xref: /freebsd/sys/netinet6/in6.c (revision 7cd2dcf07629713e5a3d60472cfe4701b705a167)
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 #include <netinet/ip_carp.h>
99 
100 #include <netinet/ip6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/nd6.h>
103 #include <netinet6/mld6_var.h>
104 #include <netinet6/ip6_mroute.h>
105 #include <netinet6/in6_ifattach.h>
106 #include <netinet6/scope6_var.h>
107 #include <netinet6/in6_pcb.h>
108 
109 /*
110  * Definitions of some costant IP6 addresses.
111  */
112 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
113 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
114 const struct in6_addr in6addr_nodelocal_allnodes =
115 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
116 const struct in6_addr in6addr_linklocal_allnodes =
117 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
118 const struct in6_addr in6addr_linklocal_allrouters =
119 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
120 const struct in6_addr in6addr_linklocal_allv2routers =
121 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
122 
123 const struct in6_addr in6mask0 = IN6MASK0;
124 const struct in6_addr in6mask32 = IN6MASK32;
125 const struct in6_addr in6mask64 = IN6MASK64;
126 const struct in6_addr in6mask96 = IN6MASK96;
127 const struct in6_addr in6mask128 = IN6MASK128;
128 
129 const struct sockaddr_in6 sa6_any =
130 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
131 
132 static int in6_lifaddr_ioctl(struct socket *, u_long, caddr_t,
133 	struct ifnet *, struct thread *);
134 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
135 	struct sockaddr_in6 *, int);
136 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
137 
138 int	(*faithprefix_p)(struct in6_addr *);
139 
140 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
141 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
142 
143 void
144 in6_ifaddloop(struct ifaddr *ifa)
145 {
146 	struct sockaddr_dl gateway;
147 	struct sockaddr_in6 mask, addr;
148 	struct rtentry rt;
149 	struct in6_ifaddr *ia;
150 	struct ifnet *ifp;
151 	struct llentry *ln;
152 
153 	ia = ifa2ia6(ifa);
154 	ifp = ifa->ifa_ifp;
155 	IF_AFDATA_LOCK(ifp);
156 	ifa->ifa_rtrequest = nd6_rtrequest;
157 	ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
158 	    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
159 	IF_AFDATA_UNLOCK(ifp);
160 	if (ln != NULL) {
161 		ln->la_expire = 0;  /* for IPv6 this means permanent */
162 		ln->ln_state = ND6_LLINFO_REACHABLE;
163 		/*
164 		 * initialize for rtmsg generation
165 		 */
166 		bzero(&gateway, sizeof(gateway));
167 		gateway.sdl_len = sizeof(gateway);
168 		gateway.sdl_family = AF_LINK;
169 		gateway.sdl_nlen = 0;
170 		gateway.sdl_alen = 6;
171 		memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
172 		    sizeof(ln->ll_addr));
173 		LLE_WUNLOCK(ln);
174 	}
175 
176 	bzero(&rt, sizeof(rt));
177 	rt.rt_gateway = (struct sockaddr *)&gateway;
178 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
179 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
180 	rt_mask(&rt) = (struct sockaddr *)&mask;
181 	rt_key(&rt) = (struct sockaddr *)&addr;
182 	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
183 	/* Announce arrival of local address to all FIBs. */
184 	rt_newaddrmsg(RTM_ADD, ifa, 0, &rt);
185 }
186 
187 void
188 in6_ifremloop(struct ifaddr *ifa)
189 {
190 	struct sockaddr_dl gateway;
191 	struct sockaddr_in6 mask, addr;
192 	struct rtentry rt0;
193 	struct in6_ifaddr *ia;
194 	struct ifnet *ifp;
195 
196 	ia = ifa2ia6(ifa);
197 	ifp = ifa->ifa_ifp;
198 	IF_AFDATA_LOCK(ifp);
199 	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
200 	    (struct sockaddr *)&ia->ia_addr);
201 	IF_AFDATA_UNLOCK(ifp);
202 
203 	/*
204 	 * initialize for rtmsg generation
205 	 */
206 	bzero(&gateway, sizeof(gateway));
207 	gateway.sdl_len = sizeof(gateway);
208 	gateway.sdl_family = AF_LINK;
209 	gateway.sdl_nlen = 0;
210 	gateway.sdl_alen = ifp->if_addrlen;
211 	bzero(&rt0, sizeof(rt0));
212 	rt0.rt_gateway = (struct sockaddr *)&gateway;
213 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
214 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
215 	rt_mask(&rt0) = (struct sockaddr *)&mask;
216 	rt_key(&rt0) = (struct sockaddr *)&addr;
217 	rt0.rt_flags = RTF_HOST | RTF_STATIC;
218 	/* Announce removal of local address to all FIBs. */
219 	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
220 }
221 
222 int
223 in6_mask2len(struct in6_addr *mask, u_char *lim0)
224 {
225 	int x = 0, y;
226 	u_char *lim = lim0, *p;
227 
228 	/* ignore the scope_id part */
229 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
230 		lim = (u_char *)mask + sizeof(*mask);
231 	for (p = (u_char *)mask; p < lim; x++, p++) {
232 		if (*p != 0xff)
233 			break;
234 	}
235 	y = 0;
236 	if (p < lim) {
237 		for (y = 0; y < 8; y++) {
238 			if ((*p & (0x80 >> y)) == 0)
239 				break;
240 		}
241 	}
242 
243 	/*
244 	 * when the limit pointer is given, do a stricter check on the
245 	 * remaining bits.
246 	 */
247 	if (p < lim) {
248 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
249 			return (-1);
250 		for (p = p + 1; p < lim; p++)
251 			if (*p != 0)
252 				return (-1);
253 	}
254 
255 	return x * 8 + y;
256 }
257 
258 #ifdef COMPAT_FREEBSD32
259 struct in6_ndifreq32 {
260 	char ifname[IFNAMSIZ];
261 	uint32_t ifindex;
262 };
263 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
264 #endif
265 
266 int
267 in6_control(struct socket *so, u_long cmd, caddr_t data,
268     struct ifnet *ifp, struct thread *td)
269 {
270 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
271 	struct	in6_ifaddr *ia = NULL;
272 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
273 	struct sockaddr_in6 *sa6;
274 	int carp_attached = 0;
275 	int error;
276 	u_long ocmd = cmd;
277 
278 	/*
279 	 * Compat to make pre-10.x ifconfig(8) operable.
280 	 */
281 	if (cmd == OSIOCAIFADDR_IN6)
282 		cmd = SIOCAIFADDR_IN6;
283 
284 	switch (cmd) {
285 	case SIOCGETSGCNT_IN6:
286 	case SIOCGETMIFCNT_IN6:
287 		/*
288 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
289 		 * We cannot see how that would be needed, so do not adjust the
290 		 * KPI blindly; more likely should clean up the IPv4 variant.
291 		 */
292 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
293 	}
294 
295 	switch(cmd) {
296 	case SIOCAADDRCTL_POLICY:
297 	case SIOCDADDRCTL_POLICY:
298 		if (td != NULL) {
299 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
300 			if (error)
301 				return (error);
302 		}
303 		return (in6_src_ioctl(cmd, data));
304 	}
305 
306 	if (ifp == NULL)
307 		return (EOPNOTSUPP);
308 
309 	switch (cmd) {
310 	case SIOCSNDFLUSH_IN6:
311 	case SIOCSPFXFLUSH_IN6:
312 	case SIOCSRTRFLUSH_IN6:
313 	case SIOCSDEFIFACE_IN6:
314 	case SIOCSIFINFO_FLAGS:
315 	case SIOCSIFINFO_IN6:
316 		if (td != NULL) {
317 			error = priv_check(td, PRIV_NETINET_ND6);
318 			if (error)
319 				return (error);
320 		}
321 		/* FALLTHROUGH */
322 	case OSIOCGIFINFO_IN6:
323 	case SIOCGIFINFO_IN6:
324 	case SIOCGDRLST_IN6:
325 	case SIOCGPRLST_IN6:
326 	case SIOCGNBRINFO_IN6:
327 	case SIOCGDEFIFACE_IN6:
328 		return (nd6_ioctl(cmd, data, ifp));
329 
330 #ifdef COMPAT_FREEBSD32
331 	case SIOCGDEFIFACE32_IN6:
332 		{
333 			struct in6_ndifreq ndif;
334 			struct in6_ndifreq32 *ndif32;
335 
336 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
337 			    ifp);
338 			if (error)
339 				return (error);
340 			ndif32 = (struct in6_ndifreq32 *)data;
341 			ndif32->ifindex = ndif.ifindex;
342 			return (0);
343 		}
344 #endif
345 	}
346 
347 	switch (cmd) {
348 	case SIOCSIFPREFIX_IN6:
349 	case SIOCDIFPREFIX_IN6:
350 	case SIOCAIFPREFIX_IN6:
351 	case SIOCCIFPREFIX_IN6:
352 	case SIOCSGIFPREFIX_IN6:
353 	case SIOCGIFPREFIX_IN6:
354 		log(LOG_NOTICE,
355 		    "prefix ioctls are now invalidated. "
356 		    "please use ifconfig.\n");
357 		return (EOPNOTSUPP);
358 	}
359 
360 	switch (cmd) {
361 	case SIOCSSCOPE6:
362 		if (td != NULL) {
363 			error = priv_check(td, PRIV_NETINET_SCOPE6);
364 			if (error)
365 				return (error);
366 		}
367 		return (scope6_set(ifp,
368 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
369 	case SIOCGSCOPE6:
370 		return (scope6_get(ifp,
371 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
372 	case SIOCGSCOPE6DEF:
373 		return (scope6_get_default((struct scope6_id *)
374 		    ifr->ifr_ifru.ifru_scope_id));
375 	}
376 
377 	switch (cmd) {
378 	case SIOCALIFADDR:
379 		if (td != NULL) {
380 			error = priv_check(td, PRIV_NET_ADDIFADDR);
381 			if (error)
382 				return (error);
383 		}
384 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
385 
386 	case SIOCDLIFADDR:
387 		if (td != NULL) {
388 			error = priv_check(td, PRIV_NET_DELIFADDR);
389 			if (error)
390 				return (error);
391 		}
392 		/* FALLTHROUGH */
393 	case SIOCGLIFADDR:
394 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
395 	}
396 
397 	/*
398 	 * Find address for this interface, if it exists.
399 	 *
400 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
401 	 * only, and used the first interface address as the target of other
402 	 * operations (without checking ifra_addr).  This was because netinet
403 	 * code/API assumed at most 1 interface address per interface.
404 	 * Since IPv6 allows a node to assign multiple addresses
405 	 * on a single interface, we almost always look and check the
406 	 * presence of ifra_addr, and reject invalid ones here.
407 	 * It also decreases duplicated code among SIOC*_IN6 operations.
408 	 */
409 	switch (cmd) {
410 	case SIOCAIFADDR_IN6:
411 	case SIOCSIFPHYADDR_IN6:
412 		sa6 = &ifra->ifra_addr;
413 		break;
414 	case SIOCSIFADDR_IN6:
415 	case SIOCGIFADDR_IN6:
416 	case SIOCSIFDSTADDR_IN6:
417 	case SIOCSIFNETMASK_IN6:
418 	case SIOCGIFDSTADDR_IN6:
419 	case SIOCGIFNETMASK_IN6:
420 	case SIOCDIFADDR_IN6:
421 	case SIOCGIFPSRCADDR_IN6:
422 	case SIOCGIFPDSTADDR_IN6:
423 	case SIOCGIFAFLAG_IN6:
424 	case SIOCSNDFLUSH_IN6:
425 	case SIOCSPFXFLUSH_IN6:
426 	case SIOCSRTRFLUSH_IN6:
427 	case SIOCGIFALIFETIME_IN6:
428 	case SIOCSIFALIFETIME_IN6:
429 	case SIOCGIFSTAT_IN6:
430 	case SIOCGIFSTAT_ICMP6:
431 		sa6 = &ifr->ifr_addr;
432 		break;
433 	default:
434 		sa6 = NULL;
435 		break;
436 	}
437 	if (sa6 && sa6->sin6_family == AF_INET6) {
438 		if (sa6->sin6_scope_id != 0)
439 			error = sa6_embedscope(sa6, 0);
440 		else
441 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
442 		if (error != 0)
443 			return (error);
444 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
445 		    &sa6->sin6_addr)) != 0)
446 			return (error);
447 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
448 	} else
449 		ia = NULL;
450 
451 	switch (cmd) {
452 	case SIOCSIFADDR_IN6:
453 	case SIOCSIFDSTADDR_IN6:
454 	case SIOCSIFNETMASK_IN6:
455 		/*
456 		 * Since IPv6 allows a node to assign multiple addresses
457 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
458 		 */
459 		/* we decided to obsolete this command (20000704) */
460 		error = EINVAL;
461 		goto out;
462 
463 	case SIOCDIFADDR_IN6:
464 		/*
465 		 * for IPv4, we look for existing in_ifaddr here to allow
466 		 * "ifconfig if0 delete" to remove the first IPv4 address on
467 		 * the interface.  For IPv6, as the spec allows multiple
468 		 * interface address from the day one, we consider "remove the
469 		 * first one" semantics to be not preferable.
470 		 */
471 		if (ia == NULL) {
472 			error = EADDRNOTAVAIL;
473 			goto out;
474 		}
475 		/* FALLTHROUGH */
476 	case SIOCAIFADDR_IN6:
477 		/*
478 		 * We always require users to specify a valid IPv6 address for
479 		 * the corresponding operation.
480 		 */
481 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
482 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
483 			error = EAFNOSUPPORT;
484 			goto out;
485 		}
486 
487 		if (td != NULL) {
488 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
489 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
490 			if (error)
491 				goto out;
492 		}
493 		break;
494 
495 	case SIOCGIFADDR_IN6:
496 		/* This interface is basically deprecated. use SIOCGIFCONF. */
497 		/* FALLTHROUGH */
498 	case SIOCGIFAFLAG_IN6:
499 	case SIOCGIFNETMASK_IN6:
500 	case SIOCGIFDSTADDR_IN6:
501 	case SIOCGIFALIFETIME_IN6:
502 		/* must think again about its semantics */
503 		if (ia == NULL) {
504 			error = EADDRNOTAVAIL;
505 			goto out;
506 		}
507 		break;
508 
509 	case SIOCSIFALIFETIME_IN6:
510 	    {
511 		struct in6_addrlifetime *lt;
512 
513 		if (td != NULL) {
514 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
515 			if (error)
516 				goto out;
517 		}
518 		if (ia == NULL) {
519 			error = EADDRNOTAVAIL;
520 			goto out;
521 		}
522 		/* sanity for overflow - beware unsigned */
523 		lt = &ifr->ifr_ifru.ifru_lifetime;
524 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
525 		    lt->ia6t_vltime + time_second < time_second) {
526 			error = EINVAL;
527 			goto out;
528 		}
529 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
530 		    lt->ia6t_pltime + time_second < time_second) {
531 			error = EINVAL;
532 			goto out;
533 		}
534 		break;
535 	    }
536 	}
537 
538 	switch (cmd) {
539 	case SIOCGIFADDR_IN6:
540 		ifr->ifr_addr = ia->ia_addr;
541 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
542 			goto out;
543 		break;
544 
545 	case SIOCGIFDSTADDR_IN6:
546 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
547 			error = EINVAL;
548 			goto out;
549 		}
550 		/*
551 		 * XXX: should we check if ifa_dstaddr is NULL and return
552 		 * an error?
553 		 */
554 		ifr->ifr_dstaddr = ia->ia_dstaddr;
555 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
556 			goto out;
557 		break;
558 
559 	case SIOCGIFNETMASK_IN6:
560 		ifr->ifr_addr = ia->ia_prefixmask;
561 		break;
562 
563 	case SIOCGIFAFLAG_IN6:
564 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
565 		break;
566 
567 	case SIOCGIFSTAT_IN6:
568 		if (ifp == NULL) {
569 			error = EINVAL;
570 			goto out;
571 		}
572 		bzero(&ifr->ifr_ifru.ifru_stat,
573 		    sizeof(ifr->ifr_ifru.ifru_stat));
574 		ifr->ifr_ifru.ifru_stat =
575 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
576 		break;
577 
578 	case SIOCGIFSTAT_ICMP6:
579 		if (ifp == NULL) {
580 			error = EINVAL;
581 			goto out;
582 		}
583 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
584 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
585 		ifr->ifr_ifru.ifru_icmp6stat =
586 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
587 		break;
588 
589 	case SIOCGIFALIFETIME_IN6:
590 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
591 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
592 			time_t maxexpire;
593 			struct in6_addrlifetime *retlt =
594 			    &ifr->ifr_ifru.ifru_lifetime;
595 
596 			/*
597 			 * XXX: adjust expiration time assuming time_t is
598 			 * signed.
599 			 */
600 			maxexpire = (-1) &
601 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
602 			if (ia->ia6_lifetime.ia6t_vltime <
603 			    maxexpire - ia->ia6_updatetime) {
604 				retlt->ia6t_expire = ia->ia6_updatetime +
605 				    ia->ia6_lifetime.ia6t_vltime;
606 			} else
607 				retlt->ia6t_expire = maxexpire;
608 		}
609 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
610 			time_t maxexpire;
611 			struct in6_addrlifetime *retlt =
612 			    &ifr->ifr_ifru.ifru_lifetime;
613 
614 			/*
615 			 * XXX: adjust expiration time assuming time_t is
616 			 * signed.
617 			 */
618 			maxexpire = (-1) &
619 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
620 			if (ia->ia6_lifetime.ia6t_pltime <
621 			    maxexpire - ia->ia6_updatetime) {
622 				retlt->ia6t_preferred = ia->ia6_updatetime +
623 				    ia->ia6_lifetime.ia6t_pltime;
624 			} else
625 				retlt->ia6t_preferred = maxexpire;
626 		}
627 		break;
628 
629 	case SIOCSIFALIFETIME_IN6:
630 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
631 		/* for sanity */
632 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
633 			ia->ia6_lifetime.ia6t_expire =
634 				time_second + ia->ia6_lifetime.ia6t_vltime;
635 		} else
636 			ia->ia6_lifetime.ia6t_expire = 0;
637 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
638 			ia->ia6_lifetime.ia6t_preferred =
639 				time_second + ia->ia6_lifetime.ia6t_pltime;
640 		} else
641 			ia->ia6_lifetime.ia6t_preferred = 0;
642 		break;
643 
644 	case SIOCAIFADDR_IN6:
645 	{
646 		int i;
647 		struct nd_prefixctl pr0;
648 		struct nd_prefix *pr;
649 
650 		/*
651 		 * first, make or update the interface address structure,
652 		 * and link it to the list.
653 		 */
654 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
655 			goto out;
656 		if (ia != NULL)
657 			ifa_free(&ia->ia_ifa);
658 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
659 		    == NULL) {
660 			/*
661 			 * this can happen when the user specify the 0 valid
662 			 * lifetime.
663 			 */
664 			break;
665 		}
666 
667 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
668 			if (carp_attach_p != NULL)
669 				error = (*carp_attach_p)(&ia->ia_ifa,
670 				    ifra->ifra_vhid);
671 			else
672 				error = EPROTONOSUPPORT;
673 			if (error)
674 				goto out;
675 			else
676 				carp_attached = 1;
677 		}
678 
679 		/*
680 		 * then, make the prefix on-link on the interface.
681 		 * XXX: we'd rather create the prefix before the address, but
682 		 * we need at least one address to install the corresponding
683 		 * interface route, so we configure the address first.
684 		 */
685 
686 		/*
687 		 * convert mask to prefix length (prefixmask has already
688 		 * been validated in in6_update_ifa().
689 		 */
690 		bzero(&pr0, sizeof(pr0));
691 		pr0.ndpr_ifp = ifp;
692 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
693 		    NULL);
694 		if (pr0.ndpr_plen == 128) {
695 			break;	/* we don't need to install a host route. */
696 		}
697 		pr0.ndpr_prefix = ifra->ifra_addr;
698 		/* apply the mask for safety. */
699 		for (i = 0; i < 4; i++) {
700 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
701 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
702 		}
703 		/*
704 		 * XXX: since we don't have an API to set prefix (not address)
705 		 * lifetimes, we just use the same lifetimes as addresses.
706 		 * The (temporarily) installed lifetimes can be overridden by
707 		 * later advertised RAs (when accept_rtadv is non 0), which is
708 		 * an intended behavior.
709 		 */
710 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
711 		pr0.ndpr_raf_auto =
712 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
713 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
714 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
715 
716 		/* add the prefix if not yet. */
717 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
718 			/*
719 			 * nd6_prelist_add will install the corresponding
720 			 * interface route.
721 			 */
722 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
723 				if (carp_attached)
724 					(*carp_detach_p)(&ia->ia_ifa);
725 				goto out;
726 			}
727 			if (pr == NULL) {
728 				if (carp_attached)
729 					(*carp_detach_p)(&ia->ia_ifa);
730 				log(LOG_ERR, "nd6_prelist_add succeeded but "
731 				    "no prefix\n");
732 				error = EINVAL;
733 				goto out;
734 			}
735 		}
736 
737 		/* relate the address to the prefix */
738 		if (ia->ia6_ndpr == NULL) {
739 			ia->ia6_ndpr = pr;
740 			pr->ndpr_refcnt++;
741 
742 			/*
743 			 * If this is the first autoconf address from the
744 			 * prefix, create a temporary address as well
745 			 * (when required).
746 			 */
747 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
748 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
749 				int e;
750 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
751 					log(LOG_NOTICE, "in6_control: failed "
752 					    "to create a temporary address, "
753 					    "errno=%d\n", e);
754 				}
755 			}
756 		}
757 
758 		/*
759 		 * this might affect the status of autoconfigured addresses,
760 		 * that is, this address might make other addresses detached.
761 		 */
762 		pfxlist_onlink_check();
763 		if (error == 0 && ia) {
764 			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
765 				/*
766 				 * Try to clear the flag when a new
767 				 * IPv6 address is added onto an
768 				 * IFDISABLED interface and it
769 				 * succeeds.
770 				 */
771 				struct in6_ndireq nd;
772 
773 				memset(&nd, 0, sizeof(nd));
774 				nd.ndi.flags = ND_IFINFO(ifp)->flags;
775 				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
776 				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
777 				    (caddr_t)&nd, ifp) < 0)
778 					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
779 					    "SIOCSIFINFO_FLAGS for -ifdisabled "
780 					    "failed.");
781 				/*
782 				 * Ignore failure of clearing the flag
783 				 * intentionally.  The failure means
784 				 * address duplication was detected.
785 				 */
786 			}
787 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
788 		}
789 		break;
790 	}
791 
792 	case SIOCDIFADDR_IN6:
793 	{
794 		struct nd_prefix *pr;
795 
796 		/*
797 		 * If the address being deleted is the only one that owns
798 		 * the corresponding prefix, expire the prefix as well.
799 		 * XXX: theoretically, we don't have to worry about such
800 		 * relationship, since we separate the address management
801 		 * and the prefix management.  We do this, however, to provide
802 		 * as much backward compatibility as possible in terms of
803 		 * the ioctl operation.
804 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
805 		 */
806 		pr = ia->ia6_ndpr;
807 		in6_purgeaddr(&ia->ia_ifa);
808 		if (pr && pr->ndpr_refcnt == 0)
809 			prelist_remove(pr);
810 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
811 		break;
812 	}
813 
814 	default:
815 		if (ifp == NULL || ifp->if_ioctl == 0) {
816 			error = EOPNOTSUPP;
817 			goto out;
818 		}
819 		error = (*ifp->if_ioctl)(ifp, cmd, data);
820 		goto out;
821 	}
822 
823 	error = 0;
824 out:
825 	if (ia != NULL)
826 		ifa_free(&ia->ia_ifa);
827 	return (error);
828 }
829 
830 
831 /*
832  * Join necessary multicast groups.  Factored out from in6_update_ifa().
833  * This entire work should only be done once, for the default FIB.
834  */
835 static int
836 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
837     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
838 {
839 	char ip6buf[INET6_ADDRSTRLEN];
840 	struct sockaddr_in6 mltaddr, mltmask;
841 	struct in6_addr llsol;
842 	struct in6_multi_mship *imm;
843 	struct rtentry *rt;
844 	int delay, error;
845 
846 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
847 
848 	/* Join solicited multicast addr for new host id. */
849 	bzero(&llsol, sizeof(struct in6_addr));
850 	llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
851 	llsol.s6_addr32[1] = 0;
852 	llsol.s6_addr32[2] = htonl(1);
853 	llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
854 	llsol.s6_addr8[12] = 0xff;
855 	if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
856 		/* XXX: should not happen */
857 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
858 		goto cleanup;
859 	}
860 	delay = 0;
861 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
862 		/*
863 		 * We need a random delay for DAD on the address being
864 		 * configured.  It also means delaying transmission of the
865 		 * corresponding MLD report to avoid report collision.
866 		 * [RFC 4861, Section 6.3.7]
867 		 */
868 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
869 	}
870 	imm = in6_joingroup(ifp, &llsol, &error, delay);
871 	if (imm == NULL) {
872 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
873 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &llsol),
874 		    if_name(ifp), error));
875 		goto cleanup;
876 	}
877 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
878 	*in6m_sol = imm->i6mm_maddr;
879 
880 	bzero(&mltmask, sizeof(mltmask));
881 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
882 	mltmask.sin6_family = AF_INET6;
883 	mltmask.sin6_addr = in6mask32;
884 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
885 
886 	/*
887 	 * Join link-local all-nodes address.
888 	 */
889 	bzero(&mltaddr, sizeof(mltaddr));
890 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
891 	mltaddr.sin6_family = AF_INET6;
892 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
893 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
894 		goto cleanup; /* XXX: should not fail */
895 
896 	/*
897 	 * XXX: do we really need this automatic routes?  We should probably
898 	 * reconsider this stuff.  Most applications actually do not need the
899 	 * routes, since they usually specify the outgoing interface.
900 	 */
901 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
902 	if (rt != NULL) {
903 		/* XXX: only works in !SCOPEDROUTING case. */
904 		if (memcmp(&mltaddr.sin6_addr,
905 		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
906 		    MLTMASK_LEN)) {
907 			RTFREE_LOCKED(rt);
908 			rt = NULL;
909 		}
910 	}
911 	if (rt == NULL) {
912 		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
913 		    (struct sockaddr *)&ia->ia_addr,
914 		    (struct sockaddr *)&mltmask, RTF_UP,
915 		    (struct rtentry **)0, RT_DEFAULT_FIB);
916 		if (error)
917 			goto cleanup;
918 	} else
919 		RTFREE_LOCKED(rt);
920 
921 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
922 	if (imm == NULL) {
923 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
924 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
925 		    &mltaddr.sin6_addr), if_name(ifp), error));
926 		goto cleanup;
927 	}
928 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
929 
930 	/*
931 	 * Join node information group address.
932 	 */
933 	delay = 0;
934 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
935 		/*
936 		 * The spec does not say anything about delay for this group,
937 		 * but the same logic should apply.
938 		 */
939 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
940 	}
941 	if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
942 		/* XXX jinmei */
943 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
944 		if (imm == NULL)
945 			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
946 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
947 			    &mltaddr.sin6_addr), if_name(ifp), error));
948 			/* XXX not very fatal, go on... */
949 		else
950 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
951 	}
952 
953 	/*
954 	 * Join interface-local all-nodes address.
955 	 * (ff01::1%ifN, and ff01::%ifN/32)
956 	 */
957 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
958 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
959 		goto cleanup; /* XXX: should not fail */
960 	/* XXX: again, do we really need the route? */
961 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
962 	if (rt != NULL) {
963 		if (memcmp(&mltaddr.sin6_addr,
964 		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
965 		    MLTMASK_LEN)) {
966 			RTFREE_LOCKED(rt);
967 			rt = NULL;
968 		}
969 	}
970 	if (rt == NULL) {
971 		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
972 		    (struct sockaddr *)&ia->ia_addr,
973 		    (struct sockaddr *)&mltmask, RTF_UP,
974 		    (struct rtentry **)0, RT_DEFAULT_FIB);
975 		if (error)
976 			goto cleanup;
977 	} else
978 		RTFREE_LOCKED(rt);
979 
980 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
981 	if (imm == NULL) {
982 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
983 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
984 		    &mltaddr.sin6_addr), if_name(ifp), error));
985 		goto cleanup;
986 	}
987 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
988 #undef	MLTMASK_LEN
989 
990 cleanup:
991 	return (error);
992 }
993 
994 /*
995  * Update parameters of an IPv6 interface address.
996  * If necessary, a new entry is created and linked into address chains.
997  * This function is separated from in6_control().
998  */
999 int
1000 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1001     struct in6_ifaddr *ia, int flags)
1002 {
1003 	int error = 0, hostIsNew = 0, plen = -1;
1004 	struct sockaddr_in6 dst6;
1005 	struct in6_addrlifetime *lt;
1006 	struct in6_multi *in6m_sol;
1007 	int delay;
1008 	char ip6buf[INET6_ADDRSTRLEN];
1009 
1010 	/* Validate parameters */
1011 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1012 		return (EINVAL);
1013 
1014 	/*
1015 	 * The destination address for a p2p link must have a family
1016 	 * of AF_UNSPEC or AF_INET6.
1017 	 */
1018 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1019 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1020 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1021 		return (EAFNOSUPPORT);
1022 	/*
1023 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1024 	 * does not carry fields other than sin6_len.
1025 	 */
1026 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1027 		return (EINVAL);
1028 	/*
1029 	 * Because the IPv6 address architecture is classless, we require
1030 	 * users to specify a (non 0) prefix length (mask) for a new address.
1031 	 * We also require the prefix (when specified) mask is valid, and thus
1032 	 * reject a non-consecutive mask.
1033 	 */
1034 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1035 		return (EINVAL);
1036 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1037 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1038 		    (u_char *)&ifra->ifra_prefixmask +
1039 		    ifra->ifra_prefixmask.sin6_len);
1040 		if (plen <= 0)
1041 			return (EINVAL);
1042 	} else {
1043 		/*
1044 		 * In this case, ia must not be NULL.  We just use its prefix
1045 		 * length.
1046 		 */
1047 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1048 	}
1049 	/*
1050 	 * If the destination address on a p2p interface is specified,
1051 	 * and the address is a scoped one, validate/set the scope
1052 	 * zone identifier.
1053 	 */
1054 	dst6 = ifra->ifra_dstaddr;
1055 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1056 	    (dst6.sin6_family == AF_INET6)) {
1057 		struct in6_addr in6_tmp;
1058 		u_int32_t zoneid;
1059 
1060 		in6_tmp = dst6.sin6_addr;
1061 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1062 			return (EINVAL); /* XXX: should be impossible */
1063 
1064 		if (dst6.sin6_scope_id != 0) {
1065 			if (dst6.sin6_scope_id != zoneid)
1066 				return (EINVAL);
1067 		} else		/* user omit to specify the ID. */
1068 			dst6.sin6_scope_id = zoneid;
1069 
1070 		/* convert into the internal form */
1071 		if (sa6_embedscope(&dst6, 0))
1072 			return (EINVAL); /* XXX: should be impossible */
1073 	}
1074 	/*
1075 	 * The destination address can be specified only for a p2p or a
1076 	 * loopback interface.  If specified, the corresponding prefix length
1077 	 * must be 128.
1078 	 */
1079 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1080 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1081 			/* XXX: noisy message */
1082 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1083 			    "be specified for a p2p or a loopback IF only\n"));
1084 			return (EINVAL);
1085 		}
1086 		if (plen != 128) {
1087 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1088 			    "be 128 when dstaddr is specified\n"));
1089 			return (EINVAL);
1090 		}
1091 	}
1092 	/* lifetime consistency check */
1093 	lt = &ifra->ifra_lifetime;
1094 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1095 		return (EINVAL);
1096 	if (lt->ia6t_vltime == 0) {
1097 		/*
1098 		 * the following log might be noisy, but this is a typical
1099 		 * configuration mistake or a tool's bug.
1100 		 */
1101 		nd6log((LOG_INFO,
1102 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1103 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1104 
1105 		if (ia == NULL)
1106 			return (0); /* there's nothing to do */
1107 	}
1108 
1109 	/*
1110 	 * If this is a new address, allocate a new ifaddr and link it
1111 	 * into chains.
1112 	 */
1113 	if (ia == NULL) {
1114 		hostIsNew = 1;
1115 		/*
1116 		 * When in6_update_ifa() is called in a process of a received
1117 		 * RA, it is called under an interrupt context.  So, we should
1118 		 * call malloc with M_NOWAIT.
1119 		 */
1120 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
1121 		    M_NOWAIT);
1122 		if (ia == NULL)
1123 			return (ENOBUFS);
1124 		bzero((caddr_t)ia, sizeof(*ia));
1125 		ifa_init(&ia->ia_ifa);
1126 		LIST_INIT(&ia->ia6_memberships);
1127 		/* Initialize the address and masks, and put time stamp */
1128 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1129 		ia->ia_addr.sin6_family = AF_INET6;
1130 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1131 		ia->ia6_createtime = time_second;
1132 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1133 			/*
1134 			 * XXX: some functions expect that ifa_dstaddr is not
1135 			 * NULL for p2p interfaces.
1136 			 */
1137 			ia->ia_ifa.ifa_dstaddr =
1138 			    (struct sockaddr *)&ia->ia_dstaddr;
1139 		} else {
1140 			ia->ia_ifa.ifa_dstaddr = NULL;
1141 		}
1142 		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1143 		ia->ia_ifp = ifp;
1144 		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1145 		IF_ADDR_WLOCK(ifp);
1146 		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1147 		IF_ADDR_WUNLOCK(ifp);
1148 
1149 		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1150 		IN6_IFADDR_WLOCK();
1151 		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1152 		IN6_IFADDR_WUNLOCK();
1153 	}
1154 
1155 	/* update timestamp */
1156 	ia->ia6_updatetime = time_second;
1157 
1158 	/* set prefix mask */
1159 	if (ifra->ifra_prefixmask.sin6_len) {
1160 		/*
1161 		 * We prohibit changing the prefix length of an existing
1162 		 * address, because
1163 		 * + such an operation should be rare in IPv6, and
1164 		 * + the operation would confuse prefix management.
1165 		 */
1166 		if (ia->ia_prefixmask.sin6_len &&
1167 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1168 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1169 			    " existing (%s) address should not be changed\n",
1170 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1171 			error = EINVAL;
1172 			goto unlink;
1173 		}
1174 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1175 	}
1176 
1177 	/*
1178 	 * If a new destination address is specified, scrub the old one and
1179 	 * install the new destination.  Note that the interface must be
1180 	 * p2p or loopback (see the check above.)
1181 	 */
1182 	if (dst6.sin6_family == AF_INET6 &&
1183 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1184 		int e;
1185 
1186 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1187 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1188 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1189 			    "a route to the old destination: %s\n",
1190 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1191 			/* proceed anyway... */
1192 		} else
1193 			ia->ia_flags &= ~IFA_ROUTE;
1194 		ia->ia_dstaddr = dst6;
1195 	}
1196 
1197 	/*
1198 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1199 	 * to see if the address is deprecated or invalidated, but initialize
1200 	 * these members for applications.
1201 	 */
1202 	ia->ia6_lifetime = ifra->ifra_lifetime;
1203 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1204 		ia->ia6_lifetime.ia6t_expire =
1205 		    time_second + ia->ia6_lifetime.ia6t_vltime;
1206 	} else
1207 		ia->ia6_lifetime.ia6t_expire = 0;
1208 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1209 		ia->ia6_lifetime.ia6t_preferred =
1210 		    time_second + ia->ia6_lifetime.ia6t_pltime;
1211 	} else
1212 		ia->ia6_lifetime.ia6t_preferred = 0;
1213 
1214 	/* reset the interface and routing table appropriately. */
1215 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1216 		goto unlink;
1217 
1218 	/*
1219 	 * configure address flags.
1220 	 */
1221 	ia->ia6_flags = ifra->ifra_flags;
1222 	/*
1223 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1224 	 * userland, make it deprecated.
1225 	 */
1226 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1227 		ia->ia6_lifetime.ia6t_pltime = 0;
1228 		ia->ia6_lifetime.ia6t_preferred = time_second;
1229 	}
1230 	/*
1231 	 * Make the address tentative before joining multicast addresses,
1232 	 * so that corresponding MLD responses would not have a tentative
1233 	 * source address.
1234 	 */
1235 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1236 	if (hostIsNew && in6if_do_dad(ifp))
1237 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1238 
1239 	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
1240 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1241 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1242 
1243 	/*
1244 	 * We are done if we have simply modified an existing address.
1245 	 */
1246 	if (!hostIsNew)
1247 		return (error);
1248 
1249 	/*
1250 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1251 	 * not just go to unlink.
1252 	 */
1253 
1254 	/* Join necessary multicast groups. */
1255 	in6m_sol = NULL;
1256 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1257 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1258 		if (error)
1259 			goto cleanup;
1260 	}
1261 
1262 	/*
1263 	 * Perform DAD, if needed.
1264 	 * XXX It may be of use, if we can administratively disable DAD.
1265 	 */
1266 	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1267 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1268 	{
1269 		int mindelay, maxdelay;
1270 
1271 		delay = 0;
1272 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1273 			/*
1274 			 * We need to impose a delay before sending an NS
1275 			 * for DAD.  Check if we also needed a delay for the
1276 			 * corresponding MLD message.  If we did, the delay
1277 			 * should be larger than the MLD delay (this could be
1278 			 * relaxed a bit, but this simple logic is at least
1279 			 * safe).
1280 			 * XXX: Break data hiding guidelines and look at
1281 			 * state for the solicited multicast group.
1282 			 */
1283 			mindelay = 0;
1284 			if (in6m_sol != NULL &&
1285 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1286 				mindelay = in6m_sol->in6m_timer;
1287 			}
1288 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1289 			if (maxdelay - mindelay == 0)
1290 				delay = 0;
1291 			else {
1292 				delay =
1293 				    (arc4random() % (maxdelay - mindelay)) +
1294 				    mindelay;
1295 			}
1296 		}
1297 		nd6_dad_start((struct ifaddr *)ia, delay);
1298 	}
1299 
1300 	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1301 	ifa_free(&ia->ia_ifa);
1302 	return (error);
1303 
1304   unlink:
1305 	/*
1306 	 * XXX: if a change of an existing address failed, keep the entry
1307 	 * anyway.
1308 	 */
1309 	if (hostIsNew) {
1310 		in6_unlink_ifa(ia, ifp);
1311 		ifa_free(&ia->ia_ifa);
1312 	}
1313 	return (error);
1314 
1315   cleanup:
1316 	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1317 	ifa_free(&ia->ia_ifa);
1318 	in6_purgeaddr(&ia->ia_ifa);
1319 	return error;
1320 }
1321 
1322 /*
1323  * Leave multicast groups.  Factored out from in6_purgeaddr().
1324  * This entire work should only be done once, for the default FIB.
1325  */
1326 static int
1327 in6_purgeaddr_mc(struct ifnet *ifp, struct in6_ifaddr *ia, struct ifaddr *ifa0)
1328 {
1329 	struct sockaddr_in6 mltaddr, mltmask;
1330 	struct in6_multi_mship *imm;
1331 	struct rtentry *rt;
1332 	struct sockaddr_in6 sin6;
1333 	int error;
1334 
1335 	/*
1336 	 * Leave from multicast groups we have joined for the interface.
1337 	 */
1338 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1339 		LIST_REMOVE(imm, i6mm_chain);
1340 		in6_leavegroup(imm);
1341 	}
1342 
1343 	/*
1344 	 * Remove the link-local all-nodes address.
1345 	 */
1346 	bzero(&mltmask, sizeof(mltmask));
1347 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1348 	mltmask.sin6_family = AF_INET6;
1349 	mltmask.sin6_addr = in6mask32;
1350 
1351 	bzero(&mltaddr, sizeof(mltaddr));
1352 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1353 	mltaddr.sin6_family = AF_INET6;
1354 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1355 
1356 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1357 		return (error);
1358 
1359 	/*
1360 	 * As for the mltaddr above, proactively prepare the sin6 to avoid
1361 	 * rtentry un- and re-locking.
1362 	 */
1363 	if (ifa0 != NULL) {
1364 		bzero(&sin6, sizeof(sin6));
1365 		sin6.sin6_len = sizeof(sin6);
1366 		sin6.sin6_family = AF_INET6;
1367 		memcpy(&sin6.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1368 		    sizeof(sin6.sin6_addr));
1369 		error = in6_setscope(&sin6.sin6_addr, ifa0->ifa_ifp, NULL);
1370 		if (error != 0)
1371 			return (error);
1372 	}
1373 
1374 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1375 	if (rt != NULL && rt->rt_gateway != NULL &&
1376 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1377 		    &ia->ia_addr.sin6_addr,
1378 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1379 		/*
1380 		 * If no more IPv6 address exists on this interface then
1381 		 * remove the multicast address route.
1382 		 */
1383 		if (ifa0 == NULL) {
1384 			memcpy(&mltaddr.sin6_addr,
1385 			    &satosin6(rt_key(rt))->sin6_addr,
1386 			    sizeof(mltaddr.sin6_addr));
1387 			RTFREE_LOCKED(rt);
1388 			error = in6_rtrequest(RTM_DELETE,
1389 			    (struct sockaddr *)&mltaddr,
1390 			    (struct sockaddr *)&ia->ia_addr,
1391 			    (struct sockaddr *)&mltmask, RTF_UP,
1392 			    (struct rtentry **)0, RT_DEFAULT_FIB);
1393 			if (error)
1394 				log(LOG_INFO, "%s: link-local all-nodes "
1395 				    "multicast address deletion error\n",
1396 				    __func__);
1397 		} else {
1398 			/*
1399 			 * Replace the gateway of the route.
1400 			 */
1401 			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1402 			RTFREE_LOCKED(rt);
1403 		}
1404 	} else {
1405 		if (rt != NULL)
1406 			RTFREE_LOCKED(rt);
1407 	}
1408 
1409 	/*
1410 	 * Remove the node-local all-nodes address.
1411 	 */
1412 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1413 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1414 		return (error);
1415 
1416 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1417 	if (rt != NULL && rt->rt_gateway != NULL &&
1418 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1419 		    &ia->ia_addr.sin6_addr,
1420 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1421 		/*
1422 		 * If no more IPv6 address exists on this interface then
1423 		 * remove the multicast address route.
1424 		 */
1425 		if (ifa0 == NULL) {
1426 			memcpy(&mltaddr.sin6_addr,
1427 			    &satosin6(rt_key(rt))->sin6_addr,
1428 			    sizeof(mltaddr.sin6_addr));
1429 
1430 			RTFREE_LOCKED(rt);
1431 			error = in6_rtrequest(RTM_DELETE,
1432 			    (struct sockaddr *)&mltaddr,
1433 			    (struct sockaddr *)&ia->ia_addr,
1434 			    (struct sockaddr *)&mltmask, RTF_UP,
1435 			    (struct rtentry **)0, RT_DEFAULT_FIB);
1436 			if (error)
1437 				log(LOG_INFO, "%s: node-local all-nodes"
1438 				    "multicast address deletion error\n",
1439 				    __func__);
1440 		} else {
1441 			/*
1442 			 * Replace the gateway of the route.
1443 			 */
1444 			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1445 			RTFREE_LOCKED(rt);
1446 		}
1447 	} else {
1448 		if (rt != NULL)
1449 			RTFREE_LOCKED(rt);
1450 	}
1451 
1452 	return (0);
1453 }
1454 
1455 void
1456 in6_purgeaddr(struct ifaddr *ifa)
1457 {
1458 	struct ifnet *ifp = ifa->ifa_ifp;
1459 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1460 	int plen, error;
1461 	struct ifaddr *ifa0;
1462 
1463 	if (ifa->ifa_carp)
1464 		(*carp_detach_p)(ifa);
1465 
1466 	/*
1467 	 * find another IPv6 address as the gateway for the
1468 	 * link-local and node-local all-nodes multicast
1469 	 * address routes
1470 	 */
1471 	IF_ADDR_RLOCK(ifp);
1472 	TAILQ_FOREACH(ifa0, &ifp->if_addrhead, ifa_link) {
1473 		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1474 		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1475 		    &ia->ia_addr.sin6_addr, sizeof(struct in6_addr)) == 0)
1476 			continue;
1477 		else
1478 			break;
1479 	}
1480 	if (ifa0 != NULL)
1481 		ifa_ref(ifa0);
1482 	IF_ADDR_RUNLOCK(ifp);
1483 
1484 	/*
1485 	 * Remove the loopback route to the interface address.
1486 	 * The check for the current setting of "nd6_useloopback"
1487 	 * is not needed.
1488 	 */
1489 	if (ia->ia_flags & IFA_RTSELF) {
1490 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1491 		    (struct sockaddr *)&ia->ia_addr);
1492 		if (error == 0)
1493 			ia->ia_flags &= ~IFA_RTSELF;
1494 	}
1495 
1496 	/* stop DAD processing */
1497 	nd6_dad_stop(ifa);
1498 
1499 	/* Remove local address entry from lltable. */
1500 	in6_ifremloop(ifa);
1501 
1502 	/* Leave multicast groups. */
1503 	error = in6_purgeaddr_mc(ifp, ia, ifa0);
1504 
1505 	if (ifa0 != NULL)
1506 		ifa_free(ifa0);
1507 
1508 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1509 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1510 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1511 		    (ia->ia_dstaddr.sin6_family == AF_INET6) ? RTF_HOST : 0);
1512 		if (error != 0)
1513 			log(LOG_INFO, "%s: err=%d, destination address delete "
1514 			    "failed\n", __func__, error);
1515 		ia->ia_flags &= ~IFA_ROUTE;
1516 	}
1517 
1518 	in6_unlink_ifa(ia, ifp);
1519 }
1520 
1521 static void
1522 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1523 {
1524 
1525 	IF_ADDR_WLOCK(ifp);
1526 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1527 	IF_ADDR_WUNLOCK(ifp);
1528 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1529 
1530 	/*
1531 	 * Defer the release of what might be the last reference to the
1532 	 * in6_ifaddr so that it can't be freed before the remainder of the
1533 	 * cleanup.
1534 	 */
1535 	IN6_IFADDR_WLOCK();
1536 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1537 	IN6_IFADDR_WUNLOCK();
1538 
1539 	/*
1540 	 * Release the reference to the base prefix.  There should be a
1541 	 * positive reference.
1542 	 */
1543 	if (ia->ia6_ndpr == NULL) {
1544 		nd6log((LOG_NOTICE,
1545 		    "in6_unlink_ifa: autoconf'ed address "
1546 		    "%p has no prefix\n", ia));
1547 	} else {
1548 		ia->ia6_ndpr->ndpr_refcnt--;
1549 		ia->ia6_ndpr = NULL;
1550 	}
1551 
1552 	/*
1553 	 * Also, if the address being removed is autoconf'ed, call
1554 	 * pfxlist_onlink_check() since the release might affect the status of
1555 	 * other (detached) addresses.
1556 	 */
1557 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1558 		pfxlist_onlink_check();
1559 	}
1560 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1561 }
1562 
1563 void
1564 in6_purgeif(struct ifnet *ifp)
1565 {
1566 	struct ifaddr *ifa, *nifa;
1567 
1568 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1569 		if (ifa->ifa_addr->sa_family != AF_INET6)
1570 			continue;
1571 		in6_purgeaddr(ifa);
1572 	}
1573 
1574 	in6_ifdetach(ifp);
1575 }
1576 
1577 /*
1578  * SIOC[GAD]LIFADDR.
1579  *	SIOCGLIFADDR: get first address. (?)
1580  *	SIOCGLIFADDR with IFLR_PREFIX:
1581  *		get first address that matches the specified prefix.
1582  *	SIOCALIFADDR: add the specified address.
1583  *	SIOCALIFADDR with IFLR_PREFIX:
1584  *		add the specified prefix, filling hostid part from
1585  *		the first link-local address.  prefixlen must be <= 64.
1586  *	SIOCDLIFADDR: delete the specified address.
1587  *	SIOCDLIFADDR with IFLR_PREFIX:
1588  *		delete the first address that matches the specified prefix.
1589  * return values:
1590  *	EINVAL on invalid parameters
1591  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1592  *	other values may be returned from in6_ioctl()
1593  *
1594  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1595  * this is to accomodate address naming scheme other than RFC2374,
1596  * in the future.
1597  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1598  * address encoding scheme. (see figure on page 8)
1599  */
1600 static int
1601 in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1602     struct ifnet *ifp, struct thread *td)
1603 {
1604 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1605 	struct ifaddr *ifa;
1606 	struct sockaddr *sa;
1607 
1608 	/* sanity checks */
1609 	if (!data || !ifp) {
1610 		panic("invalid argument to in6_lifaddr_ioctl");
1611 		/* NOTREACHED */
1612 	}
1613 
1614 	switch (cmd) {
1615 	case SIOCGLIFADDR:
1616 		/* address must be specified on GET with IFLR_PREFIX */
1617 		if ((iflr->flags & IFLR_PREFIX) == 0)
1618 			break;
1619 		/* FALLTHROUGH */
1620 	case SIOCALIFADDR:
1621 	case SIOCDLIFADDR:
1622 		/* address must be specified on ADD and DELETE */
1623 		sa = (struct sockaddr *)&iflr->addr;
1624 		if (sa->sa_family != AF_INET6)
1625 			return EINVAL;
1626 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1627 			return EINVAL;
1628 		/* XXX need improvement */
1629 		sa = (struct sockaddr *)&iflr->dstaddr;
1630 		if (sa->sa_family && sa->sa_family != AF_INET6)
1631 			return EINVAL;
1632 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1633 			return EINVAL;
1634 		break;
1635 	default: /* shouldn't happen */
1636 #if 0
1637 		panic("invalid cmd to in6_lifaddr_ioctl");
1638 		/* NOTREACHED */
1639 #else
1640 		return EOPNOTSUPP;
1641 #endif
1642 	}
1643 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1644 		return EINVAL;
1645 
1646 	switch (cmd) {
1647 	case SIOCALIFADDR:
1648 	    {
1649 		struct in6_aliasreq ifra;
1650 		struct in6_addr *hostid = NULL;
1651 		int prefixlen;
1652 
1653 		ifa = NULL;
1654 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1655 			struct sockaddr_in6 *sin6;
1656 
1657 			/*
1658 			 * hostid is to fill in the hostid part of the
1659 			 * address.  hostid points to the first link-local
1660 			 * address attached to the interface.
1661 			 */
1662 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1663 			if (!ifa)
1664 				return EADDRNOTAVAIL;
1665 			hostid = IFA_IN6(ifa);
1666 
1667 			/* prefixlen must be <= 64. */
1668 			if (64 < iflr->prefixlen) {
1669 				if (ifa != NULL)
1670 					ifa_free(ifa);
1671 				return EINVAL;
1672 			}
1673 			prefixlen = iflr->prefixlen;
1674 
1675 			/* hostid part must be zero. */
1676 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1677 			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1678 			    sin6->sin6_addr.s6_addr32[3] != 0) {
1679 				if (ifa != NULL)
1680 					ifa_free(ifa);
1681 				return EINVAL;
1682 			}
1683 		} else
1684 			prefixlen = iflr->prefixlen;
1685 
1686 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1687 		bzero(&ifra, sizeof(ifra));
1688 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1689 
1690 		bcopy(&iflr->addr, &ifra.ifra_addr,
1691 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1692 		if (hostid) {
1693 			/* fill in hostid part */
1694 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1695 			    hostid->s6_addr32[2];
1696 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1697 			    hostid->s6_addr32[3];
1698 		}
1699 
1700 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1701 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1702 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1703 			if (hostid) {
1704 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1705 				    hostid->s6_addr32[2];
1706 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1707 				    hostid->s6_addr32[3];
1708 			}
1709 		}
1710 		if (ifa != NULL)
1711 			ifa_free(ifa);
1712 
1713 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1714 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1715 
1716 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1717 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1718 	    }
1719 	case SIOCGLIFADDR:
1720 	case SIOCDLIFADDR:
1721 	    {
1722 		struct in6_ifaddr *ia;
1723 		struct in6_addr mask, candidate, match;
1724 		struct sockaddr_in6 *sin6;
1725 		int cmp;
1726 
1727 		bzero(&mask, sizeof(mask));
1728 		if (iflr->flags & IFLR_PREFIX) {
1729 			/* lookup a prefix rather than address. */
1730 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1731 
1732 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1733 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1734 			match.s6_addr32[0] &= mask.s6_addr32[0];
1735 			match.s6_addr32[1] &= mask.s6_addr32[1];
1736 			match.s6_addr32[2] &= mask.s6_addr32[2];
1737 			match.s6_addr32[3] &= mask.s6_addr32[3];
1738 
1739 			/* if you set extra bits, that's wrong */
1740 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1741 				return EINVAL;
1742 
1743 			cmp = 1;
1744 		} else {
1745 			if (cmd == SIOCGLIFADDR) {
1746 				/* on getting an address, take the 1st match */
1747 				cmp = 0;	/* XXX */
1748 			} else {
1749 				/* on deleting an address, do exact match */
1750 				in6_prefixlen2mask(&mask, 128);
1751 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1752 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1753 
1754 				cmp = 1;
1755 			}
1756 		}
1757 
1758 		IF_ADDR_RLOCK(ifp);
1759 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1760 			if (ifa->ifa_addr->sa_family != AF_INET6)
1761 				continue;
1762 			if (!cmp)
1763 				break;
1764 
1765 			/*
1766 			 * XXX: this is adhoc, but is necessary to allow
1767 			 * a user to specify fe80::/64 (not /10) for a
1768 			 * link-local address.
1769 			 */
1770 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1771 			in6_clearscope(&candidate);
1772 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1773 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1774 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1775 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1776 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1777 				break;
1778 		}
1779 		if (ifa != NULL)
1780 			ifa_ref(ifa);
1781 		IF_ADDR_RUNLOCK(ifp);
1782 		if (!ifa)
1783 			return EADDRNOTAVAIL;
1784 		ia = ifa2ia6(ifa);
1785 
1786 		if (cmd == SIOCGLIFADDR) {
1787 			int error;
1788 
1789 			/* fill in the if_laddrreq structure */
1790 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1791 			error = sa6_recoverscope(
1792 			    (struct sockaddr_in6 *)&iflr->addr);
1793 			if (error != 0) {
1794 				ifa_free(ifa);
1795 				return (error);
1796 			}
1797 
1798 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1799 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1800 				    ia->ia_dstaddr.sin6_len);
1801 				error = sa6_recoverscope(
1802 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1803 				if (error != 0) {
1804 					ifa_free(ifa);
1805 					return (error);
1806 				}
1807 			} else
1808 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1809 
1810 			iflr->prefixlen =
1811 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1812 
1813 			iflr->flags = ia->ia6_flags;	/* XXX */
1814 			ifa_free(ifa);
1815 
1816 			return 0;
1817 		} else {
1818 			struct in6_aliasreq ifra;
1819 
1820 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1821 			bzero(&ifra, sizeof(ifra));
1822 			bcopy(iflr->iflr_name, ifra.ifra_name,
1823 			    sizeof(ifra.ifra_name));
1824 
1825 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1826 			    ia->ia_addr.sin6_len);
1827 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1828 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1829 				    ia->ia_dstaddr.sin6_len);
1830 			} else {
1831 				bzero(&ifra.ifra_dstaddr,
1832 				    sizeof(ifra.ifra_dstaddr));
1833 			}
1834 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1835 			    ia->ia_prefixmask.sin6_len);
1836 
1837 			ifra.ifra_flags = ia->ia6_flags;
1838 			ifa_free(ifa);
1839 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1840 			    ifp, td);
1841 		}
1842 	    }
1843 	}
1844 
1845 	return EOPNOTSUPP;	/* just for safety */
1846 }
1847 
1848 /*
1849  * Initialize an interface's IPv6 address and routing table entry.
1850  */
1851 static int
1852 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1853     struct sockaddr_in6 *sin6, int newhost)
1854 {
1855 	int	error = 0, plen, ifacount = 0;
1856 	struct ifaddr *ifa;
1857 
1858 	/*
1859 	 * Give the interface a chance to initialize
1860 	 * if this is its first address,
1861 	 * and to validate the address if necessary.
1862 	 */
1863 	IF_ADDR_RLOCK(ifp);
1864 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1865 		if (ifa->ifa_addr->sa_family != AF_INET6)
1866 			continue;
1867 		ifacount++;
1868 	}
1869 	IF_ADDR_RUNLOCK(ifp);
1870 
1871 	ia->ia_addr = *sin6;
1872 
1873 	if (ifacount <= 1 && ifp->if_ioctl) {
1874 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1875 		if (error)
1876 			return (error);
1877 	}
1878 
1879 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1880 
1881 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1882 
1883 	/*
1884 	 * Special case:
1885 	 * If a new destination address is specified for a point-to-point
1886 	 * interface, install a route to the destination as an interface
1887 	 * direct route.
1888 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1889 	 * interface that share the same destination.
1890 	 */
1891 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1892 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1893 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1894 		int rtflags = RTF_UP | RTF_HOST;
1895 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1896 		if (error)
1897 			return (error);
1898 		ia->ia_flags |= IFA_ROUTE;
1899 		/*
1900 		 * Handle the case for ::1 .
1901 		 */
1902 		if (ifp->if_flags & IFF_LOOPBACK)
1903 			ia->ia_flags |= IFA_RTSELF;
1904 	}
1905 
1906 	/*
1907 	 * add a loopback route to self
1908 	 */
1909 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1910 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1911 		    (struct sockaddr *)&ia->ia_addr);
1912 		if (error == 0)
1913 			ia->ia_flags |= IFA_RTSELF;
1914 	}
1915 
1916 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1917 	if (newhost)
1918 		in6_ifaddloop(&(ia->ia_ifa));
1919 
1920 	return (error);
1921 }
1922 
1923 /*
1924  * Find an IPv6 interface link-local address specific to an interface.
1925  * ifaddr is returned referenced.
1926  */
1927 struct in6_ifaddr *
1928 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1929 {
1930 	struct ifaddr *ifa;
1931 
1932 	IF_ADDR_RLOCK(ifp);
1933 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1934 		if (ifa->ifa_addr->sa_family != AF_INET6)
1935 			continue;
1936 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1937 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1938 			    ignoreflags) != 0)
1939 				continue;
1940 			ifa_ref(ifa);
1941 			break;
1942 		}
1943 	}
1944 	IF_ADDR_RUNLOCK(ifp);
1945 
1946 	return ((struct in6_ifaddr *)ifa);
1947 }
1948 
1949 
1950 /*
1951  * find the internet address corresponding to a given interface and address.
1952  * ifaddr is returned referenced.
1953  */
1954 struct in6_ifaddr *
1955 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1956 {
1957 	struct ifaddr *ifa;
1958 
1959 	IF_ADDR_RLOCK(ifp);
1960 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1961 		if (ifa->ifa_addr->sa_family != AF_INET6)
1962 			continue;
1963 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1964 			ifa_ref(ifa);
1965 			break;
1966 		}
1967 	}
1968 	IF_ADDR_RUNLOCK(ifp);
1969 
1970 	return ((struct in6_ifaddr *)ifa);
1971 }
1972 
1973 /*
1974  * Convert IP6 address to printable (loggable) representation. Caller
1975  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1976  */
1977 static char digits[] = "0123456789abcdef";
1978 char *
1979 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1980 {
1981 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1982 	char *cp;
1983 	const u_int16_t *a = (const u_int16_t *)addr;
1984 	const u_int8_t *d;
1985 	int dcolon = 0, zero = 0;
1986 
1987 	cp = ip6buf;
1988 
1989 	for (i = 0; i < 8; i++) {
1990 		if (*(a + i) == 0) {
1991 			cnt++;
1992 			if (cnt == 1)
1993 				idx = i;
1994 		}
1995 		else if (maxcnt < cnt) {
1996 			maxcnt = cnt;
1997 			index = idx;
1998 			cnt = 0;
1999 		}
2000 	}
2001 	if (maxcnt < cnt) {
2002 		maxcnt = cnt;
2003 		index = idx;
2004 	}
2005 
2006 	for (i = 0; i < 8; i++) {
2007 		if (dcolon == 1) {
2008 			if (*a == 0) {
2009 				if (i == 7)
2010 					*cp++ = ':';
2011 				a++;
2012 				continue;
2013 			} else
2014 				dcolon = 2;
2015 		}
2016 		if (*a == 0) {
2017 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
2018 				if (i == 0)
2019 					*cp++ = ':';
2020 				*cp++ = ':';
2021 				dcolon = 1;
2022 			} else {
2023 				*cp++ = '0';
2024 				*cp++ = ':';
2025 			}
2026 			a++;
2027 			continue;
2028 		}
2029 		d = (const u_char *)a;
2030 		/* Try to eliminate leading zeros in printout like in :0001. */
2031 		zero = 1;
2032 		*cp = digits[*d >> 4];
2033 		if (*cp != '0') {
2034 			zero = 0;
2035 			cp++;
2036 		}
2037 		*cp = digits[*d++ & 0xf];
2038 		if (zero == 0 || (*cp != '0')) {
2039 			zero = 0;
2040 			cp++;
2041 		}
2042 		*cp = digits[*d >> 4];
2043 		if (zero == 0 || (*cp != '0')) {
2044 			zero = 0;
2045 			cp++;
2046 		}
2047 		*cp++ = digits[*d & 0xf];
2048 		*cp++ = ':';
2049 		a++;
2050 	}
2051 	*--cp = '\0';
2052 	return (ip6buf);
2053 }
2054 
2055 int
2056 in6_localaddr(struct in6_addr *in6)
2057 {
2058 	struct in6_ifaddr *ia;
2059 
2060 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2061 		return 1;
2062 
2063 	IN6_IFADDR_RLOCK();
2064 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2065 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2066 		    &ia->ia_prefixmask.sin6_addr)) {
2067 			IN6_IFADDR_RUNLOCK();
2068 			return 1;
2069 		}
2070 	}
2071 	IN6_IFADDR_RUNLOCK();
2072 
2073 	return (0);
2074 }
2075 
2076 /*
2077  * Return 1 if an internet address is for the local host and configured
2078  * on one of its interfaces.
2079  */
2080 int
2081 in6_localip(struct in6_addr *in6)
2082 {
2083 	struct in6_ifaddr *ia;
2084 
2085 	IN6_IFADDR_RLOCK();
2086 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2087 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2088 			IN6_IFADDR_RUNLOCK();
2089 			return (1);
2090 		}
2091 	}
2092 	IN6_IFADDR_RUNLOCK();
2093 	return (0);
2094 }
2095 
2096 
2097 int
2098 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2099 {
2100 	struct in6_ifaddr *ia;
2101 
2102 	IN6_IFADDR_RLOCK();
2103 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2104 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2105 		    &sa6->sin6_addr) &&
2106 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
2107 			IN6_IFADDR_RUNLOCK();
2108 			return (1); /* true */
2109 		}
2110 
2111 		/* XXX: do we still have to go thru the rest of the list? */
2112 	}
2113 	IN6_IFADDR_RUNLOCK();
2114 
2115 	return (0);		/* false */
2116 }
2117 
2118 /*
2119  * return length of part which dst and src are equal
2120  * hard coding...
2121  */
2122 int
2123 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2124 {
2125 	int match = 0;
2126 	u_char *s = (u_char *)src, *d = (u_char *)dst;
2127 	u_char *lim = s + 16, r;
2128 
2129 	while (s < lim)
2130 		if ((r = (*d++ ^ *s++)) != 0) {
2131 			while (r < 128) {
2132 				match++;
2133 				r <<= 1;
2134 			}
2135 			break;
2136 		} else
2137 			match += 8;
2138 	return match;
2139 }
2140 
2141 /* XXX: to be scope conscious */
2142 int
2143 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2144 {
2145 	int bytelen, bitlen;
2146 
2147 	/* sanity check */
2148 	if (0 > len || len > 128) {
2149 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2150 		    len);
2151 		return (0);
2152 	}
2153 
2154 	bytelen = len / 8;
2155 	bitlen = len % 8;
2156 
2157 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2158 		return (0);
2159 	if (bitlen != 0 &&
2160 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2161 	    p2->s6_addr[bytelen] >> (8 - bitlen))
2162 		return (0);
2163 
2164 	return (1);
2165 }
2166 
2167 void
2168 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2169 {
2170 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2171 	int bytelen, bitlen, i;
2172 
2173 	/* sanity check */
2174 	if (0 > len || len > 128) {
2175 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2176 		    len);
2177 		return;
2178 	}
2179 
2180 	bzero(maskp, sizeof(*maskp));
2181 	bytelen = len / 8;
2182 	bitlen = len % 8;
2183 	for (i = 0; i < bytelen; i++)
2184 		maskp->s6_addr[i] = 0xff;
2185 	if (bitlen)
2186 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2187 }
2188 
2189 /*
2190  * return the best address out of the same scope. if no address was
2191  * found, return the first valid address from designated IF.
2192  */
2193 struct in6_ifaddr *
2194 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2195 {
2196 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2197 	struct ifaddr *ifa;
2198 	struct in6_ifaddr *besta = 0;
2199 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2200 
2201 	dep[0] = dep[1] = NULL;
2202 
2203 	/*
2204 	 * We first look for addresses in the same scope.
2205 	 * If there is one, return it.
2206 	 * If two or more, return one which matches the dst longest.
2207 	 * If none, return one of global addresses assigned other ifs.
2208 	 */
2209 	IF_ADDR_RLOCK(ifp);
2210 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2211 		if (ifa->ifa_addr->sa_family != AF_INET6)
2212 			continue;
2213 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2214 			continue; /* XXX: is there any case to allow anycast? */
2215 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2216 			continue; /* don't use this interface */
2217 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2218 			continue;
2219 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2220 			if (V_ip6_use_deprecated)
2221 				dep[0] = (struct in6_ifaddr *)ifa;
2222 			continue;
2223 		}
2224 
2225 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2226 			/*
2227 			 * call in6_matchlen() as few as possible
2228 			 */
2229 			if (besta) {
2230 				if (blen == -1)
2231 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2232 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2233 				if (tlen > blen) {
2234 					blen = tlen;
2235 					besta = (struct in6_ifaddr *)ifa;
2236 				}
2237 			} else
2238 				besta = (struct in6_ifaddr *)ifa;
2239 		}
2240 	}
2241 	if (besta) {
2242 		ifa_ref(&besta->ia_ifa);
2243 		IF_ADDR_RUNLOCK(ifp);
2244 		return (besta);
2245 	}
2246 
2247 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2248 		if (ifa->ifa_addr->sa_family != AF_INET6)
2249 			continue;
2250 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2251 			continue; /* XXX: is there any case to allow anycast? */
2252 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2253 			continue; /* don't use this interface */
2254 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2255 			continue;
2256 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2257 			if (V_ip6_use_deprecated)
2258 				dep[1] = (struct in6_ifaddr *)ifa;
2259 			continue;
2260 		}
2261 
2262 		if (ifa != NULL)
2263 			ifa_ref(ifa);
2264 		IF_ADDR_RUNLOCK(ifp);
2265 		return (struct in6_ifaddr *)ifa;
2266 	}
2267 
2268 	/* use the last-resort values, that are, deprecated addresses */
2269 	if (dep[0]) {
2270 		ifa_ref((struct ifaddr *)dep[0]);
2271 		IF_ADDR_RUNLOCK(ifp);
2272 		return dep[0];
2273 	}
2274 	if (dep[1]) {
2275 		ifa_ref((struct ifaddr *)dep[1]);
2276 		IF_ADDR_RUNLOCK(ifp);
2277 		return dep[1];
2278 	}
2279 
2280 	IF_ADDR_RUNLOCK(ifp);
2281 	return NULL;
2282 }
2283 
2284 /*
2285  * perform DAD when interface becomes IFF_UP.
2286  */
2287 void
2288 in6_if_up(struct ifnet *ifp)
2289 {
2290 	struct ifaddr *ifa;
2291 	struct in6_ifaddr *ia;
2292 
2293 	IF_ADDR_RLOCK(ifp);
2294 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2295 		if (ifa->ifa_addr->sa_family != AF_INET6)
2296 			continue;
2297 		ia = (struct in6_ifaddr *)ifa;
2298 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2299 			/*
2300 			 * The TENTATIVE flag was likely set by hand
2301 			 * beforehand, implicitly indicating the need for DAD.
2302 			 * We may be able to skip the random delay in this
2303 			 * case, but we impose delays just in case.
2304 			 */
2305 			nd6_dad_start(ifa,
2306 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2307 		}
2308 	}
2309 	IF_ADDR_RUNLOCK(ifp);
2310 
2311 	/*
2312 	 * special cases, like 6to4, are handled in in6_ifattach
2313 	 */
2314 	in6_ifattach(ifp, NULL);
2315 }
2316 
2317 int
2318 in6if_do_dad(struct ifnet *ifp)
2319 {
2320 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2321 		return (0);
2322 
2323 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2324 		return (0);
2325 
2326 	switch (ifp->if_type) {
2327 #ifdef IFT_DUMMY
2328 	case IFT_DUMMY:
2329 #endif
2330 	case IFT_FAITH:
2331 		/*
2332 		 * These interfaces do not have the IFF_LOOPBACK flag,
2333 		 * but loop packets back.  We do not have to do DAD on such
2334 		 * interfaces.  We should even omit it, because loop-backed
2335 		 * NS would confuse the DAD procedure.
2336 		 */
2337 		return (0);
2338 	default:
2339 		/*
2340 		 * Our DAD routine requires the interface up and running.
2341 		 * However, some interfaces can be up before the RUNNING
2342 		 * status.  Additionaly, users may try to assign addresses
2343 		 * before the interface becomes up (or running).
2344 		 * We simply skip DAD in such a case as a work around.
2345 		 * XXX: we should rather mark "tentative" on such addresses,
2346 		 * and do DAD after the interface becomes ready.
2347 		 */
2348 		if (!((ifp->if_flags & IFF_UP) &&
2349 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2350 			return (0);
2351 
2352 		return (1);
2353 	}
2354 }
2355 
2356 /*
2357  * Calculate max IPv6 MTU through all the interfaces and store it
2358  * to in6_maxmtu.
2359  */
2360 void
2361 in6_setmaxmtu(void)
2362 {
2363 	unsigned long maxmtu = 0;
2364 	struct ifnet *ifp;
2365 
2366 	IFNET_RLOCK_NOSLEEP();
2367 	TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
2368 		/* this function can be called during ifnet initialization */
2369 		if (!ifp->if_afdata[AF_INET6])
2370 			continue;
2371 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2372 		    IN6_LINKMTU(ifp) > maxmtu)
2373 			maxmtu = IN6_LINKMTU(ifp);
2374 	}
2375 	IFNET_RUNLOCK_NOSLEEP();
2376 	if (maxmtu)	/* update only when maxmtu is positive */
2377 		V_in6_maxmtu = maxmtu;
2378 }
2379 
2380 /*
2381  * Provide the length of interface identifiers to be used for the link attached
2382  * to the given interface.  The length should be defined in "IPv6 over
2383  * xxx-link" document.  Note that address architecture might also define
2384  * the length for a particular set of address prefixes, regardless of the
2385  * link type.  As clarified in rfc2462bis, those two definitions should be
2386  * consistent, and those really are as of August 2004.
2387  */
2388 int
2389 in6_if2idlen(struct ifnet *ifp)
2390 {
2391 	switch (ifp->if_type) {
2392 	case IFT_ETHER:		/* RFC2464 */
2393 #ifdef IFT_PROPVIRTUAL
2394 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2395 #endif
2396 #ifdef IFT_L2VLAN
2397 	case IFT_L2VLAN:	/* ditto */
2398 #endif
2399 #ifdef IFT_IEEE80211
2400 	case IFT_IEEE80211:	/* ditto */
2401 #endif
2402 #ifdef IFT_MIP
2403 	case IFT_MIP:	/* ditto */
2404 #endif
2405 	case IFT_INFINIBAND:
2406 		return (64);
2407 	case IFT_FDDI:		/* RFC2467 */
2408 		return (64);
2409 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2410 		return (64);
2411 	case IFT_PPP:		/* RFC2472 */
2412 		return (64);
2413 	case IFT_ARCNET:	/* RFC2497 */
2414 		return (64);
2415 	case IFT_FRELAY:	/* RFC2590 */
2416 		return (64);
2417 	case IFT_IEEE1394:	/* RFC3146 */
2418 		return (64);
2419 	case IFT_GIF:
2420 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2421 	case IFT_LOOP:
2422 		return (64);	/* XXX: is this really correct? */
2423 	default:
2424 		/*
2425 		 * Unknown link type:
2426 		 * It might be controversial to use the today's common constant
2427 		 * of 64 for these cases unconditionally.  For full compliance,
2428 		 * we should return an error in this case.  On the other hand,
2429 		 * if we simply miss the standard for the link type or a new
2430 		 * standard is defined for a new link type, the IFID length
2431 		 * is very likely to be the common constant.  As a compromise,
2432 		 * we always use the constant, but make an explicit notice
2433 		 * indicating the "unknown" case.
2434 		 */
2435 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2436 		return (64);
2437 	}
2438 }
2439 
2440 #include <sys/sysctl.h>
2441 
2442 struct in6_llentry {
2443 	struct llentry		base;
2444 	struct sockaddr_in6	l3_addr6;
2445 };
2446 
2447 /*
2448  * Deletes an address from the address table.
2449  * This function is called by the timer functions
2450  * such as arptimer() and nd6_llinfo_timer(), and
2451  * the caller does the locking.
2452  */
2453 static void
2454 in6_lltable_free(struct lltable *llt, struct llentry *lle)
2455 {
2456 	LLE_WUNLOCK(lle);
2457 	LLE_LOCK_DESTROY(lle);
2458 	free(lle, M_LLTABLE);
2459 }
2460 
2461 static struct llentry *
2462 in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2463 {
2464 	struct in6_llentry *lle;
2465 
2466 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2467 	if (lle == NULL)		/* NB: caller generates msg */
2468 		return NULL;
2469 
2470 	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2471 	lle->base.lle_refcnt = 1;
2472 	lle->base.lle_free = in6_lltable_free;
2473 	LLE_LOCK_INIT(&lle->base);
2474 	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2475 	    CALLOUT_RETURNUNLOCKED);
2476 
2477 	return (&lle->base);
2478 }
2479 
2480 static void
2481 in6_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix,
2482     const struct sockaddr *mask, u_int flags)
2483 {
2484 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2485 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2486 	struct llentry *lle, *next;
2487 	int i;
2488 
2489 	/*
2490 	 * (flags & LLE_STATIC) means deleting all entries
2491 	 * including static ND6 entries.
2492 	 */
2493 	IF_AFDATA_WLOCK(llt->llt_ifp);
2494 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2495 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2496 			if (IN6_ARE_MASKED_ADDR_EQUAL(
2497 			    &satosin6(L3_ADDR(lle))->sin6_addr,
2498 			    &pfx->sin6_addr, &msk->sin6_addr) &&
2499 			    ((flags & LLE_STATIC) ||
2500 			    !(lle->la_flags & LLE_STATIC))) {
2501 				LLE_WLOCK(lle);
2502 				if (callout_stop(&lle->la_timer))
2503 					LLE_REMREF(lle);
2504 				llentry_free(lle);
2505 			}
2506 		}
2507 	}
2508 	IF_AFDATA_WUNLOCK(llt->llt_ifp);
2509 }
2510 
2511 static int
2512 in6_lltable_rtcheck(struct ifnet *ifp,
2513 		    u_int flags,
2514 		    const struct sockaddr *l3addr)
2515 {
2516 	struct rtentry *rt;
2517 	char ip6buf[INET6_ADDRSTRLEN];
2518 
2519 	KASSERT(l3addr->sa_family == AF_INET6,
2520 	    ("sin_family %d", l3addr->sa_family));
2521 
2522 	/* Our local addresses are always only installed on the default FIB. */
2523 	/* XXX rtalloc1 should take a const param */
2524 	rt = in6_rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0,
2525 	    RT_DEFAULT_FIB);
2526 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2527 		struct ifaddr *ifa;
2528 		/*
2529 		 * Create an ND6 cache for an IPv6 neighbor
2530 		 * that is not covered by our own prefix.
2531 		 */
2532 		/* XXX ifaof_ifpforaddr should take a const param */
2533 		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2534 		if (ifa != NULL) {
2535 			ifa_free(ifa);
2536 			if (rt != NULL)
2537 				RTFREE_LOCKED(rt);
2538 			return 0;
2539 		}
2540 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2541 		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2542 		if (rt != NULL)
2543 			RTFREE_LOCKED(rt);
2544 		return EINVAL;
2545 	}
2546 	RTFREE_LOCKED(rt);
2547 	return 0;
2548 }
2549 
2550 static struct llentry *
2551 in6_lltable_lookup(struct lltable *llt, u_int flags,
2552 	const struct sockaddr *l3addr)
2553 {
2554 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2555 	struct ifnet *ifp = llt->llt_ifp;
2556 	struct llentry *lle;
2557 	struct llentries *lleh;
2558 	u_int hashkey;
2559 
2560 	IF_AFDATA_LOCK_ASSERT(ifp);
2561 	KASSERT(l3addr->sa_family == AF_INET6,
2562 	    ("sin_family %d", l3addr->sa_family));
2563 
2564 	hashkey = sin6->sin6_addr.s6_addr32[3];
2565 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2566 	LIST_FOREACH(lle, lleh, lle_next) {
2567 		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2568 		if (lle->la_flags & LLE_DELETED)
2569 			continue;
2570 		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2571 		    sizeof(struct in6_addr)) == 0)
2572 			break;
2573 	}
2574 
2575 	if (lle == NULL) {
2576 		if (!(flags & LLE_CREATE))
2577 			return (NULL);
2578 		/*
2579 		 * A route that covers the given address must have
2580 		 * been installed 1st because we are doing a resolution,
2581 		 * verify this.
2582 		 */
2583 		if (!(flags & LLE_IFADDR) &&
2584 		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2585 			return NULL;
2586 
2587 		lle = in6_lltable_new(l3addr, flags);
2588 		if (lle == NULL) {
2589 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2590 			return NULL;
2591 		}
2592 		lle->la_flags = flags & ~LLE_CREATE;
2593 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2594 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2595 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2596 		}
2597 
2598 		lle->lle_tbl  = llt;
2599 		lle->lle_head = lleh;
2600 		lle->la_flags |= LLE_LINKED;
2601 		LIST_INSERT_HEAD(lleh, lle, lle_next);
2602 	} else if (flags & LLE_DELETE) {
2603 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2604 			LLE_WLOCK(lle);
2605 			lle->la_flags |= LLE_DELETED;
2606 			LLE_WUNLOCK(lle);
2607 #ifdef DIAGNOSTIC
2608 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2609 #endif
2610 		}
2611 		lle = (void *)-1;
2612 	}
2613 	if (LLE_IS_VALID(lle)) {
2614 		if (flags & LLE_EXCLUSIVE)
2615 			LLE_WLOCK(lle);
2616 		else
2617 			LLE_RLOCK(lle);
2618 	}
2619 	return (lle);
2620 }
2621 
2622 static int
2623 in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2624 {
2625 	struct ifnet *ifp = llt->llt_ifp;
2626 	struct llentry *lle;
2627 	/* XXX stack use */
2628 	struct {
2629 		struct rt_msghdr	rtm;
2630 		struct sockaddr_in6	sin6;
2631 		/*
2632 		 * ndp.c assumes that sdl is word aligned
2633 		 */
2634 #ifdef __LP64__
2635 		uint32_t		pad;
2636 #endif
2637 		struct sockaddr_dl	sdl;
2638 	} ndpc;
2639 	int i, error;
2640 
2641 	if (ifp->if_flags & IFF_LOOPBACK)
2642 		return 0;
2643 
2644 	LLTABLE_LOCK_ASSERT();
2645 
2646 	error = 0;
2647 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2648 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2649 			struct sockaddr_dl *sdl;
2650 
2651 			/* skip deleted or invalid entries */
2652 			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2653 				continue;
2654 			/* Skip if jailed and not a valid IP of the prison. */
2655 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2656 				continue;
2657 			/*
2658 			 * produce a msg made of:
2659 			 *  struct rt_msghdr;
2660 			 *  struct sockaddr_in6 (IPv6)
2661 			 *  struct sockaddr_dl;
2662 			 */
2663 			bzero(&ndpc, sizeof(ndpc));
2664 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2665 			ndpc.rtm.rtm_version = RTM_VERSION;
2666 			ndpc.rtm.rtm_type = RTM_GET;
2667 			ndpc.rtm.rtm_flags = RTF_UP;
2668 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2669 			ndpc.sin6.sin6_family = AF_INET6;
2670 			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2671 			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2672 
2673 			/* publish */
2674 			if (lle->la_flags & LLE_PUB)
2675 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2676 
2677 			sdl = &ndpc.sdl;
2678 			sdl->sdl_family = AF_LINK;
2679 			sdl->sdl_len = sizeof(*sdl);
2680 			sdl->sdl_alen = ifp->if_addrlen;
2681 			sdl->sdl_index = ifp->if_index;
2682 			sdl->sdl_type = ifp->if_type;
2683 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2684 			ndpc.rtm.rtm_rmx.rmx_expire =
2685 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2686 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2687 			if (lle->la_flags & LLE_STATIC)
2688 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2689 			ndpc.rtm.rtm_index = ifp->if_index;
2690 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2691 			if (error)
2692 				break;
2693 		}
2694 	}
2695 	return error;
2696 }
2697 
2698 void *
2699 in6_domifattach(struct ifnet *ifp)
2700 {
2701 	struct in6_ifextra *ext;
2702 
2703 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2704 	bzero(ext, sizeof(*ext));
2705 
2706 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2707 	    M_IFADDR, M_WAITOK);
2708 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2709 
2710 	ext->icmp6_ifstat =
2711 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2712 	    M_IFADDR, M_WAITOK);
2713 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2714 
2715 	ext->nd_ifinfo = nd6_ifattach(ifp);
2716 	ext->scope6_id = scope6_ifattach(ifp);
2717 	ext->lltable = lltable_init(ifp, AF_INET6);
2718 	if (ext->lltable != NULL) {
2719 		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2720 		ext->lltable->llt_lookup = in6_lltable_lookup;
2721 		ext->lltable->llt_dump = in6_lltable_dump;
2722 	}
2723 
2724 	ext->mld_ifinfo = mld_domifattach(ifp);
2725 
2726 	return ext;
2727 }
2728 
2729 void
2730 in6_domifdetach(struct ifnet *ifp, void *aux)
2731 {
2732 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2733 
2734 	mld_domifdetach(ifp);
2735 	scope6_ifdetach(ext->scope6_id);
2736 	nd6_ifdetach(ext->nd_ifinfo);
2737 	lltable_free(ext->lltable);
2738 	free(ext->in6_ifstat, M_IFADDR);
2739 	free(ext->icmp6_ifstat, M_IFADDR);
2740 	free(ext, M_IFADDR);
2741 }
2742 
2743 /*
2744  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2745  * v4 mapped addr or v4 compat addr
2746  */
2747 void
2748 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2749 {
2750 
2751 	bzero(sin, sizeof(*sin));
2752 	sin->sin_len = sizeof(struct sockaddr_in);
2753 	sin->sin_family = AF_INET;
2754 	sin->sin_port = sin6->sin6_port;
2755 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2756 }
2757 
2758 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2759 void
2760 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2761 {
2762 	bzero(sin6, sizeof(*sin6));
2763 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2764 	sin6->sin6_family = AF_INET6;
2765 	sin6->sin6_port = sin->sin_port;
2766 	sin6->sin6_addr.s6_addr32[0] = 0;
2767 	sin6->sin6_addr.s6_addr32[1] = 0;
2768 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2769 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2770 }
2771 
2772 /* Convert sockaddr_in6 into sockaddr_in. */
2773 void
2774 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2775 {
2776 	struct sockaddr_in *sin_p;
2777 	struct sockaddr_in6 sin6;
2778 
2779 	/*
2780 	 * Save original sockaddr_in6 addr and convert it
2781 	 * to sockaddr_in.
2782 	 */
2783 	sin6 = *(struct sockaddr_in6 *)nam;
2784 	sin_p = (struct sockaddr_in *)nam;
2785 	in6_sin6_2_sin(sin_p, &sin6);
2786 }
2787 
2788 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2789 void
2790 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2791 {
2792 	struct sockaddr_in *sin_p;
2793 	struct sockaddr_in6 *sin6_p;
2794 
2795 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2796 	sin_p = (struct sockaddr_in *)*nam;
2797 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2798 	free(*nam, M_SONAME);
2799 	*nam = (struct sockaddr *)sin6_p;
2800 }
2801