xref: /netbsd-src/sys/netinet6/in6.c (revision e89934bbf778a6d6d6894877c4da59d0c7835b0f)
1 /*	$NetBSD: in6.c,v 1.237 2017/01/23 10:19:03 ozaki-r 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 1.237 2017/01/23 10:19:03 ozaki-r 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 #include <net/net_osdep.h>
109 
110 #ifdef COMPAT_50
111 #include <compat/netinet6/in6_var.h>
112 #endif
113 
114 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
115 
116 /* enable backward compatibility code for obsoleted ioctls */
117 #define COMPAT_IN6IFIOCTL
118 
119 #ifdef	IN6_DEBUG
120 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
121 #else
122 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
123 #endif /* IN6_DEBUG */
124 
125 /*
126  * Definitions of some constant IP6 addresses.
127  */
128 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
129 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
130 const struct in6_addr in6addr_nodelocal_allnodes =
131 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
132 const struct in6_addr in6addr_linklocal_allnodes =
133 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
134 const struct in6_addr in6addr_linklocal_allrouters =
135 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
136 
137 const struct in6_addr in6mask0 = IN6MASK0;
138 const struct in6_addr in6mask32 = IN6MASK32;
139 const struct in6_addr in6mask64 = IN6MASK64;
140 const struct in6_addr in6mask96 = IN6MASK96;
141 const struct in6_addr in6mask128 = IN6MASK128;
142 
143 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
144 				     0, 0, IN6ADDR_ANY_INIT, 0};
145 
146 struct pslist_head	in6_ifaddr_list;
147 kmutex_t		in6_ifaddr_lock;
148 
149 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
150 	struct ifnet *);
151 static int in6_ifaddprefix(struct in6_ifaddr *);
152 static int in6_ifremprefix(struct in6_ifaddr *);
153 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
154 	const struct sockaddr_in6 *, int);
155 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
156 static int in6_update_ifa1(struct ifnet *, struct in6_aliasreq *,
157     struct in6_ifaddr **, struct psref *, int);
158 
159 void
160 in6_init(void)
161 {
162 
163 	PSLIST_INIT(&in6_ifaddr_list);
164 	mutex_init(&in6_ifaddr_lock, MUTEX_DEFAULT, IPL_NONE);
165 
166 	in6_sysctl_multicast_setup(NULL);
167 }
168 
169 /*
170  * Add ownaddr as loopback rtentry.  We previously add the route only if
171  * necessary (ex. on a p2p link).  However, since we now manage addresses
172  * separately from prefixes, we should always add the route.  We can't
173  * rely on the cloning mechanism from the corresponding interface route
174  * any more.
175  */
176 void
177 in6_ifaddlocal(struct ifaddr *ifa)
178 {
179 
180 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
181 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
182 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
183 	{
184 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
185 		return;
186 	}
187 
188 	rt_ifa_addlocal(ifa);
189 }
190 
191 /*
192  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
193  * if it exists.
194  */
195 void
196 in6_ifremlocal(struct ifaddr *ifa)
197 {
198 	struct in6_ifaddr *ia;
199 	struct ifaddr *alt_ifa = NULL;
200 	int ia_count = 0;
201 	struct psref psref;
202 	int s;
203 
204 	/*
205 	 * Some of BSD variants do not remove cloned routes
206 	 * from an interface direct route, when removing the direct route
207 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
208 	 * cloned routes, they could fail to remove the cloned routes when
209 	 * we handle multple addresses that share a common prefix.
210 	 * So, we should remove the route corresponding to the deleted address.
211 	 */
212 
213 	/*
214 	 * Delete the entry only if exactly one ifaddr matches the
215 	 * address, ifa->ifa_addr.
216 	 *
217 	 * If more than one ifaddr matches, replace the ifaddr in
218 	 * the routing table, rt_ifa, with a different ifaddr than
219 	 * the one we are purging, ifa.  It is important to do
220 	 * this, or else the routing table can accumulate dangling
221 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
222 	 * which will lead to crashes, later.  (More than one ifaddr
223 	 * can match if we assign the same address to multiple---probably
224 	 * p2p---interfaces.)
225 	 *
226 	 * XXX An old comment at this place said, "we should avoid
227 	 * XXX such a configuration [i.e., interfaces with the same
228 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
229 	 * XXX agree, especially now that I have fixed the dangling
230 	 * XXX ifp-pointers bug.
231 	 */
232 	s = pserialize_read_enter();
233 	IN6_ADDRLIST_READER_FOREACH(ia) {
234 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
235 			continue;
236 		if (ia->ia_ifp != ifa->ifa_ifp)
237 			alt_ifa = &ia->ia_ifa;
238 		if (++ia_count > 1 && alt_ifa != NULL)
239 			break;
240 	}
241 	if (ia_count > 1 && alt_ifa != NULL)
242 		ifa_acquire(alt_ifa, &psref);
243 	pserialize_read_exit(s);
244 
245 	if (ia_count == 0)
246 		return;
247 
248 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
249 
250 	if (ia_count > 1 && alt_ifa != NULL)
251 		ifa_release(alt_ifa, &psref);
252 }
253 
254 /* Add prefix route for the network. */
255 static int
256 in6_ifaddprefix(struct in6_ifaddr *ia)
257 {
258 	int error, flags = 0;
259 
260 	if (in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) == 128) {
261 		if (ia->ia_dstaddr.sin6_family != AF_INET6)
262 			/* We don't need to install a host route. */
263 			return 0;
264 		flags |= RTF_HOST;
265 	}
266 
267 	/* Is this a connected route for neighbour discovery? */
268 	if (nd6_need_cache(ia->ia_ifp))
269 		flags |= RTF_CONNECTED;
270 
271 	if ((error = rtinit(&ia->ia_ifa, RTM_ADD, RTF_UP | flags)) == 0)
272 		ia->ia_flags |= IFA_ROUTE;
273 	else if (error == EEXIST)
274 		/* Existance of the route is not an error. */
275 		error = 0;
276 
277 	return error;
278 }
279 
280 /* Delete network prefix route if present.
281  * Re-add it to another address if the prefix matches. */
282 static int
283 in6_ifremprefix(struct in6_ifaddr *target)
284 {
285 	int error, s;
286 	struct in6_ifaddr *ia;
287 
288 	if ((target->ia_flags & IFA_ROUTE) == 0)
289 		return 0;
290 
291 	s = pserialize_read_enter();
292 	IN6_ADDRLIST_READER_FOREACH(ia) {
293 		if (target->ia_dstaddr.sin6_len) {
294 			if (ia->ia_dstaddr.sin6_len == 0 ||
295 			    !IN6_ARE_ADDR_EQUAL(&ia->ia_dstaddr.sin6_addr,
296 			    &target->ia_dstaddr.sin6_addr))
297 				continue;
298 		} else {
299 			if (!IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
300 			    &target->ia_addr.sin6_addr,
301 			    &target->ia_prefixmask.sin6_addr))
302 				continue;
303 		}
304 
305 		/*
306 		 * if we got a matching prefix route, move IFA_ROUTE to him
307 		 */
308 		if ((ia->ia_flags & IFA_ROUTE) == 0) {
309 			struct psref psref;
310 			int bound = curlwp_bind();
311 
312 			ia6_acquire(ia, &psref);
313 			pserialize_read_exit(s);
314 
315 			rtinit(&target->ia_ifa, RTM_DELETE, 0);
316 			target->ia_flags &= ~IFA_ROUTE;
317 
318 			error = in6_ifaddprefix(ia);
319 
320 			ia6_release(ia, &psref);
321 			curlwp_bindx(bound);
322 
323 			return error;
324 		}
325 	}
326 	pserialize_read_exit(s);
327 
328 	/*
329 	 * noone seem to have prefix route.  remove it.
330 	 */
331 	rtinit(&target->ia_ifa, RTM_DELETE, 0);
332 	target->ia_flags &= ~IFA_ROUTE;
333 	return 0;
334 }
335 
336 int
337 in6_mask2len(struct in6_addr *mask, u_char *lim0)
338 {
339 	int x = 0, y;
340 	u_char *lim = lim0, *p;
341 
342 	/* ignore the scope_id part */
343 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
344 		lim = (u_char *)mask + sizeof(*mask);
345 	for (p = (u_char *)mask; p < lim; x++, p++) {
346 		if (*p != 0xff)
347 			break;
348 	}
349 	y = 0;
350 	if (p < lim) {
351 		for (y = 0; y < NBBY; y++) {
352 			if ((*p & (0x80 >> y)) == 0)
353 				break;
354 		}
355 	}
356 
357 	/*
358 	 * when the limit pointer is given, do a stricter check on the
359 	 * remaining bits.
360 	 */
361 	if (p < lim) {
362 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
363 			return -1;
364 		for (p = p + 1; p < lim; p++)
365 			if (*p != 0)
366 				return -1;
367 	}
368 
369 	return x * NBBY + y;
370 }
371 
372 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
373 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
374 
375 static int
376 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
377 {
378 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
379 	struct	in6_ifaddr *ia = NULL;
380 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
381 	struct sockaddr_in6 *sa6;
382 	int error, bound;
383 	struct psref psref;
384 
385 	switch (cmd) {
386 	case SIOCAADDRCTL_POLICY:
387 	case SIOCDADDRCTL_POLICY:
388 		/* Privileged. */
389 		return in6_src_ioctl(cmd, data);
390 	/*
391 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
392 	 */
393 	case SIOCSIFADDR:
394 	case SIOCSIFDSTADDR:
395 	case SIOCSIFBRDADDR:
396 	case SIOCSIFNETMASK:
397 		return EOPNOTSUPP;
398 	case SIOCGETSGCNT_IN6:
399 	case SIOCGETMIFCNT_IN6:
400 		return mrt6_ioctl(cmd, data);
401 	case SIOCGIFADDRPREF:
402 	case SIOCSIFADDRPREF:
403 		if (ifp == NULL)
404 			return EINVAL;
405 		return ifaddrpref_ioctl(so, cmd, data, ifp);
406 	}
407 
408 	if (ifp == NULL)
409 		return EOPNOTSUPP;
410 
411 	switch (cmd) {
412 	case SIOCSNDFLUSH_IN6:
413 	case SIOCSPFXFLUSH_IN6:
414 	case SIOCSRTRFLUSH_IN6:
415 	case SIOCSDEFIFACE_IN6:
416 	case SIOCSIFINFO_FLAGS:
417 	case SIOCSIFINFO_IN6:
418 		/* Privileged. */
419 		/* FALLTHROUGH */
420 	case OSIOCGIFINFO_IN6:
421 	case SIOCGIFINFO_IN6:
422 	case SIOCGDRLST_IN6:
423 	case SIOCGPRLST_IN6:
424 	case SIOCGNBRINFO_IN6:
425 	case SIOCGDEFIFACE_IN6:
426 		return nd6_ioctl(cmd, data, ifp);
427 	}
428 
429 	switch (cmd) {
430 	case SIOCSIFPREFIX_IN6:
431 	case SIOCDIFPREFIX_IN6:
432 	case SIOCAIFPREFIX_IN6:
433 	case SIOCCIFPREFIX_IN6:
434 	case SIOCSGIFPREFIX_IN6:
435 	case SIOCGIFPREFIX_IN6:
436 		log(LOG_NOTICE,
437 		    "prefix ioctls are now invalidated. "
438 		    "please use ifconfig.\n");
439 		return EOPNOTSUPP;
440 	}
441 
442 	switch (cmd) {
443 	case SIOCALIFADDR:
444 	case SIOCDLIFADDR:
445 		/* Privileged. */
446 		/* FALLTHROUGH */
447 	case SIOCGLIFADDR:
448 		return in6_lifaddr_ioctl(so, cmd, data, ifp);
449 	}
450 
451 	/*
452 	 * Find address for this interface, if it exists.
453 	 *
454 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
455 	 * only, and used the first interface address as the target of other
456 	 * operations (without checking ifra_addr).  This was because netinet
457 	 * code/API assumed at most 1 interface address per interface.
458 	 * Since IPv6 allows a node to assign multiple addresses
459 	 * on a single interface, we almost always look and check the
460 	 * presence of ifra_addr, and reject invalid ones here.
461 	 * It also decreases duplicated code among SIOC*_IN6 operations.
462 	 */
463 	switch (cmd) {
464 	case SIOCAIFADDR_IN6:
465 #ifdef OSIOCAIFADDR_IN6
466 	case OSIOCAIFADDR_IN6:
467 #endif
468 #ifdef OSIOCSIFPHYADDR_IN6
469 	case OSIOCSIFPHYADDR_IN6:
470 #endif
471 	case SIOCSIFPHYADDR_IN6:
472 		sa6 = &ifra->ifra_addr;
473 		break;
474 	case SIOCSIFADDR_IN6:
475 	case SIOCGIFADDR_IN6:
476 	case SIOCSIFDSTADDR_IN6:
477 	case SIOCSIFNETMASK_IN6:
478 	case SIOCGIFDSTADDR_IN6:
479 	case SIOCGIFNETMASK_IN6:
480 	case SIOCDIFADDR_IN6:
481 	case SIOCGIFPSRCADDR_IN6:
482 	case SIOCGIFPDSTADDR_IN6:
483 	case SIOCGIFAFLAG_IN6:
484 	case SIOCSNDFLUSH_IN6:
485 	case SIOCSPFXFLUSH_IN6:
486 	case SIOCSRTRFLUSH_IN6:
487 	case SIOCGIFALIFETIME_IN6:
488 #ifdef OSIOCGIFALIFETIME_IN6
489 	case OSIOCGIFALIFETIME_IN6:
490 #endif
491 	case SIOCGIFSTAT_IN6:
492 	case SIOCGIFSTAT_ICMP6:
493 		sa6 = &ifr->ifr_addr;
494 		break;
495 	default:
496 		sa6 = NULL;
497 		break;
498 	}
499 
500 	error = 0;
501 	bound = curlwp_bind();
502 	if (sa6 && sa6->sin6_family == AF_INET6) {
503 		if (sa6->sin6_scope_id != 0)
504 			error = sa6_embedscope(sa6, 0);
505 		else
506 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
507 		if (error != 0)
508 			goto out;
509 		ia = in6ifa_ifpwithaddr_psref(ifp, &sa6->sin6_addr, &psref);
510 	} else
511 		ia = NULL;
512 
513 	switch (cmd) {
514 	case SIOCSIFADDR_IN6:
515 	case SIOCSIFDSTADDR_IN6:
516 	case SIOCSIFNETMASK_IN6:
517 		/*
518 		 * Since IPv6 allows a node to assign multiple addresses
519 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
520 		 */
521 		error = EINVAL;
522 		goto release;
523 
524 	case SIOCDIFADDR_IN6:
525 		/*
526 		 * for IPv4, we look for existing in_ifaddr here to allow
527 		 * "ifconfig if0 delete" to remove the first IPv4 address on
528 		 * the interface.  For IPv6, as the spec allows multiple
529 		 * interface address from the day one, we consider "remove the
530 		 * first one" semantics to be not preferable.
531 		 */
532 		if (ia == NULL) {
533 			error = EADDRNOTAVAIL;
534 			goto out;
535 		}
536 		/* FALLTHROUGH */
537 #ifdef OSIOCAIFADDR_IN6
538 	case OSIOCAIFADDR_IN6:
539 #endif
540 	case SIOCAIFADDR_IN6:
541 		/*
542 		 * We always require users to specify a valid IPv6 address for
543 		 * the corresponding operation.
544 		 */
545 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
546 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
547 			error = EAFNOSUPPORT;
548 			goto release;
549 		}
550 		/* Privileged. */
551 
552 		break;
553 
554 	case SIOCGIFADDR_IN6:
555 		/* This interface is basically deprecated. use SIOCGIFCONF. */
556 		/* FALLTHROUGH */
557 	case SIOCGIFAFLAG_IN6:
558 	case SIOCGIFNETMASK_IN6:
559 	case SIOCGIFDSTADDR_IN6:
560 	case SIOCGIFALIFETIME_IN6:
561 #ifdef OSIOCGIFALIFETIME_IN6
562 	case OSIOCGIFALIFETIME_IN6:
563 #endif
564 		/* must think again about its semantics */
565 		if (ia == NULL) {
566 			error = EADDRNOTAVAIL;
567 			goto out;
568 		}
569 		break;
570 	}
571 
572 	switch (cmd) {
573 
574 	case SIOCGIFADDR_IN6:
575 		ifr->ifr_addr = ia->ia_addr;
576 		error = sa6_recoverscope(&ifr->ifr_addr);
577 		break;
578 
579 	case SIOCGIFDSTADDR_IN6:
580 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
581 			error = EINVAL;
582 			break;
583 		}
584 		/*
585 		 * XXX: should we check if ifa_dstaddr is NULL and return
586 		 * an error?
587 		 */
588 		ifr->ifr_dstaddr = ia->ia_dstaddr;
589 		error = sa6_recoverscope(&ifr->ifr_dstaddr);
590 		break;
591 
592 	case SIOCGIFNETMASK_IN6:
593 		ifr->ifr_addr = ia->ia_prefixmask;
594 		break;
595 
596 	case SIOCGIFAFLAG_IN6:
597 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
598 		break;
599 
600 	case SIOCGIFSTAT_IN6:
601 		if (ifp == NULL) {
602 			error = EINVAL;
603 			break;
604 		}
605 		memset(&ifr->ifr_ifru.ifru_stat, 0,
606 		    sizeof(ifr->ifr_ifru.ifru_stat));
607 		ifr->ifr_ifru.ifru_stat =
608 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
609 		break;
610 
611 	case SIOCGIFSTAT_ICMP6:
612 		if (ifp == NULL) {
613 			error = EINVAL;
614 			break;
615 		}
616 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
617 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
618 		ifr->ifr_ifru.ifru_icmp6stat =
619 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
620 		break;
621 
622 #ifdef OSIOCGIFALIFETIME_IN6
623 	case OSIOCGIFALIFETIME_IN6:
624 #endif
625 	case SIOCGIFALIFETIME_IN6:
626 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
627 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
628 			time_t maxexpire;
629 			struct in6_addrlifetime *retlt =
630 			    &ifr->ifr_ifru.ifru_lifetime;
631 
632 			/*
633 			 * XXX: adjust expiration time assuming time_t is
634 			 * signed.
635 			 */
636 			maxexpire = ((time_t)~0) &
637 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
638 			if (ia->ia6_lifetime.ia6t_vltime <
639 			    maxexpire - ia->ia6_updatetime) {
640 				retlt->ia6t_expire = ia->ia6_updatetime +
641 				    ia->ia6_lifetime.ia6t_vltime;
642 				retlt->ia6t_expire = retlt->ia6t_expire ?
643 				    time_mono_to_wall(retlt->ia6t_expire) :
644 				    0;
645 			} else
646 				retlt->ia6t_expire = maxexpire;
647 		}
648 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
649 			time_t maxexpire;
650 			struct in6_addrlifetime *retlt =
651 			    &ifr->ifr_ifru.ifru_lifetime;
652 
653 			/*
654 			 * XXX: adjust expiration time assuming time_t is
655 			 * signed.
656 			 */
657 			maxexpire = ((time_t)~0) &
658 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
659 			if (ia->ia6_lifetime.ia6t_pltime <
660 			    maxexpire - ia->ia6_updatetime) {
661 				retlt->ia6t_preferred = ia->ia6_updatetime +
662 				    ia->ia6_lifetime.ia6t_pltime;
663 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
664 				    time_mono_to_wall(retlt->ia6t_preferred) :
665 				    0;
666 			} else
667 				retlt->ia6t_preferred = maxexpire;
668 		}
669 #ifdef OSIOCFIFALIFETIME_IN6
670 		if (cmd == OSIOCFIFALIFETIME_IN6)
671 			in6_addrlifetime_to_in6_addrlifetime50(
672 			    &ifr->ifru.ifru_lifetime);
673 #endif
674 		break;
675 
676 #ifdef OSIOCAIFADDR_IN6
677 	case OSIOCAIFADDR_IN6:
678 		in6_aliasreq50_to_in6_aliasreq(ifra);
679 		/*FALLTHROUGH*/
680 #endif
681 	case SIOCAIFADDR_IN6:
682 	{
683 		struct in6_addrlifetime *lt;
684 
685 		/* reject read-only flags */
686 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
687 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
688 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
689 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
690 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
691 			error = EINVAL;
692 			break;
693 		}
694 		/*
695 		 * ia6t_expire and ia6t_preferred won't be used for now,
696 		 * so just in case.
697 		 */
698 		lt = &ifra->ifra_lifetime;
699 		if (lt->ia6t_expire != 0)
700 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
701 		if (lt->ia6t_preferred != 0)
702 			lt->ia6t_preferred =
703 			    time_wall_to_mono(lt->ia6t_preferred);
704 		/*
705 		 * make (ia == NULL) or update (ia != NULL) the interface
706 		 * address structure, and link it to the list.
707 		 */
708 		int s = splsoftnet();
709 		error = in6_update_ifa1(ifp, ifra, &ia, &psref, 0);
710 		splx(s);
711 		if (error)
712 			break;
713 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
714 		break;
715 	}
716 
717 	case SIOCDIFADDR_IN6:
718 		ia6_release(ia, &psref);
719 		ifaref(&ia->ia_ifa);
720 		in6_purgeaddr(&ia->ia_ifa);
721 		pfil_run_addrhooks(if_pfil, cmd, &ia->ia_ifa);
722 		ifafree(&ia->ia_ifa);
723 		ia = NULL;
724 		break;
725 
726 	default:
727 		error = ENOTTY;
728 	}
729 release:
730 	ia6_release(ia, &psref);
731 out:
732 	curlwp_bindx(bound);
733 	return error;
734 }
735 
736 int
737 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
738 {
739 	int error, s;
740 
741 	switch (cmd) {
742 	case SIOCSNDFLUSH_IN6:
743 	case SIOCSPFXFLUSH_IN6:
744 	case SIOCSRTRFLUSH_IN6:
745 	case SIOCSDEFIFACE_IN6:
746 	case SIOCSIFINFO_FLAGS:
747 	case SIOCSIFINFO_IN6:
748 
749 	case SIOCALIFADDR:
750 	case SIOCDLIFADDR:
751 
752 	case SIOCDIFADDR_IN6:
753 #ifdef OSIOCAIFADDR_IN6
754 	case OSIOCAIFADDR_IN6:
755 #endif
756 	case SIOCAIFADDR_IN6:
757 
758 	case SIOCAADDRCTL_POLICY:
759 	case SIOCDADDRCTL_POLICY:
760 
761 		if (kauth_authorize_network(curlwp->l_cred,
762 		    KAUTH_NETWORK_SOCKET,
763 		    KAUTH_REQ_NETWORK_SOCKET_SETPRIV,
764 		    so, NULL, NULL))
765 			return EPERM;
766 		break;
767 	}
768 
769 	s = splsoftnet();
770 #ifndef NET_MPSAFE
771 	mutex_enter(softnet_lock);
772 #endif
773 	error = in6_control1(so , cmd, data, ifp);
774 #ifndef NET_MPSAFE
775 	mutex_exit(softnet_lock);
776 #endif
777 	splx(s);
778 	return error;
779 }
780 
781 /*
782  * Update parameters of an IPv6 interface address.
783  * If necessary, a new entry is created and linked into address chains.
784  * This function is separated from in6_control().
785  * XXX: should this be performed under splsoftnet()?
786  */
787 static int
788 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
789     struct in6_ifaddr **iap, struct psref *psref, int flags)
790 {
791 	int error = 0, hostIsNew = 0, plen = -1;
792 	struct sockaddr_in6 dst6;
793 	struct in6_addrlifetime *lt;
794 	struct in6_multi_mship *imm;
795 	struct in6_multi *in6m_sol;
796 	struct rtentry *rt;
797 	int dad_delay, was_tentative;
798 	struct in6_ifaddr *ia = iap ? *iap : NULL;
799 	char ip6buf[INET6_ADDRSTRLEN];
800 
801 	KASSERT((iap == NULL && psref == NULL) ||
802 	    (iap != NULL && psref != NULL));
803 
804 	in6m_sol = NULL;
805 
806 	/* Validate parameters */
807 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
808 		return EINVAL;
809 
810 	/*
811 	 * The destination address for a p2p link must have a family
812 	 * of AF_UNSPEC or AF_INET6.
813 	 */
814 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
815 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
816 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
817 		return EAFNOSUPPORT;
818 	/*
819 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
820 	 * does not carry fields other than sin6_len.
821 	 */
822 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
823 		return EINVAL;
824 	/*
825 	 * Because the IPv6 address architecture is classless, we require
826 	 * users to specify a (non 0) prefix length (mask) for a new address.
827 	 * We also require the prefix (when specified) mask is valid, and thus
828 	 * reject a non-consecutive mask.
829 	 */
830 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
831 		return EINVAL;
832 	if (ifra->ifra_prefixmask.sin6_len != 0) {
833 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
834 		    (u_char *)&ifra->ifra_prefixmask +
835 		    ifra->ifra_prefixmask.sin6_len);
836 		if (plen <= 0)
837 			return EINVAL;
838 	} else {
839 		/*
840 		 * In this case, ia must not be NULL.  We just use its prefix
841 		 * length.
842 		 */
843 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
844 	}
845 	/*
846 	 * If the destination address on a p2p interface is specified,
847 	 * and the address is a scoped one, validate/set the scope
848 	 * zone identifier.
849 	 */
850 	dst6 = ifra->ifra_dstaddr;
851 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
852 	    (dst6.sin6_family == AF_INET6)) {
853 		struct in6_addr in6_tmp;
854 		u_int32_t zoneid;
855 
856 		in6_tmp = dst6.sin6_addr;
857 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
858 			return EINVAL; /* XXX: should be impossible */
859 
860 		if (dst6.sin6_scope_id != 0) {
861 			if (dst6.sin6_scope_id != zoneid)
862 				return EINVAL;
863 		} else		/* user omit to specify the ID. */
864 			dst6.sin6_scope_id = zoneid;
865 
866 		/* convert into the internal form */
867 		if (sa6_embedscope(&dst6, 0))
868 			return EINVAL; /* XXX: should be impossible */
869 	}
870 	/*
871 	 * The destination address can be specified only for a p2p or a
872 	 * loopback interface.  If specified, the corresponding prefix length
873 	 * must be 128.
874 	 */
875 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
876 #ifdef FORCE_P2PPLEN
877 		int i;
878 #endif
879 
880 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
881 			/* XXX: noisy message */
882 			nd6log(LOG_INFO, "a destination can "
883 			    "be specified for a p2p or a loopback IF only\n");
884 			return EINVAL;
885 		}
886 		if (plen != 128) {
887 			nd6log(LOG_INFO, "prefixlen should "
888 			    "be 128 when dstaddr is specified\n");
889 #ifdef FORCE_P2PPLEN
890 			/*
891 			 * To be compatible with old configurations,
892 			 * such as ifconfig gif0 inet6 2001::1 2001::2
893 			 * prefixlen 126, we override the specified
894 			 * prefixmask as if the prefix length was 128.
895 			 */
896 			ifra->ifra_prefixmask.sin6_len =
897 			    sizeof(struct sockaddr_in6);
898 			for (i = 0; i < 4; i++)
899 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
900 				    0xffffffff;
901 			plen = 128;
902 #else
903 			return EINVAL;
904 #endif
905 		}
906 	}
907 	/* lifetime consistency check */
908 	lt = &ifra->ifra_lifetime;
909 	if (lt->ia6t_pltime > lt->ia6t_vltime)
910 		return EINVAL;
911 	if (lt->ia6t_vltime == 0) {
912 		/*
913 		 * the following log might be noisy, but this is a typical
914 		 * configuration mistake or a tool's bug.
915 		 */
916 		nd6log(LOG_INFO, "valid lifetime is 0 for %s\n",
917 		    IN6_PRINT(ip6buf, &ifra->ifra_addr.sin6_addr));
918 
919 		if (ia == NULL)
920 			return 0; /* there's nothing to do */
921 	}
922 
923 	/*
924 	 * If this is a new address, allocate a new ifaddr and link it
925 	 * into chains.
926 	 */
927 	if (ia == NULL) {
928 		hostIsNew = 1;
929 		/*
930 		 * When in6_update_ifa() is called in a process of a received
931 		 * RA, it is called under an interrupt context.  So, we should
932 		 * call malloc with M_NOWAIT.
933 		 */
934 		ia = malloc(sizeof(*ia), M_IFADDR, M_NOWAIT|M_ZERO);
935 		if (ia == NULL)
936 			return ENOBUFS;
937 		LIST_INIT(&ia->ia6_memberships);
938 		/* Initialize the address and masks, and put time stamp */
939 		ia->ia_ifa.ifa_addr = sin6tosa(&ia->ia_addr);
940 		ia->ia_addr.sin6_family = AF_INET6;
941 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
942 		ia->ia6_createtime = time_uptime;
943 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
944 			/*
945 			 * XXX: some functions expect that ifa_dstaddr is not
946 			 * NULL for p2p interfaces.
947 			 */
948 			ia->ia_ifa.ifa_dstaddr = sin6tosa(&ia->ia_dstaddr);
949 		} else {
950 			ia->ia_ifa.ifa_dstaddr = NULL;
951 		}
952 		ia->ia_ifa.ifa_netmask = sin6tosa(&ia->ia_prefixmask);
953 
954 		ia->ia_ifp = ifp;
955 		IN6_ADDRLIST_ENTRY_INIT(ia);
956 		ifa_psref_init(&ia->ia_ifa);
957 	}
958 
959 	/* update timestamp */
960 	ia->ia6_updatetime = time_uptime;
961 
962 	/* set prefix mask */
963 	if (ifra->ifra_prefixmask.sin6_len) {
964 		if (ia->ia_prefixmask.sin6_len) {
965 			/*
966 			 * We prohibit changing the prefix length of an
967 			 * existing autoconf address, because the operation
968 			 * would confuse prefix management.
969 			 */
970 			if (ia->ia6_ndpr != NULL &&
971 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) !=
972 			    plen)
973 			{
974 				nd6log(LOG_INFO, "the prefix length of an"
975 				    " existing (%s) autoconf address should"
976 				    " not be changed\n",
977 				    IN6_PRINT(ip6buf,
978 				    &ia->ia_addr.sin6_addr));
979 				error = EINVAL;
980 				if (hostIsNew)
981 					free(ia, M_IFADDR);
982 				goto exit;
983 			}
984 
985 			if (!IN6_ARE_ADDR_EQUAL(&ia->ia_prefixmask.sin6_addr,
986 			    &ifra->ifra_prefixmask.sin6_addr))
987 				in6_ifremprefix(ia);
988 		}
989 		ia->ia_prefixmask = ifra->ifra_prefixmask;
990 	}
991 
992 	/* Set destination address. */
993 	if (dst6.sin6_family == AF_INET6) {
994 		if (!IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
995 		    &ia->ia_dstaddr.sin6_addr))
996 			in6_ifremprefix(ia);
997 		ia->ia_dstaddr = dst6;
998 	}
999 
1000 	/*
1001 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1002 	 * to see if the address is deprecated or invalidated, but initialize
1003 	 * these members for applications.
1004 	 */
1005 	ia->ia6_lifetime = ifra->ifra_lifetime;
1006 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1007 		ia->ia6_lifetime.ia6t_expire =
1008 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1009 	} else
1010 		ia->ia6_lifetime.ia6t_expire = 0;
1011 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1012 		ia->ia6_lifetime.ia6t_preferred =
1013 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1014 	} else
1015 		ia->ia6_lifetime.ia6t_preferred = 0;
1016 
1017 	/*
1018 	 * configure address flags.
1019 	 * We need to preserve tentative state so DAD works if
1020 	 * something adds the same address before DAD finishes.
1021 	 */
1022 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1023 	ia->ia6_flags = ifra->ifra_flags;
1024 
1025 	/*
1026 	 * Make the address tentative before joining multicast addresses,
1027 	 * so that corresponding MLD responses would not have a tentative
1028 	 * source address.
1029 	 */
1030 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1031 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1032 		ia->ia6_flags |= IN6_IFF_DETACHED;
1033 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1034 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp))
1035 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1036 
1037 	/*
1038 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1039 	 * userland, make it deprecated.
1040 	 */
1041 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1042 		ia->ia6_lifetime.ia6t_pltime = 0;
1043 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1044 	}
1045 
1046 	/* reset the interface and routing table appropriately. */
1047 	error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew);
1048 	if (error != 0) {
1049 		if (hostIsNew)
1050 			free(ia, M_IFADDR);
1051 		goto exit;
1052 	}
1053 
1054 	/*
1055 	 * We are done if we have simply modified an existing address.
1056 	 */
1057 	if (!hostIsNew)
1058 		return error;
1059 
1060 	/*
1061 	 * Insert ia to the global list and ifa to the interface's list.
1062 	 */
1063 	mutex_enter(&in6_ifaddr_lock);
1064 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1065 	mutex_exit(&in6_ifaddr_lock);
1066 
1067 	/* gain a refcnt for the link from in6_ifaddr */
1068 	ifaref(&ia->ia_ifa);
1069 	ifa_insert(ifp, &ia->ia_ifa);
1070 
1071 	/*
1072 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1073 	 * not just go to unlink.
1074 	 */
1075 
1076 	/* join necessary multicast groups */
1077 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1078 		struct sockaddr_in6 mltaddr, mltmask;
1079 		struct in6_addr llsol;
1080 
1081 		/* join solicited multicast addr for new host id */
1082 		memset(&llsol, 0, sizeof(struct in6_addr));
1083 		llsol.s6_addr16[0] = htons(0xff02);
1084 		llsol.s6_addr32[1] = 0;
1085 		llsol.s6_addr32[2] = htonl(1);
1086 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1087 		llsol.s6_addr8[12] = 0xff;
1088 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1089 			/* XXX: should not happen */
1090 			log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
1091 			goto cleanup;
1092 		}
1093 		dad_delay = 0;
1094 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1095 			/*
1096 			 * We need a random delay for DAD on the address
1097 			 * being configured.  It also means delaying
1098 			 * transmission of the corresponding MLD report to
1099 			 * avoid report collision.
1100 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1101 			 */
1102 			dad_delay = cprng_fast32() %
1103 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1104 		}
1105 
1106 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1107 		/* join solicited multicast addr for new host id */
1108 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
1109 		if (!imm) {
1110 			nd6log(LOG_ERR,
1111 			    "addmulti failed for %s on %s (errno=%d)\n",
1112 			    IN6_PRINT(ip6buf, &llsol), if_name(ifp), error);
1113 			goto cleanup;
1114 		}
1115 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1116 		in6m_sol = imm->i6mm_maddr;
1117 
1118 		sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
1119 
1120 		/*
1121 		 * join link-local all-nodes address
1122 		 */
1123 		sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
1124 		    0, 0, 0);
1125 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1126 			goto cleanup; /* XXX: should not fail */
1127 
1128 		/*
1129 		 * XXX: do we really need this automatic routes?
1130 		 * We should probably reconsider this stuff.  Most applications
1131 		 * actually do not need the routes, since they usually specify
1132 		 * the outgoing interface.
1133 		 */
1134 		rt = rtalloc1(sin6tosa(&mltaddr), 0);
1135 		if (rt) {
1136 			if (memcmp(&mltaddr.sin6_addr,
1137 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1138 			    MLTMASK_LEN)) {
1139 				rt_unref(rt);
1140 				rt = NULL;
1141 			} else if (rt->rt_ifp != ifp) {
1142 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1143 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1144 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1145 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1146 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1147 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1148 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1149 				rt_replace_ifa(rt, &ia->ia_ifa);
1150 				rt->rt_ifp = ifp;
1151 			}
1152 		}
1153 		if (!rt) {
1154 			struct rt_addrinfo info;
1155 
1156 			memset(&info, 0, sizeof(info));
1157 			info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
1158 			info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
1159 			info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
1160 			info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
1161 			/* XXX: we need RTF_CONNECTED to fake nd6_rtrequest */
1162 			info.rti_flags = RTF_UP | RTF_CONNECTED;
1163 			error = rtrequest1(RTM_ADD, &info, NULL);
1164 			if (error)
1165 				goto cleanup;
1166 		} else {
1167 			rt_unref(rt);
1168 		}
1169 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1170 		if (!imm) {
1171 			nd6log(LOG_WARNING,
1172 			    "addmulti failed for %s on %s (errno=%d)\n",
1173 			    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
1174 			    if_name(ifp), error);
1175 			goto cleanup;
1176 		}
1177 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1178 
1179 		/*
1180 		 * join node information group address
1181 		 */
1182 		dad_delay = 0;
1183 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1184 			/*
1185 			 * The spec doesn't say anything about delay for this
1186 			 * group, but the same logic should apply.
1187 			 */
1188 			dad_delay = cprng_fast32() %
1189 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1190 		}
1191 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
1192 			;
1193 		else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1194 		          dad_delay)) == NULL) { /* XXX jinmei */
1195 			nd6log(LOG_WARNING,
1196 			    "addmulti failed for %s on %s (errno=%d)\n",
1197 			    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
1198 			    if_name(ifp), error);
1199 			/* XXX not very fatal, go on... */
1200 		} else {
1201 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1202 		}
1203 
1204 
1205 		/*
1206 		 * join interface-local all-nodes address.
1207 		 * (ff01::1%ifN, and ff01::%ifN/32)
1208 		 */
1209 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1210 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1211 			goto cleanup; /* XXX: should not fail */
1212 
1213 		/* XXX: again, do we really need the route? */
1214 		rt = rtalloc1(sin6tosa(&mltaddr), 0);
1215 		if (rt) {
1216 			/* 32bit came from "mltmask" */
1217 			if (memcmp(&mltaddr.sin6_addr,
1218 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1219 			    32 / NBBY)) {
1220 				rt_unref(rt);
1221 				rt = NULL;
1222 			} else if (rt->rt_ifp != ifp) {
1223 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1224 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1225 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1226 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1227 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1228 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1229 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1230 				rt_replace_ifa(rt, &ia->ia_ifa);
1231 				rt->rt_ifp = ifp;
1232 			}
1233 		}
1234 		if (!rt) {
1235 			struct rt_addrinfo info;
1236 
1237 			memset(&info, 0, sizeof(info));
1238 			info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
1239 			info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
1240 			info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
1241 			info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
1242 			info.rti_flags = RTF_UP | RTF_CONNECTED;
1243 			error = rtrequest1(RTM_ADD, &info, NULL);
1244 			if (error)
1245 				goto cleanup;
1246 #undef	MLTMASK_LEN
1247 		} else {
1248 			rt_unref(rt);
1249 		}
1250 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1251 		if (!imm) {
1252 			nd6log(LOG_WARNING,
1253 			    "addmulti failed for %s on %s (errno=%d)\n",
1254 			    IN6_PRINT(ip6buf, &mltaddr.sin6_addr),
1255 			    if_name(ifp), error);
1256 			goto cleanup;
1257 		} else {
1258 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1259 		}
1260 	}
1261 
1262 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1263 	error = nd6_add_ifa_lle(ia);
1264 	if (error != 0)
1265 		goto cleanup;
1266 
1267 	/*
1268 	 * Perform DAD, if needed.
1269 	 * XXX It may be of use, if we can administratively
1270 	 * disable DAD.
1271 	 */
1272 	if (hostIsNew && if_do_dad(ifp) &&
1273 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1274 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1275 	{
1276 		int mindelay, maxdelay;
1277 
1278 		dad_delay = 0;
1279 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1280 			/*
1281 			 * We need to impose a delay before sending an NS
1282 			 * for DAD.  Check if we also needed a delay for the
1283 			 * corresponding MLD message.  If we did, the delay
1284 			 * should be larger than the MLD delay (this could be
1285 			 * relaxed a bit, but this simple logic is at least
1286 			 * safe).
1287 			 */
1288 			mindelay = 0;
1289 			if (in6m_sol != NULL &&
1290 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1291 				mindelay = in6m_sol->in6m_timer;
1292 			}
1293 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1294 			if (maxdelay - mindelay == 0)
1295 				dad_delay = 0;
1296 			else {
1297 				dad_delay =
1298 				    (cprng_fast32() % (maxdelay - mindelay)) +
1299 				    mindelay;
1300 			}
1301 		}
1302 		/* +1 ensures callout is always used */
1303 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1304 	}
1305 
1306 	if (iap != NULL) {
1307 		*iap = ia;
1308 		if (hostIsNew)
1309 			ia6_acquire(ia, psref);
1310 	}
1311 
1312 	return 0;
1313 
1314   cleanup:
1315 	in6_purgeaddr(&ia->ia_ifa);
1316   exit:
1317 	return error;
1318 }
1319 
1320 int
1321 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1322 {
1323 	int rc, s;
1324 
1325 	s = splsoftnet();
1326 	rc = in6_update_ifa1(ifp, ifra, NULL, NULL, flags);
1327 	splx(s);
1328 	return rc;
1329 }
1330 
1331 void
1332 in6_purgeaddr(struct ifaddr *ifa)
1333 {
1334 	struct ifnet *ifp = ifa->ifa_ifp;
1335 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1336 	struct in6_multi_mship *imm;
1337 
1338 	KASSERT(!ifa_held(ifa));
1339 
1340 	ifa->ifa_flags |= IFA_DESTROYING;
1341 
1342 	/* stop DAD processing */
1343 	nd6_dad_stop(ifa);
1344 
1345 	/* Delete any network route. */
1346 	in6_ifremprefix(ia);
1347 
1348 	/* Remove ownaddr's loopback rtentry, if it exists. */
1349 	in6_ifremlocal(&(ia->ia_ifa));
1350 
1351 	/*
1352 	 * leave from multicast groups we have joined for the interface
1353 	 */
1354 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1355 		LIST_REMOVE(imm, i6mm_chain);
1356 		in6_leavegroup(imm);
1357 	}
1358 
1359 	in6_unlink_ifa(ia, ifp);
1360 }
1361 
1362 static void
1363 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1364 {
1365 	int	s = splsoftnet();
1366 
1367 	mutex_enter(&in6_ifaddr_lock);
1368 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1369 	ifa_remove(ifp, &ia->ia_ifa);
1370 	mutex_exit(&in6_ifaddr_lock);
1371 
1372 	/*
1373 	 * XXX thorpej@NetBSD.org -- if the interface is going
1374 	 * XXX away, don't save the multicast entries, delete them!
1375 	 */
1376 	if (LIST_EMPTY(&ia->ia6_multiaddrs))
1377 		;
1378 	else if (if_is_deactivated(ia->ia_ifa.ifa_ifp)) {
1379 		struct in6_multi *in6m, *next;
1380 
1381 		for (in6m = LIST_FIRST(&ia->ia6_multiaddrs); in6m != NULL;
1382 		     in6m = next) {
1383 			next = LIST_NEXT(in6m, in6m_entry);
1384 			in6_delmulti(in6m);
1385 		}
1386 	} else
1387 		in6_savemkludge(ia);
1388 
1389 	/*
1390 	 * Release the reference to the ND prefix.
1391 	 */
1392 	if (ia->ia6_ndpr != NULL) {
1393 		nd6_prefix_unref(ia->ia6_ndpr);
1394 		ia->ia6_ndpr = NULL;
1395 	}
1396 
1397 	/*
1398 	 * Also, if the address being removed is autoconf'ed, call
1399 	 * nd6_pfxlist_onlink_check() since the release might affect the status of
1400 	 * other (detached) addresses.
1401 	 */
1402 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0) {
1403 		ND6_WLOCK();
1404 		nd6_pfxlist_onlink_check();
1405 		ND6_UNLOCK();
1406 	}
1407 
1408 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1409 
1410 	/*
1411 	 * release another refcnt for the link from in6_ifaddr.
1412 	 * Note that we should decrement the refcnt at least once for all *BSD.
1413 	 */
1414 	ifafree(&ia->ia_ifa);
1415 
1416 	splx(s);
1417 }
1418 
1419 void
1420 in6_purgeif(struct ifnet *ifp)
1421 {
1422 
1423 	in6_ifdetach(ifp);
1424 }
1425 
1426 /*
1427  * SIOC[GAD]LIFADDR.
1428  *	SIOCGLIFADDR: get first address. (?)
1429  *	SIOCGLIFADDR with IFLR_PREFIX:
1430  *		get first address that matches the specified prefix.
1431  *	SIOCALIFADDR: add the specified address.
1432  *	SIOCALIFADDR with IFLR_PREFIX:
1433  *		add the specified prefix, filling hostid part from
1434  *		the first link-local address.  prefixlen must be <= 64.
1435  *	SIOCDLIFADDR: delete the specified address.
1436  *	SIOCDLIFADDR with IFLR_PREFIX:
1437  *		delete the first address that matches the specified prefix.
1438  * return values:
1439  *	EINVAL on invalid parameters
1440  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1441  *	other values may be returned from in6_ioctl()
1442  *
1443  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1444  * this is to accommodate address naming scheme other than RFC2374,
1445  * in the future.
1446  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1447  * address encoding scheme. (see figure on page 8)
1448  */
1449 static int
1450 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1451 	struct ifnet *ifp)
1452 {
1453 	struct in6_ifaddr *ia = NULL; /* XXX gcc 4.8 maybe-uninitialized */
1454 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1455 	struct ifaddr *ifa;
1456 	struct sockaddr *sa;
1457 
1458 	/* sanity checks */
1459 	if (!data || !ifp) {
1460 		panic("invalid argument to in6_lifaddr_ioctl");
1461 		/* NOTREACHED */
1462 	}
1463 
1464 	switch (cmd) {
1465 	case SIOCGLIFADDR:
1466 		/* address must be specified on GET with IFLR_PREFIX */
1467 		if ((iflr->flags & IFLR_PREFIX) == 0)
1468 			break;
1469 		/* FALLTHROUGH */
1470 	case SIOCALIFADDR:
1471 	case SIOCDLIFADDR:
1472 		/* address must be specified on ADD and DELETE */
1473 		sa = (struct sockaddr *)&iflr->addr;
1474 		if (sa->sa_family != AF_INET6)
1475 			return EINVAL;
1476 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1477 			return EINVAL;
1478 		/* XXX need improvement */
1479 		sa = (struct sockaddr *)&iflr->dstaddr;
1480 		if (sa->sa_family && sa->sa_family != AF_INET6)
1481 			return EINVAL;
1482 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1483 			return EINVAL;
1484 		break;
1485 	default: /* shouldn't happen */
1486 #if 0
1487 		panic("invalid cmd to in6_lifaddr_ioctl");
1488 		/* NOTREACHED */
1489 #else
1490 		return EOPNOTSUPP;
1491 #endif
1492 	}
1493 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1494 		return EINVAL;
1495 
1496 	switch (cmd) {
1497 	case SIOCALIFADDR:
1498 	    {
1499 		struct in6_aliasreq ifra;
1500 		struct in6_addr *xhostid = NULL;
1501 		int prefixlen;
1502 		int bound = curlwp_bind();
1503 		struct psref psref;
1504 
1505 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1506 			struct sockaddr_in6 *sin6;
1507 
1508 			/*
1509 			 * xhostid is to fill in the hostid part of the
1510 			 * address.  xhostid points to the first link-local
1511 			 * address attached to the interface.
1512 			 */
1513 			ia = in6ifa_ifpforlinklocal_psref(ifp, 0, &psref);
1514 			if (ia == NULL) {
1515 				curlwp_bindx(bound);
1516 				return EADDRNOTAVAIL;
1517 			}
1518 			xhostid = IFA_IN6(&ia->ia_ifa);
1519 
1520 		 	/* prefixlen must be <= 64. */
1521 			if (64 < iflr->prefixlen) {
1522 				ia6_release(ia, &psref);
1523 				curlwp_bindx(bound);
1524 				return EINVAL;
1525 			}
1526 			prefixlen = iflr->prefixlen;
1527 
1528 			/* hostid part must be zero. */
1529 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1530 			if (sin6->sin6_addr.s6_addr32[2] != 0
1531 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1532 				ia6_release(ia, &psref);
1533 				curlwp_bindx(bound);
1534 				return EINVAL;
1535 			}
1536 		} else
1537 			prefixlen = iflr->prefixlen;
1538 
1539 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1540 		memset(&ifra, 0, sizeof(ifra));
1541 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1542 
1543 		memcpy(&ifra.ifra_addr, &iflr->addr,
1544 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1545 		if (xhostid) {
1546 			/* fill in hostid part */
1547 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1548 			    xhostid->s6_addr32[2];
1549 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1550 			    xhostid->s6_addr32[3];
1551 		}
1552 
1553 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1554 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1555 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1556 			if (xhostid) {
1557 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1558 				    xhostid->s6_addr32[2];
1559 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1560 				    xhostid->s6_addr32[3];
1561 			}
1562 		}
1563 		if (xhostid) {
1564 			ia6_release(ia, &psref);
1565 			ia = NULL;
1566 		}
1567 		curlwp_bindx(bound);
1568 
1569 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1570 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1571 
1572 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1573 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1574 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1575 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1576 	    }
1577 	case SIOCGLIFADDR:
1578 	case SIOCDLIFADDR:
1579 	    {
1580 		struct in6_addr mask, candidate, match;
1581 		struct sockaddr_in6 *sin6;
1582 		int cmp;
1583 		int error, s;
1584 
1585 		memset(&mask, 0, sizeof(mask));
1586 		if (iflr->flags & IFLR_PREFIX) {
1587 			/* lookup a prefix rather than address. */
1588 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1589 
1590 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1591 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1592 			match.s6_addr32[0] &= mask.s6_addr32[0];
1593 			match.s6_addr32[1] &= mask.s6_addr32[1];
1594 			match.s6_addr32[2] &= mask.s6_addr32[2];
1595 			match.s6_addr32[3] &= mask.s6_addr32[3];
1596 
1597 			/* if you set extra bits, that's wrong */
1598 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1599 				return EINVAL;
1600 
1601 			cmp = 1;
1602 		} else {
1603 			if (cmd == SIOCGLIFADDR) {
1604 				/* on getting an address, take the 1st match */
1605 				cmp = 0;	/* XXX */
1606 			} else {
1607 				/* on deleting an address, do exact match */
1608 				in6_prefixlen2mask(&mask, 128);
1609 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1610 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1611 
1612 				cmp = 1;
1613 			}
1614 		}
1615 
1616 		s = pserialize_read_enter();
1617 		IFADDR_READER_FOREACH(ifa, ifp) {
1618 			if (ifa->ifa_addr->sa_family != AF_INET6)
1619 				continue;
1620 			if (!cmp)
1621 				break;
1622 
1623 			/*
1624 			 * XXX: this is adhoc, but is necessary to allow
1625 			 * a user to specify fe80::/64 (not /10) for a
1626 			 * link-local address.
1627 			 */
1628 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1629 			in6_clearscope(&candidate);
1630 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1631 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1632 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1633 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1634 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1635 				break;
1636 		}
1637 		if (!ifa) {
1638 			error = EADDRNOTAVAIL;
1639 			goto error;
1640 		}
1641 		ia = ifa2ia6(ifa);
1642 
1643 		if (cmd == SIOCGLIFADDR) {
1644 			/* fill in the if_laddrreq structure */
1645 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1646 			error = sa6_recoverscope(
1647 			    (struct sockaddr_in6 *)&iflr->addr);
1648 			if (error != 0)
1649 				goto error;
1650 
1651 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1652 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1653 				    ia->ia_dstaddr.sin6_len);
1654 				error = sa6_recoverscope(
1655 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1656 				if (error != 0)
1657 					goto error;
1658 			} else
1659 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1660 
1661 			iflr->prefixlen =
1662 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1663 
1664 			iflr->flags = ia->ia6_flags;	/* XXX */
1665 
1666 			error = 0;
1667 		} else {
1668 			struct in6_aliasreq ifra;
1669 
1670 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1671 			memset(&ifra, 0, sizeof(ifra));
1672 			memcpy(ifra.ifra_name, iflr->iflr_name,
1673 			    sizeof(ifra.ifra_name));
1674 
1675 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1676 			    ia->ia_addr.sin6_len);
1677 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1678 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1679 				    ia->ia_dstaddr.sin6_len);
1680 			} else {
1681 				memset(&ifra.ifra_dstaddr, 0,
1682 				    sizeof(ifra.ifra_dstaddr));
1683 			}
1684 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1685 			    ia->ia_prefixmask.sin6_len);
1686 
1687 			ifra.ifra_flags = ia->ia6_flags;
1688 			pserialize_read_exit(s);
1689 
1690 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1691 		}
1692 	error:
1693 		pserialize_read_exit(s);
1694 		return error;
1695 	    }
1696 	}
1697 
1698 	return EOPNOTSUPP;	/* just for safety */
1699 }
1700 
1701 /*
1702  * Initialize an interface's internet6 address
1703  * and routing table entry.
1704  */
1705 static int
1706 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1707 	const struct sockaddr_in6 *sin6, int newhost)
1708 {
1709 	int	error = 0, ifacount = 0;
1710 	int	s = splsoftnet();
1711 	struct ifaddr *ifa;
1712 
1713 	/*
1714 	 * Give the interface a chance to initialize
1715 	 * if this is its first address,
1716 	 * and to validate the address if necessary.
1717 	 */
1718 	IFADDR_READER_FOREACH(ifa, ifp) {
1719 		if (ifa->ifa_addr->sa_family != AF_INET6)
1720 			continue;
1721 		ifacount++;
1722 	}
1723 
1724 	ia->ia_addr = *sin6;
1725 
1726 	if (ifacount <= 0 &&
1727 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1728 		splx(s);
1729 		return error;
1730 	}
1731 	splx(s);
1732 
1733 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1734 
1735 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1736 
1737 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1738 	if (newhost) {
1739 		/* set the rtrequest function to create llinfo */
1740 		if (ifp->if_flags & IFF_POINTOPOINT)
1741 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1742 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1743 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1744 		in6_ifaddlocal(&ia->ia_ifa);
1745 	} else {
1746 		/* Inform the routing socket of new flags/timings */
1747 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
1748 	}
1749 
1750 	/* Add the network prefix route. */
1751 	if ((error = in6_ifaddprefix(ia)) != 0) {
1752 		if (newhost)
1753 			in6_ifremlocal(&ia->ia_ifa);
1754 		return error;
1755 	}
1756 
1757 	if (ifp->if_flags & IFF_MULTICAST)
1758 		in6_restoremkludge(ia, ifp);
1759 
1760 	return error;
1761 }
1762 
1763 static struct ifaddr *
1764 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1765 {
1766 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1767 		return ifa;
1768 	return best_ifa;
1769 }
1770 
1771 /*
1772  * Find an IPv6 interface link-local address specific to an interface.
1773  */
1774 struct in6_ifaddr *
1775 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1776 {
1777 	struct ifaddr *best_ifa = NULL, *ifa;
1778 
1779 	IFADDR_READER_FOREACH(ifa, ifp) {
1780 		if (ifa->ifa_addr->sa_family != AF_INET6)
1781 			continue;
1782 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1783 			continue;
1784 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1785 			continue;
1786 		best_ifa = bestifa(best_ifa, ifa);
1787 	}
1788 
1789 	return (struct in6_ifaddr *)best_ifa;
1790 }
1791 
1792 struct in6_ifaddr *
1793 in6ifa_ifpforlinklocal_psref(const struct ifnet *ifp, const int ignoreflags,
1794     struct psref *psref)
1795 {
1796 	struct in6_ifaddr *ia;
1797 	int s = pserialize_read_enter();
1798 
1799 	ia = in6ifa_ifpforlinklocal(ifp, ignoreflags);
1800 	if (ia != NULL)
1801 		ia6_acquire(ia, psref);
1802 	pserialize_read_exit(s);
1803 
1804 	return ia;
1805 }
1806 
1807 /*
1808  * find the internet address corresponding to a given address.
1809  * ifaddr is returned referenced.
1810  */
1811 struct in6_ifaddr *
1812 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1813 {
1814 	struct in6_ifaddr *ia;
1815 	int s;
1816 
1817 	s = pserialize_read_enter();
1818 	IN6_ADDRLIST_READER_FOREACH(ia) {
1819 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1820 			if (zoneid != 0 &&
1821 			    zoneid != ia->ia_addr.sin6_scope_id)
1822 				continue;
1823 			ifaref(&ia->ia_ifa);
1824 			break;
1825 		}
1826 	}
1827 	pserialize_read_exit(s);
1828 
1829 	return ia;
1830 }
1831 
1832 /*
1833  * find the internet address corresponding to a given interface and address.
1834  */
1835 struct in6_ifaddr *
1836 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1837 {
1838 	struct ifaddr *best_ifa = NULL, *ifa;
1839 
1840 	IFADDR_READER_FOREACH(ifa, ifp) {
1841 		if (ifa->ifa_addr->sa_family != AF_INET6)
1842 			continue;
1843 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1844 			continue;
1845 		best_ifa = bestifa(best_ifa, ifa);
1846 	}
1847 
1848 	return (struct in6_ifaddr *)best_ifa;
1849 }
1850 
1851 struct in6_ifaddr *
1852 in6ifa_ifpwithaddr_psref(const struct ifnet *ifp, const struct in6_addr *addr,
1853     struct psref *psref)
1854 {
1855 	struct in6_ifaddr *ia;
1856 	int s = pserialize_read_enter();
1857 
1858 	ia = in6ifa_ifpwithaddr(ifp, addr);
1859 	if (ia != NULL)
1860 		ia6_acquire(ia, psref);
1861 	pserialize_read_exit(s);
1862 
1863 	return ia;
1864 }
1865 
1866 static struct in6_ifaddr *
1867 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1868 {
1869 	if (best_ia == NULL ||
1870 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1871 		return ia;
1872 	return best_ia;
1873 }
1874 
1875 /*
1876  * Determine if an address is on a local network.
1877  */
1878 int
1879 in6_localaddr(const struct in6_addr *in6)
1880 {
1881 	struct in6_ifaddr *ia;
1882 	int s;
1883 
1884 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1885 		return 1;
1886 
1887 	s = pserialize_read_enter();
1888 	IN6_ADDRLIST_READER_FOREACH(ia) {
1889 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1890 					      &ia->ia_prefixmask.sin6_addr)) {
1891 			pserialize_read_exit(s);
1892 			return 1;
1893 		}
1894 	}
1895 	pserialize_read_exit(s);
1896 
1897 	return 0;
1898 }
1899 
1900 int
1901 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1902 {
1903 	struct in6_ifaddr *ia;
1904 	int s;
1905 
1906 	s = pserialize_read_enter();
1907 	IN6_ADDRLIST_READER_FOREACH(ia) {
1908 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1909 		    &sa6->sin6_addr) &&
1910 #ifdef SCOPEDROUTING
1911 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1912 #endif
1913 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
1914 			pserialize_read_exit(s);
1915 			return 1; /* true */
1916 		}
1917 
1918 		/* XXX: do we still have to go thru the rest of the list? */
1919 	}
1920 	pserialize_read_exit(s);
1921 
1922 	return 0;		/* false */
1923 }
1924 
1925 /*
1926  * return length of part which dst and src are equal
1927  * hard coding...
1928  */
1929 int
1930 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1931 {
1932 	int match = 0;
1933 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1934 	u_char *lim = s + 16, r;
1935 
1936 	while (s < lim)
1937 		if ((r = (*d++ ^ *s++)) != 0) {
1938 			while (r < 128) {
1939 				match++;
1940 				r <<= 1;
1941 			}
1942 			break;
1943 		} else
1944 			match += NBBY;
1945 	return match;
1946 }
1947 
1948 /* XXX: to be scope conscious */
1949 int
1950 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1951 {
1952 	int bytelen, bitlen;
1953 
1954 	/* sanity check */
1955 	if (len < 0 || len > 128) {
1956 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1957 		    len);
1958 		return 0;
1959 	}
1960 
1961 	bytelen = len / NBBY;
1962 	bitlen = len % NBBY;
1963 
1964 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1965 		return 0;
1966 	if (bitlen != 0 &&
1967 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
1968 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
1969 		return 0;
1970 
1971 	return 1;
1972 }
1973 
1974 void
1975 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1976 {
1977 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1978 	int bytelen, bitlen, i;
1979 
1980 	/* sanity check */
1981 	if (len < 0 || len > 128) {
1982 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1983 		    len);
1984 		return;
1985 	}
1986 
1987 	memset(maskp, 0, sizeof(*maskp));
1988 	bytelen = len / NBBY;
1989 	bitlen = len % NBBY;
1990 	for (i = 0; i < bytelen; i++)
1991 		maskp->s6_addr[i] = 0xff;
1992 	if (bitlen)
1993 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1994 }
1995 
1996 /*
1997  * return the best address out of the same scope. if no address was
1998  * found, return the first valid address from designated IF.
1999  */
2000 struct in6_ifaddr *
2001 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2002 {
2003 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2004 	struct ifaddr *ifa;
2005 	struct in6_ifaddr *best_ia = NULL, *ia;
2006 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2007 
2008 	dep[0] = dep[1] = NULL;
2009 
2010 	/*
2011 	 * We first look for addresses in the same scope.
2012 	 * If there is one, return it.
2013 	 * If two or more, return one which matches the dst longest.
2014 	 * If none, return one of global addresses assigned other ifs.
2015 	 */
2016 	IFADDR_READER_FOREACH(ifa, ifp) {
2017 		if (ifa->ifa_addr->sa_family != AF_INET6)
2018 			continue;
2019 		ia = (struct in6_ifaddr *)ifa;
2020 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2021 			continue; /* XXX: is there any case to allow anycast? */
2022 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2023 			continue; /* don't use this interface */
2024 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2025 			continue;
2026 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2027 			if (ip6_use_deprecated)
2028 				dep[0] = ia;
2029 			continue;
2030 		}
2031 
2032 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
2033 			continue;
2034 		/*
2035 		 * call in6_matchlen() as few as possible
2036 		 */
2037 		if (best_ia == NULL) {
2038 			best_ia = ia;
2039 			continue;
2040 		}
2041 		if (blen == -1)
2042 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2043 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2044 		if (tlen > blen) {
2045 			blen = tlen;
2046 			best_ia = ia;
2047 		} else if (tlen == blen)
2048 			best_ia = bestia(best_ia, ia);
2049 	}
2050 	if (best_ia != NULL)
2051 		return best_ia;
2052 
2053 	IFADDR_READER_FOREACH(ifa, ifp) {
2054 		if (ifa->ifa_addr->sa_family != AF_INET6)
2055 			continue;
2056 		ia = (struct in6_ifaddr *)ifa;
2057 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2058 			continue; /* XXX: is there any case to allow anycast? */
2059 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2060 			continue; /* don't use this interface */
2061 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2062 			continue;
2063 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2064 			if (ip6_use_deprecated)
2065 				dep[1] = (struct in6_ifaddr *)ifa;
2066 			continue;
2067 		}
2068 
2069 		best_ia = bestia(best_ia, ia);
2070 	}
2071 	if (best_ia != NULL)
2072 		return best_ia;
2073 
2074 	/* use the last-resort values, that are, deprecated addresses */
2075 	if (dep[0])
2076 		return dep[0];
2077 	if (dep[1])
2078 		return dep[1];
2079 
2080 	return NULL;
2081 }
2082 
2083 /*
2084  * perform DAD when interface becomes IFF_UP.
2085  */
2086 void
2087 in6_if_link_up(struct ifnet *ifp)
2088 {
2089 	struct ifaddr *ifa;
2090 	struct in6_ifaddr *ia;
2091 	int s, bound;
2092 	char ip6buf[INET6_ADDRSTRLEN];
2093 
2094 	/* Ensure it's sane to run DAD */
2095 	if (ifp->if_link_state == LINK_STATE_DOWN)
2096 		return;
2097 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2098 		return;
2099 
2100 	bound = curlwp_bind();
2101 	s = pserialize_read_enter();
2102 	IFADDR_READER_FOREACH(ifa, ifp) {
2103 		struct psref psref;
2104 
2105 		if (ifa->ifa_addr->sa_family != AF_INET6)
2106 			continue;
2107 
2108 		ifa_acquire(ifa, &psref);
2109 		pserialize_read_exit(s);
2110 		ia = (struct in6_ifaddr *)ifa;
2111 
2112 		/* If detached then mark as tentative */
2113 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2114 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2115 			if (if_do_dad(ifp)) {
2116 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2117 				nd6log(LOG_ERR, "%s marked tentative\n",
2118 				    IN6_PRINT(ip6buf,
2119 				    &ia->ia_addr.sin6_addr));
2120 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2121 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2122 		}
2123 
2124 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2125 			int rand_delay;
2126 
2127 			/* Clear the duplicated flag as we're starting DAD. */
2128 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2129 
2130 			/*
2131 			 * The TENTATIVE flag was likely set by hand
2132 			 * beforehand, implicitly indicating the need for DAD.
2133 			 * We may be able to skip the random delay in this
2134 			 * case, but we impose delays just in case.
2135 			 */
2136 			rand_delay = cprng_fast32() %
2137 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2138 			/* +1 ensures callout is always used */
2139 			nd6_dad_start(ifa, rand_delay + 1);
2140 		}
2141 
2142 		s = pserialize_read_enter();
2143 		ifa_release(ifa, &psref);
2144 	}
2145 	pserialize_read_exit(s);
2146 	curlwp_bindx(bound);
2147 
2148 	/* Restore any detached prefixes */
2149 	ND6_WLOCK();
2150 	nd6_pfxlist_onlink_check();
2151 	ND6_UNLOCK();
2152 }
2153 
2154 void
2155 in6_if_up(struct ifnet *ifp)
2156 {
2157 
2158 	/*
2159 	 * special cases, like 6to4, are handled in in6_ifattach
2160 	 */
2161 	in6_ifattach(ifp, NULL);
2162 
2163 	/* interface may not support link state, so bring it up also */
2164 	in6_if_link_up(ifp);
2165 }
2166 
2167 /*
2168  * Mark all addresses as detached.
2169  */
2170 void
2171 in6_if_link_down(struct ifnet *ifp)
2172 {
2173 	struct ifaddr *ifa;
2174 	struct in6_ifaddr *ia;
2175 	int s, bound;
2176 	char ip6buf[INET6_ADDRSTRLEN];
2177 
2178 	/* Any prefixes on this interface should be detached as well */
2179 	ND6_WLOCK();
2180 	nd6_pfxlist_onlink_check();
2181 	ND6_UNLOCK();
2182 
2183 	bound = curlwp_bind();
2184 	s = pserialize_read_enter();
2185 	IFADDR_READER_FOREACH(ifa, ifp) {
2186 		struct psref psref;
2187 
2188 		if (ifa->ifa_addr->sa_family != AF_INET6)
2189 			continue;
2190 
2191 		ifa_acquire(ifa, &psref);
2192 		pserialize_read_exit(s);
2193 		ia = (struct in6_ifaddr *)ifa;
2194 
2195 		/* Stop DAD processing */
2196 		nd6_dad_stop(ifa);
2197 
2198 		/*
2199 		 * Mark the address as detached.
2200 		 * This satisfies RFC4862 Section 5.3, but we should apply
2201 		 * this logic to all addresses to be a good citizen and
2202 		 * avoid potential duplicated addresses.
2203 		 * When the interface comes up again, detached addresses
2204 		 * are marked tentative and DAD commences.
2205 		 */
2206 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2207 			nd6log(LOG_DEBUG, "%s marked detached\n",
2208 			    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
2209 			ia->ia6_flags |= IN6_IFF_DETACHED;
2210 			ia->ia6_flags &=
2211 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2212 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2213 		}
2214 
2215 		s = pserialize_read_enter();
2216 		ifa_release(ifa, &psref);
2217 	}
2218 	pserialize_read_exit(s);
2219 	curlwp_bindx(bound);
2220 }
2221 
2222 void
2223 in6_if_down(struct ifnet *ifp)
2224 {
2225 
2226 	in6_if_link_down(ifp);
2227 }
2228 
2229 void
2230 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2231 {
2232 
2233 	switch (link_state) {
2234 	case LINK_STATE_DOWN:
2235 		in6_if_link_down(ifp);
2236 		break;
2237 	case LINK_STATE_UP:
2238 		in6_if_link_up(ifp);
2239 		break;
2240 	}
2241 }
2242 
2243 /*
2244  * Calculate max IPv6 MTU through all the interfaces and store it
2245  * to in6_maxmtu.
2246  */
2247 void
2248 in6_setmaxmtu(void)
2249 {
2250 	unsigned long maxmtu = 0;
2251 	struct ifnet *ifp;
2252 	int s;
2253 
2254 	s = pserialize_read_enter();
2255 	IFNET_READER_FOREACH(ifp) {
2256 		/* this function can be called during ifnet initialization */
2257 		if (!ifp->if_afdata[AF_INET6])
2258 			continue;
2259 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2260 		    IN6_LINKMTU(ifp) > maxmtu)
2261 			maxmtu = IN6_LINKMTU(ifp);
2262 	}
2263 	pserialize_read_exit(s);
2264 	if (maxmtu)	     /* update only when maxmtu is positive */
2265 		in6_maxmtu = maxmtu;
2266 }
2267 
2268 /*
2269  * Provide the length of interface identifiers to be used for the link attached
2270  * to the given interface.  The length should be defined in "IPv6 over
2271  * xxx-link" document.  Note that address architecture might also define
2272  * the length for a particular set of address prefixes, regardless of the
2273  * link type.  As clarified in rfc2462bis, those two definitions should be
2274  * consistent, and those really are as of August 2004.
2275  */
2276 int
2277 in6_if2idlen(struct ifnet *ifp)
2278 {
2279 	switch (ifp->if_type) {
2280 	case IFT_ETHER:		/* RFC2464 */
2281 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2282 	case IFT_L2VLAN:	/* ditto */
2283 	case IFT_IEEE80211:	/* ditto */
2284 	case IFT_FDDI:		/* RFC2467 */
2285 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2286 	case IFT_PPP:		/* RFC2472 */
2287 	case IFT_ARCNET:	/* RFC2497 */
2288 	case IFT_FRELAY:	/* RFC2590 */
2289 	case IFT_IEEE1394:	/* RFC3146 */
2290 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2291 	case IFT_LOOP:		/* XXX: is this really correct? */
2292 		return 64;
2293 	default:
2294 		/*
2295 		 * Unknown link type:
2296 		 * It might be controversial to use the today's common constant
2297 		 * of 64 for these cases unconditionally.  For full compliance,
2298 		 * we should return an error in this case.  On the other hand,
2299 		 * if we simply miss the standard for the link type or a new
2300 		 * standard is defined for a new link type, the IFID length
2301 		 * is very likely to be the common constant.  As a compromise,
2302 		 * we always use the constant, but make an explicit notice
2303 		 * indicating the "unknown" case.
2304 		 */
2305 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2306 		return 64;
2307 	}
2308 }
2309 
2310 struct in6_llentry {
2311 	struct llentry		base;
2312 };
2313 
2314 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2315 #define	IN6_LLTBL_HASH(k, h) \
2316 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2317 
2318 /*
2319  * Do actual deallocation of @lle.
2320  * Called by LLE_FREE_LOCKED when number of references
2321  * drops to zero.
2322  */
2323 static void
2324 in6_lltable_destroy_lle(struct llentry *lle)
2325 {
2326 
2327 	LLE_WUNLOCK(lle);
2328 	LLE_LOCK_DESTROY(lle);
2329 	kmem_intr_free(lle, sizeof(struct in6_llentry));
2330 }
2331 
2332 static struct llentry *
2333 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2334 {
2335 	struct in6_llentry *lle;
2336 
2337 	lle = kmem_intr_zalloc(sizeof(struct in6_llentry), KM_NOSLEEP);
2338 	if (lle == NULL)		/* NB: caller generates msg */
2339 		return NULL;
2340 
2341 	lle->base.r_l3addr.addr6 = *addr6;
2342 	lle->base.lle_refcnt = 1;
2343 	lle->base.lle_free = in6_lltable_destroy_lle;
2344 	LLE_LOCK_INIT(&lle->base);
2345 	callout_init(&lle->base.lle_timer, CALLOUT_MPSAFE);
2346 
2347 	return &lle->base;
2348 }
2349 
2350 static int
2351 in6_lltable_match_prefix(const struct sockaddr *prefix,
2352     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2353 {
2354 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2355 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2356 
2357 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2358 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2359 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2360 		return 1;
2361 
2362 	return 0;
2363 }
2364 
2365 static void
2366 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2367 {
2368 	struct ifnet *ifp = llt->llt_ifp;
2369 
2370 	IF_AFDATA_WLOCK_ASSERT(ifp);
2371 	LLE_WLOCK_ASSERT(lle);
2372 
2373 	/* Unlink entry from table */
2374 	if ((lle->la_flags & LLE_LINKED) != 0) {
2375 
2376 		lltable_unlink_entry(llt, lle);
2377 		KASSERT((lle->la_flags & LLE_LINKED) == 0);
2378 	}
2379 	/*
2380 	 * We need to release the lock here to lle_timer proceeds;
2381 	 * lle_timer should stop immediately if LLE_LINKED isn't set.
2382 	 * Note that we cannot pass lle->lle_lock to callout_halt
2383 	 * because it's a rwlock.
2384 	 */
2385 	LLE_ADDREF(lle);
2386 	LLE_WUNLOCK(lle);
2387 	IF_AFDATA_WUNLOCK(ifp);
2388 
2389 #ifdef NET_MPSAFE
2390 	callout_halt(&lle->lle_timer, NULL);
2391 #else
2392 	if (mutex_owned(softnet_lock))
2393 		callout_halt(&lle->lle_timer, softnet_lock);
2394 	else
2395 		callout_halt(&lle->lle_timer, NULL);
2396 #endif
2397 	LLE_WLOCK(lle);
2398 	LLE_REMREF(lle);
2399 
2400 	lltable_drop_entry_queue(lle);
2401 	LLE_FREE_LOCKED(lle);
2402 
2403 	IF_AFDATA_WLOCK(ifp);
2404 }
2405 
2406 static int
2407 in6_lltable_rtcheck(struct ifnet *ifp,
2408 		    u_int flags,
2409 		    const struct sockaddr *l3addr)
2410 {
2411 	struct rtentry *rt;
2412 	char ip6buf[INET6_ADDRSTRLEN];
2413 
2414 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2415 	    "sin_family %d", l3addr->sa_family);
2416 
2417 	rt = rtalloc1(l3addr, 0);
2418 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2419 		int s;
2420 		struct ifaddr *ifa;
2421 		/*
2422 		 * Create an ND6 cache for an IPv6 neighbor
2423 		 * that is not covered by our own prefix.
2424 		 */
2425 		/* XXX ifaof_ifpforaddr should take a const param */
2426 		s = pserialize_read_enter();
2427 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2428 		if (ifa != NULL) {
2429 			pserialize_read_exit(s);
2430 			if (rt != NULL)
2431 				rt_unref(rt);
2432 			return 0;
2433 		}
2434 		pserialize_read_exit(s);
2435 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2436 		    IN6_PRINT(ip6buf,
2437 		    &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2438 		if (rt != NULL)
2439 			rt_unref(rt);
2440 		return EINVAL;
2441 	}
2442 	rt_unref(rt);
2443 	return 0;
2444 }
2445 
2446 static inline uint32_t
2447 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2448 {
2449 
2450 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2451 }
2452 
2453 static uint32_t
2454 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2455 {
2456 
2457 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2458 }
2459 
2460 static void
2461 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2462 {
2463 	struct sockaddr_in6 *sin6;
2464 
2465 	sin6 = (struct sockaddr_in6 *)sa;
2466 	bzero(sin6, sizeof(*sin6));
2467 	sin6->sin6_family = AF_INET6;
2468 	sin6->sin6_len = sizeof(*sin6);
2469 	sin6->sin6_addr = lle->r_l3addr.addr6;
2470 }
2471 
2472 static inline struct llentry *
2473 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2474 {
2475 	struct llentry *lle;
2476 	struct llentries *lleh;
2477 	u_int hashidx;
2478 
2479 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2480 	lleh = &llt->lle_head[hashidx];
2481 	LIST_FOREACH(lle, lleh, lle_next) {
2482 		if (lle->la_flags & LLE_DELETED)
2483 			continue;
2484 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2485 			break;
2486 	}
2487 
2488 	return lle;
2489 }
2490 
2491 static int
2492 in6_lltable_delete(struct lltable *llt, u_int flags,
2493 	const struct sockaddr *l3addr)
2494 {
2495 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2496 	struct llentry *lle;
2497 
2498 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2499 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2500 	    "sin_family %d", l3addr->sa_family);
2501 
2502 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2503 
2504 	if (lle == NULL)
2505 		return ENOENT;
2506 
2507 	LLE_WLOCK(lle);
2508 	lle->la_flags |= LLE_DELETED;
2509 #ifdef DIAGNOSTIC
2510 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2511 #endif
2512 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2513 		llentry_free(lle);
2514 	else
2515 		LLE_WUNLOCK(lle);
2516 
2517 	return 0;
2518 }
2519 
2520 static struct llentry *
2521 in6_lltable_create(struct lltable *llt, u_int flags,
2522 	const struct sockaddr *l3addr)
2523 {
2524 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2525 	struct ifnet *ifp = llt->llt_ifp;
2526 	struct llentry *lle;
2527 
2528 	IF_AFDATA_WLOCK_ASSERT(ifp);
2529 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2530 	    "sin_family %d", l3addr->sa_family);
2531 
2532 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2533 
2534 	if (lle != NULL) {
2535 		LLE_WLOCK(lle);
2536 		return lle;
2537 	}
2538 
2539 	/*
2540 	 * A route that covers the given address must have
2541 	 * been installed 1st because we are doing a resolution,
2542 	 * verify this.
2543 	 */
2544 	if (!(flags & LLE_IFADDR) &&
2545 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2546 		return NULL;
2547 
2548 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2549 	if (lle == NULL) {
2550 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2551 		return NULL;
2552 	}
2553 	lle->la_flags = flags;
2554 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2555 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2556 		lle->la_flags |= LLE_VALID;
2557 	}
2558 
2559 	lltable_link_entry(llt, lle);
2560 	LLE_WLOCK(lle);
2561 
2562 	return lle;
2563 }
2564 
2565 static struct llentry *
2566 in6_lltable_lookup(struct lltable *llt, u_int flags,
2567 	const struct sockaddr *l3addr)
2568 {
2569 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2570 	struct llentry *lle;
2571 
2572 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2573 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2574 	    "sin_family %d", l3addr->sa_family);
2575 
2576 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2577 
2578 	if (lle == NULL)
2579 		return NULL;
2580 
2581 	if (flags & LLE_EXCLUSIVE)
2582 		LLE_WLOCK(lle);
2583 	else
2584 		LLE_RLOCK(lle);
2585 	return lle;
2586 }
2587 
2588 static int
2589 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2590     struct rt_walkarg *w)
2591 {
2592 	struct sockaddr_in6 sin6;
2593 
2594 	LLTABLE_LOCK_ASSERT();
2595 
2596 	/* skip deleted entries */
2597 	if (lle->la_flags & LLE_DELETED)
2598 		return 0;
2599 
2600 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2601 
2602 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2603 }
2604 
2605 static struct lltable *
2606 in6_lltattach(struct ifnet *ifp)
2607 {
2608 	struct lltable *llt;
2609 
2610 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2611 	llt->llt_af = AF_INET6;
2612 	llt->llt_ifp = ifp;
2613 
2614 	llt->llt_lookup = in6_lltable_lookup;
2615 	llt->llt_create = in6_lltable_create;
2616 	llt->llt_delete = in6_lltable_delete;
2617 	llt->llt_dump_entry = in6_lltable_dump_entry;
2618 	llt->llt_hash = in6_lltable_hash;
2619 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2620 	llt->llt_free_entry = in6_lltable_free_entry;
2621 	llt->llt_match_prefix = in6_lltable_match_prefix;
2622 	lltable_link(llt);
2623 
2624 	return llt;
2625 }
2626 
2627 void *
2628 in6_domifattach(struct ifnet *ifp)
2629 {
2630 	struct in6_ifextra *ext;
2631 
2632 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2633 
2634 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2635 	    M_IFADDR, M_WAITOK|M_ZERO);
2636 
2637 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2638 	    M_IFADDR, M_WAITOK|M_ZERO);
2639 
2640 	ext->nd_ifinfo = nd6_ifattach(ifp);
2641 	ext->scope6_id = scope6_ifattach(ifp);
2642 	ext->nprefixes = 0;
2643 	ext->ndefrouters = 0;
2644 
2645 	ext->lltable = in6_lltattach(ifp);
2646 
2647 	return ext;
2648 }
2649 
2650 void
2651 in6_domifdetach(struct ifnet *ifp, void *aux)
2652 {
2653 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2654 
2655 	lltable_free(ext->lltable);
2656 	ext->lltable = NULL;
2657 	nd6_ifdetach(ifp, ext);
2658 	free(ext->in6_ifstat, M_IFADDR);
2659 	free(ext->icmp6_ifstat, M_IFADDR);
2660 	scope6_ifdetach(ext->scope6_id);
2661 	free(ext, M_IFADDR);
2662 }
2663 
2664 /*
2665  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2666  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2667  */
2668 void
2669 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2670 {
2671 	in6->s6_addr32[0] = 0;
2672 	in6->s6_addr32[1] = 0;
2673 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2674 	in6->s6_addr32[3] = in->s_addr;
2675 }
2676 
2677 /*
2678  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2679  * v4 mapped addr or v4 compat addr
2680  */
2681 void
2682 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2683 {
2684 	memset(sin, 0, sizeof(*sin));
2685 	sin->sin_len = sizeof(struct sockaddr_in);
2686 	sin->sin_family = AF_INET;
2687 	sin->sin_port = sin6->sin6_port;
2688 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2689 }
2690 
2691 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2692 void
2693 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2694 {
2695 	memset(sin6, 0, sizeof(*sin6));
2696 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2697 	sin6->sin6_family = AF_INET6;
2698 	sin6->sin6_port = sin->sin_port;
2699 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2700 }
2701 
2702 /* Convert sockaddr_in6 into sockaddr_in. */
2703 void
2704 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2705 {
2706 	struct sockaddr_in *sin_p;
2707 	struct sockaddr_in6 sin6;
2708 
2709 	/*
2710 	 * Save original sockaddr_in6 addr and convert it
2711 	 * to sockaddr_in.
2712 	 */
2713 	sin6 = *(struct sockaddr_in6 *)nam;
2714 	sin_p = (struct sockaddr_in *)nam;
2715 	in6_sin6_2_sin(sin_p, &sin6);
2716 }
2717 
2718 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2719 void
2720 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2721 {
2722 	struct sockaddr_in *sin_p;
2723 	struct sockaddr_in6 *sin6_p;
2724 
2725 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2726 	sin_p = (struct sockaddr_in *)*nam;
2727 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2728 	free(*nam, M_SONAME);
2729 	*nam = sin6tosa(sin6_p);
2730 }
2731