xref: /netbsd-src/sys/netinet6/in6.c (revision 87d689fb734c654d2486f87f7be32f1b53ecdbec)
1 /*	$NetBSD: in6.c,v 1.257 2018/01/10 10:56:30 knakahara 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 1.257 2018/01/10 10:56:30 knakahara 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 	SOFTNET_LOCK_UNLESS_NET_MPSAFE();
771 	error = in6_control1(so , cmd, data, ifp);
772 	SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();
773 	splx(s);
774 	return error;
775 }
776 
777 static int
778 in6_get_llsol_addr(struct in6_addr *llsol, struct ifnet *ifp,
779     struct in6_addr *ip6)
780 {
781 	int error;
782 
783 	memset(llsol, 0, sizeof(struct in6_addr));
784 	llsol->s6_addr16[0] = htons(0xff02);
785 	llsol->s6_addr32[1] = 0;
786 	llsol->s6_addr32[2] = htonl(1);
787 	llsol->s6_addr32[3] = ip6->s6_addr32[3];
788 	llsol->s6_addr8[12] = 0xff;
789 
790 	error = in6_setscope(llsol, ifp, NULL);
791 	if (error != 0) {
792 		/* XXX: should not happen */
793 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
794 	}
795 
796 	return error;
797 }
798 
799 static int
800 in6_join_mcastgroups(struct in6_aliasreq *ifra, struct in6_ifaddr *ia,
801     struct ifnet *ifp, int flags)
802 {
803 	int error;
804 	struct sockaddr_in6 mltaddr, mltmask;
805 	struct in6_multi_mship *imm;
806 	struct in6_addr llsol;
807 	struct rtentry *rt;
808 	int dad_delay;
809 	char ip6buf[INET6_ADDRSTRLEN];
810 
811 	/* join solicited multicast addr for new host id */
812 	error = in6_get_llsol_addr(&llsol, ifp, &ifra->ifra_addr.sin6_addr);
813 	if (error != 0)
814 		goto out;
815 	dad_delay = 0;
816 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
817 		/*
818 		 * We need a random delay for DAD on the address
819 		 * being configured.  It also means delaying
820 		 * transmission of the corresponding MLD report to
821 		 * avoid report collision.
822 		 * [draft-ietf-ipv6-rfc2462bis-02.txt]
823 		 */
824 		dad_delay = cprng_fast32() % (MAX_RTR_SOLICITATION_DELAY * hz);
825 	}
826 
827 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
828 	/* join solicited multicast addr for new host id */
829 	imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
830 	if (!imm) {
831 		nd6log(LOG_ERR,
832 		    "addmulti failed for %s on %s (errno=%d)\n",
833 		    IN6_PRINT(ip6buf, &llsol), if_name(ifp), error);
834 		goto out;
835 	}
836 	mutex_enter(&in6_ifaddr_lock);
837 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
838 	mutex_exit(&in6_ifaddr_lock);
839 
840 	sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
841 
842 	/*
843 	 * join link-local all-nodes address
844 	 */
845 	sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
846 	    0, 0, 0);
847 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
848 		goto out; /* XXX: should not fail */
849 
850 	/*
851 	 * XXX: do we really need this automatic routes?
852 	 * We should probably reconsider this stuff.  Most applications
853 	 * actually do not need the routes, since they usually specify
854 	 * the outgoing interface.
855 	 */
856 	rt = rtalloc1(sin6tosa(&mltaddr), 0);
857 	if (rt) {
858 		if (memcmp(&mltaddr.sin6_addr,
859 		    &satocsin6(rt_getkey(rt))->sin6_addr,
860 		    MLTMASK_LEN)) {
861 			rt_unref(rt);
862 			rt = NULL;
863 		} else if (rt->rt_ifp != ifp) {
864 			IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
865 			    "network %04x:%04x::/32 = %04x:%04x::/32\n",
866 			    __func__, rt->rt_ifp, ifp, ifp->if_xname,
867 			    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
868 			    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
869 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
870 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
871 			rt_replace_ifa(rt, &ia->ia_ifa);
872 			rt->rt_ifp = ifp;
873 		}
874 	}
875 	if (!rt) {
876 		struct rt_addrinfo info;
877 
878 		memset(&info, 0, sizeof(info));
879 		info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
880 		info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
881 		info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
882 		info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
883 		/* XXX: we need RTF_CONNECTED to fake nd6_rtrequest */
884 		info.rti_flags = RTF_UP | RTF_CONNECTED;
885 		error = rtrequest1(RTM_ADD, &info, NULL);
886 		if (error)
887 			goto out;
888 	} else {
889 		rt_unref(rt);
890 	}
891 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
892 	if (!imm) {
893 		nd6log(LOG_WARNING,
894 		    "addmulti failed for %s on %s (errno=%d)\n",
895 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
896 		    if_name(ifp), error);
897 		goto out;
898 	}
899 	mutex_enter(&in6_ifaddr_lock);
900 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
901 	mutex_exit(&in6_ifaddr_lock);
902 
903 	/*
904 	 * join node information group address
905 	 */
906 	dad_delay = 0;
907 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
908 		/*
909 		 * The spec doesn't say anything about delay for this
910 		 * group, but the same logic should apply.
911 		 */
912 		dad_delay = cprng_fast32() % (MAX_RTR_SOLICITATION_DELAY * hz);
913 	}
914 	if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
915 		;
916 	else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
917 		  dad_delay)) == NULL) { /* XXX jinmei */
918 		nd6log(LOG_WARNING,
919 		    "addmulti failed for %s on %s (errno=%d)\n",
920 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
921 		    if_name(ifp), error);
922 		/* XXX not very fatal, go on... */
923 	} else {
924 		mutex_enter(&in6_ifaddr_lock);
925 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
926 		mutex_exit(&in6_ifaddr_lock);
927 	}
928 
929 
930 	/*
931 	 * join interface-local all-nodes address.
932 	 * (ff01::1%ifN, and ff01::%ifN/32)
933 	 */
934 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
935 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
936 		goto out; /* XXX: should not fail */
937 
938 	/* XXX: again, do we really need the route? */
939 	rt = rtalloc1(sin6tosa(&mltaddr), 0);
940 	if (rt) {
941 		/* 32bit came from "mltmask" */
942 		if (memcmp(&mltaddr.sin6_addr,
943 		    &satocsin6(rt_getkey(rt))->sin6_addr,
944 		    32 / NBBY)) {
945 			rt_unref(rt);
946 			rt = NULL;
947 		} else if (rt->rt_ifp != ifp) {
948 			IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
949 			    "network %04x:%04x::/32 = %04x:%04x::/32\n",
950 			    __func__, rt->rt_ifp, ifp, ifp->if_xname,
951 			    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
952 			    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
953 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
954 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
955 			rt_replace_ifa(rt, &ia->ia_ifa);
956 			rt->rt_ifp = ifp;
957 		}
958 	}
959 	if (!rt) {
960 		struct rt_addrinfo info;
961 
962 		memset(&info, 0, sizeof(info));
963 		info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
964 		info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
965 		info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
966 		info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
967 		info.rti_flags = RTF_UP | RTF_CONNECTED;
968 		error = rtrequest1(RTM_ADD, &info, NULL);
969 		if (error)
970 			goto out;
971 #undef	MLTMASK_LEN
972 	} else {
973 		rt_unref(rt);
974 	}
975 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
976 	if (!imm) {
977 		nd6log(LOG_WARNING,
978 		    "addmulti failed for %s on %s (errno=%d)\n",
979 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
980 		    if_name(ifp), error);
981 		goto out;
982 	} else {
983 		mutex_enter(&in6_ifaddr_lock);
984 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
985 		mutex_exit(&in6_ifaddr_lock);
986 	}
987 	return 0;
988 
989 out:
990 	KASSERT(error != 0);
991 	return error;
992 }
993 
994 /*
995  * Update parameters of an IPv6 interface address.
996  * If necessary, a new entry is created and linked into address chains.
997  * This function is separated from in6_control().
998  * XXX: should this be performed under splsoftnet()?
999  */
1000 static int
1001 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
1002     struct in6_ifaddr **iap, struct psref *psref, int flags)
1003 {
1004 	int error = 0, hostIsNew = 0, plen = -1;
1005 	struct sockaddr_in6 dst6;
1006 	struct in6_addrlifetime *lt;
1007 	int dad_delay, was_tentative;
1008 	struct in6_ifaddr *ia = iap ? *iap : NULL;
1009 	char ip6buf[INET6_ADDRSTRLEN];
1010 
1011 	KASSERT((iap == NULL && psref == NULL) ||
1012 	    (iap != NULL && psref != NULL));
1013 
1014 	/* Validate parameters */
1015 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1016 		return EINVAL;
1017 
1018 	/*
1019 	 * The destination address for a p2p link must have a family
1020 	 * of AF_UNSPEC or AF_INET6.
1021 	 */
1022 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1023 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1024 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1025 		return EAFNOSUPPORT;
1026 	/*
1027 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1028 	 * does not carry fields other than sin6_len.
1029 	 */
1030 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1031 		return EINVAL;
1032 	/*
1033 	 * Because the IPv6 address architecture is classless, we require
1034 	 * users to specify a (non 0) prefix length (mask) for a new address.
1035 	 * We also require the prefix (when specified) mask is valid, and thus
1036 	 * reject a non-consecutive mask.
1037 	 */
1038 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1039 		return EINVAL;
1040 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1041 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1042 		    (u_char *)&ifra->ifra_prefixmask +
1043 		    ifra->ifra_prefixmask.sin6_len);
1044 		if (plen <= 0)
1045 			return EINVAL;
1046 	} else {
1047 		/*
1048 		 * In this case, ia must not be NULL.  We just use its prefix
1049 		 * length.
1050 		 */
1051 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1052 	}
1053 	/*
1054 	 * If the destination address on a p2p interface is specified,
1055 	 * and the address is a scoped one, validate/set the scope
1056 	 * zone identifier.
1057 	 */
1058 	dst6 = ifra->ifra_dstaddr;
1059 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1060 	    (dst6.sin6_family == AF_INET6)) {
1061 		struct in6_addr in6_tmp;
1062 		u_int32_t zoneid;
1063 
1064 		in6_tmp = dst6.sin6_addr;
1065 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1066 			return EINVAL; /* XXX: should be impossible */
1067 
1068 		if (dst6.sin6_scope_id != 0) {
1069 			if (dst6.sin6_scope_id != zoneid)
1070 				return EINVAL;
1071 		} else		/* user omit to specify the ID. */
1072 			dst6.sin6_scope_id = zoneid;
1073 
1074 		/* convert into the internal form */
1075 		if (sa6_embedscope(&dst6, 0))
1076 			return EINVAL; /* XXX: should be impossible */
1077 	}
1078 	/*
1079 	 * The destination address can be specified only for a p2p or a
1080 	 * loopback interface.  If specified, the corresponding prefix length
1081 	 * must be 128.
1082 	 */
1083 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1084 #ifdef FORCE_P2PPLEN
1085 		int i;
1086 #endif
1087 
1088 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1089 			/* XXX: noisy message */
1090 			nd6log(LOG_INFO, "a destination can "
1091 			    "be specified for a p2p or a loopback IF only\n");
1092 			return EINVAL;
1093 		}
1094 		if (plen != 128) {
1095 			nd6log(LOG_INFO, "prefixlen should "
1096 			    "be 128 when dstaddr is specified\n");
1097 #ifdef FORCE_P2PPLEN
1098 			/*
1099 			 * To be compatible with old configurations,
1100 			 * such as ifconfig gif0 inet6 2001::1 2001::2
1101 			 * prefixlen 126, we override the specified
1102 			 * prefixmask as if the prefix length was 128.
1103 			 */
1104 			ifra->ifra_prefixmask.sin6_len =
1105 			    sizeof(struct sockaddr_in6);
1106 			for (i = 0; i < 4; i++)
1107 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
1108 				    0xffffffff;
1109 			plen = 128;
1110 #else
1111 			return EINVAL;
1112 #endif
1113 		}
1114 	}
1115 	/* lifetime consistency check */
1116 	lt = &ifra->ifra_lifetime;
1117 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1118 		return EINVAL;
1119 	if (lt->ia6t_vltime == 0) {
1120 		/*
1121 		 * the following log might be noisy, but this is a typical
1122 		 * configuration mistake or a tool's bug.
1123 		 */
1124 		nd6log(LOG_INFO, "valid lifetime is 0 for %s\n",
1125 		    IN6_PRINT(ip6buf, &ifra->ifra_addr.sin6_addr));
1126 
1127 		if (ia == NULL)
1128 			return 0; /* there's nothing to do */
1129 	}
1130 
1131 	/*
1132 	 * If this is a new address, allocate a new ifaddr and link it
1133 	 * into chains.
1134 	 */
1135 	if (ia == NULL) {
1136 		hostIsNew = 1;
1137 		/*
1138 		 * When in6_update_ifa() is called in a process of a received
1139 		 * RA, it is called under an interrupt context.  So, we should
1140 		 * call malloc with M_NOWAIT.
1141 		 */
1142 		ia = malloc(sizeof(*ia), M_IFADDR, M_NOWAIT|M_ZERO);
1143 		if (ia == NULL)
1144 			return ENOBUFS;
1145 		LIST_INIT(&ia->ia6_memberships);
1146 		/* Initialize the address and masks, and put time stamp */
1147 		ia->ia_ifa.ifa_addr = sin6tosa(&ia->ia_addr);
1148 		ia->ia_addr.sin6_family = AF_INET6;
1149 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1150 		ia->ia6_createtime = time_uptime;
1151 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1152 			/*
1153 			 * XXX: some functions expect that ifa_dstaddr is not
1154 			 * NULL for p2p interfaces.
1155 			 */
1156 			ia->ia_ifa.ifa_dstaddr = sin6tosa(&ia->ia_dstaddr);
1157 		} else {
1158 			ia->ia_ifa.ifa_dstaddr = NULL;
1159 		}
1160 		ia->ia_ifa.ifa_netmask = sin6tosa(&ia->ia_prefixmask);
1161 
1162 		ia->ia_ifp = ifp;
1163 		IN6_ADDRLIST_ENTRY_INIT(ia);
1164 		ifa_psref_init(&ia->ia_ifa);
1165 	}
1166 
1167 	/* update timestamp */
1168 	ia->ia6_updatetime = time_uptime;
1169 
1170 	/* set prefix mask */
1171 	if (ifra->ifra_prefixmask.sin6_len) {
1172 		if (ia->ia_prefixmask.sin6_len) {
1173 			/*
1174 			 * We prohibit changing the prefix length of an
1175 			 * existing autoconf address, because the operation
1176 			 * would confuse prefix management.
1177 			 */
1178 			if (ia->ia6_ndpr != NULL &&
1179 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) !=
1180 			    plen)
1181 			{
1182 				nd6log(LOG_INFO, "the prefix length of an"
1183 				    " existing (%s) autoconf address should"
1184 				    " not be changed\n",
1185 				    IN6_PRINT(ip6buf,
1186 				    &ia->ia_addr.sin6_addr));
1187 				error = EINVAL;
1188 				if (hostIsNew)
1189 					free(ia, M_IFADDR);
1190 				return error;
1191 			}
1192 
1193 			if (!IN6_ARE_ADDR_EQUAL(&ia->ia_prefixmask.sin6_addr,
1194 			    &ifra->ifra_prefixmask.sin6_addr))
1195 				in6_ifremprefix(ia);
1196 		}
1197 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1198 	}
1199 
1200 	/* Set destination address. */
1201 	if (dst6.sin6_family == AF_INET6) {
1202 		if (!IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
1203 		    &ia->ia_dstaddr.sin6_addr))
1204 			in6_ifremprefix(ia);
1205 		ia->ia_dstaddr = dst6;
1206 	}
1207 
1208 	/*
1209 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1210 	 * to see if the address is deprecated or invalidated, but initialize
1211 	 * these members for applications.
1212 	 */
1213 	ia->ia6_lifetime = ifra->ifra_lifetime;
1214 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1215 		ia->ia6_lifetime.ia6t_expire =
1216 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1217 	} else
1218 		ia->ia6_lifetime.ia6t_expire = 0;
1219 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1220 		ia->ia6_lifetime.ia6t_preferred =
1221 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1222 	} else
1223 		ia->ia6_lifetime.ia6t_preferred = 0;
1224 
1225 	/*
1226 	 * configure address flags.
1227 	 * We need to preserve tentative state so DAD works if
1228 	 * something adds the same address before DAD finishes.
1229 	 */
1230 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1231 	ia->ia6_flags = ifra->ifra_flags;
1232 
1233 	/*
1234 	 * Make the address tentative before joining multicast addresses,
1235 	 * so that corresponding MLD responses would not have a tentative
1236 	 * source address.
1237 	 */
1238 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1239 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1240 		ia->ia6_flags |= IN6_IFF_DETACHED;
1241 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1242 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp))
1243 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1244 
1245 	/*
1246 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1247 	 * userland, make it deprecated.
1248 	 */
1249 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1250 		ia->ia6_lifetime.ia6t_pltime = 0;
1251 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1252 	}
1253 
1254 	if (hostIsNew) {
1255 		/*
1256 		 * We need a reference to ia before calling in6_ifinit.
1257 		 * Otherwise ia can be freed in in6_ifinit accidentally.
1258 		 */
1259 		ifaref(&ia->ia_ifa);
1260 	}
1261 
1262 	/* Must execute in6_ifinit and ifa_insert atomically */
1263 	mutex_enter(&in6_ifaddr_lock);
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 		mutex_exit(&in6_ifaddr_lock);
1271 		return error;
1272 	}
1273 
1274 	/*
1275 	 * We are done if we have simply modified an existing address.
1276 	 */
1277 	if (!hostIsNew) {
1278 		mutex_exit(&in6_ifaddr_lock);
1279 		return error;
1280 	}
1281 
1282 	/*
1283 	 * Insert ia to the global list and ifa to the interface's list.
1284 	 * A reference to it is already gained above.
1285 	 */
1286 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1287 	ifa_insert(ifp, &ia->ia_ifa);
1288 
1289 	mutex_exit(&in6_ifaddr_lock);
1290 
1291 	/*
1292 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1293 	 * not just go to unlink.
1294 	 */
1295 
1296 	/* join necessary multicast groups */
1297 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1298 		error = in6_join_mcastgroups(ifra, ia, ifp, flags);
1299 		if (error != 0)
1300 			goto cleanup;
1301 	}
1302 
1303 	if (nd6_need_cache(ifp)) {
1304 		/* XXX maybe unnecessary */
1305 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1306 		ia->ia_ifa.ifa_flags |= RTF_CONNECTED;
1307 	}
1308 
1309 	/*
1310 	 * Perform DAD, if needed.
1311 	 * XXX It may be of use, if we can administratively
1312 	 * disable DAD.
1313 	 */
1314 	if (hostIsNew && if_do_dad(ifp) &&
1315 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1316 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1317 	{
1318 		int mindelay, maxdelay;
1319 
1320 		dad_delay = 0;
1321 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1322 			struct in6_addr llsol;
1323 			struct in6_multi *in6m_sol = NULL;
1324 			/*
1325 			 * We need to impose a delay before sending an NS
1326 			 * for DAD.  Check if we also needed a delay for the
1327 			 * corresponding MLD message.  If we did, the delay
1328 			 * should be larger than the MLD delay (this could be
1329 			 * relaxed a bit, but this simple logic is at least
1330 			 * safe).
1331 			 */
1332 			mindelay = 0;
1333 			error = in6_get_llsol_addr(&llsol, ifp,
1334 			    &ifra->ifra_addr.sin6_addr);
1335 			in6_multi_lock(RW_READER);
1336 			if (error == 0)
1337 				in6m_sol = in6_lookup_multi(&llsol, ifp);
1338 			if (in6m_sol != NULL &&
1339 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1340 				mindelay = in6m_sol->in6m_timer;
1341 			}
1342 			in6_multi_unlock();
1343 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1344 			if (maxdelay - mindelay == 0)
1345 				dad_delay = 0;
1346 			else {
1347 				dad_delay =
1348 				    (cprng_fast32() % (maxdelay - mindelay)) +
1349 				    mindelay;
1350 			}
1351 		}
1352 		/* +1 ensures callout is always used */
1353 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1354 	}
1355 
1356 	if (iap != NULL) {
1357 		*iap = ia;
1358 		if (hostIsNew)
1359 			ia6_acquire(ia, psref);
1360 	}
1361 
1362 	return 0;
1363 
1364   cleanup:
1365 	in6_purgeaddr(&ia->ia_ifa);
1366 	return error;
1367 }
1368 
1369 int
1370 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1371 {
1372 	int rc, s;
1373 
1374 	s = splsoftnet();
1375 	rc = in6_update_ifa1(ifp, ifra, NULL, NULL, flags);
1376 	splx(s);
1377 	return rc;
1378 }
1379 
1380 void
1381 in6_purgeaddr(struct ifaddr *ifa)
1382 {
1383 	struct ifnet *ifp = ifa->ifa_ifp;
1384 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1385 	struct in6_multi_mship *imm;
1386 
1387 	/* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */
1388 	KASSERT(IFNET_LOCKED(ifp));
1389 
1390 	ifa->ifa_flags |= IFA_DESTROYING;
1391 
1392 	/* stop DAD processing */
1393 	nd6_dad_stop(ifa);
1394 
1395 	/* Delete any network route. */
1396 	in6_ifremprefix(ia);
1397 
1398 	/* Remove ownaddr's loopback rtentry, if it exists. */
1399 	in6_ifremlocal(&(ia->ia_ifa));
1400 
1401 	/*
1402 	 * leave from multicast groups we have joined for the interface
1403 	 */
1404     again:
1405 	mutex_enter(&in6_ifaddr_lock);
1406 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1407 		LIST_REMOVE(imm, i6mm_chain);
1408 		mutex_exit(&in6_ifaddr_lock);
1409 		KASSERT(imm->i6mm_maddr->in6m_ifp == ifp);
1410 		in6_leavegroup(imm);
1411 		goto again;
1412 	}
1413 	mutex_exit(&in6_ifaddr_lock);
1414 
1415 	in6_unlink_ifa(ia, ifp);
1416 }
1417 
1418 static void
1419 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1420 {
1421 	int	s = splsoftnet();
1422 
1423 	mutex_enter(&in6_ifaddr_lock);
1424 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1425 	ifa_remove(ifp, &ia->ia_ifa);
1426 	mutex_exit(&in6_ifaddr_lock);
1427 
1428 	/*
1429 	 * Release the reference to the ND prefix.
1430 	 */
1431 	if (ia->ia6_ndpr != NULL) {
1432 		nd6_prefix_unref(ia->ia6_ndpr);
1433 		ia->ia6_ndpr = NULL;
1434 	}
1435 
1436 	/*
1437 	 * Also, if the address being removed is autoconf'ed, call
1438 	 * nd6_pfxlist_onlink_check() since the release might affect the status of
1439 	 * other (detached) addresses.
1440 	 */
1441 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0) {
1442 		ND6_WLOCK();
1443 		nd6_pfxlist_onlink_check();
1444 		ND6_UNLOCK();
1445 	}
1446 
1447 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1448 
1449 	/*
1450 	 * release another refcnt for the link from in6_ifaddr.
1451 	 * Note that we should decrement the refcnt at least once for all *BSD.
1452 	 */
1453 	ifafree(&ia->ia_ifa);
1454 
1455 	splx(s);
1456 }
1457 
1458 void
1459 in6_purgeif(struct ifnet *ifp)
1460 {
1461 
1462 	IFNET_LOCK(ifp);
1463 	in6_ifdetach(ifp);
1464 	IFNET_UNLOCK(ifp);
1465 }
1466 
1467 void
1468 in6_purge_mcast_references(struct in6_multi *in6m)
1469 {
1470 	struct	in6_ifaddr *ia;
1471 
1472 	KASSERT(in6_multi_locked(RW_WRITER));
1473 
1474 	mutex_enter(&in6_ifaddr_lock);
1475 	IN6_ADDRLIST_WRITER_FOREACH(ia) {
1476 		struct in6_multi_mship *imm;
1477 		LIST_FOREACH(imm, &ia->ia6_memberships, i6mm_chain) {
1478 			if (imm->i6mm_maddr == in6m)
1479 				imm->i6mm_maddr = NULL;
1480 		}
1481 	}
1482 	mutex_exit(&in6_ifaddr_lock);
1483 }
1484 
1485 /*
1486  * SIOC[GAD]LIFADDR.
1487  *	SIOCGLIFADDR: get first address. (?)
1488  *	SIOCGLIFADDR with IFLR_PREFIX:
1489  *		get first address that matches the specified prefix.
1490  *	SIOCALIFADDR: add the specified address.
1491  *	SIOCALIFADDR with IFLR_PREFIX:
1492  *		add the specified prefix, filling hostid part from
1493  *		the first link-local address.  prefixlen must be <= 64.
1494  *	SIOCDLIFADDR: delete the specified address.
1495  *	SIOCDLIFADDR with IFLR_PREFIX:
1496  *		delete the first address that matches the specified prefix.
1497  * return values:
1498  *	EINVAL on invalid parameters
1499  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1500  *	other values may be returned from in6_ioctl()
1501  *
1502  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1503  * this is to accommodate address naming scheme other than RFC2374,
1504  * in the future.
1505  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1506  * address encoding scheme. (see figure on page 8)
1507  */
1508 static int
1509 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1510 	struct ifnet *ifp)
1511 {
1512 	struct in6_ifaddr *ia = NULL; /* XXX gcc 4.8 maybe-uninitialized */
1513 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1514 	struct ifaddr *ifa;
1515 	struct sockaddr *sa;
1516 
1517 	/* sanity checks */
1518 	if (!data || !ifp) {
1519 		panic("invalid argument to in6_lifaddr_ioctl");
1520 		/* NOTREACHED */
1521 	}
1522 
1523 	switch (cmd) {
1524 	case SIOCGLIFADDR:
1525 		/* address must be specified on GET with IFLR_PREFIX */
1526 		if ((iflr->flags & IFLR_PREFIX) == 0)
1527 			break;
1528 		/* FALLTHROUGH */
1529 	case SIOCALIFADDR:
1530 	case SIOCDLIFADDR:
1531 		/* address must be specified on ADD and DELETE */
1532 		sa = (struct sockaddr *)&iflr->addr;
1533 		if (sa->sa_family != AF_INET6)
1534 			return EINVAL;
1535 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1536 			return EINVAL;
1537 		/* XXX need improvement */
1538 		sa = (struct sockaddr *)&iflr->dstaddr;
1539 		if (sa->sa_family && sa->sa_family != AF_INET6)
1540 			return EINVAL;
1541 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1542 			return EINVAL;
1543 		break;
1544 	default: /* shouldn't happen */
1545 #if 0
1546 		panic("invalid cmd to in6_lifaddr_ioctl");
1547 		/* NOTREACHED */
1548 #else
1549 		return EOPNOTSUPP;
1550 #endif
1551 	}
1552 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1553 		return EINVAL;
1554 
1555 	switch (cmd) {
1556 	case SIOCALIFADDR:
1557 	    {
1558 		struct in6_aliasreq ifra;
1559 		struct in6_addr *xhostid = NULL;
1560 		int prefixlen;
1561 		int bound = curlwp_bind();
1562 		struct psref psref;
1563 
1564 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1565 			struct sockaddr_in6 *sin6;
1566 
1567 			/*
1568 			 * xhostid is to fill in the hostid part of the
1569 			 * address.  xhostid points to the first link-local
1570 			 * address attached to the interface.
1571 			 */
1572 			ia = in6ifa_ifpforlinklocal_psref(ifp, 0, &psref);
1573 			if (ia == NULL) {
1574 				curlwp_bindx(bound);
1575 				return EADDRNOTAVAIL;
1576 			}
1577 			xhostid = IFA_IN6(&ia->ia_ifa);
1578 
1579 		 	/* prefixlen must be <= 64. */
1580 			if (64 < iflr->prefixlen) {
1581 				ia6_release(ia, &psref);
1582 				curlwp_bindx(bound);
1583 				return EINVAL;
1584 			}
1585 			prefixlen = iflr->prefixlen;
1586 
1587 			/* hostid part must be zero. */
1588 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1589 			if (sin6->sin6_addr.s6_addr32[2] != 0
1590 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1591 				ia6_release(ia, &psref);
1592 				curlwp_bindx(bound);
1593 				return EINVAL;
1594 			}
1595 		} else
1596 			prefixlen = iflr->prefixlen;
1597 
1598 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1599 		memset(&ifra, 0, sizeof(ifra));
1600 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1601 
1602 		memcpy(&ifra.ifra_addr, &iflr->addr,
1603 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1604 		if (xhostid) {
1605 			/* fill in hostid part */
1606 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1607 			    xhostid->s6_addr32[2];
1608 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1609 			    xhostid->s6_addr32[3];
1610 		}
1611 
1612 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1613 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1614 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1615 			if (xhostid) {
1616 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1617 				    xhostid->s6_addr32[2];
1618 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1619 				    xhostid->s6_addr32[3];
1620 			}
1621 		}
1622 		if (xhostid) {
1623 			ia6_release(ia, &psref);
1624 			ia = NULL;
1625 		}
1626 		curlwp_bindx(bound);
1627 
1628 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1629 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1630 
1631 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1632 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1633 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1634 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1635 	    }
1636 	case SIOCGLIFADDR:
1637 	case SIOCDLIFADDR:
1638 	    {
1639 		struct in6_addr mask, candidate, match;
1640 		struct sockaddr_in6 *sin6;
1641 		int cmp;
1642 		int error, s;
1643 
1644 		memset(&mask, 0, sizeof(mask));
1645 		if (iflr->flags & IFLR_PREFIX) {
1646 			/* lookup a prefix rather than address. */
1647 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1648 
1649 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1650 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1651 			match.s6_addr32[0] &= mask.s6_addr32[0];
1652 			match.s6_addr32[1] &= mask.s6_addr32[1];
1653 			match.s6_addr32[2] &= mask.s6_addr32[2];
1654 			match.s6_addr32[3] &= mask.s6_addr32[3];
1655 
1656 			/* if you set extra bits, that's wrong */
1657 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1658 				return EINVAL;
1659 
1660 			cmp = 1;
1661 		} else {
1662 			if (cmd == SIOCGLIFADDR) {
1663 				/* on getting an address, take the 1st match */
1664 				cmp = 0;	/* XXX */
1665 			} else {
1666 				/* on deleting an address, do exact match */
1667 				in6_prefixlen2mask(&mask, 128);
1668 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1669 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1670 
1671 				cmp = 1;
1672 			}
1673 		}
1674 
1675 		s = pserialize_read_enter();
1676 		IFADDR_READER_FOREACH(ifa, ifp) {
1677 			if (ifa->ifa_addr->sa_family != AF_INET6)
1678 				continue;
1679 			if (!cmp)
1680 				break;
1681 
1682 			/*
1683 			 * XXX: this is adhoc, but is necessary to allow
1684 			 * a user to specify fe80::/64 (not /10) for a
1685 			 * link-local address.
1686 			 */
1687 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1688 			in6_clearscope(&candidate);
1689 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1690 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1691 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1692 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1693 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1694 				break;
1695 		}
1696 		if (!ifa) {
1697 			error = EADDRNOTAVAIL;
1698 			goto error;
1699 		}
1700 		ia = ifa2ia6(ifa);
1701 
1702 		if (cmd == SIOCGLIFADDR) {
1703 			/* fill in the if_laddrreq structure */
1704 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1705 			error = sa6_recoverscope(
1706 			    (struct sockaddr_in6 *)&iflr->addr);
1707 			if (error != 0)
1708 				goto error;
1709 
1710 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1711 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1712 				    ia->ia_dstaddr.sin6_len);
1713 				error = sa6_recoverscope(
1714 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1715 				if (error != 0)
1716 					goto error;
1717 			} else
1718 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1719 
1720 			iflr->prefixlen =
1721 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1722 
1723 			iflr->flags = ia->ia6_flags;	/* XXX */
1724 
1725 			error = 0;
1726 		} else {
1727 			struct in6_aliasreq ifra;
1728 
1729 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1730 			memset(&ifra, 0, sizeof(ifra));
1731 			memcpy(ifra.ifra_name, iflr->iflr_name,
1732 			    sizeof(ifra.ifra_name));
1733 
1734 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1735 			    ia->ia_addr.sin6_len);
1736 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1737 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1738 				    ia->ia_dstaddr.sin6_len);
1739 			} else {
1740 				memset(&ifra.ifra_dstaddr, 0,
1741 				    sizeof(ifra.ifra_dstaddr));
1742 			}
1743 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1744 			    ia->ia_prefixmask.sin6_len);
1745 
1746 			ifra.ifra_flags = ia->ia6_flags;
1747 			pserialize_read_exit(s);
1748 
1749 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1750 		}
1751 	error:
1752 		pserialize_read_exit(s);
1753 		return error;
1754 	    }
1755 	}
1756 
1757 	return EOPNOTSUPP;	/* just for safety */
1758 }
1759 
1760 /*
1761  * Initialize an interface's internet6 address
1762  * and routing table entry.
1763  */
1764 static int
1765 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1766 	const struct sockaddr_in6 *sin6, int newhost)
1767 {
1768 	int	error = 0, ifacount = 0;
1769 	int s;
1770 	struct ifaddr *ifa;
1771 
1772 	KASSERT(mutex_owned(&in6_ifaddr_lock));
1773 
1774 	/*
1775 	 * Give the interface a chance to initialize
1776 	 * if this is its first address,
1777 	 * and to validate the address if necessary.
1778 	 */
1779 	s = pserialize_read_enter();
1780 	IFADDR_READER_FOREACH(ifa, ifp) {
1781 		if (ifa->ifa_addr->sa_family != AF_INET6)
1782 			continue;
1783 		ifacount++;
1784 	}
1785 	pserialize_read_exit(s);
1786 
1787 	ia->ia_addr = *sin6;
1788 
1789 	if (ifacount == 0 &&
1790 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1791 		return error;
1792 	}
1793 
1794 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1795 
1796 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1797 
1798 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1799 	if (newhost) {
1800 		/* set the rtrequest function to create llinfo */
1801 		if (ifp->if_flags & IFF_POINTOPOINT)
1802 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1803 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1804 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1805 		in6_ifaddlocal(&ia->ia_ifa);
1806 	} else {
1807 		/* Inform the routing socket of new flags/timings */
1808 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
1809 	}
1810 
1811 	/* Add the network prefix route. */
1812 	if ((error = in6_ifaddprefix(ia)) != 0) {
1813 		if (newhost)
1814 			in6_ifremlocal(&ia->ia_ifa);
1815 		return error;
1816 	}
1817 
1818 	return error;
1819 }
1820 
1821 static struct ifaddr *
1822 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1823 {
1824 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1825 		return ifa;
1826 	return best_ifa;
1827 }
1828 
1829 /*
1830  * Find an IPv6 interface link-local address specific to an interface.
1831  */
1832 struct in6_ifaddr *
1833 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1834 {
1835 	struct ifaddr *best_ifa = NULL, *ifa;
1836 
1837 	IFADDR_READER_FOREACH(ifa, ifp) {
1838 		if (ifa->ifa_addr->sa_family != AF_INET6)
1839 			continue;
1840 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1841 			continue;
1842 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1843 			continue;
1844 		best_ifa = bestifa(best_ifa, ifa);
1845 	}
1846 
1847 	return (struct in6_ifaddr *)best_ifa;
1848 }
1849 
1850 struct in6_ifaddr *
1851 in6ifa_ifpforlinklocal_psref(const struct ifnet *ifp, const int ignoreflags,
1852     struct psref *psref)
1853 {
1854 	struct in6_ifaddr *ia;
1855 	int s = pserialize_read_enter();
1856 
1857 	ia = in6ifa_ifpforlinklocal(ifp, ignoreflags);
1858 	if (ia != NULL)
1859 		ia6_acquire(ia, psref);
1860 	pserialize_read_exit(s);
1861 
1862 	return ia;
1863 }
1864 
1865 /*
1866  * find the internet address corresponding to a given address.
1867  * ifaddr is returned referenced.
1868  */
1869 struct in6_ifaddr *
1870 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1871 {
1872 	struct in6_ifaddr *ia;
1873 	int s;
1874 
1875 	s = pserialize_read_enter();
1876 	IN6_ADDRLIST_READER_FOREACH(ia) {
1877 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1878 			if (zoneid != 0 &&
1879 			    zoneid != ia->ia_addr.sin6_scope_id)
1880 				continue;
1881 			ifaref(&ia->ia_ifa);
1882 			break;
1883 		}
1884 	}
1885 	pserialize_read_exit(s);
1886 
1887 	return ia;
1888 }
1889 
1890 /*
1891  * find the internet address corresponding to a given interface and address.
1892  */
1893 struct in6_ifaddr *
1894 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1895 {
1896 	struct ifaddr *best_ifa = NULL, *ifa;
1897 
1898 	IFADDR_READER_FOREACH(ifa, ifp) {
1899 		if (ifa->ifa_addr->sa_family != AF_INET6)
1900 			continue;
1901 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1902 			continue;
1903 		best_ifa = bestifa(best_ifa, ifa);
1904 	}
1905 
1906 	return (struct in6_ifaddr *)best_ifa;
1907 }
1908 
1909 struct in6_ifaddr *
1910 in6ifa_ifpwithaddr_psref(const struct ifnet *ifp, const struct in6_addr *addr,
1911     struct psref *psref)
1912 {
1913 	struct in6_ifaddr *ia;
1914 	int s = pserialize_read_enter();
1915 
1916 	ia = in6ifa_ifpwithaddr(ifp, addr);
1917 	if (ia != NULL)
1918 		ia6_acquire(ia, psref);
1919 	pserialize_read_exit(s);
1920 
1921 	return ia;
1922 }
1923 
1924 static struct in6_ifaddr *
1925 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1926 {
1927 	if (best_ia == NULL ||
1928 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1929 		return ia;
1930 	return best_ia;
1931 }
1932 
1933 /*
1934  * Determine if an address is on a local network.
1935  */
1936 int
1937 in6_localaddr(const struct in6_addr *in6)
1938 {
1939 	struct in6_ifaddr *ia;
1940 	int s;
1941 
1942 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1943 		return 1;
1944 
1945 	s = pserialize_read_enter();
1946 	IN6_ADDRLIST_READER_FOREACH(ia) {
1947 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1948 					      &ia->ia_prefixmask.sin6_addr)) {
1949 			pserialize_read_exit(s);
1950 			return 1;
1951 		}
1952 	}
1953 	pserialize_read_exit(s);
1954 
1955 	return 0;
1956 }
1957 
1958 int
1959 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1960 {
1961 	struct in6_ifaddr *ia;
1962 	int s;
1963 
1964 	s = pserialize_read_enter();
1965 	IN6_ADDRLIST_READER_FOREACH(ia) {
1966 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1967 		    &sa6->sin6_addr) &&
1968 #ifdef SCOPEDROUTING
1969 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1970 #endif
1971 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
1972 			pserialize_read_exit(s);
1973 			return 1; /* true */
1974 		}
1975 
1976 		/* XXX: do we still have to go thru the rest of the list? */
1977 	}
1978 	pserialize_read_exit(s);
1979 
1980 	return 0;		/* false */
1981 }
1982 
1983 /*
1984  * return length of part which dst and src are equal
1985  * hard coding...
1986  */
1987 int
1988 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1989 {
1990 	int match = 0;
1991 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1992 	u_char *lim = s + 16, r;
1993 
1994 	while (s < lim)
1995 		if ((r = (*d++ ^ *s++)) != 0) {
1996 			while (r < 128) {
1997 				match++;
1998 				r <<= 1;
1999 			}
2000 			break;
2001 		} else
2002 			match += NBBY;
2003 	return match;
2004 }
2005 
2006 /* XXX: to be scope conscious */
2007 int
2008 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2009 {
2010 	int bytelen, bitlen;
2011 
2012 	/* sanity check */
2013 	if (len < 0 || len > 128) {
2014 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2015 		    len);
2016 		return 0;
2017 	}
2018 
2019 	bytelen = len / NBBY;
2020 	bitlen = len % NBBY;
2021 
2022 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2023 		return 0;
2024 	if (bitlen != 0 &&
2025 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
2026 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
2027 		return 0;
2028 
2029 	return 1;
2030 }
2031 
2032 void
2033 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2034 {
2035 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2036 	int bytelen, bitlen, i;
2037 
2038 	/* sanity check */
2039 	if (len < 0 || len > 128) {
2040 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2041 		    len);
2042 		return;
2043 	}
2044 
2045 	memset(maskp, 0, sizeof(*maskp));
2046 	bytelen = len / NBBY;
2047 	bitlen = len % NBBY;
2048 	for (i = 0; i < bytelen; i++)
2049 		maskp->s6_addr[i] = 0xff;
2050 	if (bitlen)
2051 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2052 }
2053 
2054 /*
2055  * return the best address out of the same scope. if no address was
2056  * found, return the first valid address from designated IF.
2057  */
2058 struct in6_ifaddr *
2059 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2060 {
2061 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2062 	struct ifaddr *ifa;
2063 	struct in6_ifaddr *best_ia = NULL, *ia;
2064 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2065 
2066 	dep[0] = dep[1] = NULL;
2067 
2068 	/*
2069 	 * We first look for addresses in the same scope.
2070 	 * If there is one, return it.
2071 	 * If two or more, return one which matches the dst longest.
2072 	 * If none, return one of global addresses assigned other ifs.
2073 	 */
2074 	IFADDR_READER_FOREACH(ifa, ifp) {
2075 		if (ifa->ifa_addr->sa_family != AF_INET6)
2076 			continue;
2077 		ia = (struct in6_ifaddr *)ifa;
2078 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2079 			continue; /* XXX: is there any case to allow anycast? */
2080 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2081 			continue; /* don't use this interface */
2082 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2083 			continue;
2084 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2085 			if (ip6_use_deprecated)
2086 				dep[0] = ia;
2087 			continue;
2088 		}
2089 
2090 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
2091 			continue;
2092 		/*
2093 		 * call in6_matchlen() as few as possible
2094 		 */
2095 		if (best_ia == NULL) {
2096 			best_ia = ia;
2097 			continue;
2098 		}
2099 		if (blen == -1)
2100 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2101 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2102 		if (tlen > blen) {
2103 			blen = tlen;
2104 			best_ia = ia;
2105 		} else if (tlen == blen)
2106 			best_ia = bestia(best_ia, ia);
2107 	}
2108 	if (best_ia != NULL)
2109 		return best_ia;
2110 
2111 	IFADDR_READER_FOREACH(ifa, ifp) {
2112 		if (ifa->ifa_addr->sa_family != AF_INET6)
2113 			continue;
2114 		ia = (struct in6_ifaddr *)ifa;
2115 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2116 			continue; /* XXX: is there any case to allow anycast? */
2117 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2118 			continue; /* don't use this interface */
2119 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2120 			continue;
2121 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2122 			if (ip6_use_deprecated)
2123 				dep[1] = (struct in6_ifaddr *)ifa;
2124 			continue;
2125 		}
2126 
2127 		best_ia = bestia(best_ia, ia);
2128 	}
2129 	if (best_ia != NULL)
2130 		return best_ia;
2131 
2132 	/* use the last-resort values, that are, deprecated addresses */
2133 	if (dep[0])
2134 		return dep[0];
2135 	if (dep[1])
2136 		return dep[1];
2137 
2138 	return NULL;
2139 }
2140 
2141 /*
2142  * perform DAD when interface becomes IFF_UP.
2143  */
2144 void
2145 in6_if_link_up(struct ifnet *ifp)
2146 {
2147 	struct ifaddr *ifa;
2148 	struct in6_ifaddr *ia;
2149 	int s, bound;
2150 	char ip6buf[INET6_ADDRSTRLEN];
2151 
2152 	/* Ensure it's sane to run DAD */
2153 	if (ifp->if_link_state == LINK_STATE_DOWN)
2154 		return;
2155 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2156 		return;
2157 
2158 	bound = curlwp_bind();
2159 	s = pserialize_read_enter();
2160 	IFADDR_READER_FOREACH(ifa, ifp) {
2161 		struct psref psref;
2162 
2163 		if (ifa->ifa_addr->sa_family != AF_INET6)
2164 			continue;
2165 
2166 		ifa_acquire(ifa, &psref);
2167 		pserialize_read_exit(s);
2168 		ia = (struct in6_ifaddr *)ifa;
2169 
2170 		/* If detached then mark as tentative */
2171 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2172 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2173 			if (if_do_dad(ifp)) {
2174 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2175 				nd6log(LOG_ERR, "%s marked tentative\n",
2176 				    IN6_PRINT(ip6buf,
2177 				    &ia->ia_addr.sin6_addr));
2178 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2179 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2180 		}
2181 
2182 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2183 			int rand_delay;
2184 
2185 			/* Clear the duplicated flag as we're starting DAD. */
2186 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2187 
2188 			/*
2189 			 * The TENTATIVE flag was likely set by hand
2190 			 * beforehand, implicitly indicating the need for DAD.
2191 			 * We may be able to skip the random delay in this
2192 			 * case, but we impose delays just in case.
2193 			 */
2194 			rand_delay = cprng_fast32() %
2195 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2196 			/* +1 ensures callout is always used */
2197 			nd6_dad_start(ifa, rand_delay + 1);
2198 		}
2199 
2200 		s = pserialize_read_enter();
2201 		ifa_release(ifa, &psref);
2202 	}
2203 	pserialize_read_exit(s);
2204 	curlwp_bindx(bound);
2205 
2206 	/* Restore any detached prefixes */
2207 	ND6_WLOCK();
2208 	nd6_pfxlist_onlink_check();
2209 	ND6_UNLOCK();
2210 }
2211 
2212 void
2213 in6_if_up(struct ifnet *ifp)
2214 {
2215 
2216 	/*
2217 	 * special cases, like 6to4, are handled in in6_ifattach
2218 	 */
2219 	in6_ifattach(ifp, NULL);
2220 
2221 	/* interface may not support link state, so bring it up also */
2222 	in6_if_link_up(ifp);
2223 }
2224 
2225 /*
2226  * Mark all addresses as detached.
2227  */
2228 void
2229 in6_if_link_down(struct ifnet *ifp)
2230 {
2231 	struct ifaddr *ifa;
2232 	struct in6_ifaddr *ia;
2233 	int s, bound;
2234 	char ip6buf[INET6_ADDRSTRLEN];
2235 
2236 	/* Any prefixes on this interface should be detached as well */
2237 	ND6_WLOCK();
2238 	nd6_pfxlist_onlink_check();
2239 	ND6_UNLOCK();
2240 
2241 	bound = curlwp_bind();
2242 	s = pserialize_read_enter();
2243 	IFADDR_READER_FOREACH(ifa, ifp) {
2244 		struct psref psref;
2245 
2246 		if (ifa->ifa_addr->sa_family != AF_INET6)
2247 			continue;
2248 
2249 		ifa_acquire(ifa, &psref);
2250 		pserialize_read_exit(s);
2251 		ia = (struct in6_ifaddr *)ifa;
2252 
2253 		/* Stop DAD processing */
2254 		nd6_dad_stop(ifa);
2255 
2256 		/*
2257 		 * Mark the address as detached.
2258 		 * This satisfies RFC4862 Section 5.3, but we should apply
2259 		 * this logic to all addresses to be a good citizen and
2260 		 * avoid potential duplicated addresses.
2261 		 * When the interface comes up again, detached addresses
2262 		 * are marked tentative and DAD commences.
2263 		 */
2264 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2265 			nd6log(LOG_DEBUG, "%s marked detached\n",
2266 			    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
2267 			ia->ia6_flags |= IN6_IFF_DETACHED;
2268 			ia->ia6_flags &=
2269 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2270 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2271 		}
2272 
2273 		s = pserialize_read_enter();
2274 		ifa_release(ifa, &psref);
2275 	}
2276 	pserialize_read_exit(s);
2277 	curlwp_bindx(bound);
2278 }
2279 
2280 void
2281 in6_if_down(struct ifnet *ifp)
2282 {
2283 
2284 	in6_if_link_down(ifp);
2285 	lltable_purge_entries(LLTABLE6(ifp));
2286 }
2287 
2288 void
2289 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2290 {
2291 
2292 	switch (link_state) {
2293 	case LINK_STATE_DOWN:
2294 		in6_if_link_down(ifp);
2295 		break;
2296 	case LINK_STATE_UP:
2297 		in6_if_link_up(ifp);
2298 		break;
2299 	}
2300 }
2301 
2302 /*
2303  * Calculate max IPv6 MTU through all the interfaces and store it
2304  * to in6_maxmtu.
2305  */
2306 void
2307 in6_setmaxmtu(void)
2308 {
2309 	unsigned long maxmtu = 0;
2310 	struct ifnet *ifp;
2311 	int s;
2312 
2313 	s = pserialize_read_enter();
2314 	IFNET_READER_FOREACH(ifp) {
2315 		/* this function can be called during ifnet initialization */
2316 		if (!ifp->if_afdata[AF_INET6])
2317 			continue;
2318 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2319 		    IN6_LINKMTU(ifp) > maxmtu)
2320 			maxmtu = IN6_LINKMTU(ifp);
2321 	}
2322 	pserialize_read_exit(s);
2323 	if (maxmtu)	     /* update only when maxmtu is positive */
2324 		in6_maxmtu = maxmtu;
2325 }
2326 
2327 int
2328 in6_tunnel_validate(const struct ip6_hdr *ip6, const struct in6_addr *src,
2329     const struct in6_addr *dst)
2330 {
2331 
2332 	/* check for address match */
2333 	if (!IN6_ARE_ADDR_EQUAL(src, &ip6->ip6_dst) ||
2334 	    !IN6_ARE_ADDR_EQUAL(dst, &ip6->ip6_src))
2335 		return 0;
2336 
2337 	/* martian filters on outer source - done in ip6_input */
2338 
2339 	/* NOTE: the pakcet may be dropped by uRPF. */
2340 
2341 	/* return valid bytes length */
2342 	return sizeof(*src) + sizeof(*dst);
2343 }
2344 
2345 /*
2346  * Provide the length of interface identifiers to be used for the link attached
2347  * to the given interface.  The length should be defined in "IPv6 over
2348  * xxx-link" document.  Note that address architecture might also define
2349  * the length for a particular set of address prefixes, regardless of the
2350  * link type.  As clarified in rfc2462bis, those two definitions should be
2351  * consistent, and those really are as of August 2004.
2352  */
2353 int
2354 in6_if2idlen(struct ifnet *ifp)
2355 {
2356 	switch (ifp->if_type) {
2357 	case IFT_ETHER:		/* RFC2464 */
2358 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2359 	case IFT_L2VLAN:	/* ditto */
2360 	case IFT_IEEE80211:	/* ditto */
2361 	case IFT_FDDI:		/* RFC2467 */
2362 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2363 	case IFT_PPP:		/* RFC2472 */
2364 	case IFT_ARCNET:	/* RFC2497 */
2365 	case IFT_FRELAY:	/* RFC2590 */
2366 	case IFT_IEEE1394:	/* RFC3146 */
2367 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2368 	case IFT_LOOP:		/* XXX: is this really correct? */
2369 		return 64;
2370 	default:
2371 		/*
2372 		 * Unknown link type:
2373 		 * It might be controversial to use the today's common constant
2374 		 * of 64 for these cases unconditionally.  For full compliance,
2375 		 * we should return an error in this case.  On the other hand,
2376 		 * if we simply miss the standard for the link type or a new
2377 		 * standard is defined for a new link type, the IFID length
2378 		 * is very likely to be the common constant.  As a compromise,
2379 		 * we always use the constant, but make an explicit notice
2380 		 * indicating the "unknown" case.
2381 		 */
2382 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2383 		return 64;
2384 	}
2385 }
2386 
2387 struct in6_llentry {
2388 	struct llentry		base;
2389 };
2390 
2391 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2392 #define	IN6_LLTBL_HASH(k, h) \
2393 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2394 
2395 /*
2396  * Do actual deallocation of @lle.
2397  * Called by LLE_FREE_LOCKED when number of references
2398  * drops to zero.
2399  */
2400 static void
2401 in6_lltable_destroy_lle(struct llentry *lle)
2402 {
2403 
2404 	LLE_WUNLOCK(lle);
2405 	LLE_LOCK_DESTROY(lle);
2406 	kmem_intr_free(lle, sizeof(struct in6_llentry));
2407 }
2408 
2409 static struct llentry *
2410 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2411 {
2412 	struct in6_llentry *lle;
2413 
2414 	lle = kmem_intr_zalloc(sizeof(struct in6_llentry), KM_NOSLEEP);
2415 	if (lle == NULL)		/* NB: caller generates msg */
2416 		return NULL;
2417 
2418 	lle->base.r_l3addr.addr6 = *addr6;
2419 	lle->base.lle_refcnt = 1;
2420 	lle->base.lle_free = in6_lltable_destroy_lle;
2421 	LLE_LOCK_INIT(&lle->base);
2422 	callout_init(&lle->base.lle_timer, CALLOUT_MPSAFE);
2423 
2424 	return &lle->base;
2425 }
2426 
2427 static int
2428 in6_lltable_match_prefix(const struct sockaddr *prefix,
2429     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2430 {
2431 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2432 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2433 
2434 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2435 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2436 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2437 		return 1;
2438 
2439 	return 0;
2440 }
2441 
2442 static void
2443 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2444 {
2445 	struct ifnet *ifp = llt->llt_ifp;
2446 	bool locked = false;
2447 
2448 	LLE_WLOCK_ASSERT(lle);
2449 
2450 	/* Unlink entry from table */
2451 	if ((lle->la_flags & LLE_LINKED) != 0) {
2452 		IF_AFDATA_WLOCK_ASSERT(ifp);
2453 		lltable_unlink_entry(llt, lle);
2454 		KASSERT((lle->la_flags & LLE_LINKED) == 0);
2455 		locked = true;
2456 	}
2457 	/*
2458 	 * We need to release the lock here to lle_timer proceeds;
2459 	 * lle_timer should stop immediately if LLE_LINKED isn't set.
2460 	 * Note that we cannot pass lle->lle_lock to callout_halt
2461 	 * because it's a rwlock.
2462 	 */
2463 	LLE_ADDREF(lle);
2464 	LLE_WUNLOCK(lle);
2465 	if (locked)
2466 		IF_AFDATA_WUNLOCK(ifp);
2467 
2468 #ifdef NET_MPSAFE
2469 	callout_halt(&lle->lle_timer, NULL);
2470 #else
2471 	if (mutex_owned(softnet_lock))
2472 		callout_halt(&lle->lle_timer, softnet_lock);
2473 	else
2474 		callout_halt(&lle->lle_timer, NULL);
2475 #endif
2476 	LLE_WLOCK(lle);
2477 	LLE_REMREF(lle);
2478 
2479 	lltable_drop_entry_queue(lle);
2480 	LLE_FREE_LOCKED(lle);
2481 
2482 	if (locked)
2483 		IF_AFDATA_WLOCK(ifp);
2484 }
2485 
2486 static int
2487 in6_lltable_rtcheck(struct ifnet *ifp, u_int flags,
2488     const struct sockaddr *l3addr, const struct rtentry *rt)
2489 {
2490 	char ip6buf[INET6_ADDRSTRLEN];
2491 
2492 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2493 		int s;
2494 		struct ifaddr *ifa;
2495 		/*
2496 		 * Create an ND6 cache for an IPv6 neighbor
2497 		 * that is not covered by our own prefix.
2498 		 */
2499 		/* XXX ifaof_ifpforaddr should take a const param */
2500 		s = pserialize_read_enter();
2501 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2502 		if (ifa != NULL) {
2503 			pserialize_read_exit(s);
2504 			return 0;
2505 		}
2506 		pserialize_read_exit(s);
2507 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2508 		    IN6_PRINT(ip6buf,
2509 		    &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2510 		return EINVAL;
2511 	}
2512 	return 0;
2513 }
2514 
2515 static inline uint32_t
2516 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2517 {
2518 
2519 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2520 }
2521 
2522 static uint32_t
2523 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2524 {
2525 
2526 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2527 }
2528 
2529 static void
2530 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2531 {
2532 	struct sockaddr_in6 *sin6;
2533 
2534 	sin6 = (struct sockaddr_in6 *)sa;
2535 	bzero(sin6, sizeof(*sin6));
2536 	sin6->sin6_family = AF_INET6;
2537 	sin6->sin6_len = sizeof(*sin6);
2538 	sin6->sin6_addr = lle->r_l3addr.addr6;
2539 }
2540 
2541 static inline struct llentry *
2542 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2543 {
2544 	struct llentry *lle;
2545 	struct llentries *lleh;
2546 	u_int hashidx;
2547 
2548 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2549 	lleh = &llt->lle_head[hashidx];
2550 	LIST_FOREACH(lle, lleh, lle_next) {
2551 		if (lle->la_flags & LLE_DELETED)
2552 			continue;
2553 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2554 			break;
2555 	}
2556 
2557 	return lle;
2558 }
2559 
2560 static int
2561 in6_lltable_delete(struct lltable *llt, u_int flags,
2562 	const struct sockaddr *l3addr)
2563 {
2564 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2565 	struct llentry *lle;
2566 
2567 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2568 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2569 	    "sin_family %d", l3addr->sa_family);
2570 
2571 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2572 
2573 	if (lle == NULL) {
2574 #ifdef DEBUG
2575 		char buf[64];
2576 		sockaddr_format(l3addr, buf, sizeof(buf));
2577 		log(LOG_INFO, "%s: cache for %s is not found\n",
2578 		    __func__, buf);
2579 #endif
2580 		return ENOENT;
2581 	}
2582 
2583 	LLE_WLOCK(lle);
2584 	lle->la_flags |= LLE_DELETED;
2585 #ifdef DEBUG
2586 	{
2587 		char buf[64];
2588 		sockaddr_format(l3addr, buf, sizeof(buf));
2589 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2590 		    __func__, buf, lle);
2591 	}
2592 #endif
2593 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2594 		llentry_free(lle);
2595 	else
2596 		LLE_WUNLOCK(lle);
2597 
2598 	return 0;
2599 }
2600 
2601 static struct llentry *
2602 in6_lltable_create(struct lltable *llt, u_int flags,
2603     const struct sockaddr *l3addr, const struct rtentry *rt)
2604 {
2605 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2606 	struct ifnet *ifp = llt->llt_ifp;
2607 	struct llentry *lle;
2608 
2609 	IF_AFDATA_WLOCK_ASSERT(ifp);
2610 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2611 	    "sin_family %d", l3addr->sa_family);
2612 
2613 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2614 
2615 	if (lle != NULL) {
2616 		LLE_WLOCK(lle);
2617 		return lle;
2618 	}
2619 
2620 	/*
2621 	 * A route that covers the given address must have
2622 	 * been installed 1st because we are doing a resolution,
2623 	 * verify this.
2624 	 */
2625 	if (!(flags & LLE_IFADDR) &&
2626 	    in6_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2627 		return NULL;
2628 
2629 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2630 	if (lle == NULL) {
2631 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2632 		return NULL;
2633 	}
2634 	lle->la_flags = flags;
2635 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2636 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2637 		lle->la_flags |= LLE_VALID;
2638 	}
2639 
2640 	lltable_link_entry(llt, lle);
2641 	LLE_WLOCK(lle);
2642 
2643 	return lle;
2644 }
2645 
2646 static struct llentry *
2647 in6_lltable_lookup(struct lltable *llt, u_int flags,
2648 	const struct sockaddr *l3addr)
2649 {
2650 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2651 	struct llentry *lle;
2652 
2653 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2654 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2655 	    "sin_family %d", l3addr->sa_family);
2656 
2657 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2658 
2659 	if (lle == NULL)
2660 		return NULL;
2661 
2662 	if (flags & LLE_EXCLUSIVE)
2663 		LLE_WLOCK(lle);
2664 	else
2665 		LLE_RLOCK(lle);
2666 	return lle;
2667 }
2668 
2669 static int
2670 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2671     struct rt_walkarg *w)
2672 {
2673 	struct sockaddr_in6 sin6;
2674 
2675 	LLTABLE_LOCK_ASSERT();
2676 
2677 	/* skip deleted entries */
2678 	if (lle->la_flags & LLE_DELETED)
2679 		return 0;
2680 
2681 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2682 
2683 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2684 }
2685 
2686 static struct lltable *
2687 in6_lltattach(struct ifnet *ifp)
2688 {
2689 	struct lltable *llt;
2690 
2691 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2692 	llt->llt_af = AF_INET6;
2693 	llt->llt_ifp = ifp;
2694 
2695 	llt->llt_lookup = in6_lltable_lookup;
2696 	llt->llt_create = in6_lltable_create;
2697 	llt->llt_delete = in6_lltable_delete;
2698 	llt->llt_dump_entry = in6_lltable_dump_entry;
2699 	llt->llt_hash = in6_lltable_hash;
2700 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2701 	llt->llt_free_entry = in6_lltable_free_entry;
2702 	llt->llt_match_prefix = in6_lltable_match_prefix;
2703 	lltable_link(llt);
2704 
2705 	return llt;
2706 }
2707 
2708 void *
2709 in6_domifattach(struct ifnet *ifp)
2710 {
2711 	struct in6_ifextra *ext;
2712 
2713 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2714 
2715 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2716 	    M_IFADDR, M_WAITOK|M_ZERO);
2717 
2718 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2719 	    M_IFADDR, M_WAITOK|M_ZERO);
2720 
2721 	ext->nd_ifinfo = nd6_ifattach(ifp);
2722 	ext->scope6_id = scope6_ifattach(ifp);
2723 	ext->nprefixes = 0;
2724 	ext->ndefrouters = 0;
2725 
2726 	ext->lltable = in6_lltattach(ifp);
2727 
2728 	return ext;
2729 }
2730 
2731 void
2732 in6_domifdetach(struct ifnet *ifp, void *aux)
2733 {
2734 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2735 
2736 	lltable_free(ext->lltable);
2737 	ext->lltable = NULL;
2738 	SOFTNET_LOCK_UNLESS_NET_MPSAFE();
2739 	nd6_ifdetach(ifp, ext);
2740 	SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();
2741 	free(ext->in6_ifstat, M_IFADDR);
2742 	free(ext->icmp6_ifstat, M_IFADDR);
2743 	scope6_ifdetach(ext->scope6_id);
2744 	free(ext, M_IFADDR);
2745 }
2746 
2747 /*
2748  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2749  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2750  */
2751 void
2752 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2753 {
2754 	in6->s6_addr32[0] = 0;
2755 	in6->s6_addr32[1] = 0;
2756 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2757 	in6->s6_addr32[3] = in->s_addr;
2758 }
2759 
2760 /*
2761  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2762  * v4 mapped addr or v4 compat addr
2763  */
2764 void
2765 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2766 {
2767 	memset(sin, 0, sizeof(*sin));
2768 	sin->sin_len = sizeof(struct sockaddr_in);
2769 	sin->sin_family = AF_INET;
2770 	sin->sin_port = sin6->sin6_port;
2771 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2772 }
2773 
2774 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2775 void
2776 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2777 {
2778 	memset(sin6, 0, sizeof(*sin6));
2779 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2780 	sin6->sin6_family = AF_INET6;
2781 	sin6->sin6_port = sin->sin_port;
2782 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2783 }
2784 
2785 /* Convert sockaddr_in6 into sockaddr_in. */
2786 void
2787 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2788 {
2789 	struct sockaddr_in *sin_p;
2790 	struct sockaddr_in6 sin6;
2791 
2792 	/*
2793 	 * Save original sockaddr_in6 addr and convert it
2794 	 * to sockaddr_in.
2795 	 */
2796 	sin6 = *(struct sockaddr_in6 *)nam;
2797 	sin_p = (struct sockaddr_in *)nam;
2798 	in6_sin6_2_sin(sin_p, &sin6);
2799 }
2800 
2801 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2802 void
2803 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2804 {
2805 	struct sockaddr_in *sin_p;
2806 	struct sockaddr_in6 *sin6_p;
2807 
2808 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2809 	sin_p = (struct sockaddr_in *)*nam;
2810 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2811 	free(*nam, M_SONAME);
2812 	*nam = sin6tosa(sin6_p);
2813 }
2814