xref: /freebsd/sys/netinet6/in6.c (revision 3e65b9c6e6b7b2081d54e1dc40983c3c00eaf738)
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 __P((struct socket *, u_long, caddr_t,
133 	struct ifnet *, struct thread *));
134 static int in6_ifinit __P((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 
158 	/* XXX QL
159 	 * we need to report rt_newaddrmsg
160 	 */
161 	ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
162 	    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
163 	IF_AFDATA_UNLOCK(ifp);
164 	if (ln != NULL) {
165 		ln->la_expire = 0;  /* for IPv6 this means permanent */
166 		ln->ln_state = ND6_LLINFO_REACHABLE;
167 		/*
168 		 * initialize for rtmsg generation
169 		 */
170 		bzero(&gateway, sizeof(gateway));
171 		gateway.sdl_len = sizeof(gateway);
172 		gateway.sdl_family = AF_LINK;
173 		gateway.sdl_nlen = 0;
174 		gateway.sdl_alen = 6;
175 		memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
176 		    sizeof(ln->ll_addr));
177 		LLE_WUNLOCK(ln);
178 	}
179 
180 	bzero(&rt, sizeof(rt));
181 	rt.rt_gateway = (struct sockaddr *)&gateway;
182 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
183 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
184 	rt_mask(&rt) = (struct sockaddr *)&mask;
185 	rt_key(&rt) = (struct sockaddr *)&addr;
186 	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
187 	rt_newaddrmsg(RTM_ADD, ifa, 0, &rt);
188 }
189 
190 void
191 in6_ifremloop(struct ifaddr *ifa)
192 {
193 	struct sockaddr_dl gateway;
194 	struct sockaddr_in6 mask, addr;
195 	struct rtentry rt0;
196 	struct in6_ifaddr *ia;
197 	struct ifnet *ifp;
198 
199 	ia = ifa2ia6(ifa);
200 	ifp = ifa->ifa_ifp;
201 	IF_AFDATA_LOCK(ifp);
202 	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
203 	    (struct sockaddr *)&ia->ia_addr);
204 	IF_AFDATA_UNLOCK(ifp);
205 
206 	/*
207 	 * initialize for rtmsg generation
208 	 */
209 	bzero(&gateway, sizeof(gateway));
210 	gateway.sdl_len = sizeof(gateway);
211 	gateway.sdl_family = AF_LINK;
212 	gateway.sdl_nlen = 0;
213 	gateway.sdl_alen = ifp->if_addrlen;
214 	bzero(&rt0, sizeof(rt0));
215 	rt0.rt_gateway = (struct sockaddr *)&gateway;
216 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
217 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
218 	rt_mask(&rt0) = (struct sockaddr *)&mask;
219 	rt_key(&rt0) = (struct sockaddr *)&addr;
220 	rt0.rt_flags = RTF_HOST | RTF_STATIC;
221 	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
222 }
223 
224 int
225 in6_mask2len(struct in6_addr *mask, u_char *lim0)
226 {
227 	int x = 0, y;
228 	u_char *lim = lim0, *p;
229 
230 	/* ignore the scope_id part */
231 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
232 		lim = (u_char *)mask + sizeof(*mask);
233 	for (p = (u_char *)mask; p < lim; x++, p++) {
234 		if (*p != 0xff)
235 			break;
236 	}
237 	y = 0;
238 	if (p < lim) {
239 		for (y = 0; y < 8; y++) {
240 			if ((*p & (0x80 >> y)) == 0)
241 				break;
242 		}
243 	}
244 
245 	/*
246 	 * when the limit pointer is given, do a stricter check on the
247 	 * remaining bits.
248 	 */
249 	if (p < lim) {
250 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
251 			return (-1);
252 		for (p = p + 1; p < lim; p++)
253 			if (*p != 0)
254 				return (-1);
255 	}
256 
257 	return x * 8 + y;
258 }
259 
260 #ifdef COMPAT_FREEBSD32
261 struct in6_ndifreq32 {
262         char ifname[IFNAMSIZ];
263         uint32_t ifindex;
264 };
265 #define	SIOCGDEFIFACE32_IN6     _IOWR('i', 86, struct in6_ndifreq32)
266 #endif
267 
268 int
269 in6_control(struct socket *so, u_long cmd, caddr_t data,
270     struct ifnet *ifp, struct thread *td)
271 {
272 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
273 	struct	in6_ifaddr *ia = NULL;
274 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
275 	struct sockaddr_in6 *sa6;
276 	int carp_attached = 0;
277 	int error;
278 	u_long ocmd = cmd;
279 
280 	/*
281 	 * Compat to make pre-10.x ifconfig(8) operable.
282 	 */
283 	if (cmd == OSIOCAIFADDR_IN6)
284 		cmd = SIOCAIFADDR_IN6;
285 
286 	switch (cmd) {
287 	case SIOCGETSGCNT_IN6:
288 	case SIOCGETMIFCNT_IN6:
289 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
290 	}
291 
292 	switch(cmd) {
293 	case SIOCAADDRCTL_POLICY:
294 	case SIOCDADDRCTL_POLICY:
295 		if (td != NULL) {
296 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
297 			if (error)
298 				return (error);
299 		}
300 		return (in6_src_ioctl(cmd, data));
301 	}
302 
303 	if (ifp == NULL)
304 		return (EOPNOTSUPP);
305 
306 	switch (cmd) {
307 	case SIOCSNDFLUSH_IN6:
308 	case SIOCSPFXFLUSH_IN6:
309 	case SIOCSRTRFLUSH_IN6:
310 	case SIOCSDEFIFACE_IN6:
311 	case SIOCSIFINFO_FLAGS:
312 	case SIOCSIFINFO_IN6:
313 		if (td != NULL) {
314 			error = priv_check(td, PRIV_NETINET_ND6);
315 			if (error)
316 				return (error);
317 		}
318 		/* FALLTHROUGH */
319 	case OSIOCGIFINFO_IN6:
320 	case SIOCGIFINFO_IN6:
321 	case SIOCGDRLST_IN6:
322 	case SIOCGPRLST_IN6:
323 	case SIOCGNBRINFO_IN6:
324 	case SIOCGDEFIFACE_IN6:
325 		return (nd6_ioctl(cmd, data, ifp));
326 
327 #ifdef COMPAT_FREEBSD32
328 	case SIOCGDEFIFACE32_IN6:
329 		{
330 			struct in6_ndifreq ndif;
331 			struct in6_ndifreq32 *ndif32;
332 
333 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
334 			    ifp);
335 			if (error)
336 				return (error);
337 			ndif32 = (struct in6_ndifreq32 *)data;
338 			ndif32->ifindex = ndif.ifindex;
339 			return (0);
340 		}
341 #endif
342 	}
343 
344 	switch (cmd) {
345 	case SIOCSIFPREFIX_IN6:
346 	case SIOCDIFPREFIX_IN6:
347 	case SIOCAIFPREFIX_IN6:
348 	case SIOCCIFPREFIX_IN6:
349 	case SIOCSGIFPREFIX_IN6:
350 	case SIOCGIFPREFIX_IN6:
351 		log(LOG_NOTICE,
352 		    "prefix ioctls are now invalidated. "
353 		    "please use ifconfig.\n");
354 		return (EOPNOTSUPP);
355 	}
356 
357 	switch (cmd) {
358 	case SIOCSSCOPE6:
359 		if (td != NULL) {
360 			error = priv_check(td, PRIV_NETINET_SCOPE6);
361 			if (error)
362 				return (error);
363 		}
364 		return (scope6_set(ifp,
365 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
366 	case SIOCGSCOPE6:
367 		return (scope6_get(ifp,
368 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
369 	case SIOCGSCOPE6DEF:
370 		return (scope6_get_default((struct scope6_id *)
371 		    ifr->ifr_ifru.ifru_scope_id));
372 	}
373 
374 	switch (cmd) {
375 	case SIOCALIFADDR:
376 		if (td != NULL) {
377 			error = priv_check(td, PRIV_NET_ADDIFADDR);
378 			if (error)
379 				return (error);
380 		}
381 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
382 
383 	case SIOCDLIFADDR:
384 		if (td != NULL) {
385 			error = priv_check(td, PRIV_NET_DELIFADDR);
386 			if (error)
387 				return (error);
388 		}
389 		/* FALLTHROUGH */
390 	case SIOCGLIFADDR:
391 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
392 	}
393 
394 	/*
395 	 * Find address for this interface, if it exists.
396 	 *
397 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
398 	 * only, and used the first interface address as the target of other
399 	 * operations (without checking ifra_addr).  This was because netinet
400 	 * code/API assumed at most 1 interface address per interface.
401 	 * Since IPv6 allows a node to assign multiple addresses
402 	 * on a single interface, we almost always look and check the
403 	 * presence of ifra_addr, and reject invalid ones here.
404 	 * It also decreases duplicated code among SIOC*_IN6 operations.
405 	 */
406 	switch (cmd) {
407 	case SIOCAIFADDR_IN6:
408 	case SIOCSIFPHYADDR_IN6:
409 		sa6 = &ifra->ifra_addr;
410 		break;
411 	case SIOCSIFADDR_IN6:
412 	case SIOCGIFADDR_IN6:
413 	case SIOCSIFDSTADDR_IN6:
414 	case SIOCSIFNETMASK_IN6:
415 	case SIOCGIFDSTADDR_IN6:
416 	case SIOCGIFNETMASK_IN6:
417 	case SIOCDIFADDR_IN6:
418 	case SIOCGIFPSRCADDR_IN6:
419 	case SIOCGIFPDSTADDR_IN6:
420 	case SIOCGIFAFLAG_IN6:
421 	case SIOCSNDFLUSH_IN6:
422 	case SIOCSPFXFLUSH_IN6:
423 	case SIOCSRTRFLUSH_IN6:
424 	case SIOCGIFALIFETIME_IN6:
425 	case SIOCSIFALIFETIME_IN6:
426 	case SIOCGIFSTAT_IN6:
427 	case SIOCGIFSTAT_ICMP6:
428 		sa6 = &ifr->ifr_addr;
429 		break;
430 	default:
431 		sa6 = NULL;
432 		break;
433 	}
434 	if (sa6 && sa6->sin6_family == AF_INET6) {
435 		if (sa6->sin6_scope_id != 0)
436 			error = sa6_embedscope(sa6, 0);
437 		else
438 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
439 		if (error != 0)
440 			return (error);
441 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
442 		    &sa6->sin6_addr)) != 0)
443 			return (error);
444 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
445 	} else
446 		ia = NULL;
447 
448 	switch (cmd) {
449 	case SIOCSIFADDR_IN6:
450 	case SIOCSIFDSTADDR_IN6:
451 	case SIOCSIFNETMASK_IN6:
452 		/*
453 		 * Since IPv6 allows a node to assign multiple addresses
454 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
455 		 */
456 		/* we decided to obsolete this command (20000704) */
457 		error = EINVAL;
458 		goto out;
459 
460 	case SIOCDIFADDR_IN6:
461 		/*
462 		 * for IPv4, we look for existing in_ifaddr here to allow
463 		 * "ifconfig if0 delete" to remove the first IPv4 address on
464 		 * the interface.  For IPv6, as the spec allows multiple
465 		 * interface address from the day one, we consider "remove the
466 		 * first one" semantics to be not preferable.
467 		 */
468 		if (ia == NULL) {
469 			error = EADDRNOTAVAIL;
470 			goto out;
471 		}
472 		/* FALLTHROUGH */
473 	case SIOCAIFADDR_IN6:
474 		/*
475 		 * We always require users to specify a valid IPv6 address for
476 		 * the corresponding operation.
477 		 */
478 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
479 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
480 			error = EAFNOSUPPORT;
481 			goto out;
482 		}
483 
484 		if (td != NULL) {
485 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
486 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
487 			if (error)
488 				goto out;
489 		}
490 		break;
491 
492 	case SIOCGIFADDR_IN6:
493 		/* This interface is basically deprecated. use SIOCGIFCONF. */
494 		/* FALLTHROUGH */
495 	case SIOCGIFAFLAG_IN6:
496 	case SIOCGIFNETMASK_IN6:
497 	case SIOCGIFDSTADDR_IN6:
498 	case SIOCGIFALIFETIME_IN6:
499 		/* must think again about its semantics */
500 		if (ia == NULL) {
501 			error = EADDRNOTAVAIL;
502 			goto out;
503 		}
504 		break;
505 
506 	case SIOCSIFALIFETIME_IN6:
507 	    {
508 		struct in6_addrlifetime *lt;
509 
510 		if (td != NULL) {
511 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
512 			if (error)
513 				goto out;
514 		}
515 		if (ia == NULL) {
516 			error = EADDRNOTAVAIL;
517 			goto out;
518 		}
519 		/* sanity for overflow - beware unsigned */
520 		lt = &ifr->ifr_ifru.ifru_lifetime;
521 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
522 		    lt->ia6t_vltime + time_second < time_second) {
523 			error = EINVAL;
524 			goto out;
525 		}
526 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
527 		    lt->ia6t_pltime + time_second < time_second) {
528 			error = EINVAL;
529 			goto out;
530 		}
531 		break;
532 	    }
533 	}
534 
535 	switch (cmd) {
536 	case SIOCGIFADDR_IN6:
537 		ifr->ifr_addr = ia->ia_addr;
538 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
539 			goto out;
540 		break;
541 
542 	case SIOCGIFDSTADDR_IN6:
543 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
544 			error = EINVAL;
545 			goto out;
546 		}
547 		/*
548 		 * XXX: should we check if ifa_dstaddr is NULL and return
549 		 * an error?
550 		 */
551 		ifr->ifr_dstaddr = ia->ia_dstaddr;
552 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
553 			goto out;
554 		break;
555 
556 	case SIOCGIFNETMASK_IN6:
557 		ifr->ifr_addr = ia->ia_prefixmask;
558 		break;
559 
560 	case SIOCGIFAFLAG_IN6:
561 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
562 		break;
563 
564 	case SIOCGIFSTAT_IN6:
565 		if (ifp == NULL) {
566 			error = EINVAL;
567 			goto out;
568 		}
569 		bzero(&ifr->ifr_ifru.ifru_stat,
570 		    sizeof(ifr->ifr_ifru.ifru_stat));
571 		ifr->ifr_ifru.ifru_stat =
572 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
573 		break;
574 
575 	case SIOCGIFSTAT_ICMP6:
576 		if (ifp == NULL) {
577 			error = EINVAL;
578 			goto out;
579 		}
580 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
581 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
582 		ifr->ifr_ifru.ifru_icmp6stat =
583 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
584 		break;
585 
586 	case SIOCGIFALIFETIME_IN6:
587 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
588 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
589 			time_t maxexpire;
590 			struct in6_addrlifetime *retlt =
591 			    &ifr->ifr_ifru.ifru_lifetime;
592 
593 			/*
594 			 * XXX: adjust expiration time assuming time_t is
595 			 * signed.
596 			 */
597 			maxexpire = (-1) &
598 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
599 			if (ia->ia6_lifetime.ia6t_vltime <
600 			    maxexpire - ia->ia6_updatetime) {
601 				retlt->ia6t_expire = ia->ia6_updatetime +
602 				    ia->ia6_lifetime.ia6t_vltime;
603 			} else
604 				retlt->ia6t_expire = maxexpire;
605 		}
606 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
607 			time_t maxexpire;
608 			struct in6_addrlifetime *retlt =
609 			    &ifr->ifr_ifru.ifru_lifetime;
610 
611 			/*
612 			 * XXX: adjust expiration time assuming time_t is
613 			 * signed.
614 			 */
615 			maxexpire = (-1) &
616 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
617 			if (ia->ia6_lifetime.ia6t_pltime <
618 			    maxexpire - ia->ia6_updatetime) {
619 				retlt->ia6t_preferred = ia->ia6_updatetime +
620 				    ia->ia6_lifetime.ia6t_pltime;
621 			} else
622 				retlt->ia6t_preferred = maxexpire;
623 		}
624 		break;
625 
626 	case SIOCSIFALIFETIME_IN6:
627 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
628 		/* for sanity */
629 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
630 			ia->ia6_lifetime.ia6t_expire =
631 				time_second + ia->ia6_lifetime.ia6t_vltime;
632 		} else
633 			ia->ia6_lifetime.ia6t_expire = 0;
634 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
635 			ia->ia6_lifetime.ia6t_preferred =
636 				time_second + ia->ia6_lifetime.ia6t_pltime;
637 		} else
638 			ia->ia6_lifetime.ia6t_preferred = 0;
639 		break;
640 
641 	case SIOCAIFADDR_IN6:
642 	{
643 		int i;
644 		struct nd_prefixctl pr0;
645 		struct nd_prefix *pr;
646 
647 		/*
648 		 * first, make or update the interface address structure,
649 		 * and link it to the list.
650 		 */
651 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
652 			goto out;
653 		if (ia != NULL)
654 			ifa_free(&ia->ia_ifa);
655 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
656 		    == NULL) {
657 			/*
658 			 * this can happen when the user specify the 0 valid
659 			 * lifetime.
660 			 */
661 			break;
662 		}
663 
664 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
665 			if (carp_attach_p != NULL)
666 				error = (*carp_attach_p)(&ia->ia_ifa,
667 				    ifra->ifra_vhid);
668 			else
669 				error = EPROTONOSUPPORT;
670 			if (error)
671 				goto out;
672 			else
673 				carp_attached = 1;
674 		}
675 
676 		/*
677 		 * then, make the prefix on-link on the interface.
678 		 * XXX: we'd rather create the prefix before the address, but
679 		 * we need at least one address to install the corresponding
680 		 * interface route, so we configure the address first.
681 		 */
682 
683 		/*
684 		 * convert mask to prefix length (prefixmask has already
685 		 * been validated in in6_update_ifa().
686 		 */
687 		bzero(&pr0, sizeof(pr0));
688 		pr0.ndpr_ifp = ifp;
689 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
690 		    NULL);
691 		if (pr0.ndpr_plen == 128) {
692 			break;	/* we don't need to install a host route. */
693 		}
694 		pr0.ndpr_prefix = ifra->ifra_addr;
695 		/* apply the mask for safety. */
696 		for (i = 0; i < 4; i++) {
697 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
698 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
699 		}
700 		/*
701 		 * XXX: since we don't have an API to set prefix (not address)
702 		 * lifetimes, we just use the same lifetimes as addresses.
703 		 * The (temporarily) installed lifetimes can be overridden by
704 		 * later advertised RAs (when accept_rtadv is non 0), which is
705 		 * an intended behavior.
706 		 */
707 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
708 		pr0.ndpr_raf_auto =
709 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
710 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
711 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
712 
713 		/* add the prefix if not yet. */
714 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
715 			/*
716 			 * nd6_prelist_add will install the corresponding
717 			 * interface route.
718 			 */
719 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
720 				if (carp_attached)
721 					(*carp_detach_p)(&ia->ia_ifa);
722 				goto out;
723 			}
724 			if (pr == NULL) {
725 				if (carp_attached)
726 					(*carp_detach_p)(&ia->ia_ifa);
727 				log(LOG_ERR, "nd6_prelist_add succeeded but "
728 				    "no prefix\n");
729 				error = EINVAL;
730 				goto out;
731 			}
732 		}
733 
734 		/* relate the address to the prefix */
735 		if (ia->ia6_ndpr == NULL) {
736 			ia->ia6_ndpr = pr;
737 			pr->ndpr_refcnt++;
738 
739 			/*
740 			 * If this is the first autoconf address from the
741 			 * prefix, create a temporary address as well
742 			 * (when required).
743 			 */
744 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
745 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
746 				int e;
747 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
748 					log(LOG_NOTICE, "in6_control: failed "
749 					    "to create a temporary address, "
750 					    "errno=%d\n", e);
751 				}
752 			}
753 		}
754 
755 		/*
756 		 * this might affect the status of autoconfigured addresses,
757 		 * that is, this address might make other addresses detached.
758 		 */
759 		pfxlist_onlink_check();
760 		if (error == 0 && ia) {
761 			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
762 				/*
763 				 * Try to clear the flag when a new
764 				 * IPv6 address is added onto an
765 				 * IFDISABLED interface and it
766 				 * succeeds.
767 				 */
768 				struct in6_ndireq nd;
769 
770 				memset(&nd, 0, sizeof(nd));
771 				nd.ndi.flags = ND_IFINFO(ifp)->flags;
772 				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
773 				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
774 				    (caddr_t)&nd, ifp) < 0)
775 					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
776 					    "SIOCSIFINFO_FLAGS for -ifdisabled "
777 					    "failed.");
778 				/*
779 				 * Ignore failure of clearing the flag
780 				 * intentionally.  The failure means
781 				 * address duplication was detected.
782 				 */
783 			}
784 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
785 		}
786 		break;
787 	}
788 
789 	case SIOCDIFADDR_IN6:
790 	{
791 		struct nd_prefix *pr;
792 
793 		/*
794 		 * If the address being deleted is the only one that owns
795 		 * the corresponding prefix, expire the prefix as well.
796 		 * XXX: theoretically, we don't have to worry about such
797 		 * relationship, since we separate the address management
798 		 * and the prefix management.  We do this, however, to provide
799 		 * as much backward compatibility as possible in terms of
800 		 * the ioctl operation.
801 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
802 		 */
803 		pr = ia->ia6_ndpr;
804 		in6_purgeaddr(&ia->ia_ifa);
805 		if (pr && pr->ndpr_refcnt == 0)
806 			prelist_remove(pr);
807 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
808 		break;
809 	}
810 
811 	default:
812 		if (ifp == NULL || ifp->if_ioctl == 0) {
813 			error = EOPNOTSUPP;
814 			goto out;
815 		}
816 		error = (*ifp->if_ioctl)(ifp, cmd, data);
817 		goto out;
818 	}
819 
820 	error = 0;
821 out:
822 	if (ia != NULL)
823 		ifa_free(&ia->ia_ifa);
824 	return (error);
825 }
826 
827 /*
828  * Update parameters of an IPv6 interface address.
829  * If necessary, a new entry is created and linked into address chains.
830  * This function is separated from in6_control().
831  * XXX: should this be performed under splnet()?
832  */
833 int
834 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
835     struct in6_ifaddr *ia, int flags)
836 {
837 	int error = 0, hostIsNew = 0, plen = -1;
838 	struct sockaddr_in6 dst6;
839 	struct in6_addrlifetime *lt;
840 	struct in6_multi_mship *imm;
841 	struct in6_multi *in6m_sol;
842 	struct rtentry *rt;
843 	int delay;
844 	char ip6buf[INET6_ADDRSTRLEN];
845 
846 	/* Validate parameters */
847 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
848 		return (EINVAL);
849 
850 	/*
851 	 * The destination address for a p2p link must have a family
852 	 * of AF_UNSPEC or AF_INET6.
853 	 */
854 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
855 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
856 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
857 		return (EAFNOSUPPORT);
858 	/*
859 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
860 	 * does not carry fields other than sin6_len.
861 	 */
862 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
863 		return (EINVAL);
864 	/*
865 	 * Because the IPv6 address architecture is classless, we require
866 	 * users to specify a (non 0) prefix length (mask) for a new address.
867 	 * We also require the prefix (when specified) mask is valid, and thus
868 	 * reject a non-consecutive mask.
869 	 */
870 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
871 		return (EINVAL);
872 	if (ifra->ifra_prefixmask.sin6_len != 0) {
873 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
874 		    (u_char *)&ifra->ifra_prefixmask +
875 		    ifra->ifra_prefixmask.sin6_len);
876 		if (plen <= 0)
877 			return (EINVAL);
878 	} else {
879 		/*
880 		 * In this case, ia must not be NULL.  We just use its prefix
881 		 * length.
882 		 */
883 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
884 	}
885 	/*
886 	 * If the destination address on a p2p interface is specified,
887 	 * and the address is a scoped one, validate/set the scope
888 	 * zone identifier.
889 	 */
890 	dst6 = ifra->ifra_dstaddr;
891 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
892 	    (dst6.sin6_family == AF_INET6)) {
893 		struct in6_addr in6_tmp;
894 		u_int32_t zoneid;
895 
896 		in6_tmp = dst6.sin6_addr;
897 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
898 			return (EINVAL); /* XXX: should be impossible */
899 
900 		if (dst6.sin6_scope_id != 0) {
901 			if (dst6.sin6_scope_id != zoneid)
902 				return (EINVAL);
903 		} else		/* user omit to specify the ID. */
904 			dst6.sin6_scope_id = zoneid;
905 
906 		/* convert into the internal form */
907 		if (sa6_embedscope(&dst6, 0))
908 			return (EINVAL); /* XXX: should be impossible */
909 	}
910 	/*
911 	 * The destination address can be specified only for a p2p or a
912 	 * loopback interface.  If specified, the corresponding prefix length
913 	 * must be 128.
914 	 */
915 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
916 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
917 			/* XXX: noisy message */
918 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
919 			    "be specified for a p2p or a loopback IF only\n"));
920 			return (EINVAL);
921 		}
922 		if (plen != 128) {
923 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
924 			    "be 128 when dstaddr is specified\n"));
925 			return (EINVAL);
926 		}
927 	}
928 	/* lifetime consistency check */
929 	lt = &ifra->ifra_lifetime;
930 	if (lt->ia6t_pltime > lt->ia6t_vltime)
931 		return (EINVAL);
932 	if (lt->ia6t_vltime == 0) {
933 		/*
934 		 * the following log might be noisy, but this is a typical
935 		 * configuration mistake or a tool's bug.
936 		 */
937 		nd6log((LOG_INFO,
938 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
939 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
940 
941 		if (ia == NULL)
942 			return (0); /* there's nothing to do */
943 	}
944 
945 	/*
946 	 * If this is a new address, allocate a new ifaddr and link it
947 	 * into chains.
948 	 */
949 	if (ia == NULL) {
950 		hostIsNew = 1;
951 		/*
952 		 * When in6_update_ifa() is called in a process of a received
953 		 * RA, it is called under an interrupt context.  So, we should
954 		 * call malloc with M_NOWAIT.
955 		 */
956 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
957 		    M_NOWAIT);
958 		if (ia == NULL)
959 			return (ENOBUFS);
960 		bzero((caddr_t)ia, sizeof(*ia));
961 		ifa_init(&ia->ia_ifa);
962 		LIST_INIT(&ia->ia6_memberships);
963 		/* Initialize the address and masks, and put time stamp */
964 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
965 		ia->ia_addr.sin6_family = AF_INET6;
966 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
967 		ia->ia6_createtime = time_second;
968 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
969 			/*
970 			 * XXX: some functions expect that ifa_dstaddr is not
971 			 * NULL for p2p interfaces.
972 			 */
973 			ia->ia_ifa.ifa_dstaddr =
974 			    (struct sockaddr *)&ia->ia_dstaddr;
975 		} else {
976 			ia->ia_ifa.ifa_dstaddr = NULL;
977 		}
978 		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
979 		ia->ia_ifp = ifp;
980 		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
981 		IF_ADDR_LOCK(ifp);
982 		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
983 		IF_ADDR_UNLOCK(ifp);
984 
985 		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
986 		IN6_IFADDR_WLOCK();
987 		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
988 		IN6_IFADDR_WUNLOCK();
989 	}
990 
991 	/* update timestamp */
992 	ia->ia6_updatetime = time_second;
993 
994 	/* set prefix mask */
995 	if (ifra->ifra_prefixmask.sin6_len) {
996 		/*
997 		 * We prohibit changing the prefix length of an existing
998 		 * address, because
999 		 * + such an operation should be rare in IPv6, and
1000 		 * + the operation would confuse prefix management.
1001 		 */
1002 		if (ia->ia_prefixmask.sin6_len &&
1003 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1004 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1005 			    " existing (%s) address should not be changed\n",
1006 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1007 			error = EINVAL;
1008 			goto unlink;
1009 		}
1010 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1011 	}
1012 
1013 	/*
1014 	 * If a new destination address is specified, scrub the old one and
1015 	 * install the new destination.  Note that the interface must be
1016 	 * p2p or loopback (see the check above.)
1017 	 */
1018 	if (dst6.sin6_family == AF_INET6 &&
1019 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1020 		int e;
1021 
1022 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1023 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1024 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1025 			    "a route to the old destination: %s\n",
1026 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1027 			/* proceed anyway... */
1028 		} else
1029 			ia->ia_flags &= ~IFA_ROUTE;
1030 		ia->ia_dstaddr = dst6;
1031 	}
1032 
1033 	/*
1034 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1035 	 * to see if the address is deprecated or invalidated, but initialize
1036 	 * these members for applications.
1037 	 */
1038 	ia->ia6_lifetime = ifra->ifra_lifetime;
1039 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1040 		ia->ia6_lifetime.ia6t_expire =
1041 		    time_second + ia->ia6_lifetime.ia6t_vltime;
1042 	} else
1043 		ia->ia6_lifetime.ia6t_expire = 0;
1044 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1045 		ia->ia6_lifetime.ia6t_preferred =
1046 		    time_second + ia->ia6_lifetime.ia6t_pltime;
1047 	} else
1048 		ia->ia6_lifetime.ia6t_preferred = 0;
1049 
1050 	/* reset the interface and routing table appropriately. */
1051 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1052 		goto unlink;
1053 
1054 	/*
1055 	 * configure address flags.
1056 	 */
1057 	ia->ia6_flags = ifra->ifra_flags;
1058 	/*
1059 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1060 	 * userland, make it deprecated.
1061 	 */
1062 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1063 		ia->ia6_lifetime.ia6t_pltime = 0;
1064 		ia->ia6_lifetime.ia6t_preferred = time_second;
1065 	}
1066 	/*
1067 	 * Make the address tentative before joining multicast addresses,
1068 	 * so that corresponding MLD responses would not have a tentative
1069 	 * source address.
1070 	 */
1071 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1072 	if (hostIsNew && in6if_do_dad(ifp))
1073 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1074 
1075 	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
1076 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1077 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1078 
1079 	/*
1080 	 * We are done if we have simply modified an existing address.
1081 	 */
1082 	if (!hostIsNew)
1083 		return (error);
1084 
1085 	/*
1086 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1087 	 * not just go to unlink.
1088 	 */
1089 
1090 	/* Join necessary multicast groups */
1091 	in6m_sol = NULL;
1092 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1093 		struct sockaddr_in6 mltaddr, mltmask;
1094 		struct in6_addr llsol;
1095 
1096 		/* join solicited multicast addr for new host id */
1097 		bzero(&llsol, sizeof(struct in6_addr));
1098 		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1099 		llsol.s6_addr32[1] = 0;
1100 		llsol.s6_addr32[2] = htonl(1);
1101 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1102 		llsol.s6_addr8[12] = 0xff;
1103 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1104 			/* XXX: should not happen */
1105 			log(LOG_ERR, "in6_update_ifa: "
1106 			    "in6_setscope failed\n");
1107 			goto cleanup;
1108 		}
1109 		delay = 0;
1110 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1111 			/*
1112 			 * We need a random delay for DAD on the address
1113 			 * being configured.  It also means delaying
1114 			 * transmission of the corresponding MLD report to
1115 			 * avoid report collision.
1116 			 * [RFC 4861, Section 6.3.7]
1117 			 */
1118 			delay = arc4random() %
1119 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1120 		}
1121 		imm = in6_joingroup(ifp, &llsol, &error, delay);
1122 		if (imm == NULL) {
1123 			nd6log((LOG_WARNING,
1124 			    "in6_update_ifa: addmulti failed for "
1125 			    "%s on %s (errno=%d)\n",
1126 			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
1127 			    error));
1128 			goto cleanup;
1129 		}
1130 		LIST_INSERT_HEAD(&ia->ia6_memberships,
1131 		    imm, i6mm_chain);
1132 		in6m_sol = imm->i6mm_maddr;
1133 
1134 		bzero(&mltmask, sizeof(mltmask));
1135 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1136 		mltmask.sin6_family = AF_INET6;
1137 		mltmask.sin6_addr = in6mask32;
1138 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1139 
1140 		/*
1141 		 * join link-local all-nodes address
1142 		 */
1143 		bzero(&mltaddr, sizeof(mltaddr));
1144 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1145 		mltaddr.sin6_family = AF_INET6;
1146 		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1147 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1148 		    0)
1149 			goto cleanup; /* XXX: should not fail */
1150 
1151 		/*
1152 		 * XXX: do we really need this automatic routes?
1153 		 * We should probably reconsider this stuff.  Most applications
1154 		 * actually do not need the routes, since they usually specify
1155 		 * the outgoing interface.
1156 		 */
1157 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1158 		if (rt) {
1159 			/* XXX: only works in !SCOPEDROUTING case. */
1160 			if (memcmp(&mltaddr.sin6_addr,
1161 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1162 			    MLTMASK_LEN)) {
1163 				RTFREE_LOCKED(rt);
1164 				rt = NULL;
1165 			}
1166 		}
1167 		if (!rt) {
1168 			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1169 			    (struct sockaddr *)&ia->ia_addr,
1170 			    (struct sockaddr *)&mltmask, RTF_UP,
1171 			    (struct rtentry **)0);
1172 			if (error)
1173 				goto cleanup;
1174 		} else {
1175 			RTFREE_LOCKED(rt);
1176 		}
1177 
1178 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1179 		if (!imm) {
1180 			nd6log((LOG_WARNING,
1181 			    "in6_update_ifa: addmulti failed for "
1182 			    "%s on %s (errno=%d)\n",
1183 			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1184 			    if_name(ifp), error));
1185 			goto cleanup;
1186 		}
1187 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1188 
1189 		/*
1190 		 * join node information group address
1191 		 */
1192 		delay = 0;
1193 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1194 			/*
1195 			 * The spec doesn't say anything about delay for this
1196 			 * group, but the same logic should apply.
1197 			 */
1198 			delay = arc4random() %
1199 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1200 		}
1201 		if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
1202 			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1203 			    delay); /* XXX jinmei */
1204 			if (!imm) {
1205 				nd6log((LOG_WARNING, "in6_update_ifa: "
1206 				    "addmulti failed for %s on %s "
1207 				    "(errno=%d)\n",
1208 				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1209 				    if_name(ifp), error));
1210 				/* XXX not very fatal, go on... */
1211 			} else {
1212 				LIST_INSERT_HEAD(&ia->ia6_memberships,
1213 				    imm, i6mm_chain);
1214 			}
1215 		}
1216 
1217 		/*
1218 		 * join interface-local all-nodes address.
1219 		 * (ff01::1%ifN, and ff01::%ifN/32)
1220 		 */
1221 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1222 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1223 		    != 0)
1224 			goto cleanup; /* XXX: should not fail */
1225 		/* XXX: again, do we really need the route? */
1226 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1227 		if (rt) {
1228 			if (memcmp(&mltaddr.sin6_addr,
1229 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1230 			    MLTMASK_LEN)) {
1231 				RTFREE_LOCKED(rt);
1232 				rt = NULL;
1233 			}
1234 		}
1235 		if (!rt) {
1236 			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1237 			    (struct sockaddr *)&ia->ia_addr,
1238 			    (struct sockaddr *)&mltmask, RTF_UP,
1239 			    (struct rtentry **)0);
1240 			if (error)
1241 				goto cleanup;
1242 		} else
1243 			RTFREE_LOCKED(rt);
1244 
1245 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1246 		if (!imm) {
1247 			nd6log((LOG_WARNING, "in6_update_ifa: "
1248 			    "addmulti failed for %s on %s "
1249 			    "(errno=%d)\n",
1250 			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1251 			    if_name(ifp), error));
1252 			goto cleanup;
1253 		}
1254 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1255 #undef	MLTMASK_LEN
1256 	}
1257 
1258 	/*
1259 	 * Perform DAD, if needed.
1260 	 * XXX It may be of use, if we can administratively
1261 	 * disable DAD.
1262 	 */
1263 	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1264 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1265 	{
1266 		int mindelay, maxdelay;
1267 
1268 		delay = 0;
1269 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1270 			/*
1271 			 * We need to impose a delay before sending an NS
1272 			 * for DAD.  Check if we also needed a delay for the
1273 			 * corresponding MLD message.  If we did, the delay
1274 			 * should be larger than the MLD delay (this could be
1275 			 * relaxed a bit, but this simple logic is at least
1276 			 * safe).
1277 			 * XXX: Break data hiding guidelines and look at
1278 			 * state for the solicited multicast group.
1279 			 */
1280 			mindelay = 0;
1281 			if (in6m_sol != NULL &&
1282 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1283 				mindelay = in6m_sol->in6m_timer;
1284 			}
1285 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1286 			if (maxdelay - mindelay == 0)
1287 				delay = 0;
1288 			else {
1289 				delay =
1290 				    (arc4random() % (maxdelay - mindelay)) +
1291 				    mindelay;
1292 			}
1293 		}
1294 		nd6_dad_start((struct ifaddr *)ia, delay);
1295 	}
1296 
1297 	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1298 	ifa_free(&ia->ia_ifa);
1299 	return (error);
1300 
1301   unlink:
1302 	/*
1303 	 * XXX: if a change of an existing address failed, keep the entry
1304 	 * anyway.
1305 	 */
1306 	if (hostIsNew) {
1307 		in6_unlink_ifa(ia, ifp);
1308 		ifa_free(&ia->ia_ifa);
1309 	}
1310 	return (error);
1311 
1312   cleanup:
1313 	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1314 	ifa_free(&ia->ia_ifa);
1315 	in6_purgeaddr(&ia->ia_ifa);
1316 	return error;
1317 }
1318 
1319 void
1320 in6_purgeaddr(struct ifaddr *ifa)
1321 {
1322 	struct ifnet *ifp = ifa->ifa_ifp;
1323 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1324 	struct in6_multi_mship *imm;
1325 	struct sockaddr_in6 mltaddr, mltmask;
1326 	int plen, error;
1327 	struct rtentry *rt;
1328 	struct ifaddr *ifa0, *nifa;
1329 
1330 	if (ifa->ifa_carp)
1331 		(*carp_detach_p)(ifa);
1332 
1333 	/*
1334 	 * find another IPv6 address as the gateway for the
1335 	 * link-local and node-local all-nodes multicast
1336 	 * address routes
1337 	 */
1338 	IF_ADDR_LOCK(ifp);
1339 	TAILQ_FOREACH_SAFE(ifa0, &ifp->if_addrhead, ifa_link, nifa) {
1340 		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1341 		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1342 			   &ia->ia_addr.sin6_addr,
1343 			   sizeof(struct in6_addr)) == 0)
1344 			continue;
1345 		else
1346 			break;
1347 	}
1348 	if (ifa0 != NULL)
1349 		ifa_ref(ifa0);
1350 	IF_ADDR_UNLOCK(ifp);
1351 
1352 	/*
1353 	 * Remove the loopback route to the interface address.
1354 	 * The check for the current setting of "nd6_useloopback"
1355 	 * is not needed.
1356 	 */
1357 	if (ia->ia_flags & IFA_RTSELF) {
1358 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1359 				       (struct sockaddr *)&ia->ia_addr);
1360 		if (error == 0)
1361 			ia->ia_flags &= ~IFA_RTSELF;
1362 	}
1363 
1364 	/* stop DAD processing */
1365 	nd6_dad_stop(ifa);
1366 
1367 	in6_ifremloop(ifa);
1368 
1369 	/*
1370 	 * leave from multicast groups we have joined for the interface
1371 	 */
1372 	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1373 		LIST_REMOVE(imm, i6mm_chain);
1374 		in6_leavegroup(imm);
1375 	}
1376 
1377 	/*
1378 	 * remove the link-local all-nodes address
1379 	 */
1380 	bzero(&mltmask, sizeof(mltmask));
1381 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1382 	mltmask.sin6_family = AF_INET6;
1383 	mltmask.sin6_addr = in6mask32;
1384 
1385 	bzero(&mltaddr, sizeof(mltaddr));
1386 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1387 	mltaddr.sin6_family = AF_INET6;
1388 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1389 
1390 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1391 	    0)
1392 		goto cleanup;
1393 
1394 	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1395 	if (rt != NULL && rt->rt_gateway != NULL &&
1396 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1397 		    &ia->ia_addr.sin6_addr,
1398 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1399 		/*
1400 		 * if no more IPv6 address exists on this interface
1401 		 * then remove the multicast address route
1402 		 */
1403 		if (ifa0 == NULL) {
1404 			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1405 			       sizeof(mltaddr.sin6_addr));
1406 			RTFREE_LOCKED(rt);
1407 			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1408 					  (struct sockaddr *)&ia->ia_addr,
1409 					  (struct sockaddr *)&mltmask, RTF_UP,
1410 					  (struct rtentry **)0);
1411 			if (error)
1412 				log(LOG_INFO, "in6_purgeaddr: link-local all-nodes"
1413 				    "multicast address deletion error\n");
1414 		} else {
1415 			/*
1416 			 * replace the gateway of the route
1417 			 */
1418 			struct sockaddr_in6 sa;
1419 
1420 			bzero(&sa, sizeof(sa));
1421 			sa.sin6_len = sizeof(struct sockaddr_in6);
1422 			sa.sin6_family = AF_INET6;
1423 			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1424 			       sizeof(sa.sin6_addr));
1425 			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1426 			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1427 			RTFREE_LOCKED(rt);
1428 		}
1429 	} else {
1430 		if (rt != NULL)
1431 			RTFREE_LOCKED(rt);
1432 	}
1433 
1434 	/*
1435 	 * remove the node-local all-nodes address
1436 	 */
1437 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1438 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1439 	    0)
1440 		goto cleanup;
1441 
1442 	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1443 	if (rt != NULL && rt->rt_gateway != NULL &&
1444 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1445 		    &ia->ia_addr.sin6_addr,
1446 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1447 		/*
1448 		 * if no more IPv6 address exists on this interface
1449 		 * then remove the multicast address route
1450 		 */
1451 		if (ifa0 == NULL) {
1452 			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1453 			       sizeof(mltaddr.sin6_addr));
1454 
1455 			RTFREE_LOCKED(rt);
1456 			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1457 					  (struct sockaddr *)&ia->ia_addr,
1458 					  (struct sockaddr *)&mltmask, RTF_UP,
1459 					  (struct rtentry **)0);
1460 
1461 			if (error)
1462 				log(LOG_INFO, "in6_purgeaddr: node-local all-nodes"
1463 				    "multicast address deletion error\n");
1464 		} else {
1465 			/*
1466 			 * replace the gateway of the route
1467 			 */
1468 			struct sockaddr_in6 sa;
1469 
1470 			bzero(&sa, sizeof(sa));
1471 			sa.sin6_len = sizeof(struct sockaddr_in6);
1472 			sa.sin6_family = AF_INET6;
1473 			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1474 			       sizeof(sa.sin6_addr));
1475 			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1476 			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1477 			RTFREE_LOCKED(rt);
1478 		}
1479 	} else {
1480 		if (rt != NULL)
1481 			RTFREE_LOCKED(rt);
1482 	}
1483 
1484 cleanup:
1485 
1486 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1487 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1488 		int error;
1489 		struct sockaddr *dstaddr;
1490 
1491 		/*
1492 		 * use the interface address if configuring an
1493 		 * interface address with a /128 prefix len
1494 		 */
1495 		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1496 			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1497 		else
1498 			dstaddr = (struct sockaddr *)&ia->ia_addr;
1499 
1500 		error = rtrequest(RTM_DELETE,
1501 		    (struct sockaddr *)dstaddr,
1502 		    (struct sockaddr *)&ia->ia_addr,
1503 		    (struct sockaddr *)&ia->ia_prefixmask,
1504 		    ia->ia_flags | RTF_HOST, NULL);
1505 		if (error != 0)
1506 			return;
1507 		ia->ia_flags &= ~IFA_ROUTE;
1508 	}
1509 	if (ifa0 != NULL)
1510 		ifa_free(ifa0);
1511 
1512 	in6_unlink_ifa(ia, ifp);
1513 }
1514 
1515 static void
1516 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1517 {
1518 	int	s = splnet();
1519 
1520 	IF_ADDR_LOCK(ifp);
1521 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1522 	IF_ADDR_UNLOCK(ifp);
1523 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1524 
1525 	/*
1526 	 * Defer the release of what might be the last reference to the
1527 	 * in6_ifaddr so that it can't be freed before the remainder of the
1528 	 * cleanup.
1529 	 */
1530 	IN6_IFADDR_WLOCK();
1531 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1532 	IN6_IFADDR_WUNLOCK();
1533 
1534 	/*
1535 	 * Release the reference to the base prefix.  There should be a
1536 	 * positive reference.
1537 	 */
1538 	if (ia->ia6_ndpr == NULL) {
1539 		nd6log((LOG_NOTICE,
1540 		    "in6_unlink_ifa: autoconf'ed address "
1541 		    "%p has no prefix\n", ia));
1542 	} else {
1543 		ia->ia6_ndpr->ndpr_refcnt--;
1544 		ia->ia6_ndpr = NULL;
1545 	}
1546 
1547 	/*
1548 	 * Also, if the address being removed is autoconf'ed, call
1549 	 * pfxlist_onlink_check() since the release might affect the status of
1550 	 * other (detached) addresses.
1551 	 */
1552 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1553 		pfxlist_onlink_check();
1554 	}
1555 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1556 	splx(s);
1557 }
1558 
1559 void
1560 in6_purgeif(struct ifnet *ifp)
1561 {
1562 	struct ifaddr *ifa, *nifa;
1563 
1564 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1565 		if (ifa->ifa_addr->sa_family != AF_INET6)
1566 			continue;
1567 		in6_purgeaddr(ifa);
1568 	}
1569 
1570 	in6_ifdetach(ifp);
1571 }
1572 
1573 /*
1574  * SIOC[GAD]LIFADDR.
1575  *	SIOCGLIFADDR: get first address. (?)
1576  *	SIOCGLIFADDR with IFLR_PREFIX:
1577  *		get first address that matches the specified prefix.
1578  *	SIOCALIFADDR: add the specified address.
1579  *	SIOCALIFADDR with IFLR_PREFIX:
1580  *		add the specified prefix, filling hostid part from
1581  *		the first link-local address.  prefixlen must be <= 64.
1582  *	SIOCDLIFADDR: delete the specified address.
1583  *	SIOCDLIFADDR with IFLR_PREFIX:
1584  *		delete the first address that matches the specified prefix.
1585  * return values:
1586  *	EINVAL on invalid parameters
1587  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1588  *	other values may be returned from in6_ioctl()
1589  *
1590  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1591  * this is to accomodate address naming scheme other than RFC2374,
1592  * in the future.
1593  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1594  * address encoding scheme. (see figure on page 8)
1595  */
1596 static int
1597 in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1598     struct ifnet *ifp, struct thread *td)
1599 {
1600 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1601 	struct ifaddr *ifa;
1602 	struct sockaddr *sa;
1603 
1604 	/* sanity checks */
1605 	if (!data || !ifp) {
1606 		panic("invalid argument to in6_lifaddr_ioctl");
1607 		/* NOTREACHED */
1608 	}
1609 
1610 	switch (cmd) {
1611 	case SIOCGLIFADDR:
1612 		/* address must be specified on GET with IFLR_PREFIX */
1613 		if ((iflr->flags & IFLR_PREFIX) == 0)
1614 			break;
1615 		/* FALLTHROUGH */
1616 	case SIOCALIFADDR:
1617 	case SIOCDLIFADDR:
1618 		/* address must be specified on ADD and DELETE */
1619 		sa = (struct sockaddr *)&iflr->addr;
1620 		if (sa->sa_family != AF_INET6)
1621 			return EINVAL;
1622 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1623 			return EINVAL;
1624 		/* XXX need improvement */
1625 		sa = (struct sockaddr *)&iflr->dstaddr;
1626 		if (sa->sa_family && sa->sa_family != AF_INET6)
1627 			return EINVAL;
1628 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1629 			return EINVAL;
1630 		break;
1631 	default: /* shouldn't happen */
1632 #if 0
1633 		panic("invalid cmd to in6_lifaddr_ioctl");
1634 		/* NOTREACHED */
1635 #else
1636 		return EOPNOTSUPP;
1637 #endif
1638 	}
1639 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1640 		return EINVAL;
1641 
1642 	switch (cmd) {
1643 	case SIOCALIFADDR:
1644 	    {
1645 		struct in6_aliasreq ifra;
1646 		struct in6_addr *hostid = NULL;
1647 		int prefixlen;
1648 
1649 		ifa = NULL;
1650 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1651 			struct sockaddr_in6 *sin6;
1652 
1653 			/*
1654 			 * hostid is to fill in the hostid part of the
1655 			 * address.  hostid points to the first link-local
1656 			 * address attached to the interface.
1657 			 */
1658 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1659 			if (!ifa)
1660 				return EADDRNOTAVAIL;
1661 			hostid = IFA_IN6(ifa);
1662 
1663 			/* prefixlen must be <= 64. */
1664 			if (64 < iflr->prefixlen)
1665 				return EINVAL;
1666 			prefixlen = iflr->prefixlen;
1667 
1668 			/* hostid part must be zero. */
1669 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1670 			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1671 			    sin6->sin6_addr.s6_addr32[3] != 0) {
1672 				return EINVAL;
1673 			}
1674 		} else
1675 			prefixlen = iflr->prefixlen;
1676 
1677 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1678 		bzero(&ifra, sizeof(ifra));
1679 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1680 
1681 		bcopy(&iflr->addr, &ifra.ifra_addr,
1682 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1683 		if (hostid) {
1684 			/* fill in hostid part */
1685 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1686 			    hostid->s6_addr32[2];
1687 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1688 			    hostid->s6_addr32[3];
1689 		}
1690 
1691 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1692 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1693 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1694 			if (hostid) {
1695 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1696 				    hostid->s6_addr32[2];
1697 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1698 				    hostid->s6_addr32[3];
1699 			}
1700 		}
1701 		if (ifa != NULL)
1702 			ifa_free(ifa);
1703 
1704 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1705 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1706 
1707 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1708 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1709 	    }
1710 	case SIOCGLIFADDR:
1711 	case SIOCDLIFADDR:
1712 	    {
1713 		struct in6_ifaddr *ia;
1714 		struct in6_addr mask, candidate, match;
1715 		struct sockaddr_in6 *sin6;
1716 		int cmp;
1717 
1718 		bzero(&mask, sizeof(mask));
1719 		if (iflr->flags & IFLR_PREFIX) {
1720 			/* lookup a prefix rather than address. */
1721 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1722 
1723 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1724 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1725 			match.s6_addr32[0] &= mask.s6_addr32[0];
1726 			match.s6_addr32[1] &= mask.s6_addr32[1];
1727 			match.s6_addr32[2] &= mask.s6_addr32[2];
1728 			match.s6_addr32[3] &= mask.s6_addr32[3];
1729 
1730 			/* if you set extra bits, that's wrong */
1731 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1732 				return EINVAL;
1733 
1734 			cmp = 1;
1735 		} else {
1736 			if (cmd == SIOCGLIFADDR) {
1737 				/* on getting an address, take the 1st match */
1738 				cmp = 0;	/* XXX */
1739 			} else {
1740 				/* on deleting an address, do exact match */
1741 				in6_prefixlen2mask(&mask, 128);
1742 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1743 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1744 
1745 				cmp = 1;
1746 			}
1747 		}
1748 
1749 		IF_ADDR_LOCK(ifp);
1750 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1751 			if (ifa->ifa_addr->sa_family != AF_INET6)
1752 				continue;
1753 			if (!cmp)
1754 				break;
1755 
1756 			/*
1757 			 * XXX: this is adhoc, but is necessary to allow
1758 			 * a user to specify fe80::/64 (not /10) for a
1759 			 * link-local address.
1760 			 */
1761 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1762 			in6_clearscope(&candidate);
1763 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1764 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1765 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1766 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1767 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1768 				break;
1769 		}
1770 		IF_ADDR_UNLOCK(ifp);
1771 		if (!ifa)
1772 			return EADDRNOTAVAIL;
1773 		ia = ifa2ia6(ifa);
1774 
1775 		if (cmd == SIOCGLIFADDR) {
1776 			int error;
1777 
1778 			/* fill in the if_laddrreq structure */
1779 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1780 			error = sa6_recoverscope(
1781 			    (struct sockaddr_in6 *)&iflr->addr);
1782 			if (error != 0)
1783 				return (error);
1784 
1785 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1786 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1787 				    ia->ia_dstaddr.sin6_len);
1788 				error = sa6_recoverscope(
1789 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1790 				if (error != 0)
1791 					return (error);
1792 			} else
1793 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1794 
1795 			iflr->prefixlen =
1796 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1797 
1798 			iflr->flags = ia->ia6_flags;	/* XXX */
1799 
1800 			return 0;
1801 		} else {
1802 			struct in6_aliasreq ifra;
1803 
1804 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1805 			bzero(&ifra, sizeof(ifra));
1806 			bcopy(iflr->iflr_name, ifra.ifra_name,
1807 			    sizeof(ifra.ifra_name));
1808 
1809 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1810 			    ia->ia_addr.sin6_len);
1811 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1812 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1813 				    ia->ia_dstaddr.sin6_len);
1814 			} else {
1815 				bzero(&ifra.ifra_dstaddr,
1816 				    sizeof(ifra.ifra_dstaddr));
1817 			}
1818 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1819 			    ia->ia_prefixmask.sin6_len);
1820 
1821 			ifra.ifra_flags = ia->ia6_flags;
1822 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1823 			    ifp, td);
1824 		}
1825 	    }
1826 	}
1827 
1828 	return EOPNOTSUPP;	/* just for safety */
1829 }
1830 
1831 /*
1832  * Initialize an interface's intetnet6 address
1833  * and routing table entry.
1834  */
1835 static int
1836 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1837     struct sockaddr_in6 *sin6, int newhost)
1838 {
1839 	int	error = 0, plen, ifacount = 0;
1840 	int	s = splimp();
1841 	struct ifaddr *ifa;
1842 
1843 	/*
1844 	 * Give the interface a chance to initialize
1845 	 * if this is its first address,
1846 	 * and to validate the address if necessary.
1847 	 */
1848 	IF_ADDR_LOCK(ifp);
1849 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1850 		if (ifa->ifa_addr->sa_family != AF_INET6)
1851 			continue;
1852 		ifacount++;
1853 	}
1854 	IF_ADDR_UNLOCK(ifp);
1855 
1856 	ia->ia_addr = *sin6;
1857 
1858 	if (ifacount <= 1 && ifp->if_ioctl) {
1859 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1860 		if (error) {
1861 			splx(s);
1862 			return (error);
1863 		}
1864 	}
1865 	splx(s);
1866 
1867 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1868 
1869 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1870 
1871 	/*
1872 	 * Special case:
1873 	 * If a new destination address is specified for a point-to-point
1874 	 * interface, install a route to the destination as an interface
1875 	 * direct route.
1876 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1877 	 * interface that share the same destination.
1878 	 */
1879 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1880 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1881 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1882 		int rtflags = RTF_UP | RTF_HOST;
1883 
1884 		error = rtrequest(RTM_ADD,
1885 		    (struct sockaddr *)&ia->ia_dstaddr,
1886 		    (struct sockaddr *)&ia->ia_addr,
1887 		    (struct sockaddr *)&ia->ia_prefixmask,
1888 		    ia->ia_flags | rtflags, NULL);
1889 		if (error != 0)
1890 			return (error);
1891 		ia->ia_flags |= IFA_ROUTE;
1892 		/*
1893 		 * Handle the case for ::1 .
1894 		 */
1895 		if (ifp->if_flags & IFF_LOOPBACK)
1896 			ia->ia_flags |= IFA_RTSELF;
1897 	}
1898 
1899 	/*
1900 	 * add a loopback route to self
1901 	 */
1902 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1903 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1904 				       (struct sockaddr *)&ia->ia_addr);
1905 		if (error == 0)
1906 			ia->ia_flags |= IFA_RTSELF;
1907 	}
1908 
1909 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1910 	if (newhost)
1911 		in6_ifaddloop(&(ia->ia_ifa));
1912 
1913 	return (error);
1914 }
1915 
1916 /*
1917  * Find an IPv6 interface link-local address specific to an interface.
1918  * ifaddr is returned referenced.
1919  */
1920 struct in6_ifaddr *
1921 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1922 {
1923 	struct ifaddr *ifa;
1924 
1925 	IF_ADDR_LOCK(ifp);
1926 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1927 		if (ifa->ifa_addr->sa_family != AF_INET6)
1928 			continue;
1929 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1930 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1931 			     ignoreflags) != 0)
1932 				continue;
1933 			ifa_ref(ifa);
1934 			break;
1935 		}
1936 	}
1937 	IF_ADDR_UNLOCK(ifp);
1938 
1939 	return ((struct in6_ifaddr *)ifa);
1940 }
1941 
1942 
1943 /*
1944  * find the internet address corresponding to a given interface and address.
1945  * ifaddr is returned referenced.
1946  */
1947 struct in6_ifaddr *
1948 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1949 {
1950 	struct ifaddr *ifa;
1951 
1952 	IF_ADDR_LOCK(ifp);
1953 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1954 		if (ifa->ifa_addr->sa_family != AF_INET6)
1955 			continue;
1956 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1957 			ifa_ref(ifa);
1958 			break;
1959 		}
1960 	}
1961 	IF_ADDR_UNLOCK(ifp);
1962 
1963 	return ((struct in6_ifaddr *)ifa);
1964 }
1965 
1966 /*
1967  * Convert IP6 address to printable (loggable) representation. Caller
1968  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1969  */
1970 static char digits[] = "0123456789abcdef";
1971 char *
1972 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1973 {
1974 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1975 	char *cp;
1976 	const u_int16_t *a = (const u_int16_t *)addr;
1977 	const u_int8_t *d;
1978 	int dcolon = 0, zero = 0;
1979 
1980 	cp = ip6buf;
1981 
1982 	for (i = 0; i < 8; i++) {
1983 		if (*(a + i) == 0) {
1984 			cnt++;
1985 			if (cnt == 1)
1986 				idx = i;
1987 		}
1988 		else if (maxcnt < cnt) {
1989 			maxcnt = cnt;
1990 			index = idx;
1991 			cnt = 0;
1992 		}
1993 	}
1994 	if (maxcnt < cnt) {
1995 		maxcnt = cnt;
1996 		index = idx;
1997 	}
1998 
1999 	for (i = 0; i < 8; i++) {
2000 		if (dcolon == 1) {
2001 			if (*a == 0) {
2002 				if (i == 7)
2003 					*cp++ = ':';
2004 				a++;
2005 				continue;
2006 			} else
2007 				dcolon = 2;
2008 		}
2009 		if (*a == 0) {
2010 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
2011 				if (i == 0)
2012 					*cp++ = ':';
2013 				*cp++ = ':';
2014 				dcolon = 1;
2015 			} else {
2016 				*cp++ = '0';
2017 				*cp++ = ':';
2018 			}
2019 			a++;
2020 			continue;
2021 		}
2022 		d = (const u_char *)a;
2023 		/* Try to eliminate leading zeros in printout like in :0001. */
2024 		zero = 1;
2025 		*cp = digits[*d >> 4];
2026 		if (*cp != '0') {
2027 			zero = 0;
2028 			cp++;
2029 		}
2030 		*cp = digits[*d++ & 0xf];
2031 		if (zero == 0 || (*cp != '0')) {
2032 			zero = 0;
2033 			cp++;
2034 		}
2035 		*cp = digits[*d >> 4];
2036 		if (zero == 0 || (*cp != '0')) {
2037 			zero = 0;
2038 			cp++;
2039 		}
2040 		*cp++ = digits[*d & 0xf];
2041 		*cp++ = ':';
2042 		a++;
2043 	}
2044 	*--cp = '\0';
2045 	return (ip6buf);
2046 }
2047 
2048 int
2049 in6_localaddr(struct in6_addr *in6)
2050 {
2051 	struct in6_ifaddr *ia;
2052 
2053 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2054 		return 1;
2055 
2056 	IN6_IFADDR_RLOCK();
2057 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2058 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2059 		    &ia->ia_prefixmask.sin6_addr)) {
2060 			IN6_IFADDR_RUNLOCK();
2061 			return 1;
2062 		}
2063 	}
2064 	IN6_IFADDR_RUNLOCK();
2065 
2066 	return (0);
2067 }
2068 
2069 /*
2070  * Return 1 if an internet address is for the local host and configured
2071  * on one of its interfaces.
2072  */
2073 int
2074 in6_localip(struct in6_addr *in6)
2075 {
2076 	struct in6_ifaddr *ia;
2077 
2078 	IN6_IFADDR_RLOCK();
2079 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2080 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2081 			IN6_IFADDR_RUNLOCK();
2082 			return (1);
2083 		}
2084 	}
2085 	IN6_IFADDR_RUNLOCK();
2086 	return (0);
2087 }
2088 
2089 
2090 int
2091 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2092 {
2093 	struct in6_ifaddr *ia;
2094 
2095 	IN6_IFADDR_RLOCK();
2096 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2097 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2098 				       &sa6->sin6_addr) &&
2099 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
2100 			IN6_IFADDR_RUNLOCK();
2101 			return (1); /* true */
2102 		}
2103 
2104 		/* XXX: do we still have to go thru the rest of the list? */
2105 	}
2106 	IN6_IFADDR_RUNLOCK();
2107 
2108 	return (0);		/* false */
2109 }
2110 
2111 /*
2112  * return length of part which dst and src are equal
2113  * hard coding...
2114  */
2115 int
2116 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2117 {
2118 	int match = 0;
2119 	u_char *s = (u_char *)src, *d = (u_char *)dst;
2120 	u_char *lim = s + 16, r;
2121 
2122 	while (s < lim)
2123 		if ((r = (*d++ ^ *s++)) != 0) {
2124 			while (r < 128) {
2125 				match++;
2126 				r <<= 1;
2127 			}
2128 			break;
2129 		} else
2130 			match += 8;
2131 	return match;
2132 }
2133 
2134 /* XXX: to be scope conscious */
2135 int
2136 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2137 {
2138 	int bytelen, bitlen;
2139 
2140 	/* sanity check */
2141 	if (0 > len || len > 128) {
2142 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2143 		    len);
2144 		return (0);
2145 	}
2146 
2147 	bytelen = len / 8;
2148 	bitlen = len % 8;
2149 
2150 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2151 		return (0);
2152 	if (bitlen != 0 &&
2153 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2154 	    p2->s6_addr[bytelen] >> (8 - bitlen))
2155 		return (0);
2156 
2157 	return (1);
2158 }
2159 
2160 void
2161 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2162 {
2163 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2164 	int bytelen, bitlen, i;
2165 
2166 	/* sanity check */
2167 	if (0 > len || len > 128) {
2168 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2169 		    len);
2170 		return;
2171 	}
2172 
2173 	bzero(maskp, sizeof(*maskp));
2174 	bytelen = len / 8;
2175 	bitlen = len % 8;
2176 	for (i = 0; i < bytelen; i++)
2177 		maskp->s6_addr[i] = 0xff;
2178 	if (bitlen)
2179 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2180 }
2181 
2182 /*
2183  * return the best address out of the same scope. if no address was
2184  * found, return the first valid address from designated IF.
2185  */
2186 struct in6_ifaddr *
2187 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2188 {
2189 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2190 	struct ifaddr *ifa;
2191 	struct in6_ifaddr *besta = 0;
2192 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2193 
2194 	dep[0] = dep[1] = NULL;
2195 
2196 	/*
2197 	 * We first look for addresses in the same scope.
2198 	 * If there is one, return it.
2199 	 * If two or more, return one which matches the dst longest.
2200 	 * If none, return one of global addresses assigned other ifs.
2201 	 */
2202 	IF_ADDR_LOCK(ifp);
2203 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2204 		if (ifa->ifa_addr->sa_family != AF_INET6)
2205 			continue;
2206 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2207 			continue; /* XXX: is there any case to allow anycast? */
2208 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2209 			continue; /* don't use this interface */
2210 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2211 			continue;
2212 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2213 			if (V_ip6_use_deprecated)
2214 				dep[0] = (struct in6_ifaddr *)ifa;
2215 			continue;
2216 		}
2217 
2218 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2219 			/*
2220 			 * call in6_matchlen() as few as possible
2221 			 */
2222 			if (besta) {
2223 				if (blen == -1)
2224 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2225 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2226 				if (tlen > blen) {
2227 					blen = tlen;
2228 					besta = (struct in6_ifaddr *)ifa;
2229 				}
2230 			} else
2231 				besta = (struct in6_ifaddr *)ifa;
2232 		}
2233 	}
2234 	if (besta) {
2235 		ifa_ref(&besta->ia_ifa);
2236 		IF_ADDR_UNLOCK(ifp);
2237 		return (besta);
2238 	}
2239 	IF_ADDR_UNLOCK(ifp);
2240 
2241 	IN6_IFADDR_RLOCK();
2242 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2243 		if (ifa->ifa_addr->sa_family != AF_INET6)
2244 			continue;
2245 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2246 			continue; /* XXX: is there any case to allow anycast? */
2247 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2248 			continue; /* don't use this interface */
2249 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2250 			continue;
2251 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2252 			if (V_ip6_use_deprecated)
2253 				dep[1] = (struct in6_ifaddr *)ifa;
2254 			continue;
2255 		}
2256 
2257 		if (ifa != NULL)
2258 			ifa_ref(ifa);
2259 		IN6_IFADDR_RUNLOCK();
2260 		return (struct in6_ifaddr *)ifa;
2261 	}
2262 	IN6_IFADDR_RUNLOCK();
2263 
2264 	/* use the last-resort values, that are, deprecated addresses */
2265 	if (dep[0])
2266 		return dep[0];
2267 	if (dep[1])
2268 		return dep[1];
2269 
2270 	return NULL;
2271 }
2272 
2273 /*
2274  * perform DAD when interface becomes IFF_UP.
2275  */
2276 void
2277 in6_if_up(struct ifnet *ifp)
2278 {
2279 	struct ifaddr *ifa;
2280 	struct in6_ifaddr *ia;
2281 
2282 	IF_ADDR_LOCK(ifp);
2283 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2284 		if (ifa->ifa_addr->sa_family != AF_INET6)
2285 			continue;
2286 		ia = (struct in6_ifaddr *)ifa;
2287 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2288 			/*
2289 			 * The TENTATIVE flag was likely set by hand
2290 			 * beforehand, implicitly indicating the need for DAD.
2291 			 * We may be able to skip the random delay in this
2292 			 * case, but we impose delays just in case.
2293 			 */
2294 			nd6_dad_start(ifa,
2295 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2296 		}
2297 	}
2298 	IF_ADDR_UNLOCK(ifp);
2299 
2300 	/*
2301 	 * special cases, like 6to4, are handled in in6_ifattach
2302 	 */
2303 	in6_ifattach(ifp, NULL);
2304 }
2305 
2306 int
2307 in6if_do_dad(struct ifnet *ifp)
2308 {
2309 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2310 		return (0);
2311 
2312 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2313 		return (0);
2314 
2315 	switch (ifp->if_type) {
2316 #ifdef IFT_DUMMY
2317 	case IFT_DUMMY:
2318 #endif
2319 	case IFT_FAITH:
2320 		/*
2321 		 * These interfaces do not have the IFF_LOOPBACK flag,
2322 		 * but loop packets back.  We do not have to do DAD on such
2323 		 * interfaces.  We should even omit it, because loop-backed
2324 		 * NS would confuse the DAD procedure.
2325 		 */
2326 		return (0);
2327 	default:
2328 		/*
2329 		 * Our DAD routine requires the interface up and running.
2330 		 * However, some interfaces can be up before the RUNNING
2331 		 * status.  Additionaly, users may try to assign addresses
2332 		 * before the interface becomes up (or running).
2333 		 * We simply skip DAD in such a case as a work around.
2334 		 * XXX: we should rather mark "tentative" on such addresses,
2335 		 * and do DAD after the interface becomes ready.
2336 		 */
2337 		if (!((ifp->if_flags & IFF_UP) &&
2338 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2339 			return (0);
2340 
2341 		return (1);
2342 	}
2343 }
2344 
2345 /*
2346  * Calculate max IPv6 MTU through all the interfaces and store it
2347  * to in6_maxmtu.
2348  */
2349 void
2350 in6_setmaxmtu(void)
2351 {
2352 	unsigned long maxmtu = 0;
2353 	struct ifnet *ifp;
2354 
2355 	IFNET_RLOCK_NOSLEEP();
2356 	for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
2357 	    ifp = TAILQ_NEXT(ifp, if_list)) {
2358 		/* this function can be called during ifnet initialization */
2359 		if (!ifp->if_afdata[AF_INET6])
2360 			continue;
2361 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2362 		    IN6_LINKMTU(ifp) > maxmtu)
2363 			maxmtu = IN6_LINKMTU(ifp);
2364 	}
2365 	IFNET_RUNLOCK_NOSLEEP();
2366 	if (maxmtu)	     /* update only when maxmtu is positive */
2367 		V_in6_maxmtu = maxmtu;
2368 }
2369 
2370 /*
2371  * Provide the length of interface identifiers to be used for the link attached
2372  * to the given interface.  The length should be defined in "IPv6 over
2373  * xxx-link" document.  Note that address architecture might also define
2374  * the length for a particular set of address prefixes, regardless of the
2375  * link type.  As clarified in rfc2462bis, those two definitions should be
2376  * consistent, and those really are as of August 2004.
2377  */
2378 int
2379 in6_if2idlen(struct ifnet *ifp)
2380 {
2381 	switch (ifp->if_type) {
2382 	case IFT_ETHER:		/* RFC2464 */
2383 #ifdef IFT_PROPVIRTUAL
2384 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2385 #endif
2386 #ifdef IFT_L2VLAN
2387 	case IFT_L2VLAN:	/* ditto */
2388 #endif
2389 #ifdef IFT_IEEE80211
2390 	case IFT_IEEE80211:	/* ditto */
2391 #endif
2392 #ifdef IFT_MIP
2393 	case IFT_MIP:	/* ditto */
2394 #endif
2395 	case IFT_INFINIBAND:
2396 		return (64);
2397 	case IFT_FDDI:		/* RFC2467 */
2398 		return (64);
2399 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2400 		return (64);
2401 	case IFT_PPP:		/* RFC2472 */
2402 		return (64);
2403 	case IFT_ARCNET:	/* RFC2497 */
2404 		return (64);
2405 	case IFT_FRELAY:	/* RFC2590 */
2406 		return (64);
2407 	case IFT_IEEE1394:	/* RFC3146 */
2408 		return (64);
2409 	case IFT_GIF:
2410 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2411 	case IFT_LOOP:
2412 		return (64);	/* XXX: is this really correct? */
2413 	default:
2414 		/*
2415 		 * Unknown link type:
2416 		 * It might be controversial to use the today's common constant
2417 		 * of 64 for these cases unconditionally.  For full compliance,
2418 		 * we should return an error in this case.  On the other hand,
2419 		 * if we simply miss the standard for the link type or a new
2420 		 * standard is defined for a new link type, the IFID length
2421 		 * is very likely to be the common constant.  As a compromise,
2422 		 * we always use the constant, but make an explicit notice
2423 		 * indicating the "unknown" case.
2424 		 */
2425 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2426 		return (64);
2427 	}
2428 }
2429 
2430 #include <sys/sysctl.h>
2431 
2432 struct in6_llentry {
2433 	struct llentry		base;
2434 	struct sockaddr_in6	l3_addr6;
2435 };
2436 
2437 static struct llentry *
2438 in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2439 {
2440 	struct in6_llentry *lle;
2441 
2442 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE,
2443 	    M_DONTWAIT | M_ZERO);
2444 	if (lle == NULL)		/* NB: caller generates msg */
2445 		return NULL;
2446 
2447 	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2448 	lle->base.lle_refcnt = 1;
2449 	LLE_LOCK_INIT(&lle->base);
2450 	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2451 	    CALLOUT_RETURNUNLOCKED);
2452 
2453 	return &lle->base;
2454 }
2455 
2456 /*
2457  * Deletes an address from the address table.
2458  * This function is called by the timer functions
2459  * such as arptimer() and nd6_llinfo_timer(), and
2460  * the caller does the locking.
2461  */
2462 static void
2463 in6_lltable_free(struct lltable *llt, struct llentry *lle)
2464 {
2465 	LLE_WUNLOCK(lle);
2466 	LLE_LOCK_DESTROY(lle);
2467 	free(lle, M_LLTABLE);
2468 }
2469 
2470 static void
2471 in6_lltable_prefix_free(struct lltable *llt,
2472 			const struct sockaddr *prefix,
2473 			const struct sockaddr *mask,
2474 			u_int flags)
2475 {
2476 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2477 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2478 	struct llentry *lle, *next;
2479 	register int i;
2480 
2481 	/*
2482 	 * (flags & LLE_STATIC) means deleting all entries
2483 	 * including static ND6 entries
2484 	 */
2485 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
2486 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2487 			if (IN6_ARE_MASKED_ADDR_EQUAL(
2488 				    &((struct sockaddr_in6 *)L3_ADDR(lle))->sin6_addr,
2489 				    &pfx->sin6_addr,
2490 				    &msk->sin6_addr) &&
2491 			    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
2492 				int canceled;
2493 
2494 				canceled = callout_drain(&lle->la_timer);
2495 				LLE_WLOCK(lle);
2496 				if (canceled)
2497 					LLE_REMREF(lle);
2498 				llentry_free(lle);
2499 			}
2500 		}
2501 	}
2502 }
2503 
2504 static int
2505 in6_lltable_rtcheck(struct ifnet *ifp,
2506 		    u_int flags,
2507 		    const struct sockaddr *l3addr)
2508 {
2509 	struct rtentry *rt;
2510 	char ip6buf[INET6_ADDRSTRLEN];
2511 
2512 	KASSERT(l3addr->sa_family == AF_INET6,
2513 	    ("sin_family %d", l3addr->sa_family));
2514 
2515 	/* XXX rtalloc1 should take a const param */
2516 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
2517 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2518 		struct ifaddr *ifa;
2519 		/*
2520 		 * Create an ND6 cache for an IPv6 neighbor
2521 		 * that is not covered by our own prefix.
2522 		 */
2523 		/* XXX ifaof_ifpforaddr should take a const param */
2524 		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2525 		if (ifa != NULL) {
2526 			ifa_free(ifa);
2527 			if (rt != NULL)
2528 				RTFREE_LOCKED(rt);
2529 			return 0;
2530 		}
2531 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2532 		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2533 		if (rt != NULL)
2534 			RTFREE_LOCKED(rt);
2535 		return EINVAL;
2536 	}
2537 	RTFREE_LOCKED(rt);
2538 	return 0;
2539 }
2540 
2541 static struct llentry *
2542 in6_lltable_lookup(struct lltable *llt, u_int flags,
2543 	const struct sockaddr *l3addr)
2544 {
2545 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2546 	struct ifnet *ifp = llt->llt_ifp;
2547 	struct llentry *lle;
2548 	struct llentries *lleh;
2549 	u_int hashkey;
2550 
2551 	IF_AFDATA_LOCK_ASSERT(ifp);
2552 	KASSERT(l3addr->sa_family == AF_INET6,
2553 	    ("sin_family %d", l3addr->sa_family));
2554 
2555 	hashkey = sin6->sin6_addr.s6_addr32[3];
2556 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2557 	LIST_FOREACH(lle, lleh, lle_next) {
2558 		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2559 		if (lle->la_flags & LLE_DELETED)
2560 			continue;
2561 		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2562 			 sizeof(struct in6_addr)) == 0)
2563 			break;
2564 	}
2565 
2566 	if (lle == NULL) {
2567 		if (!(flags & LLE_CREATE))
2568 			return (NULL);
2569 		/*
2570 		 * A route that covers the given address must have
2571 		 * been installed 1st because we are doing a resolution,
2572 		 * verify this.
2573 		 */
2574 		if (!(flags & LLE_IFADDR) &&
2575 		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2576 			return NULL;
2577 
2578 		lle = in6_lltable_new(l3addr, flags);
2579 		if (lle == NULL) {
2580 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2581 			return NULL;
2582 		}
2583 		lle->la_flags = flags & ~LLE_CREATE;
2584 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2585 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2586 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2587 		}
2588 
2589 		lle->lle_tbl  = llt;
2590 		lle->lle_head = lleh;
2591 		LIST_INSERT_HEAD(lleh, lle, lle_next);
2592 	} else if (flags & LLE_DELETE) {
2593 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2594 			LLE_WLOCK(lle);
2595 			lle->la_flags = LLE_DELETED;
2596 			LLE_WUNLOCK(lle);
2597 #ifdef DIAGNOSTIC
2598 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2599 #endif
2600 		}
2601 		lle = (void *)-1;
2602 	}
2603 	if (LLE_IS_VALID(lle)) {
2604 		if (flags & LLE_EXCLUSIVE)
2605 			LLE_WLOCK(lle);
2606 		else
2607 			LLE_RLOCK(lle);
2608 	}
2609 	return (lle);
2610 }
2611 
2612 static int
2613 in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2614 {
2615 	struct ifnet *ifp = llt->llt_ifp;
2616 	struct llentry *lle;
2617 	/* XXX stack use */
2618 	struct {
2619 		struct rt_msghdr	rtm;
2620 		struct sockaddr_in6	sin6;
2621 		/*
2622 		 * ndp.c assumes that sdl is word aligned
2623 		 */
2624 #ifdef __LP64__
2625 		uint32_t		pad;
2626 #endif
2627 		struct sockaddr_dl	sdl;
2628 	} ndpc;
2629 	int i, error;
2630 
2631 	if (ifp->if_flags & IFF_LOOPBACK)
2632 		return 0;
2633 
2634 	LLTABLE_LOCK_ASSERT();
2635 
2636 	error = 0;
2637 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2638 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2639 			struct sockaddr_dl *sdl;
2640 
2641 			/* skip deleted or invalid entries */
2642 			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2643 				continue;
2644 			/* Skip if jailed and not a valid IP of the prison. */
2645 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2646 				continue;
2647 			/*
2648 			 * produce a msg made of:
2649 			 *  struct rt_msghdr;
2650 			 *  struct sockaddr_in6 (IPv6)
2651 			 *  struct sockaddr_dl;
2652 			 */
2653 			bzero(&ndpc, sizeof(ndpc));
2654 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2655 			ndpc.rtm.rtm_version = RTM_VERSION;
2656 			ndpc.rtm.rtm_type = RTM_GET;
2657 			ndpc.rtm.rtm_flags = RTF_UP;
2658 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2659 			ndpc.sin6.sin6_family = AF_INET6;
2660 			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2661 			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2662 
2663 			/* publish */
2664 			if (lle->la_flags & LLE_PUB)
2665 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2666 
2667 			sdl = &ndpc.sdl;
2668 			sdl->sdl_family = AF_LINK;
2669 			sdl->sdl_len = sizeof(*sdl);
2670 			sdl->sdl_alen = ifp->if_addrlen;
2671 			sdl->sdl_index = ifp->if_index;
2672 			sdl->sdl_type = ifp->if_type;
2673 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2674 			ndpc.rtm.rtm_rmx.rmx_expire =
2675 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2676 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2677 			if (lle->la_flags & LLE_STATIC)
2678 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2679 			ndpc.rtm.rtm_index = ifp->if_index;
2680 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2681 			if (error)
2682 				break;
2683 		}
2684 	}
2685 	return error;
2686 }
2687 
2688 void *
2689 in6_domifattach(struct ifnet *ifp)
2690 {
2691 	struct in6_ifextra *ext;
2692 
2693 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2694 	bzero(ext, sizeof(*ext));
2695 
2696 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2697 	    M_IFADDR, M_WAITOK);
2698 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2699 
2700 	ext->icmp6_ifstat =
2701 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2702 	    M_IFADDR, M_WAITOK);
2703 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2704 
2705 	ext->nd_ifinfo = nd6_ifattach(ifp);
2706 	ext->scope6_id = scope6_ifattach(ifp);
2707 	ext->lltable = lltable_init(ifp, AF_INET6);
2708 	if (ext->lltable != NULL) {
2709 		ext->lltable->llt_free = in6_lltable_free;
2710 		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2711 		ext->lltable->llt_lookup = in6_lltable_lookup;
2712 		ext->lltable->llt_dump = in6_lltable_dump;
2713 	}
2714 
2715 	ext->mld_ifinfo = mld_domifattach(ifp);
2716 
2717 	return ext;
2718 }
2719 
2720 void
2721 in6_domifdetach(struct ifnet *ifp, void *aux)
2722 {
2723 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2724 
2725 	mld_domifdetach(ifp);
2726 	scope6_ifdetach(ext->scope6_id);
2727 	nd6_ifdetach(ext->nd_ifinfo);
2728 	lltable_free(ext->lltable);
2729 	free(ext->in6_ifstat, M_IFADDR);
2730 	free(ext->icmp6_ifstat, M_IFADDR);
2731 	free(ext, M_IFADDR);
2732 }
2733 
2734 /*
2735  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2736  * v4 mapped addr or v4 compat addr
2737  */
2738 void
2739 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2740 {
2741 
2742 	bzero(sin, sizeof(*sin));
2743 	sin->sin_len = sizeof(struct sockaddr_in);
2744 	sin->sin_family = AF_INET;
2745 	sin->sin_port = sin6->sin6_port;
2746 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2747 }
2748 
2749 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2750 void
2751 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2752 {
2753 	bzero(sin6, sizeof(*sin6));
2754 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2755 	sin6->sin6_family = AF_INET6;
2756 	sin6->sin6_port = sin->sin_port;
2757 	sin6->sin6_addr.s6_addr32[0] = 0;
2758 	sin6->sin6_addr.s6_addr32[1] = 0;
2759 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2760 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2761 }
2762 
2763 /* Convert sockaddr_in6 into sockaddr_in. */
2764 void
2765 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2766 {
2767 	struct sockaddr_in *sin_p;
2768 	struct sockaddr_in6 sin6;
2769 
2770 	/*
2771 	 * Save original sockaddr_in6 addr and convert it
2772 	 * to sockaddr_in.
2773 	 */
2774 	sin6 = *(struct sockaddr_in6 *)nam;
2775 	sin_p = (struct sockaddr_in *)nam;
2776 	in6_sin6_2_sin(sin_p, &sin6);
2777 }
2778 
2779 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2780 void
2781 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2782 {
2783 	struct sockaddr_in *sin_p;
2784 	struct sockaddr_in6 *sin6_p;
2785 
2786 	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
2787 	       M_WAITOK);
2788 	sin_p = (struct sockaddr_in *)*nam;
2789 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2790 	free(*nam, M_SONAME);
2791 	*nam = (struct sockaddr *)sin6_p;
2792 }
2793