xref: /freebsd/sys/netinet6/in6.c (revision bdd1243d)
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/sysctl.h>
87 #include <sys/syslog.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/if_private.h>
92 #include <net/if_types.h>
93 #include <net/route.h>
94 #include <net/route/route_ctl.h>
95 #include <net/route/nhop.h>
96 #include <net/if_dl.h>
97 #include <net/vnet.h>
98 
99 #include <netinet/in.h>
100 #include <netinet/in_var.h>
101 #include <net/if_llatbl.h>
102 #include <netinet/if_ether.h>
103 #include <netinet/in_systm.h>
104 #include <netinet/ip.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/ip_carp.h>
107 
108 #include <netinet/ip6.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/mld6_var.h>
112 #include <netinet6/ip6_mroute.h>
113 #include <netinet6/in6_ifattach.h>
114 #include <netinet6/scope6_var.h>
115 #include <netinet6/in6_fib.h>
116 #include <netinet6/in6_pcb.h>
117 
118 /*
119  * struct in6_ifreq and struct ifreq must be type punnable for common members
120  * of ifr_ifru to allow accessors to be shared.
121  */
122 _Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
123     offsetof(struct ifreq, ifr_ifru),
124     "struct in6_ifreq and struct ifreq are not type punnable");
125 
126 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
127 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
128 
129 /*
130  * Definitions of some costant IP6 addresses.
131  */
132 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
133 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
134 const struct in6_addr in6addr_nodelocal_allnodes =
135 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
136 const struct in6_addr in6addr_linklocal_allnodes =
137 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
138 const struct in6_addr in6addr_linklocal_allrouters =
139 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
140 const struct in6_addr in6addr_linklocal_allv2routers =
141 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
142 
143 const struct in6_addr in6mask0 = IN6MASK0;
144 const struct in6_addr in6mask32 = IN6MASK32;
145 const struct in6_addr in6mask64 = IN6MASK64;
146 const struct in6_addr in6mask96 = IN6MASK96;
147 const struct in6_addr in6mask128 = IN6MASK128;
148 
149 const struct sockaddr_in6 sa6_any =
150 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
151 
152 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
153 	struct in6_aliasreq *, int);
154 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
155 
156 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
157     struct in6_ifaddr *, int);
158 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
159     struct in6_aliasreq *, int flags);
160 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
161     struct in6_ifaddr *, int, int);
162 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
163     struct in6_ifaddr *, int);
164 
165 static void in6_join_proxy_ndp_mc(struct ifnet *, const struct in6_addr *);
166 static void in6_leave_proxy_ndp_mc(struct ifnet *, const struct in6_addr *);
167 
168 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
169 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
170 
171 static struct sx in6_control_sx;
172 SX_SYSINIT(in6_control_sx, &in6_control_sx, "in6_control");
173 
174 void
175 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
176 {
177 	struct rt_addrinfo info;
178 	struct ifaddr *ifa;
179 	struct sockaddr_dl gateway;
180 	int fibnum;
181 
182 	ifa = &ia->ia_ifa;
183 
184 	/*
185 	 * Prepare info data for the host route.
186 	 * This code mimics one from ifa_maintain_loopback_route().
187 	 */
188 	bzero(&info, sizeof(struct rt_addrinfo));
189 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
190 	info.rti_info[RTAX_DST] = ifa->ifa_addr;
191 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
192 	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
193 	if (cmd != RTM_DELETE)
194 		info.rti_ifp = V_loif;
195 
196 	fibnum = ia62ifa(ia)->ifa_ifp->if_fib;
197 
198 	if (cmd == RTM_ADD) {
199 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
200 		rt_routemsg_info(cmd, &info, fibnum);
201 	} else if (cmd == RTM_DELETE) {
202 		rt_routemsg_info(cmd, &info, fibnum);
203 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
204 	}
205 }
206 
207 int
208 in6_mask2len(struct in6_addr *mask, u_char *lim0)
209 {
210 	int x = 0, y;
211 	u_char *lim = lim0, *p;
212 
213 	/* ignore the scope_id part */
214 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
215 		lim = (u_char *)mask + sizeof(*mask);
216 	for (p = (u_char *)mask; p < lim; x++, p++) {
217 		if (*p != 0xff)
218 			break;
219 	}
220 	y = 0;
221 	if (p < lim) {
222 		for (y = 0; y < 8; y++) {
223 			if ((*p & (0x80 >> y)) == 0)
224 				break;
225 		}
226 	}
227 
228 	/*
229 	 * when the limit pointer is given, do a stricter check on the
230 	 * remaining bits.
231 	 */
232 	if (p < lim) {
233 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
234 			return (-1);
235 		for (p = p + 1; p < lim; p++)
236 			if (*p != 0)
237 				return (-1);
238 	}
239 
240 	return x * 8 + y;
241 }
242 
243 #ifdef COMPAT_FREEBSD32
244 struct in6_ndifreq32 {
245 	char ifname[IFNAMSIZ];
246 	uint32_t ifindex;
247 };
248 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
249 #endif
250 
251 int
252 in6_control(struct socket *so, u_long cmd, void *data,
253     struct ifnet *ifp, struct thread *td)
254 {
255 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
256 	struct	in6_ifaddr *ia = NULL;
257 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
258 	struct sockaddr_in6 *sa6;
259 	int error;
260 	bool control_locked = false;
261 
262 	/*
263 	 * Compat to make pre-10.x ifconfig(8) operable.
264 	 */
265 	if (cmd == OSIOCAIFADDR_IN6) {
266 		cmd = SIOCAIFADDR_IN6;
267 		ifra->ifra_vhid = 0;
268 	}
269 
270 	switch (cmd) {
271 	case SIOCGETSGCNT_IN6:
272 	case SIOCGETMIFCNT_IN6:
273 		/*
274 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
275 		 * We cannot see how that would be needed, so do not adjust the
276 		 * KPI blindly; more likely should clean up the IPv4 variant.
277 		 */
278 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
279 	}
280 
281 	switch (cmd) {
282 	case SIOCAADDRCTL_POLICY:
283 	case SIOCDADDRCTL_POLICY:
284 		if (td != NULL) {
285 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
286 			if (error)
287 				return (error);
288 		}
289 		return (in6_src_ioctl(cmd, data));
290 	}
291 
292 	if (ifp == NULL)
293 		return (EOPNOTSUPP);
294 
295 	switch (cmd) {
296 	case SIOCSNDFLUSH_IN6:
297 	case SIOCSPFXFLUSH_IN6:
298 	case SIOCSRTRFLUSH_IN6:
299 	case SIOCSDEFIFACE_IN6:
300 	case SIOCSIFINFO_FLAGS:
301 	case SIOCSIFINFO_IN6:
302 		if (td != NULL) {
303 			error = priv_check(td, PRIV_NETINET_ND6);
304 			if (error)
305 				return (error);
306 		}
307 		/* FALLTHROUGH */
308 	case OSIOCGIFINFO_IN6:
309 	case SIOCGIFINFO_IN6:
310 	case SIOCGNBRINFO_IN6:
311 	case SIOCGDEFIFACE_IN6:
312 		return (nd6_ioctl(cmd, data, ifp));
313 
314 #ifdef COMPAT_FREEBSD32
315 	case SIOCGDEFIFACE32_IN6:
316 		{
317 			struct in6_ndifreq ndif;
318 			struct in6_ndifreq32 *ndif32;
319 
320 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
321 			    ifp);
322 			if (error)
323 				return (error);
324 			ndif32 = (struct in6_ndifreq32 *)data;
325 			ndif32->ifindex = ndif.ifindex;
326 			return (0);
327 		}
328 #endif
329 	}
330 
331 	switch (cmd) {
332 	case SIOCSIFPREFIX_IN6:
333 	case SIOCDIFPREFIX_IN6:
334 	case SIOCAIFPREFIX_IN6:
335 	case SIOCCIFPREFIX_IN6:
336 	case SIOCSGIFPREFIX_IN6:
337 	case SIOCGIFPREFIX_IN6:
338 		log(LOG_NOTICE,
339 		    "prefix ioctls are now invalidated. "
340 		    "please use ifconfig.\n");
341 		return (EOPNOTSUPP);
342 	}
343 
344 	switch (cmd) {
345 	case SIOCSSCOPE6:
346 		if (td != NULL) {
347 			error = priv_check(td, PRIV_NETINET_SCOPE6);
348 			if (error)
349 				return (error);
350 		}
351 		/* FALLTHROUGH */
352 	case SIOCGSCOPE6:
353 	case SIOCGSCOPE6DEF:
354 		return (scope6_ioctl(cmd, data, ifp));
355 	}
356 
357 	/*
358 	 * Find address for this interface, if it exists.
359 	 *
360 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
361 	 * only, and used the first interface address as the target of other
362 	 * operations (without checking ifra_addr).  This was because netinet
363 	 * code/API assumed at most 1 interface address per interface.
364 	 * Since IPv6 allows a node to assign multiple addresses
365 	 * on a single interface, we almost always look and check the
366 	 * presence of ifra_addr, and reject invalid ones here.
367 	 * It also decreases duplicated code among SIOC*_IN6 operations.
368 	 */
369 	switch (cmd) {
370 	case SIOCAIFADDR_IN6:
371 	case SIOCSIFPHYADDR_IN6:
372 		sa6 = &ifra->ifra_addr;
373 		break;
374 	case SIOCSIFADDR_IN6:
375 	case SIOCGIFADDR_IN6:
376 	case SIOCSIFDSTADDR_IN6:
377 	case SIOCSIFNETMASK_IN6:
378 	case SIOCGIFDSTADDR_IN6:
379 	case SIOCGIFNETMASK_IN6:
380 	case SIOCDIFADDR_IN6:
381 	case SIOCGIFPSRCADDR_IN6:
382 	case SIOCGIFPDSTADDR_IN6:
383 	case SIOCGIFAFLAG_IN6:
384 	case SIOCSNDFLUSH_IN6:
385 	case SIOCSPFXFLUSH_IN6:
386 	case SIOCSRTRFLUSH_IN6:
387 	case SIOCGIFALIFETIME_IN6:
388 	case SIOCGIFSTAT_IN6:
389 	case SIOCGIFSTAT_ICMP6:
390 		sa6 = &ifr->ifr_addr;
391 		break;
392 	case SIOCSIFADDR:
393 	case SIOCSIFBRDADDR:
394 	case SIOCSIFDSTADDR:
395 	case SIOCSIFNETMASK:
396 		/*
397 		 * Although we should pass any non-INET6 ioctl requests
398 		 * down to driver, we filter some legacy INET requests.
399 		 * Drivers trust SIOCSIFADDR et al to come from an already
400 		 * privileged layer, and do not perform any credentials
401 		 * checks or input validation.
402 		 */
403 		return (EINVAL);
404 	default:
405 		sa6 = NULL;
406 		break;
407 	}
408 	if (sa6 && sa6->sin6_family == AF_INET6) {
409 		if (sa6->sin6_scope_id != 0)
410 			error = sa6_embedscope(sa6, 0);
411 		else
412 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
413 		if (error != 0)
414 			return (error);
415 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
416 		    &sa6->sin6_addr)) != 0)
417 			return (error);
418 		sx_xlock(&in6_control_sx);
419 		control_locked = true;
420 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
421 	} else
422 		ia = NULL;
423 
424 	switch (cmd) {
425 	case SIOCSIFADDR_IN6:
426 	case SIOCSIFDSTADDR_IN6:
427 	case SIOCSIFNETMASK_IN6:
428 		/*
429 		 * Since IPv6 allows a node to assign multiple addresses
430 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
431 		 */
432 		/* we decided to obsolete this command (20000704) */
433 		error = EINVAL;
434 		goto out;
435 
436 	case SIOCDIFADDR_IN6:
437 		/*
438 		 * for IPv4, we look for existing in_ifaddr here to allow
439 		 * "ifconfig if0 delete" to remove the first IPv4 address on
440 		 * the interface.  For IPv6, as the spec allows multiple
441 		 * interface address from the day one, we consider "remove the
442 		 * first one" semantics to be not preferable.
443 		 */
444 		if (ia == NULL) {
445 			error = EADDRNOTAVAIL;
446 			goto out;
447 		}
448 		/* FALLTHROUGH */
449 	case SIOCAIFADDR_IN6:
450 		/*
451 		 * We always require users to specify a valid IPv6 address for
452 		 * the corresponding operation.
453 		 */
454 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
455 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
456 			error = EAFNOSUPPORT;
457 			goto out;
458 		}
459 
460 		if (td != NULL) {
461 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
462 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
463 			if (error)
464 				goto out;
465 		}
466 		/* FALLTHROUGH */
467 	case SIOCGIFSTAT_IN6:
468 	case SIOCGIFSTAT_ICMP6:
469 		if (ifp->if_afdata[AF_INET6] == NULL) {
470 			error = EPFNOSUPPORT;
471 			goto out;
472 		}
473 		break;
474 
475 	case SIOCGIFADDR_IN6:
476 		/* This interface is basically deprecated. use SIOCGIFCONF. */
477 		/* FALLTHROUGH */
478 	case SIOCGIFAFLAG_IN6:
479 	case SIOCGIFNETMASK_IN6:
480 	case SIOCGIFDSTADDR_IN6:
481 	case SIOCGIFALIFETIME_IN6:
482 		/* must think again about its semantics */
483 		if (ia == NULL) {
484 			error = EADDRNOTAVAIL;
485 			goto out;
486 		}
487 		break;
488 	}
489 
490 	switch (cmd) {
491 	case SIOCGIFADDR_IN6:
492 		ifr->ifr_addr = ia->ia_addr;
493 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
494 			goto out;
495 		break;
496 
497 	case SIOCGIFDSTADDR_IN6:
498 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
499 			error = EINVAL;
500 			goto out;
501 		}
502 		ifr->ifr_dstaddr = ia->ia_dstaddr;
503 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
504 			goto out;
505 		break;
506 
507 	case SIOCGIFNETMASK_IN6:
508 		ifr->ifr_addr = ia->ia_prefixmask;
509 		break;
510 
511 	case SIOCGIFAFLAG_IN6:
512 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
513 		break;
514 
515 	case SIOCGIFSTAT_IN6:
516 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
517 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
518 		    &ifr->ifr_ifru.ifru_stat,
519 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
520 		break;
521 
522 	case SIOCGIFSTAT_ICMP6:
523 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
524 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
525 		    &ifr->ifr_ifru.ifru_icmp6stat,
526 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
527 		break;
528 
529 	case SIOCGIFALIFETIME_IN6:
530 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
531 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
532 			time_t maxexpire;
533 			struct in6_addrlifetime *retlt =
534 			    &ifr->ifr_ifru.ifru_lifetime;
535 
536 			/*
537 			 * XXX: adjust expiration time assuming time_t is
538 			 * signed.
539 			 */
540 			maxexpire = (-1) &
541 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
542 			if (ia->ia6_lifetime.ia6t_vltime <
543 			    maxexpire - ia->ia6_updatetime) {
544 				retlt->ia6t_expire = ia->ia6_updatetime +
545 				    ia->ia6_lifetime.ia6t_vltime;
546 			} else
547 				retlt->ia6t_expire = maxexpire;
548 		}
549 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
550 			time_t maxexpire;
551 			struct in6_addrlifetime *retlt =
552 			    &ifr->ifr_ifru.ifru_lifetime;
553 
554 			/*
555 			 * XXX: adjust expiration time assuming time_t is
556 			 * signed.
557 			 */
558 			maxexpire = (-1) &
559 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
560 			if (ia->ia6_lifetime.ia6t_pltime <
561 			    maxexpire - ia->ia6_updatetime) {
562 				retlt->ia6t_preferred = ia->ia6_updatetime +
563 				    ia->ia6_lifetime.ia6t_pltime;
564 			} else
565 				retlt->ia6t_preferred = maxexpire;
566 		}
567 		break;
568 
569 	case SIOCAIFADDR_IN6:
570 		error = in6_addifaddr(ifp, ifra, ia);
571 		ia = NULL;
572 		break;
573 
574 	case SIOCDIFADDR_IN6:
575 		in6_purgeifaddr(ia);
576 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
577 		    IFADDR_EVENT_DEL);
578 		break;
579 
580 	default:
581 		if (ifp->if_ioctl == NULL) {
582 			error = EOPNOTSUPP;
583 			goto out;
584 		}
585 		error = (*ifp->if_ioctl)(ifp, cmd, data);
586 		goto out;
587 	}
588 
589 	error = 0;
590 out:
591 	if (control_locked)
592 		sx_xunlock(&in6_control_sx);
593 
594 	if (ia != NULL)
595 		ifa_free(&ia->ia_ifa);
596 	return (error);
597 }
598 
599 static struct in6_multi_mship *
600 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
601     int *errorp, int delay)
602 {
603 	struct in6_multi_mship *imm;
604 	int error;
605 
606 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
607 	if (imm == NULL) {
608 		*errorp = ENOBUFS;
609 		return (NULL);
610 	}
611 
612 	delay = (delay * MLD_FASTHZ) / hz;
613 
614 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
615 	if (error) {
616 		*errorp = error;
617 		free(imm, M_IP6MADDR);
618 		return (NULL);
619 	}
620 
621 	return (imm);
622 }
623 
624 static int
625 in6_solicited_node_maddr(struct in6_addr *maddr,
626     struct ifnet *ifp, const struct in6_addr *base)
627 {
628 	int error;
629 
630 	bzero(maddr, sizeof(struct in6_addr));
631 	maddr->s6_addr32[0] = IPV6_ADDR_INT32_MLL;
632 	maddr->s6_addr32[2] = htonl(1);
633 	maddr->s6_addr32[3] = base->s6_addr32[3];
634 	maddr->s6_addr8[12] = 0xff;
635 	if ((error = in6_setscope(maddr, ifp, NULL)) != 0) {
636 		/* XXX: should not happen */
637 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
638 	}
639 
640 	return error;
641 }
642 
643 /*
644  * Join necessary multicast groups.  Factored out from in6_update_ifa().
645  * This entire work should only be done once, for the default FIB.
646  */
647 static int
648 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
649     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
650 {
651 	char ip6buf[INET6_ADDRSTRLEN];
652 	struct in6_addr mltaddr;
653 	struct in6_multi_mship *imm;
654 	int delay, error;
655 
656 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
657 
658 	/* Join solicited multicast addr for new host id. */
659 	if ((error = in6_solicited_node_maddr(&mltaddr, ifp,
660 	    &ifra->ifra_addr.sin6_addr)) != 0)
661 		goto cleanup;
662 	delay = error = 0;
663 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
664 		/*
665 		 * We need a random delay for DAD on the address being
666 		 * configured.  It also means delaying transmission of the
667 		 * corresponding MLD report to avoid report collision.
668 		 * [RFC 4861, Section 6.3.7]
669 		 */
670 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
671 	}
672 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
673 	if (imm == NULL) {
674 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
675 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
676 		    if_name(ifp), error));
677 		goto cleanup;
678 	}
679 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
680 	*in6m_sol = imm->i6mm_maddr;
681 
682 	/*
683 	 * Join link-local all-nodes address.
684 	 */
685 	mltaddr = in6addr_linklocal_allnodes;
686 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
687 		goto cleanup; /* XXX: should not fail */
688 
689 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
690 	if (imm == NULL) {
691 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
692 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
693 		    if_name(ifp), error));
694 		goto cleanup;
695 	}
696 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
697 
698 	/*
699 	 * Join node information group address.
700 	 */
701 	delay = 0;
702 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
703 		/*
704 		 * The spec does not say anything about delay for this group,
705 		 * but the same logic should apply.
706 		 */
707 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
708 	}
709 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
710 		/* XXX jinmei */
711 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
712 		if (imm == NULL)
713 			nd6log((LOG_WARNING,
714 			    "%s: in6_joingroup failed for %s on %s "
715 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
716 			    &mltaddr), if_name(ifp), error));
717 			/* XXX not very fatal, go on... */
718 		else
719 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
720 	}
721 	if (V_icmp6_nodeinfo_oldmcprefix &&
722 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
723 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
724 		if (imm == NULL)
725 			nd6log((LOG_WARNING,
726 			    "%s: in6_joingroup failed for %s on %s "
727 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
728 			    &mltaddr), if_name(ifp), error));
729 			/* XXX not very fatal, go on... */
730 		else
731 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
732 	}
733 
734 	/*
735 	 * Join interface-local all-nodes address.
736 	 * (ff01::1%ifN, and ff01::%ifN/32)
737 	 */
738 	mltaddr = in6addr_nodelocal_allnodes;
739 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
740 		goto cleanup; /* XXX: should not fail */
741 
742 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
743 	if (imm == NULL) {
744 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
745 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
746 		    &mltaddr), if_name(ifp), error));
747 		goto cleanup;
748 	}
749 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
750 
751 cleanup:
752 	return (error);
753 }
754 
755 /*
756  * Update parameters of an IPv6 interface address.
757  * If necessary, a new entry is created and linked into address chains.
758  * This function is separated from in6_control().
759  */
760 int
761 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
762     struct in6_ifaddr *ia, int flags)
763 {
764 	int error, hostIsNew = 0;
765 
766 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
767 		return (error);
768 
769 	if (ia == NULL) {
770 		hostIsNew = 1;
771 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
772 			return (ENOBUFS);
773 	}
774 
775 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
776 	if (error != 0) {
777 		if (hostIsNew != 0) {
778 			in6_unlink_ifa(ia, ifp);
779 			ifa_free(&ia->ia_ifa);
780 		}
781 		return (error);
782 	}
783 
784 	if (hostIsNew)
785 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
786 
787 	return (error);
788 }
789 
790 /*
791  * Fill in basic IPv6 address request info.
792  */
793 void
794 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
795     const struct in6_addr *mask)
796 {
797 
798 	memset(ifra, 0, sizeof(struct in6_aliasreq));
799 
800 	ifra->ifra_addr.sin6_family = AF_INET6;
801 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
802 	if (addr != NULL)
803 		ifra->ifra_addr.sin6_addr = *addr;
804 
805 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
806 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
807 	if (mask != NULL)
808 		ifra->ifra_prefixmask.sin6_addr = *mask;
809 }
810 
811 static int
812 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
813     struct in6_ifaddr *ia, int flags)
814 {
815 	int plen = -1;
816 	struct sockaddr_in6 dst6;
817 	struct in6_addrlifetime *lt;
818 	char ip6buf[INET6_ADDRSTRLEN];
819 
820 	/* Validate parameters */
821 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
822 		return (EINVAL);
823 
824 	/*
825 	 * The destination address for a p2p link must have a family
826 	 * of AF_UNSPEC or AF_INET6.
827 	 */
828 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
829 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
830 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
831 		return (EAFNOSUPPORT);
832 
833 	/*
834 	 * Validate address
835 	 */
836 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
837 	    ifra->ifra_addr.sin6_family != AF_INET6)
838 		return (EINVAL);
839 
840 	/*
841 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
842 	 * does not carry fields other than sin6_len.
843 	 */
844 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
845 		return (EINVAL);
846 	/*
847 	 * Because the IPv6 address architecture is classless, we require
848 	 * users to specify a (non 0) prefix length (mask) for a new address.
849 	 * We also require the prefix (when specified) mask is valid, and thus
850 	 * reject a non-consecutive mask.
851 	 */
852 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
853 		return (EINVAL);
854 	if (ifra->ifra_prefixmask.sin6_len != 0) {
855 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
856 		    (u_char *)&ifra->ifra_prefixmask +
857 		    ifra->ifra_prefixmask.sin6_len);
858 		if (plen <= 0)
859 			return (EINVAL);
860 	} else {
861 		/*
862 		 * In this case, ia must not be NULL.  We just use its prefix
863 		 * length.
864 		 */
865 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
866 	}
867 	/*
868 	 * If the destination address on a p2p interface is specified,
869 	 * and the address is a scoped one, validate/set the scope
870 	 * zone identifier.
871 	 */
872 	dst6 = ifra->ifra_dstaddr;
873 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
874 	    (dst6.sin6_family == AF_INET6)) {
875 		struct in6_addr in6_tmp;
876 		u_int32_t zoneid;
877 
878 		in6_tmp = dst6.sin6_addr;
879 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
880 			return (EINVAL); /* XXX: should be impossible */
881 
882 		if (dst6.sin6_scope_id != 0) {
883 			if (dst6.sin6_scope_id != zoneid)
884 				return (EINVAL);
885 		} else		/* user omit to specify the ID. */
886 			dst6.sin6_scope_id = zoneid;
887 
888 		/* convert into the internal form */
889 		if (sa6_embedscope(&dst6, 0))
890 			return (EINVAL); /* XXX: should be impossible */
891 	}
892 	/* Modify original ifra_dstaddr to reflect changes */
893 	ifra->ifra_dstaddr = dst6;
894 
895 	/*
896 	 * The destination address can be specified only for a p2p or a
897 	 * loopback interface.  If specified, the corresponding prefix length
898 	 * must be 128.
899 	 */
900 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
901 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
902 			/* XXX: noisy message */
903 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
904 			    "be specified for a p2p or a loopback IF only\n"));
905 			return (EINVAL);
906 		}
907 		if (plen != 128) {
908 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
909 			    "be 128 when dstaddr is specified\n"));
910 			return (EINVAL);
911 		}
912 	}
913 	/* lifetime consistency check */
914 	lt = &ifra->ifra_lifetime;
915 	if (lt->ia6t_pltime > lt->ia6t_vltime)
916 		return (EINVAL);
917 	if (lt->ia6t_vltime == 0) {
918 		/*
919 		 * the following log might be noisy, but this is a typical
920 		 * configuration mistake or a tool's bug.
921 		 */
922 		nd6log((LOG_INFO,
923 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
924 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
925 
926 		if (ia == NULL)
927 			return (0); /* there's nothing to do */
928 	}
929 
930 	/* Check prefix mask */
931 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
932 		/*
933 		 * We prohibit changing the prefix length of an existing
934 		 * address, because
935 		 * + such an operation should be rare in IPv6, and
936 		 * + the operation would confuse prefix management.
937 		 */
938 		if (ia->ia_prefixmask.sin6_len != 0 &&
939 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
940 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
941 			    "of an existing %s address should not be changed\n",
942 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
943 
944 			return (EINVAL);
945 		}
946 	}
947 
948 	return (0);
949 }
950 
951 /*
952  * Allocate a new ifaddr and link it into chains.
953  */
954 static struct in6_ifaddr *
955 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
956 {
957 	struct in6_ifaddr *ia;
958 
959 	/*
960 	 * When in6_alloc_ifa() is called in a process of a received
961 	 * RA, it is called under an interrupt context.  So, we should
962 	 * call malloc with M_NOWAIT.
963 	 */
964 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
965 	if (ia == NULL)
966 		return (NULL);
967 	LIST_INIT(&ia->ia6_memberships);
968 	/* Initialize the address and masks, and put time stamp */
969 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
970 	ia->ia_addr.sin6_family = AF_INET6;
971 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
972 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
973 	ia->ia_addr = ifra->ifra_addr;
974 	ia->ia6_createtime = time_uptime;
975 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
976 		/*
977 		 * Some functions expect that ifa_dstaddr is not
978 		 * NULL for p2p interfaces.
979 		 */
980 		ia->ia_ifa.ifa_dstaddr =
981 		    (struct sockaddr *)&ia->ia_dstaddr;
982 	} else {
983 		ia->ia_ifa.ifa_dstaddr = NULL;
984 	}
985 
986 	/* set prefix mask if any */
987 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
988 	if (ifra->ifra_prefixmask.sin6_len != 0) {
989 		ia->ia_prefixmask.sin6_family = AF_INET6;
990 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
991 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
992 	}
993 
994 	ia->ia_ifp = ifp;
995 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
996 	IF_ADDR_WLOCK(ifp);
997 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
998 	IF_ADDR_WUNLOCK(ifp);
999 
1000 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1001 	IN6_IFADDR_WLOCK();
1002 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1003 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1004 	IN6_IFADDR_WUNLOCK();
1005 
1006 	return (ia);
1007 }
1008 
1009 /*
1010  * Update/configure interface address parameters:
1011  *
1012  * 1) Update lifetime
1013  * 2) Update interface metric ad flags
1014  * 3) Notify other subsystems
1015  */
1016 static int
1017 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1018     struct in6_ifaddr *ia, int hostIsNew, int flags)
1019 {
1020 	int error;
1021 
1022 	/* update timestamp */
1023 	ia->ia6_updatetime = time_uptime;
1024 
1025 	/*
1026 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1027 	 * to see if the address is deprecated or invalidated, but initialize
1028 	 * these members for applications.
1029 	 */
1030 	ia->ia6_lifetime = ifra->ifra_lifetime;
1031 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1032 		ia->ia6_lifetime.ia6t_expire =
1033 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1034 	} else
1035 		ia->ia6_lifetime.ia6t_expire = 0;
1036 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1037 		ia->ia6_lifetime.ia6t_preferred =
1038 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1039 	} else
1040 		ia->ia6_lifetime.ia6t_preferred = 0;
1041 
1042 	/*
1043 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1044 	 * userland, make it deprecated.
1045 	 */
1046 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1047 		ia->ia6_lifetime.ia6t_pltime = 0;
1048 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1049 	}
1050 
1051 	/*
1052 	 * configure address flags.
1053 	 */
1054 	ia->ia6_flags = ifra->ifra_flags;
1055 
1056 	/*
1057 	 * Make the address tentative before joining multicast addresses,
1058 	 * so that corresponding MLD responses would not have a tentative
1059 	 * source address.
1060 	 */
1061 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1062 
1063 	/*
1064 	 * DAD should be performed for an new address or addresses on
1065 	 * an interface with ND6_IFF_IFDISABLED.
1066 	 */
1067 	if (in6if_do_dad(ifp) &&
1068 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1069 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1070 
1071 	/* notify other subsystems */
1072 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1073 
1074 	return (error);
1075 }
1076 
1077 /*
1078  * Do link-level ifa job:
1079  * 1) Add lle entry for added address
1080  * 2) Notifies routing socket users about new address
1081  * 3) join appropriate multicast group
1082  * 4) start DAD if enabled
1083  */
1084 static int
1085 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1086     struct in6_ifaddr *ia, int flags)
1087 {
1088 	struct in6_multi *in6m_sol;
1089 	int error = 0;
1090 
1091 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1092 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1093 		in6_purgeaddr(&ia->ia_ifa);
1094 		ifa_free(&ia->ia_ifa);
1095 		return (error);
1096 	}
1097 
1098 	/* Join necessary multicast groups. */
1099 	in6m_sol = NULL;
1100 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1101 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1102 		if (error != 0) {
1103 			in6_purgeaddr(&ia->ia_ifa);
1104 			ifa_free(&ia->ia_ifa);
1105 			return (error);
1106 		}
1107 	}
1108 
1109 	/* Perform DAD, if the address is TENTATIVE. */
1110 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1111 		int delay, mindelay, maxdelay;
1112 
1113 		delay = 0;
1114 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1115 			/*
1116 			 * We need to impose a delay before sending an NS
1117 			 * for DAD.  Check if we also needed a delay for the
1118 			 * corresponding MLD message.  If we did, the delay
1119 			 * should be larger than the MLD delay (this could be
1120 			 * relaxed a bit, but this simple logic is at least
1121 			 * safe).
1122 			 * XXX: Break data hiding guidelines and look at
1123 			 * state for the solicited multicast group.
1124 			 */
1125 			mindelay = 0;
1126 			if (in6m_sol != NULL &&
1127 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1128 				mindelay = in6m_sol->in6m_timer;
1129 			}
1130 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1131 			if (maxdelay - mindelay == 0)
1132 				delay = 0;
1133 			else {
1134 				delay =
1135 				    (arc4random() % (maxdelay - mindelay)) +
1136 				    mindelay;
1137 			}
1138 		}
1139 		nd6_dad_start((struct ifaddr *)ia, delay);
1140 	}
1141 
1142 	in6_newaddrmsg(ia, RTM_ADD);
1143 	ifa_free(&ia->ia_ifa);
1144 	return (error);
1145 }
1146 
1147 /*
1148  * Adds or deletes interface route for p2p ifa.
1149  * Returns 0 on success or errno.
1150  */
1151 static int
1152 in6_handle_dstaddr_rtrequest(int cmd, struct in6_ifaddr *ia)
1153 {
1154 	struct epoch_tracker et;
1155 	struct ifaddr *ifa = &ia->ia_ifa;
1156 	int error;
1157 
1158 	/* Prepare gateway */
1159 	struct sockaddr_dl_short sdl = {
1160 		.sdl_family = AF_LINK,
1161 		.sdl_len = sizeof(struct sockaddr_dl_short),
1162 		.sdl_type = ifa->ifa_ifp->if_type,
1163 		.sdl_index = ifa->ifa_ifp->if_index,
1164 	};
1165 
1166 	struct sockaddr_in6 dst = {
1167 		.sin6_family = AF_INET6,
1168 		.sin6_len = sizeof(struct sockaddr_in6),
1169 		.sin6_addr = ia->ia_dstaddr.sin6_addr,
1170 	};
1171 
1172 	struct rt_addrinfo info = {
1173 		.rti_ifa = ifa,
1174 		.rti_ifp = ifa->ifa_ifp,
1175 		.rti_flags = RTF_PINNED | RTF_HOST,
1176 		.rti_info = {
1177 			[RTAX_DST] = (struct sockaddr *)&dst,
1178 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
1179 		},
1180 	};
1181 	/* Don't set additional per-gw filters on removal */
1182 
1183 	NET_EPOCH_ENTER(et);
1184 	error = rib_handle_ifaddr_info(ifa->ifa_ifp->if_fib, cmd, &info);
1185 	NET_EPOCH_EXIT(et);
1186 
1187 	return (error);
1188 }
1189 
1190 static bool
1191 ifa_is_p2p(struct in6_ifaddr *ia)
1192 {
1193 	int plen;
1194 
1195 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1196 
1197 	if ((plen == 128) && (ia->ia_dstaddr.sin6_family == AF_INET6) &&
1198 	    !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &ia->ia_dstaddr.sin6_addr))
1199 		return (true);
1200 
1201 	return (false);
1202 }
1203 
1204 int
1205 in6_addifaddr(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia)
1206 {
1207 	struct nd_prefixctl pr0;
1208 	struct nd_prefix *pr;
1209 	int carp_attached = 0;
1210 	int error;
1211 
1212 	/*
1213 	 * first, make or update the interface address structure,
1214 	 * and link it to the list.
1215 	 */
1216 	if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
1217 		goto out;
1218 	if (ia != NULL) {
1219 		if (ia->ia_ifa.ifa_carp)
1220 			(*carp_detach_p)(&ia->ia_ifa, true);
1221 		ifa_free(&ia->ia_ifa);
1222 	}
1223 	if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr)) == NULL) {
1224 		/*
1225 		 * this can happen when the user specify the 0 valid
1226 		 * lifetime.
1227 		 */
1228 		return (0);
1229 	}
1230 
1231 	if (ifra->ifra_vhid > 0) {
1232 		if (carp_attach_p != NULL)
1233 			error = (*carp_attach_p)(&ia->ia_ifa,
1234 			    ifra->ifra_vhid);
1235 		else
1236 			error = EPROTONOSUPPORT;
1237 		if (error)
1238 			goto out;
1239 		else
1240 			carp_attached = 1;
1241 	}
1242 
1243 	/*
1244 	 * then, make the prefix on-link on the interface.
1245 	 * XXX: we'd rather create the prefix before the address, but
1246 	 * we need at least one address to install the corresponding
1247 	 * interface route, so we configure the address first.
1248 	 */
1249 
1250 	/*
1251 	 * convert mask to prefix length (prefixmask has already
1252 	 * been validated in in6_update_ifa().
1253 	 */
1254 	bzero(&pr0, sizeof(pr0));
1255 	pr0.ndpr_ifp = ifp;
1256 	pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1257 	    NULL);
1258 	if (pr0.ndpr_plen == 128) {
1259 		/* we don't need to install a host route. */
1260 		goto aifaddr_out;
1261 	}
1262 	pr0.ndpr_prefix = ifra->ifra_addr;
1263 	/* apply the mask for safety. */
1264 	IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
1265 	    &ifra->ifra_prefixmask.sin6_addr);
1266 
1267 	/*
1268 	 * XXX: since we don't have an API to set prefix (not address)
1269 	 * lifetimes, we just use the same lifetimes as addresses.
1270 	 * The (temporarily) installed lifetimes can be overridden by
1271 	 * later advertised RAs (when accept_rtadv is non 0), which is
1272 	 * an intended behavior.
1273 	 */
1274 	pr0.ndpr_raf_onlink = 1; /* should be configurable? */
1275 	pr0.ndpr_raf_auto =
1276 	    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
1277 	pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
1278 	pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
1279 
1280 	/* add the prefix if not yet. */
1281 	if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
1282 		/*
1283 		 * nd6_prelist_add will install the corresponding
1284 		 * interface route.
1285 		 */
1286 		if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
1287 			if (carp_attached)
1288 				(*carp_detach_p)(&ia->ia_ifa, false);
1289 			goto out;
1290 		}
1291 	}
1292 
1293 	/* relate the address to the prefix */
1294 	if (ia->ia6_ndpr == NULL) {
1295 		ia->ia6_ndpr = pr;
1296 		pr->ndpr_addrcnt++;
1297 
1298 		/*
1299 		 * If this is the first autoconf address from the
1300 		 * prefix, create a temporary address as well
1301 		 * (when required).
1302 		 */
1303 		if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
1304 		    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
1305 			int e;
1306 			if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
1307 				log(LOG_NOTICE, "in6_control: failed "
1308 				    "to create a temporary address, "
1309 				    "errno=%d\n", e);
1310 			}
1311 		}
1312 	}
1313 	nd6_prefix_rele(pr);
1314 
1315 	/*
1316 	 * this might affect the status of autoconfigured addresses,
1317 	 * that is, this address might make other addresses detached.
1318 	 */
1319 	pfxlist_onlink_check();
1320 
1321 aifaddr_out:
1322 	/*
1323 	 * Try to clear the flag when a new IPv6 address is added
1324 	 * onto an IFDISABLED interface and it succeeds.
1325 	 */
1326 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1327 		struct in6_ndireq nd;
1328 
1329 		memset(&nd, 0, sizeof(nd));
1330 		nd.ndi.flags = ND_IFINFO(ifp)->flags;
1331 		nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
1332 		if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
1333 			log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
1334 			    "SIOCSIFINFO_FLAGS for -ifdisabled "
1335 			    "failed.");
1336 		/*
1337 		 * Ignore failure of clearing the flag intentionally.
1338 		 * The failure means address duplication was detected.
1339 		 */
1340 	}
1341 	error = 0;
1342 
1343 out:
1344 	if (ia != NULL)
1345 		ifa_free(&ia->ia_ifa);
1346 	return (error);
1347 }
1348 
1349 void
1350 in6_purgeaddr(struct ifaddr *ifa)
1351 {
1352 	struct ifnet *ifp = ifa->ifa_ifp;
1353 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1354 	struct in6_multi_mship *imm;
1355 	int error;
1356 
1357 	if (ifa->ifa_carp)
1358 		(*carp_detach_p)(ifa, false);
1359 
1360 	/*
1361 	 * Remove the loopback route to the interface address.
1362 	 * The check for the current setting of "nd6_useloopback"
1363 	 * is not needed.
1364 	 */
1365 	if (ia->ia_flags & IFA_RTSELF) {
1366 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1367 		    (struct sockaddr *)&ia->ia_addr);
1368 		if (error == 0)
1369 			ia->ia_flags &= ~IFA_RTSELF;
1370 	}
1371 
1372 	/* stop DAD processing */
1373 	nd6_dad_stop(ifa);
1374 
1375 	/* Leave multicast groups. */
1376 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1377 		LIST_REMOVE(imm, i6mm_chain);
1378 		if (imm->i6mm_maddr != NULL)
1379 			in6_leavegroup(imm->i6mm_maddr, NULL);
1380 		free(imm, M_IP6MADDR);
1381 	}
1382 	/* Check if we need to remove p2p route */
1383 	if ((ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1384 		error = in6_handle_dstaddr_rtrequest(RTM_DELETE, ia);
1385 		if (error != 0)
1386 			log(LOG_INFO, "%s: err=%d, destination address delete "
1387 			    "failed\n", __func__, error);
1388 		ia->ia_flags &= ~IFA_ROUTE;
1389 	}
1390 
1391 	in6_newaddrmsg(ia, RTM_DELETE);
1392 	in6_unlink_ifa(ia, ifp);
1393 }
1394 
1395 /*
1396  * Removes @ia from the corresponding interfaces and unlinks corresponding
1397  *  prefix if no addresses are using it anymore.
1398  */
1399 void
1400 in6_purgeifaddr(struct in6_ifaddr *ia)
1401 {
1402 	struct nd_prefix *pr;
1403 
1404 	/*
1405 	 * If the address being deleted is the only one that owns
1406 	 * the corresponding prefix, expire the prefix as well.
1407 	 * XXX: theoretically, we don't have to worry about such
1408 	 * relationship, since we separate the address management
1409 	 * and the prefix management.  We do this, however, to provide
1410 	 * as much backward compatibility as possible in terms of
1411 	 * the ioctl operation.
1412 	 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
1413 	 */
1414 	pr = ia->ia6_ndpr;
1415 	in6_purgeaddr(&ia->ia_ifa);
1416 	if (pr != NULL && pr->ndpr_addrcnt == 0) {
1417 		ND6_WLOCK();
1418 		nd6_prefix_unlink(pr, NULL);
1419 		ND6_WUNLOCK();
1420 		nd6_prefix_del(pr);
1421 	}
1422 }
1423 
1424 
1425 static void
1426 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1427 {
1428 	char ip6buf[INET6_ADDRSTRLEN];
1429 	int remove_lle;
1430 
1431 	IF_ADDR_WLOCK(ifp);
1432 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1433 	IF_ADDR_WUNLOCK(ifp);
1434 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1435 
1436 	/*
1437 	 * Defer the release of what might be the last reference to the
1438 	 * in6_ifaddr so that it can't be freed before the remainder of the
1439 	 * cleanup.
1440 	 */
1441 	IN6_IFADDR_WLOCK();
1442 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1443 	CK_LIST_REMOVE(ia, ia6_hash);
1444 	IN6_IFADDR_WUNLOCK();
1445 
1446 	/*
1447 	 * Release the reference to the base prefix.  There should be a
1448 	 * positive reference.
1449 	 */
1450 	remove_lle = 0;
1451 	if (ia->ia6_ndpr == NULL) {
1452 		nd6log((LOG_NOTICE,
1453 		    "in6_unlink_ifa: autoconf'ed address "
1454 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1455 	} else {
1456 		ia->ia6_ndpr->ndpr_addrcnt--;
1457 		/* Do not delete lles within prefix if refcont != 0 */
1458 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1459 			remove_lle = 1;
1460 		ia->ia6_ndpr = NULL;
1461 	}
1462 
1463 	nd6_rem_ifa_lle(ia, remove_lle);
1464 
1465 	/*
1466 	 * Also, if the address being removed is autoconf'ed, call
1467 	 * pfxlist_onlink_check() since the release might affect the status of
1468 	 * other (detached) addresses.
1469 	 */
1470 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1471 		pfxlist_onlink_check();
1472 	}
1473 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1474 }
1475 
1476 /*
1477  * Notifies other subsystems about address change/arrival:
1478  * 1) Notifies device handler on the first IPv6 address assignment
1479  * 2) Handle routing table changes for P2P links and route
1480  * 3) Handle routing table changes for address host route
1481  */
1482 static int
1483 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1484     struct in6_aliasreq *ifra, int hostIsNew)
1485 {
1486 	int	error = 0, ifacount = 0;
1487 	struct ifaddr *ifa;
1488 	struct sockaddr_in6 *pdst;
1489 	char ip6buf[INET6_ADDRSTRLEN];
1490 
1491 	/*
1492 	 * Give the interface a chance to initialize
1493 	 * if this is its first address,
1494 	 */
1495 	if (hostIsNew != 0) {
1496 		struct epoch_tracker et;
1497 
1498 		NET_EPOCH_ENTER(et);
1499 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1500 			if (ifa->ifa_addr->sa_family != AF_INET6)
1501 				continue;
1502 			ifacount++;
1503 		}
1504 		NET_EPOCH_EXIT(et);
1505 	}
1506 
1507 	if (ifacount <= 1 && ifp->if_ioctl) {
1508 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1509 		if (error)
1510 			goto done;
1511 	}
1512 
1513 	/*
1514 	 * If a new destination address is specified, scrub the old one and
1515 	 * install the new destination.  Note that the interface must be
1516 	 * p2p or loopback.
1517 	 */
1518 	pdst = &ifra->ifra_dstaddr;
1519 	if (pdst->sin6_family == AF_INET6 &&
1520 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1521 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1522 		    (in6_handle_dstaddr_rtrequest(RTM_DELETE, ia) != 0)) {
1523 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1524 			    "remove a route to the old destination: %s\n",
1525 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1526 			/* proceed anyway... */
1527 		} else
1528 			ia->ia_flags &= ~IFA_ROUTE;
1529 		ia->ia_dstaddr = *pdst;
1530 	}
1531 
1532 	/*
1533 	 * If a new destination address is specified for a point-to-point
1534 	 * interface, install a route to the destination as an interface
1535 	 * direct route.
1536 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1537 	 * interface that share the same destination.
1538 	 */
1539 	if (!(ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1540 		error = in6_handle_dstaddr_rtrequest(RTM_ADD, ia);
1541 		if (error)
1542 			goto done;
1543 		ia->ia_flags |= IFA_ROUTE;
1544 	}
1545 
1546 	/*
1547 	 * add a loopback route to self if not exists
1548 	 */
1549 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1550 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1551 		    (struct sockaddr *)&ia->ia_addr);
1552 		if (error == 0)
1553 			ia->ia_flags |= IFA_RTSELF;
1554 	}
1555 done:
1556 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1557 	    "Invoking IPv6 network device address event may sleep");
1558 
1559 	ifa_ref(&ia->ia_ifa);
1560 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1561 	    IFADDR_EVENT_ADD);
1562 	ifa_free(&ia->ia_ifa);
1563 
1564 	return (error);
1565 }
1566 
1567 /*
1568  * Find an IPv6 interface link-local address specific to an interface.
1569  * ifaddr is returned referenced.
1570  */
1571 struct in6_ifaddr *
1572 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1573 {
1574 	struct ifaddr *ifa;
1575 
1576 	NET_EPOCH_ASSERT();
1577 
1578 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1579 		if (ifa->ifa_addr->sa_family != AF_INET6)
1580 			continue;
1581 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1582 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1583 			    ignoreflags) != 0)
1584 				continue;
1585 			ifa_ref(ifa);
1586 			break;
1587 		}
1588 	}
1589 
1590 	return ((struct in6_ifaddr *)ifa);
1591 }
1592 
1593 /*
1594  * find the interface address corresponding to a given IPv6 address.
1595  * ifaddr is returned referenced if @referenced flag is set.
1596  */
1597 struct in6_ifaddr *
1598 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid, bool referenced)
1599 {
1600 	struct rm_priotracker in6_ifa_tracker;
1601 	struct in6_ifaddr *ia;
1602 
1603 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1604 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1605 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1606 			if (zoneid != 0 &&
1607 			    zoneid != ia->ia_addr.sin6_scope_id)
1608 				continue;
1609 			if (referenced)
1610 				ifa_ref(&ia->ia_ifa);
1611 			break;
1612 		}
1613 	}
1614 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1615 	return (ia);
1616 }
1617 
1618 /*
1619  * find the internet address corresponding to a given interface and address.
1620  * ifaddr is returned referenced.
1621  */
1622 struct in6_ifaddr *
1623 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1624 {
1625 	struct epoch_tracker et;
1626 	struct ifaddr *ifa;
1627 
1628 	NET_EPOCH_ENTER(et);
1629 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1630 		if (ifa->ifa_addr->sa_family != AF_INET6)
1631 			continue;
1632 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1633 			ifa_ref(ifa);
1634 			break;
1635 		}
1636 	}
1637 	NET_EPOCH_EXIT(et);
1638 
1639 	return ((struct in6_ifaddr *)ifa);
1640 }
1641 
1642 /*
1643  * Find a link-local scoped address on ifp and return it if any.
1644  */
1645 struct in6_ifaddr *
1646 in6ifa_llaonifp(struct ifnet *ifp)
1647 {
1648 	struct epoch_tracker et;
1649 	struct sockaddr_in6 *sin6;
1650 	struct ifaddr *ifa;
1651 
1652 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1653 		return (NULL);
1654 	NET_EPOCH_ENTER(et);
1655 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1656 		if (ifa->ifa_addr->sa_family != AF_INET6)
1657 			continue;
1658 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1659 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1660 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1661 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1662 			break;
1663 	}
1664 	NET_EPOCH_EXIT(et);
1665 
1666 	return ((struct in6_ifaddr *)ifa);
1667 }
1668 
1669 /*
1670  * Convert IP6 address to printable (loggable) representation. Caller
1671  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1672  */
1673 static char digits[] = "0123456789abcdef";
1674 char *
1675 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1676 {
1677 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1678 	char *cp;
1679 	const u_int16_t *a = (const u_int16_t *)addr;
1680 	const u_int8_t *d;
1681 	int dcolon = 0, zero = 0;
1682 
1683 	cp = ip6buf;
1684 
1685 	for (i = 0; i < 8; i++) {
1686 		if (*(a + i) == 0) {
1687 			cnt++;
1688 			if (cnt == 1)
1689 				idx = i;
1690 		}
1691 		else if (maxcnt < cnt) {
1692 			maxcnt = cnt;
1693 			index = idx;
1694 			cnt = 0;
1695 		}
1696 	}
1697 	if (maxcnt < cnt) {
1698 		maxcnt = cnt;
1699 		index = idx;
1700 	}
1701 
1702 	for (i = 0; i < 8; i++) {
1703 		if (dcolon == 1) {
1704 			if (*a == 0) {
1705 				if (i == 7)
1706 					*cp++ = ':';
1707 				a++;
1708 				continue;
1709 			} else
1710 				dcolon = 2;
1711 		}
1712 		if (*a == 0) {
1713 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1714 				if (i == 0)
1715 					*cp++ = ':';
1716 				*cp++ = ':';
1717 				dcolon = 1;
1718 			} else {
1719 				*cp++ = '0';
1720 				*cp++ = ':';
1721 			}
1722 			a++;
1723 			continue;
1724 		}
1725 		d = (const u_char *)a;
1726 		/* Try to eliminate leading zeros in printout like in :0001. */
1727 		zero = 1;
1728 		*cp = digits[*d >> 4];
1729 		if (*cp != '0') {
1730 			zero = 0;
1731 			cp++;
1732 		}
1733 		*cp = digits[*d++ & 0xf];
1734 		if (zero == 0 || (*cp != '0')) {
1735 			zero = 0;
1736 			cp++;
1737 		}
1738 		*cp = digits[*d >> 4];
1739 		if (zero == 0 || (*cp != '0')) {
1740 			zero = 0;
1741 			cp++;
1742 		}
1743 		*cp++ = digits[*d & 0xf];
1744 		*cp++ = ':';
1745 		a++;
1746 	}
1747 	*--cp = '\0';
1748 	return (ip6buf);
1749 }
1750 
1751 int
1752 in6_localaddr(struct in6_addr *in6)
1753 {
1754 	struct rm_priotracker in6_ifa_tracker;
1755 	struct in6_ifaddr *ia;
1756 
1757 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1758 		return 1;
1759 
1760 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1761 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1762 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1763 		    &ia->ia_prefixmask.sin6_addr)) {
1764 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1765 			return 1;
1766 		}
1767 	}
1768 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1769 
1770 	return (0);
1771 }
1772 
1773 /*
1774  * Return 1 if an internet address is for the local host and configured
1775  * on one of its interfaces.
1776  */
1777 int
1778 in6_localip(struct in6_addr *in6)
1779 {
1780 	struct rm_priotracker in6_ifa_tracker;
1781 	struct in6_ifaddr *ia;
1782 
1783 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1784 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1785 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1786 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1787 			return (1);
1788 		}
1789 	}
1790 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1791 	return (0);
1792 }
1793 
1794 /*
1795  * Like in6_localip(), but FIB-aware.
1796  */
1797 bool
1798 in6_localip_fib(struct in6_addr *in6, uint16_t fib)
1799 {
1800 	struct rm_priotracker in6_ifa_tracker;
1801 	struct in6_ifaddr *ia;
1802 
1803 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1804 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1805 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr) &&
1806 		    ia->ia_ifa.ifa_ifp->if_fib == fib) {
1807 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1808 			return (true);
1809 		}
1810 	}
1811 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1812 	return (false);
1813 }
1814 
1815 /*
1816  * Return 1 if an internet address is configured on an interface.
1817  */
1818 int
1819 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1820 {
1821 	struct in6_addr in6;
1822 	struct ifaddr *ifa;
1823 	struct in6_ifaddr *ia6;
1824 
1825 	NET_EPOCH_ASSERT();
1826 
1827 	in6 = *addr;
1828 	if (in6_clearscope(&in6))
1829 		return (0);
1830 	in6_setscope(&in6, ifp, NULL);
1831 
1832 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1833 		if (ifa->ifa_addr->sa_family != AF_INET6)
1834 			continue;
1835 		ia6 = (struct in6_ifaddr *)ifa;
1836 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1837 			return (1);
1838 	}
1839 
1840 	return (0);
1841 }
1842 
1843 int
1844 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1845 {
1846 	struct rm_priotracker in6_ifa_tracker;
1847 	struct in6_ifaddr *ia;
1848 
1849 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1850 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1851 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1852 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1853 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1854 				return (1); /* true */
1855 			}
1856 			break;
1857 		}
1858 	}
1859 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1860 
1861 	return (0);		/* false */
1862 }
1863 
1864 /*
1865  * return length of part which dst and src are equal
1866  * hard coding...
1867  */
1868 int
1869 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1870 {
1871 	int match = 0;
1872 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1873 	u_char *lim = s + 16, r;
1874 
1875 	while (s < lim)
1876 		if ((r = (*d++ ^ *s++)) != 0) {
1877 			while (r < 128) {
1878 				match++;
1879 				r <<= 1;
1880 			}
1881 			break;
1882 		} else
1883 			match += 8;
1884 	return match;
1885 }
1886 
1887 /* XXX: to be scope conscious */
1888 int
1889 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1890 {
1891 	int bytelen, bitlen;
1892 
1893 	/* sanity check */
1894 	if (0 > len || len > 128) {
1895 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1896 		    len);
1897 		return (0);
1898 	}
1899 
1900 	bytelen = len / 8;
1901 	bitlen = len % 8;
1902 
1903 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1904 		return (0);
1905 	if (bitlen != 0 &&
1906 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1907 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1908 		return (0);
1909 
1910 	return (1);
1911 }
1912 
1913 void
1914 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1915 {
1916 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1917 	int bytelen, bitlen, i;
1918 
1919 	/* sanity check */
1920 	if (0 > len || len > 128) {
1921 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1922 		    len);
1923 		return;
1924 	}
1925 
1926 	bzero(maskp, sizeof(*maskp));
1927 	bytelen = len / 8;
1928 	bitlen = len % 8;
1929 	for (i = 0; i < bytelen; i++)
1930 		maskp->s6_addr[i] = 0xff;
1931 	if (bitlen)
1932 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1933 }
1934 
1935 /*
1936  * return the best address out of the same scope. if no address was
1937  * found, return the first valid address from designated IF.
1938  */
1939 struct in6_ifaddr *
1940 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1941 {
1942 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1943 	struct ifaddr *ifa;
1944 	struct in6_ifaddr *besta = NULL;
1945 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1946 
1947 	NET_EPOCH_ASSERT();
1948 
1949 	dep[0] = dep[1] = NULL;
1950 
1951 	/*
1952 	 * We first look for addresses in the same scope.
1953 	 * If there is one, return it.
1954 	 * If two or more, return one which matches the dst longest.
1955 	 * If none, return one of global addresses assigned other ifs.
1956 	 */
1957 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1958 		if (ifa->ifa_addr->sa_family != AF_INET6)
1959 			continue;
1960 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1961 			continue; /* XXX: is there any case to allow anycast? */
1962 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1963 			continue; /* don't use this interface */
1964 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1965 			continue;
1966 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1967 			if (V_ip6_use_deprecated)
1968 				dep[0] = (struct in6_ifaddr *)ifa;
1969 			continue;
1970 		}
1971 
1972 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1973 			/*
1974 			 * call in6_matchlen() as few as possible
1975 			 */
1976 			if (besta) {
1977 				if (blen == -1)
1978 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1979 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1980 				if (tlen > blen) {
1981 					blen = tlen;
1982 					besta = (struct in6_ifaddr *)ifa;
1983 				}
1984 			} else
1985 				besta = (struct in6_ifaddr *)ifa;
1986 		}
1987 	}
1988 	if (besta)
1989 		return (besta);
1990 
1991 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1992 		if (ifa->ifa_addr->sa_family != AF_INET6)
1993 			continue;
1994 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1995 			continue; /* XXX: is there any case to allow anycast? */
1996 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1997 			continue; /* don't use this interface */
1998 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1999 			continue;
2000 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2001 			if (V_ip6_use_deprecated)
2002 				dep[1] = (struct in6_ifaddr *)ifa;
2003 			continue;
2004 		}
2005 
2006 		return (struct in6_ifaddr *)ifa;
2007 	}
2008 
2009 	/* use the last-resort values, that are, deprecated addresses */
2010 	if (dep[0])
2011 		return dep[0];
2012 	if (dep[1])
2013 		return dep[1];
2014 
2015 	return NULL;
2016 }
2017 
2018 /*
2019  * perform DAD when interface becomes IFF_UP.
2020  */
2021 void
2022 in6_if_up(struct ifnet *ifp)
2023 {
2024 	struct epoch_tracker et;
2025 	struct ifaddr *ifa;
2026 	struct in6_ifaddr *ia;
2027 
2028 	NET_EPOCH_ENTER(et);
2029 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2030 		if (ifa->ifa_addr->sa_family != AF_INET6)
2031 			continue;
2032 		ia = (struct in6_ifaddr *)ifa;
2033 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2034 			/*
2035 			 * The TENTATIVE flag was likely set by hand
2036 			 * beforehand, implicitly indicating the need for DAD.
2037 			 * We may be able to skip the random delay in this
2038 			 * case, but we impose delays just in case.
2039 			 */
2040 			nd6_dad_start(ifa,
2041 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2042 		}
2043 	}
2044 	NET_EPOCH_EXIT(et);
2045 
2046 	/*
2047 	 * special cases, like 6to4, are handled in in6_ifattach
2048 	 */
2049 	in6_ifattach(ifp, NULL);
2050 }
2051 
2052 static void
2053 in6_ifevent(void *arg __unused, struct ifnet *ifp, int event)
2054 {
2055 	if (event == IFNET_EVENT_UP)
2056 		in6_if_up(ifp);
2057 }
2058 
2059 static void
2060 in6_init(void *arg __unused)
2061 {
2062 	EVENTHANDLER_REGISTER(ifnet_event, in6_ifevent, NULL, EVENTHANDLER_PRI_ANY);
2063 }
2064 SYSINIT(in6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, in6_init, NULL);
2065 
2066 int
2067 in6if_do_dad(struct ifnet *ifp)
2068 {
2069 
2070 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2071 		return (0);
2072 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
2073 		return (0);
2074 	if ((ND_IFINFO(ifp)->flags &
2075 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
2076 		return (0);
2077 	return (1);
2078 }
2079 
2080 /*
2081  * Calculate max IPv6 MTU through all the interfaces and store it
2082  * to in6_maxmtu.
2083  */
2084 void
2085 in6_setmaxmtu(void)
2086 {
2087 	struct epoch_tracker et;
2088 	unsigned long maxmtu = 0;
2089 	struct ifnet *ifp;
2090 
2091 	NET_EPOCH_ENTER(et);
2092 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2093 		/* this function can be called during ifnet initialization */
2094 		if (!ifp->if_afdata[AF_INET6])
2095 			continue;
2096 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2097 		    IN6_LINKMTU(ifp) > maxmtu)
2098 			maxmtu = IN6_LINKMTU(ifp);
2099 	}
2100 	NET_EPOCH_EXIT(et);
2101 	if (maxmtu)	/* update only when maxmtu is positive */
2102 		V_in6_maxmtu = maxmtu;
2103 }
2104 
2105 /*
2106  * Provide the length of interface identifiers to be used for the link attached
2107  * to the given interface.  The length should be defined in "IPv6 over
2108  * xxx-link" document.  Note that address architecture might also define
2109  * the length for a particular set of address prefixes, regardless of the
2110  * link type.  As clarified in rfc2462bis, those two definitions should be
2111  * consistent, and those really are as of August 2004.
2112  */
2113 int
2114 in6_if2idlen(struct ifnet *ifp)
2115 {
2116 	switch (ifp->if_type) {
2117 	case IFT_ETHER:		/* RFC2464 */
2118 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2119 	case IFT_L2VLAN:	/* ditto */
2120 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2121 	case IFT_INFINIBAND:
2122 		return (64);
2123 	case IFT_PPP:		/* RFC2472 */
2124 		return (64);
2125 	case IFT_FRELAY:	/* RFC2590 */
2126 		return (64);
2127 	case IFT_IEEE1394:	/* RFC3146 */
2128 		return (64);
2129 	case IFT_GIF:
2130 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2131 	case IFT_LOOP:
2132 		return (64);	/* XXX: is this really correct? */
2133 	default:
2134 		/*
2135 		 * Unknown link type:
2136 		 * It might be controversial to use the today's common constant
2137 		 * of 64 for these cases unconditionally.  For full compliance,
2138 		 * we should return an error in this case.  On the other hand,
2139 		 * if we simply miss the standard for the link type or a new
2140 		 * standard is defined for a new link type, the IFID length
2141 		 * is very likely to be the common constant.  As a compromise,
2142 		 * we always use the constant, but make an explicit notice
2143 		 * indicating the "unknown" case.
2144 		 */
2145 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2146 		return (64);
2147 	}
2148 }
2149 
2150 struct in6_llentry {
2151 	struct llentry		base;
2152 };
2153 
2154 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2155 #define	IN6_LLTBL_HASH(k, h) \
2156 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2157 
2158 /*
2159  * Do actual deallocation of @lle.
2160  */
2161 static void
2162 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2163 {
2164 	struct llentry *lle;
2165 
2166 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2167 	LLE_LOCK_DESTROY(lle);
2168 	LLE_REQ_DESTROY(lle);
2169 	free(lle, M_LLTABLE);
2170 }
2171 
2172 /*
2173  * Called by LLE_FREE_LOCKED when number of references
2174  * drops to zero.
2175  */
2176 static void
2177 in6_lltable_destroy_lle(struct llentry *lle)
2178 {
2179 
2180 	LLE_WUNLOCK(lle);
2181 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2182 }
2183 
2184 static struct llentry *
2185 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2186 {
2187 	struct in6_llentry *lle;
2188 
2189 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2190 	if (lle == NULL)		/* NB: caller generates msg */
2191 		return NULL;
2192 
2193 	lle->base.r_l3addr.addr6 = *addr6;
2194 	lle->base.lle_refcnt = 1;
2195 	lle->base.lle_free = in6_lltable_destroy_lle;
2196 	LLE_LOCK_INIT(&lle->base);
2197 	LLE_REQ_INIT(&lle->base);
2198 	callout_init(&lle->base.lle_timer, 1);
2199 
2200 	return (&lle->base);
2201 }
2202 
2203 static int
2204 in6_lltable_match_prefix(const struct sockaddr *saddr,
2205     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2206 {
2207 	const struct in6_addr *addr, *mask, *lle_addr;
2208 
2209 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2210 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2211 	lle_addr = &lle->r_l3addr.addr6;
2212 
2213 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2214 		return (0);
2215 
2216 	if (lle->la_flags & LLE_IFADDR) {
2217 		/*
2218 		 * Delete LLE_IFADDR records IFF address & flag matches.
2219 		 * Note that addr is the interface address within prefix
2220 		 * being matched.
2221 		 */
2222 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2223 		    (flags & LLE_STATIC) != 0)
2224 			return (1);
2225 		return (0);
2226 	}
2227 
2228 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2229 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2230 		return (1);
2231 
2232 	return (0);
2233 }
2234 
2235 static void
2236 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2237 {
2238 	struct ifnet *ifp __diagused;
2239 
2240 	LLE_WLOCK_ASSERT(lle);
2241 	KASSERT(llt != NULL, ("lltable is NULL"));
2242 
2243 	/* Unlink entry from table */
2244 	if ((lle->la_flags & LLE_LINKED) != 0) {
2245 		ifp = llt->llt_ifp;
2246 		IF_AFDATA_WLOCK_ASSERT(ifp);
2247 		lltable_unlink_entry(llt, lle);
2248 	}
2249 
2250 	llentry_free(lle);
2251 }
2252 
2253 static int
2254 in6_lltable_rtcheck(struct ifnet *ifp,
2255 		    u_int flags,
2256 		    const struct sockaddr *l3addr)
2257 {
2258 	const struct sockaddr_in6 *sin6;
2259 	struct nhop_object *nh;
2260 	struct in6_addr dst;
2261 	uint32_t scopeid;
2262 	char ip6buf[INET6_ADDRSTRLEN];
2263 	int fibnum;
2264 
2265 	NET_EPOCH_ASSERT();
2266 	KASSERT(l3addr->sa_family == AF_INET6,
2267 	    ("sin_family %d", l3addr->sa_family));
2268 
2269 	sin6 = (const struct sockaddr_in6 *)l3addr;
2270 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2271 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2272 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2273 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2274 		struct ifaddr *ifa;
2275 		/*
2276 		 * Create an ND6 cache for an IPv6 neighbor
2277 		 * that is not covered by our own prefix.
2278 		 */
2279 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2280 		if (ifa != NULL) {
2281 			return 0;
2282 		}
2283 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2284 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2285 		return EINVAL;
2286 	}
2287 	return 0;
2288 }
2289 
2290 static inline uint32_t
2291 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2292 {
2293 
2294 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2295 }
2296 
2297 static uint32_t
2298 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2299 {
2300 
2301 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2302 }
2303 
2304 static void
2305 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2306 {
2307 	struct sockaddr_in6 *sin6;
2308 
2309 	sin6 = (struct sockaddr_in6 *)sa;
2310 	bzero(sin6, sizeof(*sin6));
2311 	sin6->sin6_family = AF_INET6;
2312 	sin6->sin6_len = sizeof(*sin6);
2313 	sin6->sin6_addr = lle->r_l3addr.addr6;
2314 }
2315 
2316 static inline struct llentry *
2317 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2318 {
2319 	struct llentry *lle;
2320 	struct llentries *lleh;
2321 	u_int hashidx;
2322 
2323 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2324 	lleh = &llt->lle_head[hashidx];
2325 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2326 		if (lle->la_flags & LLE_DELETED)
2327 			continue;
2328 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2329 			break;
2330 	}
2331 
2332 	return (lle);
2333 }
2334 
2335 static void
2336 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2337 {
2338 
2339 	lle->la_flags |= LLE_DELETED;
2340 
2341 	/* Leave the solicited multicast group. */
2342 	if ((lle->la_flags & LLE_PUB) != 0)
2343 		in6_leave_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2344 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2345 #ifdef DIAGNOSTIC
2346 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2347 #endif
2348 	llentry_free(lle);
2349 }
2350 
2351 static struct llentry *
2352 in6_lltable_alloc(struct lltable *llt, u_int flags,
2353 	const struct sockaddr *l3addr)
2354 {
2355 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2356 	struct ifnet *ifp = llt->llt_ifp;
2357 	struct llentry *lle;
2358 	char linkhdr[LLE_MAX_LINKHDR];
2359 	size_t linkhdrsize;
2360 	int lladdr_off;
2361 
2362 	KASSERT(l3addr->sa_family == AF_INET6,
2363 	    ("sin_family %d", l3addr->sa_family));
2364 
2365 	/*
2366 	 * A route that covers the given address must have
2367 	 * been installed 1st because we are doing a resolution,
2368 	 * verify this.
2369 	 */
2370 	if (!(flags & LLE_IFADDR) &&
2371 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2372 		return (NULL);
2373 
2374 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2375 	if (lle == NULL) {
2376 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2377 		return (NULL);
2378 	}
2379 	lle->la_flags = flags;
2380 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2381 		linkhdrsize = LLE_MAX_LINKHDR;
2382 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2383 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2384 			in6_lltable_free_entry(llt, lle);
2385 			return (NULL);
2386 		}
2387 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2388 		    lladdr_off);
2389 		lle->la_flags |= LLE_STATIC;
2390 	}
2391 
2392 	if ((lle->la_flags & LLE_STATIC) != 0)
2393 		lle->ln_state = ND6_LLINFO_REACHABLE;
2394 
2395 	return (lle);
2396 }
2397 
2398 static struct llentry *
2399 in6_lltable_lookup(struct lltable *llt, u_int flags,
2400 	const struct sockaddr *l3addr)
2401 {
2402 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2403 	int family = flags >> 16;
2404 	struct llentry *lle;
2405 
2406 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2407 	KASSERT(l3addr->sa_family == AF_INET6,
2408 	    ("sin_family %d", l3addr->sa_family));
2409 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2410 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2411 	    ("wrong lle request flags: %#x", flags));
2412 
2413 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2414 
2415 	if (__predict_false(family != AF_INET6))
2416 		lle = llentry_lookup_family(lle, family);
2417 
2418 	if (lle == NULL)
2419 		return (NULL);
2420 
2421 	if (flags & LLE_UNLOCKED)
2422 		return (lle);
2423 
2424 	if (flags & LLE_EXCLUSIVE)
2425 		LLE_WLOCK(lle);
2426 	else
2427 		LLE_RLOCK(lle);
2428 
2429 	/*
2430 	 * If the afdata lock is not held, the LLE may have been unlinked while
2431 	 * we were blocked on the LLE lock.  Check for this case.
2432 	 */
2433 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2434 		if (flags & LLE_EXCLUSIVE)
2435 			LLE_WUNLOCK(lle);
2436 		else
2437 			LLE_RUNLOCK(lle);
2438 		return (NULL);
2439 	}
2440 	return (lle);
2441 }
2442 
2443 static int
2444 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2445     struct sysctl_req *wr)
2446 {
2447 	struct ifnet *ifp = llt->llt_ifp;
2448 	/* XXX stack use */
2449 	struct {
2450 		struct rt_msghdr	rtm;
2451 		struct sockaddr_in6	sin6;
2452 		/*
2453 		 * ndp.c assumes that sdl is word aligned
2454 		 */
2455 #ifdef __LP64__
2456 		uint32_t		pad;
2457 #endif
2458 		struct sockaddr_dl	sdl;
2459 	} ndpc;
2460 	struct sockaddr_dl *sdl;
2461 	int error;
2462 
2463 	bzero(&ndpc, sizeof(ndpc));
2464 	/* skip deleted entries */
2465 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2466 		return (0);
2467 	/* Skip if jailed and not a valid IP of the prison. */
2468 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2469 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2470 		return (0);
2471 	/*
2472 	 * produce a msg made of:
2473 	 *  struct rt_msghdr;
2474 	 *  struct sockaddr_in6 (IPv6)
2475 	 *  struct sockaddr_dl;
2476 	 */
2477 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2478 	ndpc.rtm.rtm_version = RTM_VERSION;
2479 	ndpc.rtm.rtm_type = RTM_GET;
2480 	ndpc.rtm.rtm_flags = RTF_UP;
2481 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2482 	sa6_recoverscope(&ndpc.sin6);
2483 
2484 	/* publish */
2485 	if (lle->la_flags & LLE_PUB)
2486 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2487 
2488 	sdl = &ndpc.sdl;
2489 	sdl->sdl_family = AF_LINK;
2490 	sdl->sdl_len = sizeof(*sdl);
2491 	sdl->sdl_index = ifp->if_index;
2492 	sdl->sdl_type = ifp->if_type;
2493 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2494 		sdl->sdl_alen = ifp->if_addrlen;
2495 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2496 	} else {
2497 		sdl->sdl_alen = 0;
2498 		bzero(LLADDR(sdl), ifp->if_addrlen);
2499 	}
2500 	if (lle->la_expire != 0)
2501 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2502 		    lle->lle_remtime / hz + time_second - time_uptime;
2503 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2504 	if (lle->la_flags & LLE_STATIC)
2505 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2506 	if (lle->la_flags & LLE_IFADDR)
2507 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2508 	if (lle->ln_router != 0)
2509 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2510 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2511 	/* Store state in rmx_weight value */
2512 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2513 	ndpc.rtm.rtm_index = ifp->if_index;
2514 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2515 
2516 	return (error);
2517 }
2518 
2519 static void
2520 in6_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
2521 {
2522 	/* Join the solicited multicast group for dst. */
2523 	if ((lle->la_flags & LLE_PUB) == LLE_PUB)
2524 		in6_join_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2525 }
2526 
2527 static struct lltable *
2528 in6_lltattach(struct ifnet *ifp)
2529 {
2530 	struct lltable *llt;
2531 
2532 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2533 	llt->llt_af = AF_INET6;
2534 	llt->llt_ifp = ifp;
2535 
2536 	llt->llt_lookup = in6_lltable_lookup;
2537 	llt->llt_alloc_entry = in6_lltable_alloc;
2538 	llt->llt_delete_entry = in6_lltable_delete_entry;
2539 	llt->llt_dump_entry = in6_lltable_dump_entry;
2540 	llt->llt_hash = in6_lltable_hash;
2541 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2542 	llt->llt_free_entry = in6_lltable_free_entry;
2543 	llt->llt_match_prefix = in6_lltable_match_prefix;
2544 	llt->llt_mark_used = llentry_mark_used;
2545 	llt->llt_post_resolved = in6_lltable_post_resolved;
2546  	lltable_link(llt);
2547 
2548 	return (llt);
2549 }
2550 
2551 struct lltable *
2552 in6_lltable_get(struct ifnet *ifp)
2553 {
2554 	struct lltable *llt = NULL;
2555 
2556 	void *afdata_ptr = ifp->if_afdata[AF_INET6];
2557 	if (afdata_ptr != NULL)
2558 		llt = ((struct in6_ifextra *)afdata_ptr)->lltable;
2559 	return (llt);
2560 }
2561 
2562 void *
2563 in6_domifattach(struct ifnet *ifp)
2564 {
2565 	struct in6_ifextra *ext;
2566 
2567 	/* There are not IPv6-capable interfaces. */
2568 	switch (ifp->if_type) {
2569 	case IFT_PFLOG:
2570 	case IFT_PFSYNC:
2571 	case IFT_USB:
2572 		return (NULL);
2573 	}
2574 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2575 	bzero(ext, sizeof(*ext));
2576 
2577 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2578 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2579 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2580 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2581 
2582 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2583 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2584 	    M_WAITOK);
2585 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2586 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2587 
2588 	ext->nd_ifinfo = nd6_ifattach(ifp);
2589 	ext->scope6_id = scope6_ifattach(ifp);
2590 	ext->lltable = in6_lltattach(ifp);
2591 
2592 	ext->mld_ifinfo = mld_domifattach(ifp);
2593 
2594 	return ext;
2595 }
2596 
2597 int
2598 in6_domifmtu(struct ifnet *ifp)
2599 {
2600 	if (ifp->if_afdata[AF_INET6] == NULL)
2601 		return ifp->if_mtu;
2602 
2603 	return (IN6_LINKMTU(ifp));
2604 }
2605 
2606 void
2607 in6_domifdetach(struct ifnet *ifp, void *aux)
2608 {
2609 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2610 
2611 	mld_domifdetach(ifp);
2612 	scope6_ifdetach(ext->scope6_id);
2613 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2614 	lltable_free(ext->lltable);
2615 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2616 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2617 	free(ext->in6_ifstat, M_IFADDR);
2618 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2619 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2620 	free(ext->icmp6_ifstat, M_IFADDR);
2621 	free(ext, M_IFADDR);
2622 }
2623 
2624 /*
2625  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2626  * v4 mapped addr or v4 compat addr
2627  */
2628 void
2629 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2630 {
2631 
2632 	bzero(sin, sizeof(*sin));
2633 	sin->sin_len = sizeof(struct sockaddr_in);
2634 	sin->sin_family = AF_INET;
2635 	sin->sin_port = sin6->sin6_port;
2636 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2637 }
2638 
2639 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2640 void
2641 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2642 {
2643 	bzero(sin6, sizeof(*sin6));
2644 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2645 	sin6->sin6_family = AF_INET6;
2646 	sin6->sin6_port = sin->sin_port;
2647 	sin6->sin6_addr.s6_addr32[0] = 0;
2648 	sin6->sin6_addr.s6_addr32[1] = 0;
2649 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2650 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2651 }
2652 
2653 /* Convert sockaddr_in6 into sockaddr_in. */
2654 void
2655 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2656 {
2657 	struct sockaddr_in *sin_p;
2658 	struct sockaddr_in6 sin6;
2659 
2660 	/*
2661 	 * Save original sockaddr_in6 addr and convert it
2662 	 * to sockaddr_in.
2663 	 */
2664 	sin6 = *(struct sockaddr_in6 *)nam;
2665 	sin_p = (struct sockaddr_in *)nam;
2666 	in6_sin6_2_sin(sin_p, &sin6);
2667 }
2668 
2669 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2670 void
2671 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2672 {
2673 	struct sockaddr_in *sin_p;
2674 	struct sockaddr_in6 *sin6_p;
2675 
2676 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2677 	sin_p = (struct sockaddr_in *)*nam;
2678 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2679 	free(*nam, M_SONAME);
2680 	*nam = (struct sockaddr *)sin6_p;
2681 }
2682 
2683 /*
2684  * Join/leave the solicited multicast groups for proxy NDP entries.
2685  */
2686 static void
2687 in6_join_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2688 {
2689 	struct in6_multi *inm;
2690 	struct in6_addr mltaddr;
2691 	char ip6buf[INET6_ADDRSTRLEN];
2692 	int error;
2693 
2694 	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2695 		return;	/* error logged in in6_solicited_node_maddr. */
2696 
2697 	error = in6_joingroup(ifp, &mltaddr, NULL, &inm, 0);
2698 	if (error != 0) {
2699 		nd6log((LOG_WARNING,
2700 		    "%s: in6_joingroup failed for %s on %s (errno=%d)\n",
2701 		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp),
2702 		    error));
2703 	}
2704 }
2705 
2706 static void
2707 in6_leave_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2708 {
2709 	struct epoch_tracker et;
2710 	struct in6_multi *inm;
2711 	struct in6_addr mltaddr;
2712 	char ip6buf[INET6_ADDRSTRLEN];
2713 
2714 	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2715 		return;	/* error logged in in6_solicited_node_maddr. */
2716 
2717 	NET_EPOCH_ENTER(et);
2718 	inm = in6m_lookup(ifp, &mltaddr);
2719 	NET_EPOCH_EXIT(et);
2720 	if (inm != NULL)
2721 		in6_leavegroup(inm, NULL);
2722 	else
2723 		nd6log((LOG_WARNING, "%s: in6m_lookup failed for %s on %s\n",
2724 		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp)));
2725 }
2726 
2727 static bool
2728 in6_lle_match_pub(struct lltable *llt, struct llentry *lle, void *farg)
2729 {
2730 	return ((lle->la_flags & LLE_PUB) != 0);
2731 }
2732 
2733 void
2734 in6_purge_proxy_ndp(struct ifnet *ifp)
2735 {
2736 	struct lltable *llt;
2737 	bool need_purge;
2738 
2739 	if (ifp->if_afdata[AF_INET6] == NULL)
2740 		return;
2741 
2742 	llt = LLTABLE6(ifp);
2743 	IF_AFDATA_WLOCK(ifp);
2744 	need_purge = ((llt->llt_flags & LLT_ADDEDPROXY) != 0);
2745 	IF_AFDATA_WUNLOCK(ifp);
2746 
2747 	/*
2748 	 * Ever added proxy ndp entries, leave solicited node multicast
2749 	 * before deleting the llentry.
2750 	 */
2751 	if (need_purge)
2752 		lltable_delete_conditional(llt, in6_lle_match_pub, NULL);
2753 }
2754