xref: /netbsd-src/sys/netinet6/in6.c (revision 404fbe5fb94ca1e054339640cabb2801ce52dd30)
1 /*	$NetBSD: in6.c,v 1.143 2008/12/19 18:49:39 cegger Exp $	*/
2 /*	$KAME: in6.c,v 1.198 2001/07/18 09:12:38 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)in.c	8.2 (Berkeley) 11/15/93
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: in6.c,v 1.143 2008/12/19 18:49:39 cegger Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_pfil_hooks.h"
69 
70 #include <sys/param.h>
71 #include <sys/ioctl.h>
72 #include <sys/errno.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/proc.h>
79 #include <sys/time.h>
80 #include <sys/kernel.h>
81 #include <sys/syslog.h>
82 #include <sys/kauth.h>
83 
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/route.h>
87 #include <net/if_dl.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <net/if_ether.h>
92 
93 #include <netinet/ip6.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet6/nd6.h>
96 #include <netinet6/mld6_var.h>
97 #include <netinet6/ip6_mroute.h>
98 #include <netinet6/in6_ifattach.h>
99 #include <netinet6/scope6_var.h>
100 
101 #include <net/net_osdep.h>
102 
103 #ifdef PFIL_HOOKS
104 #include <net/pfil.h>
105 #endif
106 
107 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
108 
109 /* enable backward compatibility code for obsoleted ioctls */
110 #define COMPAT_IN6IFIOCTL
111 
112 #ifdef	IN6_DEBUG
113 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
114 #else
115 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
116 #endif /* IN6_DEBUG */
117 
118 /*
119  * Definitions of some constant IP6 addresses.
120  */
121 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
122 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
123 const struct in6_addr in6addr_nodelocal_allnodes =
124 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
125 const struct in6_addr in6addr_linklocal_allnodes =
126 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
127 const struct in6_addr in6addr_linklocal_allrouters =
128 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
129 
130 const struct in6_addr in6mask0 = IN6MASK0;
131 const struct in6_addr in6mask32 = IN6MASK32;
132 const struct in6_addr in6mask64 = IN6MASK64;
133 const struct in6_addr in6mask96 = IN6MASK96;
134 const struct in6_addr in6mask128 = IN6MASK128;
135 
136 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
137 				     0, 0, IN6ADDR_ANY_INIT, 0};
138 
139 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
140 	struct ifnet *, struct lwp *);
141 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
142 	struct sockaddr_in6 *, int);
143 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
144 
145 /*
146  * Subroutine for in6_ifaddloop() and in6_ifremloop().
147  * This routine does actual work.
148  */
149 static void
150 in6_ifloop_request(int cmd, struct ifaddr *ifa)
151 {
152 	struct sockaddr_in6 lo_sa;
153 	struct sockaddr_in6 all1_sa;
154 	struct rtentry *nrt = NULL;
155 	int e;
156 
157 	sockaddr_in6_init(&all1_sa, &in6mask128, 0, 0, 0);
158 	sockaddr_in6_init(&lo_sa, &in6addr_loopback, 0, 0, 0);
159 
160 	/*
161 	 * We specify the address itself as the gateway, and set the
162 	 * RTF_LLINFO flag, so that the corresponding host route would have
163 	 * the flag, and thus applications that assume traditional behavior
164 	 * would be happy.  Note that we assume the caller of the function
165 	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
166 	 * which changes the outgoing interface to the loopback interface.
167 	 */
168 	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr,
169 	    (struct sockaddr *)&all1_sa, RTF_UP|RTF_HOST|RTF_LLINFO, &nrt);
170 	if (e != 0) {
171 		log(LOG_ERR, "in6_ifloop_request: "
172 		    "%s operation failed for %s (errno=%d)\n",
173 		    cmd == RTM_ADD ? "ADD" : "DELETE",
174 		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
175 		    e);
176 	}
177 
178 	/*
179 	 * Make sure rt_ifa be equal to IFA, the second argument of the
180 	 * function.
181 	 * We need this because when we refer to rt_ifa->ia6_flags in
182 	 * ip6_input, we assume that the rt_ifa points to the address instead
183 	 * of the loopback address.
184 	 */
185 	if (cmd == RTM_ADD && nrt && ifa != nrt->rt_ifa)
186 		rt_replace_ifa(nrt, ifa);
187 
188 	/*
189 	 * Report the addition/removal of the address to the routing socket.
190 	 * XXX: since we called rtinit for a p2p interface with a destination,
191 	 *      we end up reporting twice in such a case.  Should we rather
192 	 *      omit the second report?
193 	 */
194 	if (nrt) {
195 		rt_newaddrmsg(cmd, ifa, e, nrt);
196 		if (cmd == RTM_DELETE) {
197 			if (nrt->rt_refcnt <= 0) {
198 				/* XXX: we should free the entry ourselves. */
199 				nrt->rt_refcnt++;
200 				rtfree(nrt);
201 			}
202 		} else {
203 			/* the cmd must be RTM_ADD here */
204 			nrt->rt_refcnt--;
205 		}
206 	}
207 }
208 
209 /*
210  * Add ownaddr as loopback rtentry.  We previously add the route only if
211  * necessary (ex. on a p2p link).  However, since we now manage addresses
212  * separately from prefixes, we should always add the route.  We can't
213  * rely on the cloning mechanism from the corresponding interface route
214  * any more.
215  */
216 void
217 in6_ifaddloop(struct ifaddr *ifa)
218 {
219 	struct rtentry *rt;
220 
221 	/* If there is no loopback entry, allocate one. */
222 	rt = rtalloc1(ifa->ifa_addr, 0);
223 	if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
224 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
225 		in6_ifloop_request(RTM_ADD, ifa);
226 	if (rt != NULL)
227 		rt->rt_refcnt--;
228 }
229 
230 /*
231  * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
232  * if it exists.
233  */
234 void
235 in6_ifremloop(struct ifaddr *ifa)
236 {
237 	struct in6_ifaddr *alt_ia = NULL, *ia;
238 	struct rtentry *rt;
239 	int ia_count = 0;
240 
241 	/*
242 	 * Some of BSD variants do not remove cloned routes
243 	 * from an interface direct route, when removing the direct route
244 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
245 	 * cloned routes, they could fail to remove the cloned routes when
246 	 * we handle multple addresses that share a common prefix.
247 	 * So, we should remove the route corresponding to the deleted address.
248 	 */
249 
250 	/*
251 	 * Delete the entry only if exactly one ifaddr matches the
252 	 * address, ifa->ifa_addr.
253 	 *
254 	 * If more than one ifaddr matches, replace the ifaddr in
255 	 * the routing table, rt_ifa, with a different ifaddr than
256 	 * the one we are purging, ifa.  It is important to do
257 	 * this, or else the routing table can accumulate dangling
258 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
259 	 * which will lead to crashes, later.  (More than one ifaddr
260 	 * can match if we assign the same address to multiple---probably
261 	 * p2p---interfaces.)
262 	 *
263 	 * XXX An old comment at this place said, "we should avoid
264 	 * XXX such a configuration [i.e., interfaces with the same
265 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
266 	 * XXX agree, especially now that I have fixed the dangling
267 	 * XXX ifp-pointers bug.
268 	 */
269 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
270 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
271 			continue;
272 		if (ia->ia_ifp != ifa->ifa_ifp)
273 			alt_ia = ia;
274 		if (++ia_count > 1 && alt_ia != NULL)
275 			break;
276 	}
277 
278 	if (ia_count == 0)
279 		return;
280 
281 	if ((rt = rtalloc1(ifa->ifa_addr, 0)) == NULL)
282 		return;
283 	rt->rt_refcnt--;
284 
285 	/*
286 	 * Before deleting, check if a corresponding loopbacked
287 	 * host route surely exists.  With this check, we can avoid
288 	 * deleting an interface direct route whose destination is
289 	 * the same as the address being removed.  This can happen
290 	 * when removing a subnet-router anycast address on an
291 	 * interface attached to a shared medium.
292 	 */
293 	if ((rt->rt_flags & RTF_HOST) == 0 ||
294 	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
295 		return;
296 
297 	/* If we cannot replace the route's ifaddr with the equivalent
298 	 * ifaddr of another interface, I believe it is safest to
299 	 * delete the route.
300 	 */
301 	if (ia_count == 1 || alt_ia == NULL)
302 		in6_ifloop_request(RTM_DELETE, ifa);
303 	else
304 		rt_replace_ifa(rt, &alt_ia->ia_ifa);
305 }
306 
307 int
308 in6_mask2len(struct in6_addr *mask, u_char *lim0)
309 {
310 	int x = 0, y;
311 	u_char *lim = lim0, *p;
312 
313 	/* ignore the scope_id part */
314 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
315 		lim = (u_char *)mask + sizeof(*mask);
316 	for (p = (u_char *)mask; p < lim; x++, p++) {
317 		if (*p != 0xff)
318 			break;
319 	}
320 	y = 0;
321 	if (p < lim) {
322 		for (y = 0; y < NBBY; y++) {
323 			if ((*p & (0x80 >> y)) == 0)
324 				break;
325 		}
326 	}
327 
328 	/*
329 	 * when the limit pointer is given, do a stricter check on the
330 	 * remaining bits.
331 	 */
332 	if (p < lim) {
333 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
334 			return -1;
335 		for (p = p + 1; p < lim; p++)
336 			if (*p != 0)
337 				return -1;
338 	}
339 
340 	return x * NBBY + y;
341 }
342 
343 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
344 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
345 
346 static int
347 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
348     struct lwp *l, int privileged)
349 {
350 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
351 	struct	in6_ifaddr *ia = NULL;
352 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
353 	struct sockaddr_in6 *sa6;
354 	int error;
355 	switch (cmd) {
356 	/*
357 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
358 	 */
359 	case SIOCSIFADDR:
360 	case SIOCSIFDSTADDR:
361 #ifdef SIOCSIFCONF_X25
362 	case SIOCSIFCONF_X25:
363 #endif
364 		return EOPNOTSUPP;
365 	case SIOCGETSGCNT_IN6:
366 	case SIOCGETMIFCNT_IN6:
367 		return mrt6_ioctl(cmd, data);
368 	}
369 
370 	if (ifp == NULL)
371 		return EOPNOTSUPP;
372 
373 	switch (cmd) {
374 	case SIOCSNDFLUSH_IN6:
375 	case SIOCSPFXFLUSH_IN6:
376 	case SIOCSRTRFLUSH_IN6:
377 	case SIOCSDEFIFACE_IN6:
378 	case SIOCSIFINFO_FLAGS:
379 	case SIOCSIFINFO_IN6:
380 		if (!privileged)
381 			return EPERM;
382 		/* FALLTHROUGH */
383 	case OSIOCGIFINFO_IN6:
384 	case SIOCGIFINFO_IN6:
385 	case SIOCGDRLST_IN6:
386 	case SIOCGPRLST_IN6:
387 	case SIOCGNBRINFO_IN6:
388 	case SIOCGDEFIFACE_IN6:
389 		return nd6_ioctl(cmd, data, ifp);
390 	}
391 
392 	switch (cmd) {
393 	case SIOCSIFPREFIX_IN6:
394 	case SIOCDIFPREFIX_IN6:
395 	case SIOCAIFPREFIX_IN6:
396 	case SIOCCIFPREFIX_IN6:
397 	case SIOCSGIFPREFIX_IN6:
398 	case SIOCGIFPREFIX_IN6:
399 		log(LOG_NOTICE,
400 		    "prefix ioctls are now invalidated. "
401 		    "please use ifconfig.\n");
402 		return EOPNOTSUPP;
403 	}
404 
405 	switch (cmd) {
406 	case SIOCALIFADDR:
407 	case SIOCDLIFADDR:
408 		if (!privileged)
409 			return EPERM;
410 		/* FALLTHROUGH */
411 	case SIOCGLIFADDR:
412 		return in6_lifaddr_ioctl(so, cmd, data, ifp, l);
413 	}
414 
415 	/*
416 	 * Find address for this interface, if it exists.
417 	 *
418 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
419 	 * only, and used the first interface address as the target of other
420 	 * operations (without checking ifra_addr).  This was because netinet
421 	 * code/API assumed at most 1 interface address per interface.
422 	 * Since IPv6 allows a node to assign multiple addresses
423 	 * on a single interface, we almost always look and check the
424 	 * presence of ifra_addr, and reject invalid ones here.
425 	 * It also decreases duplicated code among SIOC*_IN6 operations.
426 	 */
427 	switch (cmd) {
428 	case SIOCAIFADDR_IN6:
429 	case SIOCSIFPHYADDR_IN6:
430 		sa6 = &ifra->ifra_addr;
431 		break;
432 	case SIOCSIFADDR_IN6:
433 	case SIOCGIFADDR_IN6:
434 	case SIOCSIFDSTADDR_IN6:
435 	case SIOCSIFNETMASK_IN6:
436 	case SIOCGIFDSTADDR_IN6:
437 	case SIOCGIFNETMASK_IN6:
438 	case SIOCDIFADDR_IN6:
439 	case SIOCGIFPSRCADDR_IN6:
440 	case SIOCGIFPDSTADDR_IN6:
441 	case SIOCGIFAFLAG_IN6:
442 	case SIOCSNDFLUSH_IN6:
443 	case SIOCSPFXFLUSH_IN6:
444 	case SIOCSRTRFLUSH_IN6:
445 	case SIOCGIFALIFETIME_IN6:
446 	case SIOCGIFSTAT_IN6:
447 	case SIOCGIFSTAT_ICMP6:
448 		sa6 = &ifr->ifr_addr;
449 		break;
450 	default:
451 		sa6 = NULL;
452 		break;
453 	}
454 	if (sa6 && sa6->sin6_family == AF_INET6) {
455 		if (sa6->sin6_scope_id != 0)
456 			error = sa6_embedscope(sa6, 0);
457 		else
458 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
459 		if (error != 0)
460 			return error;
461 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
462 	} else
463 		ia = NULL;
464 
465 	switch (cmd) {
466 	case SIOCSIFADDR_IN6:
467 	case SIOCSIFDSTADDR_IN6:
468 	case SIOCSIFNETMASK_IN6:
469 		/*
470 		 * Since IPv6 allows a node to assign multiple addresses
471 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
472 		 */
473 		return EINVAL;
474 
475 	case SIOCDIFADDR_IN6:
476 		/*
477 		 * for IPv4, we look for existing in_ifaddr here to allow
478 		 * "ifconfig if0 delete" to remove the first IPv4 address on
479 		 * the interface.  For IPv6, as the spec allows multiple
480 		 * interface address from the day one, we consider "remove the
481 		 * first one" semantics to be not preferable.
482 		 */
483 		if (ia == NULL)
484 			return EADDRNOTAVAIL;
485 		/* FALLTHROUGH */
486 	case SIOCAIFADDR_IN6:
487 		/*
488 		 * We always require users to specify a valid IPv6 address for
489 		 * the corresponding operation.
490 		 */
491 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
492 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
493 			return EAFNOSUPPORT;
494 		if (!privileged)
495 			return EPERM;
496 
497 		break;
498 
499 	case SIOCGIFADDR_IN6:
500 		/* This interface is basically deprecated. use SIOCGIFCONF. */
501 		/* FALLTHROUGH */
502 	case SIOCGIFAFLAG_IN6:
503 	case SIOCGIFNETMASK_IN6:
504 	case SIOCGIFDSTADDR_IN6:
505 	case SIOCGIFALIFETIME_IN6:
506 		/* must think again about its semantics */
507 		if (ia == NULL)
508 			return EADDRNOTAVAIL;
509 		break;
510 	}
511 
512 	switch (cmd) {
513 
514 	case SIOCGIFADDR_IN6:
515 		ifr->ifr_addr = ia->ia_addr;
516 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
517 			return error;
518 		break;
519 
520 	case SIOCGIFDSTADDR_IN6:
521 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
522 			return EINVAL;
523 		/*
524 		 * XXX: should we check if ifa_dstaddr is NULL and return
525 		 * an error?
526 		 */
527 		ifr->ifr_dstaddr = ia->ia_dstaddr;
528 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
529 			return error;
530 		break;
531 
532 	case SIOCGIFNETMASK_IN6:
533 		ifr->ifr_addr = ia->ia_prefixmask;
534 		break;
535 
536 	case SIOCGIFAFLAG_IN6:
537 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
538 		break;
539 
540 	case SIOCGIFSTAT_IN6:
541 		if (ifp == NULL)
542 			return EINVAL;
543 		bzero(&ifr->ifr_ifru.ifru_stat,
544 		    sizeof(ifr->ifr_ifru.ifru_stat));
545 		ifr->ifr_ifru.ifru_stat =
546 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
547 		break;
548 
549 	case SIOCGIFSTAT_ICMP6:
550 		if (ifp == NULL)
551 			return EINVAL;
552 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
553 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
554 		ifr->ifr_ifru.ifru_icmp6stat =
555 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
556 		break;
557 
558 	case SIOCGIFALIFETIME_IN6:
559 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
560 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
561 			time_t maxexpire;
562 			struct in6_addrlifetime *retlt =
563 			    &ifr->ifr_ifru.ifru_lifetime;
564 
565 			/*
566 			 * XXX: adjust expiration time assuming time_t is
567 			 * signed.
568 			 */
569 			maxexpire = ((time_t)~0) &
570 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
571 			if (ia->ia6_lifetime.ia6t_vltime <
572 			    maxexpire - ia->ia6_updatetime) {
573 				retlt->ia6t_expire = ia->ia6_updatetime +
574 				    ia->ia6_lifetime.ia6t_vltime;
575 			} else
576 				retlt->ia6t_expire = maxexpire;
577 		}
578 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
579 			time_t maxexpire;
580 			struct in6_addrlifetime *retlt =
581 			    &ifr->ifr_ifru.ifru_lifetime;
582 
583 			/*
584 			 * XXX: adjust expiration time assuming time_t is
585 			 * signed.
586 			 */
587 			maxexpire = ((time_t)~0) &
588 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
589 			if (ia->ia6_lifetime.ia6t_pltime <
590 			    maxexpire - ia->ia6_updatetime) {
591 				retlt->ia6t_preferred = ia->ia6_updatetime +
592 				    ia->ia6_lifetime.ia6t_pltime;
593 			} else
594 				retlt->ia6t_preferred = maxexpire;
595 		}
596 		break;
597 
598 	case SIOCAIFADDR_IN6:
599 	{
600 		int i;
601 		struct nd_prefixctl pr0;
602 		struct nd_prefix *pr;
603 
604 		/* reject read-only flags */
605 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
606 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
607 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
608 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
609 			return EINVAL;
610 		}
611 		/*
612 		 * first, make or update the interface address structure,
613 		 * and link it to the list.
614 		 */
615 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
616 			return error;
617 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
618 		    == NULL) {
619 		    	/*
620 			 * this can happen when the user specify the 0 valid
621 			 * lifetime.
622 			 */
623 			break;
624 		}
625 
626 		/*
627 		 * then, make the prefix on-link on the interface.
628 		 * XXX: we'd rather create the prefix before the address, but
629 		 * we need at least one address to install the corresponding
630 		 * interface route, so we configure the address first.
631 		 */
632 
633 		/*
634 		 * convert mask to prefix length (prefixmask has already
635 		 * been validated in in6_update_ifa().
636 		 */
637 		bzero(&pr0, sizeof(pr0));
638 		pr0.ndpr_ifp = ifp;
639 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
640 		    NULL);
641 		if (pr0.ndpr_plen == 128) {
642 			break;	/* we don't need to install a host route. */
643 		}
644 		pr0.ndpr_prefix = ifra->ifra_addr;
645 		/* apply the mask for safety. */
646 		for (i = 0; i < 4; i++) {
647 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
648 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
649 		}
650 		/*
651 		 * XXX: since we don't have an API to set prefix (not address)
652 		 * lifetimes, we just use the same lifetimes as addresses.
653 		 * The (temporarily) installed lifetimes can be overridden by
654 		 * later advertised RAs (when accept_rtadv is non 0), which is
655 		 * an intended behavior.
656 		 */
657 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
658 		pr0.ndpr_raf_auto =
659 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
660 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
661 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
662 
663 		/* add the prefix if not yet. */
664 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
665 			/*
666 			 * nd6_prelist_add will install the corresponding
667 			 * interface route.
668 			 */
669 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
670 				return error;
671 			if (pr == NULL) {
672 				log(LOG_ERR, "nd6_prelist_add succeeded but "
673 				    "no prefix\n");
674 				return EINVAL; /* XXX panic here? */
675 			}
676 		}
677 
678 		/* relate the address to the prefix */
679 		if (ia->ia6_ndpr == NULL) {
680 			ia->ia6_ndpr = pr;
681 			pr->ndpr_refcnt++;
682 
683 			/*
684 			 * If this is the first autoconf address from the
685 			 * prefix, create a temporary address as well
686 			 * (when required).
687 			 */
688 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
689 			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
690 				int e;
691 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
692 					log(LOG_NOTICE, "in6_control: failed "
693 					    "to create a temporary address, "
694 					    "errno=%d\n", e);
695 				}
696 			}
697 		}
698 
699 		/*
700 		 * this might affect the status of autoconfigured addresses,
701 		 * that is, this address might make other addresses detached.
702 		 */
703 		pfxlist_onlink_check();
704 
705 #ifdef PFIL_HOOKS
706 		(void)pfil_run_hooks(&if_pfil, (struct mbuf **)SIOCAIFADDR_IN6,
707 		    ifp, PFIL_IFADDR);
708 #endif
709 
710 		break;
711 	}
712 
713 	case SIOCDIFADDR_IN6:
714 	{
715 		struct nd_prefix *pr;
716 
717 		/*
718 		 * If the address being deleted is the only one that owns
719 		 * the corresponding prefix, expire the prefix as well.
720 		 * XXX: theoretically, we don't have to worry about such
721 		 * relationship, since we separate the address management
722 		 * and the prefix management.  We do this, however, to provide
723 		 * as much backward compatibility as possible in terms of
724 		 * the ioctl operation.
725 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
726 		 */
727 		pr = ia->ia6_ndpr;
728 		in6_purgeaddr(&ia->ia_ifa);
729 		if (pr && pr->ndpr_refcnt == 0)
730 			prelist_remove(pr);
731 #ifdef PFIL_HOOKS
732 		(void)pfil_run_hooks(&if_pfil, (struct mbuf **)SIOCDIFADDR_IN6,
733 		    ifp, PFIL_IFADDR);
734 #endif
735 		break;
736 	}
737 
738 	default:
739 		return ENOTTY;
740 	}
741 
742 	return 0;
743 }
744 
745 int
746 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp,
747     struct lwp *l)
748 {
749 	int error, privileged, s;
750 
751 	privileged = 0;
752 	if (l && !kauth_authorize_generic(l->l_cred,
753 	    KAUTH_GENERIC_ISSUSER, NULL))
754 		privileged++;
755 
756 	s = splnet();
757 	error = in6_control1(so , cmd, data, ifp, l, privileged);
758 	splx(s);
759 	return error;
760 }
761 
762 /*
763  * Update parameters of an IPv6 interface address.
764  * If necessary, a new entry is created and linked into address chains.
765  * This function is separated from in6_control().
766  * XXX: should this be performed under splnet()?
767  */
768 static int
769 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
770     struct in6_ifaddr *ia, int flags)
771 {
772 	int error = 0, hostIsNew = 0, plen = -1;
773 	struct in6_ifaddr *oia;
774 	struct sockaddr_in6 dst6;
775 	struct in6_addrlifetime *lt;
776 	struct in6_multi_mship *imm;
777 	struct in6_multi *in6m_sol;
778 	struct rtentry *rt;
779 	int dad_delay;
780 
781 	in6m_sol = NULL;
782 
783 	/* Validate parameters */
784 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
785 		return EINVAL;
786 
787 	/*
788 	 * The destination address for a p2p link must have a family
789 	 * of AF_UNSPEC or AF_INET6.
790 	 */
791 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
792 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
793 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
794 		return EAFNOSUPPORT;
795 	/*
796 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
797 	 * does not carry fields other than sin6_len.
798 	 */
799 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
800 		return EINVAL;
801 	/*
802 	 * Because the IPv6 address architecture is classless, we require
803 	 * users to specify a (non 0) prefix length (mask) for a new address.
804 	 * We also require the prefix (when specified) mask is valid, and thus
805 	 * reject a non-consecutive mask.
806 	 */
807 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
808 		return EINVAL;
809 	if (ifra->ifra_prefixmask.sin6_len != 0) {
810 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
811 		    (u_char *)&ifra->ifra_prefixmask +
812 		    ifra->ifra_prefixmask.sin6_len);
813 		if (plen <= 0)
814 			return EINVAL;
815 	} else {
816 		/*
817 		 * In this case, ia must not be NULL.  We just use its prefix
818 		 * length.
819 		 */
820 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
821 	}
822 	/*
823 	 * If the destination address on a p2p interface is specified,
824 	 * and the address is a scoped one, validate/set the scope
825 	 * zone identifier.
826 	 */
827 	dst6 = ifra->ifra_dstaddr;
828 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
829 	    (dst6.sin6_family == AF_INET6)) {
830 		struct in6_addr in6_tmp;
831 		u_int32_t zoneid;
832 
833 		in6_tmp = dst6.sin6_addr;
834 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
835 			return EINVAL; /* XXX: should be impossible */
836 
837 		if (dst6.sin6_scope_id != 0) {
838 			if (dst6.sin6_scope_id != zoneid)
839 				return EINVAL;
840 		} else		/* user omit to specify the ID. */
841 			dst6.sin6_scope_id = zoneid;
842 
843 		/* convert into the internal form */
844 		if (sa6_embedscope(&dst6, 0))
845 			return EINVAL; /* XXX: should be impossible */
846 	}
847 	/*
848 	 * The destination address can be specified only for a p2p or a
849 	 * loopback interface.  If specified, the corresponding prefix length
850 	 * must be 128.
851 	 */
852 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
853 #ifdef FORCE_P2PPLEN
854 		int i;
855 #endif
856 
857 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
858 			/* XXX: noisy message */
859 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
860 			    "be specified for a p2p or a loopback IF only\n"));
861 			return EINVAL;
862 		}
863 		if (plen != 128) {
864 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
865 			    "be 128 when dstaddr is specified\n"));
866 #ifdef FORCE_P2PPLEN
867 			/*
868 			 * To be compatible with old configurations,
869 			 * such as ifconfig gif0 inet6 2001::1 2001::2
870 			 * prefixlen 126, we override the specified
871 			 * prefixmask as if the prefix length was 128.
872 			 */
873 			ifra->ifra_prefixmask.sin6_len =
874 			    sizeof(struct sockaddr_in6);
875 			for (i = 0; i < 4; i++)
876 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
877 				    0xffffffff;
878 			plen = 128;
879 #else
880 			return EINVAL;
881 #endif
882 		}
883 	}
884 	/* lifetime consistency check */
885 	lt = &ifra->ifra_lifetime;
886 	if (lt->ia6t_pltime > lt->ia6t_vltime)
887 		return EINVAL;
888 	if (lt->ia6t_vltime == 0) {
889 		/*
890 		 * the following log might be noisy, but this is a typical
891 		 * configuration mistake or a tool's bug.
892 		 */
893 		nd6log((LOG_INFO,
894 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
895 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr)));
896 
897 		if (ia == NULL)
898 			return 0; /* there's nothing to do */
899 	}
900 
901 	/*
902 	 * If this is a new address, allocate a new ifaddr and link it
903 	 * into chains.
904 	 */
905 	if (ia == NULL) {
906 		hostIsNew = 1;
907 		/*
908 		 * When in6_update_ifa() is called in a process of a received
909 		 * RA, it is called under an interrupt context.  So, we should
910 		 * call malloc with M_NOWAIT.
911 		 */
912 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
913 		    M_NOWAIT);
914 		if (ia == NULL)
915 			return ENOBUFS;
916 		bzero((void *)ia, sizeof(*ia));
917 		LIST_INIT(&ia->ia6_memberships);
918 		/* Initialize the address and masks, and put time stamp */
919 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
920 		ia->ia_addr.sin6_family = AF_INET6;
921 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
922 		ia->ia6_createtime = time_second;
923 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
924 			/*
925 			 * XXX: some functions expect that ifa_dstaddr is not
926 			 * NULL for p2p interfaces.
927 			 */
928 			ia->ia_ifa.ifa_dstaddr =
929 			    (struct sockaddr *)&ia->ia_dstaddr;
930 		} else {
931 			ia->ia_ifa.ifa_dstaddr = NULL;
932 		}
933 		ia->ia_ifa.ifa_netmask =
934 		    (struct sockaddr *)&ia->ia_prefixmask;
935 
936 		ia->ia_ifp = ifp;
937 		if ((oia = in6_ifaddr) != NULL) {
938 			for ( ; oia->ia_next; oia = oia->ia_next)
939 				continue;
940 			oia->ia_next = ia;
941 		} else
942 			in6_ifaddr = ia;
943 		/* gain a refcnt for the link from in6_ifaddr */
944 		IFAREF(&ia->ia_ifa);
945 
946 		ifa_insert(ifp, &ia->ia_ifa);
947 	}
948 
949 	/* update timestamp */
950 	ia->ia6_updatetime = time_second;
951 
952 	/* set prefix mask */
953 	if (ifra->ifra_prefixmask.sin6_len) {
954 		/*
955 		 * We prohibit changing the prefix length of an existing
956 		 * address, because
957 		 * + such an operation should be rare in IPv6, and
958 		 * + the operation would confuse prefix management.
959 		 */
960 		if (ia->ia_prefixmask.sin6_len &&
961 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
962 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
963 			    " existing (%s) address should not be changed\n",
964 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
965 			error = EINVAL;
966 			goto unlink;
967 		}
968 		ia->ia_prefixmask = ifra->ifra_prefixmask;
969 	}
970 
971 	/*
972 	 * If a new destination address is specified, scrub the old one and
973 	 * install the new destination.  Note that the interface must be
974 	 * p2p or loopback (see the check above.)
975 	 */
976 	if (dst6.sin6_family == AF_INET6 &&
977 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
978 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
979 		    rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0) {
980 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
981 			    "a route to the old destination: %s\n",
982 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
983 			/* proceed anyway... */
984 		} else
985 			ia->ia_flags &= ~IFA_ROUTE;
986 		ia->ia_dstaddr = dst6;
987 	}
988 
989 	/*
990 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
991 	 * to see if the address is deprecated or invalidated, but initialize
992 	 * these members for applications.
993 	 */
994 	ia->ia6_lifetime = ifra->ifra_lifetime;
995 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
996 		ia->ia6_lifetime.ia6t_expire =
997 		    time_second + ia->ia6_lifetime.ia6t_vltime;
998 	} else
999 		ia->ia6_lifetime.ia6t_expire = 0;
1000 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1001 		ia->ia6_lifetime.ia6t_preferred =
1002 		    time_second + ia->ia6_lifetime.ia6t_pltime;
1003 	} else
1004 		ia->ia6_lifetime.ia6t_preferred = 0;
1005 
1006 	/* reset the interface and routing table appropriately. */
1007 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1008 		goto unlink;
1009 
1010 	/*
1011 	 * configure address flags.
1012 	 */
1013 	ia->ia6_flags = ifra->ifra_flags;
1014 	/*
1015 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1016 	 * userland, make it deprecated.
1017 	 */
1018 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1019 		ia->ia6_lifetime.ia6t_pltime = 0;
1020 		ia->ia6_lifetime.ia6t_preferred = time_second;
1021 	}
1022 
1023 	/*
1024 	 * Make the address tentative before joining multicast addresses,
1025 	 * so that corresponding MLD responses would not have a tentative
1026 	 * source address.
1027 	 */
1028 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1029 	if (hostIsNew && in6if_do_dad(ifp))
1030 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1031 
1032 	/*
1033 	 * We are done if we have simply modified an existing address.
1034 	 */
1035 	if (!hostIsNew)
1036 		return error;
1037 
1038 	/*
1039 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1040 	 * not just go to unlink.
1041 	 */
1042 
1043 	/* join necessary multicast groups */
1044 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1045 		struct sockaddr_in6 mltaddr, mltmask;
1046 		struct in6_addr llsol;
1047 
1048 		/* join solicited multicast addr for new host id */
1049 		bzero(&llsol, sizeof(struct in6_addr));
1050 		llsol.s6_addr16[0] = htons(0xff02);
1051 		llsol.s6_addr32[1] = 0;
1052 		llsol.s6_addr32[2] = htonl(1);
1053 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1054 		llsol.s6_addr8[12] = 0xff;
1055 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1056 			/* XXX: should not happen */
1057 			log(LOG_ERR, "in6_update_ifa: "
1058 			    "in6_setscope failed\n");
1059 			goto cleanup;
1060 		}
1061 		dad_delay = 0;
1062 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1063 			/*
1064 			 * We need a random delay for DAD on the address
1065 			 * being configured.  It also means delaying
1066 			 * transmission of the corresponding MLD report to
1067 			 * avoid report collision.
1068 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1069 			 */
1070 			dad_delay = arc4random() %
1071 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1072 		}
1073 
1074 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1075 		/* join solicited multicast addr for new host id */
1076 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
1077 		if (!imm) {
1078 			nd6log((LOG_ERR,
1079 			    "in6_update_ifa: addmulti "
1080 			    "failed for %s on %s (errno=%d)\n",
1081 			    ip6_sprintf(&llsol), if_name(ifp), error));
1082 			goto cleanup;
1083 		}
1084 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1085 		in6m_sol = imm->i6mm_maddr;
1086 
1087 		sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
1088 
1089 		/*
1090 		 * join link-local all-nodes address
1091 		 */
1092 		sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
1093 		    0, 0, 0);
1094 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1095 			goto cleanup; /* XXX: should not fail */
1096 
1097 		/*
1098 		 * XXX: do we really need this automatic routes?
1099 		 * We should probably reconsider this stuff.  Most applications
1100 		 * actually do not need the routes, since they usually specify
1101 		 * the outgoing interface.
1102 		 */
1103 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1104 		if (rt) {
1105 			if (memcmp(&mltaddr.sin6_addr,
1106 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1107 			    MLTMASK_LEN)) {
1108 				RTFREE(rt);
1109 				rt = NULL;
1110 			} else if (rt->rt_ifp != ifp) {
1111 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1112 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1113 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1114 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1115 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1116 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1117 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1118 				rt_replace_ifa(rt, &ia->ia_ifa);
1119 				rt->rt_ifp = ifp;
1120 			}
1121 		}
1122 		if (!rt) {
1123 			struct rt_addrinfo info;
1124 
1125 			bzero(&info, sizeof(info));
1126 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1127 			info.rti_info[RTAX_GATEWAY] =
1128 			    (struct sockaddr *)&ia->ia_addr;
1129 			info.rti_info[RTAX_NETMASK] =
1130 			    (struct sockaddr *)&mltmask;
1131 			info.rti_info[RTAX_IFA] =
1132 			    (struct sockaddr *)&ia->ia_addr;
1133 			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1134 			info.rti_flags = RTF_UP | RTF_CLONING;
1135 			error = rtrequest1(RTM_ADD, &info, NULL);
1136 			if (error)
1137 				goto cleanup;
1138 		} else {
1139 			RTFREE(rt);
1140 		}
1141 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1142 		if (!imm) {
1143 			nd6log((LOG_WARNING,
1144 			    "in6_update_ifa: addmulti failed for "
1145 			    "%s on %s (errno=%d)\n",
1146 			    ip6_sprintf(&mltaddr.sin6_addr),
1147 			    if_name(ifp), error));
1148 			goto cleanup;
1149 		}
1150 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1151 
1152 		/*
1153 		 * join node information group address
1154 		 */
1155 		dad_delay = 0;
1156 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1157 			/*
1158 			 * The spec doesn't say anything about delay for this
1159 			 * group, but the same logic should apply.
1160 			 */
1161 			dad_delay = arc4random() %
1162 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1163 		}
1164 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
1165 			;
1166 		else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1167 		          dad_delay)) == NULL) { /* XXX jinmei */
1168 			nd6log((LOG_WARNING, "in6_update_ifa: "
1169 			    "addmulti failed for %s on %s (errno=%d)\n",
1170 			    ip6_sprintf(&mltaddr.sin6_addr),
1171 			    if_name(ifp), error));
1172 			/* XXX not very fatal, go on... */
1173 		} else {
1174 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1175 		}
1176 
1177 
1178 		/*
1179 		 * join interface-local all-nodes address.
1180 		 * (ff01::1%ifN, and ff01::%ifN/32)
1181 		 */
1182 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1183 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1184 			goto cleanup; /* XXX: should not fail */
1185 
1186 		/* XXX: again, do we really need the route? */
1187 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1188 		if (rt) {
1189 			/* 32bit came from "mltmask" */
1190 			if (memcmp(&mltaddr.sin6_addr,
1191 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1192 			    32 / NBBY)) {
1193 				RTFREE(rt);
1194 				rt = NULL;
1195 			} else if (rt->rt_ifp != ifp) {
1196 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1197 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1198 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1199 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1200 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1201 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1202 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1203 				rt_replace_ifa(rt, &ia->ia_ifa);
1204 				rt->rt_ifp = ifp;
1205 			}
1206 		}
1207 		if (!rt) {
1208 			struct rt_addrinfo info;
1209 
1210 			bzero(&info, sizeof(info));
1211 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1212 			info.rti_info[RTAX_GATEWAY] =
1213 			    (struct sockaddr *)&ia->ia_addr;
1214 			info.rti_info[RTAX_NETMASK] =
1215 			    (struct sockaddr *)&mltmask;
1216 			info.rti_info[RTAX_IFA] =
1217 			    (struct sockaddr *)&ia->ia_addr;
1218 			info.rti_flags = RTF_UP | RTF_CLONING;
1219 			error = rtrequest1(RTM_ADD, &info, NULL);
1220 			if (error)
1221 				goto cleanup;
1222 #undef	MLTMASK_LEN
1223 		} else {
1224 			RTFREE(rt);
1225 		}
1226 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1227 		if (!imm) {
1228 			nd6log((LOG_WARNING, "in6_update_ifa: "
1229 			    "addmulti failed for %s on %s (errno=%d)\n",
1230 			    ip6_sprintf(&mltaddr.sin6_addr),
1231 			    if_name(ifp), error));
1232 			goto cleanup;
1233 		} else {
1234 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1235 		}
1236 	}
1237 
1238 	/*
1239 	 * Perform DAD, if needed.
1240 	 * XXX It may be of use, if we can administratively
1241 	 * disable DAD.
1242 	 */
1243 	if (hostIsNew && in6if_do_dad(ifp) &&
1244 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1245 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1246 	{
1247 		int mindelay, maxdelay;
1248 
1249 		dad_delay = 0;
1250 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1251 			/*
1252 			 * We need to impose a delay before sending an NS
1253 			 * for DAD.  Check if we also needed a delay for the
1254 			 * corresponding MLD message.  If we did, the delay
1255 			 * should be larger than the MLD delay (this could be
1256 			 * relaxed a bit, but this simple logic is at least
1257 			 * safe).
1258 			 */
1259 			mindelay = 0;
1260 			if (in6m_sol != NULL &&
1261 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1262 				mindelay = in6m_sol->in6m_timer;
1263 			}
1264 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1265 			if (maxdelay - mindelay == 0)
1266 				dad_delay = 0;
1267 			else {
1268 				dad_delay =
1269 				    (arc4random() % (maxdelay - mindelay)) +
1270 				    mindelay;
1271 			}
1272 		}
1273 		nd6_dad_start((struct ifaddr *)ia, dad_delay);
1274 	}
1275 
1276 	return error;
1277 
1278   unlink:
1279 	/*
1280 	 * XXX: if a change of an existing address failed, keep the entry
1281 	 * anyway.
1282 	 */
1283 	if (hostIsNew)
1284 		in6_unlink_ifa(ia, ifp);
1285 	return error;
1286 
1287   cleanup:
1288 	in6_purgeaddr(&ia->ia_ifa);
1289 	return error;
1290 }
1291 
1292 int
1293 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1294     struct in6_ifaddr *ia, int flags)
1295 {
1296 	int rc, s;
1297 
1298 	s = splnet();
1299 	rc = in6_update_ifa1(ifp, ifra, ia, flags);
1300 	splx(s);
1301 	return rc;
1302 }
1303 
1304 void
1305 in6_purgeaddr(struct ifaddr *ifa)
1306 {
1307 	struct ifnet *ifp = ifa->ifa_ifp;
1308 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1309 	struct in6_multi_mship *imm;
1310 
1311 	/* stop DAD processing */
1312 	nd6_dad_stop(ifa);
1313 
1314 	/*
1315 	 * delete route to the destination of the address being purged.
1316 	 * The interface must be p2p or loopback in this case.
1317 	 */
1318 	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1319 		int e;
1320 
1321 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1322 		    != 0) {
1323 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1324 			    "a route to the p2p destination: %s on %s, "
1325 			    "errno=%d\n",
1326 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1327 			    e);
1328 			/* proceed anyway... */
1329 		} else
1330 			ia->ia_flags &= ~IFA_ROUTE;
1331 	}
1332 
1333 	/* Remove ownaddr's loopback rtentry, if it exists. */
1334 	in6_ifremloop(&(ia->ia_ifa));
1335 
1336 	/*
1337 	 * leave from multicast groups we have joined for the interface
1338 	 */
1339 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1340 		LIST_REMOVE(imm, i6mm_chain);
1341 		in6_leavegroup(imm);
1342 	}
1343 
1344 	in6_unlink_ifa(ia, ifp);
1345 }
1346 
1347 static void
1348 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1349 {
1350 	struct in6_ifaddr *oia;
1351 	int	s = splnet();
1352 
1353 	ifa_remove(ifp, &ia->ia_ifa);
1354 
1355 	oia = ia;
1356 	if (oia == (ia = in6_ifaddr))
1357 		in6_ifaddr = ia->ia_next;
1358 	else {
1359 		while (ia->ia_next && (ia->ia_next != oia))
1360 			ia = ia->ia_next;
1361 		if (ia->ia_next)
1362 			ia->ia_next = oia->ia_next;
1363 		else {
1364 			/* search failed */
1365 			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1366 		}
1367 	}
1368 
1369 	/*
1370 	 * XXX thorpej@NetBSD.org -- if the interface is going
1371 	 * XXX away, don't save the multicast entries, delete them!
1372 	 */
1373 	if (LIST_EMPTY(&oia->ia6_multiaddrs))
1374 		;
1375 	else if (oia->ia_ifa.ifa_ifp->if_output == if_nulloutput) {
1376 		struct in6_multi *in6m, *next;
1377 
1378 		for (in6m = LIST_FIRST(&oia->ia6_multiaddrs); in6m != NULL;
1379 		     in6m = next) {
1380 			next = LIST_NEXT(in6m, in6m_entry);
1381 			in6_delmulti(in6m);
1382 		}
1383 	} else
1384 		in6_savemkludge(oia);
1385 
1386 	/*
1387 	 * Release the reference to the base prefix.  There should be a
1388 	 * positive reference.
1389 	 */
1390 	if (oia->ia6_ndpr == NULL) {
1391 		nd6log((LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1392 		    "%p has no prefix\n", oia));
1393 	} else {
1394 		oia->ia6_ndpr->ndpr_refcnt--;
1395 		oia->ia6_ndpr = NULL;
1396 	}
1397 
1398 	/*
1399 	 * Also, if the address being removed is autoconf'ed, call
1400 	 * pfxlist_onlink_check() since the release might affect the status of
1401 	 * other (detached) addresses.
1402 	 */
1403 	if ((oia->ia6_flags & IN6_IFF_AUTOCONF) != 0)
1404 		pfxlist_onlink_check();
1405 
1406 	/*
1407 	 * release another refcnt for the link from in6_ifaddr.
1408 	 * Note that we should decrement the refcnt at least once for all *BSD.
1409 	 */
1410 	IFAFREE(&oia->ia_ifa);
1411 
1412 	splx(s);
1413 }
1414 
1415 void
1416 in6_purgeif(struct ifnet *ifp)
1417 {
1418 	if_purgeaddrs(ifp, AF_INET6, in6_purgeaddr);
1419 
1420 	in6_ifdetach(ifp);
1421 }
1422 
1423 /*
1424  * SIOC[GAD]LIFADDR.
1425  *	SIOCGLIFADDR: get first address. (?)
1426  *	SIOCGLIFADDR with IFLR_PREFIX:
1427  *		get first address that matches the specified prefix.
1428  *	SIOCALIFADDR: add the specified address.
1429  *	SIOCALIFADDR with IFLR_PREFIX:
1430  *		add the specified prefix, filling hostid part from
1431  *		the first link-local address.  prefixlen must be <= 64.
1432  *	SIOCDLIFADDR: delete the specified address.
1433  *	SIOCDLIFADDR with IFLR_PREFIX:
1434  *		delete the first address that matches the specified prefix.
1435  * return values:
1436  *	EINVAL on invalid parameters
1437  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1438  *	other values may be returned from in6_ioctl()
1439  *
1440  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1441  * this is to accommodate address naming scheme other than RFC2374,
1442  * in the future.
1443  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1444  * address encoding scheme. (see figure on page 8)
1445  */
1446 static int
1447 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1448 	struct ifnet *ifp, struct lwp *l)
1449 {
1450 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1451 	struct ifaddr *ifa;
1452 	struct sockaddr *sa;
1453 
1454 	/* sanity checks */
1455 	if (!data || !ifp) {
1456 		panic("invalid argument to in6_lifaddr_ioctl");
1457 		/* NOTREACHED */
1458 	}
1459 
1460 	switch (cmd) {
1461 	case SIOCGLIFADDR:
1462 		/* address must be specified on GET with IFLR_PREFIX */
1463 		if ((iflr->flags & IFLR_PREFIX) == 0)
1464 			break;
1465 		/* FALLTHROUGH */
1466 	case SIOCALIFADDR:
1467 	case SIOCDLIFADDR:
1468 		/* address must be specified on ADD and DELETE */
1469 		sa = (struct sockaddr *)&iflr->addr;
1470 		if (sa->sa_family != AF_INET6)
1471 			return EINVAL;
1472 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1473 			return EINVAL;
1474 		/* XXX need improvement */
1475 		sa = (struct sockaddr *)&iflr->dstaddr;
1476 		if (sa->sa_family && sa->sa_family != AF_INET6)
1477 			return EINVAL;
1478 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1479 			return EINVAL;
1480 		break;
1481 	default: /* shouldn't happen */
1482 #if 0
1483 		panic("invalid cmd to in6_lifaddr_ioctl");
1484 		/* NOTREACHED */
1485 #else
1486 		return EOPNOTSUPP;
1487 #endif
1488 	}
1489 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1490 		return EINVAL;
1491 
1492 	switch (cmd) {
1493 	case SIOCALIFADDR:
1494 	    {
1495 		struct in6_aliasreq ifra;
1496 		struct in6_addr *xhostid = NULL;
1497 		int prefixlen;
1498 
1499 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1500 			struct sockaddr_in6 *sin6;
1501 
1502 			/*
1503 			 * xhostid is to fill in the hostid part of the
1504 			 * address.  xhostid points to the first link-local
1505 			 * address attached to the interface.
1506 			 */
1507 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1508 			if (!ifa)
1509 				return EADDRNOTAVAIL;
1510 			xhostid = IFA_IN6(ifa);
1511 
1512 		 	/* prefixlen must be <= 64. */
1513 			if (64 < iflr->prefixlen)
1514 				return EINVAL;
1515 			prefixlen = iflr->prefixlen;
1516 
1517 			/* hostid part must be zero. */
1518 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1519 			if (sin6->sin6_addr.s6_addr32[2] != 0
1520 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1521 				return EINVAL;
1522 			}
1523 		} else
1524 			prefixlen = iflr->prefixlen;
1525 
1526 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1527 		bzero(&ifra, sizeof(ifra));
1528 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1529 
1530 		bcopy(&iflr->addr, &ifra.ifra_addr,
1531 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1532 		if (xhostid) {
1533 			/* fill in hostid part */
1534 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1535 			    xhostid->s6_addr32[2];
1536 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1537 			    xhostid->s6_addr32[3];
1538 		}
1539 
1540 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1541 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1542 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1543 			if (xhostid) {
1544 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1545 				    xhostid->s6_addr32[2];
1546 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1547 				    xhostid->s6_addr32[3];
1548 			}
1549 		}
1550 
1551 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1552 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1553 
1554 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1555 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1556 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1557 		return in6_control(so, SIOCAIFADDR_IN6, (void *)&ifra, ifp, l);
1558 	    }
1559 	case SIOCGLIFADDR:
1560 	case SIOCDLIFADDR:
1561 	    {
1562 		struct in6_ifaddr *ia;
1563 		struct in6_addr mask, candidate, match;
1564 		struct sockaddr_in6 *sin6;
1565 		int cmp;
1566 
1567 		bzero(&mask, sizeof(mask));
1568 		if (iflr->flags & IFLR_PREFIX) {
1569 			/* lookup a prefix rather than address. */
1570 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1571 
1572 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1573 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1574 			match.s6_addr32[0] &= mask.s6_addr32[0];
1575 			match.s6_addr32[1] &= mask.s6_addr32[1];
1576 			match.s6_addr32[2] &= mask.s6_addr32[2];
1577 			match.s6_addr32[3] &= mask.s6_addr32[3];
1578 
1579 			/* if you set extra bits, that's wrong */
1580 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1581 				return EINVAL;
1582 
1583 			cmp = 1;
1584 		} else {
1585 			if (cmd == SIOCGLIFADDR) {
1586 				/* on getting an address, take the 1st match */
1587 				cmp = 0;	/* XXX */
1588 			} else {
1589 				/* on deleting an address, do exact match */
1590 				in6_prefixlen2mask(&mask, 128);
1591 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1592 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1593 
1594 				cmp = 1;
1595 			}
1596 		}
1597 
1598 		IFADDR_FOREACH(ifa, ifp) {
1599 			if (ifa->ifa_addr->sa_family != AF_INET6)
1600 				continue;
1601 			if (!cmp)
1602 				break;
1603 
1604 			/*
1605 			 * XXX: this is adhoc, but is necessary to allow
1606 			 * a user to specify fe80::/64 (not /10) for a
1607 			 * link-local address.
1608 			 */
1609 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1610 			in6_clearscope(&candidate);
1611 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1612 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1613 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1614 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1615 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1616 				break;
1617 		}
1618 		if (!ifa)
1619 			return EADDRNOTAVAIL;
1620 		ia = ifa2ia6(ifa);
1621 
1622 		if (cmd == SIOCGLIFADDR) {
1623 			int error;
1624 
1625 			/* fill in the if_laddrreq structure */
1626 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1627 			error = sa6_recoverscope(
1628 			    (struct sockaddr_in6 *)&iflr->addr);
1629 			if (error != 0)
1630 				return error;
1631 
1632 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1633 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1634 				    ia->ia_dstaddr.sin6_len);
1635 				error = sa6_recoverscope(
1636 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1637 				if (error != 0)
1638 					return error;
1639 			} else
1640 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1641 
1642 			iflr->prefixlen =
1643 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1644 
1645 			iflr->flags = ia->ia6_flags;	/* XXX */
1646 
1647 			return 0;
1648 		} else {
1649 			struct in6_aliasreq ifra;
1650 
1651 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1652 			bzero(&ifra, sizeof(ifra));
1653 			bcopy(iflr->iflr_name, ifra.ifra_name,
1654 			    sizeof(ifra.ifra_name));
1655 
1656 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1657 			    ia->ia_addr.sin6_len);
1658 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1659 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1660 				    ia->ia_dstaddr.sin6_len);
1661 			} else {
1662 				bzero(&ifra.ifra_dstaddr,
1663 				    sizeof(ifra.ifra_dstaddr));
1664 			}
1665 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1666 			    ia->ia_prefixmask.sin6_len);
1667 
1668 			ifra.ifra_flags = ia->ia6_flags;
1669 			return in6_control(so, SIOCDIFADDR_IN6, (void *)&ifra,
1670 			    ifp, l);
1671 		}
1672 	    }
1673 	}
1674 
1675 	return EOPNOTSUPP;	/* just for safety */
1676 }
1677 
1678 /*
1679  * Initialize an interface's internet6 address
1680  * and routing table entry.
1681  */
1682 static int
1683 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1684 	struct sockaddr_in6 *sin6, int newhost)
1685 {
1686 	int	error = 0, plen, ifacount = 0;
1687 	int	s = splnet();
1688 	struct ifaddr *ifa;
1689 
1690 	/*
1691 	 * Give the interface a chance to initialize
1692 	 * if this is its first address,
1693 	 * and to validate the address if necessary.
1694 	 */
1695 	IFADDR_FOREACH(ifa, ifp) {
1696 		if (ifa->ifa_addr == NULL)
1697 			continue;	/* just for safety */
1698 		if (ifa->ifa_addr->sa_family != AF_INET6)
1699 			continue;
1700 		ifacount++;
1701 	}
1702 
1703 	ia->ia_addr = *sin6;
1704 
1705 	if (ifacount <= 1 &&
1706 	    (error = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ia)) != 0) {
1707 		splx(s);
1708 		return error;
1709 	}
1710 	splx(s);
1711 
1712 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1713 
1714 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1715 
1716 	/*
1717 	 * Special case:
1718 	 * If the destination address is specified for a point-to-point
1719 	 * interface, install a route to the destination as an interface
1720 	 * direct route.
1721 	 */
1722 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1723 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1724 		if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1725 				    RTF_UP | RTF_HOST)) != 0)
1726 			return error;
1727 		ia->ia_flags |= IFA_ROUTE;
1728 	}
1729 
1730 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1731 	if (newhost) {
1732 		/* set the rtrequest function to create llinfo */
1733 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1734 		in6_ifaddloop(&(ia->ia_ifa));
1735 	}
1736 
1737 	if (ifp->if_flags & IFF_MULTICAST)
1738 		in6_restoremkludge(ia, ifp);
1739 
1740 	return error;
1741 }
1742 
1743 /*
1744  * Find an IPv6 interface link-local address specific to an interface.
1745  */
1746 struct in6_ifaddr *
1747 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1748 {
1749 	struct ifaddr *ifa;
1750 
1751 	IFADDR_FOREACH(ifa, ifp) {
1752 		if (ifa->ifa_addr == NULL)
1753 			continue;	/* just for safety */
1754 		if (ifa->ifa_addr->sa_family != AF_INET6)
1755 			continue;
1756 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1757 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1758 			     ignoreflags) != 0)
1759 				continue;
1760 			break;
1761 		}
1762 	}
1763 
1764 	return (struct in6_ifaddr *)ifa;
1765 }
1766 
1767 
1768 /*
1769  * find the internet address corresponding to a given interface and address.
1770  */
1771 struct in6_ifaddr *
1772 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1773 {
1774 	struct ifaddr *ifa;
1775 
1776 	IFADDR_FOREACH(ifa, ifp) {
1777 		if (ifa->ifa_addr == NULL)
1778 			continue;	/* just for safety */
1779 		if (ifa->ifa_addr->sa_family != AF_INET6)
1780 			continue;
1781 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1782 			break;
1783 	}
1784 
1785 	return (struct in6_ifaddr *)ifa;
1786 }
1787 
1788 /*
1789  * find the internet address on a given interface corresponding to a neighbor's
1790  * address.
1791  */
1792 struct in6_ifaddr *
1793 in6ifa_ifplocaladdr(const struct ifnet *ifp, const struct in6_addr *addr)
1794 {
1795 	struct ifaddr *ifa;
1796 	struct in6_ifaddr *ia;
1797 
1798 	IFADDR_FOREACH(ifa, ifp) {
1799 		if (ifa->ifa_addr == NULL)
1800 			continue;	/* just for safety */
1801 		if (ifa->ifa_addr->sa_family != AF_INET6)
1802 			continue;
1803 		ia = (struct in6_ifaddr *)ifa;
1804 		if (IN6_ARE_MASKED_ADDR_EQUAL(addr,
1805 				&ia->ia_addr.sin6_addr,
1806 				&ia->ia_prefixmask.sin6_addr))
1807 			return ia;
1808 	}
1809 
1810 	return NULL;
1811 }
1812 
1813 /*
1814  * Convert IP6 address to printable (loggable) representation.
1815  */
1816 static int ip6round = 0;
1817 char *
1818 ip6_sprintf(const struct in6_addr *addr)
1819 {
1820 	static char ip6buf[8][48];
1821 	int i;
1822 	char *cp;
1823 	const u_int16_t *a = (const u_int16_t *)addr;
1824 	const u_int8_t *d;
1825 	int dcolon = 0;
1826 
1827 	ip6round = (ip6round + 1) & 7;
1828 	cp = ip6buf[ip6round];
1829 
1830 	for (i = 0; i < 8; i++) {
1831 		if (dcolon == 1) {
1832 			if (*a == 0) {
1833 				if (i == 7)
1834 					*cp++ = ':';
1835 				a++;
1836 				continue;
1837 			} else
1838 				dcolon = 2;
1839 		}
1840 		if (*a == 0) {
1841 			if (dcolon == 0 && *(a + 1) == 0) {
1842 				if (i == 0)
1843 					*cp++ = ':';
1844 				*cp++ = ':';
1845 				dcolon = 1;
1846 			} else {
1847 				*cp++ = '0';
1848 				*cp++ = ':';
1849 			}
1850 			a++;
1851 			continue;
1852 		}
1853 		d = (const u_char *)a;
1854 		*cp++ = hexdigits[*d >> 4];
1855 		*cp++ = hexdigits[*d++ & 0xf];
1856 		*cp++ = hexdigits[*d >> 4];
1857 		*cp++ = hexdigits[*d & 0xf];
1858 		*cp++ = ':';
1859 		a++;
1860 	}
1861 	*--cp = 0;
1862 	return ip6buf[ip6round];
1863 }
1864 
1865 /*
1866  * Determine if an address is on a local network.
1867  */
1868 int
1869 in6_localaddr(const struct in6_addr *in6)
1870 {
1871 	struct in6_ifaddr *ia;
1872 
1873 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1874 		return 1;
1875 
1876 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1877 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1878 					      &ia->ia_prefixmask.sin6_addr))
1879 			return 1;
1880 
1881 	return 0;
1882 }
1883 
1884 int
1885 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1886 {
1887 	struct in6_ifaddr *ia;
1888 
1889 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1890 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1891 		    &sa6->sin6_addr) &&
1892 #ifdef SCOPEDROUTING
1893 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1894 #endif
1895 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1896 			return 1; /* true */
1897 
1898 		/* XXX: do we still have to go thru the rest of the list? */
1899 	}
1900 
1901 	return 0;		/* false */
1902 }
1903 
1904 /*
1905  * return length of part which dst and src are equal
1906  * hard coding...
1907  */
1908 int
1909 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1910 {
1911 	int match = 0;
1912 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1913 	u_char *lim = s + 16, r;
1914 
1915 	while (s < lim)
1916 		if ((r = (*d++ ^ *s++)) != 0) {
1917 			while (r < 128) {
1918 				match++;
1919 				r <<= 1;
1920 			}
1921 			break;
1922 		} else
1923 			match += NBBY;
1924 	return match;
1925 }
1926 
1927 /* XXX: to be scope conscious */
1928 int
1929 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1930 {
1931 	int bytelen, bitlen;
1932 
1933 	/* sanity check */
1934 	if (len < 0 || len > 128) {
1935 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1936 		    len);
1937 		return 0;
1938 	}
1939 
1940 	bytelen = len / NBBY;
1941 	bitlen = len % NBBY;
1942 
1943 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1944 		return 0;
1945 	if (bitlen != 0 &&
1946 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
1947 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
1948 		return 0;
1949 
1950 	return 1;
1951 }
1952 
1953 void
1954 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1955 {
1956 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1957 	int bytelen, bitlen, i;
1958 
1959 	/* sanity check */
1960 	if (len < 0 || len > 128) {
1961 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1962 		    len);
1963 		return;
1964 	}
1965 
1966 	bzero(maskp, sizeof(*maskp));
1967 	bytelen = len / NBBY;
1968 	bitlen = len % NBBY;
1969 	for (i = 0; i < bytelen; i++)
1970 		maskp->s6_addr[i] = 0xff;
1971 	if (bitlen)
1972 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1973 }
1974 
1975 /*
1976  * return the best address out of the same scope. if no address was
1977  * found, return the first valid address from designated IF.
1978  */
1979 struct in6_ifaddr *
1980 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1981 {
1982 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1983 	struct ifaddr *ifa;
1984 	struct in6_ifaddr *besta = 0;
1985 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1986 
1987 	dep[0] = dep[1] = NULL;
1988 
1989 	/*
1990 	 * We first look for addresses in the same scope.
1991 	 * If there is one, return it.
1992 	 * If two or more, return one which matches the dst longest.
1993 	 * If none, return one of global addresses assigned other ifs.
1994 	 */
1995 	IFADDR_FOREACH(ifa, ifp) {
1996 		if (ifa->ifa_addr->sa_family != AF_INET6)
1997 			continue;
1998 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1999 			continue; /* XXX: is there any case to allow anycast? */
2000 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2001 			continue; /* don't use this interface */
2002 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2003 			continue;
2004 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2005 			if (ip6_use_deprecated)
2006 				dep[0] = (struct in6_ifaddr *)ifa;
2007 			continue;
2008 		}
2009 
2010 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2011 			/*
2012 			 * call in6_matchlen() as few as possible
2013 			 */
2014 			if (besta) {
2015 				if (blen == -1)
2016 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2017 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2018 				if (tlen > blen) {
2019 					blen = tlen;
2020 					besta = (struct in6_ifaddr *)ifa;
2021 				}
2022 			} else
2023 				besta = (struct in6_ifaddr *)ifa;
2024 		}
2025 	}
2026 	if (besta)
2027 		return besta;
2028 
2029 	IFADDR_FOREACH(ifa, ifp) {
2030 		if (ifa->ifa_addr->sa_family != AF_INET6)
2031 			continue;
2032 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2033 			continue; /* XXX: is there any case to allow anycast? */
2034 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2035 			continue; /* don't use this interface */
2036 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2037 			continue;
2038 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2039 			if (ip6_use_deprecated)
2040 				dep[1] = (struct in6_ifaddr *)ifa;
2041 			continue;
2042 		}
2043 
2044 		return (struct in6_ifaddr *)ifa;
2045 	}
2046 
2047 	/* use the last-resort values, that are, deprecated addresses */
2048 	if (dep[0])
2049 		return dep[0];
2050 	if (dep[1])
2051 		return dep[1];
2052 
2053 	return NULL;
2054 }
2055 
2056 /*
2057  * perform DAD when interface becomes IFF_UP.
2058  */
2059 void
2060 in6_if_up(struct ifnet *ifp)
2061 {
2062 	struct ifaddr *ifa;
2063 	struct in6_ifaddr *ia;
2064 
2065 	IFADDR_FOREACH(ifa, ifp) {
2066 		if (ifa->ifa_addr->sa_family != AF_INET6)
2067 			continue;
2068 		ia = (struct in6_ifaddr *)ifa;
2069 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2070 			/*
2071 			 * The TENTATIVE flag was likely set by hand
2072 			 * beforehand, implicitly indicating the need for DAD.
2073 			 * We may be able to skip the random delay in this
2074 			 * case, but we impose delays just in case.
2075 			 */
2076 			nd6_dad_start(ifa,
2077 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2078 		}
2079 	}
2080 
2081 	/*
2082 	 * special cases, like 6to4, are handled in in6_ifattach
2083 	 */
2084 	in6_ifattach(ifp, NULL);
2085 }
2086 
2087 int
2088 in6if_do_dad(struct ifnet *ifp)
2089 {
2090 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2091 		return 0;
2092 
2093 	switch (ifp->if_type) {
2094 	case IFT_FAITH:
2095 		/*
2096 		 * These interfaces do not have the IFF_LOOPBACK flag,
2097 		 * but loop packets back.  We do not have to do DAD on such
2098 		 * interfaces.  We should even omit it, because loop-backed
2099 		 * NS would confuse the DAD procedure.
2100 		 */
2101 		return 0;
2102 	default:
2103 		/*
2104 		 * Our DAD routine requires the interface up and running.
2105 		 * However, some interfaces can be up before the RUNNING
2106 		 * status.  Additionaly, users may try to assign addresses
2107 		 * before the interface becomes up (or running).
2108 		 * We simply skip DAD in such a case as a work around.
2109 		 * XXX: we should rather mark "tentative" on such addresses,
2110 		 * and do DAD after the interface becomes ready.
2111 		 */
2112 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2113 		    (IFF_UP|IFF_RUNNING))
2114 			return 0;
2115 
2116 		return 1;
2117 	}
2118 }
2119 
2120 /*
2121  * Calculate max IPv6 MTU through all the interfaces and store it
2122  * to in6_maxmtu.
2123  */
2124 void
2125 in6_setmaxmtu(void)
2126 {
2127 	unsigned long maxmtu = 0;
2128 	struct ifnet *ifp;
2129 
2130 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
2131 		/* this function can be called during ifnet initialization */
2132 		if (!ifp->if_afdata[AF_INET6])
2133 			continue;
2134 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2135 		    IN6_LINKMTU(ifp) > maxmtu)
2136 			maxmtu = IN6_LINKMTU(ifp);
2137 	}
2138 	if (maxmtu)	     /* update only when maxmtu is positive */
2139 		in6_maxmtu = maxmtu;
2140 }
2141 
2142 /*
2143  * Provide the length of interface identifiers to be used for the link attached
2144  * to the given interface.  The length should be defined in "IPv6 over
2145  * xxx-link" document.  Note that address architecture might also define
2146  * the length for a particular set of address prefixes, regardless of the
2147  * link type.  As clarified in rfc2462bis, those two definitions should be
2148  * consistent, and those really are as of August 2004.
2149  */
2150 int
2151 in6_if2idlen(struct ifnet *ifp)
2152 {
2153 	switch (ifp->if_type) {
2154 	case IFT_ETHER:		/* RFC2464 */
2155 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2156 	case IFT_L2VLAN:	/* ditto */
2157 	case IFT_IEEE80211:	/* ditto */
2158 	case IFT_FDDI:		/* RFC2467 */
2159 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2160 	case IFT_PPP:		/* RFC2472 */
2161 	case IFT_ARCNET:	/* RFC2497 */
2162 	case IFT_FRELAY:	/* RFC2590 */
2163 	case IFT_IEEE1394:	/* RFC3146 */
2164 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2165 	case IFT_LOOP:		/* XXX: is this really correct? */
2166 		return 64;
2167 	default:
2168 		/*
2169 		 * Unknown link type:
2170 		 * It might be controversial to use the today's common constant
2171 		 * of 64 for these cases unconditionally.  For full compliance,
2172 		 * we should return an error in this case.  On the other hand,
2173 		 * if we simply miss the standard for the link type or a new
2174 		 * standard is defined for a new link type, the IFID length
2175 		 * is very likely to be the common constant.  As a compromise,
2176 		 * we always use the constant, but make an explicit notice
2177 		 * indicating the "unknown" case.
2178 		 */
2179 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2180 		return 64;
2181 	}
2182 }
2183 
2184 void *
2185 in6_domifattach(struct ifnet *ifp)
2186 {
2187 	struct in6_ifextra *ext;
2188 
2189 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2190 
2191 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2192 	    M_IFADDR, M_WAITOK|M_ZERO);
2193 
2194 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2195 	    M_IFADDR, M_WAITOK|M_ZERO);
2196 
2197 	ext->nd_ifinfo = nd6_ifattach(ifp);
2198 	ext->scope6_id = scope6_ifattach(ifp);
2199 	return ext;
2200 }
2201 
2202 void
2203 in6_domifdetach(struct ifnet *ifp, void *aux)
2204 {
2205 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2206 
2207 	nd6_ifdetach(ext->nd_ifinfo);
2208 	free(ext->in6_ifstat, M_IFADDR);
2209 	free(ext->icmp6_ifstat, M_IFADDR);
2210 	scope6_ifdetach(ext->scope6_id);
2211 	free(ext, M_IFADDR);
2212 }
2213 
2214 /*
2215  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2216  * v4 mapped addr or v4 compat addr
2217  */
2218 void
2219 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2220 {
2221 	bzero(sin, sizeof(*sin));
2222 	sin->sin_len = sizeof(struct sockaddr_in);
2223 	sin->sin_family = AF_INET;
2224 	sin->sin_port = sin6->sin6_port;
2225 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2226 }
2227 
2228 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2229 void
2230 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2231 {
2232 	bzero(sin6, sizeof(*sin6));
2233 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2234 	sin6->sin6_family = AF_INET6;
2235 	sin6->sin6_port = sin->sin_port;
2236 	sin6->sin6_addr.s6_addr32[0] = 0;
2237 	sin6->sin6_addr.s6_addr32[1] = 0;
2238 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2239 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2240 }
2241 
2242 /* Convert sockaddr_in6 into sockaddr_in. */
2243 void
2244 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2245 {
2246 	struct sockaddr_in *sin_p;
2247 	struct sockaddr_in6 sin6;
2248 
2249 	/*
2250 	 * Save original sockaddr_in6 addr and convert it
2251 	 * to sockaddr_in.
2252 	 */
2253 	sin6 = *(struct sockaddr_in6 *)nam;
2254 	sin_p = (struct sockaddr_in *)nam;
2255 	in6_sin6_2_sin(sin_p, &sin6);
2256 }
2257 
2258 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2259 void
2260 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2261 {
2262 	struct sockaddr_in *sin_p;
2263 	struct sockaddr_in6 *sin6_p;
2264 
2265 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2266 	sin_p = (struct sockaddr_in *)*nam;
2267 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2268 	free(*nam, M_SONAME);
2269 	*nam = (struct sockaddr *)sin6_p;
2270 }
2271