xref: /freebsd/sys/netinet6/in6.c (revision 2008043f386721d58158e37e0d7e50df8095942d)
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 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 
69 #include <sys/param.h>
70 #include <sys/eventhandler.h>
71 #include <sys/errno.h>
72 #include <sys/jail.h>
73 #include <sys/malloc.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/sockio.h>
77 #include <sys/systm.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/lock.h>
83 #include <sys/rmlock.h>
84 #include <sys/sysctl.h>
85 #include <sys/syslog.h>
86 
87 #include <net/if.h>
88 #include <net/if_var.h>
89 #include <net/if_private.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/route/route_ctl.h>
93 #include <net/route/nhop.h>
94 #include <net/if_dl.h>
95 #include <net/vnet.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <net/if_llatbl.h>
100 #include <netinet/if_ether.h>
101 #include <netinet/in_systm.h>
102 #include <netinet/ip.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/ip_carp.h>
105 
106 #include <netinet/ip6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/nd6.h>
109 #include <netinet6/mld6_var.h>
110 #include <netinet6/ip6_mroute.h>
111 #include <netinet6/in6_ifattach.h>
112 #include <netinet6/scope6_var.h>
113 #include <netinet6/in6_fib.h>
114 #include <netinet6/in6_pcb.h>
115 
116 #ifdef MAC
117 #include <security/mac/mac_framework.h>
118 #endif
119 
120 /*
121  * struct in6_ifreq and struct ifreq must be type punnable for common members
122  * of ifr_ifru to allow accessors to be shared.
123  */
124 _Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
125     offsetof(struct ifreq, ifr_ifru),
126     "struct in6_ifreq and struct ifreq are not type punnable");
127 
128 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
129 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
130 
131 /*
132  * Definitions of some costant IP6 addresses.
133  */
134 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
135 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
136 const struct in6_addr in6addr_nodelocal_allnodes =
137 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
138 const struct in6_addr in6addr_linklocal_allnodes =
139 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
140 const struct in6_addr in6addr_linklocal_allrouters =
141 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
142 const struct in6_addr in6addr_linklocal_allv2routers =
143 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
144 
145 const struct in6_addr in6mask0 = IN6MASK0;
146 const struct in6_addr in6mask32 = IN6MASK32;
147 const struct in6_addr in6mask64 = IN6MASK64;
148 const struct in6_addr in6mask96 = IN6MASK96;
149 const struct in6_addr in6mask128 = IN6MASK128;
150 
151 const struct sockaddr_in6 sa6_any =
152 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
153 
154 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
155 	struct in6_aliasreq *, int);
156 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
157 
158 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
159     struct in6_ifaddr *, int);
160 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
161     struct in6_aliasreq *, int flags);
162 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
163     struct in6_ifaddr *, int, int);
164 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
165     struct in6_ifaddr *, int);
166 
167 static void in6_join_proxy_ndp_mc(struct ifnet *, const struct in6_addr *);
168 static void in6_leave_proxy_ndp_mc(struct ifnet *, const struct in6_addr *);
169 
170 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
171 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
172 
173 static struct sx in6_control_sx;
174 SX_SYSINIT(in6_control_sx, &in6_control_sx, "in6_control");
175 
176 void
177 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
178 {
179 	struct rt_addrinfo info;
180 	struct ifaddr *ifa;
181 	struct sockaddr_dl gateway;
182 	int fibnum;
183 
184 	ifa = &ia->ia_ifa;
185 
186 	/*
187 	 * Prepare info data for the host route.
188 	 * This code mimics one from ifa_maintain_loopback_route().
189 	 */
190 	bzero(&info, sizeof(struct rt_addrinfo));
191 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
192 	info.rti_info[RTAX_DST] = ifa->ifa_addr;
193 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
194 	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
195 	if (cmd != RTM_DELETE)
196 		info.rti_ifp = V_loif;
197 
198 	fibnum = ia62ifa(ia)->ifa_ifp->if_fib;
199 
200 	if (cmd == RTM_ADD) {
201 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
202 		rt_routemsg_info(cmd, &info, fibnum);
203 	} else if (cmd == RTM_DELETE) {
204 		rt_routemsg_info(cmd, &info, fibnum);
205 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
206 	}
207 }
208 
209 int
210 in6_mask2len(struct in6_addr *mask, u_char *lim0)
211 {
212 	int x = 0, y;
213 	u_char *lim = lim0, *p;
214 
215 	/* ignore the scope_id part */
216 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
217 		lim = (u_char *)mask + sizeof(*mask);
218 	for (p = (u_char *)mask; p < lim; x++, p++) {
219 		if (*p != 0xff)
220 			break;
221 	}
222 	y = 0;
223 	if (p < lim) {
224 		for (y = 0; y < 8; y++) {
225 			if ((*p & (0x80 >> y)) == 0)
226 				break;
227 		}
228 	}
229 
230 	/*
231 	 * when the limit pointer is given, do a stricter check on the
232 	 * remaining bits.
233 	 */
234 	if (p < lim) {
235 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
236 			return (-1);
237 		for (p = p + 1; p < lim; p++)
238 			if (*p != 0)
239 				return (-1);
240 	}
241 
242 	return x * 8 + y;
243 }
244 
245 #ifdef COMPAT_FREEBSD32
246 struct in6_ndifreq32 {
247 	char ifname[IFNAMSIZ];
248 	uint32_t ifindex;
249 };
250 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
251 #endif
252 
253 int
254 in6_control_ioctl(u_long cmd, void *data,
255     struct ifnet *ifp, struct ucred *cred)
256 {
257 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
258 	struct	in6_ifaddr *ia = NULL;
259 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
260 	struct sockaddr_in6 *sa6;
261 	int error;
262 	bool control_locked = false;
263 
264 	/*
265 	 * Compat to make pre-10.x ifconfig(8) operable.
266 	 */
267 	if (cmd == OSIOCAIFADDR_IN6) {
268 		cmd = SIOCAIFADDR_IN6;
269 		ifra->ifra_vhid = 0;
270 	}
271 
272 	switch (cmd) {
273 	case SIOCGETSGCNT_IN6:
274 	case SIOCGETMIFCNT_IN6:
275 		/*
276 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
277 		 * We cannot see how that would be needed, so do not adjust the
278 		 * KPI blindly; more likely should clean up the IPv4 variant.
279 		 */
280 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
281 	}
282 
283 	switch (cmd) {
284 	case SIOCAADDRCTL_POLICY:
285 	case SIOCDADDRCTL_POLICY:
286 		if (cred != NULL) {
287 			error = priv_check_cred(cred, PRIV_NETINET_ADDRCTRL6);
288 			if (error)
289 				return (error);
290 		}
291 		return (in6_src_ioctl(cmd, data));
292 	}
293 
294 	if (ifp == NULL)
295 		return (EOPNOTSUPP);
296 
297 	switch (cmd) {
298 	case SIOCSNDFLUSH_IN6:
299 	case SIOCSPFXFLUSH_IN6:
300 	case SIOCSRTRFLUSH_IN6:
301 	case SIOCSDEFIFACE_IN6:
302 	case SIOCSIFINFO_FLAGS:
303 	case SIOCSIFINFO_IN6:
304 		if (cred != NULL) {
305 			error = priv_check_cred(cred, PRIV_NETINET_ND6);
306 			if (error)
307 				return (error);
308 		}
309 		/* FALLTHROUGH */
310 	case OSIOCGIFINFO_IN6:
311 	case SIOCGIFINFO_IN6:
312 	case SIOCGNBRINFO_IN6:
313 	case SIOCGDEFIFACE_IN6:
314 		return (nd6_ioctl(cmd, data, ifp));
315 
316 #ifdef COMPAT_FREEBSD32
317 	case SIOCGDEFIFACE32_IN6:
318 		{
319 			struct in6_ndifreq ndif;
320 			struct in6_ndifreq32 *ndif32;
321 
322 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
323 			    ifp);
324 			if (error)
325 				return (error);
326 			ndif32 = (struct in6_ndifreq32 *)data;
327 			ndif32->ifindex = ndif.ifindex;
328 			return (0);
329 		}
330 #endif
331 	}
332 
333 	switch (cmd) {
334 	case SIOCSIFPREFIX_IN6:
335 	case SIOCDIFPREFIX_IN6:
336 	case SIOCAIFPREFIX_IN6:
337 	case SIOCCIFPREFIX_IN6:
338 	case SIOCSGIFPREFIX_IN6:
339 	case SIOCGIFPREFIX_IN6:
340 		log(LOG_NOTICE,
341 		    "prefix ioctls are now invalidated. "
342 		    "please use ifconfig.\n");
343 		return (EOPNOTSUPP);
344 	}
345 
346 	switch (cmd) {
347 	case SIOCSSCOPE6:
348 		if (cred != NULL) {
349 			error = priv_check_cred(cred, PRIV_NETINET_SCOPE6);
350 			if (error)
351 				return (error);
352 		}
353 		/* FALLTHROUGH */
354 	case SIOCGSCOPE6:
355 	case SIOCGSCOPE6DEF:
356 		return (scope6_ioctl(cmd, data, ifp));
357 	}
358 
359 	/*
360 	 * Find address for this interface, if it exists.
361 	 *
362 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
363 	 * only, and used the first interface address as the target of other
364 	 * operations (without checking ifra_addr).  This was because netinet
365 	 * code/API assumed at most 1 interface address per interface.
366 	 * Since IPv6 allows a node to assign multiple addresses
367 	 * on a single interface, we almost always look and check the
368 	 * presence of ifra_addr, and reject invalid ones here.
369 	 * It also decreases duplicated code among SIOC*_IN6 operations.
370 	 */
371 	switch (cmd) {
372 	case SIOCAIFADDR_IN6:
373 	case SIOCSIFPHYADDR_IN6:
374 		sa6 = &ifra->ifra_addr;
375 		break;
376 	case SIOCSIFADDR_IN6:
377 	case SIOCGIFADDR_IN6:
378 	case SIOCSIFDSTADDR_IN6:
379 	case SIOCSIFNETMASK_IN6:
380 	case SIOCGIFDSTADDR_IN6:
381 	case SIOCGIFNETMASK_IN6:
382 	case SIOCDIFADDR_IN6:
383 	case SIOCGIFPSRCADDR_IN6:
384 	case SIOCGIFPDSTADDR_IN6:
385 	case SIOCGIFAFLAG_IN6:
386 	case SIOCSNDFLUSH_IN6:
387 	case SIOCSPFXFLUSH_IN6:
388 	case SIOCSRTRFLUSH_IN6:
389 	case SIOCGIFALIFETIME_IN6:
390 	case SIOCGIFSTAT_IN6:
391 	case SIOCGIFSTAT_ICMP6:
392 		sa6 = &ifr->ifr_addr;
393 		break;
394 	case SIOCSIFADDR:
395 	case SIOCSIFBRDADDR:
396 	case SIOCSIFDSTADDR:
397 	case SIOCSIFNETMASK:
398 		/*
399 		 * Although we should pass any non-INET6 ioctl requests
400 		 * down to driver, we filter some legacy INET requests.
401 		 * Drivers trust SIOCSIFADDR et al to come from an already
402 		 * privileged layer, and do not perform any credentials
403 		 * checks or input validation.
404 		 */
405 		return (EINVAL);
406 	default:
407 		sa6 = NULL;
408 		break;
409 	}
410 	if (sa6 && sa6->sin6_family == AF_INET6) {
411 		if (sa6->sin6_scope_id != 0)
412 			error = sa6_embedscope(sa6, 0);
413 		else
414 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
415 		if (error != 0)
416 			return (error);
417 		if (cred != NULL && (error = prison_check_ip6(cred,
418 		    &sa6->sin6_addr)) != 0)
419 			return (error);
420 		sx_xlock(&in6_control_sx);
421 		control_locked = true;
422 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
423 	} else
424 		ia = NULL;
425 
426 	switch (cmd) {
427 	case SIOCSIFADDR_IN6:
428 	case SIOCSIFDSTADDR_IN6:
429 	case SIOCSIFNETMASK_IN6:
430 		/*
431 		 * Since IPv6 allows a node to assign multiple addresses
432 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
433 		 */
434 		/* we decided to obsolete this command (20000704) */
435 		error = EINVAL;
436 		goto out;
437 
438 	case SIOCDIFADDR_IN6:
439 		/*
440 		 * for IPv4, we look for existing in_ifaddr here to allow
441 		 * "ifconfig if0 delete" to remove the first IPv4 address on
442 		 * the interface.  For IPv6, as the spec allows multiple
443 		 * interface address from the day one, we consider "remove the
444 		 * first one" semantics to be not preferable.
445 		 */
446 		if (ia == NULL) {
447 			error = EADDRNOTAVAIL;
448 			goto out;
449 		}
450 		/* FALLTHROUGH */
451 	case SIOCAIFADDR_IN6:
452 		/*
453 		 * We always require users to specify a valid IPv6 address for
454 		 * the corresponding operation.
455 		 */
456 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
457 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
458 			error = EAFNOSUPPORT;
459 			goto out;
460 		}
461 
462 		if (cred != NULL) {
463 			error = priv_check_cred(cred, (cmd == SIOCDIFADDR_IN6) ?
464 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
465 			if (error)
466 				goto out;
467 		}
468 		/* FALLTHROUGH */
469 	case SIOCGIFSTAT_IN6:
470 	case SIOCGIFSTAT_ICMP6:
471 		if (ifp->if_afdata[AF_INET6] == NULL) {
472 			error = EPFNOSUPPORT;
473 			goto out;
474 		}
475 		break;
476 
477 	case SIOCGIFADDR_IN6:
478 		/* This interface is basically deprecated. use SIOCGIFCONF. */
479 		/* FALLTHROUGH */
480 	case SIOCGIFAFLAG_IN6:
481 	case SIOCGIFNETMASK_IN6:
482 	case SIOCGIFDSTADDR_IN6:
483 	case SIOCGIFALIFETIME_IN6:
484 		/* must think again about its semantics */
485 		if (ia == NULL) {
486 			error = EADDRNOTAVAIL;
487 			goto out;
488 		}
489 		break;
490 	}
491 
492 	switch (cmd) {
493 	case SIOCGIFADDR_IN6:
494 		ifr->ifr_addr = ia->ia_addr;
495 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
496 			goto out;
497 		break;
498 
499 	case SIOCGIFDSTADDR_IN6:
500 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
501 			error = EINVAL;
502 			goto out;
503 		}
504 		ifr->ifr_dstaddr = ia->ia_dstaddr;
505 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
506 			goto out;
507 		break;
508 
509 	case SIOCGIFNETMASK_IN6:
510 		ifr->ifr_addr = ia->ia_prefixmask;
511 		break;
512 
513 	case SIOCGIFAFLAG_IN6:
514 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
515 		break;
516 
517 	case SIOCGIFSTAT_IN6:
518 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
519 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
520 		    &ifr->ifr_ifru.ifru_stat,
521 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
522 		break;
523 
524 	case SIOCGIFSTAT_ICMP6:
525 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
526 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
527 		    &ifr->ifr_ifru.ifru_icmp6stat,
528 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
529 		break;
530 
531 	case SIOCGIFALIFETIME_IN6:
532 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
533 		if (ia->ia6_lifetime.ia6t_vltime != 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_vltime <
545 			    maxexpire - ia->ia6_updatetime) {
546 				retlt->ia6t_expire = ia->ia6_updatetime +
547 				    ia->ia6_lifetime.ia6t_vltime;
548 			} else
549 				retlt->ia6t_expire = maxexpire;
550 		}
551 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
552 			time_t maxexpire;
553 			struct in6_addrlifetime *retlt =
554 			    &ifr->ifr_ifru.ifru_lifetime;
555 
556 			/*
557 			 * XXX: adjust expiration time assuming time_t is
558 			 * signed.
559 			 */
560 			maxexpire = (-1) &
561 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
562 			if (ia->ia6_lifetime.ia6t_pltime <
563 			    maxexpire - ia->ia6_updatetime) {
564 				retlt->ia6t_preferred = ia->ia6_updatetime +
565 				    ia->ia6_lifetime.ia6t_pltime;
566 			} else
567 				retlt->ia6t_preferred = maxexpire;
568 		}
569 		break;
570 
571 	case SIOCAIFADDR_IN6:
572 #ifdef MAC
573 		/* Check if a MAC policy disallows setting the IPv6 address. */
574 		error = mac_inet6_check_add_addr(cred, &sa6->sin6_addr, ifp);
575 		if (error != 0)
576 			goto out;
577 #endif
578 		error = in6_addifaddr(ifp, ifra, ia);
579 		ia = NULL;
580 		break;
581 
582 	case SIOCDIFADDR_IN6:
583 		in6_purgeifaddr(ia);
584 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
585 		    IFADDR_EVENT_DEL);
586 		break;
587 
588 	default:
589 		if (ifp->if_ioctl == NULL) {
590 			error = EOPNOTSUPP;
591 			goto out;
592 		}
593 		error = (*ifp->if_ioctl)(ifp, cmd, data);
594 		goto out;
595 	}
596 
597 	error = 0;
598 out:
599 	if (control_locked)
600 		sx_xunlock(&in6_control_sx);
601 
602 	if (ia != NULL)
603 		ifa_free(&ia->ia_ifa);
604 	return (error);
605 }
606 
607 int
608 in6_control(struct socket *so, u_long cmd, void *data,
609     struct ifnet *ifp, struct thread *td)
610 {
611 	return (in6_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL));
612 }
613 
614 static struct in6_multi_mship *
615 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
616     int *errorp, int delay)
617 {
618 	struct in6_multi_mship *imm;
619 	int error;
620 
621 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
622 	if (imm == NULL) {
623 		*errorp = ENOBUFS;
624 		return (NULL);
625 	}
626 
627 	delay = (delay * MLD_FASTHZ) / hz;
628 
629 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
630 	if (error) {
631 		*errorp = error;
632 		free(imm, M_IP6MADDR);
633 		return (NULL);
634 	}
635 
636 	return (imm);
637 }
638 
639 static int
640 in6_solicited_node_maddr(struct in6_addr *maddr,
641     struct ifnet *ifp, const struct in6_addr *base)
642 {
643 	int error;
644 
645 	bzero(maddr, sizeof(struct in6_addr));
646 	maddr->s6_addr32[0] = IPV6_ADDR_INT32_MLL;
647 	maddr->s6_addr32[2] = htonl(1);
648 	maddr->s6_addr32[3] = base->s6_addr32[3];
649 	maddr->s6_addr8[12] = 0xff;
650 	if ((error = in6_setscope(maddr, ifp, NULL)) != 0) {
651 		/* XXX: should not happen */
652 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
653 	}
654 
655 	return error;
656 }
657 
658 /*
659  * Join necessary multicast groups.  Factored out from in6_update_ifa().
660  * This entire work should only be done once, for the default FIB.
661  */
662 static int
663 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
664     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
665 {
666 	char ip6buf[INET6_ADDRSTRLEN];
667 	struct in6_addr mltaddr;
668 	struct in6_multi_mship *imm;
669 	int delay, error;
670 
671 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
672 
673 	/* Join solicited multicast addr for new host id. */
674 	if ((error = in6_solicited_node_maddr(&mltaddr, ifp,
675 	    &ifra->ifra_addr.sin6_addr)) != 0)
676 		goto cleanup;
677 	delay = error = 0;
678 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
679 		/*
680 		 * We need a random delay for DAD on the address being
681 		 * configured.  It also means delaying transmission of the
682 		 * corresponding MLD report to avoid report collision.
683 		 * [RFC 4861, Section 6.3.7]
684 		 */
685 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
686 	}
687 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
688 	if (imm == NULL) {
689 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
690 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
691 		    if_name(ifp), error));
692 		goto cleanup;
693 	}
694 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
695 	*in6m_sol = imm->i6mm_maddr;
696 
697 	/*
698 	 * Join link-local all-nodes address.
699 	 */
700 	mltaddr = in6addr_linklocal_allnodes;
701 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
702 		goto cleanup; /* XXX: should not fail */
703 
704 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
705 	if (imm == NULL) {
706 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
707 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
708 		    if_name(ifp), error));
709 		goto cleanup;
710 	}
711 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
712 
713 	/*
714 	 * Join node information group address.
715 	 */
716 	delay = 0;
717 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
718 		/*
719 		 * The spec does not say anything about delay for this group,
720 		 * but the same logic should apply.
721 		 */
722 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
723 	}
724 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
725 		/* XXX jinmei */
726 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
727 		if (imm == NULL)
728 			nd6log((LOG_WARNING,
729 			    "%s: in6_joingroup failed for %s on %s "
730 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
731 			    &mltaddr), if_name(ifp), error));
732 			/* XXX not very fatal, go on... */
733 		else
734 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
735 	}
736 	if (V_icmp6_nodeinfo_oldmcprefix &&
737 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
738 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
739 		if (imm == NULL)
740 			nd6log((LOG_WARNING,
741 			    "%s: in6_joingroup failed for %s on %s "
742 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
743 			    &mltaddr), if_name(ifp), error));
744 			/* XXX not very fatal, go on... */
745 		else
746 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
747 	}
748 
749 	/*
750 	 * Join interface-local all-nodes address.
751 	 * (ff01::1%ifN, and ff01::%ifN/32)
752 	 */
753 	mltaddr = in6addr_nodelocal_allnodes;
754 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
755 		goto cleanup; /* XXX: should not fail */
756 
757 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
758 	if (imm == NULL) {
759 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
760 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
761 		    &mltaddr), if_name(ifp), error));
762 		goto cleanup;
763 	}
764 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
765 
766 cleanup:
767 	return (error);
768 }
769 
770 /*
771  * Update parameters of an IPv6 interface address.
772  * If necessary, a new entry is created and linked into address chains.
773  * This function is separated from in6_control().
774  */
775 int
776 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
777     struct in6_ifaddr *ia, int flags)
778 {
779 	int error, hostIsNew = 0;
780 
781 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
782 		return (error);
783 
784 	if (ia == NULL) {
785 		hostIsNew = 1;
786 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
787 			return (ENOBUFS);
788 	}
789 
790 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
791 	if (error != 0) {
792 		if (hostIsNew != 0) {
793 			in6_unlink_ifa(ia, ifp);
794 			ifa_free(&ia->ia_ifa);
795 		}
796 		return (error);
797 	}
798 
799 	if (hostIsNew)
800 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
801 
802 	return (error);
803 }
804 
805 /*
806  * Fill in basic IPv6 address request info.
807  */
808 void
809 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
810     const struct in6_addr *mask)
811 {
812 
813 	memset(ifra, 0, sizeof(struct in6_aliasreq));
814 
815 	ifra->ifra_addr.sin6_family = AF_INET6;
816 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
817 	if (addr != NULL)
818 		ifra->ifra_addr.sin6_addr = *addr;
819 
820 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
821 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
822 	if (mask != NULL)
823 		ifra->ifra_prefixmask.sin6_addr = *mask;
824 }
825 
826 static int
827 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
828     struct in6_ifaddr *ia, int flags)
829 {
830 	int plen = -1;
831 	struct sockaddr_in6 dst6;
832 	struct in6_addrlifetime *lt;
833 	char ip6buf[INET6_ADDRSTRLEN];
834 
835 	/* Validate parameters */
836 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
837 		return (EINVAL);
838 
839 	/*
840 	 * The destination address for a p2p link must have a family
841 	 * of AF_UNSPEC or AF_INET6.
842 	 */
843 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
844 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
845 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
846 		return (EAFNOSUPPORT);
847 
848 	/*
849 	 * Validate address
850 	 */
851 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
852 	    ifra->ifra_addr.sin6_family != AF_INET6)
853 		return (EINVAL);
854 
855 	/*
856 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
857 	 * does not carry fields other than sin6_len.
858 	 */
859 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
860 		return (EINVAL);
861 	/*
862 	 * Because the IPv6 address architecture is classless, we require
863 	 * users to specify a (non 0) prefix length (mask) for a new address.
864 	 * We also require the prefix (when specified) mask is valid, and thus
865 	 * reject a non-consecutive mask.
866 	 */
867 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
868 		return (EINVAL);
869 	if (ifra->ifra_prefixmask.sin6_len != 0) {
870 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
871 		    (u_char *)&ifra->ifra_prefixmask +
872 		    ifra->ifra_prefixmask.sin6_len);
873 		if (plen <= 0)
874 			return (EINVAL);
875 	} else {
876 		/*
877 		 * In this case, ia must not be NULL.  We just use its prefix
878 		 * length.
879 		 */
880 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
881 	}
882 	/*
883 	 * If the destination address on a p2p interface is specified,
884 	 * and the address is a scoped one, validate/set the scope
885 	 * zone identifier.
886 	 */
887 	dst6 = ifra->ifra_dstaddr;
888 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
889 	    (dst6.sin6_family == AF_INET6)) {
890 		struct in6_addr in6_tmp;
891 		u_int32_t zoneid;
892 
893 		in6_tmp = dst6.sin6_addr;
894 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
895 			return (EINVAL); /* XXX: should be impossible */
896 
897 		if (dst6.sin6_scope_id != 0) {
898 			if (dst6.sin6_scope_id != zoneid)
899 				return (EINVAL);
900 		} else		/* user omit to specify the ID. */
901 			dst6.sin6_scope_id = zoneid;
902 
903 		/* convert into the internal form */
904 		if (sa6_embedscope(&dst6, 0))
905 			return (EINVAL); /* XXX: should be impossible */
906 	}
907 	/* Modify original ifra_dstaddr to reflect changes */
908 	ifra->ifra_dstaddr = dst6;
909 
910 	/*
911 	 * The destination address can be specified only for a p2p or a
912 	 * loopback interface.  If specified, the corresponding prefix length
913 	 * must be 128.
914 	 */
915 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
916 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
917 			/* XXX: noisy message */
918 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
919 			    "be specified for a p2p or a loopback IF only\n"));
920 			return (EINVAL);
921 		}
922 		if (plen != 128) {
923 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
924 			    "be 128 when dstaddr is specified\n"));
925 			return (EINVAL);
926 		}
927 	}
928 	/* lifetime consistency check */
929 	lt = &ifra->ifra_lifetime;
930 	if (lt->ia6t_pltime > lt->ia6t_vltime)
931 		return (EINVAL);
932 	if (lt->ia6t_vltime == 0) {
933 		/*
934 		 * the following log might be noisy, but this is a typical
935 		 * configuration mistake or a tool's bug.
936 		 */
937 		nd6log((LOG_INFO,
938 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
939 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
940 
941 		if (ia == NULL)
942 			return (0); /* there's nothing to do */
943 	}
944 
945 	/* Check prefix mask */
946 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
947 		/*
948 		 * We prohibit changing the prefix length of an existing
949 		 * address, because
950 		 * + such an operation should be rare in IPv6, and
951 		 * + the operation would confuse prefix management.
952 		 */
953 		if (ia->ia_prefixmask.sin6_len != 0 &&
954 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
955 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
956 			    "of an existing %s address should not be changed\n",
957 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
958 
959 			return (EINVAL);
960 		}
961 	}
962 
963 	return (0);
964 }
965 
966 /*
967  * Allocate a new ifaddr and link it into chains.
968  */
969 static struct in6_ifaddr *
970 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
971 {
972 	struct in6_ifaddr *ia;
973 
974 	/*
975 	 * When in6_alloc_ifa() is called in a process of a received
976 	 * RA, it is called under an interrupt context.  So, we should
977 	 * call malloc with M_NOWAIT.
978 	 */
979 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
980 	if (ia == NULL)
981 		return (NULL);
982 	LIST_INIT(&ia->ia6_memberships);
983 	/* Initialize the address and masks, and put time stamp */
984 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
985 	ia->ia_addr.sin6_family = AF_INET6;
986 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
987 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
988 	ia->ia_addr = ifra->ifra_addr;
989 	ia->ia6_createtime = time_uptime;
990 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
991 		/*
992 		 * Some functions expect that ifa_dstaddr is not
993 		 * NULL for p2p interfaces.
994 		 */
995 		ia->ia_ifa.ifa_dstaddr =
996 		    (struct sockaddr *)&ia->ia_dstaddr;
997 	} else {
998 		ia->ia_ifa.ifa_dstaddr = NULL;
999 	}
1000 
1001 	/* set prefix mask if any */
1002 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1003 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1004 		ia->ia_prefixmask.sin6_family = AF_INET6;
1005 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1006 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1007 	}
1008 
1009 	ia->ia_ifp = ifp;
1010 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1011 	IF_ADDR_WLOCK(ifp);
1012 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1013 	IF_ADDR_WUNLOCK(ifp);
1014 
1015 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1016 	IN6_IFADDR_WLOCK();
1017 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1018 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1019 	IN6_IFADDR_WUNLOCK();
1020 
1021 	return (ia);
1022 }
1023 
1024 /*
1025  * Update/configure interface address parameters:
1026  *
1027  * 1) Update lifetime
1028  * 2) Update interface metric ad flags
1029  * 3) Notify other subsystems
1030  */
1031 static int
1032 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1033     struct in6_ifaddr *ia, int hostIsNew, int flags)
1034 {
1035 	int error;
1036 
1037 	/* update timestamp */
1038 	ia->ia6_updatetime = time_uptime;
1039 
1040 	/*
1041 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1042 	 * to see if the address is deprecated or invalidated, but initialize
1043 	 * these members for applications.
1044 	 */
1045 	ia->ia6_lifetime = ifra->ifra_lifetime;
1046 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1047 		ia->ia6_lifetime.ia6t_expire =
1048 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1049 	} else
1050 		ia->ia6_lifetime.ia6t_expire = 0;
1051 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1052 		ia->ia6_lifetime.ia6t_preferred =
1053 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1054 	} else
1055 		ia->ia6_lifetime.ia6t_preferred = 0;
1056 
1057 	/*
1058 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1059 	 * userland, make it deprecated.
1060 	 */
1061 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1062 		ia->ia6_lifetime.ia6t_pltime = 0;
1063 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1064 	}
1065 
1066 	/*
1067 	 * configure address flags.
1068 	 */
1069 	ia->ia6_flags = ifra->ifra_flags;
1070 
1071 	/*
1072 	 * Make the address tentative before joining multicast addresses,
1073 	 * so that corresponding MLD responses would not have a tentative
1074 	 * source address.
1075 	 */
1076 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1077 
1078 	/*
1079 	 * DAD should be performed for an new address or addresses on
1080 	 * an interface with ND6_IFF_IFDISABLED.
1081 	 */
1082 	if (in6if_do_dad(ifp) &&
1083 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1084 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1085 
1086 	/* notify other subsystems */
1087 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1088 
1089 	return (error);
1090 }
1091 
1092 /*
1093  * Do link-level ifa job:
1094  * 1) Add lle entry for added address
1095  * 2) Notifies routing socket users about new address
1096  * 3) join appropriate multicast group
1097  * 4) start DAD if enabled
1098  */
1099 static int
1100 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1101     struct in6_ifaddr *ia, int flags)
1102 {
1103 	struct in6_multi *in6m_sol;
1104 	int error = 0;
1105 
1106 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1107 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1108 		in6_purgeaddr(&ia->ia_ifa);
1109 		ifa_free(&ia->ia_ifa);
1110 		return (error);
1111 	}
1112 
1113 	/* Join necessary multicast groups. */
1114 	in6m_sol = NULL;
1115 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1116 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1117 		if (error != 0) {
1118 			in6_purgeaddr(&ia->ia_ifa);
1119 			ifa_free(&ia->ia_ifa);
1120 			return (error);
1121 		}
1122 	}
1123 
1124 	/* Perform DAD, if the address is TENTATIVE. */
1125 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1126 		int delay, mindelay, maxdelay;
1127 
1128 		delay = 0;
1129 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1130 			/*
1131 			 * We need to impose a delay before sending an NS
1132 			 * for DAD.  Check if we also needed a delay for the
1133 			 * corresponding MLD message.  If we did, the delay
1134 			 * should be larger than the MLD delay (this could be
1135 			 * relaxed a bit, but this simple logic is at least
1136 			 * safe).
1137 			 * XXX: Break data hiding guidelines and look at
1138 			 * state for the solicited multicast group.
1139 			 */
1140 			mindelay = 0;
1141 			if (in6m_sol != NULL &&
1142 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1143 				mindelay = in6m_sol->in6m_timer;
1144 			}
1145 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1146 			if (maxdelay - mindelay == 0)
1147 				delay = 0;
1148 			else {
1149 				delay =
1150 				    (arc4random() % (maxdelay - mindelay)) +
1151 				    mindelay;
1152 			}
1153 		}
1154 		nd6_dad_start((struct ifaddr *)ia, delay);
1155 	}
1156 
1157 	in6_newaddrmsg(ia, RTM_ADD);
1158 	ifa_free(&ia->ia_ifa);
1159 	return (error);
1160 }
1161 
1162 /*
1163  * Adds or deletes interface route for p2p ifa.
1164  * Returns 0 on success or errno.
1165  */
1166 static int
1167 in6_handle_dstaddr_rtrequest(int cmd, struct in6_ifaddr *ia)
1168 {
1169 	struct epoch_tracker et;
1170 	struct ifaddr *ifa = &ia->ia_ifa;
1171 	int error;
1172 
1173 	/* Prepare gateway */
1174 	struct sockaddr_dl_short sdl = {
1175 		.sdl_family = AF_LINK,
1176 		.sdl_len = sizeof(struct sockaddr_dl_short),
1177 		.sdl_type = ifa->ifa_ifp->if_type,
1178 		.sdl_index = ifa->ifa_ifp->if_index,
1179 	};
1180 
1181 	struct sockaddr_in6 dst = {
1182 		.sin6_family = AF_INET6,
1183 		.sin6_len = sizeof(struct sockaddr_in6),
1184 		.sin6_addr = ia->ia_dstaddr.sin6_addr,
1185 	};
1186 
1187 	struct rt_addrinfo info = {
1188 		.rti_ifa = ifa,
1189 		.rti_ifp = ifa->ifa_ifp,
1190 		.rti_flags = RTF_PINNED | RTF_HOST,
1191 		.rti_info = {
1192 			[RTAX_DST] = (struct sockaddr *)&dst,
1193 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
1194 		},
1195 	};
1196 	/* Don't set additional per-gw filters on removal */
1197 
1198 	NET_EPOCH_ENTER(et);
1199 	error = rib_handle_ifaddr_info(ifa->ifa_ifp->if_fib, cmd, &info);
1200 	NET_EPOCH_EXIT(et);
1201 
1202 	return (error);
1203 }
1204 
1205 static bool
1206 ifa_is_p2p(struct in6_ifaddr *ia)
1207 {
1208 	int plen;
1209 
1210 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1211 
1212 	if ((plen == 128) && (ia->ia_dstaddr.sin6_family == AF_INET6) &&
1213 	    !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &ia->ia_dstaddr.sin6_addr))
1214 		return (true);
1215 
1216 	return (false);
1217 }
1218 
1219 int
1220 in6_addifaddr(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia)
1221 {
1222 	struct nd_prefixctl pr0;
1223 	struct nd_prefix *pr;
1224 	int carp_attached = 0;
1225 	int error;
1226 
1227 	/*
1228 	 * first, make or update the interface address structure,
1229 	 * and link it to the list.
1230 	 */
1231 	if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
1232 		goto out;
1233 	if (ia != NULL) {
1234 		if (ia->ia_ifa.ifa_carp)
1235 			(*carp_detach_p)(&ia->ia_ifa, true);
1236 		ifa_free(&ia->ia_ifa);
1237 	}
1238 	if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr)) == NULL) {
1239 		/*
1240 		 * this can happen when the user specify the 0 valid
1241 		 * lifetime.
1242 		 */
1243 		return (0);
1244 	}
1245 
1246 	if (ifra->ifra_vhid > 0) {
1247 		if (carp_attach_p != NULL)
1248 			error = (*carp_attach_p)(&ia->ia_ifa,
1249 			    ifra->ifra_vhid);
1250 		else
1251 			error = EPROTONOSUPPORT;
1252 		if (error)
1253 			goto out;
1254 		else
1255 			carp_attached = 1;
1256 	}
1257 
1258 	/*
1259 	 * then, make the prefix on-link on the interface.
1260 	 * XXX: we'd rather create the prefix before the address, but
1261 	 * we need at least one address to install the corresponding
1262 	 * interface route, so we configure the address first.
1263 	 */
1264 
1265 	/*
1266 	 * convert mask to prefix length (prefixmask has already
1267 	 * been validated in in6_update_ifa().
1268 	 */
1269 	bzero(&pr0, sizeof(pr0));
1270 	pr0.ndpr_ifp = ifp;
1271 	pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1272 	    NULL);
1273 	if (pr0.ndpr_plen == 128) {
1274 		/* we don't need to install a host route. */
1275 		goto aifaddr_out;
1276 	}
1277 	pr0.ndpr_prefix = ifra->ifra_addr;
1278 	/* apply the mask for safety. */
1279 	IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
1280 	    &ifra->ifra_prefixmask.sin6_addr);
1281 
1282 	/*
1283 	 * XXX: since we don't have an API to set prefix (not address)
1284 	 * lifetimes, we just use the same lifetimes as addresses.
1285 	 * The (temporarily) installed lifetimes can be overridden by
1286 	 * later advertised RAs (when accept_rtadv is non 0), which is
1287 	 * an intended behavior.
1288 	 */
1289 	pr0.ndpr_raf_onlink = 1; /* should be configurable? */
1290 	pr0.ndpr_raf_auto =
1291 	    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
1292 	pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
1293 	pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
1294 
1295 	/* add the prefix if not yet. */
1296 	if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
1297 		/*
1298 		 * nd6_prelist_add will install the corresponding
1299 		 * interface route.
1300 		 */
1301 		if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
1302 			if (carp_attached)
1303 				(*carp_detach_p)(&ia->ia_ifa, false);
1304 			goto out;
1305 		}
1306 	}
1307 
1308 	/* relate the address to the prefix */
1309 	if (ia->ia6_ndpr == NULL) {
1310 		ia->ia6_ndpr = pr;
1311 		pr->ndpr_addrcnt++;
1312 
1313 		/*
1314 		 * If this is the first autoconf address from the
1315 		 * prefix, create a temporary address as well
1316 		 * (when required).
1317 		 */
1318 		if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
1319 		    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
1320 			int e;
1321 			if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
1322 				log(LOG_NOTICE, "in6_control: failed "
1323 				    "to create a temporary address, "
1324 				    "errno=%d\n", e);
1325 			}
1326 		}
1327 	}
1328 	nd6_prefix_rele(pr);
1329 
1330 	/*
1331 	 * this might affect the status of autoconfigured addresses,
1332 	 * that is, this address might make other addresses detached.
1333 	 */
1334 	pfxlist_onlink_check();
1335 
1336 aifaddr_out:
1337 	/*
1338 	 * Try to clear the flag when a new IPv6 address is added
1339 	 * onto an IFDISABLED interface and it succeeds.
1340 	 */
1341 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1342 		struct in6_ndireq nd;
1343 
1344 		memset(&nd, 0, sizeof(nd));
1345 		nd.ndi.flags = ND_IFINFO(ifp)->flags;
1346 		nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
1347 		if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
1348 			log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
1349 			    "SIOCSIFINFO_FLAGS for -ifdisabled "
1350 			    "failed.");
1351 		/*
1352 		 * Ignore failure of clearing the flag intentionally.
1353 		 * The failure means address duplication was detected.
1354 		 */
1355 	}
1356 	error = 0;
1357 
1358 out:
1359 	if (ia != NULL)
1360 		ifa_free(&ia->ia_ifa);
1361 	return (error);
1362 }
1363 
1364 void
1365 in6_purgeaddr(struct ifaddr *ifa)
1366 {
1367 	struct ifnet *ifp = ifa->ifa_ifp;
1368 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1369 	struct in6_multi_mship *imm;
1370 	int error;
1371 
1372 	if (ifa->ifa_carp)
1373 		(*carp_detach_p)(ifa, false);
1374 
1375 	/*
1376 	 * Remove the loopback route to the interface address.
1377 	 * The check for the current setting of "nd6_useloopback"
1378 	 * is not needed.
1379 	 */
1380 	if (ia->ia_flags & IFA_RTSELF) {
1381 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1382 		    (struct sockaddr *)&ia->ia_addr);
1383 		if (error == 0)
1384 			ia->ia_flags &= ~IFA_RTSELF;
1385 	}
1386 
1387 	/* stop DAD processing */
1388 	nd6_dad_stop(ifa);
1389 
1390 	/* Leave multicast groups. */
1391 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1392 		LIST_REMOVE(imm, i6mm_chain);
1393 		if (imm->i6mm_maddr != NULL)
1394 			in6_leavegroup(imm->i6mm_maddr, NULL);
1395 		free(imm, M_IP6MADDR);
1396 	}
1397 	/* Check if we need to remove p2p route */
1398 	if ((ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1399 		error = in6_handle_dstaddr_rtrequest(RTM_DELETE, ia);
1400 		if (error != 0)
1401 			log(LOG_INFO, "%s: err=%d, destination address delete "
1402 			    "failed\n", __func__, error);
1403 		ia->ia_flags &= ~IFA_ROUTE;
1404 	}
1405 
1406 	in6_newaddrmsg(ia, RTM_DELETE);
1407 	in6_unlink_ifa(ia, ifp);
1408 }
1409 
1410 /*
1411  * Removes @ia from the corresponding interfaces and unlinks corresponding
1412  *  prefix if no addresses are using it anymore.
1413  */
1414 void
1415 in6_purgeifaddr(struct in6_ifaddr *ia)
1416 {
1417 	struct nd_prefix *pr;
1418 
1419 	/*
1420 	 * If the address being deleted is the only one that owns
1421 	 * the corresponding prefix, expire the prefix as well.
1422 	 * XXX: theoretically, we don't have to worry about such
1423 	 * relationship, since we separate the address management
1424 	 * and the prefix management.  We do this, however, to provide
1425 	 * as much backward compatibility as possible in terms of
1426 	 * the ioctl operation.
1427 	 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
1428 	 */
1429 	pr = ia->ia6_ndpr;
1430 	in6_purgeaddr(&ia->ia_ifa);
1431 	if (pr != NULL && pr->ndpr_addrcnt == 0) {
1432 		ND6_WLOCK();
1433 		nd6_prefix_unlink(pr, NULL);
1434 		ND6_WUNLOCK();
1435 		nd6_prefix_del(pr);
1436 	}
1437 }
1438 
1439 
1440 static void
1441 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1442 {
1443 	char ip6buf[INET6_ADDRSTRLEN];
1444 	int remove_lle;
1445 
1446 	IF_ADDR_WLOCK(ifp);
1447 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1448 	IF_ADDR_WUNLOCK(ifp);
1449 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1450 
1451 	/*
1452 	 * Defer the release of what might be the last reference to the
1453 	 * in6_ifaddr so that it can't be freed before the remainder of the
1454 	 * cleanup.
1455 	 */
1456 	IN6_IFADDR_WLOCK();
1457 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1458 	CK_LIST_REMOVE(ia, ia6_hash);
1459 	IN6_IFADDR_WUNLOCK();
1460 
1461 	/*
1462 	 * Release the reference to the base prefix.  There should be a
1463 	 * positive reference.
1464 	 */
1465 	remove_lle = 0;
1466 	if (ia->ia6_ndpr == NULL) {
1467 		nd6log((LOG_NOTICE,
1468 		    "in6_unlink_ifa: autoconf'ed address "
1469 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1470 	} else {
1471 		ia->ia6_ndpr->ndpr_addrcnt--;
1472 		/* Do not delete lles within prefix if refcont != 0 */
1473 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1474 			remove_lle = 1;
1475 		ia->ia6_ndpr = NULL;
1476 	}
1477 
1478 	nd6_rem_ifa_lle(ia, remove_lle);
1479 
1480 	/*
1481 	 * Also, if the address being removed is autoconf'ed, call
1482 	 * pfxlist_onlink_check() since the release might affect the status of
1483 	 * other (detached) addresses.
1484 	 */
1485 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1486 		pfxlist_onlink_check();
1487 	}
1488 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1489 }
1490 
1491 /*
1492  * Notifies other subsystems about address change/arrival:
1493  * 1) Notifies device handler on the first IPv6 address assignment
1494  * 2) Handle routing table changes for P2P links and route
1495  * 3) Handle routing table changes for address host route
1496  */
1497 static int
1498 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1499     struct in6_aliasreq *ifra, int hostIsNew)
1500 {
1501 	int	error = 0, ifacount = 0;
1502 	struct ifaddr *ifa;
1503 	struct sockaddr_in6 *pdst;
1504 	char ip6buf[INET6_ADDRSTRLEN];
1505 
1506 	/*
1507 	 * Give the interface a chance to initialize
1508 	 * if this is its first address,
1509 	 */
1510 	if (hostIsNew != 0) {
1511 		struct epoch_tracker et;
1512 
1513 		NET_EPOCH_ENTER(et);
1514 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1515 			if (ifa->ifa_addr->sa_family != AF_INET6)
1516 				continue;
1517 			ifacount++;
1518 		}
1519 		NET_EPOCH_EXIT(et);
1520 	}
1521 
1522 	if (ifacount <= 1 && ifp->if_ioctl) {
1523 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1524 		if (error)
1525 			goto done;
1526 	}
1527 
1528 	/*
1529 	 * If a new destination address is specified, scrub the old one and
1530 	 * install the new destination.  Note that the interface must be
1531 	 * p2p or loopback.
1532 	 */
1533 	pdst = &ifra->ifra_dstaddr;
1534 	if (pdst->sin6_family == AF_INET6 &&
1535 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1536 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1537 		    (in6_handle_dstaddr_rtrequest(RTM_DELETE, ia) != 0)) {
1538 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1539 			    "remove a route to the old destination: %s\n",
1540 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1541 			/* proceed anyway... */
1542 		} else
1543 			ia->ia_flags &= ~IFA_ROUTE;
1544 		ia->ia_dstaddr = *pdst;
1545 	}
1546 
1547 	/*
1548 	 * If a new destination address is specified for a point-to-point
1549 	 * interface, install a route to the destination as an interface
1550 	 * direct route.
1551 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1552 	 * interface that share the same destination.
1553 	 */
1554 	if (!(ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1555 		error = in6_handle_dstaddr_rtrequest(RTM_ADD, ia);
1556 		if (error)
1557 			goto done;
1558 		ia->ia_flags |= IFA_ROUTE;
1559 	}
1560 
1561 	/*
1562 	 * add a loopback route to self if not exists
1563 	 */
1564 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1565 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1566 		    (struct sockaddr *)&ia->ia_addr);
1567 		if (error == 0)
1568 			ia->ia_flags |= IFA_RTSELF;
1569 	}
1570 done:
1571 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1572 	    "Invoking IPv6 network device address event may sleep");
1573 
1574 	ifa_ref(&ia->ia_ifa);
1575 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1576 	    IFADDR_EVENT_ADD);
1577 	ifa_free(&ia->ia_ifa);
1578 
1579 	return (error);
1580 }
1581 
1582 /*
1583  * Find an IPv6 interface link-local address specific to an interface.
1584  * ifaddr is returned referenced.
1585  */
1586 struct in6_ifaddr *
1587 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1588 {
1589 	struct ifaddr *ifa;
1590 
1591 	NET_EPOCH_ASSERT();
1592 
1593 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1594 		if (ifa->ifa_addr->sa_family != AF_INET6)
1595 			continue;
1596 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1597 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1598 			    ignoreflags) != 0)
1599 				continue;
1600 			ifa_ref(ifa);
1601 			break;
1602 		}
1603 	}
1604 
1605 	return ((struct in6_ifaddr *)ifa);
1606 }
1607 
1608 /*
1609  * find the interface address corresponding to a given IPv6 address.
1610  * ifaddr is returned referenced if @referenced flag is set.
1611  */
1612 struct in6_ifaddr *
1613 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid, bool referenced)
1614 {
1615 	struct rm_priotracker in6_ifa_tracker;
1616 	struct in6_ifaddr *ia;
1617 
1618 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1619 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1620 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1621 			if (zoneid != 0 &&
1622 			    zoneid != ia->ia_addr.sin6_scope_id)
1623 				continue;
1624 			if (referenced)
1625 				ifa_ref(&ia->ia_ifa);
1626 			break;
1627 		}
1628 	}
1629 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1630 	return (ia);
1631 }
1632 
1633 /*
1634  * find the internet address corresponding to a given interface and address.
1635  * ifaddr is returned referenced.
1636  */
1637 struct in6_ifaddr *
1638 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1639 {
1640 	struct epoch_tracker et;
1641 	struct ifaddr *ifa;
1642 
1643 	NET_EPOCH_ENTER(et);
1644 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1645 		if (ifa->ifa_addr->sa_family != AF_INET6)
1646 			continue;
1647 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1648 			ifa_ref(ifa);
1649 			break;
1650 		}
1651 	}
1652 	NET_EPOCH_EXIT(et);
1653 
1654 	return ((struct in6_ifaddr *)ifa);
1655 }
1656 
1657 /*
1658  * Find a link-local scoped address on ifp and return it if any.
1659  */
1660 struct in6_ifaddr *
1661 in6ifa_llaonifp(struct ifnet *ifp)
1662 {
1663 	struct epoch_tracker et;
1664 	struct sockaddr_in6 *sin6;
1665 	struct ifaddr *ifa;
1666 
1667 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1668 		return (NULL);
1669 	NET_EPOCH_ENTER(et);
1670 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1671 		if (ifa->ifa_addr->sa_family != AF_INET6)
1672 			continue;
1673 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1674 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1675 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1676 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1677 			break;
1678 	}
1679 	NET_EPOCH_EXIT(et);
1680 
1681 	return ((struct in6_ifaddr *)ifa);
1682 }
1683 
1684 /*
1685  * Convert IP6 address to printable (loggable) representation. Caller
1686  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1687  */
1688 static char digits[] = "0123456789abcdef";
1689 char *
1690 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1691 {
1692 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1693 	char *cp;
1694 	const u_int16_t *a = (const u_int16_t *)addr;
1695 	const u_int8_t *d;
1696 	int dcolon = 0, zero = 0;
1697 
1698 	cp = ip6buf;
1699 
1700 	for (i = 0; i < 8; i++) {
1701 		if (*(a + i) == 0) {
1702 			cnt++;
1703 			if (cnt == 1)
1704 				idx = i;
1705 		}
1706 		else if (maxcnt < cnt) {
1707 			maxcnt = cnt;
1708 			index = idx;
1709 			cnt = 0;
1710 		}
1711 	}
1712 	if (maxcnt < cnt) {
1713 		maxcnt = cnt;
1714 		index = idx;
1715 	}
1716 
1717 	for (i = 0; i < 8; i++) {
1718 		if (dcolon == 1) {
1719 			if (*a == 0) {
1720 				if (i == 7)
1721 					*cp++ = ':';
1722 				a++;
1723 				continue;
1724 			} else
1725 				dcolon = 2;
1726 		}
1727 		if (*a == 0) {
1728 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1729 				if (i == 0)
1730 					*cp++ = ':';
1731 				*cp++ = ':';
1732 				dcolon = 1;
1733 			} else {
1734 				*cp++ = '0';
1735 				*cp++ = ':';
1736 			}
1737 			a++;
1738 			continue;
1739 		}
1740 		d = (const u_char *)a;
1741 		/* Try to eliminate leading zeros in printout like in :0001. */
1742 		zero = 1;
1743 		*cp = digits[*d >> 4];
1744 		if (*cp != '0') {
1745 			zero = 0;
1746 			cp++;
1747 		}
1748 		*cp = digits[*d++ & 0xf];
1749 		if (zero == 0 || (*cp != '0')) {
1750 			zero = 0;
1751 			cp++;
1752 		}
1753 		*cp = digits[*d >> 4];
1754 		if (zero == 0 || (*cp != '0')) {
1755 			zero = 0;
1756 			cp++;
1757 		}
1758 		*cp++ = digits[*d & 0xf];
1759 		*cp++ = ':';
1760 		a++;
1761 	}
1762 	*--cp = '\0';
1763 	return (ip6buf);
1764 }
1765 
1766 int
1767 in6_localaddr(struct in6_addr *in6)
1768 {
1769 	struct rm_priotracker in6_ifa_tracker;
1770 	struct in6_ifaddr *ia;
1771 
1772 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1773 		return 1;
1774 
1775 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1776 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1777 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1778 		    &ia->ia_prefixmask.sin6_addr)) {
1779 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1780 			return 1;
1781 		}
1782 	}
1783 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1784 
1785 	return (0);
1786 }
1787 
1788 /*
1789  * Return 1 if an internet address is for the local host and configured
1790  * on one of its interfaces.
1791  */
1792 int
1793 in6_localip(struct in6_addr *in6)
1794 {
1795 	struct rm_priotracker in6_ifa_tracker;
1796 	struct in6_ifaddr *ia;
1797 
1798 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1799 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1800 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1801 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1802 			return (1);
1803 		}
1804 	}
1805 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1806 	return (0);
1807 }
1808 
1809 /*
1810  * Like in6_localip(), but FIB-aware.
1811  */
1812 bool
1813 in6_localip_fib(struct in6_addr *in6, uint16_t fib)
1814 {
1815 	struct rm_priotracker in6_ifa_tracker;
1816 	struct in6_ifaddr *ia;
1817 
1818 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1819 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1820 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr) &&
1821 		    ia->ia_ifa.ifa_ifp->if_fib == fib) {
1822 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1823 			return (true);
1824 		}
1825 	}
1826 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1827 	return (false);
1828 }
1829 
1830 /*
1831  * Return 1 if an internet address is configured on an interface.
1832  */
1833 int
1834 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1835 {
1836 	struct in6_addr in6;
1837 	struct ifaddr *ifa;
1838 	struct in6_ifaddr *ia6;
1839 
1840 	NET_EPOCH_ASSERT();
1841 
1842 	in6 = *addr;
1843 	if (in6_clearscope(&in6))
1844 		return (0);
1845 	in6_setscope(&in6, ifp, NULL);
1846 
1847 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1848 		if (ifa->ifa_addr->sa_family != AF_INET6)
1849 			continue;
1850 		ia6 = (struct in6_ifaddr *)ifa;
1851 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1852 			return (1);
1853 	}
1854 
1855 	return (0);
1856 }
1857 
1858 int
1859 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1860 {
1861 	struct rm_priotracker in6_ifa_tracker;
1862 	struct in6_ifaddr *ia;
1863 
1864 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1865 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1866 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1867 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1868 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1869 				return (1); /* true */
1870 			}
1871 			break;
1872 		}
1873 	}
1874 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1875 
1876 	return (0);		/* false */
1877 }
1878 
1879 /*
1880  * return length of part which dst and src are equal
1881  * hard coding...
1882  */
1883 int
1884 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1885 {
1886 	int match = 0;
1887 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1888 	u_char *lim = s + 16, r;
1889 
1890 	while (s < lim)
1891 		if ((r = (*d++ ^ *s++)) != 0) {
1892 			while (r < 128) {
1893 				match++;
1894 				r <<= 1;
1895 			}
1896 			break;
1897 		} else
1898 			match += 8;
1899 	return match;
1900 }
1901 
1902 /* XXX: to be scope conscious */
1903 int
1904 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1905 {
1906 	int bytelen, bitlen;
1907 
1908 	/* sanity check */
1909 	if (0 > len || len > 128) {
1910 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1911 		    len);
1912 		return (0);
1913 	}
1914 
1915 	bytelen = len / 8;
1916 	bitlen = len % 8;
1917 
1918 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1919 		return (0);
1920 	if (bitlen != 0 &&
1921 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1922 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1923 		return (0);
1924 
1925 	return (1);
1926 }
1927 
1928 void
1929 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1930 {
1931 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1932 	int bytelen, bitlen, i;
1933 
1934 	/* sanity check */
1935 	if (0 > len || len > 128) {
1936 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1937 		    len);
1938 		return;
1939 	}
1940 
1941 	bzero(maskp, sizeof(*maskp));
1942 	bytelen = len / 8;
1943 	bitlen = len % 8;
1944 	for (i = 0; i < bytelen; i++)
1945 		maskp->s6_addr[i] = 0xff;
1946 	if (bitlen)
1947 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1948 }
1949 
1950 /*
1951  * return the best address out of the same scope. if no address was
1952  * found, return the first valid address from designated IF.
1953  */
1954 struct in6_ifaddr *
1955 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1956 {
1957 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1958 	struct ifaddr *ifa;
1959 	struct in6_ifaddr *besta = NULL;
1960 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1961 
1962 	NET_EPOCH_ASSERT();
1963 
1964 	dep[0] = dep[1] = NULL;
1965 
1966 	/*
1967 	 * We first look for addresses in the same scope.
1968 	 * If there is one, return it.
1969 	 * If two or more, return one which matches the dst longest.
1970 	 * If none, return one of global addresses assigned other ifs.
1971 	 */
1972 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1973 		if (ifa->ifa_addr->sa_family != AF_INET6)
1974 			continue;
1975 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1976 			continue; /* XXX: is there any case to allow anycast? */
1977 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1978 			continue; /* don't use this interface */
1979 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1980 			continue;
1981 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1982 			if (V_ip6_use_deprecated)
1983 				dep[0] = (struct in6_ifaddr *)ifa;
1984 			continue;
1985 		}
1986 
1987 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1988 			/*
1989 			 * call in6_matchlen() as few as possible
1990 			 */
1991 			if (besta) {
1992 				if (blen == -1)
1993 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1994 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1995 				if (tlen > blen) {
1996 					blen = tlen;
1997 					besta = (struct in6_ifaddr *)ifa;
1998 				}
1999 			} else
2000 				besta = (struct in6_ifaddr *)ifa;
2001 		}
2002 	}
2003 	if (besta)
2004 		return (besta);
2005 
2006 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2007 		if (ifa->ifa_addr->sa_family != AF_INET6)
2008 			continue;
2009 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2010 			continue; /* XXX: is there any case to allow anycast? */
2011 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2012 			continue; /* don't use this interface */
2013 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2014 			continue;
2015 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2016 			if (V_ip6_use_deprecated)
2017 				dep[1] = (struct in6_ifaddr *)ifa;
2018 			continue;
2019 		}
2020 
2021 		return (struct in6_ifaddr *)ifa;
2022 	}
2023 
2024 	/* use the last-resort values, that are, deprecated addresses */
2025 	if (dep[0])
2026 		return dep[0];
2027 	if (dep[1])
2028 		return dep[1];
2029 
2030 	return NULL;
2031 }
2032 
2033 /*
2034  * perform DAD when interface becomes IFF_UP.
2035  */
2036 void
2037 in6_if_up(struct ifnet *ifp)
2038 {
2039 	struct epoch_tracker et;
2040 	struct ifaddr *ifa;
2041 	struct in6_ifaddr *ia;
2042 
2043 	NET_EPOCH_ENTER(et);
2044 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2045 		if (ifa->ifa_addr->sa_family != AF_INET6)
2046 			continue;
2047 		ia = (struct in6_ifaddr *)ifa;
2048 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2049 			/*
2050 			 * The TENTATIVE flag was likely set by hand
2051 			 * beforehand, implicitly indicating the need for DAD.
2052 			 * We may be able to skip the random delay in this
2053 			 * case, but we impose delays just in case.
2054 			 */
2055 			nd6_dad_start(ifa,
2056 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2057 		}
2058 	}
2059 	NET_EPOCH_EXIT(et);
2060 
2061 	/*
2062 	 * special cases, like 6to4, are handled in in6_ifattach
2063 	 */
2064 	in6_ifattach(ifp, NULL);
2065 }
2066 
2067 static void
2068 in6_ifevent(void *arg __unused, struct ifnet *ifp, int event)
2069 {
2070 	if (event == IFNET_EVENT_UP)
2071 		in6_if_up(ifp);
2072 }
2073 
2074 static void
2075 in6_init(void *arg __unused)
2076 {
2077 	EVENTHANDLER_REGISTER(ifnet_event, in6_ifevent, NULL, EVENTHANDLER_PRI_ANY);
2078 }
2079 SYSINIT(in6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, in6_init, NULL);
2080 
2081 int
2082 in6if_do_dad(struct ifnet *ifp)
2083 {
2084 
2085 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2086 		return (0);
2087 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
2088 		return (0);
2089 	if ((ND_IFINFO(ifp)->flags &
2090 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
2091 		return (0);
2092 	return (1);
2093 }
2094 
2095 /*
2096  * Calculate max IPv6 MTU through all the interfaces and store it
2097  * to in6_maxmtu.
2098  */
2099 void
2100 in6_setmaxmtu(void)
2101 {
2102 	struct epoch_tracker et;
2103 	unsigned long maxmtu = 0;
2104 	struct ifnet *ifp;
2105 
2106 	NET_EPOCH_ENTER(et);
2107 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2108 		/* this function can be called during ifnet initialization */
2109 		if (!ifp->if_afdata[AF_INET6])
2110 			continue;
2111 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2112 		    IN6_LINKMTU(ifp) > maxmtu)
2113 			maxmtu = IN6_LINKMTU(ifp);
2114 	}
2115 	NET_EPOCH_EXIT(et);
2116 	if (maxmtu)	/* update only when maxmtu is positive */
2117 		V_in6_maxmtu = maxmtu;
2118 }
2119 
2120 /*
2121  * Provide the length of interface identifiers to be used for the link attached
2122  * to the given interface.  The length should be defined in "IPv6 over
2123  * xxx-link" document.  Note that address architecture might also define
2124  * the length for a particular set of address prefixes, regardless of the
2125  * link type.  As clarified in rfc2462bis, those two definitions should be
2126  * consistent, and those really are as of August 2004.
2127  */
2128 int
2129 in6_if2idlen(struct ifnet *ifp)
2130 {
2131 	switch (ifp->if_type) {
2132 	case IFT_ETHER:		/* RFC2464 */
2133 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2134 	case IFT_L2VLAN:	/* ditto */
2135 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2136 	case IFT_INFINIBAND:
2137 		return (64);
2138 	case IFT_PPP:		/* RFC2472 */
2139 		return (64);
2140 	case IFT_FRELAY:	/* RFC2590 */
2141 		return (64);
2142 	case IFT_IEEE1394:	/* RFC3146 */
2143 		return (64);
2144 	case IFT_GIF:
2145 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2146 	case IFT_LOOP:
2147 		return (64);	/* XXX: is this really correct? */
2148 	default:
2149 		/*
2150 		 * Unknown link type:
2151 		 * It might be controversial to use the today's common constant
2152 		 * of 64 for these cases unconditionally.  For full compliance,
2153 		 * we should return an error in this case.  On the other hand,
2154 		 * if we simply miss the standard for the link type or a new
2155 		 * standard is defined for a new link type, the IFID length
2156 		 * is very likely to be the common constant.  As a compromise,
2157 		 * we always use the constant, but make an explicit notice
2158 		 * indicating the "unknown" case.
2159 		 */
2160 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2161 		return (64);
2162 	}
2163 }
2164 
2165 struct in6_llentry {
2166 	struct llentry		base;
2167 };
2168 
2169 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2170 #define	IN6_LLTBL_HASH(k, h) \
2171 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2172 
2173 /*
2174  * Do actual deallocation of @lle.
2175  */
2176 static void
2177 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2178 {
2179 	struct llentry *lle;
2180 
2181 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2182 	LLE_LOCK_DESTROY(lle);
2183 	LLE_REQ_DESTROY(lle);
2184 	free(lle, M_LLTABLE);
2185 }
2186 
2187 /*
2188  * Called by LLE_FREE_LOCKED when number of references
2189  * drops to zero.
2190  */
2191 static void
2192 in6_lltable_destroy_lle(struct llentry *lle)
2193 {
2194 
2195 	LLE_WUNLOCK(lle);
2196 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2197 }
2198 
2199 static struct llentry *
2200 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2201 {
2202 	struct in6_llentry *lle;
2203 
2204 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2205 	if (lle == NULL)		/* NB: caller generates msg */
2206 		return NULL;
2207 
2208 	lle->base.r_l3addr.addr6 = *addr6;
2209 	lle->base.lle_refcnt = 1;
2210 	lle->base.lle_free = in6_lltable_destroy_lle;
2211 	LLE_LOCK_INIT(&lle->base);
2212 	LLE_REQ_INIT(&lle->base);
2213 	callout_init(&lle->base.lle_timer, 1);
2214 
2215 	return (&lle->base);
2216 }
2217 
2218 static int
2219 in6_lltable_match_prefix(const struct sockaddr *saddr,
2220     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2221 {
2222 	const struct in6_addr *addr, *mask, *lle_addr;
2223 
2224 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2225 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2226 	lle_addr = &lle->r_l3addr.addr6;
2227 
2228 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2229 		return (0);
2230 
2231 	if (lle->la_flags & LLE_IFADDR) {
2232 		/*
2233 		 * Delete LLE_IFADDR records IFF address & flag matches.
2234 		 * Note that addr is the interface address within prefix
2235 		 * being matched.
2236 		 */
2237 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2238 		    (flags & LLE_STATIC) != 0)
2239 			return (1);
2240 		return (0);
2241 	}
2242 
2243 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2244 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2245 		return (1);
2246 
2247 	return (0);
2248 }
2249 
2250 static void
2251 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2252 {
2253 	struct ifnet *ifp __diagused;
2254 
2255 	LLE_WLOCK_ASSERT(lle);
2256 	KASSERT(llt != NULL, ("lltable is NULL"));
2257 
2258 	/* Unlink entry from table */
2259 	if ((lle->la_flags & LLE_LINKED) != 0) {
2260 		ifp = llt->llt_ifp;
2261 		IF_AFDATA_WLOCK_ASSERT(ifp);
2262 		lltable_unlink_entry(llt, lle);
2263 	}
2264 
2265 	llentry_free(lle);
2266 }
2267 
2268 static int
2269 in6_lltable_rtcheck(struct ifnet *ifp,
2270 		    u_int flags,
2271 		    const struct sockaddr *l3addr)
2272 {
2273 	const struct sockaddr_in6 *sin6;
2274 	struct nhop_object *nh;
2275 	struct in6_addr dst;
2276 	uint32_t scopeid;
2277 	char ip6buf[INET6_ADDRSTRLEN];
2278 	int fibnum;
2279 
2280 	NET_EPOCH_ASSERT();
2281 	KASSERT(l3addr->sa_family == AF_INET6,
2282 	    ("sin_family %d", l3addr->sa_family));
2283 
2284 	sin6 = (const struct sockaddr_in6 *)l3addr;
2285 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2286 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2287 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2288 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2289 		struct ifaddr *ifa;
2290 		/*
2291 		 * Create an ND6 cache for an IPv6 neighbor
2292 		 * that is not covered by our own prefix.
2293 		 */
2294 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2295 		if (ifa != NULL) {
2296 			return 0;
2297 		}
2298 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2299 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2300 		return EINVAL;
2301 	}
2302 	return 0;
2303 }
2304 
2305 static inline uint32_t
2306 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2307 {
2308 
2309 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2310 }
2311 
2312 static uint32_t
2313 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2314 {
2315 
2316 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2317 }
2318 
2319 static void
2320 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2321 {
2322 	struct sockaddr_in6 *sin6;
2323 
2324 	sin6 = (struct sockaddr_in6 *)sa;
2325 	bzero(sin6, sizeof(*sin6));
2326 	sin6->sin6_family = AF_INET6;
2327 	sin6->sin6_len = sizeof(*sin6);
2328 	sin6->sin6_addr = lle->r_l3addr.addr6;
2329 }
2330 
2331 static inline struct llentry *
2332 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2333 {
2334 	struct llentry *lle;
2335 	struct llentries *lleh;
2336 	u_int hashidx;
2337 
2338 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2339 	lleh = &llt->lle_head[hashidx];
2340 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2341 		if (lle->la_flags & LLE_DELETED)
2342 			continue;
2343 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2344 			break;
2345 	}
2346 
2347 	return (lle);
2348 }
2349 
2350 static void
2351 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2352 {
2353 
2354 	lle->la_flags |= LLE_DELETED;
2355 
2356 	/* Leave the solicited multicast group. */
2357 	if ((lle->la_flags & LLE_PUB) != 0)
2358 		in6_leave_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2359 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2360 #ifdef DIAGNOSTIC
2361 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2362 #endif
2363 	llentry_free(lle);
2364 }
2365 
2366 static struct llentry *
2367 in6_lltable_alloc(struct lltable *llt, u_int flags,
2368 	const struct sockaddr *l3addr)
2369 {
2370 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2371 	struct ifnet *ifp = llt->llt_ifp;
2372 	struct llentry *lle;
2373 	char linkhdr[LLE_MAX_LINKHDR];
2374 	size_t linkhdrsize;
2375 	int lladdr_off;
2376 
2377 	KASSERT(l3addr->sa_family == AF_INET6,
2378 	    ("sin_family %d", l3addr->sa_family));
2379 
2380 	/*
2381 	 * A route that covers the given address must have
2382 	 * been installed 1st because we are doing a resolution,
2383 	 * verify this.
2384 	 */
2385 	if (!(flags & LLE_IFADDR) &&
2386 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2387 		return (NULL);
2388 
2389 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2390 	if (lle == NULL) {
2391 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2392 		return (NULL);
2393 	}
2394 	lle->la_flags = flags;
2395 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2396 		linkhdrsize = LLE_MAX_LINKHDR;
2397 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2398 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2399 			in6_lltable_free_entry(llt, lle);
2400 			return (NULL);
2401 		}
2402 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2403 		    lladdr_off);
2404 		lle->la_flags |= LLE_STATIC;
2405 	}
2406 
2407 	if ((lle->la_flags & LLE_STATIC) != 0)
2408 		lle->ln_state = ND6_LLINFO_REACHABLE;
2409 
2410 	return (lle);
2411 }
2412 
2413 static struct llentry *
2414 in6_lltable_lookup(struct lltable *llt, u_int flags,
2415 	const struct sockaddr *l3addr)
2416 {
2417 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2418 	int family = flags >> 16;
2419 	struct llentry *lle;
2420 
2421 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2422 	KASSERT(l3addr->sa_family == AF_INET6,
2423 	    ("sin_family %d", l3addr->sa_family));
2424 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2425 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2426 	    ("wrong lle request flags: %#x", flags));
2427 
2428 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2429 
2430 	if (__predict_false(family != AF_INET6))
2431 		lle = llentry_lookup_family(lle, family);
2432 
2433 	if (lle == NULL)
2434 		return (NULL);
2435 
2436 	if (flags & LLE_UNLOCKED)
2437 		return (lle);
2438 
2439 	if (flags & LLE_EXCLUSIVE)
2440 		LLE_WLOCK(lle);
2441 	else
2442 		LLE_RLOCK(lle);
2443 
2444 	/*
2445 	 * If the afdata lock is not held, the LLE may have been unlinked while
2446 	 * we were blocked on the LLE lock.  Check for this case.
2447 	 */
2448 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2449 		if (flags & LLE_EXCLUSIVE)
2450 			LLE_WUNLOCK(lle);
2451 		else
2452 			LLE_RUNLOCK(lle);
2453 		return (NULL);
2454 	}
2455 	return (lle);
2456 }
2457 
2458 static int
2459 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2460     struct sysctl_req *wr)
2461 {
2462 	struct ifnet *ifp = llt->llt_ifp;
2463 	/* XXX stack use */
2464 	struct {
2465 		struct rt_msghdr	rtm;
2466 		struct sockaddr_in6	sin6;
2467 		/*
2468 		 * ndp.c assumes that sdl is word aligned
2469 		 */
2470 #ifdef __LP64__
2471 		uint32_t		pad;
2472 #endif
2473 		struct sockaddr_dl	sdl;
2474 	} ndpc;
2475 	struct sockaddr_dl *sdl;
2476 	int error;
2477 
2478 	bzero(&ndpc, sizeof(ndpc));
2479 	/* skip deleted entries */
2480 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2481 		return (0);
2482 	/* Skip if jailed and not a valid IP of the prison. */
2483 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2484 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2485 		return (0);
2486 	/*
2487 	 * produce a msg made of:
2488 	 *  struct rt_msghdr;
2489 	 *  struct sockaddr_in6 (IPv6)
2490 	 *  struct sockaddr_dl;
2491 	 */
2492 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2493 	ndpc.rtm.rtm_version = RTM_VERSION;
2494 	ndpc.rtm.rtm_type = RTM_GET;
2495 	ndpc.rtm.rtm_flags = RTF_UP;
2496 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2497 	sa6_recoverscope(&ndpc.sin6);
2498 
2499 	/* publish */
2500 	if (lle->la_flags & LLE_PUB)
2501 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2502 
2503 	sdl = &ndpc.sdl;
2504 	sdl->sdl_family = AF_LINK;
2505 	sdl->sdl_len = sizeof(*sdl);
2506 	sdl->sdl_index = ifp->if_index;
2507 	sdl->sdl_type = ifp->if_type;
2508 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2509 		sdl->sdl_alen = ifp->if_addrlen;
2510 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2511 	} else {
2512 		sdl->sdl_alen = 0;
2513 		bzero(LLADDR(sdl), ifp->if_addrlen);
2514 	}
2515 	if (lle->la_expire != 0)
2516 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2517 		    lle->lle_remtime / hz + time_second - time_uptime;
2518 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2519 	if (lle->la_flags & LLE_STATIC)
2520 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2521 	if (lle->la_flags & LLE_IFADDR)
2522 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2523 	if (lle->ln_router != 0)
2524 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2525 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2526 	/* Store state in rmx_weight value */
2527 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2528 	ndpc.rtm.rtm_index = ifp->if_index;
2529 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2530 
2531 	return (error);
2532 }
2533 
2534 static void
2535 in6_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
2536 {
2537 	/* Join the solicited multicast group for dst. */
2538 	if ((lle->la_flags & LLE_PUB) == LLE_PUB)
2539 		in6_join_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2540 }
2541 
2542 static struct lltable *
2543 in6_lltattach(struct ifnet *ifp)
2544 {
2545 	struct lltable *llt;
2546 
2547 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2548 	llt->llt_af = AF_INET6;
2549 	llt->llt_ifp = ifp;
2550 
2551 	llt->llt_lookup = in6_lltable_lookup;
2552 	llt->llt_alloc_entry = in6_lltable_alloc;
2553 	llt->llt_delete_entry = in6_lltable_delete_entry;
2554 	llt->llt_dump_entry = in6_lltable_dump_entry;
2555 	llt->llt_hash = in6_lltable_hash;
2556 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2557 	llt->llt_free_entry = in6_lltable_free_entry;
2558 	llt->llt_match_prefix = in6_lltable_match_prefix;
2559 	llt->llt_mark_used = llentry_mark_used;
2560 	llt->llt_post_resolved = in6_lltable_post_resolved;
2561  	lltable_link(llt);
2562 
2563 	return (llt);
2564 }
2565 
2566 struct lltable *
2567 in6_lltable_get(struct ifnet *ifp)
2568 {
2569 	struct lltable *llt = NULL;
2570 
2571 	void *afdata_ptr = ifp->if_afdata[AF_INET6];
2572 	if (afdata_ptr != NULL)
2573 		llt = ((struct in6_ifextra *)afdata_ptr)->lltable;
2574 	return (llt);
2575 }
2576 
2577 void *
2578 in6_domifattach(struct ifnet *ifp)
2579 {
2580 	struct in6_ifextra *ext;
2581 
2582 	/* There are not IPv6-capable interfaces. */
2583 	switch (ifp->if_type) {
2584 	case IFT_PFLOG:
2585 	case IFT_PFSYNC:
2586 	case IFT_USB:
2587 		return (NULL);
2588 	}
2589 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2590 	bzero(ext, sizeof(*ext));
2591 
2592 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2593 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2594 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2595 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2596 
2597 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2598 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2599 	    M_WAITOK);
2600 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2601 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2602 
2603 	ext->nd_ifinfo = nd6_ifattach(ifp);
2604 	ext->scope6_id = scope6_ifattach(ifp);
2605 	ext->lltable = in6_lltattach(ifp);
2606 
2607 	ext->mld_ifinfo = mld_domifattach(ifp);
2608 
2609 	return ext;
2610 }
2611 
2612 int
2613 in6_domifmtu(struct ifnet *ifp)
2614 {
2615 	if (ifp->if_afdata[AF_INET6] == NULL)
2616 		return ifp->if_mtu;
2617 
2618 	return (IN6_LINKMTU(ifp));
2619 }
2620 
2621 void
2622 in6_domifdetach(struct ifnet *ifp, void *aux)
2623 {
2624 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2625 
2626 	mld_domifdetach(ifp);
2627 	scope6_ifdetach(ext->scope6_id);
2628 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2629 	lltable_free(ext->lltable);
2630 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2631 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2632 	free(ext->in6_ifstat, M_IFADDR);
2633 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2634 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2635 	free(ext->icmp6_ifstat, M_IFADDR);
2636 	free(ext, M_IFADDR);
2637 }
2638 
2639 /*
2640  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2641  * v4 mapped addr or v4 compat addr
2642  */
2643 void
2644 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2645 {
2646 
2647 	bzero(sin, sizeof(*sin));
2648 	sin->sin_len = sizeof(struct sockaddr_in);
2649 	sin->sin_family = AF_INET;
2650 	sin->sin_port = sin6->sin6_port;
2651 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2652 }
2653 
2654 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2655 void
2656 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2657 {
2658 	bzero(sin6, sizeof(*sin6));
2659 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2660 	sin6->sin6_family = AF_INET6;
2661 	sin6->sin6_port = sin->sin_port;
2662 	sin6->sin6_addr.s6_addr32[0] = 0;
2663 	sin6->sin6_addr.s6_addr32[1] = 0;
2664 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2665 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2666 }
2667 
2668 /* Convert sockaddr_in6 into sockaddr_in. */
2669 void
2670 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2671 {
2672 	struct sockaddr_in *sin_p;
2673 	struct sockaddr_in6 sin6;
2674 
2675 	/*
2676 	 * Save original sockaddr_in6 addr and convert it
2677 	 * to sockaddr_in.
2678 	 */
2679 	sin6 = *(struct sockaddr_in6 *)nam;
2680 	sin_p = (struct sockaddr_in *)nam;
2681 	in6_sin6_2_sin(sin_p, &sin6);
2682 }
2683 
2684 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2685 void
2686 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2687 {
2688 	struct sockaddr_in *sin_p;
2689 	struct sockaddr_in6 *sin6_p;
2690 
2691 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2692 	sin_p = (struct sockaddr_in *)*nam;
2693 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2694 	free(*nam, M_SONAME);
2695 	*nam = (struct sockaddr *)sin6_p;
2696 }
2697 
2698 /*
2699  * Join/leave the solicited multicast groups for proxy NDP entries.
2700  */
2701 static void
2702 in6_join_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2703 {
2704 	struct in6_multi *inm;
2705 	struct in6_addr mltaddr;
2706 	char ip6buf[INET6_ADDRSTRLEN];
2707 	int error;
2708 
2709 	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2710 		return;	/* error logged in in6_solicited_node_maddr. */
2711 
2712 	error = in6_joingroup(ifp, &mltaddr, NULL, &inm, 0);
2713 	if (error != 0) {
2714 		nd6log((LOG_WARNING,
2715 		    "%s: in6_joingroup failed for %s on %s (errno=%d)\n",
2716 		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp),
2717 		    error));
2718 	}
2719 }
2720 
2721 static void
2722 in6_leave_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2723 {
2724 	struct epoch_tracker et;
2725 	struct in6_multi *inm;
2726 	struct in6_addr mltaddr;
2727 	char ip6buf[INET6_ADDRSTRLEN];
2728 
2729 	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2730 		return;	/* error logged in in6_solicited_node_maddr. */
2731 
2732 	NET_EPOCH_ENTER(et);
2733 	inm = in6m_lookup(ifp, &mltaddr);
2734 	NET_EPOCH_EXIT(et);
2735 	if (inm != NULL)
2736 		in6_leavegroup(inm, NULL);
2737 	else
2738 		nd6log((LOG_WARNING, "%s: in6m_lookup failed for %s on %s\n",
2739 		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp)));
2740 }
2741 
2742 static bool
2743 in6_lle_match_pub(struct lltable *llt, struct llentry *lle, void *farg)
2744 {
2745 	return ((lle->la_flags & LLE_PUB) != 0);
2746 }
2747 
2748 void
2749 in6_purge_proxy_ndp(struct ifnet *ifp)
2750 {
2751 	struct lltable *llt;
2752 	bool need_purge;
2753 
2754 	if (ifp->if_afdata[AF_INET6] == NULL)
2755 		return;
2756 
2757 	llt = LLTABLE6(ifp);
2758 	IF_AFDATA_WLOCK(ifp);
2759 	need_purge = ((llt->llt_flags & LLT_ADDEDPROXY) != 0);
2760 	IF_AFDATA_WUNLOCK(ifp);
2761 
2762 	/*
2763 	 * Ever added proxy ndp entries, leave solicited node multicast
2764 	 * before deleting the llentry.
2765 	 */
2766 	if (need_purge)
2767 		lltable_delete_conditional(llt, in6_lle_match_pub, NULL);
2768 }
2769