xref: /netbsd-src/sys/netinet6/in6.c (revision 7fa608457b817eca6e0977b37f758ae064f3c99c)
1 /*	$NetBSD: in6.c,v 1.135 2007/11/10 00:05:57 dyoung 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 1.135 2007/11/10 00:05:57 dyoung Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_pfil_hooks.h"
69 
70 #include <sys/param.h>
71 #include <sys/ioctl.h>
72 #include <sys/errno.h>
73 #include <sys/malloc.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/sockio.h>
77 #include <sys/systm.h>
78 #include <sys/proc.h>
79 #include <sys/time.h>
80 #include <sys/kernel.h>
81 #include <sys/syslog.h>
82 #include <sys/kauth.h>
83 
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/route.h>
87 #include <net/if_dl.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <net/if_ether.h>
92 
93 #include <netinet/ip6.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet6/nd6.h>
96 #include <netinet6/mld6_var.h>
97 #include <netinet6/ip6_mroute.h>
98 #include <netinet6/in6_ifattach.h>
99 #include <netinet6/scope6_var.h>
100 
101 #include <net/net_osdep.h>
102 
103 #ifdef PFIL_HOOKS
104 #include <net/pfil.h>
105 #endif
106 
107 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
108 
109 /* enable backward compatibility code for obsoleted ioctls */
110 #define COMPAT_IN6IFIOCTL
111 
112 #ifdef	IN6_DEBUG
113 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
114 #else
115 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
116 #endif /* IN6_DEBUG */
117 
118 /*
119  * Definitions of some constant IP6 addresses.
120  */
121 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
122 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
123 const struct in6_addr in6addr_nodelocal_allnodes =
124 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
125 const struct in6_addr in6addr_linklocal_allnodes =
126 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
127 const struct in6_addr in6addr_linklocal_allrouters =
128 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
129 
130 const struct in6_addr in6mask0 = IN6MASK0;
131 const struct in6_addr in6mask32 = IN6MASK32;
132 const struct in6_addr in6mask64 = IN6MASK64;
133 const struct in6_addr in6mask96 = IN6MASK96;
134 const struct in6_addr in6mask128 = IN6MASK128;
135 
136 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
137 				     0, 0, IN6ADDR_ANY_INIT, 0};
138 
139 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
140 	struct ifnet *, struct lwp *);
141 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
142 	struct sockaddr_in6 *, int);
143 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
144 
145 /*
146  * Subroutine for in6_ifaddloop() and in6_ifremloop().
147  * This routine does actual work.
148  */
149 static void
150 in6_ifloop_request(int cmd, struct ifaddr *ifa)
151 {
152 	struct sockaddr_in6 lo_sa;
153 	struct sockaddr_in6 all1_sa;
154 	struct rtentry *nrt = NULL;
155 	int e;
156 
157 	sockaddr_in6_init(&all1_sa, &in6mask128, 0, 0, 0);
158 	sockaddr_in6_init(&lo_sa, &in6addr_loopback, 0, 0, 0);
159 
160 	/*
161 	 * We specify the address itself as the gateway, and set the
162 	 * RTF_LLINFO flag, so that the corresponding host route would have
163 	 * the flag, and thus applications that assume traditional behavior
164 	 * would be happy.  Note that we assume the caller of the function
165 	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
166 	 * which changes the outgoing interface to the loopback interface.
167 	 */
168 	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr,
169 	    (struct sockaddr *)&all1_sa, RTF_UP|RTF_HOST|RTF_LLINFO, &nrt);
170 	if (e != 0) {
171 		log(LOG_ERR, "in6_ifloop_request: "
172 		    "%s operation failed for %s (errno=%d)\n",
173 		    cmd == RTM_ADD ? "ADD" : "DELETE",
174 		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
175 		    e);
176 	}
177 
178 	/*
179 	 * Make sure rt_ifa be equal to IFA, the second argument of the
180 	 * function.
181 	 * We need this because when we refer to rt_ifa->ia6_flags in
182 	 * ip6_input, we assume that the rt_ifa points to the address instead
183 	 * of the loopback address.
184 	 */
185 	if (cmd == RTM_ADD && nrt && ifa != nrt->rt_ifa)
186 		rt_replace_ifa(nrt, ifa);
187 
188 	/*
189 	 * Report the addition/removal of the address to the routing socket.
190 	 * XXX: since we called rtinit for a p2p interface with a destination,
191 	 *      we end up reporting twice in such a case.  Should we rather
192 	 *      omit the second report?
193 	 */
194 	if (nrt) {
195 		rt_newaddrmsg(cmd, ifa, e, nrt);
196 		if (cmd == RTM_DELETE) {
197 			if (nrt->rt_refcnt <= 0) {
198 				/* XXX: we should free the entry ourselves. */
199 				nrt->rt_refcnt++;
200 				rtfree(nrt);
201 			}
202 		} else {
203 			/* the cmd must be RTM_ADD here */
204 			nrt->rt_refcnt--;
205 		}
206 	}
207 }
208 
209 /*
210  * Add ownaddr as loopback rtentry.  We previously add the route only if
211  * necessary (ex. on a p2p link).  However, since we now manage addresses
212  * separately from prefixes, we should always add the route.  We can't
213  * rely on the cloning mechanism from the corresponding interface route
214  * any more.
215  */
216 void
217 in6_ifaddloop(struct ifaddr *ifa)
218 {
219 	struct rtentry *rt;
220 
221 	/* If there is no loopback entry, allocate one. */
222 	rt = rtalloc1(ifa->ifa_addr, 0);
223 	if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
224 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
225 		in6_ifloop_request(RTM_ADD, ifa);
226 	if (rt != NULL)
227 		rt->rt_refcnt--;
228 }
229 
230 /*
231  * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
232  * if it exists.
233  */
234 void
235 in6_ifremloop(struct ifaddr *ifa)
236 {
237 	struct in6_ifaddr *alt_ia = NULL, *ia;
238 	struct rtentry *rt;
239 	int ia_count = 0;
240 
241 	/*
242 	 * Some of BSD variants do not remove cloned routes
243 	 * from an interface direct route, when removing the direct route
244 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
245 	 * cloned routes, they could fail to remove the cloned routes when
246 	 * we handle multple addresses that share a common prefix.
247 	 * So, we should remove the route corresponding to the deleted address.
248 	 */
249 
250 	/*
251 	 * Delete the entry only if exactly one ifaddr matches the
252 	 * address, ifa->ifa_addr.
253 	 *
254 	 * If more than one ifaddr matches, replace the ifaddr in
255 	 * the routing table, rt_ifa, with a different ifaddr than
256 	 * the one we are purging, ifa.  It is important to do
257 	 * this, or else the routing table can accumulate dangling
258 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
259 	 * which will lead to crashes, later.  (More than one ifaddr
260 	 * can match if we assign the same address to multiple---probably
261 	 * p2p---interfaces.)
262 	 *
263 	 * XXX An old comment at this place said, "we should avoid
264 	 * XXX such a configuration [i.e., interfaces with the same
265 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
266 	 * XXX agree, especially now that I have fixed the dangling
267 	 * XXX ifp-pointers bug.
268 	 */
269 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
270 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
271 			continue;
272 		if (ia->ia_ifp != ifa->ifa_ifp)
273 			alt_ia = ia;
274 		if (++ia_count > 1 && alt_ia != NULL)
275 			break;
276 	}
277 
278 	if (ia_count == 0)
279 		return;
280 
281 	if ((rt = rtalloc1(ifa->ifa_addr, 0)) == NULL)
282 		return;
283 	rt->rt_refcnt--;
284 
285 	/*
286 	 * Before deleting, check if a corresponding loopbacked
287 	 * host route surely exists.  With this check, we can avoid
288 	 * deleting an interface direct route whose destination is
289 	 * the same as the address being removed.  This can happen
290 	 * when removing a subnet-router anycast address on an
291 	 * interface attached to a shared medium.
292 	 */
293 	if ((rt->rt_flags & RTF_HOST) == 0 ||
294 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
295 		return;
296 
297 	/* If we cannot replace the route's ifaddr with the equivalent
298 	 * ifaddr of another interface, I believe it is safest to
299 	 * delete the route.
300 	 */
301 	if (ia_count == 1 || alt_ia == NULL)
302 		in6_ifloop_request(RTM_DELETE, ifa);
303 	else
304 		rt_replace_ifa(rt, &alt_ia->ia_ifa);
305 }
306 
307 int
308 in6_mask2len(struct in6_addr *mask, u_char *lim0)
309 {
310 	int x = 0, y;
311 	u_char *lim = lim0, *p;
312 
313 	/* ignore the scope_id part */
314 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
315 		lim = (u_char *)mask + sizeof(*mask);
316 	for (p = (u_char *)mask; p < lim; x++, p++) {
317 		if (*p != 0xff)
318 			break;
319 	}
320 	y = 0;
321 	if (p < lim) {
322 		for (y = 0; y < NBBY; y++) {
323 			if ((*p & (0x80 >> y)) == 0)
324 				break;
325 		}
326 	}
327 
328 	/*
329 	 * when the limit pointer is given, do a stricter check on the
330 	 * remaining bits.
331 	 */
332 	if (p < lim) {
333 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
334 			return -1;
335 		for (p = p + 1; p < lim; p++)
336 			if (*p != 0)
337 				return -1;
338 	}
339 
340 	return x * NBBY + y;
341 }
342 
343 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
344 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
345 
346 static int
347 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
348     struct lwp *l, int privileged)
349 {
350 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
351 	struct	in6_ifaddr *ia = NULL;
352 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
353 	struct sockaddr_in6 *sa6;
354 	int error;
355 	switch (cmd) {
356 	/*
357 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
358 	 */
359 	case SIOCSIFADDR:
360 	case SIOCSIFDSTADDR:
361 #ifdef SIOCSIFCONF_X25
362 	case SIOCSIFCONF_X25:
363 #endif
364 		return EOPNOTSUPP;
365 	case SIOCGETSGCNT_IN6:
366 	case SIOCGETMIFCNT_IN6:
367 		return mrt6_ioctl(cmd, data);
368 	}
369 
370 	if (ifp == NULL)
371 		return EOPNOTSUPP;
372 
373 	switch (cmd) {
374 	case SIOCSNDFLUSH_IN6:
375 	case SIOCSPFXFLUSH_IN6:
376 	case SIOCSRTRFLUSH_IN6:
377 	case SIOCSDEFIFACE_IN6:
378 	case SIOCSIFINFO_FLAGS:
379 	case SIOCSIFINFO_IN6:
380 		if (!privileged)
381 			return EPERM;
382 		/* FALLTHROUGH */
383 	case OSIOCGIFINFO_IN6:
384 	case SIOCGIFINFO_IN6:
385 	case SIOCGDRLST_IN6:
386 	case SIOCGPRLST_IN6:
387 	case SIOCGNBRINFO_IN6:
388 	case SIOCGDEFIFACE_IN6:
389 		return nd6_ioctl(cmd, data, ifp);
390 	}
391 
392 	switch (cmd) {
393 	case SIOCSIFPREFIX_IN6:
394 	case SIOCDIFPREFIX_IN6:
395 	case SIOCAIFPREFIX_IN6:
396 	case SIOCCIFPREFIX_IN6:
397 	case SIOCSGIFPREFIX_IN6:
398 	case SIOCGIFPREFIX_IN6:
399 		log(LOG_NOTICE,
400 		    "prefix ioctls are now invalidated. "
401 		    "please use ifconfig.\n");
402 		return EOPNOTSUPP;
403 	}
404 
405 	switch (cmd) {
406 	case SIOCALIFADDR:
407 	case SIOCDLIFADDR:
408 		if (!privileged)
409 			return EPERM;
410 		/* FALLTHROUGH */
411 	case SIOCGLIFADDR:
412 		return in6_lifaddr_ioctl(so, cmd, data, ifp, l);
413 	}
414 
415 	/*
416 	 * Find address for this interface, if it exists.
417 	 *
418 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
419 	 * only, and used the first interface address as the target of other
420 	 * operations (without checking ifra_addr).  This was because netinet
421 	 * code/API assumed at most 1 interface address per interface.
422 	 * Since IPv6 allows a node to assign multiple addresses
423 	 * on a single interface, we almost always look and check the
424 	 * presence of ifra_addr, and reject invalid ones here.
425 	 * It also decreases duplicated code among SIOC*_IN6 operations.
426 	 */
427 	switch (cmd) {
428 	case SIOCAIFADDR_IN6:
429 	case SIOCSIFPHYADDR_IN6:
430 		sa6 = &ifra->ifra_addr;
431 		break;
432 	case SIOCSIFADDR_IN6:
433 	case SIOCGIFADDR_IN6:
434 	case SIOCSIFDSTADDR_IN6:
435 	case SIOCSIFNETMASK_IN6:
436 	case SIOCGIFDSTADDR_IN6:
437 	case SIOCGIFNETMASK_IN6:
438 	case SIOCDIFADDR_IN6:
439 	case SIOCGIFPSRCADDR_IN6:
440 	case SIOCGIFPDSTADDR_IN6:
441 	case SIOCGIFAFLAG_IN6:
442 	case SIOCSNDFLUSH_IN6:
443 	case SIOCSPFXFLUSH_IN6:
444 	case SIOCSRTRFLUSH_IN6:
445 	case SIOCGIFALIFETIME_IN6:
446 	case SIOCGIFSTAT_IN6:
447 	case SIOCGIFSTAT_ICMP6:
448 		sa6 = &ifr->ifr_addr;
449 		break;
450 	default:
451 		sa6 = NULL;
452 		break;
453 	}
454 	if (sa6 && sa6->sin6_family == AF_INET6) {
455 		if (sa6->sin6_scope_id != 0)
456 			error = sa6_embedscope(sa6, 0);
457 		else
458 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
459 		if (error != 0)
460 			return error;
461 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
462 	} else
463 		ia = NULL;
464 
465 	switch (cmd) {
466 	case SIOCSIFADDR_IN6:
467 	case SIOCSIFDSTADDR_IN6:
468 	case SIOCSIFNETMASK_IN6:
469 		/*
470 		 * Since IPv6 allows a node to assign multiple addresses
471 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
472 		 */
473 		return EINVAL;
474 
475 	case SIOCDIFADDR_IN6:
476 		/*
477 		 * for IPv4, we look for existing in_ifaddr here to allow
478 		 * "ifconfig if0 delete" to remove the first IPv4 address on
479 		 * the interface.  For IPv6, as the spec allows multiple
480 		 * interface address from the day one, we consider "remove the
481 		 * first one" semantics to be not preferable.
482 		 */
483 		if (ia == NULL)
484 			return EADDRNOTAVAIL;
485 		/* FALLTHROUGH */
486 	case SIOCAIFADDR_IN6:
487 		/*
488 		 * We always require users to specify a valid IPv6 address for
489 		 * the corresponding operation.
490 		 */
491 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
492 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
493 			return EAFNOSUPPORT;
494 		if (!privileged)
495 			return EPERM;
496 
497 		break;
498 
499 	case SIOCGIFADDR_IN6:
500 		/* This interface is basically deprecated. use SIOCGIFCONF. */
501 		/* FALLTHROUGH */
502 	case SIOCGIFAFLAG_IN6:
503 	case SIOCGIFNETMASK_IN6:
504 	case SIOCGIFDSTADDR_IN6:
505 	case SIOCGIFALIFETIME_IN6:
506 		/* must think again about its semantics */
507 		if (ia == NULL)
508 			return EADDRNOTAVAIL;
509 		break;
510 	}
511 
512 	switch (cmd) {
513 
514 	case SIOCGIFADDR_IN6:
515 		ifr->ifr_addr = ia->ia_addr;
516 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
517 			return error;
518 		break;
519 
520 	case SIOCGIFDSTADDR_IN6:
521 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
522 			return EINVAL;
523 		/*
524 		 * XXX: should we check if ifa_dstaddr is NULL and return
525 		 * an error?
526 		 */
527 		ifr->ifr_dstaddr = ia->ia_dstaddr;
528 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
529 			return error;
530 		break;
531 
532 	case SIOCGIFNETMASK_IN6:
533 		ifr->ifr_addr = ia->ia_prefixmask;
534 		break;
535 
536 	case SIOCGIFAFLAG_IN6:
537 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
538 		break;
539 
540 	case SIOCGIFSTAT_IN6:
541 		if (ifp == NULL)
542 			return EINVAL;
543 		bzero(&ifr->ifr_ifru.ifru_stat,
544 		    sizeof(ifr->ifr_ifru.ifru_stat));
545 		ifr->ifr_ifru.ifru_stat =
546 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
547 		break;
548 
549 	case SIOCGIFSTAT_ICMP6:
550 		if (ifp == NULL)
551 			return EINVAL;
552 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
553 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
554 		ifr->ifr_ifru.ifru_icmp6stat =
555 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
556 		break;
557 
558 	case SIOCGIFALIFETIME_IN6:
559 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
560 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
561 			time_t maxexpire;
562 			struct in6_addrlifetime *retlt =
563 			    &ifr->ifr_ifru.ifru_lifetime;
564 
565 			/*
566 			 * XXX: adjust expiration time assuming time_t is
567 			 * signed.
568 			 */
569 			maxexpire = ((time_t)~0) &
570 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
571 			if (ia->ia6_lifetime.ia6t_vltime <
572 			    maxexpire - ia->ia6_updatetime) {
573 				retlt->ia6t_expire = ia->ia6_updatetime +
574 				    ia->ia6_lifetime.ia6t_vltime;
575 			} else
576 				retlt->ia6t_expire = maxexpire;
577 		}
578 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
579 			time_t maxexpire;
580 			struct in6_addrlifetime *retlt =
581 			    &ifr->ifr_ifru.ifru_lifetime;
582 
583 			/*
584 			 * XXX: adjust expiration time assuming time_t is
585 			 * signed.
586 			 */
587 			maxexpire = ((time_t)~0) &
588 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
589 			if (ia->ia6_lifetime.ia6t_pltime <
590 			    maxexpire - ia->ia6_updatetime) {
591 				retlt->ia6t_preferred = ia->ia6_updatetime +
592 				    ia->ia6_lifetime.ia6t_pltime;
593 			} else
594 				retlt->ia6t_preferred = maxexpire;
595 		}
596 		break;
597 
598 	case SIOCAIFADDR_IN6:
599 	{
600 		int i;
601 		struct nd_prefixctl pr0;
602 		struct nd_prefix *pr;
603 
604 		/* reject read-only flags */
605 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
606 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
607 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
608 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
609 			return EINVAL;
610 		}
611 		/*
612 		 * first, make or update the interface address structure,
613 		 * and link it to the list.
614 		 */
615 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
616 			return error;
617 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
618 		    == NULL) {
619 		    	/*
620 			 * this can happen when the user specify the 0 valid
621 			 * lifetime.
622 			 */
623 			break;
624 		}
625 
626 		/*
627 		 * then, make the prefix on-link on the interface.
628 		 * XXX: we'd rather create the prefix before the address, but
629 		 * we need at least one address to install the corresponding
630 		 * interface route, so we configure the address first.
631 		 */
632 
633 		/*
634 		 * convert mask to prefix length (prefixmask has already
635 		 * been validated in in6_update_ifa().
636 		 */
637 		bzero(&pr0, sizeof(pr0));
638 		pr0.ndpr_ifp = ifp;
639 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
640 		    NULL);
641 		if (pr0.ndpr_plen == 128) {
642 			break;	/* we don't need to install a host route. */
643 		}
644 		pr0.ndpr_prefix = ifra->ifra_addr;
645 		/* apply the mask for safety. */
646 		for (i = 0; i < 4; i++) {
647 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
648 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
649 		}
650 		/*
651 		 * XXX: since we don't have an API to set prefix (not address)
652 		 * lifetimes, we just use the same lifetimes as addresses.
653 		 * The (temporarily) installed lifetimes can be overridden by
654 		 * later advertised RAs (when accept_rtadv is non 0), which is
655 		 * an intended behavior.
656 		 */
657 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
658 		pr0.ndpr_raf_auto =
659 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
660 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
661 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
662 
663 		/* add the prefix if not yet. */
664 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
665 			/*
666 			 * nd6_prelist_add will install the corresponding
667 			 * interface route.
668 			 */
669 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
670 				return error;
671 			if (pr == NULL) {
672 				log(LOG_ERR, "nd6_prelist_add succeeded but "
673 				    "no prefix\n");
674 				return EINVAL; /* XXX panic here? */
675 			}
676 		}
677 
678 		/* relate the address to the prefix */
679 		if (ia->ia6_ndpr == NULL) {
680 			ia->ia6_ndpr = pr;
681 			pr->ndpr_refcnt++;
682 
683 			/*
684 			 * If this is the first autoconf address from the
685 			 * prefix, create a temporary address as well
686 			 * (when required).
687 			 */
688 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
689 			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
690 				int e;
691 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
692 					log(LOG_NOTICE, "in6_control: failed "
693 					    "to create a temporary address, "
694 					    "errno=%d\n", e);
695 				}
696 			}
697 		}
698 
699 		/*
700 		 * this might affect the status of autoconfigured addresses,
701 		 * that is, this address might make other addresses detached.
702 		 */
703 		pfxlist_onlink_check();
704 
705 #ifdef PFIL_HOOKS
706 		(void)pfil_run_hooks(&if_pfil, (struct mbuf **)SIOCAIFADDR_IN6,
707 		    ifp, PFIL_IFADDR);
708 #endif
709 
710 		break;
711 	}
712 
713 	case SIOCDIFADDR_IN6:
714 	{
715 		struct nd_prefix *pr;
716 
717 		/*
718 		 * If the address being deleted is the only one that owns
719 		 * the corresponding prefix, expire the prefix as well.
720 		 * XXX: theoretically, we don't have to worry about such
721 		 * relationship, since we separate the address management
722 		 * and the prefix management.  We do this, however, to provide
723 		 * as much backward compatibility as possible in terms of
724 		 * the ioctl operation.
725 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
726 		 */
727 		pr = ia->ia6_ndpr;
728 		in6_purgeaddr(&ia->ia_ifa);
729 		if (pr && pr->ndpr_refcnt == 0)
730 			prelist_remove(pr);
731 #ifdef PFIL_HOOKS
732 		(void)pfil_run_hooks(&if_pfil, (struct mbuf **)SIOCDIFADDR_IN6,
733 		    ifp, PFIL_IFADDR);
734 #endif
735 		break;
736 	}
737 
738 	default:
739 		if (ifp == NULL || ifp->if_ioctl == 0)
740 			return EOPNOTSUPP;
741 		error = ((*ifp->if_ioctl)(ifp, cmd, data));
742 		return error;
743 	}
744 
745 	return 0;
746 }
747 
748 int
749 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
750     struct lwp *l)
751 {
752 	int error, privileged, s;
753 
754 	privileged = 0;
755 	if (l && !kauth_authorize_generic(l->l_cred,
756 	    KAUTH_GENERIC_ISSUSER, NULL))
757 		privileged++;
758 
759 	s = splnet();
760 	error = in6_control1(so , cmd, data, ifp, l, privileged);
761 	splx(s);
762 	return error;
763 }
764 
765 /*
766  * Update parameters of an IPv6 interface address.
767  * If necessary, a new entry is created and linked into address chains.
768  * This function is separated from in6_control().
769  * XXX: should this be performed under splnet()?
770  */
771 static int
772 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
773     struct in6_ifaddr *ia, int flags)
774 {
775 	int error = 0, hostIsNew = 0, plen = -1;
776 	struct in6_ifaddr *oia;
777 	struct sockaddr_in6 dst6;
778 	struct in6_addrlifetime *lt;
779 	struct in6_multi_mship *imm;
780 	struct in6_multi *in6m_sol;
781 	struct rtentry *rt;
782 	int dad_delay;
783 
784 	in6m_sol = NULL;
785 
786 	/* Validate parameters */
787 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
788 		return EINVAL;
789 
790 	/*
791 	 * The destination address for a p2p link must have a family
792 	 * of AF_UNSPEC or AF_INET6.
793 	 */
794 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
795 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
796 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
797 		return EAFNOSUPPORT;
798 	/*
799 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
800 	 * does not carry fields other than sin6_len.
801 	 */
802 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
803 		return EINVAL;
804 	/*
805 	 * Because the IPv6 address architecture is classless, we require
806 	 * users to specify a (non 0) prefix length (mask) for a new address.
807 	 * We also require the prefix (when specified) mask is valid, and thus
808 	 * reject a non-consecutive mask.
809 	 */
810 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
811 		return EINVAL;
812 	if (ifra->ifra_prefixmask.sin6_len != 0) {
813 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
814 		    (u_char *)&ifra->ifra_prefixmask +
815 		    ifra->ifra_prefixmask.sin6_len);
816 		if (plen <= 0)
817 			return EINVAL;
818 	} else {
819 		/*
820 		 * In this case, ia must not be NULL.  We just use its prefix
821 		 * length.
822 		 */
823 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
824 	}
825 	/*
826 	 * If the destination address on a p2p interface is specified,
827 	 * and the address is a scoped one, validate/set the scope
828 	 * zone identifier.
829 	 */
830 	dst6 = ifra->ifra_dstaddr;
831 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
832 	    (dst6.sin6_family == AF_INET6)) {
833 		struct in6_addr in6_tmp;
834 		u_int32_t zoneid;
835 
836 		in6_tmp = dst6.sin6_addr;
837 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
838 			return EINVAL; /* XXX: should be impossible */
839 
840 		if (dst6.sin6_scope_id != 0) {
841 			if (dst6.sin6_scope_id != zoneid)
842 				return EINVAL;
843 		} else		/* user omit to specify the ID. */
844 			dst6.sin6_scope_id = zoneid;
845 
846 		/* convert into the internal form */
847 		if (sa6_embedscope(&dst6, 0))
848 			return EINVAL; /* XXX: should be impossible */
849 	}
850 	/*
851 	 * The destination address can be specified only for a p2p or a
852 	 * loopback interface.  If specified, the corresponding prefix length
853 	 * must be 128.
854 	 */
855 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
856 #ifdef FORCE_P2PPLEN
857 		int i;
858 #endif
859 
860 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
861 			/* XXX: noisy message */
862 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
863 			    "be specified for a p2p or a loopback IF only\n"));
864 			return EINVAL;
865 		}
866 		if (plen != 128) {
867 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
868 			    "be 128 when dstaddr is specified\n"));
869 #ifdef FORCE_P2PPLEN
870 			/*
871 			 * To be compatible with old configurations,
872 			 * such as ifconfig gif0 inet6 2001::1 2001::2
873 			 * prefixlen 126, we override the specified
874 			 * prefixmask as if the prefix length was 128.
875 			 */
876 			ifra->ifra_prefixmask.sin6_len =
877 			    sizeof(struct sockaddr_in6);
878 			for (i = 0; i < 4; i++)
879 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
880 				    0xffffffff;
881 			plen = 128;
882 #else
883 			return EINVAL;
884 #endif
885 		}
886 	}
887 	/* lifetime consistency check */
888 	lt = &ifra->ifra_lifetime;
889 	if (lt->ia6t_pltime > lt->ia6t_vltime)
890 		return EINVAL;
891 	if (lt->ia6t_vltime == 0) {
892 		/*
893 		 * the following log might be noisy, but this is a typical
894 		 * configuration mistake or a tool's bug.
895 		 */
896 		nd6log((LOG_INFO,
897 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
898 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr)));
899 
900 		if (ia == NULL)
901 			return 0; /* there's nothing to do */
902 	}
903 
904 	/*
905 	 * If this is a new address, allocate a new ifaddr and link it
906 	 * into chains.
907 	 */
908 	if (ia == NULL) {
909 		hostIsNew = 1;
910 		/*
911 		 * When in6_update_ifa() is called in a process of a received
912 		 * RA, it is called under an interrupt context.  So, we should
913 		 * call malloc with M_NOWAIT.
914 		 */
915 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
916 		    M_NOWAIT);
917 		if (ia == NULL)
918 			return ENOBUFS;
919 		bzero((void *)ia, sizeof(*ia));
920 		LIST_INIT(&ia->ia6_memberships);
921 		/* Initialize the address and masks, and put time stamp */
922 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
923 		ia->ia_addr.sin6_family = AF_INET6;
924 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
925 		ia->ia6_createtime = time_second;
926 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
927 			/*
928 			 * XXX: some functions expect that ifa_dstaddr is not
929 			 * NULL for p2p interfaces.
930 			 */
931 			ia->ia_ifa.ifa_dstaddr =
932 			    (struct sockaddr *)&ia->ia_dstaddr;
933 		} else {
934 			ia->ia_ifa.ifa_dstaddr = NULL;
935 		}
936 		ia->ia_ifa.ifa_netmask =
937 		    (struct sockaddr *)&ia->ia_prefixmask;
938 
939 		ia->ia_ifp = ifp;
940 		if ((oia = in6_ifaddr) != NULL) {
941 			for ( ; oia->ia_next; oia = oia->ia_next)
942 				continue;
943 			oia->ia_next = ia;
944 		} else
945 			in6_ifaddr = ia;
946 		/* gain a refcnt for the link from in6_ifaddr */
947 		IFAREF(&ia->ia_ifa);
948 
949 		TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
950 		/* gain another refcnt for the link from if_addrlist */
951 		IFAREF(&ia->ia_ifa);
952 	}
953 
954 	/* update timestamp */
955 	ia->ia6_updatetime = time_second;
956 
957 	/* set prefix mask */
958 	if (ifra->ifra_prefixmask.sin6_len) {
959 		/*
960 		 * We prohibit changing the prefix length of an existing
961 		 * address, because
962 		 * + such an operation should be rare in IPv6, and
963 		 * + the operation would confuse prefix management.
964 		 */
965 		if (ia->ia_prefixmask.sin6_len &&
966 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
967 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
968 			    " existing (%s) address should not be changed\n",
969 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
970 			error = EINVAL;
971 			goto unlink;
972 		}
973 		ia->ia_prefixmask = ifra->ifra_prefixmask;
974 	}
975 
976 	/*
977 	 * If a new destination address is specified, scrub the old one and
978 	 * install the new destination.  Note that the interface must be
979 	 * p2p or loopback (see the check above.)
980 	 */
981 	if (dst6.sin6_family == AF_INET6 &&
982 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
983 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
984 		    rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0) {
985 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
986 			    "a route to the old destination: %s\n",
987 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
988 			/* proceed anyway... */
989 		} else
990 			ia->ia_flags &= ~IFA_ROUTE;
991 		ia->ia_dstaddr = dst6;
992 	}
993 
994 	/*
995 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
996 	 * to see if the address is deprecated or invalidated, but initialize
997 	 * these members for applications.
998 	 */
999 	ia->ia6_lifetime = ifra->ifra_lifetime;
1000 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1001 		ia->ia6_lifetime.ia6t_expire =
1002 		    time_second + ia->ia6_lifetime.ia6t_vltime;
1003 	} else
1004 		ia->ia6_lifetime.ia6t_expire = 0;
1005 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1006 		ia->ia6_lifetime.ia6t_preferred =
1007 		    time_second + ia->ia6_lifetime.ia6t_pltime;
1008 	} else
1009 		ia->ia6_lifetime.ia6t_preferred = 0;
1010 
1011 	/* reset the interface and routing table appropriately. */
1012 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1013 		goto unlink;
1014 
1015 	/*
1016 	 * configure address flags.
1017 	 */
1018 	ia->ia6_flags = ifra->ifra_flags;
1019 	/*
1020 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1021 	 * userland, make it deprecated.
1022 	 */
1023 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1024 		ia->ia6_lifetime.ia6t_pltime = 0;
1025 		ia->ia6_lifetime.ia6t_preferred = time_second;
1026 	}
1027 
1028 	/*
1029 	 * Make the address tentative before joining multicast addresses,
1030 	 * so that corresponding MLD responses would not have a tentative
1031 	 * source address.
1032 	 */
1033 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1034 	if (hostIsNew && in6if_do_dad(ifp))
1035 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1036 
1037 	/*
1038 	 * We are done if we have simply modified an existing address.
1039 	 */
1040 	if (!hostIsNew)
1041 		return error;
1042 
1043 	/*
1044 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1045 	 * not just go to unlink.
1046 	 */
1047 
1048 	/* join necessary multicast groups */
1049 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1050 		struct sockaddr_in6 mltaddr, mltmask;
1051 		struct in6_addr llsol;
1052 
1053 		/* join solicited multicast addr for new host id */
1054 		bzero(&llsol, sizeof(struct in6_addr));
1055 		llsol.s6_addr16[0] = htons(0xff02);
1056 		llsol.s6_addr32[1] = 0;
1057 		llsol.s6_addr32[2] = htonl(1);
1058 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1059 		llsol.s6_addr8[12] = 0xff;
1060 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1061 			/* XXX: should not happen */
1062 			log(LOG_ERR, "in6_update_ifa: "
1063 			    "in6_setscope failed\n");
1064 			goto cleanup;
1065 		}
1066 		dad_delay = 0;
1067 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1068 			/*
1069 			 * We need a random delay for DAD on the address
1070 			 * being configured.  It also means delaying
1071 			 * transmission of the corresponding MLD report to
1072 			 * avoid report collision.
1073 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1074 			 */
1075 			dad_delay = arc4random() %
1076 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1077 		}
1078 
1079 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1080 		/* join solicited multicast addr for new host id */
1081 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
1082 		if (!imm) {
1083 			nd6log((LOG_ERR,
1084 			    "in6_update_ifa: addmulti "
1085 			    "failed for %s on %s (errno=%d)\n",
1086 			    ip6_sprintf(&llsol), if_name(ifp), error));
1087 			goto cleanup;
1088 		}
1089 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1090 		in6m_sol = imm->i6mm_maddr;
1091 
1092 		sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
1093 
1094 		/*
1095 		 * join link-local all-nodes address
1096 		 */
1097 		sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
1098 		    0, 0, 0);
1099 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1100 			goto cleanup; /* XXX: should not fail */
1101 
1102 		/*
1103 		 * XXX: do we really need this automatic routes?
1104 		 * We should probably reconsider this stuff.  Most applications
1105 		 * actually do not need the routes, since they usually specify
1106 		 * the outgoing interface.
1107 		 */
1108 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1109 		if (rt) {
1110 			if (memcmp(&mltaddr.sin6_addr,
1111 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1112 			    MLTMASK_LEN)) {
1113 				RTFREE(rt);
1114 				rt = NULL;
1115 			} else if (rt->rt_ifp != ifp) {
1116 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1117 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1118 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1119 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1120 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1121 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1122 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1123 				rt_replace_ifa(rt, &ia->ia_ifa);
1124 				rt->rt_ifp = ifp;
1125 			}
1126 		}
1127 		if (!rt) {
1128 			struct rt_addrinfo info;
1129 
1130 			bzero(&info, sizeof(info));
1131 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1132 			info.rti_info[RTAX_GATEWAY] =
1133 			    (struct sockaddr *)&ia->ia_addr;
1134 			info.rti_info[RTAX_NETMASK] =
1135 			    (struct sockaddr *)&mltmask;
1136 			info.rti_info[RTAX_IFA] =
1137 			    (struct sockaddr *)&ia->ia_addr;
1138 			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1139 			info.rti_flags = RTF_UP | RTF_CLONING;
1140 			error = rtrequest1(RTM_ADD, &info, NULL);
1141 			if (error)
1142 				goto cleanup;
1143 		} else {
1144 			RTFREE(rt);
1145 		}
1146 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1147 		if (!imm) {
1148 			nd6log((LOG_WARNING,
1149 			    "in6_update_ifa: addmulti failed for "
1150 			    "%s on %s (errno=%d)\n",
1151 			    ip6_sprintf(&mltaddr.sin6_addr),
1152 			    if_name(ifp), error));
1153 			goto cleanup;
1154 		}
1155 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1156 
1157 		/*
1158 		 * join node information group address
1159 		 */
1160 		dad_delay = 0;
1161 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1162 			/*
1163 			 * The spec doesn't say anything about delay for this
1164 			 * group, but the same logic should apply.
1165 			 */
1166 			dad_delay = arc4random() %
1167 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1168 		}
1169 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
1170 			;
1171 		else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1172 		          dad_delay)) == NULL) { /* XXX jinmei */
1173 			nd6log((LOG_WARNING, "in6_update_ifa: "
1174 			    "addmulti failed for %s on %s (errno=%d)\n",
1175 			    ip6_sprintf(&mltaddr.sin6_addr),
1176 			    if_name(ifp), error));
1177 			/* XXX not very fatal, go on... */
1178 		} else {
1179 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1180 		}
1181 
1182 
1183 		/*
1184 		 * join interface-local all-nodes address.
1185 		 * (ff01::1%ifN, and ff01::%ifN/32)
1186 		 */
1187 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1188 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1189 			goto cleanup; /* XXX: should not fail */
1190 
1191 		/* XXX: again, do we really need the route? */
1192 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1193 		if (rt) {
1194 			/* 32bit came from "mltmask" */
1195 			if (memcmp(&mltaddr.sin6_addr,
1196 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1197 			    32 / NBBY)) {
1198 				RTFREE(rt);
1199 				rt = NULL;
1200 			} else if (rt->rt_ifp != ifp) {
1201 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1202 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1203 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1204 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1205 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1206 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1207 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1208 				rt_replace_ifa(rt, &ia->ia_ifa);
1209 				rt->rt_ifp = ifp;
1210 			}
1211 		}
1212 		if (!rt) {
1213 			struct rt_addrinfo info;
1214 
1215 			bzero(&info, sizeof(info));
1216 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1217 			info.rti_info[RTAX_GATEWAY] =
1218 			    (struct sockaddr *)&ia->ia_addr;
1219 			info.rti_info[RTAX_NETMASK] =
1220 			    (struct sockaddr *)&mltmask;
1221 			info.rti_info[RTAX_IFA] =
1222 			    (struct sockaddr *)&ia->ia_addr;
1223 			info.rti_flags = RTF_UP | RTF_CLONING;
1224 			error = rtrequest1(RTM_ADD, &info, NULL);
1225 			if (error)
1226 				goto cleanup;
1227 #undef	MLTMASK_LEN
1228 		} else {
1229 			RTFREE(rt);
1230 		}
1231 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1232 		if (!imm) {
1233 			nd6log((LOG_WARNING, "in6_update_ifa: "
1234 			    "addmulti failed for %s on %s (errno=%d)\n",
1235 			    ip6_sprintf(&mltaddr.sin6_addr),
1236 			    if_name(ifp), error));
1237 			goto cleanup;
1238 		} else {
1239 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1240 		}
1241 	}
1242 
1243 	/*
1244 	 * Perform DAD, if needed.
1245 	 * XXX It may be of use, if we can administratively
1246 	 * disable DAD.
1247 	 */
1248 	if (hostIsNew && in6if_do_dad(ifp) &&
1249 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1250 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1251 	{
1252 		int mindelay, maxdelay;
1253 
1254 		dad_delay = 0;
1255 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1256 			/*
1257 			 * We need to impose a delay before sending an NS
1258 			 * for DAD.  Check if we also needed a delay for the
1259 			 * corresponding MLD message.  If we did, the delay
1260 			 * should be larger than the MLD delay (this could be
1261 			 * relaxed a bit, but this simple logic is at least
1262 			 * safe).
1263 			 */
1264 			mindelay = 0;
1265 			if (in6m_sol != NULL &&
1266 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1267 				mindelay = in6m_sol->in6m_timer;
1268 			}
1269 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1270 			if (maxdelay - mindelay == 0)
1271 				dad_delay = 0;
1272 			else {
1273 				dad_delay =
1274 				    (arc4random() % (maxdelay - mindelay)) +
1275 				    mindelay;
1276 			}
1277 		}
1278 		nd6_dad_start((struct ifaddr *)ia, dad_delay);
1279 	}
1280 
1281 	return error;
1282 
1283   unlink:
1284 	/*
1285 	 * XXX: if a change of an existing address failed, keep the entry
1286 	 * anyway.
1287 	 */
1288 	if (hostIsNew)
1289 		in6_unlink_ifa(ia, ifp);
1290 	return error;
1291 
1292   cleanup:
1293 	in6_purgeaddr(&ia->ia_ifa);
1294 	return error;
1295 }
1296 
1297 int
1298 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1299     struct in6_ifaddr *ia, int flags)
1300 {
1301 	int rc, s;
1302 
1303 	s = splnet();
1304 	rc = in6_update_ifa1(ifp, ifra, ia, flags);
1305 	splx(s);
1306 	return rc;
1307 }
1308 
1309 void
1310 in6_purgeaddr(struct ifaddr *ifa)
1311 {
1312 	struct ifnet *ifp = ifa->ifa_ifp;
1313 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1314 	struct in6_multi_mship *imm;
1315 
1316 	/* stop DAD processing */
1317 	nd6_dad_stop(ifa);
1318 
1319 	/*
1320 	 * delete route to the destination of the address being purged.
1321 	 * The interface must be p2p or loopback in this case.
1322 	 */
1323 	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1324 		int e;
1325 
1326 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1327 		    != 0) {
1328 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1329 			    "a route to the p2p destination: %s on %s, "
1330 			    "errno=%d\n",
1331 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1332 			    e);
1333 			/* proceed anyway... */
1334 		} else
1335 			ia->ia_flags &= ~IFA_ROUTE;
1336 	}
1337 
1338 	/* Remove ownaddr's loopback rtentry, if it exists. */
1339 	in6_ifremloop(&(ia->ia_ifa));
1340 
1341 	/*
1342 	 * leave from multicast groups we have joined for the interface
1343 	 */
1344 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1345 		LIST_REMOVE(imm, i6mm_chain);
1346 		in6_leavegroup(imm);
1347 	}
1348 
1349 	in6_unlink_ifa(ia, ifp);
1350 }
1351 
1352 static void
1353 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1354 {
1355 	struct in6_ifaddr *oia;
1356 	int	s = splnet();
1357 
1358 	TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1359 	/* release a refcnt for the link from if_addrlist */
1360 	IFAFREE(&ia->ia_ifa);
1361 
1362 	oia = ia;
1363 	if (oia == (ia = in6_ifaddr))
1364 		in6_ifaddr = ia->ia_next;
1365 	else {
1366 		while (ia->ia_next && (ia->ia_next != oia))
1367 			ia = ia->ia_next;
1368 		if (ia->ia_next)
1369 			ia->ia_next = oia->ia_next;
1370 		else {
1371 			/* search failed */
1372 			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1373 		}
1374 	}
1375 
1376 	/*
1377 	 * XXX thorpej@NetBSD.org -- if the interface is going
1378 	 * XXX away, don't save the multicast entries, delete them!
1379 	 */
1380 	if (LIST_EMPTY(&oia->ia6_multiaddrs))
1381 		;
1382 	else if (oia->ia_ifa.ifa_ifp->if_output == if_nulloutput) {
1383 		struct in6_multi *in6m, *next;
1384 
1385 		for (in6m = LIST_FIRST(&oia->ia6_multiaddrs); in6m != NULL;
1386 		     in6m = next) {
1387 			next = LIST_NEXT(in6m, in6m_entry);
1388 			in6_delmulti(in6m);
1389 		}
1390 	} else
1391 		in6_savemkludge(oia);
1392 
1393 	/*
1394 	 * Release the reference to the base prefix.  There should be a
1395 	 * positive reference.
1396 	 */
1397 	if (oia->ia6_ndpr == NULL) {
1398 		nd6log((LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1399 		    "%p has no prefix\n", oia));
1400 	} else {
1401 		oia->ia6_ndpr->ndpr_refcnt--;
1402 		oia->ia6_ndpr = NULL;
1403 	}
1404 
1405 	/*
1406 	 * Also, if the address being removed is autoconf'ed, call
1407 	 * pfxlist_onlink_check() since the release might affect the status of
1408 	 * other (detached) addresses.
1409 	 */
1410 	if ((oia->ia6_flags & IN6_IFF_AUTOCONF) != 0)
1411 		pfxlist_onlink_check();
1412 
1413 	/*
1414 	 * release another refcnt for the link from in6_ifaddr.
1415 	 * Note that we should decrement the refcnt at least once for all *BSD.
1416 	 */
1417 	IFAFREE(&oia->ia_ifa);
1418 
1419 	splx(s);
1420 }
1421 
1422 void
1423 in6_purgeif(struct ifnet *ifp)
1424 {
1425 	struct ifaddr *ifa, *nifa;
1426 
1427 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa) {
1428 		nifa = TAILQ_NEXT(ifa, ifa_list);
1429 		if (ifa->ifa_addr->sa_family != AF_INET6)
1430 			continue;
1431 		in6_purgeaddr(ifa);
1432 	}
1433 
1434 	in6_ifdetach(ifp);
1435 }
1436 
1437 /*
1438  * SIOC[GAD]LIFADDR.
1439  *	SIOCGLIFADDR: get first address. (?)
1440  *	SIOCGLIFADDR with IFLR_PREFIX:
1441  *		get first address that matches the specified prefix.
1442  *	SIOCALIFADDR: add the specified address.
1443  *	SIOCALIFADDR with IFLR_PREFIX:
1444  *		add the specified prefix, filling hostid part from
1445  *		the first link-local address.  prefixlen must be <= 64.
1446  *	SIOCDLIFADDR: delete the specified address.
1447  *	SIOCDLIFADDR with IFLR_PREFIX:
1448  *		delete the first address that matches the specified prefix.
1449  * return values:
1450  *	EINVAL on invalid parameters
1451  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1452  *	other values may be returned from in6_ioctl()
1453  *
1454  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1455  * this is to accommodate address naming scheme other than RFC2374,
1456  * in the future.
1457  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1458  * address encoding scheme. (see figure on page 8)
1459  */
1460 static int
1461 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1462 	struct ifnet *ifp, struct lwp *l)
1463 {
1464 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1465 	struct ifaddr *ifa;
1466 	struct sockaddr *sa;
1467 
1468 	/* sanity checks */
1469 	if (!data || !ifp) {
1470 		panic("invalid argument to in6_lifaddr_ioctl");
1471 		/* NOTREACHED */
1472 	}
1473 
1474 	switch (cmd) {
1475 	case SIOCGLIFADDR:
1476 		/* address must be specified on GET with IFLR_PREFIX */
1477 		if ((iflr->flags & IFLR_PREFIX) == 0)
1478 			break;
1479 		/* FALLTHROUGH */
1480 	case SIOCALIFADDR:
1481 	case SIOCDLIFADDR:
1482 		/* address must be specified on ADD and DELETE */
1483 		sa = (struct sockaddr *)&iflr->addr;
1484 		if (sa->sa_family != AF_INET6)
1485 			return EINVAL;
1486 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1487 			return EINVAL;
1488 		/* XXX need improvement */
1489 		sa = (struct sockaddr *)&iflr->dstaddr;
1490 		if (sa->sa_family && sa->sa_family != AF_INET6)
1491 			return EINVAL;
1492 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1493 			return EINVAL;
1494 		break;
1495 	default: /* shouldn't happen */
1496 #if 0
1497 		panic("invalid cmd to in6_lifaddr_ioctl");
1498 		/* NOTREACHED */
1499 #else
1500 		return EOPNOTSUPP;
1501 #endif
1502 	}
1503 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1504 		return EINVAL;
1505 
1506 	switch (cmd) {
1507 	case SIOCALIFADDR:
1508 	    {
1509 		struct in6_aliasreq ifra;
1510 		struct in6_addr *xhostid = NULL;
1511 		int prefixlen;
1512 
1513 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1514 			struct sockaddr_in6 *sin6;
1515 
1516 			/*
1517 			 * xhostid is to fill in the hostid part of the
1518 			 * address.  xhostid points to the first link-local
1519 			 * address attached to the interface.
1520 			 */
1521 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1522 			if (!ifa)
1523 				return EADDRNOTAVAIL;
1524 			xhostid = IFA_IN6(ifa);
1525 
1526 		 	/* prefixlen must be <= 64. */
1527 			if (64 < iflr->prefixlen)
1528 				return EINVAL;
1529 			prefixlen = iflr->prefixlen;
1530 
1531 			/* hostid part must be zero. */
1532 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1533 			if (sin6->sin6_addr.s6_addr32[2] != 0
1534 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1535 				return EINVAL;
1536 			}
1537 		} else
1538 			prefixlen = iflr->prefixlen;
1539 
1540 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1541 		bzero(&ifra, sizeof(ifra));
1542 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1543 
1544 		bcopy(&iflr->addr, &ifra.ifra_addr,
1545 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1546 		if (xhostid) {
1547 			/* fill in hostid part */
1548 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1549 			    xhostid->s6_addr32[2];
1550 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1551 			    xhostid->s6_addr32[3];
1552 		}
1553 
1554 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1555 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1556 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1557 			if (xhostid) {
1558 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1559 				    xhostid->s6_addr32[2];
1560 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1561 				    xhostid->s6_addr32[3];
1562 			}
1563 		}
1564 
1565 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1566 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1567 
1568 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1569 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1570 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1571 		return in6_control(so, SIOCAIFADDR_IN6, (void *)&ifra, ifp, l);
1572 	    }
1573 	case SIOCGLIFADDR:
1574 	case SIOCDLIFADDR:
1575 	    {
1576 		struct in6_ifaddr *ia;
1577 		struct in6_addr mask, candidate, match;
1578 		struct sockaddr_in6 *sin6;
1579 		int cmp;
1580 
1581 		bzero(&mask, sizeof(mask));
1582 		if (iflr->flags & IFLR_PREFIX) {
1583 			/* lookup a prefix rather than address. */
1584 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1585 
1586 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1587 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1588 			match.s6_addr32[0] &= mask.s6_addr32[0];
1589 			match.s6_addr32[1] &= mask.s6_addr32[1];
1590 			match.s6_addr32[2] &= mask.s6_addr32[2];
1591 			match.s6_addr32[3] &= mask.s6_addr32[3];
1592 
1593 			/* if you set extra bits, that's wrong */
1594 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1595 				return EINVAL;
1596 
1597 			cmp = 1;
1598 		} else {
1599 			if (cmd == SIOCGLIFADDR) {
1600 				/* on getting an address, take the 1st match */
1601 				cmp = 0;	/* XXX */
1602 			} else {
1603 				/* on deleting an address, do exact match */
1604 				in6_prefixlen2mask(&mask, 128);
1605 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1606 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1607 
1608 				cmp = 1;
1609 			}
1610 		}
1611 
1612 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1613 			if (ifa->ifa_addr->sa_family != AF_INET6)
1614 				continue;
1615 			if (!cmp)
1616 				break;
1617 
1618 			/*
1619 			 * XXX: this is adhoc, but is necessary to allow
1620 			 * a user to specify fe80::/64 (not /10) for a
1621 			 * link-local address.
1622 			 */
1623 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1624 			in6_clearscope(&candidate);
1625 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1626 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1627 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1628 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1629 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1630 				break;
1631 		}
1632 		if (!ifa)
1633 			return EADDRNOTAVAIL;
1634 		ia = ifa2ia6(ifa);
1635 
1636 		if (cmd == SIOCGLIFADDR) {
1637 			int error;
1638 
1639 			/* fill in the if_laddrreq structure */
1640 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1641 			error = sa6_recoverscope(
1642 			    (struct sockaddr_in6 *)&iflr->addr);
1643 			if (error != 0)
1644 				return error;
1645 
1646 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1647 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1648 				    ia->ia_dstaddr.sin6_len);
1649 				error = sa6_recoverscope(
1650 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1651 				if (error != 0)
1652 					return error;
1653 			} else
1654 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1655 
1656 			iflr->prefixlen =
1657 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1658 
1659 			iflr->flags = ia->ia6_flags;	/* XXX */
1660 
1661 			return 0;
1662 		} else {
1663 			struct in6_aliasreq ifra;
1664 
1665 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1666 			bzero(&ifra, sizeof(ifra));
1667 			bcopy(iflr->iflr_name, ifra.ifra_name,
1668 			    sizeof(ifra.ifra_name));
1669 
1670 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1671 			    ia->ia_addr.sin6_len);
1672 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1673 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1674 				    ia->ia_dstaddr.sin6_len);
1675 			} else {
1676 				bzero(&ifra.ifra_dstaddr,
1677 				    sizeof(ifra.ifra_dstaddr));
1678 			}
1679 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1680 			    ia->ia_prefixmask.sin6_len);
1681 
1682 			ifra.ifra_flags = ia->ia6_flags;
1683 			return in6_control(so, SIOCDIFADDR_IN6, (void *)&ifra,
1684 			    ifp, l);
1685 		}
1686 	    }
1687 	}
1688 
1689 	return EOPNOTSUPP;	/* just for safety */
1690 }
1691 
1692 /*
1693  * Initialize an interface's internet6 address
1694  * and routing table entry.
1695  */
1696 static int
1697 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1698 	struct sockaddr_in6 *sin6, int newhost)
1699 {
1700 	int	error = 0, plen, ifacount = 0;
1701 	int	s = splnet();
1702 	struct ifaddr *ifa;
1703 
1704 	/*
1705 	 * Give the interface a chance to initialize
1706 	 * if this is its first address,
1707 	 * and to validate the address if necessary.
1708 	 */
1709 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1710 		if (ifa->ifa_addr == NULL)
1711 			continue;	/* just for safety */
1712 		if (ifa->ifa_addr->sa_family != AF_INET6)
1713 			continue;
1714 		ifacount++;
1715 	}
1716 
1717 	ia->ia_addr = *sin6;
1718 
1719 	if (ifacount <= 1 && ifp->if_ioctl &&
1720 	    (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (void *)ia))) {
1721 		splx(s);
1722 		return error;
1723 	}
1724 	splx(s);
1725 
1726 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1727 
1728 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1729 
1730 	/*
1731 	 * Special case:
1732 	 * If the destination address is specified for a point-to-point
1733 	 * interface, install a route to the destination as an interface
1734 	 * direct route.
1735 	 */
1736 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1737 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1738 		if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1739 				    RTF_UP | RTF_HOST)) != 0)
1740 			return error;
1741 		ia->ia_flags |= IFA_ROUTE;
1742 	}
1743 
1744 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1745 	if (newhost) {
1746 		/* set the rtrequest function to create llinfo */
1747 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1748 		in6_ifaddloop(&(ia->ia_ifa));
1749 	}
1750 
1751 	if (ifp->if_flags & IFF_MULTICAST)
1752 		in6_restoremkludge(ia, ifp);
1753 
1754 	return error;
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 *ifa;
1764 
1765 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1766 		if (ifa->ifa_addr == NULL)
1767 			continue;	/* just for safety */
1768 		if (ifa->ifa_addr->sa_family != AF_INET6)
1769 			continue;
1770 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1771 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1772 			     ignoreflags) != 0)
1773 				continue;
1774 			break;
1775 		}
1776 	}
1777 
1778 	return (struct in6_ifaddr *)ifa;
1779 }
1780 
1781 
1782 /*
1783  * find the internet address corresponding to a given interface and address.
1784  */
1785 struct in6_ifaddr *
1786 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1787 {
1788 	struct ifaddr *ifa;
1789 
1790 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1791 		if (ifa->ifa_addr == NULL)
1792 			continue;	/* just for safety */
1793 		if (ifa->ifa_addr->sa_family != AF_INET6)
1794 			continue;
1795 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1796 			break;
1797 	}
1798 
1799 	return (struct in6_ifaddr *)ifa;
1800 }
1801 
1802 /*
1803  * Convert IP6 address to printable (loggable) representation.
1804  */
1805 static int ip6round = 0;
1806 char *
1807 ip6_sprintf(const struct in6_addr *addr)
1808 {
1809 	static char ip6buf[8][48];
1810 	int i;
1811 	char *cp;
1812 	const u_int16_t *a = (const u_int16_t *)addr;
1813 	const u_int8_t *d;
1814 	int dcolon = 0;
1815 
1816 	ip6round = (ip6round + 1) & 7;
1817 	cp = ip6buf[ip6round];
1818 
1819 	for (i = 0; i < 8; i++) {
1820 		if (dcolon == 1) {
1821 			if (*a == 0) {
1822 				if (i == 7)
1823 					*cp++ = ':';
1824 				a++;
1825 				continue;
1826 			} else
1827 				dcolon = 2;
1828 		}
1829 		if (*a == 0) {
1830 			if (dcolon == 0 && *(a + 1) == 0) {
1831 				if (i == 0)
1832 					*cp++ = ':';
1833 				*cp++ = ':';
1834 				dcolon = 1;
1835 			} else {
1836 				*cp++ = '0';
1837 				*cp++ = ':';
1838 			}
1839 			a++;
1840 			continue;
1841 		}
1842 		d = (const u_char *)a;
1843 		*cp++ = hexdigits[*d >> 4];
1844 		*cp++ = hexdigits[*d++ & 0xf];
1845 		*cp++ = hexdigits[*d >> 4];
1846 		*cp++ = hexdigits[*d & 0xf];
1847 		*cp++ = ':';
1848 		a++;
1849 	}
1850 	*--cp = 0;
1851 	return ip6buf[ip6round];
1852 }
1853 
1854 /*
1855  * Determine if an address is on a local network.
1856  */
1857 int
1858 in6_localaddr(const struct in6_addr *in6)
1859 {
1860 	struct in6_ifaddr *ia;
1861 
1862 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1863 		return 1;
1864 
1865 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1866 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1867 					      &ia->ia_prefixmask.sin6_addr))
1868 			return 1;
1869 
1870 	return 0;
1871 }
1872 
1873 int
1874 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1875 {
1876 	struct in6_ifaddr *ia;
1877 
1878 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1879 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1880 		    &sa6->sin6_addr) &&
1881 #ifdef SCOPEDROUTING
1882 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1883 #endif
1884 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1885 			return 1; /* true */
1886 
1887 		/* XXX: do we still have to go thru the rest of the list? */
1888 	}
1889 
1890 	return 0;		/* false */
1891 }
1892 
1893 /*
1894  * return length of part which dst and src are equal
1895  * hard coding...
1896  */
1897 int
1898 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1899 {
1900 	int match = 0;
1901 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1902 	u_char *lim = s + 16, r;
1903 
1904 	while (s < lim)
1905 		if ((r = (*d++ ^ *s++)) != 0) {
1906 			while (r < 128) {
1907 				match++;
1908 				r <<= 1;
1909 			}
1910 			break;
1911 		} else
1912 			match += NBBY;
1913 	return match;
1914 }
1915 
1916 /* XXX: to be scope conscious */
1917 int
1918 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1919 {
1920 	int bytelen, bitlen;
1921 
1922 	/* sanity check */
1923 	if (len < 0 || len > 128) {
1924 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1925 		    len);
1926 		return 0;
1927 	}
1928 
1929 	bytelen = len / NBBY;
1930 	bitlen = len % NBBY;
1931 
1932 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1933 		return 0;
1934 	if (bitlen != 0 &&
1935 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
1936 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
1937 		return 0;
1938 
1939 	return 1;
1940 }
1941 
1942 void
1943 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1944 {
1945 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1946 	int bytelen, bitlen, i;
1947 
1948 	/* sanity check */
1949 	if (len < 0 || len > 128) {
1950 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1951 		    len);
1952 		return;
1953 	}
1954 
1955 	bzero(maskp, sizeof(*maskp));
1956 	bytelen = len / NBBY;
1957 	bitlen = len % NBBY;
1958 	for (i = 0; i < bytelen; i++)
1959 		maskp->s6_addr[i] = 0xff;
1960 	if (bitlen)
1961 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1962 }
1963 
1964 /*
1965  * return the best address out of the same scope. if no address was
1966  * found, return the first valid address from designated IF.
1967  */
1968 struct in6_ifaddr *
1969 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1970 {
1971 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1972 	struct ifaddr *ifa;
1973 	struct in6_ifaddr *besta = 0;
1974 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1975 
1976 	dep[0] = dep[1] = NULL;
1977 
1978 	/*
1979 	 * We first look for addresses in the same scope.
1980 	 * If there is one, return it.
1981 	 * If two or more, return one which matches the dst longest.
1982 	 * If none, return one of global addresses assigned other ifs.
1983 	 */
1984 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1985 		if (ifa->ifa_addr->sa_family != AF_INET6)
1986 			continue;
1987 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1988 			continue; /* XXX: is there any case to allow anycast? */
1989 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1990 			continue; /* don't use this interface */
1991 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1992 			continue;
1993 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1994 			if (ip6_use_deprecated)
1995 				dep[0] = (struct in6_ifaddr *)ifa;
1996 			continue;
1997 		}
1998 
1999 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2000 			/*
2001 			 * call in6_matchlen() as few as possible
2002 			 */
2003 			if (besta) {
2004 				if (blen == -1)
2005 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2006 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2007 				if (tlen > blen) {
2008 					blen = tlen;
2009 					besta = (struct in6_ifaddr *)ifa;
2010 				}
2011 			} else
2012 				besta = (struct in6_ifaddr *)ifa;
2013 		}
2014 	}
2015 	if (besta)
2016 		return besta;
2017 
2018 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2019 		if (ifa->ifa_addr->sa_family != AF_INET6)
2020 			continue;
2021 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2022 			continue; /* XXX: is there any case to allow anycast? */
2023 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2024 			continue; /* don't use this interface */
2025 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2026 			continue;
2027 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2028 			if (ip6_use_deprecated)
2029 				dep[1] = (struct in6_ifaddr *)ifa;
2030 			continue;
2031 		}
2032 
2033 		return (struct in6_ifaddr *)ifa;
2034 	}
2035 
2036 	/* use the last-resort values, that are, deprecated addresses */
2037 	if (dep[0])
2038 		return dep[0];
2039 	if (dep[1])
2040 		return dep[1];
2041 
2042 	return NULL;
2043 }
2044 
2045 /*
2046  * perform DAD when interface becomes IFF_UP.
2047  */
2048 void
2049 in6_if_up(struct ifnet *ifp)
2050 {
2051 	struct ifaddr *ifa;
2052 	struct in6_ifaddr *ia;
2053 
2054 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2055 		if (ifa->ifa_addr->sa_family != AF_INET6)
2056 			continue;
2057 		ia = (struct in6_ifaddr *)ifa;
2058 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2059 			/*
2060 			 * The TENTATIVE flag was likely set by hand
2061 			 * beforehand, implicitly indicating the need for DAD.
2062 			 * We may be able to skip the random delay in this
2063 			 * case, but we impose delays just in case.
2064 			 */
2065 			nd6_dad_start(ifa,
2066 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2067 		}
2068 	}
2069 
2070 	/*
2071 	 * special cases, like 6to4, are handled in in6_ifattach
2072 	 */
2073 	in6_ifattach(ifp, NULL);
2074 }
2075 
2076 int
2077 in6if_do_dad(struct ifnet *ifp)
2078 {
2079 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2080 		return 0;
2081 
2082 	switch (ifp->if_type) {
2083 	case IFT_FAITH:
2084 		/*
2085 		 * These interfaces do not have the IFF_LOOPBACK flag,
2086 		 * but loop packets back.  We do not have to do DAD on such
2087 		 * interfaces.  We should even omit it, because loop-backed
2088 		 * NS would confuse the DAD procedure.
2089 		 */
2090 		return 0;
2091 	default:
2092 		/*
2093 		 * Our DAD routine requires the interface up and running.
2094 		 * However, some interfaces can be up before the RUNNING
2095 		 * status.  Additionaly, users may try to assign addresses
2096 		 * before the interface becomes up (or running).
2097 		 * We simply skip DAD in such a case as a work around.
2098 		 * XXX: we should rather mark "tentative" on such addresses,
2099 		 * and do DAD after the interface becomes ready.
2100 		 */
2101 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2102 		    (IFF_UP|IFF_RUNNING))
2103 			return 0;
2104 
2105 		return 1;
2106 	}
2107 }
2108 
2109 /*
2110  * Calculate max IPv6 MTU through all the interfaces and store it
2111  * to in6_maxmtu.
2112  */
2113 void
2114 in6_setmaxmtu()
2115 {
2116 	unsigned long maxmtu = 0;
2117 	struct ifnet *ifp;
2118 
2119 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
2120 		/* this function can be called during ifnet initialization */
2121 		if (!ifp->if_afdata[AF_INET6])
2122 			continue;
2123 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2124 		    IN6_LINKMTU(ifp) > maxmtu)
2125 			maxmtu = IN6_LINKMTU(ifp);
2126 	}
2127 	if (maxmtu)	     /* update only when maxmtu is positive */
2128 		in6_maxmtu = maxmtu;
2129 }
2130 
2131 /*
2132  * Provide the length of interface identifiers to be used for the link attached
2133  * to the given interface.  The length should be defined in "IPv6 over
2134  * xxx-link" document.  Note that address architecture might also define
2135  * the length for a particular set of address prefixes, regardless of the
2136  * link type.  As clarified in rfc2462bis, those two definitions should be
2137  * consistent, and those really are as of August 2004.
2138  */
2139 int
2140 in6_if2idlen(struct ifnet *ifp)
2141 {
2142 	switch (ifp->if_type) {
2143 	case IFT_ETHER:		/* RFC2464 */
2144 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2145 	case IFT_L2VLAN:	/* ditto */
2146 	case IFT_IEEE80211:	/* ditto */
2147 	case IFT_FDDI:		/* RFC2467 */
2148 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2149 	case IFT_PPP:		/* RFC2472 */
2150 	case IFT_ARCNET:	/* RFC2497 */
2151 	case IFT_FRELAY:	/* RFC2590 */
2152 	case IFT_IEEE1394:	/* RFC3146 */
2153 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2154 	case IFT_LOOP:		/* XXX: is this really correct? */
2155 		return 64;
2156 	default:
2157 		/*
2158 		 * Unknown link type:
2159 		 * It might be controversial to use the today's common constant
2160 		 * of 64 for these cases unconditionally.  For full compliance,
2161 		 * we should return an error in this case.  On the other hand,
2162 		 * if we simply miss the standard for the link type or a new
2163 		 * standard is defined for a new link type, the IFID length
2164 		 * is very likely to be the common constant.  As a compromise,
2165 		 * we always use the constant, but make an explicit notice
2166 		 * indicating the "unknown" case.
2167 		 */
2168 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2169 		return 64;
2170 	}
2171 }
2172 
2173 void *
2174 in6_domifattach(struct ifnet *ifp)
2175 {
2176 	struct in6_ifextra *ext;
2177 
2178 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2179 	bzero(ext, sizeof(*ext));
2180 
2181 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2182 	    M_IFADDR, M_WAITOK);
2183 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2184 
2185 	ext->icmp6_ifstat =
2186 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2187 	    M_IFADDR, M_WAITOK);
2188 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2189 
2190 	ext->nd_ifinfo = nd6_ifattach(ifp);
2191 	ext->scope6_id = scope6_ifattach(ifp);
2192 	return ext;
2193 }
2194 
2195 void
2196 in6_domifdetach(struct ifnet *ifp, void *aux)
2197 {
2198 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2199 
2200 	nd6_ifdetach(ext->nd_ifinfo);
2201 	free(ext->in6_ifstat, M_IFADDR);
2202 	free(ext->icmp6_ifstat, M_IFADDR);
2203 	scope6_ifdetach(ext->scope6_id);
2204 	free(ext, M_IFADDR);
2205 }
2206 
2207 /*
2208  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2209  * v4 mapped addr or v4 compat addr
2210  */
2211 void
2212 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2213 {
2214 	bzero(sin, sizeof(*sin));
2215 	sin->sin_len = sizeof(struct sockaddr_in);
2216 	sin->sin_family = AF_INET;
2217 	sin->sin_port = sin6->sin6_port;
2218 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2219 }
2220 
2221 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2222 void
2223 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2224 {
2225 	bzero(sin6, sizeof(*sin6));
2226 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2227 	sin6->sin6_family = AF_INET6;
2228 	sin6->sin6_port = sin->sin_port;
2229 	sin6->sin6_addr.s6_addr32[0] = 0;
2230 	sin6->sin6_addr.s6_addr32[1] = 0;
2231 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2232 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2233 }
2234 
2235 /* Convert sockaddr_in6 into sockaddr_in. */
2236 void
2237 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2238 {
2239 	struct sockaddr_in *sin_p;
2240 	struct sockaddr_in6 sin6;
2241 
2242 	/*
2243 	 * Save original sockaddr_in6 addr and convert it
2244 	 * to sockaddr_in.
2245 	 */
2246 	sin6 = *(struct sockaddr_in6 *)nam;
2247 	sin_p = (struct sockaddr_in *)nam;
2248 	in6_sin6_2_sin(sin_p, &sin6);
2249 }
2250 
2251 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2252 void
2253 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2254 {
2255 	struct sockaddr_in *sin_p;
2256 	struct sockaddr_in6 *sin6_p;
2257 
2258 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2259 	sin_p = (struct sockaddr_in *)*nam;
2260 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2261 	free(*nam, M_SONAME);
2262 	*nam = (struct sockaddr *)sin6_p;
2263 }
2264