xref: /freebsd/sys/netinet6/in6.c (revision 350b7c3570aa6c87c537e54f706f1866f93a4142)
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_ioctl(u_long cmd, void *data,
253     struct ifnet *ifp, struct ucred *cred)
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 (cred != NULL) {
285 			error = priv_check_cred(cred, 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 (cred != NULL) {
303 			error = priv_check_cred(cred, 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 (cred != NULL) {
347 			error = priv_check_cred(cred, 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 (cred != NULL && (error = prison_check_ip6(cred,
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 (cred != NULL) {
461 			error = priv_check_cred(cred, (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 int
600 in6_control(struct socket *so, u_long cmd, void *data,
601     struct ifnet *ifp, struct thread *td)
602 {
603 	return (in6_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL));
604 }
605 
606 static struct in6_multi_mship *
607 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
608     int *errorp, int delay)
609 {
610 	struct in6_multi_mship *imm;
611 	int error;
612 
613 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
614 	if (imm == NULL) {
615 		*errorp = ENOBUFS;
616 		return (NULL);
617 	}
618 
619 	delay = (delay * MLD_FASTHZ) / hz;
620 
621 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
622 	if (error) {
623 		*errorp = error;
624 		free(imm, M_IP6MADDR);
625 		return (NULL);
626 	}
627 
628 	return (imm);
629 }
630 
631 static int
632 in6_solicited_node_maddr(struct in6_addr *maddr,
633     struct ifnet *ifp, const struct in6_addr *base)
634 {
635 	int error;
636 
637 	bzero(maddr, sizeof(struct in6_addr));
638 	maddr->s6_addr32[0] = IPV6_ADDR_INT32_MLL;
639 	maddr->s6_addr32[2] = htonl(1);
640 	maddr->s6_addr32[3] = base->s6_addr32[3];
641 	maddr->s6_addr8[12] = 0xff;
642 	if ((error = in6_setscope(maddr, ifp, NULL)) != 0) {
643 		/* XXX: should not happen */
644 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
645 	}
646 
647 	return error;
648 }
649 
650 /*
651  * Join necessary multicast groups.  Factored out from in6_update_ifa().
652  * This entire work should only be done once, for the default FIB.
653  */
654 static int
655 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
656     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
657 {
658 	char ip6buf[INET6_ADDRSTRLEN];
659 	struct in6_addr mltaddr;
660 	struct in6_multi_mship *imm;
661 	int delay, error;
662 
663 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
664 
665 	/* Join solicited multicast addr for new host id. */
666 	if ((error = in6_solicited_node_maddr(&mltaddr, ifp,
667 	    &ifra->ifra_addr.sin6_addr)) != 0)
668 		goto cleanup;
669 	delay = error = 0;
670 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
671 		/*
672 		 * We need a random delay for DAD on the address being
673 		 * configured.  It also means delaying transmission of the
674 		 * corresponding MLD report to avoid report collision.
675 		 * [RFC 4861, Section 6.3.7]
676 		 */
677 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
678 	}
679 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
680 	if (imm == NULL) {
681 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
682 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
683 		    if_name(ifp), error));
684 		goto cleanup;
685 	}
686 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
687 	*in6m_sol = imm->i6mm_maddr;
688 
689 	/*
690 	 * Join link-local all-nodes address.
691 	 */
692 	mltaddr = in6addr_linklocal_allnodes;
693 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
694 		goto cleanup; /* XXX: should not fail */
695 
696 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
697 	if (imm == NULL) {
698 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
699 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
700 		    if_name(ifp), error));
701 		goto cleanup;
702 	}
703 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
704 
705 	/*
706 	 * Join node information group address.
707 	 */
708 	delay = 0;
709 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
710 		/*
711 		 * The spec does not say anything about delay for this group,
712 		 * but the same logic should apply.
713 		 */
714 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
715 	}
716 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
717 		/* XXX jinmei */
718 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
719 		if (imm == NULL)
720 			nd6log((LOG_WARNING,
721 			    "%s: in6_joingroup failed for %s on %s "
722 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
723 			    &mltaddr), if_name(ifp), error));
724 			/* XXX not very fatal, go on... */
725 		else
726 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
727 	}
728 	if (V_icmp6_nodeinfo_oldmcprefix &&
729 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
730 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
731 		if (imm == NULL)
732 			nd6log((LOG_WARNING,
733 			    "%s: in6_joingroup failed for %s on %s "
734 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
735 			    &mltaddr), if_name(ifp), error));
736 			/* XXX not very fatal, go on... */
737 		else
738 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
739 	}
740 
741 	/*
742 	 * Join interface-local all-nodes address.
743 	 * (ff01::1%ifN, and ff01::%ifN/32)
744 	 */
745 	mltaddr = in6addr_nodelocal_allnodes;
746 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
747 		goto cleanup; /* XXX: should not fail */
748 
749 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
750 	if (imm == NULL) {
751 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
752 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
753 		    &mltaddr), if_name(ifp), error));
754 		goto cleanup;
755 	}
756 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
757 
758 cleanup:
759 	return (error);
760 }
761 
762 /*
763  * Update parameters of an IPv6 interface address.
764  * If necessary, a new entry is created and linked into address chains.
765  * This function is separated from in6_control().
766  */
767 int
768 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
769     struct in6_ifaddr *ia, int flags)
770 {
771 	int error, hostIsNew = 0;
772 
773 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
774 		return (error);
775 
776 	if (ia == NULL) {
777 		hostIsNew = 1;
778 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
779 			return (ENOBUFS);
780 	}
781 
782 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
783 	if (error != 0) {
784 		if (hostIsNew != 0) {
785 			in6_unlink_ifa(ia, ifp);
786 			ifa_free(&ia->ia_ifa);
787 		}
788 		return (error);
789 	}
790 
791 	if (hostIsNew)
792 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
793 
794 	return (error);
795 }
796 
797 /*
798  * Fill in basic IPv6 address request info.
799  */
800 void
801 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
802     const struct in6_addr *mask)
803 {
804 
805 	memset(ifra, 0, sizeof(struct in6_aliasreq));
806 
807 	ifra->ifra_addr.sin6_family = AF_INET6;
808 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
809 	if (addr != NULL)
810 		ifra->ifra_addr.sin6_addr = *addr;
811 
812 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
813 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
814 	if (mask != NULL)
815 		ifra->ifra_prefixmask.sin6_addr = *mask;
816 }
817 
818 static int
819 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
820     struct in6_ifaddr *ia, int flags)
821 {
822 	int plen = -1;
823 	struct sockaddr_in6 dst6;
824 	struct in6_addrlifetime *lt;
825 	char ip6buf[INET6_ADDRSTRLEN];
826 
827 	/* Validate parameters */
828 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
829 		return (EINVAL);
830 
831 	/*
832 	 * The destination address for a p2p link must have a family
833 	 * of AF_UNSPEC or AF_INET6.
834 	 */
835 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
836 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
837 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
838 		return (EAFNOSUPPORT);
839 
840 	/*
841 	 * Validate address
842 	 */
843 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
844 	    ifra->ifra_addr.sin6_family != AF_INET6)
845 		return (EINVAL);
846 
847 	/*
848 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
849 	 * does not carry fields other than sin6_len.
850 	 */
851 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
852 		return (EINVAL);
853 	/*
854 	 * Because the IPv6 address architecture is classless, we require
855 	 * users to specify a (non 0) prefix length (mask) for a new address.
856 	 * We also require the prefix (when specified) mask is valid, and thus
857 	 * reject a non-consecutive mask.
858 	 */
859 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
860 		return (EINVAL);
861 	if (ifra->ifra_prefixmask.sin6_len != 0) {
862 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
863 		    (u_char *)&ifra->ifra_prefixmask +
864 		    ifra->ifra_prefixmask.sin6_len);
865 		if (plen <= 0)
866 			return (EINVAL);
867 	} else {
868 		/*
869 		 * In this case, ia must not be NULL.  We just use its prefix
870 		 * length.
871 		 */
872 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
873 	}
874 	/*
875 	 * If the destination address on a p2p interface is specified,
876 	 * and the address is a scoped one, validate/set the scope
877 	 * zone identifier.
878 	 */
879 	dst6 = ifra->ifra_dstaddr;
880 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
881 	    (dst6.sin6_family == AF_INET6)) {
882 		struct in6_addr in6_tmp;
883 		u_int32_t zoneid;
884 
885 		in6_tmp = dst6.sin6_addr;
886 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
887 			return (EINVAL); /* XXX: should be impossible */
888 
889 		if (dst6.sin6_scope_id != 0) {
890 			if (dst6.sin6_scope_id != zoneid)
891 				return (EINVAL);
892 		} else		/* user omit to specify the ID. */
893 			dst6.sin6_scope_id = zoneid;
894 
895 		/* convert into the internal form */
896 		if (sa6_embedscope(&dst6, 0))
897 			return (EINVAL); /* XXX: should be impossible */
898 	}
899 	/* Modify original ifra_dstaddr to reflect changes */
900 	ifra->ifra_dstaddr = dst6;
901 
902 	/*
903 	 * The destination address can be specified only for a p2p or a
904 	 * loopback interface.  If specified, the corresponding prefix length
905 	 * must be 128.
906 	 */
907 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
908 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
909 			/* XXX: noisy message */
910 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
911 			    "be specified for a p2p or a loopback IF only\n"));
912 			return (EINVAL);
913 		}
914 		if (plen != 128) {
915 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
916 			    "be 128 when dstaddr is specified\n"));
917 			return (EINVAL);
918 		}
919 	}
920 	/* lifetime consistency check */
921 	lt = &ifra->ifra_lifetime;
922 	if (lt->ia6t_pltime > lt->ia6t_vltime)
923 		return (EINVAL);
924 	if (lt->ia6t_vltime == 0) {
925 		/*
926 		 * the following log might be noisy, but this is a typical
927 		 * configuration mistake or a tool's bug.
928 		 */
929 		nd6log((LOG_INFO,
930 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
931 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
932 
933 		if (ia == NULL)
934 			return (0); /* there's nothing to do */
935 	}
936 
937 	/* Check prefix mask */
938 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
939 		/*
940 		 * We prohibit changing the prefix length of an existing
941 		 * address, because
942 		 * + such an operation should be rare in IPv6, and
943 		 * + the operation would confuse prefix management.
944 		 */
945 		if (ia->ia_prefixmask.sin6_len != 0 &&
946 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
947 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
948 			    "of an existing %s address should not be changed\n",
949 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
950 
951 			return (EINVAL);
952 		}
953 	}
954 
955 	return (0);
956 }
957 
958 /*
959  * Allocate a new ifaddr and link it into chains.
960  */
961 static struct in6_ifaddr *
962 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
963 {
964 	struct in6_ifaddr *ia;
965 
966 	/*
967 	 * When in6_alloc_ifa() is called in a process of a received
968 	 * RA, it is called under an interrupt context.  So, we should
969 	 * call malloc with M_NOWAIT.
970 	 */
971 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
972 	if (ia == NULL)
973 		return (NULL);
974 	LIST_INIT(&ia->ia6_memberships);
975 	/* Initialize the address and masks, and put time stamp */
976 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
977 	ia->ia_addr.sin6_family = AF_INET6;
978 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
979 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
980 	ia->ia_addr = ifra->ifra_addr;
981 	ia->ia6_createtime = time_uptime;
982 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
983 		/*
984 		 * Some functions expect that ifa_dstaddr is not
985 		 * NULL for p2p interfaces.
986 		 */
987 		ia->ia_ifa.ifa_dstaddr =
988 		    (struct sockaddr *)&ia->ia_dstaddr;
989 	} else {
990 		ia->ia_ifa.ifa_dstaddr = NULL;
991 	}
992 
993 	/* set prefix mask if any */
994 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
995 	if (ifra->ifra_prefixmask.sin6_len != 0) {
996 		ia->ia_prefixmask.sin6_family = AF_INET6;
997 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
998 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
999 	}
1000 
1001 	ia->ia_ifp = ifp;
1002 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1003 	IF_ADDR_WLOCK(ifp);
1004 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1005 	IF_ADDR_WUNLOCK(ifp);
1006 
1007 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1008 	IN6_IFADDR_WLOCK();
1009 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1010 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1011 	IN6_IFADDR_WUNLOCK();
1012 
1013 	return (ia);
1014 }
1015 
1016 /*
1017  * Update/configure interface address parameters:
1018  *
1019  * 1) Update lifetime
1020  * 2) Update interface metric ad flags
1021  * 3) Notify other subsystems
1022  */
1023 static int
1024 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1025     struct in6_ifaddr *ia, int hostIsNew, int flags)
1026 {
1027 	int error;
1028 
1029 	/* update timestamp */
1030 	ia->ia6_updatetime = time_uptime;
1031 
1032 	/*
1033 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1034 	 * to see if the address is deprecated or invalidated, but initialize
1035 	 * these members for applications.
1036 	 */
1037 	ia->ia6_lifetime = ifra->ifra_lifetime;
1038 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1039 		ia->ia6_lifetime.ia6t_expire =
1040 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1041 	} else
1042 		ia->ia6_lifetime.ia6t_expire = 0;
1043 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1044 		ia->ia6_lifetime.ia6t_preferred =
1045 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1046 	} else
1047 		ia->ia6_lifetime.ia6t_preferred = 0;
1048 
1049 	/*
1050 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1051 	 * userland, make it deprecated.
1052 	 */
1053 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1054 		ia->ia6_lifetime.ia6t_pltime = 0;
1055 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1056 	}
1057 
1058 	/*
1059 	 * configure address flags.
1060 	 */
1061 	ia->ia6_flags = ifra->ifra_flags;
1062 
1063 	/*
1064 	 * Make the address tentative before joining multicast addresses,
1065 	 * so that corresponding MLD responses would not have a tentative
1066 	 * source address.
1067 	 */
1068 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1069 
1070 	/*
1071 	 * DAD should be performed for an new address or addresses on
1072 	 * an interface with ND6_IFF_IFDISABLED.
1073 	 */
1074 	if (in6if_do_dad(ifp) &&
1075 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1076 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1077 
1078 	/* notify other subsystems */
1079 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1080 
1081 	return (error);
1082 }
1083 
1084 /*
1085  * Do link-level ifa job:
1086  * 1) Add lle entry for added address
1087  * 2) Notifies routing socket users about new address
1088  * 3) join appropriate multicast group
1089  * 4) start DAD if enabled
1090  */
1091 static int
1092 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1093     struct in6_ifaddr *ia, int flags)
1094 {
1095 	struct in6_multi *in6m_sol;
1096 	int error = 0;
1097 
1098 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1099 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1100 		in6_purgeaddr(&ia->ia_ifa);
1101 		ifa_free(&ia->ia_ifa);
1102 		return (error);
1103 	}
1104 
1105 	/* Join necessary multicast groups. */
1106 	in6m_sol = NULL;
1107 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1108 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1109 		if (error != 0) {
1110 			in6_purgeaddr(&ia->ia_ifa);
1111 			ifa_free(&ia->ia_ifa);
1112 			return (error);
1113 		}
1114 	}
1115 
1116 	/* Perform DAD, if the address is TENTATIVE. */
1117 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1118 		int delay, mindelay, maxdelay;
1119 
1120 		delay = 0;
1121 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1122 			/*
1123 			 * We need to impose a delay before sending an NS
1124 			 * for DAD.  Check if we also needed a delay for the
1125 			 * corresponding MLD message.  If we did, the delay
1126 			 * should be larger than the MLD delay (this could be
1127 			 * relaxed a bit, but this simple logic is at least
1128 			 * safe).
1129 			 * XXX: Break data hiding guidelines and look at
1130 			 * state for the solicited multicast group.
1131 			 */
1132 			mindelay = 0;
1133 			if (in6m_sol != NULL &&
1134 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1135 				mindelay = in6m_sol->in6m_timer;
1136 			}
1137 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1138 			if (maxdelay - mindelay == 0)
1139 				delay = 0;
1140 			else {
1141 				delay =
1142 				    (arc4random() % (maxdelay - mindelay)) +
1143 				    mindelay;
1144 			}
1145 		}
1146 		nd6_dad_start((struct ifaddr *)ia, delay);
1147 	}
1148 
1149 	in6_newaddrmsg(ia, RTM_ADD);
1150 	ifa_free(&ia->ia_ifa);
1151 	return (error);
1152 }
1153 
1154 /*
1155  * Adds or deletes interface route for p2p ifa.
1156  * Returns 0 on success or errno.
1157  */
1158 static int
1159 in6_handle_dstaddr_rtrequest(int cmd, struct in6_ifaddr *ia)
1160 {
1161 	struct epoch_tracker et;
1162 	struct ifaddr *ifa = &ia->ia_ifa;
1163 	int error;
1164 
1165 	/* Prepare gateway */
1166 	struct sockaddr_dl_short sdl = {
1167 		.sdl_family = AF_LINK,
1168 		.sdl_len = sizeof(struct sockaddr_dl_short),
1169 		.sdl_type = ifa->ifa_ifp->if_type,
1170 		.sdl_index = ifa->ifa_ifp->if_index,
1171 	};
1172 
1173 	struct sockaddr_in6 dst = {
1174 		.sin6_family = AF_INET6,
1175 		.sin6_len = sizeof(struct sockaddr_in6),
1176 		.sin6_addr = ia->ia_dstaddr.sin6_addr,
1177 	};
1178 
1179 	struct rt_addrinfo info = {
1180 		.rti_ifa = ifa,
1181 		.rti_ifp = ifa->ifa_ifp,
1182 		.rti_flags = RTF_PINNED | RTF_HOST,
1183 		.rti_info = {
1184 			[RTAX_DST] = (struct sockaddr *)&dst,
1185 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
1186 		},
1187 	};
1188 	/* Don't set additional per-gw filters on removal */
1189 
1190 	NET_EPOCH_ENTER(et);
1191 	error = rib_handle_ifaddr_info(ifa->ifa_ifp->if_fib, cmd, &info);
1192 	NET_EPOCH_EXIT(et);
1193 
1194 	return (error);
1195 }
1196 
1197 static bool
1198 ifa_is_p2p(struct in6_ifaddr *ia)
1199 {
1200 	int plen;
1201 
1202 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1203 
1204 	if ((plen == 128) && (ia->ia_dstaddr.sin6_family == AF_INET6) &&
1205 	    !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &ia->ia_dstaddr.sin6_addr))
1206 		return (true);
1207 
1208 	return (false);
1209 }
1210 
1211 int
1212 in6_addifaddr(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia)
1213 {
1214 	struct nd_prefixctl pr0;
1215 	struct nd_prefix *pr;
1216 	int carp_attached = 0;
1217 	int error;
1218 
1219 	/*
1220 	 * first, make or update the interface address structure,
1221 	 * and link it to the list.
1222 	 */
1223 	if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
1224 		goto out;
1225 	if (ia != NULL) {
1226 		if (ia->ia_ifa.ifa_carp)
1227 			(*carp_detach_p)(&ia->ia_ifa, true);
1228 		ifa_free(&ia->ia_ifa);
1229 	}
1230 	if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr)) == NULL) {
1231 		/*
1232 		 * this can happen when the user specify the 0 valid
1233 		 * lifetime.
1234 		 */
1235 		return (0);
1236 	}
1237 
1238 	if (ifra->ifra_vhid > 0) {
1239 		if (carp_attach_p != NULL)
1240 			error = (*carp_attach_p)(&ia->ia_ifa,
1241 			    ifra->ifra_vhid);
1242 		else
1243 			error = EPROTONOSUPPORT;
1244 		if (error)
1245 			goto out;
1246 		else
1247 			carp_attached = 1;
1248 	}
1249 
1250 	/*
1251 	 * then, make the prefix on-link on the interface.
1252 	 * XXX: we'd rather create the prefix before the address, but
1253 	 * we need at least one address to install the corresponding
1254 	 * interface route, so we configure the address first.
1255 	 */
1256 
1257 	/*
1258 	 * convert mask to prefix length (prefixmask has already
1259 	 * been validated in in6_update_ifa().
1260 	 */
1261 	bzero(&pr0, sizeof(pr0));
1262 	pr0.ndpr_ifp = ifp;
1263 	pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1264 	    NULL);
1265 	if (pr0.ndpr_plen == 128) {
1266 		/* we don't need to install a host route. */
1267 		goto aifaddr_out;
1268 	}
1269 	pr0.ndpr_prefix = ifra->ifra_addr;
1270 	/* apply the mask for safety. */
1271 	IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
1272 	    &ifra->ifra_prefixmask.sin6_addr);
1273 
1274 	/*
1275 	 * XXX: since we don't have an API to set prefix (not address)
1276 	 * lifetimes, we just use the same lifetimes as addresses.
1277 	 * The (temporarily) installed lifetimes can be overridden by
1278 	 * later advertised RAs (when accept_rtadv is non 0), which is
1279 	 * an intended behavior.
1280 	 */
1281 	pr0.ndpr_raf_onlink = 1; /* should be configurable? */
1282 	pr0.ndpr_raf_auto =
1283 	    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
1284 	pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
1285 	pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
1286 
1287 	/* add the prefix if not yet. */
1288 	if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
1289 		/*
1290 		 * nd6_prelist_add will install the corresponding
1291 		 * interface route.
1292 		 */
1293 		if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
1294 			if (carp_attached)
1295 				(*carp_detach_p)(&ia->ia_ifa, false);
1296 			goto out;
1297 		}
1298 	}
1299 
1300 	/* relate the address to the prefix */
1301 	if (ia->ia6_ndpr == NULL) {
1302 		ia->ia6_ndpr = pr;
1303 		pr->ndpr_addrcnt++;
1304 
1305 		/*
1306 		 * If this is the first autoconf address from the
1307 		 * prefix, create a temporary address as well
1308 		 * (when required).
1309 		 */
1310 		if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
1311 		    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
1312 			int e;
1313 			if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
1314 				log(LOG_NOTICE, "in6_control: failed "
1315 				    "to create a temporary address, "
1316 				    "errno=%d\n", e);
1317 			}
1318 		}
1319 	}
1320 	nd6_prefix_rele(pr);
1321 
1322 	/*
1323 	 * this might affect the status of autoconfigured addresses,
1324 	 * that is, this address might make other addresses detached.
1325 	 */
1326 	pfxlist_onlink_check();
1327 
1328 aifaddr_out:
1329 	/*
1330 	 * Try to clear the flag when a new IPv6 address is added
1331 	 * onto an IFDISABLED interface and it succeeds.
1332 	 */
1333 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1334 		struct in6_ndireq nd;
1335 
1336 		memset(&nd, 0, sizeof(nd));
1337 		nd.ndi.flags = ND_IFINFO(ifp)->flags;
1338 		nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
1339 		if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
1340 			log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
1341 			    "SIOCSIFINFO_FLAGS for -ifdisabled "
1342 			    "failed.");
1343 		/*
1344 		 * Ignore failure of clearing the flag intentionally.
1345 		 * The failure means address duplication was detected.
1346 		 */
1347 	}
1348 	error = 0;
1349 
1350 out:
1351 	if (ia != NULL)
1352 		ifa_free(&ia->ia_ifa);
1353 	return (error);
1354 }
1355 
1356 void
1357 in6_purgeaddr(struct ifaddr *ifa)
1358 {
1359 	struct ifnet *ifp = ifa->ifa_ifp;
1360 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1361 	struct in6_multi_mship *imm;
1362 	int error;
1363 
1364 	if (ifa->ifa_carp)
1365 		(*carp_detach_p)(ifa, false);
1366 
1367 	/*
1368 	 * Remove the loopback route to the interface address.
1369 	 * The check for the current setting of "nd6_useloopback"
1370 	 * is not needed.
1371 	 */
1372 	if (ia->ia_flags & IFA_RTSELF) {
1373 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1374 		    (struct sockaddr *)&ia->ia_addr);
1375 		if (error == 0)
1376 			ia->ia_flags &= ~IFA_RTSELF;
1377 	}
1378 
1379 	/* stop DAD processing */
1380 	nd6_dad_stop(ifa);
1381 
1382 	/* Leave multicast groups. */
1383 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1384 		LIST_REMOVE(imm, i6mm_chain);
1385 		if (imm->i6mm_maddr != NULL)
1386 			in6_leavegroup(imm->i6mm_maddr, NULL);
1387 		free(imm, M_IP6MADDR);
1388 	}
1389 	/* Check if we need to remove p2p route */
1390 	if ((ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1391 		error = in6_handle_dstaddr_rtrequest(RTM_DELETE, ia);
1392 		if (error != 0)
1393 			log(LOG_INFO, "%s: err=%d, destination address delete "
1394 			    "failed\n", __func__, error);
1395 		ia->ia_flags &= ~IFA_ROUTE;
1396 	}
1397 
1398 	in6_newaddrmsg(ia, RTM_DELETE);
1399 	in6_unlink_ifa(ia, ifp);
1400 }
1401 
1402 /*
1403  * Removes @ia from the corresponding interfaces and unlinks corresponding
1404  *  prefix if no addresses are using it anymore.
1405  */
1406 void
1407 in6_purgeifaddr(struct in6_ifaddr *ia)
1408 {
1409 	struct nd_prefix *pr;
1410 
1411 	/*
1412 	 * If the address being deleted is the only one that owns
1413 	 * the corresponding prefix, expire the prefix as well.
1414 	 * XXX: theoretically, we don't have to worry about such
1415 	 * relationship, since we separate the address management
1416 	 * and the prefix management.  We do this, however, to provide
1417 	 * as much backward compatibility as possible in terms of
1418 	 * the ioctl operation.
1419 	 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
1420 	 */
1421 	pr = ia->ia6_ndpr;
1422 	in6_purgeaddr(&ia->ia_ifa);
1423 	if (pr != NULL && pr->ndpr_addrcnt == 0) {
1424 		ND6_WLOCK();
1425 		nd6_prefix_unlink(pr, NULL);
1426 		ND6_WUNLOCK();
1427 		nd6_prefix_del(pr);
1428 	}
1429 }
1430 
1431 
1432 static void
1433 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1434 {
1435 	char ip6buf[INET6_ADDRSTRLEN];
1436 	int remove_lle;
1437 
1438 	IF_ADDR_WLOCK(ifp);
1439 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1440 	IF_ADDR_WUNLOCK(ifp);
1441 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1442 
1443 	/*
1444 	 * Defer the release of what might be the last reference to the
1445 	 * in6_ifaddr so that it can't be freed before the remainder of the
1446 	 * cleanup.
1447 	 */
1448 	IN6_IFADDR_WLOCK();
1449 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1450 	CK_LIST_REMOVE(ia, ia6_hash);
1451 	IN6_IFADDR_WUNLOCK();
1452 
1453 	/*
1454 	 * Release the reference to the base prefix.  There should be a
1455 	 * positive reference.
1456 	 */
1457 	remove_lle = 0;
1458 	if (ia->ia6_ndpr == NULL) {
1459 		nd6log((LOG_NOTICE,
1460 		    "in6_unlink_ifa: autoconf'ed address "
1461 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1462 	} else {
1463 		ia->ia6_ndpr->ndpr_addrcnt--;
1464 		/* Do not delete lles within prefix if refcont != 0 */
1465 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1466 			remove_lle = 1;
1467 		ia->ia6_ndpr = NULL;
1468 	}
1469 
1470 	nd6_rem_ifa_lle(ia, remove_lle);
1471 
1472 	/*
1473 	 * Also, if the address being removed is autoconf'ed, call
1474 	 * pfxlist_onlink_check() since the release might affect the status of
1475 	 * other (detached) addresses.
1476 	 */
1477 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1478 		pfxlist_onlink_check();
1479 	}
1480 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1481 }
1482 
1483 /*
1484  * Notifies other subsystems about address change/arrival:
1485  * 1) Notifies device handler on the first IPv6 address assignment
1486  * 2) Handle routing table changes for P2P links and route
1487  * 3) Handle routing table changes for address host route
1488  */
1489 static int
1490 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1491     struct in6_aliasreq *ifra, int hostIsNew)
1492 {
1493 	int	error = 0, ifacount = 0;
1494 	struct ifaddr *ifa;
1495 	struct sockaddr_in6 *pdst;
1496 	char ip6buf[INET6_ADDRSTRLEN];
1497 
1498 	/*
1499 	 * Give the interface a chance to initialize
1500 	 * if this is its first address,
1501 	 */
1502 	if (hostIsNew != 0) {
1503 		struct epoch_tracker et;
1504 
1505 		NET_EPOCH_ENTER(et);
1506 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1507 			if (ifa->ifa_addr->sa_family != AF_INET6)
1508 				continue;
1509 			ifacount++;
1510 		}
1511 		NET_EPOCH_EXIT(et);
1512 	}
1513 
1514 	if (ifacount <= 1 && ifp->if_ioctl) {
1515 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1516 		if (error)
1517 			goto done;
1518 	}
1519 
1520 	/*
1521 	 * If a new destination address is specified, scrub the old one and
1522 	 * install the new destination.  Note that the interface must be
1523 	 * p2p or loopback.
1524 	 */
1525 	pdst = &ifra->ifra_dstaddr;
1526 	if (pdst->sin6_family == AF_INET6 &&
1527 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1528 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1529 		    (in6_handle_dstaddr_rtrequest(RTM_DELETE, ia) != 0)) {
1530 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1531 			    "remove a route to the old destination: %s\n",
1532 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1533 			/* proceed anyway... */
1534 		} else
1535 			ia->ia_flags &= ~IFA_ROUTE;
1536 		ia->ia_dstaddr = *pdst;
1537 	}
1538 
1539 	/*
1540 	 * If a new destination address is specified for a point-to-point
1541 	 * interface, install a route to the destination as an interface
1542 	 * direct route.
1543 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1544 	 * interface that share the same destination.
1545 	 */
1546 	if (!(ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1547 		error = in6_handle_dstaddr_rtrequest(RTM_ADD, ia);
1548 		if (error)
1549 			goto done;
1550 		ia->ia_flags |= IFA_ROUTE;
1551 	}
1552 
1553 	/*
1554 	 * add a loopback route to self if not exists
1555 	 */
1556 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1557 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1558 		    (struct sockaddr *)&ia->ia_addr);
1559 		if (error == 0)
1560 			ia->ia_flags |= IFA_RTSELF;
1561 	}
1562 done:
1563 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1564 	    "Invoking IPv6 network device address event may sleep");
1565 
1566 	ifa_ref(&ia->ia_ifa);
1567 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1568 	    IFADDR_EVENT_ADD);
1569 	ifa_free(&ia->ia_ifa);
1570 
1571 	return (error);
1572 }
1573 
1574 /*
1575  * Find an IPv6 interface link-local address specific to an interface.
1576  * ifaddr is returned referenced.
1577  */
1578 struct in6_ifaddr *
1579 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1580 {
1581 	struct ifaddr *ifa;
1582 
1583 	NET_EPOCH_ASSERT();
1584 
1585 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1586 		if (ifa->ifa_addr->sa_family != AF_INET6)
1587 			continue;
1588 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1589 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1590 			    ignoreflags) != 0)
1591 				continue;
1592 			ifa_ref(ifa);
1593 			break;
1594 		}
1595 	}
1596 
1597 	return ((struct in6_ifaddr *)ifa);
1598 }
1599 
1600 /*
1601  * find the interface address corresponding to a given IPv6 address.
1602  * ifaddr is returned referenced if @referenced flag is set.
1603  */
1604 struct in6_ifaddr *
1605 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid, bool referenced)
1606 {
1607 	struct rm_priotracker in6_ifa_tracker;
1608 	struct in6_ifaddr *ia;
1609 
1610 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1611 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1612 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1613 			if (zoneid != 0 &&
1614 			    zoneid != ia->ia_addr.sin6_scope_id)
1615 				continue;
1616 			if (referenced)
1617 				ifa_ref(&ia->ia_ifa);
1618 			break;
1619 		}
1620 	}
1621 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1622 	return (ia);
1623 }
1624 
1625 /*
1626  * find the internet address corresponding to a given interface and address.
1627  * ifaddr is returned referenced.
1628  */
1629 struct in6_ifaddr *
1630 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1631 {
1632 	struct epoch_tracker et;
1633 	struct ifaddr *ifa;
1634 
1635 	NET_EPOCH_ENTER(et);
1636 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1637 		if (ifa->ifa_addr->sa_family != AF_INET6)
1638 			continue;
1639 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1640 			ifa_ref(ifa);
1641 			break;
1642 		}
1643 	}
1644 	NET_EPOCH_EXIT(et);
1645 
1646 	return ((struct in6_ifaddr *)ifa);
1647 }
1648 
1649 /*
1650  * Find a link-local scoped address on ifp and return it if any.
1651  */
1652 struct in6_ifaddr *
1653 in6ifa_llaonifp(struct ifnet *ifp)
1654 {
1655 	struct epoch_tracker et;
1656 	struct sockaddr_in6 *sin6;
1657 	struct ifaddr *ifa;
1658 
1659 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1660 		return (NULL);
1661 	NET_EPOCH_ENTER(et);
1662 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1663 		if (ifa->ifa_addr->sa_family != AF_INET6)
1664 			continue;
1665 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1666 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1667 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1668 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1669 			break;
1670 	}
1671 	NET_EPOCH_EXIT(et);
1672 
1673 	return ((struct in6_ifaddr *)ifa);
1674 }
1675 
1676 /*
1677  * Convert IP6 address to printable (loggable) representation. Caller
1678  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1679  */
1680 static char digits[] = "0123456789abcdef";
1681 char *
1682 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1683 {
1684 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1685 	char *cp;
1686 	const u_int16_t *a = (const u_int16_t *)addr;
1687 	const u_int8_t *d;
1688 	int dcolon = 0, zero = 0;
1689 
1690 	cp = ip6buf;
1691 
1692 	for (i = 0; i < 8; i++) {
1693 		if (*(a + i) == 0) {
1694 			cnt++;
1695 			if (cnt == 1)
1696 				idx = i;
1697 		}
1698 		else if (maxcnt < cnt) {
1699 			maxcnt = cnt;
1700 			index = idx;
1701 			cnt = 0;
1702 		}
1703 	}
1704 	if (maxcnt < cnt) {
1705 		maxcnt = cnt;
1706 		index = idx;
1707 	}
1708 
1709 	for (i = 0; i < 8; i++) {
1710 		if (dcolon == 1) {
1711 			if (*a == 0) {
1712 				if (i == 7)
1713 					*cp++ = ':';
1714 				a++;
1715 				continue;
1716 			} else
1717 				dcolon = 2;
1718 		}
1719 		if (*a == 0) {
1720 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1721 				if (i == 0)
1722 					*cp++ = ':';
1723 				*cp++ = ':';
1724 				dcolon = 1;
1725 			} else {
1726 				*cp++ = '0';
1727 				*cp++ = ':';
1728 			}
1729 			a++;
1730 			continue;
1731 		}
1732 		d = (const u_char *)a;
1733 		/* Try to eliminate leading zeros in printout like in :0001. */
1734 		zero = 1;
1735 		*cp = digits[*d >> 4];
1736 		if (*cp != '0') {
1737 			zero = 0;
1738 			cp++;
1739 		}
1740 		*cp = digits[*d++ & 0xf];
1741 		if (zero == 0 || (*cp != '0')) {
1742 			zero = 0;
1743 			cp++;
1744 		}
1745 		*cp = digits[*d >> 4];
1746 		if (zero == 0 || (*cp != '0')) {
1747 			zero = 0;
1748 			cp++;
1749 		}
1750 		*cp++ = digits[*d & 0xf];
1751 		*cp++ = ':';
1752 		a++;
1753 	}
1754 	*--cp = '\0';
1755 	return (ip6buf);
1756 }
1757 
1758 int
1759 in6_localaddr(struct in6_addr *in6)
1760 {
1761 	struct rm_priotracker in6_ifa_tracker;
1762 	struct in6_ifaddr *ia;
1763 
1764 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1765 		return 1;
1766 
1767 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1768 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1769 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1770 		    &ia->ia_prefixmask.sin6_addr)) {
1771 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1772 			return 1;
1773 		}
1774 	}
1775 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1776 
1777 	return (0);
1778 }
1779 
1780 /*
1781  * Return 1 if an internet address is for the local host and configured
1782  * on one of its interfaces.
1783  */
1784 int
1785 in6_localip(struct in6_addr *in6)
1786 {
1787 	struct rm_priotracker in6_ifa_tracker;
1788 	struct in6_ifaddr *ia;
1789 
1790 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1791 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1792 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1793 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1794 			return (1);
1795 		}
1796 	}
1797 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1798 	return (0);
1799 }
1800 
1801 /*
1802  * Like in6_localip(), but FIB-aware.
1803  */
1804 bool
1805 in6_localip_fib(struct in6_addr *in6, uint16_t fib)
1806 {
1807 	struct rm_priotracker in6_ifa_tracker;
1808 	struct in6_ifaddr *ia;
1809 
1810 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1811 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1812 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr) &&
1813 		    ia->ia_ifa.ifa_ifp->if_fib == fib) {
1814 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1815 			return (true);
1816 		}
1817 	}
1818 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1819 	return (false);
1820 }
1821 
1822 /*
1823  * Return 1 if an internet address is configured on an interface.
1824  */
1825 int
1826 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1827 {
1828 	struct in6_addr in6;
1829 	struct ifaddr *ifa;
1830 	struct in6_ifaddr *ia6;
1831 
1832 	NET_EPOCH_ASSERT();
1833 
1834 	in6 = *addr;
1835 	if (in6_clearscope(&in6))
1836 		return (0);
1837 	in6_setscope(&in6, ifp, NULL);
1838 
1839 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1840 		if (ifa->ifa_addr->sa_family != AF_INET6)
1841 			continue;
1842 		ia6 = (struct in6_ifaddr *)ifa;
1843 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1844 			return (1);
1845 	}
1846 
1847 	return (0);
1848 }
1849 
1850 int
1851 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1852 {
1853 	struct rm_priotracker in6_ifa_tracker;
1854 	struct in6_ifaddr *ia;
1855 
1856 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1857 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1858 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1859 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1860 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1861 				return (1); /* true */
1862 			}
1863 			break;
1864 		}
1865 	}
1866 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1867 
1868 	return (0);		/* false */
1869 }
1870 
1871 /*
1872  * return length of part which dst and src are equal
1873  * hard coding...
1874  */
1875 int
1876 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1877 {
1878 	int match = 0;
1879 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1880 	u_char *lim = s + 16, r;
1881 
1882 	while (s < lim)
1883 		if ((r = (*d++ ^ *s++)) != 0) {
1884 			while (r < 128) {
1885 				match++;
1886 				r <<= 1;
1887 			}
1888 			break;
1889 		} else
1890 			match += 8;
1891 	return match;
1892 }
1893 
1894 /* XXX: to be scope conscious */
1895 int
1896 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1897 {
1898 	int bytelen, bitlen;
1899 
1900 	/* sanity check */
1901 	if (0 > len || len > 128) {
1902 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1903 		    len);
1904 		return (0);
1905 	}
1906 
1907 	bytelen = len / 8;
1908 	bitlen = len % 8;
1909 
1910 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1911 		return (0);
1912 	if (bitlen != 0 &&
1913 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1914 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1915 		return (0);
1916 
1917 	return (1);
1918 }
1919 
1920 void
1921 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1922 {
1923 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1924 	int bytelen, bitlen, i;
1925 
1926 	/* sanity check */
1927 	if (0 > len || len > 128) {
1928 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1929 		    len);
1930 		return;
1931 	}
1932 
1933 	bzero(maskp, sizeof(*maskp));
1934 	bytelen = len / 8;
1935 	bitlen = len % 8;
1936 	for (i = 0; i < bytelen; i++)
1937 		maskp->s6_addr[i] = 0xff;
1938 	if (bitlen)
1939 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1940 }
1941 
1942 /*
1943  * return the best address out of the same scope. if no address was
1944  * found, return the first valid address from designated IF.
1945  */
1946 struct in6_ifaddr *
1947 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1948 {
1949 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1950 	struct ifaddr *ifa;
1951 	struct in6_ifaddr *besta = NULL;
1952 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1953 
1954 	NET_EPOCH_ASSERT();
1955 
1956 	dep[0] = dep[1] = NULL;
1957 
1958 	/*
1959 	 * We first look for addresses in the same scope.
1960 	 * If there is one, return it.
1961 	 * If two or more, return one which matches the dst longest.
1962 	 * If none, return one of global addresses assigned other ifs.
1963 	 */
1964 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1965 		if (ifa->ifa_addr->sa_family != AF_INET6)
1966 			continue;
1967 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1968 			continue; /* XXX: is there any case to allow anycast? */
1969 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1970 			continue; /* don't use this interface */
1971 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1972 			continue;
1973 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1974 			if (V_ip6_use_deprecated)
1975 				dep[0] = (struct in6_ifaddr *)ifa;
1976 			continue;
1977 		}
1978 
1979 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1980 			/*
1981 			 * call in6_matchlen() as few as possible
1982 			 */
1983 			if (besta) {
1984 				if (blen == -1)
1985 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1986 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1987 				if (tlen > blen) {
1988 					blen = tlen;
1989 					besta = (struct in6_ifaddr *)ifa;
1990 				}
1991 			} else
1992 				besta = (struct in6_ifaddr *)ifa;
1993 		}
1994 	}
1995 	if (besta)
1996 		return (besta);
1997 
1998 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1999 		if (ifa->ifa_addr->sa_family != AF_INET6)
2000 			continue;
2001 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2002 			continue; /* XXX: is there any case to allow anycast? */
2003 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2004 			continue; /* don't use this interface */
2005 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2006 			continue;
2007 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2008 			if (V_ip6_use_deprecated)
2009 				dep[1] = (struct in6_ifaddr *)ifa;
2010 			continue;
2011 		}
2012 
2013 		return (struct in6_ifaddr *)ifa;
2014 	}
2015 
2016 	/* use the last-resort values, that are, deprecated addresses */
2017 	if (dep[0])
2018 		return dep[0];
2019 	if (dep[1])
2020 		return dep[1];
2021 
2022 	return NULL;
2023 }
2024 
2025 /*
2026  * perform DAD when interface becomes IFF_UP.
2027  */
2028 void
2029 in6_if_up(struct ifnet *ifp)
2030 {
2031 	struct epoch_tracker et;
2032 	struct ifaddr *ifa;
2033 	struct in6_ifaddr *ia;
2034 
2035 	NET_EPOCH_ENTER(et);
2036 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2037 		if (ifa->ifa_addr->sa_family != AF_INET6)
2038 			continue;
2039 		ia = (struct in6_ifaddr *)ifa;
2040 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2041 			/*
2042 			 * The TENTATIVE flag was likely set by hand
2043 			 * beforehand, implicitly indicating the need for DAD.
2044 			 * We may be able to skip the random delay in this
2045 			 * case, but we impose delays just in case.
2046 			 */
2047 			nd6_dad_start(ifa,
2048 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2049 		}
2050 	}
2051 	NET_EPOCH_EXIT(et);
2052 
2053 	/*
2054 	 * special cases, like 6to4, are handled in in6_ifattach
2055 	 */
2056 	in6_ifattach(ifp, NULL);
2057 }
2058 
2059 static void
2060 in6_ifevent(void *arg __unused, struct ifnet *ifp, int event)
2061 {
2062 	if (event == IFNET_EVENT_UP)
2063 		in6_if_up(ifp);
2064 }
2065 
2066 static void
2067 in6_init(void *arg __unused)
2068 {
2069 	EVENTHANDLER_REGISTER(ifnet_event, in6_ifevent, NULL, EVENTHANDLER_PRI_ANY);
2070 }
2071 SYSINIT(in6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, in6_init, NULL);
2072 
2073 int
2074 in6if_do_dad(struct ifnet *ifp)
2075 {
2076 
2077 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2078 		return (0);
2079 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
2080 		return (0);
2081 	if ((ND_IFINFO(ifp)->flags &
2082 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
2083 		return (0);
2084 	return (1);
2085 }
2086 
2087 /*
2088  * Calculate max IPv6 MTU through all the interfaces and store it
2089  * to in6_maxmtu.
2090  */
2091 void
2092 in6_setmaxmtu(void)
2093 {
2094 	struct epoch_tracker et;
2095 	unsigned long maxmtu = 0;
2096 	struct ifnet *ifp;
2097 
2098 	NET_EPOCH_ENTER(et);
2099 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2100 		/* this function can be called during ifnet initialization */
2101 		if (!ifp->if_afdata[AF_INET6])
2102 			continue;
2103 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2104 		    IN6_LINKMTU(ifp) > maxmtu)
2105 			maxmtu = IN6_LINKMTU(ifp);
2106 	}
2107 	NET_EPOCH_EXIT(et);
2108 	if (maxmtu)	/* update only when maxmtu is positive */
2109 		V_in6_maxmtu = maxmtu;
2110 }
2111 
2112 /*
2113  * Provide the length of interface identifiers to be used for the link attached
2114  * to the given interface.  The length should be defined in "IPv6 over
2115  * xxx-link" document.  Note that address architecture might also define
2116  * the length for a particular set of address prefixes, regardless of the
2117  * link type.  As clarified in rfc2462bis, those two definitions should be
2118  * consistent, and those really are as of August 2004.
2119  */
2120 int
2121 in6_if2idlen(struct ifnet *ifp)
2122 {
2123 	switch (ifp->if_type) {
2124 	case IFT_ETHER:		/* RFC2464 */
2125 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2126 	case IFT_L2VLAN:	/* ditto */
2127 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2128 	case IFT_INFINIBAND:
2129 		return (64);
2130 	case IFT_PPP:		/* RFC2472 */
2131 		return (64);
2132 	case IFT_FRELAY:	/* RFC2590 */
2133 		return (64);
2134 	case IFT_IEEE1394:	/* RFC3146 */
2135 		return (64);
2136 	case IFT_GIF:
2137 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2138 	case IFT_LOOP:
2139 		return (64);	/* XXX: is this really correct? */
2140 	default:
2141 		/*
2142 		 * Unknown link type:
2143 		 * It might be controversial to use the today's common constant
2144 		 * of 64 for these cases unconditionally.  For full compliance,
2145 		 * we should return an error in this case.  On the other hand,
2146 		 * if we simply miss the standard for the link type or a new
2147 		 * standard is defined for a new link type, the IFID length
2148 		 * is very likely to be the common constant.  As a compromise,
2149 		 * we always use the constant, but make an explicit notice
2150 		 * indicating the "unknown" case.
2151 		 */
2152 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2153 		return (64);
2154 	}
2155 }
2156 
2157 struct in6_llentry {
2158 	struct llentry		base;
2159 };
2160 
2161 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2162 #define	IN6_LLTBL_HASH(k, h) \
2163 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2164 
2165 /*
2166  * Do actual deallocation of @lle.
2167  */
2168 static void
2169 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2170 {
2171 	struct llentry *lle;
2172 
2173 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2174 	LLE_LOCK_DESTROY(lle);
2175 	LLE_REQ_DESTROY(lle);
2176 	free(lle, M_LLTABLE);
2177 }
2178 
2179 /*
2180  * Called by LLE_FREE_LOCKED when number of references
2181  * drops to zero.
2182  */
2183 static void
2184 in6_lltable_destroy_lle(struct llentry *lle)
2185 {
2186 
2187 	LLE_WUNLOCK(lle);
2188 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2189 }
2190 
2191 static struct llentry *
2192 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2193 {
2194 	struct in6_llentry *lle;
2195 
2196 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2197 	if (lle == NULL)		/* NB: caller generates msg */
2198 		return NULL;
2199 
2200 	lle->base.r_l3addr.addr6 = *addr6;
2201 	lle->base.lle_refcnt = 1;
2202 	lle->base.lle_free = in6_lltable_destroy_lle;
2203 	LLE_LOCK_INIT(&lle->base);
2204 	LLE_REQ_INIT(&lle->base);
2205 	callout_init(&lle->base.lle_timer, 1);
2206 
2207 	return (&lle->base);
2208 }
2209 
2210 static int
2211 in6_lltable_match_prefix(const struct sockaddr *saddr,
2212     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2213 {
2214 	const struct in6_addr *addr, *mask, *lle_addr;
2215 
2216 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2217 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2218 	lle_addr = &lle->r_l3addr.addr6;
2219 
2220 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2221 		return (0);
2222 
2223 	if (lle->la_flags & LLE_IFADDR) {
2224 		/*
2225 		 * Delete LLE_IFADDR records IFF address & flag matches.
2226 		 * Note that addr is the interface address within prefix
2227 		 * being matched.
2228 		 */
2229 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2230 		    (flags & LLE_STATIC) != 0)
2231 			return (1);
2232 		return (0);
2233 	}
2234 
2235 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2236 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2237 		return (1);
2238 
2239 	return (0);
2240 }
2241 
2242 static void
2243 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2244 {
2245 	struct ifnet *ifp __diagused;
2246 
2247 	LLE_WLOCK_ASSERT(lle);
2248 	KASSERT(llt != NULL, ("lltable is NULL"));
2249 
2250 	/* Unlink entry from table */
2251 	if ((lle->la_flags & LLE_LINKED) != 0) {
2252 		ifp = llt->llt_ifp;
2253 		IF_AFDATA_WLOCK_ASSERT(ifp);
2254 		lltable_unlink_entry(llt, lle);
2255 	}
2256 
2257 	llentry_free(lle);
2258 }
2259 
2260 static int
2261 in6_lltable_rtcheck(struct ifnet *ifp,
2262 		    u_int flags,
2263 		    const struct sockaddr *l3addr)
2264 {
2265 	const struct sockaddr_in6 *sin6;
2266 	struct nhop_object *nh;
2267 	struct in6_addr dst;
2268 	uint32_t scopeid;
2269 	char ip6buf[INET6_ADDRSTRLEN];
2270 	int fibnum;
2271 
2272 	NET_EPOCH_ASSERT();
2273 	KASSERT(l3addr->sa_family == AF_INET6,
2274 	    ("sin_family %d", l3addr->sa_family));
2275 
2276 	sin6 = (const struct sockaddr_in6 *)l3addr;
2277 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2278 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2279 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2280 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2281 		struct ifaddr *ifa;
2282 		/*
2283 		 * Create an ND6 cache for an IPv6 neighbor
2284 		 * that is not covered by our own prefix.
2285 		 */
2286 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2287 		if (ifa != NULL) {
2288 			return 0;
2289 		}
2290 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2291 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2292 		return EINVAL;
2293 	}
2294 	return 0;
2295 }
2296 
2297 static inline uint32_t
2298 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2299 {
2300 
2301 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2302 }
2303 
2304 static uint32_t
2305 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2306 {
2307 
2308 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2309 }
2310 
2311 static void
2312 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2313 {
2314 	struct sockaddr_in6 *sin6;
2315 
2316 	sin6 = (struct sockaddr_in6 *)sa;
2317 	bzero(sin6, sizeof(*sin6));
2318 	sin6->sin6_family = AF_INET6;
2319 	sin6->sin6_len = sizeof(*sin6);
2320 	sin6->sin6_addr = lle->r_l3addr.addr6;
2321 }
2322 
2323 static inline struct llentry *
2324 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2325 {
2326 	struct llentry *lle;
2327 	struct llentries *lleh;
2328 	u_int hashidx;
2329 
2330 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2331 	lleh = &llt->lle_head[hashidx];
2332 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2333 		if (lle->la_flags & LLE_DELETED)
2334 			continue;
2335 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2336 			break;
2337 	}
2338 
2339 	return (lle);
2340 }
2341 
2342 static void
2343 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2344 {
2345 
2346 	lle->la_flags |= LLE_DELETED;
2347 
2348 	/* Leave the solicited multicast group. */
2349 	if ((lle->la_flags & LLE_PUB) != 0)
2350 		in6_leave_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2351 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2352 #ifdef DIAGNOSTIC
2353 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2354 #endif
2355 	llentry_free(lle);
2356 }
2357 
2358 static struct llentry *
2359 in6_lltable_alloc(struct lltable *llt, u_int flags,
2360 	const struct sockaddr *l3addr)
2361 {
2362 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2363 	struct ifnet *ifp = llt->llt_ifp;
2364 	struct llentry *lle;
2365 	char linkhdr[LLE_MAX_LINKHDR];
2366 	size_t linkhdrsize;
2367 	int lladdr_off;
2368 
2369 	KASSERT(l3addr->sa_family == AF_INET6,
2370 	    ("sin_family %d", l3addr->sa_family));
2371 
2372 	/*
2373 	 * A route that covers the given address must have
2374 	 * been installed 1st because we are doing a resolution,
2375 	 * verify this.
2376 	 */
2377 	if (!(flags & LLE_IFADDR) &&
2378 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2379 		return (NULL);
2380 
2381 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2382 	if (lle == NULL) {
2383 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2384 		return (NULL);
2385 	}
2386 	lle->la_flags = flags;
2387 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2388 		linkhdrsize = LLE_MAX_LINKHDR;
2389 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2390 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2391 			in6_lltable_free_entry(llt, lle);
2392 			return (NULL);
2393 		}
2394 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2395 		    lladdr_off);
2396 		lle->la_flags |= LLE_STATIC;
2397 	}
2398 
2399 	if ((lle->la_flags & LLE_STATIC) != 0)
2400 		lle->ln_state = ND6_LLINFO_REACHABLE;
2401 
2402 	return (lle);
2403 }
2404 
2405 static struct llentry *
2406 in6_lltable_lookup(struct lltable *llt, u_int flags,
2407 	const struct sockaddr *l3addr)
2408 {
2409 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2410 	int family = flags >> 16;
2411 	struct llentry *lle;
2412 
2413 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2414 	KASSERT(l3addr->sa_family == AF_INET6,
2415 	    ("sin_family %d", l3addr->sa_family));
2416 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2417 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2418 	    ("wrong lle request flags: %#x", flags));
2419 
2420 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2421 
2422 	if (__predict_false(family != AF_INET6))
2423 		lle = llentry_lookup_family(lle, family);
2424 
2425 	if (lle == NULL)
2426 		return (NULL);
2427 
2428 	if (flags & LLE_UNLOCKED)
2429 		return (lle);
2430 
2431 	if (flags & LLE_EXCLUSIVE)
2432 		LLE_WLOCK(lle);
2433 	else
2434 		LLE_RLOCK(lle);
2435 
2436 	/*
2437 	 * If the afdata lock is not held, the LLE may have been unlinked while
2438 	 * we were blocked on the LLE lock.  Check for this case.
2439 	 */
2440 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2441 		if (flags & LLE_EXCLUSIVE)
2442 			LLE_WUNLOCK(lle);
2443 		else
2444 			LLE_RUNLOCK(lle);
2445 		return (NULL);
2446 	}
2447 	return (lle);
2448 }
2449 
2450 static int
2451 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2452     struct sysctl_req *wr)
2453 {
2454 	struct ifnet *ifp = llt->llt_ifp;
2455 	/* XXX stack use */
2456 	struct {
2457 		struct rt_msghdr	rtm;
2458 		struct sockaddr_in6	sin6;
2459 		/*
2460 		 * ndp.c assumes that sdl is word aligned
2461 		 */
2462 #ifdef __LP64__
2463 		uint32_t		pad;
2464 #endif
2465 		struct sockaddr_dl	sdl;
2466 	} ndpc;
2467 	struct sockaddr_dl *sdl;
2468 	int error;
2469 
2470 	bzero(&ndpc, sizeof(ndpc));
2471 	/* skip deleted entries */
2472 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2473 		return (0);
2474 	/* Skip if jailed and not a valid IP of the prison. */
2475 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2476 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2477 		return (0);
2478 	/*
2479 	 * produce a msg made of:
2480 	 *  struct rt_msghdr;
2481 	 *  struct sockaddr_in6 (IPv6)
2482 	 *  struct sockaddr_dl;
2483 	 */
2484 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2485 	ndpc.rtm.rtm_version = RTM_VERSION;
2486 	ndpc.rtm.rtm_type = RTM_GET;
2487 	ndpc.rtm.rtm_flags = RTF_UP;
2488 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2489 	sa6_recoverscope(&ndpc.sin6);
2490 
2491 	/* publish */
2492 	if (lle->la_flags & LLE_PUB)
2493 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2494 
2495 	sdl = &ndpc.sdl;
2496 	sdl->sdl_family = AF_LINK;
2497 	sdl->sdl_len = sizeof(*sdl);
2498 	sdl->sdl_index = ifp->if_index;
2499 	sdl->sdl_type = ifp->if_type;
2500 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2501 		sdl->sdl_alen = ifp->if_addrlen;
2502 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2503 	} else {
2504 		sdl->sdl_alen = 0;
2505 		bzero(LLADDR(sdl), ifp->if_addrlen);
2506 	}
2507 	if (lle->la_expire != 0)
2508 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2509 		    lle->lle_remtime / hz + time_second - time_uptime;
2510 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2511 	if (lle->la_flags & LLE_STATIC)
2512 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2513 	if (lle->la_flags & LLE_IFADDR)
2514 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2515 	if (lle->ln_router != 0)
2516 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2517 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2518 	/* Store state in rmx_weight value */
2519 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2520 	ndpc.rtm.rtm_index = ifp->if_index;
2521 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2522 
2523 	return (error);
2524 }
2525 
2526 static void
2527 in6_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
2528 {
2529 	/* Join the solicited multicast group for dst. */
2530 	if ((lle->la_flags & LLE_PUB) == LLE_PUB)
2531 		in6_join_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2532 }
2533 
2534 static struct lltable *
2535 in6_lltattach(struct ifnet *ifp)
2536 {
2537 	struct lltable *llt;
2538 
2539 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2540 	llt->llt_af = AF_INET6;
2541 	llt->llt_ifp = ifp;
2542 
2543 	llt->llt_lookup = in6_lltable_lookup;
2544 	llt->llt_alloc_entry = in6_lltable_alloc;
2545 	llt->llt_delete_entry = in6_lltable_delete_entry;
2546 	llt->llt_dump_entry = in6_lltable_dump_entry;
2547 	llt->llt_hash = in6_lltable_hash;
2548 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2549 	llt->llt_free_entry = in6_lltable_free_entry;
2550 	llt->llt_match_prefix = in6_lltable_match_prefix;
2551 	llt->llt_mark_used = llentry_mark_used;
2552 	llt->llt_post_resolved = in6_lltable_post_resolved;
2553  	lltable_link(llt);
2554 
2555 	return (llt);
2556 }
2557 
2558 struct lltable *
2559 in6_lltable_get(struct ifnet *ifp)
2560 {
2561 	struct lltable *llt = NULL;
2562 
2563 	void *afdata_ptr = ifp->if_afdata[AF_INET6];
2564 	if (afdata_ptr != NULL)
2565 		llt = ((struct in6_ifextra *)afdata_ptr)->lltable;
2566 	return (llt);
2567 }
2568 
2569 void *
2570 in6_domifattach(struct ifnet *ifp)
2571 {
2572 	struct in6_ifextra *ext;
2573 
2574 	/* There are not IPv6-capable interfaces. */
2575 	switch (ifp->if_type) {
2576 	case IFT_PFLOG:
2577 	case IFT_PFSYNC:
2578 	case IFT_USB:
2579 		return (NULL);
2580 	}
2581 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2582 	bzero(ext, sizeof(*ext));
2583 
2584 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2585 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2586 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2587 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2588 
2589 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2590 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2591 	    M_WAITOK);
2592 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2593 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2594 
2595 	ext->nd_ifinfo = nd6_ifattach(ifp);
2596 	ext->scope6_id = scope6_ifattach(ifp);
2597 	ext->lltable = in6_lltattach(ifp);
2598 
2599 	ext->mld_ifinfo = mld_domifattach(ifp);
2600 
2601 	return ext;
2602 }
2603 
2604 int
2605 in6_domifmtu(struct ifnet *ifp)
2606 {
2607 	if (ifp->if_afdata[AF_INET6] == NULL)
2608 		return ifp->if_mtu;
2609 
2610 	return (IN6_LINKMTU(ifp));
2611 }
2612 
2613 void
2614 in6_domifdetach(struct ifnet *ifp, void *aux)
2615 {
2616 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2617 
2618 	mld_domifdetach(ifp);
2619 	scope6_ifdetach(ext->scope6_id);
2620 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2621 	lltable_free(ext->lltable);
2622 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2623 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2624 	free(ext->in6_ifstat, M_IFADDR);
2625 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2626 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2627 	free(ext->icmp6_ifstat, M_IFADDR);
2628 	free(ext, M_IFADDR);
2629 }
2630 
2631 /*
2632  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2633  * v4 mapped addr or v4 compat addr
2634  */
2635 void
2636 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2637 {
2638 
2639 	bzero(sin, sizeof(*sin));
2640 	sin->sin_len = sizeof(struct sockaddr_in);
2641 	sin->sin_family = AF_INET;
2642 	sin->sin_port = sin6->sin6_port;
2643 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2644 }
2645 
2646 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2647 void
2648 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2649 {
2650 	bzero(sin6, sizeof(*sin6));
2651 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2652 	sin6->sin6_family = AF_INET6;
2653 	sin6->sin6_port = sin->sin_port;
2654 	sin6->sin6_addr.s6_addr32[0] = 0;
2655 	sin6->sin6_addr.s6_addr32[1] = 0;
2656 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2657 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2658 }
2659 
2660 /* Convert sockaddr_in6 into sockaddr_in. */
2661 void
2662 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2663 {
2664 	struct sockaddr_in *sin_p;
2665 	struct sockaddr_in6 sin6;
2666 
2667 	/*
2668 	 * Save original sockaddr_in6 addr and convert it
2669 	 * to sockaddr_in.
2670 	 */
2671 	sin6 = *(struct sockaddr_in6 *)nam;
2672 	sin_p = (struct sockaddr_in *)nam;
2673 	in6_sin6_2_sin(sin_p, &sin6);
2674 }
2675 
2676 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2677 void
2678 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2679 {
2680 	struct sockaddr_in *sin_p;
2681 	struct sockaddr_in6 *sin6_p;
2682 
2683 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2684 	sin_p = (struct sockaddr_in *)*nam;
2685 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2686 	free(*nam, M_SONAME);
2687 	*nam = (struct sockaddr *)sin6_p;
2688 }
2689 
2690 /*
2691  * Join/leave the solicited multicast groups for proxy NDP entries.
2692  */
2693 static void
2694 in6_join_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2695 {
2696 	struct in6_multi *inm;
2697 	struct in6_addr mltaddr;
2698 	char ip6buf[INET6_ADDRSTRLEN];
2699 	int error;
2700 
2701 	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2702 		return;	/* error logged in in6_solicited_node_maddr. */
2703 
2704 	error = in6_joingroup(ifp, &mltaddr, NULL, &inm, 0);
2705 	if (error != 0) {
2706 		nd6log((LOG_WARNING,
2707 		    "%s: in6_joingroup failed for %s on %s (errno=%d)\n",
2708 		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp),
2709 		    error));
2710 	}
2711 }
2712 
2713 static void
2714 in6_leave_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2715 {
2716 	struct epoch_tracker et;
2717 	struct in6_multi *inm;
2718 	struct in6_addr mltaddr;
2719 	char ip6buf[INET6_ADDRSTRLEN];
2720 
2721 	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2722 		return;	/* error logged in in6_solicited_node_maddr. */
2723 
2724 	NET_EPOCH_ENTER(et);
2725 	inm = in6m_lookup(ifp, &mltaddr);
2726 	NET_EPOCH_EXIT(et);
2727 	if (inm != NULL)
2728 		in6_leavegroup(inm, NULL);
2729 	else
2730 		nd6log((LOG_WARNING, "%s: in6m_lookup failed for %s on %s\n",
2731 		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp)));
2732 }
2733 
2734 static bool
2735 in6_lle_match_pub(struct lltable *llt, struct llentry *lle, void *farg)
2736 {
2737 	return ((lle->la_flags & LLE_PUB) != 0);
2738 }
2739 
2740 void
2741 in6_purge_proxy_ndp(struct ifnet *ifp)
2742 {
2743 	struct lltable *llt;
2744 	bool need_purge;
2745 
2746 	if (ifp->if_afdata[AF_INET6] == NULL)
2747 		return;
2748 
2749 	llt = LLTABLE6(ifp);
2750 	IF_AFDATA_WLOCK(ifp);
2751 	need_purge = ((llt->llt_flags & LLT_ADDEDPROXY) != 0);
2752 	IF_AFDATA_WUNLOCK(ifp);
2753 
2754 	/*
2755 	 * Ever added proxy ndp entries, leave solicited node multicast
2756 	 * before deleting the llentry.
2757 	 */
2758 	if (need_purge)
2759 		lltable_delete_conditional(llt, in6_lle_match_pub, NULL);
2760 }
2761