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