xref: /freebsd/sys/netinet6/in6.c (revision bcce9a2b33a8e9187a63f435726a7a801e89f326)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1991, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)in.c	8.2 (Berkeley) 11/15/93
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 
71 #include <sys/param.h>
72 #include <sys/eventhandler.h>
73 #include <sys/errno.h>
74 #include <sys/jail.h>
75 #include <sys/malloc.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sockio.h>
79 #include <sys/systm.h>
80 #include <sys/priv.h>
81 #include <sys/proc.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/lock.h>
85 #include <sys/rmlock.h>
86 #include <sys/syslog.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/if_dl.h>
93 #include <net/vnet.h>
94 
95 #include <netinet/in.h>
96 #include <netinet/in_var.h>
97 #include <net/if_llatbl.h>
98 #include <netinet/if_ether.h>
99 #include <netinet/in_systm.h>
100 #include <netinet/ip.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/ip_carp.h>
103 
104 #include <netinet/ip6.h>
105 #include <netinet6/ip6_var.h>
106 #include <netinet6/nd6.h>
107 #include <netinet6/mld6_var.h>
108 #include <netinet6/ip6_mroute.h>
109 #include <netinet6/in6_ifattach.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/in6_fib.h>
112 #include <netinet6/in6_pcb.h>
113 
114 /*
115  * struct in6_ifreq and struct ifreq must be type punnable for common members
116  * of ifr_ifru to allow accessors to be shared.
117  */
118 _Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
119     offsetof(struct ifreq, ifr_ifru),
120     "struct in6_ifreq and struct ifreq are not type punnable");
121 
122 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
123 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
124 
125 /*
126  * Definitions of some costant 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 const struct in6_addr in6addr_linklocal_allv2routers =
137 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
138 
139 const struct in6_addr in6mask0 = IN6MASK0;
140 const struct in6_addr in6mask32 = IN6MASK32;
141 const struct in6_addr in6mask64 = IN6MASK64;
142 const struct in6_addr in6mask96 = IN6MASK96;
143 const struct in6_addr in6mask128 = IN6MASK128;
144 
145 const struct sockaddr_in6 sa6_any =
146 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
147 
148 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
149 	struct in6_aliasreq *, int);
150 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
151 
152 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
153     struct in6_ifaddr *, int);
154 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
155     struct in6_aliasreq *, int flags);
156 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
157     struct in6_ifaddr *, int, int);
158 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
159     struct in6_ifaddr *, int);
160 
161 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
162 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
163 
164 
165 void
166 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
167 {
168 	struct sockaddr_dl gateway;
169 	struct sockaddr_in6 mask, addr;
170 	struct rtentry rt;
171 	int fibnum;
172 
173 	/*
174 	 * initialize for rtmsg generation
175 	 */
176 	bzero(&gateway, sizeof(gateway));
177 	gateway.sdl_len = sizeof(gateway);
178 	gateway.sdl_family = AF_LINK;
179 
180 	bzero(&rt, sizeof(rt));
181 	rt.rt_gateway = (struct sockaddr *)&gateway;
182 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
183 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
184 	rt_mask(&rt) = (struct sockaddr *)&mask;
185 	rt_key(&rt) = (struct sockaddr *)&addr;
186 	rt.rt_flags = RTF_HOST | RTF_STATIC;
187 	if (cmd == RTM_ADD)
188 		rt.rt_flags |= RTF_UP;
189 	fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ia62ifa(ia)->ifa_ifp->if_fib;
190 	/* Announce arrival of local address to this FIB. */
191 	rt_newaddrmsg_fib(cmd, &ia->ia_ifa, 0, &rt, fibnum);
192 }
193 
194 int
195 in6_mask2len(struct in6_addr *mask, u_char *lim0)
196 {
197 	int x = 0, y;
198 	u_char *lim = lim0, *p;
199 
200 	/* ignore the scope_id part */
201 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
202 		lim = (u_char *)mask + sizeof(*mask);
203 	for (p = (u_char *)mask; p < lim; x++, p++) {
204 		if (*p != 0xff)
205 			break;
206 	}
207 	y = 0;
208 	if (p < lim) {
209 		for (y = 0; y < 8; y++) {
210 			if ((*p & (0x80 >> y)) == 0)
211 				break;
212 		}
213 	}
214 
215 	/*
216 	 * when the limit pointer is given, do a stricter check on the
217 	 * remaining bits.
218 	 */
219 	if (p < lim) {
220 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
221 			return (-1);
222 		for (p = p + 1; p < lim; p++)
223 			if (*p != 0)
224 				return (-1);
225 	}
226 
227 	return x * 8 + y;
228 }
229 
230 #ifdef COMPAT_FREEBSD32
231 struct in6_ndifreq32 {
232 	char ifname[IFNAMSIZ];
233 	uint32_t ifindex;
234 };
235 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
236 #endif
237 
238 int
239 in6_control(struct socket *so, u_long cmd, caddr_t data,
240     struct ifnet *ifp, struct thread *td)
241 {
242 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
243 	struct	in6_ifaddr *ia = NULL;
244 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
245 	struct sockaddr_in6 *sa6;
246 	int carp_attached = 0;
247 	int error;
248 	u_long ocmd = cmd;
249 
250 	/*
251 	 * Compat to make pre-10.x ifconfig(8) operable.
252 	 */
253 	if (cmd == OSIOCAIFADDR_IN6)
254 		cmd = SIOCAIFADDR_IN6;
255 
256 	switch (cmd) {
257 	case SIOCGETSGCNT_IN6:
258 	case SIOCGETMIFCNT_IN6:
259 		/*
260 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
261 		 * We cannot see how that would be needed, so do not adjust the
262 		 * KPI blindly; more likely should clean up the IPv4 variant.
263 		 */
264 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
265 	}
266 
267 	switch (cmd) {
268 	case SIOCAADDRCTL_POLICY:
269 	case SIOCDADDRCTL_POLICY:
270 		if (td != NULL) {
271 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
272 			if (error)
273 				return (error);
274 		}
275 		return (in6_src_ioctl(cmd, data));
276 	}
277 
278 	if (ifp == NULL)
279 		return (EOPNOTSUPP);
280 
281 	switch (cmd) {
282 	case SIOCSNDFLUSH_IN6:
283 	case SIOCSPFXFLUSH_IN6:
284 	case SIOCSRTRFLUSH_IN6:
285 	case SIOCSDEFIFACE_IN6:
286 	case SIOCSIFINFO_FLAGS:
287 	case SIOCSIFINFO_IN6:
288 		if (td != NULL) {
289 			error = priv_check(td, PRIV_NETINET_ND6);
290 			if (error)
291 				return (error);
292 		}
293 		/* FALLTHROUGH */
294 	case OSIOCGIFINFO_IN6:
295 	case SIOCGIFINFO_IN6:
296 	case SIOCGNBRINFO_IN6:
297 	case SIOCGDEFIFACE_IN6:
298 		return (nd6_ioctl(cmd, data, ifp));
299 
300 #ifdef COMPAT_FREEBSD32
301 	case SIOCGDEFIFACE32_IN6:
302 		{
303 			struct in6_ndifreq ndif;
304 			struct in6_ndifreq32 *ndif32;
305 
306 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
307 			    ifp);
308 			if (error)
309 				return (error);
310 			ndif32 = (struct in6_ndifreq32 *)data;
311 			ndif32->ifindex = ndif.ifindex;
312 			return (0);
313 		}
314 #endif
315 	}
316 
317 	switch (cmd) {
318 	case SIOCSIFPREFIX_IN6:
319 	case SIOCDIFPREFIX_IN6:
320 	case SIOCAIFPREFIX_IN6:
321 	case SIOCCIFPREFIX_IN6:
322 	case SIOCSGIFPREFIX_IN6:
323 	case SIOCGIFPREFIX_IN6:
324 		log(LOG_NOTICE,
325 		    "prefix ioctls are now invalidated. "
326 		    "please use ifconfig.\n");
327 		return (EOPNOTSUPP);
328 	}
329 
330 	switch (cmd) {
331 	case SIOCSSCOPE6:
332 		if (td != NULL) {
333 			error = priv_check(td, PRIV_NETINET_SCOPE6);
334 			if (error)
335 				return (error);
336 		}
337 		/* FALLTHROUGH */
338 	case SIOCGSCOPE6:
339 	case SIOCGSCOPE6DEF:
340 		return (scope6_ioctl(cmd, data, ifp));
341 	}
342 
343 	/*
344 	 * Find address for this interface, if it exists.
345 	 *
346 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
347 	 * only, and used the first interface address as the target of other
348 	 * operations (without checking ifra_addr).  This was because netinet
349 	 * code/API assumed at most 1 interface address per interface.
350 	 * Since IPv6 allows a node to assign multiple addresses
351 	 * on a single interface, we almost always look and check the
352 	 * presence of ifra_addr, and reject invalid ones here.
353 	 * It also decreases duplicated code among SIOC*_IN6 operations.
354 	 */
355 	switch (cmd) {
356 	case SIOCAIFADDR_IN6:
357 	case SIOCSIFPHYADDR_IN6:
358 		sa6 = &ifra->ifra_addr;
359 		break;
360 	case SIOCSIFADDR_IN6:
361 	case SIOCGIFADDR_IN6:
362 	case SIOCSIFDSTADDR_IN6:
363 	case SIOCSIFNETMASK_IN6:
364 	case SIOCGIFDSTADDR_IN6:
365 	case SIOCGIFNETMASK_IN6:
366 	case SIOCDIFADDR_IN6:
367 	case SIOCGIFPSRCADDR_IN6:
368 	case SIOCGIFPDSTADDR_IN6:
369 	case SIOCGIFAFLAG_IN6:
370 	case SIOCSNDFLUSH_IN6:
371 	case SIOCSPFXFLUSH_IN6:
372 	case SIOCSRTRFLUSH_IN6:
373 	case SIOCGIFALIFETIME_IN6:
374 	case SIOCGIFSTAT_IN6:
375 	case SIOCGIFSTAT_ICMP6:
376 		sa6 = &ifr->ifr_addr;
377 		break;
378 	case SIOCSIFADDR:
379 	case SIOCSIFBRDADDR:
380 	case SIOCSIFDSTADDR:
381 	case SIOCSIFNETMASK:
382 		/*
383 		 * Although we should pass any non-INET6 ioctl requests
384 		 * down to driver, we filter some legacy INET requests.
385 		 * Drivers trust SIOCSIFADDR et al to come from an already
386 		 * privileged layer, and do not perform any credentials
387 		 * checks or input validation.
388 		 */
389 		return (EINVAL);
390 	default:
391 		sa6 = NULL;
392 		break;
393 	}
394 	if (sa6 && sa6->sin6_family == AF_INET6) {
395 		if (sa6->sin6_scope_id != 0)
396 			error = sa6_embedscope(sa6, 0);
397 		else
398 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
399 		if (error != 0)
400 			return (error);
401 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
402 		    &sa6->sin6_addr)) != 0)
403 			return (error);
404 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
405 	} else
406 		ia = NULL;
407 
408 	switch (cmd) {
409 	case SIOCSIFADDR_IN6:
410 	case SIOCSIFDSTADDR_IN6:
411 	case SIOCSIFNETMASK_IN6:
412 		/*
413 		 * Since IPv6 allows a node to assign multiple addresses
414 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
415 		 */
416 		/* we decided to obsolete this command (20000704) */
417 		error = EINVAL;
418 		goto out;
419 
420 	case SIOCDIFADDR_IN6:
421 		/*
422 		 * for IPv4, we look for existing in_ifaddr here to allow
423 		 * "ifconfig if0 delete" to remove the first IPv4 address on
424 		 * the interface.  For IPv6, as the spec allows multiple
425 		 * interface address from the day one, we consider "remove the
426 		 * first one" semantics to be not preferable.
427 		 */
428 		if (ia == NULL) {
429 			error = EADDRNOTAVAIL;
430 			goto out;
431 		}
432 		/* FALLTHROUGH */
433 	case SIOCAIFADDR_IN6:
434 		/*
435 		 * We always require users to specify a valid IPv6 address for
436 		 * the corresponding operation.
437 		 */
438 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
439 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
440 			error = EAFNOSUPPORT;
441 			goto out;
442 		}
443 
444 		if (td != NULL) {
445 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
446 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
447 			if (error)
448 				goto out;
449 		}
450 		/* FALLTHROUGH */
451 	case SIOCGIFSTAT_IN6:
452 	case SIOCGIFSTAT_ICMP6:
453 		if (ifp->if_afdata[AF_INET6] == NULL) {
454 			error = EPFNOSUPPORT;
455 			goto out;
456 		}
457 		break;
458 
459 	case SIOCGIFADDR_IN6:
460 		/* This interface is basically deprecated. use SIOCGIFCONF. */
461 		/* FALLTHROUGH */
462 	case SIOCGIFAFLAG_IN6:
463 	case SIOCGIFNETMASK_IN6:
464 	case SIOCGIFDSTADDR_IN6:
465 	case SIOCGIFALIFETIME_IN6:
466 		/* must think again about its semantics */
467 		if (ia == NULL) {
468 			error = EADDRNOTAVAIL;
469 			goto out;
470 		}
471 		break;
472 	}
473 
474 	switch (cmd) {
475 	case SIOCGIFADDR_IN6:
476 		ifr->ifr_addr = ia->ia_addr;
477 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
478 			goto out;
479 		break;
480 
481 	case SIOCGIFDSTADDR_IN6:
482 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
483 			error = EINVAL;
484 			goto out;
485 		}
486 		ifr->ifr_dstaddr = ia->ia_dstaddr;
487 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
488 			goto out;
489 		break;
490 
491 	case SIOCGIFNETMASK_IN6:
492 		ifr->ifr_addr = ia->ia_prefixmask;
493 		break;
494 
495 	case SIOCGIFAFLAG_IN6:
496 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
497 		break;
498 
499 	case SIOCGIFSTAT_IN6:
500 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
501 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
502 		    &ifr->ifr_ifru.ifru_stat,
503 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
504 		break;
505 
506 	case SIOCGIFSTAT_ICMP6:
507 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
508 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
509 		    &ifr->ifr_ifru.ifru_icmp6stat,
510 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
511 		break;
512 
513 	case SIOCGIFALIFETIME_IN6:
514 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
515 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
516 			time_t maxexpire;
517 			struct in6_addrlifetime *retlt =
518 			    &ifr->ifr_ifru.ifru_lifetime;
519 
520 			/*
521 			 * XXX: adjust expiration time assuming time_t is
522 			 * signed.
523 			 */
524 			maxexpire = (-1) &
525 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
526 			if (ia->ia6_lifetime.ia6t_vltime <
527 			    maxexpire - ia->ia6_updatetime) {
528 				retlt->ia6t_expire = ia->ia6_updatetime +
529 				    ia->ia6_lifetime.ia6t_vltime;
530 			} else
531 				retlt->ia6t_expire = maxexpire;
532 		}
533 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
534 			time_t maxexpire;
535 			struct in6_addrlifetime *retlt =
536 			    &ifr->ifr_ifru.ifru_lifetime;
537 
538 			/*
539 			 * XXX: adjust expiration time assuming time_t is
540 			 * signed.
541 			 */
542 			maxexpire = (-1) &
543 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
544 			if (ia->ia6_lifetime.ia6t_pltime <
545 			    maxexpire - ia->ia6_updatetime) {
546 				retlt->ia6t_preferred = ia->ia6_updatetime +
547 				    ia->ia6_lifetime.ia6t_pltime;
548 			} else
549 				retlt->ia6t_preferred = maxexpire;
550 		}
551 		break;
552 
553 	case SIOCAIFADDR_IN6:
554 	{
555 		struct nd_prefixctl pr0;
556 		struct nd_prefix *pr;
557 
558 		/*
559 		 * first, make or update the interface address structure,
560 		 * and link it to the list.
561 		 */
562 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
563 			goto out;
564 		if (ia != NULL) {
565 			if (ia->ia_ifa.ifa_carp)
566 				(*carp_detach_p)(&ia->ia_ifa, true);
567 			ifa_free(&ia->ia_ifa);
568 		}
569 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
570 		    == NULL) {
571 			/*
572 			 * this can happen when the user specify the 0 valid
573 			 * lifetime.
574 			 */
575 			break;
576 		}
577 
578 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
579 			if (carp_attach_p != NULL)
580 				error = (*carp_attach_p)(&ia->ia_ifa,
581 				    ifra->ifra_vhid);
582 			else
583 				error = EPROTONOSUPPORT;
584 			if (error)
585 				goto out;
586 			else
587 				carp_attached = 1;
588 		}
589 
590 		/*
591 		 * then, make the prefix on-link on the interface.
592 		 * XXX: we'd rather create the prefix before the address, but
593 		 * we need at least one address to install the corresponding
594 		 * interface route, so we configure the address first.
595 		 */
596 
597 		/*
598 		 * convert mask to prefix length (prefixmask has already
599 		 * been validated in in6_update_ifa().
600 		 */
601 		bzero(&pr0, sizeof(pr0));
602 		pr0.ndpr_ifp = ifp;
603 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
604 		    NULL);
605 		if (pr0.ndpr_plen == 128) {
606 			/* we don't need to install a host route. */
607 			goto aifaddr_out;
608 		}
609 		pr0.ndpr_prefix = ifra->ifra_addr;
610 		/* apply the mask for safety. */
611 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
612 		    &ifra->ifra_prefixmask.sin6_addr);
613 
614 		/*
615 		 * XXX: since we don't have an API to set prefix (not address)
616 		 * lifetimes, we just use the same lifetimes as addresses.
617 		 * The (temporarily) installed lifetimes can be overridden by
618 		 * later advertised RAs (when accept_rtadv is non 0), which is
619 		 * an intended behavior.
620 		 */
621 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
622 		pr0.ndpr_raf_auto =
623 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
624 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
625 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
626 
627 		/* add the prefix if not yet. */
628 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
629 			/*
630 			 * nd6_prelist_add will install the corresponding
631 			 * interface route.
632 			 */
633 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
634 				if (carp_attached)
635 					(*carp_detach_p)(&ia->ia_ifa, false);
636 				goto out;
637 			}
638 		}
639 
640 		/* relate the address to the prefix */
641 		if (ia->ia6_ndpr == NULL) {
642 			ia->ia6_ndpr = pr;
643 			pr->ndpr_addrcnt++;
644 
645 			/*
646 			 * If this is the first autoconf address from the
647 			 * prefix, create a temporary address as well
648 			 * (when required).
649 			 */
650 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
651 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
652 				int e;
653 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
654 					log(LOG_NOTICE, "in6_control: failed "
655 					    "to create a temporary address, "
656 					    "errno=%d\n", e);
657 				}
658 			}
659 		}
660 		nd6_prefix_rele(pr);
661 
662 		/*
663 		 * this might affect the status of autoconfigured addresses,
664 		 * that is, this address might make other addresses detached.
665 		 */
666 		pfxlist_onlink_check();
667 
668 aifaddr_out:
669 		/*
670 		 * Try to clear the flag when a new IPv6 address is added
671 		 * onto an IFDISABLED interface and it succeeds.
672 		 */
673 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
674 			struct in6_ndireq nd;
675 
676 			memset(&nd, 0, sizeof(nd));
677 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
678 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
679 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
680 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
681 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
682 				    "failed.");
683 			/*
684 			 * Ignore failure of clearing the flag intentionally.
685 			 * The failure means address duplication was detected.
686 			 */
687 		}
688 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
689 		break;
690 	}
691 
692 	case SIOCDIFADDR_IN6:
693 	{
694 		struct nd_prefix *pr;
695 
696 		/*
697 		 * If the address being deleted is the only one that owns
698 		 * the corresponding prefix, expire the prefix as well.
699 		 * XXX: theoretically, we don't have to worry about such
700 		 * relationship, since we separate the address management
701 		 * and the prefix management.  We do this, however, to provide
702 		 * as much backward compatibility as possible in terms of
703 		 * the ioctl operation.
704 		 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
705 		 */
706 		pr = ia->ia6_ndpr;
707 		in6_purgeaddr(&ia->ia_ifa);
708 		if (pr != NULL && pr->ndpr_addrcnt == 0) {
709 			ND6_WLOCK();
710 			nd6_prefix_unlink(pr, NULL);
711 			ND6_WUNLOCK();
712 			nd6_prefix_del(pr);
713 		}
714 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
715 		break;
716 	}
717 
718 	default:
719 		if (ifp->if_ioctl == NULL) {
720 			error = EOPNOTSUPP;
721 			goto out;
722 		}
723 		error = (*ifp->if_ioctl)(ifp, cmd, data);
724 		goto out;
725 	}
726 
727 	error = 0;
728 out:
729 	if (ia != NULL)
730 		ifa_free(&ia->ia_ifa);
731 	return (error);
732 }
733 
734 
735 /*
736  * Join necessary multicast groups.  Factored out from in6_update_ifa().
737  * This entire work should only be done once, for the default FIB.
738  */
739 static int
740 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
741     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
742 {
743 	char ip6buf[INET6_ADDRSTRLEN];
744 	struct in6_addr mltaddr;
745 	struct in6_multi_mship *imm;
746 	int delay, error;
747 
748 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
749 
750 	/* Join solicited multicast addr for new host id. */
751 	bzero(&mltaddr, sizeof(struct in6_addr));
752 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
753 	mltaddr.s6_addr32[2] = htonl(1);
754 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
755 	mltaddr.s6_addr8[12] = 0xff;
756 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
757 		/* XXX: should not happen */
758 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
759 		goto cleanup;
760 	}
761 	delay = error = 0;
762 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
763 		/*
764 		 * We need a random delay for DAD on the address being
765 		 * configured.  It also means delaying transmission of the
766 		 * corresponding MLD report to avoid report collision.
767 		 * [RFC 4861, Section 6.3.7]
768 		 */
769 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
770 	}
771 	imm = in6_joingroup(ifp, &mltaddr, &error, delay);
772 	if (imm == NULL) {
773 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
774 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
775 		    if_name(ifp), error));
776 		goto cleanup;
777 	}
778 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
779 	*in6m_sol = imm->i6mm_maddr;
780 
781 	/*
782 	 * Join link-local all-nodes address.
783 	 */
784 	mltaddr = in6addr_linklocal_allnodes;
785 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
786 		goto cleanup; /* XXX: should not fail */
787 
788 	imm = in6_joingroup(ifp, &mltaddr, &error, 0);
789 	if (imm == NULL) {
790 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
791 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
792 		    if_name(ifp), error));
793 		goto cleanup;
794 	}
795 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
796 
797 	/*
798 	 * Join node information group address.
799 	 */
800 	delay = 0;
801 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
802 		/*
803 		 * The spec does not say anything about delay for this group,
804 		 * but the same logic should apply.
805 		 */
806 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
807 	}
808 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
809 		/* XXX jinmei */
810 		imm = in6_joingroup(ifp, &mltaddr, &error, delay);
811 		if (imm == NULL)
812 			nd6log((LOG_WARNING,
813 			    "%s: in6_joingroup failed for %s on %s "
814 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
815 			    &mltaddr), if_name(ifp), error));
816 			/* XXX not very fatal, go on... */
817 		else
818 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
819 	}
820 	if (V_icmp6_nodeinfo_oldmcprefix &&
821 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
822 		imm = in6_joingroup(ifp, &mltaddr, &error, delay);
823 		if (imm == NULL)
824 			nd6log((LOG_WARNING,
825 			    "%s: in6_joingroup failed for %s on %s "
826 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
827 			    &mltaddr), if_name(ifp), error));
828 			/* XXX not very fatal, go on... */
829 		else
830 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
831 	}
832 
833 	/*
834 	 * Join interface-local all-nodes address.
835 	 * (ff01::1%ifN, and ff01::%ifN/32)
836 	 */
837 	mltaddr = in6addr_nodelocal_allnodes;
838 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
839 		goto cleanup; /* XXX: should not fail */
840 
841 	imm = in6_joingroup(ifp, &mltaddr, &error, 0);
842 	if (imm == NULL) {
843 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
844 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
845 		    &mltaddr), if_name(ifp), error));
846 		goto cleanup;
847 	}
848 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
849 
850 cleanup:
851 	return (error);
852 }
853 
854 /*
855  * Update parameters of an IPv6 interface address.
856  * If necessary, a new entry is created and linked into address chains.
857  * This function is separated from in6_control().
858  */
859 int
860 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
861     struct in6_ifaddr *ia, int flags)
862 {
863 	int error, hostIsNew = 0;
864 
865 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
866 		return (error);
867 
868 	if (ia == NULL) {
869 		hostIsNew = 1;
870 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
871 			return (ENOBUFS);
872 	}
873 
874 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
875 	if (error != 0) {
876 		if (hostIsNew != 0) {
877 			in6_unlink_ifa(ia, ifp);
878 			ifa_free(&ia->ia_ifa);
879 		}
880 		return (error);
881 	}
882 
883 	if (hostIsNew)
884 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
885 
886 	return (error);
887 }
888 
889 /*
890  * Fill in basic IPv6 address request info.
891  */
892 void
893 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
894     const struct in6_addr *mask)
895 {
896 
897 	memset(ifra, 0, sizeof(struct in6_aliasreq));
898 
899 	ifra->ifra_addr.sin6_family = AF_INET6;
900 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
901 	if (addr != NULL)
902 		ifra->ifra_addr.sin6_addr = *addr;
903 
904 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
905 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
906 	if (mask != NULL)
907 		ifra->ifra_prefixmask.sin6_addr = *mask;
908 }
909 
910 static int
911 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
912     struct in6_ifaddr *ia, int flags)
913 {
914 	int plen = -1;
915 	struct sockaddr_in6 dst6;
916 	struct in6_addrlifetime *lt;
917 	char ip6buf[INET6_ADDRSTRLEN];
918 
919 	/* Validate parameters */
920 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
921 		return (EINVAL);
922 
923 	/*
924 	 * The destination address for a p2p link must have a family
925 	 * of AF_UNSPEC or AF_INET6.
926 	 */
927 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
928 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
929 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
930 		return (EAFNOSUPPORT);
931 
932 	/*
933 	 * Validate address
934 	 */
935 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
936 	    ifra->ifra_addr.sin6_family != AF_INET6)
937 		return (EINVAL);
938 
939 	/*
940 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
941 	 * does not carry fields other than sin6_len.
942 	 */
943 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
944 		return (EINVAL);
945 	/*
946 	 * Because the IPv6 address architecture is classless, we require
947 	 * users to specify a (non 0) prefix length (mask) for a new address.
948 	 * We also require the prefix (when specified) mask is valid, and thus
949 	 * reject a non-consecutive mask.
950 	 */
951 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
952 		return (EINVAL);
953 	if (ifra->ifra_prefixmask.sin6_len != 0) {
954 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
955 		    (u_char *)&ifra->ifra_prefixmask +
956 		    ifra->ifra_prefixmask.sin6_len);
957 		if (plen <= 0)
958 			return (EINVAL);
959 	} else {
960 		/*
961 		 * In this case, ia must not be NULL.  We just use its prefix
962 		 * length.
963 		 */
964 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
965 	}
966 	/*
967 	 * If the destination address on a p2p interface is specified,
968 	 * and the address is a scoped one, validate/set the scope
969 	 * zone identifier.
970 	 */
971 	dst6 = ifra->ifra_dstaddr;
972 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
973 	    (dst6.sin6_family == AF_INET6)) {
974 		struct in6_addr in6_tmp;
975 		u_int32_t zoneid;
976 
977 		in6_tmp = dst6.sin6_addr;
978 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
979 			return (EINVAL); /* XXX: should be impossible */
980 
981 		if (dst6.sin6_scope_id != 0) {
982 			if (dst6.sin6_scope_id != zoneid)
983 				return (EINVAL);
984 		} else		/* user omit to specify the ID. */
985 			dst6.sin6_scope_id = zoneid;
986 
987 		/* convert into the internal form */
988 		if (sa6_embedscope(&dst6, 0))
989 			return (EINVAL); /* XXX: should be impossible */
990 	}
991 	/* Modify original ifra_dstaddr to reflect changes */
992 	ifra->ifra_dstaddr = dst6;
993 
994 	/*
995 	 * The destination address can be specified only for a p2p or a
996 	 * loopback interface.  If specified, the corresponding prefix length
997 	 * must be 128.
998 	 */
999 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1000 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1001 			/* XXX: noisy message */
1002 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1003 			    "be specified for a p2p or a loopback IF only\n"));
1004 			return (EINVAL);
1005 		}
1006 		if (plen != 128) {
1007 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1008 			    "be 128 when dstaddr is specified\n"));
1009 			return (EINVAL);
1010 		}
1011 	}
1012 	/* lifetime consistency check */
1013 	lt = &ifra->ifra_lifetime;
1014 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1015 		return (EINVAL);
1016 	if (lt->ia6t_vltime == 0) {
1017 		/*
1018 		 * the following log might be noisy, but this is a typical
1019 		 * configuration mistake or a tool's bug.
1020 		 */
1021 		nd6log((LOG_INFO,
1022 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1023 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1024 
1025 		if (ia == NULL)
1026 			return (0); /* there's nothing to do */
1027 	}
1028 
1029 	/* Check prefix mask */
1030 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1031 		/*
1032 		 * We prohibit changing the prefix length of an existing
1033 		 * address, because
1034 		 * + such an operation should be rare in IPv6, and
1035 		 * + the operation would confuse prefix management.
1036 		 */
1037 		if (ia->ia_prefixmask.sin6_len != 0 &&
1038 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1039 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1040 			    "of an existing %s address should not be changed\n",
1041 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1042 
1043 			return (EINVAL);
1044 		}
1045 	}
1046 
1047 	return (0);
1048 }
1049 
1050 
1051 /*
1052  * Allocate a new ifaddr and link it into chains.
1053  */
1054 static struct in6_ifaddr *
1055 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1056 {
1057 	struct in6_ifaddr *ia;
1058 
1059 	/*
1060 	 * When in6_alloc_ifa() is called in a process of a received
1061 	 * RA, it is called under an interrupt context.  So, we should
1062 	 * call malloc with M_NOWAIT.
1063 	 */
1064 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1065 	if (ia == NULL)
1066 		return (NULL);
1067 	LIST_INIT(&ia->ia6_memberships);
1068 	/* Initialize the address and masks, and put time stamp */
1069 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1070 	ia->ia_addr.sin6_family = AF_INET6;
1071 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1072 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1073 	ia->ia_addr = ifra->ifra_addr;
1074 	ia->ia6_createtime = time_uptime;
1075 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1076 		/*
1077 		 * Some functions expect that ifa_dstaddr is not
1078 		 * NULL for p2p interfaces.
1079 		 */
1080 		ia->ia_ifa.ifa_dstaddr =
1081 		    (struct sockaddr *)&ia->ia_dstaddr;
1082 	} else {
1083 		ia->ia_ifa.ifa_dstaddr = NULL;
1084 	}
1085 
1086 	/* set prefix mask if any */
1087 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1088 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1089 		ia->ia_prefixmask.sin6_family = AF_INET6;
1090 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1091 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1092 	}
1093 
1094 	ia->ia_ifp = ifp;
1095 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1096 	IF_ADDR_WLOCK(ifp);
1097 	TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1098 	IF_ADDR_WUNLOCK(ifp);
1099 
1100 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1101 	IN6_IFADDR_WLOCK();
1102 	TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1103 	LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1104 	IN6_IFADDR_WUNLOCK();
1105 
1106 	return (ia);
1107 }
1108 
1109 /*
1110  * Update/configure interface address parameters:
1111  *
1112  * 1) Update lifetime
1113  * 2) Update interface metric ad flags
1114  * 3) Notify other subsystems
1115  */
1116 static int
1117 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1118     struct in6_ifaddr *ia, int hostIsNew, int flags)
1119 {
1120 	int error;
1121 
1122 	/* update timestamp */
1123 	ia->ia6_updatetime = time_uptime;
1124 
1125 	/*
1126 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1127 	 * to see if the address is deprecated or invalidated, but initialize
1128 	 * these members for applications.
1129 	 */
1130 	ia->ia6_lifetime = ifra->ifra_lifetime;
1131 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1132 		ia->ia6_lifetime.ia6t_expire =
1133 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1134 	} else
1135 		ia->ia6_lifetime.ia6t_expire = 0;
1136 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1137 		ia->ia6_lifetime.ia6t_preferred =
1138 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1139 	} else
1140 		ia->ia6_lifetime.ia6t_preferred = 0;
1141 
1142 	/*
1143 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1144 	 * userland, make it deprecated.
1145 	 */
1146 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1147 		ia->ia6_lifetime.ia6t_pltime = 0;
1148 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1149 	}
1150 
1151 	/*
1152 	 * configure address flags.
1153 	 */
1154 	ia->ia6_flags = ifra->ifra_flags;
1155 
1156 	/*
1157 	 * Make the address tentative before joining multicast addresses,
1158 	 * so that corresponding MLD responses would not have a tentative
1159 	 * source address.
1160 	 */
1161 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1162 
1163 	/*
1164 	 * DAD should be performed for an new address or addresses on
1165 	 * an interface with ND6_IFF_IFDISABLED.
1166 	 */
1167 	if (in6if_do_dad(ifp) &&
1168 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1169 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1170 
1171 	/* notify other subsystems */
1172 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1173 
1174 	return (error);
1175 }
1176 
1177 /*
1178  * Do link-level ifa job:
1179  * 1) Add lle entry for added address
1180  * 2) Notifies routing socket users about new address
1181  * 3) join appropriate multicast group
1182  * 4) start DAD if enabled
1183  */
1184 static int
1185 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1186     struct in6_ifaddr *ia, int flags)
1187 {
1188 	struct in6_multi *in6m_sol;
1189 	int error = 0;
1190 
1191 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1192 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1193 		in6_purgeaddr(&ia->ia_ifa);
1194 		ifa_free(&ia->ia_ifa);
1195 		return (error);
1196 	}
1197 
1198 	/* Join necessary multicast groups. */
1199 	in6m_sol = NULL;
1200 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1201 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1202 		if (error != 0) {
1203 			in6_purgeaddr(&ia->ia_ifa);
1204 			ifa_free(&ia->ia_ifa);
1205 			return (error);
1206 		}
1207 	}
1208 
1209 	/* Perform DAD, if the address is TENTATIVE. */
1210 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1211 		int delay, mindelay, maxdelay;
1212 
1213 		delay = 0;
1214 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1215 			/*
1216 			 * We need to impose a delay before sending an NS
1217 			 * for DAD.  Check if we also needed a delay for the
1218 			 * corresponding MLD message.  If we did, the delay
1219 			 * should be larger than the MLD delay (this could be
1220 			 * relaxed a bit, but this simple logic is at least
1221 			 * safe).
1222 			 * XXX: Break data hiding guidelines and look at
1223 			 * state for the solicited multicast group.
1224 			 */
1225 			mindelay = 0;
1226 			if (in6m_sol != NULL &&
1227 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1228 				mindelay = in6m_sol->in6m_timer;
1229 			}
1230 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1231 			if (maxdelay - mindelay == 0)
1232 				delay = 0;
1233 			else {
1234 				delay =
1235 				    (arc4random() % (maxdelay - mindelay)) +
1236 				    mindelay;
1237 			}
1238 		}
1239 		nd6_dad_start((struct ifaddr *)ia, delay);
1240 	}
1241 
1242 	in6_newaddrmsg(ia, RTM_ADD);
1243 	ifa_free(&ia->ia_ifa);
1244 	return (error);
1245 }
1246 
1247 void
1248 in6_purgeaddr(struct ifaddr *ifa)
1249 {
1250 	struct ifnet *ifp = ifa->ifa_ifp;
1251 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1252 	struct in6_multi_mship *imm;
1253 	int plen, error;
1254 
1255 	if (ifa->ifa_carp)
1256 		(*carp_detach_p)(ifa, false);
1257 
1258 	/*
1259 	 * Remove the loopback route to the interface address.
1260 	 * The check for the current setting of "nd6_useloopback"
1261 	 * is not needed.
1262 	 */
1263 	if (ia->ia_flags & IFA_RTSELF) {
1264 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1265 		    (struct sockaddr *)&ia->ia_addr);
1266 		if (error == 0)
1267 			ia->ia_flags &= ~IFA_RTSELF;
1268 	}
1269 
1270 	/* stop DAD processing */
1271 	nd6_dad_stop(ifa);
1272 
1273 	/* Leave multicast groups. */
1274 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1275 		LIST_REMOVE(imm, i6mm_chain);
1276 		in6_leavegroup(imm);
1277 	}
1278 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1279 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1280 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1281 		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1282 		if (error != 0)
1283 			log(LOG_INFO, "%s: err=%d, destination address delete "
1284 			    "failed\n", __func__, error);
1285 		ia->ia_flags &= ~IFA_ROUTE;
1286 	}
1287 
1288 	in6_newaddrmsg(ia, RTM_DELETE);
1289 	in6_unlink_ifa(ia, ifp);
1290 }
1291 
1292 static void
1293 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1294 {
1295 	char ip6buf[INET6_ADDRSTRLEN];
1296 	int remove_lle;
1297 
1298 	IF_ADDR_WLOCK(ifp);
1299 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1300 	IF_ADDR_WUNLOCK(ifp);
1301 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1302 
1303 	/*
1304 	 * Defer the release of what might be the last reference to the
1305 	 * in6_ifaddr so that it can't be freed before the remainder of the
1306 	 * cleanup.
1307 	 */
1308 	IN6_IFADDR_WLOCK();
1309 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1310 	LIST_REMOVE(ia, ia6_hash);
1311 	IN6_IFADDR_WUNLOCK();
1312 
1313 	/*
1314 	 * Release the reference to the base prefix.  There should be a
1315 	 * positive reference.
1316 	 */
1317 	remove_lle = 0;
1318 	if (ia->ia6_ndpr == NULL) {
1319 		nd6log((LOG_NOTICE,
1320 		    "in6_unlink_ifa: autoconf'ed address "
1321 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1322 	} else {
1323 		ia->ia6_ndpr->ndpr_addrcnt--;
1324 		/* Do not delete lles within prefix if refcont != 0 */
1325 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1326 			remove_lle = 1;
1327 		ia->ia6_ndpr = NULL;
1328 	}
1329 
1330 	nd6_rem_ifa_lle(ia, remove_lle);
1331 
1332 	/*
1333 	 * Also, if the address being removed is autoconf'ed, call
1334 	 * pfxlist_onlink_check() since the release might affect the status of
1335 	 * other (detached) addresses.
1336 	 */
1337 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1338 		pfxlist_onlink_check();
1339 	}
1340 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1341 }
1342 
1343 /*
1344  * Notifies other subsystems about address change/arrival:
1345  * 1) Notifies device handler on the first IPv6 address assignment
1346  * 2) Handle routing table changes for P2P links and route
1347  * 3) Handle routing table changes for address host route
1348  */
1349 static int
1350 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1351     struct in6_aliasreq *ifra, int hostIsNew)
1352 {
1353 	int	error = 0, plen, ifacount = 0;
1354 	struct ifaddr *ifa;
1355 	struct sockaddr_in6 *pdst;
1356 	char ip6buf[INET6_ADDRSTRLEN];
1357 
1358 	/*
1359 	 * Give the interface a chance to initialize
1360 	 * if this is its first address,
1361 	 */
1362 	if (hostIsNew != 0) {
1363 		IF_ADDR_RLOCK(ifp);
1364 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1365 			if (ifa->ifa_addr->sa_family != AF_INET6)
1366 				continue;
1367 			ifacount++;
1368 		}
1369 		IF_ADDR_RUNLOCK(ifp);
1370 	}
1371 
1372 	if (ifacount <= 1 && ifp->if_ioctl) {
1373 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1374 		if (error)
1375 			return (error);
1376 	}
1377 
1378 	/*
1379 	 * If a new destination address is specified, scrub the old one and
1380 	 * install the new destination.  Note that the interface must be
1381 	 * p2p or loopback.
1382 	 */
1383 	pdst = &ifra->ifra_dstaddr;
1384 	if (pdst->sin6_family == AF_INET6 &&
1385 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1386 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1387 		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1388 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1389 			    "remove a route to the old destination: %s\n",
1390 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1391 			/* proceed anyway... */
1392 		} else
1393 			ia->ia_flags &= ~IFA_ROUTE;
1394 		ia->ia_dstaddr = *pdst;
1395 	}
1396 
1397 	/*
1398 	 * If a new destination address is specified for a point-to-point
1399 	 * interface, install a route to the destination as an interface
1400 	 * direct route.
1401 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1402 	 * interface that share the same destination.
1403 	 */
1404 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1405 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1406 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1407 		int rtflags = RTF_UP | RTF_HOST;
1408 		/*
1409 		 * Handle the case for ::1 .
1410 		 */
1411 		if (ifp->if_flags & IFF_LOOPBACK)
1412 			ia->ia_flags |= IFA_RTSELF;
1413 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1414 		if (error)
1415 			return (error);
1416 		ia->ia_flags |= IFA_ROUTE;
1417 	}
1418 
1419 	/*
1420 	 * add a loopback route to self if not exists
1421 	 */
1422 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1423 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1424 		    (struct sockaddr *)&ia->ia_addr);
1425 		if (error == 0)
1426 			ia->ia_flags |= IFA_RTSELF;
1427 	}
1428 
1429 	return (error);
1430 }
1431 
1432 /*
1433  * Find an IPv6 interface link-local address specific to an interface.
1434  * ifaddr is returned referenced.
1435  */
1436 struct in6_ifaddr *
1437 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1438 {
1439 	struct ifaddr *ifa;
1440 
1441 	IF_ADDR_RLOCK(ifp);
1442 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1443 		if (ifa->ifa_addr->sa_family != AF_INET6)
1444 			continue;
1445 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1446 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1447 			    ignoreflags) != 0)
1448 				continue;
1449 			ifa_ref(ifa);
1450 			break;
1451 		}
1452 	}
1453 	IF_ADDR_RUNLOCK(ifp);
1454 
1455 	return ((struct in6_ifaddr *)ifa);
1456 }
1457 
1458 
1459 /*
1460  * find the interface address corresponding to a given IPv6 address.
1461  * ifaddr is returned referenced.
1462  */
1463 struct in6_ifaddr *
1464 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1465 {
1466 	struct rm_priotracker in6_ifa_tracker;
1467 	struct in6_ifaddr *ia;
1468 
1469 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1470 	LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1471 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1472 			if (zoneid != 0 &&
1473 			    zoneid != ia->ia_addr.sin6_scope_id)
1474 				continue;
1475 			ifa_ref(&ia->ia_ifa);
1476 			break;
1477 		}
1478 	}
1479 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1480 	return (ia);
1481 }
1482 
1483 /*
1484  * find the internet address corresponding to a given interface and address.
1485  * ifaddr is returned referenced.
1486  */
1487 struct in6_ifaddr *
1488 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1489 {
1490 	struct ifaddr *ifa;
1491 
1492 	IF_ADDR_RLOCK(ifp);
1493 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1494 		if (ifa->ifa_addr->sa_family != AF_INET6)
1495 			continue;
1496 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1497 			ifa_ref(ifa);
1498 			break;
1499 		}
1500 	}
1501 	IF_ADDR_RUNLOCK(ifp);
1502 
1503 	return ((struct in6_ifaddr *)ifa);
1504 }
1505 
1506 /*
1507  * Find a link-local scoped address on ifp and return it if any.
1508  */
1509 struct in6_ifaddr *
1510 in6ifa_llaonifp(struct ifnet *ifp)
1511 {
1512 	struct sockaddr_in6 *sin6;
1513 	struct ifaddr *ifa;
1514 
1515 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1516 		return (NULL);
1517 	IF_ADDR_RLOCK(ifp);
1518 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1519 		if (ifa->ifa_addr->sa_family != AF_INET6)
1520 			continue;
1521 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1522 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1523 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1524 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1525 			break;
1526 	}
1527 	IF_ADDR_RUNLOCK(ifp);
1528 
1529 	return ((struct in6_ifaddr *)ifa);
1530 }
1531 
1532 /*
1533  * Convert IP6 address to printable (loggable) representation. Caller
1534  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1535  */
1536 static char digits[] = "0123456789abcdef";
1537 char *
1538 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1539 {
1540 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1541 	char *cp;
1542 	const u_int16_t *a = (const u_int16_t *)addr;
1543 	const u_int8_t *d;
1544 	int dcolon = 0, zero = 0;
1545 
1546 	cp = ip6buf;
1547 
1548 	for (i = 0; i < 8; i++) {
1549 		if (*(a + i) == 0) {
1550 			cnt++;
1551 			if (cnt == 1)
1552 				idx = i;
1553 		}
1554 		else if (maxcnt < cnt) {
1555 			maxcnt = cnt;
1556 			index = idx;
1557 			cnt = 0;
1558 		}
1559 	}
1560 	if (maxcnt < cnt) {
1561 		maxcnt = cnt;
1562 		index = idx;
1563 	}
1564 
1565 	for (i = 0; i < 8; i++) {
1566 		if (dcolon == 1) {
1567 			if (*a == 0) {
1568 				if (i == 7)
1569 					*cp++ = ':';
1570 				a++;
1571 				continue;
1572 			} else
1573 				dcolon = 2;
1574 		}
1575 		if (*a == 0) {
1576 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1577 				if (i == 0)
1578 					*cp++ = ':';
1579 				*cp++ = ':';
1580 				dcolon = 1;
1581 			} else {
1582 				*cp++ = '0';
1583 				*cp++ = ':';
1584 			}
1585 			a++;
1586 			continue;
1587 		}
1588 		d = (const u_char *)a;
1589 		/* Try to eliminate leading zeros in printout like in :0001. */
1590 		zero = 1;
1591 		*cp = digits[*d >> 4];
1592 		if (*cp != '0') {
1593 			zero = 0;
1594 			cp++;
1595 		}
1596 		*cp = digits[*d++ & 0xf];
1597 		if (zero == 0 || (*cp != '0')) {
1598 			zero = 0;
1599 			cp++;
1600 		}
1601 		*cp = digits[*d >> 4];
1602 		if (zero == 0 || (*cp != '0')) {
1603 			zero = 0;
1604 			cp++;
1605 		}
1606 		*cp++ = digits[*d & 0xf];
1607 		*cp++ = ':';
1608 		a++;
1609 	}
1610 	*--cp = '\0';
1611 	return (ip6buf);
1612 }
1613 
1614 int
1615 in6_localaddr(struct in6_addr *in6)
1616 {
1617 	struct rm_priotracker in6_ifa_tracker;
1618 	struct in6_ifaddr *ia;
1619 
1620 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1621 		return 1;
1622 
1623 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1624 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1625 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1626 		    &ia->ia_prefixmask.sin6_addr)) {
1627 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1628 			return 1;
1629 		}
1630 	}
1631 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1632 
1633 	return (0);
1634 }
1635 
1636 /*
1637  * Return 1 if an internet address is for the local host and configured
1638  * on one of its interfaces.
1639  */
1640 int
1641 in6_localip(struct in6_addr *in6)
1642 {
1643 	struct rm_priotracker in6_ifa_tracker;
1644 	struct in6_ifaddr *ia;
1645 
1646 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1647 	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1648 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1649 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1650 			return (1);
1651 		}
1652 	}
1653 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1654 	return (0);
1655 }
1656 
1657 /*
1658  * Return 1 if an internet address is configured on an interface.
1659  */
1660 int
1661 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1662 {
1663 	struct in6_addr in6;
1664 	struct ifaddr *ifa;
1665 	struct in6_ifaddr *ia6;
1666 
1667 	in6 = *addr;
1668 	if (in6_clearscope(&in6))
1669 		return (0);
1670 	in6_setscope(&in6, ifp, NULL);
1671 
1672 	IF_ADDR_RLOCK(ifp);
1673 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1674 		if (ifa->ifa_addr->sa_family != AF_INET6)
1675 			continue;
1676 		ia6 = (struct in6_ifaddr *)ifa;
1677 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6)) {
1678 			IF_ADDR_RUNLOCK(ifp);
1679 			return (1);
1680 		}
1681 	}
1682 	IF_ADDR_RUNLOCK(ifp);
1683 
1684 	return (0);
1685 }
1686 
1687 int
1688 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1689 {
1690 	struct rm_priotracker in6_ifa_tracker;
1691 	struct in6_ifaddr *ia;
1692 
1693 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1694 	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1695 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1696 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1697 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1698 				return (1); /* true */
1699 			}
1700 			break;
1701 		}
1702 	}
1703 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1704 
1705 	return (0);		/* false */
1706 }
1707 
1708 /*
1709  * return length of part which dst and src are equal
1710  * hard coding...
1711  */
1712 int
1713 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1714 {
1715 	int match = 0;
1716 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1717 	u_char *lim = s + 16, r;
1718 
1719 	while (s < lim)
1720 		if ((r = (*d++ ^ *s++)) != 0) {
1721 			while (r < 128) {
1722 				match++;
1723 				r <<= 1;
1724 			}
1725 			break;
1726 		} else
1727 			match += 8;
1728 	return match;
1729 }
1730 
1731 /* XXX: to be scope conscious */
1732 int
1733 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1734 {
1735 	int bytelen, bitlen;
1736 
1737 	/* sanity check */
1738 	if (0 > len || len > 128) {
1739 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1740 		    len);
1741 		return (0);
1742 	}
1743 
1744 	bytelen = len / 8;
1745 	bitlen = len % 8;
1746 
1747 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1748 		return (0);
1749 	if (bitlen != 0 &&
1750 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1751 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1752 		return (0);
1753 
1754 	return (1);
1755 }
1756 
1757 void
1758 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1759 {
1760 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1761 	int bytelen, bitlen, i;
1762 
1763 	/* sanity check */
1764 	if (0 > len || len > 128) {
1765 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1766 		    len);
1767 		return;
1768 	}
1769 
1770 	bzero(maskp, sizeof(*maskp));
1771 	bytelen = len / 8;
1772 	bitlen = len % 8;
1773 	for (i = 0; i < bytelen; i++)
1774 		maskp->s6_addr[i] = 0xff;
1775 	if (bitlen)
1776 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1777 }
1778 
1779 /*
1780  * return the best address out of the same scope. if no address was
1781  * found, return the first valid address from designated IF.
1782  */
1783 struct in6_ifaddr *
1784 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1785 {
1786 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1787 	struct ifaddr *ifa;
1788 	struct in6_ifaddr *besta = NULL;
1789 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1790 
1791 	dep[0] = dep[1] = NULL;
1792 
1793 	/*
1794 	 * We first look for addresses in the same scope.
1795 	 * If there is one, return it.
1796 	 * If two or more, return one which matches the dst longest.
1797 	 * If none, return one of global addresses assigned other ifs.
1798 	 */
1799 	IF_ADDR_RLOCK(ifp);
1800 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1801 		if (ifa->ifa_addr->sa_family != AF_INET6)
1802 			continue;
1803 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1804 			continue; /* XXX: is there any case to allow anycast? */
1805 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1806 			continue; /* don't use this interface */
1807 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1808 			continue;
1809 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1810 			if (V_ip6_use_deprecated)
1811 				dep[0] = (struct in6_ifaddr *)ifa;
1812 			continue;
1813 		}
1814 
1815 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1816 			/*
1817 			 * call in6_matchlen() as few as possible
1818 			 */
1819 			if (besta) {
1820 				if (blen == -1)
1821 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1822 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1823 				if (tlen > blen) {
1824 					blen = tlen;
1825 					besta = (struct in6_ifaddr *)ifa;
1826 				}
1827 			} else
1828 				besta = (struct in6_ifaddr *)ifa;
1829 		}
1830 	}
1831 	if (besta) {
1832 		ifa_ref(&besta->ia_ifa);
1833 		IF_ADDR_RUNLOCK(ifp);
1834 		return (besta);
1835 	}
1836 
1837 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1838 		if (ifa->ifa_addr->sa_family != AF_INET6)
1839 			continue;
1840 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1841 			continue; /* XXX: is there any case to allow anycast? */
1842 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1843 			continue; /* don't use this interface */
1844 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1845 			continue;
1846 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1847 			if (V_ip6_use_deprecated)
1848 				dep[1] = (struct in6_ifaddr *)ifa;
1849 			continue;
1850 		}
1851 
1852 		if (ifa != NULL)
1853 			ifa_ref(ifa);
1854 		IF_ADDR_RUNLOCK(ifp);
1855 		return (struct in6_ifaddr *)ifa;
1856 	}
1857 
1858 	/* use the last-resort values, that are, deprecated addresses */
1859 	if (dep[0]) {
1860 		ifa_ref((struct ifaddr *)dep[0]);
1861 		IF_ADDR_RUNLOCK(ifp);
1862 		return dep[0];
1863 	}
1864 	if (dep[1]) {
1865 		ifa_ref((struct ifaddr *)dep[1]);
1866 		IF_ADDR_RUNLOCK(ifp);
1867 		return dep[1];
1868 	}
1869 
1870 	IF_ADDR_RUNLOCK(ifp);
1871 	return NULL;
1872 }
1873 
1874 /*
1875  * perform DAD when interface becomes IFF_UP.
1876  */
1877 void
1878 in6_if_up(struct ifnet *ifp)
1879 {
1880 	struct ifaddr *ifa;
1881 	struct in6_ifaddr *ia;
1882 
1883 	IF_ADDR_RLOCK(ifp);
1884 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1885 		if (ifa->ifa_addr->sa_family != AF_INET6)
1886 			continue;
1887 		ia = (struct in6_ifaddr *)ifa;
1888 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1889 			/*
1890 			 * The TENTATIVE flag was likely set by hand
1891 			 * beforehand, implicitly indicating the need for DAD.
1892 			 * We may be able to skip the random delay in this
1893 			 * case, but we impose delays just in case.
1894 			 */
1895 			nd6_dad_start(ifa,
1896 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1897 		}
1898 	}
1899 	IF_ADDR_RUNLOCK(ifp);
1900 
1901 	/*
1902 	 * special cases, like 6to4, are handled in in6_ifattach
1903 	 */
1904 	in6_ifattach(ifp, NULL);
1905 }
1906 
1907 int
1908 in6if_do_dad(struct ifnet *ifp)
1909 {
1910 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1911 		return (0);
1912 
1913 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) ||
1914 	    (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD))
1915 		return (0);
1916 
1917 	/*
1918 	 * Our DAD routine requires the interface up and running.
1919 	 * However, some interfaces can be up before the RUNNING
1920 	 * status.  Additionally, users may try to assign addresses
1921 	 * before the interface becomes up (or running).
1922 	 * This function returns EAGAIN in that case.
1923 	 * The caller should mark "tentative" on the address instead of
1924 	 * performing DAD immediately.
1925 	 */
1926 	if (!((ifp->if_flags & IFF_UP) &&
1927 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
1928 		return (EAGAIN);
1929 
1930 	return (1);
1931 }
1932 
1933 /*
1934  * Calculate max IPv6 MTU through all the interfaces and store it
1935  * to in6_maxmtu.
1936  */
1937 void
1938 in6_setmaxmtu(void)
1939 {
1940 	unsigned long maxmtu = 0;
1941 	struct ifnet *ifp;
1942 
1943 	IFNET_RLOCK_NOSLEEP();
1944 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1945 		/* this function can be called during ifnet initialization */
1946 		if (!ifp->if_afdata[AF_INET6])
1947 			continue;
1948 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1949 		    IN6_LINKMTU(ifp) > maxmtu)
1950 			maxmtu = IN6_LINKMTU(ifp);
1951 	}
1952 	IFNET_RUNLOCK_NOSLEEP();
1953 	if (maxmtu)	/* update only when maxmtu is positive */
1954 		V_in6_maxmtu = maxmtu;
1955 }
1956 
1957 /*
1958  * Provide the length of interface identifiers to be used for the link attached
1959  * to the given interface.  The length should be defined in "IPv6 over
1960  * xxx-link" document.  Note that address architecture might also define
1961  * the length for a particular set of address prefixes, regardless of the
1962  * link type.  As clarified in rfc2462bis, those two definitions should be
1963  * consistent, and those really are as of August 2004.
1964  */
1965 int
1966 in6_if2idlen(struct ifnet *ifp)
1967 {
1968 	switch (ifp->if_type) {
1969 	case IFT_ETHER:		/* RFC2464 */
1970 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
1971 	case IFT_L2VLAN:	/* ditto */
1972 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
1973 	case IFT_INFINIBAND:
1974 		return (64);
1975 	case IFT_PPP:		/* RFC2472 */
1976 		return (64);
1977 	case IFT_FRELAY:	/* RFC2590 */
1978 		return (64);
1979 	case IFT_IEEE1394:	/* RFC3146 */
1980 		return (64);
1981 	case IFT_GIF:
1982 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
1983 	case IFT_LOOP:
1984 		return (64);	/* XXX: is this really correct? */
1985 	default:
1986 		/*
1987 		 * Unknown link type:
1988 		 * It might be controversial to use the today's common constant
1989 		 * of 64 for these cases unconditionally.  For full compliance,
1990 		 * we should return an error in this case.  On the other hand,
1991 		 * if we simply miss the standard for the link type or a new
1992 		 * standard is defined for a new link type, the IFID length
1993 		 * is very likely to be the common constant.  As a compromise,
1994 		 * we always use the constant, but make an explicit notice
1995 		 * indicating the "unknown" case.
1996 		 */
1997 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
1998 		return (64);
1999 	}
2000 }
2001 
2002 #include <sys/sysctl.h>
2003 
2004 struct in6_llentry {
2005 	struct llentry		base;
2006 };
2007 
2008 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2009 #define	IN6_LLTBL_HASH(k, h) \
2010 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2011 
2012 /*
2013  * Do actual deallocation of @lle.
2014  */
2015 static void
2016 in6_lltable_destroy_lle_unlocked(struct llentry *lle)
2017 {
2018 
2019 	LLE_LOCK_DESTROY(lle);
2020 	LLE_REQ_DESTROY(lle);
2021 	free(lle, M_LLTABLE);
2022 }
2023 
2024 /*
2025  * Called by LLE_FREE_LOCKED when number of references
2026  * drops to zero.
2027  */
2028 static void
2029 in6_lltable_destroy_lle(struct llentry *lle)
2030 {
2031 
2032 	LLE_WUNLOCK(lle);
2033 	in6_lltable_destroy_lle_unlocked(lle);
2034 }
2035 
2036 static struct llentry *
2037 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2038 {
2039 	struct in6_llentry *lle;
2040 
2041 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2042 	if (lle == NULL)		/* NB: caller generates msg */
2043 		return NULL;
2044 
2045 	lle->base.r_l3addr.addr6 = *addr6;
2046 	lle->base.lle_refcnt = 1;
2047 	lle->base.lle_free = in6_lltable_destroy_lle;
2048 	LLE_LOCK_INIT(&lle->base);
2049 	LLE_REQ_INIT(&lle->base);
2050 	callout_init(&lle->base.lle_timer, 1);
2051 
2052 	return (&lle->base);
2053 }
2054 
2055 static int
2056 in6_lltable_match_prefix(const struct sockaddr *saddr,
2057     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2058 {
2059 	const struct in6_addr *addr, *mask, *lle_addr;
2060 
2061 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2062 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2063 	lle_addr = &lle->r_l3addr.addr6;
2064 
2065 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2066 		return (0);
2067 
2068 	if (lle->la_flags & LLE_IFADDR) {
2069 
2070 		/*
2071 		 * Delete LLE_IFADDR records IFF address & flag matches.
2072 		 * Note that addr is the interface address within prefix
2073 		 * being matched.
2074 		 */
2075 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2076 		    (flags & LLE_STATIC) != 0)
2077 			return (1);
2078 		return (0);
2079 	}
2080 
2081 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2082 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2083 		return (1);
2084 
2085 	return (0);
2086 }
2087 
2088 static void
2089 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2090 {
2091 	struct ifnet *ifp;
2092 
2093 	LLE_WLOCK_ASSERT(lle);
2094 	KASSERT(llt != NULL, ("lltable is NULL"));
2095 
2096 	/* Unlink entry from table */
2097 	if ((lle->la_flags & LLE_LINKED) != 0) {
2098 
2099 		ifp = llt->llt_ifp;
2100 		IF_AFDATA_WLOCK_ASSERT(ifp);
2101 		lltable_unlink_entry(llt, lle);
2102 	}
2103 
2104 	if (callout_stop(&lle->lle_timer) > 0)
2105 		LLE_REMREF(lle);
2106 
2107 	llentry_free(lle);
2108 }
2109 
2110 static int
2111 in6_lltable_rtcheck(struct ifnet *ifp,
2112 		    u_int flags,
2113 		    const struct sockaddr *l3addr)
2114 {
2115 	const struct sockaddr_in6 *sin6;
2116 	struct nhop6_basic nh6;
2117 	struct in6_addr dst;
2118 	uint32_t scopeid;
2119 	int error;
2120 	char ip6buf[INET6_ADDRSTRLEN];
2121 	int fibnum;
2122 
2123 	KASSERT(l3addr->sa_family == AF_INET6,
2124 	    ("sin_family %d", l3addr->sa_family));
2125 
2126 	sin6 = (const struct sockaddr_in6 *)l3addr;
2127 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2128 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2129 	error = fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6);
2130 	if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.nh_ifp != ifp) {
2131 		struct ifaddr *ifa;
2132 		/*
2133 		 * Create an ND6 cache for an IPv6 neighbor
2134 		 * that is not covered by our own prefix.
2135 		 */
2136 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2137 		if (ifa != NULL) {
2138 			ifa_free(ifa);
2139 			return 0;
2140 		}
2141 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2142 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2143 		return EINVAL;
2144 	}
2145 	return 0;
2146 }
2147 
2148 /*
2149  * Called by the datapath to indicate that the entry was used.
2150  */
2151 static void
2152 in6_lltable_mark_used(struct llentry *lle)
2153 {
2154 
2155 	LLE_REQ_LOCK(lle);
2156 	lle->r_skip_req = 0;
2157 
2158 	/*
2159 	 * Set the hit time so the callback function
2160 	 * can determine the remaining time before
2161 	 * transiting to the DELAY state.
2162 	 */
2163 	lle->lle_hittime = time_uptime;
2164 	LLE_REQ_UNLOCK(lle);
2165 }
2166 
2167 static inline uint32_t
2168 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2169 {
2170 
2171 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2172 }
2173 
2174 static uint32_t
2175 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2176 {
2177 
2178 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2179 }
2180 
2181 static void
2182 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2183 {
2184 	struct sockaddr_in6 *sin6;
2185 
2186 	sin6 = (struct sockaddr_in6 *)sa;
2187 	bzero(sin6, sizeof(*sin6));
2188 	sin6->sin6_family = AF_INET6;
2189 	sin6->sin6_len = sizeof(*sin6);
2190 	sin6->sin6_addr = lle->r_l3addr.addr6;
2191 }
2192 
2193 static inline struct llentry *
2194 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2195 {
2196 	struct llentry *lle;
2197 	struct llentries *lleh;
2198 	u_int hashidx;
2199 
2200 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2201 	lleh = &llt->lle_head[hashidx];
2202 	LIST_FOREACH(lle, lleh, lle_next) {
2203 		if (lle->la_flags & LLE_DELETED)
2204 			continue;
2205 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2206 			break;
2207 	}
2208 
2209 	return (lle);
2210 }
2211 
2212 static void
2213 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2214 {
2215 
2216 	lle->la_flags |= LLE_DELETED;
2217 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2218 #ifdef DIAGNOSTIC
2219 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2220 #endif
2221 	llentry_free(lle);
2222 }
2223 
2224 static struct llentry *
2225 in6_lltable_alloc(struct lltable *llt, u_int flags,
2226 	const struct sockaddr *l3addr)
2227 {
2228 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2229 	struct ifnet *ifp = llt->llt_ifp;
2230 	struct llentry *lle;
2231 	char linkhdr[LLE_MAX_LINKHDR];
2232 	size_t linkhdrsize;
2233 	int lladdr_off;
2234 
2235 	KASSERT(l3addr->sa_family == AF_INET6,
2236 	    ("sin_family %d", l3addr->sa_family));
2237 
2238 	/*
2239 	 * A route that covers the given address must have
2240 	 * been installed 1st because we are doing a resolution,
2241 	 * verify this.
2242 	 */
2243 	if (!(flags & LLE_IFADDR) &&
2244 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2245 		return (NULL);
2246 
2247 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2248 	if (lle == NULL) {
2249 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2250 		return (NULL);
2251 	}
2252 	lle->la_flags = flags;
2253 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2254 		linkhdrsize = LLE_MAX_LINKHDR;
2255 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2256 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2257 			in6_lltable_destroy_lle_unlocked(lle);
2258 			return (NULL);
2259 		}
2260 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2261 		    lladdr_off);
2262 		lle->la_flags |= LLE_STATIC;
2263 	}
2264 
2265 	if ((lle->la_flags & LLE_STATIC) != 0)
2266 		lle->ln_state = ND6_LLINFO_REACHABLE;
2267 
2268 	return (lle);
2269 }
2270 
2271 static struct llentry *
2272 in6_lltable_lookup(struct lltable *llt, u_int flags,
2273 	const struct sockaddr *l3addr)
2274 {
2275 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2276 	struct llentry *lle;
2277 
2278 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2279 	KASSERT(l3addr->sa_family == AF_INET6,
2280 	    ("sin_family %d", l3addr->sa_family));
2281 
2282 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2283 
2284 	if (lle == NULL)
2285 		return (NULL);
2286 
2287 	KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) !=
2288 	    (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X",
2289 	    flags));
2290 
2291 	if (flags & LLE_UNLOCKED)
2292 		return (lle);
2293 
2294 	if (flags & LLE_EXCLUSIVE)
2295 		LLE_WLOCK(lle);
2296 	else
2297 		LLE_RLOCK(lle);
2298 	return (lle);
2299 }
2300 
2301 static int
2302 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2303     struct sysctl_req *wr)
2304 {
2305 	struct ifnet *ifp = llt->llt_ifp;
2306 	/* XXX stack use */
2307 	struct {
2308 		struct rt_msghdr	rtm;
2309 		struct sockaddr_in6	sin6;
2310 		/*
2311 		 * ndp.c assumes that sdl is word aligned
2312 		 */
2313 #ifdef __LP64__
2314 		uint32_t		pad;
2315 #endif
2316 		struct sockaddr_dl	sdl;
2317 	} ndpc;
2318 	struct sockaddr_dl *sdl;
2319 	int error;
2320 
2321 	bzero(&ndpc, sizeof(ndpc));
2322 			/* skip deleted entries */
2323 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2324 				return (0);
2325 			/* Skip if jailed and not a valid IP of the prison. */
2326 			lltable_fill_sa_entry(lle,
2327 			    (struct sockaddr *)&ndpc.sin6);
2328 			if (prison_if(wr->td->td_ucred,
2329 			    (struct sockaddr *)&ndpc.sin6) != 0)
2330 				return (0);
2331 			/*
2332 			 * produce a msg made of:
2333 			 *  struct rt_msghdr;
2334 			 *  struct sockaddr_in6 (IPv6)
2335 			 *  struct sockaddr_dl;
2336 			 */
2337 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2338 			ndpc.rtm.rtm_version = RTM_VERSION;
2339 			ndpc.rtm.rtm_type = RTM_GET;
2340 			ndpc.rtm.rtm_flags = RTF_UP;
2341 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2342 			if (V_deembed_scopeid)
2343 				sa6_recoverscope(&ndpc.sin6);
2344 
2345 			/* publish */
2346 			if (lle->la_flags & LLE_PUB)
2347 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2348 
2349 			sdl = &ndpc.sdl;
2350 			sdl->sdl_family = AF_LINK;
2351 			sdl->sdl_len = sizeof(*sdl);
2352 			sdl->sdl_index = ifp->if_index;
2353 			sdl->sdl_type = ifp->if_type;
2354 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2355 				sdl->sdl_alen = ifp->if_addrlen;
2356 				bcopy(lle->ll_addr, LLADDR(sdl),
2357 				    ifp->if_addrlen);
2358 			} else {
2359 				sdl->sdl_alen = 0;
2360 				bzero(LLADDR(sdl), ifp->if_addrlen);
2361 			}
2362 			if (lle->la_expire != 0)
2363 				ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2364 				    lle->lle_remtime / hz +
2365 				    time_second - time_uptime;
2366 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2367 			if (lle->la_flags & LLE_STATIC)
2368 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2369 			if (lle->la_flags & LLE_IFADDR)
2370 				ndpc.rtm.rtm_flags |= RTF_PINNED;
2371 			if (lle->ln_router != 0)
2372 				ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2373 			ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2374 			/* Store state in rmx_weight value */
2375 			ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2376 			ndpc.rtm.rtm_index = ifp->if_index;
2377 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2378 
2379 	return (error);
2380 }
2381 
2382 static struct lltable *
2383 in6_lltattach(struct ifnet *ifp)
2384 {
2385 	struct lltable *llt;
2386 
2387 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2388 	llt->llt_af = AF_INET6;
2389 	llt->llt_ifp = ifp;
2390 
2391 	llt->llt_lookup = in6_lltable_lookup;
2392 	llt->llt_alloc_entry = in6_lltable_alloc;
2393 	llt->llt_delete_entry = in6_lltable_delete_entry;
2394 	llt->llt_dump_entry = in6_lltable_dump_entry;
2395 	llt->llt_hash = in6_lltable_hash;
2396 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2397 	llt->llt_free_entry = in6_lltable_free_entry;
2398 	llt->llt_match_prefix = in6_lltable_match_prefix;
2399 	llt->llt_mark_used = in6_lltable_mark_used;
2400  	lltable_link(llt);
2401 
2402 	return (llt);
2403 }
2404 
2405 void *
2406 in6_domifattach(struct ifnet *ifp)
2407 {
2408 	struct in6_ifextra *ext;
2409 
2410 	/* There are not IPv6-capable interfaces. */
2411 	switch (ifp->if_type) {
2412 	case IFT_PFLOG:
2413 	case IFT_PFSYNC:
2414 	case IFT_USB:
2415 		return (NULL);
2416 	}
2417 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2418 	bzero(ext, sizeof(*ext));
2419 
2420 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2421 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2422 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2423 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2424 
2425 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2426 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2427 	    M_WAITOK);
2428 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2429 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2430 
2431 	ext->nd_ifinfo = nd6_ifattach(ifp);
2432 	ext->scope6_id = scope6_ifattach(ifp);
2433 	ext->lltable = in6_lltattach(ifp);
2434 
2435 	ext->mld_ifinfo = mld_domifattach(ifp);
2436 
2437 	return ext;
2438 }
2439 
2440 int
2441 in6_domifmtu(struct ifnet *ifp)
2442 {
2443 	if (ifp->if_afdata[AF_INET6] == NULL)
2444 		return ifp->if_mtu;
2445 
2446 	return (IN6_LINKMTU(ifp));
2447 }
2448 
2449 void
2450 in6_domifdetach(struct ifnet *ifp, void *aux)
2451 {
2452 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2453 
2454 	mld_domifdetach(ifp);
2455 	scope6_ifdetach(ext->scope6_id);
2456 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2457 	lltable_free(ext->lltable);
2458 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2459 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2460 	free(ext->in6_ifstat, M_IFADDR);
2461 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2462 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2463 	free(ext->icmp6_ifstat, M_IFADDR);
2464 	free(ext, M_IFADDR);
2465 }
2466 
2467 /*
2468  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2469  * v4 mapped addr or v4 compat addr
2470  */
2471 void
2472 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2473 {
2474 
2475 	bzero(sin, sizeof(*sin));
2476 	sin->sin_len = sizeof(struct sockaddr_in);
2477 	sin->sin_family = AF_INET;
2478 	sin->sin_port = sin6->sin6_port;
2479 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2480 }
2481 
2482 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2483 void
2484 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2485 {
2486 	bzero(sin6, sizeof(*sin6));
2487 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2488 	sin6->sin6_family = AF_INET6;
2489 	sin6->sin6_port = sin->sin_port;
2490 	sin6->sin6_addr.s6_addr32[0] = 0;
2491 	sin6->sin6_addr.s6_addr32[1] = 0;
2492 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2493 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2494 }
2495 
2496 /* Convert sockaddr_in6 into sockaddr_in. */
2497 void
2498 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2499 {
2500 	struct sockaddr_in *sin_p;
2501 	struct sockaddr_in6 sin6;
2502 
2503 	/*
2504 	 * Save original sockaddr_in6 addr and convert it
2505 	 * to sockaddr_in.
2506 	 */
2507 	sin6 = *(struct sockaddr_in6 *)nam;
2508 	sin_p = (struct sockaddr_in *)nam;
2509 	in6_sin6_2_sin(sin_p, &sin6);
2510 }
2511 
2512 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2513 void
2514 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2515 {
2516 	struct sockaddr_in *sin_p;
2517 	struct sockaddr_in6 *sin6_p;
2518 
2519 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2520 	sin_p = (struct sockaddr_in *)*nam;
2521 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2522 	free(*nam, M_SONAME);
2523 	*nam = (struct sockaddr *)sin6_p;
2524 }
2525