xref: /freebsd/sys/netinet6/in6.c (revision 5861f9665471e98e544f6fa3ce73c4912229ff82)
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_inet.h"
67 #include "opt_inet6.h"
68 
69 #include <sys/param.h>
70 #include <sys/errno.h>
71 #include <sys/jail.h>
72 #include <sys/malloc.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/sockio.h>
76 #include <sys/systm.h>
77 #include <sys/priv.h>
78 #include <sys/proc.h>
79 #include <sys/time.h>
80 #include <sys/kernel.h>
81 #include <sys/syslog.h>
82 #include <sys/vimage.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 #include <net/vnet.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_var.h>
92 #include <net/if_llatbl.h>
93 #include <netinet/if_ether.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/ip.h>
96 #include <netinet/in_pcb.h>
97 
98 #include <netinet/ip6.h>
99 #include <netinet6/ip6_var.h>
100 #include <netinet6/nd6.h>
101 #include <netinet6/mld6_var.h>
102 #include <netinet6/ip6_mroute.h>
103 #include <netinet6/in6_ifattach.h>
104 #include <netinet6/scope6_var.h>
105 #include <netinet6/in6_pcb.h>
106 #include <netinet6/vinet6.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 int
181 in6_control(struct socket *so, u_long cmd, caddr_t data,
182     struct ifnet *ifp, struct thread *td)
183 {
184 	INIT_VNET_INET6(curvnet);
185 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
186 	struct	in6_ifaddr *ia = NULL;
187 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
188 	struct sockaddr_in6 *sa6;
189 	int error;
190 
191 	switch (cmd) {
192 	case SIOCGETSGCNT_IN6:
193 	case SIOCGETMIFCNT_IN6:
194 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
195 	}
196 
197 	switch(cmd) {
198 	case SIOCAADDRCTL_POLICY:
199 	case SIOCDADDRCTL_POLICY:
200 		if (td != NULL) {
201 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
202 			if (error)
203 				return (error);
204 		}
205 		return (in6_src_ioctl(cmd, data));
206 	}
207 
208 	if (ifp == NULL)
209 		return (EOPNOTSUPP);
210 
211 	switch (cmd) {
212 	case SIOCSNDFLUSH_IN6:
213 	case SIOCSPFXFLUSH_IN6:
214 	case SIOCSRTRFLUSH_IN6:
215 	case SIOCSDEFIFACE_IN6:
216 	case SIOCSIFINFO_FLAGS:
217 	case SIOCSIFINFO_IN6:
218 		if (td != NULL) {
219 			error = priv_check(td, PRIV_NETINET_ND6);
220 			if (error)
221 				return (error);
222 		}
223 		/* FALLTHROUGH */
224 	case OSIOCGIFINFO_IN6:
225 	case SIOCGIFINFO_IN6:
226 	case SIOCGDRLST_IN6:
227 	case SIOCGPRLST_IN6:
228 	case SIOCGNBRINFO_IN6:
229 	case SIOCGDEFIFACE_IN6:
230 		return (nd6_ioctl(cmd, data, ifp));
231 	}
232 
233 	switch (cmd) {
234 	case SIOCSIFPREFIX_IN6:
235 	case SIOCDIFPREFIX_IN6:
236 	case SIOCAIFPREFIX_IN6:
237 	case SIOCCIFPREFIX_IN6:
238 	case SIOCSGIFPREFIX_IN6:
239 	case SIOCGIFPREFIX_IN6:
240 		log(LOG_NOTICE,
241 		    "prefix ioctls are now invalidated. "
242 		    "please use ifconfig.\n");
243 		return (EOPNOTSUPP);
244 	}
245 
246 	switch (cmd) {
247 	case SIOCSSCOPE6:
248 		if (td != NULL) {
249 			error = priv_check(td, PRIV_NETINET_SCOPE6);
250 			if (error)
251 				return (error);
252 		}
253 		return (scope6_set(ifp,
254 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
255 	case SIOCGSCOPE6:
256 		return (scope6_get(ifp,
257 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
258 	case SIOCGSCOPE6DEF:
259 		return (scope6_get_default((struct scope6_id *)
260 		    ifr->ifr_ifru.ifru_scope_id));
261 	}
262 
263 	switch (cmd) {
264 	case SIOCALIFADDR:
265 		if (td != NULL) {
266 			error = priv_check(td, PRIV_NET_ADDIFADDR);
267 			if (error)
268 				return (error);
269 		}
270 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
271 
272 	case SIOCDLIFADDR:
273 		if (td != NULL) {
274 			error = priv_check(td, PRIV_NET_DELIFADDR);
275 			if (error)
276 				return (error);
277 		}
278 		/* FALLTHROUGH */
279 	case SIOCGLIFADDR:
280 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
281 	}
282 
283 	/*
284 	 * Find address for this interface, if it exists.
285 	 *
286 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
287 	 * only, and used the first interface address as the target of other
288 	 * operations (without checking ifra_addr).  This was because netinet
289 	 * code/API assumed at most 1 interface address per interface.
290 	 * Since IPv6 allows a node to assign multiple addresses
291 	 * on a single interface, we almost always look and check the
292 	 * presence of ifra_addr, and reject invalid ones here.
293 	 * It also decreases duplicated code among SIOC*_IN6 operations.
294 	 */
295 	switch (cmd) {
296 	case SIOCAIFADDR_IN6:
297 	case SIOCSIFPHYADDR_IN6:
298 		sa6 = &ifra->ifra_addr;
299 		break;
300 	case SIOCSIFADDR_IN6:
301 	case SIOCGIFADDR_IN6:
302 	case SIOCSIFDSTADDR_IN6:
303 	case SIOCSIFNETMASK_IN6:
304 	case SIOCGIFDSTADDR_IN6:
305 	case SIOCGIFNETMASK_IN6:
306 	case SIOCDIFADDR_IN6:
307 	case SIOCGIFPSRCADDR_IN6:
308 	case SIOCGIFPDSTADDR_IN6:
309 	case SIOCGIFAFLAG_IN6:
310 	case SIOCSNDFLUSH_IN6:
311 	case SIOCSPFXFLUSH_IN6:
312 	case SIOCSRTRFLUSH_IN6:
313 	case SIOCGIFALIFETIME_IN6:
314 	case SIOCSIFALIFETIME_IN6:
315 	case SIOCGIFSTAT_IN6:
316 	case SIOCGIFSTAT_ICMP6:
317 		sa6 = &ifr->ifr_addr;
318 		break;
319 	default:
320 		sa6 = NULL;
321 		break;
322 	}
323 	if (sa6 && sa6->sin6_family == AF_INET6) {
324 		if (sa6->sin6_scope_id != 0)
325 			error = sa6_embedscope(sa6, 0);
326 		else
327 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
328 		if (error != 0)
329 			return (error);
330 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
331 		    &sa6->sin6_addr)) != 0)
332 			return (error);
333 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
334 	} else
335 		ia = NULL;
336 
337 	switch (cmd) {
338 	case SIOCSIFADDR_IN6:
339 	case SIOCSIFDSTADDR_IN6:
340 	case SIOCSIFNETMASK_IN6:
341 		/*
342 		 * Since IPv6 allows a node to assign multiple addresses
343 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
344 		 */
345 		/* we decided to obsolete this command (20000704) */
346 		error = EINVAL;
347 		goto out;
348 
349 	case SIOCDIFADDR_IN6:
350 		/*
351 		 * for IPv4, we look for existing in_ifaddr here to allow
352 		 * "ifconfig if0 delete" to remove the first IPv4 address on
353 		 * the interface.  For IPv6, as the spec allows multiple
354 		 * interface address from the day one, we consider "remove the
355 		 * first one" semantics to be not preferable.
356 		 */
357 		if (ia == NULL) {
358 			error = EADDRNOTAVAIL;
359 			goto out;
360 		}
361 		/* FALLTHROUGH */
362 	case SIOCAIFADDR_IN6:
363 		/*
364 		 * We always require users to specify a valid IPv6 address for
365 		 * the corresponding operation.
366 		 */
367 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
368 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
369 			error = EAFNOSUPPORT;
370 			goto out;
371 		}
372 
373 		if (td != NULL) {
374 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
375 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
376 			if (error)
377 				goto out;
378 		}
379 		break;
380 
381 	case SIOCGIFADDR_IN6:
382 		/* This interface is basically deprecated. use SIOCGIFCONF. */
383 		/* FALLTHROUGH */
384 	case SIOCGIFAFLAG_IN6:
385 	case SIOCGIFNETMASK_IN6:
386 	case SIOCGIFDSTADDR_IN6:
387 	case SIOCGIFALIFETIME_IN6:
388 		/* must think again about its semantics */
389 		if (ia == NULL) {
390 			error = EADDRNOTAVAIL;
391 			goto out;
392 		}
393 		break;
394 
395 	case SIOCSIFALIFETIME_IN6:
396 	    {
397 		struct in6_addrlifetime *lt;
398 
399 		if (td != NULL) {
400 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
401 			if (error)
402 				goto out;
403 		}
404 		if (ia == NULL) {
405 			error = EADDRNOTAVAIL;
406 			goto out;
407 		}
408 		/* sanity for overflow - beware unsigned */
409 		lt = &ifr->ifr_ifru.ifru_lifetime;
410 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
411 		    lt->ia6t_vltime + time_second < time_second) {
412 			error = EINVAL;
413 			goto out;
414 		}
415 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
416 		    lt->ia6t_pltime + time_second < time_second) {
417 			error = EINVAL;
418 			goto out;
419 		}
420 		break;
421 	    }
422 	}
423 
424 	switch (cmd) {
425 	case SIOCGIFADDR_IN6:
426 		ifr->ifr_addr = ia->ia_addr;
427 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
428 			goto out;
429 		break;
430 
431 	case SIOCGIFDSTADDR_IN6:
432 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
433 			error = EINVAL;
434 			goto out;
435 		}
436 		/*
437 		 * XXX: should we check if ifa_dstaddr is NULL and return
438 		 * an error?
439 		 */
440 		ifr->ifr_dstaddr = ia->ia_dstaddr;
441 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
442 			goto out;
443 		break;
444 
445 	case SIOCGIFNETMASK_IN6:
446 		ifr->ifr_addr = ia->ia_prefixmask;
447 		break;
448 
449 	case SIOCGIFAFLAG_IN6:
450 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
451 		break;
452 
453 	case SIOCGIFSTAT_IN6:
454 		if (ifp == NULL) {
455 			error = EINVAL;
456 			goto out;
457 		}
458 		bzero(&ifr->ifr_ifru.ifru_stat,
459 		    sizeof(ifr->ifr_ifru.ifru_stat));
460 		ifr->ifr_ifru.ifru_stat =
461 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
462 		break;
463 
464 	case SIOCGIFSTAT_ICMP6:
465 		if (ifp == NULL) {
466 			error = EINVAL;
467 			goto out;
468 		}
469 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
470 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
471 		ifr->ifr_ifru.ifru_icmp6stat =
472 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
473 		break;
474 
475 	case SIOCGIFALIFETIME_IN6:
476 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
477 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
478 			time_t maxexpire;
479 			struct in6_addrlifetime *retlt =
480 			    &ifr->ifr_ifru.ifru_lifetime;
481 
482 			/*
483 			 * XXX: adjust expiration time assuming time_t is
484 			 * signed.
485 			 */
486 			maxexpire = (-1) &
487 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
488 			if (ia->ia6_lifetime.ia6t_vltime <
489 			    maxexpire - ia->ia6_updatetime) {
490 				retlt->ia6t_expire = ia->ia6_updatetime +
491 				    ia->ia6_lifetime.ia6t_vltime;
492 			} else
493 				retlt->ia6t_expire = maxexpire;
494 		}
495 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
496 			time_t maxexpire;
497 			struct in6_addrlifetime *retlt =
498 			    &ifr->ifr_ifru.ifru_lifetime;
499 
500 			/*
501 			 * XXX: adjust expiration time assuming time_t is
502 			 * signed.
503 			 */
504 			maxexpire = (-1) &
505 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
506 			if (ia->ia6_lifetime.ia6t_pltime <
507 			    maxexpire - ia->ia6_updatetime) {
508 				retlt->ia6t_preferred = ia->ia6_updatetime +
509 				    ia->ia6_lifetime.ia6t_pltime;
510 			} else
511 				retlt->ia6t_preferred = maxexpire;
512 		}
513 		break;
514 
515 	case SIOCSIFALIFETIME_IN6:
516 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
517 		/* for sanity */
518 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
519 			ia->ia6_lifetime.ia6t_expire =
520 				time_second + ia->ia6_lifetime.ia6t_vltime;
521 		} else
522 			ia->ia6_lifetime.ia6t_expire = 0;
523 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
524 			ia->ia6_lifetime.ia6t_preferred =
525 				time_second + ia->ia6_lifetime.ia6t_pltime;
526 		} else
527 			ia->ia6_lifetime.ia6t_preferred = 0;
528 		break;
529 
530 	case SIOCAIFADDR_IN6:
531 	{
532 		int i;
533 		struct nd_prefixctl pr0;
534 		struct nd_prefix *pr;
535 
536 		/*
537 		 * first, make or update the interface address structure,
538 		 * and link it to the list.
539 		 */
540 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
541 			goto out;
542 		if (ia != NULL)
543 			ifa_free(&ia->ia_ifa);
544 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
545 		    == NULL) {
546 			/*
547 			 * this can happen when the user specify the 0 valid
548 			 * lifetime.
549 			 */
550 			break;
551 		}
552 
553 		/*
554 		 * then, make the prefix on-link on the interface.
555 		 * XXX: we'd rather create the prefix before the address, but
556 		 * we need at least one address to install the corresponding
557 		 * interface route, so we configure the address first.
558 		 */
559 
560 		/*
561 		 * convert mask to prefix length (prefixmask has already
562 		 * been validated in in6_update_ifa().
563 		 */
564 		bzero(&pr0, sizeof(pr0));
565 		pr0.ndpr_ifp = ifp;
566 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
567 		    NULL);
568 		if (pr0.ndpr_plen == 128) {
569 			break;	/* we don't need to install a host route. */
570 		}
571 		pr0.ndpr_prefix = ifra->ifra_addr;
572 		/* apply the mask for safety. */
573 		for (i = 0; i < 4; i++) {
574 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
575 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
576 		}
577 		/*
578 		 * XXX: since we don't have an API to set prefix (not address)
579 		 * lifetimes, we just use the same lifetimes as addresses.
580 		 * The (temporarily) installed lifetimes can be overridden by
581 		 * later advertised RAs (when accept_rtadv is non 0), which is
582 		 * an intended behavior.
583 		 */
584 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
585 		pr0.ndpr_raf_auto =
586 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
587 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
588 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
589 
590 		/* add the prefix if not yet. */
591 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
592 			/*
593 			 * nd6_prelist_add will install the corresponding
594 			 * interface route.
595 			 */
596 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
597 				goto out;
598 			if (pr == NULL) {
599 				log(LOG_ERR, "nd6_prelist_add succeeded but "
600 				    "no prefix\n");
601 				error = EINVAL;
602 				goto out;
603 			}
604 		}
605 
606 		/* relate the address to the prefix */
607 		if (ia->ia6_ndpr == NULL) {
608 			ia->ia6_ndpr = pr;
609 			pr->ndpr_refcnt++;
610 
611 			/*
612 			 * If this is the first autoconf address from the
613 			 * prefix, create a temporary address as well
614 			 * (when required).
615 			 */
616 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
617 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
618 				int e;
619 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
620 					log(LOG_NOTICE, "in6_control: failed "
621 					    "to create a temporary address, "
622 					    "errno=%d\n", e);
623 				}
624 			}
625 		}
626 
627 		/*
628 		 * this might affect the status of autoconfigured addresses,
629 		 * that is, this address might make other addresses detached.
630 		 */
631 		pfxlist_onlink_check();
632 		if (error == 0 && ia)
633 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
634 		break;
635 	}
636 
637 	case SIOCDIFADDR_IN6:
638 	{
639 		struct nd_prefix *pr;
640 
641 		/*
642 		 * If the address being deleted is the only one that owns
643 		 * the corresponding prefix, expire the prefix as well.
644 		 * XXX: theoretically, we don't have to worry about such
645 		 * relationship, since we separate the address management
646 		 * and the prefix management.  We do this, however, to provide
647 		 * as much backward compatibility as possible in terms of
648 		 * the ioctl operation.
649 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
650 		 */
651 		pr = ia->ia6_ndpr;
652 		in6_purgeaddr(&ia->ia_ifa);
653 		if (pr && pr->ndpr_refcnt == 0)
654 			prelist_remove(pr);
655 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
656 		break;
657 	}
658 
659 	default:
660 		if (ifp == NULL || ifp->if_ioctl == 0) {
661 			error = EOPNOTSUPP;
662 			goto out;
663 		}
664 		error = (*ifp->if_ioctl)(ifp, cmd, data);
665 		goto out;
666 	}
667 
668 	error = 0;
669 out:
670 	if (ia != NULL)
671 		ifa_free(&ia->ia_ifa);
672 	return (error);
673 }
674 
675 /*
676  * Update parameters of an IPv6 interface address.
677  * If necessary, a new entry is created and linked into address chains.
678  * This function is separated from in6_control().
679  * XXX: should this be performed under splnet()?
680  */
681 int
682 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
683     struct in6_ifaddr *ia, int flags)
684 {
685 	INIT_VNET_INET6(ifp->if_vnet);
686 	int error = 0, hostIsNew = 0, plen = -1;
687 	struct sockaddr_in6 dst6;
688 	struct in6_addrlifetime *lt;
689 	struct in6_multi_mship *imm;
690 	struct in6_multi *in6m_sol;
691 	struct rtentry *rt;
692 	int delay;
693 	char ip6buf[INET6_ADDRSTRLEN];
694 
695 	/* Validate parameters */
696 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
697 		return (EINVAL);
698 
699 	/*
700 	 * The destination address for a p2p link must have a family
701 	 * of AF_UNSPEC or AF_INET6.
702 	 */
703 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
704 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
705 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
706 		return (EAFNOSUPPORT);
707 	/*
708 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
709 	 * does not carry fields other than sin6_len.
710 	 */
711 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
712 		return (EINVAL);
713 	/*
714 	 * Because the IPv6 address architecture is classless, we require
715 	 * users to specify a (non 0) prefix length (mask) for a new address.
716 	 * We also require the prefix (when specified) mask is valid, and thus
717 	 * reject a non-consecutive mask.
718 	 */
719 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
720 		return (EINVAL);
721 	if (ifra->ifra_prefixmask.sin6_len != 0) {
722 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
723 		    (u_char *)&ifra->ifra_prefixmask +
724 		    ifra->ifra_prefixmask.sin6_len);
725 		if (plen <= 0)
726 			return (EINVAL);
727 	} else {
728 		/*
729 		 * In this case, ia must not be NULL.  We just use its prefix
730 		 * length.
731 		 */
732 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
733 	}
734 	/*
735 	 * If the destination address on a p2p interface is specified,
736 	 * and the address is a scoped one, validate/set the scope
737 	 * zone identifier.
738 	 */
739 	dst6 = ifra->ifra_dstaddr;
740 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
741 	    (dst6.sin6_family == AF_INET6)) {
742 		struct in6_addr in6_tmp;
743 		u_int32_t zoneid;
744 
745 		in6_tmp = dst6.sin6_addr;
746 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
747 			return (EINVAL); /* XXX: should be impossible */
748 
749 		if (dst6.sin6_scope_id != 0) {
750 			if (dst6.sin6_scope_id != zoneid)
751 				return (EINVAL);
752 		} else		/* user omit to specify the ID. */
753 			dst6.sin6_scope_id = zoneid;
754 
755 		/* convert into the internal form */
756 		if (sa6_embedscope(&dst6, 0))
757 			return (EINVAL); /* XXX: should be impossible */
758 	}
759 	/*
760 	 * The destination address can be specified only for a p2p or a
761 	 * loopback interface.  If specified, the corresponding prefix length
762 	 * must be 128.
763 	 */
764 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
765 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
766 			/* XXX: noisy message */
767 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
768 			    "be specified for a p2p or a loopback IF only\n"));
769 			return (EINVAL);
770 		}
771 		if (plen != 128) {
772 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
773 			    "be 128 when dstaddr is specified\n"));
774 			return (EINVAL);
775 		}
776 	}
777 	/* lifetime consistency check */
778 	lt = &ifra->ifra_lifetime;
779 	if (lt->ia6t_pltime > lt->ia6t_vltime)
780 		return (EINVAL);
781 	if (lt->ia6t_vltime == 0) {
782 		/*
783 		 * the following log might be noisy, but this is a typical
784 		 * configuration mistake or a tool's bug.
785 		 */
786 		nd6log((LOG_INFO,
787 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
788 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
789 
790 		if (ia == NULL)
791 			return (0); /* there's nothing to do */
792 	}
793 
794 	/*
795 	 * If this is a new address, allocate a new ifaddr and link it
796 	 * into chains.
797 	 */
798 	if (ia == NULL) {
799 		hostIsNew = 1;
800 		/*
801 		 * When in6_update_ifa() is called in a process of a received
802 		 * RA, it is called under an interrupt context.  So, we should
803 		 * call malloc with M_NOWAIT.
804 		 */
805 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
806 		    M_NOWAIT);
807 		if (ia == NULL)
808 			return (ENOBUFS);
809 		bzero((caddr_t)ia, sizeof(*ia));
810 		ifa_init(&ia->ia_ifa);
811 		LIST_INIT(&ia->ia6_memberships);
812 		/* Initialize the address and masks, and put time stamp */
813 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
814 		ia->ia_addr.sin6_family = AF_INET6;
815 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
816 		ia->ia6_createtime = time_second;
817 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
818 			/*
819 			 * XXX: some functions expect that ifa_dstaddr is not
820 			 * NULL for p2p interfaces.
821 			 */
822 			ia->ia_ifa.ifa_dstaddr =
823 			    (struct sockaddr *)&ia->ia_dstaddr;
824 		} else {
825 			ia->ia_ifa.ifa_dstaddr = NULL;
826 		}
827 		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
828 		ia->ia_ifp = ifp;
829 		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
830 		IF_ADDR_LOCK(ifp);
831 		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
832 		IF_ADDR_UNLOCK(ifp);
833 
834 		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
835 		IN6_IFADDR_WLOCK();
836 		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
837 		IN6_IFADDR_WUNLOCK();
838 	}
839 
840 	/* update timestamp */
841 	ia->ia6_updatetime = time_second;
842 
843 	/* set prefix mask */
844 	if (ifra->ifra_prefixmask.sin6_len) {
845 		/*
846 		 * We prohibit changing the prefix length of an existing
847 		 * address, because
848 		 * + such an operation should be rare in IPv6, and
849 		 * + the operation would confuse prefix management.
850 		 */
851 		if (ia->ia_prefixmask.sin6_len &&
852 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
853 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
854 			    " existing (%s) address should not be changed\n",
855 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
856 			error = EINVAL;
857 			goto unlink;
858 		}
859 		ia->ia_prefixmask = ifra->ifra_prefixmask;
860 	}
861 
862 	/*
863 	 * If a new destination address is specified, scrub the old one and
864 	 * install the new destination.  Note that the interface must be
865 	 * p2p or loopback (see the check above.)
866 	 */
867 	if (dst6.sin6_family == AF_INET6 &&
868 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
869 		int e;
870 
871 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
872 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
873 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
874 			    "a route to the old destination: %s\n",
875 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
876 			/* proceed anyway... */
877 		} else
878 			ia->ia_flags &= ~IFA_ROUTE;
879 		ia->ia_dstaddr = dst6;
880 	}
881 
882 	/*
883 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
884 	 * to see if the address is deprecated or invalidated, but initialize
885 	 * these members for applications.
886 	 */
887 	ia->ia6_lifetime = ifra->ifra_lifetime;
888 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
889 		ia->ia6_lifetime.ia6t_expire =
890 		    time_second + ia->ia6_lifetime.ia6t_vltime;
891 	} else
892 		ia->ia6_lifetime.ia6t_expire = 0;
893 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
894 		ia->ia6_lifetime.ia6t_preferred =
895 		    time_second + ia->ia6_lifetime.ia6t_pltime;
896 	} else
897 		ia->ia6_lifetime.ia6t_preferred = 0;
898 
899 	/* reset the interface and routing table appropriately. */
900 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
901 		goto unlink;
902 
903 	/*
904 	 * configure address flags.
905 	 */
906 	ia->ia6_flags = ifra->ifra_flags;
907 	/*
908 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
909 	 * userland, make it deprecated.
910 	 */
911 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
912 		ia->ia6_lifetime.ia6t_pltime = 0;
913 		ia->ia6_lifetime.ia6t_preferred = time_second;
914 	}
915 	/*
916 	 * Make the address tentative before joining multicast addresses,
917 	 * so that corresponding MLD responses would not have a tentative
918 	 * source address.
919 	 */
920 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
921 	if (hostIsNew && in6if_do_dad(ifp))
922 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
923 
924 	/*
925 	 * We are done if we have simply modified an existing address.
926 	 */
927 	if (!hostIsNew)
928 		return (error);
929 
930 	/*
931 	 * Beyond this point, we should call in6_purgeaddr upon an error,
932 	 * not just go to unlink.
933 	 */
934 
935 	/* Join necessary multicast groups */
936 	in6m_sol = NULL;
937 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
938 		struct sockaddr_in6 mltaddr, mltmask;
939 		struct in6_addr llsol;
940 
941 		/* join solicited multicast addr for new host id */
942 		bzero(&llsol, sizeof(struct in6_addr));
943 		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
944 		llsol.s6_addr32[1] = 0;
945 		llsol.s6_addr32[2] = htonl(1);
946 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
947 		llsol.s6_addr8[12] = 0xff;
948 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
949 			/* XXX: should not happen */
950 			log(LOG_ERR, "in6_update_ifa: "
951 			    "in6_setscope failed\n");
952 			goto cleanup;
953 		}
954 		delay = 0;
955 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
956 			/*
957 			 * We need a random delay for DAD on the address
958 			 * being configured.  It also means delaying
959 			 * transmission of the corresponding MLD report to
960 			 * avoid report collision.
961 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
962 			 */
963 			delay = arc4random() %
964 			    (MAX_RTR_SOLICITATION_DELAY * hz);
965 		}
966 		imm = in6_joingroup(ifp, &llsol, &error, delay);
967 		if (imm == NULL) {
968 			nd6log((LOG_WARNING,
969 			    "in6_update_ifa: addmulti failed for "
970 			    "%s on %s (errno=%d)\n",
971 			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
972 			    error));
973 			goto cleanup;
974 		}
975 		LIST_INSERT_HEAD(&ia->ia6_memberships,
976 		    imm, i6mm_chain);
977 		in6m_sol = imm->i6mm_maddr;
978 
979 		bzero(&mltmask, sizeof(mltmask));
980 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
981 		mltmask.sin6_family = AF_INET6;
982 		mltmask.sin6_addr = in6mask32;
983 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
984 
985 		/*
986 		 * join link-local all-nodes address
987 		 */
988 		bzero(&mltaddr, sizeof(mltaddr));
989 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
990 		mltaddr.sin6_family = AF_INET6;
991 		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
992 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
993 		    0)
994 			goto cleanup; /* XXX: should not fail */
995 
996 		/*
997 		 * XXX: do we really need this automatic routes?
998 		 * We should probably reconsider this stuff.  Most applications
999 		 * actually do not need the routes, since they usually specify
1000 		 * the outgoing interface.
1001 		 */
1002 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1003 		if (rt) {
1004 			/* XXX: only works in !SCOPEDROUTING case. */
1005 			if (memcmp(&mltaddr.sin6_addr,
1006 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1007 			    MLTMASK_LEN)) {
1008 				RTFREE_LOCKED(rt);
1009 				rt = NULL;
1010 			}
1011 		}
1012 		if (!rt) {
1013 			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1014 			    (struct sockaddr *)&ia->ia_addr,
1015 			    (struct sockaddr *)&mltmask, RTF_UP,
1016 			    (struct rtentry **)0);
1017 			if (error)
1018 				goto cleanup;
1019 		} else {
1020 			RTFREE_LOCKED(rt);
1021 		}
1022 
1023 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1024 		if (!imm) {
1025 			nd6log((LOG_WARNING,
1026 			    "in6_update_ifa: addmulti failed for "
1027 			    "%s on %s (errno=%d)\n",
1028 			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1029 			    if_name(ifp), error));
1030 			goto cleanup;
1031 		}
1032 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1033 
1034 		/*
1035 		 * join node information group address
1036 		 */
1037 		delay = 0;
1038 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1039 			/*
1040 			 * The spec doesn't say anything about delay for this
1041 			 * group, but the same logic should apply.
1042 			 */
1043 			delay = arc4random() %
1044 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1045 		}
1046 		if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
1047 			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1048 			    delay); /* XXX jinmei */
1049 			if (!imm) {
1050 				nd6log((LOG_WARNING, "in6_update_ifa: "
1051 				    "addmulti failed for %s on %s "
1052 				    "(errno=%d)\n",
1053 				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1054 				    if_name(ifp), error));
1055 				/* XXX not very fatal, go on... */
1056 			} else {
1057 				LIST_INSERT_HEAD(&ia->ia6_memberships,
1058 				    imm, i6mm_chain);
1059 			}
1060 		}
1061 
1062 		/*
1063 		 * join interface-local all-nodes address.
1064 		 * (ff01::1%ifN, and ff01::%ifN/32)
1065 		 */
1066 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1067 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1068 		    != 0)
1069 			goto cleanup; /* XXX: should not fail */
1070 		/* XXX: again, do we really need the route? */
1071 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1072 		if (rt) {
1073 			if (memcmp(&mltaddr.sin6_addr,
1074 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1075 			    MLTMASK_LEN)) {
1076 				RTFREE_LOCKED(rt);
1077 				rt = NULL;
1078 			}
1079 		}
1080 		if (!rt) {
1081 			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1082 			    (struct sockaddr *)&ia->ia_addr,
1083 			    (struct sockaddr *)&mltmask, RTF_UP,
1084 			    (struct rtentry **)0);
1085 			if (error)
1086 				goto cleanup;
1087 		} else
1088 			RTFREE_LOCKED(rt);
1089 
1090 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1091 		if (!imm) {
1092 			nd6log((LOG_WARNING, "in6_update_ifa: "
1093 			    "addmulti failed for %s on %s "
1094 			    "(errno=%d)\n",
1095 			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1096 			    if_name(ifp), error));
1097 			goto cleanup;
1098 		}
1099 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1100 #undef	MLTMASK_LEN
1101 	}
1102 
1103 	/*
1104 	 * Perform DAD, if needed.
1105 	 * XXX It may be of use, if we can administratively
1106 	 * disable DAD.
1107 	 */
1108 	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1109 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1110 	{
1111 		int mindelay, maxdelay;
1112 
1113 		delay = 0;
1114 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1115 			/*
1116 			 * We need to impose a delay before sending an NS
1117 			 * for DAD.  Check if we also needed a delay for the
1118 			 * corresponding MLD message.  If we did, the delay
1119 			 * should be larger than the MLD delay (this could be
1120 			 * relaxed a bit, but this simple logic is at least
1121 			 * safe).
1122 			 * XXX: Break data hiding guidelines and look at
1123 			 * state for the solicited multicast group.
1124 			 */
1125 			mindelay = 0;
1126 			if (in6m_sol != NULL &&
1127 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1128 				mindelay = in6m_sol->in6m_timer;
1129 			}
1130 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1131 			if (maxdelay - mindelay == 0)
1132 				delay = 0;
1133 			else {
1134 				delay =
1135 				    (arc4random() % (maxdelay - mindelay)) +
1136 				    mindelay;
1137 			}
1138 		}
1139 		nd6_dad_start((struct ifaddr *)ia, delay);
1140 	}
1141 
1142 	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1143 	ifa_free(&ia->ia_ifa);
1144 	return (error);
1145 
1146   unlink:
1147 	/*
1148 	 * XXX: if a change of an existing address failed, keep the entry
1149 	 * anyway.
1150 	 */
1151 	if (hostIsNew) {
1152 		in6_unlink_ifa(ia, ifp);
1153 		ifa_free(&ia->ia_ifa);
1154 	}
1155 	return (error);
1156 
1157   cleanup:
1158 	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1159 	ifa_free(&ia->ia_ifa);
1160 	in6_purgeaddr(&ia->ia_ifa);
1161 	return error;
1162 }
1163 
1164 void
1165 in6_purgeaddr(struct ifaddr *ifa)
1166 {
1167 	struct ifnet *ifp = ifa->ifa_ifp;
1168 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1169 	struct in6_multi_mship *imm;
1170 	struct sockaddr_in6 mltaddr, mltmask;
1171 	struct rtentry rt0;
1172 	struct sockaddr_dl gateway;
1173 	struct sockaddr_in6 mask, addr;
1174 	int plen, error;
1175 	struct rtentry *rt;
1176 	struct ifaddr *ifa0, *nifa;
1177 
1178 	/*
1179 	 * find another IPv6 address as the gateway for the
1180 	 * link-local and node-local all-nodes multicast
1181 	 * address routes
1182 	 */
1183 	IF_ADDR_LOCK(ifp);
1184 	TAILQ_FOREACH_SAFE(ifa0, &ifp->if_addrhead, ifa_link, nifa) {
1185 		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1186 		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1187 			   &ia->ia_addr.sin6_addr,
1188 			   sizeof(struct in6_addr)) == 0)
1189 			continue;
1190 		else
1191 			break;
1192 	}
1193 	if (ifa0 != NULL)
1194 		ifa_ref(ifa0);
1195 	IF_ADDR_UNLOCK(ifp);
1196 
1197 	/* stop DAD processing */
1198 	nd6_dad_stop(ifa);
1199 
1200 	IF_AFDATA_LOCK(ifp);
1201 	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
1202 	    (struct sockaddr *)&ia->ia_addr);
1203 	IF_AFDATA_UNLOCK(ifp);
1204 
1205 	/*
1206 	 * initialize for rtmsg generation
1207 	 */
1208 	bzero(&gateway, sizeof(gateway));
1209 	gateway.sdl_len = sizeof(gateway);
1210 	gateway.sdl_family = AF_LINK;
1211 	gateway.sdl_nlen = 0;
1212 	gateway.sdl_alen = ifp->if_addrlen;
1213 	/* */
1214 	bzero(&rt0, sizeof(rt0));
1215 	rt0.rt_gateway = (struct sockaddr *)&gateway;
1216 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
1217 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
1218 	rt_mask(&rt0) = (struct sockaddr *)&mask;
1219 	rt_key(&rt0) = (struct sockaddr *)&addr;
1220 	rt0.rt_flags = RTF_HOST | RTF_STATIC;
1221 	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
1222 
1223 	/*
1224 	 * leave from multicast groups we have joined for the interface
1225 	 */
1226 	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1227 		LIST_REMOVE(imm, i6mm_chain);
1228 		in6_leavegroup(imm);
1229 	}
1230 
1231 	/*
1232 	 * remove the link-local all-nodes address
1233 	 */
1234 	bzero(&mltmask, sizeof(mltmask));
1235 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1236 	mltmask.sin6_family = AF_INET6;
1237 	mltmask.sin6_addr = in6mask32;
1238 
1239 	bzero(&mltaddr, sizeof(mltaddr));
1240 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1241 	mltaddr.sin6_family = AF_INET6;
1242 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1243 
1244 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1245 	    0)
1246 		goto cleanup;
1247 
1248 	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1249 	if (rt != NULL && rt->rt_gateway != NULL &&
1250 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1251 		    &ia->ia_addr.sin6_addr,
1252 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1253 		/*
1254 		 * if no more IPv6 address exists on this interface
1255 		 * then remove the multicast address route
1256 		 */
1257 		if (ifa0 == NULL) {
1258 			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1259 			       sizeof(mltaddr.sin6_addr));
1260 			RTFREE_LOCKED(rt);
1261 			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1262 					  (struct sockaddr *)&ia->ia_addr,
1263 					  (struct sockaddr *)&mltmask, RTF_UP,
1264 					  (struct rtentry **)0);
1265 			if (error)
1266 				log(LOG_INFO, "in6_purgeaddr: link-local all-nodes"
1267 				    "multicast address deletion error\n");
1268 		} else {
1269 			/*
1270 			 * replace the gateway of the route
1271 			 */
1272 			struct sockaddr_in6 sa;
1273 
1274 			bzero(&sa, sizeof(sa));
1275 			sa.sin6_len = sizeof(struct sockaddr_in6);
1276 			sa.sin6_family = AF_INET6;
1277 			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1278 			       sizeof(sa.sin6_addr));
1279 			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1280 			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1281 			RTFREE_LOCKED(rt);
1282 		}
1283 	} else {
1284 		if (rt != NULL)
1285 			RTFREE_LOCKED(rt);
1286 	}
1287 
1288 	/*
1289 	 * remove the node-local all-nodes address
1290 	 */
1291 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1292 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1293 	    0)
1294 		goto cleanup;
1295 
1296 	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1297 	if (rt != NULL && rt->rt_gateway != NULL &&
1298 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1299 		    &ia->ia_addr.sin6_addr,
1300 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1301 		/*
1302 		 * if no more IPv6 address exists on this interface
1303 		 * then remove the multicast address route
1304 		 */
1305 		if (ifa0 == NULL) {
1306 			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1307 			       sizeof(mltaddr.sin6_addr));
1308 
1309 			RTFREE_LOCKED(rt);
1310 			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1311 					  (struct sockaddr *)&ia->ia_addr,
1312 					  (struct sockaddr *)&mltmask, RTF_UP,
1313 					  (struct rtentry **)0);
1314 
1315 			if (error)
1316 				log(LOG_INFO, "in6_purgeaddr: node-local all-nodes"
1317 				    "multicast address deletion error\n");
1318 		} else {
1319 			/*
1320 			 * replace the gateway of the route
1321 			 */
1322 			struct sockaddr_in6 sa;
1323 
1324 			bzero(&sa, sizeof(sa));
1325 			sa.sin6_len = sizeof(struct sockaddr_in6);
1326 			sa.sin6_family = AF_INET6;
1327 			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1328 			       sizeof(sa.sin6_addr));
1329 			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1330 			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1331 			RTFREE_LOCKED(rt);
1332 		}
1333 	} else {
1334 		if (rt != NULL)
1335 			RTFREE_LOCKED(rt);
1336 	}
1337 
1338 cleanup:
1339 
1340 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1341 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1342 		int error;
1343 		struct sockaddr *dstaddr;
1344 
1345 		/*
1346 		 * use the interface address if configuring an
1347 		 * interface address with a /128 prefix len
1348 		 */
1349 		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1350 			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1351 		else
1352 			dstaddr = (struct sockaddr *)&ia->ia_addr;
1353 
1354 		error = rtrequest(RTM_DELETE,
1355 		    (struct sockaddr *)dstaddr,
1356 		    (struct sockaddr *)&ia->ia_addr,
1357 		    (struct sockaddr *)&ia->ia_prefixmask,
1358 		    ia->ia_flags | RTF_HOST, NULL);
1359 		if (error != 0)
1360 			return;
1361 		ia->ia_flags &= ~IFA_ROUTE;
1362 	}
1363 	if (ifa0 != NULL)
1364 		ifa_free(ifa0);
1365 
1366 	in6_unlink_ifa(ia, ifp);
1367 }
1368 
1369 static void
1370 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1371 {
1372 	INIT_VNET_INET6(ifp->if_vnet);
1373 	int	s = splnet();
1374 
1375 	IF_ADDR_LOCK(ifp);
1376 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1377 	IF_ADDR_UNLOCK(ifp);
1378 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1379 
1380 	/*
1381 	 * Defer the release of what might be the last reference to the
1382 	 * in6_ifaddr so that it can't be freed before the remainder of the
1383 	 * cleanup.
1384 	 */
1385 	IN6_IFADDR_WLOCK();
1386 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1387 	IN6_IFADDR_WUNLOCK();
1388 
1389 	/*
1390 	 * Release the reference to the base prefix.  There should be a
1391 	 * positive reference.
1392 	 */
1393 	if (ia->ia6_ndpr == NULL) {
1394 		nd6log((LOG_NOTICE,
1395 		    "in6_unlink_ifa: autoconf'ed address "
1396 		    "%p has no prefix\n", ia));
1397 	} else {
1398 		ia->ia6_ndpr->ndpr_refcnt--;
1399 		ia->ia6_ndpr = NULL;
1400 	}
1401 
1402 	/*
1403 	 * Also, if the address being removed is autoconf'ed, call
1404 	 * pfxlist_onlink_check() since the release might affect the status of
1405 	 * other (detached) addresses.
1406 	 */
1407 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1408 		pfxlist_onlink_check();
1409 	}
1410 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1411 	splx(s);
1412 }
1413 
1414 void
1415 in6_purgeif(struct ifnet *ifp)
1416 {
1417 	struct ifaddr *ifa, *nifa;
1418 
1419 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
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(struct socket *so, u_long cmd, caddr_t data,
1453     struct ifnet *ifp, struct thread *td)
1454 {
1455 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1456 	struct ifaddr *ifa;
1457 	struct sockaddr *sa;
1458 
1459 	/* sanity checks */
1460 	if (!data || !ifp) {
1461 		panic("invalid argument to in6_lifaddr_ioctl");
1462 		/* NOTREACHED */
1463 	}
1464 
1465 	switch (cmd) {
1466 	case SIOCGLIFADDR:
1467 		/* address must be specified on GET with IFLR_PREFIX */
1468 		if ((iflr->flags & IFLR_PREFIX) == 0)
1469 			break;
1470 		/* FALLTHROUGH */
1471 	case SIOCALIFADDR:
1472 	case SIOCDLIFADDR:
1473 		/* address must be specified on ADD and DELETE */
1474 		sa = (struct sockaddr *)&iflr->addr;
1475 		if (sa->sa_family != AF_INET6)
1476 			return EINVAL;
1477 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1478 			return EINVAL;
1479 		/* XXX need improvement */
1480 		sa = (struct sockaddr *)&iflr->dstaddr;
1481 		if (sa->sa_family && sa->sa_family != AF_INET6)
1482 			return EINVAL;
1483 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1484 			return EINVAL;
1485 		break;
1486 	default: /* shouldn't happen */
1487 #if 0
1488 		panic("invalid cmd to in6_lifaddr_ioctl");
1489 		/* NOTREACHED */
1490 #else
1491 		return EOPNOTSUPP;
1492 #endif
1493 	}
1494 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1495 		return EINVAL;
1496 
1497 	switch (cmd) {
1498 	case SIOCALIFADDR:
1499 	    {
1500 		struct in6_aliasreq ifra;
1501 		struct in6_addr *hostid = NULL;
1502 		int prefixlen;
1503 
1504 		ifa = NULL;
1505 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1506 			struct sockaddr_in6 *sin6;
1507 
1508 			/*
1509 			 * hostid is to fill in the hostid part of the
1510 			 * address.  hostid points to the first link-local
1511 			 * address attached to the interface.
1512 			 */
1513 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1514 			if (!ifa)
1515 				return EADDRNOTAVAIL;
1516 			hostid = IFA_IN6(ifa);
1517 
1518 			/* prefixlen must be <= 64. */
1519 			if (64 < iflr->prefixlen)
1520 				return EINVAL;
1521 			prefixlen = iflr->prefixlen;
1522 
1523 			/* hostid part must be zero. */
1524 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1525 			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1526 			    sin6->sin6_addr.s6_addr32[3] != 0) {
1527 				return EINVAL;
1528 			}
1529 		} else
1530 			prefixlen = iflr->prefixlen;
1531 
1532 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1533 		bzero(&ifra, sizeof(ifra));
1534 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1535 
1536 		bcopy(&iflr->addr, &ifra.ifra_addr,
1537 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1538 		if (hostid) {
1539 			/* fill in hostid part */
1540 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1541 			    hostid->s6_addr32[2];
1542 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1543 			    hostid->s6_addr32[3];
1544 		}
1545 
1546 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1547 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1548 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1549 			if (hostid) {
1550 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1551 				    hostid->s6_addr32[2];
1552 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1553 				    hostid->s6_addr32[3];
1554 			}
1555 		}
1556 		if (ifa != NULL)
1557 			ifa_free(ifa);
1558 
1559 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1560 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1561 
1562 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1563 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1564 	    }
1565 	case SIOCGLIFADDR:
1566 	case SIOCDLIFADDR:
1567 	    {
1568 		struct in6_ifaddr *ia;
1569 		struct in6_addr mask, candidate, match;
1570 		struct sockaddr_in6 *sin6;
1571 		int cmp;
1572 
1573 		bzero(&mask, sizeof(mask));
1574 		if (iflr->flags & IFLR_PREFIX) {
1575 			/* lookup a prefix rather than address. */
1576 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1577 
1578 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1579 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1580 			match.s6_addr32[0] &= mask.s6_addr32[0];
1581 			match.s6_addr32[1] &= mask.s6_addr32[1];
1582 			match.s6_addr32[2] &= mask.s6_addr32[2];
1583 			match.s6_addr32[3] &= mask.s6_addr32[3];
1584 
1585 			/* if you set extra bits, that's wrong */
1586 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1587 				return EINVAL;
1588 
1589 			cmp = 1;
1590 		} else {
1591 			if (cmd == SIOCGLIFADDR) {
1592 				/* on getting an address, take the 1st match */
1593 				cmp = 0;	/* XXX */
1594 			} else {
1595 				/* on deleting an address, do exact match */
1596 				in6_prefixlen2mask(&mask, 128);
1597 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1598 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1599 
1600 				cmp = 1;
1601 			}
1602 		}
1603 
1604 		IF_ADDR_LOCK(ifp);
1605 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1606 			if (ifa->ifa_addr->sa_family != AF_INET6)
1607 				continue;
1608 			if (!cmp)
1609 				break;
1610 
1611 			/*
1612 			 * XXX: this is adhoc, but is necessary to allow
1613 			 * a user to specify fe80::/64 (not /10) for a
1614 			 * link-local address.
1615 			 */
1616 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1617 			in6_clearscope(&candidate);
1618 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1619 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1620 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1621 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1622 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1623 				break;
1624 		}
1625 		IF_ADDR_UNLOCK(ifp);
1626 		if (!ifa)
1627 			return EADDRNOTAVAIL;
1628 		ia = ifa2ia6(ifa);
1629 
1630 		if (cmd == SIOCGLIFADDR) {
1631 			int error;
1632 
1633 			/* fill in the if_laddrreq structure */
1634 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1635 			error = sa6_recoverscope(
1636 			    (struct sockaddr_in6 *)&iflr->addr);
1637 			if (error != 0)
1638 				return (error);
1639 
1640 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1641 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1642 				    ia->ia_dstaddr.sin6_len);
1643 				error = sa6_recoverscope(
1644 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1645 				if (error != 0)
1646 					return (error);
1647 			} else
1648 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1649 
1650 			iflr->prefixlen =
1651 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1652 
1653 			iflr->flags = ia->ia6_flags;	/* XXX */
1654 
1655 			return 0;
1656 		} else {
1657 			struct in6_aliasreq ifra;
1658 
1659 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1660 			bzero(&ifra, sizeof(ifra));
1661 			bcopy(iflr->iflr_name, ifra.ifra_name,
1662 			    sizeof(ifra.ifra_name));
1663 
1664 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1665 			    ia->ia_addr.sin6_len);
1666 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1667 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1668 				    ia->ia_dstaddr.sin6_len);
1669 			} else {
1670 				bzero(&ifra.ifra_dstaddr,
1671 				    sizeof(ifra.ifra_dstaddr));
1672 			}
1673 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1674 			    ia->ia_prefixmask.sin6_len);
1675 
1676 			ifra.ifra_flags = ia->ia6_flags;
1677 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1678 			    ifp, td);
1679 		}
1680 	    }
1681 	}
1682 
1683 	return EOPNOTSUPP;	/* just for safety */
1684 }
1685 
1686 /*
1687  * Initialize an interface's intetnet6 address
1688  * and routing table entry.
1689  */
1690 static int
1691 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1692     struct sockaddr_in6 *sin6, int newhost)
1693 {
1694 	int	error = 0, plen, ifacount = 0;
1695 	int	s = splimp();
1696 	struct ifaddr *ifa;
1697 
1698 	/*
1699 	 * Give the interface a chance to initialize
1700 	 * if this is its first address,
1701 	 * and to validate the address if necessary.
1702 	 */
1703 	IF_ADDR_LOCK(ifp);
1704 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1705 		if (ifa->ifa_addr->sa_family != AF_INET6)
1706 			continue;
1707 		ifacount++;
1708 	}
1709 	IF_ADDR_UNLOCK(ifp);
1710 
1711 	ia->ia_addr = *sin6;
1712 
1713 	if (ifacount <= 1 && ifp->if_ioctl) {
1714 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1715 		if (error) {
1716 			splx(s);
1717 			return (error);
1718 		}
1719 	}
1720 	splx(s);
1721 
1722 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1723 
1724 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1725 
1726 	/*
1727 	 * Special case:
1728 	 * If a new destination address is specified for a point-to-point
1729 	 * interface, install a route to the destination as an interface
1730 	 * direct route.
1731 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1732 	 * interface that share the same destination.
1733 	 */
1734 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1735 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128) {
1736 		struct sockaddr *dstaddr;
1737 		int rtflags = RTF_UP | RTF_HOST;
1738 
1739 		/*
1740 		 * use the interface address if configuring an
1741 		 * interface address with a /128 prefix len
1742 		 */
1743 		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1744 			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1745 		else
1746 			dstaddr = (struct sockaddr *)&ia->ia_addr;
1747 
1748 		error = rtrequest(RTM_ADD,
1749 		    (struct sockaddr *)dstaddr,
1750 		    (struct sockaddr *)&ia->ia_addr,
1751 		    (struct sockaddr *)&ia->ia_prefixmask,
1752 		    ia->ia_flags | rtflags, NULL);
1753 		if (error != 0)
1754 			return (error);
1755 		ia->ia_flags |= IFA_ROUTE;
1756 	}
1757 
1758 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1759 	if (newhost) {
1760 		struct llentry *ln;
1761 		struct rtentry rt;
1762 		struct sockaddr_dl gateway;
1763 		struct sockaddr_in6 mask, addr;
1764 
1765 		IF_AFDATA_LOCK(ifp);
1766 		ia->ia_ifa.ifa_rtrequest = NULL;
1767 
1768 		/* XXX QL
1769 		 * we need to report rt_newaddrmsg
1770 		 */
1771 		ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR | LLE_EXCLUSIVE),
1772 		    (struct sockaddr *)&ia->ia_addr);
1773 		IF_AFDATA_UNLOCK(ifp);
1774 		if (ln != NULL) {
1775 			ln->la_expire = 0;  /* for IPv6 this means permanent */
1776 			ln->ln_state = ND6_LLINFO_REACHABLE;
1777 			/*
1778 			 * initialize for rtmsg generation
1779 			 */
1780 			bzero(&gateway, sizeof(gateway));
1781 			gateway.sdl_len = sizeof(gateway);
1782 			gateway.sdl_family = AF_LINK;
1783 			gateway.sdl_nlen = 0;
1784 			gateway.sdl_alen = 6;
1785 			memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned, sizeof(ln->ll_addr));
1786 			/* */
1787 			LLE_WUNLOCK(ln);
1788 		}
1789 
1790 		bzero(&rt, sizeof(rt));
1791 		rt.rt_gateway = (struct sockaddr *)&gateway;
1792 		memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
1793 		memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
1794 		rt_mask(&rt) = (struct sockaddr *)&mask;
1795 		rt_key(&rt) = (struct sockaddr *)&addr;
1796 		rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
1797 		rt_newaddrmsg(RTM_ADD, &ia->ia_ifa, 0, &rt);
1798 	}
1799 
1800 	return (error);
1801 }
1802 
1803 /*
1804  * Find an IPv6 interface link-local address specific to an interface.
1805  * ifaddr is returned referenced.
1806  */
1807 struct in6_ifaddr *
1808 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1809 {
1810 	struct ifaddr *ifa;
1811 
1812 	IF_ADDR_LOCK(ifp);
1813 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1814 		if (ifa->ifa_addr->sa_family != AF_INET6)
1815 			continue;
1816 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1817 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1818 			     ignoreflags) != 0)
1819 				continue;
1820 			ifa_ref(ifa);
1821 			break;
1822 		}
1823 	}
1824 	IF_ADDR_UNLOCK(ifp);
1825 
1826 	return ((struct in6_ifaddr *)ifa);
1827 }
1828 
1829 
1830 /*
1831  * find the internet address corresponding to a given interface and address.
1832  * ifaddr is returned referenced.
1833  */
1834 struct in6_ifaddr *
1835 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1836 {
1837 	struct ifaddr *ifa;
1838 
1839 	IF_ADDR_LOCK(ifp);
1840 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1841 		if (ifa->ifa_addr->sa_family != AF_INET6)
1842 			continue;
1843 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1844 			ifa_ref(ifa);
1845 			break;
1846 		}
1847 	}
1848 	IF_ADDR_UNLOCK(ifp);
1849 
1850 	return ((struct in6_ifaddr *)ifa);
1851 }
1852 
1853 /*
1854  * Convert IP6 address to printable (loggable) representation. Caller
1855  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1856  */
1857 static char digits[] = "0123456789abcdef";
1858 char *
1859 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1860 {
1861 	int i;
1862 	char *cp;
1863 	const u_int16_t *a = (const u_int16_t *)addr;
1864 	const u_int8_t *d;
1865 	int dcolon = 0, zero = 0;
1866 
1867 	cp = ip6buf;
1868 
1869 	for (i = 0; i < 8; i++) {
1870 		if (dcolon == 1) {
1871 			if (*a == 0) {
1872 				if (i == 7)
1873 					*cp++ = ':';
1874 				a++;
1875 				continue;
1876 			} else
1877 				dcolon = 2;
1878 		}
1879 		if (*a == 0) {
1880 			if (dcolon == 0 && *(a + 1) == 0) {
1881 				if (i == 0)
1882 					*cp++ = ':';
1883 				*cp++ = ':';
1884 				dcolon = 1;
1885 			} else {
1886 				*cp++ = '0';
1887 				*cp++ = ':';
1888 			}
1889 			a++;
1890 			continue;
1891 		}
1892 		d = (const u_char *)a;
1893 		/* Try to eliminate leading zeros in printout like in :0001. */
1894 		zero = 1;
1895 		*cp = digits[*d >> 4];
1896 		if (*cp != '0') {
1897 			zero = 0;
1898 			cp++;
1899 		}
1900 		*cp = digits[*d++ & 0xf];
1901 		if (zero == 0 || (*cp != '0')) {
1902 			zero = 0;
1903 			cp++;
1904 		}
1905 		*cp = digits[*d >> 4];
1906 		if (zero == 0 || (*cp != '0')) {
1907 			zero = 0;
1908 			cp++;
1909 		}
1910 		*cp++ = digits[*d & 0xf];
1911 		*cp++ = ':';
1912 		a++;
1913 	}
1914 	*--cp = '\0';
1915 	return (ip6buf);
1916 }
1917 
1918 int
1919 in6_localaddr(struct in6_addr *in6)
1920 {
1921 	INIT_VNET_INET6(curvnet);
1922 	struct in6_ifaddr *ia;
1923 
1924 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1925 		return 1;
1926 
1927 	IN6_IFADDR_RLOCK();
1928 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1929 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1930 		    &ia->ia_prefixmask.sin6_addr)) {
1931 			IN6_IFADDR_RUNLOCK();
1932 			return 1;
1933 		}
1934 	}
1935 	IN6_IFADDR_RUNLOCK();
1936 
1937 	return (0);
1938 }
1939 
1940 int
1941 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1942 {
1943 	INIT_VNET_INET6(curvnet);
1944 	struct in6_ifaddr *ia;
1945 
1946 	IN6_IFADDR_RLOCK();
1947 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1948 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1949 				       &sa6->sin6_addr) &&
1950 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
1951 			IN6_IFADDR_RUNLOCK();
1952 			return (1); /* true */
1953 		}
1954 
1955 		/* XXX: do we still have to go thru the rest of the list? */
1956 	}
1957 	IN6_IFADDR_RUNLOCK();
1958 
1959 	return (0);		/* false */
1960 }
1961 
1962 /*
1963  * return length of part which dst and src are equal
1964  * hard coding...
1965  */
1966 int
1967 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1968 {
1969 	int match = 0;
1970 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1971 	u_char *lim = s + 16, r;
1972 
1973 	while (s < lim)
1974 		if ((r = (*d++ ^ *s++)) != 0) {
1975 			while (r < 128) {
1976 				match++;
1977 				r <<= 1;
1978 			}
1979 			break;
1980 		} else
1981 			match += 8;
1982 	return match;
1983 }
1984 
1985 /* XXX: to be scope conscious */
1986 int
1987 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1988 {
1989 	int bytelen, bitlen;
1990 
1991 	/* sanity check */
1992 	if (0 > len || len > 128) {
1993 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1994 		    len);
1995 		return (0);
1996 	}
1997 
1998 	bytelen = len / 8;
1999 	bitlen = len % 8;
2000 
2001 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2002 		return (0);
2003 	if (bitlen != 0 &&
2004 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2005 	    p2->s6_addr[bytelen] >> (8 - bitlen))
2006 		return (0);
2007 
2008 	return (1);
2009 }
2010 
2011 void
2012 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2013 {
2014 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2015 	int bytelen, bitlen, i;
2016 
2017 	/* sanity check */
2018 	if (0 > len || len > 128) {
2019 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2020 		    len);
2021 		return;
2022 	}
2023 
2024 	bzero(maskp, sizeof(*maskp));
2025 	bytelen = len / 8;
2026 	bitlen = len % 8;
2027 	for (i = 0; i < bytelen; i++)
2028 		maskp->s6_addr[i] = 0xff;
2029 	if (bitlen)
2030 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2031 }
2032 
2033 /*
2034  * return the best address out of the same scope. if no address was
2035  * found, return the first valid address from designated IF.
2036  */
2037 struct in6_ifaddr *
2038 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2039 {
2040 	INIT_VNET_INET6(curvnet);
2041 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2042 	struct ifaddr *ifa;
2043 	struct in6_ifaddr *besta = 0;
2044 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2045 
2046 	dep[0] = dep[1] = NULL;
2047 
2048 	/*
2049 	 * We first look for addresses in the same scope.
2050 	 * If there is one, return it.
2051 	 * If two or more, return one which matches the dst longest.
2052 	 * If none, return one of global addresses assigned other ifs.
2053 	 */
2054 	IF_ADDR_LOCK(ifp);
2055 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2056 		if (ifa->ifa_addr->sa_family != AF_INET6)
2057 			continue;
2058 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2059 			continue; /* XXX: is there any case to allow anycast? */
2060 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2061 			continue; /* don't use this interface */
2062 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2063 			continue;
2064 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2065 			if (V_ip6_use_deprecated)
2066 				dep[0] = (struct in6_ifaddr *)ifa;
2067 			continue;
2068 		}
2069 
2070 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2071 			/*
2072 			 * call in6_matchlen() as few as possible
2073 			 */
2074 			if (besta) {
2075 				if (blen == -1)
2076 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2077 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2078 				if (tlen > blen) {
2079 					blen = tlen;
2080 					besta = (struct in6_ifaddr *)ifa;
2081 				}
2082 			} else
2083 				besta = (struct in6_ifaddr *)ifa;
2084 		}
2085 	}
2086 	if (besta) {
2087 		ifa_ref(&besta->ia_ifa);
2088 		IF_ADDR_UNLOCK(ifp);
2089 		return (besta);
2090 	}
2091 	IF_ADDR_UNLOCK(ifp);
2092 
2093 	IN6_IFADDR_RLOCK();
2094 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2095 		if (ifa->ifa_addr->sa_family != AF_INET6)
2096 			continue;
2097 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2098 			continue; /* XXX: is there any case to allow anycast? */
2099 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2100 			continue; /* don't use this interface */
2101 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2102 			continue;
2103 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2104 			if (V_ip6_use_deprecated)
2105 				dep[1] = (struct in6_ifaddr *)ifa;
2106 			continue;
2107 		}
2108 
2109 		if (ifa != NULL)
2110 			ifa_ref(ifa);
2111 		IN6_IFADDR_RUNLOCK();
2112 		return (struct in6_ifaddr *)ifa;
2113 	}
2114 	IN6_IFADDR_RUNLOCK();
2115 
2116 	/* use the last-resort values, that are, deprecated addresses */
2117 	if (dep[0])
2118 		return dep[0];
2119 	if (dep[1])
2120 		return dep[1];
2121 
2122 	return NULL;
2123 }
2124 
2125 /*
2126  * perform DAD when interface becomes IFF_UP.
2127  */
2128 void
2129 in6_if_up(struct ifnet *ifp)
2130 {
2131 	struct ifaddr *ifa;
2132 	struct in6_ifaddr *ia;
2133 
2134 	IF_ADDR_LOCK(ifp);
2135 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2136 		if (ifa->ifa_addr->sa_family != AF_INET6)
2137 			continue;
2138 		ia = (struct in6_ifaddr *)ifa;
2139 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2140 			/*
2141 			 * The TENTATIVE flag was likely set by hand
2142 			 * beforehand, implicitly indicating the need for DAD.
2143 			 * We may be able to skip the random delay in this
2144 			 * case, but we impose delays just in case.
2145 			 */
2146 			nd6_dad_start(ifa,
2147 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2148 		}
2149 	}
2150 	IF_ADDR_UNLOCK(ifp);
2151 
2152 	/*
2153 	 * special cases, like 6to4, are handled in in6_ifattach
2154 	 */
2155 	in6_ifattach(ifp, NULL);
2156 }
2157 
2158 int
2159 in6if_do_dad(struct ifnet *ifp)
2160 {
2161 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2162 		return (0);
2163 
2164 	switch (ifp->if_type) {
2165 #ifdef IFT_DUMMY
2166 	case IFT_DUMMY:
2167 #endif
2168 	case IFT_FAITH:
2169 		/*
2170 		 * These interfaces do not have the IFF_LOOPBACK flag,
2171 		 * but loop packets back.  We do not have to do DAD on such
2172 		 * interfaces.  We should even omit it, because loop-backed
2173 		 * NS would confuse the DAD procedure.
2174 		 */
2175 		return (0);
2176 	default:
2177 		/*
2178 		 * Our DAD routine requires the interface up and running.
2179 		 * However, some interfaces can be up before the RUNNING
2180 		 * status.  Additionaly, users may try to assign addresses
2181 		 * before the interface becomes up (or running).
2182 		 * We simply skip DAD in such a case as a work around.
2183 		 * XXX: we should rather mark "tentative" on such addresses,
2184 		 * and do DAD after the interface becomes ready.
2185 		 */
2186 		if (!((ifp->if_flags & IFF_UP) &&
2187 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2188 			return (0);
2189 
2190 		return (1);
2191 	}
2192 }
2193 
2194 /*
2195  * Calculate max IPv6 MTU through all the interfaces and store it
2196  * to in6_maxmtu.
2197  */
2198 void
2199 in6_setmaxmtu(void)
2200 {
2201 	INIT_VNET_NET(curvnet);
2202 	INIT_VNET_INET6(curvnet);
2203 	unsigned long maxmtu = 0;
2204 	struct ifnet *ifp;
2205 
2206 	IFNET_RLOCK();
2207 	for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
2208 	    ifp = TAILQ_NEXT(ifp, if_list)) {
2209 		/* this function can be called during ifnet initialization */
2210 		if (!ifp->if_afdata[AF_INET6])
2211 			continue;
2212 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2213 		    IN6_LINKMTU(ifp) > maxmtu)
2214 			maxmtu = IN6_LINKMTU(ifp);
2215 	}
2216 	IFNET_RUNLOCK();
2217 	if (maxmtu)	     /* update only when maxmtu is positive */
2218 		V_in6_maxmtu = maxmtu;
2219 }
2220 
2221 /*
2222  * Provide the length of interface identifiers to be used for the link attached
2223  * to the given interface.  The length should be defined in "IPv6 over
2224  * xxx-link" document.  Note that address architecture might also define
2225  * the length for a particular set of address prefixes, regardless of the
2226  * link type.  As clarified in rfc2462bis, those two definitions should be
2227  * consistent, and those really are as of August 2004.
2228  */
2229 int
2230 in6_if2idlen(struct ifnet *ifp)
2231 {
2232 	switch (ifp->if_type) {
2233 	case IFT_ETHER:		/* RFC2464 */
2234 #ifdef IFT_PROPVIRTUAL
2235 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2236 #endif
2237 #ifdef IFT_L2VLAN
2238 	case IFT_L2VLAN:	/* ditto */
2239 #endif
2240 #ifdef IFT_IEEE80211
2241 	case IFT_IEEE80211:	/* ditto */
2242 #endif
2243 #ifdef IFT_MIP
2244 	case IFT_MIP:	/* ditto */
2245 #endif
2246 		return (64);
2247 	case IFT_FDDI:		/* RFC2467 */
2248 		return (64);
2249 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2250 		return (64);
2251 	case IFT_PPP:		/* RFC2472 */
2252 		return (64);
2253 	case IFT_ARCNET:	/* RFC2497 */
2254 		return (64);
2255 	case IFT_FRELAY:	/* RFC2590 */
2256 		return (64);
2257 	case IFT_IEEE1394:	/* RFC3146 */
2258 		return (64);
2259 	case IFT_GIF:
2260 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2261 	case IFT_LOOP:
2262 		return (64);	/* XXX: is this really correct? */
2263 	default:
2264 		/*
2265 		 * Unknown link type:
2266 		 * It might be controversial to use the today's common constant
2267 		 * of 64 for these cases unconditionally.  For full compliance,
2268 		 * we should return an error in this case.  On the other hand,
2269 		 * if we simply miss the standard for the link type or a new
2270 		 * standard is defined for a new link type, the IFID length
2271 		 * is very likely to be the common constant.  As a compromise,
2272 		 * we always use the constant, but make an explicit notice
2273 		 * indicating the "unknown" case.
2274 		 */
2275 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2276 		return (64);
2277 	}
2278 }
2279 
2280 #include <sys/sysctl.h>
2281 
2282 struct in6_llentry {
2283 	struct llentry		base;
2284 	struct sockaddr_in6	l3_addr6;
2285 };
2286 
2287 static struct llentry *
2288 in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2289 {
2290 	struct in6_llentry *lle;
2291 
2292 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE,
2293 	    M_DONTWAIT | M_ZERO);
2294 	if (lle == NULL)		/* NB: caller generates msg */
2295 		return NULL;
2296 
2297 	callout_init(&lle->base.ln_timer_ch, CALLOUT_MPSAFE);
2298 	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2299 	lle->base.lle_refcnt = 1;
2300 	LLE_LOCK_INIT(&lle->base);
2301 	return &lle->base;
2302 }
2303 
2304 /*
2305  * Deletes an address from the address table.
2306  * This function is called by the timer functions
2307  * such as arptimer() and nd6_llinfo_timer(), and
2308  * the caller does the locking.
2309  */
2310 static void
2311 in6_lltable_free(struct lltable *llt, struct llentry *lle)
2312 {
2313 	LLE_WUNLOCK(lle);
2314 	LLE_LOCK_DESTROY(lle);
2315 	free(lle, M_LLTABLE);
2316 }
2317 
2318 static void
2319 in6_lltable_prefix_free(struct lltable *llt,
2320 			const struct sockaddr *prefix,
2321 			const struct sockaddr *mask)
2322 {
2323 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2324 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2325 	struct llentry *lle, *next;
2326 	register int i;
2327 
2328 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
2329 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2330 			if (IN6_ARE_MASKED_ADDR_EQUAL(
2331 				    &((struct sockaddr_in6 *)L3_ADDR(lle))->sin6_addr,
2332 				    &pfx->sin6_addr,
2333 				    &msk->sin6_addr)) {
2334 				callout_drain(&lle->la_timer);
2335 				LLE_WLOCK(lle);
2336 				llentry_free(lle);
2337 			}
2338 		}
2339 	}
2340 }
2341 
2342 static int
2343 in6_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr)
2344 {
2345 	struct rtentry *rt;
2346 	char ip6buf[INET6_ADDRSTRLEN];
2347 
2348 	KASSERT(l3addr->sa_family == AF_INET6,
2349 	    ("sin_family %d", l3addr->sa_family));
2350 
2351 	/* XXX rtalloc1 should take a const param */
2352 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
2353 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2354 		struct ifaddr *ifa;
2355 		/*
2356 		 * Create an ND6 cache for an IPv6 neighbor
2357 		 * that is not covered by our own prefix.
2358 		 */
2359 		/* XXX ifaof_ifpforaddr should take a const param */
2360 		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2361 		if (ifa != NULL) {
2362 			ifa_free(ifa);
2363 			if (rt != NULL)
2364 				RTFREE_LOCKED(rt);
2365 			return 0;
2366 		}
2367 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2368 		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2369 		if (rt != NULL)
2370 			RTFREE_LOCKED(rt);
2371 		return EINVAL;
2372 	}
2373 	RTFREE_LOCKED(rt);
2374 	return 0;
2375 }
2376 
2377 static struct llentry *
2378 in6_lltable_lookup(struct lltable *llt, u_int flags,
2379 	const struct sockaddr *l3addr)
2380 {
2381 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2382 	struct ifnet *ifp = llt->llt_ifp;
2383 	struct llentry *lle;
2384 	struct llentries *lleh;
2385 	u_int hashkey;
2386 
2387 	IF_AFDATA_LOCK_ASSERT(ifp);
2388 	KASSERT(l3addr->sa_family == AF_INET6,
2389 	    ("sin_family %d", l3addr->sa_family));
2390 
2391 	hashkey = sin6->sin6_addr.s6_addr32[3];
2392 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2393 	LIST_FOREACH(lle, lleh, lle_next) {
2394 		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2395 		if (lle->la_flags & LLE_DELETED)
2396 			continue;
2397 		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2398 			 sizeof(struct in6_addr)) == 0)
2399 			break;
2400 	}
2401 
2402 	if (lle == NULL) {
2403 		if (!(flags & LLE_CREATE))
2404 			return (NULL);
2405 		/*
2406 		 * A route that covers the given address must have
2407 		 * been installed 1st because we are doing a resolution,
2408 		 * verify this.
2409 		 */
2410 		if (!(flags & LLE_IFADDR) &&
2411 		    in6_lltable_rtcheck(ifp, l3addr) != 0)
2412 			return NULL;
2413 
2414 		lle = in6_lltable_new(l3addr, flags);
2415 		if (lle == NULL) {
2416 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2417 			return NULL;
2418 		}
2419 		lle->la_flags = flags & ~LLE_CREATE;
2420 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2421 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2422 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2423 		}
2424 
2425 		lle->lle_tbl  = llt;
2426 		lle->lle_head = lleh;
2427 		LIST_INSERT_HEAD(lleh, lle, lle_next);
2428 	} else if (flags & LLE_DELETE) {
2429 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2430 			LLE_WLOCK(lle);
2431 			lle->la_flags = LLE_DELETED;
2432 			LLE_WUNLOCK(lle);
2433 #ifdef DIAGNOSTICS
2434 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2435 #endif
2436 		}
2437 		lle = (void *)-1;
2438 	}
2439 	if (LLE_IS_VALID(lle)) {
2440 		if (flags & LLE_EXCLUSIVE)
2441 			LLE_WLOCK(lle);
2442 		else
2443 			LLE_RLOCK(lle);
2444 	}
2445 	return (lle);
2446 }
2447 
2448 static int
2449 in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2450 {
2451 	struct ifnet *ifp = llt->llt_ifp;
2452 	struct llentry *lle;
2453 	/* XXX stack use */
2454 	struct {
2455 		struct rt_msghdr	rtm;
2456 		struct sockaddr_in6	sin6;
2457 		/*
2458 		 * ndp.c assumes that sdl is word aligned
2459 		 */
2460 #ifdef __LP64__
2461 		uint32_t		pad;
2462 #endif
2463 		struct sockaddr_dl	sdl;
2464 	} ndpc;
2465 	int i, error;
2466 
2467 	/* XXXXX
2468 	 * current IFNET_RLOCK() is mapped to IFNET_WLOCK()
2469 	 * so it is okay to use this ASSERT, change it when
2470 	 * IFNET lock is finalized
2471 	 */
2472 	IFNET_WLOCK_ASSERT();
2473 
2474 	error = 0;
2475 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2476 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2477 			struct sockaddr_dl *sdl;
2478 
2479 			/* skip deleted or invalid entries */
2480 			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2481 				continue;
2482 			/* Skip if jailed and not a valid IP of the prison. */
2483 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2484 				continue;
2485 			/*
2486 			 * produce a msg made of:
2487 			 *  struct rt_msghdr;
2488 			 *  struct sockaddr_in6 (IPv6)
2489 			 *  struct sockaddr_dl;
2490 			 */
2491 			bzero(&ndpc, sizeof(ndpc));
2492 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2493 			ndpc.rtm.rtm_version = RTM_VERSION;
2494 			ndpc.rtm.rtm_type = RTM_GET;
2495 			ndpc.rtm.rtm_flags = RTF_UP;
2496 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2497 			ndpc.sin6.sin6_family = AF_INET6;
2498 			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2499 			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2500 
2501 			/* publish */
2502 			if (lle->la_flags & LLE_PUB)
2503 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2504 
2505 			sdl = &ndpc.sdl;
2506 			sdl->sdl_family = AF_LINK;
2507 			sdl->sdl_len = sizeof(*sdl);
2508 			sdl->sdl_alen = ifp->if_addrlen;
2509 			sdl->sdl_index = ifp->if_index;
2510 			sdl->sdl_type = ifp->if_type;
2511 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2512 			ndpc.rtm.rtm_rmx.rmx_expire =
2513 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2514 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2515 			if (lle->la_flags & LLE_STATIC)
2516 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2517 			ndpc.rtm.rtm_index = ifp->if_index;
2518 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2519 			if (error)
2520 				break;
2521 		}
2522 	}
2523 	return error;
2524 }
2525 
2526 void *
2527 in6_domifattach(struct ifnet *ifp)
2528 {
2529 	struct in6_ifextra *ext;
2530 
2531 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2532 	bzero(ext, sizeof(*ext));
2533 
2534 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2535 	    M_IFADDR, M_WAITOK);
2536 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2537 
2538 	ext->icmp6_ifstat =
2539 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2540 	    M_IFADDR, M_WAITOK);
2541 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2542 
2543 	ext->nd_ifinfo = nd6_ifattach(ifp);
2544 	ext->scope6_id = scope6_ifattach(ifp);
2545 	ext->lltable = lltable_init(ifp, AF_INET6);
2546 	if (ext->lltable != NULL) {
2547 		ext->lltable->llt_new = in6_lltable_new;
2548 		ext->lltable->llt_free = in6_lltable_free;
2549 		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2550 		ext->lltable->llt_rtcheck = in6_lltable_rtcheck;
2551 		ext->lltable->llt_lookup = in6_lltable_lookup;
2552 		ext->lltable->llt_dump = in6_lltable_dump;
2553 	}
2554 
2555 	ext->mld_ifinfo = mld_domifattach(ifp);
2556 
2557 	return ext;
2558 }
2559 
2560 void
2561 in6_domifdetach(struct ifnet *ifp, void *aux)
2562 {
2563 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2564 
2565 	mld_domifdetach(ifp);
2566 	scope6_ifdetach(ext->scope6_id);
2567 	nd6_ifdetach(ext->nd_ifinfo);
2568 	lltable_free(ext->lltable);
2569 	free(ext->in6_ifstat, M_IFADDR);
2570 	free(ext->icmp6_ifstat, M_IFADDR);
2571 	free(ext, M_IFADDR);
2572 }
2573 
2574 /*
2575  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2576  * v4 mapped addr or v4 compat addr
2577  */
2578 void
2579 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2580 {
2581 
2582 	bzero(sin, sizeof(*sin));
2583 	sin->sin_len = sizeof(struct sockaddr_in);
2584 	sin->sin_family = AF_INET;
2585 	sin->sin_port = sin6->sin6_port;
2586 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2587 }
2588 
2589 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2590 void
2591 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2592 {
2593 	bzero(sin6, sizeof(*sin6));
2594 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2595 	sin6->sin6_family = AF_INET6;
2596 	sin6->sin6_port = sin->sin_port;
2597 	sin6->sin6_addr.s6_addr32[0] = 0;
2598 	sin6->sin6_addr.s6_addr32[1] = 0;
2599 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2600 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2601 }
2602 
2603 /* Convert sockaddr_in6 into sockaddr_in. */
2604 void
2605 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2606 {
2607 	struct sockaddr_in *sin_p;
2608 	struct sockaddr_in6 sin6;
2609 
2610 	/*
2611 	 * Save original sockaddr_in6 addr and convert it
2612 	 * to sockaddr_in.
2613 	 */
2614 	sin6 = *(struct sockaddr_in6 *)nam;
2615 	sin_p = (struct sockaddr_in *)nam;
2616 	in6_sin6_2_sin(sin_p, &sin6);
2617 }
2618 
2619 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2620 void
2621 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2622 {
2623 	struct sockaddr_in *sin_p;
2624 	struct sockaddr_in6 *sin6_p;
2625 
2626 	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
2627 	       M_WAITOK);
2628 	sin_p = (struct sockaddr_in *)*nam;
2629 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2630 	free(*nam, M_SONAME);
2631 	*nam = (struct sockaddr *)sin6_p;
2632 }
2633