xref: /netbsd-src/sys/netinet6/in6.c (revision 9616dacfef448e70e3fbbd865bddf60d54b656c5)
1 /*	$NetBSD: in6.c,v 1.231 2017/01/10 05:42:34 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 1.231 2017/01/10 05:42:34 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 = splnet();
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 = splnet();
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 splnet()?
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 
800 	in6m_sol = NULL;
801 
802 	/* Validate parameters */
803 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
804 		return EINVAL;
805 
806 	/*
807 	 * The destination address for a p2p link must have a family
808 	 * of AF_UNSPEC or AF_INET6.
809 	 */
810 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
811 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
812 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
813 		return EAFNOSUPPORT;
814 	/*
815 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
816 	 * does not carry fields other than sin6_len.
817 	 */
818 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
819 		return EINVAL;
820 	/*
821 	 * Because the IPv6 address architecture is classless, we require
822 	 * users to specify a (non 0) prefix length (mask) for a new address.
823 	 * We also require the prefix (when specified) mask is valid, and thus
824 	 * reject a non-consecutive mask.
825 	 */
826 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
827 		return EINVAL;
828 	if (ifra->ifra_prefixmask.sin6_len != 0) {
829 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
830 		    (u_char *)&ifra->ifra_prefixmask +
831 		    ifra->ifra_prefixmask.sin6_len);
832 		if (plen <= 0)
833 			return EINVAL;
834 	} else {
835 		/*
836 		 * In this case, ia must not be NULL.  We just use its prefix
837 		 * length.
838 		 */
839 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
840 	}
841 	/*
842 	 * If the destination address on a p2p interface is specified,
843 	 * and the address is a scoped one, validate/set the scope
844 	 * zone identifier.
845 	 */
846 	dst6 = ifra->ifra_dstaddr;
847 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
848 	    (dst6.sin6_family == AF_INET6)) {
849 		struct in6_addr in6_tmp;
850 		u_int32_t zoneid;
851 
852 		in6_tmp = dst6.sin6_addr;
853 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
854 			return EINVAL; /* XXX: should be impossible */
855 
856 		if (dst6.sin6_scope_id != 0) {
857 			if (dst6.sin6_scope_id != zoneid)
858 				return EINVAL;
859 		} else		/* user omit to specify the ID. */
860 			dst6.sin6_scope_id = zoneid;
861 
862 		/* convert into the internal form */
863 		if (sa6_embedscope(&dst6, 0))
864 			return EINVAL; /* XXX: should be impossible */
865 	}
866 	/*
867 	 * The destination address can be specified only for a p2p or a
868 	 * loopback interface.  If specified, the corresponding prefix length
869 	 * must be 128.
870 	 */
871 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
872 #ifdef FORCE_P2PPLEN
873 		int i;
874 #endif
875 
876 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
877 			/* XXX: noisy message */
878 			nd6log(LOG_INFO, "a destination can "
879 			    "be specified for a p2p or a loopback IF only\n");
880 			return EINVAL;
881 		}
882 		if (plen != 128) {
883 			nd6log(LOG_INFO, "prefixlen should "
884 			    "be 128 when dstaddr is specified\n");
885 #ifdef FORCE_P2PPLEN
886 			/*
887 			 * To be compatible with old configurations,
888 			 * such as ifconfig gif0 inet6 2001::1 2001::2
889 			 * prefixlen 126, we override the specified
890 			 * prefixmask as if the prefix length was 128.
891 			 */
892 			ifra->ifra_prefixmask.sin6_len =
893 			    sizeof(struct sockaddr_in6);
894 			for (i = 0; i < 4; i++)
895 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
896 				    0xffffffff;
897 			plen = 128;
898 #else
899 			return EINVAL;
900 #endif
901 		}
902 	}
903 	/* lifetime consistency check */
904 	lt = &ifra->ifra_lifetime;
905 	if (lt->ia6t_pltime > lt->ia6t_vltime)
906 		return EINVAL;
907 	if (lt->ia6t_vltime == 0) {
908 		/*
909 		 * the following log might be noisy, but this is a typical
910 		 * configuration mistake or a tool's bug.
911 		 */
912 		nd6log(LOG_INFO, "valid lifetime is 0 for %s\n",
913 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr));
914 
915 		if (ia == NULL)
916 			return 0; /* there's nothing to do */
917 	}
918 
919 	/*
920 	 * If this is a new address, allocate a new ifaddr and link it
921 	 * into chains.
922 	 */
923 	if (ia == NULL) {
924 		hostIsNew = 1;
925 		/*
926 		 * When in6_update_ifa() is called in a process of a received
927 		 * RA, it is called under an interrupt context.  So, we should
928 		 * call malloc with M_NOWAIT.
929 		 */
930 		ia = malloc(sizeof(*ia), M_IFADDR, M_NOWAIT|M_ZERO);
931 		if (ia == NULL)
932 			return ENOBUFS;
933 		LIST_INIT(&ia->ia6_memberships);
934 		/* Initialize the address and masks, and put time stamp */
935 		ia->ia_ifa.ifa_addr = sin6tosa(&ia->ia_addr);
936 		ia->ia_addr.sin6_family = AF_INET6;
937 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
938 		ia->ia6_createtime = time_uptime;
939 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
940 			/*
941 			 * XXX: some functions expect that ifa_dstaddr is not
942 			 * NULL for p2p interfaces.
943 			 */
944 			ia->ia_ifa.ifa_dstaddr = sin6tosa(&ia->ia_dstaddr);
945 		} else {
946 			ia->ia_ifa.ifa_dstaddr = NULL;
947 		}
948 		ia->ia_ifa.ifa_netmask = sin6tosa(&ia->ia_prefixmask);
949 
950 		ia->ia_ifp = ifp;
951 		IN6_ADDRLIST_ENTRY_INIT(ia);
952 		ifa_psref_init(&ia->ia_ifa);
953 		if (psref)
954 			ia6_acquire(ia, psref);
955 	}
956 
957 	/* update timestamp */
958 	ia->ia6_updatetime = time_uptime;
959 
960 	/* set prefix mask */
961 	if (ifra->ifra_prefixmask.sin6_len) {
962 		if (ia->ia_prefixmask.sin6_len) {
963 			/*
964 			 * We prohibit changing the prefix length of an
965 			 * existing autoconf address, because the operation
966 			 * would confuse prefix management.
967 			 */
968 			if (ia->ia6_ndpr != NULL &&
969 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) !=
970 			    plen)
971 			{
972 				nd6log(LOG_INFO, "the prefix length of an"
973 				    " existing (%s) autoconf address should"
974 				    " not be changed\n",
975 				    ip6_sprintf(&ia->ia_addr.sin6_addr));
976 				error = EINVAL;
977 				if (hostIsNew)
978 					free(ia, M_IFADDR);
979 				goto exit;
980 			}
981 
982 			if (!IN6_ARE_ADDR_EQUAL(&ia->ia_prefixmask.sin6_addr,
983 			    &ifra->ifra_prefixmask.sin6_addr))
984 				in6_ifremprefix(ia);
985 		}
986 		ia->ia_prefixmask = ifra->ifra_prefixmask;
987 	}
988 
989 	/* Set destination address. */
990 	if (dst6.sin6_family == AF_INET6) {
991 		if (!IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
992 		    &ia->ia_dstaddr.sin6_addr))
993 			in6_ifremprefix(ia);
994 		ia->ia_dstaddr = dst6;
995 	}
996 
997 	/*
998 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
999 	 * to see if the address is deprecated or invalidated, but initialize
1000 	 * these members for applications.
1001 	 */
1002 	ia->ia6_lifetime = ifra->ifra_lifetime;
1003 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1004 		ia->ia6_lifetime.ia6t_expire =
1005 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1006 	} else
1007 		ia->ia6_lifetime.ia6t_expire = 0;
1008 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1009 		ia->ia6_lifetime.ia6t_preferred =
1010 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1011 	} else
1012 		ia->ia6_lifetime.ia6t_preferred = 0;
1013 
1014 	/*
1015 	 * configure address flags.
1016 	 * We need to preserve tentative state so DAD works if
1017 	 * something adds the same address before DAD finishes.
1018 	 */
1019 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1020 	ia->ia6_flags = ifra->ifra_flags;
1021 
1022 	/*
1023 	 * Make the address tentative before joining multicast addresses,
1024 	 * so that corresponding MLD responses would not have a tentative
1025 	 * source address.
1026 	 */
1027 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1028 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1029 		ia->ia6_flags |= IN6_IFF_DETACHED;
1030 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1031 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp))
1032 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1033 
1034 	/*
1035 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1036 	 * userland, make it deprecated.
1037 	 */
1038 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1039 		ia->ia6_lifetime.ia6t_pltime = 0;
1040 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1041 	}
1042 
1043 	/* reset the interface and routing table appropriately. */
1044 	error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew);
1045 	if (error != 0) {
1046 		if (hostIsNew)
1047 			free(ia, M_IFADDR);
1048 		goto exit;
1049 	}
1050 
1051 	/*
1052 	 * We are done if we have simply modified an existing address.
1053 	 */
1054 	if (!hostIsNew)
1055 		return error;
1056 
1057 	/*
1058 	 * Insert ia to the global list and ifa to the interface's list.
1059 	 */
1060 	mutex_enter(&in6_ifaddr_lock);
1061 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1062 	mutex_exit(&in6_ifaddr_lock);
1063 
1064 	/* gain a refcnt for the link from in6_ifaddr */
1065 	ifaref(&ia->ia_ifa);
1066 	ifa_insert(ifp, &ia->ia_ifa);
1067 
1068 	/*
1069 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1070 	 * not just go to unlink.
1071 	 */
1072 
1073 	/* join necessary multicast groups */
1074 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1075 		struct sockaddr_in6 mltaddr, mltmask;
1076 		struct in6_addr llsol;
1077 
1078 		/* join solicited multicast addr for new host id */
1079 		memset(&llsol, 0, sizeof(struct in6_addr));
1080 		llsol.s6_addr16[0] = htons(0xff02);
1081 		llsol.s6_addr32[1] = 0;
1082 		llsol.s6_addr32[2] = htonl(1);
1083 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1084 		llsol.s6_addr8[12] = 0xff;
1085 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1086 			/* XXX: should not happen */
1087 			log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
1088 			goto cleanup;
1089 		}
1090 		dad_delay = 0;
1091 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1092 			/*
1093 			 * We need a random delay for DAD on the address
1094 			 * being configured.  It also means delaying
1095 			 * transmission of the corresponding MLD report to
1096 			 * avoid report collision.
1097 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1098 			 */
1099 			dad_delay = cprng_fast32() %
1100 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1101 		}
1102 
1103 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1104 		/* join solicited multicast addr for new host id */
1105 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
1106 		if (!imm) {
1107 			nd6log(LOG_ERR,
1108 			    "addmulti failed for %s on %s (errno=%d)\n",
1109 			    ip6_sprintf(&llsol), if_name(ifp), error);
1110 			goto cleanup;
1111 		}
1112 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1113 		in6m_sol = imm->i6mm_maddr;
1114 
1115 		sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
1116 
1117 		/*
1118 		 * join link-local all-nodes address
1119 		 */
1120 		sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
1121 		    0, 0, 0);
1122 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1123 			goto cleanup; /* XXX: should not fail */
1124 
1125 		/*
1126 		 * XXX: do we really need this automatic routes?
1127 		 * We should probably reconsider this stuff.  Most applications
1128 		 * actually do not need the routes, since they usually specify
1129 		 * the outgoing interface.
1130 		 */
1131 		rt = rtalloc1(sin6tosa(&mltaddr), 0);
1132 		if (rt) {
1133 			if (memcmp(&mltaddr.sin6_addr,
1134 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1135 			    MLTMASK_LEN)) {
1136 				rt_unref(rt);
1137 				rt = NULL;
1138 			} else if (rt->rt_ifp != ifp) {
1139 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1140 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1141 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1142 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1143 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1144 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1145 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1146 				rt_replace_ifa(rt, &ia->ia_ifa);
1147 				rt->rt_ifp = ifp;
1148 			}
1149 		}
1150 		if (!rt) {
1151 			struct rt_addrinfo info;
1152 
1153 			memset(&info, 0, sizeof(info));
1154 			info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
1155 			info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
1156 			info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
1157 			info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
1158 			/* XXX: we need RTF_CONNECTED to fake nd6_rtrequest */
1159 			info.rti_flags = RTF_UP | RTF_CONNECTED;
1160 			error = rtrequest1(RTM_ADD, &info, NULL);
1161 			if (error)
1162 				goto cleanup;
1163 		} else {
1164 			rt_unref(rt);
1165 		}
1166 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1167 		if (!imm) {
1168 			nd6log(LOG_WARNING,
1169 			    "addmulti failed for %s on %s (errno=%d)\n",
1170 			    ip6_sprintf(&mltaddr.sin6_addr),
1171 			    if_name(ifp), error);
1172 			goto cleanup;
1173 		}
1174 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1175 
1176 		/*
1177 		 * join node information group address
1178 		 */
1179 		dad_delay = 0;
1180 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1181 			/*
1182 			 * The spec doesn't say anything about delay for this
1183 			 * group, but the same logic should apply.
1184 			 */
1185 			dad_delay = cprng_fast32() %
1186 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1187 		}
1188 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
1189 			;
1190 		else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1191 		          dad_delay)) == NULL) { /* XXX jinmei */
1192 			nd6log(LOG_WARNING,
1193 			    "addmulti failed for %s on %s (errno=%d)\n",
1194 			    ip6_sprintf(&mltaddr.sin6_addr),
1195 			    if_name(ifp), error);
1196 			/* XXX not very fatal, go on... */
1197 		} else {
1198 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1199 		}
1200 
1201 
1202 		/*
1203 		 * join interface-local all-nodes address.
1204 		 * (ff01::1%ifN, and ff01::%ifN/32)
1205 		 */
1206 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1207 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1208 			goto cleanup; /* XXX: should not fail */
1209 
1210 		/* XXX: again, do we really need the route? */
1211 		rt = rtalloc1(sin6tosa(&mltaddr), 0);
1212 		if (rt) {
1213 			/* 32bit came from "mltmask" */
1214 			if (memcmp(&mltaddr.sin6_addr,
1215 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1216 			    32 / NBBY)) {
1217 				rt_unref(rt);
1218 				rt = NULL;
1219 			} else if (rt->rt_ifp != ifp) {
1220 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1221 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1222 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1223 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1224 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1225 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1226 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1227 				rt_replace_ifa(rt, &ia->ia_ifa);
1228 				rt->rt_ifp = ifp;
1229 			}
1230 		}
1231 		if (!rt) {
1232 			struct rt_addrinfo info;
1233 
1234 			memset(&info, 0, sizeof(info));
1235 			info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
1236 			info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia->ia_addr);
1237 			info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
1238 			info.rti_info[RTAX_IFA] = sin6tosa(&ia->ia_addr);
1239 			info.rti_flags = RTF_UP | RTF_CONNECTED;
1240 			error = rtrequest1(RTM_ADD, &info, NULL);
1241 			if (error)
1242 				goto cleanup;
1243 #undef	MLTMASK_LEN
1244 		} else {
1245 			rt_unref(rt);
1246 		}
1247 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1248 		if (!imm) {
1249 			nd6log(LOG_WARNING,
1250 			    "addmulti failed for %s on %s (errno=%d)\n",
1251 			    ip6_sprintf(&mltaddr.sin6_addr),
1252 			    if_name(ifp), error);
1253 			goto cleanup;
1254 		} else {
1255 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1256 		}
1257 	}
1258 
1259 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1260 	error = nd6_add_ifa_lle(ia);
1261 	if (error != 0)
1262 		goto cleanup;
1263 
1264 	/*
1265 	 * Perform DAD, if needed.
1266 	 * XXX It may be of use, if we can administratively
1267 	 * disable DAD.
1268 	 */
1269 	if (hostIsNew && if_do_dad(ifp) &&
1270 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1271 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1272 	{
1273 		int mindelay, maxdelay;
1274 
1275 		dad_delay = 0;
1276 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1277 			/*
1278 			 * We need to impose a delay before sending an NS
1279 			 * for DAD.  Check if we also needed a delay for the
1280 			 * corresponding MLD message.  If we did, the delay
1281 			 * should be larger than the MLD delay (this could be
1282 			 * relaxed a bit, but this simple logic is at least
1283 			 * safe).
1284 			 */
1285 			mindelay = 0;
1286 			if (in6m_sol != NULL &&
1287 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1288 				mindelay = in6m_sol->in6m_timer;
1289 			}
1290 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1291 			if (maxdelay - mindelay == 0)
1292 				dad_delay = 0;
1293 			else {
1294 				dad_delay =
1295 				    (cprng_fast32() % (maxdelay - mindelay)) +
1296 				    mindelay;
1297 			}
1298 		}
1299 		/* +1 ensures callout is always used */
1300 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1301 	}
1302 
1303 	if (iap)
1304 		*iap = ia;
1305 
1306 	return 0;
1307 
1308   cleanup:
1309 	in6_purgeaddr(&ia->ia_ifa);
1310   exit:
1311 	return error;
1312 }
1313 
1314 int
1315 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1316 {
1317 	int rc, s;
1318 
1319 	s = splnet();
1320 	rc = in6_update_ifa1(ifp, ifra, NULL, NULL, flags);
1321 	splx(s);
1322 	return rc;
1323 }
1324 
1325 void
1326 in6_purgeaddr(struct ifaddr *ifa)
1327 {
1328 	struct ifnet *ifp = ifa->ifa_ifp;
1329 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1330 	struct in6_multi_mship *imm;
1331 
1332 	KASSERT(!ifa_held(ifa));
1333 
1334 	ifa->ifa_flags |= IFA_DESTROYING;
1335 
1336 	/* stop DAD processing */
1337 	nd6_dad_stop(ifa);
1338 
1339 	/* Delete any network route. */
1340 	in6_ifremprefix(ia);
1341 
1342 	/* Remove ownaddr's loopback rtentry, if it exists. */
1343 	in6_ifremlocal(&(ia->ia_ifa));
1344 
1345 	/*
1346 	 * leave from multicast groups we have joined for the interface
1347 	 */
1348 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1349 		LIST_REMOVE(imm, i6mm_chain);
1350 		in6_leavegroup(imm);
1351 	}
1352 
1353 	in6_unlink_ifa(ia, ifp);
1354 }
1355 
1356 static void
1357 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1358 {
1359 	int	s = splnet();
1360 
1361 	mutex_enter(&in6_ifaddr_lock);
1362 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1363 	ifa_remove(ifp, &ia->ia_ifa);
1364 	mutex_exit(&in6_ifaddr_lock);
1365 
1366 	/*
1367 	 * XXX thorpej@NetBSD.org -- if the interface is going
1368 	 * XXX away, don't save the multicast entries, delete them!
1369 	 */
1370 	if (LIST_EMPTY(&ia->ia6_multiaddrs))
1371 		;
1372 	else if (if_is_deactivated(ia->ia_ifa.ifa_ifp)) {
1373 		struct in6_multi *in6m, *next;
1374 
1375 		for (in6m = LIST_FIRST(&ia->ia6_multiaddrs); in6m != NULL;
1376 		     in6m = next) {
1377 			next = LIST_NEXT(in6m, in6m_entry);
1378 			in6_delmulti(in6m);
1379 		}
1380 	} else
1381 		in6_savemkludge(ia);
1382 
1383 	/*
1384 	 * Release the reference to the ND prefix.
1385 	 */
1386 	if (ia->ia6_ndpr != NULL) {
1387 		nd6_prefix_unref(ia->ia6_ndpr);
1388 		ia->ia6_ndpr = NULL;
1389 	}
1390 
1391 	/*
1392 	 * Also, if the address being removed is autoconf'ed, call
1393 	 * nd6_pfxlist_onlink_check() since the release might affect the status of
1394 	 * other (detached) addresses.
1395 	 */
1396 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0) {
1397 		ND6_WLOCK();
1398 		nd6_pfxlist_onlink_check();
1399 		ND6_UNLOCK();
1400 	}
1401 
1402 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1403 
1404 	/*
1405 	 * release another refcnt for the link from in6_ifaddr.
1406 	 * Note that we should decrement the refcnt at least once for all *BSD.
1407 	 */
1408 	ifafree(&ia->ia_ifa);
1409 
1410 	splx(s);
1411 }
1412 
1413 void
1414 in6_purgeif(struct ifnet *ifp)
1415 {
1416 
1417 	in6_ifdetach(ifp);
1418 }
1419 
1420 /*
1421  * SIOC[GAD]LIFADDR.
1422  *	SIOCGLIFADDR: get first address. (?)
1423  *	SIOCGLIFADDR with IFLR_PREFIX:
1424  *		get first address that matches the specified prefix.
1425  *	SIOCALIFADDR: add the specified address.
1426  *	SIOCALIFADDR with IFLR_PREFIX:
1427  *		add the specified prefix, filling hostid part from
1428  *		the first link-local address.  prefixlen must be <= 64.
1429  *	SIOCDLIFADDR: delete the specified address.
1430  *	SIOCDLIFADDR with IFLR_PREFIX:
1431  *		delete the first address that matches the specified prefix.
1432  * return values:
1433  *	EINVAL on invalid parameters
1434  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1435  *	other values may be returned from in6_ioctl()
1436  *
1437  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1438  * this is to accommodate address naming scheme other than RFC2374,
1439  * in the future.
1440  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1441  * address encoding scheme. (see figure on page 8)
1442  */
1443 static int
1444 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1445 	struct ifnet *ifp)
1446 {
1447 	struct in6_ifaddr *ia = NULL; /* XXX gcc 4.8 maybe-uninitialized */
1448 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1449 	struct ifaddr *ifa;
1450 	struct sockaddr *sa;
1451 
1452 	/* sanity checks */
1453 	if (!data || !ifp) {
1454 		panic("invalid argument to in6_lifaddr_ioctl");
1455 		/* NOTREACHED */
1456 	}
1457 
1458 	switch (cmd) {
1459 	case SIOCGLIFADDR:
1460 		/* address must be specified on GET with IFLR_PREFIX */
1461 		if ((iflr->flags & IFLR_PREFIX) == 0)
1462 			break;
1463 		/* FALLTHROUGH */
1464 	case SIOCALIFADDR:
1465 	case SIOCDLIFADDR:
1466 		/* address must be specified on ADD and DELETE */
1467 		sa = (struct sockaddr *)&iflr->addr;
1468 		if (sa->sa_family != AF_INET6)
1469 			return EINVAL;
1470 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1471 			return EINVAL;
1472 		/* XXX need improvement */
1473 		sa = (struct sockaddr *)&iflr->dstaddr;
1474 		if (sa->sa_family && sa->sa_family != AF_INET6)
1475 			return EINVAL;
1476 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1477 			return EINVAL;
1478 		break;
1479 	default: /* shouldn't happen */
1480 #if 0
1481 		panic("invalid cmd to in6_lifaddr_ioctl");
1482 		/* NOTREACHED */
1483 #else
1484 		return EOPNOTSUPP;
1485 #endif
1486 	}
1487 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1488 		return EINVAL;
1489 
1490 	switch (cmd) {
1491 	case SIOCALIFADDR:
1492 	    {
1493 		struct in6_aliasreq ifra;
1494 		struct in6_addr *xhostid = NULL;
1495 		int prefixlen;
1496 		int bound = curlwp_bind();
1497 		struct psref psref;
1498 
1499 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1500 			struct sockaddr_in6 *sin6;
1501 
1502 			/*
1503 			 * xhostid is to fill in the hostid part of the
1504 			 * address.  xhostid points to the first link-local
1505 			 * address attached to the interface.
1506 			 */
1507 			ia = in6ifa_ifpforlinklocal_psref(ifp, 0, &psref);
1508 			if (ia == NULL) {
1509 				curlwp_bindx(bound);
1510 				return EADDRNOTAVAIL;
1511 			}
1512 			xhostid = IFA_IN6(&ia->ia_ifa);
1513 
1514 		 	/* prefixlen must be <= 64. */
1515 			if (64 < iflr->prefixlen) {
1516 				ia6_release(ia, &psref);
1517 				curlwp_bindx(bound);
1518 				return EINVAL;
1519 			}
1520 			prefixlen = iflr->prefixlen;
1521 
1522 			/* hostid part must be zero. */
1523 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1524 			if (sin6->sin6_addr.s6_addr32[2] != 0
1525 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1526 				ia6_release(ia, &psref);
1527 				curlwp_bindx(bound);
1528 				return EINVAL;
1529 			}
1530 		} else
1531 			prefixlen = iflr->prefixlen;
1532 
1533 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1534 		memset(&ifra, 0, sizeof(ifra));
1535 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1536 
1537 		memcpy(&ifra.ifra_addr, &iflr->addr,
1538 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1539 		if (xhostid) {
1540 			/* fill in hostid part */
1541 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1542 			    xhostid->s6_addr32[2];
1543 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1544 			    xhostid->s6_addr32[3];
1545 		}
1546 
1547 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1548 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1549 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1550 			if (xhostid) {
1551 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1552 				    xhostid->s6_addr32[2];
1553 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1554 				    xhostid->s6_addr32[3];
1555 			}
1556 		}
1557 		if (xhostid) {
1558 			ia6_release(ia, &psref);
1559 			ia = NULL;
1560 		}
1561 		curlwp_bindx(bound);
1562 
1563 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1564 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1565 
1566 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1567 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1568 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1569 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1570 	    }
1571 	case SIOCGLIFADDR:
1572 	case SIOCDLIFADDR:
1573 	    {
1574 		struct in6_addr mask, candidate, match;
1575 		struct sockaddr_in6 *sin6;
1576 		int cmp;
1577 		int error, s;
1578 
1579 		memset(&mask, 0, sizeof(mask));
1580 		if (iflr->flags & IFLR_PREFIX) {
1581 			/* lookup a prefix rather than address. */
1582 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1583 
1584 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1585 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1586 			match.s6_addr32[0] &= mask.s6_addr32[0];
1587 			match.s6_addr32[1] &= mask.s6_addr32[1];
1588 			match.s6_addr32[2] &= mask.s6_addr32[2];
1589 			match.s6_addr32[3] &= mask.s6_addr32[3];
1590 
1591 			/* if you set extra bits, that's wrong */
1592 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1593 				return EINVAL;
1594 
1595 			cmp = 1;
1596 		} else {
1597 			if (cmd == SIOCGLIFADDR) {
1598 				/* on getting an address, take the 1st match */
1599 				cmp = 0;	/* XXX */
1600 			} else {
1601 				/* on deleting an address, do exact match */
1602 				in6_prefixlen2mask(&mask, 128);
1603 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1604 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1605 
1606 				cmp = 1;
1607 			}
1608 		}
1609 
1610 		s = pserialize_read_enter();
1611 		IFADDR_READER_FOREACH(ifa, ifp) {
1612 			if (ifa->ifa_addr->sa_family != AF_INET6)
1613 				continue;
1614 			if (!cmp)
1615 				break;
1616 
1617 			/*
1618 			 * XXX: this is adhoc, but is necessary to allow
1619 			 * a user to specify fe80::/64 (not /10) for a
1620 			 * link-local address.
1621 			 */
1622 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1623 			in6_clearscope(&candidate);
1624 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1625 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1626 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1627 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1628 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1629 				break;
1630 		}
1631 		if (!ifa) {
1632 			error = EADDRNOTAVAIL;
1633 			goto error;
1634 		}
1635 		ia = ifa2ia6(ifa);
1636 
1637 		if (cmd == SIOCGLIFADDR) {
1638 			/* fill in the if_laddrreq structure */
1639 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1640 			error = sa6_recoverscope(
1641 			    (struct sockaddr_in6 *)&iflr->addr);
1642 			if (error != 0)
1643 				goto error;
1644 
1645 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1646 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1647 				    ia->ia_dstaddr.sin6_len);
1648 				error = sa6_recoverscope(
1649 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1650 				if (error != 0)
1651 					goto error;
1652 			} else
1653 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1654 
1655 			iflr->prefixlen =
1656 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1657 
1658 			iflr->flags = ia->ia6_flags;	/* XXX */
1659 
1660 			error = 0;
1661 		} else {
1662 			struct in6_aliasreq ifra;
1663 
1664 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1665 			memset(&ifra, 0, sizeof(ifra));
1666 			memcpy(ifra.ifra_name, iflr->iflr_name,
1667 			    sizeof(ifra.ifra_name));
1668 
1669 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1670 			    ia->ia_addr.sin6_len);
1671 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1672 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1673 				    ia->ia_dstaddr.sin6_len);
1674 			} else {
1675 				memset(&ifra.ifra_dstaddr, 0,
1676 				    sizeof(ifra.ifra_dstaddr));
1677 			}
1678 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1679 			    ia->ia_prefixmask.sin6_len);
1680 
1681 			ifra.ifra_flags = ia->ia6_flags;
1682 			pserialize_read_exit(s);
1683 
1684 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1685 		}
1686 	error:
1687 		pserialize_read_exit(s);
1688 		return error;
1689 	    }
1690 	}
1691 
1692 	return EOPNOTSUPP;	/* just for safety */
1693 }
1694 
1695 /*
1696  * Initialize an interface's internet6 address
1697  * and routing table entry.
1698  */
1699 static int
1700 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1701 	const struct sockaddr_in6 *sin6, int newhost)
1702 {
1703 	int	error = 0, ifacount = 0;
1704 	int	s = splnet();
1705 	struct ifaddr *ifa;
1706 
1707 	/*
1708 	 * Give the interface a chance to initialize
1709 	 * if this is its first address,
1710 	 * and to validate the address if necessary.
1711 	 */
1712 	IFADDR_READER_FOREACH(ifa, ifp) {
1713 		if (ifa->ifa_addr->sa_family != AF_INET6)
1714 			continue;
1715 		ifacount++;
1716 	}
1717 
1718 	ia->ia_addr = *sin6;
1719 
1720 	if (ifacount <= 0 &&
1721 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1722 		splx(s);
1723 		return error;
1724 	}
1725 	splx(s);
1726 
1727 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1728 
1729 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1730 
1731 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1732 	if (newhost) {
1733 		/* set the rtrequest function to create llinfo */
1734 		if (ifp->if_flags & IFF_POINTOPOINT)
1735 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1736 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1737 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1738 		in6_ifaddlocal(&ia->ia_ifa);
1739 	} else {
1740 		/* Inform the routing socket of new flags/timings */
1741 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
1742 	}
1743 
1744 	/* Add the network prefix route. */
1745 	if ((error = in6_ifaddprefix(ia)) != 0) {
1746 		if (newhost)
1747 			in6_ifremlocal(&ia->ia_ifa);
1748 		return error;
1749 	}
1750 
1751 	if (ifp->if_flags & IFF_MULTICAST)
1752 		in6_restoremkludge(ia, ifp);
1753 
1754 	return error;
1755 }
1756 
1757 static struct ifaddr *
1758 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1759 {
1760 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1761 		return ifa;
1762 	return best_ifa;
1763 }
1764 
1765 /*
1766  * Find an IPv6 interface link-local address specific to an interface.
1767  */
1768 struct in6_ifaddr *
1769 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1770 {
1771 	struct ifaddr *best_ifa = NULL, *ifa;
1772 
1773 	IFADDR_READER_FOREACH(ifa, ifp) {
1774 		if (ifa->ifa_addr->sa_family != AF_INET6)
1775 			continue;
1776 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1777 			continue;
1778 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1779 			continue;
1780 		best_ifa = bestifa(best_ifa, ifa);
1781 	}
1782 
1783 	return (struct in6_ifaddr *)best_ifa;
1784 }
1785 
1786 struct in6_ifaddr *
1787 in6ifa_ifpforlinklocal_psref(const struct ifnet *ifp, const int ignoreflags,
1788     struct psref *psref)
1789 {
1790 	struct in6_ifaddr *ia;
1791 	int s = pserialize_read_enter();
1792 
1793 	ia = in6ifa_ifpforlinklocal(ifp, ignoreflags);
1794 	if (ia != NULL)
1795 		ia6_acquire(ia, psref);
1796 	pserialize_read_exit(s);
1797 
1798 	return ia;
1799 }
1800 
1801 /*
1802  * find the internet address corresponding to a given address.
1803  * ifaddr is returned referenced.
1804  */
1805 struct in6_ifaddr *
1806 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1807 {
1808 	struct in6_ifaddr *ia;
1809 	int s;
1810 
1811 	s = pserialize_read_enter();
1812 	IN6_ADDRLIST_READER_FOREACH(ia) {
1813 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1814 			if (zoneid != 0 &&
1815 			    zoneid != ia->ia_addr.sin6_scope_id)
1816 				continue;
1817 			ifaref(&ia->ia_ifa);
1818 			break;
1819 		}
1820 	}
1821 	pserialize_read_exit(s);
1822 
1823 	return ia;
1824 }
1825 
1826 /*
1827  * find the internet address corresponding to a given interface and address.
1828  */
1829 struct in6_ifaddr *
1830 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
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_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1838 			continue;
1839 		best_ifa = bestifa(best_ifa, ifa);
1840 	}
1841 
1842 	return (struct in6_ifaddr *)best_ifa;
1843 }
1844 
1845 struct in6_ifaddr *
1846 in6ifa_ifpwithaddr_psref(const struct ifnet *ifp, const struct in6_addr *addr,
1847     struct psref *psref)
1848 {
1849 	struct in6_ifaddr *ia;
1850 	int s = pserialize_read_enter();
1851 
1852 	ia = in6ifa_ifpwithaddr(ifp, addr);
1853 	if (ia != NULL)
1854 		ia6_acquire(ia, psref);
1855 	pserialize_read_exit(s);
1856 
1857 	return ia;
1858 }
1859 
1860 static struct in6_ifaddr *
1861 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1862 {
1863 	if (best_ia == NULL ||
1864 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1865 		return ia;
1866 	return best_ia;
1867 }
1868 
1869 /*
1870  * Convert IP6 address to printable (loggable) representation.
1871  */
1872 char *
1873 ip6_sprintf(const struct in6_addr *addr)
1874 {
1875 	static int ip6round = 0;
1876 	static char ip6buf[8][INET6_ADDRSTRLEN];
1877 	char *cp = ip6buf[ip6round++ & 7];
1878 
1879 	in6_print(cp, INET6_ADDRSTRLEN, addr);
1880 	return cp;
1881 }
1882 
1883 /*
1884  * Determine if an address is on a local network.
1885  */
1886 int
1887 in6_localaddr(const struct in6_addr *in6)
1888 {
1889 	struct in6_ifaddr *ia;
1890 	int s;
1891 
1892 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1893 		return 1;
1894 
1895 	s = pserialize_read_enter();
1896 	IN6_ADDRLIST_READER_FOREACH(ia) {
1897 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1898 					      &ia->ia_prefixmask.sin6_addr)) {
1899 			pserialize_read_exit(s);
1900 			return 1;
1901 		}
1902 	}
1903 	pserialize_read_exit(s);
1904 
1905 	return 0;
1906 }
1907 
1908 int
1909 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1910 {
1911 	struct in6_ifaddr *ia;
1912 	int s;
1913 
1914 	s = pserialize_read_enter();
1915 	IN6_ADDRLIST_READER_FOREACH(ia) {
1916 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1917 		    &sa6->sin6_addr) &&
1918 #ifdef SCOPEDROUTING
1919 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1920 #endif
1921 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
1922 			pserialize_read_exit(s);
1923 			return 1; /* true */
1924 		}
1925 
1926 		/* XXX: do we still have to go thru the rest of the list? */
1927 	}
1928 	pserialize_read_exit(s);
1929 
1930 	return 0;		/* false */
1931 }
1932 
1933 /*
1934  * return length of part which dst and src are equal
1935  * hard coding...
1936  */
1937 int
1938 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1939 {
1940 	int match = 0;
1941 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1942 	u_char *lim = s + 16, r;
1943 
1944 	while (s < lim)
1945 		if ((r = (*d++ ^ *s++)) != 0) {
1946 			while (r < 128) {
1947 				match++;
1948 				r <<= 1;
1949 			}
1950 			break;
1951 		} else
1952 			match += NBBY;
1953 	return match;
1954 }
1955 
1956 /* XXX: to be scope conscious */
1957 int
1958 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1959 {
1960 	int bytelen, bitlen;
1961 
1962 	/* sanity check */
1963 	if (len < 0 || len > 128) {
1964 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1965 		    len);
1966 		return 0;
1967 	}
1968 
1969 	bytelen = len / NBBY;
1970 	bitlen = len % NBBY;
1971 
1972 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1973 		return 0;
1974 	if (bitlen != 0 &&
1975 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
1976 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
1977 		return 0;
1978 
1979 	return 1;
1980 }
1981 
1982 void
1983 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1984 {
1985 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1986 	int bytelen, bitlen, i;
1987 
1988 	/* sanity check */
1989 	if (len < 0 || len > 128) {
1990 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1991 		    len);
1992 		return;
1993 	}
1994 
1995 	memset(maskp, 0, sizeof(*maskp));
1996 	bytelen = len / NBBY;
1997 	bitlen = len % NBBY;
1998 	for (i = 0; i < bytelen; i++)
1999 		maskp->s6_addr[i] = 0xff;
2000 	if (bitlen)
2001 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2002 }
2003 
2004 /*
2005  * return the best address out of the same scope. if no address was
2006  * found, return the first valid address from designated IF.
2007  */
2008 struct in6_ifaddr *
2009 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2010 {
2011 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2012 	struct ifaddr *ifa;
2013 	struct in6_ifaddr *best_ia = NULL, *ia;
2014 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2015 
2016 	dep[0] = dep[1] = NULL;
2017 
2018 	/*
2019 	 * We first look for addresses in the same scope.
2020 	 * If there is one, return it.
2021 	 * If two or more, return one which matches the dst longest.
2022 	 * If none, return one of global addresses assigned other ifs.
2023 	 */
2024 	IFADDR_READER_FOREACH(ifa, ifp) {
2025 		if (ifa->ifa_addr->sa_family != AF_INET6)
2026 			continue;
2027 		ia = (struct in6_ifaddr *)ifa;
2028 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2029 			continue; /* XXX: is there any case to allow anycast? */
2030 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2031 			continue; /* don't use this interface */
2032 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2033 			continue;
2034 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2035 			if (ip6_use_deprecated)
2036 				dep[0] = ia;
2037 			continue;
2038 		}
2039 
2040 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
2041 			continue;
2042 		/*
2043 		 * call in6_matchlen() as few as possible
2044 		 */
2045 		if (best_ia == NULL) {
2046 			best_ia = ia;
2047 			continue;
2048 		}
2049 		if (blen == -1)
2050 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2051 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2052 		if (tlen > blen) {
2053 			blen = tlen;
2054 			best_ia = ia;
2055 		} else if (tlen == blen)
2056 			best_ia = bestia(best_ia, ia);
2057 	}
2058 	if (best_ia != NULL)
2059 		return best_ia;
2060 
2061 	IFADDR_READER_FOREACH(ifa, ifp) {
2062 		if (ifa->ifa_addr->sa_family != AF_INET6)
2063 			continue;
2064 		ia = (struct in6_ifaddr *)ifa;
2065 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2066 			continue; /* XXX: is there any case to allow anycast? */
2067 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2068 			continue; /* don't use this interface */
2069 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2070 			continue;
2071 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2072 			if (ip6_use_deprecated)
2073 				dep[1] = (struct in6_ifaddr *)ifa;
2074 			continue;
2075 		}
2076 
2077 		best_ia = bestia(best_ia, ia);
2078 	}
2079 	if (best_ia != NULL)
2080 		return best_ia;
2081 
2082 	/* use the last-resort values, that are, deprecated addresses */
2083 	if (dep[0])
2084 		return dep[0];
2085 	if (dep[1])
2086 		return dep[1];
2087 
2088 	return NULL;
2089 }
2090 
2091 /*
2092  * perform DAD when interface becomes IFF_UP.
2093  */
2094 void
2095 in6_if_link_up(struct ifnet *ifp)
2096 {
2097 	struct ifaddr *ifa;
2098 	struct in6_ifaddr *ia;
2099 	int s, bound;
2100 
2101 	/* Ensure it's sane to run DAD */
2102 	if (ifp->if_link_state == LINK_STATE_DOWN)
2103 		return;
2104 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2105 		return;
2106 
2107 	bound = curlwp_bind();
2108 	s = pserialize_read_enter();
2109 	IFADDR_READER_FOREACH(ifa, ifp) {
2110 		struct psref psref;
2111 
2112 		if (ifa->ifa_addr->sa_family != AF_INET6)
2113 			continue;
2114 
2115 		ifa_acquire(ifa, &psref);
2116 		pserialize_read_exit(s);
2117 		ia = (struct in6_ifaddr *)ifa;
2118 
2119 		/* If detached then mark as tentative */
2120 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2121 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2122 			if (if_do_dad(ifp)) {
2123 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2124 				nd6log(LOG_ERR, "%s marked tentative\n",
2125 				    ip6_sprintf(&ia->ia_addr.sin6_addr));
2126 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2127 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2128 		}
2129 
2130 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2131 			int rand_delay;
2132 
2133 			/* Clear the duplicated flag as we're starting DAD. */
2134 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2135 
2136 			/*
2137 			 * The TENTATIVE flag was likely set by hand
2138 			 * beforehand, implicitly indicating the need for DAD.
2139 			 * We may be able to skip the random delay in this
2140 			 * case, but we impose delays just in case.
2141 			 */
2142 			rand_delay = cprng_fast32() %
2143 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2144 			/* +1 ensures callout is always used */
2145 			nd6_dad_start(ifa, rand_delay + 1);
2146 		}
2147 
2148 		s = pserialize_read_enter();
2149 		ifa_release(ifa, &psref);
2150 	}
2151 	pserialize_read_exit(s);
2152 	curlwp_bindx(bound);
2153 
2154 	/* Restore any detached prefixes */
2155 	ND6_WLOCK();
2156 	nd6_pfxlist_onlink_check();
2157 	ND6_UNLOCK();
2158 }
2159 
2160 void
2161 in6_if_up(struct ifnet *ifp)
2162 {
2163 
2164 	/*
2165 	 * special cases, like 6to4, are handled in in6_ifattach
2166 	 */
2167 	in6_ifattach(ifp, NULL);
2168 
2169 	/* interface may not support link state, so bring it up also */
2170 	in6_if_link_up(ifp);
2171 }
2172 
2173 /*
2174  * Mark all addresses as detached.
2175  */
2176 void
2177 in6_if_link_down(struct ifnet *ifp)
2178 {
2179 	struct ifaddr *ifa;
2180 	struct in6_ifaddr *ia;
2181 	int s, bound;
2182 
2183 	/* Any prefixes on this interface should be detached as well */
2184 	ND6_WLOCK();
2185 	nd6_pfxlist_onlink_check();
2186 	ND6_UNLOCK();
2187 
2188 	bound = curlwp_bind();
2189 	s = pserialize_read_enter();
2190 	IFADDR_READER_FOREACH(ifa, ifp) {
2191 		struct psref psref;
2192 
2193 		if (ifa->ifa_addr->sa_family != AF_INET6)
2194 			continue;
2195 
2196 		ifa_acquire(ifa, &psref);
2197 		pserialize_read_exit(s);
2198 		ia = (struct in6_ifaddr *)ifa;
2199 
2200 		/* Stop DAD processing */
2201 		nd6_dad_stop(ifa);
2202 
2203 		/*
2204 		 * Mark the address as detached.
2205 		 * This satisfies RFC4862 Section 5.3, but we should apply
2206 		 * this logic to all addresses to be a good citizen and
2207 		 * avoid potential duplicated addresses.
2208 		 * When the interface comes up again, detached addresses
2209 		 * are marked tentative and DAD commences.
2210 		 */
2211 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2212 			nd6log(LOG_DEBUG, "%s marked detached\n",
2213 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
2214 			ia->ia6_flags |= IN6_IFF_DETACHED;
2215 			ia->ia6_flags &=
2216 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2217 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2218 		}
2219 
2220 		s = pserialize_read_enter();
2221 		ifa_release(ifa, &psref);
2222 	}
2223 	pserialize_read_exit(s);
2224 	curlwp_bindx(bound);
2225 }
2226 
2227 void
2228 in6_if_down(struct ifnet *ifp)
2229 {
2230 
2231 	in6_if_link_down(ifp);
2232 }
2233 
2234 void
2235 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2236 {
2237 
2238 	switch (link_state) {
2239 	case LINK_STATE_DOWN:
2240 		in6_if_link_down(ifp);
2241 		break;
2242 	case LINK_STATE_UP:
2243 		in6_if_link_up(ifp);
2244 		break;
2245 	}
2246 }
2247 
2248 /*
2249  * Calculate max IPv6 MTU through all the interfaces and store it
2250  * to in6_maxmtu.
2251  */
2252 void
2253 in6_setmaxmtu(void)
2254 {
2255 	unsigned long maxmtu = 0;
2256 	struct ifnet *ifp;
2257 	int s;
2258 
2259 	s = pserialize_read_enter();
2260 	IFNET_READER_FOREACH(ifp) {
2261 		/* this function can be called during ifnet initialization */
2262 		if (!ifp->if_afdata[AF_INET6])
2263 			continue;
2264 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2265 		    IN6_LINKMTU(ifp) > maxmtu)
2266 			maxmtu = IN6_LINKMTU(ifp);
2267 	}
2268 	pserialize_read_exit(s);
2269 	if (maxmtu)	     /* update only when maxmtu is positive */
2270 		in6_maxmtu = maxmtu;
2271 }
2272 
2273 /*
2274  * Provide the length of interface identifiers to be used for the link attached
2275  * to the given interface.  The length should be defined in "IPv6 over
2276  * xxx-link" document.  Note that address architecture might also define
2277  * the length for a particular set of address prefixes, regardless of the
2278  * link type.  As clarified in rfc2462bis, those two definitions should be
2279  * consistent, and those really are as of August 2004.
2280  */
2281 int
2282 in6_if2idlen(struct ifnet *ifp)
2283 {
2284 	switch (ifp->if_type) {
2285 	case IFT_ETHER:		/* RFC2464 */
2286 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2287 	case IFT_L2VLAN:	/* ditto */
2288 	case IFT_IEEE80211:	/* ditto */
2289 	case IFT_FDDI:		/* RFC2467 */
2290 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2291 	case IFT_PPP:		/* RFC2472 */
2292 	case IFT_ARCNET:	/* RFC2497 */
2293 	case IFT_FRELAY:	/* RFC2590 */
2294 	case IFT_IEEE1394:	/* RFC3146 */
2295 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2296 	case IFT_LOOP:		/* XXX: is this really correct? */
2297 		return 64;
2298 	default:
2299 		/*
2300 		 * Unknown link type:
2301 		 * It might be controversial to use the today's common constant
2302 		 * of 64 for these cases unconditionally.  For full compliance,
2303 		 * we should return an error in this case.  On the other hand,
2304 		 * if we simply miss the standard for the link type or a new
2305 		 * standard is defined for a new link type, the IFID length
2306 		 * is very likely to be the common constant.  As a compromise,
2307 		 * we always use the constant, but make an explicit notice
2308 		 * indicating the "unknown" case.
2309 		 */
2310 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2311 		return 64;
2312 	}
2313 }
2314 
2315 struct in6_llentry {
2316 	struct llentry		base;
2317 };
2318 
2319 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2320 #define	IN6_LLTBL_HASH(k, h) \
2321 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2322 
2323 /*
2324  * Do actual deallocation of @lle.
2325  * Called by LLE_FREE_LOCKED when number of references
2326  * drops to zero.
2327  */
2328 static void
2329 in6_lltable_destroy_lle(struct llentry *lle)
2330 {
2331 
2332 	LLE_WUNLOCK(lle);
2333 	LLE_LOCK_DESTROY(lle);
2334 	kmem_intr_free(lle, sizeof(struct in6_llentry));
2335 }
2336 
2337 static struct llentry *
2338 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2339 {
2340 	struct in6_llentry *lle;
2341 
2342 	lle = kmem_intr_zalloc(sizeof(struct in6_llentry), KM_NOSLEEP);
2343 	if (lle == NULL)		/* NB: caller generates msg */
2344 		return NULL;
2345 
2346 	lle->base.r_l3addr.addr6 = *addr6;
2347 	lle->base.lle_refcnt = 1;
2348 	lle->base.lle_free = in6_lltable_destroy_lle;
2349 	LLE_LOCK_INIT(&lle->base);
2350 	callout_init(&lle->base.lle_timer, CALLOUT_MPSAFE);
2351 
2352 	return &lle->base;
2353 }
2354 
2355 static int
2356 in6_lltable_match_prefix(const struct sockaddr *prefix,
2357     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2358 {
2359 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2360 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2361 
2362 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2363 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2364 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2365 		return 1;
2366 
2367 	return 0;
2368 }
2369 
2370 static void
2371 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2372 {
2373 	struct ifnet *ifp = llt->llt_ifp;
2374 
2375 	IF_AFDATA_WLOCK_ASSERT(ifp);
2376 	LLE_WLOCK_ASSERT(lle);
2377 
2378 	/* Unlink entry from table */
2379 	if ((lle->la_flags & LLE_LINKED) != 0) {
2380 
2381 		lltable_unlink_entry(llt, lle);
2382 		KASSERT((lle->la_flags & LLE_LINKED) == 0);
2383 	}
2384 	/*
2385 	 * We need to release the lock here to lle_timer proceeds;
2386 	 * lle_timer should stop immediately if LLE_LINKED isn't set.
2387 	 * Note that we cannot pass lle->lle_lock to callout_halt
2388 	 * because it's a rwlock.
2389 	 */
2390 	LLE_ADDREF(lle);
2391 	LLE_WUNLOCK(lle);
2392 	IF_AFDATA_WUNLOCK(ifp);
2393 
2394 #ifdef NET_MPSAFE
2395 	callout_halt(&lle->lle_timer, NULL);
2396 #else
2397 	if (mutex_owned(softnet_lock))
2398 		callout_halt(&lle->lle_timer, softnet_lock);
2399 	else
2400 		callout_halt(&lle->lle_timer, NULL);
2401 #endif
2402 	LLE_WLOCK(lle);
2403 	LLE_REMREF(lle);
2404 
2405 	lltable_drop_entry_queue(lle);
2406 	LLE_FREE_LOCKED(lle);
2407 
2408 	IF_AFDATA_WLOCK(ifp);
2409 }
2410 
2411 static int
2412 in6_lltable_rtcheck(struct ifnet *ifp,
2413 		    u_int flags,
2414 		    const struct sockaddr *l3addr)
2415 {
2416 	struct rtentry *rt;
2417 
2418 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2419 	    "sin_family %d", l3addr->sa_family);
2420 
2421 	rt = rtalloc1(l3addr, 0);
2422 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2423 		int s;
2424 		struct ifaddr *ifa;
2425 		/*
2426 		 * Create an ND6 cache for an IPv6 neighbor
2427 		 * that is not covered by our own prefix.
2428 		 */
2429 		/* XXX ifaof_ifpforaddr should take a const param */
2430 		s = pserialize_read_enter();
2431 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2432 		if (ifa != NULL) {
2433 			pserialize_read_exit(s);
2434 			if (rt != NULL)
2435 				rt_unref(rt);
2436 			return 0;
2437 		}
2438 		pserialize_read_exit(s);
2439 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2440 		    ip6_sprintf(&((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2441 		if (rt != NULL)
2442 			rt_unref(rt);
2443 		return EINVAL;
2444 	}
2445 	rt_unref(rt);
2446 	return 0;
2447 }
2448 
2449 static inline uint32_t
2450 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2451 {
2452 
2453 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2454 }
2455 
2456 static uint32_t
2457 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2458 {
2459 
2460 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2461 }
2462 
2463 static void
2464 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2465 {
2466 	struct sockaddr_in6 *sin6;
2467 
2468 	sin6 = (struct sockaddr_in6 *)sa;
2469 	bzero(sin6, sizeof(*sin6));
2470 	sin6->sin6_family = AF_INET6;
2471 	sin6->sin6_len = sizeof(*sin6);
2472 	sin6->sin6_addr = lle->r_l3addr.addr6;
2473 }
2474 
2475 static inline struct llentry *
2476 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2477 {
2478 	struct llentry *lle;
2479 	struct llentries *lleh;
2480 	u_int hashidx;
2481 
2482 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2483 	lleh = &llt->lle_head[hashidx];
2484 	LIST_FOREACH(lle, lleh, lle_next) {
2485 		if (lle->la_flags & LLE_DELETED)
2486 			continue;
2487 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2488 			break;
2489 	}
2490 
2491 	return lle;
2492 }
2493 
2494 static int
2495 in6_lltable_delete(struct lltable *llt, u_int flags,
2496 	const struct sockaddr *l3addr)
2497 {
2498 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2499 	struct llentry *lle;
2500 
2501 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2502 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2503 	    "sin_family %d", l3addr->sa_family);
2504 
2505 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2506 
2507 	if (lle == NULL)
2508 		return ENOENT;
2509 
2510 	LLE_WLOCK(lle);
2511 	lle->la_flags |= LLE_DELETED;
2512 #ifdef DIAGNOSTIC
2513 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2514 #endif
2515 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2516 		llentry_free(lle);
2517 	else
2518 		LLE_WUNLOCK(lle);
2519 
2520 	return 0;
2521 }
2522 
2523 static struct llentry *
2524 in6_lltable_create(struct lltable *llt, u_int flags,
2525 	const struct sockaddr *l3addr)
2526 {
2527 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2528 	struct ifnet *ifp = llt->llt_ifp;
2529 	struct llentry *lle;
2530 
2531 	IF_AFDATA_WLOCK_ASSERT(ifp);
2532 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2533 	    "sin_family %d", l3addr->sa_family);
2534 
2535 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2536 
2537 	if (lle != NULL) {
2538 		LLE_WLOCK(lle);
2539 		return lle;
2540 	}
2541 
2542 	/*
2543 	 * A route that covers the given address must have
2544 	 * been installed 1st because we are doing a resolution,
2545 	 * verify this.
2546 	 */
2547 	if (!(flags & LLE_IFADDR) &&
2548 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2549 		return NULL;
2550 
2551 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2552 	if (lle == NULL) {
2553 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2554 		return NULL;
2555 	}
2556 	lle->la_flags = flags;
2557 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2558 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2559 		lle->la_flags |= LLE_VALID;
2560 	}
2561 
2562 	lltable_link_entry(llt, lle);
2563 	LLE_WLOCK(lle);
2564 
2565 	return lle;
2566 }
2567 
2568 static struct llentry *
2569 in6_lltable_lookup(struct lltable *llt, u_int flags,
2570 	const struct sockaddr *l3addr)
2571 {
2572 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2573 	struct llentry *lle;
2574 
2575 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2576 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2577 	    "sin_family %d", l3addr->sa_family);
2578 
2579 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2580 
2581 	if (lle == NULL)
2582 		return NULL;
2583 
2584 	if (flags & LLE_EXCLUSIVE)
2585 		LLE_WLOCK(lle);
2586 	else
2587 		LLE_RLOCK(lle);
2588 	return lle;
2589 }
2590 
2591 static int
2592 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2593     struct rt_walkarg *w)
2594 {
2595 	struct sockaddr_in6 sin6;
2596 
2597 	LLTABLE_LOCK_ASSERT();
2598 
2599 	/* skip deleted entries */
2600 	if (lle->la_flags & LLE_DELETED)
2601 		return 0;
2602 
2603 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2604 
2605 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2606 }
2607 
2608 static struct lltable *
2609 in6_lltattach(struct ifnet *ifp)
2610 {
2611 	struct lltable *llt;
2612 
2613 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2614 	llt->llt_af = AF_INET6;
2615 	llt->llt_ifp = ifp;
2616 
2617 	llt->llt_lookup = in6_lltable_lookup;
2618 	llt->llt_create = in6_lltable_create;
2619 	llt->llt_delete = in6_lltable_delete;
2620 	llt->llt_dump_entry = in6_lltable_dump_entry;
2621 	llt->llt_hash = in6_lltable_hash;
2622 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2623 	llt->llt_free_entry = in6_lltable_free_entry;
2624 	llt->llt_match_prefix = in6_lltable_match_prefix;
2625 	lltable_link(llt);
2626 
2627 	return llt;
2628 }
2629 
2630 void *
2631 in6_domifattach(struct ifnet *ifp)
2632 {
2633 	struct in6_ifextra *ext;
2634 
2635 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2636 
2637 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2638 	    M_IFADDR, M_WAITOK|M_ZERO);
2639 
2640 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2641 	    M_IFADDR, M_WAITOK|M_ZERO);
2642 
2643 	ext->nd_ifinfo = nd6_ifattach(ifp);
2644 	ext->scope6_id = scope6_ifattach(ifp);
2645 	ext->nprefixes = 0;
2646 	ext->ndefrouters = 0;
2647 
2648 	ext->lltable = in6_lltattach(ifp);
2649 
2650 	return ext;
2651 }
2652 
2653 void
2654 in6_domifdetach(struct ifnet *ifp, void *aux)
2655 {
2656 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2657 
2658 	lltable_free(ext->lltable);
2659 	ext->lltable = NULL;
2660 	nd6_ifdetach(ifp, ext);
2661 	free(ext->in6_ifstat, M_IFADDR);
2662 	free(ext->icmp6_ifstat, M_IFADDR);
2663 	scope6_ifdetach(ext->scope6_id);
2664 	free(ext, M_IFADDR);
2665 }
2666 
2667 /*
2668  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2669  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2670  */
2671 void
2672 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2673 {
2674 	in6->s6_addr32[0] = 0;
2675 	in6->s6_addr32[1] = 0;
2676 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2677 	in6->s6_addr32[3] = in->s_addr;
2678 }
2679 
2680 /*
2681  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2682  * v4 mapped addr or v4 compat addr
2683  */
2684 void
2685 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2686 {
2687 	memset(sin, 0, sizeof(*sin));
2688 	sin->sin_len = sizeof(struct sockaddr_in);
2689 	sin->sin_family = AF_INET;
2690 	sin->sin_port = sin6->sin6_port;
2691 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2692 }
2693 
2694 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2695 void
2696 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2697 {
2698 	memset(sin6, 0, sizeof(*sin6));
2699 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2700 	sin6->sin6_family = AF_INET6;
2701 	sin6->sin6_port = sin->sin_port;
2702 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2703 }
2704 
2705 /* Convert sockaddr_in6 into sockaddr_in. */
2706 void
2707 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2708 {
2709 	struct sockaddr_in *sin_p;
2710 	struct sockaddr_in6 sin6;
2711 
2712 	/*
2713 	 * Save original sockaddr_in6 addr and convert it
2714 	 * to sockaddr_in.
2715 	 */
2716 	sin6 = *(struct sockaddr_in6 *)nam;
2717 	sin_p = (struct sockaddr_in *)nam;
2718 	in6_sin6_2_sin(sin_p, &sin6);
2719 }
2720 
2721 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2722 void
2723 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2724 {
2725 	struct sockaddr_in *sin_p;
2726 	struct sockaddr_in6 *sin6_p;
2727 
2728 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2729 	sin_p = (struct sockaddr_in *)*nam;
2730 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2731 	free(*nam, M_SONAME);
2732 	*nam = sin6tosa(sin6_p);
2733 }
2734