xref: /freebsd/sys/netinet6/in6.c (revision 87c1627502a5dde91e5284118eec8682b60f27a2)
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 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
110 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
111 
112 /*
113  * Definitions of some costant IP6 addresses.
114  */
115 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
116 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
117 const struct in6_addr in6addr_nodelocal_allnodes =
118 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
119 const struct in6_addr in6addr_linklocal_allnodes =
120 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
121 const struct in6_addr in6addr_linklocal_allrouters =
122 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
123 const struct in6_addr in6addr_linklocal_allv2routers =
124 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
125 
126 const struct in6_addr in6mask0 = IN6MASK0;
127 const struct in6_addr in6mask32 = IN6MASK32;
128 const struct in6_addr in6mask64 = IN6MASK64;
129 const struct in6_addr in6mask96 = IN6MASK96;
130 const struct in6_addr in6mask128 = IN6MASK128;
131 
132 const struct sockaddr_in6 sa6_any =
133 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
134 
135 static int in6_lifaddr_ioctl(struct socket *, u_long, caddr_t,
136 	struct ifnet *, struct thread *);
137 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
138 	struct sockaddr_in6 *, int);
139 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
140 
141 int	(*faithprefix_p)(struct in6_addr *);
142 
143 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
144 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
145 
146 void
147 in6_ifaddloop(struct ifaddr *ifa)
148 {
149 	struct sockaddr_dl gateway;
150 	struct sockaddr_in6 mask, addr;
151 	struct rtentry rt;
152 	struct in6_ifaddr *ia;
153 	struct ifnet *ifp;
154 	struct llentry *ln;
155 
156 	ia = ifa2ia6(ifa);
157 	ifp = ifa->ifa_ifp;
158 	IF_AFDATA_LOCK(ifp);
159 	ifa->ifa_rtrequest = nd6_rtrequest;
160 	ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
161 	    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
162 	IF_AFDATA_UNLOCK(ifp);
163 	if (ln != NULL) {
164 		ln->la_expire = 0;  /* for IPv6 this means permanent */
165 		ln->ln_state = ND6_LLINFO_REACHABLE;
166 		/*
167 		 * initialize for rtmsg generation
168 		 */
169 		bzero(&gateway, sizeof(gateway));
170 		gateway.sdl_len = sizeof(gateway);
171 		gateway.sdl_family = AF_LINK;
172 		gateway.sdl_nlen = 0;
173 		gateway.sdl_alen = 6;
174 		memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
175 		    sizeof(ln->ll_addr));
176 		LLE_WUNLOCK(ln);
177 	}
178 
179 	bzero(&rt, sizeof(rt));
180 	rt.rt_gateway = (struct sockaddr *)&gateway;
181 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
182 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
183 	rt_mask(&rt) = (struct sockaddr *)&mask;
184 	rt_key(&rt) = (struct sockaddr *)&addr;
185 	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
186 	/* Announce arrival of local address to all FIBs. */
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 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
202 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
203 	lltable_prefix_free(AF_INET6, (struct sockaddr *)&addr,
204 	            (struct sockaddr *)&mask, LLE_STATIC);
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 	rt_mask(&rt0) = (struct sockaddr *)&mask;
217 	rt_key(&rt0) = (struct sockaddr *)&addr;
218 	rt0.rt_flags = RTF_HOST | RTF_STATIC;
219 	/* Announce removal of local address to all FIBs. */
220 	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
221 }
222 
223 int
224 in6_mask2len(struct in6_addr *mask, u_char *lim0)
225 {
226 	int x = 0, y;
227 	u_char *lim = lim0, *p;
228 
229 	/* ignore the scope_id part */
230 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
231 		lim = (u_char *)mask + sizeof(*mask);
232 	for (p = (u_char *)mask; p < lim; x++, p++) {
233 		if (*p != 0xff)
234 			break;
235 	}
236 	y = 0;
237 	if (p < lim) {
238 		for (y = 0; y < 8; y++) {
239 			if ((*p & (0x80 >> y)) == 0)
240 				break;
241 		}
242 	}
243 
244 	/*
245 	 * when the limit pointer is given, do a stricter check on the
246 	 * remaining bits.
247 	 */
248 	if (p < lim) {
249 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
250 			return (-1);
251 		for (p = p + 1; p < lim; p++)
252 			if (*p != 0)
253 				return (-1);
254 	}
255 
256 	return x * 8 + y;
257 }
258 
259 #ifdef COMPAT_FREEBSD32
260 struct in6_ndifreq32 {
261 	char ifname[IFNAMSIZ];
262 	uint32_t ifindex;
263 };
264 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
265 #endif
266 
267 int
268 in6_control(struct socket *so, u_long cmd, caddr_t data,
269     struct ifnet *ifp, struct thread *td)
270 {
271 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
272 	struct	in6_ifaddr *ia = NULL;
273 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
274 	struct sockaddr_in6 *sa6;
275 	int carp_attached = 0;
276 	int error;
277 	u_long ocmd = cmd;
278 
279 	/*
280 	 * Compat to make pre-10.x ifconfig(8) operable.
281 	 */
282 	if (cmd == OSIOCAIFADDR_IN6)
283 		cmd = SIOCAIFADDR_IN6;
284 
285 	switch (cmd) {
286 	case SIOCGETSGCNT_IN6:
287 	case SIOCGETMIFCNT_IN6:
288 		/*
289 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
290 		 * We cannot see how that would be needed, so do not adjust the
291 		 * KPI blindly; more likely should clean up the IPv4 variant.
292 		 */
293 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
294 	}
295 
296 	switch(cmd) {
297 	case SIOCAADDRCTL_POLICY:
298 	case SIOCDADDRCTL_POLICY:
299 		if (td != NULL) {
300 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
301 			if (error)
302 				return (error);
303 		}
304 		return (in6_src_ioctl(cmd, data));
305 	}
306 
307 	if (ifp == NULL)
308 		return (EOPNOTSUPP);
309 
310 	switch (cmd) {
311 	case SIOCSNDFLUSH_IN6:
312 	case SIOCSPFXFLUSH_IN6:
313 	case SIOCSRTRFLUSH_IN6:
314 	case SIOCSDEFIFACE_IN6:
315 	case SIOCSIFINFO_FLAGS:
316 	case SIOCSIFINFO_IN6:
317 		if (td != NULL) {
318 			error = priv_check(td, PRIV_NETINET_ND6);
319 			if (error)
320 				return (error);
321 		}
322 		/* FALLTHROUGH */
323 	case OSIOCGIFINFO_IN6:
324 	case SIOCGIFINFO_IN6:
325 	case SIOCGDRLST_IN6:
326 	case SIOCGPRLST_IN6:
327 	case SIOCGNBRINFO_IN6:
328 	case SIOCGDEFIFACE_IN6:
329 		return (nd6_ioctl(cmd, data, ifp));
330 
331 #ifdef COMPAT_FREEBSD32
332 	case SIOCGDEFIFACE32_IN6:
333 		{
334 			struct in6_ndifreq ndif;
335 			struct in6_ndifreq32 *ndif32;
336 
337 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
338 			    ifp);
339 			if (error)
340 				return (error);
341 			ndif32 = (struct in6_ndifreq32 *)data;
342 			ndif32->ifindex = ndif.ifindex;
343 			return (0);
344 		}
345 #endif
346 	}
347 
348 	switch (cmd) {
349 	case SIOCSIFPREFIX_IN6:
350 	case SIOCDIFPREFIX_IN6:
351 	case SIOCAIFPREFIX_IN6:
352 	case SIOCCIFPREFIX_IN6:
353 	case SIOCSGIFPREFIX_IN6:
354 	case SIOCGIFPREFIX_IN6:
355 		log(LOG_NOTICE,
356 		    "prefix ioctls are now invalidated. "
357 		    "please use ifconfig.\n");
358 		return (EOPNOTSUPP);
359 	}
360 
361 	switch (cmd) {
362 	case SIOCSSCOPE6:
363 		if (td != NULL) {
364 			error = priv_check(td, PRIV_NETINET_SCOPE6);
365 			if (error)
366 				return (error);
367 		}
368 		return (scope6_set(ifp,
369 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
370 	case SIOCGSCOPE6:
371 		return (scope6_get(ifp,
372 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
373 	case SIOCGSCOPE6DEF:
374 		return (scope6_get_default((struct scope6_id *)
375 		    ifr->ifr_ifru.ifru_scope_id));
376 	}
377 
378 	switch (cmd) {
379 	case SIOCALIFADDR:
380 		if (td != NULL) {
381 			error = priv_check(td, PRIV_NET_ADDIFADDR);
382 			if (error)
383 				return (error);
384 		}
385 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
386 
387 	case SIOCDLIFADDR:
388 		if (td != NULL) {
389 			error = priv_check(td, PRIV_NET_DELIFADDR);
390 			if (error)
391 				return (error);
392 		}
393 		/* FALLTHROUGH */
394 	case SIOCGLIFADDR:
395 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
396 	}
397 
398 	/*
399 	 * Find address for this interface, if it exists.
400 	 *
401 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
402 	 * only, and used the first interface address as the target of other
403 	 * operations (without checking ifra_addr).  This was because netinet
404 	 * code/API assumed at most 1 interface address per interface.
405 	 * Since IPv6 allows a node to assign multiple addresses
406 	 * on a single interface, we almost always look and check the
407 	 * presence of ifra_addr, and reject invalid ones here.
408 	 * It also decreases duplicated code among SIOC*_IN6 operations.
409 	 */
410 	switch (cmd) {
411 	case SIOCAIFADDR_IN6:
412 	case SIOCSIFPHYADDR_IN6:
413 		sa6 = &ifra->ifra_addr;
414 		break;
415 	case SIOCSIFADDR_IN6:
416 	case SIOCGIFADDR_IN6:
417 	case SIOCSIFDSTADDR_IN6:
418 	case SIOCSIFNETMASK_IN6:
419 	case SIOCGIFDSTADDR_IN6:
420 	case SIOCGIFNETMASK_IN6:
421 	case SIOCDIFADDR_IN6:
422 	case SIOCGIFPSRCADDR_IN6:
423 	case SIOCGIFPDSTADDR_IN6:
424 	case SIOCGIFAFLAG_IN6:
425 	case SIOCSNDFLUSH_IN6:
426 	case SIOCSPFXFLUSH_IN6:
427 	case SIOCSRTRFLUSH_IN6:
428 	case SIOCGIFALIFETIME_IN6:
429 	case SIOCSIFALIFETIME_IN6:
430 	case SIOCGIFSTAT_IN6:
431 	case SIOCGIFSTAT_ICMP6:
432 		sa6 = &ifr->ifr_addr;
433 		break;
434 	default:
435 		sa6 = NULL;
436 		break;
437 	}
438 	if (sa6 && sa6->sin6_family == AF_INET6) {
439 		if (sa6->sin6_scope_id != 0)
440 			error = sa6_embedscope(sa6, 0);
441 		else
442 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
443 		if (error != 0)
444 			return (error);
445 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
446 		    &sa6->sin6_addr)) != 0)
447 			return (error);
448 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
449 	} else
450 		ia = NULL;
451 
452 	switch (cmd) {
453 	case SIOCSIFADDR_IN6:
454 	case SIOCSIFDSTADDR_IN6:
455 	case SIOCSIFNETMASK_IN6:
456 		/*
457 		 * Since IPv6 allows a node to assign multiple addresses
458 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
459 		 */
460 		/* we decided to obsolete this command (20000704) */
461 		error = EINVAL;
462 		goto out;
463 
464 	case SIOCDIFADDR_IN6:
465 		/*
466 		 * for IPv4, we look for existing in_ifaddr here to allow
467 		 * "ifconfig if0 delete" to remove the first IPv4 address on
468 		 * the interface.  For IPv6, as the spec allows multiple
469 		 * interface address from the day one, we consider "remove the
470 		 * first one" semantics to be not preferable.
471 		 */
472 		if (ia == NULL) {
473 			error = EADDRNOTAVAIL;
474 			goto out;
475 		}
476 		/* FALLTHROUGH */
477 	case SIOCAIFADDR_IN6:
478 		/*
479 		 * We always require users to specify a valid IPv6 address for
480 		 * the corresponding operation.
481 		 */
482 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
483 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
484 			error = EAFNOSUPPORT;
485 			goto out;
486 		}
487 
488 		if (td != NULL) {
489 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
490 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
491 			if (error)
492 				goto out;
493 		}
494 		break;
495 
496 	case SIOCGIFADDR_IN6:
497 		/* This interface is basically deprecated. use SIOCGIFCONF. */
498 		/* FALLTHROUGH */
499 	case SIOCGIFAFLAG_IN6:
500 	case SIOCGIFNETMASK_IN6:
501 	case SIOCGIFDSTADDR_IN6:
502 	case SIOCGIFALIFETIME_IN6:
503 		/* must think again about its semantics */
504 		if (ia == NULL) {
505 			error = EADDRNOTAVAIL;
506 			goto out;
507 		}
508 		break;
509 
510 	case SIOCSIFALIFETIME_IN6:
511 	    {
512 		struct in6_addrlifetime *lt;
513 
514 		if (td != NULL) {
515 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
516 			if (error)
517 				goto out;
518 		}
519 		if (ia == NULL) {
520 			error = EADDRNOTAVAIL;
521 			goto out;
522 		}
523 		/* sanity for overflow - beware unsigned */
524 		lt = &ifr->ifr_ifru.ifru_lifetime;
525 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
526 		    lt->ia6t_vltime + time_second < time_second) {
527 			error = EINVAL;
528 			goto out;
529 		}
530 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
531 		    lt->ia6t_pltime + time_second < time_second) {
532 			error = EINVAL;
533 			goto out;
534 		}
535 		break;
536 	    }
537 	}
538 
539 	switch (cmd) {
540 	case SIOCGIFADDR_IN6:
541 		ifr->ifr_addr = ia->ia_addr;
542 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
543 			goto out;
544 		break;
545 
546 	case SIOCGIFDSTADDR_IN6:
547 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
548 			error = EINVAL;
549 			goto out;
550 		}
551 		/*
552 		 * XXX: should we check if ifa_dstaddr is NULL and return
553 		 * an error?
554 		 */
555 		ifr->ifr_dstaddr = ia->ia_dstaddr;
556 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
557 			goto out;
558 		break;
559 
560 	case SIOCGIFNETMASK_IN6:
561 		ifr->ifr_addr = ia->ia_prefixmask;
562 		break;
563 
564 	case SIOCGIFAFLAG_IN6:
565 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
566 		break;
567 
568 	case SIOCGIFSTAT_IN6:
569 		if (ifp == NULL) {
570 			error = EINVAL;
571 			goto out;
572 		}
573 		bzero(&ifr->ifr_ifru.ifru_stat,
574 		    sizeof(ifr->ifr_ifru.ifru_stat));
575 		ifr->ifr_ifru.ifru_stat =
576 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
577 		break;
578 
579 	case SIOCGIFSTAT_ICMP6:
580 		if (ifp == NULL) {
581 			error = EINVAL;
582 			goto out;
583 		}
584 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
585 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
586 		ifr->ifr_ifru.ifru_icmp6stat =
587 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
588 		break;
589 
590 	case SIOCGIFALIFETIME_IN6:
591 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
592 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
593 			time_t maxexpire;
594 			struct in6_addrlifetime *retlt =
595 			    &ifr->ifr_ifru.ifru_lifetime;
596 
597 			/*
598 			 * XXX: adjust expiration time assuming time_t is
599 			 * signed.
600 			 */
601 			maxexpire = (-1) &
602 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
603 			if (ia->ia6_lifetime.ia6t_vltime <
604 			    maxexpire - ia->ia6_updatetime) {
605 				retlt->ia6t_expire = ia->ia6_updatetime +
606 				    ia->ia6_lifetime.ia6t_vltime;
607 			} else
608 				retlt->ia6t_expire = maxexpire;
609 		}
610 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
611 			time_t maxexpire;
612 			struct in6_addrlifetime *retlt =
613 			    &ifr->ifr_ifru.ifru_lifetime;
614 
615 			/*
616 			 * XXX: adjust expiration time assuming time_t is
617 			 * signed.
618 			 */
619 			maxexpire = (-1) &
620 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
621 			if (ia->ia6_lifetime.ia6t_pltime <
622 			    maxexpire - ia->ia6_updatetime) {
623 				retlt->ia6t_preferred = ia->ia6_updatetime +
624 				    ia->ia6_lifetime.ia6t_pltime;
625 			} else
626 				retlt->ia6t_preferred = maxexpire;
627 		}
628 		break;
629 
630 	case SIOCSIFALIFETIME_IN6:
631 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
632 		/* for sanity */
633 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
634 			ia->ia6_lifetime.ia6t_expire =
635 				time_second + ia->ia6_lifetime.ia6t_vltime;
636 		} else
637 			ia->ia6_lifetime.ia6t_expire = 0;
638 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
639 			ia->ia6_lifetime.ia6t_preferred =
640 				time_second + ia->ia6_lifetime.ia6t_pltime;
641 		} else
642 			ia->ia6_lifetime.ia6t_preferred = 0;
643 		break;
644 
645 	case SIOCAIFADDR_IN6:
646 	{
647 		int i;
648 		struct nd_prefixctl pr0;
649 		struct nd_prefix *pr;
650 
651 		/*
652 		 * first, make or update the interface address structure,
653 		 * and link it to the list.
654 		 */
655 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
656 			goto out;
657 		if (ia != NULL)
658 			ifa_free(&ia->ia_ifa);
659 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
660 		    == NULL) {
661 			/*
662 			 * this can happen when the user specify the 0 valid
663 			 * lifetime.
664 			 */
665 			break;
666 		}
667 
668 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
669 			if (carp_attach_p != NULL)
670 				error = (*carp_attach_p)(&ia->ia_ifa,
671 				    ifra->ifra_vhid);
672 			else
673 				error = EPROTONOSUPPORT;
674 			if (error)
675 				goto out;
676 			else
677 				carp_attached = 1;
678 		}
679 
680 		/*
681 		 * then, make the prefix on-link on the interface.
682 		 * XXX: we'd rather create the prefix before the address, but
683 		 * we need at least one address to install the corresponding
684 		 * interface route, so we configure the address first.
685 		 */
686 
687 		/*
688 		 * convert mask to prefix length (prefixmask has already
689 		 * been validated in in6_update_ifa().
690 		 */
691 		bzero(&pr0, sizeof(pr0));
692 		pr0.ndpr_ifp = ifp;
693 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
694 		    NULL);
695 		if (pr0.ndpr_plen == 128) {
696 			break;	/* we don't need to install a host route. */
697 		}
698 		pr0.ndpr_prefix = ifra->ifra_addr;
699 		/* apply the mask for safety. */
700 		for (i = 0; i < 4; i++) {
701 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
702 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
703 		}
704 		/*
705 		 * XXX: since we don't have an API to set prefix (not address)
706 		 * lifetimes, we just use the same lifetimes as addresses.
707 		 * The (temporarily) installed lifetimes can be overridden by
708 		 * later advertised RAs (when accept_rtadv is non 0), which is
709 		 * an intended behavior.
710 		 */
711 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
712 		pr0.ndpr_raf_auto =
713 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
714 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
715 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
716 
717 		/* add the prefix if not yet. */
718 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
719 			/*
720 			 * nd6_prelist_add will install the corresponding
721 			 * interface route.
722 			 */
723 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
724 				if (carp_attached)
725 					(*carp_detach_p)(&ia->ia_ifa);
726 				goto out;
727 			}
728 			if (pr == NULL) {
729 				if (carp_attached)
730 					(*carp_detach_p)(&ia->ia_ifa);
731 				log(LOG_ERR, "nd6_prelist_add succeeded but "
732 				    "no prefix\n");
733 				error = EINVAL;
734 				goto out;
735 			}
736 		}
737 
738 		/* relate the address to the prefix */
739 		if (ia->ia6_ndpr == NULL) {
740 			ia->ia6_ndpr = pr;
741 			pr->ndpr_refcnt++;
742 
743 			/*
744 			 * If this is the first autoconf address from the
745 			 * prefix, create a temporary address as well
746 			 * (when required).
747 			 */
748 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
749 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
750 				int e;
751 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
752 					log(LOG_NOTICE, "in6_control: failed "
753 					    "to create a temporary address, "
754 					    "errno=%d\n", e);
755 				}
756 			}
757 		}
758 
759 		/*
760 		 * this might affect the status of autoconfigured addresses,
761 		 * that is, this address might make other addresses detached.
762 		 */
763 		pfxlist_onlink_check();
764 		if (error == 0 && ia) {
765 			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
766 				/*
767 				 * Try to clear the flag when a new
768 				 * IPv6 address is added onto an
769 				 * IFDISABLED interface and it
770 				 * succeeds.
771 				 */
772 				struct in6_ndireq nd;
773 
774 				memset(&nd, 0, sizeof(nd));
775 				nd.ndi.flags = ND_IFINFO(ifp)->flags;
776 				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
777 				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
778 				    (caddr_t)&nd, ifp) < 0)
779 					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
780 					    "SIOCSIFINFO_FLAGS for -ifdisabled "
781 					    "failed.");
782 				/*
783 				 * Ignore failure of clearing the flag
784 				 * intentionally.  The failure means
785 				 * address duplication was detected.
786 				 */
787 			}
788 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
789 		}
790 		break;
791 	}
792 
793 	case SIOCDIFADDR_IN6:
794 	{
795 		struct nd_prefix *pr;
796 
797 		/*
798 		 * If the address being deleted is the only one that owns
799 		 * the corresponding prefix, expire the prefix as well.
800 		 * XXX: theoretically, we don't have to worry about such
801 		 * relationship, since we separate the address management
802 		 * and the prefix management.  We do this, however, to provide
803 		 * as much backward compatibility as possible in terms of
804 		 * the ioctl operation.
805 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
806 		 */
807 		pr = ia->ia6_ndpr;
808 		in6_purgeaddr(&ia->ia_ifa);
809 		if (pr && pr->ndpr_refcnt == 0)
810 			prelist_remove(pr);
811 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
812 		break;
813 	}
814 
815 	default:
816 		if (ifp == NULL || ifp->if_ioctl == 0) {
817 			error = EOPNOTSUPP;
818 			goto out;
819 		}
820 		error = (*ifp->if_ioctl)(ifp, cmd, data);
821 		goto out;
822 	}
823 
824 	error = 0;
825 out:
826 	if (ia != NULL)
827 		ifa_free(&ia->ia_ifa);
828 	return (error);
829 }
830 
831 
832 /*
833  * Join necessary multicast groups.  Factored out from in6_update_ifa().
834  * This entire work should only be done once, for the default FIB.
835  */
836 static int
837 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
838     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
839 {
840 	char ip6buf[INET6_ADDRSTRLEN];
841 	struct sockaddr_in6 mltaddr, mltmask;
842 	struct in6_addr llsol;
843 	struct in6_multi_mship *imm;
844 	struct rtentry *rt;
845 	int delay, error;
846 
847 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
848 
849 	/* Join solicited multicast addr for new host id. */
850 	bzero(&llsol, sizeof(struct in6_addr));
851 	llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
852 	llsol.s6_addr32[1] = 0;
853 	llsol.s6_addr32[2] = htonl(1);
854 	llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
855 	llsol.s6_addr8[12] = 0xff;
856 	if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
857 		/* XXX: should not happen */
858 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
859 		goto cleanup;
860 	}
861 	delay = 0;
862 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
863 		/*
864 		 * We need a random delay for DAD on the address being
865 		 * configured.  It also means delaying transmission of the
866 		 * corresponding MLD report to avoid report collision.
867 		 * [RFC 4861, Section 6.3.7]
868 		 */
869 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
870 	}
871 	imm = in6_joingroup(ifp, &llsol, &error, delay);
872 	if (imm == NULL) {
873 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
874 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &llsol),
875 		    if_name(ifp), error));
876 		goto cleanup;
877 	}
878 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
879 	*in6m_sol = imm->i6mm_maddr;
880 
881 	bzero(&mltmask, sizeof(mltmask));
882 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
883 	mltmask.sin6_family = AF_INET6;
884 	mltmask.sin6_addr = in6mask32;
885 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
886 
887 	/*
888 	 * Join link-local all-nodes address.
889 	 */
890 	bzero(&mltaddr, sizeof(mltaddr));
891 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
892 	mltaddr.sin6_family = AF_INET6;
893 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
894 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
895 		goto cleanup; /* XXX: should not fail */
896 
897 	/*
898 	 * XXX: do we really need this automatic routes?  We should probably
899 	 * reconsider this stuff.  Most applications actually do not need the
900 	 * routes, since they usually specify the outgoing interface.
901 	 */
902 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
903 	if (rt != NULL) {
904 		/* XXX: only works in !SCOPEDROUTING case. */
905 		if (memcmp(&mltaddr.sin6_addr,
906 		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
907 		    MLTMASK_LEN)) {
908 			RTFREE_LOCKED(rt);
909 			rt = NULL;
910 		}
911 	}
912 	if (rt == NULL) {
913 		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
914 		    (struct sockaddr *)&ia->ia_addr,
915 		    (struct sockaddr *)&mltmask, RTF_UP,
916 		    (struct rtentry **)0, RT_DEFAULT_FIB);
917 		if (error)
918 			goto cleanup;
919 	} else
920 		RTFREE_LOCKED(rt);
921 
922 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
923 	if (imm == NULL) {
924 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
925 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
926 		    &mltaddr.sin6_addr), if_name(ifp), error));
927 		goto cleanup;
928 	}
929 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
930 
931 	/*
932 	 * Join node information group address.
933 	 */
934 	delay = 0;
935 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
936 		/*
937 		 * The spec does not say anything about delay for this group,
938 		 * but the same logic should apply.
939 		 */
940 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
941 	}
942 	if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
943 		/* XXX jinmei */
944 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
945 		if (imm == NULL)
946 			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
947 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
948 			    &mltaddr.sin6_addr), if_name(ifp), error));
949 			/* XXX not very fatal, go on... */
950 		else
951 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
952 	}
953 	if (V_icmp6_nodeinfo_oldmcprefix &&
954 	     in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
955 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
956 		if (imm == NULL)
957 			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
958 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
959 			    &mltaddr.sin6_addr), if_name(ifp), error));
960 			/* XXX not very fatal, go on... */
961 		else
962 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
963 	}
964 
965 	/*
966 	 * Join interface-local all-nodes address.
967 	 * (ff01::1%ifN, and ff01::%ifN/32)
968 	 */
969 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
970 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
971 		goto cleanup; /* XXX: should not fail */
972 	/* XXX: again, do we really need the route? */
973 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
974 	if (rt != NULL) {
975 		if (memcmp(&mltaddr.sin6_addr,
976 		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
977 		    MLTMASK_LEN)) {
978 			RTFREE_LOCKED(rt);
979 			rt = NULL;
980 		}
981 	}
982 	if (rt == NULL) {
983 		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
984 		    (struct sockaddr *)&ia->ia_addr,
985 		    (struct sockaddr *)&mltmask, RTF_UP,
986 		    (struct rtentry **)0, RT_DEFAULT_FIB);
987 		if (error)
988 			goto cleanup;
989 	} else
990 		RTFREE_LOCKED(rt);
991 
992 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
993 	if (imm == NULL) {
994 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
995 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
996 		    &mltaddr.sin6_addr), if_name(ifp), error));
997 		goto cleanup;
998 	}
999 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1000 #undef	MLTMASK_LEN
1001 
1002 cleanup:
1003 	return (error);
1004 }
1005 
1006 /*
1007  * Update parameters of an IPv6 interface address.
1008  * If necessary, a new entry is created and linked into address chains.
1009  * This function is separated from in6_control().
1010  */
1011 int
1012 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1013     struct in6_ifaddr *ia, int flags)
1014 {
1015 	int error = 0, hostIsNew = 0, plen = -1;
1016 	struct sockaddr_in6 dst6;
1017 	struct in6_addrlifetime *lt;
1018 	struct in6_multi *in6m_sol;
1019 	int delay;
1020 	char ip6buf[INET6_ADDRSTRLEN];
1021 
1022 	/* Validate parameters */
1023 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1024 		return (EINVAL);
1025 
1026 	/*
1027 	 * The destination address for a p2p link must have a family
1028 	 * of AF_UNSPEC or AF_INET6.
1029 	 */
1030 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1031 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1032 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1033 		return (EAFNOSUPPORT);
1034 	/*
1035 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1036 	 * does not carry fields other than sin6_len.
1037 	 */
1038 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1039 		return (EINVAL);
1040 	/*
1041 	 * Because the IPv6 address architecture is classless, we require
1042 	 * users to specify a (non 0) prefix length (mask) for a new address.
1043 	 * We also require the prefix (when specified) mask is valid, and thus
1044 	 * reject a non-consecutive mask.
1045 	 */
1046 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1047 		return (EINVAL);
1048 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1049 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1050 		    (u_char *)&ifra->ifra_prefixmask +
1051 		    ifra->ifra_prefixmask.sin6_len);
1052 		if (plen <= 0)
1053 			return (EINVAL);
1054 	} else {
1055 		/*
1056 		 * In this case, ia must not be NULL.  We just use its prefix
1057 		 * length.
1058 		 */
1059 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1060 	}
1061 	/*
1062 	 * If the destination address on a p2p interface is specified,
1063 	 * and the address is a scoped one, validate/set the scope
1064 	 * zone identifier.
1065 	 */
1066 	dst6 = ifra->ifra_dstaddr;
1067 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1068 	    (dst6.sin6_family == AF_INET6)) {
1069 		struct in6_addr in6_tmp;
1070 		u_int32_t zoneid;
1071 
1072 		in6_tmp = dst6.sin6_addr;
1073 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1074 			return (EINVAL); /* XXX: should be impossible */
1075 
1076 		if (dst6.sin6_scope_id != 0) {
1077 			if (dst6.sin6_scope_id != zoneid)
1078 				return (EINVAL);
1079 		} else		/* user omit to specify the ID. */
1080 			dst6.sin6_scope_id = zoneid;
1081 
1082 		/* convert into the internal form */
1083 		if (sa6_embedscope(&dst6, 0))
1084 			return (EINVAL); /* XXX: should be impossible */
1085 	}
1086 	/*
1087 	 * The destination address can be specified only for a p2p or a
1088 	 * loopback interface.  If specified, the corresponding prefix length
1089 	 * must be 128.
1090 	 */
1091 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1092 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1093 			/* XXX: noisy message */
1094 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1095 			    "be specified for a p2p or a loopback IF only\n"));
1096 			return (EINVAL);
1097 		}
1098 		if (plen != 128) {
1099 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1100 			    "be 128 when dstaddr is specified\n"));
1101 			return (EINVAL);
1102 		}
1103 	}
1104 	/* lifetime consistency check */
1105 	lt = &ifra->ifra_lifetime;
1106 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1107 		return (EINVAL);
1108 	if (lt->ia6t_vltime == 0) {
1109 		/*
1110 		 * the following log might be noisy, but this is a typical
1111 		 * configuration mistake or a tool's bug.
1112 		 */
1113 		nd6log((LOG_INFO,
1114 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1115 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1116 
1117 		if (ia == NULL)
1118 			return (0); /* there's nothing to do */
1119 	}
1120 
1121 	/*
1122 	 * If this is a new address, allocate a new ifaddr and link it
1123 	 * into chains.
1124 	 */
1125 	if (ia == NULL) {
1126 		hostIsNew = 1;
1127 		/*
1128 		 * When in6_update_ifa() is called in a process of a received
1129 		 * RA, it is called under an interrupt context.  So, we should
1130 		 * call malloc with M_NOWAIT.
1131 		 */
1132 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
1133 		    M_NOWAIT);
1134 		if (ia == NULL)
1135 			return (ENOBUFS);
1136 		bzero((caddr_t)ia, sizeof(*ia));
1137 		ifa_init(&ia->ia_ifa);
1138 		LIST_INIT(&ia->ia6_memberships);
1139 		/* Initialize the address and masks, and put time stamp */
1140 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1141 		ia->ia_addr.sin6_family = AF_INET6;
1142 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1143 		ia->ia6_createtime = time_second;
1144 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1145 			/*
1146 			 * XXX: some functions expect that ifa_dstaddr is not
1147 			 * NULL for p2p interfaces.
1148 			 */
1149 			ia->ia_ifa.ifa_dstaddr =
1150 			    (struct sockaddr *)&ia->ia_dstaddr;
1151 		} else {
1152 			ia->ia_ifa.ifa_dstaddr = NULL;
1153 		}
1154 		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1155 		ia->ia_ifp = ifp;
1156 		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1157 		IF_ADDR_WLOCK(ifp);
1158 		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1159 		IF_ADDR_WUNLOCK(ifp);
1160 
1161 		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1162 		IN6_IFADDR_WLOCK();
1163 		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1164 		LIST_INSERT_HEAD(IN6ADDR_HASH(&ifra->ifra_addr.sin6_addr),
1165 		    ia, ia6_hash);
1166 		IN6_IFADDR_WUNLOCK();
1167 	}
1168 
1169 	/* update timestamp */
1170 	ia->ia6_updatetime = time_second;
1171 
1172 	/* set prefix mask */
1173 	if (ifra->ifra_prefixmask.sin6_len) {
1174 		/*
1175 		 * We prohibit changing the prefix length of an existing
1176 		 * address, because
1177 		 * + such an operation should be rare in IPv6, and
1178 		 * + the operation would confuse prefix management.
1179 		 */
1180 		if (ia->ia_prefixmask.sin6_len &&
1181 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1182 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1183 			    " existing (%s) address should not be changed\n",
1184 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1185 			error = EINVAL;
1186 			goto unlink;
1187 		}
1188 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1189 	}
1190 
1191 	/*
1192 	 * If a new destination address is specified, scrub the old one and
1193 	 * install the new destination.  Note that the interface must be
1194 	 * p2p or loopback (see the check above.)
1195 	 */
1196 	if (dst6.sin6_family == AF_INET6 &&
1197 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1198 		int e;
1199 
1200 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1201 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1202 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1203 			    "a route to the old destination: %s\n",
1204 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1205 			/* proceed anyway... */
1206 		} else
1207 			ia->ia_flags &= ~IFA_ROUTE;
1208 		ia->ia_dstaddr = dst6;
1209 	}
1210 
1211 	/*
1212 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1213 	 * to see if the address is deprecated or invalidated, but initialize
1214 	 * these members for applications.
1215 	 */
1216 	ia->ia6_lifetime = ifra->ifra_lifetime;
1217 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1218 		ia->ia6_lifetime.ia6t_expire =
1219 		    time_second + ia->ia6_lifetime.ia6t_vltime;
1220 	} else
1221 		ia->ia6_lifetime.ia6t_expire = 0;
1222 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1223 		ia->ia6_lifetime.ia6t_preferred =
1224 		    time_second + ia->ia6_lifetime.ia6t_pltime;
1225 	} else
1226 		ia->ia6_lifetime.ia6t_preferred = 0;
1227 
1228 	/* reset the interface and routing table appropriately. */
1229 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1230 		goto unlink;
1231 
1232 	/*
1233 	 * configure address flags.
1234 	 */
1235 	ia->ia6_flags = ifra->ifra_flags;
1236 	/*
1237 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1238 	 * userland, make it deprecated.
1239 	 */
1240 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1241 		ia->ia6_lifetime.ia6t_pltime = 0;
1242 		ia->ia6_lifetime.ia6t_preferred = time_second;
1243 	}
1244 	/*
1245 	 * Make the address tentative before joining multicast addresses,
1246 	 * so that corresponding MLD responses would not have a tentative
1247 	 * source address.
1248 	 */
1249 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1250 	if (hostIsNew && in6if_do_dad(ifp))
1251 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1252 
1253 	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
1254 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1255 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1256 
1257 	/*
1258 	 * We are done if we have simply modified an existing address.
1259 	 */
1260 	if (!hostIsNew)
1261 		return (error);
1262 
1263 	/*
1264 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1265 	 * not just go to unlink.
1266 	 */
1267 
1268 	/* Join necessary multicast groups. */
1269 	in6m_sol = NULL;
1270 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1271 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1272 		if (error)
1273 			goto cleanup;
1274 	}
1275 
1276 	/*
1277 	 * Perform DAD, if needed.
1278 	 * XXX It may be of use, if we can administratively disable DAD.
1279 	 */
1280 	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1281 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1282 	{
1283 		int mindelay, maxdelay;
1284 
1285 		delay = 0;
1286 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1287 			/*
1288 			 * We need to impose a delay before sending an NS
1289 			 * for DAD.  Check if we also needed a delay for the
1290 			 * corresponding MLD message.  If we did, the delay
1291 			 * should be larger than the MLD delay (this could be
1292 			 * relaxed a bit, but this simple logic is at least
1293 			 * safe).
1294 			 * XXX: Break data hiding guidelines and look at
1295 			 * state for the solicited multicast group.
1296 			 */
1297 			mindelay = 0;
1298 			if (in6m_sol != NULL &&
1299 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1300 				mindelay = in6m_sol->in6m_timer;
1301 			}
1302 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1303 			if (maxdelay - mindelay == 0)
1304 				delay = 0;
1305 			else {
1306 				delay =
1307 				    (arc4random() % (maxdelay - mindelay)) +
1308 				    mindelay;
1309 			}
1310 		}
1311 		nd6_dad_start((struct ifaddr *)ia, delay);
1312 	}
1313 
1314 	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1315 	ifa_free(&ia->ia_ifa);
1316 	return (error);
1317 
1318   unlink:
1319 	/*
1320 	 * XXX: if a change of an existing address failed, keep the entry
1321 	 * anyway.
1322 	 */
1323 	if (hostIsNew) {
1324 		in6_unlink_ifa(ia, ifp);
1325 		ifa_free(&ia->ia_ifa);
1326 	}
1327 	return (error);
1328 
1329   cleanup:
1330 	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1331 	ifa_free(&ia->ia_ifa);
1332 	in6_purgeaddr(&ia->ia_ifa);
1333 	return error;
1334 }
1335 
1336 /*
1337  * Leave multicast groups.  Factored out from in6_purgeaddr().
1338  * This entire work should only be done once, for the default FIB.
1339  */
1340 static int
1341 in6_purgeaddr_mc(struct ifnet *ifp, struct in6_ifaddr *ia, struct ifaddr *ifa0)
1342 {
1343 	struct sockaddr_in6 mltaddr, mltmask;
1344 	struct in6_multi_mship *imm;
1345 	struct rtentry *rt;
1346 	struct sockaddr_in6 sin6;
1347 	int error;
1348 
1349 	/*
1350 	 * Leave from multicast groups we have joined for the interface.
1351 	 */
1352 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1353 		LIST_REMOVE(imm, i6mm_chain);
1354 		in6_leavegroup(imm);
1355 	}
1356 
1357 	/*
1358 	 * Remove the link-local all-nodes address.
1359 	 */
1360 	bzero(&mltmask, sizeof(mltmask));
1361 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1362 	mltmask.sin6_family = AF_INET6;
1363 	mltmask.sin6_addr = in6mask32;
1364 
1365 	bzero(&mltaddr, sizeof(mltaddr));
1366 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1367 	mltaddr.sin6_family = AF_INET6;
1368 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1369 
1370 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1371 		return (error);
1372 
1373 	/*
1374 	 * As for the mltaddr above, proactively prepare the sin6 to avoid
1375 	 * rtentry un- and re-locking.
1376 	 */
1377 	if (ifa0 != NULL) {
1378 		bzero(&sin6, sizeof(sin6));
1379 		sin6.sin6_len = sizeof(sin6);
1380 		sin6.sin6_family = AF_INET6;
1381 		memcpy(&sin6.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1382 		    sizeof(sin6.sin6_addr));
1383 		error = in6_setscope(&sin6.sin6_addr, ifa0->ifa_ifp, NULL);
1384 		if (error != 0)
1385 			return (error);
1386 	}
1387 
1388 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1389 	if (rt != NULL && rt->rt_gateway != NULL &&
1390 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1391 		    &ia->ia_addr.sin6_addr,
1392 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1393 		/*
1394 		 * If no more IPv6 address exists on this interface then
1395 		 * remove the multicast address route.
1396 		 */
1397 		if (ifa0 == NULL) {
1398 			memcpy(&mltaddr.sin6_addr,
1399 			    &satosin6(rt_key(rt))->sin6_addr,
1400 			    sizeof(mltaddr.sin6_addr));
1401 			RTFREE_LOCKED(rt);
1402 			error = in6_rtrequest(RTM_DELETE,
1403 			    (struct sockaddr *)&mltaddr,
1404 			    (struct sockaddr *)&ia->ia_addr,
1405 			    (struct sockaddr *)&mltmask, RTF_UP,
1406 			    (struct rtentry **)0, RT_DEFAULT_FIB);
1407 			if (error)
1408 				log(LOG_INFO, "%s: link-local all-nodes "
1409 				    "multicast address deletion error\n",
1410 				    __func__);
1411 		} else {
1412 			/*
1413 			 * Replace the gateway of the route.
1414 			 */
1415 			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1416 			RTFREE_LOCKED(rt);
1417 		}
1418 	} else {
1419 		if (rt != NULL)
1420 			RTFREE_LOCKED(rt);
1421 	}
1422 
1423 	/*
1424 	 * Remove the node-local all-nodes address.
1425 	 */
1426 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1427 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1428 		return (error);
1429 
1430 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1431 	if (rt != NULL && rt->rt_gateway != NULL &&
1432 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1433 		    &ia->ia_addr.sin6_addr,
1434 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1435 		/*
1436 		 * If no more IPv6 address exists on this interface then
1437 		 * remove the multicast address route.
1438 		 */
1439 		if (ifa0 == NULL) {
1440 			memcpy(&mltaddr.sin6_addr,
1441 			    &satosin6(rt_key(rt))->sin6_addr,
1442 			    sizeof(mltaddr.sin6_addr));
1443 
1444 			RTFREE_LOCKED(rt);
1445 			error = in6_rtrequest(RTM_DELETE,
1446 			    (struct sockaddr *)&mltaddr,
1447 			    (struct sockaddr *)&ia->ia_addr,
1448 			    (struct sockaddr *)&mltmask, RTF_UP,
1449 			    (struct rtentry **)0, RT_DEFAULT_FIB);
1450 			if (error)
1451 				log(LOG_INFO, "%s: node-local all-nodes"
1452 				    "multicast address deletion error\n",
1453 				    __func__);
1454 		} else {
1455 			/*
1456 			 * Replace the gateway of the route.
1457 			 */
1458 			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1459 			RTFREE_LOCKED(rt);
1460 		}
1461 	} else {
1462 		if (rt != NULL)
1463 			RTFREE_LOCKED(rt);
1464 	}
1465 
1466 	return (0);
1467 }
1468 
1469 void
1470 in6_purgeaddr(struct ifaddr *ifa)
1471 {
1472 	struct ifnet *ifp = ifa->ifa_ifp;
1473 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1474 	int plen, error;
1475 	struct ifaddr *ifa0;
1476 
1477 	if (ifa->ifa_carp)
1478 		(*carp_detach_p)(ifa);
1479 
1480 	/*
1481 	 * find another IPv6 address as the gateway for the
1482 	 * link-local and node-local all-nodes multicast
1483 	 * address routes
1484 	 */
1485 	IF_ADDR_RLOCK(ifp);
1486 	TAILQ_FOREACH(ifa0, &ifp->if_addrhead, ifa_link) {
1487 		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1488 		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1489 		    &ia->ia_addr.sin6_addr, sizeof(struct in6_addr)) == 0)
1490 			continue;
1491 		else
1492 			break;
1493 	}
1494 	if (ifa0 != NULL)
1495 		ifa_ref(ifa0);
1496 	IF_ADDR_RUNLOCK(ifp);
1497 
1498 	/*
1499 	 * Remove the loopback route to the interface address.
1500 	 * The check for the current setting of "nd6_useloopback"
1501 	 * is not needed.
1502 	 */
1503 	if (ia->ia_flags & IFA_RTSELF) {
1504 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1505 		    (struct sockaddr *)&ia->ia_addr);
1506 		if (error == 0)
1507 			ia->ia_flags &= ~IFA_RTSELF;
1508 	}
1509 
1510 	/* stop DAD processing */
1511 	nd6_dad_stop(ifa);
1512 
1513 	/* Remove local address entry from lltable. */
1514 	in6_ifremloop(ifa);
1515 
1516 	/* Leave multicast groups. */
1517 	error = in6_purgeaddr_mc(ifp, ia, ifa0);
1518 
1519 	if (ifa0 != NULL)
1520 		ifa_free(ifa0);
1521 
1522 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1523 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1524 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1525 		    (ia->ia_dstaddr.sin6_family == AF_INET6) ? RTF_HOST : 0);
1526 		if (error != 0)
1527 			log(LOG_INFO, "%s: err=%d, destination address delete "
1528 			    "failed\n", __func__, error);
1529 		ia->ia_flags &= ~IFA_ROUTE;
1530 	}
1531 
1532 	in6_unlink_ifa(ia, ifp);
1533 }
1534 
1535 static void
1536 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1537 {
1538 
1539 	IF_ADDR_WLOCK(ifp);
1540 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1541 	IF_ADDR_WUNLOCK(ifp);
1542 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1543 
1544 	/*
1545 	 * Defer the release of what might be the last reference to the
1546 	 * in6_ifaddr so that it can't be freed before the remainder of the
1547 	 * cleanup.
1548 	 */
1549 	IN6_IFADDR_WLOCK();
1550 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1551 	LIST_REMOVE(ia, ia6_hash);
1552 	IN6_IFADDR_WUNLOCK();
1553 
1554 	/*
1555 	 * Release the reference to the base prefix.  There should be a
1556 	 * positive reference.
1557 	 */
1558 	if (ia->ia6_ndpr == NULL) {
1559 		nd6log((LOG_NOTICE,
1560 		    "in6_unlink_ifa: autoconf'ed address "
1561 		    "%p has no prefix\n", ia));
1562 	} else {
1563 		ia->ia6_ndpr->ndpr_refcnt--;
1564 		ia->ia6_ndpr = NULL;
1565 	}
1566 
1567 	/*
1568 	 * Also, if the address being removed is autoconf'ed, call
1569 	 * pfxlist_onlink_check() since the release might affect the status of
1570 	 * other (detached) addresses.
1571 	 */
1572 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1573 		pfxlist_onlink_check();
1574 	}
1575 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1576 }
1577 
1578 void
1579 in6_purgeif(struct ifnet *ifp)
1580 {
1581 	struct ifaddr *ifa, *nifa;
1582 
1583 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1584 		if (ifa->ifa_addr->sa_family != AF_INET6)
1585 			continue;
1586 		in6_purgeaddr(ifa);
1587 	}
1588 
1589 	in6_ifdetach(ifp);
1590 }
1591 
1592 /*
1593  * SIOC[GAD]LIFADDR.
1594  *	SIOCGLIFADDR: get first address. (?)
1595  *	SIOCGLIFADDR with IFLR_PREFIX:
1596  *		get first address that matches the specified prefix.
1597  *	SIOCALIFADDR: add the specified address.
1598  *	SIOCALIFADDR with IFLR_PREFIX:
1599  *		add the specified prefix, filling hostid part from
1600  *		the first link-local address.  prefixlen must be <= 64.
1601  *	SIOCDLIFADDR: delete the specified address.
1602  *	SIOCDLIFADDR with IFLR_PREFIX:
1603  *		delete the first address that matches the specified prefix.
1604  * return values:
1605  *	EINVAL on invalid parameters
1606  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1607  *	other values may be returned from in6_ioctl()
1608  *
1609  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1610  * this is to accomodate address naming scheme other than RFC2374,
1611  * in the future.
1612  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1613  * address encoding scheme. (see figure on page 8)
1614  */
1615 static int
1616 in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1617     struct ifnet *ifp, struct thread *td)
1618 {
1619 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1620 	struct ifaddr *ifa;
1621 	struct sockaddr *sa;
1622 
1623 	/* sanity checks */
1624 	if (!data || !ifp) {
1625 		panic("invalid argument to in6_lifaddr_ioctl");
1626 		/* NOTREACHED */
1627 	}
1628 
1629 	switch (cmd) {
1630 	case SIOCGLIFADDR:
1631 		/* address must be specified on GET with IFLR_PREFIX */
1632 		if ((iflr->flags & IFLR_PREFIX) == 0)
1633 			break;
1634 		/* FALLTHROUGH */
1635 	case SIOCALIFADDR:
1636 	case SIOCDLIFADDR:
1637 		/* address must be specified on ADD and DELETE */
1638 		sa = (struct sockaddr *)&iflr->addr;
1639 		if (sa->sa_family != AF_INET6)
1640 			return EINVAL;
1641 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1642 			return EINVAL;
1643 		/* XXX need improvement */
1644 		sa = (struct sockaddr *)&iflr->dstaddr;
1645 		if (sa->sa_family && sa->sa_family != AF_INET6)
1646 			return EINVAL;
1647 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1648 			return EINVAL;
1649 		break;
1650 	default: /* shouldn't happen */
1651 #if 0
1652 		panic("invalid cmd to in6_lifaddr_ioctl");
1653 		/* NOTREACHED */
1654 #else
1655 		return EOPNOTSUPP;
1656 #endif
1657 	}
1658 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1659 		return EINVAL;
1660 
1661 	switch (cmd) {
1662 	case SIOCALIFADDR:
1663 	    {
1664 		struct in6_aliasreq ifra;
1665 		struct in6_addr *hostid = NULL;
1666 		int prefixlen;
1667 
1668 		ifa = NULL;
1669 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1670 			struct sockaddr_in6 *sin6;
1671 
1672 			/*
1673 			 * hostid is to fill in the hostid part of the
1674 			 * address.  hostid points to the first link-local
1675 			 * address attached to the interface.
1676 			 */
1677 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1678 			if (!ifa)
1679 				return EADDRNOTAVAIL;
1680 			hostid = IFA_IN6(ifa);
1681 
1682 			/* prefixlen must be <= 64. */
1683 			if (64 < iflr->prefixlen) {
1684 				if (ifa != NULL)
1685 					ifa_free(ifa);
1686 				return EINVAL;
1687 			}
1688 			prefixlen = iflr->prefixlen;
1689 
1690 			/* hostid part must be zero. */
1691 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1692 			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1693 			    sin6->sin6_addr.s6_addr32[3] != 0) {
1694 				if (ifa != NULL)
1695 					ifa_free(ifa);
1696 				return EINVAL;
1697 			}
1698 		} else
1699 			prefixlen = iflr->prefixlen;
1700 
1701 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1702 		bzero(&ifra, sizeof(ifra));
1703 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1704 
1705 		bcopy(&iflr->addr, &ifra.ifra_addr,
1706 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1707 		if (hostid) {
1708 			/* fill in hostid part */
1709 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1710 			    hostid->s6_addr32[2];
1711 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1712 			    hostid->s6_addr32[3];
1713 		}
1714 
1715 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1716 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1717 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1718 			if (hostid) {
1719 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1720 				    hostid->s6_addr32[2];
1721 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1722 				    hostid->s6_addr32[3];
1723 			}
1724 		}
1725 		if (ifa != NULL)
1726 			ifa_free(ifa);
1727 
1728 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1729 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1730 
1731 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1732 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1733 	    }
1734 	case SIOCGLIFADDR:
1735 	case SIOCDLIFADDR:
1736 	    {
1737 		struct in6_ifaddr *ia;
1738 		struct in6_addr mask, candidate, match;
1739 		struct sockaddr_in6 *sin6;
1740 		int cmp;
1741 
1742 		bzero(&mask, sizeof(mask));
1743 		if (iflr->flags & IFLR_PREFIX) {
1744 			/* lookup a prefix rather than address. */
1745 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1746 
1747 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1748 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1749 			match.s6_addr32[0] &= mask.s6_addr32[0];
1750 			match.s6_addr32[1] &= mask.s6_addr32[1];
1751 			match.s6_addr32[2] &= mask.s6_addr32[2];
1752 			match.s6_addr32[3] &= mask.s6_addr32[3];
1753 
1754 			/* if you set extra bits, that's wrong */
1755 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1756 				return EINVAL;
1757 
1758 			cmp = 1;
1759 		} else {
1760 			if (cmd == SIOCGLIFADDR) {
1761 				/* on getting an address, take the 1st match */
1762 				cmp = 0;	/* XXX */
1763 			} else {
1764 				/* on deleting an address, do exact match */
1765 				in6_prefixlen2mask(&mask, 128);
1766 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1767 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1768 
1769 				cmp = 1;
1770 			}
1771 		}
1772 
1773 		IF_ADDR_RLOCK(ifp);
1774 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1775 			if (ifa->ifa_addr->sa_family != AF_INET6)
1776 				continue;
1777 			if (!cmp)
1778 				break;
1779 
1780 			/*
1781 			 * XXX: this is adhoc, but is necessary to allow
1782 			 * a user to specify fe80::/64 (not /10) for a
1783 			 * link-local address.
1784 			 */
1785 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1786 			in6_clearscope(&candidate);
1787 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1788 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1789 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1790 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1791 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1792 				break;
1793 		}
1794 		if (ifa != NULL)
1795 			ifa_ref(ifa);
1796 		IF_ADDR_RUNLOCK(ifp);
1797 		if (!ifa)
1798 			return EADDRNOTAVAIL;
1799 		ia = ifa2ia6(ifa);
1800 
1801 		if (cmd == SIOCGLIFADDR) {
1802 			int error;
1803 
1804 			/* fill in the if_laddrreq structure */
1805 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1806 			error = sa6_recoverscope(
1807 			    (struct sockaddr_in6 *)&iflr->addr);
1808 			if (error != 0) {
1809 				ifa_free(ifa);
1810 				return (error);
1811 			}
1812 
1813 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1814 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1815 				    ia->ia_dstaddr.sin6_len);
1816 				error = sa6_recoverscope(
1817 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1818 				if (error != 0) {
1819 					ifa_free(ifa);
1820 					return (error);
1821 				}
1822 			} else
1823 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1824 
1825 			iflr->prefixlen =
1826 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1827 
1828 			iflr->flags = ia->ia6_flags;	/* XXX */
1829 			ifa_free(ifa);
1830 
1831 			return 0;
1832 		} else {
1833 			struct in6_aliasreq ifra;
1834 
1835 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1836 			bzero(&ifra, sizeof(ifra));
1837 			bcopy(iflr->iflr_name, ifra.ifra_name,
1838 			    sizeof(ifra.ifra_name));
1839 
1840 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1841 			    ia->ia_addr.sin6_len);
1842 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1843 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1844 				    ia->ia_dstaddr.sin6_len);
1845 			} else {
1846 				bzero(&ifra.ifra_dstaddr,
1847 				    sizeof(ifra.ifra_dstaddr));
1848 			}
1849 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1850 			    ia->ia_prefixmask.sin6_len);
1851 
1852 			ifra.ifra_flags = ia->ia6_flags;
1853 			ifa_free(ifa);
1854 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1855 			    ifp, td);
1856 		}
1857 	    }
1858 	}
1859 
1860 	return EOPNOTSUPP;	/* just for safety */
1861 }
1862 
1863 /*
1864  * Initialize an interface's IPv6 address and routing table entry.
1865  */
1866 static int
1867 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1868     struct sockaddr_in6 *sin6, int newhost)
1869 {
1870 	int	error = 0, plen, ifacount = 0;
1871 	struct ifaddr *ifa;
1872 
1873 	/*
1874 	 * Give the interface a chance to initialize
1875 	 * if this is its first address,
1876 	 * and to validate the address if necessary.
1877 	 */
1878 	IF_ADDR_RLOCK(ifp);
1879 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1880 		if (ifa->ifa_addr->sa_family != AF_INET6)
1881 			continue;
1882 		ifacount++;
1883 	}
1884 	IF_ADDR_RUNLOCK(ifp);
1885 
1886 	ia->ia_addr = *sin6;
1887 
1888 	if (ifacount <= 1 && ifp->if_ioctl) {
1889 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1890 		if (error)
1891 			return (error);
1892 	}
1893 
1894 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1895 
1896 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1897 
1898 	/*
1899 	 * Special case:
1900 	 * If a new destination address is specified for a point-to-point
1901 	 * interface, install a route to the destination as an interface
1902 	 * direct route.
1903 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1904 	 * interface that share the same destination.
1905 	 */
1906 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1907 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1908 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1909 		int rtflags = RTF_UP | RTF_HOST;
1910 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1911 		if (error)
1912 			return (error);
1913 		ia->ia_flags |= IFA_ROUTE;
1914 		/*
1915 		 * Handle the case for ::1 .
1916 		 */
1917 		if (ifp->if_flags & IFF_LOOPBACK)
1918 			ia->ia_flags |= IFA_RTSELF;
1919 	}
1920 
1921 	/*
1922 	 * add a loopback route to self
1923 	 */
1924 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1925 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1926 		    (struct sockaddr *)&ia->ia_addr);
1927 		if (error == 0)
1928 			ia->ia_flags |= IFA_RTSELF;
1929 	}
1930 
1931 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1932 	if (newhost)
1933 		in6_ifaddloop(&(ia->ia_ifa));
1934 
1935 	return (error);
1936 }
1937 
1938 /*
1939  * Find an IPv6 interface link-local address specific to an interface.
1940  * ifaddr is returned referenced.
1941  */
1942 struct in6_ifaddr *
1943 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1944 {
1945 	struct ifaddr *ifa;
1946 
1947 	IF_ADDR_RLOCK(ifp);
1948 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1949 		if (ifa->ifa_addr->sa_family != AF_INET6)
1950 			continue;
1951 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1952 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1953 			    ignoreflags) != 0)
1954 				continue;
1955 			ifa_ref(ifa);
1956 			break;
1957 		}
1958 	}
1959 	IF_ADDR_RUNLOCK(ifp);
1960 
1961 	return ((struct in6_ifaddr *)ifa);
1962 }
1963 
1964 
1965 /*
1966  * find the internet address corresponding to a given interface and address.
1967  * ifaddr is returned referenced.
1968  */
1969 struct in6_ifaddr *
1970 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1971 {
1972 	struct ifaddr *ifa;
1973 
1974 	IF_ADDR_RLOCK(ifp);
1975 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1976 		if (ifa->ifa_addr->sa_family != AF_INET6)
1977 			continue;
1978 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1979 			ifa_ref(ifa);
1980 			break;
1981 		}
1982 	}
1983 	IF_ADDR_RUNLOCK(ifp);
1984 
1985 	return ((struct in6_ifaddr *)ifa);
1986 }
1987 
1988 /*
1989  * Convert IP6 address to printable (loggable) representation. Caller
1990  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1991  */
1992 static char digits[] = "0123456789abcdef";
1993 char *
1994 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1995 {
1996 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1997 	char *cp;
1998 	const u_int16_t *a = (const u_int16_t *)addr;
1999 	const u_int8_t *d;
2000 	int dcolon = 0, zero = 0;
2001 
2002 	cp = ip6buf;
2003 
2004 	for (i = 0; i < 8; i++) {
2005 		if (*(a + i) == 0) {
2006 			cnt++;
2007 			if (cnt == 1)
2008 				idx = i;
2009 		}
2010 		else if (maxcnt < cnt) {
2011 			maxcnt = cnt;
2012 			index = idx;
2013 			cnt = 0;
2014 		}
2015 	}
2016 	if (maxcnt < cnt) {
2017 		maxcnt = cnt;
2018 		index = idx;
2019 	}
2020 
2021 	for (i = 0; i < 8; i++) {
2022 		if (dcolon == 1) {
2023 			if (*a == 0) {
2024 				if (i == 7)
2025 					*cp++ = ':';
2026 				a++;
2027 				continue;
2028 			} else
2029 				dcolon = 2;
2030 		}
2031 		if (*a == 0) {
2032 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
2033 				if (i == 0)
2034 					*cp++ = ':';
2035 				*cp++ = ':';
2036 				dcolon = 1;
2037 			} else {
2038 				*cp++ = '0';
2039 				*cp++ = ':';
2040 			}
2041 			a++;
2042 			continue;
2043 		}
2044 		d = (const u_char *)a;
2045 		/* Try to eliminate leading zeros in printout like in :0001. */
2046 		zero = 1;
2047 		*cp = digits[*d >> 4];
2048 		if (*cp != '0') {
2049 			zero = 0;
2050 			cp++;
2051 		}
2052 		*cp = digits[*d++ & 0xf];
2053 		if (zero == 0 || (*cp != '0')) {
2054 			zero = 0;
2055 			cp++;
2056 		}
2057 		*cp = digits[*d >> 4];
2058 		if (zero == 0 || (*cp != '0')) {
2059 			zero = 0;
2060 			cp++;
2061 		}
2062 		*cp++ = digits[*d & 0xf];
2063 		*cp++ = ':';
2064 		a++;
2065 	}
2066 	*--cp = '\0';
2067 	return (ip6buf);
2068 }
2069 
2070 int
2071 in6_localaddr(struct in6_addr *in6)
2072 {
2073 	struct in6_ifaddr *ia;
2074 
2075 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2076 		return 1;
2077 
2078 	IN6_IFADDR_RLOCK();
2079 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2080 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2081 		    &ia->ia_prefixmask.sin6_addr)) {
2082 			IN6_IFADDR_RUNLOCK();
2083 			return 1;
2084 		}
2085 	}
2086 	IN6_IFADDR_RUNLOCK();
2087 
2088 	return (0);
2089 }
2090 
2091 /*
2092  * Return 1 if an internet address is for the local host and configured
2093  * on one of its interfaces.
2094  */
2095 int
2096 in6_localip(struct in6_addr *in6)
2097 {
2098 	struct in6_ifaddr *ia;
2099 
2100 	IN6_IFADDR_RLOCK();
2101 	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
2102 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2103 			IN6_IFADDR_RUNLOCK();
2104 			return (1);
2105 		}
2106 	}
2107 	IN6_IFADDR_RUNLOCK();
2108 	return (0);
2109 }
2110 
2111 int
2112 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2113 {
2114 	struct in6_ifaddr *ia;
2115 
2116 	IN6_IFADDR_RLOCK();
2117 	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
2118 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
2119 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2120 				IN6_IFADDR_RUNLOCK();
2121 				return (1); /* true */
2122 			}
2123 			break;
2124 		}
2125 	}
2126 	IN6_IFADDR_RUNLOCK();
2127 
2128 	return (0);		/* false */
2129 }
2130 
2131 /*
2132  * return length of part which dst and src are equal
2133  * hard coding...
2134  */
2135 int
2136 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2137 {
2138 	int match = 0;
2139 	u_char *s = (u_char *)src, *d = (u_char *)dst;
2140 	u_char *lim = s + 16, r;
2141 
2142 	while (s < lim)
2143 		if ((r = (*d++ ^ *s++)) != 0) {
2144 			while (r < 128) {
2145 				match++;
2146 				r <<= 1;
2147 			}
2148 			break;
2149 		} else
2150 			match += 8;
2151 	return match;
2152 }
2153 
2154 /* XXX: to be scope conscious */
2155 int
2156 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2157 {
2158 	int bytelen, bitlen;
2159 
2160 	/* sanity check */
2161 	if (0 > len || len > 128) {
2162 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2163 		    len);
2164 		return (0);
2165 	}
2166 
2167 	bytelen = len / 8;
2168 	bitlen = len % 8;
2169 
2170 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2171 		return (0);
2172 	if (bitlen != 0 &&
2173 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2174 	    p2->s6_addr[bytelen] >> (8 - bitlen))
2175 		return (0);
2176 
2177 	return (1);
2178 }
2179 
2180 void
2181 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2182 {
2183 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2184 	int bytelen, bitlen, i;
2185 
2186 	/* sanity check */
2187 	if (0 > len || len > 128) {
2188 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2189 		    len);
2190 		return;
2191 	}
2192 
2193 	bzero(maskp, sizeof(*maskp));
2194 	bytelen = len / 8;
2195 	bitlen = len % 8;
2196 	for (i = 0; i < bytelen; i++)
2197 		maskp->s6_addr[i] = 0xff;
2198 	if (bitlen)
2199 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2200 }
2201 
2202 /*
2203  * return the best address out of the same scope. if no address was
2204  * found, return the first valid address from designated IF.
2205  */
2206 struct in6_ifaddr *
2207 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2208 {
2209 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2210 	struct ifaddr *ifa;
2211 	struct in6_ifaddr *besta = 0;
2212 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2213 
2214 	dep[0] = dep[1] = NULL;
2215 
2216 	/*
2217 	 * We first look for addresses in the same scope.
2218 	 * If there is one, return it.
2219 	 * If two or more, return one which matches the dst longest.
2220 	 * If none, return one of global addresses assigned other ifs.
2221 	 */
2222 	IF_ADDR_RLOCK(ifp);
2223 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2224 		if (ifa->ifa_addr->sa_family != AF_INET6)
2225 			continue;
2226 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2227 			continue; /* XXX: is there any case to allow anycast? */
2228 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2229 			continue; /* don't use this interface */
2230 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2231 			continue;
2232 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2233 			if (V_ip6_use_deprecated)
2234 				dep[0] = (struct in6_ifaddr *)ifa;
2235 			continue;
2236 		}
2237 
2238 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2239 			/*
2240 			 * call in6_matchlen() as few as possible
2241 			 */
2242 			if (besta) {
2243 				if (blen == -1)
2244 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2245 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2246 				if (tlen > blen) {
2247 					blen = tlen;
2248 					besta = (struct in6_ifaddr *)ifa;
2249 				}
2250 			} else
2251 				besta = (struct in6_ifaddr *)ifa;
2252 		}
2253 	}
2254 	if (besta) {
2255 		ifa_ref(&besta->ia_ifa);
2256 		IF_ADDR_RUNLOCK(ifp);
2257 		return (besta);
2258 	}
2259 
2260 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2261 		if (ifa->ifa_addr->sa_family != AF_INET6)
2262 			continue;
2263 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2264 			continue; /* XXX: is there any case to allow anycast? */
2265 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2266 			continue; /* don't use this interface */
2267 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2268 			continue;
2269 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2270 			if (V_ip6_use_deprecated)
2271 				dep[1] = (struct in6_ifaddr *)ifa;
2272 			continue;
2273 		}
2274 
2275 		if (ifa != NULL)
2276 			ifa_ref(ifa);
2277 		IF_ADDR_RUNLOCK(ifp);
2278 		return (struct in6_ifaddr *)ifa;
2279 	}
2280 
2281 	/* use the last-resort values, that are, deprecated addresses */
2282 	if (dep[0]) {
2283 		ifa_ref((struct ifaddr *)dep[0]);
2284 		IF_ADDR_RUNLOCK(ifp);
2285 		return dep[0];
2286 	}
2287 	if (dep[1]) {
2288 		ifa_ref((struct ifaddr *)dep[1]);
2289 		IF_ADDR_RUNLOCK(ifp);
2290 		return dep[1];
2291 	}
2292 
2293 	IF_ADDR_RUNLOCK(ifp);
2294 	return NULL;
2295 }
2296 
2297 /*
2298  * perform DAD when interface becomes IFF_UP.
2299  */
2300 void
2301 in6_if_up(struct ifnet *ifp)
2302 {
2303 	struct ifaddr *ifa;
2304 	struct in6_ifaddr *ia;
2305 
2306 	IF_ADDR_RLOCK(ifp);
2307 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2308 		if (ifa->ifa_addr->sa_family != AF_INET6)
2309 			continue;
2310 		ia = (struct in6_ifaddr *)ifa;
2311 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2312 			/*
2313 			 * The TENTATIVE flag was likely set by hand
2314 			 * beforehand, implicitly indicating the need for DAD.
2315 			 * We may be able to skip the random delay in this
2316 			 * case, but we impose delays just in case.
2317 			 */
2318 			nd6_dad_start(ifa,
2319 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2320 		}
2321 	}
2322 	IF_ADDR_RUNLOCK(ifp);
2323 
2324 	/*
2325 	 * special cases, like 6to4, are handled in in6_ifattach
2326 	 */
2327 	in6_ifattach(ifp, NULL);
2328 }
2329 
2330 int
2331 in6if_do_dad(struct ifnet *ifp)
2332 {
2333 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2334 		return (0);
2335 
2336 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2337 		return (0);
2338 
2339 	switch (ifp->if_type) {
2340 #ifdef IFT_DUMMY
2341 	case IFT_DUMMY:
2342 #endif
2343 	case IFT_FAITH:
2344 		/*
2345 		 * These interfaces do not have the IFF_LOOPBACK flag,
2346 		 * but loop packets back.  We do not have to do DAD on such
2347 		 * interfaces.  We should even omit it, because loop-backed
2348 		 * NS would confuse the DAD procedure.
2349 		 */
2350 		return (0);
2351 	default:
2352 		/*
2353 		 * Our DAD routine requires the interface up and running.
2354 		 * However, some interfaces can be up before the RUNNING
2355 		 * status.  Additionaly, users may try to assign addresses
2356 		 * before the interface becomes up (or running).
2357 		 * We simply skip DAD in such a case as a work around.
2358 		 * XXX: we should rather mark "tentative" on such addresses,
2359 		 * and do DAD after the interface becomes ready.
2360 		 */
2361 		if (!((ifp->if_flags & IFF_UP) &&
2362 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2363 			return (0);
2364 
2365 		return (1);
2366 	}
2367 }
2368 
2369 /*
2370  * Calculate max IPv6 MTU through all the interfaces and store it
2371  * to in6_maxmtu.
2372  */
2373 void
2374 in6_setmaxmtu(void)
2375 {
2376 	unsigned long maxmtu = 0;
2377 	struct ifnet *ifp;
2378 
2379 	IFNET_RLOCK_NOSLEEP();
2380 	TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
2381 		/* this function can be called during ifnet initialization */
2382 		if (!ifp->if_afdata[AF_INET6])
2383 			continue;
2384 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2385 		    IN6_LINKMTU(ifp) > maxmtu)
2386 			maxmtu = IN6_LINKMTU(ifp);
2387 	}
2388 	IFNET_RUNLOCK_NOSLEEP();
2389 	if (maxmtu)	/* update only when maxmtu is positive */
2390 		V_in6_maxmtu = maxmtu;
2391 }
2392 
2393 /*
2394  * Provide the length of interface identifiers to be used for the link attached
2395  * to the given interface.  The length should be defined in "IPv6 over
2396  * xxx-link" document.  Note that address architecture might also define
2397  * the length for a particular set of address prefixes, regardless of the
2398  * link type.  As clarified in rfc2462bis, those two definitions should be
2399  * consistent, and those really are as of August 2004.
2400  */
2401 int
2402 in6_if2idlen(struct ifnet *ifp)
2403 {
2404 	switch (ifp->if_type) {
2405 	case IFT_ETHER:		/* RFC2464 */
2406 #ifdef IFT_PROPVIRTUAL
2407 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2408 #endif
2409 #ifdef IFT_L2VLAN
2410 	case IFT_L2VLAN:	/* ditto */
2411 #endif
2412 #ifdef IFT_IEEE80211
2413 	case IFT_IEEE80211:	/* ditto */
2414 #endif
2415 #ifdef IFT_MIP
2416 	case IFT_MIP:	/* ditto */
2417 #endif
2418 	case IFT_INFINIBAND:
2419 		return (64);
2420 	case IFT_FDDI:		/* RFC2467 */
2421 		return (64);
2422 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2423 		return (64);
2424 	case IFT_PPP:		/* RFC2472 */
2425 		return (64);
2426 	case IFT_ARCNET:	/* RFC2497 */
2427 		return (64);
2428 	case IFT_FRELAY:	/* RFC2590 */
2429 		return (64);
2430 	case IFT_IEEE1394:	/* RFC3146 */
2431 		return (64);
2432 	case IFT_GIF:
2433 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2434 	case IFT_LOOP:
2435 		return (64);	/* XXX: is this really correct? */
2436 	default:
2437 		/*
2438 		 * Unknown link type:
2439 		 * It might be controversial to use the today's common constant
2440 		 * of 64 for these cases unconditionally.  For full compliance,
2441 		 * we should return an error in this case.  On the other hand,
2442 		 * if we simply miss the standard for the link type or a new
2443 		 * standard is defined for a new link type, the IFID length
2444 		 * is very likely to be the common constant.  As a compromise,
2445 		 * we always use the constant, but make an explicit notice
2446 		 * indicating the "unknown" case.
2447 		 */
2448 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2449 		return (64);
2450 	}
2451 }
2452 
2453 #include <sys/sysctl.h>
2454 
2455 struct in6_llentry {
2456 	struct llentry		base;
2457 	struct sockaddr_in6	l3_addr6;
2458 };
2459 
2460 /*
2461  * Deletes an address from the address table.
2462  * This function is called by the timer functions
2463  * such as arptimer() and nd6_llinfo_timer(), and
2464  * the caller does the locking.
2465  */
2466 static void
2467 in6_lltable_free(struct lltable *llt, struct llentry *lle)
2468 {
2469 	LLE_WUNLOCK(lle);
2470 	LLE_LOCK_DESTROY(lle);
2471 	free(lle, M_LLTABLE);
2472 }
2473 
2474 static struct llentry *
2475 in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2476 {
2477 	struct in6_llentry *lle;
2478 
2479 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2480 	if (lle == NULL)		/* NB: caller generates msg */
2481 		return NULL;
2482 
2483 	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2484 	lle->base.lle_refcnt = 1;
2485 	lle->base.lle_free = in6_lltable_free;
2486 	LLE_LOCK_INIT(&lle->base);
2487 	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2488 	    CALLOUT_RETURNUNLOCKED);
2489 
2490 	return (&lle->base);
2491 }
2492 
2493 static void
2494 in6_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix,
2495     const struct sockaddr *mask, u_int flags)
2496 {
2497 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2498 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2499 	struct llentry *lle, *next;
2500 	int i;
2501 
2502 	/*
2503 	 * (flags & LLE_STATIC) means deleting all entries
2504 	 * including static ND6 entries.
2505 	 */
2506 	IF_AFDATA_WLOCK(llt->llt_ifp);
2507 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2508 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2509 			if (IN6_ARE_MASKED_ADDR_EQUAL(
2510 			    &satosin6(L3_ADDR(lle))->sin6_addr,
2511 			    &pfx->sin6_addr, &msk->sin6_addr) &&
2512 			    ((flags & LLE_STATIC) ||
2513 			    !(lle->la_flags & LLE_STATIC))) {
2514 				LLE_WLOCK(lle);
2515 				if (callout_stop(&lle->la_timer))
2516 					LLE_REMREF(lle);
2517 				llentry_free(lle);
2518 			}
2519 		}
2520 	}
2521 	IF_AFDATA_WUNLOCK(llt->llt_ifp);
2522 }
2523 
2524 static int
2525 in6_lltable_rtcheck(struct ifnet *ifp,
2526 		    u_int flags,
2527 		    const struct sockaddr *l3addr)
2528 {
2529 	struct rtentry *rt;
2530 	char ip6buf[INET6_ADDRSTRLEN];
2531 
2532 	KASSERT(l3addr->sa_family == AF_INET6,
2533 	    ("sin_family %d", l3addr->sa_family));
2534 
2535 	/* Our local addresses are always only installed on the default FIB. */
2536 	/* XXX rtalloc1 should take a const param */
2537 	rt = in6_rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0,
2538 	    RT_DEFAULT_FIB);
2539 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2540 		struct ifaddr *ifa;
2541 		/*
2542 		 * Create an ND6 cache for an IPv6 neighbor
2543 		 * that is not covered by our own prefix.
2544 		 */
2545 		/* XXX ifaof_ifpforaddr should take a const param */
2546 		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2547 		if (ifa != NULL) {
2548 			ifa_free(ifa);
2549 			if (rt != NULL)
2550 				RTFREE_LOCKED(rt);
2551 			return 0;
2552 		}
2553 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2554 		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2555 		if (rt != NULL)
2556 			RTFREE_LOCKED(rt);
2557 		return EINVAL;
2558 	}
2559 	RTFREE_LOCKED(rt);
2560 	return 0;
2561 }
2562 
2563 static struct llentry *
2564 in6_lltable_lookup(struct lltable *llt, u_int flags,
2565 	const struct sockaddr *l3addr)
2566 {
2567 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2568 	struct ifnet *ifp = llt->llt_ifp;
2569 	struct llentry *lle;
2570 	struct llentries *lleh;
2571 	u_int hashkey;
2572 
2573 	IF_AFDATA_LOCK_ASSERT(ifp);
2574 	KASSERT(l3addr->sa_family == AF_INET6,
2575 	    ("sin_family %d", l3addr->sa_family));
2576 
2577 	hashkey = sin6->sin6_addr.s6_addr32[3];
2578 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2579 	LIST_FOREACH(lle, lleh, lle_next) {
2580 		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2581 		if (lle->la_flags & LLE_DELETED)
2582 			continue;
2583 		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2584 		    sizeof(struct in6_addr)) == 0)
2585 			break;
2586 	}
2587 
2588 	if (lle == NULL) {
2589 		if (!(flags & LLE_CREATE))
2590 			return (NULL);
2591 		/*
2592 		 * A route that covers the given address must have
2593 		 * been installed 1st because we are doing a resolution,
2594 		 * verify this.
2595 		 */
2596 		if (!(flags & LLE_IFADDR) &&
2597 		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2598 			return NULL;
2599 
2600 		lle = in6_lltable_new(l3addr, flags);
2601 		if (lle == NULL) {
2602 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2603 			return NULL;
2604 		}
2605 		lle->la_flags = flags & ~LLE_CREATE;
2606 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2607 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2608 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2609 		}
2610 
2611 		lle->lle_tbl  = llt;
2612 		lle->lle_head = lleh;
2613 		lle->la_flags |= LLE_LINKED;
2614 		LIST_INSERT_HEAD(lleh, lle, lle_next);
2615 	} else if (flags & LLE_DELETE) {
2616 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2617 			LLE_WLOCK(lle);
2618 			lle->la_flags |= LLE_DELETED;
2619 #ifdef DIAGNOSTIC
2620 			log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2621 #endif
2622 			if ((lle->la_flags &
2623 			    (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2624 				llentry_free(lle);
2625 			else
2626 				LLE_WUNLOCK(lle);
2627 		}
2628 		lle = (void *)-1;
2629 	}
2630 	if (LLE_IS_VALID(lle)) {
2631 		if (flags & LLE_EXCLUSIVE)
2632 			LLE_WLOCK(lle);
2633 		else
2634 			LLE_RLOCK(lle);
2635 	}
2636 	return (lle);
2637 }
2638 
2639 static int
2640 in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2641 {
2642 	struct ifnet *ifp = llt->llt_ifp;
2643 	struct llentry *lle;
2644 	/* XXX stack use */
2645 	struct {
2646 		struct rt_msghdr	rtm;
2647 		struct sockaddr_in6	sin6;
2648 		/*
2649 		 * ndp.c assumes that sdl is word aligned
2650 		 */
2651 #ifdef __LP64__
2652 		uint32_t		pad;
2653 #endif
2654 		struct sockaddr_dl	sdl;
2655 	} ndpc;
2656 	int i, error;
2657 
2658 	if (ifp->if_flags & IFF_LOOPBACK)
2659 		return 0;
2660 
2661 	LLTABLE_LOCK_ASSERT();
2662 
2663 	error = 0;
2664 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2665 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2666 			struct sockaddr_dl *sdl;
2667 
2668 			/* skip deleted or invalid entries */
2669 			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2670 				continue;
2671 			/* Skip if jailed and not a valid IP of the prison. */
2672 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2673 				continue;
2674 			/*
2675 			 * produce a msg made of:
2676 			 *  struct rt_msghdr;
2677 			 *  struct sockaddr_in6 (IPv6)
2678 			 *  struct sockaddr_dl;
2679 			 */
2680 			bzero(&ndpc, sizeof(ndpc));
2681 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2682 			ndpc.rtm.rtm_version = RTM_VERSION;
2683 			ndpc.rtm.rtm_type = RTM_GET;
2684 			ndpc.rtm.rtm_flags = RTF_UP;
2685 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2686 			ndpc.sin6.sin6_family = AF_INET6;
2687 			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2688 			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2689 			if (V_deembed_scopeid)
2690 				sa6_recoverscope(&ndpc.sin6);
2691 
2692 			/* publish */
2693 			if (lle->la_flags & LLE_PUB)
2694 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2695 
2696 			sdl = &ndpc.sdl;
2697 			sdl->sdl_family = AF_LINK;
2698 			sdl->sdl_len = sizeof(*sdl);
2699 			sdl->sdl_alen = ifp->if_addrlen;
2700 			sdl->sdl_index = ifp->if_index;
2701 			sdl->sdl_type = ifp->if_type;
2702 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2703 			ndpc.rtm.rtm_rmx.rmx_expire =
2704 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2705 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2706 			if (lle->la_flags & LLE_STATIC)
2707 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2708 			ndpc.rtm.rtm_index = ifp->if_index;
2709 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2710 			if (error)
2711 				break;
2712 		}
2713 	}
2714 	return error;
2715 }
2716 
2717 void *
2718 in6_domifattach(struct ifnet *ifp)
2719 {
2720 	struct in6_ifextra *ext;
2721 
2722 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2723 	bzero(ext, sizeof(*ext));
2724 
2725 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2726 	    M_IFADDR, M_WAITOK);
2727 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2728 
2729 	ext->icmp6_ifstat =
2730 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2731 	    M_IFADDR, M_WAITOK);
2732 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2733 
2734 	ext->nd_ifinfo = nd6_ifattach(ifp);
2735 	ext->scope6_id = scope6_ifattach(ifp);
2736 	ext->lltable = lltable_init(ifp, AF_INET6);
2737 	if (ext->lltable != NULL) {
2738 		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2739 		ext->lltable->llt_lookup = in6_lltable_lookup;
2740 		ext->lltable->llt_dump = in6_lltable_dump;
2741 	}
2742 
2743 	ext->mld_ifinfo = mld_domifattach(ifp);
2744 
2745 	return ext;
2746 }
2747 
2748 void
2749 in6_domifdetach(struct ifnet *ifp, void *aux)
2750 {
2751 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2752 
2753 	mld_domifdetach(ifp);
2754 	scope6_ifdetach(ext->scope6_id);
2755 	nd6_ifdetach(ext->nd_ifinfo);
2756 	lltable_free(ext->lltable);
2757 	free(ext->in6_ifstat, M_IFADDR);
2758 	free(ext->icmp6_ifstat, M_IFADDR);
2759 	free(ext, M_IFADDR);
2760 }
2761 
2762 /*
2763  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2764  * v4 mapped addr or v4 compat addr
2765  */
2766 void
2767 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2768 {
2769 
2770 	bzero(sin, sizeof(*sin));
2771 	sin->sin_len = sizeof(struct sockaddr_in);
2772 	sin->sin_family = AF_INET;
2773 	sin->sin_port = sin6->sin6_port;
2774 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2775 }
2776 
2777 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2778 void
2779 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2780 {
2781 	bzero(sin6, sizeof(*sin6));
2782 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2783 	sin6->sin6_family = AF_INET6;
2784 	sin6->sin6_port = sin->sin_port;
2785 	sin6->sin6_addr.s6_addr32[0] = 0;
2786 	sin6->sin6_addr.s6_addr32[1] = 0;
2787 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2788 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2789 }
2790 
2791 /* Convert sockaddr_in6 into sockaddr_in. */
2792 void
2793 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2794 {
2795 	struct sockaddr_in *sin_p;
2796 	struct sockaddr_in6 sin6;
2797 
2798 	/*
2799 	 * Save original sockaddr_in6 addr and convert it
2800 	 * to sockaddr_in.
2801 	 */
2802 	sin6 = *(struct sockaddr_in6 *)nam;
2803 	sin_p = (struct sockaddr_in *)nam;
2804 	in6_sin6_2_sin(sin_p, &sin6);
2805 }
2806 
2807 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2808 void
2809 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2810 {
2811 	struct sockaddr_in *sin_p;
2812 	struct sockaddr_in6 *sin6_p;
2813 
2814 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2815 	sin_p = (struct sockaddr_in *)*nam;
2816 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2817 	free(*nam, M_SONAME);
2818 	*nam = (struct sockaddr *)sin6_p;
2819 }
2820