553 		return EOPNOTSUPP;
1554 #endif
1555 	}
1556 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1557 		return EINVAL;
1558 
1559 	switch (cmd) {
1560 	case SIOCALIFADDR:
1561 	    {
1562 		struct in6_aliasreq ifra;
1563 		struct in6_addr *xhostid = NULL;
1564 		int prefixlen;
1565 		int bound = curlwp_bind();
1566 		struct psref psref;
1567 
1568 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1569 			struct sockaddr_in6 *sin6;
1570 
1571 			/*
1572 			 * xhostid is to fill in the hostid part of the
1573 			 * address.  xhostid points to the first link-local
1574 			 * address attached to the interface.
1575 			 */
1576 			ia = in6ifa_ifpforlinklocal_psref(ifp, 0, &psref);
1577 			if (ia == NULL) {
1578 				curlwp_bindx(bound);
1579 				return EADDRNOTAVAIL;
1580 			}
1581 			xhostid = IFA_IN6(&ia->ia_ifa);
1582 
1583 		 	/* prefixlen must be <= 64. */
1584 			if (64 < iflr->prefixlen) {
1585 				ia6_release(ia, &psref);
1586 				curlwp_bindx(bound);
1587 				return EINVAL;
1588 			}
1589 			prefixlen = iflr->prefixlen;
1590 
1591 			/* hostid part must be zero. */
1592 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1593 			if (sin6->sin6_addr.s6_addr32[2] != 0
1594 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1595 				ia6_release(ia, &psref);
1596 				curlwp_bindx(bound);
1597 				return EINVAL;
1598 			}
1599 		} else
1600 			prefixlen = iflr->prefixlen;
1601 
1602 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1603 		memset(&ifra, 0, sizeof(ifra));
1604 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1605 
1606 		memcpy(&ifra.ifra_addr, &iflr->addr,
1607 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1608 		if (xhostid) {
1609 			/* fill in hostid part */
1610 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1611 			    xhostid->s6_addr32[2];
1612 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1613 			    xhostid->s6_addr32[3];
1614 		}
1615 
1616 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1617 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1618 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1619 			if (xhostid) {
1620 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1621 				    xhostid->s6_addr32[2];
1622 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1623 				    xhostid->s6_addr32[3];
1624 			}
1625 		}
1626 		if (xhostid) {
1627 			ia6_release(ia, &psref);
1628 			ia = NULL;
1629 		}
1630 		curlwp_bindx(bound);
1631 
1632 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1633 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1634 
1635 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1636 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1637 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1638 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1639 	    }
1640 	case SIOCGLIFADDR:
1641 	case SIOCDLIFADDR:
1642 	    {
1643 		struct in6_addr mask, candidate, match;
1644 		struct sockaddr_in6 *sin6;
1645 		int cmp;
1646 		int error, s;
1647 
1648 		memset(&mask, 0, sizeof(mask));
1649 		if (iflr->flags & IFLR_PREFIX) {
1650 			/* lookup a prefix rather than address. */
1651 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1652 
1653 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1654 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1655 			match.s6_addr32[0] &= mask.s6_addr32[0];
1656 			match.s6_addr32[1] &= mask.s6_addr32[1];
1657 			match.s6_addr32[2] &= mask.s6_addr32[2];
1658 			match.s6_addr32[3] &= mask.s6_addr32[3];
1659 
1660 			/* if you set extra bits, that's wrong */
1661 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1662 				return EINVAL;
1663 
1664 			cmp = 1;
1665 		} else {
1666 			if (cmd == SIOCGLIFADDR) {
1667 				/* on getting an address, take the 1st match */
1668 				cmp = 0;	/* XXX */
1669 			} else {
1670 				/* on deleting an address, do exact match */
1671 				in6_prefixlen2mask(&mask, 128);
1672 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1673 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1674 
1675 				cmp = 1;
1676 			}
1677 		}
1678 
1679 		s = pserialize_read_enter();
1680 		IFADDR_READER_FOREACH(ifa, ifp) {
1681 			if (ifa->ifa_addr->sa_family != AF_INET6)
1682 				continue;
1683 			if (!cmp)
1684 				break;
1685 
1686 			/*
1687 			 * XXX: this is adhoc, but is necessary to allow
1688 			 * a user to specify fe80::/64 (not /10) for a
1689 			 * link-local address.
1690 			 */
1691 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1692 			in6_clearscope(&candidate);
1693 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1694 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1695 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1696 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1697 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1698 				break;
1699 		}
1700 		if (!ifa) {
1701 			error = EADDRNOTAVAIL;
1702 			goto error;
1703 		}
1704 		ia = ifa2ia6(ifa);
1705 
1706 		if (cmd == SIOCGLIFADDR) {
1707 			/* fill in the if_laddrreq structure */
1708 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1709 			error = sa6_recoverscope(
1710 			    (struct sockaddr_in6 *)&iflr->addr);
1711 			if (error != 0)
1712 				goto error;
1713 
1714 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1715 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1716 				    ia->ia_dstaddr.sin6_len);
1717 				error = sa6_recoverscope(
1718 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1719 				if (error != 0)
1720 					goto error;
1721 			} else
1722 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1723 
1724 			iflr->prefixlen =
1725 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1726 
1727 			iflr->flags = ia->ia6_flags;	/* XXX */
1728 
1729 			error = 0;
1730 		} else {
1731 			struct in6_aliasreq ifra;
1732 
1733 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1734 			memset(&ifra, 0, sizeof(ifra));
1735 			memcpy(ifra.ifra_name, iflr->iflr_name,
1736 			    sizeof(ifra.ifra_name));
1737 
1738 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1739 			    ia->ia_addr.sin6_len);
1740 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1741 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1742 				    ia->ia_dstaddr.sin6_len);
1743 			} else {
1744 				memset(&ifra.ifra_dstaddr, 0,
1745 				    sizeof(ifra.ifra_dstaddr));
1746 			}
1747 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1748 			    ia->ia_prefixmask.sin6_len);
1749 
1750 			ifra.ifra_flags = ia->ia6_flags;
1751 			pserialize_read_exit(s);
1752 
1753 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1754 		}
1755 	error:
1756 		pserialize_read_exit(s);
1757 		return error;
1758 	    }
1759 	}
1760 
1761 	return EOPNOTSUPP;	/* just for safety */
1762 }
1763 
1764 /*
1765  * Initialize an interface's internet6 address
1766  * and routing table entry.
1767  */
1768 static int
1769 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1770 	const struct sockaddr_in6 *sin6, int newhost)
1771 {
1772 	int	error = 0, ifacount = 0;
1773 	int s;
1774 	struct ifaddr *ifa;
1775 
1776 	KASSERT(mutex_owned(&in6_ifaddr_lock));
1777 
1778 	/*
1779 	 * Give the interface a chance to initialize
1780 	 * if this is its first address,
1781 	 * and to validate the address if necessary.
1782 	 */
1783 	s = pserialize_read_enter();
1784 	IFADDR_READER_FOREACH(ifa, ifp) {
1785 		if (ifa->ifa_addr->sa_family != AF_INET6)
1786 			continue;
1787 		ifacount++;
1788 	}
1789 	pserialize_read_exit(s);
1790 
1791 	ia->ia_addr = *sin6;
1792 
1793 	if (ifacount == 0 &&
1794 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1795 		return error;
1796 	}
1797 
1798 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1799 
1800 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1801 
1802 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1803 	if (newhost) {
1804 		/* set the rtrequest function to create llinfo */
1805 		if (ifp->if_flags & IFF_POINTOPOINT)
1806 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1807 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1808 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1809 		in6_ifaddlocal(&ia->ia_ifa);
1810 	} else {
1811 		/* Inform the routing socket of new flags/timings */
1812 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
1813 	}
1814 
1815 	/* Add the network prefix route. */
1816 	if ((error = in6_ifaddprefix(ia)) != 0) {
1817 		if (newhost)
1818 			in6_ifremlocal(&ia->ia_ifa);
1819 		return error;
1820 	}
1821 
1822 	return error;
1823 }
1824 
1825 static struct ifaddr *
1826 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1827 {
1828 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1829 		return ifa;
1830 	return best_ifa;
1831 }
1832 
1833 /*
1834  * Find an IPv6 interface link-local address specific to an interface.
1835  */
1836 struct in6_ifaddr *
1837 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1838 {
1839 	struct ifaddr *best_ifa = NULL, *ifa;
1840 
1841 	IFADDR_READER_FOREACH(ifa, ifp) {
1842 		if (ifa->ifa_addr->sa_family != AF_INET6)
1843 			continue;
1844 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1845 			continue;
1846 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1847 			continue;
1848 		best_ifa = bestifa(best_ifa, ifa);
1849 	}
1850 
1851 	return (struct in6_ifaddr *)best_ifa;
1852 }
1853 
1854 struct in6_ifaddr *
1855 in6ifa_ifpforlinklocal_psref(const struct ifnet *ifp, const int ignoreflags,
1856     struct psref *psref)
1857 {
1858 	struct in6_ifaddr *ia;
1859 	int s = pserialize_read_enter();
1860 
1861 	ia = in6ifa_ifpforlinklocal(ifp, ignoreflags);
1862 	if (ia != NULL)
1863 		ia6_acquire(ia, psref);
1864 	pserialize_read_exit(s);
1865 
1866 	return ia;
1867 }
1868 
1869 /*
1870  * find the internet address corresponding to a given address.
1871  * ifaddr is returned referenced.
1872  */
1873 struct in6_ifaddr *
1874 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1875 {
1876 	struct in6_ifaddr *ia;
1877 	int s;
1878 
1879 	s = pserialize_read_enter();
1880 	IN6_ADDRLIST_READER_FOREACH(ia) {
1881 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1882 			if (zoneid != 0 &&
1883 			    zoneid != ia->ia_addr.sin6_scope_id)
1884 				continue;
1885 			ifaref(&ia->ia_ifa);
1886 			break;
1887 		}
1888 	}
1889 	pserialize_read_exit(s);
1890 
1891 	return ia;
1892 }
1893 
1894 /*
1895  * find the internet address corresponding to a given interface and address.
1896  */
1897 struct in6_ifaddr *
1898 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1899 {
1900 	struct ifaddr *best_ifa = NULL, *ifa;
1901 
1902 	IFADDR_READER_FOREACH(ifa, ifp) {
1903 		if (ifa->ifa_addr->sa_family != AF_INET6)
1904 			continue;
1905 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1906 			continue;
1907 		best_ifa = bestifa(best_ifa, ifa);
1908 	}
1909 
1910 	return (struct in6_ifaddr *)best_ifa;
1911 }
1912 
1913 struct in6_ifaddr *
1914 in6ifa_ifpwithaddr_psref(const struct ifnet *ifp, const struct in6_addr *addr,
1915     struct psref *psref)
1916 {
1917 	struct in6_ifaddr *ia;
1918 	int s = pserialize_read_enter();
1919 
1920 	ia = in6ifa_ifpwithaddr(ifp, addr);
1921 	if (ia != NULL)
1922 		ia6_acquire(ia, psref);
1923 	pserialize_read_exit(s);
1924 
1925 	return ia;
1926 }
1927 
1928 static struct in6_ifaddr *
1929 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1930 {
1931 	if (best_ia == NULL ||
1932 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1933 		return ia;
1934 	return best_ia;
1935 }
1936 
1937 /*
1938  * Determine if an address is on a local network.
1939  */
1940 int
1941 in6_localaddr(const struct in6_addr *in6)
1942 {
1943 	struct in6_ifaddr *ia;
1944 	int s;
1945 
1946 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1947 		return 1;
1948 
1949 	s = pserialize_read_enter();
1950 	IN6_ADDRLIST_READER_FOREACH(ia) {
1951 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1952 					      &ia->ia_prefixmask.sin6_addr)) {
1953 			pserialize_read_exit(s);
1954 			return 1;
1955 		}
1956 	}
1957 	pserialize_read_exit(s);
1958 
1959 	return 0;
1960 }
1961 
1962 int
1963 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1964 {
1965 	struct in6_ifaddr *ia;
1966 	int s;
1967 
1968 	s = pserialize_read_enter();
1969 	IN6_ADDRLIST_READER_FOREACH(ia) {
1970 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1971 		    &sa6->sin6_addr) &&
1972 #ifdef SCOPEDROUTING
1973 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1974 #endif
1975 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
1976 			pserialize_read_exit(s);
1977 			return 1; /* true */
1978 		}
1979 
1980 		/* XXX: do we still have to go thru the rest of the list? */
1981 	}
1982 	pserialize_read_exit(s);
1983 
1984 	return 0;		/* false */
1985 }
1986 
1987 /*
1988  * return length of part which dst and src are equal
1989  * hard coding...
1990  */
1991 int
1992 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1993 {
1994 	int match = 0;
1995 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1996 	u_char *lim = s + 16, r;
1997 
1998 	while (s < lim)
1999 		if ((r = (*d++ ^ *s++)) != 0) {
2000 			while (r < 128) {
2001 				match++;
2002 				r <<= 1;
2003 			}
2004 			break;
2005 		} else
2006 			match += NBBY;
2007 	return match;
2008 }
2009 
2010 /* XXX: to be scope conscious */
2011 int
2012 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2013 {
2014 	int bytelen, bitlen;
2015 
2016 	/* sanity check */
2017 	if (len < 0 || len > 128) {
2018 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2019 		    len);
2020 		return 0;
2021 	}
2022 
2023 	bytelen = len / NBBY;
2024 	bitlen = len % NBBY;
2025 
2026 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2027 		return 0;
2028 	if (bitlen != 0 &&
2029 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
2030 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
2031 		return 0;
2032 
2033 	return 1;
2034 }
2035 
2036 void
2037 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2038 {
2039 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2040 	int bytelen, bitlen, i;
2041 
2042 	/* sanity check */
2043 	if (len < 0 || len > 128) {
2044 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2045 		    len);
2046 		return;
2047 	}
2048 
2049 	memset(maskp, 0, sizeof(*maskp));
2050 	bytelen = len / NBBY;
2051 	bitlen = len % NBBY;
2052 	for (i = 0; i < bytelen; i++)
2053 		maskp->s6_addr[i] = 0xff;
2054 	if (bitlen)
2055 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2056 }
2057 
2058 /*
2059  * return the best address out of the same scope. if no address was
2060  * found, return the first valid address from designated IF.
2061  */
2062 struct in6_ifaddr *
2063 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2064 {
2065 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2066 	struct ifaddr *ifa;
2067 	struct in6_ifaddr *best_ia = NULL, *ia;
2068 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2069 
2070 	dep[0] = dep[1] = NULL;
2071 
2072 	/*
2073 	 * We first look for addresses in the same scope.
2074 	 * If there is one, return it.
2075 	 * If two or more, return one which matches the dst longest.
2076 	 * If none, return one of global addresses assigned other ifs.
2077 	 */
2078 	IFADDR_READER_FOREACH(ifa, ifp) {
2079 		if (ifa->ifa_addr->sa_family != AF_INET6)
2080 			continue;
2081 		ia = (struct in6_ifaddr *)ifa;
2082 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2083 			continue; /* XXX: is there any case to allow anycast? */
2084 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2085 			continue; /* don't use this interface */
2086 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2087 			continue;
2088 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2089 			if (ip6_use_deprecated)
2090 				dep[0] = ia;
2091 			continue;
2092 		}
2093 
2094 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
2095 			continue;
2096 		/*
2097 		 * call in6_matchlen() as few as possible
2098 		 */
2099 		if (best_ia == NULL) {
2100 			best_ia = ia;
2101 			continue;
2102 		}
2103 		if (blen == -1)
2104 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2105 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2106 		if (tlen > blen) {
2107 			blen = tlen;
2108 			best_ia = ia;
2109 		} else if (tlen == blen)
2110 			best_ia = bestia(best_ia, ia);
2111 	}
2112 	if (best_ia != NULL)
2113 		return best_ia;
2114 
2115 	IFADDR_READER_FOREACH(ifa, ifp) {
2116 		if (ifa->ifa_addr->sa_family != AF_INET6)
2117 			continue;
2118 		ia = (struct in6_ifaddr *)ifa;
2119 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2120 			continue; /* XXX: is there any case to allow anycast? */
2121 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2122 			continue; /* don't use this interface */
2123 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2124 			continue;
2125 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2126 			if (ip6_use_deprecated)
2127 				dep[1] = (struct in6_ifaddr *)ifa;
2128 			continue;
2129 		}
2130 
2131 		best_ia = bestia(best_ia, ia);
2132 	}
2133 	if (best_ia != NULL)
2134 		return best_ia;
2135 
2136 	/* use the last-resort values, that are, deprecated addresses */
2137 	if (dep[0])
2138 		return dep[0];
2139 	if (dep[1])
2140 		return dep[1];
2141 
2142 	return NULL;
2143 }
2144 
2145 /*
2146  * perform DAD when interface becomes IFF_UP.
2147  */
2148 void
2149 in6_if_link_up(struct ifnet *ifp)
2150 {
2151 	struct ifaddr *ifa;
2152 	struct in6_ifaddr *ia;
2153 	int s, bound;
2154 	char ip6buf[INET6_ADDRSTRLEN];
2155 
2156 	/* Ensure it's sane to run DAD */
2157 	if (ifp->if_link_state == LINK_STATE_DOWN)
2158 		return;
2159 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2160 		return;
2161 
2162 	bound = curlwp_bind();
2163 	s = pserialize_read_enter();
2164 	IFADDR_READER_FOREACH(ifa, ifp) {
2165 		struct psref psref;
2166 
2167 		if (ifa->ifa_addr->sa_family != AF_INET6)
2168 			continue;
2169 
2170 		ifa_acquire(ifa, &psref);
2171 		pserialize_read_exit(s);
2172 		ia = (struct in6_ifaddr *)ifa;
2173 
2174 		/* If detached then mark as tentative */
2175 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2176 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2177 			if (if_do_dad(ifp)) {
2178 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2179 				nd6log(LOG_ERR, "%s marked tentative\n",
2180 				    IN6_PRINT(ip6buf,
2181 				    &ia->ia_addr.sin6_addr));
2182 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2183 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2184 		}
2185 
2186 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2187 			int rand_delay;
2188 
2189 			/* Clear the duplicated flag as we're starting DAD. */
2190 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2191 
2192 			/*
2193 			 * The TENTATIVE flag was likely set by hand
2194 			 * beforehand, implicitly indicating the need for DAD.
2195 			 * We may be able to skip the random delay in this
2196 			 * case, but we impose delays just in case.
2197 			 */
2198 			rand_delay = cprng_fast32() %
2199 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2200 			/* +1 ensures callout is always used */
2201 			nd6_dad_start(ifa, rand_delay + 1);
2202 		}
2203 
2204 		s = pserialize_read_enter();
2205 		ifa_release(ifa, &psref);
2206 	}
2207 	pserialize_read_exit(s);
2208 	curlwp_bindx(bound);
2209 
2210 	/* Restore any detached prefixes */
2211 	ND6_WLOCK();
2212 	nd6_pfxlist_onlink_check();
2213 	ND6_UNLOCK();
2214 }
2215 
2216 void
2217 in6_if_up(struct ifnet *ifp)
2218 {
2219 
2220 	/*
2221 	 * special cases, like 6to4, are handled in in6_ifattach
2222 	 */
2223 	in6_ifattach(ifp, NULL);
2224 
2225 	/* interface may not support link state, so bring it up also */
2226 	in6_if_link_up(ifp);
2227 }
2228 
2229 /*
2230  * Mark all addresses as detached.
2231  */
2232 void
2233 in6_if_link_down(struct ifnet *ifp)
2234 {
2235 	struct ifaddr *ifa;
2236 	struct in6_ifaddr *ia;
2237 	int s, bound;
2238 	char ip6buf[INET6_ADDRSTRLEN];
2239 
2240 	/* Any prefixes on this interface should be detached as well */
2241 	ND6_WLOCK();
2242 	nd6_pfxlist_onlink_check();
2243 	ND6_UNLOCK();
2244 
2245 	bound = curlwp_bind();
2246 	s = pserialize_read_enter();
2247 	IFADDR_READER_FOREACH(ifa, ifp) {
2248 		struct psref psref;
2249 
2250 		if (ifa->ifa_addr->sa_family != AF_INET6)
2251 			continue;
2252 
2253 		ifa_acquire(ifa, &psref);
2254 		pserialize_read_exit(s);
2255 		ia = (struct in6_ifaddr *)ifa;
2256 
2257 		/* Stop DAD processing */
2258 		nd6_dad_stop(ifa);
2259 
2260 		/*
2261 		 * Mark the address as detached.
2262 		 * This satisfies RFC4862 Section 5.3, but we should apply
2263 		 * this logic to all addresses to be a good citizen and
2264 		 * avoid potential duplicated addresses.
2265 		 * When the interface comes up again, detached addresses
2266 		 * are marked tentative and DAD commences.
2267 		 */
2268 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2269 			nd6log(LOG_DEBUG, "%s marked detached\n",
2270 			    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
2271 			ia->ia6_flags |= IN6_IFF_DETACHED;
2272 			ia->ia6_flags &=
2273 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2274 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2275 		}
2276 
2277 		s = pserialize_read_enter();
2278 		ifa_release(ifa, &psref);
2279 	}
2280 	pserialize_read_exit(s);
2281 	curlwp_bindx(bound);
2282 }
2283 
2284 void
2285 in6_if_down(struct ifnet *ifp)
2286 {
2287 
2288 	in6_if_link_down(ifp);
2289 	lltable_purge_entries(LLTABLE6(ifp));
2290 }
2291 
2292 void
2293 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2294 {
2295 
2296 	switch (link_state) {
2297 	case LINK_STATE_DOWN:
2298 		in6_if_link_down(ifp);
2299 		break;
2300 	case LINK_STATE_UP:
2301 		in6_if_link_up(ifp);
2302 		break;
2303 	}
2304 }
2305 
2306 /*
2307  * Calculate max IPv6 MTU through all the interfaces and store it
2308  * to in6_maxmtu.
2309  */
2310 void
2311 in6_setmaxmtu(void)
2312 {
2313 	unsigned long maxmtu = 0;
2314 	struct ifnet *ifp;
2315 	int s;
2316 
2317 	s = pserialize_read_enter();
2318 	IFNET_READER_FOREACH(ifp) {
2319 		/* this function can be called during ifnet initialization */
2320 		if (!ifp->if_afdata[AF_INET6])
2321 			continue;
2322 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2323 		    IN6_LINKMTU(ifp) > maxmtu)
2324 			maxmtu = IN6_LINKMTU(ifp);
2325 	}
2326 	pserialize_read_exit(s);
2327 	if (maxmtu)	     /* update only when maxmtu is positive */
2328 		in6_maxmtu = maxmtu;
2329 }
2330 
2331 int
2332 in6_tunnel_validate(const struct ip6_hdr *ip6, const struct in6_addr *src,
2333     const struct in6_addr *dst)
2334 {
2335 
2336 	/* check for address match */
2337 	if (!IN6_ARE_ADDR_EQUAL(src, &ip6->ip6_dst) ||
2338 	    !IN6_ARE_ADDR_EQUAL(dst, &ip6->ip6_src))
2339 		return 0;
2340 
2341 	/* martian filters on outer source - done in ip6_input */
2342 
2343 	/* NOTE: the pakcet may be dropped by uRPF. */
2344 
2345 	/* return valid bytes length */
2346 	return sizeof(*src) + sizeof(*dst);
2347 }
2348 
2349 /*
2350  * Provide the length of interface identifiers to be used for the link attached
2351  * to the given interface.  The length should be defined in "IPv6 over
2352  * xxx-link" document.  Note that address architecture might also define
2353  * the length for a particular set of address prefixes, regardless of the
2354  * link type.  As clarified in rfc2462bis, those two definitions should be
2355  * consistent, and those really are as of August 2004.
2356  */
2357 int
2358 in6_if2idlen(struct ifnet *ifp)
2359 {
2360 	switch (ifp->if_type) {
2361 	case IFT_ETHER:		/* RFC2464 */
2362 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2363 	case IFT_L2VLAN:	/* ditto */
2364 	case IFT_IEEE80211:	/* ditto */
2365 	case IFT_FDDI:		/* RFC2467 */
2366 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2367 	case IFT_PPP:		/* RFC2472 */
2368 	case IFT_ARCNET:	/* RFC2497 */
2369 	case IFT_FRELAY:	/* RFC2590 */
2370 	case IFT_IEEE1394:	/* RFC3146 */
2371 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2372 	case IFT_LOOP:		/* XXX: is this really correct? */
2373 		return 64;
2374 	default:
2375 		/*
2376 		 * Unknown link type:
2377 		 * It might be controversial to use the today's common constant
2378 		 * of 64 for these cases unconditionally.  For full compliance,
2379 		 * we should return an error in this case.  On the other hand,
2380 		 * if we simply miss the standard for the link type or a new
2381 		 * standard is defined for a new link type, the IFID length
2382 		 * is very likely to be the common constant.  As a compromise,
2383 		 * we always use the constant, but make an explicit notice
2384 		 * indicating the "unknown" case.
2385 		 */
2386 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2387 		return 64;
2388 	}
2389 }
2390 
2391 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2392 #define	IN6_LLTBL_HASH(k, h) \
2393 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2394 
2395 /*
2396  * Do actual deallocation of @lle.
2397  * Called by LLE_FREE_LOCKED when number of references
2398  * drops to zero.
2399  */
2400 static void
2401 in6_lltable_destroy_lle(struct llentry *lle)
2402 {
2403 
2404 	KASSERT(lle->la_numheld == 0);
2405 
2406 	LLE_WUNLOCK(lle);
2407 	LLE_LOCK_DESTROY(lle);
2408 	llentry_pool_put(lle);
2409 }
2410 
2411 static struct llentry *
2412 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2413 {
2414 	struct llentry *lle;
2415 
2416 	lle = llentry_pool_get(PR_NOWAIT);
2417 	if (lle == NULL)		/* NB: caller generates msg */
2418 		return NULL;
2419 
2420 	lle->r_l3addr.addr6 = *addr6;
2421 	lle->lle_refcnt = 1;
2422 	lle->lle_free = in6_lltable_destroy_lle;
2423 	LLE_LOCK_INIT(lle);
2424 	callout_init(&lle->lle_timer, CALLOUT_MPSAFE);
2425 
2426 	return lle;
2427 }
2428 
2429 static int
2430 in6_lltable_match_prefix(const struct sockaddr *prefix,
2431     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2432 {
2433 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2434 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2435 
2436 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2437 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2438 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2439 		return 1;
2440 
2441 	return 0;
2442 }
2443 
2444 static void
2445 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2446 {
2447 
2448 	LLE_WLOCK_ASSERT(lle);
2449 	(void) llentry_free(lle);
2450 }
2451 
2452 static int
2453 in6_lltable_rtcheck(struct ifnet *ifp, u_int flags,
2454     const struct sockaddr *l3addr, const struct rtentry *rt)
2455 {
2456 	char ip6buf[INET6_ADDRSTRLEN];
2457 
2458 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2459 		int s;
2460 		struct ifaddr *ifa;
2461 		/*
2462 		 * Create an ND6 cache for an IPv6 neighbor
2463 		 * that is not covered by our own prefix.
2464 		 */
2465 		/* XXX ifaof_ifpforaddr should take a const param */
2466 		s = pserialize_read_enter();
2467 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2468 		if (ifa != NULL) {
2469 			pserialize_read_exit(s);
2470 			return 0;
2471 		}
2472 		pserialize_read_exit(s);
2473 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2474 		    IN6_PRINT(ip6buf,
2475 		    &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2476 		return EINVAL;
2477 	}
2478 	return 0;
2479 }
2480 
2481 static inline uint32_t
2482 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2483 {
2484 
2485 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2486 }
2487 
2488 static uint32_t
2489 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2490 {
2491 
2492 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2493 }
2494 
2495 static void
2496 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2497 {
2498 	struct sockaddr_in6 *sin6;
2499 
2500 	sin6 = (struct sockaddr_in6 *)sa;
2501 	bzero(sin6, sizeof(*sin6));
2502 	sin6->sin6_family = AF_INET6;
2503 	sin6->sin6_len = sizeof(*sin6);
2504 	sin6->sin6_addr = lle->r_l3addr.addr6;
2505 }
2506 
2507 static inline struct llentry *
2508 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2509 {
2510 	struct llentry *lle;
2511 	struct llentries *lleh;
2512 	u_int hashidx;
2513 
2514 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2515 	lleh = &llt->lle_head[hashidx];
2516 	LIST_FOREACH(lle, lleh, lle_next) {
2517 		if (lle->la_flags & LLE_DELETED)
2518 			continue;
2519 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2520 			break;
2521 	}
2522 
2523 	return lle;
2524 }
2525 
2526 static int
2527 in6_lltable_delete(struct lltable *llt, u_int flags,
2528 	const struct sockaddr *l3addr)
2529 {
2530 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2531 	struct llentry *lle;
2532 
2533 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2534 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2535 	    "sin_family %d", l3addr->sa_family);
2536 
2537 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2538 
2539 	if (lle == NULL) {
2540 #ifdef LLTABLE_DEBUG
2541 		char buf[64];
2542 		sockaddr_format(l3addr, buf, sizeof(buf));
2543 		log(LOG_INFO, "%s: cache for %s is not found\n",
2544 		    __func__, buf);
2545 #endif
2546 		return ENOENT;
2547 	}
2548 
2549 	LLE_WLOCK(lle);
2550 #ifdef LLTABLE_DEBUG
2551 	{
2552 		char buf[64];
2553 		sockaddr_format(l3addr, buf, sizeof(buf));
2554 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2555 		    __func__, buf, lle);
2556 	}
2557 #endif
2558 	llentry_free(lle);
2559 
2560 	return 0;
2561 }
2562 
2563 static struct llentry *
2564 in6_lltable_create(struct lltable *llt, u_int flags,
2565     const struct sockaddr *l3addr, const struct rtentry *rt)
2566 {
2567 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2568 	struct ifnet *ifp = llt->llt_ifp;
2569 	struct llentry *lle;
2570 
2571 	IF_AFDATA_WLOCK_ASSERT(ifp);
2572 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2573 	    "sin_family %d", l3addr->sa_family);
2574 
2575 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2576 
2577 	if (lle != NULL) {
2578 		LLE_WLOCK(lle);
2579 		return lle;
2580 	}
2581 
2582 	/*
2583 	 * A route that covers the given address must have
2584 	 * been installed 1st because we are doing a resolution,
2585 	 * verify this.
2586 	 */
2587 	if (!(flags & LLE_IFADDR) &&
2588 	    in6_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2589 		return NULL;
2590 
2591 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2592 	if (lle == NULL) {
2593 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2594 		return NULL;
2595 	}
2596 	lle->la_flags = flags;
2597 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2598 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2599 		lle->la_flags |= LLE_VALID;
2600 	}
2601 
2602 	lltable_link_entry(llt, lle);
2603 	LLE_WLOCK(lle);
2604 
2605 	return lle;
2606 }
2607 
2608 static struct llentry *
2609 in6_lltable_lookup(struct lltable *llt, u_int flags,
2610 	const struct sockaddr *l3addr)
2611 {
2612 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2613 	struct llentry *lle;
2614 
2615 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2616 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2617 	    "sin_family %d", l3addr->sa_family);
2618 
2619 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2620 
2621 	if (lle == NULL)
2622 		return NULL;
2623 
2624 	if (flags & LLE_EXCLUSIVE)
2625 		LLE_WLOCK(lle);
2626 	else
2627 		LLE_RLOCK(lle);
2628 	return lle;
2629 }
2630 
2631 static int
2632 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2633     struct rt_walkarg *w)
2634 {
2635 	struct sockaddr_in6 sin6;
2636 
2637 	LLTABLE_LOCK_ASSERT();
2638 
2639 	/* skip deleted entries */
2640 	if (lle->la_flags & LLE_DELETED)
2641 		return 0;
2642 
2643 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2644 
2645 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2646 }
2647 
2648 static struct lltable *
2649 in6_lltattach(struct ifnet *ifp)
2650 {
2651 	struct lltable *llt;
2652 
2653 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2654 	llt->llt_af = AF_INET6;
2655 	llt->llt_ifp = ifp;
2656 
2657 	llt->llt_lookup = in6_lltable_lookup;
2658 	llt->llt_create = in6_lltable_create;
2659 	llt->llt_delete = in6_lltable_delete;
2660 	llt->llt_dump_entry = in6_lltable_dump_entry;
2661 	llt->llt_hash = in6_lltable_hash;
2662 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2663 	llt->llt_free_entry = in6_lltable_free_entry;
2664 	llt->llt_match_prefix = in6_lltable_match_prefix;
2665 	lltable_link(llt);
2666 
2667 	return llt;
2668 }
2669 
2670 void *
2671 in6_domifattach(struct ifnet *ifp)
2672 {
2673 	struct in6_ifextra *ext;
2674 
2675 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2676 
2677 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2678 	    M_IFADDR, M_WAITOK|M_ZERO);
2679 
2680 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2681 	    M_IFADDR, M_WAITOK|M_ZERO);
2682 
2683 	ext->nd_ifinfo = nd6_ifattach(ifp);
2684 	ext->scope6_id = scope6_ifattach(ifp);
2685 	ext->nprefixes = 0;
2686 	ext->ndefrouters = 0;
2687 
2688 	ext->lltable = in6_lltattach(ifp);
2689 
2690 	return ext;
2691 }
2692 
2693 void
2694 in6_domifdetach(struct ifnet *ifp, void *aux)
2695 {
2696 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2697 
2698 	lltable_free(ext->lltable);
2699 	ext->lltable = NULL;
2700 	SOFTNET_LOCK_UNLESS_NET_MPSAFE();
2701 	nd6_ifdetach(ifp, ext);
2702 	SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();
2703 	free(ext->in6_ifstat, M_IFADDR);
2704 	free(ext->icmp6_ifstat, M_IFADDR);
2705 	scope6_ifdetach(ext->scope6_id);
2706 	free(ext, M_IFADDR);
2707 }
2708 
2709 /*
2710  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2711  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2712  */
2713 void
2714 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2715 {
2716 	in6->s6_addr32[0] = 0;
2717 	in6->s6_addr32[1] = 0;
2718 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2719 	in6->s6_addr32[3] = in->s_addr;
2720 }
2721 
2722 /*
2723  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2724  * v4 mapped addr or v4 compat addr
2725  */
2726 void
2727 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2728 {
2729 	memset(sin, 0, sizeof(*sin));
2730 	sin->sin_len = sizeof(struct sockaddr_in);
2731 	sin->sin_family = AF_INET;
2732 	sin->sin_port = sin6->sin6_port;
2733 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2734 }
2735 
2736 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2737 void
2738 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2739 {
2740 	memset(sin6, 0, sizeof(*sin6));
2741 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2742 	sin6->sin6_family = AF_INET6;
2743 	sin6->sin6_port = sin->sin_port;
2744 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2745 }
2746 
2747 /* Convert sockaddr_in6 into sockaddr_in. */
2748 void
2749 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2750 {
2751 	struct sockaddr_in *sin_p;
2752 	struct sockaddr_in6 sin6;
2753 
2754 	/*
2755 	 * Save original sockaddr_in6 addr and convert it
2756 	 * to sockaddr_in.
2757 	 */
2758 	sin6 = *(struct sockaddr_in6 *)nam;
2759 	sin_p = (struct sockaddr_in *)nam;
2760 	in6_sin6_2_sin(sin_p, &sin6);
2761 }
2762 
2763 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2764 void
2765 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2766 {
2767 	struct sockaddr_in *sin_p;
2768 	struct sockaddr_in6 *sin6_p;
2769 
2770 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2771 	sin_p = (struct sockaddr_in *)*nam;
2772 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2773 	free(*nam, M_SONAME);
2774 	*nam = sin6tosa(sin6_p);
2775 }
2776