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
63#include <sys/cdefs.h>
64#include "opt_inet.h"
65#include "opt_inet6.h"
66
67#include <sys/param.h>
68#include <sys/eventhandler.h>
69#include <sys/errno.h>
70#include <sys/jail.h>
71#include <sys/malloc.h>
72#include <sys/socket.h>
73#include <sys/socketvar.h>
74#include <sys/sockio.h>
75#include <sys/systm.h>
76#include <sys/priv.h>
77#include <sys/proc.h>
78#include <sys/time.h>
79#include <sys/kernel.h>
80#include <sys/lock.h>
81#include <sys/rmlock.h>
82#include <sys/sysctl.h>
83#include <sys/syslog.h>
84
85#include <net/if.h>
86#include <net/if_var.h>
87#include <net/if_private.h>
88#include <net/if_types.h>
89#include <net/route.h>
90#include <net/route/route_ctl.h>
91#include <net/route/nhop.h>
92#include <net/if_dl.h>
93#include <net/vnet.h>
94
95#include <netinet/in.h>
96#include <netinet/in_var.h>
97#include <net/if_llatbl.h>
98#include <netinet/if_ether.h>
99#include <netinet/in_systm.h>
100#include <netinet/ip.h>
101#include <netinet/in_pcb.h>
102#include <netinet/ip_carp.h>
103#include <netinet/icmp6.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#ifdef MAC
116#include <security/mac/mac_framework.h>
117#endif
118
119/*
120 * struct in6_ifreq and struct ifreq must be type punnable for common members
121 * of ifr_ifru to allow accessors to be shared.
122 */
123_Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
124    offsetof(struct ifreq, ifr_ifru),
125    "struct in6_ifreq and struct ifreq are not type punnable");
126
127VNET_DEFINE_STATIC(int, icmp6_nodeinfo_oldmcprefix) = 1;
128#define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
129SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_NODEINFO_OLDMCPREFIX,
130    nodeinfo_oldmcprefix, CTLFLAG_VNET | CTLFLAG_RW,
131    &VNET_NAME(icmp6_nodeinfo_oldmcprefix), 0,
132    "Join old IPv6 NI group address in draft-ietf-ipngwg-icmp-name-lookup "
133    "for compatibility with KAME implementation");
134
135VNET_DEFINE_STATIC(int, nd6_useloopback) = 1;
136#define	V_nd6_useloopback	VNET(nd6_useloopback)
137SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_USELOOPBACK, nd6_useloopback,
138    CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_useloopback), 0,
139    "Create a loopback route when configuring an IPv6 address");
140
141/*
142 * Definitions of some costant IP6 addresses.
143 */
144const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
145const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
146const struct in6_addr in6addr_nodelocal_allnodes =
147	IN6ADDR_NODELOCAL_ALLNODES_INIT;
148const struct in6_addr in6addr_linklocal_allnodes =
149	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
150const struct in6_addr in6addr_linklocal_allrouters =
151	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
152const struct in6_addr in6addr_linklocal_allv2routers =
153	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
154
155const struct in6_addr in6mask0 = IN6MASK0;
156const struct in6_addr in6mask32 = IN6MASK32;
157const struct in6_addr in6mask64 = IN6MASK64;
158const struct in6_addr in6mask96 = IN6MASK96;
159const struct in6_addr in6mask128 = IN6MASK128;
160
161const struct sockaddr_in6 sa6_any =
162	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
163
164static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
165	struct in6_aliasreq *, int);
166static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
167
168static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
169    struct in6_ifaddr *, int);
170static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
171    struct in6_aliasreq *, int flags);
172static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
173    struct in6_ifaddr *, int, int);
174static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
175    struct in6_ifaddr *, int);
176
177static void in6_join_proxy_ndp_mc(struct ifnet *, const struct in6_addr *);
178static void in6_leave_proxy_ndp_mc(struct ifnet *, const struct in6_addr *);
179
180#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
181#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
182
183static struct sx in6_control_sx;
184SX_SYSINIT(in6_control_sx, &in6_control_sx, "in6_control");
185
186void
187in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
188{
189	struct rt_addrinfo info;
190	struct ifaddr *ifa;
191	struct sockaddr_dl gateway;
192	int fibnum;
193
194	ifa = &ia->ia_ifa;
195
196	/*
197	 * Prepare info data for the host route.
198	 * This code mimics one from ifa_maintain_loopback_route().
199	 */
200	bzero(&info, sizeof(struct rt_addrinfo));
201	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
202	info.rti_info[RTAX_DST] = ifa->ifa_addr;
203	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
204	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
205	if (cmd != RTM_DELETE)
206		info.rti_ifp = V_loif;
207
208	fibnum = ia62ifa(ia)->ifa_ifp->if_fib;
209
210	if (cmd == RTM_ADD) {
211		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
212		rt_routemsg_info(cmd, &info, fibnum);
213	} else if (cmd == RTM_DELETE) {
214		rt_routemsg_info(cmd, &info, fibnum);
215		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
216	}
217}
218
219int
220in6_mask2len(struct in6_addr *mask, u_char *lim0)
221{
222	int x = 0, y;
223	u_char *lim = lim0, *p;
224
225	/* ignore the scope_id part */
226	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
227		lim = (u_char *)mask + sizeof(*mask);
228	for (p = (u_char *)mask; p < lim; x++, p++) {
229		if (*p != 0xff)
230			break;
231	}
232	y = 0;
233	if (p < lim) {
234		for (y = 0; y < 8; y++) {
235			if ((*p & (0x80 >> y)) == 0)
236				break;
237		}
238	}
239
240	/*
241	 * when the limit pointer is given, do a stricter check on the
242	 * remaining bits.
243	 */
244	if (p < lim) {
245		if (y != 0 && (*p & (0x00ff >> y)) != 0)
246			return (-1);
247		for (p = p + 1; p < lim; p++)
248			if (*p != 0)
249				return (-1);
250	}
251
252	return x * 8 + y;
253}
254
255#ifdef COMPAT_FREEBSD32
256struct in6_ndifreq32 {
257	char ifname[IFNAMSIZ];
258	uint32_t ifindex;
259};
260#define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
261#endif
262
263int
264in6_control_ioctl(u_long cmd, void *data,
265    struct ifnet *ifp, struct ucred *cred)
266{
267	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
268	struct	in6_ifaddr *ia = NULL;
269	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
270	struct sockaddr_in6 *sa6;
271	int error;
272	bool control_locked = false;
273
274	/*
275	 * Compat to make pre-10.x ifconfig(8) operable.
276	 */
277	if (cmd == OSIOCAIFADDR_IN6) {
278		cmd = SIOCAIFADDR_IN6;
279		ifra->ifra_vhid = 0;
280	}
281
282	switch (cmd) {
283	case SIOCGETSGCNT_IN6:
284	case SIOCGETMIFCNT_IN6:
285		/*
286		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
287		 * We cannot see how that would be needed, so do not adjust the
288		 * KPI blindly; more likely should clean up the IPv4 variant.
289		 */
290		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
291	}
292
293	switch (cmd) {
294	case SIOCAADDRCTL_POLICY:
295	case SIOCDADDRCTL_POLICY:
296		if (cred != NULL) {
297			error = priv_check_cred(cred, PRIV_NETINET_ADDRCTRL6);
298			if (error)
299				return (error);
300		}
301		return (in6_src_ioctl(cmd, data));
302	}
303
304	if (ifp == NULL)
305		return (EOPNOTSUPP);
306
307	switch (cmd) {
308	case SIOCSNDFLUSH_IN6:
309	case SIOCSPFXFLUSH_IN6:
310	case SIOCSRTRFLUSH_IN6:
311	case SIOCSDEFIFACE_IN6:
312	case SIOCSIFINFO_FLAGS:
313	case SIOCSIFINFO_IN6:
314		if (cred != NULL) {
315			error = priv_check_cred(cred, PRIV_NETINET_ND6);
316			if (error)
317				return (error);
318		}
319		/* FALLTHROUGH */
320	case OSIOCGIFINFO_IN6:
321	case SIOCGIFINFO_IN6:
322	case SIOCGNBRINFO_IN6:
323	case SIOCGDEFIFACE_IN6:
324		return (nd6_ioctl(cmd, data, ifp));
325
326#ifdef COMPAT_FREEBSD32
327	case SIOCGDEFIFACE32_IN6:
328		{
329			struct in6_ndifreq ndif;
330			struct in6_ndifreq32 *ndif32;
331
332			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
333			    ifp);
334			if (error)
335				return (error);
336			ndif32 = (struct in6_ndifreq32 *)data;
337			ndif32->ifindex = ndif.ifindex;
338			return (0);
339		}
340#endif
341	}
342
343	switch (cmd) {
344	case SIOCSIFPREFIX_IN6:
345	case SIOCDIFPREFIX_IN6:
346	case SIOCAIFPREFIX_IN6:
347	case SIOCCIFPREFIX_IN6:
348	case SIOCSGIFPREFIX_IN6:
349	case SIOCGIFPREFIX_IN6:
350		log(LOG_NOTICE,
351		    "prefix ioctls are now invalidated. "
352		    "please use ifconfig.\n");
353		return (EOPNOTSUPP);
354	}
355
356	switch (cmd) {
357	case SIOCSSCOPE6:
358		if (cred != NULL) {
359			error = priv_check_cred(cred, PRIV_NETINET_SCOPE6);
360			if (error)
361				return (error);
362		}
363		/* FALLTHROUGH */
364	case SIOCGSCOPE6:
365	case SIOCGSCOPE6DEF:
366		return (scope6_ioctl(cmd, data, ifp));
367	}
368
369	/*
370	 * Find address for this interface, if it exists.
371	 *
372	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
373	 * only, and used the first interface address as the target of other
374	 * operations (without checking ifra_addr).  This was because netinet
375	 * code/API assumed at most 1 interface address per interface.
376	 * Since IPv6 allows a node to assign multiple addresses
377	 * on a single interface, we almost always look and check the
378	 * presence of ifra_addr, and reject invalid ones here.
379	 * It also decreases duplicated code among SIOC*_IN6 operations.
380	 */
381	switch (cmd) {
382	case SIOCAIFADDR_IN6:
383	case SIOCSIFPHYADDR_IN6:
384		sa6 = &ifra->ifra_addr;
385		break;
386	case SIOCSIFADDR_IN6:
387	case SIOCGIFADDR_IN6:
388	case SIOCSIFDSTADDR_IN6:
389	case SIOCSIFNETMASK_IN6:
390	case SIOCGIFDSTADDR_IN6:
391	case SIOCGIFNETMASK_IN6:
392	case SIOCDIFADDR_IN6:
393	case SIOCGIFPSRCADDR_IN6:
394	case SIOCGIFPDSTADDR_IN6:
395	case SIOCGIFAFLAG_IN6:
396	case SIOCSNDFLUSH_IN6:
397	case SIOCSPFXFLUSH_IN6:
398	case SIOCSRTRFLUSH_IN6:
399	case SIOCGIFALIFETIME_IN6:
400	case SIOCGIFSTAT_IN6:
401	case SIOCGIFSTAT_ICMP6:
402		sa6 = &ifr->ifr_addr;
403		break;
404	case SIOCSIFADDR:
405	case SIOCSIFBRDADDR:
406	case SIOCSIFDSTADDR:
407	case SIOCSIFNETMASK:
408		/*
409		 * Although we should pass any non-INET6 ioctl requests
410		 * down to driver, we filter some legacy INET requests.
411		 * Drivers trust SIOCSIFADDR et al to come from an already
412		 * privileged layer, and do not perform any credentials
413		 * checks or input validation.
414		 */
415		return (EINVAL);
416	default:
417		sa6 = NULL;
418		break;
419	}
420	if (sa6 && sa6->sin6_family == AF_INET6) {
421		if (sa6->sin6_scope_id != 0)
422			error = sa6_embedscope(sa6, 0);
423		else
424			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
425		if (error != 0)
426			return (error);
427		if (cred != NULL && (error = prison_check_ip6(cred,
428		    &sa6->sin6_addr)) != 0)
429			return (error);
430		sx_xlock(&in6_control_sx);
431		control_locked = true;
432		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
433	} else
434		ia = NULL;
435
436	switch (cmd) {
437	case SIOCSIFADDR_IN6:
438	case SIOCSIFDSTADDR_IN6:
439	case SIOCSIFNETMASK_IN6:
440		/*
441		 * Since IPv6 allows a node to assign multiple addresses
442		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
443		 */
444		/* we decided to obsolete this command (20000704) */
445		error = EINVAL;
446		goto out;
447
448	case SIOCDIFADDR_IN6:
449		/*
450		 * for IPv4, we look for existing in_ifaddr here to allow
451		 * "ifconfig if0 delete" to remove the first IPv4 address on
452		 * the interface.  For IPv6, as the spec allows multiple
453		 * interface address from the day one, we consider "remove the
454		 * first one" semantics to be not preferable.
455		 */
456		if (ia == NULL) {
457			error = EADDRNOTAVAIL;
458			goto out;
459		}
460		/* FALLTHROUGH */
461	case SIOCAIFADDR_IN6:
462		/*
463		 * We always require users to specify a valid IPv6 address for
464		 * the corresponding operation.
465		 */
466		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
467		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
468			error = EAFNOSUPPORT;
469			goto out;
470		}
471
472		if (cred != NULL) {
473			error = priv_check_cred(cred, (cmd == SIOCDIFADDR_IN6) ?
474			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
475			if (error)
476				goto out;
477		}
478		/* FALLTHROUGH */
479	case SIOCGIFSTAT_IN6:
480	case SIOCGIFSTAT_ICMP6:
481		if (ifp->if_afdata[AF_INET6] == NULL) {
482			error = EPFNOSUPPORT;
483			goto out;
484		}
485		break;
486
487	case SIOCGIFADDR_IN6:
488		/* This interface is basically deprecated. use SIOCGIFCONF. */
489		/* FALLTHROUGH */
490	case SIOCGIFAFLAG_IN6:
491	case SIOCGIFNETMASK_IN6:
492	case SIOCGIFDSTADDR_IN6:
493	case SIOCGIFALIFETIME_IN6:
494		/* must think again about its semantics */
495		if (ia == NULL) {
496			error = EADDRNOTAVAIL;
497			goto out;
498		}
499		break;
500	}
501
502	switch (cmd) {
503	case SIOCGIFADDR_IN6:
504		ifr->ifr_addr = ia->ia_addr;
505		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
506			goto out;
507		break;
508
509	case SIOCGIFDSTADDR_IN6:
510		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
511			error = EINVAL;
512			goto out;
513		}
514		ifr->ifr_dstaddr = ia->ia_dstaddr;
515		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
516			goto out;
517		break;
518
519	case SIOCGIFNETMASK_IN6:
520		ifr->ifr_addr = ia->ia_prefixmask;
521		break;
522
523	case SIOCGIFAFLAG_IN6:
524		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
525		break;
526
527	case SIOCGIFSTAT_IN6:
528		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
529		    ifp->if_afdata[AF_INET6])->in6_ifstat,
530		    &ifr->ifr_ifru.ifru_stat,
531		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
532		break;
533
534	case SIOCGIFSTAT_ICMP6:
535		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
536		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
537		    &ifr->ifr_ifru.ifru_icmp6stat,
538		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
539		break;
540
541	case SIOCGIFALIFETIME_IN6:
542		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
543		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
544			time_t maxexpire;
545			struct in6_addrlifetime *retlt =
546			    &ifr->ifr_ifru.ifru_lifetime;
547
548			/*
549			 * XXX: adjust expiration time assuming time_t is
550			 * signed.
551			 */
552			maxexpire = (-1) &
553			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
554			if (ia->ia6_lifetime.ia6t_vltime <
555			    maxexpire - ia->ia6_updatetime) {
556				retlt->ia6t_expire = ia->ia6_updatetime +
557				    ia->ia6_lifetime.ia6t_vltime;
558			} else
559				retlt->ia6t_expire = maxexpire;
560		}
561		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
562			time_t maxexpire;
563			struct in6_addrlifetime *retlt =
564			    &ifr->ifr_ifru.ifru_lifetime;
565
566			/*
567			 * XXX: adjust expiration time assuming time_t is
568			 * signed.
569			 */
570			maxexpire = (-1) &
571			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
572			if (ia->ia6_lifetime.ia6t_pltime <
573			    maxexpire - ia->ia6_updatetime) {
574				retlt->ia6t_preferred = ia->ia6_updatetime +
575				    ia->ia6_lifetime.ia6t_pltime;
576			} else
577				retlt->ia6t_preferred = maxexpire;
578		}
579		break;
580
581	case SIOCAIFADDR_IN6:
582#ifdef MAC
583		/* Check if a MAC policy disallows setting the IPv6 address. */
584		error = mac_inet6_check_add_addr(cred, &sa6->sin6_addr, ifp);
585		if (error != 0)
586			goto out;
587#endif
588		error = in6_addifaddr(ifp, ifra, ia);
589		ia = NULL;
590		break;
591
592	case SIOCDIFADDR_IN6:
593		in6_purgeifaddr(ia);
594		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
595		    IFADDR_EVENT_DEL);
596		break;
597
598	default:
599		if (ifp->if_ioctl == NULL) {
600			error = EOPNOTSUPP;
601			goto out;
602		}
603		error = (*ifp->if_ioctl)(ifp, cmd, data);
604		goto out;
605	}
606
607	error = 0;
608out:
609	if (control_locked)
610		sx_xunlock(&in6_control_sx);
611
612	if (ia != NULL)
613		ifa_free(&ia->ia_ifa);
614	return (error);
615}
616
617int
618in6_control(struct socket *so, u_long cmd, void *data,
619    struct ifnet *ifp, struct thread *td)
620{
621	return (in6_control_ioctl(cmd, data, ifp, td ? td->td_ucred : NULL));
622}
623
624static struct in6_multi_mship *
625in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
626    int *errorp, int delay)
627{
628	struct in6_multi_mship *imm;
629	int error;
630
631	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
632	if (imm == NULL) {
633		*errorp = ENOBUFS;
634		return (NULL);
635	}
636
637	delay = (delay * MLD_FASTHZ) / hz;
638
639	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
640	if (error) {
641		*errorp = error;
642		free(imm, M_IP6MADDR);
643		return (NULL);
644	}
645
646	return (imm);
647}
648
649static int
650in6_solicited_node_maddr(struct in6_addr *maddr,
651    struct ifnet *ifp, const struct in6_addr *base)
652{
653	int error;
654
655	bzero(maddr, sizeof(struct in6_addr));
656	maddr->s6_addr32[0] = IPV6_ADDR_INT32_MLL;
657	maddr->s6_addr32[2] = htonl(1);
658	maddr->s6_addr32[3] = base->s6_addr32[3];
659	maddr->s6_addr8[12] = 0xff;
660	if ((error = in6_setscope(maddr, ifp, NULL)) != 0) {
661		/* XXX: should not happen */
662		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
663	}
664
665	return error;
666}
667
668/*
669 * Join necessary multicast groups.  Factored out from in6_update_ifa().
670 * This entire work should only be done once, for the default FIB.
671 */
672static int
673in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
674    struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
675{
676	char ip6buf[INET6_ADDRSTRLEN];
677	struct in6_addr mltaddr;
678	struct in6_multi_mship *imm;
679	int delay, error;
680
681	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
682
683	/* Join solicited multicast addr for new host id. */
684	if ((error = in6_solicited_node_maddr(&mltaddr, ifp,
685	    &ifra->ifra_addr.sin6_addr)) != 0)
686		goto cleanup;
687	delay = error = 0;
688	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
689		/*
690		 * We need a random delay for DAD on the address being
691		 * configured.  It also means delaying transmission of the
692		 * corresponding MLD report to avoid report collision.
693		 * [RFC 4861, Section 6.3.7]
694		 */
695		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
696	}
697	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
698	if (imm == NULL) {
699		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
700		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
701		    if_name(ifp), error));
702		goto cleanup;
703	}
704	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
705	*in6m_sol = imm->i6mm_maddr;
706
707	/*
708	 * Join link-local all-nodes address.
709	 */
710	mltaddr = in6addr_linklocal_allnodes;
711	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
712		goto cleanup; /* XXX: should not fail */
713
714	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
715	if (imm == NULL) {
716		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
717		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
718		    if_name(ifp), error));
719		goto cleanup;
720	}
721	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
722
723	/*
724	 * Join node information group address.
725	 */
726	delay = 0;
727	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
728		/*
729		 * The spec does not say anything about delay for this group,
730		 * but the same logic should apply.
731		 */
732		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
733	}
734	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
735		/* XXX jinmei */
736		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
737		if (imm == NULL)
738			nd6log((LOG_WARNING,
739			    "%s: in6_joingroup failed for %s on %s "
740			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
741			    &mltaddr), if_name(ifp), error));
742			/* XXX not very fatal, go on... */
743		else
744			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
745	}
746	if (V_icmp6_nodeinfo_oldmcprefix &&
747	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
748		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
749		if (imm == NULL)
750			nd6log((LOG_WARNING,
751			    "%s: in6_joingroup failed for %s on %s "
752			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
753			    &mltaddr), if_name(ifp), error));
754			/* XXX not very fatal, go on... */
755		else
756			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
757	}
758
759	/*
760	 * Join interface-local all-nodes address.
761	 * (ff01::1%ifN, and ff01::%ifN/32)
762	 */
763	mltaddr = in6addr_nodelocal_allnodes;
764	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
765		goto cleanup; /* XXX: should not fail */
766
767	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
768	if (imm == NULL) {
769		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
770		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
771		    &mltaddr), if_name(ifp), error));
772		goto cleanup;
773	}
774	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
775
776cleanup:
777	return (error);
778}
779
780/*
781 * Update parameters of an IPv6 interface address.
782 * If necessary, a new entry is created and linked into address chains.
783 * This function is separated from in6_control().
784 */
785int
786in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
787    struct in6_ifaddr *ia, int flags)
788{
789	int error, hostIsNew = 0;
790
791	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
792		return (error);
793
794	if (ia == NULL) {
795		hostIsNew = 1;
796		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
797			return (ENOBUFS);
798	}
799
800	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
801	if (error != 0) {
802		if (hostIsNew != 0) {
803			in6_unlink_ifa(ia, ifp);
804			ifa_free(&ia->ia_ifa);
805		}
806		return (error);
807	}
808
809	if (hostIsNew)
810		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
811
812	return (error);
813}
814
815/*
816 * Fill in basic IPv6 address request info.
817 */
818void
819in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
820    const struct in6_addr *mask)
821{
822
823	memset(ifra, 0, sizeof(struct in6_aliasreq));
824
825	ifra->ifra_addr.sin6_family = AF_INET6;
826	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
827	if (addr != NULL)
828		ifra->ifra_addr.sin6_addr = *addr;
829
830	ifra->ifra_prefixmask.sin6_family = AF_INET6;
831	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
832	if (mask != NULL)
833		ifra->ifra_prefixmask.sin6_addr = *mask;
834}
835
836static int
837in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
838    struct in6_ifaddr *ia, int flags)
839{
840	int plen = -1;
841	struct sockaddr_in6 dst6;
842	struct in6_addrlifetime *lt;
843	char ip6buf[INET6_ADDRSTRLEN];
844
845	/* Validate parameters */
846	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
847		return (EINVAL);
848
849	/*
850	 * The destination address for a p2p link must have a family
851	 * of AF_UNSPEC or AF_INET6.
852	 */
853	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
854	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
855	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
856		return (EAFNOSUPPORT);
857
858	/*
859	 * Validate address
860	 */
861	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
862	    ifra->ifra_addr.sin6_family != AF_INET6)
863		return (EINVAL);
864
865	/*
866	 * validate ifra_prefixmask.  don't check sin6_family, netmask
867	 * does not carry fields other than sin6_len.
868	 */
869	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
870		return (EINVAL);
871	/*
872	 * Because the IPv6 address architecture is classless, we require
873	 * users to specify a (non 0) prefix length (mask) for a new address.
874	 * We also require the prefix (when specified) mask is valid, and thus
875	 * reject a non-consecutive mask.
876	 */
877	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
878		return (EINVAL);
879	if (ifra->ifra_prefixmask.sin6_len != 0) {
880		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
881		    (u_char *)&ifra->ifra_prefixmask +
882		    ifra->ifra_prefixmask.sin6_len);
883		if (plen <= 0)
884			return (EINVAL);
885	} else {
886		/*
887		 * In this case, ia must not be NULL.  We just use its prefix
888		 * length.
889		 */
890		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
891	}
892	/*
893	 * If the destination address on a p2p interface is specified,
894	 * and the address is a scoped one, validate/set the scope
895	 * zone identifier.
896	 */
897	dst6 = ifra->ifra_dstaddr;
898	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
899	    (dst6.sin6_family == AF_INET6)) {
900		struct in6_addr in6_tmp;
901		u_int32_t zoneid;
902
903		in6_tmp = dst6.sin6_addr;
904		if (in6_setscope(&in6_tmp, ifp, &zoneid))
905			return (EINVAL); /* XXX: should be impossible */
906
907		if (dst6.sin6_scope_id != 0) {
908			if (dst6.sin6_scope_id != zoneid)
909				return (EINVAL);
910		} else		/* user omit to specify the ID. */
911			dst6.sin6_scope_id = zoneid;
912
913		/* convert into the internal form */
914		if (sa6_embedscope(&dst6, 0))
915			return (EINVAL); /* XXX: should be impossible */
916	}
917	/* Modify original ifra_dstaddr to reflect changes */
918	ifra->ifra_dstaddr = dst6;
919
920	/*
921	 * The destination address can be specified only for a p2p or a
922	 * loopback interface.  If specified, the corresponding prefix length
923	 * must be 128.
924	 */
925	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
926		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
927			/* XXX: noisy message */
928			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
929			    "be specified for a p2p or a loopback IF only\n"));
930			return (EINVAL);
931		}
932		if (plen != 128) {
933			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
934			    "be 128 when dstaddr is specified\n"));
935			return (EINVAL);
936		}
937	}
938	/* lifetime consistency check */
939	lt = &ifra->ifra_lifetime;
940	if (lt->ia6t_pltime > lt->ia6t_vltime)
941		return (EINVAL);
942	if (lt->ia6t_vltime == 0) {
943		/*
944		 * the following log might be noisy, but this is a typical
945		 * configuration mistake or a tool's bug.
946		 */
947		nd6log((LOG_INFO,
948		    "in6_update_ifa: valid lifetime is 0 for %s\n",
949		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
950
951		if (ia == NULL)
952			return (0); /* there's nothing to do */
953	}
954
955	/* Check prefix mask */
956	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
957		/*
958		 * We prohibit changing the prefix length of an existing
959		 * address, because
960		 * + such an operation should be rare in IPv6, and
961		 * + the operation would confuse prefix management.
962		 */
963		if (ia->ia_prefixmask.sin6_len != 0 &&
964		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
965			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
966			    "of an existing %s address should not be changed\n",
967			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
968
969			return (EINVAL);
970		}
971	}
972
973	return (0);
974}
975
976/*
977 * Allocate a new ifaddr and link it into chains.
978 */
979static struct in6_ifaddr *
980in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
981{
982	struct in6_ifaddr *ia;
983
984	/*
985	 * When in6_alloc_ifa() is called in a process of a received
986	 * RA, it is called under an interrupt context.  So, we should
987	 * call malloc with M_NOWAIT.
988	 */
989	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
990	if (ia == NULL)
991		return (NULL);
992	LIST_INIT(&ia->ia6_memberships);
993	/* Initialize the address and masks, and put time stamp */
994	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
995	ia->ia_addr.sin6_family = AF_INET6;
996	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
997	/* XXX: Can we assign ,sin6_addr and skip the rest? */
998	ia->ia_addr = ifra->ifra_addr;
999	ia->ia6_createtime = time_uptime;
1000	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1001		/*
1002		 * Some functions expect that ifa_dstaddr is not
1003		 * NULL for p2p interfaces.
1004		 */
1005		ia->ia_ifa.ifa_dstaddr =
1006		    (struct sockaddr *)&ia->ia_dstaddr;
1007	} else {
1008		ia->ia_ifa.ifa_dstaddr = NULL;
1009	}
1010
1011	/* set prefix mask if any */
1012	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1013	if (ifra->ifra_prefixmask.sin6_len != 0) {
1014		ia->ia_prefixmask.sin6_family = AF_INET6;
1015		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1016		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1017	}
1018
1019	ia->ia_ifp = ifp;
1020	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1021	IF_ADDR_WLOCK(ifp);
1022	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1023	IF_ADDR_WUNLOCK(ifp);
1024
1025	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1026	IN6_IFADDR_WLOCK();
1027	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1028	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1029	IN6_IFADDR_WUNLOCK();
1030
1031	return (ia);
1032}
1033
1034/*
1035 * Update/configure interface address parameters:
1036 *
1037 * 1) Update lifetime
1038 * 2) Update interface metric ad flags
1039 * 3) Notify other subsystems
1040 */
1041static int
1042in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1043    struct in6_ifaddr *ia, int hostIsNew, int flags)
1044{
1045	int error;
1046
1047	/* update timestamp */
1048	ia->ia6_updatetime = time_uptime;
1049
1050	/*
1051	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1052	 * to see if the address is deprecated or invalidated, but initialize
1053	 * these members for applications.
1054	 */
1055	ia->ia6_lifetime = ifra->ifra_lifetime;
1056	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1057		ia->ia6_lifetime.ia6t_expire =
1058		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1059	} else
1060		ia->ia6_lifetime.ia6t_expire = 0;
1061	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1062		ia->ia6_lifetime.ia6t_preferred =
1063		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1064	} else
1065		ia->ia6_lifetime.ia6t_preferred = 0;
1066
1067	/*
1068	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1069	 * userland, make it deprecated.
1070	 */
1071	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1072		ia->ia6_lifetime.ia6t_pltime = 0;
1073		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1074	}
1075
1076	/*
1077	 * configure address flags.
1078	 */
1079	ia->ia6_flags = ifra->ifra_flags;
1080
1081	/*
1082	 * Make the address tentative before joining multicast addresses,
1083	 * so that corresponding MLD responses would not have a tentative
1084	 * source address.
1085	 */
1086	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1087
1088	/*
1089	 * DAD should be performed for an new address or addresses on
1090	 * an interface with ND6_IFF_IFDISABLED.
1091	 */
1092	if (in6if_do_dad(ifp) &&
1093	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1094		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1095
1096	/* notify other subsystems */
1097	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1098
1099	return (error);
1100}
1101
1102/*
1103 * Do link-level ifa job:
1104 * 1) Add lle entry for added address
1105 * 2) Notifies routing socket users about new address
1106 * 3) join appropriate multicast group
1107 * 4) start DAD if enabled
1108 */
1109static int
1110in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1111    struct in6_ifaddr *ia, int flags)
1112{
1113	struct in6_multi *in6m_sol;
1114	int error = 0;
1115
1116	/* Add local address to lltable, if necessary (ex. on p2p link). */
1117	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1118		in6_purgeaddr(&ia->ia_ifa);
1119		ifa_free(&ia->ia_ifa);
1120		return (error);
1121	}
1122
1123	/* Join necessary multicast groups. */
1124	in6m_sol = NULL;
1125	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1126		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1127		if (error != 0) {
1128			in6_purgeaddr(&ia->ia_ifa);
1129			ifa_free(&ia->ia_ifa);
1130			return (error);
1131		}
1132	}
1133
1134	/* Perform DAD, if the address is TENTATIVE. */
1135	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1136		int delay, mindelay, maxdelay;
1137
1138		delay = 0;
1139		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1140			/*
1141			 * We need to impose a delay before sending an NS
1142			 * for DAD.  Check if we also needed a delay for the
1143			 * corresponding MLD message.  If we did, the delay
1144			 * should be larger than the MLD delay (this could be
1145			 * relaxed a bit, but this simple logic is at least
1146			 * safe).
1147			 * XXX: Break data hiding guidelines and look at
1148			 * state for the solicited multicast group.
1149			 */
1150			mindelay = 0;
1151			if (in6m_sol != NULL &&
1152			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1153				mindelay = in6m_sol->in6m_timer;
1154			}
1155			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1156			if (maxdelay - mindelay == 0)
1157				delay = 0;
1158			else {
1159				delay =
1160				    (arc4random() % (maxdelay - mindelay)) +
1161				    mindelay;
1162			}
1163		}
1164		nd6_dad_start((struct ifaddr *)ia, delay);
1165	}
1166
1167	in6_newaddrmsg(ia, RTM_ADD);
1168	ifa_free(&ia->ia_ifa);
1169	return (error);
1170}
1171
1172/*
1173 * Adds or deletes interface route for p2p ifa.
1174 * Returns 0 on success or errno.
1175 */
1176static int
1177in6_handle_dstaddr_rtrequest(int cmd, struct in6_ifaddr *ia)
1178{
1179	struct epoch_tracker et;
1180	struct ifaddr *ifa = &ia->ia_ifa;
1181	int error;
1182
1183	/* Prepare gateway */
1184	struct sockaddr_dl_short sdl = {
1185		.sdl_family = AF_LINK,
1186		.sdl_len = sizeof(struct sockaddr_dl_short),
1187		.sdl_type = ifa->ifa_ifp->if_type,
1188		.sdl_index = ifa->ifa_ifp->if_index,
1189	};
1190
1191	struct sockaddr_in6 dst = {
1192		.sin6_family = AF_INET6,
1193		.sin6_len = sizeof(struct sockaddr_in6),
1194		.sin6_addr = ia->ia_dstaddr.sin6_addr,
1195	};
1196
1197	struct rt_addrinfo info = {
1198		.rti_ifa = ifa,
1199		.rti_ifp = ifa->ifa_ifp,
1200		.rti_flags = RTF_PINNED | RTF_HOST,
1201		.rti_info = {
1202			[RTAX_DST] = (struct sockaddr *)&dst,
1203			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
1204		},
1205	};
1206	/* Don't set additional per-gw filters on removal */
1207
1208	NET_EPOCH_ENTER(et);
1209	error = rib_handle_ifaddr_info(ifa->ifa_ifp->if_fib, cmd, &info);
1210	NET_EPOCH_EXIT(et);
1211
1212	return (error);
1213}
1214
1215static bool
1216ifa_is_p2p(struct in6_ifaddr *ia)
1217{
1218	int plen;
1219
1220	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1221
1222	if ((plen == 128) && (ia->ia_dstaddr.sin6_family == AF_INET6) &&
1223	    !IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &ia->ia_dstaddr.sin6_addr))
1224		return (true);
1225
1226	return (false);
1227}
1228
1229int
1230in6_addifaddr(struct ifnet *ifp, struct in6_aliasreq *ifra, struct in6_ifaddr *ia)
1231{
1232	struct nd_prefixctl pr0;
1233	struct nd_prefix *pr;
1234	int carp_attached = 0;
1235	int error;
1236
1237	/*
1238	 * first, make or update the interface address structure,
1239	 * and link it to the list.
1240	 */
1241	if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
1242		goto out;
1243	if (ia != NULL) {
1244		if (ia->ia_ifa.ifa_carp)
1245			(*carp_detach_p)(&ia->ia_ifa, true);
1246		ifa_free(&ia->ia_ifa);
1247	}
1248	if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr)) == NULL) {
1249		/*
1250		 * this can happen when the user specify the 0 valid
1251		 * lifetime.
1252		 */
1253		return (0);
1254	}
1255
1256	if (ifra->ifra_vhid > 0) {
1257		if (carp_attach_p != NULL)
1258			error = (*carp_attach_p)(&ia->ia_ifa,
1259			    ifra->ifra_vhid);
1260		else
1261			error = EPROTONOSUPPORT;
1262		if (error)
1263			goto out;
1264		else
1265			carp_attached = 1;
1266	}
1267
1268	/*
1269	 * then, make the prefix on-link on the interface.
1270	 * XXX: we'd rather create the prefix before the address, but
1271	 * we need at least one address to install the corresponding
1272	 * interface route, so we configure the address first.
1273	 */
1274
1275	/*
1276	 * convert mask to prefix length (prefixmask has already
1277	 * been validated in in6_update_ifa().
1278	 */
1279	bzero(&pr0, sizeof(pr0));
1280	pr0.ndpr_ifp = ifp;
1281	pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1282	    NULL);
1283	if (pr0.ndpr_plen == 128) {
1284		/* we don't need to install a host route. */
1285		goto aifaddr_out;
1286	}
1287	pr0.ndpr_prefix = ifra->ifra_addr;
1288	/* apply the mask for safety. */
1289	IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
1290	    &ifra->ifra_prefixmask.sin6_addr);
1291
1292	/*
1293	 * XXX: since we don't have an API to set prefix (not address)
1294	 * lifetimes, we just use the same lifetimes as addresses.
1295	 * The (temporarily) installed lifetimes can be overridden by
1296	 * later advertised RAs (when accept_rtadv is non 0), which is
1297	 * an intended behavior.
1298	 */
1299	pr0.ndpr_raf_onlink = 1; /* should be configurable? */
1300	pr0.ndpr_raf_auto =
1301	    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
1302	pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
1303	pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
1304
1305	/* add the prefix if not yet. */
1306	if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
1307		/*
1308		 * nd6_prelist_add will install the corresponding
1309		 * interface route.
1310		 */
1311		if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
1312			if (carp_attached)
1313				(*carp_detach_p)(&ia->ia_ifa, false);
1314			goto out;
1315		}
1316	}
1317
1318	/* relate the address to the prefix */
1319	if (ia->ia6_ndpr == NULL) {
1320		ia->ia6_ndpr = pr;
1321		pr->ndpr_addrcnt++;
1322
1323		/*
1324		 * If this is the first autoconf address from the
1325		 * prefix, create a temporary address as well
1326		 * (when required).
1327		 */
1328		if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
1329		    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
1330			int e;
1331			if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
1332				log(LOG_NOTICE, "in6_control: failed "
1333				    "to create a temporary address, "
1334				    "errno=%d\n", e);
1335			}
1336		}
1337	}
1338	nd6_prefix_rele(pr);
1339
1340	/*
1341	 * this might affect the status of autoconfigured addresses,
1342	 * that is, this address might make other addresses detached.
1343	 */
1344	pfxlist_onlink_check();
1345
1346aifaddr_out:
1347	/*
1348	 * Try to clear the flag when a new IPv6 address is added
1349	 * onto an IFDISABLED interface and it succeeds.
1350	 */
1351	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1352		struct in6_ndireq nd;
1353
1354		memset(&nd, 0, sizeof(nd));
1355		nd.ndi.flags = ND_IFINFO(ifp)->flags;
1356		nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
1357		if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
1358			log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
1359			    "SIOCSIFINFO_FLAGS for -ifdisabled "
1360			    "failed.");
1361		/*
1362		 * Ignore failure of clearing the flag intentionally.
1363		 * The failure means address duplication was detected.
1364		 */
1365	}
1366	error = 0;
1367
1368out:
1369	if (ia != NULL)
1370		ifa_free(&ia->ia_ifa);
1371	return (error);
1372}
1373
1374void
1375in6_purgeaddr(struct ifaddr *ifa)
1376{
1377	struct ifnet *ifp = ifa->ifa_ifp;
1378	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1379	struct in6_multi_mship *imm;
1380	int error;
1381
1382	if (ifa->ifa_carp)
1383		(*carp_detach_p)(ifa, false);
1384
1385	/*
1386	 * Remove the loopback route to the interface address.
1387	 * The check for the current setting of "nd6_useloopback"
1388	 * is not needed.
1389	 */
1390	if (ia->ia_flags & IFA_RTSELF) {
1391		error = ifa_del_loopback_route((struct ifaddr *)ia,
1392		    (struct sockaddr *)&ia->ia_addr);
1393		if (error == 0)
1394			ia->ia_flags &= ~IFA_RTSELF;
1395	}
1396
1397	/* stop DAD processing */
1398	nd6_dad_stop(ifa);
1399
1400	/* Leave multicast groups. */
1401	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1402		LIST_REMOVE(imm, i6mm_chain);
1403		if (imm->i6mm_maddr != NULL)
1404			in6_leavegroup(imm->i6mm_maddr, NULL);
1405		free(imm, M_IP6MADDR);
1406	}
1407	/* Check if we need to remove p2p route */
1408	if ((ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1409		error = in6_handle_dstaddr_rtrequest(RTM_DELETE, ia);
1410		if (error != 0)
1411			log(LOG_INFO, "%s: err=%d, destination address delete "
1412			    "failed\n", __func__, error);
1413		ia->ia_flags &= ~IFA_ROUTE;
1414	}
1415
1416	in6_newaddrmsg(ia, RTM_DELETE);
1417	in6_unlink_ifa(ia, ifp);
1418}
1419
1420/*
1421 * Removes @ia from the corresponding interfaces and unlinks corresponding
1422 *  prefix if no addresses are using it anymore.
1423 */
1424void
1425in6_purgeifaddr(struct in6_ifaddr *ia)
1426{
1427	struct nd_prefix *pr;
1428
1429	/*
1430	 * If the address being deleted is the only one that owns
1431	 * the corresponding prefix, expire the prefix as well.
1432	 * XXX: theoretically, we don't have to worry about such
1433	 * relationship, since we separate the address management
1434	 * and the prefix management.  We do this, however, to provide
1435	 * as much backward compatibility as possible in terms of
1436	 * the ioctl operation.
1437	 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
1438	 */
1439	pr = ia->ia6_ndpr;
1440	in6_purgeaddr(&ia->ia_ifa);
1441	if (pr != NULL && pr->ndpr_addrcnt == 0) {
1442		ND6_WLOCK();
1443		nd6_prefix_unlink(pr, NULL);
1444		ND6_WUNLOCK();
1445		nd6_prefix_del(pr);
1446	}
1447}
1448
1449
1450static void
1451in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1452{
1453	char ip6buf[INET6_ADDRSTRLEN];
1454	int remove_lle;
1455
1456	IF_ADDR_WLOCK(ifp);
1457	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1458	IF_ADDR_WUNLOCK(ifp);
1459	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1460
1461	/*
1462	 * Defer the release of what might be the last reference to the
1463	 * in6_ifaddr so that it can't be freed before the remainder of the
1464	 * cleanup.
1465	 */
1466	IN6_IFADDR_WLOCK();
1467	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1468	CK_LIST_REMOVE(ia, ia6_hash);
1469	IN6_IFADDR_WUNLOCK();
1470
1471	/*
1472	 * Release the reference to the base prefix.  There should be a
1473	 * positive reference.
1474	 */
1475	remove_lle = 0;
1476	if (ia->ia6_ndpr == NULL) {
1477		nd6log((LOG_NOTICE,
1478		    "in6_unlink_ifa: autoconf'ed address "
1479		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1480	} else {
1481		ia->ia6_ndpr->ndpr_addrcnt--;
1482		/* Do not delete lles within prefix if refcont != 0 */
1483		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1484			remove_lle = 1;
1485		ia->ia6_ndpr = NULL;
1486	}
1487
1488	nd6_rem_ifa_lle(ia, remove_lle);
1489
1490	/*
1491	 * Also, if the address being removed is autoconf'ed, call
1492	 * pfxlist_onlink_check() since the release might affect the status of
1493	 * other (detached) addresses.
1494	 */
1495	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1496		pfxlist_onlink_check();
1497	}
1498	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1499}
1500
1501/*
1502 * Notifies other subsystems about address change/arrival:
1503 * 1) Notifies device handler on the first IPv6 address assignment
1504 * 2) Handle routing table changes for P2P links and route
1505 * 3) Handle routing table changes for address host route
1506 */
1507static int
1508in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1509    struct in6_aliasreq *ifra, int hostIsNew)
1510{
1511	int	error = 0, ifacount = 0;
1512	struct ifaddr *ifa;
1513	struct sockaddr_in6 *pdst;
1514	char ip6buf[INET6_ADDRSTRLEN];
1515
1516	/*
1517	 * Give the interface a chance to initialize
1518	 * if this is its first address,
1519	 */
1520	if (hostIsNew != 0) {
1521		struct epoch_tracker et;
1522
1523		NET_EPOCH_ENTER(et);
1524		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1525			if (ifa->ifa_addr->sa_family != AF_INET6)
1526				continue;
1527			ifacount++;
1528		}
1529		NET_EPOCH_EXIT(et);
1530	}
1531
1532	if (ifacount <= 1 && ifp->if_ioctl) {
1533		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1534		if (error)
1535			goto done;
1536	}
1537
1538	/*
1539	 * If a new destination address is specified, scrub the old one and
1540	 * install the new destination.  Note that the interface must be
1541	 * p2p or loopback.
1542	 */
1543	pdst = &ifra->ifra_dstaddr;
1544	if (pdst->sin6_family == AF_INET6 &&
1545	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1546		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1547		    (in6_handle_dstaddr_rtrequest(RTM_DELETE, ia) != 0)) {
1548			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1549			    "remove a route to the old destination: %s\n",
1550			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1551			/* proceed anyway... */
1552		} else
1553			ia->ia_flags &= ~IFA_ROUTE;
1554		ia->ia_dstaddr = *pdst;
1555	}
1556
1557	/*
1558	 * If a new destination address is specified for a point-to-point
1559	 * interface, install a route to the destination as an interface
1560	 * direct route.
1561	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1562	 * interface that share the same destination.
1563	 */
1564	if (!(ia->ia_flags & IFA_ROUTE) && ifa_is_p2p(ia)) {
1565		error = in6_handle_dstaddr_rtrequest(RTM_ADD, ia);
1566		if (error)
1567			goto done;
1568		ia->ia_flags |= IFA_ROUTE;
1569	}
1570
1571	/*
1572	 * add a loopback route to self if not exists
1573	 */
1574	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1575		error = ifa_add_loopback_route((struct ifaddr *)ia,
1576		    (struct sockaddr *)&ia->ia_addr);
1577		if (error == 0)
1578			ia->ia_flags |= IFA_RTSELF;
1579	}
1580done:
1581	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1582	    "Invoking IPv6 network device address event may sleep");
1583
1584	ifa_ref(&ia->ia_ifa);
1585	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1586	    IFADDR_EVENT_ADD);
1587	ifa_free(&ia->ia_ifa);
1588
1589	return (error);
1590}
1591
1592/*
1593 * Find an IPv6 interface link-local address specific to an interface.
1594 * ifaddr is returned referenced.
1595 */
1596struct in6_ifaddr *
1597in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1598{
1599	struct ifaddr *ifa;
1600
1601	NET_EPOCH_ASSERT();
1602
1603	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1604		if (ifa->ifa_addr->sa_family != AF_INET6)
1605			continue;
1606		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1607			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1608			    ignoreflags) != 0)
1609				continue;
1610			ifa_ref(ifa);
1611			break;
1612		}
1613	}
1614
1615	return ((struct in6_ifaddr *)ifa);
1616}
1617
1618/*
1619 * find the interface address corresponding to a given IPv6 address.
1620 * ifaddr is returned referenced if @referenced flag is set.
1621 */
1622struct in6_ifaddr *
1623in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid, bool referenced)
1624{
1625	struct rm_priotracker in6_ifa_tracker;
1626	struct in6_ifaddr *ia;
1627
1628	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1629	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1630		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1631			if (zoneid != 0 &&
1632			    zoneid != ia->ia_addr.sin6_scope_id)
1633				continue;
1634			if (referenced)
1635				ifa_ref(&ia->ia_ifa);
1636			break;
1637		}
1638	}
1639	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1640	return (ia);
1641}
1642
1643/*
1644 * find the internet address corresponding to a given interface and address.
1645 * ifaddr is returned referenced.
1646 */
1647struct in6_ifaddr *
1648in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1649{
1650	struct epoch_tracker et;
1651	struct ifaddr *ifa;
1652
1653	NET_EPOCH_ENTER(et);
1654	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1655		if (ifa->ifa_addr->sa_family != AF_INET6)
1656			continue;
1657		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1658			ifa_ref(ifa);
1659			break;
1660		}
1661	}
1662	NET_EPOCH_EXIT(et);
1663
1664	return ((struct in6_ifaddr *)ifa);
1665}
1666
1667/*
1668 * Find a link-local scoped address on ifp and return it if any.
1669 */
1670struct in6_ifaddr *
1671in6ifa_llaonifp(struct ifnet *ifp)
1672{
1673	struct epoch_tracker et;
1674	struct sockaddr_in6 *sin6;
1675	struct ifaddr *ifa;
1676
1677	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1678		return (NULL);
1679	NET_EPOCH_ENTER(et);
1680	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1681		if (ifa->ifa_addr->sa_family != AF_INET6)
1682			continue;
1683		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1684		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1685		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1686		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1687			break;
1688	}
1689	NET_EPOCH_EXIT(et);
1690
1691	return ((struct in6_ifaddr *)ifa);
1692}
1693
1694/*
1695 * Convert IP6 address to printable (loggable) representation. Caller
1696 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1697 */
1698static char digits[] = "0123456789abcdef";
1699char *
1700ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1701{
1702	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1703	char *cp;
1704	const u_int16_t *a = (const u_int16_t *)addr;
1705	const u_int8_t *d;
1706	int dcolon = 0, zero = 0;
1707
1708	cp = ip6buf;
1709
1710	for (i = 0; i < 8; i++) {
1711		if (*(a + i) == 0) {
1712			cnt++;
1713			if (cnt == 1)
1714				idx = i;
1715		}
1716		else if (maxcnt < cnt) {
1717			maxcnt = cnt;
1718			index = idx;
1719			cnt = 0;
1720		}
1721	}
1722	if (maxcnt < cnt) {
1723		maxcnt = cnt;
1724		index = idx;
1725	}
1726
1727	for (i = 0; i < 8; i++) {
1728		if (dcolon == 1) {
1729			if (*a == 0) {
1730				if (i == 7)
1731					*cp++ = ':';
1732				a++;
1733				continue;
1734			} else
1735				dcolon = 2;
1736		}
1737		if (*a == 0) {
1738			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1739				if (i == 0)
1740					*cp++ = ':';
1741				*cp++ = ':';
1742				dcolon = 1;
1743			} else {
1744				*cp++ = '0';
1745				*cp++ = ':';
1746			}
1747			a++;
1748			continue;
1749		}
1750		d = (const u_char *)a;
1751		/* Try to eliminate leading zeros in printout like in :0001. */
1752		zero = 1;
1753		*cp = digits[*d >> 4];
1754		if (*cp != '0') {
1755			zero = 0;
1756			cp++;
1757		}
1758		*cp = digits[*d++ & 0xf];
1759		if (zero == 0 || (*cp != '0')) {
1760			zero = 0;
1761			cp++;
1762		}
1763		*cp = digits[*d >> 4];
1764		if (zero == 0 || (*cp != '0')) {
1765			zero = 0;
1766			cp++;
1767		}
1768		*cp++ = digits[*d & 0xf];
1769		*cp++ = ':';
1770		a++;
1771	}
1772	*--cp = '\0';
1773	return (ip6buf);
1774}
1775
1776int
1777in6_localaddr(struct in6_addr *in6)
1778{
1779	struct rm_priotracker in6_ifa_tracker;
1780	struct in6_ifaddr *ia;
1781
1782	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1783		return 1;
1784
1785	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1786	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1787		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1788		    &ia->ia_prefixmask.sin6_addr)) {
1789			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1790			return 1;
1791		}
1792	}
1793	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1794
1795	return (0);
1796}
1797
1798/*
1799 * Return 1 if an internet address is for the local host and configured
1800 * on one of its interfaces.
1801 */
1802int
1803in6_localip(struct in6_addr *in6)
1804{
1805	struct rm_priotracker in6_ifa_tracker;
1806	struct in6_ifaddr *ia;
1807
1808	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1809	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1810		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1811			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1812			return (1);
1813		}
1814	}
1815	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1816	return (0);
1817}
1818
1819/*
1820 * Like in6_localip(), but FIB-aware and carp(4)-aware.
1821 */
1822bool
1823in6_localip_fib(struct in6_addr *in6, uint16_t fib)
1824{
1825	struct rm_priotracker in6_ifa_tracker;
1826	struct in6_ifaddr *ia;
1827
1828	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1829	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1830		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr) &&
1831		    (ia->ia_ifa.ifa_carp == NULL ||
1832		    carp_master_p(&ia->ia_ifa)) &&
1833		    ia->ia_ifa.ifa_ifp->if_fib == fib) {
1834			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1835			return (true);
1836		}
1837	}
1838	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1839	return (false);
1840}
1841
1842/*
1843 * Return 1 if an internet address is configured on an interface.
1844 */
1845int
1846in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1847{
1848	struct in6_addr in6;
1849	struct ifaddr *ifa;
1850	struct in6_ifaddr *ia6;
1851
1852	NET_EPOCH_ASSERT();
1853
1854	in6 = *addr;
1855	if (in6_clearscope(&in6))
1856		return (0);
1857	in6_setscope(&in6, ifp, NULL);
1858
1859	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1860		if (ifa->ifa_addr->sa_family != AF_INET6)
1861			continue;
1862		ia6 = (struct in6_ifaddr *)ifa;
1863		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1864			return (1);
1865	}
1866
1867	return (0);
1868}
1869
1870int
1871in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1872{
1873	struct rm_priotracker in6_ifa_tracker;
1874	struct in6_ifaddr *ia;
1875
1876	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1877	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1878		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1879			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1880				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1881				return (1); /* true */
1882			}
1883			break;
1884		}
1885	}
1886	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1887
1888	return (0);		/* false */
1889}
1890
1891/*
1892 * return length of part which dst and src are equal
1893 * hard coding...
1894 */
1895int
1896in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1897{
1898	int match = 0;
1899	u_char *s = (u_char *)src, *d = (u_char *)dst;
1900	u_char *lim = s + 16, r;
1901
1902	while (s < lim)
1903		if ((r = (*d++ ^ *s++)) != 0) {
1904			while (r < 128) {
1905				match++;
1906				r <<= 1;
1907			}
1908			break;
1909		} else
1910			match += 8;
1911	return match;
1912}
1913
1914/* XXX: to be scope conscious */
1915int
1916in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1917{
1918	int bytelen, bitlen;
1919
1920	/* sanity check */
1921	if (0 > len || len > 128) {
1922		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1923		    len);
1924		return (0);
1925	}
1926
1927	bytelen = len / 8;
1928	bitlen = len % 8;
1929
1930	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1931		return (0);
1932	if (bitlen != 0 &&
1933	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1934	    p2->s6_addr[bytelen] >> (8 - bitlen))
1935		return (0);
1936
1937	return (1);
1938}
1939
1940void
1941in6_prefixlen2mask(struct in6_addr *maskp, int len)
1942{
1943	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1944	int bytelen, bitlen, i;
1945
1946	/* sanity check */
1947	if (0 > len || len > 128) {
1948		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1949		    len);
1950		return;
1951	}
1952
1953	bzero(maskp, sizeof(*maskp));
1954	bytelen = len / 8;
1955	bitlen = len % 8;
1956	for (i = 0; i < bytelen; i++)
1957		maskp->s6_addr[i] = 0xff;
1958	if (bitlen)
1959		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1960}
1961
1962/*
1963 * return the best address out of the same scope. if no address was
1964 * found, return the first valid address from designated IF.
1965 */
1966struct in6_ifaddr *
1967in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1968{
1969	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1970	struct ifaddr *ifa;
1971	struct in6_ifaddr *besta = NULL;
1972	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1973
1974	NET_EPOCH_ASSERT();
1975
1976	dep[0] = dep[1] = NULL;
1977
1978	/*
1979	 * We first look for addresses in the same scope.
1980	 * If there is one, return it.
1981	 * If two or more, return one which matches the dst longest.
1982	 * If none, return one of global addresses assigned other ifs.
1983	 */
1984	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1985		if (ifa->ifa_addr->sa_family != AF_INET6)
1986			continue;
1987		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1988			continue; /* XXX: is there any case to allow anycast? */
1989		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1990			continue; /* don't use this interface */
1991		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1992			continue;
1993		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1994			if (V_ip6_use_deprecated)
1995				dep[0] = (struct in6_ifaddr *)ifa;
1996			continue;
1997		}
1998
1999		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2000			/*
2001			 * call in6_matchlen() as few as possible
2002			 */
2003			if (besta) {
2004				if (blen == -1)
2005					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2006				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2007				if (tlen > blen) {
2008					blen = tlen;
2009					besta = (struct in6_ifaddr *)ifa;
2010				}
2011			} else
2012				besta = (struct in6_ifaddr *)ifa;
2013		}
2014	}
2015	if (besta)
2016		return (besta);
2017
2018	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2019		if (ifa->ifa_addr->sa_family != AF_INET6)
2020			continue;
2021		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2022			continue; /* XXX: is there any case to allow anycast? */
2023		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2024			continue; /* don't use this interface */
2025		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2026			continue;
2027		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2028			if (V_ip6_use_deprecated)
2029				dep[1] = (struct in6_ifaddr *)ifa;
2030			continue;
2031		}
2032
2033		return (struct in6_ifaddr *)ifa;
2034	}
2035
2036	/* use the last-resort values, that are, deprecated addresses */
2037	if (dep[0])
2038		return dep[0];
2039	if (dep[1])
2040		return dep[1];
2041
2042	return NULL;
2043}
2044
2045/*
2046 * perform DAD when interface becomes IFF_UP.
2047 */
2048void
2049in6_if_up(struct ifnet *ifp)
2050{
2051	struct epoch_tracker et;
2052	struct ifaddr *ifa;
2053	struct in6_ifaddr *ia;
2054
2055	NET_EPOCH_ENTER(et);
2056	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2057		if (ifa->ifa_addr->sa_family != AF_INET6)
2058			continue;
2059		ia = (struct in6_ifaddr *)ifa;
2060		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2061			/*
2062			 * The TENTATIVE flag was likely set by hand
2063			 * beforehand, implicitly indicating the need for DAD.
2064			 * We may be able to skip the random delay in this
2065			 * case, but we impose delays just in case.
2066			 */
2067			nd6_dad_start(ifa,
2068			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2069		}
2070	}
2071	NET_EPOCH_EXIT(et);
2072
2073	/*
2074	 * special cases, like 6to4, are handled in in6_ifattach
2075	 */
2076	in6_ifattach(ifp, NULL);
2077}
2078
2079static void
2080in6_ifevent(void *arg __unused, struct ifnet *ifp, int event)
2081{
2082	if (event == IFNET_EVENT_UP)
2083		in6_if_up(ifp);
2084}
2085
2086static void
2087in6_init(void *arg __unused)
2088{
2089	EVENTHANDLER_REGISTER(ifnet_event, in6_ifevent, NULL, EVENTHANDLER_PRI_ANY);
2090}
2091SYSINIT(in6_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, in6_init, NULL);
2092
2093int
2094in6if_do_dad(struct ifnet *ifp)
2095{
2096
2097	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2098		return (0);
2099	if ((ifp->if_flags & IFF_MULTICAST) == 0)
2100		return (0);
2101	if ((ND_IFINFO(ifp)->flags &
2102	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
2103		return (0);
2104	return (1);
2105}
2106
2107/*
2108 * Calculate max IPv6 MTU through all the interfaces and store it
2109 * to in6_maxmtu.
2110 */
2111void
2112in6_setmaxmtu(void)
2113{
2114	struct epoch_tracker et;
2115	unsigned long maxmtu = 0;
2116	struct ifnet *ifp;
2117
2118	NET_EPOCH_ENTER(et);
2119	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2120		/* this function can be called during ifnet initialization */
2121		if (!ifp->if_afdata[AF_INET6])
2122			continue;
2123		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2124		    IN6_LINKMTU(ifp) > maxmtu)
2125			maxmtu = IN6_LINKMTU(ifp);
2126	}
2127	NET_EPOCH_EXIT(et);
2128	if (maxmtu)	/* update only when maxmtu is positive */
2129		V_in6_maxmtu = maxmtu;
2130}
2131
2132/*
2133 * Provide the length of interface identifiers to be used for the link attached
2134 * to the given interface.  The length should be defined in "IPv6 over
2135 * xxx-link" document.  Note that address architecture might also define
2136 * the length for a particular set of address prefixes, regardless of the
2137 * link type.  As clarified in rfc2462bis, those two definitions should be
2138 * consistent, and those really are as of August 2004.
2139 */
2140int
2141in6_if2idlen(struct ifnet *ifp)
2142{
2143	switch (ifp->if_type) {
2144	case IFT_ETHER:		/* RFC2464 */
2145	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2146	case IFT_L2VLAN:	/* ditto */
2147	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2148	case IFT_INFINIBAND:
2149		return (64);
2150	case IFT_PPP:		/* RFC2472 */
2151		return (64);
2152	case IFT_FRELAY:	/* RFC2590 */
2153		return (64);
2154	case IFT_IEEE1394:	/* RFC3146 */
2155		return (64);
2156	case IFT_GIF:
2157		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2158	case IFT_LOOP:
2159		return (64);	/* XXX: is this really correct? */
2160	default:
2161		/*
2162		 * Unknown link type:
2163		 * It might be controversial to use the today's common constant
2164		 * of 64 for these cases unconditionally.  For full compliance,
2165		 * we should return an error in this case.  On the other hand,
2166		 * if we simply miss the standard for the link type or a new
2167		 * standard is defined for a new link type, the IFID length
2168		 * is very likely to be the common constant.  As a compromise,
2169		 * we always use the constant, but make an explicit notice
2170		 * indicating the "unknown" case.
2171		 */
2172		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2173		return (64);
2174	}
2175}
2176
2177struct in6_llentry {
2178	struct llentry		base;
2179};
2180
2181#define	IN6_LLTBL_DEFAULT_HSIZE	32
2182#define	IN6_LLTBL_HASH(k, h) \
2183	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2184
2185/*
2186 * Do actual deallocation of @lle.
2187 */
2188static void
2189in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2190{
2191	struct llentry *lle;
2192
2193	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2194	LLE_LOCK_DESTROY(lle);
2195	LLE_REQ_DESTROY(lle);
2196	free(lle, M_LLTABLE);
2197}
2198
2199/*
2200 * Called by LLE_FREE_LOCKED when number of references
2201 * drops to zero.
2202 */
2203static void
2204in6_lltable_destroy_lle(struct llentry *lle)
2205{
2206
2207	LLE_WUNLOCK(lle);
2208	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2209}
2210
2211static struct llentry *
2212in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2213{
2214	struct in6_llentry *lle;
2215
2216	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2217	if (lle == NULL)		/* NB: caller generates msg */
2218		return NULL;
2219
2220	lle->base.r_l3addr.addr6 = *addr6;
2221	lle->base.lle_refcnt = 1;
2222	lle->base.lle_free = in6_lltable_destroy_lle;
2223	LLE_LOCK_INIT(&lle->base);
2224	LLE_REQ_INIT(&lle->base);
2225	callout_init(&lle->base.lle_timer, 1);
2226
2227	return (&lle->base);
2228}
2229
2230static int
2231in6_lltable_match_prefix(const struct sockaddr *saddr,
2232    const struct sockaddr *smask, u_int flags, struct llentry *lle)
2233{
2234	const struct in6_addr *addr, *mask, *lle_addr;
2235
2236	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2237	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2238	lle_addr = &lle->r_l3addr.addr6;
2239
2240	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2241		return (0);
2242
2243	if (lle->la_flags & LLE_IFADDR) {
2244		/*
2245		 * Delete LLE_IFADDR records IFF address & flag matches.
2246		 * Note that addr is the interface address within prefix
2247		 * being matched.
2248		 */
2249		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2250		    (flags & LLE_STATIC) != 0)
2251			return (1);
2252		return (0);
2253	}
2254
2255	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2256	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2257		return (1);
2258
2259	return (0);
2260}
2261
2262static void
2263in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2264{
2265	struct ifnet *ifp __diagused;
2266
2267	LLE_WLOCK_ASSERT(lle);
2268	KASSERT(llt != NULL, ("lltable is NULL"));
2269
2270	/* Unlink entry from table */
2271	if ((lle->la_flags & LLE_LINKED) != 0) {
2272		ifp = llt->llt_ifp;
2273		IF_AFDATA_WLOCK_ASSERT(ifp);
2274		lltable_unlink_entry(llt, lle);
2275	}
2276
2277	llentry_free(lle);
2278}
2279
2280static int
2281in6_lltable_rtcheck(struct ifnet *ifp,
2282		    u_int flags,
2283		    const struct sockaddr *l3addr)
2284{
2285	const struct sockaddr_in6 *sin6;
2286	struct nhop_object *nh;
2287	struct in6_addr dst;
2288	uint32_t scopeid;
2289	char ip6buf[INET6_ADDRSTRLEN];
2290	int fibnum;
2291
2292	NET_EPOCH_ASSERT();
2293	KASSERT(l3addr->sa_family == AF_INET6,
2294	    ("sin_family %d", l3addr->sa_family));
2295
2296	sin6 = (const struct sockaddr_in6 *)l3addr;
2297	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2298	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2299	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2300	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2301		struct ifaddr *ifa;
2302		/*
2303		 * Create an ND6 cache for an IPv6 neighbor
2304		 * that is not covered by our own prefix.
2305		 */
2306		ifa = ifaof_ifpforaddr(l3addr, ifp);
2307		if (ifa != NULL) {
2308			return 0;
2309		}
2310		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2311		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2312		return EINVAL;
2313	}
2314	return 0;
2315}
2316
2317static inline uint32_t
2318in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2319{
2320
2321	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2322}
2323
2324static uint32_t
2325in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2326{
2327
2328	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2329}
2330
2331static void
2332in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2333{
2334	struct sockaddr_in6 *sin6;
2335
2336	sin6 = (struct sockaddr_in6 *)sa;
2337	bzero(sin6, sizeof(*sin6));
2338	sin6->sin6_family = AF_INET6;
2339	sin6->sin6_len = sizeof(*sin6);
2340	sin6->sin6_addr = lle->r_l3addr.addr6;
2341}
2342
2343static inline struct llentry *
2344in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2345{
2346	struct llentry *lle;
2347	struct llentries *lleh;
2348	u_int hashidx;
2349
2350	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2351	lleh = &llt->lle_head[hashidx];
2352	CK_LIST_FOREACH(lle, lleh, lle_next) {
2353		if (lle->la_flags & LLE_DELETED)
2354			continue;
2355		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2356			break;
2357	}
2358
2359	return (lle);
2360}
2361
2362static void
2363in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2364{
2365
2366	lle->la_flags |= LLE_DELETED;
2367
2368	/* Leave the solicited multicast group. */
2369	if ((lle->la_flags & LLE_PUB) != 0)
2370		in6_leave_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2371	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2372#ifdef DIAGNOSTIC
2373	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2374#endif
2375	llentry_free(lle);
2376}
2377
2378static struct llentry *
2379in6_lltable_alloc(struct lltable *llt, u_int flags,
2380	const struct sockaddr *l3addr)
2381{
2382	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2383	struct ifnet *ifp = llt->llt_ifp;
2384	struct llentry *lle;
2385	char linkhdr[LLE_MAX_LINKHDR];
2386	size_t linkhdrsize;
2387	int lladdr_off;
2388
2389	KASSERT(l3addr->sa_family == AF_INET6,
2390	    ("sin_family %d", l3addr->sa_family));
2391
2392	/*
2393	 * A route that covers the given address must have
2394	 * been installed 1st because we are doing a resolution,
2395	 * verify this.
2396	 */
2397	if (!(flags & LLE_IFADDR) &&
2398	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2399		return (NULL);
2400
2401	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2402	if (lle == NULL) {
2403		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2404		return (NULL);
2405	}
2406	lle->la_flags = flags;
2407	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2408		linkhdrsize = LLE_MAX_LINKHDR;
2409		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2410		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2411			in6_lltable_free_entry(llt, lle);
2412			return (NULL);
2413		}
2414		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2415		    lladdr_off);
2416		lle->la_flags |= LLE_STATIC;
2417	}
2418
2419	if ((lle->la_flags & LLE_STATIC) != 0)
2420		lle->ln_state = ND6_LLINFO_REACHABLE;
2421
2422	return (lle);
2423}
2424
2425static struct llentry *
2426in6_lltable_lookup(struct lltable *llt, u_int flags,
2427	const struct sockaddr *l3addr)
2428{
2429	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2430	int family = flags >> 16;
2431	struct llentry *lle;
2432
2433	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2434	KASSERT(l3addr->sa_family == AF_INET6,
2435	    ("sin_family %d", l3addr->sa_family));
2436	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2437	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2438	    ("wrong lle request flags: %#x", flags));
2439
2440	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2441
2442	if (__predict_false(family != AF_INET6))
2443		lle = llentry_lookup_family(lle, family);
2444
2445	if (lle == NULL)
2446		return (NULL);
2447
2448	if (flags & LLE_UNLOCKED)
2449		return (lle);
2450
2451	if (flags & LLE_EXCLUSIVE)
2452		LLE_WLOCK(lle);
2453	else
2454		LLE_RLOCK(lle);
2455
2456	/*
2457	 * If the afdata lock is not held, the LLE may have been unlinked while
2458	 * we were blocked on the LLE lock.  Check for this case.
2459	 */
2460	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2461		if (flags & LLE_EXCLUSIVE)
2462			LLE_WUNLOCK(lle);
2463		else
2464			LLE_RUNLOCK(lle);
2465		return (NULL);
2466	}
2467	return (lle);
2468}
2469
2470static int
2471in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2472    struct sysctl_req *wr)
2473{
2474	struct ifnet *ifp = llt->llt_ifp;
2475	/* XXX stack use */
2476	struct {
2477		struct rt_msghdr	rtm;
2478		struct sockaddr_in6	sin6;
2479		/*
2480		 * ndp.c assumes that sdl is word aligned
2481		 */
2482#ifdef __LP64__
2483		uint32_t		pad;
2484#endif
2485		struct sockaddr_dl	sdl;
2486	} ndpc;
2487	struct sockaddr_dl *sdl;
2488	int error;
2489
2490	bzero(&ndpc, sizeof(ndpc));
2491	/* skip deleted entries */
2492	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2493		return (0);
2494	/* Skip if jailed and not a valid IP of the prison. */
2495	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2496	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2497		return (0);
2498	/*
2499	 * produce a msg made of:
2500	 *  struct rt_msghdr;
2501	 *  struct sockaddr_in6 (IPv6)
2502	 *  struct sockaddr_dl;
2503	 */
2504	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2505	ndpc.rtm.rtm_version = RTM_VERSION;
2506	ndpc.rtm.rtm_type = RTM_GET;
2507	ndpc.rtm.rtm_flags = RTF_UP;
2508	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2509	sa6_recoverscope(&ndpc.sin6);
2510
2511	/* publish */
2512	if (lle->la_flags & LLE_PUB)
2513		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2514
2515	sdl = &ndpc.sdl;
2516	sdl->sdl_family = AF_LINK;
2517	sdl->sdl_len = sizeof(*sdl);
2518	sdl->sdl_index = ifp->if_index;
2519	sdl->sdl_type = ifp->if_type;
2520	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2521		sdl->sdl_alen = ifp->if_addrlen;
2522		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2523	} else {
2524		sdl->sdl_alen = 0;
2525		bzero(LLADDR(sdl), ifp->if_addrlen);
2526	}
2527	if (lle->la_expire != 0)
2528		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2529		    lle->lle_remtime / hz + time_second - time_uptime;
2530	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2531	if (lle->la_flags & LLE_STATIC)
2532		ndpc.rtm.rtm_flags |= RTF_STATIC;
2533	if (lle->la_flags & LLE_IFADDR)
2534		ndpc.rtm.rtm_flags |= RTF_PINNED;
2535	if (lle->ln_router != 0)
2536		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2537	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2538	/* Store state in rmx_weight value */
2539	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2540	ndpc.rtm.rtm_index = ifp->if_index;
2541	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2542
2543	return (error);
2544}
2545
2546static void
2547in6_lltable_post_resolved(struct lltable *llt, struct llentry *lle)
2548{
2549	/* Join the solicited multicast group for dst. */
2550	if ((lle->la_flags & LLE_PUB) == LLE_PUB)
2551		in6_join_proxy_ndp_mc(llt->llt_ifp, &lle->r_l3addr.addr6);
2552}
2553
2554static struct lltable *
2555in6_lltattach(struct ifnet *ifp)
2556{
2557	struct lltable *llt;
2558
2559	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2560	llt->llt_af = AF_INET6;
2561	llt->llt_ifp = ifp;
2562
2563	llt->llt_lookup = in6_lltable_lookup;
2564	llt->llt_alloc_entry = in6_lltable_alloc;
2565	llt->llt_delete_entry = in6_lltable_delete_entry;
2566	llt->llt_dump_entry = in6_lltable_dump_entry;
2567	llt->llt_hash = in6_lltable_hash;
2568	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2569	llt->llt_free_entry = in6_lltable_free_entry;
2570	llt->llt_match_prefix = in6_lltable_match_prefix;
2571	llt->llt_mark_used = llentry_mark_used;
2572	llt->llt_post_resolved = in6_lltable_post_resolved;
2573 	lltable_link(llt);
2574
2575	return (llt);
2576}
2577
2578struct lltable *
2579in6_lltable_get(struct ifnet *ifp)
2580{
2581	struct lltable *llt = NULL;
2582
2583	void *afdata_ptr = ifp->if_afdata[AF_INET6];
2584	if (afdata_ptr != NULL)
2585		llt = ((struct in6_ifextra *)afdata_ptr)->lltable;
2586	return (llt);
2587}
2588
2589void *
2590in6_domifattach(struct ifnet *ifp)
2591{
2592	struct in6_ifextra *ext;
2593
2594	/* There are not IPv6-capable interfaces. */
2595	switch (ifp->if_type) {
2596	case IFT_PFLOG:
2597	case IFT_PFSYNC:
2598	case IFT_USB:
2599		return (NULL);
2600	}
2601	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2602	bzero(ext, sizeof(*ext));
2603
2604	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2605	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2606	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2607	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2608
2609	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2610	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2611	    M_WAITOK);
2612	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2613	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2614
2615	ext->nd_ifinfo = nd6_ifattach(ifp);
2616	ext->scope6_id = scope6_ifattach(ifp);
2617	ext->lltable = in6_lltattach(ifp);
2618
2619	ext->mld_ifinfo = mld_domifattach(ifp);
2620
2621	return ext;
2622}
2623
2624int
2625in6_domifmtu(struct ifnet *ifp)
2626{
2627	if (ifp->if_afdata[AF_INET6] == NULL)
2628		return ifp->if_mtu;
2629
2630	return (IN6_LINKMTU(ifp));
2631}
2632
2633void
2634in6_domifdetach(struct ifnet *ifp, void *aux)
2635{
2636	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2637
2638	mld_domifdetach(ifp);
2639	scope6_ifdetach(ext->scope6_id);
2640	nd6_ifdetach(ifp, ext->nd_ifinfo);
2641	lltable_free(ext->lltable);
2642	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2643	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2644	free(ext->in6_ifstat, M_IFADDR);
2645	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2646	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2647	free(ext->icmp6_ifstat, M_IFADDR);
2648	free(ext, M_IFADDR);
2649}
2650
2651/*
2652 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2653 * v4 mapped addr or v4 compat addr
2654 */
2655void
2656in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2657{
2658
2659	bzero(sin, sizeof(*sin));
2660	sin->sin_len = sizeof(struct sockaddr_in);
2661	sin->sin_family = AF_INET;
2662	sin->sin_port = sin6->sin6_port;
2663	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2664}
2665
2666/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2667void
2668in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2669{
2670	bzero(sin6, sizeof(*sin6));
2671	sin6->sin6_len = sizeof(struct sockaddr_in6);
2672	sin6->sin6_family = AF_INET6;
2673	sin6->sin6_port = sin->sin_port;
2674	sin6->sin6_addr.s6_addr32[0] = 0;
2675	sin6->sin6_addr.s6_addr32[1] = 0;
2676	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2677	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2678}
2679
2680/* Convert sockaddr_in6 into sockaddr_in. */
2681void
2682in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2683{
2684	struct sockaddr_in *sin_p;
2685	struct sockaddr_in6 sin6;
2686
2687	/*
2688	 * Save original sockaddr_in6 addr and convert it
2689	 * to sockaddr_in.
2690	 */
2691	sin6 = *(struct sockaddr_in6 *)nam;
2692	sin_p = (struct sockaddr_in *)nam;
2693	in6_sin6_2_sin(sin_p, &sin6);
2694}
2695
2696/*
2697 * Join/leave the solicited multicast groups for proxy NDP entries.
2698 */
2699static void
2700in6_join_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2701{
2702	struct in6_multi *inm;
2703	struct in6_addr mltaddr;
2704	char ip6buf[INET6_ADDRSTRLEN];
2705	int error;
2706
2707	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2708		return;	/* error logged in in6_solicited_node_maddr. */
2709
2710	error = in6_joingroup(ifp, &mltaddr, NULL, &inm, 0);
2711	if (error != 0) {
2712		nd6log((LOG_WARNING,
2713		    "%s: in6_joingroup failed for %s on %s (errno=%d)\n",
2714		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp),
2715		    error));
2716	}
2717}
2718
2719static void
2720in6_leave_proxy_ndp_mc(struct ifnet *ifp, const struct in6_addr *dst)
2721{
2722	struct epoch_tracker et;
2723	struct in6_multi *inm;
2724	struct in6_addr mltaddr;
2725	char ip6buf[INET6_ADDRSTRLEN];
2726
2727	if (in6_solicited_node_maddr(&mltaddr, ifp, dst) != 0)
2728		return;	/* error logged in in6_solicited_node_maddr. */
2729
2730	NET_EPOCH_ENTER(et);
2731	inm = in6m_lookup(ifp, &mltaddr);
2732	NET_EPOCH_EXIT(et);
2733	if (inm != NULL)
2734		in6_leavegroup(inm, NULL);
2735	else
2736		nd6log((LOG_WARNING, "%s: in6m_lookup failed for %s on %s\n",
2737		    __func__, ip6_sprintf(ip6buf, &mltaddr), if_name(ifp)));
2738}
2739
2740static bool
2741in6_lle_match_pub(struct lltable *llt, struct llentry *lle, void *farg)
2742{
2743	return ((lle->la_flags & LLE_PUB) != 0);
2744}
2745
2746void
2747in6_purge_proxy_ndp(struct ifnet *ifp)
2748{
2749	struct lltable *llt;
2750	bool need_purge;
2751
2752	if (ifp->if_afdata[AF_INET6] == NULL)
2753		return;
2754
2755	llt = LLTABLE6(ifp);
2756	IF_AFDATA_WLOCK(ifp);
2757	need_purge = ((llt->llt_flags & LLT_ADDEDPROXY) != 0);
2758	IF_AFDATA_WUNLOCK(ifp);
2759
2760	/*
2761	 * Ever added proxy ndp entries, leave solicited node multicast
2762	 * before deleting the llentry.
2763	 */
2764	if (need_purge)
2765		lltable_delete_conditional(llt, in6_lle_match_pub, NULL);
2766}
2767