xref: /netbsd-src/sys/netinet6/in6.c (revision f3cfa6f6ce31685c6c4a758bc430e69eb99f50a4)
1 /*	$NetBSD: in6.c,v 1.275 2019/04/29 11:57:22 roy 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 1.275 2019/04/29 11:57:22 roy Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_compat_netbsd.h"
70 #include "opt_net_mpsafe.h"
71 #endif
72 
73 #include <sys/param.h>
74 #include <sys/ioctl.h>
75 #include <sys/errno.h>
76 #include <sys/malloc.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/sockio.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/syslog.h>
85 #include <sys/kauth.h>
86 #include <sys/cprng.h>
87 #include <sys/kmem.h>
88 
89 #include <net/if.h>
90 #include <net/if_types.h>
91 #include <net/if_llatbl.h>
92 #include <net/if_ether.h>
93 #include <net/if_dl.h>
94 #include <net/pfil.h>
95 #include <net/route.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.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 
108 #ifdef COMPAT_50
109 #include <compat/netinet6/in6_var.h>
110 #endif
111 
112 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
113 
114 /* enable backward compatibility code for obsoleted ioctls */
115 #define COMPAT_IN6IFIOCTL
116 
117 #ifdef	IN6_DEBUG
118 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
119 #else
120 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
121 #endif /* IN6_DEBUG */
122 
123 /*
124  * Definitions of some constant IP6 addresses.
125  */
126 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
127 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
128 const struct in6_addr in6addr_nodelocal_allnodes =
129 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
130 const struct in6_addr in6addr_linklocal_allnodes =
131 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
132 const struct in6_addr in6addr_linklocal_allrouters =
133 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
134 
135 const struct in6_addr in6mask0 = IN6MASK0;
136 const struct in6_addr in6mask32 = IN6MASK32;
137 const struct in6_addr in6mask64 = IN6MASK64;
138 const struct in6_addr in6mask96 = IN6MASK96;
139 const struct in6_addr in6mask128 = IN6MASK128;
140 
141 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
142 				     0, 0, IN6ADDR_ANY_INIT, 0};
143 
144 struct pslist_head	in6_ifaddr_list;
145 kmutex_t		in6_ifaddr_lock;
146 
147 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
148 	struct ifnet *);
149 static int in6_ifaddprefix(struct in6_ifaddr *);
150 static int in6_ifremprefix(struct in6_ifaddr *);
151 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
152 	const struct sockaddr_in6 *, int);
153 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
154 static int in6_update_ifa1(struct ifnet *, struct in6_aliasreq *,
155     struct in6_ifaddr **, struct psref *, int);
156 
157 void
158 in6_init(void)
159 {
160 
161 	PSLIST_INIT(&in6_ifaddr_list);
162 	mutex_init(&in6_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
163 
164 	in6_sysctl_multicast_setup(NULL);
165 }
166 
167 /*
168  * Add ownaddr as loopback rtentry.  We previously add the route only if
169  * necessary (ex. on a p2p link).  However, since we now manage addresses
170  * separately from prefixes, we should always add the route.  We can't
171  * rely on the cloning mechanism from the corresponding interface route
172  * any more.
173  */
174 void
175 in6_ifaddlocal(struct ifaddr *ifa)
176 {
177 
178 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
179 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
180 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
181 	{
182 		rt_addrmsg(RTM_NEWADDR, ifa);
183 		return;
184 	}
185 
186 	rt_ifa_addlocal(ifa);
187 }
188 
189 /*
190  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
191  * if it exists.
192  */
193 void
194 in6_ifremlocal(struct ifaddr *ifa)
195 {
196 	struct in6_ifaddr *ia;
197 	struct ifaddr *alt_ifa = NULL;
198 	int ia_count = 0;
199 	struct psref psref;
200 	int s;
201 
202 	/*
203 	 * Some of BSD variants do not remove cloned routes
204 	 * from an interface direct route, when removing the direct route
205 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
206 	 * cloned routes, they could fail to remove the cloned routes when
207 	 * we handle multple addresses that share a common prefix.
208 	 * So, we should remove the route corresponding to the deleted address.
209 	 */
210 
211 	/*
212 	 * Delete the entry only if exactly one ifaddr matches the
213 	 * address, ifa->ifa_addr.
214 	 *
215 	 * If more than one ifaddr matches, replace the ifaddr in
216 	 * the routing table, rt_ifa, with a different ifaddr than
217 	 * the one we are purging, ifa.  It is important to do
218 	 * this, or else the routing table can accumulate dangling
219 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
220 	 * which will lead to crashes, later.  (More than one ifaddr
221 	 * can match if we assign the same address to multiple---probably
222 	 * p2p---interfaces.)
223 	 *
224 	 * XXX An old comment at this place said, "we should avoid
225 	 * XXX such a configuration [i.e., interfaces with the same
226 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
227 	 * XXX agree, especially now that I have fixed the dangling
228 	 * XXX ifp-pointers bug.
229 	 */
230 	s = pserialize_read_enter();
231 	IN6_ADDRLIST_READER_FOREACH(ia) {
232 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
233 			continue;
234 		if (ia->ia_ifp != ifa->ifa_ifp)
235 			alt_ifa = &ia->ia_ifa;
236 		if (++ia_count > 1 && alt_ifa != NULL)
237 			break;
238 	}
239 	if (ia_count > 1 && alt_ifa != NULL)
240 		ifa_acquire(alt_ifa, &psref);
241 	pserialize_read_exit(s);
242 
243 	if (ia_count == 0)
244 		return;
245 
246 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
247 
248 	if (ia_count > 1 && alt_ifa != NULL)
249 		ifa_release(alt_ifa, &psref);
250 }
251 
252 /* Add prefix route for the network. */
253 static int
254 in6_ifaddprefix(struct in6_ifaddr *ia)
255 {
256 	int error, flags = 0;
257 
258 	if (in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) == 128) {
259 		if (ia->ia_dstaddr.sin6_family != AF_INET6)
260 			/* We don't need to install a host route. */
261 			return 0;
262 		flags |= RTF_HOST;
263 	}
264 
265 	/* Is this a connected route for neighbour discovery? */
266 	if (nd6_need_cache(ia->ia_ifp))
267 		flags |= RTF_CONNECTED;
268 
269 	if ((error = rtinit(&ia->ia_ifa, RTM_ADD, RTF_UP | flags)) == 0)
270 		ia->ia_flags |= IFA_ROUTE;
271 	else if (error == EEXIST)
272 		/* Existance of the route is not an error. */
273 		error = 0;
274 
275 	return error;
276 }
277 
278 /* Delete network prefix route if present.
279  * Re-add it to another address if the prefix matches. */
280 static int
281 in6_ifremprefix(struct in6_ifaddr *target)
282 {
283 	int error, s;
284 	struct in6_ifaddr *ia;
285 
286 	if ((target->ia_flags & IFA_ROUTE) == 0)
287 		return 0;
288 
289 	s = pserialize_read_enter();
290 	IN6_ADDRLIST_READER_FOREACH(ia) {
291 		if (target->ia_dstaddr.sin6_len) {
292 			if (ia->ia_dstaddr.sin6_len == 0 ||
293 			    !IN6_ARE_ADDR_EQUAL(&ia->ia_dstaddr.sin6_addr,
294 			    &target->ia_dstaddr.sin6_addr))
295 				continue;
296 		} else {
297 			if (!IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
298 			    &target->ia_addr.sin6_addr,
299 			    &target->ia_prefixmask.sin6_addr))
300 				continue;
301 		}
302 
303 		/*
304 		 * if we got a matching prefix route, move IFA_ROUTE to him
305 		 */
306 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
307 			struct psref psref;
308 			int bound = curlwp_bind();
309 
310 			ia6_acquire(ia, &psref);
311 			pserialize_read_exit(s);
312 
313 			rtinit(&target->ia_ifa, RTM_DELETE, 0);
314 			target->ia_flags &= ~IFA_ROUTE;
315 
316 			error = in6_ifaddprefix(ia);
317 
318 			ia6_release(ia, &psref);
319 			curlwp_bindx(bound);
320 
321 			return error;
322 		}
323 	}
324 	pserialize_read_exit(s);
325 
326 	/*
327 	 * noone seem to have prefix route.  remove it.
328 	 */
329 	rtinit(&target->ia_ifa, RTM_DELETE, 0);
330 	target->ia_flags &= ~IFA_ROUTE;
331 	return 0;
332 }
333 
334 int
335 in6_mask2len(struct in6_addr *mask, u_char *lim0)
336 {
337 	int x = 0, y;
338 	u_char *lim = lim0, *p;
339 
340 	/* ignore the scope_id part */
341 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
342 		lim = (u_char *)mask + sizeof(*mask);
343 	for (p = (u_char *)mask; p < lim; x++, p++) {
344 		if (*p != 0xff)
345 			break;
346 	}
347 	y = 0;
348 	if (p < lim) {
349 		for (y = 0; y < NBBY; y++) {
350 			if ((*p & (0x80 >> y)) == 0)
351 				break;
352 		}
353 	}
354 
355 	/*
356 	 * when the limit pointer is given, do a stricter check on the
357 	 * remaining bits.
358 	 */
359 	if (p < lim) {
360 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
361 			return -1;
362 		for (p = p + 1; p < lim; p++)
363 			if (*p != 0)
364 				return -1;
365 	}
366 
367 	return x * NBBY + y;
368 }
369 
370 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
371 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
372 
373 static int
374 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
375 {
376 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
377 	struct	in6_ifaddr *ia = NULL;
378 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
379 	struct sockaddr_in6 *sa6;
380 	int error, bound;
381 	struct psref psref;
382 
383 	switch (cmd) {
384 	case SIOCAADDRCTL_POLICY:
385 	case SIOCDADDRCTL_POLICY:
386 		/* Privileged. */
387 		return in6_src_ioctl(cmd, data);
388 	/*
389 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
390 	 */
391 	case SIOCSIFADDR:
392 	case SIOCSIFDSTADDR:
393 	case SIOCSIFBRDADDR:
394 	case SIOCSIFNETMASK:
395 		return EOPNOTSUPP;
396 	case SIOCGETSGCNT_IN6:
397 	case SIOCGETMIFCNT_IN6:
398 		return mrt6_ioctl(cmd, data);
399 	case SIOCGIFADDRPREF:
400 	case SIOCSIFADDRPREF:
401 		if (ifp == NULL)
402 			return EINVAL;
403 		return ifaddrpref_ioctl(so, cmd, data, ifp);
404 	}
405 
406 	if (ifp == NULL)
407 		return EOPNOTSUPP;
408 
409 	switch (cmd) {
410 	case SIOCSNDFLUSH_IN6:
411 	case SIOCSPFXFLUSH_IN6:
412 	case SIOCSRTRFLUSH_IN6:
413 	case SIOCSDEFIFACE_IN6:
414 	case SIOCSIFINFO_FLAGS:
415 	case SIOCSIFINFO_IN6:
416 		/* Privileged. */
417 		/* FALLTHROUGH */
418 	case OSIOCGIFINFO_IN6:
419 	case SIOCGIFINFO_IN6:
420 	case SIOCGDRLST_IN6:
421 	case SIOCGPRLST_IN6:
422 	case SIOCGNBRINFO_IN6:
423 	case SIOCGDEFIFACE_IN6:
424 		return nd6_ioctl(cmd, data, ifp);
425 	}
426 
427 	switch (cmd) {
428 	case SIOCSIFPREFIX_IN6:
429 	case SIOCDIFPREFIX_IN6:
430 	case SIOCAIFPREFIX_IN6:
431 	case SIOCCIFPREFIX_IN6:
432 	case SIOCSGIFPREFIX_IN6:
433 	case SIOCGIFPREFIX_IN6:
434 		log(LOG_NOTICE,
435 		    "prefix ioctls are now invalidated. "
436 		    "please use ifconfig.\n");
437 		return EOPNOTSUPP;
438 	}
439 
440 	switch (cmd) {
441 	case SIOCALIFADDR:
442 	case SIOCDLIFADDR:
443 		/* Privileged. */
444 		/* FALLTHROUGH */
445 	case SIOCGLIFADDR:
446 		return in6_lifaddr_ioctl(so, cmd, data, ifp);
447 	}
448 
449 	/*
450 	 * Find address for this interface, if it exists.
451 	 *
452 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
453 	 * only, and used the first interface address as the target of other
454 	 * operations (without checking ifra_addr).  This was because netinet
455 	 * code/API assumed at most 1 interface address per interface.
456 	 * Since IPv6 allows a node to assign multiple addresses
457 	 * on a single interface, we almost always look and check the
458 	 * presence of ifra_addr, and reject invalid ones here.
459 	 * It also decreases duplicated code among SIOC*_IN6 operations.
460 	 */
461 	switch (cmd) {
462 	case SIOCAIFADDR_IN6:
463 #ifdef OSIOCAIFADDR_IN6
464 	case OSIOCAIFADDR_IN6:
465 #endif
466 #ifdef OSIOCSIFPHYADDR_IN6
467 	case OSIOCSIFPHYADDR_IN6:
468 #endif
469 	case SIOCSIFPHYADDR_IN6:
470 		sa6 = &ifra->ifra_addr;
471 		break;
472 	case SIOCSIFADDR_IN6:
473 	case SIOCGIFADDR_IN6:
474 	case SIOCSIFDSTADDR_IN6:
475 	case SIOCSIFNETMASK_IN6:
476 	case SIOCGIFDSTADDR_IN6:
477 	case SIOCGIFNETMASK_IN6:
478 	case SIOCDIFADDR_IN6:
479 	case SIOCGIFPSRCADDR_IN6:
480 	case SIOCGIFPDSTADDR_IN6:
481 	case SIOCGIFAFLAG_IN6:
482 	case SIOCSNDFLUSH_IN6:
483 	case SIOCSPFXFLUSH_IN6:
484 	case SIOCSRTRFLUSH_IN6:
485 	case SIOCGIFALIFETIME_IN6:
486 #ifdef OSIOCGIFALIFETIME_IN6
487 	case OSIOCGIFALIFETIME_IN6:
488 #endif
489 	case SIOCGIFSTAT_IN6:
490 	case SIOCGIFSTAT_ICMP6:
491 		sa6 = &ifr->ifr_addr;
492 		break;
493 	default:
494 		sa6 = NULL;
495 		break;
496 	}
497 
498 	error = 0;
499 	bound = curlwp_bind();
500 	if (sa6 && sa6->sin6_family == AF_INET6) {
501 		if (sa6->sin6_scope_id != 0)
502 			error = sa6_embedscope(sa6, 0);
503 		else
504 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
505 		if (error != 0)
506 			goto out;
507 		ia = in6ifa_ifpwithaddr_psref(ifp, &sa6->sin6_addr, &psref);
508 	} else
509 		ia = NULL;
510 
511 	switch (cmd) {
512 	case SIOCSIFADDR_IN6:
513 	case SIOCSIFDSTADDR_IN6:
514 	case SIOCSIFNETMASK_IN6:
515 		/*
516 		 * Since IPv6 allows a node to assign multiple addresses
517 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
518 		 */
519 		error = EINVAL;
520 		goto release;
521 
522 	case SIOCDIFADDR_IN6:
523 		/*
524 		 * for IPv4, we look for existing in_ifaddr here to allow
525 		 * "ifconfig if0 delete" to remove the first IPv4 address on
526 		 * the interface.  For IPv6, as the spec allows multiple
527 		 * interface address from the day one, we consider "remove the
528 		 * first one" semantics to be not preferable.
529 		 */
530 		if (ia == NULL) {
531 			error = EADDRNOTAVAIL;
532 			goto out;
533 		}
534 #ifdef OSIOCAIFADDR_IN6
535 		/* FALLTHROUGH */
536 	case OSIOCAIFADDR_IN6:
537 #endif
538 		/* FALLTHROUGH */
539 	case SIOCAIFADDR_IN6:
540 		/*
541 		 * We always require users to specify a valid IPv6 address for
542 		 * the corresponding operation.
543 		 */
544 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
545 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
546 			error = EAFNOSUPPORT;
547 			goto release;
548 		}
549 		/* Privileged. */
550 
551 		break;
552 
553 	case SIOCGIFADDR_IN6:
554 		/* This interface is basically deprecated. use SIOCGIFCONF. */
555 		/* FALLTHROUGH */
556 	case SIOCGIFAFLAG_IN6:
557 	case SIOCGIFNETMASK_IN6:
558 	case SIOCGIFDSTADDR_IN6:
559 	case SIOCGIFALIFETIME_IN6:
560 #ifdef OSIOCGIFALIFETIME_IN6
561 	case OSIOCGIFALIFETIME_IN6:
562 #endif
563 		/* must think again about its semantics */
564 		if (ia == NULL) {
565 			error = EADDRNOTAVAIL;
566 			goto out;
567 		}
568 		break;
569 	}
570 
571 	switch (cmd) {
572 
573 	case SIOCGIFADDR_IN6:
574 		ifr->ifr_addr = ia->ia_addr;
575 		error = sa6_recoverscope(&ifr->ifr_addr);
576 		break;
577 
578 	case SIOCGIFDSTADDR_IN6:
579 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
580 			error = EINVAL;
581 			break;
582 		}
583 		/*
584 		 * XXX: should we check if ifa_dstaddr is NULL and return
585 		 * an error?
586 		 */
587 		ifr->ifr_dstaddr = ia->ia_dstaddr;
588 		error = sa6_recoverscope(&ifr->ifr_dstaddr);
589 		break;
590 
591 	case SIOCGIFNETMASK_IN6:
592 		ifr->ifr_addr = ia->ia_prefixmask;
593 		break;
594 
595 	case SIOCGIFAFLAG_IN6:
596 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
597 		break;
598 
599 	case SIOCGIFSTAT_IN6:
600 		if (ifp == NULL) {
601 			error = EINVAL;
602 			break;
603 		}
604 		memset(&ifr->ifr_ifru.ifru_stat, 0,
605 		    sizeof(ifr->ifr_ifru.ifru_stat));
606 		ifr->ifr_ifru.ifru_stat =
607 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
608 		break;
609 
610 	case SIOCGIFSTAT_ICMP6:
611 		if (ifp == NULL) {
612 			error = EINVAL;
613 			break;
614 		}
615 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
616 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
617 		ifr->ifr_ifru.ifru_icmp6stat =
618 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
619 		break;
620 
621 #ifdef OSIOCGIFALIFETIME_IN6
622 	case OSIOCGIFALIFETIME_IN6:
623 #endif
624 	case SIOCGIFALIFETIME_IN6:
625 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
626 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
627 			time_t maxexpire;
628 			struct in6_addrlifetime *retlt =
629 			    &ifr->ifr_ifru.ifru_lifetime;
630 
631 			/*
632 			 * XXX: adjust expiration time assuming time_t is
633 			 * signed.
634 			 */
635 			maxexpire = ((time_t)~0) &
636 			    (time_t)~(1ULL << ((sizeof(maxexpire) * NBBY) - 1));
637 			if (ia->ia6_lifetime.ia6t_vltime <
638 			    maxexpire - ia->ia6_updatetime) {
639 				retlt->ia6t_expire = ia->ia6_updatetime +
640 				    ia->ia6_lifetime.ia6t_vltime;
641 				retlt->ia6t_expire = retlt->ia6t_expire ?
642 				    time_mono_to_wall(retlt->ia6t_expire) :
643 				    0;
644 			} else
645 				retlt->ia6t_expire = maxexpire;
646 		}
647 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
648 			time_t maxexpire;
649 			struct in6_addrlifetime *retlt =
650 			    &ifr->ifr_ifru.ifru_lifetime;
651 
652 			/*
653 			 * XXX: adjust expiration time assuming time_t is
654 			 * signed.
655 			 */
656 			maxexpire = ((time_t)~0) &
657 			    (time_t)~(1ULL << ((sizeof(maxexpire) * NBBY) - 1));
658 			if (ia->ia6_lifetime.ia6t_pltime <
659 			    maxexpire - ia->ia6_updatetime) {
660 				retlt->ia6t_preferred = ia->ia6_updatetime +
661 				    ia->ia6_lifetime.ia6t_pltime;
662 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
663 				    time_mono_to_wall(retlt->ia6t_preferred) :
664 				    0;
665 			} else
666 				retlt->ia6t_preferred = maxexpire;
667 		}
668 #ifdef OSIOCFIFALIFETIME_IN6
669 		if (cmd == OSIOCFIFALIFETIME_IN6)
670 			in6_addrlifetime_to_in6_addrlifetime50(
671 			    &ifr->ifru.ifru_lifetime);
672 #endif
673 		break;
674 
675 #ifdef OSIOCAIFADDR_IN6
676 	case OSIOCAIFADDR_IN6:
677 		in6_aliasreq50_to_in6_aliasreq(ifra);
678 #endif
679 		/*FALLTHROUGH*/
680 	case SIOCAIFADDR_IN6:
681 	{
682 		struct in6_addrlifetime *lt;
683 
684 		/* reject read-only flags */
685 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
686 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
687 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
688 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
689 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
690 			error = EINVAL;
691 			break;
692 		}
693 		/*
694 		 * ia6t_expire and ia6t_preferred won't be used for now,
695 		 * so just in case.
696 		 */
697 		lt = &ifra->ifra_lifetime;
698 		if (lt->ia6t_expire != 0)
699 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
700 		if (lt->ia6t_preferred != 0)
701 			lt->ia6t_preferred =
702 			    time_wall_to_mono(lt->ia6t_preferred);
703 		/*
704 		 * make (ia == NULL) or update (ia != NULL) the interface
705 		 * address structure, and link it to the list.
706 		 */
707 		int s = splsoftnet();
708 		error = in6_update_ifa1(ifp, ifra, &ia, &psref, 0);
709 		splx(s);
710 		if (error)
711 			break;
712 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
713 		break;
714 	}
715 
716 	case SIOCDIFADDR_IN6:
717 		ia6_release(ia, &psref);
718 		ifaref(&ia->ia_ifa);
719 		in6_purgeaddr(&ia->ia_ifa);
720 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
721 		ifafree(&ia->ia_ifa);
722 		ia = NULL;
723 		break;
724 
725 	default:
726 		error = ENOTTY;
727 	}
728 release:
729 	ia6_release(ia, &psref);
730 out:
731 	curlwp_bindx(bound);
732 	return error;
733 }
734 
735 int
736 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
737 {
738 	int error, s;
739 
740 	switch (cmd) {
741 	case SIOCSNDFLUSH_IN6:
742 	case SIOCSPFXFLUSH_IN6:
743 	case SIOCSRTRFLUSH_IN6:
744 	case SIOCSDEFIFACE_IN6:
745 	case SIOCSIFINFO_FLAGS:
746 	case SIOCSIFINFO_IN6:
747 
748 	case SIOCALIFADDR:
749 	case SIOCDLIFADDR:
750 
751 	case SIOCDIFADDR_IN6:
752 #ifdef OSIOCAIFADDR_IN6
753 	case OSIOCAIFADDR_IN6:
754 #endif
755 	case SIOCAIFADDR_IN6:
756 
757 	case SIOCAADDRCTL_POLICY:
758 	case SIOCDADDRCTL_POLICY:
759 
760 		if (kauth_authorize_network(curlwp->l_cred,
761 		    KAUTH_NETWORK_SOCKET,
762 		    KAUTH_REQ_NETWORK_SOCKET_SETPRIV,
763 		    so, NULL, NULL))
764 			return EPERM;
765 		break;
766 	}
767 
768 	s = splsoftnet();
769 #ifndef NET_MPSAFE
770 	KASSERT(KERNEL_LOCKED_P());
771 #endif
772 	error = in6_control1(so , cmd, data, ifp);
773 	splx(s);
774 	return error;
775 }
776 
777 static int
778 in6_get_llsol_addr(struct in6_addr *llsol, struct ifnet *ifp,
779     struct in6_addr *ip6)
780 {
781 	int error;
782 
783 	memset(llsol, 0, sizeof(struct in6_addr));
784 	llsol->s6_addr16[0] = htons(0xff02);
785 	llsol->s6_addr32[1] = 0;
786 	llsol->s6_addr32[2] = htonl(1);
787 	llsol->s6_addr32[3] = ip6->s6_addr32[3];
788 	llsol->s6_addr8[12] = 0xff;
789 
790 	error = in6_setscope(llsol, ifp, NULL);
791 	if (error != 0) {
792 		/* XXX: should not happen */
793 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
794 	}
795 
796 	return error;
797 }
798 
799 static int
800 in6_join_mcastgroups(struct in6_aliasreq *ifra, struct in6_ifaddr *ia,
801     struct ifnet *ifp, int flags)
802 {
803 	int error;
804 	struct sockaddr_in6 mltaddr, mltmask;
805 	struct in6_multi_mship *imm;
806 	struct in6_addr llsol;
807 	struct rtentry *rt;
808 	int dad_delay;
809 	char ip6buf[INET6_ADDRSTRLEN];
810 
811 	/* join solicited multicast addr for new host id */
812 	error = in6_get_llsol_addr(&llsol, ifp, &ifra->ifra_addr.sin6_addr);
813 	if (error != 0)
814 		goto out;
815 	dad_delay = 0;
816 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
817 		/*
818 		 * We need a random delay for DAD on the address
819 		 * being configured.  It also means delaying
820 		 * transmission of the corresponding MLD report to
821 		 * avoid report collision.
822 		 * [draft-ietf-ipv6-rfc2462bis-02.txt]
823 		 */
824 		dad_delay = cprng_fast32() % (MAX_RTR_SOLICITATION_DELAY * hz);
825 	}
826 
827 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
828 	/* join solicited multicast addr for new host id */
829 	imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
830 	if (!imm) {
831 		nd6log(LOG_ERR,
832 		    "addmulti failed for %s on %s (errno=%d)\n",
833 		    IN6_PRINT(ip6buf, &llsol), if_name(ifp), error);
834 		goto out;
835 	}
836 	mutex_enter(&in6_ifaddr_lock);
837 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
838 	mutex_exit(&in6_ifaddr_lock);
839 
840 	sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
841 
842 	/*
843 	 * join link-local all-nodes address
844 	 */
845 	sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
846 	    0, 0, 0);
847 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
848 		goto out; /* XXX: should not fail */
849 
850 	/*
851 	 * XXX: do we really need this automatic routes?
852 	 * We should probably reconsider this stuff.  Most applications
853 	 * actually do not need the routes, since they usually specify
854 	 * the outgoing interface.
855 	 */
856 	rt = rtalloc1(sin6tosa(&mltaddr), 0);
857 	if (rt) {
858 		if (memcmp(&mltaddr.sin6_addr,
859 		    &satocsin6(rt_getkey(rt))->sin6_addr,
860 		    MLTMASK_LEN)) {
861 			rt_unref(rt);
862 			rt = NULL;
863 		} else if (rt->rt_ifp != ifp) {
864 			IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
865 			    "network %04x:%04x::/32 = %04x:%04x::/32\n",
866 			    __func__, rt->rt_ifp, ifp, ifp->if_xname,
867 			    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
868 			    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
869 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
870 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
871 #ifdef NET_MPSAFE
872 			error = rt_update_prepare(rt);
873 			if (error == 0) {
874 				rt_replace_ifa(rt, &ia->ia_ifa);
875 				rt->rt_ifp = ifp;
876 				rt_update_finish(rt);
877 			} else {
878 				/*
879 				 * If error != 0, the rtentry is being
880 				 * destroyed, so doing nothing doesn't
881 				 * matter.
882 				 */
883 			}
884 #else
885 			rt_replace_ifa(rt, &ia->ia_ifa);
886 			rt->rt_ifp = ifp;
887 #endif
888 		}
889 	}
890 	if (!rt) {
891 		struct rt_addrinfo info;
892 
893 		memset(&info, 0, sizeof(info));
894 		info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
895 		info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
896 		info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
897 		info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
898 		/* XXX: we need RTF_CONNECTED to fake nd6_rtrequest */
899 		info.rti_flags = RTF_UP | RTF_CONNECTED;
900 		error = rtrequest1(RTM_ADD, &info, NULL);
901 		if (error)
902 			goto out;
903 	} else {
904 		rt_unref(rt);
905 	}
906 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
907 	if (!imm) {
908 		nd6log(LOG_WARNING,
909 		    "addmulti failed for %s on %s (errno=%d)\n",
910 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
911 		    if_name(ifp), error);
912 		goto out;
913 	}
914 	mutex_enter(&in6_ifaddr_lock);
915 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
916 	mutex_exit(&in6_ifaddr_lock);
917 
918 	/*
919 	 * join node information group address
920 	 */
921 	dad_delay = 0;
922 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
923 		/*
924 		 * The spec doesn't say anything about delay for this
925 		 * group, but the same logic should apply.
926 		 */
927 		dad_delay = cprng_fast32() % (MAX_RTR_SOLICITATION_DELAY * hz);
928 	}
929 	if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
930 		;
931 	else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
932 		  dad_delay)) == NULL) { /* XXX jinmei */
933 		nd6log(LOG_WARNING,
934 		    "addmulti failed for %s on %s (errno=%d)\n",
935 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
936 		    if_name(ifp), error);
937 		/* XXX not very fatal, go on... */
938 	} else {
939 		mutex_enter(&in6_ifaddr_lock);
940 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
941 		mutex_exit(&in6_ifaddr_lock);
942 	}
943 
944 
945 	/*
946 	 * join interface-local all-nodes address.
947 	 * (ff01::1%ifN, and ff01::%ifN/32)
948 	 */
949 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
950 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
951 		goto out; /* XXX: should not fail */
952 
953 	/* XXX: again, do we really need the route? */
954 	rt = rtalloc1(sin6tosa(&mltaddr), 0);
955 	if (rt) {
956 		/* 32bit came from "mltmask" */
957 		if (memcmp(&mltaddr.sin6_addr,
958 		    &satocsin6(rt_getkey(rt))->sin6_addr,
959 		    32 / NBBY)) {
960 			rt_unref(rt);
961 			rt = NULL;
962 		} else if (rt->rt_ifp != ifp) {
963 			IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
964 			    "network %04x:%04x::/32 = %04x:%04x::/32\n",
965 			    __func__, rt->rt_ifp, ifp, ifp->if_xname,
966 			    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
967 			    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
968 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
969 			    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
970 #ifdef NET_MPSAFE
971 			error = rt_update_prepare(rt);
972 			if (error == 0) {
973 				rt_replace_ifa(rt, &ia->ia_ifa);
974 				rt->rt_ifp = ifp;
975 				rt_update_finish(rt);
976 			} else {
977 				/*
978 				 * If error != 0, the rtentry is being
979 				 * destroyed, so doing nothing doesn't
980 				 * matter.
981 				 */
982 			}
983 #else
984 			rt_replace_ifa(rt, &ia->ia_ifa);
985 			rt->rt_ifp = ifp;
986 #endif
987 		}
988 	}
989 	if (!rt) {
990 		struct rt_addrinfo info;
991 
992 		memset(&info, 0, sizeof(info));
993 		info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
994 		info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
995 		info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
996 		info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
997 		info.rti_flags = RTF_UP | RTF_CONNECTED;
998 		error = rtrequest1(RTM_ADD, &info, NULL);
999 		if (error)
1000 			goto out;
1001 #undef	MLTMASK_LEN
1002 	} else {
1003 		rt_unref(rt);
1004 	}
1005 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1006 	if (!imm) {
1007 		nd6log(LOG_WARNING,
1008 		    "addmulti failed for %s on %s (errno=%d)\n",
1009 		    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
1010 		    if_name(ifp), error);
1011 		goto out;
1012 	} else {
1013 		mutex_enter(&in6_ifaddr_lock);
1014 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1015 		mutex_exit(&in6_ifaddr_lock);
1016 	}
1017 	return 0;
1018 
1019 out:
1020 	KASSERT(error != 0);
1021 	return error;
1022 }
1023 
1024 /*
1025  * Update parameters of an IPv6 interface address.
1026  * If necessary, a new entry is created and linked into address chains.
1027  * This function is separated from in6_control().
1028  * XXX: should this be performed under splsoftnet()?
1029  */
1030 static int
1031 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
1032     struct in6_ifaddr **iap, struct psref *psref, int flags)
1033 {
1034 	int error = 0, hostIsNew = 0, plen = -1;
1035 	struct sockaddr_in6 dst6;
1036 	struct in6_addrlifetime *lt;
1037 	int dad_delay, was_tentative;
1038 	struct in6_ifaddr *ia = iap ? *iap : NULL;
1039 	char ip6buf[INET6_ADDRSTRLEN];
1040 
1041 	KASSERT((iap == NULL && psref == NULL) ||
1042 	    (iap != NULL && psref != NULL));
1043 
1044 	/* Validate parameters */
1045 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1046 		return EINVAL;
1047 
1048 	/*
1049 	 * The destination address for a p2p link must have a family
1050 	 * of AF_UNSPEC or AF_INET6.
1051 	 */
1052 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1053 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1054 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1055 		return EAFNOSUPPORT;
1056 	/*
1057 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1058 	 * does not carry fields other than sin6_len.
1059 	 */
1060 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1061 		return EINVAL;
1062 	/*
1063 	 * Because the IPv6 address architecture is classless, we require
1064 	 * users to specify a (non 0) prefix length (mask) for a new address.
1065 	 * We also require the prefix (when specified) mask is valid, and thus
1066 	 * reject a non-consecutive mask.
1067 	 */
1068 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1069 		return EINVAL;
1070 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1071 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1072 		    (u_char *)&ifra->ifra_prefixmask +
1073 		    ifra->ifra_prefixmask.sin6_len);
1074 		if (plen <= 0)
1075 			return EINVAL;
1076 	} else {
1077 		/*
1078 		 * In this case, ia must not be NULL.  We just use its prefix
1079 		 * length.
1080 		 */
1081 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1082 	}
1083 	/*
1084 	 * If the destination address on a p2p interface is specified,
1085 	 * and the address is a scoped one, validate/set the scope
1086 	 * zone identifier.
1087 	 */
1088 	dst6 = ifra->ifra_dstaddr;
1089 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1090 	    (dst6.sin6_family == AF_INET6)) {
1091 		struct in6_addr in6_tmp;
1092 		u_int32_t zoneid;
1093 
1094 		in6_tmp = dst6.sin6_addr;
1095 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1096 			return EINVAL; /* XXX: should be impossible */
1097 
1098 		if (dst6.sin6_scope_id != 0) {
1099 			if (dst6.sin6_scope_id != zoneid)
1100 				return EINVAL;
1101 		} else		/* user omit to specify the ID. */
1102 			dst6.sin6_scope_id = zoneid;
1103 
1104 		/* convert into the internal form */
1105 		if (sa6_embedscope(&dst6, 0))
1106 			return EINVAL; /* XXX: should be impossible */
1107 	}
1108 	/*
1109 	 * The destination address can be specified only for a p2p or a
1110 	 * loopback interface.  If specified, the corresponding prefix length
1111 	 * must be 128.
1112 	 */
1113 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1114 #ifdef FORCE_P2PPLEN
1115 		int i;
1116 #endif
1117 
1118 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1119 			/* XXX: noisy message */
1120 			nd6log(LOG_INFO, "a destination can "
1121 			    "be specified for a p2p or a loopback IF only\n");
1122 			return EINVAL;
1123 		}
1124 		if (plen != 128) {
1125 			nd6log(LOG_INFO, "prefixlen should "
1126 			    "be 128 when dstaddr is specified\n");
1127 #ifdef FORCE_P2PPLEN
1128 			/*
1129 			 * To be compatible with old configurations,
1130 			 * such as ifconfig gif0 inet6 2001::1 2001::2
1131 			 * prefixlen 126, we override the specified
1132 			 * prefixmask as if the prefix length was 128.
1133 			 */
1134 			ifra->ifra_prefixmask.sin6_len =
1135 			    sizeof(struct sockaddr_in6);
1136 			for (i = 0; i < 4; i++)
1137 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
1138 				    0xffffffff;
1139 			plen = 128;
1140 #else
1141 			return EINVAL;
1142 #endif
1143 		}
1144 	}
1145 	/* lifetime consistency check */
1146 	lt = &ifra->ifra_lifetime;
1147 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1148 		return EINVAL;
1149 	if (lt->ia6t_vltime == 0) {
1150 		/*
1151 		 * the following log might be noisy, but this is a typical
1152 		 * configuration mistake or a tool's bug.
1153 		 */
1154 		nd6log(LOG_INFO, "valid lifetime is 0 for %s\n",
1155 		    IN6_PRINT(ip6buf, &ifra->ifra_addr.sin6_addr));
1156 
1157 		if (ia == NULL)
1158 			return 0; /* there's nothing to do */
1159 	}
1160 
1161 	/*
1162 	 * If this is a new address, allocate a new ifaddr and link it
1163 	 * into chains.
1164 	 */
1165 	if (ia == NULL) {
1166 		hostIsNew = 1;
1167 		/*
1168 		 * When in6_update_ifa() is called in a process of a received
1169 		 * RA, it is called under an interrupt context.  So, we should
1170 		 * call malloc with M_NOWAIT.
1171 		 */
1172 		ia = malloc(sizeof(*ia), M_IFADDR, M_NOWAIT|M_ZERO);
1173 		if (ia == NULL)
1174 			return ENOBUFS;
1175 		LIST_INIT(&ia->ia6_memberships);
1176 		/* Initialize the address and masks, and put time stamp */
1177 		ia->ia_ifa.ifa_addr = sin6tosa(&ia->ia_addr);
1178 		ia->ia_addr.sin6_family = AF_INET6;
1179 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1180 		ia->ia6_createtime = time_uptime;
1181 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1182 			/*
1183 			 * XXX: some functions expect that ifa_dstaddr is not
1184 			 * NULL for p2p interfaces.
1185 			 */
1186 			ia->ia_ifa.ifa_dstaddr = sin6tosa(&ia->ia_dstaddr);
1187 		} else {
1188 			ia->ia_ifa.ifa_dstaddr = NULL;
1189 		}
1190 		ia->ia_ifa.ifa_netmask = sin6tosa(&ia->ia_prefixmask);
1191 
1192 		ia->ia_ifp = ifp;
1193 		IN6_ADDRLIST_ENTRY_INIT(ia);
1194 		ifa_psref_init(&ia->ia_ifa);
1195 	}
1196 
1197 	/* update timestamp */
1198 	ia->ia6_updatetime = time_uptime;
1199 
1200 	/* set prefix mask */
1201 	if (ifra->ifra_prefixmask.sin6_len) {
1202 		if (ia->ia_prefixmask.sin6_len) {
1203 			/*
1204 			 * We prohibit changing the prefix length of an
1205 			 * existing autoconf address, because the operation
1206 			 * would confuse prefix management.
1207 			 */
1208 			if (ia->ia6_ndpr != NULL &&
1209 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) !=
1210 			    plen)
1211 			{
1212 				nd6log(LOG_INFO, "the prefix length of an"
1213 				    " existing (%s) autoconf address should"
1214 				    " not be changed\n",
1215 				    IN6_PRINT(ip6buf,
1216 				    &ia->ia_addr.sin6_addr));
1217 				error = EINVAL;
1218 				if (hostIsNew)
1219 					free(ia, M_IFADDR);
1220 				return error;
1221 			}
1222 
1223 			if (!IN6_ARE_ADDR_EQUAL(&ia->ia_prefixmask.sin6_addr,
1224 			    &ifra->ifra_prefixmask.sin6_addr))
1225 				in6_ifremprefix(ia);
1226 		}
1227 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1228 	}
1229 
1230 	/* Set destination address. */
1231 	if (dst6.sin6_family == AF_INET6) {
1232 		if (!IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
1233 		    &ia->ia_dstaddr.sin6_addr))
1234 			in6_ifremprefix(ia);
1235 		ia->ia_dstaddr = dst6;
1236 	}
1237 
1238 	/*
1239 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1240 	 * to see if the address is deprecated or invalidated, but initialize
1241 	 * these members for applications.
1242 	 */
1243 	ia->ia6_lifetime = ifra->ifra_lifetime;
1244 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1245 		ia->ia6_lifetime.ia6t_expire =
1246 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1247 	} else
1248 		ia->ia6_lifetime.ia6t_expire = 0;
1249 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1250 		ia->ia6_lifetime.ia6t_preferred =
1251 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1252 	} else
1253 		ia->ia6_lifetime.ia6t_preferred = 0;
1254 
1255 	/*
1256 	 * configure address flags.
1257 	 * We need to preserve tentative state so DAD works if
1258 	 * something adds the same address before DAD finishes.
1259 	 */
1260 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1261 	ia->ia6_flags = ifra->ifra_flags;
1262 
1263 	/*
1264 	 * Make the address tentative before joining multicast addresses,
1265 	 * so that corresponding MLD responses would not have a tentative
1266 	 * source address.
1267 	 */
1268 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1269 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1270 		ia->ia6_flags |= IN6_IFF_DETACHED;
1271 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1272 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp) &&
1273 	           ip6_dad_enabled()) {
1274 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1275 	}
1276 
1277 	/*
1278 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1279 	 * userland, make it deprecated.
1280 	 */
1281 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1282 		ia->ia6_lifetime.ia6t_pltime = 0;
1283 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1284 	}
1285 
1286 	if (hostIsNew) {
1287 		/*
1288 		 * We need a reference to ia before calling in6_ifinit.
1289 		 * Otherwise ia can be freed in in6_ifinit accidentally.
1290 		 */
1291 		ifaref(&ia->ia_ifa);
1292 	}
1293 
1294 	/* Must execute in6_ifinit and ifa_insert atomically */
1295 	mutex_enter(&in6_ifaddr_lock);
1296 
1297 	/* reset the interface and routing table appropriately. */
1298 	error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew);
1299 	if (error != 0) {
1300 		if (hostIsNew)
1301 			free(ia, M_IFADDR);
1302 		mutex_exit(&in6_ifaddr_lock);
1303 		return error;
1304 	}
1305 
1306 	/*
1307 	 * We are done if we have simply modified an existing address.
1308 	 */
1309 	if (!hostIsNew) {
1310 		mutex_exit(&in6_ifaddr_lock);
1311 		return error;
1312 	}
1313 
1314 	/*
1315 	 * Insert ia to the global list and ifa to the interface's list.
1316 	 * A reference to it is already gained above.
1317 	 */
1318 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1319 	ifa_insert(ifp, &ia->ia_ifa);
1320 
1321 	mutex_exit(&in6_ifaddr_lock);
1322 
1323 	/*
1324 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1325 	 * not just go to unlink.
1326 	 */
1327 
1328 	/* join necessary multicast groups */
1329 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1330 		error = in6_join_mcastgroups(ifra, ia, ifp, flags);
1331 		if (error != 0)
1332 			goto cleanup;
1333 	}
1334 
1335 	if (nd6_need_cache(ifp)) {
1336 		/* XXX maybe unnecessary */
1337 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1338 		ia->ia_ifa.ifa_flags |= RTF_CONNECTED;
1339 	}
1340 
1341 	/*
1342 	 * Perform DAD, if needed.
1343 	 * XXX It may be of use, if we can administratively
1344 	 * disable DAD.
1345 	 */
1346 	if (hostIsNew && if_do_dad(ifp) &&
1347 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1348 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1349 	{
1350 		int mindelay, maxdelay;
1351 
1352 		dad_delay = 0;
1353 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1354 			struct in6_addr llsol;
1355 			struct in6_multi *in6m_sol = NULL;
1356 			/*
1357 			 * We need to impose a delay before sending an NS
1358 			 * for DAD.  Check if we also needed a delay for the
1359 			 * corresponding MLD message.  If we did, the delay
1360 			 * should be larger than the MLD delay (this could be
1361 			 * relaxed a bit, but this simple logic is at least
1362 			 * safe).
1363 			 */
1364 			mindelay = 0;
1365 			error = in6_get_llsol_addr(&llsol, ifp,
1366 			    &ifra->ifra_addr.sin6_addr);
1367 			in6_multi_lock(RW_READER);
1368 			if (error == 0)
1369 				in6m_sol = in6_lookup_multi(&llsol, ifp);
1370 			if (in6m_sol != NULL &&
1371 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1372 				mindelay = in6m_sol->in6m_timer;
1373 			}
1374 			in6_multi_unlock();
1375 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1376 			if (maxdelay - mindelay == 0)
1377 				dad_delay = 0;
1378 			else {
1379 				dad_delay =
1380 				    (cprng_fast32() % (maxdelay - mindelay)) +
1381 				    mindelay;
1382 			}
1383 		}
1384 		/* +1 ensures callout is always used */
1385 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1386 	}
1387 
1388 	if (iap != NULL) {
1389 		*iap = ia;
1390 		if (hostIsNew)
1391 			ia6_acquire(ia, psref);
1392 	}
1393 
1394 	return 0;
1395 
1396   cleanup:
1397 	in6_purgeaddr(&ia->ia_ifa);
1398 	return error;
1399 }
1400 
1401 int
1402 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1403 {
1404 	int rc, s;
1405 
1406 	s = splsoftnet();
1407 	rc = in6_update_ifa1(ifp, ifra, NULL, NULL, flags);
1408 	splx(s);
1409 	return rc;
1410 }
1411 
1412 void
1413 in6_purgeaddr(struct ifaddr *ifa)
1414 {
1415 	struct ifnet *ifp = ifa->ifa_ifp;
1416 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1417 	struct in6_multi_mship *imm;
1418 
1419 	/* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */
1420 	KASSERT(IFNET_LOCKED(ifp));
1421 
1422 	ifa->ifa_flags |= IFA_DESTROYING;
1423 
1424 	/* stop DAD processing */
1425 	nd6_dad_stop(ifa);
1426 
1427 	/* Delete any network route. */
1428 	in6_ifremprefix(ia);
1429 
1430 	/* Remove ownaddr's loopback rtentry, if it exists. */
1431 	in6_ifremlocal(&(ia->ia_ifa));
1432 
1433 	/*
1434 	 * leave from multicast groups we have joined for the interface
1435 	 */
1436     again:
1437 	mutex_enter(&in6_ifaddr_lock);
1438 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1439 		struct in6_multi *in6m __diagused = imm->i6mm_maddr;
1440 		KASSERT(in6m == NULL || in6m->in6m_ifp == ifp);
1441 		LIST_REMOVE(imm, i6mm_chain);
1442 		mutex_exit(&in6_ifaddr_lock);
1443 
1444 		in6_leavegroup(imm);
1445 		goto again;
1446 	}
1447 	mutex_exit(&in6_ifaddr_lock);
1448 
1449 	in6_unlink_ifa(ia, ifp);
1450 }
1451 
1452 static void
1453 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1454 {
1455 	int	s = splsoftnet();
1456 
1457 	mutex_enter(&in6_ifaddr_lock);
1458 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1459 	ifa_remove(ifp, &ia->ia_ifa);
1460 	/* Assume ifa_remove called pserialize_perform and psref_destroy */
1461 	mutex_exit(&in6_ifaddr_lock);
1462 
1463 	/*
1464 	 * Release the reference to the ND prefix.
1465 	 */
1466 	if (ia->ia6_ndpr != NULL) {
1467 		nd6_prefix_unref(ia->ia6_ndpr);
1468 		ia->ia6_ndpr = NULL;
1469 	}
1470 
1471 	/*
1472 	 * Also, if the address being removed is autoconf'ed, call
1473 	 * nd6_pfxlist_onlink_check() since the release might affect the status of
1474 	 * other (detached) addresses.
1475 	 */
1476 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0) {
1477 		ND6_WLOCK();
1478 		nd6_pfxlist_onlink_check();
1479 		ND6_UNLOCK();
1480 	}
1481 
1482 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1483 
1484 	/*
1485 	 * release another refcnt for the link from in6_ifaddr.
1486 	 * Note that we should decrement the refcnt at least once for all *BSD.
1487 	 */
1488 	ifafree(&ia->ia_ifa);
1489 
1490 	splx(s);
1491 }
1492 
1493 void
1494 in6_purgeif(struct ifnet *ifp)
1495 {
1496 
1497 	IFNET_LOCK(ifp);
1498 	in6_ifdetach(ifp);
1499 	IFNET_UNLOCK(ifp);
1500 }
1501 
1502 void
1503 in6_purge_mcast_references(struct in6_multi *in6m)
1504 {
1505 	struct	in6_ifaddr *ia;
1506 
1507 	KASSERT(in6_multi_locked(RW_WRITER));
1508 
1509 	mutex_enter(&in6_ifaddr_lock);
1510 	IN6_ADDRLIST_WRITER_FOREACH(ia) {
1511 		struct in6_multi_mship *imm;
1512 		LIST_FOREACH(imm, &ia->ia6_memberships, i6mm_chain) {
1513 			if (imm->i6mm_maddr == in6m)
1514 				imm->i6mm_maddr = NULL;
1515 		}
1516 	}
1517 	mutex_exit(&in6_ifaddr_lock);
1518 }
1519 
1520 /*
1521  * SIOC[GAD]LIFADDR.
1522  *	SIOCGLIFADDR: get first address. (?)
1523  *	SIOCGLIFADDR with IFLR_PREFIX:
1524  *		get first address that matches the specified prefix.
1525  *	SIOCALIFADDR: add the specified address.
1526  *	SIOCALIFADDR with IFLR_PREFIX:
1527  *		add the specified prefix, filling hostid part from
1528  *		the first link-local address.  prefixlen must be <= 64.
1529  *	SIOCDLIFADDR: delete the specified address.
1530  *	SIOCDLIFADDR with IFLR_PREFIX:
1531  *		delete the first address that matches the specified prefix.
1532  * return values:
1533  *	EINVAL on invalid parameters
1534  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1535  *	other values may be returned from in6_ioctl()
1536  *
1537  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1538  * this is to accommodate address naming scheme other than RFC2374,
1539  * in the future.
1540  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1541  * address encoding scheme. (see figure on page 8)
1542  */
1543 static int
1544 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1545 	struct ifnet *ifp)
1546 {
1547 	struct in6_ifaddr *ia = NULL; /* XXX gcc 4.8 maybe-uninitialized */
1548 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1549 	struct ifaddr *ifa;
1550 	struct sockaddr *sa;
1551 
1552 	/* sanity checks */
1553 	if (!data || !ifp) {
1554 		panic("invalid argument to in6_lifaddr_ioctl");
1555 		/* NOTREACHED */
1556 	}
1557 
1558 	switch (cmd) {
1559 	case SIOCGLIFADDR:
1560 		/* address must be specified on GET with IFLR_PREFIX */
1561 		if ((iflr->flags & IFLR_PREFIX) == 0)
1562 			break;
1563 		/* FALLTHROUGH */
1564 	case SIOCALIFADDR:
1565 	case SIOCDLIFADDR:
1566 		/* address must be specified on ADD and DELETE */
1567 		sa = (struct sockaddr *)&iflr->addr;
1568 		if (sa->sa_family != AF_INET6)
1569 			return EINVAL;
1570 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1571 			return EINVAL;
1572 		/* XXX need improvement */
1573 		sa = (struct sockaddr *)&iflr->dstaddr;
1574 		if (sa->sa_family && sa->sa_family != AF_INET6)
1575 			return EINVAL;
1576 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1577 			return EINVAL;
1578 		break;
1579 	default: /* shouldn't happen */
1580 #if 0
1581 		panic("invalid cmd to in6_lifaddr_ioctl");
1582 		/* NOTREACHED */
1583 #else
1584 		return EOPNOTSUPP;
1585 #endif
1586 	}
1587 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1588 		return EINVAL;
1589 
1590 	switch (cmd) {
1591 	case SIOCALIFADDR:
1592 	    {
1593 		struct in6_aliasreq ifra;
1594 		struct in6_addr *xhostid = NULL;
1595 		int prefixlen;
1596 		int bound = curlwp_bind();
1597 		struct psref psref;
1598 
1599 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1600 			struct sockaddr_in6 *sin6;
1601 
1602 			/*
1603 			 * xhostid is to fill in the hostid part of the
1604 			 * address.  xhostid points to the first link-local
1605 			 * address attached to the interface.
1606 			 */
1607 			ia = in6ifa_ifpforlinklocal_psref(ifp, 0, &psref);
1608 			if (ia == NULL) {
1609 				curlwp_bindx(bound);
1610 				return EADDRNOTAVAIL;
1611 			}
1612 			xhostid = IFA_IN6(&ia->ia_ifa);
1613 
1614 		 	/* prefixlen must be <= 64. */
1615 			if (64 < iflr->prefixlen) {
1616 				ia6_release(ia, &psref);
1617 				curlwp_bindx(bound);
1618 				return EINVAL;
1619 			}
1620 			prefixlen = iflr->prefixlen;
1621 
1622 			/* hostid part must be zero. */
1623 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1624 			if (sin6->sin6_addr.s6_addr32[2] != 0
1625 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1626 				ia6_release(ia, &psref);
1627 				curlwp_bindx(bound);
1628 				return EINVAL;
1629 			}
1630 		} else
1631 			prefixlen = iflr->prefixlen;
1632 
1633 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1634 		memset(&ifra, 0, sizeof(ifra));
1635 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1636 
1637 		memcpy(&ifra.ifra_addr, &iflr->addr,
1638 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1639 		if (xhostid) {
1640 			/* fill in hostid part */
1641 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1642 			    xhostid->s6_addr32[2];
1643 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1644 			    xhostid->s6_addr32[3];
1645 		}
1646 
1647 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1648 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1649 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1650 			if (xhostid) {
1651 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1652 				    xhostid->s6_addr32[2];
1653 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1654 				    xhostid->s6_addr32[3];
1655 			}
1656 		}
1657 		if (xhostid) {
1658 			ia6_release(ia, &psref);
1659 			ia = NULL;
1660 		}
1661 		curlwp_bindx(bound);
1662 
1663 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1664 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1665 
1666 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1667 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1668 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1669 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1670 	    }
1671 	case SIOCGLIFADDR:
1672 	case SIOCDLIFADDR:
1673 	    {
1674 		struct in6_addr mask, candidate, match;
1675 		struct sockaddr_in6 *sin6;
1676 		int cmp;
1677 		int error, s;
1678 
1679 		memset(&mask, 0, sizeof(mask));
1680 		if (iflr->flags & IFLR_PREFIX) {
1681 			/* lookup a prefix rather than address. */
1682 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1683 
1684 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1685 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1686 			match.s6_addr32[0] &= mask.s6_addr32[0];
1687 			match.s6_addr32[1] &= mask.s6_addr32[1];
1688 			match.s6_addr32[2] &= mask.s6_addr32[2];
1689 			match.s6_addr32[3] &= mask.s6_addr32[3];
1690 
1691 			/* if you set extra bits, that's wrong */
1692 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1693 				return EINVAL;
1694 
1695 			cmp = 1;
1696 		} else {
1697 			if (cmd == SIOCGLIFADDR) {
1698 				/* on getting an address, take the 1st match */
1699 				cmp = 0;	/* XXX */
1700 			} else {
1701 				/* on deleting an address, do exact match */
1702 				in6_prefixlen2mask(&mask, 128);
1703 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1704 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1705 
1706 				cmp = 1;
1707 			}
1708 		}
1709 
1710 		s = pserialize_read_enter();
1711 		IFADDR_READER_FOREACH(ifa, ifp) {
1712 			if (ifa->ifa_addr->sa_family != AF_INET6)
1713 				continue;
1714 			if (!cmp)
1715 				break;
1716 
1717 			/*
1718 			 * XXX: this is adhoc, but is necessary to allow
1719 			 * a user to specify fe80::/64 (not /10) for a
1720 			 * link-local address.
1721 			 */
1722 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1723 			in6_clearscope(&candidate);
1724 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1725 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1726 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1727 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1728 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1729 				break;
1730 		}
1731 		if (!ifa) {
1732 			error = EADDRNOTAVAIL;
1733 			goto error;
1734 		}
1735 		ia = ifa2ia6(ifa);
1736 
1737 		if (cmd == SIOCGLIFADDR) {
1738 			/* fill in the if_laddrreq structure */
1739 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1740 			error = sa6_recoverscope(
1741 			    (struct sockaddr_in6 *)&iflr->addr);
1742 			if (error != 0)
1743 				goto error;
1744 
1745 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1746 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1747 				    ia->ia_dstaddr.sin6_len);
1748 				error = sa6_recoverscope(
1749 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1750 				if (error != 0)
1751 					goto error;
1752 			} else
1753 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1754 
1755 			iflr->prefixlen =
1756 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1757 
1758 			iflr->flags = ia->ia6_flags;	/* XXX */
1759 
1760 			error = 0;
1761 		} else {
1762 			struct in6_aliasreq ifra;
1763 
1764 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1765 			memset(&ifra, 0, sizeof(ifra));
1766 			memcpy(ifra.ifra_name, iflr->iflr_name,
1767 			    sizeof(ifra.ifra_name));
1768 
1769 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1770 			    ia->ia_addr.sin6_len);
1771 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1772 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1773 				    ia->ia_dstaddr.sin6_len);
1774 			} else {
1775 				memset(&ifra.ifra_dstaddr, 0,
1776 				    sizeof(ifra.ifra_dstaddr));
1777 			}
1778 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1779 			    ia->ia_prefixmask.sin6_len);
1780 
1781 			ifra.ifra_flags = ia->ia6_flags;
1782 			pserialize_read_exit(s);
1783 
1784 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1785 		}
1786 	error:
1787 		pserialize_read_exit(s);
1788 		return error;
1789 	    }
1790 	}
1791 
1792 	return EOPNOTSUPP;	/* just for safety */
1793 }
1794 
1795 /*
1796  * Initialize an interface's internet6 address
1797  * and routing table entry.
1798  */
1799 static int
1800 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1801 	const struct sockaddr_in6 *sin6, int newhost)
1802 {
1803 	int	error = 0, ifacount = 0;
1804 	int s;
1805 	struct ifaddr *ifa;
1806 
1807 	KASSERT(mutex_owned(&in6_ifaddr_lock));
1808 
1809 	/*
1810 	 * Give the interface a chance to initialize
1811 	 * if this is its first address,
1812 	 * and to validate the address if necessary.
1813 	 */
1814 	s = pserialize_read_enter();
1815 	IFADDR_READER_FOREACH(ifa, ifp) {
1816 		if (ifa->ifa_addr->sa_family != AF_INET6)
1817 			continue;
1818 		ifacount++;
1819 	}
1820 	pserialize_read_exit(s);
1821 
1822 	ia->ia_addr = *sin6;
1823 
1824 	if (ifacount == 0 &&
1825 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1826 		return error;
1827 	}
1828 
1829 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1830 
1831 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1832 
1833 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1834 	if (newhost) {
1835 		/* set the rtrequest function to create llinfo */
1836 		if (ifp->if_flags & IFF_POINTOPOINT)
1837 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1838 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1839 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1840 		in6_ifaddlocal(&ia->ia_ifa);
1841 	} else {
1842 		/* Inform the routing socket of new flags/timings */
1843 		rt_addrmsg(RTM_NEWADDR, &ia->ia_ifa);
1844 	}
1845 
1846 	/* Add the network prefix route. */
1847 	if ((error = in6_ifaddprefix(ia)) != 0) {
1848 		if (newhost)
1849 			in6_ifremlocal(&ia->ia_ifa);
1850 		return error;
1851 	}
1852 
1853 	return error;
1854 }
1855 
1856 static struct ifaddr *
1857 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1858 {
1859 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1860 		return ifa;
1861 	return best_ifa;
1862 }
1863 
1864 /*
1865  * Find an IPv6 interface link-local address specific to an interface.
1866  */
1867 struct in6_ifaddr *
1868 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1869 {
1870 	struct ifaddr *best_ifa = NULL, *ifa;
1871 
1872 	IFADDR_READER_FOREACH(ifa, ifp) {
1873 		if (ifa->ifa_addr->sa_family != AF_INET6)
1874 			continue;
1875 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1876 			continue;
1877 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1878 			continue;
1879 		best_ifa = bestifa(best_ifa, ifa);
1880 	}
1881 
1882 	return (struct in6_ifaddr *)best_ifa;
1883 }
1884 
1885 struct in6_ifaddr *
1886 in6ifa_ifpforlinklocal_psref(const struct ifnet *ifp, const int ignoreflags,
1887     struct psref *psref)
1888 {
1889 	struct in6_ifaddr *ia;
1890 	int s = pserialize_read_enter();
1891 
1892 	ia = in6ifa_ifpforlinklocal(ifp, ignoreflags);
1893 	if (ia != NULL)
1894 		ia6_acquire(ia, psref);
1895 	pserialize_read_exit(s);
1896 
1897 	return ia;
1898 }
1899 
1900 /*
1901  * find the internet address corresponding to a given address.
1902  * ifaddr is returned referenced.
1903  */
1904 struct in6_ifaddr *
1905 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1906 {
1907 	struct in6_ifaddr *ia;
1908 	int s;
1909 
1910 	s = pserialize_read_enter();
1911 	IN6_ADDRLIST_READER_FOREACH(ia) {
1912 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1913 			if (zoneid != 0 &&
1914 			    zoneid != ia->ia_addr.sin6_scope_id)
1915 				continue;
1916 			ifaref(&ia->ia_ifa);
1917 			break;
1918 		}
1919 	}
1920 	pserialize_read_exit(s);
1921 
1922 	return ia;
1923 }
1924 
1925 /*
1926  * find the internet address corresponding to a given interface and address.
1927  */
1928 struct in6_ifaddr *
1929 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1930 {
1931 	struct ifaddr *best_ifa = NULL, *ifa;
1932 
1933 	IFADDR_READER_FOREACH(ifa, ifp) {
1934 		if (ifa->ifa_addr->sa_family != AF_INET6)
1935 			continue;
1936 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1937 			continue;
1938 		best_ifa = bestifa(best_ifa, ifa);
1939 	}
1940 
1941 	return (struct in6_ifaddr *)best_ifa;
1942 }
1943 
1944 struct in6_ifaddr *
1945 in6ifa_ifpwithaddr_psref(const struct ifnet *ifp, const struct in6_addr *addr,
1946     struct psref *psref)
1947 {
1948 	struct in6_ifaddr *ia;
1949 	int s = pserialize_read_enter();
1950 
1951 	ia = in6ifa_ifpwithaddr(ifp, addr);
1952 	if (ia != NULL)
1953 		ia6_acquire(ia, psref);
1954 	pserialize_read_exit(s);
1955 
1956 	return ia;
1957 }
1958 
1959 static struct in6_ifaddr *
1960 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1961 {
1962 	if (best_ia == NULL ||
1963 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1964 		return ia;
1965 	return best_ia;
1966 }
1967 
1968 /*
1969  * Determine if an address is on a local network.
1970  */
1971 int
1972 in6_localaddr(const struct in6_addr *in6)
1973 {
1974 	struct in6_ifaddr *ia;
1975 	int s;
1976 
1977 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1978 		return 1;
1979 
1980 	s = pserialize_read_enter();
1981 	IN6_ADDRLIST_READER_FOREACH(ia) {
1982 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1983 					      &ia->ia_prefixmask.sin6_addr)) {
1984 			pserialize_read_exit(s);
1985 			return 1;
1986 		}
1987 	}
1988 	pserialize_read_exit(s);
1989 
1990 	return 0;
1991 }
1992 
1993 int
1994 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1995 {
1996 	struct in6_ifaddr *ia;
1997 	int s;
1998 
1999 	s = pserialize_read_enter();
2000 	IN6_ADDRLIST_READER_FOREACH(ia) {
2001 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2002 		    &sa6->sin6_addr) &&
2003 #ifdef SCOPEDROUTING
2004 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
2005 #endif
2006 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
2007 			pserialize_read_exit(s);
2008 			return 1; /* true */
2009 		}
2010 
2011 		/* XXX: do we still have to go thru the rest of the list? */
2012 	}
2013 	pserialize_read_exit(s);
2014 
2015 	return 0;		/* false */
2016 }
2017 
2018 /*
2019  * return length of part which dst and src are equal
2020  * hard coding...
2021  */
2022 int
2023 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2024 {
2025 	int match = 0;
2026 	u_char *s = (u_char *)src, *d = (u_char *)dst;
2027 	u_char *lim = s + 16, r;
2028 
2029 	while (s < lim)
2030 		if ((r = (*d++ ^ *s++)) != 0) {
2031 			while (r < 128) {
2032 				match++;
2033 				r <<= 1;
2034 			}
2035 			break;
2036 		} else
2037 			match += NBBY;
2038 	return match;
2039 }
2040 
2041 /* XXX: to be scope conscious */
2042 int
2043 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2044 {
2045 	int bytelen, bitlen;
2046 
2047 	/* sanity check */
2048 	if (len < 0 || len > 128) {
2049 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2050 		    len);
2051 		return 0;
2052 	}
2053 
2054 	bytelen = len / NBBY;
2055 	bitlen = len % NBBY;
2056 
2057 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2058 		return 0;
2059 	if (bitlen != 0 &&
2060 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
2061 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
2062 		return 0;
2063 
2064 	return 1;
2065 }
2066 
2067 void
2068 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2069 {
2070 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2071 	int bytelen, bitlen, i;
2072 
2073 	/* sanity check */
2074 	if (len < 0 || len > 128) {
2075 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2076 		    len);
2077 		return;
2078 	}
2079 
2080 	memset(maskp, 0, sizeof(*maskp));
2081 	bytelen = len / NBBY;
2082 	bitlen = len % NBBY;
2083 	for (i = 0; i < bytelen; i++)
2084 		maskp->s6_addr[i] = 0xff;
2085 	if (bitlen)
2086 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2087 }
2088 
2089 /*
2090  * return the best address out of the same scope. if no address was
2091  * found, return the first valid address from designated IF.
2092  */
2093 struct in6_ifaddr *
2094 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2095 {
2096 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2097 	struct ifaddr *ifa;
2098 	struct in6_ifaddr *best_ia = NULL, *ia;
2099 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2100 
2101 	dep[0] = dep[1] = NULL;
2102 
2103 	/*
2104 	 * We first look for addresses in the same scope.
2105 	 * If there is one, return it.
2106 	 * If two or more, return one which matches the dst longest.
2107 	 * If none, return one of global addresses assigned other ifs.
2108 	 */
2109 	IFADDR_READER_FOREACH(ifa, ifp) {
2110 		if (ifa->ifa_addr->sa_family != AF_INET6)
2111 			continue;
2112 		ia = (struct in6_ifaddr *)ifa;
2113 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2114 			continue; /* XXX: is there any case to allow anycast? */
2115 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2116 			continue; /* don't use this interface */
2117 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2118 			continue;
2119 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2120 			if (ip6_use_deprecated)
2121 				dep[0] = ia;
2122 			continue;
2123 		}
2124 
2125 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
2126 			continue;
2127 		/*
2128 		 * call in6_matchlen() as few as possible
2129 		 */
2130 		if (best_ia == NULL) {
2131 			best_ia = ia;
2132 			continue;
2133 		}
2134 		if (blen == -1)
2135 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2136 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2137 		if (tlen > blen) {
2138 			blen = tlen;
2139 			best_ia = ia;
2140 		} else if (tlen == blen)
2141 			best_ia = bestia(best_ia, ia);
2142 	}
2143 	if (best_ia != NULL)
2144 		return best_ia;
2145 
2146 	IFADDR_READER_FOREACH(ifa, ifp) {
2147 		if (ifa->ifa_addr->sa_family != AF_INET6)
2148 			continue;
2149 		ia = (struct in6_ifaddr *)ifa;
2150 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2151 			continue; /* XXX: is there any case to allow anycast? */
2152 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2153 			continue; /* don't use this interface */
2154 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2155 			continue;
2156 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2157 			if (ip6_use_deprecated)
2158 				dep[1] = (struct in6_ifaddr *)ifa;
2159 			continue;
2160 		}
2161 
2162 		best_ia = bestia(best_ia, ia);
2163 	}
2164 	if (best_ia != NULL)
2165 		return best_ia;
2166 
2167 	/* use the last-resort values, that are, deprecated addresses */
2168 	if (dep[0])
2169 		return dep[0];
2170 	if (dep[1])
2171 		return dep[1];
2172 
2173 	return NULL;
2174 }
2175 
2176 /*
2177  * perform DAD when interface becomes IFF_UP.
2178  */
2179 void
2180 in6_if_link_up(struct ifnet *ifp)
2181 {
2182 	struct ifaddr *ifa;
2183 	struct in6_ifaddr *ia;
2184 	int s, bound;
2185 	char ip6buf[INET6_ADDRSTRLEN];
2186 
2187 	/* Ensure it's sane to run DAD */
2188 	if (ifp->if_link_state == LINK_STATE_DOWN)
2189 		return;
2190 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2191 		return;
2192 
2193 	bound = curlwp_bind();
2194 	s = pserialize_read_enter();
2195 	IFADDR_READER_FOREACH(ifa, ifp) {
2196 		struct psref psref;
2197 
2198 		if (ifa->ifa_addr->sa_family != AF_INET6)
2199 			continue;
2200 
2201 		ifa_acquire(ifa, &psref);
2202 		pserialize_read_exit(s);
2203 		ia = (struct in6_ifaddr *)ifa;
2204 
2205 		/* If detached then mark as tentative */
2206 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2207 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2208 			if (ip6_dad_enabled() && if_do_dad(ifp)) {
2209 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2210 				nd6log(LOG_ERR, "%s marked tentative\n",
2211 				    IN6_PRINT(ip6buf,
2212 				    &ia->ia_addr.sin6_addr));
2213 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2214 				rt_addrmsg(RTM_NEWADDR, ifa);
2215 		}
2216 
2217 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2218 			int rand_delay;
2219 
2220 			/* Clear the duplicated flag as we're starting DAD. */
2221 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2222 
2223 			/*
2224 			 * The TENTATIVE flag was likely set by hand
2225 			 * beforehand, implicitly indicating the need for DAD.
2226 			 * We may be able to skip the random delay in this
2227 			 * case, but we impose delays just in case.
2228 			 */
2229 			rand_delay = cprng_fast32() %
2230 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2231 			/* +1 ensures callout is always used */
2232 			nd6_dad_start(ifa, rand_delay + 1);
2233 		}
2234 
2235 		s = pserialize_read_enter();
2236 		ifa_release(ifa, &psref);
2237 	}
2238 	pserialize_read_exit(s);
2239 	curlwp_bindx(bound);
2240 
2241 	/* Restore any detached prefixes */
2242 	ND6_WLOCK();
2243 	nd6_pfxlist_onlink_check();
2244 	ND6_UNLOCK();
2245 }
2246 
2247 void
2248 in6_if_up(struct ifnet *ifp)
2249 {
2250 
2251 	/*
2252 	 * special cases, like 6to4, are handled in in6_ifattach
2253 	 */
2254 	in6_ifattach(ifp, NULL);
2255 
2256 	/* interface may not support link state, so bring it up also */
2257 	in6_if_link_up(ifp);
2258 }
2259 
2260 /*
2261  * Mark all addresses as detached.
2262  */
2263 void
2264 in6_if_link_down(struct ifnet *ifp)
2265 {
2266 	struct ifaddr *ifa;
2267 	struct in6_ifaddr *ia;
2268 	int s, bound;
2269 	char ip6buf[INET6_ADDRSTRLEN];
2270 
2271 	/* Any prefixes on this interface should be detached as well */
2272 	ND6_WLOCK();
2273 	nd6_pfxlist_onlink_check();
2274 	ND6_UNLOCK();
2275 
2276 	bound = curlwp_bind();
2277 	s = pserialize_read_enter();
2278 	IFADDR_READER_FOREACH(ifa, ifp) {
2279 		struct psref psref;
2280 
2281 		if (ifa->ifa_addr->sa_family != AF_INET6)
2282 			continue;
2283 
2284 		ifa_acquire(ifa, &psref);
2285 		pserialize_read_exit(s);
2286 		ia = (struct in6_ifaddr *)ifa;
2287 
2288 		/* Stop DAD processing */
2289 		nd6_dad_stop(ifa);
2290 
2291 		/*
2292 		 * Mark the address as detached.
2293 		 * This satisfies RFC4862 Section 5.3, but we should apply
2294 		 * this logic to all addresses to be a good citizen and
2295 		 * avoid potential duplicated addresses.
2296 		 * When the interface comes up again, detached addresses
2297 		 * are marked tentative and DAD commences.
2298 		 */
2299 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2300 			nd6log(LOG_DEBUG, "%s marked detached\n",
2301 			    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
2302 			ia->ia6_flags |= IN6_IFF_DETACHED;
2303 			ia->ia6_flags &=
2304 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2305 			rt_addrmsg(RTM_NEWADDR, ifa);
2306 		}
2307 
2308 		s = pserialize_read_enter();
2309 		ifa_release(ifa, &psref);
2310 	}
2311 	pserialize_read_exit(s);
2312 	curlwp_bindx(bound);
2313 }
2314 
2315 void
2316 in6_if_down(struct ifnet *ifp)
2317 {
2318 
2319 	in6_if_link_down(ifp);
2320 	lltable_purge_entries(LLTABLE6(ifp));
2321 }
2322 
2323 void
2324 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2325 {
2326 
2327 	switch (link_state) {
2328 	case LINK_STATE_DOWN:
2329 		in6_if_link_down(ifp);
2330 		break;
2331 	case LINK_STATE_UP:
2332 		in6_if_link_up(ifp);
2333 		break;
2334 	}
2335 }
2336 
2337 /*
2338  * Calculate max IPv6 MTU through all the interfaces and store it
2339  * to in6_maxmtu.
2340  */
2341 void
2342 in6_setmaxmtu(void)
2343 {
2344 	unsigned long maxmtu = 0;
2345 	struct ifnet *ifp;
2346 	int s;
2347 
2348 	s = pserialize_read_enter();
2349 	IFNET_READER_FOREACH(ifp) {
2350 		/* this function can be called during ifnet initialization */
2351 		if (!ifp->if_afdata[AF_INET6])
2352 			continue;
2353 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2354 		    IN6_LINKMTU(ifp) > maxmtu)
2355 			maxmtu = IN6_LINKMTU(ifp);
2356 	}
2357 	pserialize_read_exit(s);
2358 	if (maxmtu)	     /* update only when maxmtu is positive */
2359 		in6_maxmtu = maxmtu;
2360 }
2361 
2362 int
2363 in6_tunnel_validate(const struct ip6_hdr *ip6, const struct in6_addr *src,
2364     const struct in6_addr *dst)
2365 {
2366 
2367 	/* check for address match */
2368 	if (!IN6_ARE_ADDR_EQUAL(src, &ip6->ip6_dst) ||
2369 	    !IN6_ARE_ADDR_EQUAL(dst, &ip6->ip6_src))
2370 		return 0;
2371 
2372 	/* martian filters on outer source - done in ip6_input */
2373 
2374 	/* NOTE: the packet may be dropped by uRPF. */
2375 
2376 	/* return valid bytes length */
2377 	return sizeof(*src) + sizeof(*dst);
2378 }
2379 
2380 /*
2381  * Provide the length of interface identifiers to be used for the link attached
2382  * to the given interface.  The length should be defined in "IPv6 over
2383  * xxx-link" document.  Note that address architecture might also define
2384  * the length for a particular set of address prefixes, regardless of the
2385  * link type.  As clarified in rfc2462bis, those two definitions should be
2386  * consistent, and those really are as of August 2004.
2387  */
2388 int
2389 in6_if2idlen(struct ifnet *ifp)
2390 {
2391 	switch (ifp->if_type) {
2392 	case IFT_ETHER:		/* RFC2464 */
2393 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2394 	case IFT_L2VLAN:	/* ditto */
2395 	case IFT_IEEE80211:	/* ditto */
2396 	case IFT_FDDI:		/* RFC2467 */
2397 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2398 	case IFT_PPP:		/* RFC2472 */
2399 	case IFT_ARCNET:	/* RFC2497 */
2400 	case IFT_FRELAY:	/* RFC2590 */
2401 	case IFT_IEEE1394:	/* RFC3146 */
2402 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2403 	case IFT_LOOP:		/* XXX: is this really correct? */
2404 		return 64;
2405 	default:
2406 		/*
2407 		 * Unknown link type:
2408 		 * It might be controversial to use the today's common constant
2409 		 * of 64 for these cases unconditionally.  For full compliance,
2410 		 * we should return an error in this case.  On the other hand,
2411 		 * if we simply miss the standard for the link type or a new
2412 		 * standard is defined for a new link type, the IFID length
2413 		 * is very likely to be the common constant.  As a compromise,
2414 		 * we always use the constant, but make an explicit notice
2415 		 * indicating the "unknown" case.
2416 		 */
2417 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2418 		return 64;
2419 	}
2420 }
2421 
2422 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2423 #define	IN6_LLTBL_HASH(k, h) \
2424 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2425 
2426 /*
2427  * Do actual deallocation of @lle.
2428  * Called by LLE_FREE_LOCKED when number of references
2429  * drops to zero.
2430  */
2431 static void
2432 in6_lltable_destroy_lle(struct llentry *lle)
2433 {
2434 
2435 	KASSERT(lle->la_numheld == 0);
2436 
2437 	LLE_WUNLOCK(lle);
2438 	LLE_LOCK_DESTROY(lle);
2439 	llentry_pool_put(lle);
2440 }
2441 
2442 static struct llentry *
2443 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2444 {
2445 	struct llentry *lle;
2446 
2447 	lle = llentry_pool_get(PR_NOWAIT);
2448 	if (lle == NULL)		/* NB: caller generates msg */
2449 		return NULL;
2450 
2451 	lle->r_l3addr.addr6 = *addr6;
2452 	lle->lle_refcnt = 1;
2453 	lle->lle_free = in6_lltable_destroy_lle;
2454 	LLE_LOCK_INIT(lle);
2455 	callout_init(&lle->lle_timer, CALLOUT_MPSAFE);
2456 
2457 	return lle;
2458 }
2459 
2460 static int
2461 in6_lltable_match_prefix(const struct sockaddr *prefix,
2462     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2463 {
2464 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2465 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2466 
2467 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2468 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2469 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2470 		return 1;
2471 
2472 	return 0;
2473 }
2474 
2475 static void
2476 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2477 {
2478 
2479 	LLE_WLOCK_ASSERT(lle);
2480 	(void) llentry_free(lle);
2481 }
2482 
2483 static int
2484 in6_lltable_rtcheck(struct ifnet *ifp, u_int flags,
2485     const struct sockaddr *l3addr, const struct rtentry *rt)
2486 {
2487 	char ip6buf[INET6_ADDRSTRLEN];
2488 
2489 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2490 		int s;
2491 		struct ifaddr *ifa;
2492 		/*
2493 		 * Create an ND6 cache for an IPv6 neighbor
2494 		 * that is not covered by our own prefix.
2495 		 */
2496 		/* XXX ifaof_ifpforaddr should take a const param */
2497 		s = pserialize_read_enter();
2498 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2499 		if (ifa != NULL) {
2500 			pserialize_read_exit(s);
2501 			return 0;
2502 		}
2503 		pserialize_read_exit(s);
2504 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2505 		    IN6_PRINT(ip6buf,
2506 		    &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2507 		return EINVAL;
2508 	}
2509 	return 0;
2510 }
2511 
2512 static inline uint32_t
2513 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2514 {
2515 
2516 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2517 }
2518 
2519 static uint32_t
2520 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2521 {
2522 
2523 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2524 }
2525 
2526 static void
2527 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2528 {
2529 	struct sockaddr_in6 *sin6;
2530 
2531 	sin6 = (struct sockaddr_in6 *)sa;
2532 	bzero(sin6, sizeof(*sin6));
2533 	sin6->sin6_family = AF_INET6;
2534 	sin6->sin6_len = sizeof(*sin6);
2535 	sin6->sin6_addr = lle->r_l3addr.addr6;
2536 }
2537 
2538 static inline struct llentry *
2539 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2540 {
2541 	struct llentry *lle;
2542 	struct llentries *lleh;
2543 	u_int hashidx;
2544 
2545 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2546 	lleh = &llt->lle_head[hashidx];
2547 	LIST_FOREACH(lle, lleh, lle_next) {
2548 		if (lle->la_flags & LLE_DELETED)
2549 			continue;
2550 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2551 			break;
2552 	}
2553 
2554 	return lle;
2555 }
2556 
2557 static int
2558 in6_lltable_delete(struct lltable *llt, u_int flags,
2559 	const struct sockaddr *l3addr)
2560 {
2561 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2562 	struct llentry *lle;
2563 
2564 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2565 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2566 	    "sin_family %d", l3addr->sa_family);
2567 
2568 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2569 
2570 	if (lle == NULL) {
2571 #ifdef LLTABLE_DEBUG
2572 		char buf[64];
2573 		sockaddr_format(l3addr, buf, sizeof(buf));
2574 		log(LOG_INFO, "%s: cache for %s is not found\n",
2575 		    __func__, buf);
2576 #endif
2577 		return ENOENT;
2578 	}
2579 
2580 	LLE_WLOCK(lle);
2581 #ifdef LLTABLE_DEBUG
2582 	{
2583 		char buf[64];
2584 		sockaddr_format(l3addr, buf, sizeof(buf));
2585 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2586 		    __func__, buf, lle);
2587 	}
2588 #endif
2589 	llentry_free(lle);
2590 
2591 	return 0;
2592 }
2593 
2594 static struct llentry *
2595 in6_lltable_create(struct lltable *llt, u_int flags,
2596     const struct sockaddr *l3addr, const struct rtentry *rt)
2597 {
2598 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2599 	struct ifnet *ifp = llt->llt_ifp;
2600 	struct llentry *lle;
2601 
2602 	IF_AFDATA_WLOCK_ASSERT(ifp);
2603 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2604 	    "sin_family %d", l3addr->sa_family);
2605 
2606 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2607 
2608 	if (lle != NULL) {
2609 		LLE_WLOCK(lle);
2610 		return lle;
2611 	}
2612 
2613 	/*
2614 	 * A route that covers the given address must have
2615 	 * been installed 1st because we are doing a resolution,
2616 	 * verify this.
2617 	 */
2618 	if (!(flags & LLE_IFADDR) &&
2619 	    in6_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2620 		return NULL;
2621 
2622 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2623 	if (lle == NULL) {
2624 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2625 		return NULL;
2626 	}
2627 	lle->la_flags = flags;
2628 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2629 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2630 		lle->la_flags |= LLE_VALID;
2631 	}
2632 
2633 	lltable_link_entry(llt, lle);
2634 	LLE_WLOCK(lle);
2635 
2636 	return lle;
2637 }
2638 
2639 static struct llentry *
2640 in6_lltable_lookup(struct lltable *llt, u_int flags,
2641 	const struct sockaddr *l3addr)
2642 {
2643 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2644 	struct llentry *lle;
2645 
2646 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2647 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2648 	    "sin_family %d", l3addr->sa_family);
2649 
2650 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2651 
2652 	if (lle == NULL)
2653 		return NULL;
2654 
2655 	if (flags & LLE_EXCLUSIVE)
2656 		LLE_WLOCK(lle);
2657 	else
2658 		LLE_RLOCK(lle);
2659 	return lle;
2660 }
2661 
2662 static int
2663 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2664     struct rt_walkarg *w)
2665 {
2666 	struct sockaddr_in6 sin6;
2667 
2668 	LLTABLE_LOCK_ASSERT();
2669 
2670 	/* skip deleted entries */
2671 	if (lle->la_flags & LLE_DELETED)
2672 		return 0;
2673 
2674 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2675 
2676 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2677 }
2678 
2679 static struct lltable *
2680 in6_lltattach(struct ifnet *ifp)
2681 {
2682 	struct lltable *llt;
2683 
2684 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2685 	llt->llt_af = AF_INET6;
2686 	llt->llt_ifp = ifp;
2687 
2688 	llt->llt_lookup = in6_lltable_lookup;
2689 	llt->llt_create = in6_lltable_create;
2690 	llt->llt_delete = in6_lltable_delete;
2691 	llt->llt_dump_entry = in6_lltable_dump_entry;
2692 	llt->llt_hash = in6_lltable_hash;
2693 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2694 	llt->llt_free_entry = in6_lltable_free_entry;
2695 	llt->llt_match_prefix = in6_lltable_match_prefix;
2696 	lltable_link(llt);
2697 
2698 	return llt;
2699 }
2700 
2701 void *
2702 in6_domifattach(struct ifnet *ifp)
2703 {
2704 	struct in6_ifextra *ext;
2705 
2706 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2707 
2708 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2709 	    M_IFADDR, M_WAITOK|M_ZERO);
2710 
2711 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2712 	    M_IFADDR, M_WAITOK|M_ZERO);
2713 
2714 	ext->nd_ifinfo = nd6_ifattach(ifp);
2715 	ext->scope6_id = scope6_ifattach(ifp);
2716 	ext->nprefixes = 0;
2717 	ext->ndefrouters = 0;
2718 
2719 	ext->lltable = in6_lltattach(ifp);
2720 
2721 	return ext;
2722 }
2723 
2724 void
2725 in6_domifdetach(struct ifnet *ifp, void *aux)
2726 {
2727 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2728 
2729 	lltable_free(ext->lltable);
2730 	ext->lltable = NULL;
2731 	SOFTNET_LOCK_UNLESS_NET_MPSAFE();
2732 	nd6_ifdetach(ifp, ext);
2733 	SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();
2734 	free(ext->in6_ifstat, M_IFADDR);
2735 	free(ext->icmp6_ifstat, M_IFADDR);
2736 	scope6_ifdetach(ext->scope6_id);
2737 	free(ext, M_IFADDR);
2738 }
2739 
2740 /*
2741  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2742  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2743  */
2744 void
2745 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2746 {
2747 	in6->s6_addr32[0] = 0;
2748 	in6->s6_addr32[1] = 0;
2749 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2750 	in6->s6_addr32[3] = in->s_addr;
2751 }
2752 
2753 /*
2754  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2755  * v4 mapped addr or v4 compat addr
2756  */
2757 void
2758 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2759 {
2760 	memset(sin, 0, sizeof(*sin));
2761 	sin->sin_len = sizeof(struct sockaddr_in);
2762 	sin->sin_family = AF_INET;
2763 	sin->sin_port = sin6->sin6_port;
2764 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2765 }
2766 
2767 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2768 void
2769 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2770 {
2771 	memset(sin6, 0, sizeof(*sin6));
2772 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2773 	sin6->sin6_family = AF_INET6;
2774 	sin6->sin6_port = sin->sin_port;
2775 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2776 }
2777 
2778 /* Convert sockaddr_in6 into sockaddr_in. */
2779 void
2780 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2781 {
2782 	struct sockaddr_in *sin_p;
2783 	struct sockaddr_in6 sin6;
2784 
2785 	/*
2786 	 * Save original sockaddr_in6 addr and convert it
2787 	 * to sockaddr_in.
2788 	 */
2789 	sin6 = *(struct sockaddr_in6 *)nam;
2790 	sin_p = (struct sockaddr_in *)nam;
2791 	in6_sin6_2_sin(sin_p, &sin6);
2792 }
2793 
2794 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2795 void
2796 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2797 {
2798 	struct sockaddr_in *sin_p;
2799 	struct sockaddr_in6 *sin6_p;
2800 
2801 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2802 	sin_p = (struct sockaddr_in *)*nam;
2803 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2804 	free(*nam, M_SONAME);
2805 	*nam = sin6tosa(sin6_p);
2806 }
2807