xref: /netbsd-src/sys/netinet6/in6.c (revision 181254a7b1bdde6873432bffef2d2decc4b5c22f)
1 /*	$NetBSD: in6.c,v 1.281 2020/06/16 17:12:18 maxv 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 1.281 2020/06/16 17:12:18 maxv 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 #ifdef COMPAT_90
112 #include <compat/netinet6/in6_var.h>
113 #include <compat/netinet6/nd6.h>
114 #endif
115 
116 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
117 
118 /* enable backward compatibility code for obsoleted ioctls */
119 #define COMPAT_IN6IFIOCTL
120 
121 #ifdef	IN6_DEBUG
122 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
123 #else
124 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
125 #endif /* IN6_DEBUG */
126 
127 /*
128  * Definitions of some constant IP6 addresses.
129  */
130 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
131 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
132 const struct in6_addr in6addr_nodelocal_allnodes =
133 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
134 const struct in6_addr in6addr_linklocal_allnodes =
135 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
136 const struct in6_addr in6addr_linklocal_allrouters =
137 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
138 
139 const struct in6_addr in6mask0 = IN6MASK0;
140 const struct in6_addr in6mask32 = IN6MASK32;
141 const struct in6_addr in6mask64 = IN6MASK64;
142 const struct in6_addr in6mask96 = IN6MASK96;
143 const struct in6_addr in6mask128 = IN6MASK128;
144 
145 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
146 				     0, 0, IN6ADDR_ANY_INIT, 0};
147 
148 struct pslist_head	in6_ifaddr_list;
149 kmutex_t		in6_ifaddr_lock;
150 
151 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
152 	struct ifnet *);
153 static int in6_ifaddprefix(struct in6_ifaddr *);
154 static int in6_ifremprefix(struct in6_ifaddr *);
155 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
156 	const struct sockaddr_in6 *, int);
157 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
158 static int in6_update_ifa1(struct ifnet *, struct in6_aliasreq *,
159     struct in6_ifaddr **, struct psref *, int);
160 
161 void
162 in6_init(void)
163 {
164 
165 	PSLIST_INIT(&in6_ifaddr_list);
166 	mutex_init(&in6_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
167 
168 	in6_sysctl_multicast_setup(NULL);
169 }
170 
171 /*
172  * Add ownaddr as loopback rtentry.  We previously add the route only if
173  * necessary (ex. on a p2p link).  However, since we now manage addresses
174  * separately from prefixes, we should always add the route.  We can't
175  * rely on the cloning mechanism from the corresponding interface route
176  * any more.
177  */
178 void
179 in6_ifaddlocal(struct ifaddr *ifa)
180 {
181 
182 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
183 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
184 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
185 	{
186 		rt_addrmsg(RTM_NEWADDR, ifa);
187 		return;
188 	}
189 
190 	rt_ifa_addlocal(ifa);
191 }
192 
193 /*
194  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
195  * if it exists.
196  */
197 void
198 in6_ifremlocal(struct ifaddr *ifa)
199 {
200 	struct in6_ifaddr *ia;
201 	struct ifaddr *alt_ifa = NULL;
202 	int ia_count = 0;
203 	struct psref psref;
204 	int s;
205 
206 	/*
207 	 * Some of BSD variants do not remove cloned routes
208 	 * from an interface direct route, when removing the direct route
209 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
210 	 * cloned routes, they could fail to remove the cloned routes when
211 	 * we handle multple addresses that share a common prefix.
212 	 * So, we should remove the route corresponding to the deleted address.
213 	 */
214 
215 	/*
216 	 * Delete the entry only if exactly one ifaddr matches the
217 	 * address, ifa->ifa_addr.
218 	 *
219 	 * If more than one ifaddr matches, replace the ifaddr in
220 	 * the routing table, rt_ifa, with a different ifaddr than
221 	 * the one we are purging, ifa.  It is important to do
222 	 * this, or else the routing table can accumulate dangling
223 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
224 	 * which will lead to crashes, later.  (More than one ifaddr
225 	 * can match if we assign the same address to multiple---probably
226 	 * p2p---interfaces.)
227 	 *
228 	 * XXX An old comment at this place said, "we should avoid
229 	 * XXX such a configuration [i.e., interfaces with the same
230 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
231 	 * XXX agree, especially now that I have fixed the dangling
232 	 * XXX ifp-pointers bug.
233 	 */
234 	s = pserialize_read_enter();
235 	IN6_ADDRLIST_READER_FOREACH(ia) {
236 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
237 			continue;
238 		if (ia->ia_ifp != ifa->ifa_ifp)
239 			alt_ifa = &ia->ia_ifa;
240 		if (++ia_count > 1 && alt_ifa != NULL)
241 			break;
242 	}
243 	if (ia_count > 1 && alt_ifa != NULL)
244 		ifa_acquire(alt_ifa, &psref);
245 	pserialize_read_exit(s);
246 
247 	if (ia_count == 0)
248 		return;
249 
250 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
251 
252 	if (ia_count > 1 && alt_ifa != NULL)
253 		ifa_release(alt_ifa, &psref);
254 }
255 
256 /* Add prefix route for the network. */
257 static int
258 in6_ifaddprefix(struct in6_ifaddr *ia)
259 {
260 	int error, flags = 0;
261 
262 	if (in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) == 128) {
263 		if (ia->ia_dstaddr.sin6_family != AF_INET6)
264 			/* We don't need to install a host route. */
265 			return 0;
266 		flags |= RTF_HOST;
267 	}
268 
269 	/* Is this a connected route for neighbour discovery? */
270 	if (nd6_need_cache(ia->ia_ifp))
271 		flags |= RTF_CONNECTED;
272 
273 	if ((error = rtinit(&ia->ia_ifa, RTM_ADD, RTF_UP | flags)) == 0)
274 		ia->ia_flags |= IFA_ROUTE;
275 	else if (error == EEXIST)
276 		/* Existance of the route is not an error. */
277 		error = 0;
278 
279 	return error;
280 }
281 
282 /* Delete network prefix route if present.
283  * Re-add it to another address if the prefix matches. */
284 static int
285 in6_ifremprefix(struct in6_ifaddr *target)
286 {
287 	int error, s;
288 	struct in6_ifaddr *ia;
289 
290 	if ((target->ia_flags & IFA_ROUTE) == 0)
291 		return 0;
292 
293 	s = pserialize_read_enter();
294 	IN6_ADDRLIST_READER_FOREACH(ia) {
295 		if (target->ia_dstaddr.sin6_len) {
296 			if (ia->ia_dstaddr.sin6_len == 0 ||
297 			    !IN6_ARE_ADDR_EQUAL(&ia->ia_dstaddr.sin6_addr,
298 			    &target->ia_dstaddr.sin6_addr))
299 				continue;
300 		} else {
301 			if (!IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
302 			    &target->ia_addr.sin6_addr,
303 			    &target->ia_prefixmask.sin6_addr))
304 				continue;
305 		}
306 
307 		/*
308 		 * if we got a matching prefix route, move IFA_ROUTE to him
309 		 */
310 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
311 			struct psref psref;
312 			int bound = curlwp_bind();
313 
314 			ia6_acquire(ia, &psref);
315 			pserialize_read_exit(s);
316 
317 			rtinit(&target->ia_ifa, RTM_DELETE, 0);
318 			target->ia_flags &= ~IFA_ROUTE;
319 
320 			error = in6_ifaddprefix(ia);
321 
322 			ia6_release(ia, &psref);
323 			curlwp_bindx(bound);
324 
325 			return error;
326 		}
327 	}
328 	pserialize_read_exit(s);
329 
330 	/*
331 	 * noone seem to have prefix route.  remove it.
332 	 */
333 	rtinit(&target->ia_ifa, RTM_DELETE, 0);
334 	target->ia_flags &= ~IFA_ROUTE;
335 	return 0;
336 }
337 
338 int
339 in6_mask2len(struct in6_addr *mask, u_char *lim0)
340 {
341 	int x = 0, y;
342 	u_char *lim = lim0, *p;
343 
344 	/* ignore the scope_id part */
345 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
346 		lim = (u_char *)mask + sizeof(*mask);
347 	for (p = (u_char *)mask; p < lim; x++, p++) {
348 		if (*p != 0xff)
349 			break;
350 	}
351 	y = 0;
352 	if (p < lim) {
353 		for (y = 0; y < NBBY; y++) {
354 			if ((*p & (0x80 >> y)) == 0)
355 				break;
356 		}
357 	}
358 
359 	/*
360 	 * when the limit pointer is given, do a stricter check on the
361 	 * remaining bits.
362 	 */
363 	if (p < lim) {
364 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
365 			return -1;
366 		for (p = p + 1; p < lim; p++)
367 			if (*p != 0)
368 				return -1;
369 	}
370 
371 	return x * NBBY + y;
372 }
373 
374 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
375 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
376 
377 static int
378 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
379 {
380 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
381 	struct	in6_ifaddr *ia = NULL;
382 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
383 	struct sockaddr_in6 *sa6;
384 	int error, bound;
385 	struct psref psref;
386 
387 	switch (cmd) {
388 	case SIOCAADDRCTL_POLICY:
389 	case SIOCDADDRCTL_POLICY:
390 		/* Privileged. */
391 		return in6_src_ioctl(cmd, data);
392 	/*
393 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
394 	 */
395 	case SIOCSIFADDR:
396 	case SIOCSIFDSTADDR:
397 	case SIOCSIFBRDADDR:
398 	case SIOCSIFNETMASK:
399 		return EOPNOTSUPP;
400 	case SIOCGETSGCNT_IN6:
401 	case SIOCGETMIFCNT_IN6:
402 		return mrt6_ioctl(cmd, data);
403 	case SIOCGIFADDRPREF:
404 	case SIOCSIFADDRPREF:
405 		if (ifp == NULL)
406 			return EINVAL;
407 		return ifaddrpref_ioctl(so, cmd, data, ifp);
408 	}
409 
410 	if (ifp == NULL)
411 		return EOPNOTSUPP;
412 
413 	switch (cmd) {
414 #ifdef OSIOCSIFINFO_IN6_90
415 	case OSIOCSIFINFO_FLAGS_90:
416 	case OSIOCSIFINFO_IN6_90:
417 	case OSIOCSDEFIFACE_IN6:
418 	case OSIOCSNDFLUSH_IN6:
419 	case OSIOCSPFXFLUSH_IN6:
420 	case OSIOCSRTRFLUSH_IN6:
421 #endif
422 	case SIOCSIFINFO_FLAGS:
423 	case SIOCSIFINFO_IN6:
424 		/* Privileged. */
425 		/* FALLTHROUGH */
426 #ifdef OSIOCGIFINFO_IN6
427 	case OSIOCGIFINFO_IN6:
428 #endif
429 #ifdef OSIOCGIFINFO_IN6_90
430 	case OSIOCGDRLST_IN6:
431 	case OSIOCGPRLST_IN6:
432 	case OSIOCGIFINFO_IN6_90:
433 	case OSIOCGDEFIFACE_IN6:
434 #endif
435 	case SIOCGIFINFO_IN6:
436 	case SIOCGNBRINFO_IN6:
437 		return nd6_ioctl(cmd, data, ifp);
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 SIOCGIFALIFETIME_IN6:
483 #ifdef OSIOCGIFALIFETIME_IN6
484 	case OSIOCGIFALIFETIME_IN6:
485 #endif
486 	case SIOCGIFSTAT_IN6:
487 	case SIOCGIFSTAT_ICMP6:
488 		sa6 = &ifr->ifr_addr;
489 		break;
490 	default:
491 		sa6 = NULL;
492 		break;
493 	}
494 
495 	error = 0;
496 	bound = curlwp_bind();
497 	if (sa6 && sa6->sin6_family == AF_INET6) {
498 		if (sa6->sin6_scope_id != 0)
499 			error = sa6_embedscope(sa6, 0);
500 		else
501 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
502 		if (error != 0)
503 			goto out;
504 		ia = in6ifa_ifpwithaddr_psref(ifp, &sa6->sin6_addr, &psref);
505 	} else
506 		ia = NULL;
507 
508 	switch (cmd) {
509 	case SIOCSIFADDR_IN6:
510 	case SIOCSIFDSTADDR_IN6:
511 	case SIOCSIFNETMASK_IN6:
512 		/*
513 		 * Since IPv6 allows a node to assign multiple addresses
514 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
515 		 */
516 		error = EINVAL;
517 		goto release;
518 
519 	case SIOCDIFADDR_IN6:
520 		/*
521 		 * for IPv4, we look for existing in_ifaddr here to allow
522 		 * "ifconfig if0 delete" to remove the first IPv4 address on
523 		 * the interface.  For IPv6, as the spec allows multiple
524 		 * interface address from the day one, we consider "remove the
525 		 * first one" semantics to be not preferable.
526 		 */
527 		if (ia == NULL) {
528 			error = EADDRNOTAVAIL;
529 			goto out;
530 		}
531 #ifdef OSIOCAIFADDR_IN6
532 		/* FALLTHROUGH */
533 	case OSIOCAIFADDR_IN6:
534 #endif
535 		/* FALLTHROUGH */
536 	case SIOCAIFADDR_IN6:
537 		/*
538 		 * We always require users to specify a valid IPv6 address for
539 		 * the corresponding operation.
540 		 */
541 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
542 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
543 			error = EAFNOSUPPORT;
544 			goto release;
545 		}
546 		/* Privileged. */
547 
548 		break;
549 
550 	case SIOCGIFADDR_IN6:
551 		/* This interface is basically deprecated. use SIOCGIFCONF. */
552 		/* FALLTHROUGH */
553 	case SIOCGIFAFLAG_IN6:
554 	case SIOCGIFNETMASK_IN6:
555 	case SIOCGIFDSTADDR_IN6:
556 	case SIOCGIFALIFETIME_IN6:
557 #ifdef OSIOCGIFALIFETIME_IN6
558 	case OSIOCGIFALIFETIME_IN6:
559 #endif
560 		/* must think again about its semantics */
561 		if (ia == NULL) {
562 			error = EADDRNOTAVAIL;
563 			goto out;
564 		}
565 		break;
566 	}
567 
568 	switch (cmd) {
569 
570 	case SIOCGIFADDR_IN6:
571 		ifr->ifr_addr = ia->ia_addr;
572 		error = sa6_recoverscope(&ifr->ifr_addr);
573 		break;
574 
575 	case SIOCGIFDSTADDR_IN6:
576 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
577 			error = EINVAL;
578 			break;
579 		}
580 		/*
581 		 * XXX: should we check if ifa_dstaddr is NULL and return
582 		 * an error?
583 		 */
584 		ifr->ifr_dstaddr = ia->ia_dstaddr;
585 		error = sa6_recoverscope(&ifr->ifr_dstaddr);
586 		break;
587 
588 	case SIOCGIFNETMASK_IN6:
589 		ifr->ifr_addr = ia->ia_prefixmask;
590 		break;
591 
592 	case SIOCGIFAFLAG_IN6:
593 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
594 		break;
595 
596 	case SIOCGIFSTAT_IN6:
597 		if (ifp == NULL) {
598 			error = EINVAL;
599 			break;
600 		}
601 		memset(&ifr->ifr_ifru.ifru_stat, 0,
602 		    sizeof(ifr->ifr_ifru.ifru_stat));
603 		ifr->ifr_ifru.ifru_stat =
604 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
605 		break;
606 
607 	case SIOCGIFSTAT_ICMP6:
608 		if (ifp == NULL) {
609 			error = EINVAL;
610 			break;
611 		}
612 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
613 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
614 		ifr->ifr_ifru.ifru_icmp6stat =
615 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
616 		break;
617 
618 #ifdef OSIOCGIFALIFETIME_IN6
619 	case OSIOCGIFALIFETIME_IN6:
620 #endif
621 	case SIOCGIFALIFETIME_IN6:
622 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
623 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
624 			time_t maxexpire;
625 			struct in6_addrlifetime *retlt =
626 			    &ifr->ifr_ifru.ifru_lifetime;
627 
628 			/*
629 			 * XXX: adjust expiration time assuming time_t is
630 			 * signed.
631 			 */
632 			maxexpire = ((time_t)~0) &
633 			    (time_t)~(1ULL << ((sizeof(maxexpire) * NBBY) - 1));
634 			if (ia->ia6_lifetime.ia6t_vltime <
635 			    maxexpire - ia->ia6_updatetime) {
636 				retlt->ia6t_expire = ia->ia6_updatetime +
637 				    ia->ia6_lifetime.ia6t_vltime;
638 				retlt->ia6t_expire = retlt->ia6t_expire ?
639 				    time_mono_to_wall(retlt->ia6t_expire) :
640 				    0;
641 			} else
642 				retlt->ia6t_expire = maxexpire;
643 		}
644 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
645 			time_t maxexpire;
646 			struct in6_addrlifetime *retlt =
647 			    &ifr->ifr_ifru.ifru_lifetime;
648 
649 			/*
650 			 * XXX: adjust expiration time assuming time_t is
651 			 * signed.
652 			 */
653 			maxexpire = ((time_t)~0) &
654 			    (time_t)~(1ULL << ((sizeof(maxexpire) * NBBY) - 1));
655 			if (ia->ia6_lifetime.ia6t_pltime <
656 			    maxexpire - ia->ia6_updatetime) {
657 				retlt->ia6t_preferred = ia->ia6_updatetime +
658 				    ia->ia6_lifetime.ia6t_pltime;
659 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
660 				    time_mono_to_wall(retlt->ia6t_preferred) :
661 				    0;
662 			} else
663 				retlt->ia6t_preferred = maxexpire;
664 		}
665 #ifdef OSIOCFIFALIFETIME_IN6
666 		if (cmd == OSIOCFIFALIFETIME_IN6)
667 			in6_addrlifetime_to_in6_addrlifetime50(
668 			    &ifr->ifru.ifru_lifetime);
669 #endif
670 		break;
671 
672 #ifdef OSIOCAIFADDR_IN6
673 	case OSIOCAIFADDR_IN6:
674 		in6_aliasreq50_to_in6_aliasreq(ifra);
675 #endif
676 		/*FALLTHROUGH*/
677 	case SIOCAIFADDR_IN6:
678 	{
679 		struct in6_addrlifetime *lt;
680 
681 		/* reject read-only flags */
682 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
683 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
684 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
685 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0) {
686 			error = EINVAL;
687 			break;
688 		}
689 		/*
690 		 * ia6t_expire and ia6t_preferred won't be used for now,
691 		 * so just in case.
692 		 */
693 		lt = &ifra->ifra_lifetime;
694 		if (lt->ia6t_expire != 0)
695 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
696 		if (lt->ia6t_preferred != 0)
697 			lt->ia6t_preferred =
698 			    time_wall_to_mono(lt->ia6t_preferred);
699 		/*
700 		 * make (ia == NULL) or update (ia != NULL) the interface
701 		 * address structure, and link it to the list.
702 		 */
703 		int s = splsoftnet();
704 		error = in6_update_ifa1(ifp, ifra, &ia, &psref, 0);
705 		splx(s);
706 		if (error)
707 			break;
708 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
709 		break;
710 	}
711 
712 	case SIOCDIFADDR_IN6:
713 		ia6_release(ia, &psref);
714 		ifaref(&ia->ia_ifa);
715 		in6_purgeaddr(&ia->ia_ifa);
716 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
717 		ifafree(&ia->ia_ifa);
718 		ia = NULL;
719 		break;
720 
721 	default:
722 		error = ENOTTY;
723 	}
724 release:
725 	ia6_release(ia, &psref);
726 out:
727 	curlwp_bindx(bound);
728 	return error;
729 }
730 
731 int
732 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
733 {
734 	int error, s;
735 
736 	switch (cmd) {
737 #ifdef OSIOCSIFINFO_IN6_90
738 	case OSIOCSIFINFO_FLAGS_90:
739 	case OSIOCSIFINFO_IN6_90:
740 	case OSIOCSDEFIFACE_IN6:
741 	case OSIOCSNDFLUSH_IN6:
742 	case OSIOCSPFXFLUSH_IN6:
743 	case OSIOCSRTRFLUSH_IN6:
744 #endif
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 			if (!IN6_ARE_ADDR_EQUAL(&ia->ia_prefixmask.sin6_addr,
1204 			    &ifra->ifra_prefixmask.sin6_addr))
1205 				in6_ifremprefix(ia);
1206 		}
1207 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1208 	}
1209 
1210 	/* Set destination address. */
1211 	if (dst6.sin6_family == AF_INET6) {
1212 		if (!IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
1213 		    &ia->ia_dstaddr.sin6_addr))
1214 			in6_ifremprefix(ia);
1215 		ia->ia_dstaddr = dst6;
1216 	}
1217 
1218 	/*
1219 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1220 	 * to see if the address is deprecated or invalidated, but initialize
1221 	 * these members for applications.
1222 	 */
1223 	ia->ia6_lifetime = ifra->ifra_lifetime;
1224 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1225 		ia->ia6_lifetime.ia6t_expire =
1226 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1227 	} else
1228 		ia->ia6_lifetime.ia6t_expire = 0;
1229 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1230 		ia->ia6_lifetime.ia6t_preferred =
1231 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1232 	} else
1233 		ia->ia6_lifetime.ia6t_preferred = 0;
1234 
1235 	/*
1236 	 * configure address flags.
1237 	 * We need to preserve tentative state so DAD works if
1238 	 * something adds the same address before DAD finishes.
1239 	 */
1240 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1241 	ia->ia6_flags = ifra->ifra_flags;
1242 
1243 	/*
1244 	 * Make the address tentative before joining multicast addresses,
1245 	 * so that corresponding MLD responses would not have a tentative
1246 	 * source address.
1247 	 */
1248 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1249 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1250 		ia->ia6_flags |= IN6_IFF_DETACHED;
1251 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1252 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp) &&
1253 	           ip6_dad_enabled()) {
1254 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1255 	}
1256 
1257 	/*
1258 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1259 	 * userland, make it deprecated.
1260 	 */
1261 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1262 		ia->ia6_lifetime.ia6t_pltime = 0;
1263 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1264 	}
1265 
1266 	if (hostIsNew) {
1267 		/*
1268 		 * We need a reference to ia before calling in6_ifinit.
1269 		 * Otherwise ia can be freed in in6_ifinit accidentally.
1270 		 */
1271 		ifaref(&ia->ia_ifa);
1272 	}
1273 
1274 	/* Must execute in6_ifinit and ifa_insert atomically */
1275 	mutex_enter(&in6_ifaddr_lock);
1276 
1277 	/* reset the interface and routing table appropriately. */
1278 	error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew);
1279 	if (error != 0) {
1280 		if (hostIsNew)
1281 			free(ia, M_IFADDR);
1282 		mutex_exit(&in6_ifaddr_lock);
1283 		return error;
1284 	}
1285 
1286 	/*
1287 	 * We are done if we have simply modified an existing address.
1288 	 */
1289 	if (!hostIsNew) {
1290 		mutex_exit(&in6_ifaddr_lock);
1291 		return error;
1292 	}
1293 
1294 	/*
1295 	 * Insert ia to the global list and ifa to the interface's list.
1296 	 * A reference to it is already gained above.
1297 	 */
1298 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1299 	ifa_insert(ifp, &ia->ia_ifa);
1300 
1301 	mutex_exit(&in6_ifaddr_lock);
1302 
1303 	/*
1304 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1305 	 * not just go to unlink.
1306 	 */
1307 
1308 	/* join necessary multicast groups */
1309 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1310 		error = in6_join_mcastgroups(ifra, ia, ifp, flags);
1311 		if (error != 0)
1312 			goto cleanup;
1313 	}
1314 
1315 	if (nd6_need_cache(ifp)) {
1316 		/* XXX maybe unnecessary */
1317 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1318 		ia->ia_ifa.ifa_flags |= RTF_CONNECTED;
1319 	}
1320 
1321 	/*
1322 	 * Perform DAD, if needed.
1323 	 * XXX It may be of use, if we can administratively
1324 	 * disable DAD.
1325 	 */
1326 	if (hostIsNew && if_do_dad(ifp) &&
1327 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1328 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1329 	{
1330 		int mindelay, maxdelay;
1331 
1332 		dad_delay = 0;
1333 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1334 			struct in6_addr llsol;
1335 			struct in6_multi *in6m_sol = NULL;
1336 			/*
1337 			 * We need to impose a delay before sending an NS
1338 			 * for DAD.  Check if we also needed a delay for the
1339 			 * corresponding MLD message.  If we did, the delay
1340 			 * should be larger than the MLD delay (this could be
1341 			 * relaxed a bit, but this simple logic is at least
1342 			 * safe).
1343 			 */
1344 			mindelay = 0;
1345 			error = in6_get_llsol_addr(&llsol, ifp,
1346 			    &ifra->ifra_addr.sin6_addr);
1347 			in6_multi_lock(RW_READER);
1348 			if (error == 0)
1349 				in6m_sol = in6_lookup_multi(&llsol, ifp);
1350 			if (in6m_sol != NULL &&
1351 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1352 				mindelay = in6m_sol->in6m_timer;
1353 			}
1354 			in6_multi_unlock();
1355 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1356 			if (maxdelay - mindelay == 0)
1357 				dad_delay = 0;
1358 			else {
1359 				dad_delay =
1360 				    (cprng_fast32() % (maxdelay - mindelay)) +
1361 				    mindelay;
1362 			}
1363 		}
1364 		/* +1 ensures callout is always used */
1365 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1366 	}
1367 
1368 	if (iap != NULL) {
1369 		*iap = ia;
1370 		if (hostIsNew)
1371 			ia6_acquire(ia, psref);
1372 	}
1373 
1374 	return 0;
1375 
1376   cleanup:
1377 	in6_purgeaddr(&ia->ia_ifa);
1378 	return error;
1379 }
1380 
1381 int
1382 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1383 {
1384 	int rc, s;
1385 
1386 	s = splsoftnet();
1387 	rc = in6_update_ifa1(ifp, ifra, NULL, NULL, flags);
1388 	splx(s);
1389 	return rc;
1390 }
1391 
1392 void
1393 in6_purgeaddr(struct ifaddr *ifa)
1394 {
1395 	struct ifnet *ifp = ifa->ifa_ifp;
1396 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1397 	struct in6_multi_mship *imm;
1398 
1399 	/* KASSERT(!ifa_held(ifa)); XXX need ifa_not_held (psref_not_held) */
1400 	KASSERT(IFNET_LOCKED(ifp));
1401 
1402 	ifa->ifa_flags |= IFA_DESTROYING;
1403 
1404 	/* stop DAD processing */
1405 	nd6_dad_stop(ifa);
1406 
1407 	/* Delete any network route. */
1408 	in6_ifremprefix(ia);
1409 
1410 	/* Remove ownaddr's loopback rtentry, if it exists. */
1411 	in6_ifremlocal(&(ia->ia_ifa));
1412 
1413 	/*
1414 	 * leave from multicast groups we have joined for the interface
1415 	 */
1416     again:
1417 	mutex_enter(&in6_ifaddr_lock);
1418 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1419 		struct in6_multi *in6m __diagused = imm->i6mm_maddr;
1420 		KASSERTMSG(in6m == NULL || in6m->in6m_ifp == ifp,
1421 		    "in6m_ifp=%s ifp=%s", in6m ? in6m->in6m_ifp->if_xname : NULL,
1422 		    ifp->if_xname);
1423 		LIST_REMOVE(imm, i6mm_chain);
1424 		mutex_exit(&in6_ifaddr_lock);
1425 
1426 		in6_leavegroup(imm);
1427 		goto again;
1428 	}
1429 	mutex_exit(&in6_ifaddr_lock);
1430 
1431 	in6_unlink_ifa(ia, ifp);
1432 }
1433 
1434 static void
1435 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1436 {
1437 	int	s = splsoftnet();
1438 
1439 	mutex_enter(&in6_ifaddr_lock);
1440 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1441 	ifa_remove(ifp, &ia->ia_ifa);
1442 	/* Assume ifa_remove called pserialize_perform and psref_destroy */
1443 	mutex_exit(&in6_ifaddr_lock);
1444 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1445 
1446 	/*
1447 	 * release another refcnt for the link from in6_ifaddr.
1448 	 * Note that we should decrement the refcnt at least once for all *BSD.
1449 	 */
1450 	ifafree(&ia->ia_ifa);
1451 
1452 	splx(s);
1453 }
1454 
1455 void
1456 in6_purgeif(struct ifnet *ifp)
1457 {
1458 
1459 	IFNET_LOCK(ifp);
1460 	in6_ifdetach(ifp);
1461 	IFNET_UNLOCK(ifp);
1462 }
1463 
1464 void
1465 in6_purge_mcast_references(struct in6_multi *in6m)
1466 {
1467 	struct	in6_ifaddr *ia;
1468 
1469 	KASSERT(in6_multi_locked(RW_WRITER));
1470 
1471 	mutex_enter(&in6_ifaddr_lock);
1472 	IN6_ADDRLIST_WRITER_FOREACH(ia) {
1473 		struct in6_multi_mship *imm;
1474 		LIST_FOREACH(imm, &ia->ia6_memberships, i6mm_chain) {
1475 			if (imm->i6mm_maddr == in6m)
1476 				imm->i6mm_maddr = NULL;
1477 		}
1478 	}
1479 	mutex_exit(&in6_ifaddr_lock);
1480 }
1481 
1482 /*
1483  * SIOC[GAD]LIFADDR.
1484  *	SIOCGLIFADDR: get first address. (?)
1485  *	SIOCGLIFADDR with IFLR_PREFIX:
1486  *		get first address that matches the specified prefix.
1487  *	SIOCALIFADDR: add the specified address.
1488  *	SIOCALIFADDR with IFLR_PREFIX:
1489  *		add the specified prefix, filling hostid part from
1490  *		the first link-local address.  prefixlen must be <= 64.
1491  *	SIOCDLIFADDR: delete the specified address.
1492  *	SIOCDLIFADDR with IFLR_PREFIX:
1493  *		delete the first address that matches the specified prefix.
1494  * return values:
1495  *	EINVAL on invalid parameters
1496  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1497  *	other values may be returned from in6_ioctl()
1498  *
1499  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1500  * this is to accommodate address naming scheme other than RFC2374,
1501  * in the future.
1502  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1503  * address encoding scheme. (see figure on page 8)
1504  */
1505 static int
1506 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1507 	struct ifnet *ifp)
1508 {
1509 	struct in6_ifaddr *ia = NULL; /* XXX gcc 4.8 maybe-uninitialized */
1510 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1511 	struct ifaddr *ifa;
1512 	struct sockaddr *sa;
1513 
1514 	/* sanity checks */
1515 	if (!data || !ifp) {
1516 		panic("invalid argument to in6_lifaddr_ioctl");
1517 		/* NOTREACHED */
1518 	}
1519 
1520 	switch (cmd) {
1521 	case SIOCGLIFADDR:
1522 		/* address must be specified on GET with IFLR_PREFIX */
1523 		if ((iflr->flags & IFLR_PREFIX) == 0)
1524 			break;
1525 		/* FALLTHROUGH */
1526 	case SIOCALIFADDR:
1527 	case SIOCDLIFADDR:
1528 		/* address must be specified on ADD and DELETE */
1529 		sa = (struct sockaddr *)&iflr->addr;
1530 		if (sa->sa_family != AF_INET6)
1531 			return EINVAL;
1532 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1533 			return EINVAL;
1534 		/* XXX need improvement */
1535 		sa = (struct sockaddr *)&iflr->dstaddr;
1536 		if (sa->sa_family && sa->sa_family != AF_INET6)
1537 			return EINVAL;
1538 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1539 			return EINVAL;
1540 		break;
1541 	default: /* shouldn't happen */
1542 #if 0
1543 		panic("invalid cmd to in6_lifaddr_ioctl");
1544 		/* NOTREACHED */
1545 #else
1546 		return EOPNOTSUPP;
1547 #endif
1548 	}
1549 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1550 		return EINVAL;
1551 
1552 	switch (cmd) {
1553 	case SIOCALIFADDR:
1554 	    {
1555 		struct in6_aliasreq ifra;
1556 		struct in6_addr *xhostid = NULL;
1557 		int prefixlen;
1558 		int bound = curlwp_bind();
1559 		struct psref psref;
1560 
1561 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1562 			struct sockaddr_in6 *sin6;
1563 
1564 			/*
1565 			 * xhostid is to fill in the hostid part of the
1566 			 * address.  xhostid points to the first link-local
1567 			 * address attached to the interface.
1568 			 */
1569 			ia = in6ifa_ifpforlinklocal_psref(ifp, 0, &psref);
1570 			if (ia == NULL) {
1571 				curlwp_bindx(bound);
1572 				return EADDRNOTAVAIL;
1573 			}
1574 			xhostid = IFA_IN6(&ia->ia_ifa);
1575 
1576 		 	/* prefixlen must be <= 64. */
1577 			if (64 < iflr->prefixlen) {
1578 				ia6_release(ia, &psref);
1579 				curlwp_bindx(bound);
1580 				return EINVAL;
1581 			}
1582 			prefixlen = iflr->prefixlen;
1583 
1584 			/* hostid part must be zero. */
1585 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1586 			if (sin6->sin6_addr.s6_addr32[2] != 0
1587 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1588 				ia6_release(ia, &psref);
1589 				curlwp_bindx(bound);
1590 				return EINVAL;
1591 			}
1592 		} else
1593 			prefixlen = iflr->prefixlen;
1594 
1595 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1596 		memset(&ifra, 0, sizeof(ifra));
1597 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1598 
1599 		memcpy(&ifra.ifra_addr, &iflr->addr,
1600 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1601 		if (xhostid) {
1602 			/* fill in hostid part */
1603 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1604 			    xhostid->s6_addr32[2];
1605 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1606 			    xhostid->s6_addr32[3];
1607 		}
1608 
1609 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1610 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1611 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1612 			if (xhostid) {
1613 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1614 				    xhostid->s6_addr32[2];
1615 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1616 				    xhostid->s6_addr32[3];
1617 			}
1618 		}
1619 		if (xhostid) {
1620 			ia6_release(ia, &psref);
1621 			ia = NULL;
1622 		}
1623 		curlwp_bindx(bound);
1624 
1625 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1626 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1627 
1628 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1629 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1630 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1631 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1632 	    }
1633 	case SIOCGLIFADDR:
1634 	case SIOCDLIFADDR:
1635 	    {
1636 		struct in6_addr mask, candidate, match;
1637 		struct sockaddr_in6 *sin6;
1638 		int cmp;
1639 		int error, s;
1640 
1641 		memset(&mask, 0, sizeof(mask));
1642 		if (iflr->flags & IFLR_PREFIX) {
1643 			/* lookup a prefix rather than address. */
1644 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1645 
1646 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1647 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1648 			match.s6_addr32[0] &= mask.s6_addr32[0];
1649 			match.s6_addr32[1] &= mask.s6_addr32[1];
1650 			match.s6_addr32[2] &= mask.s6_addr32[2];
1651 			match.s6_addr32[3] &= mask.s6_addr32[3];
1652 
1653 			/* if you set extra bits, that's wrong */
1654 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1655 				return EINVAL;
1656 
1657 			cmp = 1;
1658 		} else {
1659 			if (cmd == SIOCGLIFADDR) {
1660 				/* on getting an address, take the 1st match */
1661 				cmp = 0;	/* XXX */
1662 			} else {
1663 				/* on deleting an address, do exact match */
1664 				in6_prefixlen2mask(&mask, 128);
1665 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1666 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1667 
1668 				cmp = 1;
1669 			}
1670 		}
1671 
1672 		s = pserialize_read_enter();
1673 		IFADDR_READER_FOREACH(ifa, ifp) {
1674 			if (ifa->ifa_addr->sa_family != AF_INET6)
1675 				continue;
1676 			if (!cmp)
1677 				break;
1678 
1679 			/*
1680 			 * XXX: this is adhoc, but is necessary to allow
1681 			 * a user to specify fe80::/64 (not /10) for a
1682 			 * link-local address.
1683 			 */
1684 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1685 			in6_clearscope(&candidate);
1686 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1687 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1688 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1689 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1690 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1691 				break;
1692 		}
1693 		if (!ifa) {
1694 			error = EADDRNOTAVAIL;
1695 			goto error;
1696 		}
1697 		ia = ifa2ia6(ifa);
1698 
1699 		if (cmd == SIOCGLIFADDR) {
1700 			/* fill in the if_laddrreq structure */
1701 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1702 			error = sa6_recoverscope(
1703 			    (struct sockaddr_in6 *)&iflr->addr);
1704 			if (error != 0)
1705 				goto error;
1706 
1707 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1708 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1709 				    ia->ia_dstaddr.sin6_len);
1710 				error = sa6_recoverscope(
1711 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1712 				if (error != 0)
1713 					goto error;
1714 			} else
1715 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1716 
1717 			iflr->prefixlen =
1718 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1719 
1720 			iflr->flags = ia->ia6_flags;	/* XXX */
1721 
1722 			error = 0;
1723 		} else {
1724 			struct in6_aliasreq ifra;
1725 
1726 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1727 			memset(&ifra, 0, sizeof(ifra));
1728 			memcpy(ifra.ifra_name, iflr->iflr_name,
1729 			    sizeof(ifra.ifra_name));
1730 
1731 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1732 			    ia->ia_addr.sin6_len);
1733 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1734 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1735 				    ia->ia_dstaddr.sin6_len);
1736 			} else {
1737 				memset(&ifra.ifra_dstaddr, 0,
1738 				    sizeof(ifra.ifra_dstaddr));
1739 			}
1740 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1741 			    ia->ia_prefixmask.sin6_len);
1742 
1743 			ifra.ifra_flags = ia->ia6_flags;
1744 			pserialize_read_exit(s);
1745 
1746 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1747 		}
1748 	error:
1749 		pserialize_read_exit(s);
1750 		return error;
1751 	    }
1752 	}
1753 
1754 	return EOPNOTSUPP;	/* just for safety */
1755 }
1756 
1757 /*
1758  * Initialize an interface's internet6 address
1759  * and routing table entry.
1760  */
1761 static int
1762 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1763 	const struct sockaddr_in6 *sin6, int newhost)
1764 {
1765 	int	error = 0, ifacount = 0;
1766 	int s;
1767 	struct ifaddr *ifa;
1768 
1769 	KASSERT(mutex_owned(&in6_ifaddr_lock));
1770 
1771 	/*
1772 	 * Give the interface a chance to initialize
1773 	 * if this is its first address,
1774 	 * and to validate the address if necessary.
1775 	 */
1776 	s = pserialize_read_enter();
1777 	IFADDR_READER_FOREACH(ifa, ifp) {
1778 		if (ifa->ifa_addr->sa_family != AF_INET6)
1779 			continue;
1780 		ifacount++;
1781 	}
1782 	pserialize_read_exit(s);
1783 
1784 	ia->ia_addr = *sin6;
1785 
1786 	if (ifacount == 0 &&
1787 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1788 		return error;
1789 	}
1790 
1791 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1792 
1793 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1794 
1795 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1796 	if (newhost) {
1797 		/* set the rtrequest function to create llinfo */
1798 		if (ifp->if_flags & IFF_POINTOPOINT)
1799 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1800 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1801 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1802 		in6_ifaddlocal(&ia->ia_ifa);
1803 	} else {
1804 		/* Inform the routing socket of new flags/timings */
1805 		rt_addrmsg(RTM_NEWADDR, &ia->ia_ifa);
1806 	}
1807 
1808 	/* Add the network prefix route. */
1809 	if ((error = in6_ifaddprefix(ia)) != 0) {
1810 		if (newhost)
1811 			in6_ifremlocal(&ia->ia_ifa);
1812 		return error;
1813 	}
1814 
1815 	return error;
1816 }
1817 
1818 static struct ifaddr *
1819 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1820 {
1821 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1822 		return ifa;
1823 	return best_ifa;
1824 }
1825 
1826 /*
1827  * Find an IPv6 interface link-local address specific to an interface.
1828  */
1829 struct in6_ifaddr *
1830 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1831 {
1832 	struct ifaddr *best_ifa = NULL, *ifa;
1833 
1834 	IFADDR_READER_FOREACH(ifa, ifp) {
1835 		if (ifa->ifa_addr->sa_family != AF_INET6)
1836 			continue;
1837 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1838 			continue;
1839 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1840 			continue;
1841 		best_ifa = bestifa(best_ifa, ifa);
1842 	}
1843 
1844 	return (struct in6_ifaddr *)best_ifa;
1845 }
1846 
1847 struct in6_ifaddr *
1848 in6ifa_ifpforlinklocal_psref(const struct ifnet *ifp, const int ignoreflags,
1849     struct psref *psref)
1850 {
1851 	struct in6_ifaddr *ia;
1852 	int s = pserialize_read_enter();
1853 
1854 	ia = in6ifa_ifpforlinklocal(ifp, ignoreflags);
1855 	if (ia != NULL)
1856 		ia6_acquire(ia, psref);
1857 	pserialize_read_exit(s);
1858 
1859 	return ia;
1860 }
1861 
1862 /*
1863  * find the internet address corresponding to a given address.
1864  * ifaddr is returned referenced.
1865  */
1866 struct in6_ifaddr *
1867 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1868 {
1869 	struct in6_ifaddr *ia;
1870 	int s;
1871 
1872 	s = pserialize_read_enter();
1873 	IN6_ADDRLIST_READER_FOREACH(ia) {
1874 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1875 			if (zoneid != 0 &&
1876 			    zoneid != ia->ia_addr.sin6_scope_id)
1877 				continue;
1878 			ifaref(&ia->ia_ifa);
1879 			break;
1880 		}
1881 	}
1882 	pserialize_read_exit(s);
1883 
1884 	return ia;
1885 }
1886 
1887 /*
1888  * find the internet address corresponding to a given interface and address.
1889  */
1890 struct in6_ifaddr *
1891 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1892 {
1893 	struct ifaddr *best_ifa = NULL, *ifa;
1894 
1895 	IFADDR_READER_FOREACH(ifa, ifp) {
1896 		if (ifa->ifa_addr->sa_family != AF_INET6)
1897 			continue;
1898 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1899 			continue;
1900 		best_ifa = bestifa(best_ifa, ifa);
1901 	}
1902 
1903 	return (struct in6_ifaddr *)best_ifa;
1904 }
1905 
1906 struct in6_ifaddr *
1907 in6ifa_ifpwithaddr_psref(const struct ifnet *ifp, const struct in6_addr *addr,
1908     struct psref *psref)
1909 {
1910 	struct in6_ifaddr *ia;
1911 	int s = pserialize_read_enter();
1912 
1913 	ia = in6ifa_ifpwithaddr(ifp, addr);
1914 	if (ia != NULL)
1915 		ia6_acquire(ia, psref);
1916 	pserialize_read_exit(s);
1917 
1918 	return ia;
1919 }
1920 
1921 static struct in6_ifaddr *
1922 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1923 {
1924 	if (best_ia == NULL ||
1925 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1926 		return ia;
1927 	return best_ia;
1928 }
1929 
1930 /*
1931  * Determine if an address is on a local network.
1932  */
1933 int
1934 in6_localaddr(const struct in6_addr *in6)
1935 {
1936 	struct in6_ifaddr *ia;
1937 	int s;
1938 
1939 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1940 		return 1;
1941 
1942 	s = pserialize_read_enter();
1943 	IN6_ADDRLIST_READER_FOREACH(ia) {
1944 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1945 					      &ia->ia_prefixmask.sin6_addr)) {
1946 			pserialize_read_exit(s);
1947 			return 1;
1948 		}
1949 	}
1950 	pserialize_read_exit(s);
1951 
1952 	return 0;
1953 }
1954 
1955 int
1956 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1957 {
1958 	struct in6_ifaddr *ia;
1959 	int s;
1960 
1961 	s = pserialize_read_enter();
1962 	IN6_ADDRLIST_READER_FOREACH(ia) {
1963 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1964 		    &sa6->sin6_addr) &&
1965 #ifdef SCOPEDROUTING
1966 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1967 #endif
1968 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
1969 			pserialize_read_exit(s);
1970 			return 1; /* true */
1971 		}
1972 
1973 		/* XXX: do we still have to go thru the rest of the list? */
1974 	}
1975 	pserialize_read_exit(s);
1976 
1977 	return 0;		/* false */
1978 }
1979 
1980 /*
1981  * return length of part which dst and src are equal
1982  * hard coding...
1983  */
1984 int
1985 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1986 {
1987 	int match = 0;
1988 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1989 	u_char *lim = s + 16, r;
1990 
1991 	while (s < lim)
1992 		if ((r = (*d++ ^ *s++)) != 0) {
1993 			while (r < 128) {
1994 				match++;
1995 				r <<= 1;
1996 			}
1997 			break;
1998 		} else
1999 			match += NBBY;
2000 	return match;
2001 }
2002 
2003 void
2004 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2005 {
2006 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2007 	int bytelen, bitlen, i;
2008 
2009 	/* sanity check */
2010 	if (len < 0 || len > 128) {
2011 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2012 		    len);
2013 		return;
2014 	}
2015 
2016 	memset(maskp, 0, sizeof(*maskp));
2017 	bytelen = len / NBBY;
2018 	bitlen = len % NBBY;
2019 	for (i = 0; i < bytelen; i++)
2020 		maskp->s6_addr[i] = 0xff;
2021 	if (bitlen)
2022 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2023 }
2024 
2025 /*
2026  * return the best address out of the same scope. if no address was
2027  * found, return the first valid address from designated IF.
2028  */
2029 struct in6_ifaddr *
2030 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2031 {
2032 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2033 	struct ifaddr *ifa;
2034 	struct in6_ifaddr *best_ia = NULL, *ia;
2035 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2036 
2037 	dep[0] = dep[1] = NULL;
2038 
2039 	/*
2040 	 * We first look for addresses in the same scope.
2041 	 * If there is one, return it.
2042 	 * If two or more, return one which matches the dst longest.
2043 	 * If none, return one of global addresses assigned other ifs.
2044 	 */
2045 	IFADDR_READER_FOREACH(ifa, ifp) {
2046 		if (ifa->ifa_addr->sa_family != AF_INET6)
2047 			continue;
2048 		ia = (struct in6_ifaddr *)ifa;
2049 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2050 			continue; /* XXX: is there any case to allow anycast? */
2051 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2052 			continue; /* don't use this interface */
2053 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2054 			continue;
2055 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2056 			if (ip6_use_deprecated)
2057 				dep[0] = ia;
2058 			continue;
2059 		}
2060 
2061 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
2062 			continue;
2063 		/*
2064 		 * call in6_matchlen() as few as possible
2065 		 */
2066 		if (best_ia == NULL) {
2067 			best_ia = ia;
2068 			continue;
2069 		}
2070 		if (blen == -1)
2071 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2072 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2073 		if (tlen > blen) {
2074 			blen = tlen;
2075 			best_ia = ia;
2076 		} else if (tlen == blen)
2077 			best_ia = bestia(best_ia, ia);
2078 	}
2079 	if (best_ia != NULL)
2080 		return best_ia;
2081 
2082 	IFADDR_READER_FOREACH(ifa, ifp) {
2083 		if (ifa->ifa_addr->sa_family != AF_INET6)
2084 			continue;
2085 		ia = (struct in6_ifaddr *)ifa;
2086 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2087 			continue; /* XXX: is there any case to allow anycast? */
2088 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2089 			continue; /* don't use this interface */
2090 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2091 			continue;
2092 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2093 			if (ip6_use_deprecated)
2094 				dep[1] = (struct in6_ifaddr *)ifa;
2095 			continue;
2096 		}
2097 
2098 		best_ia = bestia(best_ia, ia);
2099 	}
2100 	if (best_ia != NULL)
2101 		return best_ia;
2102 
2103 	/* use the last-resort values, that are, deprecated addresses */
2104 	if (dep[0])
2105 		return dep[0];
2106 	if (dep[1])
2107 		return dep[1];
2108 
2109 	return NULL;
2110 }
2111 
2112 /*
2113  * perform DAD when interface becomes IFF_UP.
2114  */
2115 void
2116 in6_if_link_up(struct ifnet *ifp)
2117 {
2118 	struct ifaddr *ifa;
2119 	struct in6_ifaddr *ia;
2120 	int s, bound;
2121 	char ip6buf[INET6_ADDRSTRLEN];
2122 
2123 	/* Ensure it's sane to run DAD */
2124 	if (ifp->if_link_state == LINK_STATE_DOWN)
2125 		return;
2126 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2127 		return;
2128 
2129 	bound = curlwp_bind();
2130 	s = pserialize_read_enter();
2131 	IFADDR_READER_FOREACH(ifa, ifp) {
2132 		struct psref psref;
2133 
2134 		if (ifa->ifa_addr->sa_family != AF_INET6)
2135 			continue;
2136 
2137 		ifa_acquire(ifa, &psref);
2138 		pserialize_read_exit(s);
2139 		ia = (struct in6_ifaddr *)ifa;
2140 
2141 		/* If detached then mark as tentative */
2142 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2143 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2144 			if (ip6_dad_enabled() && if_do_dad(ifp)) {
2145 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2146 				nd6log(LOG_ERR, "%s marked tentative\n",
2147 				    IN6_PRINT(ip6buf,
2148 				    &ia->ia_addr.sin6_addr));
2149 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2150 				rt_addrmsg(RTM_NEWADDR, ifa);
2151 		}
2152 
2153 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2154 			int rand_delay;
2155 
2156 			/* Clear the duplicated flag as we're starting DAD. */
2157 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2158 
2159 			/*
2160 			 * The TENTATIVE flag was likely set by hand
2161 			 * beforehand, implicitly indicating the need for DAD.
2162 			 * We may be able to skip the random delay in this
2163 			 * case, but we impose delays just in case.
2164 			 */
2165 			rand_delay = cprng_fast32() %
2166 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2167 			/* +1 ensures callout is always used */
2168 			nd6_dad_start(ifa, rand_delay + 1);
2169 		}
2170 
2171 		s = pserialize_read_enter();
2172 		ifa_release(ifa, &psref);
2173 	}
2174 	pserialize_read_exit(s);
2175 	curlwp_bindx(bound);
2176 }
2177 
2178 void
2179 in6_if_up(struct ifnet *ifp)
2180 {
2181 
2182 	/*
2183 	 * special cases, like 6to4, are handled in in6_ifattach
2184 	 */
2185 	in6_ifattach(ifp, NULL);
2186 
2187 	/* interface may not support link state, so bring it up also */
2188 	in6_if_link_up(ifp);
2189 }
2190 
2191 /*
2192  * Mark all addresses as detached.
2193  */
2194 void
2195 in6_if_link_down(struct ifnet *ifp)
2196 {
2197 	struct ifaddr *ifa;
2198 	struct in6_ifaddr *ia;
2199 	int s, bound;
2200 	char ip6buf[INET6_ADDRSTRLEN];
2201 
2202 	bound = curlwp_bind();
2203 	s = pserialize_read_enter();
2204 	IFADDR_READER_FOREACH(ifa, ifp) {
2205 		struct psref psref;
2206 
2207 		if (ifa->ifa_addr->sa_family != AF_INET6)
2208 			continue;
2209 
2210 		ifa_acquire(ifa, &psref);
2211 		pserialize_read_exit(s);
2212 		ia = (struct in6_ifaddr *)ifa;
2213 
2214 		/* Stop DAD processing */
2215 		nd6_dad_stop(ifa);
2216 
2217 		/*
2218 		 * Mark the address as detached.
2219 		 * This satisfies RFC4862 Section 5.3, but we should apply
2220 		 * this logic to all addresses to be a good citizen and
2221 		 * avoid potential duplicated addresses.
2222 		 * When the interface comes up again, detached addresses
2223 		 * are marked tentative and DAD commences.
2224 		 */
2225 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2226 			nd6log(LOG_DEBUG, "%s marked detached\n",
2227 			    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
2228 			ia->ia6_flags |= IN6_IFF_DETACHED;
2229 			ia->ia6_flags &=
2230 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2231 			rt_addrmsg(RTM_NEWADDR, ifa);
2232 		}
2233 
2234 		s = pserialize_read_enter();
2235 		ifa_release(ifa, &psref);
2236 	}
2237 	pserialize_read_exit(s);
2238 	curlwp_bindx(bound);
2239 }
2240 
2241 void
2242 in6_if_down(struct ifnet *ifp)
2243 {
2244 
2245 	in6_if_link_down(ifp);
2246 	lltable_purge_entries(LLTABLE6(ifp));
2247 }
2248 
2249 void
2250 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2251 {
2252 
2253 	switch (link_state) {
2254 	case LINK_STATE_DOWN:
2255 		in6_if_link_down(ifp);
2256 		break;
2257 	case LINK_STATE_UP:
2258 		in6_if_link_up(ifp);
2259 		break;
2260 	}
2261 }
2262 
2263 int
2264 in6_tunnel_validate(const struct ip6_hdr *ip6, const struct in6_addr *src,
2265     const struct in6_addr *dst)
2266 {
2267 
2268 	/* check for address match */
2269 	if (!IN6_ARE_ADDR_EQUAL(src, &ip6->ip6_dst) ||
2270 	    !IN6_ARE_ADDR_EQUAL(dst, &ip6->ip6_src))
2271 		return 0;
2272 
2273 	/* martian filters on outer source - done in ip6_input */
2274 
2275 	/* NOTE: the packet may be dropped by uRPF. */
2276 
2277 	/* return valid bytes length */
2278 	return sizeof(*src) + sizeof(*dst);
2279 }
2280 
2281 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2282 #define	IN6_LLTBL_HASH(k, h) \
2283 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2284 
2285 /*
2286  * Do actual deallocation of @lle.
2287  * Called by LLE_FREE_LOCKED when number of references
2288  * drops to zero.
2289  */
2290 static void
2291 in6_lltable_destroy_lle(struct llentry *lle)
2292 {
2293 
2294 	KASSERTMSG(lle->la_numheld == 0, "la_numheld=%d", lle->la_numheld);
2295 
2296 	LLE_WUNLOCK(lle);
2297 	LLE_LOCK_DESTROY(lle);
2298 	llentry_pool_put(lle);
2299 }
2300 
2301 static struct llentry *
2302 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2303 {
2304 	struct llentry *lle;
2305 
2306 	lle = llentry_pool_get(PR_NOWAIT);
2307 	if (lle == NULL)		/* NB: caller generates msg */
2308 		return NULL;
2309 
2310 	lle->r_l3addr.addr6 = *addr6;
2311 	lle->lle_refcnt = 1;
2312 	lle->lle_free = in6_lltable_destroy_lle;
2313 	LLE_LOCK_INIT(lle);
2314 	callout_init(&lle->lle_timer, CALLOUT_MPSAFE);
2315 
2316 	return lle;
2317 }
2318 
2319 static int
2320 in6_lltable_match_prefix(const struct sockaddr *prefix,
2321     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2322 {
2323 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2324 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2325 
2326 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2327 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2328 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2329 		return 1;
2330 
2331 	return 0;
2332 }
2333 
2334 static void
2335 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2336 {
2337 
2338 	LLE_WLOCK_ASSERT(lle);
2339 	(void) llentry_free(lle);
2340 }
2341 
2342 static int
2343 in6_lltable_rtcheck(struct ifnet *ifp, u_int flags,
2344     const struct sockaddr *l3addr, const struct rtentry *rt)
2345 {
2346 	char ip6buf[INET6_ADDRSTRLEN];
2347 
2348 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2349 		int s;
2350 		struct ifaddr *ifa;
2351 		/*
2352 		 * Create an ND6 cache for an IPv6 neighbor
2353 		 * that is not covered by our own prefix.
2354 		 */
2355 		/* XXX ifaof_ifpforaddr should take a const param */
2356 		s = pserialize_read_enter();
2357 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2358 		if (ifa != NULL) {
2359 			pserialize_read_exit(s);
2360 			return 0;
2361 		}
2362 		pserialize_read_exit(s);
2363 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2364 		    IN6_PRINT(ip6buf,
2365 		    &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2366 		return EINVAL;
2367 	}
2368 	return 0;
2369 }
2370 
2371 static inline uint32_t
2372 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2373 {
2374 
2375 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2376 }
2377 
2378 static uint32_t
2379 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2380 {
2381 
2382 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2383 }
2384 
2385 static void
2386 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2387 {
2388 	struct sockaddr_in6 *sin6;
2389 
2390 	sin6 = (struct sockaddr_in6 *)sa;
2391 	bzero(sin6, sizeof(*sin6));
2392 	sin6->sin6_family = AF_INET6;
2393 	sin6->sin6_len = sizeof(*sin6);
2394 	sin6->sin6_addr = lle->r_l3addr.addr6;
2395 }
2396 
2397 static inline struct llentry *
2398 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2399 {
2400 	struct llentry *lle;
2401 	struct llentries *lleh;
2402 	u_int hashidx;
2403 
2404 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2405 	lleh = &llt->lle_head[hashidx];
2406 	LIST_FOREACH(lle, lleh, lle_next) {
2407 		if (lle->la_flags & LLE_DELETED)
2408 			continue;
2409 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2410 			break;
2411 	}
2412 
2413 	return lle;
2414 }
2415 
2416 static int
2417 in6_lltable_delete(struct lltable *llt, u_int flags,
2418 	const struct sockaddr *l3addr)
2419 {
2420 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2421 	struct llentry *lle;
2422 
2423 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2424 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2425 	    "sin_family %d", l3addr->sa_family);
2426 
2427 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2428 
2429 	if (lle == NULL) {
2430 #ifdef LLTABLE_DEBUG
2431 		char buf[64];
2432 		sockaddr_format(l3addr, buf, sizeof(buf));
2433 		log(LOG_INFO, "%s: cache for %s is not found\n",
2434 		    __func__, buf);
2435 #endif
2436 		return ENOENT;
2437 	}
2438 
2439 	LLE_WLOCK(lle);
2440 #ifdef LLTABLE_DEBUG
2441 	{
2442 		char buf[64];
2443 		sockaddr_format(l3addr, buf, sizeof(buf));
2444 		log(LOG_INFO, "%s: cache for %s (%p) is deleted\n",
2445 		    __func__, buf, lle);
2446 	}
2447 #endif
2448 	llentry_free(lle);
2449 
2450 	return 0;
2451 }
2452 
2453 static struct llentry *
2454 in6_lltable_create(struct lltable *llt, u_int flags,
2455     const struct sockaddr *l3addr, const struct rtentry *rt)
2456 {
2457 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2458 	struct ifnet *ifp = llt->llt_ifp;
2459 	struct llentry *lle;
2460 
2461 	IF_AFDATA_WLOCK_ASSERT(ifp);
2462 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2463 	    "sin_family %d", l3addr->sa_family);
2464 
2465 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2466 
2467 	if (lle != NULL) {
2468 		LLE_WLOCK(lle);
2469 		return lle;
2470 	}
2471 
2472 	/*
2473 	 * A route that covers the given address must have
2474 	 * been installed 1st because we are doing a resolution,
2475 	 * verify this.
2476 	 */
2477 	if (!(flags & LLE_IFADDR) &&
2478 	    in6_lltable_rtcheck(ifp, flags, l3addr, rt) != 0)
2479 		return NULL;
2480 
2481 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2482 	if (lle == NULL) {
2483 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2484 		return NULL;
2485 	}
2486 	lle->la_flags = flags;
2487 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2488 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2489 		lle->la_flags |= LLE_VALID;
2490 	}
2491 
2492 	lltable_link_entry(llt, lle);
2493 	LLE_WLOCK(lle);
2494 
2495 	return lle;
2496 }
2497 
2498 static struct llentry *
2499 in6_lltable_lookup(struct lltable *llt, u_int flags,
2500 	const struct sockaddr *l3addr)
2501 {
2502 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2503 	struct llentry *lle;
2504 
2505 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2506 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2507 	    "sin_family %d", l3addr->sa_family);
2508 
2509 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2510 
2511 	if (lle == NULL)
2512 		return NULL;
2513 
2514 	if (flags & LLE_EXCLUSIVE)
2515 		LLE_WLOCK(lle);
2516 	else
2517 		LLE_RLOCK(lle);
2518 	return lle;
2519 }
2520 
2521 static int
2522 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2523     struct rt_walkarg *w)
2524 {
2525 	struct sockaddr_in6 sin6;
2526 
2527 	LLTABLE_LOCK_ASSERT();
2528 
2529 	/* skip deleted entries */
2530 	if (lle->la_flags & LLE_DELETED)
2531 		return 0;
2532 
2533 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2534 
2535 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2536 }
2537 
2538 static struct lltable *
2539 in6_lltattach(struct ifnet *ifp)
2540 {
2541 	struct lltable *llt;
2542 
2543 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2544 	llt->llt_af = AF_INET6;
2545 	llt->llt_ifp = ifp;
2546 
2547 	llt->llt_lookup = in6_lltable_lookup;
2548 	llt->llt_create = in6_lltable_create;
2549 	llt->llt_delete = in6_lltable_delete;
2550 	llt->llt_dump_entry = in6_lltable_dump_entry;
2551 	llt->llt_hash = in6_lltable_hash;
2552 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2553 	llt->llt_free_entry = in6_lltable_free_entry;
2554 	llt->llt_match_prefix = in6_lltable_match_prefix;
2555 	lltable_link(llt);
2556 
2557 	return llt;
2558 }
2559 
2560 void *
2561 in6_domifattach(struct ifnet *ifp)
2562 {
2563 	struct in6_ifextra *ext;
2564 
2565 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2566 
2567 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2568 	    M_IFADDR, M_WAITOK|M_ZERO);
2569 
2570 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2571 	    M_IFADDR, M_WAITOK|M_ZERO);
2572 
2573 	ext->nd_ifinfo = nd6_ifattach(ifp);
2574 	ext->scope6_id = scope6_ifattach(ifp);
2575 	ext->lltable = in6_lltattach(ifp);
2576 
2577 	return ext;
2578 }
2579 
2580 void
2581 in6_domifdetach(struct ifnet *ifp, void *aux)
2582 {
2583 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2584 
2585 	lltable_free(ext->lltable);
2586 	ext->lltable = NULL;
2587 	SOFTNET_LOCK_UNLESS_NET_MPSAFE();
2588 	nd6_ifdetach(ifp, ext);
2589 	SOFTNET_UNLOCK_UNLESS_NET_MPSAFE();
2590 	free(ext->in6_ifstat, M_IFADDR);
2591 	free(ext->icmp6_ifstat, M_IFADDR);
2592 	scope6_ifdetach(ext->scope6_id);
2593 	free(ext, M_IFADDR);
2594 }
2595 
2596 /*
2597  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2598  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2599  */
2600 void
2601 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2602 {
2603 	in6->s6_addr32[0] = 0;
2604 	in6->s6_addr32[1] = 0;
2605 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2606 	in6->s6_addr32[3] = in->s_addr;
2607 }
2608 
2609 /*
2610  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2611  * v4 mapped addr or v4 compat addr
2612  */
2613 void
2614 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2615 {
2616 	memset(sin, 0, sizeof(*sin));
2617 	sin->sin_len = sizeof(struct sockaddr_in);
2618 	sin->sin_family = AF_INET;
2619 	sin->sin_port = sin6->sin6_port;
2620 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2621 }
2622 
2623 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2624 void
2625 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2626 {
2627 	memset(sin6, 0, sizeof(*sin6));
2628 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2629 	sin6->sin6_family = AF_INET6;
2630 	sin6->sin6_port = sin->sin_port;
2631 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2632 }
2633 
2634 /* Convert sockaddr_in6 into sockaddr_in. */
2635 void
2636 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2637 {
2638 	struct sockaddr_in *sin_p;
2639 	struct sockaddr_in6 sin6;
2640 
2641 	/*
2642 	 * Save original sockaddr_in6 addr and convert it
2643 	 * to sockaddr_in.
2644 	 */
2645 	sin6 = *(struct sockaddr_in6 *)nam;
2646 	sin_p = (struct sockaddr_in *)nam;
2647 	in6_sin6_2_sin(sin_p, &sin6);
2648 }
2649 
2650 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2651 void
2652 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2653 {
2654 	struct sockaddr_in *sin_p;
2655 	struct sockaddr_in6 *sin6_p;
2656 
2657 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2658 	sin_p = (struct sockaddr_in *)*nam;
2659 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2660 	free(*nam, M_SONAME);
2661 	*nam = sin6tosa(sin6_p);
2662 }
2663