xref: /netbsd-src/sys/netinet6/in6.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: in6.c,v 1.90 2004/07/26 13:44:35 yamt 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 1.90 2004/07/26 13:44:35 yamt 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 
83 #include <net/if.h>
84 #include <net/if_types.h>
85 #include <net/route.h>
86 #include <net/if_dl.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <net/if_ether.h>
91 
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet6/nd6.h>
95 #include <netinet6/mld6_var.h>
96 #include <netinet6/ip6_mroute.h>
97 #include <netinet6/in6_ifattach.h>
98 
99 #include <net/net_osdep.h>
100 
101 #ifdef PFIL_HOOKS
102 #include <net/pfil.h>
103 #endif
104 
105 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
106 
107 /* enable backward compatibility code for obsoleted ioctls */
108 #define COMPAT_IN6IFIOCTL
109 
110 /*
111  * Definitions of some constant IP6 addresses.
112  */
113 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
114 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
115 const struct in6_addr in6addr_nodelocal_allnodes =
116 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
117 const struct in6_addr in6addr_linklocal_allnodes =
118 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
119 const struct in6_addr in6addr_linklocal_allrouters =
120 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
121 
122 const struct in6_addr in6mask0 = IN6MASK0;
123 const struct in6_addr in6mask32 = IN6MASK32;
124 const struct in6_addr in6mask64 = IN6MASK64;
125 const struct in6_addr in6mask96 = IN6MASK96;
126 const struct in6_addr in6mask128 = IN6MASK128;
127 
128 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
129 				     0, 0, IN6ADDR_ANY_INIT, 0};
130 
131 static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
132 	struct ifnet *, struct proc *));
133 static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
134 	struct sockaddr_in6 *, int));
135 static void in6_unlink_ifa __P((struct in6_ifaddr *, struct ifnet *));
136 
137 /*
138  * This structure is used to keep track of in6_multi chains which belong to
139  * deleted interface addresses.
140  */
141 static LIST_HEAD(, multi6_kludge) in6_mk; /* XXX BSS initialization */
142 
143 struct multi6_kludge {
144 	LIST_ENTRY(multi6_kludge) mk_entry;
145 	struct ifnet *mk_ifp;
146 	struct in6_multihead mk_head;
147 };
148 
149 /*
150  * Subroutine for in6_ifaddloop() and in6_ifremloop().
151  * This routine does actual work.
152  */
153 static void
154 in6_ifloop_request(int cmd, struct ifaddr *ifa)
155 {
156 	struct sockaddr_in6 lo_sa;
157 	struct sockaddr_in6 all1_sa;
158 	struct rtentry *nrt = NULL;
159 	int e;
160 
161 	bzero(&lo_sa, sizeof(lo_sa));
162 	bzero(&all1_sa, sizeof(all1_sa));
163 	lo_sa.sin6_family = all1_sa.sin6_family = AF_INET6;
164 	lo_sa.sin6_len = all1_sa.sin6_len = sizeof(struct sockaddr_in6);
165 	lo_sa.sin6_addr = in6addr_loopback;
166 	all1_sa.sin6_addr = in6mask128;
167 
168 	/*
169 	 * We specify the address itself as the gateway, and set the
170 	 * RTF_LLINFO flag, so that the corresponding host route would have
171 	 * the flag, and thus applications that assume traditional behavior
172 	 * would be happy.  Note that we assume the caller of the function
173 	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
174 	 * which changes the outgoing interface to the loopback interface.
175 	 */
176 	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr,
177 	    (struct sockaddr *)&all1_sa, RTF_UP|RTF_HOST|RTF_LLINFO, &nrt);
178 	if (e != 0) {
179 		log(LOG_ERR, "in6_ifloop_request: "
180 		    "%s operation failed for %s (errno=%d)\n",
181 		    cmd == RTM_ADD ? "ADD" : "DELETE",
182 		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
183 		    e);
184 	}
185 
186 	/*
187 	 * Make sure rt_ifa be equal to IFA, the second argument of the
188 	 * function.
189 	 * We need this because when we refer to rt_ifa->ia6_flags in
190 	 * ip6_input, we assume that the rt_ifa points to the address instead
191 	 * of the loopback address.
192 	 */
193 	if (cmd == RTM_ADD && nrt && ifa != nrt->rt_ifa) {
194 		IFAFREE(nrt->rt_ifa);
195 		IFAREF(ifa);
196 		nrt->rt_ifa = ifa;
197 	}
198 
199 	/*
200 	 * Report the addition/removal of the address to the routing socket.
201 	 * XXX: since we called rtinit for a p2p interface with a destination,
202 	 *      we end up reporting twice in such a case.  Should we rather
203 	 *      omit the second report?
204 	 */
205 	if (nrt) {
206 		rt_newaddrmsg(cmd, ifa, e, nrt);
207 		if (cmd == RTM_DELETE) {
208 			if (nrt->rt_refcnt <= 0) {
209 				/* XXX: we should free the entry ourselves. */
210 				nrt->rt_refcnt++;
211 				rtfree(nrt);
212 			}
213 		} else {
214 			/* the cmd must be RTM_ADD here */
215 			nrt->rt_refcnt--;
216 		}
217 	}
218 }
219 
220 /*
221  * Add ownaddr as loopback rtentry.  We previously add the route only if
222  * necessary (ex. on a p2p link).  However, since we now manage addresses
223  * separately from prefixes, we should always add the route.  We can't
224  * rely on the cloning mechanism from the corresponding interface route
225  * any more.
226  */
227 static void
228 in6_ifaddloop(struct ifaddr *ifa)
229 {
230 	struct rtentry *rt;
231 
232 	/* If there is no loopback entry, allocate one. */
233 	rt = rtalloc1(ifa->ifa_addr, 0);
234 	if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
235 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
236 		in6_ifloop_request(RTM_ADD, ifa);
237 	if (rt)
238 		rt->rt_refcnt--;
239 }
240 
241 /*
242  * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
243  * if it exists.
244  */
245 static void
246 in6_ifremloop(struct ifaddr *ifa)
247 {
248 	struct in6_ifaddr *ia;
249 	struct rtentry *rt;
250 	int ia_count = 0;
251 
252 	/*
253 	 * Some of BSD variants do not remove cloned routes
254 	 * from an interface direct route, when removing the direct route
255 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
256 	 * cloned routes, they could fail to remove the cloned routes when
257 	 * we handle multple addresses that share a common prefix.
258 	 * So, we should remove the route corresponding to the deleted address.
259 	 */
260 
261 	/*
262 	 * Delete the entry only if exact one ifa exists.  More than one ifa
263 	 * can exist if we assign a same single address to multiple
264 	 * (probably p2p) interfaces.
265 	 * XXX: we should avoid such a configuration in IPv6...
266 	 */
267 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
268 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
269 			ia_count++;
270 			if (ia_count > 1)
271 				break;
272 		}
273 	}
274 
275 	if (ia_count == 1) {
276 		/*
277 		 * Before deleting, check if a corresponding loopbacked host
278 		 * route surely exists.  With this check, we can avoid to
279 		 * delete an interface direct route whose destination is same
280 		 * as the address being removed.  This can happen when removing
281 		 * a subnet-router anycast address on an interface attahced
282 		 * to a shared medium.
283 		 */
284 		rt = rtalloc1(ifa->ifa_addr, 0);
285 		if (rt != NULL && (rt->rt_flags & RTF_HOST) != 0 &&
286 		    (rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
287 			rt->rt_refcnt--;
288 			in6_ifloop_request(RTM_DELETE, ifa);
289 		}
290 	}
291 }
292 
293 int
294 in6_ifindex2scopeid(idx)
295 	int idx;
296 {
297 	struct ifnet *ifp;
298 	struct ifaddr *ifa;
299 	struct sockaddr_in6 *sin6;
300 
301 	if (idx < 0 || if_indexlim <= idx)
302 		return -1;
303 	ifp = ifindex2ifnet[idx];
304 	if (!ifp)
305 		return -1;
306 
307 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
308 	{
309 		if (ifa->ifa_addr->sa_family != AF_INET6)
310 			continue;
311 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
312 		if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))
313 			return sin6->sin6_scope_id & 0xffff;
314 	}
315 
316 	return -1;
317 }
318 
319 int
320 in6_mask2len(mask, lim0)
321 	struct in6_addr *mask;
322 	u_char *lim0;
323 {
324 	int x = 0, y;
325 	u_char *lim = lim0, *p;
326 
327 	/* ignore the scope_id part */
328 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
329 		lim = (u_char *)mask + sizeof(*mask);
330 	for (p = (u_char *)mask; p < lim; x++, p++) {
331 		if (*p != 0xff)
332 			break;
333 	}
334 	y = 0;
335 	if (p < lim) {
336 		for (y = 0; y < 8; y++) {
337 			if ((*p & (0x80 >> y)) == 0)
338 				break;
339 		}
340 	}
341 
342 	/*
343 	 * when the limit pointer is given, do a stricter check on the
344 	 * remaining bits.
345 	 */
346 	if (p < lim) {
347 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
348 			return (-1);
349 		for (p = p + 1; p < lim; p++)
350 			if (*p != 0)
351 				return (-1);
352 	}
353 
354 	return x * 8 + y;
355 }
356 
357 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
358 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
359 
360 int
361 in6_control(so, cmd, data, ifp, p)
362 	struct	socket *so;
363 	u_long cmd;
364 	caddr_t	data;
365 	struct ifnet *ifp;
366 	struct proc *p;
367 {
368 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
369 	struct	in6_ifaddr *ia = NULL;
370 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
371 	struct sockaddr_in6 *sa6;
372 	int privileged;
373 
374 	privileged = 0;
375 	if (p && !suser(p->p_ucred, &p->p_acflag))
376 		privileged++;
377 
378 	switch (cmd) {
379 	case SIOCGETSGCNT_IN6:
380 	case SIOCGETMIFCNT_IN6:
381 		return (mrt6_ioctl(cmd, data));
382 	}
383 
384 	if (ifp == NULL)
385 		return (EOPNOTSUPP);
386 
387 	switch (cmd) {
388 	case SIOCSNDFLUSH_IN6:
389 	case SIOCSPFXFLUSH_IN6:
390 	case SIOCSRTRFLUSH_IN6:
391 	case SIOCSDEFIFACE_IN6:
392 	case SIOCSIFINFO_FLAGS:
393 		if (!privileged)
394 			return (EPERM);
395 		/* FALLTHROUGH */
396 	case OSIOCGIFINFO_IN6:
397 	case SIOCGIFINFO_IN6:
398 	case SIOCGDRLST_IN6:
399 	case SIOCGPRLST_IN6:
400 	case SIOCGNBRINFO_IN6:
401 	case SIOCGDEFIFACE_IN6:
402 		return (nd6_ioctl(cmd, data, ifp));
403 	}
404 
405 	switch (cmd) {
406 	case SIOCSIFPREFIX_IN6:
407 	case SIOCDIFPREFIX_IN6:
408 	case SIOCAIFPREFIX_IN6:
409 	case SIOCCIFPREFIX_IN6:
410 	case SIOCSGIFPREFIX_IN6:
411 	case SIOCGIFPREFIX_IN6:
412 		log(LOG_NOTICE,
413 		    "prefix ioctls are now invalidated. "
414 		    "please use ifconfig.\n");
415 		return (EOPNOTSUPP);
416 	}
417 
418 	switch (cmd) {
419 	case SIOCALIFADDR:
420 	case SIOCDLIFADDR:
421 		if (!privileged)
422 			return (EPERM);
423 		/* FALLTHROUGH */
424 	case SIOCGLIFADDR:
425 		return in6_lifaddr_ioctl(so, cmd, data, ifp, p);
426 	}
427 
428 	/*
429 	 * Find address for this interface, if it exists.
430 	 *
431 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
432 	 * only, and used the first interface address as the target of other
433 	 * operations (without checking ifra_addr).  This was because netinet
434 	 * code/API assumed at most 1 interface address per interface.
435 	 * Since IPv6 allows a node to assign multiple addresses
436 	 * on a single interface, we almost always look and check the
437 	 * presence of ifra_addr, and reject invalid ones here.
438 	 * It also decreases duplicated code among SIOC*_IN6 operations.
439 	 */
440 	switch (cmd) {
441 	case SIOCAIFADDR_IN6:
442 	case SIOCSIFPHYADDR_IN6:
443 		sa6 = &ifra->ifra_addr;
444 		break;
445 	case SIOCSIFADDR_IN6:
446 	case SIOCGIFADDR_IN6:
447 	case SIOCSIFDSTADDR_IN6:
448 	case SIOCSIFNETMASK_IN6:
449 	case SIOCGIFDSTADDR_IN6:
450 	case SIOCGIFNETMASK_IN6:
451 	case SIOCDIFADDR_IN6:
452 	case SIOCGIFPSRCADDR_IN6:
453 	case SIOCGIFPDSTADDR_IN6:
454 	case SIOCGIFAFLAG_IN6:
455 	case SIOCSNDFLUSH_IN6:
456 	case SIOCSPFXFLUSH_IN6:
457 	case SIOCSRTRFLUSH_IN6:
458 	case SIOCGIFALIFETIME_IN6:
459 	case SIOCSIFALIFETIME_IN6:
460 	case SIOCGIFSTAT_IN6:
461 	case SIOCGIFSTAT_ICMP6:
462 		sa6 = &ifr->ifr_addr;
463 		break;
464 	default:
465 		sa6 = NULL;
466 		break;
467 	}
468 	if (sa6 && sa6->sin6_family == AF_INET6) {
469 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) {
470 			if (sa6->sin6_addr.s6_addr16[1] == 0) {
471 				/* link ID is not embedded by the user */
472 				sa6->sin6_addr.s6_addr16[1] =
473 				    htons(ifp->if_index);
474 			} else if (sa6->sin6_addr.s6_addr16[1] !=
475 			    htons(ifp->if_index)) {
476 				return (EINVAL);	/* link ID contradicts */
477 			}
478 			if (sa6->sin6_scope_id) {
479 				if (sa6->sin6_scope_id !=
480 				    (u_int32_t)ifp->if_index)
481 					return (EINVAL);
482 				sa6->sin6_scope_id = 0; /* XXX: good way? */
483 			}
484 		}
485 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
486 	} else
487 		ia = NULL;
488 
489 	switch (cmd) {
490 	case SIOCSIFADDR_IN6:
491 	case SIOCSIFDSTADDR_IN6:
492 	case SIOCSIFNETMASK_IN6:
493 		/*
494 		 * Since IPv6 allows a node to assign multiple addresses
495 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
496 		 */
497 		return (EINVAL);
498 
499 	case SIOCDIFADDR_IN6:
500 		/*
501 		 * for IPv4, we look for existing in_ifaddr here to allow
502 		 * "ifconfig if0 delete" to remove the first IPv4 address on
503 		 * the interface.  For IPv6, as the spec allows multiple
504 		 * interface address from the day one, we consider "remove the
505 		 * first one" semantics to be not preferable.
506 		 */
507 		if (ia == NULL)
508 			return (EADDRNOTAVAIL);
509 		/* FALLTHROUGH */
510 	case SIOCAIFADDR_IN6:
511 		/*
512 		 * We always require users to specify a valid IPv6 address for
513 		 * the corresponding operation.
514 		 */
515 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
516 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
517 			return (EAFNOSUPPORT);
518 		if (!privileged)
519 			return (EPERM);
520 
521 		break;
522 
523 	case SIOCGIFADDR_IN6:
524 		/* This interface is basically deprecated. use SIOCGIFCONF. */
525 		/* FALLTHROUGH */
526 	case SIOCGIFAFLAG_IN6:
527 	case SIOCGIFNETMASK_IN6:
528 	case SIOCGIFDSTADDR_IN6:
529 	case SIOCGIFALIFETIME_IN6:
530 		/* must think again about its semantics */
531 		if (ia == NULL)
532 			return (EADDRNOTAVAIL);
533 		break;
534 	case SIOCSIFALIFETIME_IN6:
535 	    {
536 		struct in6_addrlifetime *lt;
537 
538 		if (!privileged)
539 			return (EPERM);
540 		if (ia == NULL)
541 			return (EADDRNOTAVAIL);
542 		/* sanity for overflow - beware unsigned */
543 		lt = &ifr->ifr_ifru.ifru_lifetime;
544 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
545 		 && lt->ia6t_vltime + time.tv_sec < time.tv_sec) {
546 			return EINVAL;
547 		}
548 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
549 		 && lt->ia6t_pltime + time.tv_sec < time.tv_sec) {
550 			return EINVAL;
551 		}
552 		break;
553 	    }
554 	}
555 
556 	switch (cmd) {
557 
558 	case SIOCGIFADDR_IN6:
559 		ifr->ifr_addr = ia->ia_addr;
560 		break;
561 
562 	case SIOCGIFDSTADDR_IN6:
563 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
564 			return (EINVAL);
565 		/*
566 		 * XXX: should we check if ifa_dstaddr is NULL and return
567 		 * an error?
568 		 */
569 		ifr->ifr_dstaddr = ia->ia_dstaddr;
570 		break;
571 
572 	case SIOCGIFNETMASK_IN6:
573 		ifr->ifr_addr = ia->ia_prefixmask;
574 		break;
575 
576 	case SIOCGIFAFLAG_IN6:
577 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
578 		break;
579 
580 	case SIOCGIFSTAT_IN6:
581 		if (ifp == NULL)
582 			return EINVAL;
583 		bzero(&ifr->ifr_ifru.ifru_stat,
584 		    sizeof(ifr->ifr_ifru.ifru_stat));
585 		ifr->ifr_ifru.ifru_stat =
586 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
587 		break;
588 
589 	case SIOCGIFSTAT_ICMP6:
590 		if (ifp == NULL)
591 			return EINVAL;
592 		bzero(&ifr->ifr_ifru.ifru_stat,
593 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
594 		ifr->ifr_ifru.ifru_icmp6stat =
595 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
596 		break;
597 
598 	case SIOCGIFALIFETIME_IN6:
599 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
600 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
601 			time_t maxexpire;
602 			struct in6_addrlifetime *retlt =
603 			    &ifr->ifr_ifru.ifru_lifetime;
604 
605 			/*
606 			 * XXX: adjust expiration time assuming time_t is
607 			 * signed.
608 			 */
609 			maxexpire = (-1) &
610 			    ~(1 << ((sizeof(maxexpire) * 8) - 1));
611 			if (ia->ia6_lifetime.ia6t_vltime <
612 			    maxexpire - ia->ia6_updatetime) {
613 				retlt->ia6t_expire = ia->ia6_updatetime +
614 				    ia->ia6_lifetime.ia6t_vltime;
615 			} else
616 				retlt->ia6t_expire = maxexpire;
617 		}
618 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
619 			time_t maxexpire;
620 			struct in6_addrlifetime *retlt =
621 			    &ifr->ifr_ifru.ifru_lifetime;
622 
623 			/*
624 			 * XXX: adjust expiration time assuming time_t is
625 			 * signed.
626 			 */
627 			maxexpire = (-1) &
628 			    ~(1 << ((sizeof(maxexpire) * 8) - 1));
629 			if (ia->ia6_lifetime.ia6t_pltime <
630 			    maxexpire - ia->ia6_updatetime) {
631 				retlt->ia6t_preferred = ia->ia6_updatetime +
632 				    ia->ia6_lifetime.ia6t_pltime;
633 			} else
634 				retlt->ia6t_preferred = maxexpire;
635 		}
636 		break;
637 
638 	case SIOCSIFALIFETIME_IN6:
639 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
640 		/* for sanity */
641 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
642 			ia->ia6_lifetime.ia6t_expire =
643 				time.tv_sec + ia->ia6_lifetime.ia6t_vltime;
644 		} else
645 			ia->ia6_lifetime.ia6t_expire = 0;
646 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
647 			ia->ia6_lifetime.ia6t_preferred =
648 				time.tv_sec + ia->ia6_lifetime.ia6t_pltime;
649 		} else
650 			ia->ia6_lifetime.ia6t_preferred = 0;
651 		break;
652 
653 	case SIOCAIFADDR_IN6:
654 	{
655 		int i, error = 0;
656 		struct nd_prefix pr0, *pr;
657 
658 		/* reject read-only flags */
659 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
660 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
661 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
662 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
663 			return (EINVAL);
664 		}
665 		/*
666 		 * first, make or update the interface address structure,
667 		 * and link it to the list.
668 		 */
669 		if ((error = in6_update_ifa(ifp, ifra, ia)) != 0)
670 			return (error);
671 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
672 		    == NULL) {
673 		    	/*
674 			 * this can happen when the user specify the 0 valid
675 			 * lifetime.
676 			 */
677 			break;
678 		}
679 
680 		/*
681 		 * then, make the prefix on-link on the interface.
682 		 * XXX: we'd rather create the prefix before the address, but
683 		 * we need at least one address to install the corresponding
684 		 * interface route, so we configure the address first.
685 		 */
686 
687 		/*
688 		 * convert mask to prefix length (prefixmask has already
689 		 * been validated in in6_update_ifa().
690 		 */
691 		bzero(&pr0, sizeof(pr0));
692 		pr0.ndpr_ifp = ifp;
693 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
694 		    NULL);
695 		if (pr0.ndpr_plen == 128) {
696 			break;	/* we don't need to install a host route. */
697 		}
698 		pr0.ndpr_prefix = ifra->ifra_addr;
699 		pr0.ndpr_mask = ifra->ifra_prefixmask.sin6_addr;
700 		/* apply the mask for safety. */
701 		for (i = 0; i < 4; i++) {
702 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
703 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
704 		}
705 		/*
706 		 * XXX: since we don't have an API to set prefix (not address)
707 		 * lifetimes, we just use the same lifetimes as addresses.
708 		 * The (temporarily) installed lifetimes can be overridden by
709 		 * later advertised RAs (when accept_rtadv is non 0), which is
710 		 * an intended behavior.
711 		 */
712 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
713 		pr0.ndpr_raf_auto =
714 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
715 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
716 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
717 
718 		/* add the prefix if not yet. */
719 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
720 			/*
721 			 * nd6_prelist_add will install the corresponding
722 			 * interface route.
723 			 */
724 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
725 				return (error);
726 			if (pr == NULL) {
727 				log(LOG_ERR, "nd6_prelist_add succeeded but "
728 				    "no prefix\n");
729 				return (EINVAL); /* XXX panic here? */
730 			}
731 		}
732 
733 		/* relate the address to the prefix */
734 		if (ia->ia6_ndpr == NULL) {
735 			ia->ia6_ndpr = pr;
736 			pr->ndpr_refcnt++;
737 		}
738 
739 		/*
740 		 * this might affect the status of autoconfigured addresses,
741 		 * that is, this address might make other addresses detached.
742 		 */
743 		pfxlist_onlink_check();
744 
745 #ifdef PFIL_HOOKS
746 		(void)pfil_run_hooks(&if_pfil, (struct mbuf **)SIOCAIFADDR_IN6,
747 		    ifp, PFIL_IFADDR);
748 #endif
749 
750 		break;
751 	}
752 
753 	case SIOCDIFADDR_IN6:
754 	{
755 		int i = 0, purgeprefix = 0;
756 		struct nd_prefix pr0, *pr = NULL;
757 
758 		/*
759 		 * If the address being deleted is the only one that owns
760 		 * the corresponding prefix, expire the prefix as well.
761 		 * XXX: theoretically, we don't have to worry about such
762 		 * relationship, since we separate the address management
763 		 * and the prefix management.  We do this, however, to provide
764 		 * as much backward compatibility as possible in terms of
765 		 * the ioctl operation.
766 		 */
767 		bzero(&pr0, sizeof(pr0));
768 		pr0.ndpr_ifp = ifp;
769 		pr0.ndpr_plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr,
770 		    NULL);
771 		if (pr0.ndpr_plen == 128)
772 			goto purgeaddr;
773 		pr0.ndpr_prefix = ia->ia_addr;
774 		pr0.ndpr_mask = ia->ia_prefixmask.sin6_addr;
775 		for (i = 0; i < 4; i++) {
776 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
777 			    ia->ia_prefixmask.sin6_addr.s6_addr32[i];
778 		}
779 		if ((pr = nd6_prefix_lookup(&pr0)) != NULL &&
780 		    pr == ia->ia6_ndpr) {
781 			pr->ndpr_refcnt--;
782 			if (pr->ndpr_refcnt == 0)
783 				purgeprefix = 1;
784 		}
785 
786 	  purgeaddr:
787 		in6_purgeaddr(&ia->ia_ifa);
788 		if (pr && purgeprefix)
789 			prelist_remove(pr);
790 #ifdef PFIL_HOOKS
791 		(void)pfil_run_hooks(&if_pfil, (struct mbuf **)SIOCDIFADDR_IN6,
792 		    ifp, PFIL_IFADDR);
793 #endif
794 		break;
795 	}
796 
797 	default:
798 		if (ifp == NULL || ifp->if_ioctl == 0)
799 			return (EOPNOTSUPP);
800 		return ((*ifp->if_ioctl)(ifp, cmd, data));
801 	}
802 
803 	return (0);
804 }
805 
806 /*
807  * Update parameters of an IPv6 interface address.
808  * If necessary, a new entry is created and linked into address chains.
809  * This function is separated from in6_control().
810  * XXX: should this be performed under splnet()?
811  */
812 int
813 in6_update_ifa(ifp, ifra, ia)
814 	struct ifnet *ifp;
815 	struct in6_aliasreq *ifra;
816 	struct in6_ifaddr *ia;
817 {
818 	int error = 0, hostIsNew = 0, plen = -1;
819 	struct in6_ifaddr *oia;
820 	struct sockaddr_in6 dst6;
821 	struct in6_addrlifetime *lt;
822 	struct in6_multi_mship *imm;
823 	struct rtentry *rt;
824 
825 	/* Validate parameters */
826 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
827 		return (EINVAL);
828 
829 	/*
830 	 * The destination address for a p2p link must have a family
831 	 * of AF_UNSPEC or AF_INET6.
832 	 */
833 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
834 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
835 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
836 		return (EAFNOSUPPORT);
837 	/*
838 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
839 	 * does not carry fields other than sin6_len.
840 	 */
841 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
842 		return (EINVAL);
843 	/*
844 	 * Because the IPv6 address architecture is classless, we require
845 	 * users to specify a (non 0) prefix length (mask) for a new address.
846 	 * We also require the prefix (when specified) mask is valid, and thus
847 	 * reject a non-consecutive mask.
848 	 */
849 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
850 		return (EINVAL);
851 	if (ifra->ifra_prefixmask.sin6_len != 0) {
852 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
853 		    (u_char *)&ifra->ifra_prefixmask +
854 		    ifra->ifra_prefixmask.sin6_len);
855 		if (plen <= 0)
856 			return (EINVAL);
857 	} else {
858 		/*
859 		 * In this case, ia must not be NULL.  We just use its prefix
860 		 * length.
861 		 */
862 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
863 	}
864 	/*
865 	 * If the destination address on a p2p interface is specified,
866 	 * and the address is a scoped one, validate/set the scope
867 	 * zone identifier.
868 	 */
869 	dst6 = ifra->ifra_dstaddr;
870 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
871 	    (dst6.sin6_family == AF_INET6)) {
872 		/* link-local index check: should be a separate function? */
873 		if (IN6_IS_ADDR_LINKLOCAL(&dst6.sin6_addr)) {
874 			if (dst6.sin6_addr.s6_addr16[1] == 0) {
875 				/*
876 				 * interface ID is not embedded by
877 				 * the user
878 				 */
879 				dst6.sin6_addr.s6_addr16[1] =
880 				    htons(ifp->if_index);
881 			} else if (dst6.sin6_addr.s6_addr16[1] !=
882 			    htons(ifp->if_index)) {
883 				return (EINVAL);	/* ifid contradicts */
884 			}
885 		}
886 	}
887 	/*
888 	 * The destination address can be specified only for a p2p or a
889 	 * loopback interface.  If specified, the corresponding prefix length
890 	 * must be 128.
891 	 */
892 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
893 #ifdef FORCE_P2PPLEN
894 		int i;
895 #endif
896 
897 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
898 			/* XXX: noisy message */
899 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
900 			    "be specified for a p2p or a loopback IF only\n"));
901 			return (EINVAL);
902 		}
903 		if (plen != 128) {
904 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
905 			    "be 128 when dstaddr is specified\n"));
906 #ifdef FORCE_P2PPLEN
907 			/*
908 			 * To be compatible with old configurations,
909 			 * such as ifconfig gif0 inet6 2001::1 2001::2
910 			 * prefixlen 126, we override the specified
911 			 * prefixmask as if the prefix length was 128.
912 			 */
913 			ifra->ifra_prefixmask.sin6_len =
914 			    sizeof(struct sockaddr_in6);
915 			for (i = 0; i < 4; i++)
916 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
917 				    0xffffffff;
918 			plen = 128;
919 #else
920 			return (EINVAL);
921 #endif
922 		}
923 	}
924 	/* lifetime consistency check */
925 	lt = &ifra->ifra_lifetime;
926 	if (lt->ia6t_pltime > lt->ia6t_vltime)
927 		return (EINVAL);
928 	if (lt->ia6t_vltime == 0) {
929 		/*
930 		 * the following log might be noisy, but this is a typical
931 		 * configuration mistake or a tool's bug.
932 		 */
933 		nd6log((LOG_INFO,
934 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
935 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr)));
936 
937 		if (ia == NULL)
938 			return (0); /* there's nothing to do */
939 	}
940 
941 	/*
942 	 * If this is a new address, allocate a new ifaddr and link it
943 	 * into chains.
944 	 */
945 	if (ia == NULL) {
946 		hostIsNew = 1;
947 		/*
948 		 * When in6_update_ifa() is called in a process of a received
949 		 * RA, it is called under an interrupt context.  So, we should
950 		 * call malloc with M_NOWAIT.
951 		 */
952 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
953 		    M_NOWAIT);
954 		if (ia == NULL)
955 			return (ENOBUFS);
956 		bzero((caddr_t)ia, sizeof(*ia));
957 		LIST_INIT(&ia->ia6_memberships);
958 		/* Initialize the address and masks, and put time stamp */
959 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
960 		ia->ia_addr.sin6_family = AF_INET6;
961 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
962 		ia->ia6_createtime = ia->ia6_updatetime = time.tv_sec;
963 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
964 			/*
965 			 * XXX: some functions expect that ifa_dstaddr is not
966 			 * NULL for p2p interfaces.
967 			 */
968 			ia->ia_ifa.ifa_dstaddr =
969 			    (struct sockaddr *)&ia->ia_dstaddr;
970 		} else {
971 			ia->ia_ifa.ifa_dstaddr = NULL;
972 		}
973 		ia->ia_ifa.ifa_netmask =
974 		    (struct sockaddr *)&ia->ia_prefixmask;
975 
976 		ia->ia_ifp = ifp;
977 		if ((oia = in6_ifaddr) != NULL) {
978 			for ( ; oia->ia_next; oia = oia->ia_next)
979 				continue;
980 			oia->ia_next = ia;
981 		} else
982 			in6_ifaddr = ia;
983 		/* gain a refcnt for the link from in6_ifaddr */
984 		IFAREF(&ia->ia_ifa);
985 
986 		TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa,
987 				  ifa_list);
988 		/* gain another refcnt for the link from if_addrlist */
989 		IFAREF(&ia->ia_ifa);
990 	}
991 
992 	/* set prefix mask */
993 	if (ifra->ifra_prefixmask.sin6_len) {
994 		/*
995 		 * We prohibit changing the prefix length of an existing
996 		 * address, because
997 		 * + such an operation should be rare in IPv6, and
998 		 * + the operation would confuse prefix management.
999 		 */
1000 		if (ia->ia_prefixmask.sin6_len &&
1001 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1002 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1003 			    " existing (%s) address should not be changed\n",
1004 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
1005 			error = EINVAL;
1006 			goto unlink;
1007 		}
1008 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1009 	}
1010 
1011 	/*
1012 	 * If a new destination address is specified, scrub the old one and
1013 	 * install the new destination.  Note that the interface must be
1014 	 * p2p or loopback (see the check above.)
1015 	 */
1016 	if (dst6.sin6_family == AF_INET6 &&
1017 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1018 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1019 		    rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0) {
1020 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1021 			    "a route to the old destination: %s\n",
1022 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
1023 			/* proceed anyway... */
1024 		} else
1025 			ia->ia_flags &= ~IFA_ROUTE;
1026 		ia->ia_dstaddr = dst6;
1027 	}
1028 
1029 	/*
1030 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1031 	 * to see if the address is deprecated or invalidated, but initialize
1032 	 * these members for applications.
1033 	 */
1034 	ia->ia6_lifetime = ifra->ifra_lifetime;
1035 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1036 		ia->ia6_lifetime.ia6t_expire =
1037 		    time.tv_sec + ia->ia6_lifetime.ia6t_vltime;
1038 	} else
1039 		ia->ia6_lifetime.ia6t_expire = 0;
1040 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1041 		ia->ia6_lifetime.ia6t_preferred =
1042 		    time.tv_sec + ia->ia6_lifetime.ia6t_pltime;
1043 	} else
1044 		ia->ia6_lifetime.ia6t_preferred = 0;
1045 
1046 	/* reset the interface and routing table appropriately. */
1047 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1048 		goto unlink;
1049 
1050 	/*
1051 	 * configure address flags.
1052 	 */
1053 	ia->ia6_flags = ifra->ifra_flags;
1054 	/*
1055 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1056 	 * userland, make it deprecated.
1057 	 */
1058 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1059 		ia->ia6_lifetime.ia6t_pltime = 0;
1060 		ia->ia6_lifetime.ia6t_preferred = time.tv_sec;
1061 	}
1062 	/*
1063 	 * Make the address tentative before joining multicast addresses,
1064 	 * so that corresponding MLD responses would not have a tentative
1065 	 * source address.
1066 	 */
1067 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1068 	if (hostIsNew && in6if_do_dad(ifp))
1069 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1070 
1071 	/*
1072 	 * We are done if we have simply modified an existing address.
1073 	 */
1074 	if (!hostIsNew)
1075 		return (error);
1076 
1077 	/*
1078 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1079 	 * not just go to unlink.
1080 	 */
1081 
1082 	/* join necessary multiast groups */
1083 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1084 		struct sockaddr_in6 mltaddr, mltmask;
1085 #ifndef SCOPEDROUTING
1086 		u_int32_t zoneid = 0;
1087 #endif
1088 
1089 		/* join solicited multicast addr for new host id */
1090 		struct sockaddr_in6 llsol;
1091 
1092 		bzero(&llsol, sizeof(llsol));
1093 		llsol.sin6_family = AF_INET6;
1094 		llsol.sin6_len = sizeof(llsol);
1095 		llsol.sin6_addr.s6_addr16[0] = htons(0xff02);
1096 		llsol.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1097 		llsol.sin6_addr.s6_addr32[1] = 0;
1098 		llsol.sin6_addr.s6_addr32[2] = htonl(1);
1099 		llsol.sin6_addr.s6_addr32[3] =
1100 		    ifra->ifra_addr.sin6_addr.s6_addr32[3];
1101 		llsol.sin6_addr.s6_addr8[12] = 0xff;
1102 		imm = in6_joingroup(ifp, &llsol.sin6_addr, &error);
1103 		if (!imm) {
1104 			nd6log((LOG_ERR,
1105 			    "in6_update_ifa: addmulti "
1106 			    "failed for %s on %s (errno=%d)\n",
1107 			    ip6_sprintf(&llsol.sin6_addr),
1108 			    if_name(ifp), error));
1109 			goto cleanup;
1110 		}
1111 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1112 
1113 		bzero(&mltmask, sizeof(mltmask));
1114 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1115 		mltmask.sin6_family = AF_INET6;
1116 		mltmask.sin6_addr = in6mask32;
1117 
1118 		/*
1119 		 * join link-local all-nodes address
1120 		 */
1121 		bzero(&mltaddr, sizeof(mltaddr));
1122 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1123 		mltaddr.sin6_family = AF_INET6;
1124 		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1125 		mltaddr.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1126 
1127 		/*
1128 		 * XXX: do we really need this automatic routes?
1129 		 * We should probably reconsider this stuff.  Most applications
1130 		 * actually do not need the routes, since they usually specify
1131 		 * the outgoing interface.
1132 		 */
1133 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1134 		if (rt) {
1135 			/*
1136 			 * 32bit came from "mltmask"
1137 			 * XXX: only works in !SCOPEDROUTING case.
1138 			 */
1139 			if (memcmp(&mltaddr.sin6_addr,
1140 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1141 			    32 / 8)) {
1142 				RTFREE(rt);
1143 				rt = NULL;
1144 			}
1145 		}
1146 		if (!rt) {
1147 			struct rt_addrinfo info;
1148 
1149 			bzero(&info, sizeof(info));
1150 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1151 			info.rti_info[RTAX_GATEWAY] =
1152 			    (struct sockaddr *)&ia->ia_addr;
1153 			info.rti_info[RTAX_NETMASK] =
1154 			    (struct sockaddr *)&mltmask;
1155 			info.rti_info[RTAX_IFA] =
1156 			    (struct sockaddr *)&ia->ia_addr;
1157 			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1158 			info.rti_flags = RTF_UP | RTF_CLONING;
1159 			error = rtrequest1(RTM_ADD, &info, NULL);
1160 			if (error)
1161 				goto cleanup;
1162 		} else {
1163 			RTFREE(rt);
1164 		}
1165 #ifndef SCOPEDROUTING
1166 		mltaddr.sin6_scope_id = zoneid;	/* XXX */
1167 #endif
1168 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error);
1169 		if (!imm) {
1170 			nd6log((LOG_WARNING,
1171 			    "in6_update_ifa: addmulti failed for "
1172 			    "%s on %s (errno=%d)\n",
1173 			    ip6_sprintf(&mltaddr.sin6_addr),
1174 			    if_name(ifp), error));
1175 			goto cleanup;
1176 		}
1177 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1178 
1179 		/*
1180 		 * join node information group address
1181 		 */
1182 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) == 0) {
1183 			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error);
1184 			if (!imm) {
1185 				nd6log((LOG_WARNING, "in6_update_ifa: "
1186 				    "addmulti failed for %s on %s (errno=%d)\n",
1187 				    ip6_sprintf(&mltaddr.sin6_addr),
1188 				    if_name(ifp), error));
1189 				/* XXX not very fatal, go on... */
1190 			} else {
1191 				LIST_INSERT_HEAD(&ia->ia6_memberships,
1192 				    imm, i6mm_chain);
1193 			}
1194 		}
1195 
1196 		if (ifp->if_flags & IFF_LOOPBACK) {
1197 			/*
1198 			 * join node-local all-nodes address, on loopback.
1199 			 * (ff01::1%ifN, and ff01::%ifN/32)
1200 			 */
1201 			mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1202 
1203 			/* XXX: again, do we really need the route? */
1204 			rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1205 			if (rt) {
1206 				/* 32bit came from "mltmask" */
1207 				if (memcmp(&mltaddr.sin6_addr,
1208 				    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1209 				    32 / 8)) {
1210 					RTFREE(rt);
1211 					rt = NULL;
1212 				}
1213 			}
1214 			if (!rt) {
1215 				struct rt_addrinfo info;
1216 
1217 				bzero(&info, sizeof(info));
1218 				info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1219 				info.rti_info[RTAX_GATEWAY] =
1220 				    (struct sockaddr *)&ia->ia_addr;
1221 				info.rti_info[RTAX_NETMASK] =
1222 				    (struct sockaddr *)&mltmask;
1223 				info.rti_info[RTAX_IFA] =
1224 				    (struct sockaddr *)&ia->ia_addr;
1225 				info.rti_flags = RTF_UP | RTF_CLONING;
1226 				error = rtrequest1(RTM_ADD, &info, NULL);
1227 				if (error)
1228 					goto cleanup;
1229 			} else {
1230 				RTFREE(rt);
1231 			}
1232 			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error);
1233 			if (!imm) {
1234 				nd6log((LOG_WARNING, "in6_update_ifa: "
1235 				    "addmulti failed for %s on %s (errno=%d)\n",
1236 				    ip6_sprintf(&mltaddr.sin6_addr),
1237 				    if_name(ifp), error));
1238 				goto cleanup;
1239 			}
1240 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1241 		}
1242 	}
1243 
1244 	/*
1245 	 * Perform DAD, if needed.
1246 	 * XXX It may be of use, if we can administratively
1247 	 * disable DAD.
1248 	 */
1249 	if (hostIsNew && in6if_do_dad(ifp) &&
1250 	    (ifra->ifra_flags & IN6_IFF_NODAD) == 0)
1251 	{
1252 		nd6_dad_start((struct ifaddr *)ia, NULL);
1253 	}
1254 
1255 	return (error);
1256 
1257   unlink:
1258 	/*
1259 	 * XXX: if a change of an existing address failed, keep the entry
1260 	 * anyway.
1261 	 */
1262 	if (hostIsNew)
1263 		in6_unlink_ifa(ia, ifp);
1264 	return (error);
1265 
1266   cleanup:
1267 	in6_purgeaddr(&ia->ia_ifa);
1268 	return error;
1269 }
1270 
1271 void
1272 in6_purgeaddr(ifa)
1273 	struct ifaddr *ifa;
1274 {
1275 	struct ifnet *ifp = ifa->ifa_ifp;
1276 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1277 	struct in6_multi_mship *imm;
1278 
1279 	/* stop DAD processing */
1280 	nd6_dad_stop(ifa);
1281 
1282 	/*
1283 	 * delete route to the destination of the address being purged.
1284 	 * The interface must be p2p or loopback in this case.
1285 	 */
1286 	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1287 		int e;
1288 
1289 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1290 		    != 0) {
1291 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1292 			    "a route to the p2p destination: %s on %s, "
1293 			    "errno=%d\n",
1294 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1295 			    e);
1296 			/* proceed anyway... */
1297 		} else
1298 			ia->ia_flags &= ~IFA_ROUTE;
1299 	}
1300 
1301 	/* Remove ownaddr's loopback rtentry, if it exists. */
1302 	in6_ifremloop(&(ia->ia_ifa));
1303 
1304 	/*
1305 	 * leave from multicast groups we have joined for the interface
1306 	 */
1307 	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1308 		LIST_REMOVE(imm, i6mm_chain);
1309 		in6_leavegroup(imm);
1310 	}
1311 
1312 	in6_unlink_ifa(ia, ifp);
1313 }
1314 
1315 static void
1316 in6_unlink_ifa(ia, ifp)
1317 	struct in6_ifaddr *ia;
1318 	struct ifnet *ifp;
1319 {
1320 	struct in6_ifaddr *oia;
1321 	int	s = splnet();
1322 
1323 	TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1324 	/* release a refcnt for the link from if_addrlist */
1325 	IFAFREE(&ia->ia_ifa);
1326 
1327 	oia = ia;
1328 	if (oia == (ia = in6_ifaddr))
1329 		in6_ifaddr = ia->ia_next;
1330 	else {
1331 		while (ia->ia_next && (ia->ia_next != oia))
1332 			ia = ia->ia_next;
1333 		if (ia->ia_next)
1334 			ia->ia_next = oia->ia_next;
1335 		else {
1336 			/* search failed */
1337 			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1338 		}
1339 	}
1340 
1341 	if (oia->ia6_multiaddrs.lh_first != NULL) {
1342 		/*
1343 		 * XXX thorpej@NetBSD.org -- if the interface is going
1344 		 * XXX away, don't save the multicast entries, delete them!
1345 		 */
1346 		if (oia->ia_ifa.ifa_ifp->if_output == if_nulloutput) {
1347 			struct in6_multi *in6m;
1348 
1349 			while ((in6m =
1350 			    LIST_FIRST(&oia->ia6_multiaddrs)) != NULL)
1351 				in6_delmulti(in6m);
1352 		} else
1353 			in6_savemkludge(oia);
1354 	}
1355 
1356 	/*
1357 	 * When an autoconfigured address is being removed, release the
1358 	 * reference to the base prefix.  Also, since the release might
1359 	 * affect the status of other (detached) addresses, call
1360 	 * pfxlist_onlink_check().
1361 	 */
1362 	if ((oia->ia6_flags & IN6_IFF_AUTOCONF) != 0) {
1363 		if (oia->ia6_ndpr == NULL) {
1364 			log(LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1365 			    "%p has no prefix\n", oia);
1366 		} else {
1367 			oia->ia6_ndpr->ndpr_refcnt--;
1368 			oia->ia6_flags &= ~IN6_IFF_AUTOCONF;
1369 			oia->ia6_ndpr = NULL;
1370 		}
1371 
1372 		pfxlist_onlink_check();
1373 	}
1374 
1375 	/*
1376 	 * release another refcnt for the link from in6_ifaddr.
1377 	 * Note that we should decrement the refcnt at least once for all *BSD.
1378 	 */
1379 	IFAFREE(&oia->ia_ifa);
1380 
1381 	splx(s);
1382 }
1383 
1384 void
1385 in6_purgeif(ifp)
1386 	struct ifnet *ifp;
1387 {
1388 	struct ifaddr *ifa, *nifa;
1389 
1390 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa)
1391 	{
1392 		nifa = TAILQ_NEXT(ifa, ifa_list);
1393 		if (ifa->ifa_addr->sa_family != AF_INET6)
1394 			continue;
1395 		in6_purgeaddr(ifa);
1396 	}
1397 
1398 	in6_ifdetach(ifp);
1399 }
1400 
1401 /*
1402  * SIOC[GAD]LIFADDR.
1403  *	SIOCGLIFADDR: get first address. (?)
1404  *	SIOCGLIFADDR with IFLR_PREFIX:
1405  *		get first address that matches the specified prefix.
1406  *	SIOCALIFADDR: add the specified address.
1407  *	SIOCALIFADDR with IFLR_PREFIX:
1408  *		add the specified prefix, filling hostid part from
1409  *		the first link-local address.  prefixlen must be <= 64.
1410  *	SIOCDLIFADDR: delete the specified address.
1411  *	SIOCDLIFADDR with IFLR_PREFIX:
1412  *		delete the first address that matches the specified prefix.
1413  * return values:
1414  *	EINVAL on invalid parameters
1415  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1416  *	other values may be returned from in6_ioctl()
1417  *
1418  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1419  * this is to accomodate address naming scheme other than RFC2374,
1420  * in the future.
1421  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1422  * address encoding scheme. (see figure on page 8)
1423  */
1424 static int
1425 in6_lifaddr_ioctl(so, cmd, data, ifp, p)
1426 	struct socket *so;
1427 	u_long cmd;
1428 	caddr_t	data;
1429 	struct ifnet *ifp;
1430 	struct proc *p;
1431 {
1432 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1433 	struct ifaddr *ifa;
1434 	struct sockaddr *sa;
1435 
1436 	/* sanity checks */
1437 	if (!data || !ifp) {
1438 		panic("invalid argument to in6_lifaddr_ioctl");
1439 		/* NOTREACHED */
1440 	}
1441 
1442 	switch (cmd) {
1443 	case SIOCGLIFADDR:
1444 		/* address must be specified on GET with IFLR_PREFIX */
1445 		if ((iflr->flags & IFLR_PREFIX) == 0)
1446 			break;
1447 		/* FALLTHROUGH */
1448 	case SIOCALIFADDR:
1449 	case SIOCDLIFADDR:
1450 		/* address must be specified on ADD and DELETE */
1451 		sa = (struct sockaddr *)&iflr->addr;
1452 		if (sa->sa_family != AF_INET6)
1453 			return EINVAL;
1454 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1455 			return EINVAL;
1456 		/* XXX need improvement */
1457 		sa = (struct sockaddr *)&iflr->dstaddr;
1458 		if (sa->sa_family && sa->sa_family != AF_INET6)
1459 			return EINVAL;
1460 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1461 			return EINVAL;
1462 		break;
1463 	default: /* shouldn't happen */
1464 #if 0
1465 		panic("invalid cmd to in6_lifaddr_ioctl");
1466 		/* NOTREACHED */
1467 #else
1468 		return EOPNOTSUPP;
1469 #endif
1470 	}
1471 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1472 		return EINVAL;
1473 
1474 	switch (cmd) {
1475 	case SIOCALIFADDR:
1476 	    {
1477 		struct in6_aliasreq ifra;
1478 		struct in6_addr *hostid = NULL;
1479 		int prefixlen;
1480 
1481 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1482 			struct sockaddr_in6 *sin6;
1483 
1484 			/*
1485 			 * hostid is to fill in the hostid part of the
1486 			 * address.  hostid points to the first link-local
1487 			 * address attached to the interface.
1488 			 */
1489 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1490 			if (!ifa)
1491 				return EADDRNOTAVAIL;
1492 			hostid = IFA_IN6(ifa);
1493 
1494 		 	/* prefixlen must be <= 64. */
1495 			if (64 < iflr->prefixlen)
1496 				return EINVAL;
1497 			prefixlen = iflr->prefixlen;
1498 
1499 			/* hostid part must be zero. */
1500 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1501 			if (sin6->sin6_addr.s6_addr32[2] != 0
1502 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1503 				return EINVAL;
1504 			}
1505 		} else
1506 			prefixlen = iflr->prefixlen;
1507 
1508 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1509 		bzero(&ifra, sizeof(ifra));
1510 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1511 
1512 		bcopy(&iflr->addr, &ifra.ifra_addr,
1513 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1514 		if (hostid) {
1515 			/* fill in hostid part */
1516 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1517 			    hostid->s6_addr32[2];
1518 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1519 			    hostid->s6_addr32[3];
1520 		}
1521 
1522 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1523 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1524 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1525 			if (hostid) {
1526 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1527 				    hostid->s6_addr32[2];
1528 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1529 				    hostid->s6_addr32[3];
1530 			}
1531 		}
1532 
1533 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1534 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1535 
1536 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1537 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, p);
1538 	    }
1539 	case SIOCGLIFADDR:
1540 	case SIOCDLIFADDR:
1541 	    {
1542 		struct in6_ifaddr *ia;
1543 		struct in6_addr mask, candidate, match;
1544 		struct sockaddr_in6 *sin6;
1545 		int cmp;
1546 
1547 		bzero(&mask, sizeof(mask));
1548 		if (iflr->flags & IFLR_PREFIX) {
1549 			/* lookup a prefix rather than address. */
1550 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1551 
1552 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1553 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1554 			match.s6_addr32[0] &= mask.s6_addr32[0];
1555 			match.s6_addr32[1] &= mask.s6_addr32[1];
1556 			match.s6_addr32[2] &= mask.s6_addr32[2];
1557 			match.s6_addr32[3] &= mask.s6_addr32[3];
1558 
1559 			/* if you set extra bits, that's wrong */
1560 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1561 				return EINVAL;
1562 
1563 			cmp = 1;
1564 		} else {
1565 			if (cmd == SIOCGLIFADDR) {
1566 				/* on getting an address, take the 1st match */
1567 				cmp = 0;	/* XXX */
1568 			} else {
1569 				/* on deleting an address, do exact match */
1570 				in6_prefixlen2mask(&mask, 128);
1571 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1572 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1573 
1574 				cmp = 1;
1575 			}
1576 		}
1577 
1578 		for (ifa = ifp->if_addrlist.tqh_first;
1579 		     ifa;
1580 		     ifa = ifa->ifa_list.tqe_next)
1581 		{
1582 			if (ifa->ifa_addr->sa_family != AF_INET6)
1583 				continue;
1584 			if (!cmp)
1585 				break;
1586 
1587 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1588 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1589 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1590 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1591 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1592 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1593 				break;
1594 		}
1595 		if (!ifa)
1596 			return EADDRNOTAVAIL;
1597 		ia = ifa2ia6(ifa);
1598 
1599 		if (cmd == SIOCGLIFADDR) {
1600 			/* fill in the if_laddrreq structure */
1601 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1602 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1603 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1604 				    ia->ia_dstaddr.sin6_len);
1605 			} else
1606 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1607 
1608 			iflr->prefixlen =
1609 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1610 
1611 			iflr->flags = ia->ia6_flags;	/* XXX */
1612 
1613 			return 0;
1614 		} else {
1615 			struct in6_aliasreq ifra;
1616 
1617 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1618 			bzero(&ifra, sizeof(ifra));
1619 			bcopy(iflr->iflr_name, ifra.ifra_name,
1620 			    sizeof(ifra.ifra_name));
1621 
1622 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1623 			    ia->ia_addr.sin6_len);
1624 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1625 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1626 				    ia->ia_dstaddr.sin6_len);
1627 			} else {
1628 				bzero(&ifra.ifra_dstaddr,
1629 				    sizeof(ifra.ifra_dstaddr));
1630 			}
1631 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1632 			    ia->ia_prefixmask.sin6_len);
1633 
1634 			ifra.ifra_flags = ia->ia6_flags;
1635 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1636 			    ifp, p);
1637 		}
1638 	    }
1639 	}
1640 
1641 	return EOPNOTSUPP;	/* just for safety */
1642 }
1643 
1644 /*
1645  * Initialize an interface's intetnet6 address
1646  * and routing table entry.
1647  */
1648 static int
1649 in6_ifinit(ifp, ia, sin6, newhost)
1650 	struct ifnet *ifp;
1651 	struct in6_ifaddr *ia;
1652 	struct sockaddr_in6 *sin6;
1653 	int newhost;
1654 {
1655 	int	error = 0, plen, ifacount = 0;
1656 	int	s = splnet();
1657 	struct ifaddr *ifa;
1658 
1659 	/*
1660 	 * Give the interface a chance to initialize
1661 	 * if this is its first address,
1662 	 * and to validate the address if necessary.
1663 	 */
1664 	for (ifa = ifp->if_addrlist.tqh_first; ifa;
1665 	     ifa = ifa->ifa_list.tqe_next)
1666 	{
1667 		if (ifa->ifa_addr == NULL)
1668 			continue;	/* just for safety */
1669 		if (ifa->ifa_addr->sa_family != AF_INET6)
1670 			continue;
1671 		ifacount++;
1672 	}
1673 
1674 	ia->ia_addr = *sin6;
1675 
1676 	if (ifacount <= 1 && ifp->if_ioctl &&
1677 	    (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia))) {
1678 		splx(s);
1679 		return (error);
1680 	}
1681 	splx(s);
1682 
1683 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1684 
1685 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1686 
1687 	/*
1688 	 * Special case:
1689 	 * If the destination address is specified for a point-to-point
1690 	 * interface, install a route to the destination as an interface
1691 	 * direct route.
1692 	 */
1693 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1694 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1695 		if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1696 				    RTF_UP | RTF_HOST)) != 0)
1697 			return (error);
1698 		ia->ia_flags |= IFA_ROUTE;
1699 	}
1700 
1701 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1702 	if (newhost) {
1703 		/* set the rtrequest function to create llinfo */
1704 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1705 		in6_ifaddloop(&(ia->ia_ifa));
1706 	}
1707 
1708 	if (ifp->if_flags & IFF_MULTICAST)
1709 		in6_restoremkludge(ia, ifp);
1710 
1711 	return (error);
1712 }
1713 
1714 /*
1715  * Multicast address kludge:
1716  * If there were any multicast addresses attached to this interface address,
1717  * either move them to another address on this interface, or save them until
1718  * such time as this interface is reconfigured for IPv6.
1719  */
1720 void
1721 in6_savemkludge(oia)
1722 	struct in6_ifaddr *oia;
1723 {
1724 	struct in6_ifaddr *ia;
1725 	struct in6_multi *in6m, *next;
1726 
1727 	IFP_TO_IA6(oia->ia_ifp, ia);
1728 	if (ia) {	/* there is another address */
1729 		for (in6m = oia->ia6_multiaddrs.lh_first; in6m; in6m = next){
1730 			next = in6m->in6m_entry.le_next;
1731 			IFAFREE(&in6m->in6m_ia->ia_ifa);
1732 			IFAREF(&ia->ia_ifa);
1733 			in6m->in6m_ia = ia;
1734 			LIST_INSERT_HEAD(&ia->ia6_multiaddrs, in6m, in6m_entry);
1735 		}
1736 	} else {	/* last address on this if deleted, save */
1737 		struct multi6_kludge *mk;
1738 
1739 		for (mk = in6_mk.lh_first; mk; mk = mk->mk_entry.le_next) {
1740 			if (mk->mk_ifp == oia->ia_ifp)
1741 				break;
1742 		}
1743 		if (mk == NULL) /* this should not happen! */
1744 			panic("in6_savemkludge: no kludge space");
1745 
1746 		for (in6m = oia->ia6_multiaddrs.lh_first; in6m; in6m = next){
1747 			next = in6m->in6m_entry.le_next;
1748 			IFAFREE(&in6m->in6m_ia->ia_ifa); /* release reference */
1749 			in6m->in6m_ia = NULL;
1750 			LIST_INSERT_HEAD(&mk->mk_head, in6m, in6m_entry);
1751 		}
1752 	}
1753 }
1754 
1755 /*
1756  * Continuation of multicast address hack:
1757  * If there was a multicast group list previously saved for this interface,
1758  * then we re-attach it to the first address configured on the i/f.
1759  */
1760 void
1761 in6_restoremkludge(ia, ifp)
1762 	struct in6_ifaddr *ia;
1763 	struct ifnet *ifp;
1764 {
1765 	struct multi6_kludge *mk;
1766 
1767 	for (mk = in6_mk.lh_first; mk; mk = mk->mk_entry.le_next) {
1768 		if (mk->mk_ifp == ifp) {
1769 			struct in6_multi *in6m, *next;
1770 
1771 			for (in6m = mk->mk_head.lh_first; in6m; in6m = next) {
1772 				next = in6m->in6m_entry.le_next;
1773 				in6m->in6m_ia = ia;
1774 				IFAREF(&ia->ia_ifa);
1775 				LIST_INSERT_HEAD(&ia->ia6_multiaddrs,
1776 				    in6m, in6m_entry);
1777 			}
1778 			LIST_INIT(&mk->mk_head);
1779 			break;
1780 		}
1781 	}
1782 }
1783 
1784 /*
1785  * Allocate space for the kludge at interface initialization time.
1786  * Formerly, we dynamically allocated the space in in6_savemkludge() with
1787  * malloc(M_WAITOK).  However, it was wrong since the function could be called
1788  * under an interrupt context (software timer on address lifetime expiration).
1789  * Also, we cannot just give up allocating the strucutre, since the group
1790  * membership structure is very complex and we need to keep it anyway.
1791  * Of course, this function MUST NOT be called under an interrupt context.
1792  * Specifically, it is expected to be called only from in6_ifattach(), though
1793  * it is a global function.
1794  */
1795 void
1796 in6_createmkludge(ifp)
1797 	struct ifnet *ifp;
1798 {
1799 	struct multi6_kludge *mk;
1800 
1801 	for (mk = in6_mk.lh_first; mk; mk = mk->mk_entry.le_next) {
1802 		/* If we've already had one, do not allocate. */
1803 		if (mk->mk_ifp == ifp)
1804 			return;
1805 	}
1806 
1807 	mk = malloc(sizeof(*mk), M_IPMADDR, M_WAITOK);
1808 
1809 	bzero(mk, sizeof(*mk));
1810 	LIST_INIT(&mk->mk_head);
1811 	mk->mk_ifp = ifp;
1812 	LIST_INSERT_HEAD(&in6_mk, mk, mk_entry);
1813 }
1814 
1815 void
1816 in6_purgemkludge(ifp)
1817 	struct ifnet *ifp;
1818 {
1819 	struct multi6_kludge *mk;
1820 	struct in6_multi *in6m;
1821 
1822 	for (mk = in6_mk.lh_first; mk; mk = mk->mk_entry.le_next) {
1823 		if (mk->mk_ifp != ifp)
1824 			continue;
1825 
1826 		/* leave from all multicast groups joined */
1827 		while ((in6m = LIST_FIRST(&mk->mk_head)) != NULL)
1828 			in6_delmulti(in6m);
1829 		LIST_REMOVE(mk, mk_entry);
1830 		free(mk, M_IPMADDR);
1831 		break;
1832 	}
1833 }
1834 
1835 /*
1836  * Add an address to the list of IP6 multicast addresses for a
1837  * given interface.
1838  */
1839 struct	in6_multi *
1840 in6_addmulti(maddr6, ifp, errorp)
1841 	struct in6_addr *maddr6;
1842 	struct ifnet *ifp;
1843 	int *errorp;
1844 {
1845 	struct	in6_ifaddr *ia;
1846 	struct	in6_ifreq ifr;
1847 	struct	in6_multi *in6m;
1848 	int	s = splsoftnet();
1849 
1850 	*errorp = 0;
1851 	/*
1852 	 * See if address already in list.
1853 	 */
1854 	IN6_LOOKUP_MULTI(*maddr6, ifp, in6m);
1855 	if (in6m != NULL) {
1856 		/*
1857 		 * Found it; just increment the refrence count.
1858 		 */
1859 		in6m->in6m_refcount++;
1860 	} else {
1861 		/*
1862 		 * New address; allocate a new multicast record
1863 		 * and link it into the interface's multicast list.
1864 		 */
1865 		in6m = (struct in6_multi *)
1866 			malloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT);
1867 		if (in6m == NULL) {
1868 			splx(s);
1869 			*errorp = ENOBUFS;
1870 			return (NULL);
1871 		}
1872 		in6m->in6m_addr = *maddr6;
1873 		in6m->in6m_ifp = ifp;
1874 		in6m->in6m_refcount = 1;
1875 		IFP_TO_IA6(ifp, ia);
1876 		if (ia == NULL) {
1877 			free(in6m, M_IPMADDR);
1878 			splx(s);
1879 			*errorp = EADDRNOTAVAIL; /* appropriate? */
1880 			return (NULL);
1881 		}
1882 		in6m->in6m_ia = ia;
1883 		IFAREF(&ia->ia_ifa); /* gain a reference */
1884 		LIST_INSERT_HEAD(&ia->ia6_multiaddrs, in6m, in6m_entry);
1885 
1886 		/*
1887 		 * Ask the network driver to update its multicast reception
1888 		 * filter appropriately for the new address.
1889 		 */
1890 		bzero(&ifr.ifr_addr, sizeof(struct sockaddr_in6));
1891 		ifr.ifr_addr.sin6_len = sizeof(struct sockaddr_in6);
1892 		ifr.ifr_addr.sin6_family = AF_INET6;
1893 		ifr.ifr_addr.sin6_addr = *maddr6;
1894 		if (ifp->if_ioctl == NULL)
1895 			*errorp = ENXIO; /* XXX: appropriate? */
1896 		else
1897 			*errorp = (*ifp->if_ioctl)(ifp, SIOCADDMULTI,
1898 			    (caddr_t)&ifr);
1899 		if (*errorp) {
1900 			LIST_REMOVE(in6m, in6m_entry);
1901 			free(in6m, M_IPMADDR);
1902 			IFAFREE(&ia->ia_ifa);
1903 			splx(s);
1904 			return (NULL);
1905 		}
1906 		/*
1907 		 * Let MLD6 know that we have joined a new IP6 multicast
1908 		 * group.
1909 		 */
1910 		mld6_start_listening(in6m);
1911 	}
1912 	splx(s);
1913 	return (in6m);
1914 }
1915 
1916 /*
1917  * Delete a multicast address record.
1918  */
1919 void
1920 in6_delmulti(in6m)
1921 	struct in6_multi *in6m;
1922 {
1923 	struct	in6_ifreq ifr;
1924 	struct	in6_ifaddr *ia;
1925 	int	s = splsoftnet();
1926 
1927 	if (--in6m->in6m_refcount == 0) {
1928 		/*
1929 		 * No remaining claims to this record; let MLD6 know
1930 		 * that we are leaving the multicast group.
1931 		 */
1932 		mld6_stop_listening(in6m);
1933 
1934 		/*
1935 		 * Unlink from list.
1936 		 */
1937 		LIST_REMOVE(in6m, in6m_entry);
1938 		if (in6m->in6m_ia) {
1939 			IFAFREE(&in6m->in6m_ia->ia_ifa); /* release reference */
1940 		}
1941 		/*
1942 		 * Delete all references of this multicasting group from
1943 		 * the membership arrays
1944 		 */
1945 		for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1946 			struct in6_multi_mship *imm;
1947 			LIST_FOREACH(imm, &ia->ia6_memberships,
1948 			    i6mm_chain) {
1949 				if (imm->i6mm_maddr == in6m)
1950 					imm->i6mm_maddr = NULL;
1951 			}
1952 		}
1953 
1954 		/*
1955 		 * Notify the network driver to update its multicast
1956 		 * reception filter.
1957 		 */
1958 		bzero(&ifr.ifr_addr, sizeof(struct sockaddr_in6));
1959 		ifr.ifr_addr.sin6_len = sizeof(struct sockaddr_in6);
1960 		ifr.ifr_addr.sin6_family = AF_INET6;
1961 		ifr.ifr_addr.sin6_addr = in6m->in6m_addr;
1962 		(*in6m->in6m_ifp->if_ioctl)(in6m->in6m_ifp,
1963 					    SIOCDELMULTI, (caddr_t)&ifr);
1964 		free(in6m, M_IPMADDR);
1965 	}
1966 	splx(s);
1967 }
1968 
1969 struct in6_multi_mship *
1970 in6_joingroup(ifp, addr, errorp)
1971 	struct ifnet *ifp;
1972 	struct in6_addr *addr;
1973 	int *errorp;
1974 {
1975 	struct in6_multi_mship *imm;
1976 
1977 	imm = malloc(sizeof(*imm), M_IPMADDR, M_NOWAIT);
1978 	if (!imm) {
1979 		*errorp = ENOBUFS;
1980 		return NULL;
1981 	}
1982 	imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp);
1983 	if (!imm->i6mm_maddr) {
1984 		/* *errorp is alrady set */
1985 		free(imm, M_IPMADDR);
1986 		return NULL;
1987 	}
1988 	return imm;
1989 }
1990 
1991 int
1992 in6_leavegroup(imm)
1993 	struct in6_multi_mship *imm;
1994 {
1995 
1996 	if (imm->i6mm_maddr)
1997 		in6_delmulti(imm->i6mm_maddr);
1998 	free(imm,  M_IPMADDR);
1999 	return 0;
2000 }
2001 
2002 /*
2003  * Find an IPv6 interface link-local address specific to an interface.
2004  */
2005 struct in6_ifaddr *
2006 in6ifa_ifpforlinklocal(ifp, ignoreflags)
2007 	struct ifnet *ifp;
2008 	int ignoreflags;
2009 {
2010 	struct ifaddr *ifa;
2011 
2012 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
2013 	{
2014 		if (ifa->ifa_addr == NULL)
2015 			continue;	/* just for safety */
2016 		if (ifa->ifa_addr->sa_family != AF_INET6)
2017 			continue;
2018 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
2019 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
2020 			     ignoreflags) != 0)
2021 				continue;
2022 			break;
2023 		}
2024 	}
2025 
2026 	return ((struct in6_ifaddr *)ifa);
2027 }
2028 
2029 
2030 /*
2031  * find the internet address corresponding to a given interface and address.
2032  */
2033 struct in6_ifaddr *
2034 in6ifa_ifpwithaddr(ifp, addr)
2035 	struct ifnet *ifp;
2036 	struct in6_addr *addr;
2037 {
2038 	struct ifaddr *ifa;
2039 
2040 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
2041 	{
2042 		if (ifa->ifa_addr == NULL)
2043 			continue;	/* just for safety */
2044 		if (ifa->ifa_addr->sa_family != AF_INET6)
2045 			continue;
2046 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
2047 			break;
2048 	}
2049 
2050 	return ((struct in6_ifaddr *)ifa);
2051 }
2052 
2053 /*
2054  * Convert IP6 address to printable (loggable) representation.
2055  */
2056 static char digits[] = "0123456789abcdef";
2057 static int ip6round = 0;
2058 char *
2059 ip6_sprintf(addr)
2060 	const struct in6_addr *addr;
2061 {
2062 	static char ip6buf[8][48];
2063 	int i;
2064 	char *cp;
2065 	const u_int16_t *a = (const u_int16_t *)addr;
2066 	const u_int8_t *d;
2067 	int dcolon = 0;
2068 
2069 	ip6round = (ip6round + 1) & 7;
2070 	cp = ip6buf[ip6round];
2071 
2072 	for (i = 0; i < 8; i++) {
2073 		if (dcolon == 1) {
2074 			if (*a == 0) {
2075 				if (i == 7)
2076 					*cp++ = ':';
2077 				a++;
2078 				continue;
2079 			} else
2080 				dcolon = 2;
2081 		}
2082 		if (*a == 0) {
2083 			if (dcolon == 0 && *(a + 1) == 0) {
2084 				if (i == 0)
2085 					*cp++ = ':';
2086 				*cp++ = ':';
2087 				dcolon = 1;
2088 			} else {
2089 				*cp++ = '0';
2090 				*cp++ = ':';
2091 			}
2092 			a++;
2093 			continue;
2094 		}
2095 		d = (const u_char *)a;
2096 		*cp++ = digits[*d >> 4];
2097 		*cp++ = digits[*d++ & 0xf];
2098 		*cp++ = digits[*d >> 4];
2099 		*cp++ = digits[*d & 0xf];
2100 		*cp++ = ':';
2101 		a++;
2102 	}
2103 	*--cp = 0;
2104 	return (ip6buf[ip6round]);
2105 }
2106 
2107 /*
2108  * Determine if an address is on a local network.
2109  */
2110 int
2111 in6_localaddr(in6)
2112 	struct in6_addr *in6;
2113 {
2114 	struct in6_ifaddr *ia;
2115 
2116 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2117 		return (1);
2118 
2119 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2120 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2121 					      &ia->ia_prefixmask.sin6_addr))
2122 			return (1);
2123 
2124 	return (0);
2125 }
2126 
2127 /*
2128  * Get a scope of the address. Node-local, link-local, site-local or global.
2129  */
2130 int
2131 in6_addrscope (addr)
2132 struct in6_addr *addr;
2133 {
2134 	int scope;
2135 
2136 	if (addr->s6_addr8[0] == 0xfe) {
2137 		scope = addr->s6_addr8[1] & 0xc0;
2138 
2139 		switch (scope) {
2140 		case 0x80:
2141 			return IPV6_ADDR_SCOPE_LINKLOCAL;
2142 		case 0xc0:
2143 			return IPV6_ADDR_SCOPE_SITELOCAL;
2144 		default:
2145 			return IPV6_ADDR_SCOPE_GLOBAL; /* just in case */
2146 		}
2147 	}
2148 
2149 
2150 	if (addr->s6_addr8[0] == 0xff) {
2151 		scope = addr->s6_addr8[1] & 0x0f;
2152 
2153 		/*
2154 		 * due to other scope such as reserved,
2155 		 * return scope doesn't work.
2156 		 */
2157 		switch (scope) {
2158 		case IPV6_ADDR_SCOPE_NODELOCAL:
2159 			return IPV6_ADDR_SCOPE_NODELOCAL;
2160 		case IPV6_ADDR_SCOPE_LINKLOCAL:
2161 			return IPV6_ADDR_SCOPE_LINKLOCAL;
2162 		case IPV6_ADDR_SCOPE_SITELOCAL:
2163 			return IPV6_ADDR_SCOPE_SITELOCAL;
2164 		default:
2165 			return IPV6_ADDR_SCOPE_GLOBAL;
2166 		}
2167 	}
2168 
2169 	if (bcmp(&in6addr_loopback, addr, sizeof(*addr) - 1) == 0) {
2170 		if (addr->s6_addr8[15] == 1) /* loopback */
2171 			return IPV6_ADDR_SCOPE_NODELOCAL;
2172 		if (addr->s6_addr8[15] == 0) /* unspecified */
2173 			return IPV6_ADDR_SCOPE_LINKLOCAL;
2174 	}
2175 
2176 	return IPV6_ADDR_SCOPE_GLOBAL;
2177 }
2178 
2179 int
2180 in6_addr2scopeid(ifp, addr)
2181 	struct ifnet *ifp;	/* must not be NULL */
2182 	struct in6_addr *addr;	/* must not be NULL */
2183 {
2184 	int scope = in6_addrscope(addr);
2185 
2186 	switch (scope) {
2187 	case IPV6_ADDR_SCOPE_NODELOCAL:
2188 		return (-1);	/* XXX: is this an appropriate value? */
2189 
2190 	case IPV6_ADDR_SCOPE_LINKLOCAL:
2191 		/* XXX: we do not distinguish between a link and an I/F. */
2192 		return (ifp->if_index);
2193 
2194 	case IPV6_ADDR_SCOPE_SITELOCAL:
2195 		return (0);	/* XXX: invalid. */
2196 
2197 	default:
2198 		return (0);	/* XXX: treat as global. */
2199 	}
2200 }
2201 
2202 int
2203 in6_is_addr_deprecated(sa6)
2204 	struct sockaddr_in6 *sa6;
2205 {
2206 	struct in6_ifaddr *ia;
2207 
2208 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
2209 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2210 		    &sa6->sin6_addr) &&
2211 #ifdef SCOPEDROUTING
2212 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
2213 #endif
2214 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
2215 			return (1); /* true */
2216 
2217 		/* XXX: do we still have to go thru the rest of the list? */
2218 	}
2219 
2220 	return (0);		/* false */
2221 }
2222 
2223 /*
2224  * return length of part which dst and src are equal
2225  * hard coding...
2226  */
2227 int
2228 in6_matchlen(src, dst)
2229 struct in6_addr *src, *dst;
2230 {
2231 	int match = 0;
2232 	u_char *s = (u_char *)src, *d = (u_char *)dst;
2233 	u_char *lim = s + 16, r;
2234 
2235 	while (s < lim)
2236 		if ((r = (*d++ ^ *s++)) != 0) {
2237 			while (r < 128) {
2238 				match++;
2239 				r <<= 1;
2240 			}
2241 			break;
2242 		} else
2243 			match += 8;
2244 	return match;
2245 }
2246 
2247 /* XXX: to be scope conscious */
2248 int
2249 in6_are_prefix_equal(p1, p2, len)
2250 	struct in6_addr *p1, *p2;
2251 	int len;
2252 {
2253 	int bytelen, bitlen;
2254 
2255 	/* sanity check */
2256 	if (0 > len || len > 128) {
2257 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2258 		    len);
2259 		return (0);
2260 	}
2261 
2262 	bytelen = len / 8;
2263 	bitlen = len % 8;
2264 
2265 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2266 		return (0);
2267 	if (bitlen != 0 &&
2268 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2269 	    p2->s6_addr[bytelen] >> (8 - bitlen))
2270 		return (0);
2271 
2272 	return (1);
2273 }
2274 
2275 void
2276 in6_prefixlen2mask(maskp, len)
2277 	struct in6_addr *maskp;
2278 	int len;
2279 {
2280 	static const u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2281 	int bytelen, bitlen, i;
2282 
2283 	/* sanity check */
2284 	if (0 > len || len > 128) {
2285 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2286 		    len);
2287 		return;
2288 	}
2289 
2290 	bzero(maskp, sizeof(*maskp));
2291 	bytelen = len / 8;
2292 	bitlen = len % 8;
2293 	for (i = 0; i < bytelen; i++)
2294 		maskp->s6_addr[i] = 0xff;
2295 	if (bitlen)
2296 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2297 }
2298 
2299 /*
2300  * return the best address out of the same scope
2301  */
2302 struct in6_ifaddr *
2303 in6_ifawithscope(oifp, dst)
2304 	struct ifnet *oifp;
2305 	struct in6_addr *dst;
2306 {
2307 	int dst_scope =	in6_addrscope(dst), src_scope, best_scope = 0;
2308 	int blen = -1;
2309 	struct ifaddr *ifa;
2310 	struct ifnet *ifp;
2311 	struct in6_ifaddr *ifa_best = NULL;
2312 
2313 	if (oifp == NULL) {
2314 		printf("in6_ifawithscope: output interface is not specified\n");
2315 		return (NULL);
2316 	}
2317 
2318 	/*
2319 	 * We search for all addresses on all interfaces from the beginning.
2320 	 * Comparing an interface with the outgoing interface will be done
2321 	 * only at the final stage of tiebreaking.
2322 	 */
2323 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2324 	{
2325 		/*
2326 		 * We can never take an address that breaks the scope zone
2327 		 * of the destination.
2328 		 */
2329 		if (in6_addr2scopeid(ifp, dst) != in6_addr2scopeid(oifp, dst))
2330 			continue;
2331 
2332 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
2333 		     ifa = ifa->ifa_list.tqe_next)
2334 		{
2335 			int tlen = -1, dscopecmp, bscopecmp, matchcmp;
2336 
2337 			if (ifa->ifa_addr->sa_family != AF_INET6)
2338 				continue;
2339 
2340 			src_scope = in6_addrscope(IFA_IN6(ifa));
2341 
2342 #ifdef ADDRSELECT_DEBUG		/* should be removed after stabilization */
2343 			dscopecmp = IN6_ARE_SCOPE_CMP(src_scope, dst_scope);
2344 			printf("in6_ifawithscope: dst=%s bestaddr=%s, "
2345 			       "newaddr=%s, scope=%x, dcmp=%d, bcmp=%d, "
2346 			       "matchlen=%d, flgs=%x\n",
2347 			       ip6_sprintf(dst),
2348 			       ifa_best ? ip6_sprintf(&ifa_best->ia_addr.sin6_addr) : "none",
2349 			       ip6_sprintf(IFA_IN6(ifa)), src_scope,
2350 			       dscopecmp,
2351 			       ifa_best ? IN6_ARE_SCOPE_CMP(src_scope, best_scope) : -1,
2352 			       in6_matchlen(IFA_IN6(ifa), dst),
2353 			       ((struct in6_ifaddr *)ifa)->ia6_flags);
2354 #endif
2355 
2356 			/*
2357 			 * Don't use an address before completing DAD
2358 			 * nor a duplicated address.
2359 			 */
2360 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2361 			    IN6_IFF_NOTREADY)
2362 				continue;
2363 
2364 			/* XXX: is there any case to allow anycasts? */
2365 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2366 			    IN6_IFF_ANYCAST)
2367 				continue;
2368 
2369 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2370 			    IN6_IFF_DETACHED)
2371 				continue;
2372 
2373 			/*
2374 			 * If this is the first address we find,
2375 			 * keep it anyway.
2376 			 */
2377 			if (ifa_best == NULL)
2378 				goto replace;
2379 
2380 			/*
2381 			 * ifa_best is never NULL beyond this line except
2382 			 * within the block labeled "replace".
2383 			 */
2384 
2385 			/*
2386 			 * If ifa_best has a smaller scope than dst and
2387 			 * the current address has a larger one than
2388 			 * (or equal to) dst, always replace ifa_best.
2389 			 * Also, if the current address has a smaller scope
2390 			 * than dst, ignore it unless ifa_best also has a
2391 			 * smaller scope.
2392 			 */
2393 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0 &&
2394 			    IN6_ARE_SCOPE_CMP(src_scope, dst_scope) >= 0)
2395 				goto replace;
2396 			if (IN6_ARE_SCOPE_CMP(src_scope, dst_scope) < 0 &&
2397 			    IN6_ARE_SCOPE_CMP(best_scope, dst_scope) >= 0)
2398 				continue;
2399 
2400 			/*
2401 			 * A deprecated address SHOULD NOT be used in new
2402 			 * communications if an alternate (non-deprecated)
2403 			 * address is available and has sufficient scope.
2404 			 * RFC 2462, Section 5.5.4.
2405 			 */
2406 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2407 			    IN6_IFF_DEPRECATED) {
2408 				/*
2409 				 * Ignore any deprecated addresses if
2410 				 * specified by configuration.
2411 				 */
2412 				if (!ip6_use_deprecated)
2413 					continue;
2414 
2415 				/*
2416 				 * If we have already found a non-deprecated
2417 				 * candidate, just ignore deprecated addresses.
2418 				 */
2419 				if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED)
2420 				    == 0)
2421 					continue;
2422 			}
2423 
2424 			/*
2425 			 * A non-deprecated address is always preferred
2426 			 * to a deprecated one regardless of scopes and
2427 			 * address matching.
2428 			 */
2429 			if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED) &&
2430 			    (((struct in6_ifaddr *)ifa)->ia6_flags &
2431 			     IN6_IFF_DEPRECATED) == 0)
2432 				goto replace;
2433 
2434 			/*
2435 			 * At this point, we have two cases:
2436 			 * 1. we are looking at a non-deprecated address,
2437 			 *    and ifa_best is also non-deprecated.
2438 			 * 2. we are looking at a deprecated address,
2439 			 *    and ifa_best is also deprecated.
2440 			 * Also, we do not have to consider a case where
2441 			 * the scope of if_best is larger(smaller) than dst and
2442 			 * the scope of the current address is smaller(larger)
2443 			 * than dst. Such a case has already been covered.
2444 			 * Tiebreaking is done according to the following
2445 			 * items:
2446 			 * - the scope comparison between the address and
2447 			 *   dst (dscopecmp)
2448 			 * - the scope comparison between the address and
2449 			 *   ifa_best (bscopecmp)
2450 			 * - if the address match dst longer than ifa_best
2451 			 *   (matchcmp)
2452 			 * - if the address is on the outgoing I/F (outI/F)
2453 			 *
2454 			 * Roughly speaking, the selection policy is
2455 			 * - the most important item is scope. The same scope
2456 			 *   is best. Then search for a larger scope.
2457 			 *   Smaller scopes are the last resort.
2458 			 * - A deprecated address is chosen only when we have
2459 			 *   no address that has an enough scope, but is
2460 			 *   prefered to any addresses of smaller scopes.
2461 			 * - Longest address match against dst is considered
2462 			 *   only for addresses that has the same scope of dst.
2463 			 * - If there is no other reasons to choose one,
2464 			 *   addresses on the outgoing I/F are preferred.
2465 			 *
2466 			 * The precise decision table is as follows:
2467 			 * dscopecmp bscopecmp matchcmp outI/F | replace?
2468 			 *    !equal     equal      N/A    Yes |      Yes (1)
2469 			 *    !equal     equal      N/A     No |       No (2)
2470 			 *    larger    larger      N/A    N/A |       No (3)
2471 			 *    larger   smaller      N/A    N/A |      Yes (4)
2472 			 *   smaller    larger      N/A    N/A |      Yes (5)
2473 			 *   smaller   smaller      N/A    N/A |       No (6)
2474 			 *     equal   smaller      N/A    N/A |      Yes (7)
2475 			 *     equal    larger       (already done)
2476 			 *     equal     equal   larger    N/A |      Yes (8)
2477 			 *     equal     equal  smaller    N/A |       No (9)
2478 			 *     equal     equal    equal    Yes |      Yes (a)
2479 			 *     eaual     eqaul    equal     No |       No (b)
2480 			 */
2481 			dscopecmp = IN6_ARE_SCOPE_CMP(src_scope, dst_scope);
2482 			bscopecmp = IN6_ARE_SCOPE_CMP(src_scope, best_scope);
2483 
2484 			if (dscopecmp && bscopecmp == 0) {
2485 				if (oifp == ifp) /* (1) */
2486 					goto replace;
2487 				continue; /* (2) */
2488 			}
2489 			if (dscopecmp > 0) {
2490 				if (bscopecmp > 0) /* (3) */
2491 					continue;
2492 				goto replace; /* (4) */
2493 			}
2494 			if (dscopecmp < 0) {
2495 				if (bscopecmp > 0) /* (5) */
2496 					goto replace;
2497 				continue; /* (6) */
2498 			}
2499 
2500 			/* now dscopecmp must be 0 */
2501 			if (bscopecmp < 0)
2502 				goto replace; /* (7) */
2503 
2504 			/*
2505 			 * At last both dscopecmp and bscopecmp must be 0.
2506 			 * We need address matching against dst for
2507 			 * tiebreaking.
2508 			 */
2509 			tlen = in6_matchlen(IFA_IN6(ifa), dst);
2510 			matchcmp = tlen - blen;
2511 			if (matchcmp > 0) /* (8) */
2512 				goto replace;
2513 			if (matchcmp < 0) /* (9) */
2514 				continue;
2515 			if (oifp == ifp) /* (a) */
2516 				goto replace;
2517 			continue; /* (b) */
2518 
2519 		  replace:
2520 			ifa_best = (struct in6_ifaddr *)ifa;
2521 			blen = tlen >= 0 ? tlen :
2522 				in6_matchlen(IFA_IN6(ifa), dst);
2523 			best_scope = in6_addrscope(&ifa_best->ia_addr.sin6_addr);
2524 		}
2525 	}
2526 
2527 	/* count statistics for future improvements */
2528 	if (ifa_best == NULL)
2529 		ip6stat.ip6s_sources_none++;
2530 	else {
2531 		if (oifp == ifa_best->ia_ifp)
2532 			ip6stat.ip6s_sources_sameif[best_scope]++;
2533 		else
2534 			ip6stat.ip6s_sources_otherif[best_scope]++;
2535 
2536 		if (best_scope == dst_scope)
2537 			ip6stat.ip6s_sources_samescope[best_scope]++;
2538 		else
2539 			ip6stat.ip6s_sources_otherscope[best_scope]++;
2540 
2541 		if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED) != 0)
2542 			ip6stat.ip6s_sources_deprecated[best_scope]++;
2543 	}
2544 
2545 	return (ifa_best);
2546 }
2547 
2548 /*
2549  * return the best address out of the same scope. if no address was
2550  * found, return the first valid address from designated IF.
2551  */
2552 struct in6_ifaddr *
2553 in6_ifawithifp(ifp, dst)
2554 	struct ifnet *ifp;
2555 	struct in6_addr *dst;
2556 {
2557 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2558 	struct ifaddr *ifa;
2559 	struct in6_ifaddr *besta = 0;
2560 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2561 
2562 	dep[0] = dep[1] = NULL;
2563 
2564 	/*
2565 	 * We first look for addresses in the same scope.
2566 	 * If there is one, return it.
2567 	 * If two or more, return one which matches the dst longest.
2568 	 * If none, return one of global addresses assigned other ifs.
2569 	 */
2570 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
2571 	{
2572 		if (ifa->ifa_addr->sa_family != AF_INET6)
2573 			continue;
2574 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2575 			continue; /* XXX: is there any case to allow anycast? */
2576 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2577 			continue; /* don't use this interface */
2578 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2579 			continue;
2580 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2581 			if (ip6_use_deprecated)
2582 				dep[0] = (struct in6_ifaddr *)ifa;
2583 			continue;
2584 		}
2585 
2586 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2587 			/*
2588 			 * call in6_matchlen() as few as possible
2589 			 */
2590 			if (besta) {
2591 				if (blen == -1)
2592 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2593 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2594 				if (tlen > blen) {
2595 					blen = tlen;
2596 					besta = (struct in6_ifaddr *)ifa;
2597 				}
2598 			} else
2599 				besta = (struct in6_ifaddr *)ifa;
2600 		}
2601 	}
2602 	if (besta)
2603 		return (besta);
2604 
2605 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
2606 	{
2607 		if (ifa->ifa_addr->sa_family != AF_INET6)
2608 			continue;
2609 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2610 			continue; /* XXX: is there any case to allow anycast? */
2611 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2612 			continue; /* don't use this interface */
2613 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2614 			continue;
2615 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2616 			if (ip6_use_deprecated)
2617 				dep[1] = (struct in6_ifaddr *)ifa;
2618 			continue;
2619 		}
2620 
2621 		return (struct in6_ifaddr *)ifa;
2622 	}
2623 
2624 	/* use the last-resort values, that are, deprecated addresses */
2625 	if (dep[0])
2626 		return dep[0];
2627 	if (dep[1])
2628 		return dep[1];
2629 
2630 	return NULL;
2631 }
2632 
2633 /*
2634  * perform DAD when interface becomes IFF_UP.
2635  */
2636 void
2637 in6_if_up(ifp)
2638 	struct ifnet *ifp;
2639 {
2640 	struct ifaddr *ifa;
2641 	struct in6_ifaddr *ia;
2642 	int dad_delay;		/* delay ticks before DAD output */
2643 
2644 	/*
2645 	 * special cases, like 6to4, are handled in in6_ifattach
2646 	 */
2647 	in6_ifattach(ifp, NULL);
2648 
2649 	dad_delay = 0;
2650 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
2651 	{
2652 		if (ifa->ifa_addr->sa_family != AF_INET6)
2653 			continue;
2654 		ia = (struct in6_ifaddr *)ifa;
2655 		if (ia->ia6_flags & IN6_IFF_TENTATIVE)
2656 			nd6_dad_start(ifa, &dad_delay);
2657 	}
2658 }
2659 
2660 int
2661 in6if_do_dad(ifp)
2662 	struct ifnet *ifp;
2663 {
2664 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2665 		return (0);
2666 
2667 	switch (ifp->if_type) {
2668 	case IFT_FAITH:
2669 		/*
2670 		 * These interfaces do not have the IFF_LOOPBACK flag,
2671 		 * but loop packets back.  We do not have to do DAD on such
2672 		 * interfaces.  We should even omit it, because loop-backed
2673 		 * NS would confuse the DAD procedure.
2674 		 */
2675 		return (0);
2676 	default:
2677 		/*
2678 		 * Our DAD routine requires the interface up and running.
2679 		 * However, some interfaces can be up before the RUNNING
2680 		 * status.  Additionaly, users may try to assign addresses
2681 		 * before the interface becomes up (or running).
2682 		 * We simply skip DAD in such a case as a work around.
2683 		 * XXX: we should rather mark "tentative" on such addresses,
2684 		 * and do DAD after the interface becomes ready.
2685 		 */
2686 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2687 		    (IFF_UP|IFF_RUNNING))
2688 			return (0);
2689 
2690 		return (1);
2691 	}
2692 }
2693 
2694 /*
2695  * Calculate max IPv6 MTU through all the interfaces and store it
2696  * to in6_maxmtu.
2697  */
2698 void
2699 in6_setmaxmtu()
2700 {
2701 	unsigned long maxmtu = 0;
2702 	struct ifnet *ifp;
2703 
2704 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2705 	{
2706 		/* this function can be called during ifnet initialization */
2707 		if (!ifp->if_afdata[AF_INET6])
2708 			continue;
2709 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2710 		    IN6_LINKMTU(ifp) > maxmtu)
2711 			maxmtu = IN6_LINKMTU(ifp);
2712 	}
2713 	if (maxmtu)	     /* update only when maxmtu is positive */
2714 		in6_maxmtu = maxmtu;
2715 }
2716 
2717 void *
2718 in6_domifattach(ifp)
2719 	struct ifnet *ifp;
2720 {
2721 	struct in6_ifextra *ext;
2722 
2723 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2724 	bzero(ext, sizeof(*ext));
2725 
2726 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2727 	    M_IFADDR, M_WAITOK);
2728 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2729 
2730 	ext->icmp6_ifstat =
2731 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2732 	    M_IFADDR, M_WAITOK);
2733 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2734 
2735 	ext->nd_ifinfo = nd6_ifattach(ifp);
2736 	return ext;
2737 }
2738 
2739 void
2740 in6_domifdetach(ifp, aux)
2741 	struct ifnet *ifp;
2742 	void *aux;
2743 {
2744 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2745 
2746 	nd6_ifdetach(ext->nd_ifinfo);
2747 	free(ext->in6_ifstat, M_IFADDR);
2748 	free(ext->icmp6_ifstat, M_IFADDR);
2749 	free(ext, M_IFADDR);
2750 }
2751