xref: /freebsd/sys/netinet6/in6.c (revision 3157ba21)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)in.c	8.2 (Berkeley) 11/15/93
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_compat.h"
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 
70 #include <sys/param.h>
71 #include <sys/errno.h>
72 #include <sys/jail.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/priv.h>
79 #include <sys/proc.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 
84 #include <net/if.h>
85 #include <net/if_var.h>
86 #include <net/if_types.h>
87 #include <net/route.h>
88 #include <net/if_dl.h>
89 #include <net/vnet.h>
90 
91 #include <netinet/in.h>
92 #include <netinet/in_var.h>
93 #include <net/if_llatbl.h>
94 #include <netinet/if_ether.h>
95 #include <netinet/in_systm.h>
96 #include <netinet/ip.h>
97 #include <netinet/in_pcb.h>
98 
99 #include <netinet/ip6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/nd6.h>
102 #include <netinet6/mld6_var.h>
103 #include <netinet6/ip6_mroute.h>
104 #include <netinet6/in6_ifattach.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/in6_pcb.h>
107 
108 /*
109  * Definitions of some costant IP6 addresses.
110  */
111 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
112 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
113 const struct in6_addr in6addr_nodelocal_allnodes =
114 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
115 const struct in6_addr in6addr_linklocal_allnodes =
116 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
117 const struct in6_addr in6addr_linklocal_allrouters =
118 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
119 const struct in6_addr in6addr_linklocal_allv2routers =
120 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_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 =
129 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
130 
131 static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
132 	struct ifnet *, struct thread *));
133 static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
134 	struct sockaddr_in6 *, int));
135 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
136 
137 int	(*faithprefix_p)(struct in6_addr *);
138 
139 
140 
141 int
142 in6_mask2len(struct in6_addr *mask, u_char *lim0)
143 {
144 	int x = 0, y;
145 	u_char *lim = lim0, *p;
146 
147 	/* ignore the scope_id part */
148 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
149 		lim = (u_char *)mask + sizeof(*mask);
150 	for (p = (u_char *)mask; p < lim; x++, p++) {
151 		if (*p != 0xff)
152 			break;
153 	}
154 	y = 0;
155 	if (p < lim) {
156 		for (y = 0; y < 8; y++) {
157 			if ((*p & (0x80 >> y)) == 0)
158 				break;
159 		}
160 	}
161 
162 	/*
163 	 * when the limit pointer is given, do a stricter check on the
164 	 * remaining bits.
165 	 */
166 	if (p < lim) {
167 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
168 			return (-1);
169 		for (p = p + 1; p < lim; p++)
170 			if (*p != 0)
171 				return (-1);
172 	}
173 
174 	return x * 8 + y;
175 }
176 
177 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
178 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
179 
180 #ifdef COMPAT_FREEBSD32
181 struct in6_ndifreq32 {
182         char ifname[IFNAMSIZ];
183         uint32_t ifindex;
184 };
185 #define	SIOCGDEFIFACE32_IN6     _IOWR('i', 86, struct in6_ndifreq32)
186 #endif
187 
188 int
189 in6_control(struct socket *so, u_long cmd, caddr_t data,
190     struct ifnet *ifp, struct thread *td)
191 {
192 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
193 	struct	in6_ifaddr *ia = NULL;
194 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
195 	struct sockaddr_in6 *sa6;
196 	int error;
197 
198 	switch (cmd) {
199 	case SIOCGETSGCNT_IN6:
200 	case SIOCGETMIFCNT_IN6:
201 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
202 	}
203 
204 	switch(cmd) {
205 	case SIOCAADDRCTL_POLICY:
206 	case SIOCDADDRCTL_POLICY:
207 		if (td != NULL) {
208 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
209 			if (error)
210 				return (error);
211 		}
212 		return (in6_src_ioctl(cmd, data));
213 	}
214 
215 	if (ifp == NULL)
216 		return (EOPNOTSUPP);
217 
218 	switch (cmd) {
219 	case SIOCSNDFLUSH_IN6:
220 	case SIOCSPFXFLUSH_IN6:
221 	case SIOCSRTRFLUSH_IN6:
222 	case SIOCSDEFIFACE_IN6:
223 	case SIOCSIFINFO_FLAGS:
224 	case SIOCSIFINFO_IN6:
225 		if (td != NULL) {
226 			error = priv_check(td, PRIV_NETINET_ND6);
227 			if (error)
228 				return (error);
229 		}
230 		/* FALLTHROUGH */
231 	case OSIOCGIFINFO_IN6:
232 	case SIOCGIFINFO_IN6:
233 	case SIOCGDRLST_IN6:
234 	case SIOCGPRLST_IN6:
235 	case SIOCGNBRINFO_IN6:
236 	case SIOCGDEFIFACE_IN6:
237 		return (nd6_ioctl(cmd, data, ifp));
238 
239 #ifdef COMPAT_FREEBSD32
240 	case SIOCGDEFIFACE32_IN6:
241 		{
242 			struct in6_ndifreq ndif;
243 			struct in6_ndifreq32 *ndif32;
244 
245 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
246 			    ifp);
247 			if (error)
248 				return (error);
249 			ndif32 = (struct in6_ndifreq32 *)data;
250 			ndif32->ifindex = ndif.ifindex;
251 			return (0);
252 		}
253 #endif
254 	}
255 
256 	switch (cmd) {
257 	case SIOCSIFPREFIX_IN6:
258 	case SIOCDIFPREFIX_IN6:
259 	case SIOCAIFPREFIX_IN6:
260 	case SIOCCIFPREFIX_IN6:
261 	case SIOCSGIFPREFIX_IN6:
262 	case SIOCGIFPREFIX_IN6:
263 		log(LOG_NOTICE,
264 		    "prefix ioctls are now invalidated. "
265 		    "please use ifconfig.\n");
266 		return (EOPNOTSUPP);
267 	}
268 
269 	switch (cmd) {
270 	case SIOCSSCOPE6:
271 		if (td != NULL) {
272 			error = priv_check(td, PRIV_NETINET_SCOPE6);
273 			if (error)
274 				return (error);
275 		}
276 		return (scope6_set(ifp,
277 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
278 	case SIOCGSCOPE6:
279 		return (scope6_get(ifp,
280 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
281 	case SIOCGSCOPE6DEF:
282 		return (scope6_get_default((struct scope6_id *)
283 		    ifr->ifr_ifru.ifru_scope_id));
284 	}
285 
286 	switch (cmd) {
287 	case SIOCALIFADDR:
288 		if (td != NULL) {
289 			error = priv_check(td, PRIV_NET_ADDIFADDR);
290 			if (error)
291 				return (error);
292 		}
293 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
294 
295 	case SIOCDLIFADDR:
296 		if (td != NULL) {
297 			error = priv_check(td, PRIV_NET_DELIFADDR);
298 			if (error)
299 				return (error);
300 		}
301 		/* FALLTHROUGH */
302 	case SIOCGLIFADDR:
303 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
304 	}
305 
306 	/*
307 	 * Find address for this interface, if it exists.
308 	 *
309 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
310 	 * only, and used the first interface address as the target of other
311 	 * operations (without checking ifra_addr).  This was because netinet
312 	 * code/API assumed at most 1 interface address per interface.
313 	 * Since IPv6 allows a node to assign multiple addresses
314 	 * on a single interface, we almost always look and check the
315 	 * presence of ifra_addr, and reject invalid ones here.
316 	 * It also decreases duplicated code among SIOC*_IN6 operations.
317 	 */
318 	switch (cmd) {
319 	case SIOCAIFADDR_IN6:
320 	case SIOCSIFPHYADDR_IN6:
321 		sa6 = &ifra->ifra_addr;
322 		break;
323 	case SIOCSIFADDR_IN6:
324 	case SIOCGIFADDR_IN6:
325 	case SIOCSIFDSTADDR_IN6:
326 	case SIOCSIFNETMASK_IN6:
327 	case SIOCGIFDSTADDR_IN6:
328 	case SIOCGIFNETMASK_IN6:
329 	case SIOCDIFADDR_IN6:
330 	case SIOCGIFPSRCADDR_IN6:
331 	case SIOCGIFPDSTADDR_IN6:
332 	case SIOCGIFAFLAG_IN6:
333 	case SIOCSNDFLUSH_IN6:
334 	case SIOCSPFXFLUSH_IN6:
335 	case SIOCSRTRFLUSH_IN6:
336 	case SIOCGIFALIFETIME_IN6:
337 	case SIOCSIFALIFETIME_IN6:
338 	case SIOCGIFSTAT_IN6:
339 	case SIOCGIFSTAT_ICMP6:
340 		sa6 = &ifr->ifr_addr;
341 		break;
342 	default:
343 		sa6 = NULL;
344 		break;
345 	}
346 	if (sa6 && sa6->sin6_family == AF_INET6) {
347 		if (sa6->sin6_scope_id != 0)
348 			error = sa6_embedscope(sa6, 0);
349 		else
350 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
351 		if (error != 0)
352 			return (error);
353 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
354 		    &sa6->sin6_addr)) != 0)
355 			return (error);
356 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
357 	} else
358 		ia = NULL;
359 
360 	switch (cmd) {
361 	case SIOCSIFADDR_IN6:
362 	case SIOCSIFDSTADDR_IN6:
363 	case SIOCSIFNETMASK_IN6:
364 		/*
365 		 * Since IPv6 allows a node to assign multiple addresses
366 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
367 		 */
368 		/* we decided to obsolete this command (20000704) */
369 		error = EINVAL;
370 		goto out;
371 
372 	case SIOCDIFADDR_IN6:
373 		/*
374 		 * for IPv4, we look for existing in_ifaddr here to allow
375 		 * "ifconfig if0 delete" to remove the first IPv4 address on
376 		 * the interface.  For IPv6, as the spec allows multiple
377 		 * interface address from the day one, we consider "remove the
378 		 * first one" semantics to be not preferable.
379 		 */
380 		if (ia == NULL) {
381 			error = EADDRNOTAVAIL;
382 			goto out;
383 		}
384 		/* FALLTHROUGH */
385 	case SIOCAIFADDR_IN6:
386 		/*
387 		 * We always require users to specify a valid IPv6 address for
388 		 * the corresponding operation.
389 		 */
390 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
391 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
392 			error = EAFNOSUPPORT;
393 			goto out;
394 		}
395 
396 		if (td != NULL) {
397 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
398 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
399 			if (error)
400 				goto out;
401 		}
402 		break;
403 
404 	case SIOCGIFADDR_IN6:
405 		/* This interface is basically deprecated. use SIOCGIFCONF. */
406 		/* FALLTHROUGH */
407 	case SIOCGIFAFLAG_IN6:
408 	case SIOCGIFNETMASK_IN6:
409 	case SIOCGIFDSTADDR_IN6:
410 	case SIOCGIFALIFETIME_IN6:
411 		/* must think again about its semantics */
412 		if (ia == NULL) {
413 			error = EADDRNOTAVAIL;
414 			goto out;
415 		}
416 		break;
417 
418 	case SIOCSIFALIFETIME_IN6:
419 	    {
420 		struct in6_addrlifetime *lt;
421 
422 		if (td != NULL) {
423 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
424 			if (error)
425 				goto out;
426 		}
427 		if (ia == NULL) {
428 			error = EADDRNOTAVAIL;
429 			goto out;
430 		}
431 		/* sanity for overflow - beware unsigned */
432 		lt = &ifr->ifr_ifru.ifru_lifetime;
433 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
434 		    lt->ia6t_vltime + time_second < time_second) {
435 			error = EINVAL;
436 			goto out;
437 		}
438 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
439 		    lt->ia6t_pltime + time_second < time_second) {
440 			error = EINVAL;
441 			goto out;
442 		}
443 		break;
444 	    }
445 	}
446 
447 	switch (cmd) {
448 	case SIOCGIFADDR_IN6:
449 		ifr->ifr_addr = ia->ia_addr;
450 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
451 			goto out;
452 		break;
453 
454 	case SIOCGIFDSTADDR_IN6:
455 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
456 			error = EINVAL;
457 			goto out;
458 		}
459 		/*
460 		 * XXX: should we check if ifa_dstaddr is NULL and return
461 		 * an error?
462 		 */
463 		ifr->ifr_dstaddr = ia->ia_dstaddr;
464 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
465 			goto out;
466 		break;
467 
468 	case SIOCGIFNETMASK_IN6:
469 		ifr->ifr_addr = ia->ia_prefixmask;
470 		break;
471 
472 	case SIOCGIFAFLAG_IN6:
473 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
474 		break;
475 
476 	case SIOCGIFSTAT_IN6:
477 		if (ifp == NULL) {
478 			error = EINVAL;
479 			goto out;
480 		}
481 		bzero(&ifr->ifr_ifru.ifru_stat,
482 		    sizeof(ifr->ifr_ifru.ifru_stat));
483 		ifr->ifr_ifru.ifru_stat =
484 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
485 		break;
486 
487 	case SIOCGIFSTAT_ICMP6:
488 		if (ifp == NULL) {
489 			error = EINVAL;
490 			goto out;
491 		}
492 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
493 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
494 		ifr->ifr_ifru.ifru_icmp6stat =
495 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
496 		break;
497 
498 	case SIOCGIFALIFETIME_IN6:
499 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
500 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
501 			time_t maxexpire;
502 			struct in6_addrlifetime *retlt =
503 			    &ifr->ifr_ifru.ifru_lifetime;
504 
505 			/*
506 			 * XXX: adjust expiration time assuming time_t is
507 			 * signed.
508 			 */
509 			maxexpire = (-1) &
510 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
511 			if (ia->ia6_lifetime.ia6t_vltime <
512 			    maxexpire - ia->ia6_updatetime) {
513 				retlt->ia6t_expire = ia->ia6_updatetime +
514 				    ia->ia6_lifetime.ia6t_vltime;
515 			} else
516 				retlt->ia6t_expire = maxexpire;
517 		}
518 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
519 			time_t maxexpire;
520 			struct in6_addrlifetime *retlt =
521 			    &ifr->ifr_ifru.ifru_lifetime;
522 
523 			/*
524 			 * XXX: adjust expiration time assuming time_t is
525 			 * signed.
526 			 */
527 			maxexpire = (-1) &
528 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
529 			if (ia->ia6_lifetime.ia6t_pltime <
530 			    maxexpire - ia->ia6_updatetime) {
531 				retlt->ia6t_preferred = ia->ia6_updatetime +
532 				    ia->ia6_lifetime.ia6t_pltime;
533 			} else
534 				retlt->ia6t_preferred = maxexpire;
535 		}
536 		break;
537 
538 	case SIOCSIFALIFETIME_IN6:
539 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
540 		/* for sanity */
541 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
542 			ia->ia6_lifetime.ia6t_expire =
543 				time_second + ia->ia6_lifetime.ia6t_vltime;
544 		} else
545 			ia->ia6_lifetime.ia6t_expire = 0;
546 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
547 			ia->ia6_lifetime.ia6t_preferred =
548 				time_second + ia->ia6_lifetime.ia6t_pltime;
549 		} else
550 			ia->ia6_lifetime.ia6t_preferred = 0;
551 		break;
552 
553 	case SIOCAIFADDR_IN6:
554 	{
555 		int i;
556 		struct nd_prefixctl pr0;
557 		struct nd_prefix *pr;
558 
559 		/*
560 		 * first, make or update the interface address structure,
561 		 * and link it to the list.
562 		 */
563 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
564 			goto out;
565 		if (ia != NULL)
566 			ifa_free(&ia->ia_ifa);
567 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
568 		    == NULL) {
569 			/*
570 			 * this can happen when the user specify the 0 valid
571 			 * lifetime.
572 			 */
573 			break;
574 		}
575 
576 		/*
577 		 * then, make the prefix on-link on the interface.
578 		 * XXX: we'd rather create the prefix before the address, but
579 		 * we need at least one address to install the corresponding
580 		 * interface route, so we configure the address first.
581 		 */
582 
583 		/*
584 		 * convert mask to prefix length (prefixmask has already
585 		 * been validated in in6_update_ifa().
586 		 */
587 		bzero(&pr0, sizeof(pr0));
588 		pr0.ndpr_ifp = ifp;
589 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
590 		    NULL);
591 		if (pr0.ndpr_plen == 128) {
592 			break;	/* we don't need to install a host route. */
593 		}
594 		pr0.ndpr_prefix = ifra->ifra_addr;
595 		/* apply the mask for safety. */
596 		for (i = 0; i < 4; i++) {
597 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
598 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
599 		}
600 		/*
601 		 * XXX: since we don't have an API to set prefix (not address)
602 		 * lifetimes, we just use the same lifetimes as addresses.
603 		 * The (temporarily) installed lifetimes can be overridden by
604 		 * later advertised RAs (when accept_rtadv is non 0), which is
605 		 * an intended behavior.
606 		 */
607 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
608 		pr0.ndpr_raf_auto =
609 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
610 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
611 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
612 
613 		/* add the prefix if not yet. */
614 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
615 			/*
616 			 * nd6_prelist_add will install the corresponding
617 			 * interface route.
618 			 */
619 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
620 				goto out;
621 			if (pr == NULL) {
622 				log(LOG_ERR, "nd6_prelist_add succeeded but "
623 				    "no prefix\n");
624 				error = EINVAL;
625 				goto out;
626 			}
627 		}
628 
629 		/* relate the address to the prefix */
630 		if (ia->ia6_ndpr == NULL) {
631 			ia->ia6_ndpr = pr;
632 			pr->ndpr_refcnt++;
633 
634 			/*
635 			 * If this is the first autoconf address from the
636 			 * prefix, create a temporary address as well
637 			 * (when required).
638 			 */
639 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
640 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
641 				int e;
642 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
643 					log(LOG_NOTICE, "in6_control: failed "
644 					    "to create a temporary address, "
645 					    "errno=%d\n", e);
646 				}
647 			}
648 		}
649 
650 		/*
651 		 * this might affect the status of autoconfigured addresses,
652 		 * that is, this address might make other addresses detached.
653 		 */
654 		pfxlist_onlink_check();
655 		if (error == 0 && ia)
656 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
657 		break;
658 	}
659 
660 	case SIOCDIFADDR_IN6:
661 	{
662 		struct nd_prefix *pr;
663 
664 		/*
665 		 * If the address being deleted is the only one that owns
666 		 * the corresponding prefix, expire the prefix as well.
667 		 * XXX: theoretically, we don't have to worry about such
668 		 * relationship, since we separate the address management
669 		 * and the prefix management.  We do this, however, to provide
670 		 * as much backward compatibility as possible in terms of
671 		 * the ioctl operation.
672 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
673 		 */
674 		pr = ia->ia6_ndpr;
675 		in6_purgeaddr(&ia->ia_ifa);
676 		if (pr && pr->ndpr_refcnt == 0)
677 			prelist_remove(pr);
678 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
679 		break;
680 	}
681 
682 	default:
683 		if (ifp == NULL || ifp->if_ioctl == 0) {
684 			error = EOPNOTSUPP;
685 			goto out;
686 		}
687 		error = (*ifp->if_ioctl)(ifp, cmd, data);
688 		goto out;
689 	}
690 
691 	error = 0;
692 out:
693 	if (ia != NULL)
694 		ifa_free(&ia->ia_ifa);
695 	return (error);
696 }
697 
698 /*
699  * Update parameters of an IPv6 interface address.
700  * If necessary, a new entry is created and linked into address chains.
701  * This function is separated from in6_control().
702  * XXX: should this be performed under splnet()?
703  */
704 int
705 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
706     struct in6_ifaddr *ia, int flags)
707 {
708 	int error = 0, hostIsNew = 0, plen = -1;
709 	struct sockaddr_in6 dst6;
710 	struct in6_addrlifetime *lt;
711 	struct in6_multi_mship *imm;
712 	struct in6_multi *in6m_sol;
713 	struct rtentry *rt;
714 	int delay;
715 	char ip6buf[INET6_ADDRSTRLEN];
716 
717 	/* Validate parameters */
718 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
719 		return (EINVAL);
720 
721 	/*
722 	 * The destination address for a p2p link must have a family
723 	 * of AF_UNSPEC or AF_INET6.
724 	 */
725 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
726 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
727 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
728 		return (EAFNOSUPPORT);
729 	/*
730 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
731 	 * does not carry fields other than sin6_len.
732 	 */
733 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
734 		return (EINVAL);
735 	/*
736 	 * Because the IPv6 address architecture is classless, we require
737 	 * users to specify a (non 0) prefix length (mask) for a new address.
738 	 * We also require the prefix (when specified) mask is valid, and thus
739 	 * reject a non-consecutive mask.
740 	 */
741 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
742 		return (EINVAL);
743 	if (ifra->ifra_prefixmask.sin6_len != 0) {
744 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
745 		    (u_char *)&ifra->ifra_prefixmask +
746 		    ifra->ifra_prefixmask.sin6_len);
747 		if (plen <= 0)
748 			return (EINVAL);
749 	} else {
750 		/*
751 		 * In this case, ia must not be NULL.  We just use its prefix
752 		 * length.
753 		 */
754 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
755 	}
756 	/*
757 	 * If the destination address on a p2p interface is specified,
758 	 * and the address is a scoped one, validate/set the scope
759 	 * zone identifier.
760 	 */
761 	dst6 = ifra->ifra_dstaddr;
762 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
763 	    (dst6.sin6_family == AF_INET6)) {
764 		struct in6_addr in6_tmp;
765 		u_int32_t zoneid;
766 
767 		in6_tmp = dst6.sin6_addr;
768 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
769 			return (EINVAL); /* XXX: should be impossible */
770 
771 		if (dst6.sin6_scope_id != 0) {
772 			if (dst6.sin6_scope_id != zoneid)
773 				return (EINVAL);
774 		} else		/* user omit to specify the ID. */
775 			dst6.sin6_scope_id = zoneid;
776 
777 		/* convert into the internal form */
778 		if (sa6_embedscope(&dst6, 0))
779 			return (EINVAL); /* XXX: should be impossible */
780 	}
781 	/*
782 	 * The destination address can be specified only for a p2p or a
783 	 * loopback interface.  If specified, the corresponding prefix length
784 	 * must be 128.
785 	 */
786 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
787 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
788 			/* XXX: noisy message */
789 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
790 			    "be specified for a p2p or a loopback IF only\n"));
791 			return (EINVAL);
792 		}
793 		if (plen != 128) {
794 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
795 			    "be 128 when dstaddr is specified\n"));
796 			return (EINVAL);
797 		}
798 	}
799 	/* lifetime consistency check */
800 	lt = &ifra->ifra_lifetime;
801 	if (lt->ia6t_pltime > lt->ia6t_vltime)
802 		return (EINVAL);
803 	if (lt->ia6t_vltime == 0) {
804 		/*
805 		 * the following log might be noisy, but this is a typical
806 		 * configuration mistake or a tool's bug.
807 		 */
808 		nd6log((LOG_INFO,
809 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
810 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
811 
812 		if (ia == NULL)
813 			return (0); /* there's nothing to do */
814 	}
815 
816 	/*
817 	 * If this is a new address, allocate a new ifaddr and link it
818 	 * into chains.
819 	 */
820 	if (ia == NULL) {
821 		hostIsNew = 1;
822 		/*
823 		 * When in6_update_ifa() is called in a process of a received
824 		 * RA, it is called under an interrupt context.  So, we should
825 		 * call malloc with M_NOWAIT.
826 		 */
827 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
828 		    M_NOWAIT);
829 		if (ia == NULL)
830 			return (ENOBUFS);
831 		bzero((caddr_t)ia, sizeof(*ia));
832 		ifa_init(&ia->ia_ifa);
833 		LIST_INIT(&ia->ia6_memberships);
834 		/* Initialize the address and masks, and put time stamp */
835 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
836 		ia->ia_addr.sin6_family = AF_INET6;
837 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
838 		ia->ia6_createtime = time_second;
839 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
840 			/*
841 			 * XXX: some functions expect that ifa_dstaddr is not
842 			 * NULL for p2p interfaces.
843 			 */
844 			ia->ia_ifa.ifa_dstaddr =
845 			    (struct sockaddr *)&ia->ia_dstaddr;
846 		} else {
847 			ia->ia_ifa.ifa_dstaddr = NULL;
848 		}
849 		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
850 		ia->ia_ifp = ifp;
851 		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
852 		IF_ADDR_LOCK(ifp);
853 		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
854 		IF_ADDR_UNLOCK(ifp);
855 
856 		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
857 		IN6_IFADDR_WLOCK();
858 		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
859 		IN6_IFADDR_WUNLOCK();
860 	}
861 
862 	/* update timestamp */
863 	ia->ia6_updatetime = time_second;
864 
865 	/* set prefix mask */
866 	if (ifra->ifra_prefixmask.sin6_len) {
867 		/*
868 		 * We prohibit changing the prefix length of an existing
869 		 * address, because
870 		 * + such an operation should be rare in IPv6, and
871 		 * + the operation would confuse prefix management.
872 		 */
873 		if (ia->ia_prefixmask.sin6_len &&
874 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
875 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
876 			    " existing (%s) address should not be changed\n",
877 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
878 			error = EINVAL;
879 			goto unlink;
880 		}
881 		ia->ia_prefixmask = ifra->ifra_prefixmask;
882 	}
883 
884 	/*
885 	 * If a new destination address is specified, scrub the old one and
886 	 * install the new destination.  Note that the interface must be
887 	 * p2p or loopback (see the check above.)
888 	 */
889 	if (dst6.sin6_family == AF_INET6 &&
890 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
891 		int e;
892 
893 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
894 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
895 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
896 			    "a route to the old destination: %s\n",
897 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
898 			/* proceed anyway... */
899 		} else
900 			ia->ia_flags &= ~IFA_ROUTE;
901 		ia->ia_dstaddr = dst6;
902 	}
903 
904 	/*
905 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
906 	 * to see if the address is deprecated or invalidated, but initialize
907 	 * these members for applications.
908 	 */
909 	ia->ia6_lifetime = ifra->ifra_lifetime;
910 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
911 		ia->ia6_lifetime.ia6t_expire =
912 		    time_second + ia->ia6_lifetime.ia6t_vltime;
913 	} else
914 		ia->ia6_lifetime.ia6t_expire = 0;
915 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
916 		ia->ia6_lifetime.ia6t_preferred =
917 		    time_second + ia->ia6_lifetime.ia6t_pltime;
918 	} else
919 		ia->ia6_lifetime.ia6t_preferred = 0;
920 
921 	/* reset the interface and routing table appropriately. */
922 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
923 		goto unlink;
924 
925 	/*
926 	 * configure address flags.
927 	 */
928 	ia->ia6_flags = ifra->ifra_flags;
929 	/*
930 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
931 	 * userland, make it deprecated.
932 	 */
933 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
934 		ia->ia6_lifetime.ia6t_pltime = 0;
935 		ia->ia6_lifetime.ia6t_preferred = time_second;
936 	}
937 	/*
938 	 * Make the address tentative before joining multicast addresses,
939 	 * so that corresponding MLD responses would not have a tentative
940 	 * source address.
941 	 */
942 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
943 	if (hostIsNew && in6if_do_dad(ifp))
944 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
945 
946 	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
947 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
948 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
949 
950 	/*
951 	 * We are done if we have simply modified an existing address.
952 	 */
953 	if (!hostIsNew)
954 		return (error);
955 
956 	/*
957 	 * Beyond this point, we should call in6_purgeaddr upon an error,
958 	 * not just go to unlink.
959 	 */
960 
961 	/* Join necessary multicast groups */
962 	in6m_sol = NULL;
963 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
964 		struct sockaddr_in6 mltaddr, mltmask;
965 		struct in6_addr llsol;
966 
967 		/* join solicited multicast addr for new host id */
968 		bzero(&llsol, sizeof(struct in6_addr));
969 		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
970 		llsol.s6_addr32[1] = 0;
971 		llsol.s6_addr32[2] = htonl(1);
972 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
973 		llsol.s6_addr8[12] = 0xff;
974 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
975 			/* XXX: should not happen */
976 			log(LOG_ERR, "in6_update_ifa: "
977 			    "in6_setscope failed\n");
978 			goto cleanup;
979 		}
980 		delay = 0;
981 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
982 			/*
983 			 * We need a random delay for DAD on the address
984 			 * being configured.  It also means delaying
985 			 * transmission of the corresponding MLD report to
986 			 * avoid report collision.
987 			 * [RFC 4861, Section 6.3.7]
988 			 */
989 			delay = arc4random() %
990 			    (MAX_RTR_SOLICITATION_DELAY * hz);
991 		}
992 		imm = in6_joingroup(ifp, &llsol, &error, delay);
993 		if (imm == NULL) {
994 			nd6log((LOG_WARNING,
995 			    "in6_update_ifa: addmulti failed for "
996 			    "%s on %s (errno=%d)\n",
997 			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
998 			    error));
999 			goto cleanup;
1000 		}
1001 		LIST_INSERT_HEAD(&ia->ia6_memberships,
1002 		    imm, i6mm_chain);
1003 		in6m_sol = imm->i6mm_maddr;
1004 
1005 		bzero(&mltmask, sizeof(mltmask));
1006 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1007 		mltmask.sin6_family = AF_INET6;
1008 		mltmask.sin6_addr = in6mask32;
1009 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1010 
1011 		/*
1012 		 * join link-local all-nodes address
1013 		 */
1014 		bzero(&mltaddr, sizeof(mltaddr));
1015 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1016 		mltaddr.sin6_family = AF_INET6;
1017 		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1018 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1019 		    0)
1020 			goto cleanup; /* XXX: should not fail */
1021 
1022 		/*
1023 		 * XXX: do we really need this automatic routes?
1024 		 * We should probably reconsider this stuff.  Most applications
1025 		 * actually do not need the routes, since they usually specify
1026 		 * the outgoing interface.
1027 		 */
1028 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1029 		if (rt) {
1030 			/* XXX: only works in !SCOPEDROUTING case. */
1031 			if (memcmp(&mltaddr.sin6_addr,
1032 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1033 			    MLTMASK_LEN)) {
1034 				RTFREE_LOCKED(rt);
1035 				rt = NULL;
1036 			}
1037 		}
1038 		if (!rt) {
1039 			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1040 			    (struct sockaddr *)&ia->ia_addr,
1041 			    (struct sockaddr *)&mltmask, RTF_UP,
1042 			    (struct rtentry **)0);
1043 			if (error)
1044 				goto cleanup;
1045 		} else {
1046 			RTFREE_LOCKED(rt);
1047 		}
1048 
1049 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1050 		if (!imm) {
1051 			nd6log((LOG_WARNING,
1052 			    "in6_update_ifa: addmulti failed for "
1053 			    "%s on %s (errno=%d)\n",
1054 			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1055 			    if_name(ifp), error));
1056 			goto cleanup;
1057 		}
1058 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1059 
1060 		/*
1061 		 * join node information group address
1062 		 */
1063 		delay = 0;
1064 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1065 			/*
1066 			 * The spec doesn't say anything about delay for this
1067 			 * group, but the same logic should apply.
1068 			 */
1069 			delay = arc4random() %
1070 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1071 		}
1072 		if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
1073 			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1074 			    delay); /* XXX jinmei */
1075 			if (!imm) {
1076 				nd6log((LOG_WARNING, "in6_update_ifa: "
1077 				    "addmulti failed for %s on %s "
1078 				    "(errno=%d)\n",
1079 				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1080 				    if_name(ifp), error));
1081 				/* XXX not very fatal, go on... */
1082 			} else {
1083 				LIST_INSERT_HEAD(&ia->ia6_memberships,
1084 				    imm, i6mm_chain);
1085 			}
1086 		}
1087 
1088 		/*
1089 		 * join interface-local all-nodes address.
1090 		 * (ff01::1%ifN, and ff01::%ifN/32)
1091 		 */
1092 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1093 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1094 		    != 0)
1095 			goto cleanup; /* XXX: should not fail */
1096 		/* XXX: again, do we really need the route? */
1097 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1098 		if (rt) {
1099 			if (memcmp(&mltaddr.sin6_addr,
1100 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1101 			    MLTMASK_LEN)) {
1102 				RTFREE_LOCKED(rt);
1103 				rt = NULL;
1104 			}
1105 		}
1106 		if (!rt) {
1107 			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1108 			    (struct sockaddr *)&ia->ia_addr,
1109 			    (struct sockaddr *)&mltmask, RTF_UP,
1110 			    (struct rtentry **)0);
1111 			if (error)
1112 				goto cleanup;
1113 		} else
1114 			RTFREE_LOCKED(rt);
1115 
1116 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1117 		if (!imm) {
1118 			nd6log((LOG_WARNING, "in6_update_ifa: "
1119 			    "addmulti failed for %s on %s "
1120 			    "(errno=%d)\n",
1121 			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1122 			    if_name(ifp), error));
1123 			goto cleanup;
1124 		}
1125 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1126 #undef	MLTMASK_LEN
1127 	}
1128 
1129 	/*
1130 	 * Perform DAD, if needed.
1131 	 * XXX It may be of use, if we can administratively
1132 	 * disable DAD.
1133 	 */
1134 	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1135 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1136 	{
1137 		int mindelay, maxdelay;
1138 
1139 		delay = 0;
1140 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1141 			/*
1142 			 * We need to impose a delay before sending an NS
1143 			 * for DAD.  Check if we also needed a delay for the
1144 			 * corresponding MLD message.  If we did, the delay
1145 			 * should be larger than the MLD delay (this could be
1146 			 * relaxed a bit, but this simple logic is at least
1147 			 * safe).
1148 			 * XXX: Break data hiding guidelines and look at
1149 			 * state for the solicited multicast group.
1150 			 */
1151 			mindelay = 0;
1152 			if (in6m_sol != NULL &&
1153 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1154 				mindelay = in6m_sol->in6m_timer;
1155 			}
1156 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1157 			if (maxdelay - mindelay == 0)
1158 				delay = 0;
1159 			else {
1160 				delay =
1161 				    (arc4random() % (maxdelay - mindelay)) +
1162 				    mindelay;
1163 			}
1164 		}
1165 		nd6_dad_start((struct ifaddr *)ia, delay);
1166 	}
1167 
1168 	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1169 	ifa_free(&ia->ia_ifa);
1170 	return (error);
1171 
1172   unlink:
1173 	/*
1174 	 * XXX: if a change of an existing address failed, keep the entry
1175 	 * anyway.
1176 	 */
1177 	if (hostIsNew) {
1178 		in6_unlink_ifa(ia, ifp);
1179 		ifa_free(&ia->ia_ifa);
1180 	}
1181 	return (error);
1182 
1183   cleanup:
1184 	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1185 	ifa_free(&ia->ia_ifa);
1186 	in6_purgeaddr(&ia->ia_ifa);
1187 	return error;
1188 }
1189 
1190 void
1191 in6_purgeaddr(struct ifaddr *ifa)
1192 {
1193 	struct ifnet *ifp = ifa->ifa_ifp;
1194 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1195 	struct in6_multi_mship *imm;
1196 	struct sockaddr_in6 mltaddr, mltmask;
1197 	struct rtentry rt0;
1198 	struct sockaddr_dl gateway;
1199 	struct sockaddr_in6 mask, addr;
1200 	int plen, error;
1201 	struct rtentry *rt;
1202 	struct ifaddr *ifa0, *nifa;
1203 
1204 	/*
1205 	 * find another IPv6 address as the gateway for the
1206 	 * link-local and node-local all-nodes multicast
1207 	 * address routes
1208 	 */
1209 	IF_ADDR_LOCK(ifp);
1210 	TAILQ_FOREACH_SAFE(ifa0, &ifp->if_addrhead, ifa_link, nifa) {
1211 		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1212 		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1213 			   &ia->ia_addr.sin6_addr,
1214 			   sizeof(struct in6_addr)) == 0)
1215 			continue;
1216 		else
1217 			break;
1218 	}
1219 	if (ifa0 != NULL)
1220 		ifa_ref(ifa0);
1221 	IF_ADDR_UNLOCK(ifp);
1222 
1223 	/*
1224 	 * Remove the loopback route to the interface address.
1225 	 * The check for the current setting of "nd6_useloopback"
1226 	 * is not needed.
1227 	 */
1228 	if (ia->ia_flags & IFA_RTSELF) {
1229 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1230 				       (struct sockaddr *)&ia->ia_addr);
1231 		if (error == 0)
1232 			ia->ia_flags &= ~IFA_RTSELF;
1233 	}
1234 
1235 	/* stop DAD processing */
1236 	nd6_dad_stop(ifa);
1237 
1238 	IF_AFDATA_LOCK(ifp);
1239 	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
1240 	    (struct sockaddr *)&ia->ia_addr);
1241 	IF_AFDATA_UNLOCK(ifp);
1242 
1243 	/*
1244 	 * initialize for rtmsg generation
1245 	 */
1246 	bzero(&gateway, sizeof(gateway));
1247 	gateway.sdl_len = sizeof(gateway);
1248 	gateway.sdl_family = AF_LINK;
1249 	gateway.sdl_nlen = 0;
1250 	gateway.sdl_alen = ifp->if_addrlen;
1251 	/* */
1252 	bzero(&rt0, sizeof(rt0));
1253 	rt0.rt_gateway = (struct sockaddr *)&gateway;
1254 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
1255 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
1256 	rt_mask(&rt0) = (struct sockaddr *)&mask;
1257 	rt_key(&rt0) = (struct sockaddr *)&addr;
1258 	rt0.rt_flags = RTF_HOST | RTF_STATIC;
1259 	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
1260 
1261 	/*
1262 	 * leave from multicast groups we have joined for the interface
1263 	 */
1264 	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1265 		LIST_REMOVE(imm, i6mm_chain);
1266 		in6_leavegroup(imm);
1267 	}
1268 
1269 	/*
1270 	 * remove the link-local all-nodes address
1271 	 */
1272 	bzero(&mltmask, sizeof(mltmask));
1273 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1274 	mltmask.sin6_family = AF_INET6;
1275 	mltmask.sin6_addr = in6mask32;
1276 
1277 	bzero(&mltaddr, sizeof(mltaddr));
1278 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1279 	mltaddr.sin6_family = AF_INET6;
1280 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1281 
1282 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1283 	    0)
1284 		goto cleanup;
1285 
1286 	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1287 	if (rt != NULL && rt->rt_gateway != NULL &&
1288 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1289 		    &ia->ia_addr.sin6_addr,
1290 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1291 		/*
1292 		 * if no more IPv6 address exists on this interface
1293 		 * then remove the multicast address route
1294 		 */
1295 		if (ifa0 == NULL) {
1296 			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1297 			       sizeof(mltaddr.sin6_addr));
1298 			RTFREE_LOCKED(rt);
1299 			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1300 					  (struct sockaddr *)&ia->ia_addr,
1301 					  (struct sockaddr *)&mltmask, RTF_UP,
1302 					  (struct rtentry **)0);
1303 			if (error)
1304 				log(LOG_INFO, "in6_purgeaddr: link-local all-nodes"
1305 				    "multicast address deletion error\n");
1306 		} else {
1307 			/*
1308 			 * replace the gateway of the route
1309 			 */
1310 			struct sockaddr_in6 sa;
1311 
1312 			bzero(&sa, sizeof(sa));
1313 			sa.sin6_len = sizeof(struct sockaddr_in6);
1314 			sa.sin6_family = AF_INET6;
1315 			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1316 			       sizeof(sa.sin6_addr));
1317 			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1318 			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1319 			RTFREE_LOCKED(rt);
1320 		}
1321 	} else {
1322 		if (rt != NULL)
1323 			RTFREE_LOCKED(rt);
1324 	}
1325 
1326 	/*
1327 	 * remove the node-local all-nodes address
1328 	 */
1329 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1330 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1331 	    0)
1332 		goto cleanup;
1333 
1334 	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1335 	if (rt != NULL && rt->rt_gateway != NULL &&
1336 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1337 		    &ia->ia_addr.sin6_addr,
1338 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1339 		/*
1340 		 * if no more IPv6 address exists on this interface
1341 		 * then remove the multicast address route
1342 		 */
1343 		if (ifa0 == NULL) {
1344 			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1345 			       sizeof(mltaddr.sin6_addr));
1346 
1347 			RTFREE_LOCKED(rt);
1348 			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1349 					  (struct sockaddr *)&ia->ia_addr,
1350 					  (struct sockaddr *)&mltmask, RTF_UP,
1351 					  (struct rtentry **)0);
1352 
1353 			if (error)
1354 				log(LOG_INFO, "in6_purgeaddr: node-local all-nodes"
1355 				    "multicast address deletion error\n");
1356 		} else {
1357 			/*
1358 			 * replace the gateway of the route
1359 			 */
1360 			struct sockaddr_in6 sa;
1361 
1362 			bzero(&sa, sizeof(sa));
1363 			sa.sin6_len = sizeof(struct sockaddr_in6);
1364 			sa.sin6_family = AF_INET6;
1365 			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1366 			       sizeof(sa.sin6_addr));
1367 			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1368 			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1369 			RTFREE_LOCKED(rt);
1370 		}
1371 	} else {
1372 		if (rt != NULL)
1373 			RTFREE_LOCKED(rt);
1374 	}
1375 
1376 cleanup:
1377 
1378 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1379 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1380 		int error;
1381 		struct sockaddr *dstaddr;
1382 
1383 		/*
1384 		 * use the interface address if configuring an
1385 		 * interface address with a /128 prefix len
1386 		 */
1387 		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1388 			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1389 		else
1390 			dstaddr = (struct sockaddr *)&ia->ia_addr;
1391 
1392 		error = rtrequest(RTM_DELETE,
1393 		    (struct sockaddr *)dstaddr,
1394 		    (struct sockaddr *)&ia->ia_addr,
1395 		    (struct sockaddr *)&ia->ia_prefixmask,
1396 		    ia->ia_flags | RTF_HOST, NULL);
1397 		if (error != 0)
1398 			return;
1399 		ia->ia_flags &= ~IFA_ROUTE;
1400 	}
1401 	if (ifa0 != NULL)
1402 		ifa_free(ifa0);
1403 
1404 	in6_unlink_ifa(ia, ifp);
1405 }
1406 
1407 static void
1408 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1409 {
1410 	int	s = splnet();
1411 
1412 	IF_ADDR_LOCK(ifp);
1413 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1414 	IF_ADDR_UNLOCK(ifp);
1415 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1416 
1417 	/*
1418 	 * Defer the release of what might be the last reference to the
1419 	 * in6_ifaddr so that it can't be freed before the remainder of the
1420 	 * cleanup.
1421 	 */
1422 	IN6_IFADDR_WLOCK();
1423 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1424 	IN6_IFADDR_WUNLOCK();
1425 
1426 	/*
1427 	 * Release the reference to the base prefix.  There should be a
1428 	 * positive reference.
1429 	 */
1430 	if (ia->ia6_ndpr == NULL) {
1431 		nd6log((LOG_NOTICE,
1432 		    "in6_unlink_ifa: autoconf'ed address "
1433 		    "%p has no prefix\n", ia));
1434 	} else {
1435 		ia->ia6_ndpr->ndpr_refcnt--;
1436 		ia->ia6_ndpr = NULL;
1437 	}
1438 
1439 	/*
1440 	 * Also, if the address being removed is autoconf'ed, call
1441 	 * pfxlist_onlink_check() since the release might affect the status of
1442 	 * other (detached) addresses.
1443 	 */
1444 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1445 		pfxlist_onlink_check();
1446 	}
1447 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1448 	splx(s);
1449 }
1450 
1451 void
1452 in6_purgeif(struct ifnet *ifp)
1453 {
1454 	struct ifaddr *ifa, *nifa;
1455 
1456 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1457 		if (ifa->ifa_addr->sa_family != AF_INET6)
1458 			continue;
1459 		in6_purgeaddr(ifa);
1460 	}
1461 
1462 	in6_ifdetach(ifp);
1463 }
1464 
1465 /*
1466  * SIOC[GAD]LIFADDR.
1467  *	SIOCGLIFADDR: get first address. (?)
1468  *	SIOCGLIFADDR with IFLR_PREFIX:
1469  *		get first address that matches the specified prefix.
1470  *	SIOCALIFADDR: add the specified address.
1471  *	SIOCALIFADDR with IFLR_PREFIX:
1472  *		add the specified prefix, filling hostid part from
1473  *		the first link-local address.  prefixlen must be <= 64.
1474  *	SIOCDLIFADDR: delete the specified address.
1475  *	SIOCDLIFADDR with IFLR_PREFIX:
1476  *		delete the first address that matches the specified prefix.
1477  * return values:
1478  *	EINVAL on invalid parameters
1479  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1480  *	other values may be returned from in6_ioctl()
1481  *
1482  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1483  * this is to accomodate address naming scheme other than RFC2374,
1484  * in the future.
1485  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1486  * address encoding scheme. (see figure on page 8)
1487  */
1488 static int
1489 in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1490     struct ifnet *ifp, struct thread *td)
1491 {
1492 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1493 	struct ifaddr *ifa;
1494 	struct sockaddr *sa;
1495 
1496 	/* sanity checks */
1497 	if (!data || !ifp) {
1498 		panic("invalid argument to in6_lifaddr_ioctl");
1499 		/* NOTREACHED */
1500 	}
1501 
1502 	switch (cmd) {
1503 	case SIOCGLIFADDR:
1504 		/* address must be specified on GET with IFLR_PREFIX */
1505 		if ((iflr->flags & IFLR_PREFIX) == 0)
1506 			break;
1507 		/* FALLTHROUGH */
1508 	case SIOCALIFADDR:
1509 	case SIOCDLIFADDR:
1510 		/* address must be specified on ADD and DELETE */
1511 		sa = (struct sockaddr *)&iflr->addr;
1512 		if (sa->sa_family != AF_INET6)
1513 			return EINVAL;
1514 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1515 			return EINVAL;
1516 		/* XXX need improvement */
1517 		sa = (struct sockaddr *)&iflr->dstaddr;
1518 		if (sa->sa_family && sa->sa_family != AF_INET6)
1519 			return EINVAL;
1520 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1521 			return EINVAL;
1522 		break;
1523 	default: /* shouldn't happen */
1524 #if 0
1525 		panic("invalid cmd to in6_lifaddr_ioctl");
1526 		/* NOTREACHED */
1527 #else
1528 		return EOPNOTSUPP;
1529 #endif
1530 	}
1531 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1532 		return EINVAL;
1533 
1534 	switch (cmd) {
1535 	case SIOCALIFADDR:
1536 	    {
1537 		struct in6_aliasreq ifra;
1538 		struct in6_addr *hostid = NULL;
1539 		int prefixlen;
1540 
1541 		ifa = NULL;
1542 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1543 			struct sockaddr_in6 *sin6;
1544 
1545 			/*
1546 			 * hostid is to fill in the hostid part of the
1547 			 * address.  hostid points to the first link-local
1548 			 * address attached to the interface.
1549 			 */
1550 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1551 			if (!ifa)
1552 				return EADDRNOTAVAIL;
1553 			hostid = IFA_IN6(ifa);
1554 
1555 			/* prefixlen must be <= 64. */
1556 			if (64 < iflr->prefixlen)
1557 				return EINVAL;
1558 			prefixlen = iflr->prefixlen;
1559 
1560 			/* hostid part must be zero. */
1561 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1562 			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1563 			    sin6->sin6_addr.s6_addr32[3] != 0) {
1564 				return EINVAL;
1565 			}
1566 		} else
1567 			prefixlen = iflr->prefixlen;
1568 
1569 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1570 		bzero(&ifra, sizeof(ifra));
1571 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1572 
1573 		bcopy(&iflr->addr, &ifra.ifra_addr,
1574 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1575 		if (hostid) {
1576 			/* fill in hostid part */
1577 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1578 			    hostid->s6_addr32[2];
1579 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1580 			    hostid->s6_addr32[3];
1581 		}
1582 
1583 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1584 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1585 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1586 			if (hostid) {
1587 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1588 				    hostid->s6_addr32[2];
1589 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1590 				    hostid->s6_addr32[3];
1591 			}
1592 		}
1593 		if (ifa != NULL)
1594 			ifa_free(ifa);
1595 
1596 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1597 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1598 
1599 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1600 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1601 	    }
1602 	case SIOCGLIFADDR:
1603 	case SIOCDLIFADDR:
1604 	    {
1605 		struct in6_ifaddr *ia;
1606 		struct in6_addr mask, candidate, match;
1607 		struct sockaddr_in6 *sin6;
1608 		int cmp;
1609 
1610 		bzero(&mask, sizeof(mask));
1611 		if (iflr->flags & IFLR_PREFIX) {
1612 			/* lookup a prefix rather than address. */
1613 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1614 
1615 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1616 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1617 			match.s6_addr32[0] &= mask.s6_addr32[0];
1618 			match.s6_addr32[1] &= mask.s6_addr32[1];
1619 			match.s6_addr32[2] &= mask.s6_addr32[2];
1620 			match.s6_addr32[3] &= mask.s6_addr32[3];
1621 
1622 			/* if you set extra bits, that's wrong */
1623 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1624 				return EINVAL;
1625 
1626 			cmp = 1;
1627 		} else {
1628 			if (cmd == SIOCGLIFADDR) {
1629 				/* on getting an address, take the 1st match */
1630 				cmp = 0;	/* XXX */
1631 			} else {
1632 				/* on deleting an address, do exact match */
1633 				in6_prefixlen2mask(&mask, 128);
1634 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1635 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1636 
1637 				cmp = 1;
1638 			}
1639 		}
1640 
1641 		IF_ADDR_LOCK(ifp);
1642 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1643 			if (ifa->ifa_addr->sa_family != AF_INET6)
1644 				continue;
1645 			if (!cmp)
1646 				break;
1647 
1648 			/*
1649 			 * XXX: this is adhoc, but is necessary to allow
1650 			 * a user to specify fe80::/64 (not /10) for a
1651 			 * link-local address.
1652 			 */
1653 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1654 			in6_clearscope(&candidate);
1655 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1656 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1657 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1658 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1659 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1660 				break;
1661 		}
1662 		IF_ADDR_UNLOCK(ifp);
1663 		if (!ifa)
1664 			return EADDRNOTAVAIL;
1665 		ia = ifa2ia6(ifa);
1666 
1667 		if (cmd == SIOCGLIFADDR) {
1668 			int error;
1669 
1670 			/* fill in the if_laddrreq structure */
1671 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1672 			error = sa6_recoverscope(
1673 			    (struct sockaddr_in6 *)&iflr->addr);
1674 			if (error != 0)
1675 				return (error);
1676 
1677 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1678 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1679 				    ia->ia_dstaddr.sin6_len);
1680 				error = sa6_recoverscope(
1681 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1682 				if (error != 0)
1683 					return (error);
1684 			} else
1685 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1686 
1687 			iflr->prefixlen =
1688 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1689 
1690 			iflr->flags = ia->ia6_flags;	/* XXX */
1691 
1692 			return 0;
1693 		} else {
1694 			struct in6_aliasreq ifra;
1695 
1696 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1697 			bzero(&ifra, sizeof(ifra));
1698 			bcopy(iflr->iflr_name, ifra.ifra_name,
1699 			    sizeof(ifra.ifra_name));
1700 
1701 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1702 			    ia->ia_addr.sin6_len);
1703 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1704 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1705 				    ia->ia_dstaddr.sin6_len);
1706 			} else {
1707 				bzero(&ifra.ifra_dstaddr,
1708 				    sizeof(ifra.ifra_dstaddr));
1709 			}
1710 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1711 			    ia->ia_prefixmask.sin6_len);
1712 
1713 			ifra.ifra_flags = ia->ia6_flags;
1714 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1715 			    ifp, td);
1716 		}
1717 	    }
1718 	}
1719 
1720 	return EOPNOTSUPP;	/* just for safety */
1721 }
1722 
1723 /*
1724  * Initialize an interface's intetnet6 address
1725  * and routing table entry.
1726  */
1727 static int
1728 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1729     struct sockaddr_in6 *sin6, int newhost)
1730 {
1731 	int	error = 0, plen, ifacount = 0;
1732 	int	s = splimp();
1733 	struct ifaddr *ifa;
1734 
1735 	/*
1736 	 * Give the interface a chance to initialize
1737 	 * if this is its first address,
1738 	 * and to validate the address if necessary.
1739 	 */
1740 	IF_ADDR_LOCK(ifp);
1741 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1742 		if (ifa->ifa_addr->sa_family != AF_INET6)
1743 			continue;
1744 		ifacount++;
1745 	}
1746 	IF_ADDR_UNLOCK(ifp);
1747 
1748 	ia->ia_addr = *sin6;
1749 
1750 	if (ifacount <= 1 && ifp->if_ioctl) {
1751 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1752 		if (error) {
1753 			splx(s);
1754 			return (error);
1755 		}
1756 	}
1757 	splx(s);
1758 
1759 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1760 
1761 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1762 
1763 	/*
1764 	 * Special case:
1765 	 * If a new destination address is specified for a point-to-point
1766 	 * interface, install a route to the destination as an interface
1767 	 * direct route.
1768 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1769 	 * interface that share the same destination.
1770 	 */
1771 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1772 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1773 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1774 		int rtflags = RTF_UP | RTF_HOST;
1775 
1776 		error = rtrequest(RTM_ADD,
1777 		    (struct sockaddr *)&ia->ia_dstaddr,
1778 		    (struct sockaddr *)&ia->ia_addr,
1779 		    (struct sockaddr *)&ia->ia_prefixmask,
1780 		    ia->ia_flags | rtflags, NULL);
1781 		if (error != 0)
1782 			return (error);
1783 		ia->ia_flags |= IFA_ROUTE;
1784 	}
1785 
1786 	/*
1787 	 * add a loopback route to self
1788 	 */
1789 	if (!(ia->ia_flags & IFA_ROUTE)
1790 	    && (V_nd6_useloopback
1791 		|| (ifp->if_flags & IFF_LOOPBACK))) {
1792 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1793 				       (struct sockaddr *)&ia->ia_addr);
1794 		if (error == 0)
1795 			ia->ia_flags |= IFA_RTSELF;
1796 	}
1797 
1798 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1799 	if (newhost) {
1800 		struct llentry *ln;
1801 		struct rtentry rt;
1802 		struct sockaddr_dl gateway;
1803 		struct sockaddr_in6 mask, addr;
1804 
1805 		IF_AFDATA_LOCK(ifp);
1806 		ia->ia_ifa.ifa_rtrequest = NULL;
1807 
1808 		/* XXX QL
1809 		 * we need to report rt_newaddrmsg
1810 		 */
1811 		ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR | LLE_EXCLUSIVE),
1812 		    (struct sockaddr *)&ia->ia_addr);
1813 		IF_AFDATA_UNLOCK(ifp);
1814 		if (ln != NULL) {
1815 			ln->la_expire = 0;  /* for IPv6 this means permanent */
1816 			ln->ln_state = ND6_LLINFO_REACHABLE;
1817 			/*
1818 			 * initialize for rtmsg generation
1819 			 */
1820 			bzero(&gateway, sizeof(gateway));
1821 			gateway.sdl_len = sizeof(gateway);
1822 			gateway.sdl_family = AF_LINK;
1823 			gateway.sdl_nlen = 0;
1824 			gateway.sdl_alen = 6;
1825 			memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned, sizeof(ln->ll_addr));
1826 			/* */
1827 			LLE_WUNLOCK(ln);
1828 		}
1829 
1830 		bzero(&rt, sizeof(rt));
1831 		rt.rt_gateway = (struct sockaddr *)&gateway;
1832 		memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
1833 		memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
1834 		rt_mask(&rt) = (struct sockaddr *)&mask;
1835 		rt_key(&rt) = (struct sockaddr *)&addr;
1836 		rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
1837 		rt_newaddrmsg(RTM_ADD, &ia->ia_ifa, 0, &rt);
1838 	}
1839 
1840 	return (error);
1841 }
1842 
1843 /*
1844  * Find an IPv6 interface link-local address specific to an interface.
1845  * ifaddr is returned referenced.
1846  */
1847 struct in6_ifaddr *
1848 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1849 {
1850 	struct ifaddr *ifa;
1851 
1852 	IF_ADDR_LOCK(ifp);
1853 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1854 		if (ifa->ifa_addr->sa_family != AF_INET6)
1855 			continue;
1856 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1857 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1858 			     ignoreflags) != 0)
1859 				continue;
1860 			ifa_ref(ifa);
1861 			break;
1862 		}
1863 	}
1864 	IF_ADDR_UNLOCK(ifp);
1865 
1866 	return ((struct in6_ifaddr *)ifa);
1867 }
1868 
1869 
1870 /*
1871  * find the internet address corresponding to a given interface and address.
1872  * ifaddr is returned referenced.
1873  */
1874 struct in6_ifaddr *
1875 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1876 {
1877 	struct ifaddr *ifa;
1878 
1879 	IF_ADDR_LOCK(ifp);
1880 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1881 		if (ifa->ifa_addr->sa_family != AF_INET6)
1882 			continue;
1883 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1884 			ifa_ref(ifa);
1885 			break;
1886 		}
1887 	}
1888 	IF_ADDR_UNLOCK(ifp);
1889 
1890 	return ((struct in6_ifaddr *)ifa);
1891 }
1892 
1893 /*
1894  * Convert IP6 address to printable (loggable) representation. Caller
1895  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1896  */
1897 static char digits[] = "0123456789abcdef";
1898 char *
1899 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1900 {
1901 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1902 	char *cp;
1903 	const u_int16_t *a = (const u_int16_t *)addr;
1904 	const u_int8_t *d;
1905 	int dcolon = 0, zero = 0;
1906 
1907 	cp = ip6buf;
1908 
1909 	for (i = 0; i < 8; i++) {
1910 		if (*(a + i) == 0) {
1911 			cnt++;
1912 			if (cnt == 1)
1913 				idx = i;
1914 		}
1915 		else if (maxcnt < cnt) {
1916 			maxcnt = cnt;
1917 			index = idx;
1918 			cnt = 0;
1919 		}
1920 	}
1921 	if (maxcnt < cnt) {
1922 		maxcnt = cnt;
1923 		index = idx;
1924 	}
1925 
1926 	for (i = 0; i < 8; i++) {
1927 		if (dcolon == 1) {
1928 			if (*a == 0) {
1929 				if (i == 7)
1930 					*cp++ = ':';
1931 				a++;
1932 				continue;
1933 			} else
1934 				dcolon = 2;
1935 		}
1936 		if (*a == 0) {
1937 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1938 				if (i == 0)
1939 					*cp++ = ':';
1940 				*cp++ = ':';
1941 				dcolon = 1;
1942 			} else {
1943 				*cp++ = '0';
1944 				*cp++ = ':';
1945 			}
1946 			a++;
1947 			continue;
1948 		}
1949 		d = (const u_char *)a;
1950 		/* Try to eliminate leading zeros in printout like in :0001. */
1951 		zero = 1;
1952 		*cp = digits[*d >> 4];
1953 		if (*cp != '0') {
1954 			zero = 0;
1955 			cp++;
1956 		}
1957 		*cp = digits[*d++ & 0xf];
1958 		if (zero == 0 || (*cp != '0')) {
1959 			zero = 0;
1960 			cp++;
1961 		}
1962 		*cp = digits[*d >> 4];
1963 		if (zero == 0 || (*cp != '0')) {
1964 			zero = 0;
1965 			cp++;
1966 		}
1967 		*cp++ = digits[*d & 0xf];
1968 		*cp++ = ':';
1969 		a++;
1970 	}
1971 	*--cp = '\0';
1972 	return (ip6buf);
1973 }
1974 
1975 int
1976 in6_localaddr(struct in6_addr *in6)
1977 {
1978 	struct in6_ifaddr *ia;
1979 
1980 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1981 		return 1;
1982 
1983 	IN6_IFADDR_RLOCK();
1984 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1985 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1986 		    &ia->ia_prefixmask.sin6_addr)) {
1987 			IN6_IFADDR_RUNLOCK();
1988 			return 1;
1989 		}
1990 	}
1991 	IN6_IFADDR_RUNLOCK();
1992 
1993 	return (0);
1994 }
1995 
1996 int
1997 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1998 {
1999 	struct in6_ifaddr *ia;
2000 
2001 	IN6_IFADDR_RLOCK();
2002 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2003 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2004 				       &sa6->sin6_addr) &&
2005 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
2006 			IN6_IFADDR_RUNLOCK();
2007 			return (1); /* true */
2008 		}
2009 
2010 		/* XXX: do we still have to go thru the rest of the list? */
2011 	}
2012 	IN6_IFADDR_RUNLOCK();
2013 
2014 	return (0);		/* false */
2015 }
2016 
2017 /*
2018  * return length of part which dst and src are equal
2019  * hard coding...
2020  */
2021 int
2022 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2023 {
2024 	int match = 0;
2025 	u_char *s = (u_char *)src, *d = (u_char *)dst;
2026 	u_char *lim = s + 16, r;
2027 
2028 	while (s < lim)
2029 		if ((r = (*d++ ^ *s++)) != 0) {
2030 			while (r < 128) {
2031 				match++;
2032 				r <<= 1;
2033 			}
2034 			break;
2035 		} else
2036 			match += 8;
2037 	return match;
2038 }
2039 
2040 /* XXX: to be scope conscious */
2041 int
2042 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2043 {
2044 	int bytelen, bitlen;
2045 
2046 	/* sanity check */
2047 	if (0 > len || len > 128) {
2048 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2049 		    len);
2050 		return (0);
2051 	}
2052 
2053 	bytelen = len / 8;
2054 	bitlen = len % 8;
2055 
2056 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2057 		return (0);
2058 	if (bitlen != 0 &&
2059 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2060 	    p2->s6_addr[bytelen] >> (8 - bitlen))
2061 		return (0);
2062 
2063 	return (1);
2064 }
2065 
2066 void
2067 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2068 {
2069 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2070 	int bytelen, bitlen, i;
2071 
2072 	/* sanity check */
2073 	if (0 > len || len > 128) {
2074 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2075 		    len);
2076 		return;
2077 	}
2078 
2079 	bzero(maskp, sizeof(*maskp));
2080 	bytelen = len / 8;
2081 	bitlen = len % 8;
2082 	for (i = 0; i < bytelen; i++)
2083 		maskp->s6_addr[i] = 0xff;
2084 	if (bitlen)
2085 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2086 }
2087 
2088 /*
2089  * return the best address out of the same scope. if no address was
2090  * found, return the first valid address from designated IF.
2091  */
2092 struct in6_ifaddr *
2093 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2094 {
2095 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2096 	struct ifaddr *ifa;
2097 	struct in6_ifaddr *besta = 0;
2098 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2099 
2100 	dep[0] = dep[1] = NULL;
2101 
2102 	/*
2103 	 * We first look for addresses in the same scope.
2104 	 * If there is one, return it.
2105 	 * If two or more, return one which matches the dst longest.
2106 	 * If none, return one of global addresses assigned other ifs.
2107 	 */
2108 	IF_ADDR_LOCK(ifp);
2109 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2110 		if (ifa->ifa_addr->sa_family != AF_INET6)
2111 			continue;
2112 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2113 			continue; /* XXX: is there any case to allow anycast? */
2114 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2115 			continue; /* don't use this interface */
2116 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2117 			continue;
2118 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2119 			if (V_ip6_use_deprecated)
2120 				dep[0] = (struct in6_ifaddr *)ifa;
2121 			continue;
2122 		}
2123 
2124 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2125 			/*
2126 			 * call in6_matchlen() as few as possible
2127 			 */
2128 			if (besta) {
2129 				if (blen == -1)
2130 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2131 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2132 				if (tlen > blen) {
2133 					blen = tlen;
2134 					besta = (struct in6_ifaddr *)ifa;
2135 				}
2136 			} else
2137 				besta = (struct in6_ifaddr *)ifa;
2138 		}
2139 	}
2140 	if (besta) {
2141 		ifa_ref(&besta->ia_ifa);
2142 		IF_ADDR_UNLOCK(ifp);
2143 		return (besta);
2144 	}
2145 	IF_ADDR_UNLOCK(ifp);
2146 
2147 	IN6_IFADDR_RLOCK();
2148 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2149 		if (ifa->ifa_addr->sa_family != AF_INET6)
2150 			continue;
2151 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2152 			continue; /* XXX: is there any case to allow anycast? */
2153 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2154 			continue; /* don't use this interface */
2155 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2156 			continue;
2157 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2158 			if (V_ip6_use_deprecated)
2159 				dep[1] = (struct in6_ifaddr *)ifa;
2160 			continue;
2161 		}
2162 
2163 		if (ifa != NULL)
2164 			ifa_ref(ifa);
2165 		IN6_IFADDR_RUNLOCK();
2166 		return (struct in6_ifaddr *)ifa;
2167 	}
2168 	IN6_IFADDR_RUNLOCK();
2169 
2170 	/* use the last-resort values, that are, deprecated addresses */
2171 	if (dep[0])
2172 		return dep[0];
2173 	if (dep[1])
2174 		return dep[1];
2175 
2176 	return NULL;
2177 }
2178 
2179 /*
2180  * perform DAD when interface becomes IFF_UP.
2181  */
2182 void
2183 in6_if_up(struct ifnet *ifp)
2184 {
2185 	struct ifaddr *ifa;
2186 	struct in6_ifaddr *ia;
2187 
2188 	IF_ADDR_LOCK(ifp);
2189 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2190 		if (ifa->ifa_addr->sa_family != AF_INET6)
2191 			continue;
2192 		ia = (struct in6_ifaddr *)ifa;
2193 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2194 			/*
2195 			 * The TENTATIVE flag was likely set by hand
2196 			 * beforehand, implicitly indicating the need for DAD.
2197 			 * We may be able to skip the random delay in this
2198 			 * case, but we impose delays just in case.
2199 			 */
2200 			nd6_dad_start(ifa,
2201 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2202 		}
2203 	}
2204 	IF_ADDR_UNLOCK(ifp);
2205 
2206 	/*
2207 	 * special cases, like 6to4, are handled in in6_ifattach
2208 	 */
2209 	in6_ifattach(ifp, NULL);
2210 }
2211 
2212 int
2213 in6if_do_dad(struct ifnet *ifp)
2214 {
2215 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2216 		return (0);
2217 
2218 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2219 		return (0);
2220 
2221 	switch (ifp->if_type) {
2222 #ifdef IFT_DUMMY
2223 	case IFT_DUMMY:
2224 #endif
2225 	case IFT_FAITH:
2226 		/*
2227 		 * These interfaces do not have the IFF_LOOPBACK flag,
2228 		 * but loop packets back.  We do not have to do DAD on such
2229 		 * interfaces.  We should even omit it, because loop-backed
2230 		 * NS would confuse the DAD procedure.
2231 		 */
2232 		return (0);
2233 	default:
2234 		/*
2235 		 * Our DAD routine requires the interface up and running.
2236 		 * However, some interfaces can be up before the RUNNING
2237 		 * status.  Additionaly, users may try to assign addresses
2238 		 * before the interface becomes up (or running).
2239 		 * We simply skip DAD in such a case as a work around.
2240 		 * XXX: we should rather mark "tentative" on such addresses,
2241 		 * and do DAD after the interface becomes ready.
2242 		 */
2243 		if (!((ifp->if_flags & IFF_UP) &&
2244 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2245 			return (0);
2246 
2247 		return (1);
2248 	}
2249 }
2250 
2251 /*
2252  * Calculate max IPv6 MTU through all the interfaces and store it
2253  * to in6_maxmtu.
2254  */
2255 void
2256 in6_setmaxmtu(void)
2257 {
2258 	unsigned long maxmtu = 0;
2259 	struct ifnet *ifp;
2260 
2261 	IFNET_RLOCK_NOSLEEP();
2262 	for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
2263 	    ifp = TAILQ_NEXT(ifp, if_list)) {
2264 		/* this function can be called during ifnet initialization */
2265 		if (!ifp->if_afdata[AF_INET6])
2266 			continue;
2267 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2268 		    IN6_LINKMTU(ifp) > maxmtu)
2269 			maxmtu = IN6_LINKMTU(ifp);
2270 	}
2271 	IFNET_RUNLOCK_NOSLEEP();
2272 	if (maxmtu)	     /* update only when maxmtu is positive */
2273 		V_in6_maxmtu = maxmtu;
2274 }
2275 
2276 /*
2277  * Provide the length of interface identifiers to be used for the link attached
2278  * to the given interface.  The length should be defined in "IPv6 over
2279  * xxx-link" document.  Note that address architecture might also define
2280  * the length for a particular set of address prefixes, regardless of the
2281  * link type.  As clarified in rfc2462bis, those two definitions should be
2282  * consistent, and those really are as of August 2004.
2283  */
2284 int
2285 in6_if2idlen(struct ifnet *ifp)
2286 {
2287 	switch (ifp->if_type) {
2288 	case IFT_ETHER:		/* RFC2464 */
2289 #ifdef IFT_PROPVIRTUAL
2290 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2291 #endif
2292 #ifdef IFT_L2VLAN
2293 	case IFT_L2VLAN:	/* ditto */
2294 #endif
2295 #ifdef IFT_IEEE80211
2296 	case IFT_IEEE80211:	/* ditto */
2297 #endif
2298 #ifdef IFT_MIP
2299 	case IFT_MIP:	/* ditto */
2300 #endif
2301 		return (64);
2302 	case IFT_FDDI:		/* RFC2467 */
2303 		return (64);
2304 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2305 		return (64);
2306 	case IFT_PPP:		/* RFC2472 */
2307 		return (64);
2308 	case IFT_ARCNET:	/* RFC2497 */
2309 		return (64);
2310 	case IFT_FRELAY:	/* RFC2590 */
2311 		return (64);
2312 	case IFT_IEEE1394:	/* RFC3146 */
2313 		return (64);
2314 	case IFT_GIF:
2315 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2316 	case IFT_LOOP:
2317 		return (64);	/* XXX: is this really correct? */
2318 	default:
2319 		/*
2320 		 * Unknown link type:
2321 		 * It might be controversial to use the today's common constant
2322 		 * of 64 for these cases unconditionally.  For full compliance,
2323 		 * we should return an error in this case.  On the other hand,
2324 		 * if we simply miss the standard for the link type or a new
2325 		 * standard is defined for a new link type, the IFID length
2326 		 * is very likely to be the common constant.  As a compromise,
2327 		 * we always use the constant, but make an explicit notice
2328 		 * indicating the "unknown" case.
2329 		 */
2330 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2331 		return (64);
2332 	}
2333 }
2334 
2335 #include <sys/sysctl.h>
2336 
2337 struct in6_llentry {
2338 	struct llentry		base;
2339 	struct sockaddr_in6	l3_addr6;
2340 };
2341 
2342 static struct llentry *
2343 in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2344 {
2345 	struct in6_llentry *lle;
2346 
2347 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE,
2348 	    M_DONTWAIT | M_ZERO);
2349 	if (lle == NULL)		/* NB: caller generates msg */
2350 		return NULL;
2351 
2352 	callout_init(&lle->base.ln_timer_ch, CALLOUT_MPSAFE);
2353 	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2354 	lle->base.lle_refcnt = 1;
2355 	LLE_LOCK_INIT(&lle->base);
2356 	return &lle->base;
2357 }
2358 
2359 /*
2360  * Deletes an address from the address table.
2361  * This function is called by the timer functions
2362  * such as arptimer() and nd6_llinfo_timer(), and
2363  * the caller does the locking.
2364  */
2365 static void
2366 in6_lltable_free(struct lltable *llt, struct llentry *lle)
2367 {
2368 	LLE_WUNLOCK(lle);
2369 	LLE_LOCK_DESTROY(lle);
2370 	free(lle, M_LLTABLE);
2371 }
2372 
2373 static void
2374 in6_lltable_prefix_free(struct lltable *llt,
2375 			const struct sockaddr *prefix,
2376 			const struct sockaddr *mask)
2377 {
2378 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2379 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2380 	struct llentry *lle, *next;
2381 	register int i;
2382 
2383 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
2384 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2385 			if (IN6_ARE_MASKED_ADDR_EQUAL(
2386 				    &((struct sockaddr_in6 *)L3_ADDR(lle))->sin6_addr,
2387 				    &pfx->sin6_addr,
2388 				    &msk->sin6_addr)) {
2389 				int canceled;
2390 
2391 				canceled = callout_drain(&lle->la_timer);
2392 				LLE_WLOCK(lle);
2393 				if (canceled)
2394 					LLE_REMREF(lle);
2395 				llentry_free(lle);
2396 			}
2397 		}
2398 	}
2399 }
2400 
2401 static int
2402 in6_lltable_rtcheck(struct ifnet *ifp,
2403 		    u_int flags,
2404 		    const struct sockaddr *l3addr)
2405 {
2406 	struct rtentry *rt;
2407 	char ip6buf[INET6_ADDRSTRLEN];
2408 
2409 	KASSERT(l3addr->sa_family == AF_INET6,
2410 	    ("sin_family %d", l3addr->sa_family));
2411 
2412 	/* XXX rtalloc1 should take a const param */
2413 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
2414 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2415 		struct ifaddr *ifa;
2416 		/*
2417 		 * Create an ND6 cache for an IPv6 neighbor
2418 		 * that is not covered by our own prefix.
2419 		 */
2420 		/* XXX ifaof_ifpforaddr should take a const param */
2421 		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2422 		if (ifa != NULL) {
2423 			ifa_free(ifa);
2424 			if (rt != NULL)
2425 				RTFREE_LOCKED(rt);
2426 			return 0;
2427 		}
2428 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2429 		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2430 		if (rt != NULL)
2431 			RTFREE_LOCKED(rt);
2432 		return EINVAL;
2433 	}
2434 	RTFREE_LOCKED(rt);
2435 	return 0;
2436 }
2437 
2438 static struct llentry *
2439 in6_lltable_lookup(struct lltable *llt, u_int flags,
2440 	const struct sockaddr *l3addr)
2441 {
2442 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2443 	struct ifnet *ifp = llt->llt_ifp;
2444 	struct llentry *lle;
2445 	struct llentries *lleh;
2446 	u_int hashkey;
2447 
2448 	IF_AFDATA_LOCK_ASSERT(ifp);
2449 	KASSERT(l3addr->sa_family == AF_INET6,
2450 	    ("sin_family %d", l3addr->sa_family));
2451 
2452 	hashkey = sin6->sin6_addr.s6_addr32[3];
2453 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2454 	LIST_FOREACH(lle, lleh, lle_next) {
2455 		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2456 		if (lle->la_flags & LLE_DELETED)
2457 			continue;
2458 		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2459 			 sizeof(struct in6_addr)) == 0)
2460 			break;
2461 	}
2462 
2463 	if (lle == NULL) {
2464 		if (!(flags & LLE_CREATE))
2465 			return (NULL);
2466 		/*
2467 		 * A route that covers the given address must have
2468 		 * been installed 1st because we are doing a resolution,
2469 		 * verify this.
2470 		 */
2471 		if (!(flags & LLE_IFADDR) &&
2472 		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2473 			return NULL;
2474 
2475 		lle = in6_lltable_new(l3addr, flags);
2476 		if (lle == NULL) {
2477 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2478 			return NULL;
2479 		}
2480 		lle->la_flags = flags & ~LLE_CREATE;
2481 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2482 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2483 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2484 		}
2485 
2486 		lle->lle_tbl  = llt;
2487 		lle->lle_head = lleh;
2488 		LIST_INSERT_HEAD(lleh, lle, lle_next);
2489 	} else if (flags & LLE_DELETE) {
2490 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2491 			LLE_WLOCK(lle);
2492 			lle->la_flags = LLE_DELETED;
2493 			LLE_WUNLOCK(lle);
2494 #ifdef DIAGNOSTIC
2495 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2496 #endif
2497 		}
2498 		lle = (void *)-1;
2499 	}
2500 	if (LLE_IS_VALID(lle)) {
2501 		if (flags & LLE_EXCLUSIVE)
2502 			LLE_WLOCK(lle);
2503 		else
2504 			LLE_RLOCK(lle);
2505 	}
2506 	return (lle);
2507 }
2508 
2509 static int
2510 in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2511 {
2512 	struct ifnet *ifp = llt->llt_ifp;
2513 	struct llentry *lle;
2514 	/* XXX stack use */
2515 	struct {
2516 		struct rt_msghdr	rtm;
2517 		struct sockaddr_in6	sin6;
2518 		/*
2519 		 * ndp.c assumes that sdl is word aligned
2520 		 */
2521 #ifdef __LP64__
2522 		uint32_t		pad;
2523 #endif
2524 		struct sockaddr_dl	sdl;
2525 	} ndpc;
2526 	int i, error;
2527 
2528 	if (ifp->if_flags & IFF_LOOPBACK)
2529 		return 0;
2530 
2531 	LLTABLE_LOCK_ASSERT();
2532 
2533 	error = 0;
2534 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2535 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2536 			struct sockaddr_dl *sdl;
2537 
2538 			/* skip deleted or invalid entries */
2539 			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2540 				continue;
2541 			/* Skip if jailed and not a valid IP of the prison. */
2542 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2543 				continue;
2544 			/*
2545 			 * produce a msg made of:
2546 			 *  struct rt_msghdr;
2547 			 *  struct sockaddr_in6 (IPv6)
2548 			 *  struct sockaddr_dl;
2549 			 */
2550 			bzero(&ndpc, sizeof(ndpc));
2551 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2552 			ndpc.rtm.rtm_version = RTM_VERSION;
2553 			ndpc.rtm.rtm_type = RTM_GET;
2554 			ndpc.rtm.rtm_flags = RTF_UP;
2555 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2556 			ndpc.sin6.sin6_family = AF_INET6;
2557 			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2558 			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2559 
2560 			/* publish */
2561 			if (lle->la_flags & LLE_PUB)
2562 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2563 
2564 			sdl = &ndpc.sdl;
2565 			sdl->sdl_family = AF_LINK;
2566 			sdl->sdl_len = sizeof(*sdl);
2567 			sdl->sdl_alen = ifp->if_addrlen;
2568 			sdl->sdl_index = ifp->if_index;
2569 			sdl->sdl_type = ifp->if_type;
2570 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2571 			ndpc.rtm.rtm_rmx.rmx_expire =
2572 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2573 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2574 			if (lle->la_flags & LLE_STATIC)
2575 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2576 			ndpc.rtm.rtm_index = ifp->if_index;
2577 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2578 			if (error)
2579 				break;
2580 		}
2581 	}
2582 	return error;
2583 }
2584 
2585 void *
2586 in6_domifattach(struct ifnet *ifp)
2587 {
2588 	struct in6_ifextra *ext;
2589 
2590 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2591 	bzero(ext, sizeof(*ext));
2592 
2593 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2594 	    M_IFADDR, M_WAITOK);
2595 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2596 
2597 	ext->icmp6_ifstat =
2598 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2599 	    M_IFADDR, M_WAITOK);
2600 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2601 
2602 	ext->nd_ifinfo = nd6_ifattach(ifp);
2603 	ext->scope6_id = scope6_ifattach(ifp);
2604 	ext->lltable = lltable_init(ifp, AF_INET6);
2605 	if (ext->lltable != NULL) {
2606 		ext->lltable->llt_new = in6_lltable_new;
2607 		ext->lltable->llt_free = in6_lltable_free;
2608 		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2609 		ext->lltable->llt_rtcheck = in6_lltable_rtcheck;
2610 		ext->lltable->llt_lookup = in6_lltable_lookup;
2611 		ext->lltable->llt_dump = in6_lltable_dump;
2612 	}
2613 
2614 	ext->mld_ifinfo = mld_domifattach(ifp);
2615 
2616 	return ext;
2617 }
2618 
2619 void
2620 in6_domifdetach(struct ifnet *ifp, void *aux)
2621 {
2622 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2623 
2624 	mld_domifdetach(ifp);
2625 	scope6_ifdetach(ext->scope6_id);
2626 	nd6_ifdetach(ext->nd_ifinfo);
2627 	lltable_free(ext->lltable);
2628 	free(ext->in6_ifstat, M_IFADDR);
2629 	free(ext->icmp6_ifstat, M_IFADDR);
2630 	free(ext, M_IFADDR);
2631 }
2632 
2633 /*
2634  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2635  * v4 mapped addr or v4 compat addr
2636  */
2637 void
2638 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2639 {
2640 
2641 	bzero(sin, sizeof(*sin));
2642 	sin->sin_len = sizeof(struct sockaddr_in);
2643 	sin->sin_family = AF_INET;
2644 	sin->sin_port = sin6->sin6_port;
2645 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2646 }
2647 
2648 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2649 void
2650 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2651 {
2652 	bzero(sin6, sizeof(*sin6));
2653 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2654 	sin6->sin6_family = AF_INET6;
2655 	sin6->sin6_port = sin->sin_port;
2656 	sin6->sin6_addr.s6_addr32[0] = 0;
2657 	sin6->sin6_addr.s6_addr32[1] = 0;
2658 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2659 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2660 }
2661 
2662 /* Convert sockaddr_in6 into sockaddr_in. */
2663 void
2664 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2665 {
2666 	struct sockaddr_in *sin_p;
2667 	struct sockaddr_in6 sin6;
2668 
2669 	/*
2670 	 * Save original sockaddr_in6 addr and convert it
2671 	 * to sockaddr_in.
2672 	 */
2673 	sin6 = *(struct sockaddr_in6 *)nam;
2674 	sin_p = (struct sockaddr_in *)nam;
2675 	in6_sin6_2_sin(sin_p, &sin6);
2676 }
2677 
2678 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2679 void
2680 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2681 {
2682 	struct sockaddr_in *sin_p;
2683 	struct sockaddr_in6 *sin6_p;
2684 
2685 	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
2686 	       M_WAITOK);
2687 	sin_p = (struct sockaddr_in *)*nam;
2688 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2689 	free(*nam, M_SONAME);
2690 	*nam = (struct sockaddr *)sin6_p;
2691 }
2692