xref: /netbsd-src/sys/netinet6/in6.c (revision f21b7d7f2cbdd5c14b3882c4e8a3d43580d460a6)
1 /*	$NetBSD: in6.c,v 1.220 2016/09/13 15:57:50 christos 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 1.220 2016/09/13 15:57:50 christos Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_compat_netbsd.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/ioctl.h>
74 #include <sys/errno.h>
75 #include <sys/malloc.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sockio.h>
79 #include <sys/systm.h>
80 #include <sys/proc.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/syslog.h>
84 #include <sys/kauth.h>
85 #include <sys/cprng.h>
86 #include <sys/kmem.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/if_llatbl.h>
91 #include <net/if_ether.h>
92 #include <net/if_dl.h>
93 #include <net/pfil.h>
94 #include <net/route.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_var.h>
98 
99 #include <netinet/ip6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/nd6.h>
102 #include <netinet6/mld6_var.h>
103 #include <netinet6/ip6_mroute.h>
104 #include <netinet6/in6_ifattach.h>
105 #include <netinet6/scope6_var.h>
106 
107 #include <net/net_osdep.h>
108 
109 #ifdef COMPAT_50
110 #include <compat/netinet6/in6_var.h>
111 #endif
112 
113 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
114 
115 /* enable backward compatibility code for obsoleted ioctls */
116 #define COMPAT_IN6IFIOCTL
117 
118 #ifdef	IN6_DEBUG
119 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
120 #else
121 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
122 #endif /* IN6_DEBUG */
123 
124 /*
125  * Definitions of some constant IP6 addresses.
126  */
127 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
128 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
129 const struct in6_addr in6addr_nodelocal_allnodes =
130 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
131 const struct in6_addr in6addr_linklocal_allnodes =
132 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
133 const struct in6_addr in6addr_linklocal_allrouters =
134 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
135 
136 const struct in6_addr in6mask0 = IN6MASK0;
137 const struct in6_addr in6mask32 = IN6MASK32;
138 const struct in6_addr in6mask64 = IN6MASK64;
139 const struct in6_addr in6mask96 = IN6MASK96;
140 const struct in6_addr in6mask128 = IN6MASK128;
141 
142 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
143 				     0, 0, IN6ADDR_ANY_INIT, 0};
144 
145 struct pslist_head	in6_ifaddr_list;
146 kmutex_t		in6_ifaddr_lock;
147 
148 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
149 	struct ifnet *);
150 static int in6_ifaddprefix(struct in6_ifaddr *);
151 static int in6_ifremprefix(struct in6_ifaddr *);
152 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
153 	const struct sockaddr_in6 *, int);
154 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
155 
156 void
157 in6_init(void)
158 {
159 
160 	PSLIST_INIT(&in6_ifaddr_list);
161 	mutex_init(&in6_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
162 }
163 
164 /*
165  * Add ownaddr as loopback rtentry.  We previously add the route only if
166  * necessary (ex. on a p2p link).  However, since we now manage addresses
167  * separately from prefixes, we should always add the route.  We can't
168  * rely on the cloning mechanism from the corresponding interface route
169  * any more.
170  */
171 void
172 in6_ifaddlocal(struct ifaddr *ifa)
173 {
174 
175 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
176 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
177 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
178 	{
179 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
180 		return;
181 	}
182 
183 	rt_ifa_addlocal(ifa);
184 }
185 
186 /*
187  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
188  * if it exists.
189  */
190 void
191 in6_ifremlocal(struct ifaddr *ifa)
192 {
193 	struct in6_ifaddr *ia;
194 	struct ifaddr *alt_ifa = NULL;
195 	int ia_count = 0;
196 	struct psref psref;
197 	int s;
198 
199 	/*
200 	 * Some of BSD variants do not remove cloned routes
201 	 * from an interface direct route, when removing the direct route
202 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
203 	 * cloned routes, they could fail to remove the cloned routes when
204 	 * we handle multple addresses that share a common prefix.
205 	 * So, we should remove the route corresponding to the deleted address.
206 	 */
207 
208 	/*
209 	 * Delete the entry only if exactly one ifaddr matches the
210 	 * address, ifa->ifa_addr.
211 	 *
212 	 * If more than one ifaddr matches, replace the ifaddr in
213 	 * the routing table, rt_ifa, with a different ifaddr than
214 	 * the one we are purging, ifa.  It is important to do
215 	 * this, or else the routing table can accumulate dangling
216 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
217 	 * which will lead to crashes, later.  (More than one ifaddr
218 	 * can match if we assign the same address to multiple---probably
219 	 * p2p---interfaces.)
220 	 *
221 	 * XXX An old comment at this place said, "we should avoid
222 	 * XXX such a configuration [i.e., interfaces with the same
223 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
224 	 * XXX agree, especially now that I have fixed the dangling
225 	 * XXX ifp-pointers bug.
226 	 */
227 	s = pserialize_read_enter();
228 	IN6_ADDRLIST_READER_FOREACH(ia) {
229 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
230 			continue;
231 		if (ia->ia_ifp != ifa->ifa_ifp)
232 			alt_ifa = &ia->ia_ifa;
233 		if (++ia_count > 1 && alt_ifa != NULL)
234 			break;
235 	}
236 	if (ia_count > 1 && alt_ifa != NULL)
237 		ifa_acquire(alt_ifa, &psref);
238 	pserialize_read_exit(s);
239 
240 	if (ia_count == 0)
241 		return;
242 
243 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
244 
245 	if (ia_count > 1 && alt_ifa != NULL)
246 		ifa_release(alt_ifa, &psref);
247 }
248 
249 /* Add prefix route for the network. */
250 static int
251 in6_ifaddprefix(struct in6_ifaddr *ia)
252 {
253 	int error, flags = 0;
254 
255 	if (in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) == 128) {
256 		if (ia->ia_dstaddr.sin6_family != AF_INET6)
257 			/* We don't need to install a host route. */
258 			return 0;
259 		flags |= RTF_HOST;
260 	}
261 
262 	/* Is this a connected route for neighbour discovery? */
263 	if (nd6_need_cache(ia->ia_ifp))
264 		flags |= RTF_CONNECTED;
265 
266 	if ((error = rtinit(&ia->ia_ifa, RTM_ADD, RTF_UP | flags)) == 0)
267 		ia->ia_flags |= IFA_ROUTE;
268 	else if (error == EEXIST)
269 		/* Existance of the route is not an error. */
270 		error = 0;
271 
272 	return error;
273 }
274 
275 /* Delete network prefix route if present.
276  * Re-add it to another address if the prefix matches. */
277 static int
278 in6_ifremprefix(struct in6_ifaddr *target)
279 {
280 	int error, s;
281 	struct in6_ifaddr *ia;
282 
283 	if ((target->ia_flags & IFA_ROUTE) == 0)
284 		return 0;
285 
286 	s = pserialize_read_enter();
287 	IN6_ADDRLIST_READER_FOREACH(ia) {
288 		if (target->ia_dstaddr.sin6_len) {
289 			if (ia->ia_dstaddr.sin6_len == 0 ||
290 			    !IN6_ARE_ADDR_EQUAL(&ia->ia_dstaddr.sin6_addr,
291 			    &target->ia_dstaddr.sin6_addr))
292 				continue;
293 		} else {
294 			if (!IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
295 			    &target->ia_addr.sin6_addr,
296 			    &target->ia_prefixmask.sin6_addr))
297 				continue;
298 		}
299 
300 		/*
301 		 * if we got a matching prefix route, move IFA_ROUTE to him
302 		 */
303 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
304 			struct psref psref;
305 			int bound = curlwp_bind();
306 
307 			ia6_acquire(ia, &psref);
308 			pserialize_read_exit(s);
309 
310 			rtinit(&target->ia_ifa, RTM_DELETE, 0);
311 			target->ia_flags &= ~IFA_ROUTE;
312 
313 			error = in6_ifaddprefix(ia);
314 
315 			ia6_release(ia, &psref);
316 			curlwp_bindx(bound);
317 
318 			return error;
319 		}
320 	}
321 	pserialize_read_exit(s);
322 
323 	/*
324 	 * noone seem to have prefix route.  remove it.
325 	 */
326 	rtinit(&target->ia_ifa, RTM_DELETE, 0);
327 	target->ia_flags &= ~IFA_ROUTE;
328 	return 0;
329 }
330 
331 int
332 in6_mask2len(struct in6_addr *mask, u_char *lim0)
333 {
334 	int x = 0, y;
335 	u_char *lim = lim0, *p;
336 
337 	/* ignore the scope_id part */
338 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
339 		lim = (u_char *)mask + sizeof(*mask);
340 	for (p = (u_char *)mask; p < lim; x++, p++) {
341 		if (*p != 0xff)
342 			break;
343 	}
344 	y = 0;
345 	if (p < lim) {
346 		for (y = 0; y < NBBY; y++) {
347 			if ((*p & (0x80 >> y)) == 0)
348 				break;
349 		}
350 	}
351 
352 	/*
353 	 * when the limit pointer is given, do a stricter check on the
354 	 * remaining bits.
355 	 */
356 	if (p < lim) {
357 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
358 			return -1;
359 		for (p = p + 1; p < lim; p++)
360 			if (*p != 0)
361 				return -1;
362 	}
363 
364 	return x * NBBY + y;
365 }
366 
367 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
368 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
369 
370 static int
371 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
372 {
373 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
374 	struct	in6_ifaddr *ia = NULL;
375 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
376 	struct sockaddr_in6 *sa6;
377 	int error, bound;
378 	struct psref psref;
379 	bool run_hooks = false;
380 
381 	switch (cmd) {
382 	case SIOCAADDRCTL_POLICY:
383 	case SIOCDADDRCTL_POLICY:
384 		/* Privileged. */
385 		return in6_src_ioctl(cmd, data);
386 	/*
387 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
388 	 */
389 	case SIOCSIFADDR:
390 	case SIOCSIFDSTADDR:
391 	case SIOCSIFBRDADDR:
392 	case SIOCSIFNETMASK:
393 		return EOPNOTSUPP;
394 	case SIOCGETSGCNT_IN6:
395 	case SIOCGETMIFCNT_IN6:
396 		return mrt6_ioctl(cmd, data);
397 	case SIOCGIFADDRPREF:
398 	case SIOCSIFADDRPREF:
399 		if (ifp == NULL)
400 			return EINVAL;
401 		return ifaddrpref_ioctl(so, cmd, data, ifp);
402 	}
403 
404 	if (ifp == NULL)
405 		return EOPNOTSUPP;
406 
407 	switch (cmd) {
408 	case SIOCSNDFLUSH_IN6:
409 	case SIOCSPFXFLUSH_IN6:
410 	case SIOCSRTRFLUSH_IN6:
411 	case SIOCSDEFIFACE_IN6:
412 	case SIOCSIFINFO_FLAGS:
413 	case SIOCSIFINFO_IN6:
414 		/* Privileged. */
415 		/* FALLTHROUGH */
416 	case OSIOCGIFINFO_IN6:
417 	case SIOCGIFINFO_IN6:
418 	case SIOCGDRLST_IN6:
419 	case SIOCGPRLST_IN6:
420 	case SIOCGNBRINFO_IN6:
421 	case SIOCGDEFIFACE_IN6:
422 		return nd6_ioctl(cmd, data, ifp);
423 	}
424 
425 	switch (cmd) {
426 	case SIOCSIFPREFIX_IN6:
427 	case SIOCDIFPREFIX_IN6:
428 	case SIOCAIFPREFIX_IN6:
429 	case SIOCCIFPREFIX_IN6:
430 	case SIOCSGIFPREFIX_IN6:
431 	case SIOCGIFPREFIX_IN6:
432 		log(LOG_NOTICE,
433 		    "prefix ioctls are now invalidated. "
434 		    "please use ifconfig.\n");
435 		return EOPNOTSUPP;
436 	}
437 
438 	switch (cmd) {
439 	case SIOCALIFADDR:
440 	case SIOCDLIFADDR:
441 		/* Privileged. */
442 		/* FALLTHROUGH */
443 	case SIOCGLIFADDR:
444 		return in6_lifaddr_ioctl(so, cmd, data, ifp);
445 	}
446 
447 	/*
448 	 * Find address for this interface, if it exists.
449 	 *
450 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
451 	 * only, and used the first interface address as the target of other
452 	 * operations (without checking ifra_addr).  This was because netinet
453 	 * code/API assumed at most 1 interface address per interface.
454 	 * Since IPv6 allows a node to assign multiple addresses
455 	 * on a single interface, we almost always look and check the
456 	 * presence of ifra_addr, and reject invalid ones here.
457 	 * It also decreases duplicated code among SIOC*_IN6 operations.
458 	 */
459 	switch (cmd) {
460 	case SIOCAIFADDR_IN6:
461 #ifdef OSIOCAIFADDR_IN6
462 	case OSIOCAIFADDR_IN6:
463 #endif
464 #ifdef OSIOCSIFPHYADDR_IN6
465 	case OSIOCSIFPHYADDR_IN6:
466 #endif
467 	case SIOCSIFPHYADDR_IN6:
468 		sa6 = &ifra->ifra_addr;
469 		break;
470 	case SIOCSIFADDR_IN6:
471 	case SIOCGIFADDR_IN6:
472 	case SIOCSIFDSTADDR_IN6:
473 	case SIOCSIFNETMASK_IN6:
474 	case SIOCGIFDSTADDR_IN6:
475 	case SIOCGIFNETMASK_IN6:
476 	case SIOCDIFADDR_IN6:
477 	case SIOCGIFPSRCADDR_IN6:
478 	case SIOCGIFPDSTADDR_IN6:
479 	case SIOCGIFAFLAG_IN6:
480 	case SIOCSNDFLUSH_IN6:
481 	case SIOCSPFXFLUSH_IN6:
482 	case SIOCSRTRFLUSH_IN6:
483 	case SIOCGIFALIFETIME_IN6:
484 #ifdef OSIOCGIFALIFETIME_IN6
485 	case OSIOCGIFALIFETIME_IN6:
486 #endif
487 	case SIOCGIFSTAT_IN6:
488 	case SIOCGIFSTAT_ICMP6:
489 		sa6 = &ifr->ifr_addr;
490 		break;
491 	default:
492 		sa6 = NULL;
493 		break;
494 	}
495 
496 	error = 0;
497 	bound = curlwp_bind();
498 	if (sa6 && sa6->sin6_family == AF_INET6) {
499 		if (sa6->sin6_scope_id != 0)
500 			error = sa6_embedscope(sa6, 0);
501 		else
502 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
503 		if (error != 0)
504 			goto out;
505 		ia = in6ifa_ifpwithaddr_psref(ifp, &sa6->sin6_addr, &psref);
506 	} else
507 		ia = NULL;
508 
509 	switch (cmd) {
510 	case SIOCSIFADDR_IN6:
511 	case SIOCSIFDSTADDR_IN6:
512 	case SIOCSIFNETMASK_IN6:
513 		/*
514 		 * Since IPv6 allows a node to assign multiple addresses
515 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
516 		 */
517 		error = EINVAL;
518 		goto release;
519 
520 	case SIOCDIFADDR_IN6:
521 		/*
522 		 * for IPv4, we look for existing in_ifaddr here to allow
523 		 * "ifconfig if0 delete" to remove the first IPv4 address on
524 		 * the interface.  For IPv6, as the spec allows multiple
525 		 * interface address from the day one, we consider "remove the
526 		 * first one" semantics to be not preferable.
527 		 */
528 		if (ia == NULL) {
529 			error = EADDRNOTAVAIL;
530 			goto out;
531 		}
532 		/* FALLTHROUGH */
533 #ifdef OSIOCAIFADDR_IN6
534 	case OSIOCAIFADDR_IN6:
535 #endif
536 	case SIOCAIFADDR_IN6:
537 		/*
538 		 * We always require users to specify a valid IPv6 address for
539 		 * the corresponding operation.
540 		 */
541 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
542 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
543 			error = EAFNOSUPPORT;
544 			goto release;
545 		}
546 		/* Privileged. */
547 
548 		break;
549 
550 	case SIOCGIFADDR_IN6:
551 		/* This interface is basically deprecated. use SIOCGIFCONF. */
552 		/* FALLTHROUGH */
553 	case SIOCGIFAFLAG_IN6:
554 	case SIOCGIFNETMASK_IN6:
555 	case SIOCGIFDSTADDR_IN6:
556 	case SIOCGIFALIFETIME_IN6:
557 #ifdef OSIOCGIFALIFETIME_IN6
558 	case OSIOCGIFALIFETIME_IN6:
559 #endif
560 		/* must think again about its semantics */
561 		if (ia == NULL) {
562 			error = EADDRNOTAVAIL;
563 			goto out;
564 		}
565 		break;
566 	}
567 
568 	switch (cmd) {
569 
570 	case SIOCGIFADDR_IN6:
571 		ifr->ifr_addr = ia->ia_addr;
572 		error = sa6_recoverscope(&ifr->ifr_addr);
573 		break;
574 
575 	case SIOCGIFDSTADDR_IN6:
576 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
577 			error = EINVAL;
578 			break;
579 		}
580 		/*
581 		 * XXX: should we check if ifa_dstaddr is NULL and return
582 		 * an error?
583 		 */
584 		ifr->ifr_dstaddr = ia->ia_dstaddr;
585 		error = sa6_recoverscope(&ifr->ifr_dstaddr);
586 		break;
587 
588 	case SIOCGIFNETMASK_IN6:
589 		ifr->ifr_addr = ia->ia_prefixmask;
590 		break;
591 
592 	case SIOCGIFAFLAG_IN6:
593 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
594 		break;
595 
596 	case SIOCGIFSTAT_IN6:
597 		if (ifp == NULL) {
598 			error = EINVAL;
599 			break;
600 		}
601 		memset(&ifr->ifr_ifru.ifru_stat, 0,
602 		    sizeof(ifr->ifr_ifru.ifru_stat));
603 		ifr->ifr_ifru.ifru_stat =
604 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
605 		break;
606 
607 	case SIOCGIFSTAT_ICMP6:
608 		if (ifp == NULL) {
609 			error = EINVAL;
610 			break;
611 		}
612 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
613 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
614 		ifr->ifr_ifru.ifru_icmp6stat =
615 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
616 		break;
617 
618 #ifdef OSIOCGIFALIFETIME_IN6
619 	case OSIOCGIFALIFETIME_IN6:
620 #endif
621 	case SIOCGIFALIFETIME_IN6:
622 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
623 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
624 			time_t maxexpire;
625 			struct in6_addrlifetime *retlt =
626 			    &ifr->ifr_ifru.ifru_lifetime;
627 
628 			/*
629 			 * XXX: adjust expiration time assuming time_t is
630 			 * signed.
631 			 */
632 			maxexpire = ((time_t)~0) &
633 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
634 			if (ia->ia6_lifetime.ia6t_vltime <
635 			    maxexpire - ia->ia6_updatetime) {
636 				retlt->ia6t_expire = ia->ia6_updatetime +
637 				    ia->ia6_lifetime.ia6t_vltime;
638 				retlt->ia6t_expire = retlt->ia6t_expire ?
639 				    time_mono_to_wall(retlt->ia6t_expire) :
640 				    0;
641 			} else
642 				retlt->ia6t_expire = maxexpire;
643 		}
644 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
645 			time_t maxexpire;
646 			struct in6_addrlifetime *retlt =
647 			    &ifr->ifr_ifru.ifru_lifetime;
648 
649 			/*
650 			 * XXX: adjust expiration time assuming time_t is
651 			 * signed.
652 			 */
653 			maxexpire = ((time_t)~0) &
654 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
655 			if (ia->ia6_lifetime.ia6t_pltime <
656 			    maxexpire - ia->ia6_updatetime) {
657 				retlt->ia6t_preferred = ia->ia6_updatetime +
658 				    ia->ia6_lifetime.ia6t_pltime;
659 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
660 				    time_mono_to_wall(retlt->ia6t_preferred) :
661 				    0;
662 			} else
663 				retlt->ia6t_preferred = maxexpire;
664 		}
665 #ifdef OSIOCFIFALIFETIME_IN6
666 		if (cmd == OSIOCFIFALIFETIME_IN6)
667 			in6_addrlifetime_to_in6_addrlifetime50(
668 			    &ifr->ifru.ifru_lifetime);
669 #endif
670 		break;
671 
672 #ifdef OSIOCAIFADDR_IN6
673 	case OSIOCAIFADDR_IN6:
674 		in6_aliasreq50_to_in6_aliasreq(ifra);
675 		/*FALLTHROUGH*/
676 #endif
677 	case SIOCAIFADDR_IN6:
678 	{
679 		struct in6_addrlifetime *lt;
680 
681 		/* reject read-only flags */
682 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
683 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
684 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
685 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
686 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
687 			error = EINVAL;
688 			break;
689 		}
690 		/*
691 		 * ia6t_expire and ia6t_preferred won't be used for now,
692 		 * so just in case.
693 		 */
694 		lt = &ifra->ifra_lifetime;
695 		if (lt->ia6t_expire != 0)
696 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
697 		if (lt->ia6t_preferred != 0)
698 			lt->ia6t_preferred =
699 			    time_wall_to_mono(lt->ia6t_preferred);
700 		/*
701 		 * make (ia == NULL) or update (ia != NULL) the interface
702 		 * address structure, and link it to the list.
703 		 */
704 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
705 			break;
706 		run_hooks = true;
707 		break;
708 	}
709 
710 	case SIOCDIFADDR_IN6:
711 	{
712 		struct nd_prefix *pr;
713 
714 		/*
715 		 * If the address being deleted is the only one that owns
716 		 * the corresponding prefix, expire the prefix as well.
717 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
718 		 */
719 		pr = ia->ia6_ndpr;
720 		ia6_release(ia, &psref);
721 		in6_purgeaddr(&ia->ia_ifa);
722 		ia = NULL;
723 		if (pr && pr->ndpr_refcnt == 0)
724 			prelist_remove(pr);
725 		run_hooks = true;
726 		break;
727 	}
728 
729 	default:
730 		error = ENOTTY;
731 	}
732 release:
733 	ia6_release(ia, &psref);
734 
735 	if (run_hooks)
736 		pfil_run_hooks(if_pfil, (struct mbuf **)cmd, ifp, PFIL_IFADDR);
737 out:
738 	curlwp_bindx(bound);
739 	return error;
740 }
741 
742 int
743 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
744 {
745 	int error, s;
746 
747 	switch (cmd) {
748 	case SIOCSNDFLUSH_IN6:
749 	case SIOCSPFXFLUSH_IN6:
750 	case SIOCSRTRFLUSH_IN6:
751 	case SIOCSDEFIFACE_IN6:
752 	case SIOCSIFINFO_FLAGS:
753 	case SIOCSIFINFO_IN6:
754 
755 	case SIOCALIFADDR:
756 	case SIOCDLIFADDR:
757 
758 	case SIOCDIFADDR_IN6:
759 #ifdef OSIOCAIFADDR_IN6
760 	case OSIOCAIFADDR_IN6:
761 #endif
762 	case SIOCAIFADDR_IN6:
763 
764 	case SIOCAADDRCTL_POLICY:
765 	case SIOCDADDRCTL_POLICY:
766 
767 		if (kauth_authorize_network(curlwp->l_cred,
768 		    KAUTH_NETWORK_SOCKET,
769 		    KAUTH_REQ_NETWORK_SOCKET_SETPRIV,
770 		    so, NULL, NULL))
771 			return EPERM;
772 		break;
773 	}
774 
775 	s = splnet();
776 	mutex_enter(softnet_lock);
777 	error = in6_control1(so , cmd, data, ifp);
778 	mutex_exit(softnet_lock);
779 	splx(s);
780 	return error;
781 }
782 
783 /*
784  * Update parameters of an IPv6 interface address.
785  * If necessary, a new entry is created and linked into address chains.
786  * This function is separated from in6_control().
787  * XXX: should this be performed under splnet()?
788  */
789 static int
790 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
791     struct in6_ifaddr *ia, int flags)
792 {
793 	int error = 0, hostIsNew = 0, plen = -1;
794 	struct sockaddr_in6 dst6;
795 	struct in6_addrlifetime *lt;
796 	struct in6_multi_mship *imm;
797 	struct in6_multi *in6m_sol;
798 	struct rtentry *rt;
799 	int dad_delay, was_tentative;
800 
801 	in6m_sol = NULL;
802 
803 	/* Validate parameters */
804 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
805 		return EINVAL;
806 
807 	/*
808 	 * The destination address for a p2p link must have a family
809 	 * of AF_UNSPEC or AF_INET6.
810 	 */
811 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
812 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
813 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
814 		return EAFNOSUPPORT;
815 	/*
816 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
817 	 * does not carry fields other than sin6_len.
818 	 */
819 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
820 		return EINVAL;
821 	/*
822 	 * Because the IPv6 address architecture is classless, we require
823 	 * users to specify a (non 0) prefix length (mask) for a new address.
824 	 * We also require the prefix (when specified) mask is valid, and thus
825 	 * reject a non-consecutive mask.
826 	 */
827 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
828 		return EINVAL;
829 	if (ifra->ifra_prefixmask.sin6_len != 0) {
830 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
831 		    (u_char *)&ifra->ifra_prefixmask +
832 		    ifra->ifra_prefixmask.sin6_len);
833 		if (plen <= 0)
834 			return EINVAL;
835 	} else {
836 		/*
837 		 * In this case, ia must not be NULL.  We just use its prefix
838 		 * length.
839 		 */
840 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
841 	}
842 	/*
843 	 * If the destination address on a p2p interface is specified,
844 	 * and the address is a scoped one, validate/set the scope
845 	 * zone identifier.
846 	 */
847 	dst6 = ifra->ifra_dstaddr;
848 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
849 	    (dst6.sin6_family == AF_INET6)) {
850 		struct in6_addr in6_tmp;
851 		u_int32_t zoneid;
852 
853 		in6_tmp = dst6.sin6_addr;
854 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
855 			return EINVAL; /* XXX: should be impossible */
856 
857 		if (dst6.sin6_scope_id != 0) {
858 			if (dst6.sin6_scope_id != zoneid)
859 				return EINVAL;
860 		} else		/* user omit to specify the ID. */
861 			dst6.sin6_scope_id = zoneid;
862 
863 		/* convert into the internal form */
864 		if (sa6_embedscope(&dst6, 0))
865 			return EINVAL; /* XXX: should be impossible */
866 	}
867 	/*
868 	 * The destination address can be specified only for a p2p or a
869 	 * loopback interface.  If specified, the corresponding prefix length
870 	 * must be 128.
871 	 */
872 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
873 #ifdef FORCE_P2PPLEN
874 		int i;
875 #endif
876 
877 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
878 			/* XXX: noisy message */
879 			nd6log(LOG_INFO, "a destination can "
880 			    "be specified for a p2p or a loopback IF only\n");
881 			return EINVAL;
882 		}
883 		if (plen != 128) {
884 			nd6log(LOG_INFO, "prefixlen should "
885 			    "be 128 when dstaddr is specified\n");
886 #ifdef FORCE_P2PPLEN
887 			/*
888 			 * To be compatible with old configurations,
889 			 * such as ifconfig gif0 inet6 2001::1 2001::2
890 			 * prefixlen 126, we override the specified
891 			 * prefixmask as if the prefix length was 128.
892 			 */
893 			ifra->ifra_prefixmask.sin6_len =
894 			    sizeof(struct sockaddr_in6);
895 			for (i = 0; i < 4; i++)
896 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
897 				    0xffffffff;
898 			plen = 128;
899 #else
900 			return EINVAL;
901 #endif
902 		}
903 	}
904 	/* lifetime consistency check */
905 	lt = &ifra->ifra_lifetime;
906 	if (lt->ia6t_pltime > lt->ia6t_vltime)
907 		return EINVAL;
908 	if (lt->ia6t_vltime == 0) {
909 		/*
910 		 * the following log might be noisy, but this is a typical
911 		 * configuration mistake or a tool's bug.
912 		 */
913 		nd6log(LOG_INFO, "valid lifetime is 0 for %s\n",
914 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr));
915 
916 		if (ia == NULL)
917 			return 0; /* there's nothing to do */
918 	}
919 
920 	/*
921 	 * If this is a new address, allocate a new ifaddr and link it
922 	 * into chains.
923 	 */
924 	if (ia == NULL) {
925 		hostIsNew = 1;
926 		/*
927 		 * When in6_update_ifa() is called in a process of a received
928 		 * RA, it is called under an interrupt context.  So, we should
929 		 * call malloc with M_NOWAIT.
930 		 */
931 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
932 		    M_NOWAIT);
933 		if (ia == NULL)
934 			return ENOBUFS;
935 		memset(ia, 0, sizeof(*ia));
936 		LIST_INIT(&ia->ia6_memberships);
937 		/* Initialize the address and masks, and put time stamp */
938 		ia->ia_ifa.ifa_addr = sin6tosa(&ia->ia_addr);
939 		ia->ia_addr.sin6_family = AF_INET6;
940 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
941 		ia->ia6_createtime = time_uptime;
942 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
943 			/*
944 			 * XXX: some functions expect that ifa_dstaddr is not
945 			 * NULL for p2p interfaces.
946 			 */
947 			ia->ia_ifa.ifa_dstaddr = sin6tosa(&ia->ia_dstaddr);
948 		} else {
949 			ia->ia_ifa.ifa_dstaddr = NULL;
950 		}
951 		ia->ia_ifa.ifa_netmask = sin6tosa(&ia->ia_prefixmask);
952 
953 		ia->ia_ifp = ifp;
954 		IN6_ADDRLIST_ENTRY_INIT(ia);
955 		ifa_psref_init(&ia->ia_ifa);
956 	}
957 
958 	/* update timestamp */
959 	ia->ia6_updatetime = time_uptime;
960 
961 	/* set prefix mask */
962 	if (ifra->ifra_prefixmask.sin6_len) {
963 		if (ia->ia_prefixmask.sin6_len) {
964 			/*
965 			 * We prohibit changing the prefix length of an
966 			 * existing autoconf address, because the operation
967 			 * would confuse prefix management.
968 			 */
969 			if (ia->ia6_ndpr != NULL &&
970 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) !=
971 			    plen)
972 			{
973 				nd6log(LOG_INFO, "the prefix length of an"
974 				    " existing (%s) autoconf address should"
975 				    " not be changed\n",
976 				    ip6_sprintf(&ia->ia_addr.sin6_addr));
977 				error = EINVAL;
978 				if (hostIsNew)
979 					free(ia, M_IFADDR);
980 				goto exit;
981 			}
982 
983 			if (!IN6_ARE_ADDR_EQUAL(&ia->ia_prefixmask.sin6_addr,
984 			    &ifra->ifra_prefixmask.sin6_addr))
985 				in6_ifremprefix(ia);
986 		}
987 		ia->ia_prefixmask = ifra->ifra_prefixmask;
988 	}
989 
990 	/* Set destination address. */
991 	if (dst6.sin6_family == AF_INET6) {
992 		if (!IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
993 		    &ia->ia_dstaddr.sin6_addr))
994 			in6_ifremprefix(ia);
995 		ia->ia_dstaddr = dst6;
996 	}
997 
998 	/*
999 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1000 	 * to see if the address is deprecated or invalidated, but initialize
1001 	 * these members for applications.
1002 	 */
1003 	ia->ia6_lifetime = ifra->ifra_lifetime;
1004 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1005 		ia->ia6_lifetime.ia6t_expire =
1006 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1007 	} else
1008 		ia->ia6_lifetime.ia6t_expire = 0;
1009 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1010 		ia->ia6_lifetime.ia6t_preferred =
1011 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1012 	} else
1013 		ia->ia6_lifetime.ia6t_preferred = 0;
1014 
1015 	/*
1016 	 * configure address flags.
1017 	 * We need to preserve tentative state so DAD works if
1018 	 * something adds the same address before DAD finishes.
1019 	 */
1020 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1021 	ia->ia6_flags = ifra->ifra_flags;
1022 
1023 	/*
1024 	 * Make the address tentative before joining multicast addresses,
1025 	 * so that corresponding MLD responses would not have a tentative
1026 	 * source address.
1027 	 */
1028 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1029 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1030 		ia->ia6_flags |= IN6_IFF_DETACHED;
1031 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1032 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp))
1033 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1034 
1035 	/*
1036 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1037 	 * userland, make it deprecated.
1038 	 */
1039 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1040 		ia->ia6_lifetime.ia6t_pltime = 0;
1041 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1042 	}
1043 
1044 	/* reset the interface and routing table appropriately. */
1045 	error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew);
1046 	if (error != 0) {
1047 		if (hostIsNew)
1048 			free(ia, M_IFADDR);
1049 		goto exit;
1050 	}
1051 
1052 	/*
1053 	 * We are done if we have simply modified an existing address.
1054 	 */
1055 	if (!hostIsNew)
1056 		return error;
1057 
1058 	/*
1059 	 * Insert ia to the global list and ifa to the interface's list.
1060 	 */
1061 	mutex_enter(&in6_ifaddr_lock);
1062 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1063 	mutex_exit(&in6_ifaddr_lock);
1064 
1065 	/* gain a refcnt for the link from in6_ifaddr */
1066 	ifaref(&ia->ia_ifa);
1067 	ifa_insert(ifp, &ia->ia_ifa);
1068 
1069 	/*
1070 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1071 	 * not just go to unlink.
1072 	 */
1073 
1074 	/* join necessary multicast groups */
1075 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1076 		struct sockaddr_in6 mltaddr, mltmask;
1077 		struct in6_addr llsol;
1078 
1079 		/* join solicited multicast addr for new host id */
1080 		memset(&llsol, 0, sizeof(struct in6_addr));
1081 		llsol.s6_addr16[0] = htons(0xff02);
1082 		llsol.s6_addr32[1] = 0;
1083 		llsol.s6_addr32[2] = htonl(1);
1084 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1085 		llsol.s6_addr8[12] = 0xff;
1086 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1087 			/* XXX: should not happen */
1088 			log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
1089 			goto cleanup;
1090 		}
1091 		dad_delay = 0;
1092 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1093 			/*
1094 			 * We need a random delay for DAD on the address
1095 			 * being configured.  It also means delaying
1096 			 * transmission of the corresponding MLD report to
1097 			 * avoid report collision.
1098 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1099 			 */
1100 			dad_delay = cprng_fast32() %
1101 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1102 		}
1103 
1104 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1105 		/* join solicited multicast addr for new host id */
1106 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
1107 		if (!imm) {
1108 			nd6log(LOG_ERR,
1109 			    "addmulti failed for %s on %s (errno=%d)\n",
1110 			    ip6_sprintf(&llsol), if_name(ifp), error);
1111 			goto cleanup;
1112 		}
1113 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1114 		in6m_sol = imm->i6mm_maddr;
1115 
1116 		sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
1117 
1118 		/*
1119 		 * join link-local all-nodes address
1120 		 */
1121 		sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
1122 		    0, 0, 0);
1123 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1124 			goto cleanup; /* XXX: should not fail */
1125 
1126 		/*
1127 		 * XXX: do we really need this automatic routes?
1128 		 * We should probably reconsider this stuff.  Most applications
1129 		 * actually do not need the routes, since they usually specify
1130 		 * the outgoing interface.
1131 		 */
1132 		rt = rtalloc1(sin6tosa(&mltaddr), 0);
1133 		if (rt) {
1134 			if (memcmp(&mltaddr.sin6_addr,
1135 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1136 			    MLTMASK_LEN)) {
1137 				rtfree(rt);
1138 				rt = NULL;
1139 			} else if (rt->rt_ifp != ifp) {
1140 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1141 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1142 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1143 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1144 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1145 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1146 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1147 				rt_replace_ifa(rt, &ia->ia_ifa);
1148 				rt->rt_ifp = ifp;
1149 			}
1150 		}
1151 		if (!rt) {
1152 			struct rt_addrinfo info;
1153 
1154 			memset(&info, 0, sizeof(info));
1155 			info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
1156 			info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
1157 			info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
1158 			info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
1159 			/* XXX: we need RTF_CONNECTED to fake nd6_rtrequest */
1160 			info.rti_flags = RTF_UP | RTF_CONNECTED;
1161 			error = rtrequest1(RTM_ADD, &info, NULL);
1162 			if (error)
1163 				goto cleanup;
1164 		} else {
1165 			rtfree(rt);
1166 		}
1167 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1168 		if (!imm) {
1169 			nd6log(LOG_WARNING,
1170 			    "addmulti failed for %s on %s (errno=%d)\n",
1171 			    ip6_sprintf(&mltaddr.sin6_addr),
1172 			    if_name(ifp), error);
1173 			goto cleanup;
1174 		}
1175 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1176 
1177 		/*
1178 		 * join node information group address
1179 		 */
1180 		dad_delay = 0;
1181 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1182 			/*
1183 			 * The spec doesn't say anything about delay for this
1184 			 * group, but the same logic should apply.
1185 			 */
1186 			dad_delay = cprng_fast32() %
1187 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1188 		}
1189 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
1190 			;
1191 		else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1192 		          dad_delay)) == NULL) { /* XXX jinmei */
1193 			nd6log(LOG_WARNING,
1194 			    "addmulti failed for %s on %s (errno=%d)\n",
1195 			    ip6_sprintf(&mltaddr.sin6_addr),
1196 			    if_name(ifp), error);
1197 			/* XXX not very fatal, go on... */
1198 		} else {
1199 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1200 		}
1201 
1202 
1203 		/*
1204 		 * join interface-local all-nodes address.
1205 		 * (ff01::1%ifN, and ff01::%ifN/32)
1206 		 */
1207 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1208 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1209 			goto cleanup; /* XXX: should not fail */
1210 
1211 		/* XXX: again, do we really need the route? */
1212 		rt = rtalloc1(sin6tosa(&mltaddr), 0);
1213 		if (rt) {
1214 			/* 32bit came from "mltmask" */
1215 			if (memcmp(&mltaddr.sin6_addr,
1216 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1217 			    32 / NBBY)) {
1218 				rtfree(rt);
1219 				rt = NULL;
1220 			} else if (rt->rt_ifp != ifp) {
1221 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1222 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1223 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1224 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1225 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1226 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1227 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1228 				rt_replace_ifa(rt, &ia->ia_ifa);
1229 				rt->rt_ifp = ifp;
1230 			}
1231 		}
1232 		if (!rt) {
1233 			struct rt_addrinfo info;
1234 
1235 			memset(&info, 0, sizeof(info));
1236 			info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
1237 			info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
1238 			info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
1239 			info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
1240 			info.rti_flags = RTF_UP | RTF_CONNECTED;
1241 			error = rtrequest1(RTM_ADD, &info, NULL);
1242 			if (error)
1243 				goto cleanup;
1244 #undef	MLTMASK_LEN
1245 		} else {
1246 			rtfree(rt);
1247 		}
1248 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1249 		if (!imm) {
1250 			nd6log(LOG_WARNING,
1251 			    "addmulti failed for %s on %s (errno=%d)\n",
1252 			    ip6_sprintf(&mltaddr.sin6_addr),
1253 			    if_name(ifp), error);
1254 			goto cleanup;
1255 		} else {
1256 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1257 		}
1258 	}
1259 
1260 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1261 	error = nd6_add_ifa_lle(ia);
1262 	if (error != 0)
1263 		goto cleanup;
1264 
1265 	/*
1266 	 * Perform DAD, if needed.
1267 	 * XXX It may be of use, if we can administratively
1268 	 * disable DAD.
1269 	 */
1270 	if (hostIsNew && if_do_dad(ifp) &&
1271 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1272 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1273 	{
1274 		int mindelay, maxdelay;
1275 
1276 		dad_delay = 0;
1277 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1278 			/*
1279 			 * We need to impose a delay before sending an NS
1280 			 * for DAD.  Check if we also needed a delay for the
1281 			 * corresponding MLD message.  If we did, the delay
1282 			 * should be larger than the MLD delay (this could be
1283 			 * relaxed a bit, but this simple logic is at least
1284 			 * safe).
1285 			 */
1286 			mindelay = 0;
1287 			if (in6m_sol != NULL &&
1288 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1289 				mindelay = in6m_sol->in6m_timer;
1290 			}
1291 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1292 			if (maxdelay - mindelay == 0)
1293 				dad_delay = 0;
1294 			else {
1295 				dad_delay =
1296 				    (cprng_fast32() % (maxdelay - mindelay)) +
1297 				    mindelay;
1298 			}
1299 		}
1300 		/* +1 ensures callout is always used */
1301 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1302 	}
1303 
1304 	return 0;
1305 
1306   cleanup:
1307 	in6_purgeaddr(&ia->ia_ifa);
1308   exit:
1309 	return error;
1310 }
1311 
1312 int
1313 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1314     struct in6_ifaddr *ia, int flags)
1315 {
1316 	int rc, s;
1317 
1318 	s = splnet();
1319 	rc = in6_update_ifa1(ifp, ifra, ia, flags);
1320 	splx(s);
1321 	return rc;
1322 }
1323 
1324 void
1325 in6_purgeaddr(struct ifaddr *ifa)
1326 {
1327 	struct ifnet *ifp = ifa->ifa_ifp;
1328 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1329 	struct in6_multi_mship *imm;
1330 
1331 	/* stop DAD processing */
1332 	nd6_dad_stop(ifa);
1333 
1334 	/* Delete any network route. */
1335 	in6_ifremprefix(ia);
1336 
1337 	/* Remove ownaddr's loopback rtentry, if it exists. */
1338 	in6_ifremlocal(&(ia->ia_ifa));
1339 
1340 	/*
1341 	 * leave from multicast groups we have joined for the interface
1342 	 */
1343 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1344 		LIST_REMOVE(imm, i6mm_chain);
1345 		in6_leavegroup(imm);
1346 	}
1347 
1348 	in6_unlink_ifa(ia, ifp);
1349 }
1350 
1351 static void
1352 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1353 {
1354 	int	s = splnet();
1355 
1356 	mutex_enter(&in6_ifaddr_lock);
1357 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1358 	ifa_remove(ifp, &ia->ia_ifa);
1359 	mutex_exit(&in6_ifaddr_lock);
1360 
1361 	/*
1362 	 * XXX thorpej@NetBSD.org -- if the interface is going
1363 	 * XXX away, don't save the multicast entries, delete them!
1364 	 */
1365 	if (LIST_EMPTY(&ia->ia6_multiaddrs))
1366 		;
1367 	else if (if_is_deactivated(ia->ia_ifa.ifa_ifp)) {
1368 		struct in6_multi *in6m, *next;
1369 
1370 		for (in6m = LIST_FIRST(&ia->ia6_multiaddrs); in6m != NULL;
1371 		     in6m = next) {
1372 			next = LIST_NEXT(in6m, in6m_entry);
1373 			in6_delmulti(in6m);
1374 		}
1375 	} else
1376 		in6_savemkludge(ia);
1377 
1378 	/*
1379 	 * Release the reference to the ND prefix.
1380 	 */
1381 	if (ia->ia6_ndpr != NULL) {
1382 		ia->ia6_ndpr->ndpr_refcnt--;
1383 		ia->ia6_ndpr = NULL;
1384 	}
1385 
1386 	/*
1387 	 * Also, if the address being removed is autoconf'ed, call
1388 	 * pfxlist_onlink_check() since the release might affect the status of
1389 	 * other (detached) addresses.
1390 	 */
1391 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0)
1392 		pfxlist_onlink_check();
1393 
1394 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1395 
1396 	/*
1397 	 * release another refcnt for the link from in6_ifaddr.
1398 	 * Note that we should decrement the refcnt at least once for all *BSD.
1399 	 */
1400 	ifafree(&ia->ia_ifa);
1401 
1402 	splx(s);
1403 }
1404 
1405 void
1406 in6_purgeif(struct ifnet *ifp)
1407 {
1408 
1409 	in6_ifdetach(ifp);
1410 }
1411 
1412 /*
1413  * SIOC[GAD]LIFADDR.
1414  *	SIOCGLIFADDR: get first address. (?)
1415  *	SIOCGLIFADDR with IFLR_PREFIX:
1416  *		get first address that matches the specified prefix.
1417  *	SIOCALIFADDR: add the specified address.
1418  *	SIOCALIFADDR with IFLR_PREFIX:
1419  *		add the specified prefix, filling hostid part from
1420  *		the first link-local address.  prefixlen must be <= 64.
1421  *	SIOCDLIFADDR: delete the specified address.
1422  *	SIOCDLIFADDR with IFLR_PREFIX:
1423  *		delete the first address that matches the specified prefix.
1424  * return values:
1425  *	EINVAL on invalid parameters
1426  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1427  *	other values may be returned from in6_ioctl()
1428  *
1429  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1430  * this is to accommodate address naming scheme other than RFC2374,
1431  * in the future.
1432  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1433  * address encoding scheme. (see figure on page 8)
1434  */
1435 static int
1436 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1437 	struct ifnet *ifp)
1438 {
1439 	struct in6_ifaddr *ia = NULL; /* XXX gcc 4.8 maybe-uninitialized */
1440 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1441 	struct ifaddr *ifa;
1442 	struct sockaddr *sa;
1443 
1444 	/* sanity checks */
1445 	if (!data || !ifp) {
1446 		panic("invalid argument to in6_lifaddr_ioctl");
1447 		/* NOTREACHED */
1448 	}
1449 
1450 	switch (cmd) {
1451 	case SIOCGLIFADDR:
1452 		/* address must be specified on GET with IFLR_PREFIX */
1453 		if ((iflr->flags & IFLR_PREFIX) == 0)
1454 			break;
1455 		/* FALLTHROUGH */
1456 	case SIOCALIFADDR:
1457 	case SIOCDLIFADDR:
1458 		/* address must be specified on ADD and DELETE */
1459 		sa = (struct sockaddr *)&iflr->addr;
1460 		if (sa->sa_family != AF_INET6)
1461 			return EINVAL;
1462 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1463 			return EINVAL;
1464 		/* XXX need improvement */
1465 		sa = (struct sockaddr *)&iflr->dstaddr;
1466 		if (sa->sa_family && sa->sa_family != AF_INET6)
1467 			return EINVAL;
1468 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1469 			return EINVAL;
1470 		break;
1471 	default: /* shouldn't happen */
1472 #if 0
1473 		panic("invalid cmd to in6_lifaddr_ioctl");
1474 		/* NOTREACHED */
1475 #else
1476 		return EOPNOTSUPP;
1477 #endif
1478 	}
1479 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1480 		return EINVAL;
1481 
1482 	switch (cmd) {
1483 	case SIOCALIFADDR:
1484 	    {
1485 		struct in6_aliasreq ifra;
1486 		struct in6_addr *xhostid = NULL;
1487 		int prefixlen;
1488 		int bound = curlwp_bind();
1489 		struct psref psref;
1490 
1491 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1492 			struct sockaddr_in6 *sin6;
1493 
1494 			/*
1495 			 * xhostid is to fill in the hostid part of the
1496 			 * address.  xhostid points to the first link-local
1497 			 * address attached to the interface.
1498 			 */
1499 			ia = in6ifa_ifpforlinklocal_psref(ifp, 0, &psref);
1500 			if (ia == NULL) {
1501 				curlwp_bindx(bound);
1502 				return EADDRNOTAVAIL;
1503 			}
1504 			xhostid = IFA_IN6(&ia->ia_ifa);
1505 
1506 		 	/* prefixlen must be <= 64. */
1507 			if (64 < iflr->prefixlen) {
1508 				ia6_release(ia, &psref);
1509 				curlwp_bindx(bound);
1510 				return EINVAL;
1511 			}
1512 			prefixlen = iflr->prefixlen;
1513 
1514 			/* hostid part must be zero. */
1515 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1516 			if (sin6->sin6_addr.s6_addr32[2] != 0
1517 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1518 				ia6_release(ia, &psref);
1519 				curlwp_bindx(bound);
1520 				return EINVAL;
1521 			}
1522 		} else
1523 			prefixlen = iflr->prefixlen;
1524 
1525 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1526 		memset(&ifra, 0, sizeof(ifra));
1527 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1528 
1529 		memcpy(&ifra.ifra_addr, &iflr->addr,
1530 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1531 		if (xhostid) {
1532 			/* fill in hostid part */
1533 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1534 			    xhostid->s6_addr32[2];
1535 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1536 			    xhostid->s6_addr32[3];
1537 		}
1538 
1539 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1540 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1541 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1542 			if (xhostid) {
1543 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1544 				    xhostid->s6_addr32[2];
1545 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1546 				    xhostid->s6_addr32[3];
1547 			}
1548 		}
1549 		if (xhostid) {
1550 			ia6_release(ia, &psref);
1551 			ia = NULL;
1552 		}
1553 		curlwp_bindx(bound);
1554 
1555 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1556 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1557 
1558 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1559 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1560 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1561 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1562 	    }
1563 	case SIOCGLIFADDR:
1564 	case SIOCDLIFADDR:
1565 	    {
1566 		struct in6_addr mask, candidate, match;
1567 		struct sockaddr_in6 *sin6;
1568 		int cmp;
1569 		int error, s;
1570 
1571 		memset(&mask, 0, sizeof(mask));
1572 		if (iflr->flags & IFLR_PREFIX) {
1573 			/* lookup a prefix rather than address. */
1574 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1575 
1576 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1577 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1578 			match.s6_addr32[0] &= mask.s6_addr32[0];
1579 			match.s6_addr32[1] &= mask.s6_addr32[1];
1580 			match.s6_addr32[2] &= mask.s6_addr32[2];
1581 			match.s6_addr32[3] &= mask.s6_addr32[3];
1582 
1583 			/* if you set extra bits, that's wrong */
1584 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1585 				return EINVAL;
1586 
1587 			cmp = 1;
1588 		} else {
1589 			if (cmd == SIOCGLIFADDR) {
1590 				/* on getting an address, take the 1st match */
1591 				cmp = 0;	/* XXX */
1592 			} else {
1593 				/* on deleting an address, do exact match */
1594 				in6_prefixlen2mask(&mask, 128);
1595 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1596 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1597 
1598 				cmp = 1;
1599 			}
1600 		}
1601 
1602 		s = pserialize_read_enter();
1603 		IFADDR_READER_FOREACH(ifa, ifp) {
1604 			if (ifa->ifa_addr->sa_family != AF_INET6)
1605 				continue;
1606 			if (!cmp)
1607 				break;
1608 
1609 			/*
1610 			 * XXX: this is adhoc, but is necessary to allow
1611 			 * a user to specify fe80::/64 (not /10) for a
1612 			 * link-local address.
1613 			 */
1614 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1615 			in6_clearscope(&candidate);
1616 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1617 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1618 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1619 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1620 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1621 				break;
1622 		}
1623 		if (!ifa) {
1624 			error = EADDRNOTAVAIL;
1625 			goto error;
1626 		}
1627 		ia = ifa2ia6(ifa);
1628 
1629 		if (cmd == SIOCGLIFADDR) {
1630 			/* fill in the if_laddrreq structure */
1631 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1632 			error = sa6_recoverscope(
1633 			    (struct sockaddr_in6 *)&iflr->addr);
1634 			if (error != 0)
1635 				goto error;
1636 
1637 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1638 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1639 				    ia->ia_dstaddr.sin6_len);
1640 				error = sa6_recoverscope(
1641 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1642 				if (error != 0)
1643 					goto error;
1644 			} else
1645 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1646 
1647 			iflr->prefixlen =
1648 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1649 
1650 			iflr->flags = ia->ia6_flags;	/* XXX */
1651 
1652 			error = 0;
1653 		} else {
1654 			struct in6_aliasreq ifra;
1655 
1656 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1657 			memset(&ifra, 0, sizeof(ifra));
1658 			memcpy(ifra.ifra_name, iflr->iflr_name,
1659 			    sizeof(ifra.ifra_name));
1660 
1661 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1662 			    ia->ia_addr.sin6_len);
1663 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1664 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1665 				    ia->ia_dstaddr.sin6_len);
1666 			} else {
1667 				memset(&ifra.ifra_dstaddr, 0,
1668 				    sizeof(ifra.ifra_dstaddr));
1669 			}
1670 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1671 			    ia->ia_prefixmask.sin6_len);
1672 
1673 			ifra.ifra_flags = ia->ia6_flags;
1674 			pserialize_read_exit(s);
1675 
1676 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1677 		}
1678 	error:
1679 		pserialize_read_exit(s);
1680 		return error;
1681 	    }
1682 	}
1683 
1684 	return EOPNOTSUPP;	/* just for safety */
1685 }
1686 
1687 /*
1688  * Initialize an interface's internet6 address
1689  * and routing table entry.
1690  */
1691 static int
1692 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1693 	const struct sockaddr_in6 *sin6, int newhost)
1694 {
1695 	int	error = 0, ifacount = 0;
1696 	int	s = splnet();
1697 	struct ifaddr *ifa;
1698 
1699 	/*
1700 	 * Give the interface a chance to initialize
1701 	 * if this is its first address,
1702 	 * and to validate the address if necessary.
1703 	 */
1704 	IFADDR_READER_FOREACH(ifa, ifp) {
1705 		if (ifa->ifa_addr->sa_family != AF_INET6)
1706 			continue;
1707 		ifacount++;
1708 	}
1709 
1710 	ia->ia_addr = *sin6;
1711 
1712 	if (ifacount <= 0 &&
1713 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1714 		splx(s);
1715 		return error;
1716 	}
1717 	splx(s);
1718 
1719 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1720 
1721 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1722 
1723 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1724 	if (newhost) {
1725 		/* set the rtrequest function to create llinfo */
1726 		if (ifp->if_flags & IFF_POINTOPOINT)
1727 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1728 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1729 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1730 		in6_ifaddlocal(&ia->ia_ifa);
1731 	} else {
1732 		/* Inform the routing socket of new flags/timings */
1733 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
1734 	}
1735 
1736 	/* Add the network prefix route. */
1737 	if ((error = in6_ifaddprefix(ia)) != 0) {
1738 		if (newhost)
1739 			in6_ifremlocal(&ia->ia_ifa);
1740 		return error;
1741 	}
1742 
1743 	if (ifp->if_flags & IFF_MULTICAST)
1744 		in6_restoremkludge(ia, ifp);
1745 
1746 	return error;
1747 }
1748 
1749 static struct ifaddr *
1750 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1751 {
1752 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1753 		return ifa;
1754 	return best_ifa;
1755 }
1756 
1757 /*
1758  * Find an IPv6 interface link-local address specific to an interface.
1759  */
1760 struct in6_ifaddr *
1761 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1762 {
1763 	struct ifaddr *best_ifa = NULL, *ifa;
1764 
1765 	IFADDR_READER_FOREACH(ifa, ifp) {
1766 		if (ifa->ifa_addr->sa_family != AF_INET6)
1767 			continue;
1768 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1769 			continue;
1770 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1771 			continue;
1772 		best_ifa = bestifa(best_ifa, ifa);
1773 	}
1774 
1775 	return (struct in6_ifaddr *)best_ifa;
1776 }
1777 
1778 struct in6_ifaddr *
1779 in6ifa_ifpforlinklocal_psref(const struct ifnet *ifp, const int ignoreflags,
1780     struct psref *psref)
1781 {
1782 	struct in6_ifaddr *ia;
1783 	int s = pserialize_read_enter();
1784 
1785 	ia = in6ifa_ifpforlinklocal(ifp, ignoreflags);
1786 	if (ia != NULL)
1787 		ia6_acquire(ia, psref);
1788 	pserialize_read_exit(s);
1789 
1790 	return ia;
1791 }
1792 
1793 /*
1794  * find the internet address corresponding to a given address.
1795  * ifaddr is returned referenced.
1796  */
1797 struct in6_ifaddr *
1798 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1799 {
1800 	struct in6_ifaddr *ia;
1801 	int s;
1802 
1803 	s = pserialize_read_enter();
1804 	IN6_ADDRLIST_READER_FOREACH(ia) {
1805 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1806 			if (zoneid != 0 &&
1807 			    zoneid != ia->ia_addr.sin6_scope_id)
1808 				continue;
1809 			ifaref(&ia->ia_ifa);
1810 			break;
1811 		}
1812 	}
1813 	pserialize_read_exit(s);
1814 
1815 	return ia;
1816 }
1817 
1818 /*
1819  * find the internet address corresponding to a given interface and address.
1820  */
1821 struct in6_ifaddr *
1822 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1823 {
1824 	struct ifaddr *best_ifa = NULL, *ifa;
1825 
1826 	IFADDR_READER_FOREACH(ifa, ifp) {
1827 		if (ifa->ifa_addr->sa_family != AF_INET6)
1828 			continue;
1829 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1830 			continue;
1831 		best_ifa = bestifa(best_ifa, ifa);
1832 	}
1833 
1834 	return (struct in6_ifaddr *)best_ifa;
1835 }
1836 
1837 struct in6_ifaddr *
1838 in6ifa_ifpwithaddr_psref(const struct ifnet *ifp, const struct in6_addr *addr,
1839     struct psref *psref)
1840 {
1841 	struct in6_ifaddr *ia;
1842 	int s = pserialize_read_enter();
1843 
1844 	ia = in6ifa_ifpwithaddr(ifp, addr);
1845 	if (ia != NULL)
1846 		ia6_acquire(ia, psref);
1847 	pserialize_read_exit(s);
1848 
1849 	return ia;
1850 }
1851 
1852 static struct in6_ifaddr *
1853 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1854 {
1855 	if (best_ia == NULL ||
1856 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1857 		return ia;
1858 	return best_ia;
1859 }
1860 
1861 /*
1862  * Convert IP6 address to printable (loggable) representation.
1863  */
1864 char *
1865 ip6_sprintf(const struct in6_addr *addr)
1866 {
1867 	static int ip6round = 0;
1868 	static char ip6buf[8][INET6_ADDRSTRLEN];
1869 	char *cp = ip6buf[ip6round++ & 7];
1870 
1871 	in6_print(cp, INET6_ADDRSTRLEN, addr);
1872 	return cp;
1873 }
1874 
1875 /*
1876  * Determine if an address is on a local network.
1877  */
1878 int
1879 in6_localaddr(const struct in6_addr *in6)
1880 {
1881 	struct in6_ifaddr *ia;
1882 	int s;
1883 
1884 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1885 		return 1;
1886 
1887 	s = pserialize_read_enter();
1888 	IN6_ADDRLIST_READER_FOREACH(ia) {
1889 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1890 					      &ia->ia_prefixmask.sin6_addr)) {
1891 			pserialize_read_exit(s);
1892 			return 1;
1893 		}
1894 	}
1895 	pserialize_read_exit(s);
1896 
1897 	return 0;
1898 }
1899 
1900 int
1901 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1902 {
1903 	struct in6_ifaddr *ia;
1904 	int s;
1905 
1906 	s = pserialize_read_enter();
1907 	IN6_ADDRLIST_READER_FOREACH(ia) {
1908 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1909 		    &sa6->sin6_addr) &&
1910 #ifdef SCOPEDROUTING
1911 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1912 #endif
1913 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
1914 			pserialize_read_exit(s);
1915 			return 1; /* true */
1916 		}
1917 
1918 		/* XXX: do we still have to go thru the rest of the list? */
1919 	}
1920 	pserialize_read_exit(s);
1921 
1922 	return 0;		/* false */
1923 }
1924 
1925 /*
1926  * return length of part which dst and src are equal
1927  * hard coding...
1928  */
1929 int
1930 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1931 {
1932 	int match = 0;
1933 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1934 	u_char *lim = s + 16, r;
1935 
1936 	while (s < lim)
1937 		if ((r = (*d++ ^ *s++)) != 0) {
1938 			while (r < 128) {
1939 				match++;
1940 				r <<= 1;
1941 			}
1942 			break;
1943 		} else
1944 			match += NBBY;
1945 	return match;
1946 }
1947 
1948 /* XXX: to be scope conscious */
1949 int
1950 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1951 {
1952 	int bytelen, bitlen;
1953 
1954 	/* sanity check */
1955 	if (len < 0 || len > 128) {
1956 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1957 		    len);
1958 		return 0;
1959 	}
1960 
1961 	bytelen = len / NBBY;
1962 	bitlen = len % NBBY;
1963 
1964 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1965 		return 0;
1966 	if (bitlen != 0 &&
1967 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
1968 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
1969 		return 0;
1970 
1971 	return 1;
1972 }
1973 
1974 void
1975 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1976 {
1977 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1978 	int bytelen, bitlen, i;
1979 
1980 	/* sanity check */
1981 	if (len < 0 || len > 128) {
1982 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1983 		    len);
1984 		return;
1985 	}
1986 
1987 	memset(maskp, 0, sizeof(*maskp));
1988 	bytelen = len / NBBY;
1989 	bitlen = len % NBBY;
1990 	for (i = 0; i < bytelen; i++)
1991 		maskp->s6_addr[i] = 0xff;
1992 	if (bitlen)
1993 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1994 }
1995 
1996 /*
1997  * return the best address out of the same scope. if no address was
1998  * found, return the first valid address from designated IF.
1999  */
2000 struct in6_ifaddr *
2001 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2002 {
2003 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2004 	struct ifaddr *ifa;
2005 	struct in6_ifaddr *best_ia = NULL, *ia;
2006 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2007 
2008 	dep[0] = dep[1] = NULL;
2009 
2010 	/*
2011 	 * We first look for addresses in the same scope.
2012 	 * If there is one, return it.
2013 	 * If two or more, return one which matches the dst longest.
2014 	 * If none, return one of global addresses assigned other ifs.
2015 	 */
2016 	IFADDR_READER_FOREACH(ifa, ifp) {
2017 		if (ifa->ifa_addr->sa_family != AF_INET6)
2018 			continue;
2019 		ia = (struct in6_ifaddr *)ifa;
2020 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2021 			continue; /* XXX: is there any case to allow anycast? */
2022 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2023 			continue; /* don't use this interface */
2024 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2025 			continue;
2026 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2027 			if (ip6_use_deprecated)
2028 				dep[0] = ia;
2029 			continue;
2030 		}
2031 
2032 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
2033 			continue;
2034 		/*
2035 		 * call in6_matchlen() as few as possible
2036 		 */
2037 		if (best_ia == NULL) {
2038 			best_ia = ia;
2039 			continue;
2040 		}
2041 		if (blen == -1)
2042 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2043 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2044 		if (tlen > blen) {
2045 			blen = tlen;
2046 			best_ia = ia;
2047 		} else if (tlen == blen)
2048 			best_ia = bestia(best_ia, ia);
2049 	}
2050 	if (best_ia != NULL)
2051 		return best_ia;
2052 
2053 	IFADDR_READER_FOREACH(ifa, ifp) {
2054 		if (ifa->ifa_addr->sa_family != AF_INET6)
2055 			continue;
2056 		ia = (struct in6_ifaddr *)ifa;
2057 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2058 			continue; /* XXX: is there any case to allow anycast? */
2059 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2060 			continue; /* don't use this interface */
2061 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2062 			continue;
2063 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2064 			if (ip6_use_deprecated)
2065 				dep[1] = (struct in6_ifaddr *)ifa;
2066 			continue;
2067 		}
2068 
2069 		best_ia = bestia(best_ia, ia);
2070 	}
2071 	if (best_ia != NULL)
2072 		return best_ia;
2073 
2074 	/* use the last-resort values, that are, deprecated addresses */
2075 	if (dep[0])
2076 		return dep[0];
2077 	if (dep[1])
2078 		return dep[1];
2079 
2080 	return NULL;
2081 }
2082 
2083 /*
2084  * perform DAD when interface becomes IFF_UP.
2085  */
2086 void
2087 in6_if_link_up(struct ifnet *ifp)
2088 {
2089 	struct ifaddr *ifa;
2090 	struct in6_ifaddr *ia;
2091 	int s, bound;
2092 
2093 	/* Ensure it's sane to run DAD */
2094 	if (ifp->if_link_state == LINK_STATE_DOWN)
2095 		return;
2096 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2097 		return;
2098 
2099 	bound = curlwp_bind();
2100 	s = pserialize_read_enter();
2101 	IFADDR_READER_FOREACH(ifa, ifp) {
2102 		struct psref psref;
2103 
2104 		if (ifa->ifa_addr->sa_family != AF_INET6)
2105 			continue;
2106 
2107 		ifa_acquire(ifa, &psref);
2108 		pserialize_read_exit(s);
2109 		ia = (struct in6_ifaddr *)ifa;
2110 
2111 		/* If detached then mark as tentative */
2112 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2113 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2114 			if (if_do_dad(ifp)) {
2115 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2116 				nd6log(LOG_ERR, "%s marked tentative\n",
2117 				    ip6_sprintf(&ia->ia_addr.sin6_addr));
2118 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2119 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2120 		}
2121 
2122 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2123 			int rand_delay;
2124 
2125 			/* Clear the duplicated flag as we're starting DAD. */
2126 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2127 
2128 			/*
2129 			 * The TENTATIVE flag was likely set by hand
2130 			 * beforehand, implicitly indicating the need for DAD.
2131 			 * We may be able to skip the random delay in this
2132 			 * case, but we impose delays just in case.
2133 			 */
2134 			rand_delay = cprng_fast32() %
2135 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2136 			/* +1 ensures callout is always used */
2137 			nd6_dad_start(ifa, rand_delay + 1);
2138 		}
2139 
2140 		s = pserialize_read_enter();
2141 		ifa_release(ifa, &psref);
2142 	}
2143 	pserialize_read_exit(s);
2144 	curlwp_bindx(bound);
2145 
2146 	/* Restore any detached prefixes */
2147 	pfxlist_onlink_check();
2148 }
2149 
2150 void
2151 in6_if_up(struct ifnet *ifp)
2152 {
2153 
2154 	/*
2155 	 * special cases, like 6to4, are handled in in6_ifattach
2156 	 */
2157 	in6_ifattach(ifp, NULL);
2158 
2159 	/* interface may not support link state, so bring it up also */
2160 	in6_if_link_up(ifp);
2161 }
2162 
2163 /*
2164  * Mark all addresses as detached.
2165  */
2166 void
2167 in6_if_link_down(struct ifnet *ifp)
2168 {
2169 	struct ifaddr *ifa;
2170 	struct in6_ifaddr *ia;
2171 	int s, bound;
2172 
2173 	/* Any prefixes on this interface should be detached as well */
2174 	pfxlist_onlink_check();
2175 
2176 	bound = curlwp_bind();
2177 	s = pserialize_read_enter();
2178 	IFADDR_READER_FOREACH(ifa, ifp) {
2179 		struct psref psref;
2180 
2181 		if (ifa->ifa_addr->sa_family != AF_INET6)
2182 			continue;
2183 
2184 		ifa_acquire(ifa, &psref);
2185 		pserialize_read_exit(s);
2186 		ia = (struct in6_ifaddr *)ifa;
2187 
2188 		/* Stop DAD processing */
2189 		nd6_dad_stop(ifa);
2190 
2191 		/*
2192 		 * Mark the address as detached.
2193 		 * This satisfies RFC4862 Section 5.3, but we should apply
2194 		 * this logic to all addresses to be a good citizen and
2195 		 * avoid potential duplicated addresses.
2196 		 * When the interface comes up again, detached addresses
2197 		 * are marked tentative and DAD commences.
2198 		 */
2199 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2200 			nd6log(LOG_DEBUG, "%s marked detached\n",
2201 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
2202 			ia->ia6_flags |= IN6_IFF_DETACHED;
2203 			ia->ia6_flags &=
2204 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2205 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2206 		}
2207 
2208 		s = pserialize_read_enter();
2209 		ifa_release(ifa, &psref);
2210 	}
2211 	pserialize_read_exit(s);
2212 	curlwp_bindx(bound);
2213 }
2214 
2215 void
2216 in6_if_down(struct ifnet *ifp)
2217 {
2218 
2219 	in6_if_link_down(ifp);
2220 }
2221 
2222 void
2223 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2224 {
2225 
2226 	switch (link_state) {
2227 	case LINK_STATE_DOWN:
2228 		in6_if_link_down(ifp);
2229 		break;
2230 	case LINK_STATE_UP:
2231 		in6_if_link_up(ifp);
2232 		break;
2233 	}
2234 }
2235 
2236 /*
2237  * Calculate max IPv6 MTU through all the interfaces and store it
2238  * to in6_maxmtu.
2239  */
2240 void
2241 in6_setmaxmtu(void)
2242 {
2243 	unsigned long maxmtu = 0;
2244 	struct ifnet *ifp;
2245 	int s;
2246 
2247 	s = pserialize_read_enter();
2248 	IFNET_READER_FOREACH(ifp) {
2249 		/* this function can be called during ifnet initialization */
2250 		if (!ifp->if_afdata[AF_INET6])
2251 			continue;
2252 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2253 		    IN6_LINKMTU(ifp) > maxmtu)
2254 			maxmtu = IN6_LINKMTU(ifp);
2255 	}
2256 	pserialize_read_exit(s);
2257 	if (maxmtu)	     /* update only when maxmtu is positive */
2258 		in6_maxmtu = maxmtu;
2259 }
2260 
2261 /*
2262  * Provide the length of interface identifiers to be used for the link attached
2263  * to the given interface.  The length should be defined in "IPv6 over
2264  * xxx-link" document.  Note that address architecture might also define
2265  * the length for a particular set of address prefixes, regardless of the
2266  * link type.  As clarified in rfc2462bis, those two definitions should be
2267  * consistent, and those really are as of August 2004.
2268  */
2269 int
2270 in6_if2idlen(struct ifnet *ifp)
2271 {
2272 	switch (ifp->if_type) {
2273 	case IFT_ETHER:		/* RFC2464 */
2274 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2275 	case IFT_L2VLAN:	/* ditto */
2276 	case IFT_IEEE80211:	/* ditto */
2277 	case IFT_FDDI:		/* RFC2467 */
2278 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2279 	case IFT_PPP:		/* RFC2472 */
2280 	case IFT_ARCNET:	/* RFC2497 */
2281 	case IFT_FRELAY:	/* RFC2590 */
2282 	case IFT_IEEE1394:	/* RFC3146 */
2283 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2284 	case IFT_LOOP:		/* XXX: is this really correct? */
2285 		return 64;
2286 	default:
2287 		/*
2288 		 * Unknown link type:
2289 		 * It might be controversial to use the today's common constant
2290 		 * of 64 for these cases unconditionally.  For full compliance,
2291 		 * we should return an error in this case.  On the other hand,
2292 		 * if we simply miss the standard for the link type or a new
2293 		 * standard is defined for a new link type, the IFID length
2294 		 * is very likely to be the common constant.  As a compromise,
2295 		 * we always use the constant, but make an explicit notice
2296 		 * indicating the "unknown" case.
2297 		 */
2298 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2299 		return 64;
2300 	}
2301 }
2302 
2303 struct in6_llentry {
2304 	struct llentry		base;
2305 };
2306 
2307 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2308 #define	IN6_LLTBL_HASH(k, h) \
2309 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2310 
2311 /*
2312  * Do actual deallocation of @lle.
2313  * Called by LLE_FREE_LOCKED when number of references
2314  * drops to zero.
2315  */
2316 static void
2317 in6_lltable_destroy_lle(struct llentry *lle)
2318 {
2319 
2320 	LLE_WUNLOCK(lle);
2321 	LLE_LOCK_DESTROY(lle);
2322 	kmem_intr_free(lle, sizeof(struct in6_llentry));
2323 }
2324 
2325 static struct llentry *
2326 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2327 {
2328 	struct in6_llentry *lle;
2329 
2330 	lle = kmem_intr_zalloc(sizeof(struct in6_llentry), KM_NOSLEEP);
2331 	if (lle == NULL)		/* NB: caller generates msg */
2332 		return NULL;
2333 
2334 	lle->base.r_l3addr.addr6 = *addr6;
2335 	lle->base.lle_refcnt = 1;
2336 	lle->base.lle_free = in6_lltable_destroy_lle;
2337 	LLE_LOCK_INIT(&lle->base);
2338 	callout_init(&lle->base.lle_timer, CALLOUT_MPSAFE);
2339 
2340 	return &lle->base;
2341 }
2342 
2343 static int
2344 in6_lltable_match_prefix(const struct sockaddr *prefix,
2345     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2346 {
2347 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2348 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2349 
2350 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2351 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2352 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2353 		return 1;
2354 
2355 	return 0;
2356 }
2357 
2358 static void
2359 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2360 {
2361 	struct ifnet *ifp __diagused;
2362 
2363 	LLE_WLOCK_ASSERT(lle);
2364 	KASSERT(llt != NULL);
2365 
2366 	/* Unlink entry from table */
2367 	if ((lle->la_flags & LLE_LINKED) != 0) {
2368 
2369 		ifp = llt->llt_ifp;
2370 		IF_AFDATA_WLOCK_ASSERT(ifp);
2371 		lltable_unlink_entry(llt, lle);
2372 	}
2373 
2374 	KASSERT(mutex_owned(softnet_lock));
2375 	callout_halt(&lle->lle_timer, softnet_lock);
2376 	LLE_REMREF(lle);
2377 
2378 	llentry_free(lle);
2379 }
2380 
2381 static int
2382 in6_lltable_rtcheck(struct ifnet *ifp,
2383 		    u_int flags,
2384 		    const struct sockaddr *l3addr)
2385 {
2386 	struct rtentry *rt;
2387 
2388 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2389 	    "sin_family %d", l3addr->sa_family);
2390 
2391 	rt = rtalloc1(l3addr, 0);
2392 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2393 		int s;
2394 		struct ifaddr *ifa;
2395 		/*
2396 		 * Create an ND6 cache for an IPv6 neighbor
2397 		 * that is not covered by our own prefix.
2398 		 */
2399 		/* XXX ifaof_ifpforaddr should take a const param */
2400 		s = pserialize_read_enter();
2401 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2402 		if (ifa != NULL) {
2403 			pserialize_read_exit(s);
2404 			if (rt != NULL)
2405 				rtfree(rt);
2406 			return 0;
2407 		}
2408 		pserialize_read_exit(s);
2409 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2410 		    ip6_sprintf(&((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2411 		if (rt != NULL)
2412 			rtfree(rt);
2413 		return EINVAL;
2414 	}
2415 	rtfree(rt);
2416 	return 0;
2417 }
2418 
2419 static inline uint32_t
2420 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2421 {
2422 
2423 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2424 }
2425 
2426 static uint32_t
2427 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2428 {
2429 
2430 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2431 }
2432 
2433 static void
2434 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2435 {
2436 	struct sockaddr_in6 *sin6;
2437 
2438 	sin6 = (struct sockaddr_in6 *)sa;
2439 	bzero(sin6, sizeof(*sin6));
2440 	sin6->sin6_family = AF_INET6;
2441 	sin6->sin6_len = sizeof(*sin6);
2442 	sin6->sin6_addr = lle->r_l3addr.addr6;
2443 }
2444 
2445 static inline struct llentry *
2446 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2447 {
2448 	struct llentry *lle;
2449 	struct llentries *lleh;
2450 	u_int hashidx;
2451 
2452 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2453 	lleh = &llt->lle_head[hashidx];
2454 	LIST_FOREACH(lle, lleh, lle_next) {
2455 		if (lle->la_flags & LLE_DELETED)
2456 			continue;
2457 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2458 			break;
2459 	}
2460 
2461 	return lle;
2462 }
2463 
2464 static int
2465 in6_lltable_delete(struct lltable *llt, u_int flags,
2466 	const struct sockaddr *l3addr)
2467 {
2468 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2469 	struct llentry *lle;
2470 
2471 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2472 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2473 	    "sin_family %d", l3addr->sa_family);
2474 
2475 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2476 
2477 	if (lle == NULL)
2478 		return ENOENT;
2479 
2480 	LLE_WLOCK(lle);
2481 	lle->la_flags |= LLE_DELETED;
2482 #ifdef DIAGNOSTIC
2483 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2484 #endif
2485 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2486 		llentry_free(lle);
2487 	else
2488 		LLE_WUNLOCK(lle);
2489 
2490 	return 0;
2491 }
2492 
2493 static struct llentry *
2494 in6_lltable_create(struct lltable *llt, u_int flags,
2495 	const struct sockaddr *l3addr)
2496 {
2497 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2498 	struct ifnet *ifp = llt->llt_ifp;
2499 	struct llentry *lle;
2500 
2501 	IF_AFDATA_WLOCK_ASSERT(ifp);
2502 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2503 	    "sin_family %d", l3addr->sa_family);
2504 
2505 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2506 
2507 	if (lle != NULL) {
2508 		LLE_WLOCK(lle);
2509 		return lle;
2510 	}
2511 
2512 	/*
2513 	 * A route that covers the given address must have
2514 	 * been installed 1st because we are doing a resolution,
2515 	 * verify this.
2516 	 */
2517 	if (!(flags & LLE_IFADDR) &&
2518 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2519 		return NULL;
2520 
2521 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2522 	if (lle == NULL) {
2523 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2524 		return NULL;
2525 	}
2526 	lle->la_flags = flags;
2527 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2528 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2529 		lle->la_flags |= LLE_VALID;
2530 	}
2531 
2532 	lltable_link_entry(llt, lle);
2533 	LLE_WLOCK(lle);
2534 
2535 	return lle;
2536 }
2537 
2538 static struct llentry *
2539 in6_lltable_lookup(struct lltable *llt, u_int flags,
2540 	const struct sockaddr *l3addr)
2541 {
2542 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2543 	struct llentry *lle;
2544 
2545 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2546 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2547 	    "sin_family %d", l3addr->sa_family);
2548 
2549 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2550 
2551 	if (lle == NULL)
2552 		return NULL;
2553 
2554 	if (flags & LLE_EXCLUSIVE)
2555 		LLE_WLOCK(lle);
2556 	else
2557 		LLE_RLOCK(lle);
2558 	return lle;
2559 }
2560 
2561 static int
2562 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2563     struct rt_walkarg *w)
2564 {
2565 	struct sockaddr_in6 sin6;
2566 
2567 	LLTABLE_LOCK_ASSERT();
2568 
2569 	/* skip deleted entries */
2570 	if (lle->la_flags & LLE_DELETED)
2571 		return 0;
2572 
2573 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2574 
2575 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2576 }
2577 
2578 static struct lltable *
2579 in6_lltattach(struct ifnet *ifp)
2580 {
2581 	struct lltable *llt;
2582 
2583 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2584 	llt->llt_af = AF_INET6;
2585 	llt->llt_ifp = ifp;
2586 
2587 	llt->llt_lookup = in6_lltable_lookup;
2588 	llt->llt_create = in6_lltable_create;
2589 	llt->llt_delete = in6_lltable_delete;
2590 	llt->llt_dump_entry = in6_lltable_dump_entry;
2591 	llt->llt_hash = in6_lltable_hash;
2592 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2593 	llt->llt_free_entry = in6_lltable_free_entry;
2594 	llt->llt_match_prefix = in6_lltable_match_prefix;
2595 	lltable_link(llt);
2596 
2597 	return llt;
2598 }
2599 
2600 void *
2601 in6_domifattach(struct ifnet *ifp)
2602 {
2603 	struct in6_ifextra *ext;
2604 
2605 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2606 
2607 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2608 	    M_IFADDR, M_WAITOK|M_ZERO);
2609 
2610 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2611 	    M_IFADDR, M_WAITOK|M_ZERO);
2612 
2613 	ext->nd_ifinfo = nd6_ifattach(ifp);
2614 	ext->scope6_id = scope6_ifattach(ifp);
2615 	ext->nprefixes = 0;
2616 	ext->ndefrouters = 0;
2617 
2618 	ext->lltable = in6_lltattach(ifp);
2619 
2620 	return ext;
2621 }
2622 
2623 void
2624 in6_domifdetach(struct ifnet *ifp, void *aux)
2625 {
2626 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2627 
2628 	lltable_free(ext->lltable);
2629 	ext->lltable = NULL;
2630 	nd6_ifdetach(ifp, ext);
2631 	free(ext->in6_ifstat, M_IFADDR);
2632 	free(ext->icmp6_ifstat, M_IFADDR);
2633 	scope6_ifdetach(ext->scope6_id);
2634 	free(ext, M_IFADDR);
2635 }
2636 
2637 /*
2638  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2639  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2640  */
2641 void
2642 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2643 {
2644 	in6->s6_addr32[0] = 0;
2645 	in6->s6_addr32[1] = 0;
2646 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2647 	in6->s6_addr32[3] = in->s_addr;
2648 }
2649 
2650 /*
2651  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2652  * v4 mapped addr or v4 compat addr
2653  */
2654 void
2655 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2656 {
2657 	memset(sin, 0, sizeof(*sin));
2658 	sin->sin_len = sizeof(struct sockaddr_in);
2659 	sin->sin_family = AF_INET;
2660 	sin->sin_port = sin6->sin6_port;
2661 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2662 }
2663 
2664 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2665 void
2666 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2667 {
2668 	memset(sin6, 0, sizeof(*sin6));
2669 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2670 	sin6->sin6_family = AF_INET6;
2671 	sin6->sin6_port = sin->sin_port;
2672 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2673 }
2674 
2675 /* Convert sockaddr_in6 into sockaddr_in. */
2676 void
2677 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2678 {
2679 	struct sockaddr_in *sin_p;
2680 	struct sockaddr_in6 sin6;
2681 
2682 	/*
2683 	 * Save original sockaddr_in6 addr and convert it
2684 	 * to sockaddr_in.
2685 	 */
2686 	sin6 = *(struct sockaddr_in6 *)nam;
2687 	sin_p = (struct sockaddr_in *)nam;
2688 	in6_sin6_2_sin(sin_p, &sin6);
2689 }
2690 
2691 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2692 void
2693 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2694 {
2695 	struct sockaddr_in *sin_p;
2696 	struct sockaddr_in6 *sin6_p;
2697 
2698 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2699 	sin_p = (struct sockaddr_in *)*nam;
2700 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2701 	free(*nam, M_SONAME);
2702 	*nam = sin6tosa(sin6_p);
2703 }
2704