xref: /netbsd-src/sys/netinet6/in6.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: in6.c,v 1.113 2006/10/15 07:00:44 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 1.113 2006/10/15 07:00:44 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 /*
113  * Definitions of some constant IP6 addresses.
114  */
115 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
116 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
117 const struct in6_addr in6addr_nodelocal_allnodes =
118 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
119 const struct in6_addr in6addr_linklocal_allnodes =
120 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
121 const struct in6_addr in6addr_linklocal_allrouters =
122 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
123 
124 const struct in6_addr in6mask0 = IN6MASK0;
125 const struct in6_addr in6mask32 = IN6MASK32;
126 const struct in6_addr in6mask64 = IN6MASK64;
127 const struct in6_addr in6mask96 = IN6MASK96;
128 const struct in6_addr in6mask128 = IN6MASK128;
129 
130 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
131 				     0, 0, IN6ADDR_ANY_INIT, 0};
132 
133 static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
134 	struct ifnet *, struct lwp *));
135 static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
136 	struct sockaddr_in6 *, int));
137 static void in6_unlink_ifa __P((struct in6_ifaddr *, struct ifnet *));
138 
139 /*
140  * Subroutine for in6_ifaddloop() and in6_ifremloop().
141  * This routine does actual work.
142  */
143 static void
144 in6_ifloop_request(int cmd, struct ifaddr *ifa)
145 {
146 	struct sockaddr_in6 lo_sa;
147 	struct sockaddr_in6 all1_sa;
148 	struct rtentry *nrt = NULL;
149 	int e;
150 
151 	bzero(&lo_sa, sizeof(lo_sa));
152 	bzero(&all1_sa, sizeof(all1_sa));
153 	lo_sa.sin6_family = all1_sa.sin6_family = AF_INET6;
154 	lo_sa.sin6_len = all1_sa.sin6_len = sizeof(struct sockaddr_in6);
155 	lo_sa.sin6_addr = in6addr_loopback;
156 	all1_sa.sin6_addr = in6mask128;
157 
158 	/*
159 	 * We specify the address itself as the gateway, and set the
160 	 * RTF_LLINFO flag, so that the corresponding host route would have
161 	 * the flag, and thus applications that assume traditional behavior
162 	 * would be happy.  Note that we assume the caller of the function
163 	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
164 	 * which changes the outgoing interface to the loopback interface.
165 	 */
166 	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr,
167 	    (struct sockaddr *)&all1_sa, RTF_UP|RTF_HOST|RTF_LLINFO, &nrt);
168 	if (e != 0) {
169 		log(LOG_ERR, "in6_ifloop_request: "
170 		    "%s operation failed for %s (errno=%d)\n",
171 		    cmd == RTM_ADD ? "ADD" : "DELETE",
172 		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
173 		    e);
174 	}
175 
176 	/*
177 	 * Make sure rt_ifa be equal to IFA, the second argument of the
178 	 * function.
179 	 * We need this because when we refer to rt_ifa->ia6_flags in
180 	 * ip6_input, we assume that the rt_ifa points to the address instead
181 	 * of the loopback address.
182 	 */
183 	if (cmd == RTM_ADD && nrt && ifa != nrt->rt_ifa) {
184 		IFAFREE(nrt->rt_ifa);
185 		IFAREF(ifa);
186 		nrt->rt_ifa = ifa;
187 	}
188 
189 	/*
190 	 * Report the addition/removal of the address to the routing socket.
191 	 * XXX: since we called rtinit for a p2p interface with a destination,
192 	 *      we end up reporting twice in such a case.  Should we rather
193 	 *      omit the second report?
194 	 */
195 	if (nrt) {
196 		rt_newaddrmsg(cmd, ifa, e, nrt);
197 		if (cmd == RTM_DELETE) {
198 			if (nrt->rt_refcnt <= 0) {
199 				/* XXX: we should free the entry ourselves. */
200 				nrt->rt_refcnt++;
201 				rtfree(nrt);
202 			}
203 		} else {
204 			/* the cmd must be RTM_ADD here */
205 			nrt->rt_refcnt--;
206 		}
207 	}
208 }
209 
210 /*
211  * Add ownaddr as loopback rtentry.  We previously add the route only if
212  * necessary (ex. on a p2p link).  However, since we now manage addresses
213  * separately from prefixes, we should always add the route.  We can't
214  * rely on the cloning mechanism from the corresponding interface route
215  * any more.
216  */
217 void
218 in6_ifaddloop(struct ifaddr *ifa)
219 {
220 	struct rtentry *rt;
221 
222 	/* If there is no loopback entry, allocate one. */
223 	rt = rtalloc1(ifa->ifa_addr, 0);
224 	if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
225 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
226 		in6_ifloop_request(RTM_ADD, ifa);
227 	if (rt)
228 		rt->rt_refcnt--;
229 }
230 
231 /*
232  * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
233  * if it exists.
234  */
235 void
236 in6_ifremloop(struct ifaddr *ifa)
237 {
238 	struct in6_ifaddr *ia;
239 	struct rtentry *rt;
240 	int ia_count = 0;
241 
242 	/*
243 	 * Some of BSD variants do not remove cloned routes
244 	 * from an interface direct route, when removing the direct route
245 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
246 	 * cloned routes, they could fail to remove the cloned routes when
247 	 * we handle multple addresses that share a common prefix.
248 	 * So, we should remove the route corresponding to the deleted address.
249 	 */
250 
251 	/*
252 	 * Delete the entry only if exact one ifa exists.  More than one ifa
253 	 * can exist if we assign a same single address to multiple
254 	 * (probably p2p) interfaces.
255 	 * XXX: we should avoid such a configuration in IPv6...
256 	 */
257 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
258 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
259 			ia_count++;
260 			if (ia_count > 1)
261 				break;
262 		}
263 	}
264 
265 	if (ia_count == 1) {
266 		/*
267 		 * Before deleting, check if a corresponding loopbacked host
268 		 * route surely exists.  With this check, we can avoid to
269 		 * delete an interface direct route whose destination is same
270 		 * as the address being removed.  This can happen when removing
271 		 * a subnet-router anycast address on an interface attahced
272 		 * to a shared medium.
273 		 */
274 		rt = rtalloc1(ifa->ifa_addr, 0);
275 		if (rt != NULL && (rt->rt_flags & RTF_HOST) != 0 &&
276 		    (rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
277 			rt->rt_refcnt--;
278 			in6_ifloop_request(RTM_DELETE, ifa);
279 		}
280 	}
281 }
282 
283 int
284 in6_mask2len(mask, lim0)
285 	struct in6_addr *mask;
286 	u_char *lim0;
287 {
288 	int x = 0, y;
289 	u_char *lim = lim0, *p;
290 
291 	/* ignore the scope_id part */
292 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
293 		lim = (u_char *)mask + sizeof(*mask);
294 	for (p = (u_char *)mask; p < lim; x++, p++) {
295 		if (*p != 0xff)
296 			break;
297 	}
298 	y = 0;
299 	if (p < lim) {
300 		for (y = 0; y < 8; y++) {
301 			if ((*p & (0x80 >> y)) == 0)
302 				break;
303 		}
304 	}
305 
306 	/*
307 	 * when the limit pointer is given, do a stricter check on the
308 	 * remaining bits.
309 	 */
310 	if (p < lim) {
311 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
312 			return (-1);
313 		for (p = p + 1; p < lim; p++)
314 			if (*p != 0)
315 				return (-1);
316 	}
317 
318 	return x * 8 + y;
319 }
320 
321 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
322 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
323 
324 int
325 in6_control(so, cmd, data, ifp, l)
326 	struct	socket *so;
327 	u_long cmd;
328 	caddr_t	data;
329 	struct ifnet *ifp;
330 	struct lwp *l;
331 {
332 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
333 	struct	in6_ifaddr *ia = NULL;
334 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
335 	struct sockaddr_in6 *sa6;
336 	int error, privileged;
337 
338 	privileged = 0;
339 	if (l && !kauth_authorize_generic(l->l_cred,
340 	    KAUTH_GENERIC_ISSUSER, &l->l_acflag))
341 		privileged++;
342 
343 	switch (cmd) {
344 	/*
345 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
346 	 */
347 	case SIOCSIFADDR:
348 	case SIOCSIFDSTADDR:
349 #ifdef SIOCSIFCONF_X26
350 	case SIOCSIFCONF_X25:
351 #endif
352 		return EOPNOTSUPP;
353 	case SIOCGETSGCNT_IN6:
354 	case SIOCGETMIFCNT_IN6:
355 		return (mrt6_ioctl(cmd, data));
356 	}
357 
358 	if (ifp == NULL)
359 		return (EOPNOTSUPP);
360 
361 	switch (cmd) {
362 	case SIOCSNDFLUSH_IN6:
363 	case SIOCSPFXFLUSH_IN6:
364 	case SIOCSRTRFLUSH_IN6:
365 	case SIOCSDEFIFACE_IN6:
366 	case SIOCSIFINFO_FLAGS:
367 	case SIOCSIFINFO_IN6:
368 		if (!privileged)
369 			return (EPERM);
370 		/* FALLTHROUGH */
371 	case OSIOCGIFINFO_IN6:
372 	case SIOCGIFINFO_IN6:
373 	case SIOCGDRLST_IN6:
374 	case SIOCGPRLST_IN6:
375 	case SIOCGNBRINFO_IN6:
376 	case SIOCGDEFIFACE_IN6:
377 		return (nd6_ioctl(cmd, data, ifp));
378 	}
379 
380 	switch (cmd) {
381 	case SIOCSIFPREFIX_IN6:
382 	case SIOCDIFPREFIX_IN6:
383 	case SIOCAIFPREFIX_IN6:
384 	case SIOCCIFPREFIX_IN6:
385 	case SIOCSGIFPREFIX_IN6:
386 	case SIOCGIFPREFIX_IN6:
387 		log(LOG_NOTICE,
388 		    "prefix ioctls are now invalidated. "
389 		    "please use ifconfig.\n");
390 		return (EOPNOTSUPP);
391 	}
392 
393 	switch (cmd) {
394 	case SIOCALIFADDR:
395 	case SIOCDLIFADDR:
396 		if (!privileged)
397 			return (EPERM);
398 		/* FALLTHROUGH */
399 	case SIOCGLIFADDR:
400 		return in6_lifaddr_ioctl(so, cmd, data, ifp, l);
401 	}
402 
403 	/*
404 	 * Find address for this interface, if it exists.
405 	 *
406 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
407 	 * only, and used the first interface address as the target of other
408 	 * operations (without checking ifra_addr).  This was because netinet
409 	 * code/API assumed at most 1 interface address per interface.
410 	 * Since IPv6 allows a node to assign multiple addresses
411 	 * on a single interface, we almost always look and check the
412 	 * presence of ifra_addr, and reject invalid ones here.
413 	 * It also decreases duplicated code among SIOC*_IN6 operations.
414 	 */
415 	switch (cmd) {
416 	case SIOCAIFADDR_IN6:
417 	case SIOCSIFPHYADDR_IN6:
418 		sa6 = &ifra->ifra_addr;
419 		break;
420 	case SIOCSIFADDR_IN6:
421 	case SIOCGIFADDR_IN6:
422 	case SIOCSIFDSTADDR_IN6:
423 	case SIOCSIFNETMASK_IN6:
424 	case SIOCGIFDSTADDR_IN6:
425 	case SIOCGIFNETMASK_IN6:
426 	case SIOCDIFADDR_IN6:
427 	case SIOCGIFPSRCADDR_IN6:
428 	case SIOCGIFPDSTADDR_IN6:
429 	case SIOCGIFAFLAG_IN6:
430 	case SIOCSNDFLUSH_IN6:
431 	case SIOCSPFXFLUSH_IN6:
432 	case SIOCSRTRFLUSH_IN6:
433 	case SIOCGIFALIFETIME_IN6:
434 	case SIOCSIFALIFETIME_IN6:
435 	case SIOCGIFSTAT_IN6:
436 	case SIOCGIFSTAT_ICMP6:
437 		sa6 = &ifr->ifr_addr;
438 		break;
439 	default:
440 		sa6 = NULL;
441 		break;
442 	}
443 	if (sa6 && sa6->sin6_family == AF_INET6) {
444 		if (sa6->sin6_scope_id != 0)
445 			error = sa6_embedscope(sa6, 0);
446 		else
447 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
448 		if (error != 0)
449 			return (error);
450 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
451 	} else
452 		ia = NULL;
453 
454 	switch (cmd) {
455 	case SIOCSIFADDR_IN6:
456 	case SIOCSIFDSTADDR_IN6:
457 	case SIOCSIFNETMASK_IN6:
458 		/*
459 		 * Since IPv6 allows a node to assign multiple addresses
460 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
461 		 */
462 		return (EINVAL);
463 
464 	case SIOCDIFADDR_IN6:
465 		/*
466 		 * for IPv4, we look for existing in_ifaddr here to allow
467 		 * "ifconfig if0 delete" to remove the first IPv4 address on
468 		 * the interface.  For IPv6, as the spec allows multiple
469 		 * interface address from the day one, we consider "remove the
470 		 * first one" semantics to be not preferable.
471 		 */
472 		if (ia == NULL)
473 			return (EADDRNOTAVAIL);
474 		/* FALLTHROUGH */
475 	case SIOCAIFADDR_IN6:
476 		/*
477 		 * We always require users to specify a valid IPv6 address for
478 		 * the corresponding operation.
479 		 */
480 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
481 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
482 			return (EAFNOSUPPORT);
483 		if (!privileged)
484 			return (EPERM);
485 
486 		break;
487 
488 	case SIOCGIFADDR_IN6:
489 		/* This interface is basically deprecated. use SIOCGIFCONF. */
490 		/* FALLTHROUGH */
491 	case SIOCGIFAFLAG_IN6:
492 	case SIOCGIFNETMASK_IN6:
493 	case SIOCGIFDSTADDR_IN6:
494 	case SIOCGIFALIFETIME_IN6:
495 		/* must think again about its semantics */
496 		if (ia == NULL)
497 			return (EADDRNOTAVAIL);
498 		break;
499 	case SIOCSIFALIFETIME_IN6:
500 	    {
501 		struct in6_addrlifetime *lt;
502 
503 		if (!privileged)
504 			return (EPERM);
505 		if (ia == NULL)
506 			return (EADDRNOTAVAIL);
507 		/* sanity for overflow - beware unsigned */
508 		lt = &ifr->ifr_ifru.ifru_lifetime;
509 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
510 		 && lt->ia6t_vltime + time_second < time_second) {
511 			return EINVAL;
512 		}
513 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
514 		 && lt->ia6t_pltime + time_second < time_second) {
515 			return EINVAL;
516 		}
517 		break;
518 	    }
519 	}
520 
521 	switch (cmd) {
522 
523 	case SIOCGIFADDR_IN6:
524 		ifr->ifr_addr = ia->ia_addr;
525 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
526 			return (error);
527 		break;
528 
529 	case SIOCGIFDSTADDR_IN6:
530 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
531 			return (EINVAL);
532 		/*
533 		 * XXX: should we check if ifa_dstaddr is NULL and return
534 		 * an error?
535 		 */
536 		ifr->ifr_dstaddr = ia->ia_dstaddr;
537 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
538 			return (error);
539 		break;
540 
541 	case SIOCGIFNETMASK_IN6:
542 		ifr->ifr_addr = ia->ia_prefixmask;
543 		break;
544 
545 	case SIOCGIFAFLAG_IN6:
546 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
547 		break;
548 
549 	case SIOCGIFSTAT_IN6:
550 		if (ifp == NULL)
551 			return EINVAL;
552 		bzero(&ifr->ifr_ifru.ifru_stat,
553 		    sizeof(ifr->ifr_ifru.ifru_stat));
554 		ifr->ifr_ifru.ifru_stat =
555 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
556 		break;
557 
558 	case SIOCGIFSTAT_ICMP6:
559 		if (ifp == NULL)
560 			return EINVAL;
561 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
562 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
563 		ifr->ifr_ifru.ifru_icmp6stat =
564 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
565 		break;
566 
567 	case SIOCGIFALIFETIME_IN6:
568 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
569 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
570 			time_t maxexpire;
571 			struct in6_addrlifetime *retlt =
572 			    &ifr->ifr_ifru.ifru_lifetime;
573 
574 			/*
575 			 * XXX: adjust expiration time assuming time_t is
576 			 * signed.
577 			 */
578 			maxexpire = ((time_t)~0) &
579 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
580 			if (ia->ia6_lifetime.ia6t_vltime <
581 			    maxexpire - ia->ia6_updatetime) {
582 				retlt->ia6t_expire = ia->ia6_updatetime +
583 				    ia->ia6_lifetime.ia6t_vltime;
584 			} else
585 				retlt->ia6t_expire = maxexpire;
586 		}
587 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
588 			time_t maxexpire;
589 			struct in6_addrlifetime *retlt =
590 			    &ifr->ifr_ifru.ifru_lifetime;
591 
592 			/*
593 			 * XXX: adjust expiration time assuming time_t is
594 			 * signed.
595 			 */
596 			maxexpire = ((time_t)~0) &
597 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
598 			if (ia->ia6_lifetime.ia6t_pltime <
599 			    maxexpire - ia->ia6_updatetime) {
600 				retlt->ia6t_preferred = ia->ia6_updatetime +
601 				    ia->ia6_lifetime.ia6t_pltime;
602 			} else
603 				retlt->ia6t_preferred = maxexpire;
604 		}
605 		break;
606 
607 	case SIOCSIFALIFETIME_IN6:
608 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
609 		/* for sanity */
610 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
611 			ia->ia6_lifetime.ia6t_expire =
612 				time_second + ia->ia6_lifetime.ia6t_vltime;
613 		} else
614 			ia->ia6_lifetime.ia6t_expire = 0;
615 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
616 			ia->ia6_lifetime.ia6t_preferred =
617 				time_second + ia->ia6_lifetime.ia6t_pltime;
618 		} else
619 			ia->ia6_lifetime.ia6t_preferred = 0;
620 		break;
621 
622 	case SIOCAIFADDR_IN6:
623 	{
624 		int i;
625 		struct nd_prefixctl pr0;
626 		struct nd_prefix *pr;
627 
628 		/* reject read-only flags */
629 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
630 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
631 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
632 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
633 			return (EINVAL);
634 		}
635 		/*
636 		 * first, make or update the interface address structure,
637 		 * and link it to the list.
638 		 */
639 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
640 			return (error);
641 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
642 		    == NULL) {
643 		    	/*
644 			 * this can happen when the user specify the 0 valid
645 			 * lifetime.
646 			 */
647 			break;
648 		}
649 
650 		/*
651 		 * then, make the prefix on-link on the interface.
652 		 * XXX: we'd rather create the prefix before the address, but
653 		 * we need at least one address to install the corresponding
654 		 * interface route, so we configure the address first.
655 		 */
656 
657 		/*
658 		 * convert mask to prefix length (prefixmask has already
659 		 * been validated in in6_update_ifa().
660 		 */
661 		bzero(&pr0, sizeof(pr0));
662 		pr0.ndpr_ifp = ifp;
663 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
664 		    NULL);
665 		if (pr0.ndpr_plen == 128) {
666 			break;	/* we don't need to install a host route. */
667 		}
668 		pr0.ndpr_prefix = ifra->ifra_addr;
669 		/* apply the mask for safety. */
670 		for (i = 0; i < 4; i++) {
671 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
672 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
673 		}
674 		/*
675 		 * XXX: since we don't have an API to set prefix (not address)
676 		 * lifetimes, we just use the same lifetimes as addresses.
677 		 * The (temporarily) installed lifetimes can be overridden by
678 		 * later advertised RAs (when accept_rtadv is non 0), which is
679 		 * an intended behavior.
680 		 */
681 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
682 		pr0.ndpr_raf_auto =
683 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
684 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
685 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
686 
687 		/* add the prefix if not yet. */
688 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
689 			/*
690 			 * nd6_prelist_add will install the corresponding
691 			 * interface route.
692 			 */
693 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
694 				return (error);
695 			if (pr == NULL) {
696 				log(LOG_ERR, "nd6_prelist_add succeeded but "
697 				    "no prefix\n");
698 				return (EINVAL); /* XXX panic here? */
699 			}
700 		}
701 
702 		/* relate the address to the prefix */
703 		if (ia->ia6_ndpr == NULL) {
704 			ia->ia6_ndpr = pr;
705 			pr->ndpr_refcnt++;
706 
707 			/*
708 			 * If this is the first autoconf address from the
709 			 * prefix, create a temporary address as well
710 			 * (when required).
711 			 */
712 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
713 			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
714 				int e;
715 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
716 					log(LOG_NOTICE, "in6_control: failed "
717 					    "to create a temporary address, "
718 					    "errno=%d\n", e);
719 				}
720 			}
721 		}
722 
723 		/*
724 		 * this might affect the status of autoconfigured addresses,
725 		 * that is, this address might make other addresses detached.
726 		 */
727 		pfxlist_onlink_check();
728 
729 #ifdef PFIL_HOOKS
730 		(void)pfil_run_hooks(&if_pfil, (struct mbuf **)SIOCAIFADDR_IN6,
731 		    ifp, PFIL_IFADDR);
732 #endif
733 
734 		break;
735 	}
736 
737 	case SIOCDIFADDR_IN6:
738 	{
739 		struct nd_prefix *pr;
740 
741 		/*
742 		 * If the address being deleted is the only one that owns
743 		 * the corresponding prefix, expire the prefix as well.
744 		 * XXX: theoretically, we don't have to worry about such
745 		 * relationship, since we separate the address management
746 		 * and the prefix management.  We do this, however, to provide
747 		 * as much backward compatibility as possible in terms of
748 		 * the ioctl operation.
749 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
750 		 */
751 		pr = ia->ia6_ndpr;
752 		in6_purgeaddr(&ia->ia_ifa);
753 		if (pr && pr->ndpr_refcnt == 0)
754 			prelist_remove(pr);
755 #ifdef PFIL_HOOKS
756 		(void)pfil_run_hooks(&if_pfil, (struct mbuf **)SIOCDIFADDR_IN6,
757 		    ifp, PFIL_IFADDR);
758 #endif
759 		break;
760 	}
761 
762 	default:
763 		if (ifp == NULL || ifp->if_ioctl == 0)
764 			return (EOPNOTSUPP);
765 		error = ((*ifp->if_ioctl)(ifp, cmd, data));
766 		return error;
767 	}
768 
769 	return (0);
770 }
771 
772 /*
773  * Update parameters of an IPv6 interface address.
774  * If necessary, a new entry is created and linked into address chains.
775  * This function is separated from in6_control().
776  * XXX: should this be performed under splnet()?
777  */
778 int
779 in6_update_ifa(ifp, ifra, ia, flags)
780 	struct ifnet *ifp;
781 	struct in6_aliasreq *ifra;
782 	struct in6_ifaddr *ia;
783 	int flags;
784 {
785 	int error = 0, hostIsNew = 0, plen = -1;
786 	struct in6_ifaddr *oia;
787 	struct sockaddr_in6 dst6;
788 	struct in6_addrlifetime *lt;
789 	struct in6_multi_mship *imm;
790 	struct in6_multi *in6m_sol;
791 	struct rtentry *rt;
792 	int dad_delay;
793 
794 	in6m_sol = NULL;
795 
796 	/* Validate parameters */
797 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
798 		return (EINVAL);
799 
800 	/*
801 	 * The destination address for a p2p link must have a family
802 	 * of AF_UNSPEC or AF_INET6.
803 	 */
804 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
805 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
806 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
807 		return (EAFNOSUPPORT);
808 	/*
809 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
810 	 * does not carry fields other than sin6_len.
811 	 */
812 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
813 		return (EINVAL);
814 	/*
815 	 * Because the IPv6 address architecture is classless, we require
816 	 * users to specify a (non 0) prefix length (mask) for a new address.
817 	 * We also require the prefix (when specified) mask is valid, and thus
818 	 * reject a non-consecutive mask.
819 	 */
820 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
821 		return (EINVAL);
822 	if (ifra->ifra_prefixmask.sin6_len != 0) {
823 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
824 		    (u_char *)&ifra->ifra_prefixmask +
825 		    ifra->ifra_prefixmask.sin6_len);
826 		if (plen <= 0)
827 			return (EINVAL);
828 	} else {
829 		/*
830 		 * In this case, ia must not be NULL.  We just use its prefix
831 		 * length.
832 		 */
833 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
834 	}
835 	/*
836 	 * If the destination address on a p2p interface is specified,
837 	 * and the address is a scoped one, validate/set the scope
838 	 * zone identifier.
839 	 */
840 	dst6 = ifra->ifra_dstaddr;
841 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
842 	    (dst6.sin6_family == AF_INET6)) {
843 		struct in6_addr in6_tmp;
844 		u_int32_t zoneid;
845 
846 		in6_tmp = dst6.sin6_addr;
847 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
848 			return (EINVAL); /* XXX: should be impossible */
849 
850 		if (dst6.sin6_scope_id != 0) {
851 			if (dst6.sin6_scope_id != zoneid)
852 				return (EINVAL);
853 		} else		/* user omit to specify the ID. */
854 			dst6.sin6_scope_id = zoneid;
855 
856 		/* convert into the internal form */
857 		if (sa6_embedscope(&dst6, 0))
858 			return (EINVAL); /* XXX: should be impossible */
859 	}
860 	/*
861 	 * The destination address can be specified only for a p2p or a
862 	 * loopback interface.  If specified, the corresponding prefix length
863 	 * must be 128.
864 	 */
865 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
866 #ifdef FORCE_P2PPLEN
867 		int i;
868 #endif
869 
870 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
871 			/* XXX: noisy message */
872 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
873 			    "be specified for a p2p or a loopback IF only\n"));
874 			return (EINVAL);
875 		}
876 		if (plen != 128) {
877 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
878 			    "be 128 when dstaddr is specified\n"));
879 #ifdef FORCE_P2PPLEN
880 			/*
881 			 * To be compatible with old configurations,
882 			 * such as ifconfig gif0 inet6 2001::1 2001::2
883 			 * prefixlen 126, we override the specified
884 			 * prefixmask as if the prefix length was 128.
885 			 */
886 			ifra->ifra_prefixmask.sin6_len =
887 			    sizeof(struct sockaddr_in6);
888 			for (i = 0; i < 4; i++)
889 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
890 				    0xffffffff;
891 			plen = 128;
892 #else
893 			return (EINVAL);
894 #endif
895 		}
896 	}
897 	/* lifetime consistency check */
898 	lt = &ifra->ifra_lifetime;
899 	if (lt->ia6t_pltime > lt->ia6t_vltime)
900 		return (EINVAL);
901 	if (lt->ia6t_vltime == 0) {
902 		/*
903 		 * the following log might be noisy, but this is a typical
904 		 * configuration mistake or a tool's bug.
905 		 */
906 		nd6log((LOG_INFO,
907 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
908 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr)));
909 
910 		if (ia == NULL)
911 			return (0); /* there's nothing to do */
912 	}
913 
914 	/*
915 	 * If this is a new address, allocate a new ifaddr and link it
916 	 * into chains.
917 	 */
918 	if (ia == NULL) {
919 		hostIsNew = 1;
920 		/*
921 		 * When in6_update_ifa() is called in a process of a received
922 		 * RA, it is called under an interrupt context.  So, we should
923 		 * call malloc with M_NOWAIT.
924 		 */
925 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
926 		    M_NOWAIT);
927 		if (ia == NULL)
928 			return (ENOBUFS);
929 		bzero((caddr_t)ia, sizeof(*ia));
930 		LIST_INIT(&ia->ia6_memberships);
931 		/* Initialize the address and masks, and put time stamp */
932 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
933 		ia->ia_addr.sin6_family = AF_INET6;
934 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
935 		ia->ia6_createtime = time_second;
936 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
937 			/*
938 			 * XXX: some functions expect that ifa_dstaddr is not
939 			 * NULL for p2p interfaces.
940 			 */
941 			ia->ia_ifa.ifa_dstaddr =
942 			    (struct sockaddr *)&ia->ia_dstaddr;
943 		} else {
944 			ia->ia_ifa.ifa_dstaddr = NULL;
945 		}
946 		ia->ia_ifa.ifa_netmask =
947 		    (struct sockaddr *)&ia->ia_prefixmask;
948 
949 		ia->ia_ifp = ifp;
950 		if ((oia = in6_ifaddr) != NULL) {
951 			for ( ; oia->ia_next; oia = oia->ia_next)
952 				continue;
953 			oia->ia_next = ia;
954 		} else
955 			in6_ifaddr = ia;
956 		/* gain a refcnt for the link from in6_ifaddr */
957 		IFAREF(&ia->ia_ifa);
958 
959 		TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa,
960 				  ifa_list);
961 		/* gain another refcnt for the link from if_addrlist */
962 		IFAREF(&ia->ia_ifa);
963 	}
964 
965 	/* update timestamp */
966 	ia->ia6_updatetime = time_second;
967 
968 	/* set prefix mask */
969 	if (ifra->ifra_prefixmask.sin6_len) {
970 		/*
971 		 * We prohibit changing the prefix length of an existing
972 		 * address, because
973 		 * + such an operation should be rare in IPv6, and
974 		 * + the operation would confuse prefix management.
975 		 */
976 		if (ia->ia_prefixmask.sin6_len &&
977 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
978 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
979 			    " existing (%s) address should not be changed\n",
980 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
981 			error = EINVAL;
982 			goto unlink;
983 		}
984 		ia->ia_prefixmask = ifra->ifra_prefixmask;
985 	}
986 
987 	/*
988 	 * If a new destination address is specified, scrub the old one and
989 	 * install the new destination.  Note that the interface must be
990 	 * p2p or loopback (see the check above.)
991 	 */
992 	if (dst6.sin6_family == AF_INET6 &&
993 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
994 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
995 		    rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0) {
996 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
997 			    "a route to the old destination: %s\n",
998 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
999 			/* proceed anyway... */
1000 		} else
1001 			ia->ia_flags &= ~IFA_ROUTE;
1002 		ia->ia_dstaddr = dst6;
1003 	}
1004 
1005 	/*
1006 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1007 	 * to see if the address is deprecated or invalidated, but initialize
1008 	 * these members for applications.
1009 	 */
1010 	ia->ia6_lifetime = ifra->ifra_lifetime;
1011 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1012 		ia->ia6_lifetime.ia6t_expire =
1013 		    time_second + ia->ia6_lifetime.ia6t_vltime;
1014 	} else
1015 		ia->ia6_lifetime.ia6t_expire = 0;
1016 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1017 		ia->ia6_lifetime.ia6t_preferred =
1018 		    time_second + ia->ia6_lifetime.ia6t_pltime;
1019 	} else
1020 		ia->ia6_lifetime.ia6t_preferred = 0;
1021 
1022 	/* reset the interface and routing table appropriately. */
1023 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1024 		goto unlink;
1025 
1026 	/*
1027 	 * configure address flags.
1028 	 */
1029 	ia->ia6_flags = ifra->ifra_flags;
1030 	/*
1031 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1032 	 * userland, make it deprecated.
1033 	 */
1034 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1035 		ia->ia6_lifetime.ia6t_pltime = 0;
1036 		ia->ia6_lifetime.ia6t_preferred = time_second;
1037 	}
1038 
1039 	/*
1040 	 * Make the address tentative before joining multicast addresses,
1041 	 * so that corresponding MLD responses would not have a tentative
1042 	 * source address.
1043 	 */
1044 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1045 	if (hostIsNew && in6if_do_dad(ifp))
1046 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1047 
1048 	/*
1049 	 * We are done if we have simply modified an existing address.
1050 	 */
1051 	if (!hostIsNew)
1052 		return (error);
1053 
1054 	/*
1055 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1056 	 * not just go to unlink.
1057 	 */
1058 
1059 	/* join necessary multicast groups */
1060 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1061 		struct sockaddr_in6 mltaddr, mltmask;
1062 		struct in6_addr llsol;
1063 
1064 		/* join solicited multicast addr for new host id */
1065 		bzero(&llsol, sizeof(struct in6_addr));
1066 		llsol.s6_addr16[0] = htons(0xff02);
1067 		llsol.s6_addr32[1] = 0;
1068 		llsol.s6_addr32[2] = htonl(1);
1069 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1070 		llsol.s6_addr8[12] = 0xff;
1071 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1072 			/* XXX: should not happen */
1073 			log(LOG_ERR, "in6_update_ifa: "
1074 			    "in6_setscope failed\n");
1075 			goto cleanup;
1076 		}
1077 		dad_delay = 0;
1078 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1079 			/*
1080 			 * We need a random delay for DAD on the address
1081 			 * being configured.  It also means delaying
1082 			 * transmission of the corresponding MLD report to
1083 			 * avoid report collision.
1084 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1085 			 */
1086 			dad_delay = arc4random() %
1087 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1088 		}
1089 
1090 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1091 		/* join solicited multicast addr for new host id */
1092 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
1093 		if (!imm) {
1094 			nd6log((LOG_ERR,
1095 			    "in6_update_ifa: addmulti "
1096 			    "failed for %s on %s (errno=%d)\n",
1097 			    ip6_sprintf(&llsol), if_name(ifp), error));
1098 			goto cleanup;
1099 		}
1100 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1101 		in6m_sol = imm->i6mm_maddr;
1102 
1103 		bzero(&mltmask, sizeof(mltmask));
1104 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1105 		mltmask.sin6_family = AF_INET6;
1106 		mltmask.sin6_addr = in6mask32;
1107 
1108 		/*
1109 		 * join link-local all-nodes address
1110 		 */
1111 		bzero(&mltaddr, sizeof(mltaddr));
1112 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1113 		mltaddr.sin6_family = AF_INET6;
1114 		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1115 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1116 		    0)
1117 			goto cleanup; /* XXX: should not fail */
1118 
1119 		/*
1120 		 * XXX: do we really need this automatic routes?
1121 		 * We should probably reconsider this stuff.  Most applications
1122 		 * actually do not need the routes, since they usually specify
1123 		 * the outgoing interface.
1124 		 */
1125 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1126 		if (rt) {
1127 			if (memcmp(&mltaddr.sin6_addr,
1128 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1129 			    MLTMASK_LEN)) {
1130 				RTFREE(rt);
1131 				rt = NULL;
1132 			}
1133 		}
1134 		if (!rt) {
1135 			struct rt_addrinfo info;
1136 
1137 			bzero(&info, sizeof(info));
1138 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1139 			info.rti_info[RTAX_GATEWAY] =
1140 			    (struct sockaddr *)&ia->ia_addr;
1141 			info.rti_info[RTAX_NETMASK] =
1142 			    (struct sockaddr *)&mltmask;
1143 			info.rti_info[RTAX_IFA] =
1144 			    (struct sockaddr *)&ia->ia_addr;
1145 			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1146 			info.rti_flags = RTF_UP | RTF_CLONING;
1147 			error = rtrequest1(RTM_ADD, &info, NULL);
1148 			if (error)
1149 				goto cleanup;
1150 		} else {
1151 			RTFREE(rt);
1152 		}
1153 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1154 		if (!imm) {
1155 			nd6log((LOG_WARNING,
1156 			    "in6_update_ifa: addmulti failed for "
1157 			    "%s on %s (errno=%d)\n",
1158 			    ip6_sprintf(&mltaddr.sin6_addr),
1159 			    if_name(ifp), error));
1160 			goto cleanup;
1161 		}
1162 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1163 
1164 		/*
1165 		 * join node information group address
1166 		 */
1167 		dad_delay = 0;
1168 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1169 			/*
1170 			 * The spec doesn't say anything about delay for this
1171 			 * group, but the same logic should apply.
1172 			 */
1173 			dad_delay = arc4random() %
1174 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1175 		}
1176 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) == 0) {
1177 			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1178 			    dad_delay); /* XXX jinmei */
1179 			if (!imm) {
1180 				nd6log((LOG_WARNING, "in6_update_ifa: "
1181 				    "addmulti failed for %s on %s (errno=%d)\n",
1182 				    ip6_sprintf(&mltaddr.sin6_addr),
1183 				    if_name(ifp), error));
1184 				/* XXX not very fatal, go on... */
1185 			} else {
1186 				LIST_INSERT_HEAD(&ia->ia6_memberships,
1187 				    imm, i6mm_chain);
1188 			}
1189 		}
1190 
1191 
1192 		/*
1193 		 * join interface-local all-nodes address.
1194 		 * (ff01::1%ifN, and ff01::%ifN/32)
1195 		 */
1196 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1197 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1198 		    != 0)
1199 			goto cleanup; /* XXX: should not fail */
1200 
1201 		/* XXX: again, do we really need the route? */
1202 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1203 		if (rt) {
1204 			/* 32bit came from "mltmask" */
1205 			if (memcmp(&mltaddr.sin6_addr,
1206 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1207 			    32 / 8)) {
1208 				RTFREE(rt);
1209 				rt = NULL;
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 void
1298 in6_purgeaddr(ifa)
1299 	struct ifaddr *ifa;
1300 {
1301 	struct ifnet *ifp = ifa->ifa_ifp;
1302 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1303 	struct in6_multi_mship *imm;
1304 
1305 	/* stop DAD processing */
1306 	nd6_dad_stop(ifa);
1307 
1308 	/*
1309 	 * delete route to the destination of the address being purged.
1310 	 * The interface must be p2p or loopback in this case.
1311 	 */
1312 	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1313 		int e;
1314 
1315 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1316 		    != 0) {
1317 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1318 			    "a route to the p2p destination: %s on %s, "
1319 			    "errno=%d\n",
1320 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1321 			    e);
1322 			/* proceed anyway... */
1323 		} else
1324 			ia->ia_flags &= ~IFA_ROUTE;
1325 	}
1326 
1327 	/* Remove ownaddr's loopback rtentry, if it exists. */
1328 	in6_ifremloop(&(ia->ia_ifa));
1329 
1330 	/*
1331 	 * leave from multicast groups we have joined for the interface
1332 	 */
1333 	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1334 		LIST_REMOVE(imm, i6mm_chain);
1335 		in6_leavegroup(imm);
1336 	}
1337 
1338 	in6_unlink_ifa(ia, ifp);
1339 }
1340 
1341 static void
1342 in6_unlink_ifa(ia, ifp)
1343 	struct in6_ifaddr *ia;
1344 	struct ifnet *ifp;
1345 {
1346 	struct in6_ifaddr *oia;
1347 	int	s = splnet();
1348 
1349 	TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1350 	/* release a refcnt for the link from if_addrlist */
1351 	IFAFREE(&ia->ia_ifa);
1352 
1353 	oia = ia;
1354 	if (oia == (ia = in6_ifaddr))
1355 		in6_ifaddr = ia->ia_next;
1356 	else {
1357 		while (ia->ia_next && (ia->ia_next != oia))
1358 			ia = ia->ia_next;
1359 		if (ia->ia_next)
1360 			ia->ia_next = oia->ia_next;
1361 		else {
1362 			/* search failed */
1363 			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1364 		}
1365 	}
1366 
1367 	if (oia->ia6_multiaddrs.lh_first != NULL) {
1368 		/*
1369 		 * XXX thorpej@NetBSD.org -- if the interface is going
1370 		 * XXX away, don't save the multicast entries, delete them!
1371 		 */
1372 		if (oia->ia_ifa.ifa_ifp->if_output == if_nulloutput) {
1373 			struct in6_multi *in6m;
1374 
1375 			while ((in6m =
1376 			    LIST_FIRST(&oia->ia6_multiaddrs)) != NULL)
1377 				in6_delmulti(in6m);
1378 		} else
1379 			in6_savemkludge(oia);
1380 	}
1381 
1382 	/*
1383 	 * Release the reference to the base prefix.  There should be a
1384 	 * positive reference.
1385 	 */
1386 	if (oia->ia6_ndpr == NULL) {
1387 		nd6log((LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1388 		    "%p has no prefix\n", oia));
1389 	} else {
1390 		oia->ia6_ndpr->ndpr_refcnt--;
1391 		oia->ia6_ndpr = NULL;
1392 	}
1393 
1394 	/*
1395 	 * Also, if the address being removed is autoconf'ed, call
1396 	 * pfxlist_onlink_check() since the release might affect the status of
1397 	 * other (detached) addresses.
1398 	 */
1399 	if ((oia->ia6_flags & IN6_IFF_AUTOCONF) != 0)
1400 		pfxlist_onlink_check();
1401 
1402 	/*
1403 	 * release another refcnt for the link from in6_ifaddr.
1404 	 * Note that we should decrement the refcnt at least once for all *BSD.
1405 	 */
1406 	IFAFREE(&oia->ia_ifa);
1407 
1408 	splx(s);
1409 }
1410 
1411 void
1412 in6_purgeif(ifp)
1413 	struct ifnet *ifp;
1414 {
1415 	struct ifaddr *ifa, *nifa;
1416 
1417 	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa)
1418 	{
1419 		nifa = TAILQ_NEXT(ifa, ifa_list);
1420 		if (ifa->ifa_addr->sa_family != AF_INET6)
1421 			continue;
1422 		in6_purgeaddr(ifa);
1423 	}
1424 
1425 	in6_ifdetach(ifp);
1426 }
1427 
1428 /*
1429  * SIOC[GAD]LIFADDR.
1430  *	SIOCGLIFADDR: get first address. (?)
1431  *	SIOCGLIFADDR with IFLR_PREFIX:
1432  *		get first address that matches the specified prefix.
1433  *	SIOCALIFADDR: add the specified address.
1434  *	SIOCALIFADDR with IFLR_PREFIX:
1435  *		add the specified prefix, filling hostid part from
1436  *		the first link-local address.  prefixlen must be <= 64.
1437  *	SIOCDLIFADDR: delete the specified address.
1438  *	SIOCDLIFADDR with IFLR_PREFIX:
1439  *		delete the first address that matches the specified prefix.
1440  * return values:
1441  *	EINVAL on invalid parameters
1442  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1443  *	other values may be returned from in6_ioctl()
1444  *
1445  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1446  * this is to accomodate address naming scheme other than RFC2374,
1447  * in the future.
1448  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1449  * address encoding scheme. (see figure on page 8)
1450  */
1451 static int
1452 in6_lifaddr_ioctl(so, cmd, data, ifp, l)
1453 	struct socket *so;
1454 	u_long cmd;
1455 	caddr_t	data;
1456 	struct ifnet *ifp;
1457 	struct lwp *l;
1458 {
1459 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1460 	struct ifaddr *ifa;
1461 	struct sockaddr *sa;
1462 
1463 	/* sanity checks */
1464 	if (!data || !ifp) {
1465 		panic("invalid argument to in6_lifaddr_ioctl");
1466 		/* NOTREACHED */
1467 	}
1468 
1469 	switch (cmd) {
1470 	case SIOCGLIFADDR:
1471 		/* address must be specified on GET with IFLR_PREFIX */
1472 		if ((iflr->flags & IFLR_PREFIX) == 0)
1473 			break;
1474 		/* FALLTHROUGH */
1475 	case SIOCALIFADDR:
1476 	case SIOCDLIFADDR:
1477 		/* address must be specified on ADD and DELETE */
1478 		sa = (struct sockaddr *)&iflr->addr;
1479 		if (sa->sa_family != AF_INET6)
1480 			return EINVAL;
1481 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1482 			return EINVAL;
1483 		/* XXX need improvement */
1484 		sa = (struct sockaddr *)&iflr->dstaddr;
1485 		if (sa->sa_family && sa->sa_family != AF_INET6)
1486 			return EINVAL;
1487 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1488 			return EINVAL;
1489 		break;
1490 	default: /* shouldn't happen */
1491 #if 0
1492 		panic("invalid cmd to in6_lifaddr_ioctl");
1493 		/* NOTREACHED */
1494 #else
1495 		return EOPNOTSUPP;
1496 #endif
1497 	}
1498 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1499 		return EINVAL;
1500 
1501 	switch (cmd) {
1502 	case SIOCALIFADDR:
1503 	    {
1504 		struct in6_aliasreq ifra;
1505 		struct in6_addr *xhostid = NULL;
1506 		int prefixlen;
1507 
1508 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1509 			struct sockaddr_in6 *sin6;
1510 
1511 			/*
1512 			 * xhostid is to fill in the hostid part of the
1513 			 * address.  xhostid points to the first link-local
1514 			 * address attached to the interface.
1515 			 */
1516 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1517 			if (!ifa)
1518 				return EADDRNOTAVAIL;
1519 			xhostid = IFA_IN6(ifa);
1520 
1521 		 	/* prefixlen must be <= 64. */
1522 			if (64 < iflr->prefixlen)
1523 				return EINVAL;
1524 			prefixlen = iflr->prefixlen;
1525 
1526 			/* hostid part must be zero. */
1527 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1528 			if (sin6->sin6_addr.s6_addr32[2] != 0
1529 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1530 				return EINVAL;
1531 			}
1532 		} else
1533 			prefixlen = iflr->prefixlen;
1534 
1535 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1536 		bzero(&ifra, sizeof(ifra));
1537 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1538 
1539 		bcopy(&iflr->addr, &ifra.ifra_addr,
1540 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1541 		if (xhostid) {
1542 			/* fill in hostid part */
1543 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1544 			    xhostid->s6_addr32[2];
1545 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1546 			    xhostid->s6_addr32[3];
1547 		}
1548 
1549 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1550 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1551 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1552 			if (xhostid) {
1553 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1554 				    xhostid->s6_addr32[2];
1555 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1556 				    xhostid->s6_addr32[3];
1557 			}
1558 		}
1559 
1560 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1561 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1562 
1563 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1564 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1565 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1566 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, l);
1567 	    }
1568 	case SIOCGLIFADDR:
1569 	case SIOCDLIFADDR:
1570 	    {
1571 		struct in6_ifaddr *ia;
1572 		struct in6_addr mask, candidate, match;
1573 		struct sockaddr_in6 *sin6;
1574 		int cmp;
1575 
1576 		bzero(&mask, sizeof(mask));
1577 		if (iflr->flags & IFLR_PREFIX) {
1578 			/* lookup a prefix rather than address. */
1579 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1580 
1581 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1582 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1583 			match.s6_addr32[0] &= mask.s6_addr32[0];
1584 			match.s6_addr32[1] &= mask.s6_addr32[1];
1585 			match.s6_addr32[2] &= mask.s6_addr32[2];
1586 			match.s6_addr32[3] &= mask.s6_addr32[3];
1587 
1588 			/* if you set extra bits, that's wrong */
1589 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1590 				return EINVAL;
1591 
1592 			cmp = 1;
1593 		} else {
1594 			if (cmd == SIOCGLIFADDR) {
1595 				/* on getting an address, take the 1st match */
1596 				cmp = 0;	/* XXX */
1597 			} else {
1598 				/* on deleting an address, do exact match */
1599 				in6_prefixlen2mask(&mask, 128);
1600 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1601 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1602 
1603 				cmp = 1;
1604 			}
1605 		}
1606 
1607 		for (ifa = ifp->if_addrlist.tqh_first;
1608 		     ifa;
1609 		     ifa = ifa->ifa_list.tqe_next)
1610 		{
1611 			if (ifa->ifa_addr->sa_family != AF_INET6)
1612 				continue;
1613 			if (!cmp)
1614 				break;
1615 
1616 			/*
1617 			 * XXX: this is adhoc, but is necessary to allow
1618 			 * a user to specify fe80::/64 (not /10) for a
1619 			 * link-local address.
1620 			 */
1621 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1622 			in6_clearscope(&candidate);
1623 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1624 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1625 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1626 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1627 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1628 				break;
1629 		}
1630 		if (!ifa)
1631 			return EADDRNOTAVAIL;
1632 		ia = ifa2ia6(ifa);
1633 
1634 		if (cmd == SIOCGLIFADDR) {
1635 			int error;
1636 
1637 			/* fill in the if_laddrreq structure */
1638 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1639 			error = sa6_recoverscope(
1640 			    (struct sockaddr_in6 *)&iflr->addr);
1641 			if (error != 0)
1642 				return (error);
1643 
1644 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1645 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1646 				    ia->ia_dstaddr.sin6_len);
1647 				error = sa6_recoverscope(
1648 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1649 				if (error != 0)
1650 					return (error);
1651 			} else
1652 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1653 
1654 			iflr->prefixlen =
1655 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1656 
1657 			iflr->flags = ia->ia6_flags;	/* XXX */
1658 
1659 			return 0;
1660 		} else {
1661 			struct in6_aliasreq ifra;
1662 
1663 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1664 			bzero(&ifra, sizeof(ifra));
1665 			bcopy(iflr->iflr_name, ifra.ifra_name,
1666 			    sizeof(ifra.ifra_name));
1667 
1668 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1669 			    ia->ia_addr.sin6_len);
1670 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1671 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1672 				    ia->ia_dstaddr.sin6_len);
1673 			} else {
1674 				bzero(&ifra.ifra_dstaddr,
1675 				    sizeof(ifra.ifra_dstaddr));
1676 			}
1677 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1678 			    ia->ia_prefixmask.sin6_len);
1679 
1680 			ifra.ifra_flags = ia->ia6_flags;
1681 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1682 			    ifp, l);
1683 		}
1684 	    }
1685 	}
1686 
1687 	return EOPNOTSUPP;	/* just for safety */
1688 }
1689 
1690 /*
1691  * Initialize an interface's internet6 address
1692  * and routing table entry.
1693  */
1694 static int
1695 in6_ifinit(ifp, ia, sin6, newhost)
1696 	struct ifnet *ifp;
1697 	struct in6_ifaddr *ia;
1698 	struct sockaddr_in6 *sin6;
1699 	int newhost;
1700 {
1701 	int	error = 0, plen, ifacount = 0;
1702 	int	s = splnet();
1703 	struct ifaddr *ifa;
1704 
1705 	/*
1706 	 * Give the interface a chance to initialize
1707 	 * if this is its first address,
1708 	 * and to validate the address if necessary.
1709 	 */
1710 	for (ifa = ifp->if_addrlist.tqh_first; ifa;
1711 	     ifa = ifa->ifa_list.tqe_next)
1712 	{
1713 		if (ifa->ifa_addr == NULL)
1714 			continue;	/* just for safety */
1715 		if (ifa->ifa_addr->sa_family != AF_INET6)
1716 			continue;
1717 		ifacount++;
1718 	}
1719 
1720 	ia->ia_addr = *sin6;
1721 
1722 	if (ifacount <= 1 && ifp->if_ioctl &&
1723 	    (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia))) {
1724 		splx(s);
1725 		return (error);
1726 	}
1727 	splx(s);
1728 
1729 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1730 
1731 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1732 
1733 	/*
1734 	 * Special case:
1735 	 * If the destination address is specified for a point-to-point
1736 	 * interface, install a route to the destination as an interface
1737 	 * direct route.
1738 	 */
1739 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1740 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1741 		if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1742 				    RTF_UP | RTF_HOST)) != 0)
1743 			return (error);
1744 		ia->ia_flags |= IFA_ROUTE;
1745 	}
1746 
1747 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1748 	if (newhost) {
1749 		/* set the rtrequest function to create llinfo */
1750 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1751 		in6_ifaddloop(&(ia->ia_ifa));
1752 	}
1753 
1754 	if (ifp->if_flags & IFF_MULTICAST)
1755 		in6_restoremkludge(ia, ifp);
1756 
1757 	return (error);
1758 }
1759 
1760 /*
1761  * Find an IPv6 interface link-local address specific to an interface.
1762  */
1763 struct in6_ifaddr *
1764 in6ifa_ifpforlinklocal(ifp, ignoreflags)
1765 	struct ifnet *ifp;
1766 	int ignoreflags;
1767 {
1768 	struct ifaddr *ifa;
1769 
1770 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
1771 	{
1772 		if (ifa->ifa_addr == NULL)
1773 			continue;	/* just for safety */
1774 		if (ifa->ifa_addr->sa_family != AF_INET6)
1775 			continue;
1776 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1777 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1778 			     ignoreflags) != 0)
1779 				continue;
1780 			break;
1781 		}
1782 	}
1783 
1784 	return ((struct in6_ifaddr *)ifa);
1785 }
1786 
1787 
1788 /*
1789  * find the internet address corresponding to a given interface and address.
1790  */
1791 struct in6_ifaddr *
1792 in6ifa_ifpwithaddr(ifp, addr)
1793 	struct ifnet *ifp;
1794 	struct in6_addr *addr;
1795 {
1796 	struct ifaddr *ifa;
1797 
1798 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
1799 	{
1800 		if (ifa->ifa_addr == NULL)
1801 			continue;	/* just for safety */
1802 		if (ifa->ifa_addr->sa_family != AF_INET6)
1803 			continue;
1804 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1805 			break;
1806 	}
1807 
1808 	return ((struct in6_ifaddr *)ifa);
1809 }
1810 
1811 /*
1812  * Convert IP6 address to printable (loggable) representation.
1813  */
1814 static int ip6round = 0;
1815 char *
1816 ip6_sprintf(addr)
1817 	const struct in6_addr *addr;
1818 {
1819 	static char ip6buf[8][48];
1820 	int i;
1821 	char *cp;
1822 	const u_int16_t *a = (const u_int16_t *)addr;
1823 	const u_int8_t *d;
1824 	int dcolon = 0;
1825 
1826 	ip6round = (ip6round + 1) & 7;
1827 	cp = ip6buf[ip6round];
1828 
1829 	for (i = 0; i < 8; i++) {
1830 		if (dcolon == 1) {
1831 			if (*a == 0) {
1832 				if (i == 7)
1833 					*cp++ = ':';
1834 				a++;
1835 				continue;
1836 			} else
1837 				dcolon = 2;
1838 		}
1839 		if (*a == 0) {
1840 			if (dcolon == 0 && *(a + 1) == 0) {
1841 				if (i == 0)
1842 					*cp++ = ':';
1843 				*cp++ = ':';
1844 				dcolon = 1;
1845 			} else {
1846 				*cp++ = '0';
1847 				*cp++ = ':';
1848 			}
1849 			a++;
1850 			continue;
1851 		}
1852 		d = (const u_char *)a;
1853 		*cp++ = hexdigits[*d >> 4];
1854 		*cp++ = hexdigits[*d++ & 0xf];
1855 		*cp++ = hexdigits[*d >> 4];
1856 		*cp++ = hexdigits[*d & 0xf];
1857 		*cp++ = ':';
1858 		a++;
1859 	}
1860 	*--cp = 0;
1861 	return (ip6buf[ip6round]);
1862 }
1863 
1864 /*
1865  * Determine if an address is on a local network.
1866  */
1867 int
1868 in6_localaddr(in6)
1869 	struct in6_addr *in6;
1870 {
1871 	struct in6_ifaddr *ia;
1872 
1873 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1874 		return (1);
1875 
1876 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1877 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1878 					      &ia->ia_prefixmask.sin6_addr))
1879 			return (1);
1880 
1881 	return (0);
1882 }
1883 
1884 int
1885 in6_is_addr_deprecated(sa6)
1886 	struct sockaddr_in6 *sa6;
1887 {
1888 	struct in6_ifaddr *ia;
1889 
1890 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1891 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1892 		    &sa6->sin6_addr) &&
1893 #ifdef SCOPEDROUTING
1894 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1895 #endif
1896 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1897 			return (1); /* true */
1898 
1899 		/* XXX: do we still have to go thru the rest of the list? */
1900 	}
1901 
1902 	return (0);		/* false */
1903 }
1904 
1905 /*
1906  * return length of part which dst and src are equal
1907  * hard coding...
1908  */
1909 int
1910 in6_matchlen(src, dst)
1911 struct in6_addr *src, *dst;
1912 {
1913 	int match = 0;
1914 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1915 	u_char *lim = s + 16, r;
1916 
1917 	while (s < lim)
1918 		if ((r = (*d++ ^ *s++)) != 0) {
1919 			while (r < 128) {
1920 				match++;
1921 				r <<= 1;
1922 			}
1923 			break;
1924 		} else
1925 			match += 8;
1926 	return match;
1927 }
1928 
1929 /* XXX: to be scope conscious */
1930 int
1931 in6_are_prefix_equal(p1, p2, len)
1932 	struct in6_addr *p1, *p2;
1933 	int len;
1934 {
1935 	int bytelen, bitlen;
1936 
1937 	/* sanity check */
1938 	if (len < 0 || len > 128) {
1939 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1940 		    len);
1941 		return (0);
1942 	}
1943 
1944 	bytelen = len / 8;
1945 	bitlen = len % 8;
1946 
1947 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1948 		return (0);
1949 	if (bitlen != 0 &&
1950 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1951 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1952 		return (0);
1953 
1954 	return (1);
1955 }
1956 
1957 void
1958 in6_prefixlen2mask(maskp, len)
1959 	struct in6_addr *maskp;
1960 	int len;
1961 {
1962 	static const u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1963 	int bytelen, bitlen, i;
1964 
1965 	/* sanity check */
1966 	if (len < 0 || len > 128) {
1967 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1968 		    len);
1969 		return;
1970 	}
1971 
1972 	bzero(maskp, sizeof(*maskp));
1973 	bytelen = len / 8;
1974 	bitlen = len % 8;
1975 	for (i = 0; i < bytelen; i++)
1976 		maskp->s6_addr[i] = 0xff;
1977 	if (bitlen)
1978 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1979 }
1980 
1981 /*
1982  * return the best address out of the same scope. if no address was
1983  * found, return the first valid address from designated IF.
1984  */
1985 struct in6_ifaddr *
1986 in6_ifawithifp(ifp, dst)
1987 	struct ifnet *ifp;
1988 	struct in6_addr *dst;
1989 {
1990 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1991 	struct ifaddr *ifa;
1992 	struct in6_ifaddr *besta = 0;
1993 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1994 
1995 	dep[0] = dep[1] = NULL;
1996 
1997 	/*
1998 	 * We first look for addresses in the same scope.
1999 	 * If there is one, return it.
2000 	 * If two or more, return one which matches the dst longest.
2001 	 * If none, return one of global addresses assigned other ifs.
2002 	 */
2003 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
2004 	{
2005 		if (ifa->ifa_addr->sa_family != AF_INET6)
2006 			continue;
2007 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2008 			continue; /* XXX: is there any case to allow anycast? */
2009 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2010 			continue; /* don't use this interface */
2011 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2012 			continue;
2013 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2014 			if (ip6_use_deprecated)
2015 				dep[0] = (struct in6_ifaddr *)ifa;
2016 			continue;
2017 		}
2018 
2019 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2020 			/*
2021 			 * call in6_matchlen() as few as possible
2022 			 */
2023 			if (besta) {
2024 				if (blen == -1)
2025 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2026 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2027 				if (tlen > blen) {
2028 					blen = tlen;
2029 					besta = (struct in6_ifaddr *)ifa;
2030 				}
2031 			} else
2032 				besta = (struct in6_ifaddr *)ifa;
2033 		}
2034 	}
2035 	if (besta)
2036 		return (besta);
2037 
2038 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
2039 	{
2040 		if (ifa->ifa_addr->sa_family != AF_INET6)
2041 			continue;
2042 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2043 			continue; /* XXX: is there any case to allow anycast? */
2044 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2045 			continue; /* don't use this interface */
2046 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2047 			continue;
2048 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2049 			if (ip6_use_deprecated)
2050 				dep[1] = (struct in6_ifaddr *)ifa;
2051 			continue;
2052 		}
2053 
2054 		return (struct in6_ifaddr *)ifa;
2055 	}
2056 
2057 	/* use the last-resort values, that are, deprecated addresses */
2058 	if (dep[0])
2059 		return dep[0];
2060 	if (dep[1])
2061 		return dep[1];
2062 
2063 	return NULL;
2064 }
2065 
2066 /*
2067  * perform DAD when interface becomes IFF_UP.
2068  */
2069 void
2070 in6_if_up(ifp)
2071 	struct ifnet *ifp;
2072 {
2073 	struct ifaddr *ifa;
2074 	struct in6_ifaddr *ia;
2075 
2076 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
2077 	{
2078 		if (ifa->ifa_addr->sa_family != AF_INET6)
2079 			continue;
2080 		ia = (struct in6_ifaddr *)ifa;
2081 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2082 			/*
2083 			 * The TENTATIVE flag was likely set by hand
2084 			 * beforehand, implicitly indicating the need for DAD.
2085 			 * We may be able to skip the random delay in this
2086 			 * case, but we impose delays just in case.
2087 			 */
2088 			nd6_dad_start(ifa,
2089 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2090 		}
2091 	}
2092 
2093 	/*
2094 	 * special cases, like 6to4, are handled in in6_ifattach
2095 	 */
2096 	in6_ifattach(ifp, NULL);
2097 }
2098 
2099 int
2100 in6if_do_dad(ifp)
2101 	struct ifnet *ifp;
2102 {
2103 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2104 		return (0);
2105 
2106 	switch (ifp->if_type) {
2107 	case IFT_FAITH:
2108 		/*
2109 		 * These interfaces do not have the IFF_LOOPBACK flag,
2110 		 * but loop packets back.  We do not have to do DAD on such
2111 		 * interfaces.  We should even omit it, because loop-backed
2112 		 * NS would confuse the DAD procedure.
2113 		 */
2114 		return (0);
2115 	default:
2116 		/*
2117 		 * Our DAD routine requires the interface up and running.
2118 		 * However, some interfaces can be up before the RUNNING
2119 		 * status.  Additionaly, users may try to assign addresses
2120 		 * before the interface becomes up (or running).
2121 		 * We simply skip DAD in such a case as a work around.
2122 		 * XXX: we should rather mark "tentative" on such addresses,
2123 		 * and do DAD after the interface becomes ready.
2124 		 */
2125 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2126 		    (IFF_UP|IFF_RUNNING))
2127 			return (0);
2128 
2129 		return (1);
2130 	}
2131 }
2132 
2133 /*
2134  * Calculate max IPv6 MTU through all the interfaces and store it
2135  * to in6_maxmtu.
2136  */
2137 void
2138 in6_setmaxmtu()
2139 {
2140 	unsigned long maxmtu = 0;
2141 	struct ifnet *ifp;
2142 
2143 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2144 	{
2145 		/* this function can be called during ifnet initialization */
2146 		if (!ifp->if_afdata[AF_INET6])
2147 			continue;
2148 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2149 		    IN6_LINKMTU(ifp) > maxmtu)
2150 			maxmtu = IN6_LINKMTU(ifp);
2151 	}
2152 	if (maxmtu)	     /* update only when maxmtu is positive */
2153 		in6_maxmtu = maxmtu;
2154 }
2155 
2156 /*
2157  * Provide the length of interface identifiers to be used for the link attached
2158  * to the given interface.  The length should be defined in "IPv6 over
2159  * xxx-link" document.  Note that address architecture might also define
2160  * the length for a particular set of address prefixes, regardless of the
2161  * link type.  As clarified in rfc2462bis, those two definitions should be
2162  * consistent, and those really are as of August 2004.
2163  */
2164 int
2165 in6_if2idlen(ifp)
2166 	struct ifnet *ifp;
2167 {
2168 	switch (ifp->if_type) {
2169 	case IFT_ETHER:		/* RFC2464 */
2170 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2171 	case IFT_L2VLAN:	/* ditto */
2172 	case IFT_IEEE80211:	/* ditto */
2173 	case IFT_FDDI:		/* RFC2467 */
2174 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2175 	case IFT_PPP:		/* RFC2472 */
2176 	case IFT_ARCNET:	/* RFC2497 */
2177 	case IFT_FRELAY:	/* RFC2590 */
2178 	case IFT_IEEE1394:	/* RFC3146 */
2179 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2180 	case IFT_LOOP:		/* XXX: is this really correct? */
2181 		return 64;
2182 	default:
2183 		/*
2184 		 * Unknown link type:
2185 		 * It might be controversial to use the today's common constant
2186 		 * of 64 for these cases unconditionally.  For full compliance,
2187 		 * we should return an error in this case.  On the other hand,
2188 		 * if we simply miss the standard for the link type or a new
2189 		 * standard is defined for a new link type, the IFID length
2190 		 * is very likely to be the common constant.  As a compromise,
2191 		 * we always use the constant, but make an explicit notice
2192 		 * indicating the "unknown" case.
2193 		 */
2194 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2195 		return 64;
2196 	}
2197 }
2198 
2199 void *
2200 in6_domifattach(ifp)
2201 	struct ifnet *ifp;
2202 {
2203 	struct in6_ifextra *ext;
2204 
2205 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2206 	bzero(ext, sizeof(*ext));
2207 
2208 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2209 	    M_IFADDR, M_WAITOK);
2210 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2211 
2212 	ext->icmp6_ifstat =
2213 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2214 	    M_IFADDR, M_WAITOK);
2215 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2216 
2217 	ext->nd_ifinfo = nd6_ifattach(ifp);
2218 	ext->scope6_id = scope6_ifattach(ifp);
2219 	return ext;
2220 }
2221 
2222 void
2223 in6_domifdetach(struct ifnet *ifp __unused, void *aux)
2224 {
2225 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2226 
2227 	nd6_ifdetach(ext->nd_ifinfo);
2228 	free(ext->in6_ifstat, M_IFADDR);
2229 	free(ext->icmp6_ifstat, M_IFADDR);
2230 	scope6_ifdetach(ext->scope6_id);
2231 	free(ext, M_IFADDR);
2232 }
2233