xref: /netbsd-src/sys/netinet6/in6.c (revision 7e30e94394d0994ab9534f68a8f91665045c91ce)
1 /*	$NetBSD: in6.c,v 1.244 2017/03/02 09:48:20 ozaki-r 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 1.244 2017/03/02 09:48:20 ozaki-r 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 #include <net/net_osdep.h>
109 
110 #ifdef COMPAT_50
111 #include <compat/netinet6/in6_var.h>
112 #endif
113 
114 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
115 
116 /* enable backward compatibility code for obsoleted ioctls */
117 #define COMPAT_IN6IFIOCTL
118 
119 #ifdef	IN6_DEBUG
120 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
121 #else
122 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
123 #endif /* IN6_DEBUG */
124 
125 /*
126  * Definitions of some constant IP6 addresses.
127  */
128 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
129 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
130 const struct in6_addr in6addr_nodelocal_allnodes =
131 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
132 const struct in6_addr in6addr_linklocal_allnodes =
133 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
134 const struct in6_addr in6addr_linklocal_allrouters =
135 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
136 
137 const struct in6_addr in6mask0 = IN6MASK0;
138 const struct in6_addr in6mask32 = IN6MASK32;
139 const struct in6_addr in6mask64 = IN6MASK64;
140 const struct in6_addr in6mask96 = IN6MASK96;
141 const struct in6_addr in6mask128 = IN6MASK128;
142 
143 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
144 				     0, 0, IN6ADDR_ANY_INIT, 0};
145 
146 struct pslist_head	in6_ifaddr_list;
147 kmutex_t		in6_ifaddr_lock;
148 
149 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
150 	struct ifnet *);
151 static int in6_ifaddprefix(struct in6_ifaddr *);
152 static int in6_ifremprefix(struct in6_ifaddr *);
153 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
154 	const struct sockaddr_in6 *, int);
155 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
156 static int in6_update_ifa1(struct ifnet *, struct in6_aliasreq *,
157     struct in6_ifaddr **, struct psref *, int);
158 
159 void
160 in6_init(void)
161 {
162 
163 	PSLIST_INIT(&in6_ifaddr_list);
164 	mutex_init(&in6_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
165 
166 	in6_sysctl_multicast_setup(NULL);
167 }
168 
169 /*
170  * Add ownaddr as loopback rtentry.  We previously add the route only if
171  * necessary (ex. on a p2p link).  However, since we now manage addresses
172  * separately from prefixes, we should always add the route.  We can't
173  * rely on the cloning mechanism from the corresponding interface route
174  * any more.
175  */
176 void
177 in6_ifaddlocal(struct ifaddr *ifa)
178 {
179 
180 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
181 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
182 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
183 	{
184 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
185 		return;
186 	}
187 
188 	rt_ifa_addlocal(ifa);
189 }
190 
191 /*
192  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
193  * if it exists.
194  */
195 void
196 in6_ifremlocal(struct ifaddr *ifa)
197 {
198 	struct in6_ifaddr *ia;
199 	struct ifaddr *alt_ifa = NULL;
200 	int ia_count = 0;
201 	struct psref psref;
202 	int s;
203 
204 	/*
205 	 * Some of BSD variants do not remove cloned routes
206 	 * from an interface direct route, when removing the direct route
207 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
208 	 * cloned routes, they could fail to remove the cloned routes when
209 	 * we handle multple addresses that share a common prefix.
210 	 * So, we should remove the route corresponding to the deleted address.
211 	 */
212 
213 	/*
214 	 * Delete the entry only if exactly one ifaddr matches the
215 	 * address, ifa->ifa_addr.
216 	 *
217 	 * If more than one ifaddr matches, replace the ifaddr in
218 	 * the routing table, rt_ifa, with a different ifaddr than
219 	 * the one we are purging, ifa.  It is important to do
220 	 * this, or else the routing table can accumulate dangling
221 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
222 	 * which will lead to crashes, later.  (More than one ifaddr
223 	 * can match if we assign the same address to multiple---probably
224 	 * p2p---interfaces.)
225 	 *
226 	 * XXX An old comment at this place said, "we should avoid
227 	 * XXX such a configuration [i.e., interfaces with the same
228 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
229 	 * XXX agree, especially now that I have fixed the dangling
230 	 * XXX ifp-pointers bug.
231 	 */
232 	s = pserialize_read_enter();
233 	IN6_ADDRLIST_READER_FOREACH(ia) {
234 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
235 			continue;
236 		if (ia->ia_ifp != ifa->ifa_ifp)
237 			alt_ifa = &ia->ia_ifa;
238 		if (++ia_count > 1 && alt_ifa != NULL)
239 			break;
240 	}
241 	if (ia_count > 1 && alt_ifa != NULL)
242 		ifa_acquire(alt_ifa, &psref);
243 	pserialize_read_exit(s);
244 
245 	if (ia_count == 0)
246 		return;
247 
248 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
249 
250 	if (ia_count > 1 && alt_ifa != NULL)
251 		ifa_release(alt_ifa, &psref);
252 }
253 
254 /* Add prefix route for the network. */
255 static int
256 in6_ifaddprefix(struct in6_ifaddr *ia)
257 {
258 	int error, flags = 0;
259 
260 	if (in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) == 128) {
261 		if (ia->ia_dstaddr.sin6_family != AF_INET6)
262 			/* We don't need to install a host route. */
263 			return 0;
264 		flags |= RTF_HOST;
265 	}
266 
267 	/* Is this a connected route for neighbour discovery? */
268 	if (nd6_need_cache(ia->ia_ifp))
269 		flags |= RTF_CONNECTED;
270 
271 	if ((error = rtinit(&ia->ia_ifa, RTM_ADD, RTF_UP | flags)) == 0)
272 		ia->ia_flags |= IFA_ROUTE;
273 	else if (error == EEXIST)
274 		/* Existance of the route is not an error. */
275 		error = 0;
276 
277 	return error;
278 }
279 
280 /* Delete network prefix route if present.
281  * Re-add it to another address if the prefix matches. */
282 static int
283 in6_ifremprefix(struct in6_ifaddr *target)
284 {
285 	int error, s;
286 	struct in6_ifaddr *ia;
287 
288 	if ((target->ia_flags & IFA_ROUTE) == 0)
289 		return 0;
290 
291 	s = pserialize_read_enter();
292 	IN6_ADDRLIST_READER_FOREACH(ia) {
293 		if (target->ia_dstaddr.sin6_len) {
294 			if (ia->ia_dstaddr.sin6_len == 0 ||
295 			    !IN6_ARE_ADDR_EQUAL(&ia->ia_dstaddr.sin6_addr,
296 			    &target->ia_dstaddr.sin6_addr))
297 				continue;
298 		} else {
299 			if (!IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
300 			    &target->ia_addr.sin6_addr,
301 			    &target->ia_prefixmask.sin6_addr))
302 				continue;
303 		}
304 
305 		/*
306 		 * if we got a matching prefix route, move IFA_ROUTE to him
307 		 */
308 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
309 			struct psref psref;
310 			int bound = curlwp_bind();
311 
312 			ia6_acquire(ia, &psref);
313 			pserialize_read_exit(s);
314 
315 			rtinit(&target->ia_ifa, RTM_DELETE, 0);
316 			target->ia_flags &= ~IFA_ROUTE;
317 
318 			error = in6_ifaddprefix(ia);
319 
320 			ia6_release(ia, &psref);
321 			curlwp_bindx(bound);
322 
323 			return error;
324 		}
325 	}
326 	pserialize_read_exit(s);
327 
328 	/*
329 	 * noone seem to have prefix route.  remove it.
330 	 */
331 	rtinit(&target->ia_ifa, RTM_DELETE, 0);
332 	target->ia_flags &= ~IFA_ROUTE;
333 	return 0;
334 }
335 
336 int
337 in6_mask2len(struct in6_addr *mask, u_char *lim0)
338 {
339 	int x = 0, y;
340 	u_char *lim = lim0, *p;
341 
342 	/* ignore the scope_id part */
343 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
344 		lim = (u_char *)mask + sizeof(*mask);
345 	for (p = (u_char *)mask; p < lim; x++, p++) {
346 		if (*p != 0xff)
347 			break;
348 	}
349 	y = 0;
350 	if (p < lim) {
351 		for (y = 0; y < NBBY; y++) {
352 			if ((*p & (0x80 >> y)) == 0)
353 				break;
354 		}
355 	}
356 
357 	/*
358 	 * when the limit pointer is given, do a stricter check on the
359 	 * remaining bits.
360 	 */
361 	if (p < lim) {
362 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
363 			return -1;
364 		for (p = p + 1; p < lim; p++)
365 			if (*p != 0)
366 				return -1;
367 	}
368 
369 	return x * NBBY + y;
370 }
371 
372 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
373 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
374 
375 static int
376 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
377 {
378 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
379 	struct	in6_ifaddr *ia = NULL;
380 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
381 	struct sockaddr_in6 *sa6;
382 	int error, bound;
383 	struct psref psref;
384 
385 	switch (cmd) {
386 	case SIOCAADDRCTL_POLICY:
387 	case SIOCDADDRCTL_POLICY:
388 		/* Privileged. */
389 		return in6_src_ioctl(cmd, data);
390 	/*
391 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
392 	 */
393 	case SIOCSIFADDR:
394 	case SIOCSIFDSTADDR:
395 	case SIOCSIFBRDADDR:
396 	case SIOCSIFNETMASK:
397 		return EOPNOTSUPP;
398 	case SIOCGETSGCNT_IN6:
399 	case SIOCGETMIFCNT_IN6:
400 		return mrt6_ioctl(cmd, data);
401 	case SIOCGIFADDRPREF:
402 	case SIOCSIFADDRPREF:
403 		if (ifp == NULL)
404 			return EINVAL;
405 		return ifaddrpref_ioctl(so, cmd, data, ifp);
406 	}
407 
408 	if (ifp == NULL)
409 		return EOPNOTSUPP;
410 
411 	switch (cmd) {
412 	case SIOCSNDFLUSH_IN6:
413 	case SIOCSPFXFLUSH_IN6:
414 	case SIOCSRTRFLUSH_IN6:
415 	case SIOCSDEFIFACE_IN6:
416 	case SIOCSIFINFO_FLAGS:
417 	case SIOCSIFINFO_IN6:
418 		/* Privileged. */
419 		/* FALLTHROUGH */
420 	case OSIOCGIFINFO_IN6:
421 	case SIOCGIFINFO_IN6:
422 	case SIOCGDRLST_IN6:
423 	case SIOCGPRLST_IN6:
424 	case SIOCGNBRINFO_IN6:
425 	case SIOCGDEFIFACE_IN6:
426 		return nd6_ioctl(cmd, data, ifp);
427 	}
428 
429 	switch (cmd) {
430 	case SIOCSIFPREFIX_IN6:
431 	case SIOCDIFPREFIX_IN6:
432 	case SIOCAIFPREFIX_IN6:
433 	case SIOCCIFPREFIX_IN6:
434 	case SIOCSGIFPREFIX_IN6:
435 	case SIOCGIFPREFIX_IN6:
436 		log(LOG_NOTICE,
437 		    "prefix ioctls are now invalidated. "
438 		    "please use ifconfig.\n");
439 		return EOPNOTSUPP;
440 	}
441 
442 	switch (cmd) {
443 	case SIOCALIFADDR:
444 	case SIOCDLIFADDR:
445 		/* Privileged. */
446 		/* FALLTHROUGH */
447 	case SIOCGLIFADDR:
448 		return in6_lifaddr_ioctl(so, cmd, data, ifp);
449 	}
450 
451 	/*
452 	 * Find address for this interface, if it exists.
453 	 *
454 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
455 	 * only, and used the first interface address as the target of other
456 	 * operations (without checking ifra_addr).  This was because netinet
457 	 * code/API assumed at most 1 interface address per interface.
458 	 * Since IPv6 allows a node to assign multiple addresses
459 	 * on a single interface, we almost always look and check the
460 	 * presence of ifra_addr, and reject invalid ones here.
461 	 * It also decreases duplicated code among SIOC*_IN6 operations.
462 	 */
463 	switch (cmd) {
464 	case SIOCAIFADDR_IN6:
465 #ifdef OSIOCAIFADDR_IN6
466 	case OSIOCAIFADDR_IN6:
467 #endif
468 #ifdef OSIOCSIFPHYADDR_IN6
469 	case OSIOCSIFPHYADDR_IN6:
470 #endif
471 	case SIOCSIFPHYADDR_IN6:
472 		sa6 = &ifra->ifra_addr;
473 		break;
474 	case SIOCSIFADDR_IN6:
475 	case SIOCGIFADDR_IN6:
476 	case SIOCSIFDSTADDR_IN6:
477 	case SIOCSIFNETMASK_IN6:
478 	case SIOCGIFDSTADDR_IN6:
479 	case SIOCGIFNETMASK_IN6:
480 	case SIOCDIFADDR_IN6:
481 	case SIOCGIFPSRCADDR_IN6:
482 	case SIOCGIFPDSTADDR_IN6:
483 	case SIOCGIFAFLAG_IN6:
484 	case SIOCSNDFLUSH_IN6:
485 	case SIOCSPFXFLUSH_IN6:
486 	case SIOCSRTRFLUSH_IN6:
487 	case SIOCGIFALIFETIME_IN6:
488 #ifdef OSIOCGIFALIFETIME_IN6
489 	case OSIOCGIFALIFETIME_IN6:
490 #endif
491 	case SIOCGIFSTAT_IN6:
492 	case SIOCGIFSTAT_ICMP6:
493 		sa6 = &ifr->ifr_addr;
494 		break;
495 	default:
496 		sa6 = NULL;
497 		break;
498 	}
499 
500 	error = 0;
501 	bound = curlwp_bind();
502 	if (sa6 && sa6->sin6_family == AF_INET6) {
503 		if (sa6->sin6_scope_id != 0)
504 			error = sa6_embedscope(sa6, 0);
505 		else
506 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
507 		if (error != 0)
508 			goto out;
509 		ia = in6ifa_ifpwithaddr_psref(ifp, &sa6->sin6_addr, &psref);
510 	} else
511 		ia = NULL;
512 
513 	switch (cmd) {
514 	case SIOCSIFADDR_IN6:
515 	case SIOCSIFDSTADDR_IN6:
516 	case SIOCSIFNETMASK_IN6:
517 		/*
518 		 * Since IPv6 allows a node to assign multiple addresses
519 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
520 		 */
521 		error = EINVAL;
522 		goto release;
523 
524 	case SIOCDIFADDR_IN6:
525 		/*
526 		 * for IPv4, we look for existing in_ifaddr here to allow
527 		 * "ifconfig if0 delete" to remove the first IPv4 address on
528 		 * the interface.  For IPv6, as the spec allows multiple
529 		 * interface address from the day one, we consider "remove the
530 		 * first one" semantics to be not preferable.
531 		 */
532 		if (ia == NULL) {
533 			error = EADDRNOTAVAIL;
534 			goto out;
535 		}
536 		/* FALLTHROUGH */
537 #ifdef OSIOCAIFADDR_IN6
538 	case OSIOCAIFADDR_IN6:
539 #endif
540 	case SIOCAIFADDR_IN6:
541 		/*
542 		 * We always require users to specify a valid IPv6 address for
543 		 * the corresponding operation.
544 		 */
545 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
546 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
547 			error = EAFNOSUPPORT;
548 			goto release;
549 		}
550 		/* Privileged. */
551 
552 		break;
553 
554 	case SIOCGIFADDR_IN6:
555 		/* This interface is basically deprecated. use SIOCGIFCONF. */
556 		/* FALLTHROUGH */
557 	case SIOCGIFAFLAG_IN6:
558 	case SIOCGIFNETMASK_IN6:
559 	case SIOCGIFDSTADDR_IN6:
560 	case SIOCGIFALIFETIME_IN6:
561 #ifdef OSIOCGIFALIFETIME_IN6
562 	case OSIOCGIFALIFETIME_IN6:
563 #endif
564 		/* must think again about its semantics */
565 		if (ia == NULL) {
566 			error = EADDRNOTAVAIL;
567 			goto out;
568 		}
569 		break;
570 	}
571 
572 	switch (cmd) {
573 
574 	case SIOCGIFADDR_IN6:
575 		ifr->ifr_addr = ia->ia_addr;
576 		error = sa6_recoverscope(&ifr->ifr_addr);
577 		break;
578 
579 	case SIOCGIFDSTADDR_IN6:
580 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
581 			error = EINVAL;
582 			break;
583 		}
584 		/*
585 		 * XXX: should we check if ifa_dstaddr is NULL and return
586 		 * an error?
587 		 */
588 		ifr->ifr_dstaddr = ia->ia_dstaddr;
589 		error = sa6_recoverscope(&ifr->ifr_dstaddr);
590 		break;
591 
592 	case SIOCGIFNETMASK_IN6:
593 		ifr->ifr_addr = ia->ia_prefixmask;
594 		break;
595 
596 	case SIOCGIFAFLAG_IN6:
597 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
598 		break;
599 
600 	case SIOCGIFSTAT_IN6:
601 		if (ifp == NULL) {
602 			error = EINVAL;
603 			break;
604 		}
605 		memset(&ifr->ifr_ifru.ifru_stat, 0,
606 		    sizeof(ifr->ifr_ifru.ifru_stat));
607 		ifr->ifr_ifru.ifru_stat =
608 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
609 		break;
610 
611 	case SIOCGIFSTAT_ICMP6:
612 		if (ifp == NULL) {
613 			error = EINVAL;
614 			break;
615 		}
616 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
617 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
618 		ifr->ifr_ifru.ifru_icmp6stat =
619 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
620 		break;
621 
622 #ifdef OSIOCGIFALIFETIME_IN6
623 	case OSIOCGIFALIFETIME_IN6:
624 #endif
625 	case SIOCGIFALIFETIME_IN6:
626 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
627 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
628 			time_t maxexpire;
629 			struct in6_addrlifetime *retlt =
630 			    &ifr->ifr_ifru.ifru_lifetime;
631 
632 			/*
633 			 * XXX: adjust expiration time assuming time_t is
634 			 * signed.
635 			 */
636 			maxexpire = ((time_t)~0) &
637 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
638 			if (ia->ia6_lifetime.ia6t_vltime <
639 			    maxexpire - ia->ia6_updatetime) {
640 				retlt->ia6t_expire = ia->ia6_updatetime +
641 				    ia->ia6_lifetime.ia6t_vltime;
642 				retlt->ia6t_expire = retlt->ia6t_expire ?
643 				    time_mono_to_wall(retlt->ia6t_expire) :
644 				    0;
645 			} else
646 				retlt->ia6t_expire = maxexpire;
647 		}
648 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
649 			time_t maxexpire;
650 			struct in6_addrlifetime *retlt =
651 			    &ifr->ifr_ifru.ifru_lifetime;
652 
653 			/*
654 			 * XXX: adjust expiration time assuming time_t is
655 			 * signed.
656 			 */
657 			maxexpire = ((time_t)~0) &
658 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
659 			if (ia->ia6_lifetime.ia6t_pltime <
660 			    maxexpire - ia->ia6_updatetime) {
661 				retlt->ia6t_preferred = ia->ia6_updatetime +
662 				    ia->ia6_lifetime.ia6t_pltime;
663 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
664 				    time_mono_to_wall(retlt->ia6t_preferred) :
665 				    0;
666 			} else
667 				retlt->ia6t_preferred = maxexpire;
668 		}
669 #ifdef OSIOCFIFALIFETIME_IN6
670 		if (cmd == OSIOCFIFALIFETIME_IN6)
671 			in6_addrlifetime_to_in6_addrlifetime50(
672 			    &ifr->ifru.ifru_lifetime);
673 #endif
674 		break;
675 
676 #ifdef OSIOCAIFADDR_IN6
677 	case OSIOCAIFADDR_IN6:
678 		in6_aliasreq50_to_in6_aliasreq(ifra);
679 		/*FALLTHROUGH*/
680 #endif
681 	case SIOCAIFADDR_IN6:
682 	{
683 		struct in6_addrlifetime *lt;
684 
685 		/* reject read-only flags */
686 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
687 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
688 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
689 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
690 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
691 			error = EINVAL;
692 			break;
693 		}
694 		/*
695 		 * ia6t_expire and ia6t_preferred won't be used for now,
696 		 * so just in case.
697 		 */
698 		lt = &ifra->ifra_lifetime;
699 		if (lt->ia6t_expire != 0)
700 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
701 		if (lt->ia6t_preferred != 0)
702 			lt->ia6t_preferred =
703 			    time_wall_to_mono(lt->ia6t_preferred);
704 		/*
705 		 * make (ia == NULL) or update (ia != NULL) the interface
706 		 * address structure, and link it to the list.
707 		 */
708 		int s = splsoftnet();
709 		error = in6_update_ifa1(ifp, ifra, &ia, &psref, 0);
710 		splx(s);
711 		if (error)
712 			break;
713 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
714 		break;
715 	}
716 
717 	case SIOCDIFADDR_IN6:
718 		ia6_release(ia, &psref);
719 		ifaref(&ia->ia_ifa);
720 		in6_purgeaddr(&ia->ia_ifa);
721 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
722 		ifafree(&ia->ia_ifa);
723 		ia = NULL;
724 		break;
725 
726 	default:
727 		error = ENOTTY;
728 	}
729 release:
730 	ia6_release(ia, &psref);
731 out:
732 	curlwp_bindx(bound);
733 	return error;
734 }
735 
736 int
737 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
738 {
739 	int error, s;
740 
741 	switch (cmd) {
742 	case SIOCSNDFLUSH_IN6:
743 	case SIOCSPFXFLUSH_IN6:
744 	case SIOCSRTRFLUSH_IN6:
745 	case SIOCSDEFIFACE_IN6:
746 	case SIOCSIFINFO_FLAGS:
747 	case SIOCSIFINFO_IN6:
748 
749 	case SIOCALIFADDR:
750 	case SIOCDLIFADDR:
751 
752 	case SIOCDIFADDR_IN6:
753 #ifdef OSIOCAIFADDR_IN6
754 	case OSIOCAIFADDR_IN6:
755 #endif
756 	case SIOCAIFADDR_IN6:
757 
758 	case SIOCAADDRCTL_POLICY:
759 	case SIOCDADDRCTL_POLICY:
760 
761 		if (kauth_authorize_network(curlwp->l_cred,
762 		    KAUTH_NETWORK_SOCKET,
763 		    KAUTH_REQ_NETWORK_SOCKET_SETPRIV,
764 		    so, NULL, NULL))
765 			return EPERM;
766 		break;
767 	}
768 
769 	s = splsoftnet();
770 #ifndef NET_MPSAFE
771 	mutex_enter(softnet_lock);
772 #endif
773 	error = in6_control1(so , cmd, data, ifp);
774 #ifndef NET_MPSAFE
775 	mutex_exit(softnet_lock);
776 #endif
777 	splx(s);
778 	return error;
779 }
780 
781 static int
782 in6_get_llsol_addr(struct in6_addr *llsol, struct ifnet *ifp,
783     struct in6_addr *ip6)
784 {
785 	int error;
786 
787 	memset(llsol, 0, sizeof(struct in6_addr));
788 	llsol->s6_addr16[0] = htons(0xff02);
789 	llsol->s6_addr32[1] = 0;
790 	llsol->s6_addr32[2] = htonl(1);
791 	llsol->s6_addr32[3] = ip6->s6_addr32[3];
792 	llsol->s6_addr8[12] = 0xff;
793 
794 	error = in6_setscope(llsol, ifp, NULL);
795 	if (error != 0) {
796 		/* XXX: should not happen */
797 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
798 	}
799 
800 	return error;
801 }
802 
803 static int
804 in6_join_mcastgroups(struct in6_aliasreq *ifra, struct in6_ifaddr *ia,
805     struct ifnet *ifp, int flags)
806 {
807 	int error;
808 	struct sockaddr_in6 mltaddr, mltmask;
809 	struct in6_multi_mship *imm;
810 	struct in6_addr llsol;
811 	struct rtentry *rt;
812 	int dad_delay;
813 	char ip6buf[INET6_ADDRSTRLEN];
814 
815 	/* join solicited multicast addr for new host id */
816 	error = in6_get_llsol_addr(&llsol, ifp, &ifra->ifra_addr.sin6_addr);
817 	if (error != 0)
818 		goto out;
819 	dad_delay = 0;
820 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
821 		/*
822 		 * We need a random delay for DAD on the address
823 		 * being configured.  It also means delaying
824 		 * transmission of the corresponding MLD report to
825 		 * avoid report collision.
826 		 * [draft-ietf-ipv6-rfc2462bis-02.txt]
827 		 */
828 		dad_delay = cprng_fast32() % (MAX_RTR_SOLICITATION_DELAY * hz);
829 	}
830 
831 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
832 	/* join solicited multicast addr for new host id */
833 	imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
834 	if (!imm) {
835 		nd6log(LOG_ERR,
836 		    "addmulti failed for %s on %s (errno=%d)\n",
837 		    IN6_PRINT(ip6buf, &llsol), if_name(ifp), error);
838 		goto out;
839 	}
840 	mutex_enter(&in6_ifaddr_lock);
841 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
842 	mutex_exit(&in6_ifaddr_lock);
843 
844 	sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
845 
846 	/*
847 	 * join link-local all-nodes address
848 	 */
849 	sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
850 	    0, 0, 0);
851 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
852 		goto out; /* XXX: should not fail */
853 
854 	/*
855 	 * XXX: do we really need this automatic routes?
856 	 * We should probably reconsider this stuff.  Most applications
857 	 * actually do not need the routes, since they usually specify
858 	 * the outgoing interface.
859 	 */
860 	rt = rtalloc1(sin6tosa(&mltaddr), 0);
861 	if (rt) {
862 		if (memcmp(&mltaddr.sin6_addr,
863 		    &satocsin6(rt_getkey(rt))->sin6_addr,
864 		    MLTMASK_LEN)) {
865 			rt_unref(rt);
866 			rt = NULL;
867 		} else if (rt->rt_ifp != ifp) {
868 			IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
869 			    "network %04x:%04x::/32 = %04x:%04x::/32\n",
870 			    __func__, rt->rt_ifp, ifp, ifp->if_xname,
871 			    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
872 			    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
873 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
874 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
875 			rt_replace_ifa(rt, &ia->ia_ifa);
876 			rt->rt_ifp = ifp;
877 		}
878 	}
879 	if (!rt) {
880 		struct rt_addrinfo info;
881 
882 		memset(&info, 0, sizeof(info));
883 		info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
884 		info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
885 		info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
886 		info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
887 		/* XXX: we need RTF_CONNECTED to fake nd6_rtrequest */
888 		info.rti_flags = RTF_UP | RTF_CONNECTED;
889 		error = rtrequest1(RTM_ADD, &info, NULL);
890 		if (error)
891 			goto out;
892 	} else {
893 		rt_unref(rt);
894 	}
895 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
896 	if (!imm) {
897 		nd6log(LOG_WARNING,
898 		    "addmulti failed for %s on %s (errno=%d)\n",
899 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
900 		    if_name(ifp), error);
901 		goto out;
902 	}
903 	mutex_enter(&in6_ifaddr_lock);
904 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
905 	mutex_exit(&in6_ifaddr_lock);
906 
907 	/*
908 	 * join node information group address
909 	 */
910 	dad_delay = 0;
911 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
912 		/*
913 		 * The spec doesn't say anything about delay for this
914 		 * group, but the same logic should apply.
915 		 */
916 		dad_delay = cprng_fast32() % (MAX_RTR_SOLICITATION_DELAY * hz);
917 	}
918 	if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
919 		;
920 	else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
921 		  dad_delay)) == NULL) { /* XXX jinmei */
922 		nd6log(LOG_WARNING,
923 		    "addmulti failed for %s on %s (errno=%d)\n",
924 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
925 		    if_name(ifp), error);
926 		/* XXX not very fatal, go on... */
927 	} else {
928 		mutex_enter(&in6_ifaddr_lock);
929 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
930 		mutex_exit(&in6_ifaddr_lock);
931 	}
932 
933 
934 	/*
935 	 * join interface-local all-nodes address.
936 	 * (ff01::1%ifN, and ff01::%ifN/32)
937 	 */
938 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
939 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
940 		goto out; /* XXX: should not fail */
941 
942 	/* XXX: again, do we really need the route? */
943 	rt = rtalloc1(sin6tosa(&mltaddr), 0);
944 	if (rt) {
945 		/* 32bit came from "mltmask" */
946 		if (memcmp(&mltaddr.sin6_addr,
947 		    &satocsin6(rt_getkey(rt))->sin6_addr,
948 		    32 / NBBY)) {
949 			rt_unref(rt);
950 			rt = NULL;
951 		} else if (rt->rt_ifp != ifp) {
952 			IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
953 			    "network %04x:%04x::/32 = %04x:%04x::/32\n",
954 			    __func__, rt->rt_ifp, ifp, ifp->if_xname,
955 			    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
956 			    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
957 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
958 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
959 			rt_replace_ifa(rt, &ia->ia_ifa);
960 			rt->rt_ifp = ifp;
961 		}
962 	}
963 	if (!rt) {
964 		struct rt_addrinfo info;
965 
966 		memset(&info, 0, sizeof(info));
967 		info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
968 		info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
969 		info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
970 		info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
971 		info.rti_flags = RTF_UP | RTF_CONNECTED;
972 		error = rtrequest1(RTM_ADD, &info, NULL);
973 		if (error)
974 			goto out;
975 #undef	MLTMASK_LEN
976 	} else {
977 		rt_unref(rt);
978 	}
979 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
980 	if (!imm) {
981 		nd6log(LOG_WARNING,
982 		    "addmulti failed for %s on %s (errno=%d)\n",
983 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
984 		    if_name(ifp), error);
985 		goto out;
986 	} else {
987 		mutex_enter(&in6_ifaddr_lock);
988 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
989 		mutex_exit(&in6_ifaddr_lock);
990 	}
991 	return 0;
992 
993 out:
994 	KASSERT(error != 0);
995 	return error;
996 }
997 
998 /*
999  * Update parameters of an IPv6 interface address.
1000  * If necessary, a new entry is created and linked into address chains.
1001  * This function is separated from in6_control().
1002  * XXX: should this be performed under splsoftnet()?
1003  */
1004 static int
1005 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
1006     struct in6_ifaddr **iap, struct psref *psref, int flags)
1007 {
1008 	int error = 0, hostIsNew = 0, plen = -1;
1009 	struct sockaddr_in6 dst6;
1010 	struct in6_addrlifetime *lt;
1011 	int dad_delay, was_tentative;
1012 	struct in6_ifaddr *ia = iap ? *iap : NULL;
1013 	char ip6buf[INET6_ADDRSTRLEN];
1014 
1015 	KASSERT((iap == NULL && psref == NULL) ||
1016 	    (iap != NULL && psref != NULL));
1017 
1018 	/* Validate parameters */
1019 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1020 		return EINVAL;
1021 
1022 	/*
1023 	 * The destination address for a p2p link must have a family
1024 	 * of AF_UNSPEC or AF_INET6.
1025 	 */
1026 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1027 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1028 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1029 		return EAFNOSUPPORT;
1030 	/*
1031 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1032 	 * does not carry fields other than sin6_len.
1033 	 */
1034 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1035 		return EINVAL;
1036 	/*
1037 	 * Because the IPv6 address architecture is classless, we require
1038 	 * users to specify a (non 0) prefix length (mask) for a new address.
1039 	 * We also require the prefix (when specified) mask is valid, and thus
1040 	 * reject a non-consecutive mask.
1041 	 */
1042 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1043 		return EINVAL;
1044 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1045 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1046 		    (u_char *)&ifra->ifra_prefixmask +
1047 		    ifra->ifra_prefixmask.sin6_len);
1048 		if (plen <= 0)
1049 			return EINVAL;
1050 	} else {
1051 		/*
1052 		 * In this case, ia must not be NULL.  We just use its prefix
1053 		 * length.
1054 		 */
1055 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1056 	}
1057 	/*
1058 	 * If the destination address on a p2p interface is specified,
1059 	 * and the address is a scoped one, validate/set the scope
1060 	 * zone identifier.
1061 	 */
1062 	dst6 = ifra->ifra_dstaddr;
1063 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1064 	    (dst6.sin6_family == AF_INET6)) {
1065 		struct in6_addr in6_tmp;
1066 		u_int32_t zoneid;
1067 
1068 		in6_tmp = dst6.sin6_addr;
1069 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1070 			return EINVAL; /* XXX: should be impossible */
1071 
1072 		if (dst6.sin6_scope_id != 0) {
1073 			if (dst6.sin6_scope_id != zoneid)
1074 				return EINVAL;
1075 		} else		/* user omit to specify the ID. */
1076 			dst6.sin6_scope_id = zoneid;
1077 
1078 		/* convert into the internal form */
1079 		if (sa6_embedscope(&dst6, 0))
1080 			return EINVAL; /* XXX: should be impossible */
1081 	}
1082 	/*
1083 	 * The destination address can be specified only for a p2p or a
1084 	 * loopback interface.  If specified, the corresponding prefix length
1085 	 * must be 128.
1086 	 */
1087 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1088 #ifdef FORCE_P2PPLEN
1089 		int i;
1090 #endif
1091 
1092 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1093 			/* XXX: noisy message */
1094 			nd6log(LOG_INFO, "a destination can "
1095 			    "be specified for a p2p or a loopback IF only\n");
1096 			return EINVAL;
1097 		}
1098 		if (plen != 128) {
1099 			nd6log(LOG_INFO, "prefixlen should "
1100 			    "be 128 when dstaddr is specified\n");
1101 #ifdef FORCE_P2PPLEN
1102 			/*
1103 			 * To be compatible with old configurations,
1104 			 * such as ifconfig gif0 inet6 2001::1 2001::2
1105 			 * prefixlen 126, we override the specified
1106 			 * prefixmask as if the prefix length was 128.
1107 			 */
1108 			ifra->ifra_prefixmask.sin6_len =
1109 			    sizeof(struct sockaddr_in6);
1110 			for (i = 0; i < 4; i++)
1111 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
1112 				    0xffffffff;
1113 			plen = 128;
1114 #else
1115 			return EINVAL;
1116 #endif
1117 		}
1118 	}
1119 	/* lifetime consistency check */
1120 	lt = &ifra->ifra_lifetime;
1121 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1122 		return EINVAL;
1123 	if (lt->ia6t_vltime == 0) {
1124 		/*
1125 		 * the following log might be noisy, but this is a typical
1126 		 * configuration mistake or a tool's bug.
1127 		 */
1128 		nd6log(LOG_INFO, "valid lifetime is 0 for %s\n",
1129 		    IN6_PRINT(ip6buf, &ifra->ifra_addr.sin6_addr));
1130 
1131 		if (ia == NULL)
1132 			return 0; /* there's nothing to do */
1133 	}
1134 
1135 	/*
1136 	 * If this is a new address, allocate a new ifaddr and link it
1137 	 * into chains.
1138 	 */
1139 	if (ia == NULL) {
1140 		hostIsNew = 1;
1141 		/*
1142 		 * When in6_update_ifa() is called in a process of a received
1143 		 * RA, it is called under an interrupt context.  So, we should
1144 		 * call malloc with M_NOWAIT.
1145 		 */
1146 		ia = malloc(sizeof(*ia), M_IFADDR, M_NOWAIT|M_ZERO);
1147 		if (ia == NULL)
1148 			return ENOBUFS;
1149 		LIST_INIT(&ia->ia6_memberships);
1150 		/* Initialize the address and masks, and put time stamp */
1151 		ia->ia_ifa.ifa_addr = sin6tosa(&ia->ia_addr);
1152 		ia->ia_addr.sin6_family = AF_INET6;
1153 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1154 		ia->ia6_createtime = time_uptime;
1155 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1156 			/*
1157 			 * XXX: some functions expect that ifa_dstaddr is not
1158 			 * NULL for p2p interfaces.
1159 			 */
1160 			ia->ia_ifa.ifa_dstaddr = sin6tosa(&ia->ia_dstaddr);
1161 		} else {
1162 			ia->ia_ifa.ifa_dstaddr = NULL;
1163 		}
1164 		ia->ia_ifa.ifa_netmask = sin6tosa(&ia->ia_prefixmask);
1165 
1166 		ia->ia_ifp = ifp;
1167 		IN6_ADDRLIST_ENTRY_INIT(ia);
1168 		ifa_psref_init(&ia->ia_ifa);
1169 	}
1170 
1171 	/* update timestamp */
1172 	ia->ia6_updatetime = time_uptime;
1173 
1174 	/* set prefix mask */
1175 	if (ifra->ifra_prefixmask.sin6_len) {
1176 		if (ia->ia_prefixmask.sin6_len) {
1177 			/*
1178 			 * We prohibit changing the prefix length of an
1179 			 * existing autoconf address, because the operation
1180 			 * would confuse prefix management.
1181 			 */
1182 			if (ia->ia6_ndpr != NULL &&
1183 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) !=
1184 			    plen)
1185 			{
1186 				nd6log(LOG_INFO, "the prefix length of an"
1187 				    " existing (%s) autoconf address should"
1188 				    " not be changed\n",
1189 				    IN6_PRINT(ip6buf,
1190 				    &ia->ia_addr.sin6_addr));
1191 				error = EINVAL;
1192 				if (hostIsNew)
1193 					free(ia, M_IFADDR);
1194 				goto exit;
1195 			}
1196 
1197 			if (!IN6_ARE_ADDR_EQUAL(&ia->ia_prefixmask.sin6_addr,
1198 			    &ifra->ifra_prefixmask.sin6_addr))
1199 				in6_ifremprefix(ia);
1200 		}
1201 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1202 	}
1203 
1204 	/* Set destination address. */
1205 	if (dst6.sin6_family == AF_INET6) {
1206 		if (!IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
1207 		    &ia->ia_dstaddr.sin6_addr))
1208 			in6_ifremprefix(ia);
1209 		ia->ia_dstaddr = dst6;
1210 	}
1211 
1212 	/*
1213 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1214 	 * to see if the address is deprecated or invalidated, but initialize
1215 	 * these members for applications.
1216 	 */
1217 	ia->ia6_lifetime = ifra->ifra_lifetime;
1218 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1219 		ia->ia6_lifetime.ia6t_expire =
1220 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1221 	} else
1222 		ia->ia6_lifetime.ia6t_expire = 0;
1223 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1224 		ia->ia6_lifetime.ia6t_preferred =
1225 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1226 	} else
1227 		ia->ia6_lifetime.ia6t_preferred = 0;
1228 
1229 	/*
1230 	 * configure address flags.
1231 	 * We need to preserve tentative state so DAD works if
1232 	 * something adds the same address before DAD finishes.
1233 	 */
1234 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1235 	ia->ia6_flags = ifra->ifra_flags;
1236 
1237 	/*
1238 	 * Make the address tentative before joining multicast addresses,
1239 	 * so that corresponding MLD responses would not have a tentative
1240 	 * source address.
1241 	 */
1242 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1243 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1244 		ia->ia6_flags |= IN6_IFF_DETACHED;
1245 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1246 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp))
1247 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1248 
1249 	/*
1250 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1251 	 * userland, make it deprecated.
1252 	 */
1253 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1254 		ia->ia6_lifetime.ia6t_pltime = 0;
1255 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1256 	}
1257 
1258 	if (hostIsNew) {
1259 		/*
1260 		 * We need a reference to ia before calling in6_ifinit.
1261 		 * Otherwise ia can be freed in in6_ifinit accidentally.
1262 		 */
1263 		ifaref(&ia->ia_ifa);
1264 	}
1265 	/* reset the interface and routing table appropriately. */
1266 	error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew);
1267 	if (error != 0) {
1268 		if (hostIsNew)
1269 			free(ia, M_IFADDR);
1270 		goto exit;
1271 	}
1272 
1273 	/*
1274 	 * We are done if we have simply modified an existing address.
1275 	 */
1276 	if (!hostIsNew)
1277 		return error;
1278 
1279 	/*
1280 	 * Insert ia to the global list and ifa to the interface's list.
1281 	 * A reference to it is already gained above.
1282 	 */
1283 	mutex_enter(&in6_ifaddr_lock);
1284 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1285 	mutex_exit(&in6_ifaddr_lock);
1286 
1287 	ifa_insert(ifp, &ia->ia_ifa);
1288 
1289 	/*
1290 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1291 	 * not just go to unlink.
1292 	 */
1293 
1294 	/* join necessary multicast groups */
1295 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1296 		error = in6_join_mcastgroups(ifra, ia, ifp, flags);
1297 		if (error != 0)
1298 			goto cleanup;
1299 	}
1300 
1301 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1302 	error = nd6_add_ifa_lle(ia);
1303 	if (error != 0)
1304 		goto cleanup;
1305 
1306 	/*
1307 	 * Perform DAD, if needed.
1308 	 * XXX It may be of use, if we can administratively
1309 	 * disable DAD.
1310 	 */
1311 	if (hostIsNew && if_do_dad(ifp) &&
1312 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1313 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1314 	{
1315 		int mindelay, maxdelay;
1316 
1317 		dad_delay = 0;
1318 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1319 			struct in6_addr llsol;
1320 			struct in6_multi *in6m_sol = NULL;
1321 			/*
1322 			 * We need to impose a delay before sending an NS
1323 			 * for DAD.  Check if we also needed a delay for the
1324 			 * corresponding MLD message.  If we did, the delay
1325 			 * should be larger than the MLD delay (this could be
1326 			 * relaxed a bit, but this simple logic is at least
1327 			 * safe).
1328 			 */
1329 			mindelay = 0;
1330 			error = in6_get_llsol_addr(&llsol, ifp,
1331 			    &ifra->ifra_addr.sin6_addr);
1332 			in6_multi_lock(RW_READER);
1333 			if (error == 0)
1334 				in6m_sol = in6_lookup_multi(&llsol, ifp);
1335 			if (in6m_sol != NULL &&
1336 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1337 				mindelay = in6m_sol->in6m_timer;
1338 			}
1339 			in6_multi_unlock();
1340 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1341 			if (maxdelay - mindelay == 0)
1342 				dad_delay = 0;
1343 			else {
1344 				dad_delay =
1345 				    (cprng_fast32() % (maxdelay - mindelay)) +
1346 				    mindelay;
1347 			}
1348 		}
1349 		/* +1 ensures callout is always used */
1350 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1351 	}
1352 
1353 	if (iap != NULL) {
1354 		*iap = ia;
1355 		if (hostIsNew)
1356 			ia6_acquire(ia, psref);
1357 	}
1358 
1359 	return 0;
1360 
1361   cleanup:
1362 	in6_purgeaddr(&ia->ia_ifa);
1363   exit:
1364 	return error;
1365 }
1366 
1367 int
1368 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1369 {
1370 	int rc, s;
1371 
1372 	s = splsoftnet();
1373 	rc = in6_update_ifa1(ifp, ifra, NULL, NULL, flags);
1374 	splx(s);
1375 	return rc;
1376 }
1377 
1378 void
1379 in6_purgeaddr(struct ifaddr *ifa)
1380 {
1381 	struct ifnet *ifp = ifa->ifa_ifp;
1382 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1383 	struct in6_multi_mship *imm;
1384 
1385 	KASSERT(!ifa_held(ifa));
1386 
1387 	ifa->ifa_flags |= IFA_DESTROYING;
1388 
1389 	/* stop DAD processing */
1390 	nd6_dad_stop(ifa);
1391 
1392 	/* Delete any network route. */
1393 	in6_ifremprefix(ia);
1394 
1395 	/* Remove ownaddr's loopback rtentry, if it exists. */
1396 	in6_ifremlocal(&(ia->ia_ifa));
1397 
1398 	/*
1399 	 * leave from multicast groups we have joined for the interface
1400 	 */
1401 	mutex_enter(&in6_ifaddr_lock);
1402 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1403 		LIST_REMOVE(imm, i6mm_chain);
1404 		mutex_exit(&in6_ifaddr_lock);
1405 		in6_leavegroup(imm);
1406 		mutex_enter(&in6_ifaddr_lock);
1407 	}
1408 	mutex_exit(&in6_ifaddr_lock);
1409 
1410 	in6_unlink_ifa(ia, ifp);
1411 }
1412 
1413 static void
1414 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1415 {
1416 	int	s = splsoftnet();
1417 
1418 	mutex_enter(&in6_ifaddr_lock);
1419 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1420 	ifa_remove(ifp, &ia->ia_ifa);
1421 	mutex_exit(&in6_ifaddr_lock);
1422 
1423 	/*
1424 	 * Release the reference to the ND prefix.
1425 	 */
1426 	if (ia->ia6_ndpr != NULL) {
1427 		nd6_prefix_unref(ia->ia6_ndpr);
1428 		ia->ia6_ndpr = NULL;
1429 	}
1430 
1431 	/*
1432 	 * Also, if the address being removed is autoconf'ed, call
1433 	 * nd6_pfxlist_onlink_check() since the release might affect the status of
1434 	 * other (detached) addresses.
1435 	 */
1436 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0) {
1437 		ND6_WLOCK();
1438 		nd6_pfxlist_onlink_check();
1439 		ND6_UNLOCK();
1440 	}
1441 
1442 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1443 
1444 	/*
1445 	 * release another refcnt for the link from in6_ifaddr.
1446 	 * Note that we should decrement the refcnt at least once for all *BSD.
1447 	 */
1448 	ifafree(&ia->ia_ifa);
1449 
1450 	splx(s);
1451 }
1452 
1453 void
1454 in6_purgeif(struct ifnet *ifp)
1455 {
1456 
1457 	in6_ifdetach(ifp);
1458 }
1459 
1460 void
1461 in6_purge_mcast_references(struct in6_multi *in6m)
1462 {
1463 	struct	in6_ifaddr *ia;
1464 
1465 	KASSERT(in6_multi_locked(RW_WRITER));
1466 
1467 	mutex_enter(&in6_ifaddr_lock);
1468 	IN6_ADDRLIST_WRITER_FOREACH(ia) {
1469 		struct in6_multi_mship *imm;
1470 		LIST_FOREACH(imm, &ia->ia6_memberships, i6mm_chain) {
1471 			if (imm->i6mm_maddr == in6m)
1472 				imm->i6mm_maddr = NULL;
1473 		}
1474 	}
1475 	mutex_exit(&in6_ifaddr_lock);
1476 }
1477 
1478 /*
1479  * SIOC[GAD]LIFADDR.
1480  *	SIOCGLIFADDR: get first address. (?)
1481  *	SIOCGLIFADDR with IFLR_PREFIX:
1482  *		get first address that matches the specified prefix.
1483  *	SIOCALIFADDR: add the specified address.
1484  *	SIOCALIFADDR with IFLR_PREFIX:
1485  *		add the specified prefix, filling hostid part from
1486  *		the first link-local address.  prefixlen must be <= 64.
1487  *	SIOCDLIFADDR: delete the specified address.
1488  *	SIOCDLIFADDR with IFLR_PREFIX:
1489  *		delete the first address that matches the specified prefix.
1490  * return values:
1491  *	EINVAL on invalid parameters
1492  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1493  *	other values may be returned from in6_ioctl()
1494  *
1495  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1496  * this is to accommodate address naming scheme other than RFC2374,
1497  * in the future.
1498  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1499  * address encoding scheme. (see figure on page 8)
1500  */
1501 static int
1502 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1503 	struct ifnet *ifp)
1504 {
1505 	struct in6_ifaddr *ia = NULL; /* XXX gcc 4.8 maybe-uninitialized */
1506 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1507 	struct ifaddr *ifa;
1508 	struct sockaddr *sa;
1509 
1510 	/* sanity checks */
1511 	if (!data || !ifp) {
1512 		panic("invalid argument to in6_lifaddr_ioctl");
1513 		/* NOTREACHED */
1514 	}
1515 
1516 	switch (cmd) {
1517 	case SIOCGLIFADDR:
1518 		/* address must be specified on GET with IFLR_PREFIX */
1519 		if ((iflr->flags & IFLR_PREFIX) == 0)
1520 			break;
1521 		/* FALLTHROUGH */
1522 	case SIOCALIFADDR:
1523 	case SIOCDLIFADDR:
1524 		/* address must be specified on ADD and DELETE */
1525 		sa = (struct sockaddr *)&iflr->addr;
1526 		if (sa->sa_family != AF_INET6)
1527 			return EINVAL;
1528 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1529 			return EINVAL;
1530 		/* XXX need improvement */
1531 		sa = (struct sockaddr *)&iflr->dstaddr;
1532 		if (sa->sa_family && sa->sa_family != AF_INET6)
1533 			return EINVAL;
1534 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1535 			return EINVAL;
1536 		break;
1537 	default: /* shouldn't happen */
1538 #if 0
1539 		panic("invalid cmd to in6_lifaddr_ioctl");
1540 		/* NOTREACHED */
1541 #else
1542 		return EOPNOTSUPP;
1543 #endif
1544 	}
1545 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1546 		return EINVAL;
1547 
1548 	switch (cmd) {
1549 	case SIOCALIFADDR:
1550 	    {
1551 		struct in6_aliasreq ifra;
1552 		struct in6_addr *xhostid = NULL;
1553 		int prefixlen;
1554 		int bound = curlwp_bind();
1555 		struct psref psref;
1556 
1557 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1558 			struct sockaddr_in6 *sin6;
1559 
1560 			/*
1561 			 * xhostid is to fill in the hostid part of the
1562 			 * address.  xhostid points to the first link-local
1563 			 * address attached to the interface.
1564 			 */
1565 			ia = in6ifa_ifpforlinklocal_psref(ifp, 0, &psref);
1566 			if (ia == NULL) {
1567 				curlwp_bindx(bound);
1568 				return EADDRNOTAVAIL;
1569 			}
1570 			xhostid = IFA_IN6(&ia->ia_ifa);
1571 
1572 		 	/* prefixlen must be <= 64. */
1573 			if (64 < iflr->prefixlen) {
1574 				ia6_release(ia, &psref);
1575 				curlwp_bindx(bound);
1576 				return EINVAL;
1577 			}
1578 			prefixlen = iflr->prefixlen;
1579 
1580 			/* hostid part must be zero. */
1581 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1582 			if (sin6->sin6_addr.s6_addr32[2] != 0
1583 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1584 				ia6_release(ia, &psref);
1585 				curlwp_bindx(bound);
1586 				return EINVAL;
1587 			}
1588 		} else
1589 			prefixlen = iflr->prefixlen;
1590 
1591 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1592 		memset(&ifra, 0, sizeof(ifra));
1593 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1594 
1595 		memcpy(&ifra.ifra_addr, &iflr->addr,
1596 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1597 		if (xhostid) {
1598 			/* fill in hostid part */
1599 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1600 			    xhostid->s6_addr32[2];
1601 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1602 			    xhostid->s6_addr32[3];
1603 		}
1604 
1605 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1606 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1607 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1608 			if (xhostid) {
1609 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1610 				    xhostid->s6_addr32[2];
1611 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1612 				    xhostid->s6_addr32[3];
1613 			}
1614 		}
1615 		if (xhostid) {
1616 			ia6_release(ia, &psref);
1617 			ia = NULL;
1618 		}
1619 		curlwp_bindx(bound);
1620 
1621 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1622 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1623 
1624 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1625 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1626 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1627 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1628 	    }
1629 	case SIOCGLIFADDR:
1630 	case SIOCDLIFADDR:
1631 	    {
1632 		struct in6_addr mask, candidate, match;
1633 		struct sockaddr_in6 *sin6;
1634 		int cmp;
1635 		int error, s;
1636 
1637 		memset(&mask, 0, sizeof(mask));
1638 		if (iflr->flags & IFLR_PREFIX) {
1639 			/* lookup a prefix rather than address. */
1640 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1641 
1642 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1643 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1644 			match.s6_addr32[0] &= mask.s6_addr32[0];
1645 			match.s6_addr32[1] &= mask.s6_addr32[1];
1646 			match.s6_addr32[2] &= mask.s6_addr32[2];
1647 			match.s6_addr32[3] &= mask.s6_addr32[3];
1648 
1649 			/* if you set extra bits, that's wrong */
1650 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1651 				return EINVAL;
1652 
1653 			cmp = 1;
1654 		} else {
1655 			if (cmd == SIOCGLIFADDR) {
1656 				/* on getting an address, take the 1st match */
1657 				cmp = 0;	/* XXX */
1658 			} else {
1659 				/* on deleting an address, do exact match */
1660 				in6_prefixlen2mask(&mask, 128);
1661 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1662 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1663 
1664 				cmp = 1;
1665 			}
1666 		}
1667 
1668 		s = pserialize_read_enter();
1669 		IFADDR_READER_FOREACH(ifa, ifp) {
1670 			if (ifa->ifa_addr->sa_family != AF_INET6)
1671 				continue;
1672 			if (!cmp)
1673 				break;
1674 
1675 			/*
1676 			 * XXX: this is adhoc, but is necessary to allow
1677 			 * a user to specify fe80::/64 (not /10) for a
1678 			 * link-local address.
1679 			 */
1680 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1681 			in6_clearscope(&candidate);
1682 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1683 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1684 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1685 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1686 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1687 				break;
1688 		}
1689 		if (!ifa) {
1690 			error = EADDRNOTAVAIL;
1691 			goto error;
1692 		}
1693 		ia = ifa2ia6(ifa);
1694 
1695 		if (cmd == SIOCGLIFADDR) {
1696 			/* fill in the if_laddrreq structure */
1697 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1698 			error = sa6_recoverscope(
1699 			    (struct sockaddr_in6 *)&iflr->addr);
1700 			if (error != 0)
1701 				goto error;
1702 
1703 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1704 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1705 				    ia->ia_dstaddr.sin6_len);
1706 				error = sa6_recoverscope(
1707 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1708 				if (error != 0)
1709 					goto error;
1710 			} else
1711 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1712 
1713 			iflr->prefixlen =
1714 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1715 
1716 			iflr->flags = ia->ia6_flags;	/* XXX */
1717 
1718 			error = 0;
1719 		} else {
1720 			struct in6_aliasreq ifra;
1721 
1722 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1723 			memset(&ifra, 0, sizeof(ifra));
1724 			memcpy(ifra.ifra_name, iflr->iflr_name,
1725 			    sizeof(ifra.ifra_name));
1726 
1727 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1728 			    ia->ia_addr.sin6_len);
1729 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1730 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1731 				    ia->ia_dstaddr.sin6_len);
1732 			} else {
1733 				memset(&ifra.ifra_dstaddr, 0,
1734 				    sizeof(ifra.ifra_dstaddr));
1735 			}
1736 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1737 			    ia->ia_prefixmask.sin6_len);
1738 
1739 			ifra.ifra_flags = ia->ia6_flags;
1740 			pserialize_read_exit(s);
1741 
1742 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1743 		}
1744 	error:
1745 		pserialize_read_exit(s);
1746 		return error;
1747 	    }
1748 	}
1749 
1750 	return EOPNOTSUPP;	/* just for safety */
1751 }
1752 
1753 /*
1754  * Initialize an interface's internet6 address
1755  * and routing table entry.
1756  */
1757 static int
1758 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1759 	const struct sockaddr_in6 *sin6, int newhost)
1760 {
1761 	int	error = 0, ifacount = 0;
1762 	int	s = splsoftnet();
1763 	struct ifaddr *ifa;
1764 
1765 	/*
1766 	 * Give the interface a chance to initialize
1767 	 * if this is its first address,
1768 	 * and to validate the address if necessary.
1769 	 */
1770 	IFADDR_READER_FOREACH(ifa, ifp) {
1771 		if (ifa->ifa_addr->sa_family != AF_INET6)
1772 			continue;
1773 		ifacount++;
1774 	}
1775 
1776 	ia->ia_addr = *sin6;
1777 
1778 	if (ifacount <= 0 &&
1779 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1780 		splx(s);
1781 		return error;
1782 	}
1783 	splx(s);
1784 
1785 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1786 
1787 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1788 
1789 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1790 	if (newhost) {
1791 		/* set the rtrequest function to create llinfo */
1792 		if (ifp->if_flags & IFF_POINTOPOINT)
1793 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1794 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1795 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1796 		in6_ifaddlocal(&ia->ia_ifa);
1797 	} else {
1798 		/* Inform the routing socket of new flags/timings */
1799 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
1800 	}
1801 
1802 	/* Add the network prefix route. */
1803 	if ((error = in6_ifaddprefix(ia)) != 0) {
1804 		if (newhost)
1805 			in6_ifremlocal(&ia->ia_ifa);
1806 		return error;
1807 	}
1808 
1809 	return error;
1810 }
1811 
1812 static struct ifaddr *
1813 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1814 {
1815 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1816 		return ifa;
1817 	return best_ifa;
1818 }
1819 
1820 /*
1821  * Find an IPv6 interface link-local address specific to an interface.
1822  */
1823 struct in6_ifaddr *
1824 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1825 {
1826 	struct ifaddr *best_ifa = NULL, *ifa;
1827 
1828 	IFADDR_READER_FOREACH(ifa, ifp) {
1829 		if (ifa->ifa_addr->sa_family != AF_INET6)
1830 			continue;
1831 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1832 			continue;
1833 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1834 			continue;
1835 		best_ifa = bestifa(best_ifa, ifa);
1836 	}
1837 
1838 	return (struct in6_ifaddr *)best_ifa;
1839 }
1840 
1841 struct in6_ifaddr *
1842 in6ifa_ifpforlinklocal_psref(const struct ifnet *ifp, const int ignoreflags,
1843     struct psref *psref)
1844 {
1845 	struct in6_ifaddr *ia;
1846 	int s = pserialize_read_enter();
1847 
1848 	ia = in6ifa_ifpforlinklocal(ifp, ignoreflags);
1849 	if (ia != NULL)
1850 		ia6_acquire(ia, psref);
1851 	pserialize_read_exit(s);
1852 
1853 	return ia;
1854 }
1855 
1856 /*
1857  * find the internet address corresponding to a given address.
1858  * ifaddr is returned referenced.
1859  */
1860 struct in6_ifaddr *
1861 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1862 {
1863 	struct in6_ifaddr *ia;
1864 	int s;
1865 
1866 	s = pserialize_read_enter();
1867 	IN6_ADDRLIST_READER_FOREACH(ia) {
1868 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1869 			if (zoneid != 0 &&
1870 			    zoneid != ia->ia_addr.sin6_scope_id)
1871 				continue;
1872 			ifaref(&ia->ia_ifa);
1873 			break;
1874 		}
1875 	}
1876 	pserialize_read_exit(s);
1877 
1878 	return ia;
1879 }
1880 
1881 /*
1882  * find the internet address corresponding to a given interface and address.
1883  */
1884 struct in6_ifaddr *
1885 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1886 {
1887 	struct ifaddr *best_ifa = NULL, *ifa;
1888 
1889 	IFADDR_READER_FOREACH(ifa, ifp) {
1890 		if (ifa->ifa_addr->sa_family != AF_INET6)
1891 			continue;
1892 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1893 			continue;
1894 		best_ifa = bestifa(best_ifa, ifa);
1895 	}
1896 
1897 	return (struct in6_ifaddr *)best_ifa;
1898 }
1899 
1900 struct in6_ifaddr *
1901 in6ifa_ifpwithaddr_psref(const struct ifnet *ifp, const struct in6_addr *addr,
1902     struct psref *psref)
1903 {
1904 	struct in6_ifaddr *ia;
1905 	int s = pserialize_read_enter();
1906 
1907 	ia = in6ifa_ifpwithaddr(ifp, addr);
1908 	if (ia != NULL)
1909 		ia6_acquire(ia, psref);
1910 	pserialize_read_exit(s);
1911 
1912 	return ia;
1913 }
1914 
1915 static struct in6_ifaddr *
1916 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1917 {
1918 	if (best_ia == NULL ||
1919 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1920 		return ia;
1921 	return best_ia;
1922 }
1923 
1924 /*
1925  * Determine if an address is on a local network.
1926  */
1927 int
1928 in6_localaddr(const struct in6_addr *in6)
1929 {
1930 	struct in6_ifaddr *ia;
1931 	int s;
1932 
1933 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1934 		return 1;
1935 
1936 	s = pserialize_read_enter();
1937 	IN6_ADDRLIST_READER_FOREACH(ia) {
1938 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1939 					      &ia->ia_prefixmask.sin6_addr)) {
1940 			pserialize_read_exit(s);
1941 			return 1;
1942 		}
1943 	}
1944 	pserialize_read_exit(s);
1945 
1946 	return 0;
1947 }
1948 
1949 int
1950 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1951 {
1952 	struct in6_ifaddr *ia;
1953 	int s;
1954 
1955 	s = pserialize_read_enter();
1956 	IN6_ADDRLIST_READER_FOREACH(ia) {
1957 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1958 		    &sa6->sin6_addr) &&
1959 #ifdef SCOPEDROUTING
1960 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1961 #endif
1962 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
1963 			pserialize_read_exit(s);
1964 			return 1; /* true */
1965 		}
1966 
1967 		/* XXX: do we still have to go thru the rest of the list? */
1968 	}
1969 	pserialize_read_exit(s);
1970 
1971 	return 0;		/* false */
1972 }
1973 
1974 /*
1975  * return length of part which dst and src are equal
1976  * hard coding...
1977  */
1978 int
1979 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1980 {
1981 	int match = 0;
1982 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1983 	u_char *lim = s + 16, r;
1984 
1985 	while (s < lim)
1986 		if ((r = (*d++ ^ *s++)) != 0) {
1987 			while (r < 128) {
1988 				match++;
1989 				r <<= 1;
1990 			}
1991 			break;
1992 		} else
1993 			match += NBBY;
1994 	return match;
1995 }
1996 
1997 /* XXX: to be scope conscious */
1998 int
1999 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2000 {
2001 	int bytelen, bitlen;
2002 
2003 	/* sanity check */
2004 	if (len < 0 || len > 128) {
2005 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2006 		    len);
2007 		return 0;
2008 	}
2009 
2010 	bytelen = len / NBBY;
2011 	bitlen = len % NBBY;
2012 
2013 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2014 		return 0;
2015 	if (bitlen != 0 &&
2016 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
2017 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
2018 		return 0;
2019 
2020 	return 1;
2021 }
2022 
2023 void
2024 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2025 {
2026 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2027 	int bytelen, bitlen, i;
2028 
2029 	/* sanity check */
2030 	if (len < 0 || len > 128) {
2031 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2032 		    len);
2033 		return;
2034 	}
2035 
2036 	memset(maskp, 0, sizeof(*maskp));
2037 	bytelen = len / NBBY;
2038 	bitlen = len % NBBY;
2039 	for (i = 0; i < bytelen; i++)
2040 		maskp->s6_addr[i] = 0xff;
2041 	if (bitlen)
2042 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2043 }
2044 
2045 /*
2046  * return the best address out of the same scope. if no address was
2047  * found, return the first valid address from designated IF.
2048  */
2049 struct in6_ifaddr *
2050 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2051 {
2052 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2053 	struct ifaddr *ifa;
2054 	struct in6_ifaddr *best_ia = NULL, *ia;
2055 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2056 
2057 	dep[0] = dep[1] = NULL;
2058 
2059 	/*
2060 	 * We first look for addresses in the same scope.
2061 	 * If there is one, return it.
2062 	 * If two or more, return one which matches the dst longest.
2063 	 * If none, return one of global addresses assigned other ifs.
2064 	 */
2065 	IFADDR_READER_FOREACH(ifa, ifp) {
2066 		if (ifa->ifa_addr->sa_family != AF_INET6)
2067 			continue;
2068 		ia = (struct in6_ifaddr *)ifa;
2069 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2070 			continue; /* XXX: is there any case to allow anycast? */
2071 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2072 			continue; /* don't use this interface */
2073 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2074 			continue;
2075 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2076 			if (ip6_use_deprecated)
2077 				dep[0] = ia;
2078 			continue;
2079 		}
2080 
2081 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
2082 			continue;
2083 		/*
2084 		 * call in6_matchlen() as few as possible
2085 		 */
2086 		if (best_ia == NULL) {
2087 			best_ia = ia;
2088 			continue;
2089 		}
2090 		if (blen == -1)
2091 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2092 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2093 		if (tlen > blen) {
2094 			blen = tlen;
2095 			best_ia = ia;
2096 		} else if (tlen == blen)
2097 			best_ia = bestia(best_ia, ia);
2098 	}
2099 	if (best_ia != NULL)
2100 		return best_ia;
2101 
2102 	IFADDR_READER_FOREACH(ifa, ifp) {
2103 		if (ifa->ifa_addr->sa_family != AF_INET6)
2104 			continue;
2105 		ia = (struct in6_ifaddr *)ifa;
2106 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2107 			continue; /* XXX: is there any case to allow anycast? */
2108 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2109 			continue; /* don't use this interface */
2110 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2111 			continue;
2112 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2113 			if (ip6_use_deprecated)
2114 				dep[1] = (struct in6_ifaddr *)ifa;
2115 			continue;
2116 		}
2117 
2118 		best_ia = bestia(best_ia, ia);
2119 	}
2120 	if (best_ia != NULL)
2121 		return best_ia;
2122 
2123 	/* use the last-resort values, that are, deprecated addresses */
2124 	if (dep[0])
2125 		return dep[0];
2126 	if (dep[1])
2127 		return dep[1];
2128 
2129 	return NULL;
2130 }
2131 
2132 /*
2133  * perform DAD when interface becomes IFF_UP.
2134  */
2135 void
2136 in6_if_link_up(struct ifnet *ifp)
2137 {
2138 	struct ifaddr *ifa;
2139 	struct in6_ifaddr *ia;
2140 	int s, bound;
2141 	char ip6buf[INET6_ADDRSTRLEN];
2142 
2143 	/* Ensure it's sane to run DAD */
2144 	if (ifp->if_link_state == LINK_STATE_DOWN)
2145 		return;
2146 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2147 		return;
2148 
2149 	bound = curlwp_bind();
2150 	s = pserialize_read_enter();
2151 	IFADDR_READER_FOREACH(ifa, ifp) {
2152 		struct psref psref;
2153 
2154 		if (ifa->ifa_addr->sa_family != AF_INET6)
2155 			continue;
2156 
2157 		ifa_acquire(ifa, &psref);
2158 		pserialize_read_exit(s);
2159 		ia = (struct in6_ifaddr *)ifa;
2160 
2161 		/* If detached then mark as tentative */
2162 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2163 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2164 			if (if_do_dad(ifp)) {
2165 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2166 				nd6log(LOG_ERR, "%s marked tentative\n",
2167 				    IN6_PRINT(ip6buf,
2168 				    &ia->ia_addr.sin6_addr));
2169 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2170 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2171 		}
2172 
2173 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2174 			int rand_delay;
2175 
2176 			/* Clear the duplicated flag as we're starting DAD. */
2177 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2178 
2179 			/*
2180 			 * The TENTATIVE flag was likely set by hand
2181 			 * beforehand, implicitly indicating the need for DAD.
2182 			 * We may be able to skip the random delay in this
2183 			 * case, but we impose delays just in case.
2184 			 */
2185 			rand_delay = cprng_fast32() %
2186 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2187 			/* +1 ensures callout is always used */
2188 			nd6_dad_start(ifa, rand_delay + 1);
2189 		}
2190 
2191 		s = pserialize_read_enter();
2192 		ifa_release(ifa, &psref);
2193 	}
2194 	pserialize_read_exit(s);
2195 	curlwp_bindx(bound);
2196 
2197 	/* Restore any detached prefixes */
2198 	ND6_WLOCK();
2199 	nd6_pfxlist_onlink_check();
2200 	ND6_UNLOCK();
2201 }
2202 
2203 void
2204 in6_if_up(struct ifnet *ifp)
2205 {
2206 
2207 	/*
2208 	 * special cases, like 6to4, are handled in in6_ifattach
2209 	 */
2210 	in6_ifattach(ifp, NULL);
2211 
2212 	/* interface may not support link state, so bring it up also */
2213 	in6_if_link_up(ifp);
2214 }
2215 
2216 /*
2217  * Mark all addresses as detached.
2218  */
2219 void
2220 in6_if_link_down(struct ifnet *ifp)
2221 {
2222 	struct ifaddr *ifa;
2223 	struct in6_ifaddr *ia;
2224 	int s, bound;
2225 	char ip6buf[INET6_ADDRSTRLEN];
2226 
2227 	/* Any prefixes on this interface should be detached as well */
2228 	ND6_WLOCK();
2229 	nd6_pfxlist_onlink_check();
2230 	ND6_UNLOCK();
2231 
2232 	bound = curlwp_bind();
2233 	s = pserialize_read_enter();
2234 	IFADDR_READER_FOREACH(ifa, ifp) {
2235 		struct psref psref;
2236 
2237 		if (ifa->ifa_addr->sa_family != AF_INET6)
2238 			continue;
2239 
2240 		ifa_acquire(ifa, &psref);
2241 		pserialize_read_exit(s);
2242 		ia = (struct in6_ifaddr *)ifa;
2243 
2244 		/* Stop DAD processing */
2245 		nd6_dad_stop(ifa);
2246 
2247 		/*
2248 		 * Mark the address as detached.
2249 		 * This satisfies RFC4862 Section 5.3, but we should apply
2250 		 * this logic to all addresses to be a good citizen and
2251 		 * avoid potential duplicated addresses.
2252 		 * When the interface comes up again, detached addresses
2253 		 * are marked tentative and DAD commences.
2254 		 */
2255 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2256 			nd6log(LOG_DEBUG, "%s marked detached\n",
2257 			    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
2258 			ia->ia6_flags |= IN6_IFF_DETACHED;
2259 			ia->ia6_flags &=
2260 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2261 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2262 		}
2263 
2264 		s = pserialize_read_enter();
2265 		ifa_release(ifa, &psref);
2266 	}
2267 	pserialize_read_exit(s);
2268 	curlwp_bindx(bound);
2269 }
2270 
2271 void
2272 in6_if_down(struct ifnet *ifp)
2273 {
2274 
2275 	in6_if_link_down(ifp);
2276 }
2277 
2278 void
2279 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2280 {
2281 
2282 	switch (link_state) {
2283 	case LINK_STATE_DOWN:
2284 		in6_if_link_down(ifp);
2285 		break;
2286 	case LINK_STATE_UP:
2287 		in6_if_link_up(ifp);
2288 		break;
2289 	}
2290 }
2291 
2292 /*
2293  * Calculate max IPv6 MTU through all the interfaces and store it
2294  * to in6_maxmtu.
2295  */
2296 void
2297 in6_setmaxmtu(void)
2298 {
2299 	unsigned long maxmtu = 0;
2300 	struct ifnet *ifp;
2301 	int s;
2302 
2303 	s = pserialize_read_enter();
2304 	IFNET_READER_FOREACH(ifp) {
2305 		/* this function can be called during ifnet initialization */
2306 		if (!ifp->if_afdata[AF_INET6])
2307 			continue;
2308 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2309 		    IN6_LINKMTU(ifp) > maxmtu)
2310 			maxmtu = IN6_LINKMTU(ifp);
2311 	}
2312 	pserialize_read_exit(s);
2313 	if (maxmtu)	     /* update only when maxmtu is positive */
2314 		in6_maxmtu = maxmtu;
2315 }
2316 
2317 /*
2318  * Provide the length of interface identifiers to be used for the link attached
2319  * to the given interface.  The length should be defined in "IPv6 over
2320  * xxx-link" document.  Note that address architecture might also define
2321  * the length for a particular set of address prefixes, regardless of the
2322  * link type.  As clarified in rfc2462bis, those two definitions should be
2323  * consistent, and those really are as of August 2004.
2324  */
2325 int
2326 in6_if2idlen(struct ifnet *ifp)
2327 {
2328 	switch (ifp->if_type) {
2329 	case IFT_ETHER:		/* RFC2464 */
2330 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2331 	case IFT_L2VLAN:	/* ditto */
2332 	case IFT_IEEE80211:	/* ditto */
2333 	case IFT_FDDI:		/* RFC2467 */
2334 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2335 	case IFT_PPP:		/* RFC2472 */
2336 	case IFT_ARCNET:	/* RFC2497 */
2337 	case IFT_FRELAY:	/* RFC2590 */
2338 	case IFT_IEEE1394:	/* RFC3146 */
2339 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2340 	case IFT_LOOP:		/* XXX: is this really correct? */
2341 		return 64;
2342 	default:
2343 		/*
2344 		 * Unknown link type:
2345 		 * It might be controversial to use the today's common constant
2346 		 * of 64 for these cases unconditionally.  For full compliance,
2347 		 * we should return an error in this case.  On the other hand,
2348 		 * if we simply miss the standard for the link type or a new
2349 		 * standard is defined for a new link type, the IFID length
2350 		 * is very likely to be the common constant.  As a compromise,
2351 		 * we always use the constant, but make an explicit notice
2352 		 * indicating the "unknown" case.
2353 		 */
2354 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2355 		return 64;
2356 	}
2357 }
2358 
2359 struct in6_llentry {
2360 	struct llentry		base;
2361 };
2362 
2363 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2364 #define	IN6_LLTBL_HASH(k, h) \
2365 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2366 
2367 /*
2368  * Do actual deallocation of @lle.
2369  * Called by LLE_FREE_LOCKED when number of references
2370  * drops to zero.
2371  */
2372 static void
2373 in6_lltable_destroy_lle(struct llentry *lle)
2374 {
2375 
2376 	LLE_WUNLOCK(lle);
2377 	LLE_LOCK_DESTROY(lle);
2378 	kmem_intr_free(lle, sizeof(struct in6_llentry));
2379 }
2380 
2381 static struct llentry *
2382 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2383 {
2384 	struct in6_llentry *lle;
2385 
2386 	lle = kmem_intr_zalloc(sizeof(struct in6_llentry), KM_NOSLEEP);
2387 	if (lle == NULL)		/* NB: caller generates msg */
2388 		return NULL;
2389 
2390 	lle->base.r_l3addr.addr6 = *addr6;
2391 	lle->base.lle_refcnt = 1;
2392 	lle->base.lle_free = in6_lltable_destroy_lle;
2393 	LLE_LOCK_INIT(&lle->base);
2394 	callout_init(&lle->base.lle_timer, CALLOUT_MPSAFE);
2395 
2396 	return &lle->base;
2397 }
2398 
2399 static int
2400 in6_lltable_match_prefix(const struct sockaddr *prefix,
2401     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2402 {
2403 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2404 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2405 
2406 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2407 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2408 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2409 		return 1;
2410 
2411 	return 0;
2412 }
2413 
2414 static void
2415 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2416 {
2417 	struct ifnet *ifp = llt->llt_ifp;
2418 
2419 	IF_AFDATA_WLOCK_ASSERT(ifp);
2420 	LLE_WLOCK_ASSERT(lle);
2421 
2422 	/* Unlink entry from table */
2423 	if ((lle->la_flags & LLE_LINKED) != 0) {
2424 
2425 		lltable_unlink_entry(llt, lle);
2426 		KASSERT((lle->la_flags & LLE_LINKED) == 0);
2427 	}
2428 	/*
2429 	 * We need to release the lock here to lle_timer proceeds;
2430 	 * lle_timer should stop immediately if LLE_LINKED isn't set.
2431 	 * Note that we cannot pass lle->lle_lock to callout_halt
2432 	 * because it's a rwlock.
2433 	 */
2434 	LLE_ADDREF(lle);
2435 	LLE_WUNLOCK(lle);
2436 	IF_AFDATA_WUNLOCK(ifp);
2437 
2438 #ifdef NET_MPSAFE
2439 	callout_halt(&lle->lle_timer, NULL);
2440 #else
2441 	if (mutex_owned(softnet_lock))
2442 		callout_halt(&lle->lle_timer, softnet_lock);
2443 	else
2444 		callout_halt(&lle->lle_timer, NULL);
2445 #endif
2446 	LLE_WLOCK(lle);
2447 	LLE_REMREF(lle);
2448 
2449 	lltable_drop_entry_queue(lle);
2450 	LLE_FREE_LOCKED(lle);
2451 
2452 	IF_AFDATA_WLOCK(ifp);
2453 }
2454 
2455 static int
2456 in6_lltable_rtcheck(struct ifnet *ifp,
2457 		    u_int flags,
2458 		    const struct sockaddr *l3addr)
2459 {
2460 	struct rtentry *rt;
2461 	char ip6buf[INET6_ADDRSTRLEN];
2462 
2463 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2464 	    "sin_family %d", l3addr->sa_family);
2465 
2466 	rt = rtalloc1(l3addr, 0);
2467 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2468 		int s;
2469 		struct ifaddr *ifa;
2470 		/*
2471 		 * Create an ND6 cache for an IPv6 neighbor
2472 		 * that is not covered by our own prefix.
2473 		 */
2474 		/* XXX ifaof_ifpforaddr should take a const param */
2475 		s = pserialize_read_enter();
2476 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2477 		if (ifa != NULL) {
2478 			pserialize_read_exit(s);
2479 			if (rt != NULL)
2480 				rt_unref(rt);
2481 			return 0;
2482 		}
2483 		pserialize_read_exit(s);
2484 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2485 		    IN6_PRINT(ip6buf,
2486 		    &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2487 		if (rt != NULL)
2488 			rt_unref(rt);
2489 		return EINVAL;
2490 	}
2491 	rt_unref(rt);
2492 	return 0;
2493 }
2494 
2495 static inline uint32_t
2496 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2497 {
2498 
2499 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2500 }
2501 
2502 static uint32_t
2503 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2504 {
2505 
2506 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2507 }
2508 
2509 static void
2510 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2511 {
2512 	struct sockaddr_in6 *sin6;
2513 
2514 	sin6 = (struct sockaddr_in6 *)sa;
2515 	bzero(sin6, sizeof(*sin6));
2516 	sin6->sin6_family = AF_INET6;
2517 	sin6->sin6_len = sizeof(*sin6);
2518 	sin6->sin6_addr = lle->r_l3addr.addr6;
2519 }
2520 
2521 static inline struct llentry *
2522 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2523 {
2524 	struct llentry *lle;
2525 	struct llentries *lleh;
2526 	u_int hashidx;
2527 
2528 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2529 	lleh = &llt->lle_head[hashidx];
2530 	LIST_FOREACH(lle, lleh, lle_next) {
2531 		if (lle->la_flags & LLE_DELETED)
2532 			continue;
2533 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2534 			break;
2535 	}
2536 
2537 	return lle;
2538 }
2539 
2540 static int
2541 in6_lltable_delete(struct lltable *llt, u_int flags,
2542 	const struct sockaddr *l3addr)
2543 {
2544 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2545 	struct llentry *lle;
2546 
2547 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2548 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2549 	    "sin_family %d", l3addr->sa_family);
2550 
2551 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2552 
2553 	if (lle == NULL)
2554 		return ENOENT;
2555 
2556 	LLE_WLOCK(lle);
2557 	lle->la_flags |= LLE_DELETED;
2558 #ifdef DIAGNOSTIC
2559 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2560 #endif
2561 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2562 		llentry_free(lle);
2563 	else
2564 		LLE_WUNLOCK(lle);
2565 
2566 	return 0;
2567 }
2568 
2569 static struct llentry *
2570 in6_lltable_create(struct lltable *llt, u_int flags,
2571 	const struct sockaddr *l3addr)
2572 {
2573 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2574 	struct ifnet *ifp = llt->llt_ifp;
2575 	struct llentry *lle;
2576 
2577 	IF_AFDATA_WLOCK_ASSERT(ifp);
2578 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2579 	    "sin_family %d", l3addr->sa_family);
2580 
2581 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2582 
2583 	if (lle != NULL) {
2584 		LLE_WLOCK(lle);
2585 		return lle;
2586 	}
2587 
2588 	/*
2589 	 * A route that covers the given address must have
2590 	 * been installed 1st because we are doing a resolution,
2591 	 * verify this.
2592 	 */
2593 	if (!(flags & LLE_IFADDR) &&
2594 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2595 		return NULL;
2596 
2597 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2598 	if (lle == NULL) {
2599 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2600 		return NULL;
2601 	}
2602 	lle->la_flags = flags;
2603 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2604 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2605 		lle->la_flags |= LLE_VALID;
2606 	}
2607 
2608 	lltable_link_entry(llt, lle);
2609 	LLE_WLOCK(lle);
2610 
2611 	return lle;
2612 }
2613 
2614 static struct llentry *
2615 in6_lltable_lookup(struct lltable *llt, u_int flags,
2616 	const struct sockaddr *l3addr)
2617 {
2618 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2619 	struct llentry *lle;
2620 
2621 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2622 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2623 	    "sin_family %d", l3addr->sa_family);
2624 
2625 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2626 
2627 	if (lle == NULL)
2628 		return NULL;
2629 
2630 	if (flags & LLE_EXCLUSIVE)
2631 		LLE_WLOCK(lle);
2632 	else
2633 		LLE_RLOCK(lle);
2634 	return lle;
2635 }
2636 
2637 static int
2638 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2639     struct rt_walkarg *w)
2640 {
2641 	struct sockaddr_in6 sin6;
2642 
2643 	LLTABLE_LOCK_ASSERT();
2644 
2645 	/* skip deleted entries */
2646 	if (lle->la_flags & LLE_DELETED)
2647 		return 0;
2648 
2649 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2650 
2651 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2652 }
2653 
2654 static struct lltable *
2655 in6_lltattach(struct ifnet *ifp)
2656 {
2657 	struct lltable *llt;
2658 
2659 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2660 	llt->llt_af = AF_INET6;
2661 	llt->llt_ifp = ifp;
2662 
2663 	llt->llt_lookup = in6_lltable_lookup;
2664 	llt->llt_create = in6_lltable_create;
2665 	llt->llt_delete = in6_lltable_delete;
2666 	llt->llt_dump_entry = in6_lltable_dump_entry;
2667 	llt->llt_hash = in6_lltable_hash;
2668 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2669 	llt->llt_free_entry = in6_lltable_free_entry;
2670 	llt->llt_match_prefix = in6_lltable_match_prefix;
2671 	lltable_link(llt);
2672 
2673 	return llt;
2674 }
2675 
2676 void *
2677 in6_domifattach(struct ifnet *ifp)
2678 {
2679 	struct in6_ifextra *ext;
2680 
2681 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2682 
2683 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2684 	    M_IFADDR, M_WAITOK|M_ZERO);
2685 
2686 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2687 	    M_IFADDR, M_WAITOK|M_ZERO);
2688 
2689 	ext->nd_ifinfo = nd6_ifattach(ifp);
2690 	ext->scope6_id = scope6_ifattach(ifp);
2691 	ext->nprefixes = 0;
2692 	ext->ndefrouters = 0;
2693 
2694 	ext->lltable = in6_lltattach(ifp);
2695 
2696 	return ext;
2697 }
2698 
2699 void
2700 in6_domifdetach(struct ifnet *ifp, void *aux)
2701 {
2702 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2703 
2704 	lltable_free(ext->lltable);
2705 	ext->lltable = NULL;
2706 #ifndef NET_MPSAFE
2707 	mutex_enter(softnet_lock);
2708 #endif
2709 	nd6_ifdetach(ifp, ext);
2710 #ifndef NET_MPSAFE
2711 	mutex_exit(softnet_lock);
2712 #endif
2713 	free(ext->in6_ifstat, M_IFADDR);
2714 	free(ext->icmp6_ifstat, M_IFADDR);
2715 	scope6_ifdetach(ext->scope6_id);
2716 	free(ext, M_IFADDR);
2717 }
2718 
2719 /*
2720  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2721  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2722  */
2723 void
2724 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2725 {
2726 	in6->s6_addr32[0] = 0;
2727 	in6->s6_addr32[1] = 0;
2728 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2729 	in6->s6_addr32[3] = in->s_addr;
2730 }
2731 
2732 /*
2733  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2734  * v4 mapped addr or v4 compat addr
2735  */
2736 void
2737 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2738 {
2739 	memset(sin, 0, sizeof(*sin));
2740 	sin->sin_len = sizeof(struct sockaddr_in);
2741 	sin->sin_family = AF_INET;
2742 	sin->sin_port = sin6->sin6_port;
2743 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2744 }
2745 
2746 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2747 void
2748 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2749 {
2750 	memset(sin6, 0, sizeof(*sin6));
2751 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2752 	sin6->sin6_family = AF_INET6;
2753 	sin6->sin6_port = sin->sin_port;
2754 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2755 }
2756 
2757 /* Convert sockaddr_in6 into sockaddr_in. */
2758 void
2759 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2760 {
2761 	struct sockaddr_in *sin_p;
2762 	struct sockaddr_in6 sin6;
2763 
2764 	/*
2765 	 * Save original sockaddr_in6 addr and convert it
2766 	 * to sockaddr_in.
2767 	 */
2768 	sin6 = *(struct sockaddr_in6 *)nam;
2769 	sin_p = (struct sockaddr_in *)nam;
2770 	in6_sin6_2_sin(sin_p, &sin6);
2771 }
2772 
2773 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2774 void
2775 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2776 {
2777 	struct sockaddr_in *sin_p;
2778 	struct sockaddr_in6 *sin6_p;
2779 
2780 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2781 	sin_p = (struct sockaddr_in *)*nam;
2782 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2783 	free(*nam, M_SONAME);
2784 	*nam = sin6tosa(sin6_p);
2785 }
2786