in6.c revision f5d8b3c62e7892fddb7c372282c702cbe7198b30
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/protosw.h>
83#include <sys/time.h>
84#include <sys/kernel.h>
85#include <sys/lock.h>
86#include <sys/rmlock.h>
87#include <sys/syslog.h>
88
89#include <net/if.h>
90#include <net/if_var.h>
91#include <net/if_types.h>
92#include <net/route.h>
93#include <net/if_dl.h>
94#include <net/vnet.h>
95
96#include <netinet/in.h>
97#include <netinet/in_var.h>
98#include <net/if_llatbl.h>
99#include <netinet/if_ether.h>
100#include <netinet/in_systm.h>
101#include <netinet/ip.h>
102#include <netinet/in_pcb.h>
103#include <netinet/ip_carp.h>
104
105#include <netinet/ip6.h>
106#include <netinet6/ip6_var.h>
107#include <netinet6/nd6.h>
108#include <netinet6/mld6_var.h>
109#include <netinet6/ip6_mroute.h>
110#include <netinet6/in6_ifattach.h>
111#include <netinet6/scope6_var.h>
112#include <netinet6/in6_fib.h>
113#include <netinet6/in6_pcb.h>
114
115/*
116 * struct in6_ifreq and struct ifreq must be type punnable for common members
117 * of ifr_ifru to allow accessors to be shared.
118 */
119_Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
120    offsetof(struct ifreq, ifr_ifru),
121    "struct in6_ifreq and struct ifreq are not type punnable");
122
123VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
124#define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
125
126/*
127 * Definitions of some costant IP6 addresses.
128 */
129const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
130const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
131const struct in6_addr in6addr_nodelocal_allnodes =
132	IN6ADDR_NODELOCAL_ALLNODES_INIT;
133const struct in6_addr in6addr_linklocal_allnodes =
134	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
135const struct in6_addr in6addr_linklocal_allrouters =
136	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
137const struct in6_addr in6addr_linklocal_allv2routers =
138	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
139
140const struct in6_addr in6mask0 = IN6MASK0;
141const struct in6_addr in6mask32 = IN6MASK32;
142const struct in6_addr in6mask64 = IN6MASK64;
143const struct in6_addr in6mask96 = IN6MASK96;
144const struct in6_addr in6mask128 = IN6MASK128;
145
146const struct sockaddr_in6 sa6_any =
147	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
148
149static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
150	struct in6_aliasreq *, int);
151static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
152
153static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
154    struct in6_ifaddr *, int);
155static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
156    struct in6_aliasreq *, int flags);
157static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
158    struct in6_ifaddr *, int, int);
159static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
160    struct in6_ifaddr *, int);
161
162#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
163#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
164
165
166void
167in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
168{
169	struct sockaddr_dl gateway;
170	struct sockaddr_in6 mask, addr;
171	struct rtentry rt;
172	int fibnum;
173
174	/*
175	 * initialize for rtmsg generation
176	 */
177	bzero(&gateway, sizeof(gateway));
178	gateway.sdl_len = sizeof(gateway);
179	gateway.sdl_family = AF_LINK;
180
181	bzero(&rt, sizeof(rt));
182	rt.rt_gateway = (struct sockaddr *)&gateway;
183	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
184	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
185	rt_mask(&rt) = (struct sockaddr *)&mask;
186	rt_key(&rt) = (struct sockaddr *)&addr;
187	rt.rt_flags = RTF_HOST | RTF_STATIC;
188	if (cmd == RTM_ADD)
189		rt.rt_flags |= RTF_UP;
190	fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ia62ifa(ia)->ifa_ifp->if_fib;
191	/* Announce arrival of local address to this FIB. */
192	rt_newaddrmsg_fib(cmd, &ia->ia_ifa, 0, &rt, fibnum);
193}
194
195int
196in6_mask2len(struct in6_addr *mask, u_char *lim0)
197{
198	int x = 0, y;
199	u_char *lim = lim0, *p;
200
201	/* ignore the scope_id part */
202	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
203		lim = (u_char *)mask + sizeof(*mask);
204	for (p = (u_char *)mask; p < lim; x++, p++) {
205		if (*p != 0xff)
206			break;
207	}
208	y = 0;
209	if (p < lim) {
210		for (y = 0; y < 8; y++) {
211			if ((*p & (0x80 >> y)) == 0)
212				break;
213		}
214	}
215
216	/*
217	 * when the limit pointer is given, do a stricter check on the
218	 * remaining bits.
219	 */
220	if (p < lim) {
221		if (y != 0 && (*p & (0x00ff >> y)) != 0)
222			return (-1);
223		for (p = p + 1; p < lim; p++)
224			if (*p != 0)
225				return (-1);
226	}
227
228	return x * 8 + y;
229}
230
231#ifdef COMPAT_FREEBSD32
232struct in6_ndifreq32 {
233	char ifname[IFNAMSIZ];
234	uint32_t ifindex;
235};
236#define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
237#endif
238
239int
240in6_control(struct socket *so, u_long cmd, caddr_t data,
241    struct ifnet *ifp, struct thread *td)
242{
243	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
244	struct	in6_ifaddr *ia = NULL;
245	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
246	struct sockaddr_in6 *sa6;
247	int carp_attached = 0;
248	int error;
249	u_long ocmd = cmd;
250
251	/*
252	 * Compat to make pre-10.x ifconfig(8) operable.
253	 */
254	if (cmd == OSIOCAIFADDR_IN6)
255		cmd = SIOCAIFADDR_IN6;
256
257	switch (cmd) {
258	case SIOCGETSGCNT_IN6:
259	case SIOCGETMIFCNT_IN6:
260		/*
261		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
262		 * We cannot see how that would be needed, so do not adjust the
263		 * KPI blindly; more likely should clean up the IPv4 variant.
264		 */
265		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
266	}
267
268	switch (cmd) {
269	case SIOCAADDRCTL_POLICY:
270	case SIOCDADDRCTL_POLICY:
271		if (td != NULL) {
272			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
273			if (error)
274				return (error);
275		}
276		return (in6_src_ioctl(cmd, data));
277	}
278
279	if (ifp == NULL)
280		return (EOPNOTSUPP);
281
282	switch (cmd) {
283	case SIOCSNDFLUSH_IN6:
284	case SIOCSPFXFLUSH_IN6:
285	case SIOCSRTRFLUSH_IN6:
286	case SIOCSDEFIFACE_IN6:
287	case SIOCSIFINFO_FLAGS:
288	case SIOCSIFINFO_IN6:
289		if (td != NULL) {
290			error = priv_check(td, PRIV_NETINET_ND6);
291			if (error)
292				return (error);
293		}
294		/* FALLTHROUGH */
295	case OSIOCGIFINFO_IN6:
296	case SIOCGIFINFO_IN6:
297	case SIOCGNBRINFO_IN6:
298	case SIOCGDEFIFACE_IN6:
299		return (nd6_ioctl(cmd, data, ifp));
300
301#ifdef COMPAT_FREEBSD32
302	case SIOCGDEFIFACE32_IN6:
303		{
304			struct in6_ndifreq ndif;
305			struct in6_ndifreq32 *ndif32;
306
307			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
308			    ifp);
309			if (error)
310				return (error);
311			ndif32 = (struct in6_ndifreq32 *)data;
312			ndif32->ifindex = ndif.ifindex;
313			return (0);
314		}
315#endif
316	}
317
318	switch (cmd) {
319	case SIOCSIFPREFIX_IN6:
320	case SIOCDIFPREFIX_IN6:
321	case SIOCAIFPREFIX_IN6:
322	case SIOCCIFPREFIX_IN6:
323	case SIOCSGIFPREFIX_IN6:
324	case SIOCGIFPREFIX_IN6:
325		log(LOG_NOTICE,
326		    "prefix ioctls are now invalidated. "
327		    "please use ifconfig.\n");
328		return (EOPNOTSUPP);
329	}
330
331	switch (cmd) {
332	case SIOCSSCOPE6:
333		if (td != NULL) {
334			error = priv_check(td, PRIV_NETINET_SCOPE6);
335			if (error)
336				return (error);
337		}
338		/* FALLTHROUGH */
339	case SIOCGSCOPE6:
340	case SIOCGSCOPE6DEF:
341		return (scope6_ioctl(cmd, data, ifp));
342	}
343
344	/*
345	 * Find address for this interface, if it exists.
346	 *
347	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
348	 * only, and used the first interface address as the target of other
349	 * operations (without checking ifra_addr).  This was because netinet
350	 * code/API assumed at most 1 interface address per interface.
351	 * Since IPv6 allows a node to assign multiple addresses
352	 * on a single interface, we almost always look and check the
353	 * presence of ifra_addr, and reject invalid ones here.
354	 * It also decreases duplicated code among SIOC*_IN6 operations.
355	 */
356	switch (cmd) {
357	case SIOCAIFADDR_IN6:
358	case SIOCSIFPHYADDR_IN6:
359		sa6 = &ifra->ifra_addr;
360		break;
361	case SIOCSIFADDR_IN6:
362	case SIOCGIFADDR_IN6:
363	case SIOCSIFDSTADDR_IN6:
364	case SIOCSIFNETMASK_IN6:
365	case SIOCGIFDSTADDR_IN6:
366	case SIOCGIFNETMASK_IN6:
367	case SIOCDIFADDR_IN6:
368	case SIOCGIFPSRCADDR_IN6:
369	case SIOCGIFPDSTADDR_IN6:
370	case SIOCGIFAFLAG_IN6:
371	case SIOCSNDFLUSH_IN6:
372	case SIOCSPFXFLUSH_IN6:
373	case SIOCSRTRFLUSH_IN6:
374	case SIOCGIFALIFETIME_IN6:
375	case SIOCGIFSTAT_IN6:
376	case SIOCGIFSTAT_ICMP6:
377		sa6 = &ifr->ifr_addr;
378		break;
379	case SIOCSIFADDR:
380	case SIOCSIFBRDADDR:
381	case SIOCSIFDSTADDR:
382	case SIOCSIFNETMASK:
383		/*
384		 * Although we should pass any non-INET6 ioctl requests
385		 * down to driver, we filter some legacy INET requests.
386		 * Drivers trust SIOCSIFADDR et al to come from an already
387		 * privileged layer, and do not perform any credentials
388		 * checks or input validation.
389		 */
390		return (EINVAL);
391	default:
392		sa6 = NULL;
393		break;
394	}
395	if (sa6 && sa6->sin6_family == AF_INET6) {
396		if (sa6->sin6_scope_id != 0)
397			error = sa6_embedscope(sa6, 0);
398		else
399			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
400		if (error != 0)
401			return (error);
402		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
403		    &sa6->sin6_addr)) != 0)
404			return (error);
405		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
406	} else
407		ia = NULL;
408
409	switch (cmd) {
410	case SIOCSIFADDR_IN6:
411	case SIOCSIFDSTADDR_IN6:
412	case SIOCSIFNETMASK_IN6:
413		/*
414		 * Since IPv6 allows a node to assign multiple addresses
415		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
416		 */
417		/* we decided to obsolete this command (20000704) */
418		error = EINVAL;
419		goto out;
420
421	case SIOCDIFADDR_IN6:
422		/*
423		 * for IPv4, we look for existing in_ifaddr here to allow
424		 * "ifconfig if0 delete" to remove the first IPv4 address on
425		 * the interface.  For IPv6, as the spec allows multiple
426		 * interface address from the day one, we consider "remove the
427		 * first one" semantics to be not preferable.
428		 */
429		if (ia == NULL) {
430			error = EADDRNOTAVAIL;
431			goto out;
432		}
433		/* FALLTHROUGH */
434	case SIOCAIFADDR_IN6:
435		/*
436		 * We always require users to specify a valid IPv6 address for
437		 * the corresponding operation.
438		 */
439		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
440		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
441			error = EAFNOSUPPORT;
442			goto out;
443		}
444
445		if (td != NULL) {
446			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
447			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
448			if (error)
449				goto out;
450		}
451		/* FALLTHROUGH */
452	case SIOCGIFSTAT_IN6:
453	case SIOCGIFSTAT_ICMP6:
454		if (ifp->if_afdata[AF_INET6] == NULL) {
455			error = EPFNOSUPPORT;
456			goto out;
457		}
458		break;
459
460	case SIOCGIFADDR_IN6:
461		/* This interface is basically deprecated. use SIOCGIFCONF. */
462		/* FALLTHROUGH */
463	case SIOCGIFAFLAG_IN6:
464	case SIOCGIFNETMASK_IN6:
465	case SIOCGIFDSTADDR_IN6:
466	case SIOCGIFALIFETIME_IN6:
467		/* must think again about its semantics */
468		if (ia == NULL) {
469			error = EADDRNOTAVAIL;
470			goto out;
471		}
472		break;
473	}
474
475	switch (cmd) {
476	case SIOCGIFADDR_IN6:
477		ifr->ifr_addr = ia->ia_addr;
478		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
479			goto out;
480		break;
481
482	case SIOCGIFDSTADDR_IN6:
483		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
484			error = EINVAL;
485			goto out;
486		}
487		ifr->ifr_dstaddr = ia->ia_dstaddr;
488		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
489			goto out;
490		break;
491
492	case SIOCGIFNETMASK_IN6:
493		ifr->ifr_addr = ia->ia_prefixmask;
494		break;
495
496	case SIOCGIFAFLAG_IN6:
497		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
498		break;
499
500	case SIOCGIFSTAT_IN6:
501		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
502		    ifp->if_afdata[AF_INET6])->in6_ifstat,
503		    &ifr->ifr_ifru.ifru_stat,
504		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
505		break;
506
507	case SIOCGIFSTAT_ICMP6:
508		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
509		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
510		    &ifr->ifr_ifru.ifru_icmp6stat,
511		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
512		break;
513
514	case SIOCGIFALIFETIME_IN6:
515		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
516		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
517			time_t maxexpire;
518			struct in6_addrlifetime *retlt =
519			    &ifr->ifr_ifru.ifru_lifetime;
520
521			/*
522			 * XXX: adjust expiration time assuming time_t is
523			 * signed.
524			 */
525			maxexpire = (-1) &
526			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
527			if (ia->ia6_lifetime.ia6t_vltime <
528			    maxexpire - ia->ia6_updatetime) {
529				retlt->ia6t_expire = ia->ia6_updatetime +
530				    ia->ia6_lifetime.ia6t_vltime;
531			} else
532				retlt->ia6t_expire = maxexpire;
533		}
534		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
535			time_t maxexpire;
536			struct in6_addrlifetime *retlt =
537			    &ifr->ifr_ifru.ifru_lifetime;
538
539			/*
540			 * XXX: adjust expiration time assuming time_t is
541			 * signed.
542			 */
543			maxexpire = (-1) &
544			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
545			if (ia->ia6_lifetime.ia6t_pltime <
546			    maxexpire - ia->ia6_updatetime) {
547				retlt->ia6t_preferred = ia->ia6_updatetime +
548				    ia->ia6_lifetime.ia6t_pltime;
549			} else
550				retlt->ia6t_preferred = maxexpire;
551		}
552		break;
553
554	case SIOCAIFADDR_IN6:
555	{
556		struct nd_prefixctl pr0;
557		struct nd_prefix *pr;
558
559		/*
560		 * first, make or update the interface address structure,
561		 * and link it to the list.
562		 */
563		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
564			goto out;
565		if (ia != NULL) {
566			if (ia->ia_ifa.ifa_carp)
567				(*carp_detach_p)(&ia->ia_ifa, true);
568			ifa_free(&ia->ia_ifa);
569		}
570		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
571		    == NULL) {
572			/*
573			 * this can happen when the user specify the 0 valid
574			 * lifetime.
575			 */
576			break;
577		}
578
579		if (cmd == ocmd && ifra->ifra_vhid > 0) {
580			if (carp_attach_p != NULL)
581				error = (*carp_attach_p)(&ia->ia_ifa,
582				    ifra->ifra_vhid);
583			else
584				error = EPROTONOSUPPORT;
585			if (error)
586				goto out;
587			else
588				carp_attached = 1;
589		}
590
591		/*
592		 * then, make the prefix on-link on the interface.
593		 * XXX: we'd rather create the prefix before the address, but
594		 * we need at least one address to install the corresponding
595		 * interface route, so we configure the address first.
596		 */
597
598		/*
599		 * convert mask to prefix length (prefixmask has already
600		 * been validated in in6_update_ifa().
601		 */
602		bzero(&pr0, sizeof(pr0));
603		pr0.ndpr_ifp = ifp;
604		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
605		    NULL);
606		if (pr0.ndpr_plen == 128) {
607			/* we don't need to install a host route. */
608			goto aifaddr_out;
609		}
610		pr0.ndpr_prefix = ifra->ifra_addr;
611		/* apply the mask for safety. */
612		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
613		    &ifra->ifra_prefixmask.sin6_addr);
614
615		/*
616		 * XXX: since we don't have an API to set prefix (not address)
617		 * lifetimes, we just use the same lifetimes as addresses.
618		 * The (temporarily) installed lifetimes can be overridden by
619		 * later advertised RAs (when accept_rtadv is non 0), which is
620		 * an intended behavior.
621		 */
622		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
623		pr0.ndpr_raf_auto =
624		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
625		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
626		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
627
628		/* add the prefix if not yet. */
629		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
630			/*
631			 * nd6_prelist_add will install the corresponding
632			 * interface route.
633			 */
634			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
635				if (carp_attached)
636					(*carp_detach_p)(&ia->ia_ifa, false);
637				goto out;
638			}
639		}
640
641		/* relate the address to the prefix */
642		if (ia->ia6_ndpr == NULL) {
643			ia->ia6_ndpr = pr;
644			pr->ndpr_addrcnt++;
645
646			/*
647			 * If this is the first autoconf address from the
648			 * prefix, create a temporary address as well
649			 * (when required).
650			 */
651			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
652			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
653				int e;
654				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
655					log(LOG_NOTICE, "in6_control: failed "
656					    "to create a temporary address, "
657					    "errno=%d\n", e);
658				}
659			}
660		}
661		nd6_prefix_rele(pr);
662
663		/*
664		 * this might affect the status of autoconfigured addresses,
665		 * that is, this address might make other addresses detached.
666		 */
667		pfxlist_onlink_check();
668
669aifaddr_out:
670		/*
671		 * Try to clear the flag when a new IPv6 address is added
672		 * onto an IFDISABLED interface and it succeeds.
673		 */
674		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
675			struct in6_ndireq nd;
676
677			memset(&nd, 0, sizeof(nd));
678			nd.ndi.flags = ND_IFINFO(ifp)->flags;
679			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
680			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
681				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
682				    "SIOCSIFINFO_FLAGS for -ifdisabled "
683				    "failed.");
684			/*
685			 * Ignore failure of clearing the flag intentionally.
686			 * The failure means address duplication was detected.
687			 */
688		}
689		break;
690	}
691
692	case SIOCDIFADDR_IN6:
693	{
694		struct nd_prefix *pr;
695
696		/*
697		 * If the address being deleted is the only one that owns
698		 * the corresponding prefix, expire the prefix as well.
699		 * XXX: theoretically, we don't have to worry about such
700		 * relationship, since we separate the address management
701		 * and the prefix management.  We do this, however, to provide
702		 * as much backward compatibility as possible in terms of
703		 * the ioctl operation.
704		 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
705		 */
706		pr = ia->ia6_ndpr;
707		in6_purgeaddr(&ia->ia_ifa);
708		if (pr != NULL && pr->ndpr_addrcnt == 0) {
709			ND6_WLOCK();
710			nd6_prefix_unlink(pr, NULL);
711			ND6_WUNLOCK();
712			nd6_prefix_del(pr);
713		}
714		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
715		break;
716	}
717
718	default:
719		if (ifp->if_ioctl == NULL) {
720			error = EOPNOTSUPP;
721			goto out;
722		}
723		error = (*ifp->if_ioctl)(ifp, cmd, data);
724		goto out;
725	}
726
727	error = 0;
728out:
729	if (ia != NULL)
730		ifa_free(&ia->ia_ifa);
731	return (error);
732}
733
734
735static struct in6_multi_mship *
736in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
737    int *errorp, int delay)
738{
739	struct in6_multi_mship *imm;
740	int error;
741
742	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
743	if (imm == NULL) {
744		*errorp = ENOBUFS;
745		return (NULL);
746	}
747
748	delay = (delay * PR_FASTHZ) / hz;
749
750	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
751	if (error) {
752		*errorp = error;
753		free(imm, M_IP6MADDR);
754		return (NULL);
755	}
756
757	return (imm);
758}
759/*
760 * Join necessary multicast groups.  Factored out from in6_update_ifa().
761 * This entire work should only be done once, for the default FIB.
762 */
763static int
764in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
765    struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
766{
767	char ip6buf[INET6_ADDRSTRLEN];
768	struct in6_addr mltaddr;
769	struct in6_multi_mship *imm;
770	int delay, error;
771
772	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
773
774	/* Join solicited multicast addr for new host id. */
775	bzero(&mltaddr, sizeof(struct in6_addr));
776	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
777	mltaddr.s6_addr32[2] = htonl(1);
778	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
779	mltaddr.s6_addr8[12] = 0xff;
780	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
781		/* XXX: should not happen */
782		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
783		goto cleanup;
784	}
785	delay = error = 0;
786	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
787		/*
788		 * We need a random delay for DAD on the address being
789		 * configured.  It also means delaying transmission of the
790		 * corresponding MLD report to avoid report collision.
791		 * [RFC 4861, Section 6.3.7]
792		 */
793		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
794	}
795	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
796	if (imm == NULL) {
797		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
798		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
799		    if_name(ifp), error));
800		goto cleanup;
801	}
802	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
803	*in6m_sol = imm->i6mm_maddr;
804
805	/*
806	 * Join link-local all-nodes address.
807	 */
808	mltaddr = in6addr_linklocal_allnodes;
809	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
810		goto cleanup; /* XXX: should not fail */
811
812	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
813	if (imm == NULL) {
814		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
815		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
816		    if_name(ifp), error));
817		goto cleanup;
818	}
819	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
820
821	/*
822	 * Join node information group address.
823	 */
824	delay = 0;
825	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
826		/*
827		 * The spec does not say anything about delay for this group,
828		 * but the same logic should apply.
829		 */
830		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
831	}
832	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
833		/* XXX jinmei */
834		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
835		if (imm == NULL)
836			nd6log((LOG_WARNING,
837			    "%s: in6_joingroup failed for %s on %s "
838			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
839			    &mltaddr), if_name(ifp), error));
840			/* XXX not very fatal, go on... */
841		else
842			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
843	}
844	if (V_icmp6_nodeinfo_oldmcprefix &&
845	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
846		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
847		if (imm == NULL)
848			nd6log((LOG_WARNING,
849			    "%s: in6_joingroup failed for %s on %s "
850			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
851			    &mltaddr), if_name(ifp), error));
852			/* XXX not very fatal, go on... */
853		else
854			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
855	}
856
857	/*
858	 * Join interface-local all-nodes address.
859	 * (ff01::1%ifN, and ff01::%ifN/32)
860	 */
861	mltaddr = in6addr_nodelocal_allnodes;
862	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
863		goto cleanup; /* XXX: should not fail */
864
865	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
866	if (imm == NULL) {
867		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
868		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
869		    &mltaddr), if_name(ifp), error));
870		goto cleanup;
871	}
872	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
873
874cleanup:
875	return (error);
876}
877
878/*
879 * Update parameters of an IPv6 interface address.
880 * If necessary, a new entry is created and linked into address chains.
881 * This function is separated from in6_control().
882 */
883int
884in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
885    struct in6_ifaddr *ia, int flags)
886{
887	int error, hostIsNew = 0;
888
889	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
890		return (error);
891
892	if (ia == NULL) {
893		hostIsNew = 1;
894		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
895			return (ENOBUFS);
896	}
897
898	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
899	if (error != 0) {
900		if (hostIsNew != 0) {
901			in6_unlink_ifa(ia, ifp);
902			ifa_free(&ia->ia_ifa);
903		}
904		return (error);
905	}
906
907	if (hostIsNew)
908		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
909
910	return (error);
911}
912
913/*
914 * Fill in basic IPv6 address request info.
915 */
916void
917in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
918    const struct in6_addr *mask)
919{
920
921	memset(ifra, 0, sizeof(struct in6_aliasreq));
922
923	ifra->ifra_addr.sin6_family = AF_INET6;
924	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
925	if (addr != NULL)
926		ifra->ifra_addr.sin6_addr = *addr;
927
928	ifra->ifra_prefixmask.sin6_family = AF_INET6;
929	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
930	if (mask != NULL)
931		ifra->ifra_prefixmask.sin6_addr = *mask;
932}
933
934static int
935in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
936    struct in6_ifaddr *ia, int flags)
937{
938	int plen = -1;
939	struct sockaddr_in6 dst6;
940	struct in6_addrlifetime *lt;
941	char ip6buf[INET6_ADDRSTRLEN];
942
943	/* Validate parameters */
944	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
945		return (EINVAL);
946
947	/*
948	 * The destination address for a p2p link must have a family
949	 * of AF_UNSPEC or AF_INET6.
950	 */
951	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
952	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
953	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
954		return (EAFNOSUPPORT);
955
956	/*
957	 * Validate address
958	 */
959	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
960	    ifra->ifra_addr.sin6_family != AF_INET6)
961		return (EINVAL);
962
963	/*
964	 * validate ifra_prefixmask.  don't check sin6_family, netmask
965	 * does not carry fields other than sin6_len.
966	 */
967	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
968		return (EINVAL);
969	/*
970	 * Because the IPv6 address architecture is classless, we require
971	 * users to specify a (non 0) prefix length (mask) for a new address.
972	 * We also require the prefix (when specified) mask is valid, and thus
973	 * reject a non-consecutive mask.
974	 */
975	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
976		return (EINVAL);
977	if (ifra->ifra_prefixmask.sin6_len != 0) {
978		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
979		    (u_char *)&ifra->ifra_prefixmask +
980		    ifra->ifra_prefixmask.sin6_len);
981		if (plen <= 0)
982			return (EINVAL);
983	} else {
984		/*
985		 * In this case, ia must not be NULL.  We just use its prefix
986		 * length.
987		 */
988		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
989	}
990	/*
991	 * If the destination address on a p2p interface is specified,
992	 * and the address is a scoped one, validate/set the scope
993	 * zone identifier.
994	 */
995	dst6 = ifra->ifra_dstaddr;
996	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
997	    (dst6.sin6_family == AF_INET6)) {
998		struct in6_addr in6_tmp;
999		u_int32_t zoneid;
1000
1001		in6_tmp = dst6.sin6_addr;
1002		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1003			return (EINVAL); /* XXX: should be impossible */
1004
1005		if (dst6.sin6_scope_id != 0) {
1006			if (dst6.sin6_scope_id != zoneid)
1007				return (EINVAL);
1008		} else		/* user omit to specify the ID. */
1009			dst6.sin6_scope_id = zoneid;
1010
1011		/* convert into the internal form */
1012		if (sa6_embedscope(&dst6, 0))
1013			return (EINVAL); /* XXX: should be impossible */
1014	}
1015	/* Modify original ifra_dstaddr to reflect changes */
1016	ifra->ifra_dstaddr = dst6;
1017
1018	/*
1019	 * The destination address can be specified only for a p2p or a
1020	 * loopback interface.  If specified, the corresponding prefix length
1021	 * must be 128.
1022	 */
1023	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1024		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1025			/* XXX: noisy message */
1026			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1027			    "be specified for a p2p or a loopback IF only\n"));
1028			return (EINVAL);
1029		}
1030		if (plen != 128) {
1031			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1032			    "be 128 when dstaddr is specified\n"));
1033			return (EINVAL);
1034		}
1035	}
1036	/* lifetime consistency check */
1037	lt = &ifra->ifra_lifetime;
1038	if (lt->ia6t_pltime > lt->ia6t_vltime)
1039		return (EINVAL);
1040	if (lt->ia6t_vltime == 0) {
1041		/*
1042		 * the following log might be noisy, but this is a typical
1043		 * configuration mistake or a tool's bug.
1044		 */
1045		nd6log((LOG_INFO,
1046		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1047		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1048
1049		if (ia == NULL)
1050			return (0); /* there's nothing to do */
1051	}
1052
1053	/* Check prefix mask */
1054	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1055		/*
1056		 * We prohibit changing the prefix length of an existing
1057		 * address, because
1058		 * + such an operation should be rare in IPv6, and
1059		 * + the operation would confuse prefix management.
1060		 */
1061		if (ia->ia_prefixmask.sin6_len != 0 &&
1062		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1063			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1064			    "of an existing %s address should not be changed\n",
1065			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1066
1067			return (EINVAL);
1068		}
1069	}
1070
1071	return (0);
1072}
1073
1074
1075/*
1076 * Allocate a new ifaddr and link it into chains.
1077 */
1078static struct in6_ifaddr *
1079in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1080{
1081	struct in6_ifaddr *ia;
1082
1083	/*
1084	 * When in6_alloc_ifa() is called in a process of a received
1085	 * RA, it is called under an interrupt context.  So, we should
1086	 * call malloc with M_NOWAIT.
1087	 */
1088	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1089	if (ia == NULL)
1090		return (NULL);
1091	LIST_INIT(&ia->ia6_memberships);
1092	/* Initialize the address and masks, and put time stamp */
1093	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1094	ia->ia_addr.sin6_family = AF_INET6;
1095	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1096	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1097	ia->ia_addr = ifra->ifra_addr;
1098	ia->ia6_createtime = time_uptime;
1099	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1100		/*
1101		 * Some functions expect that ifa_dstaddr is not
1102		 * NULL for p2p interfaces.
1103		 */
1104		ia->ia_ifa.ifa_dstaddr =
1105		    (struct sockaddr *)&ia->ia_dstaddr;
1106	} else {
1107		ia->ia_ifa.ifa_dstaddr = NULL;
1108	}
1109
1110	/* set prefix mask if any */
1111	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1112	if (ifra->ifra_prefixmask.sin6_len != 0) {
1113		ia->ia_prefixmask.sin6_family = AF_INET6;
1114		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1115		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1116	}
1117
1118	ia->ia_ifp = ifp;
1119	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1120	IF_ADDR_WLOCK(ifp);
1121	TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1122	IF_ADDR_WUNLOCK(ifp);
1123
1124	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1125	IN6_IFADDR_WLOCK();
1126	TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1127	LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1128	IN6_IFADDR_WUNLOCK();
1129
1130	return (ia);
1131}
1132
1133/*
1134 * Update/configure interface address parameters:
1135 *
1136 * 1) Update lifetime
1137 * 2) Update interface metric ad flags
1138 * 3) Notify other subsystems
1139 */
1140static int
1141in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1142    struct in6_ifaddr *ia, int hostIsNew, int flags)
1143{
1144	int error;
1145
1146	/* update timestamp */
1147	ia->ia6_updatetime = time_uptime;
1148
1149	/*
1150	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1151	 * to see if the address is deprecated or invalidated, but initialize
1152	 * these members for applications.
1153	 */
1154	ia->ia6_lifetime = ifra->ifra_lifetime;
1155	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1156		ia->ia6_lifetime.ia6t_expire =
1157		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1158	} else
1159		ia->ia6_lifetime.ia6t_expire = 0;
1160	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1161		ia->ia6_lifetime.ia6t_preferred =
1162		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1163	} else
1164		ia->ia6_lifetime.ia6t_preferred = 0;
1165
1166	/*
1167	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1168	 * userland, make it deprecated.
1169	 */
1170	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1171		ia->ia6_lifetime.ia6t_pltime = 0;
1172		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1173	}
1174
1175	/*
1176	 * configure address flags.
1177	 */
1178	ia->ia6_flags = ifra->ifra_flags;
1179
1180	/*
1181	 * Make the address tentative before joining multicast addresses,
1182	 * so that corresponding MLD responses would not have a tentative
1183	 * source address.
1184	 */
1185	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1186
1187	/*
1188	 * DAD should be performed for an new address or addresses on
1189	 * an interface with ND6_IFF_IFDISABLED.
1190	 */
1191	if (in6if_do_dad(ifp) &&
1192	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1193		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1194
1195	/* notify other subsystems */
1196	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1197
1198	return (error);
1199}
1200
1201/*
1202 * Do link-level ifa job:
1203 * 1) Add lle entry for added address
1204 * 2) Notifies routing socket users about new address
1205 * 3) join appropriate multicast group
1206 * 4) start DAD if enabled
1207 */
1208static int
1209in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1210    struct in6_ifaddr *ia, int flags)
1211{
1212	struct in6_multi *in6m_sol;
1213	int error = 0;
1214
1215	/* Add local address to lltable, if necessary (ex. on p2p link). */
1216	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1217		in6_purgeaddr(&ia->ia_ifa);
1218		ifa_free(&ia->ia_ifa);
1219		return (error);
1220	}
1221
1222	/* Join necessary multicast groups. */
1223	in6m_sol = NULL;
1224	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1225		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1226		if (error != 0) {
1227			in6_purgeaddr(&ia->ia_ifa);
1228			ifa_free(&ia->ia_ifa);
1229			return (error);
1230		}
1231	}
1232
1233	/* Perform DAD, if the address is TENTATIVE. */
1234	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1235		int delay, mindelay, maxdelay;
1236
1237		delay = 0;
1238		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1239			/*
1240			 * We need to impose a delay before sending an NS
1241			 * for DAD.  Check if we also needed a delay for the
1242			 * corresponding MLD message.  If we did, the delay
1243			 * should be larger than the MLD delay (this could be
1244			 * relaxed a bit, but this simple logic is at least
1245			 * safe).
1246			 * XXX: Break data hiding guidelines and look at
1247			 * state for the solicited multicast group.
1248			 */
1249			mindelay = 0;
1250			if (in6m_sol != NULL &&
1251			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1252				mindelay = in6m_sol->in6m_timer;
1253			}
1254			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1255			if (maxdelay - mindelay == 0)
1256				delay = 0;
1257			else {
1258				delay =
1259				    (arc4random() % (maxdelay - mindelay)) +
1260				    mindelay;
1261			}
1262		}
1263		nd6_dad_start((struct ifaddr *)ia, delay);
1264	}
1265
1266	in6_newaddrmsg(ia, RTM_ADD);
1267	ifa_free(&ia->ia_ifa);
1268	return (error);
1269}
1270
1271void
1272in6_purgeaddr(struct ifaddr *ifa)
1273{
1274	struct ifnet *ifp = ifa->ifa_ifp;
1275	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1276	struct in6_multi_mship *imm;
1277	int plen, error;
1278
1279	if (ifa->ifa_carp)
1280		(*carp_detach_p)(ifa, false);
1281
1282	/*
1283	 * Remove the loopback route to the interface address.
1284	 * The check for the current setting of "nd6_useloopback"
1285	 * is not needed.
1286	 */
1287	if (ia->ia_flags & IFA_RTSELF) {
1288		error = ifa_del_loopback_route((struct ifaddr *)ia,
1289		    (struct sockaddr *)&ia->ia_addr);
1290		if (error == 0)
1291			ia->ia_flags &= ~IFA_RTSELF;
1292	}
1293
1294	/* stop DAD processing */
1295	nd6_dad_stop(ifa);
1296
1297	/* Leave multicast groups. */
1298	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1299		LIST_REMOVE(imm, i6mm_chain);
1300		if (imm->i6mm_maddr != NULL)
1301			in6_leavegroup(imm->i6mm_maddr, NULL);
1302		free(imm, M_IP6MADDR);
1303	}
1304	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1305	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1306		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1307		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1308		if (error != 0)
1309			log(LOG_INFO, "%s: err=%d, destination address delete "
1310			    "failed\n", __func__, error);
1311		ia->ia_flags &= ~IFA_ROUTE;
1312	}
1313
1314	in6_newaddrmsg(ia, RTM_DELETE);
1315	in6_unlink_ifa(ia, ifp);
1316}
1317
1318static void
1319in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1320{
1321	char ip6buf[INET6_ADDRSTRLEN];
1322	int remove_lle;
1323
1324	IF_ADDR_WLOCK(ifp);
1325	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1326	IF_ADDR_WUNLOCK(ifp);
1327	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1328
1329	/*
1330	 * Defer the release of what might be the last reference to the
1331	 * in6_ifaddr so that it can't be freed before the remainder of the
1332	 * cleanup.
1333	 */
1334	IN6_IFADDR_WLOCK();
1335	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1336	LIST_REMOVE(ia, ia6_hash);
1337	IN6_IFADDR_WUNLOCK();
1338
1339	/*
1340	 * Release the reference to the base prefix.  There should be a
1341	 * positive reference.
1342	 */
1343	remove_lle = 0;
1344	if (ia->ia6_ndpr == NULL) {
1345		nd6log((LOG_NOTICE,
1346		    "in6_unlink_ifa: autoconf'ed address "
1347		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1348	} else {
1349		ia->ia6_ndpr->ndpr_addrcnt--;
1350		/* Do not delete lles within prefix if refcont != 0 */
1351		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1352			remove_lle = 1;
1353		ia->ia6_ndpr = NULL;
1354	}
1355
1356	nd6_rem_ifa_lle(ia, remove_lle);
1357
1358	/*
1359	 * Also, if the address being removed is autoconf'ed, call
1360	 * pfxlist_onlink_check() since the release might affect the status of
1361	 * other (detached) addresses.
1362	 */
1363	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1364		pfxlist_onlink_check();
1365	}
1366	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1367}
1368
1369/*
1370 * Notifies other subsystems about address change/arrival:
1371 * 1) Notifies device handler on the first IPv6 address assignment
1372 * 2) Handle routing table changes for P2P links and route
1373 * 3) Handle routing table changes for address host route
1374 */
1375static int
1376in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1377    struct in6_aliasreq *ifra, int hostIsNew)
1378{
1379	int	error = 0, plen, ifacount = 0;
1380	struct ifaddr *ifa;
1381	struct sockaddr_in6 *pdst;
1382	char ip6buf[INET6_ADDRSTRLEN];
1383
1384	/*
1385	 * Give the interface a chance to initialize
1386	 * if this is its first address,
1387	 */
1388	if (hostIsNew != 0) {
1389		IF_ADDR_RLOCK(ifp);
1390		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1391			if (ifa->ifa_addr->sa_family != AF_INET6)
1392				continue;
1393			ifacount++;
1394		}
1395		IF_ADDR_RUNLOCK(ifp);
1396	}
1397
1398	if (ifacount <= 1 && ifp->if_ioctl) {
1399		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1400		if (error)
1401			goto done;
1402	}
1403
1404	/*
1405	 * If a new destination address is specified, scrub the old one and
1406	 * install the new destination.  Note that the interface must be
1407	 * p2p or loopback.
1408	 */
1409	pdst = &ifra->ifra_dstaddr;
1410	if (pdst->sin6_family == AF_INET6 &&
1411	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1412		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1413		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1414			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1415			    "remove a route to the old destination: %s\n",
1416			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1417			/* proceed anyway... */
1418		} else
1419			ia->ia_flags &= ~IFA_ROUTE;
1420		ia->ia_dstaddr = *pdst;
1421	}
1422
1423	/*
1424	 * If a new destination address is specified for a point-to-point
1425	 * interface, install a route to the destination as an interface
1426	 * direct route.
1427	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1428	 * interface that share the same destination.
1429	 */
1430	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1431	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1432	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1433		int rtflags = RTF_UP | RTF_HOST;
1434		/*
1435		 * Handle the case for ::1 .
1436		 */
1437		if (ifp->if_flags & IFF_LOOPBACK)
1438			ia->ia_flags |= IFA_RTSELF;
1439		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1440		if (error)
1441			goto done;
1442		ia->ia_flags |= IFA_ROUTE;
1443	}
1444
1445	/*
1446	 * add a loopback route to self if not exists
1447	 */
1448	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1449		error = ifa_add_loopback_route((struct ifaddr *)ia,
1450		    (struct sockaddr *)&ia->ia_addr);
1451		if (error == 0)
1452			ia->ia_flags |= IFA_RTSELF;
1453	}
1454done:
1455	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1456	    "Invoking IPv6 network device address event may sleep");
1457
1458	EVENTHANDLER_INVOKE(ifaddr_event, ifp);
1459
1460	return (error);
1461}
1462
1463/*
1464 * Find an IPv6 interface link-local address specific to an interface.
1465 * ifaddr is returned referenced.
1466 */
1467struct in6_ifaddr *
1468in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1469{
1470	struct ifaddr *ifa;
1471
1472	IF_ADDR_RLOCK(ifp);
1473	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1474		if (ifa->ifa_addr->sa_family != AF_INET6)
1475			continue;
1476		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1477			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1478			    ignoreflags) != 0)
1479				continue;
1480			ifa_ref(ifa);
1481			break;
1482		}
1483	}
1484	IF_ADDR_RUNLOCK(ifp);
1485
1486	return ((struct in6_ifaddr *)ifa);
1487}
1488
1489
1490/*
1491 * find the interface address corresponding to a given IPv6 address.
1492 * ifaddr is returned referenced.
1493 */
1494struct in6_ifaddr *
1495in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1496{
1497	struct rm_priotracker in6_ifa_tracker;
1498	struct in6_ifaddr *ia;
1499
1500	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1501	LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1502		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1503			if (zoneid != 0 &&
1504			    zoneid != ia->ia_addr.sin6_scope_id)
1505				continue;
1506			ifa_ref(&ia->ia_ifa);
1507			break;
1508		}
1509	}
1510	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1511	return (ia);
1512}
1513
1514/*
1515 * find the internet address corresponding to a given interface and address.
1516 * ifaddr is returned referenced.
1517 */
1518struct in6_ifaddr *
1519in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1520{
1521	struct ifaddr *ifa;
1522
1523	IF_ADDR_RLOCK(ifp);
1524	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1525		if (ifa->ifa_addr->sa_family != AF_INET6)
1526			continue;
1527		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1528			ifa_ref(ifa);
1529			break;
1530		}
1531	}
1532	IF_ADDR_RUNLOCK(ifp);
1533
1534	return ((struct in6_ifaddr *)ifa);
1535}
1536
1537/*
1538 * Find a link-local scoped address on ifp and return it if any.
1539 */
1540struct in6_ifaddr *
1541in6ifa_llaonifp(struct ifnet *ifp)
1542{
1543	struct sockaddr_in6 *sin6;
1544	struct ifaddr *ifa;
1545
1546	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1547		return (NULL);
1548	IF_ADDR_RLOCK(ifp);
1549	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1550		if (ifa->ifa_addr->sa_family != AF_INET6)
1551			continue;
1552		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1553		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1554		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1555		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1556			break;
1557	}
1558	IF_ADDR_RUNLOCK(ifp);
1559
1560	return ((struct in6_ifaddr *)ifa);
1561}
1562
1563/*
1564 * Convert IP6 address to printable (loggable) representation. Caller
1565 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1566 */
1567static char digits[] = "0123456789abcdef";
1568char *
1569ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1570{
1571	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1572	char *cp;
1573	const u_int16_t *a = (const u_int16_t *)addr;
1574	const u_int8_t *d;
1575	int dcolon = 0, zero = 0;
1576
1577	cp = ip6buf;
1578
1579	for (i = 0; i < 8; i++) {
1580		if (*(a + i) == 0) {
1581			cnt++;
1582			if (cnt == 1)
1583				idx = i;
1584		}
1585		else if (maxcnt < cnt) {
1586			maxcnt = cnt;
1587			index = idx;
1588			cnt = 0;
1589		}
1590	}
1591	if (maxcnt < cnt) {
1592		maxcnt = cnt;
1593		index = idx;
1594	}
1595
1596	for (i = 0; i < 8; i++) {
1597		if (dcolon == 1) {
1598			if (*a == 0) {
1599				if (i == 7)
1600					*cp++ = ':';
1601				a++;
1602				continue;
1603			} else
1604				dcolon = 2;
1605		}
1606		if (*a == 0) {
1607			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1608				if (i == 0)
1609					*cp++ = ':';
1610				*cp++ = ':';
1611				dcolon = 1;
1612			} else {
1613				*cp++ = '0';
1614				*cp++ = ':';
1615			}
1616			a++;
1617			continue;
1618		}
1619		d = (const u_char *)a;
1620		/* Try to eliminate leading zeros in printout like in :0001. */
1621		zero = 1;
1622		*cp = digits[*d >> 4];
1623		if (*cp != '0') {
1624			zero = 0;
1625			cp++;
1626		}
1627		*cp = digits[*d++ & 0xf];
1628		if (zero == 0 || (*cp != '0')) {
1629			zero = 0;
1630			cp++;
1631		}
1632		*cp = digits[*d >> 4];
1633		if (zero == 0 || (*cp != '0')) {
1634			zero = 0;
1635			cp++;
1636		}
1637		*cp++ = digits[*d & 0xf];
1638		*cp++ = ':';
1639		a++;
1640	}
1641	*--cp = '\0';
1642	return (ip6buf);
1643}
1644
1645int
1646in6_localaddr(struct in6_addr *in6)
1647{
1648	struct rm_priotracker in6_ifa_tracker;
1649	struct in6_ifaddr *ia;
1650
1651	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1652		return 1;
1653
1654	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1655	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1656		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1657		    &ia->ia_prefixmask.sin6_addr)) {
1658			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1659			return 1;
1660		}
1661	}
1662	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1663
1664	return (0);
1665}
1666
1667/*
1668 * Return 1 if an internet address is for the local host and configured
1669 * on one of its interfaces.
1670 */
1671int
1672in6_localip(struct in6_addr *in6)
1673{
1674	struct rm_priotracker in6_ifa_tracker;
1675	struct in6_ifaddr *ia;
1676
1677	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1678	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1679		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1680			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1681			return (1);
1682		}
1683	}
1684	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1685	return (0);
1686}
1687
1688/*
1689 * Return 1 if an internet address is configured on an interface.
1690 */
1691int
1692in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1693{
1694	struct in6_addr in6;
1695	struct ifaddr *ifa;
1696	struct in6_ifaddr *ia6;
1697
1698	in6 = *addr;
1699	if (in6_clearscope(&in6))
1700		return (0);
1701	in6_setscope(&in6, ifp, NULL);
1702
1703	IF_ADDR_RLOCK(ifp);
1704	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1705		if (ifa->ifa_addr->sa_family != AF_INET6)
1706			continue;
1707		ia6 = (struct in6_ifaddr *)ifa;
1708		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6)) {
1709			IF_ADDR_RUNLOCK(ifp);
1710			return (1);
1711		}
1712	}
1713	IF_ADDR_RUNLOCK(ifp);
1714
1715	return (0);
1716}
1717
1718int
1719in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1720{
1721	struct rm_priotracker in6_ifa_tracker;
1722	struct in6_ifaddr *ia;
1723
1724	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1725	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1726		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1727			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1728				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1729				return (1); /* true */
1730			}
1731			break;
1732		}
1733	}
1734	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1735
1736	return (0);		/* false */
1737}
1738
1739/*
1740 * return length of part which dst and src are equal
1741 * hard coding...
1742 */
1743int
1744in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1745{
1746	int match = 0;
1747	u_char *s = (u_char *)src, *d = (u_char *)dst;
1748	u_char *lim = s + 16, r;
1749
1750	while (s < lim)
1751		if ((r = (*d++ ^ *s++)) != 0) {
1752			while (r < 128) {
1753				match++;
1754				r <<= 1;
1755			}
1756			break;
1757		} else
1758			match += 8;
1759	return match;
1760}
1761
1762/* XXX: to be scope conscious */
1763int
1764in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1765{
1766	int bytelen, bitlen;
1767
1768	/* sanity check */
1769	if (0 > len || len > 128) {
1770		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1771		    len);
1772		return (0);
1773	}
1774
1775	bytelen = len / 8;
1776	bitlen = len % 8;
1777
1778	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1779		return (0);
1780	if (bitlen != 0 &&
1781	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1782	    p2->s6_addr[bytelen] >> (8 - bitlen))
1783		return (0);
1784
1785	return (1);
1786}
1787
1788void
1789in6_prefixlen2mask(struct in6_addr *maskp, int len)
1790{
1791	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1792	int bytelen, bitlen, i;
1793
1794	/* sanity check */
1795	if (0 > len || len > 128) {
1796		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1797		    len);
1798		return;
1799	}
1800
1801	bzero(maskp, sizeof(*maskp));
1802	bytelen = len / 8;
1803	bitlen = len % 8;
1804	for (i = 0; i < bytelen; i++)
1805		maskp->s6_addr[i] = 0xff;
1806	if (bitlen)
1807		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1808}
1809
1810/*
1811 * return the best address out of the same scope. if no address was
1812 * found, return the first valid address from designated IF.
1813 */
1814struct in6_ifaddr *
1815in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1816{
1817	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1818	struct ifaddr *ifa;
1819	struct in6_ifaddr *besta = NULL;
1820	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1821
1822	dep[0] = dep[1] = NULL;
1823
1824	/*
1825	 * We first look for addresses in the same scope.
1826	 * If there is one, return it.
1827	 * If two or more, return one which matches the dst longest.
1828	 * If none, return one of global addresses assigned other ifs.
1829	 */
1830	IF_ADDR_RLOCK(ifp);
1831	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1832		if (ifa->ifa_addr->sa_family != AF_INET6)
1833			continue;
1834		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1835			continue; /* XXX: is there any case to allow anycast? */
1836		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1837			continue; /* don't use this interface */
1838		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1839			continue;
1840		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1841			if (V_ip6_use_deprecated)
1842				dep[0] = (struct in6_ifaddr *)ifa;
1843			continue;
1844		}
1845
1846		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1847			/*
1848			 * call in6_matchlen() as few as possible
1849			 */
1850			if (besta) {
1851				if (blen == -1)
1852					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1853				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1854				if (tlen > blen) {
1855					blen = tlen;
1856					besta = (struct in6_ifaddr *)ifa;
1857				}
1858			} else
1859				besta = (struct in6_ifaddr *)ifa;
1860		}
1861	}
1862	if (besta) {
1863		ifa_ref(&besta->ia_ifa);
1864		IF_ADDR_RUNLOCK(ifp);
1865		return (besta);
1866	}
1867
1868	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1869		if (ifa->ifa_addr->sa_family != AF_INET6)
1870			continue;
1871		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1872			continue; /* XXX: is there any case to allow anycast? */
1873		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1874			continue; /* don't use this interface */
1875		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1876			continue;
1877		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1878			if (V_ip6_use_deprecated)
1879				dep[1] = (struct in6_ifaddr *)ifa;
1880			continue;
1881		}
1882
1883		if (ifa != NULL)
1884			ifa_ref(ifa);
1885		IF_ADDR_RUNLOCK(ifp);
1886		return (struct in6_ifaddr *)ifa;
1887	}
1888
1889	/* use the last-resort values, that are, deprecated addresses */
1890	if (dep[0]) {
1891		ifa_ref((struct ifaddr *)dep[0]);
1892		IF_ADDR_RUNLOCK(ifp);
1893		return dep[0];
1894	}
1895	if (dep[1]) {
1896		ifa_ref((struct ifaddr *)dep[1]);
1897		IF_ADDR_RUNLOCK(ifp);
1898		return dep[1];
1899	}
1900
1901	IF_ADDR_RUNLOCK(ifp);
1902	return NULL;
1903}
1904
1905/*
1906 * perform DAD when interface becomes IFF_UP.
1907 */
1908void
1909in6_if_up(struct ifnet *ifp)
1910{
1911	struct ifaddr *ifa;
1912	struct in6_ifaddr *ia;
1913
1914	IF_ADDR_RLOCK(ifp);
1915	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1916		if (ifa->ifa_addr->sa_family != AF_INET6)
1917			continue;
1918		ia = (struct in6_ifaddr *)ifa;
1919		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1920			/*
1921			 * The TENTATIVE flag was likely set by hand
1922			 * beforehand, implicitly indicating the need for DAD.
1923			 * We may be able to skip the random delay in this
1924			 * case, but we impose delays just in case.
1925			 */
1926			nd6_dad_start(ifa,
1927			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1928		}
1929	}
1930	IF_ADDR_RUNLOCK(ifp);
1931
1932	/*
1933	 * special cases, like 6to4, are handled in in6_ifattach
1934	 */
1935	in6_ifattach(ifp, NULL);
1936}
1937
1938int
1939in6if_do_dad(struct ifnet *ifp)
1940{
1941	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1942		return (0);
1943
1944	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) ||
1945	    (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD))
1946		return (0);
1947
1948	/*
1949	 * Our DAD routine requires the interface up and running.
1950	 * However, some interfaces can be up before the RUNNING
1951	 * status.  Additionally, users may try to assign addresses
1952	 * before the interface becomes up (or running).
1953	 * This function returns EAGAIN in that case.
1954	 * The caller should mark "tentative" on the address instead of
1955	 * performing DAD immediately.
1956	 */
1957	if (!((ifp->if_flags & IFF_UP) &&
1958	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
1959		return (EAGAIN);
1960
1961	return (1);
1962}
1963
1964/*
1965 * Calculate max IPv6 MTU through all the interfaces and store it
1966 * to in6_maxmtu.
1967 */
1968void
1969in6_setmaxmtu(void)
1970{
1971	unsigned long maxmtu = 0;
1972	struct ifnet *ifp;
1973
1974	IFNET_RLOCK_NOSLEEP();
1975	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1976		/* this function can be called during ifnet initialization */
1977		if (!ifp->if_afdata[AF_INET6])
1978			continue;
1979		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1980		    IN6_LINKMTU(ifp) > maxmtu)
1981			maxmtu = IN6_LINKMTU(ifp);
1982	}
1983	IFNET_RUNLOCK_NOSLEEP();
1984	if (maxmtu)	/* update only when maxmtu is positive */
1985		V_in6_maxmtu = maxmtu;
1986}
1987
1988/*
1989 * Provide the length of interface identifiers to be used for the link attached
1990 * to the given interface.  The length should be defined in "IPv6 over
1991 * xxx-link" document.  Note that address architecture might also define
1992 * the length for a particular set of address prefixes, regardless of the
1993 * link type.  As clarified in rfc2462bis, those two definitions should be
1994 * consistent, and those really are as of August 2004.
1995 */
1996int
1997in6_if2idlen(struct ifnet *ifp)
1998{
1999	switch (ifp->if_type) {
2000	case IFT_ETHER:		/* RFC2464 */
2001	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2002	case IFT_L2VLAN:	/* ditto */
2003	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2004	case IFT_INFINIBAND:
2005		return (64);
2006	case IFT_PPP:		/* RFC2472 */
2007		return (64);
2008	case IFT_FRELAY:	/* RFC2590 */
2009		return (64);
2010	case IFT_IEEE1394:	/* RFC3146 */
2011		return (64);
2012	case IFT_GIF:
2013		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2014	case IFT_LOOP:
2015		return (64);	/* XXX: is this really correct? */
2016	default:
2017		/*
2018		 * Unknown link type:
2019		 * It might be controversial to use the today's common constant
2020		 * of 64 for these cases unconditionally.  For full compliance,
2021		 * we should return an error in this case.  On the other hand,
2022		 * if we simply miss the standard for the link type or a new
2023		 * standard is defined for a new link type, the IFID length
2024		 * is very likely to be the common constant.  As a compromise,
2025		 * we always use the constant, but make an explicit notice
2026		 * indicating the "unknown" case.
2027		 */
2028		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2029		return (64);
2030	}
2031}
2032
2033#include <sys/sysctl.h>
2034
2035struct in6_llentry {
2036	struct llentry		base;
2037};
2038
2039#define	IN6_LLTBL_DEFAULT_HSIZE	32
2040#define	IN6_LLTBL_HASH(k, h) \
2041	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2042
2043/*
2044 * Do actual deallocation of @lle.
2045 */
2046static void
2047in6_lltable_destroy_lle_unlocked(struct llentry *lle)
2048{
2049
2050	LLE_LOCK_DESTROY(lle);
2051	LLE_REQ_DESTROY(lle);
2052	free(lle, M_LLTABLE);
2053}
2054
2055/*
2056 * Called by LLE_FREE_LOCKED when number of references
2057 * drops to zero.
2058 */
2059static void
2060in6_lltable_destroy_lle(struct llentry *lle)
2061{
2062
2063	LLE_WUNLOCK(lle);
2064	in6_lltable_destroy_lle_unlocked(lle);
2065}
2066
2067static struct llentry *
2068in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2069{
2070	struct in6_llentry *lle;
2071
2072	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2073	if (lle == NULL)		/* NB: caller generates msg */
2074		return NULL;
2075
2076	lle->base.r_l3addr.addr6 = *addr6;
2077	lle->base.lle_refcnt = 1;
2078	lle->base.lle_free = in6_lltable_destroy_lle;
2079	LLE_LOCK_INIT(&lle->base);
2080	LLE_REQ_INIT(&lle->base);
2081	callout_init(&lle->base.lle_timer, 1);
2082
2083	return (&lle->base);
2084}
2085
2086static int
2087in6_lltable_match_prefix(const struct sockaddr *saddr,
2088    const struct sockaddr *smask, u_int flags, struct llentry *lle)
2089{
2090	const struct in6_addr *addr, *mask, *lle_addr;
2091
2092	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2093	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2094	lle_addr = &lle->r_l3addr.addr6;
2095
2096	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2097		return (0);
2098
2099	if (lle->la_flags & LLE_IFADDR) {
2100
2101		/*
2102		 * Delete LLE_IFADDR records IFF address & flag matches.
2103		 * Note that addr is the interface address within prefix
2104		 * being matched.
2105		 */
2106		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2107		    (flags & LLE_STATIC) != 0)
2108			return (1);
2109		return (0);
2110	}
2111
2112	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2113	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2114		return (1);
2115
2116	return (0);
2117}
2118
2119static void
2120in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2121{
2122	struct ifnet *ifp;
2123
2124	LLE_WLOCK_ASSERT(lle);
2125	KASSERT(llt != NULL, ("lltable is NULL"));
2126
2127	/* Unlink entry from table */
2128	if ((lle->la_flags & LLE_LINKED) != 0) {
2129
2130		ifp = llt->llt_ifp;
2131		IF_AFDATA_WLOCK_ASSERT(ifp);
2132		lltable_unlink_entry(llt, lle);
2133	}
2134
2135	if (callout_stop(&lle->lle_timer) > 0)
2136		LLE_REMREF(lle);
2137
2138	llentry_free(lle);
2139}
2140
2141static int
2142in6_lltable_rtcheck(struct ifnet *ifp,
2143		    u_int flags,
2144		    const struct sockaddr *l3addr)
2145{
2146	const struct sockaddr_in6 *sin6;
2147	struct nhop6_basic nh6;
2148	struct in6_addr dst;
2149	uint32_t scopeid;
2150	int error;
2151	char ip6buf[INET6_ADDRSTRLEN];
2152	int fibnum;
2153
2154	KASSERT(l3addr->sa_family == AF_INET6,
2155	    ("sin_family %d", l3addr->sa_family));
2156
2157	sin6 = (const struct sockaddr_in6 *)l3addr;
2158	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2159	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2160	error = fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6);
2161	if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.nh_ifp != ifp) {
2162		struct ifaddr *ifa;
2163		/*
2164		 * Create an ND6 cache for an IPv6 neighbor
2165		 * that is not covered by our own prefix.
2166		 */
2167		ifa = ifaof_ifpforaddr(l3addr, ifp);
2168		if (ifa != NULL) {
2169			ifa_free(ifa);
2170			return 0;
2171		}
2172		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2173		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2174		return EINVAL;
2175	}
2176	return 0;
2177}
2178
2179/*
2180 * Called by the datapath to indicate that the entry was used.
2181 */
2182static void
2183in6_lltable_mark_used(struct llentry *lle)
2184{
2185
2186	LLE_REQ_LOCK(lle);
2187	lle->r_skip_req = 0;
2188
2189	/*
2190	 * Set the hit time so the callback function
2191	 * can determine the remaining time before
2192	 * transiting to the DELAY state.
2193	 */
2194	lle->lle_hittime = time_uptime;
2195	LLE_REQ_UNLOCK(lle);
2196}
2197
2198static inline uint32_t
2199in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2200{
2201
2202	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2203}
2204
2205static uint32_t
2206in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2207{
2208
2209	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2210}
2211
2212static void
2213in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2214{
2215	struct sockaddr_in6 *sin6;
2216
2217	sin6 = (struct sockaddr_in6 *)sa;
2218	bzero(sin6, sizeof(*sin6));
2219	sin6->sin6_family = AF_INET6;
2220	sin6->sin6_len = sizeof(*sin6);
2221	sin6->sin6_addr = lle->r_l3addr.addr6;
2222}
2223
2224static inline struct llentry *
2225in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2226{
2227	struct llentry *lle;
2228	struct llentries *lleh;
2229	u_int hashidx;
2230
2231	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2232	lleh = &llt->lle_head[hashidx];
2233	LIST_FOREACH(lle, lleh, lle_next) {
2234		if (lle->la_flags & LLE_DELETED)
2235			continue;
2236		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2237			break;
2238	}
2239
2240	return (lle);
2241}
2242
2243static void
2244in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2245{
2246
2247	lle->la_flags |= LLE_DELETED;
2248	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2249#ifdef DIAGNOSTIC
2250	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2251#endif
2252	llentry_free(lle);
2253}
2254
2255static struct llentry *
2256in6_lltable_alloc(struct lltable *llt, u_int flags,
2257	const struct sockaddr *l3addr)
2258{
2259	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2260	struct ifnet *ifp = llt->llt_ifp;
2261	struct llentry *lle;
2262	char linkhdr[LLE_MAX_LINKHDR];
2263	size_t linkhdrsize;
2264	int lladdr_off;
2265
2266	KASSERT(l3addr->sa_family == AF_INET6,
2267	    ("sin_family %d", l3addr->sa_family));
2268
2269	/*
2270	 * A route that covers the given address must have
2271	 * been installed 1st because we are doing a resolution,
2272	 * verify this.
2273	 */
2274	if (!(flags & LLE_IFADDR) &&
2275	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2276		return (NULL);
2277
2278	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2279	if (lle == NULL) {
2280		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2281		return (NULL);
2282	}
2283	lle->la_flags = flags;
2284	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2285		linkhdrsize = LLE_MAX_LINKHDR;
2286		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2287		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2288			in6_lltable_destroy_lle_unlocked(lle);
2289			return (NULL);
2290		}
2291		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2292		    lladdr_off);
2293		lle->la_flags |= LLE_STATIC;
2294	}
2295
2296	if ((lle->la_flags & LLE_STATIC) != 0)
2297		lle->ln_state = ND6_LLINFO_REACHABLE;
2298
2299	return (lle);
2300}
2301
2302static struct llentry *
2303in6_lltable_lookup(struct lltable *llt, u_int flags,
2304	const struct sockaddr *l3addr)
2305{
2306	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2307	struct llentry *lle;
2308
2309	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2310	KASSERT(l3addr->sa_family == AF_INET6,
2311	    ("sin_family %d", l3addr->sa_family));
2312
2313	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2314
2315	if (lle == NULL)
2316		return (NULL);
2317
2318	KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) !=
2319	    (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X",
2320	    flags));
2321
2322	if (flags & LLE_UNLOCKED)
2323		return (lle);
2324
2325	if (flags & LLE_EXCLUSIVE)
2326		LLE_WLOCK(lle);
2327	else
2328		LLE_RLOCK(lle);
2329	return (lle);
2330}
2331
2332static int
2333in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2334    struct sysctl_req *wr)
2335{
2336	struct ifnet *ifp = llt->llt_ifp;
2337	/* XXX stack use */
2338	struct {
2339		struct rt_msghdr	rtm;
2340		struct sockaddr_in6	sin6;
2341		/*
2342		 * ndp.c assumes that sdl is word aligned
2343		 */
2344#ifdef __LP64__
2345		uint32_t		pad;
2346#endif
2347		struct sockaddr_dl	sdl;
2348	} ndpc;
2349	struct sockaddr_dl *sdl;
2350	int error;
2351
2352	bzero(&ndpc, sizeof(ndpc));
2353	/* skip deleted entries */
2354	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2355		return (0);
2356	/* Skip if jailed and not a valid IP of the prison. */
2357	lltable_fill_sa_entry(lle,
2358	    (struct sockaddr *)&ndpc.sin6);
2359	if (prison_if(wr->td->td_ucred,
2360	    (struct sockaddr *)&ndpc.sin6) != 0)
2361		return (0);
2362	/*
2363	 * produce a msg made of:
2364	 *  struct rt_msghdr;
2365	 *  struct sockaddr_in6 (IPv6)
2366	 *  struct sockaddr_dl;
2367	 */
2368	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2369	ndpc.rtm.rtm_version = RTM_VERSION;
2370	ndpc.rtm.rtm_type = RTM_GET;
2371	ndpc.rtm.rtm_flags = RTF_UP;
2372	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2373	if (V_deembed_scopeid)
2374		sa6_recoverscope(&ndpc.sin6);
2375
2376	/* publish */
2377	if (lle->la_flags & LLE_PUB)
2378		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2379
2380	sdl = &ndpc.sdl;
2381	sdl->sdl_family = AF_LINK;
2382	sdl->sdl_len = sizeof(*sdl);
2383	sdl->sdl_index = ifp->if_index;
2384	sdl->sdl_type = ifp->if_type;
2385	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2386		sdl->sdl_alen = ifp->if_addrlen;
2387		bcopy(lle->ll_addr, LLADDR(sdl),
2388		    ifp->if_addrlen);
2389	} else {
2390		sdl->sdl_alen = 0;
2391		bzero(LLADDR(sdl), ifp->if_addrlen);
2392	}
2393	if (lle->la_expire != 0)
2394		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2395		    lle->lle_remtime / hz +
2396		    time_second - time_uptime;
2397	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2398	if (lle->la_flags & LLE_STATIC)
2399		ndpc.rtm.rtm_flags |= RTF_STATIC;
2400	if (lle->la_flags & LLE_IFADDR)
2401		ndpc.rtm.rtm_flags |= RTF_PINNED;
2402	if (lle->ln_router != 0)
2403		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2404	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2405	/* Store state in rmx_weight value */
2406	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2407	ndpc.rtm.rtm_index = ifp->if_index;
2408	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2409
2410	return (error);
2411}
2412
2413static struct lltable *
2414in6_lltattach(struct ifnet *ifp)
2415{
2416	struct lltable *llt;
2417
2418	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2419	llt->llt_af = AF_INET6;
2420	llt->llt_ifp = ifp;
2421
2422	llt->llt_lookup = in6_lltable_lookup;
2423	llt->llt_alloc_entry = in6_lltable_alloc;
2424	llt->llt_delete_entry = in6_lltable_delete_entry;
2425	llt->llt_dump_entry = in6_lltable_dump_entry;
2426	llt->llt_hash = in6_lltable_hash;
2427	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2428	llt->llt_free_entry = in6_lltable_free_entry;
2429	llt->llt_match_prefix = in6_lltable_match_prefix;
2430	llt->llt_mark_used = in6_lltable_mark_used;
2431 	lltable_link(llt);
2432
2433	return (llt);
2434}
2435
2436void *
2437in6_domifattach(struct ifnet *ifp)
2438{
2439	struct in6_ifextra *ext;
2440
2441	/* There are not IPv6-capable interfaces. */
2442	switch (ifp->if_type) {
2443	case IFT_PFLOG:
2444	case IFT_PFSYNC:
2445	case IFT_USB:
2446		return (NULL);
2447	}
2448	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2449	bzero(ext, sizeof(*ext));
2450
2451	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2452	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2453	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2454	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2455
2456	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2457	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2458	    M_WAITOK);
2459	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2460	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2461
2462	ext->nd_ifinfo = nd6_ifattach(ifp);
2463	ext->scope6_id = scope6_ifattach(ifp);
2464	ext->lltable = in6_lltattach(ifp);
2465
2466	ext->mld_ifinfo = mld_domifattach(ifp);
2467
2468	return ext;
2469}
2470
2471int
2472in6_domifmtu(struct ifnet *ifp)
2473{
2474	if (ifp->if_afdata[AF_INET6] == NULL)
2475		return ifp->if_mtu;
2476
2477	return (IN6_LINKMTU(ifp));
2478}
2479
2480void
2481in6_domifdetach(struct ifnet *ifp, void *aux)
2482{
2483	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2484
2485	mld_domifdetach(ifp);
2486	scope6_ifdetach(ext->scope6_id);
2487	nd6_ifdetach(ifp, ext->nd_ifinfo);
2488	lltable_free(ext->lltable);
2489	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2490	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2491	free(ext->in6_ifstat, M_IFADDR);
2492	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2493	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2494	free(ext->icmp6_ifstat, M_IFADDR);
2495	free(ext, M_IFADDR);
2496}
2497
2498/*
2499 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2500 * v4 mapped addr or v4 compat addr
2501 */
2502void
2503in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2504{
2505
2506	bzero(sin, sizeof(*sin));
2507	sin->sin_len = sizeof(struct sockaddr_in);
2508	sin->sin_family = AF_INET;
2509	sin->sin_port = sin6->sin6_port;
2510	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2511}
2512
2513/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2514void
2515in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2516{
2517	bzero(sin6, sizeof(*sin6));
2518	sin6->sin6_len = sizeof(struct sockaddr_in6);
2519	sin6->sin6_family = AF_INET6;
2520	sin6->sin6_port = sin->sin_port;
2521	sin6->sin6_addr.s6_addr32[0] = 0;
2522	sin6->sin6_addr.s6_addr32[1] = 0;
2523	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2524	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2525}
2526
2527/* Convert sockaddr_in6 into sockaddr_in. */
2528void
2529in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2530{
2531	struct sockaddr_in *sin_p;
2532	struct sockaddr_in6 sin6;
2533
2534	/*
2535	 * Save original sockaddr_in6 addr and convert it
2536	 * to sockaddr_in.
2537	 */
2538	sin6 = *(struct sockaddr_in6 *)nam;
2539	sin_p = (struct sockaddr_in *)nam;
2540	in6_sin6_2_sin(sin_p, &sin6);
2541}
2542
2543/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2544void
2545in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2546{
2547	struct sockaddr_in *sin_p;
2548	struct sockaddr_in6 *sin6_p;
2549
2550	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2551	sin_p = (struct sockaddr_in *)*nam;
2552	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2553	free(*nam, M_SONAME);
2554	*nam = (struct sockaddr *)sin6_p;
2555}
2556