in6.c revision 53cb168f80fa70688bd35678b681684102548ae5
1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $
30 */
31
32/*-
33 * Copyright (c) 1982, 1986, 1991, 1993
34 *	The Regents of the University of California.  All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 *    notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 *    notice, this list of conditions and the following disclaimer in the
43 *    documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 *    may be used to endorse or promote products derived from this software
46 *    without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 *	@(#)in.c	8.2 (Berkeley) 11/15/93
61 */
62
63#include <sys/cdefs.h>
64__FBSDID("$FreeBSD$");
65
66#include "opt_compat.h"
67#include "opt_inet.h"
68#include "opt_inet6.h"
69
70#include <sys/param.h>
71#include <sys/errno.h>
72#include <sys/jail.h>
73#include <sys/malloc.h>
74#include <sys/socket.h>
75#include <sys/socketvar.h>
76#include <sys/sockio.h>
77#include <sys/systm.h>
78#include <sys/priv.h>
79#include <sys/proc.h>
80#include <sys/time.h>
81#include <sys/kernel.h>
82#include <sys/syslog.h>
83
84#include <net/if.h>
85#include <net/if_var.h>
86#include <net/if_types.h>
87#include <net/route.h>
88#include <net/if_dl.h>
89#include <net/vnet.h>
90
91#include <netinet/in.h>
92#include <netinet/in_var.h>
93#include <net/if_llatbl.h>
94#include <netinet/if_ether.h>
95#include <netinet/in_systm.h>
96#include <netinet/ip.h>
97#include <netinet/in_pcb.h>
98#include <netinet/ip_carp.h>
99
100#include <netinet/ip6.h>
101#include <netinet6/ip6_var.h>
102#include <netinet6/nd6.h>
103#include <netinet6/mld6_var.h>
104#include <netinet6/ip6_mroute.h>
105#include <netinet6/in6_ifattach.h>
106#include <netinet6/scope6_var.h>
107#include <netinet6/in6_pcb.h>
108
109/*
110 * Definitions of some costant IP6 addresses.
111 */
112const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
113const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
114const struct in6_addr in6addr_nodelocal_allnodes =
115	IN6ADDR_NODELOCAL_ALLNODES_INIT;
116const struct in6_addr in6addr_linklocal_allnodes =
117	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
118const struct in6_addr in6addr_linklocal_allrouters =
119	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
120const struct in6_addr in6addr_linklocal_allv2routers =
121	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
122
123const struct in6_addr in6mask0 = IN6MASK0;
124const struct in6_addr in6mask32 = IN6MASK32;
125const struct in6_addr in6mask64 = IN6MASK64;
126const struct in6_addr in6mask96 = IN6MASK96;
127const struct in6_addr in6mask128 = IN6MASK128;
128
129const struct sockaddr_in6 sa6_any =
130	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
131
132static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
133	struct ifnet *, struct thread *));
134static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
135	struct sockaddr_in6 *, int));
136static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
137
138int	(*faithprefix_p)(struct in6_addr *);
139
140#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
141#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
142
143void
144in6_ifaddloop(struct ifaddr *ifa)
145{
146	struct sockaddr_dl gateway;
147	struct sockaddr_in6 mask, addr;
148	struct rtentry rt;
149	struct in6_ifaddr *ia;
150	struct ifnet *ifp;
151	struct llentry *ln;
152
153	ia = ifa2ia6(ifa);
154	ifp = ifa->ifa_ifp;
155	IF_AFDATA_LOCK(ifp);
156	ifa->ifa_rtrequest = nd6_rtrequest;
157	ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
158	    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
159	IF_AFDATA_UNLOCK(ifp);
160	if (ln != NULL) {
161		ln->la_expire = 0;  /* for IPv6 this means permanent */
162		ln->ln_state = ND6_LLINFO_REACHABLE;
163		/*
164		 * initialize for rtmsg generation
165		 */
166		bzero(&gateway, sizeof(gateway));
167		gateway.sdl_len = sizeof(gateway);
168		gateway.sdl_family = AF_LINK;
169		gateway.sdl_nlen = 0;
170		gateway.sdl_alen = 6;
171		memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
172		    sizeof(ln->ll_addr));
173		LLE_WUNLOCK(ln);
174	}
175
176	bzero(&rt, sizeof(rt));
177	rt.rt_gateway = (struct sockaddr *)&gateway;
178	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
179	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
180	rt_mask(&rt) = (struct sockaddr *)&mask;
181	rt_key(&rt) = (struct sockaddr *)&addr;
182	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
183	/* Announce arrival of local address to all FIBs. */
184	rt_newaddrmsg(RTM_ADD, ifa, 0, &rt);
185}
186
187void
188in6_ifremloop(struct ifaddr *ifa)
189{
190	struct sockaddr_dl gateway;
191	struct sockaddr_in6 mask, addr;
192	struct rtentry rt0;
193	struct in6_ifaddr *ia;
194	struct ifnet *ifp;
195
196	ia = ifa2ia6(ifa);
197	ifp = ifa->ifa_ifp;
198	IF_AFDATA_LOCK(ifp);
199	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
200	    (struct sockaddr *)&ia->ia_addr);
201	IF_AFDATA_UNLOCK(ifp);
202
203	/*
204	 * initialize for rtmsg generation
205	 */
206	bzero(&gateway, sizeof(gateway));
207	gateway.sdl_len = sizeof(gateway);
208	gateway.sdl_family = AF_LINK;
209	gateway.sdl_nlen = 0;
210	gateway.sdl_alen = ifp->if_addrlen;
211	bzero(&rt0, sizeof(rt0));
212	rt0.rt_gateway = (struct sockaddr *)&gateway;
213	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
214	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
215	rt_mask(&rt0) = (struct sockaddr *)&mask;
216	rt_key(&rt0) = (struct sockaddr *)&addr;
217	rt0.rt_flags = RTF_HOST | RTF_STATIC;
218	/* Announce removal of local address to all FIBs. */
219	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
220}
221
222int
223in6_mask2len(struct in6_addr *mask, u_char *lim0)
224{
225	int x = 0, y;
226	u_char *lim = lim0, *p;
227
228	/* ignore the scope_id part */
229	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
230		lim = (u_char *)mask + sizeof(*mask);
231	for (p = (u_char *)mask; p < lim; x++, p++) {
232		if (*p != 0xff)
233			break;
234	}
235	y = 0;
236	if (p < lim) {
237		for (y = 0; y < 8; y++) {
238			if ((*p & (0x80 >> y)) == 0)
239				break;
240		}
241	}
242
243	/*
244	 * when the limit pointer is given, do a stricter check on the
245	 * remaining bits.
246	 */
247	if (p < lim) {
248		if (y != 0 && (*p & (0x00ff >> y)) != 0)
249			return (-1);
250		for (p = p + 1; p < lim; p++)
251			if (*p != 0)
252				return (-1);
253	}
254
255	return x * 8 + y;
256}
257
258#ifdef COMPAT_FREEBSD32
259struct in6_ndifreq32 {
260	char ifname[IFNAMSIZ];
261	uint32_t ifindex;
262};
263#define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
264#endif
265
266int
267in6_control(struct socket *so, u_long cmd, caddr_t data,
268    struct ifnet *ifp, struct thread *td)
269{
270	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
271	struct	in6_ifaddr *ia = NULL;
272	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
273	struct sockaddr_in6 *sa6;
274	int carp_attached = 0;
275	int error;
276	u_long ocmd = cmd;
277
278	/*
279	 * Compat to make pre-10.x ifconfig(8) operable.
280	 */
281	if (cmd == OSIOCAIFADDR_IN6)
282		cmd = SIOCAIFADDR_IN6;
283
284	switch (cmd) {
285	case SIOCGETSGCNT_IN6:
286	case SIOCGETMIFCNT_IN6:
287		/*
288		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
289		 * We cannot see how that would be needed, so do not adjust the
290		 * KPI blindly; more likely should clean up the IPv4 variant.
291		 */
292		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
293	}
294
295	switch(cmd) {
296	case SIOCAADDRCTL_POLICY:
297	case SIOCDADDRCTL_POLICY:
298		if (td != NULL) {
299			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
300			if (error)
301				return (error);
302		}
303		return (in6_src_ioctl(cmd, data));
304	}
305
306	if (ifp == NULL)
307		return (EOPNOTSUPP);
308
309	switch (cmd) {
310	case SIOCSNDFLUSH_IN6:
311	case SIOCSPFXFLUSH_IN6:
312	case SIOCSRTRFLUSH_IN6:
313	case SIOCSDEFIFACE_IN6:
314	case SIOCSIFINFO_FLAGS:
315	case SIOCSIFINFO_IN6:
316		if (td != NULL) {
317			error = priv_check(td, PRIV_NETINET_ND6);
318			if (error)
319				return (error);
320		}
321		/* FALLTHROUGH */
322	case OSIOCGIFINFO_IN6:
323	case SIOCGIFINFO_IN6:
324	case SIOCGDRLST_IN6:
325	case SIOCGPRLST_IN6:
326	case SIOCGNBRINFO_IN6:
327	case SIOCGDEFIFACE_IN6:
328		return (nd6_ioctl(cmd, data, ifp));
329
330#ifdef COMPAT_FREEBSD32
331	case SIOCGDEFIFACE32_IN6:
332		{
333			struct in6_ndifreq ndif;
334			struct in6_ndifreq32 *ndif32;
335
336			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
337			    ifp);
338			if (error)
339				return (error);
340			ndif32 = (struct in6_ndifreq32 *)data;
341			ndif32->ifindex = ndif.ifindex;
342			return (0);
343		}
344#endif
345	}
346
347	switch (cmd) {
348	case SIOCSIFPREFIX_IN6:
349	case SIOCDIFPREFIX_IN6:
350	case SIOCAIFPREFIX_IN6:
351	case SIOCCIFPREFIX_IN6:
352	case SIOCSGIFPREFIX_IN6:
353	case SIOCGIFPREFIX_IN6:
354		log(LOG_NOTICE,
355		    "prefix ioctls are now invalidated. "
356		    "please use ifconfig.\n");
357		return (EOPNOTSUPP);
358	}
359
360	switch (cmd) {
361	case SIOCSSCOPE6:
362		if (td != NULL) {
363			error = priv_check(td, PRIV_NETINET_SCOPE6);
364			if (error)
365				return (error);
366		}
367		return (scope6_set(ifp,
368		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
369	case SIOCGSCOPE6:
370		return (scope6_get(ifp,
371		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
372	case SIOCGSCOPE6DEF:
373		return (scope6_get_default((struct scope6_id *)
374		    ifr->ifr_ifru.ifru_scope_id));
375	}
376
377	switch (cmd) {
378	case SIOCALIFADDR:
379		if (td != NULL) {
380			error = priv_check(td, PRIV_NET_ADDIFADDR);
381			if (error)
382				return (error);
383		}
384		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
385
386	case SIOCDLIFADDR:
387		if (td != NULL) {
388			error = priv_check(td, PRIV_NET_DELIFADDR);
389			if (error)
390				return (error);
391		}
392		/* FALLTHROUGH */
393	case SIOCGLIFADDR:
394		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
395	}
396
397	/*
398	 * Find address for this interface, if it exists.
399	 *
400	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
401	 * only, and used the first interface address as the target of other
402	 * operations (without checking ifra_addr).  This was because netinet
403	 * code/API assumed at most 1 interface address per interface.
404	 * Since IPv6 allows a node to assign multiple addresses
405	 * on a single interface, we almost always look and check the
406	 * presence of ifra_addr, and reject invalid ones here.
407	 * It also decreases duplicated code among SIOC*_IN6 operations.
408	 */
409	switch (cmd) {
410	case SIOCAIFADDR_IN6:
411	case SIOCSIFPHYADDR_IN6:
412		sa6 = &ifra->ifra_addr;
413		break;
414	case SIOCSIFADDR_IN6:
415	case SIOCGIFADDR_IN6:
416	case SIOCSIFDSTADDR_IN6:
417	case SIOCSIFNETMASK_IN6:
418	case SIOCGIFDSTADDR_IN6:
419	case SIOCGIFNETMASK_IN6:
420	case SIOCDIFADDR_IN6:
421	case SIOCGIFPSRCADDR_IN6:
422	case SIOCGIFPDSTADDR_IN6:
423	case SIOCGIFAFLAG_IN6:
424	case SIOCSNDFLUSH_IN6:
425	case SIOCSPFXFLUSH_IN6:
426	case SIOCSRTRFLUSH_IN6:
427	case SIOCGIFALIFETIME_IN6:
428	case SIOCSIFALIFETIME_IN6:
429	case SIOCGIFSTAT_IN6:
430	case SIOCGIFSTAT_ICMP6:
431		sa6 = &ifr->ifr_addr;
432		break;
433	default:
434		sa6 = NULL;
435		break;
436	}
437	if (sa6 && sa6->sin6_family == AF_INET6) {
438		if (sa6->sin6_scope_id != 0)
439			error = sa6_embedscope(sa6, 0);
440		else
441			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
442		if (error != 0)
443			return (error);
444		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
445		    &sa6->sin6_addr)) != 0)
446			return (error);
447		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
448	} else
449		ia = NULL;
450
451	switch (cmd) {
452	case SIOCSIFADDR_IN6:
453	case SIOCSIFDSTADDR_IN6:
454	case SIOCSIFNETMASK_IN6:
455		/*
456		 * Since IPv6 allows a node to assign multiple addresses
457		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
458		 */
459		/* we decided to obsolete this command (20000704) */
460		error = EINVAL;
461		goto out;
462
463	case SIOCDIFADDR_IN6:
464		/*
465		 * for IPv4, we look for existing in_ifaddr here to allow
466		 * "ifconfig if0 delete" to remove the first IPv4 address on
467		 * the interface.  For IPv6, as the spec allows multiple
468		 * interface address from the day one, we consider "remove the
469		 * first one" semantics to be not preferable.
470		 */
471		if (ia == NULL) {
472			error = EADDRNOTAVAIL;
473			goto out;
474		}
475		/* FALLTHROUGH */
476	case SIOCAIFADDR_IN6:
477		/*
478		 * We always require users to specify a valid IPv6 address for
479		 * the corresponding operation.
480		 */
481		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
482		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
483			error = EAFNOSUPPORT;
484			goto out;
485		}
486
487		if (td != NULL) {
488			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
489			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
490			if (error)
491				goto out;
492		}
493		break;
494
495	case SIOCGIFADDR_IN6:
496		/* This interface is basically deprecated. use SIOCGIFCONF. */
497		/* FALLTHROUGH */
498	case SIOCGIFAFLAG_IN6:
499	case SIOCGIFNETMASK_IN6:
500	case SIOCGIFDSTADDR_IN6:
501	case SIOCGIFALIFETIME_IN6:
502		/* must think again about its semantics */
503		if (ia == NULL) {
504			error = EADDRNOTAVAIL;
505			goto out;
506		}
507		break;
508
509	case SIOCSIFALIFETIME_IN6:
510	    {
511		struct in6_addrlifetime *lt;
512
513		if (td != NULL) {
514			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
515			if (error)
516				goto out;
517		}
518		if (ia == NULL) {
519			error = EADDRNOTAVAIL;
520			goto out;
521		}
522		/* sanity for overflow - beware unsigned */
523		lt = &ifr->ifr_ifru.ifru_lifetime;
524		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
525		    lt->ia6t_vltime + time_second < time_second) {
526			error = EINVAL;
527			goto out;
528		}
529		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
530		    lt->ia6t_pltime + time_second < time_second) {
531			error = EINVAL;
532			goto out;
533		}
534		break;
535	    }
536	}
537
538	switch (cmd) {
539	case SIOCGIFADDR_IN6:
540		ifr->ifr_addr = ia->ia_addr;
541		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
542			goto out;
543		break;
544
545	case SIOCGIFDSTADDR_IN6:
546		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
547			error = EINVAL;
548			goto out;
549		}
550		/*
551		 * XXX: should we check if ifa_dstaddr is NULL and return
552		 * an error?
553		 */
554		ifr->ifr_dstaddr = ia->ia_dstaddr;
555		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
556			goto out;
557		break;
558
559	case SIOCGIFNETMASK_IN6:
560		ifr->ifr_addr = ia->ia_prefixmask;
561		break;
562
563	case SIOCGIFAFLAG_IN6:
564		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
565		break;
566
567	case SIOCGIFSTAT_IN6:
568		if (ifp == NULL) {
569			error = EINVAL;
570			goto out;
571		}
572		bzero(&ifr->ifr_ifru.ifru_stat,
573		    sizeof(ifr->ifr_ifru.ifru_stat));
574		ifr->ifr_ifru.ifru_stat =
575		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
576		break;
577
578	case SIOCGIFSTAT_ICMP6:
579		if (ifp == NULL) {
580			error = EINVAL;
581			goto out;
582		}
583		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
584		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
585		ifr->ifr_ifru.ifru_icmp6stat =
586		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
587		break;
588
589	case SIOCGIFALIFETIME_IN6:
590		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
591		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
592			time_t maxexpire;
593			struct in6_addrlifetime *retlt =
594			    &ifr->ifr_ifru.ifru_lifetime;
595
596			/*
597			 * XXX: adjust expiration time assuming time_t is
598			 * signed.
599			 */
600			maxexpire = (-1) &
601			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
602			if (ia->ia6_lifetime.ia6t_vltime <
603			    maxexpire - ia->ia6_updatetime) {
604				retlt->ia6t_expire = ia->ia6_updatetime +
605				    ia->ia6_lifetime.ia6t_vltime;
606			} else
607				retlt->ia6t_expire = maxexpire;
608		}
609		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
610			time_t maxexpire;
611			struct in6_addrlifetime *retlt =
612			    &ifr->ifr_ifru.ifru_lifetime;
613
614			/*
615			 * XXX: adjust expiration time assuming time_t is
616			 * signed.
617			 */
618			maxexpire = (-1) &
619			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
620			if (ia->ia6_lifetime.ia6t_pltime <
621			    maxexpire - ia->ia6_updatetime) {
622				retlt->ia6t_preferred = ia->ia6_updatetime +
623				    ia->ia6_lifetime.ia6t_pltime;
624			} else
625				retlt->ia6t_preferred = maxexpire;
626		}
627		break;
628
629	case SIOCSIFALIFETIME_IN6:
630		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
631		/* for sanity */
632		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
633			ia->ia6_lifetime.ia6t_expire =
634				time_second + ia->ia6_lifetime.ia6t_vltime;
635		} else
636			ia->ia6_lifetime.ia6t_expire = 0;
637		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
638			ia->ia6_lifetime.ia6t_preferred =
639				time_second + ia->ia6_lifetime.ia6t_pltime;
640		} else
641			ia->ia6_lifetime.ia6t_preferred = 0;
642		break;
643
644	case SIOCAIFADDR_IN6:
645	{
646		int i;
647		struct nd_prefixctl pr0;
648		struct nd_prefix *pr;
649
650		/*
651		 * first, make or update the interface address structure,
652		 * and link it to the list.
653		 */
654		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
655			goto out;
656		if (ia != NULL)
657			ifa_free(&ia->ia_ifa);
658		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
659		    == NULL) {
660			/*
661			 * this can happen when the user specify the 0 valid
662			 * lifetime.
663			 */
664			break;
665		}
666
667		if (cmd == ocmd && ifra->ifra_vhid > 0) {
668			if (carp_attach_p != NULL)
669				error = (*carp_attach_p)(&ia->ia_ifa,
670				    ifra->ifra_vhid);
671			else
672				error = EPROTONOSUPPORT;
673			if (error)
674				goto out;
675			else
676				carp_attached = 1;
677		}
678
679		/*
680		 * then, make the prefix on-link on the interface.
681		 * XXX: we'd rather create the prefix before the address, but
682		 * we need at least one address to install the corresponding
683		 * interface route, so we configure the address first.
684		 */
685
686		/*
687		 * convert mask to prefix length (prefixmask has already
688		 * been validated in in6_update_ifa().
689		 */
690		bzero(&pr0, sizeof(pr0));
691		pr0.ndpr_ifp = ifp;
692		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
693		    NULL);
694		if (pr0.ndpr_plen == 128) {
695			break;	/* we don't need to install a host route. */
696		}
697		pr0.ndpr_prefix = ifra->ifra_addr;
698		/* apply the mask for safety. */
699		for (i = 0; i < 4; i++) {
700			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
701			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
702		}
703		/*
704		 * XXX: since we don't have an API to set prefix (not address)
705		 * lifetimes, we just use the same lifetimes as addresses.
706		 * The (temporarily) installed lifetimes can be overridden by
707		 * later advertised RAs (when accept_rtadv is non 0), which is
708		 * an intended behavior.
709		 */
710		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
711		pr0.ndpr_raf_auto =
712		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
713		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
714		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
715
716		/* add the prefix if not yet. */
717		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
718			/*
719			 * nd6_prelist_add will install the corresponding
720			 * interface route.
721			 */
722			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
723				if (carp_attached)
724					(*carp_detach_p)(&ia->ia_ifa);
725				goto out;
726			}
727			if (pr == NULL) {
728				if (carp_attached)
729					(*carp_detach_p)(&ia->ia_ifa);
730				log(LOG_ERR, "nd6_prelist_add succeeded but "
731				    "no prefix\n");
732				error = EINVAL;
733				goto out;
734			}
735		}
736
737		/* relate the address to the prefix */
738		if (ia->ia6_ndpr == NULL) {
739			ia->ia6_ndpr = pr;
740			pr->ndpr_refcnt++;
741
742			/*
743			 * If this is the first autoconf address from the
744			 * prefix, create a temporary address as well
745			 * (when required).
746			 */
747			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
748			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
749				int e;
750				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
751					log(LOG_NOTICE, "in6_control: failed "
752					    "to create a temporary address, "
753					    "errno=%d\n", e);
754				}
755			}
756		}
757
758		/*
759		 * this might affect the status of autoconfigured addresses,
760		 * that is, this address might make other addresses detached.
761		 */
762		pfxlist_onlink_check();
763		if (error == 0 && ia) {
764			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
765				/*
766				 * Try to clear the flag when a new
767				 * IPv6 address is added onto an
768				 * IFDISABLED interface and it
769				 * succeeds.
770				 */
771				struct in6_ndireq nd;
772
773				memset(&nd, 0, sizeof(nd));
774				nd.ndi.flags = ND_IFINFO(ifp)->flags;
775				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
776				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
777				    (caddr_t)&nd, ifp) < 0)
778					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
779					    "SIOCSIFINFO_FLAGS for -ifdisabled "
780					    "failed.");
781				/*
782				 * Ignore failure of clearing the flag
783				 * intentionally.  The failure means
784				 * address duplication was detected.
785				 */
786			}
787			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
788		}
789		break;
790	}
791
792	case SIOCDIFADDR_IN6:
793	{
794		struct nd_prefix *pr;
795
796		/*
797		 * If the address being deleted is the only one that owns
798		 * the corresponding prefix, expire the prefix as well.
799		 * XXX: theoretically, we don't have to worry about such
800		 * relationship, since we separate the address management
801		 * and the prefix management.  We do this, however, to provide
802		 * as much backward compatibility as possible in terms of
803		 * the ioctl operation.
804		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
805		 */
806		pr = ia->ia6_ndpr;
807		in6_purgeaddr(&ia->ia_ifa);
808		if (pr && pr->ndpr_refcnt == 0)
809			prelist_remove(pr);
810		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
811		break;
812	}
813
814	default:
815		if (ifp == NULL || ifp->if_ioctl == 0) {
816			error = EOPNOTSUPP;
817			goto out;
818		}
819		error = (*ifp->if_ioctl)(ifp, cmd, data);
820		goto out;
821	}
822
823	error = 0;
824out:
825	if (ia != NULL)
826		ifa_free(&ia->ia_ifa);
827	return (error);
828}
829
830
831/*
832 * Join necessary multicast groups.  Factored out from in6_update_ifa().
833 * This entire work should only be done once, for the default FIB.
834 */
835static int
836in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
837    struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
838{
839	char ip6buf[INET6_ADDRSTRLEN];
840	struct sockaddr_in6 mltaddr, mltmask;
841	struct in6_addr llsol;
842	struct in6_multi_mship *imm;
843	struct rtentry *rt;
844	int delay, error;
845
846	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
847
848	/* Join solicited multicast addr for new host id. */
849	bzero(&llsol, sizeof(struct in6_addr));
850	llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
851	llsol.s6_addr32[1] = 0;
852	llsol.s6_addr32[2] = htonl(1);
853	llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
854	llsol.s6_addr8[12] = 0xff;
855	if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
856		/* XXX: should not happen */
857		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
858		goto cleanup;
859	}
860	delay = 0;
861	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
862		/*
863		 * We need a random delay for DAD on the address being
864		 * configured.  It also means delaying transmission of the
865		 * corresponding MLD report to avoid report collision.
866		 * [RFC 4861, Section 6.3.7]
867		 */
868		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
869	}
870	imm = in6_joingroup(ifp, &llsol, &error, delay);
871	if (imm == NULL) {
872		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
873		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &llsol),
874		    if_name(ifp), error));
875		goto cleanup;
876	}
877	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
878	*in6m_sol = imm->i6mm_maddr;
879
880	bzero(&mltmask, sizeof(mltmask));
881	mltmask.sin6_len = sizeof(struct sockaddr_in6);
882	mltmask.sin6_family = AF_INET6;
883	mltmask.sin6_addr = in6mask32;
884#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
885
886	/*
887	 * Join link-local all-nodes address.
888	 */
889	bzero(&mltaddr, sizeof(mltaddr));
890	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
891	mltaddr.sin6_family = AF_INET6;
892	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
893	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
894		goto cleanup; /* XXX: should not fail */
895
896	/*
897	 * XXX: do we really need this automatic routes?  We should probably
898	 * reconsider this stuff.  Most applications actually do not need the
899	 * routes, since they usually specify the outgoing interface.
900	 */
901	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
902	if (rt != NULL) {
903		/* XXX: only works in !SCOPEDROUTING case. */
904		if (memcmp(&mltaddr.sin6_addr,
905		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
906		    MLTMASK_LEN)) {
907			RTFREE_LOCKED(rt);
908			rt = NULL;
909		}
910	}
911	if (rt == NULL) {
912		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
913		    (struct sockaddr *)&ia->ia_addr,
914		    (struct sockaddr *)&mltmask, RTF_UP,
915		    (struct rtentry **)0, RT_DEFAULT_FIB);
916		if (error)
917			goto cleanup;
918	} else
919		RTFREE_LOCKED(rt);
920
921	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
922	if (imm == NULL) {
923		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
924		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
925		    &mltaddr.sin6_addr), if_name(ifp), error));
926		goto cleanup;
927	}
928	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
929
930	/*
931	 * Join node information group address.
932	 */
933	delay = 0;
934	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
935		/*
936		 * The spec does not say anything about delay for this group,
937		 * but the same logic should apply.
938		 */
939		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
940	}
941	if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
942		/* XXX jinmei */
943		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
944		if (imm == NULL)
945			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
946			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
947			    &mltaddr.sin6_addr), if_name(ifp), error));
948			/* XXX not very fatal, go on... */
949		else
950			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
951	}
952
953	/*
954	 * Join interface-local all-nodes address.
955	 * (ff01::1%ifN, and ff01::%ifN/32)
956	 */
957	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
958	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
959		goto cleanup; /* XXX: should not fail */
960	/* XXX: again, do we really need the route? */
961	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
962	if (rt != NULL) {
963		if (memcmp(&mltaddr.sin6_addr,
964		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
965		    MLTMASK_LEN)) {
966			RTFREE_LOCKED(rt);
967			rt = NULL;
968		}
969	}
970	if (rt == NULL) {
971		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
972		    (struct sockaddr *)&ia->ia_addr,
973		    (struct sockaddr *)&mltmask, RTF_UP,
974		    (struct rtentry **)0, RT_DEFAULT_FIB);
975		if (error)
976			goto cleanup;
977	} else
978		RTFREE_LOCKED(rt);
979
980	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
981	if (imm == NULL) {
982		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
983		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
984		    &mltaddr.sin6_addr), if_name(ifp), error));
985		goto cleanup;
986	}
987	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
988#undef	MLTMASK_LEN
989
990cleanup:
991	return (error);
992}
993
994/*
995 * Update parameters of an IPv6 interface address.
996 * If necessary, a new entry is created and linked into address chains.
997 * This function is separated from in6_control().
998 * XXX: should this be performed under splnet()?
999 */
1000int
1001in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1002    struct in6_ifaddr *ia, int flags)
1003{
1004	int error = 0, hostIsNew = 0, plen = -1;
1005	struct sockaddr_in6 dst6;
1006	struct in6_addrlifetime *lt;
1007	struct in6_multi *in6m_sol;
1008	int delay;
1009	char ip6buf[INET6_ADDRSTRLEN];
1010
1011	/* Validate parameters */
1012	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1013		return (EINVAL);
1014
1015	/*
1016	 * The destination address for a p2p link must have a family
1017	 * of AF_UNSPEC or AF_INET6.
1018	 */
1019	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1020	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1021	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1022		return (EAFNOSUPPORT);
1023	/*
1024	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1025	 * does not carry fields other than sin6_len.
1026	 */
1027	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1028		return (EINVAL);
1029	/*
1030	 * Because the IPv6 address architecture is classless, we require
1031	 * users to specify a (non 0) prefix length (mask) for a new address.
1032	 * We also require the prefix (when specified) mask is valid, and thus
1033	 * reject a non-consecutive mask.
1034	 */
1035	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1036		return (EINVAL);
1037	if (ifra->ifra_prefixmask.sin6_len != 0) {
1038		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1039		    (u_char *)&ifra->ifra_prefixmask +
1040		    ifra->ifra_prefixmask.sin6_len);
1041		if (plen <= 0)
1042			return (EINVAL);
1043	} else {
1044		/*
1045		 * In this case, ia must not be NULL.  We just use its prefix
1046		 * length.
1047		 */
1048		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1049	}
1050	/*
1051	 * If the destination address on a p2p interface is specified,
1052	 * and the address is a scoped one, validate/set the scope
1053	 * zone identifier.
1054	 */
1055	dst6 = ifra->ifra_dstaddr;
1056	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1057	    (dst6.sin6_family == AF_INET6)) {
1058		struct in6_addr in6_tmp;
1059		u_int32_t zoneid;
1060
1061		in6_tmp = dst6.sin6_addr;
1062		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1063			return (EINVAL); /* XXX: should be impossible */
1064
1065		if (dst6.sin6_scope_id != 0) {
1066			if (dst6.sin6_scope_id != zoneid)
1067				return (EINVAL);
1068		} else		/* user omit to specify the ID. */
1069			dst6.sin6_scope_id = zoneid;
1070
1071		/* convert into the internal form */
1072		if (sa6_embedscope(&dst6, 0))
1073			return (EINVAL); /* XXX: should be impossible */
1074	}
1075	/*
1076	 * The destination address can be specified only for a p2p or a
1077	 * loopback interface.  If specified, the corresponding prefix length
1078	 * must be 128.
1079	 */
1080	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1081		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1082			/* XXX: noisy message */
1083			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1084			    "be specified for a p2p or a loopback IF only\n"));
1085			return (EINVAL);
1086		}
1087		if (plen != 128) {
1088			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1089			    "be 128 when dstaddr is specified\n"));
1090			return (EINVAL);
1091		}
1092	}
1093	/* lifetime consistency check */
1094	lt = &ifra->ifra_lifetime;
1095	if (lt->ia6t_pltime > lt->ia6t_vltime)
1096		return (EINVAL);
1097	if (lt->ia6t_vltime == 0) {
1098		/*
1099		 * the following log might be noisy, but this is a typical
1100		 * configuration mistake or a tool's bug.
1101		 */
1102		nd6log((LOG_INFO,
1103		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1104		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1105
1106		if (ia == NULL)
1107			return (0); /* there's nothing to do */
1108	}
1109
1110	/*
1111	 * If this is a new address, allocate a new ifaddr and link it
1112	 * into chains.
1113	 */
1114	if (ia == NULL) {
1115		hostIsNew = 1;
1116		/*
1117		 * When in6_update_ifa() is called in a process of a received
1118		 * RA, it is called under an interrupt context.  So, we should
1119		 * call malloc with M_NOWAIT.
1120		 */
1121		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
1122		    M_NOWAIT);
1123		if (ia == NULL)
1124			return (ENOBUFS);
1125		bzero((caddr_t)ia, sizeof(*ia));
1126		ifa_init(&ia->ia_ifa);
1127		LIST_INIT(&ia->ia6_memberships);
1128		/* Initialize the address and masks, and put time stamp */
1129		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1130		ia->ia_addr.sin6_family = AF_INET6;
1131		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1132		ia->ia6_createtime = time_second;
1133		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1134			/*
1135			 * XXX: some functions expect that ifa_dstaddr is not
1136			 * NULL for p2p interfaces.
1137			 */
1138			ia->ia_ifa.ifa_dstaddr =
1139			    (struct sockaddr *)&ia->ia_dstaddr;
1140		} else {
1141			ia->ia_ifa.ifa_dstaddr = NULL;
1142		}
1143		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1144		ia->ia_ifp = ifp;
1145		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1146		IF_ADDR_WLOCK(ifp);
1147		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1148		IF_ADDR_WUNLOCK(ifp);
1149
1150		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1151		IN6_IFADDR_WLOCK();
1152		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1153		IN6_IFADDR_WUNLOCK();
1154	}
1155
1156	/* update timestamp */
1157	ia->ia6_updatetime = time_second;
1158
1159	/* set prefix mask */
1160	if (ifra->ifra_prefixmask.sin6_len) {
1161		/*
1162		 * We prohibit changing the prefix length of an existing
1163		 * address, because
1164		 * + such an operation should be rare in IPv6, and
1165		 * + the operation would confuse prefix management.
1166		 */
1167		if (ia->ia_prefixmask.sin6_len &&
1168		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1169			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1170			    " existing (%s) address should not be changed\n",
1171			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1172			error = EINVAL;
1173			goto unlink;
1174		}
1175		ia->ia_prefixmask = ifra->ifra_prefixmask;
1176	}
1177
1178	/*
1179	 * If a new destination address is specified, scrub the old one and
1180	 * install the new destination.  Note that the interface must be
1181	 * p2p or loopback (see the check above.)
1182	 */
1183	if (dst6.sin6_family == AF_INET6 &&
1184	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1185		int e;
1186
1187		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1188		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1189			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1190			    "a route to the old destination: %s\n",
1191			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1192			/* proceed anyway... */
1193		} else
1194			ia->ia_flags &= ~IFA_ROUTE;
1195		ia->ia_dstaddr = dst6;
1196	}
1197
1198	/*
1199	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1200	 * to see if the address is deprecated or invalidated, but initialize
1201	 * these members for applications.
1202	 */
1203	ia->ia6_lifetime = ifra->ifra_lifetime;
1204	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1205		ia->ia6_lifetime.ia6t_expire =
1206		    time_second + ia->ia6_lifetime.ia6t_vltime;
1207	} else
1208		ia->ia6_lifetime.ia6t_expire = 0;
1209	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1210		ia->ia6_lifetime.ia6t_preferred =
1211		    time_second + ia->ia6_lifetime.ia6t_pltime;
1212	} else
1213		ia->ia6_lifetime.ia6t_preferred = 0;
1214
1215	/* reset the interface and routing table appropriately. */
1216	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1217		goto unlink;
1218
1219	/*
1220	 * configure address flags.
1221	 */
1222	ia->ia6_flags = ifra->ifra_flags;
1223	/*
1224	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1225	 * userland, make it deprecated.
1226	 */
1227	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1228		ia->ia6_lifetime.ia6t_pltime = 0;
1229		ia->ia6_lifetime.ia6t_preferred = time_second;
1230	}
1231	/*
1232	 * Make the address tentative before joining multicast addresses,
1233	 * so that corresponding MLD responses would not have a tentative
1234	 * source address.
1235	 */
1236	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1237	if (hostIsNew && in6if_do_dad(ifp))
1238		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1239
1240	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
1241	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1242		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1243
1244	/*
1245	 * We are done if we have simply modified an existing address.
1246	 */
1247	if (!hostIsNew)
1248		return (error);
1249
1250	/*
1251	 * Beyond this point, we should call in6_purgeaddr upon an error,
1252	 * not just go to unlink.
1253	 */
1254
1255	/* Join necessary multicast groups. */
1256	in6m_sol = NULL;
1257	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1258		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1259		if (error)
1260			goto cleanup;
1261	}
1262
1263	/*
1264	 * Perform DAD, if needed.
1265	 * XXX It may be of use, if we can administratively disable DAD.
1266	 */
1267	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1268	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1269	{
1270		int mindelay, maxdelay;
1271
1272		delay = 0;
1273		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1274			/*
1275			 * We need to impose a delay before sending an NS
1276			 * for DAD.  Check if we also needed a delay for the
1277			 * corresponding MLD message.  If we did, the delay
1278			 * should be larger than the MLD delay (this could be
1279			 * relaxed a bit, but this simple logic is at least
1280			 * safe).
1281			 * XXX: Break data hiding guidelines and look at
1282			 * state for the solicited multicast group.
1283			 */
1284			mindelay = 0;
1285			if (in6m_sol != NULL &&
1286			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1287				mindelay = in6m_sol->in6m_timer;
1288			}
1289			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1290			if (maxdelay - mindelay == 0)
1291				delay = 0;
1292			else {
1293				delay =
1294				    (arc4random() % (maxdelay - mindelay)) +
1295				    mindelay;
1296			}
1297		}
1298		nd6_dad_start((struct ifaddr *)ia, delay);
1299	}
1300
1301	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1302	ifa_free(&ia->ia_ifa);
1303	return (error);
1304
1305  unlink:
1306	/*
1307	 * XXX: if a change of an existing address failed, keep the entry
1308	 * anyway.
1309	 */
1310	if (hostIsNew) {
1311		in6_unlink_ifa(ia, ifp);
1312		ifa_free(&ia->ia_ifa);
1313	}
1314	return (error);
1315
1316  cleanup:
1317	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1318	ifa_free(&ia->ia_ifa);
1319	in6_purgeaddr(&ia->ia_ifa);
1320	return error;
1321}
1322
1323/*
1324 * Leave multicast groups.  Factored out from in6_purgeaddr().
1325 * This entire work should only be done once, for the default FIB.
1326 */
1327static int
1328in6_purgeaddr_mc(struct ifnet *ifp, struct in6_ifaddr *ia, struct ifaddr *ifa0)
1329{
1330	struct sockaddr_in6 mltaddr, mltmask;
1331	struct in6_multi_mship *imm;
1332	struct rtentry *rt;
1333	struct sockaddr_in6 sin6;
1334	int error;
1335
1336	/*
1337	 * Leave from multicast groups we have joined for the interface.
1338	 */
1339	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1340		LIST_REMOVE(imm, i6mm_chain);
1341		in6_leavegroup(imm);
1342	}
1343
1344	/*
1345	 * Remove the link-local all-nodes address.
1346	 */
1347	bzero(&mltmask, sizeof(mltmask));
1348	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1349	mltmask.sin6_family = AF_INET6;
1350	mltmask.sin6_addr = in6mask32;
1351
1352	bzero(&mltaddr, sizeof(mltaddr));
1353	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1354	mltaddr.sin6_family = AF_INET6;
1355	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1356
1357	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1358		return (error);
1359
1360	/*
1361	 * As for the mltaddr above, proactively prepare the sin6 to avoid
1362	 * rtentry un- and re-locking.
1363	 */
1364	if (ifa0 != NULL) {
1365		bzero(&sin6, sizeof(sin6));
1366		sin6.sin6_len = sizeof(sin6);
1367		sin6.sin6_family = AF_INET6;
1368		memcpy(&sin6.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1369		    sizeof(sin6.sin6_addr));
1370		error = in6_setscope(&sin6.sin6_addr, ifa0->ifa_ifp, NULL);
1371		if (error != 0)
1372			return (error);
1373	}
1374
1375	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1376	if (rt != NULL && rt->rt_gateway != NULL &&
1377	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1378		    &ia->ia_addr.sin6_addr,
1379		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1380		/*
1381		 * If no more IPv6 address exists on this interface then
1382		 * remove the multicast address route.
1383		 */
1384		if (ifa0 == NULL) {
1385			memcpy(&mltaddr.sin6_addr,
1386			    &satosin6(rt_key(rt))->sin6_addr,
1387			    sizeof(mltaddr.sin6_addr));
1388			RTFREE_LOCKED(rt);
1389			error = in6_rtrequest(RTM_DELETE,
1390			    (struct sockaddr *)&mltaddr,
1391			    (struct sockaddr *)&ia->ia_addr,
1392			    (struct sockaddr *)&mltmask, RTF_UP,
1393			    (struct rtentry **)0, RT_DEFAULT_FIB);
1394			if (error)
1395				log(LOG_INFO, "%s: link-local all-nodes "
1396				    "multicast address deletion error\n",
1397				    __func__);
1398		} else {
1399			/*
1400			 * Replace the gateway of the route.
1401			 */
1402			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1403			RTFREE_LOCKED(rt);
1404		}
1405	} else {
1406		if (rt != NULL)
1407			RTFREE_LOCKED(rt);
1408	}
1409
1410	/*
1411	 * Remove the node-local all-nodes address.
1412	 */
1413	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1414	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1415		return (error);
1416
1417	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1418	if (rt != NULL && rt->rt_gateway != NULL &&
1419	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1420		    &ia->ia_addr.sin6_addr,
1421		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1422		/*
1423		 * If no more IPv6 address exists on this interface then
1424		 * remove the multicast address route.
1425		 */
1426		if (ifa0 == NULL) {
1427			memcpy(&mltaddr.sin6_addr,
1428			    &satosin6(rt_key(rt))->sin6_addr,
1429			    sizeof(mltaddr.sin6_addr));
1430
1431			RTFREE_LOCKED(rt);
1432			error = in6_rtrequest(RTM_DELETE,
1433			    (struct sockaddr *)&mltaddr,
1434			    (struct sockaddr *)&ia->ia_addr,
1435			    (struct sockaddr *)&mltmask, RTF_UP,
1436			    (struct rtentry **)0, RT_DEFAULT_FIB);
1437			if (error)
1438				log(LOG_INFO, "%s: node-local all-nodes"
1439				    "multicast address deletion error\n",
1440				    __func__);
1441		} else {
1442			/*
1443			 * Replace the gateway of the route.
1444			 */
1445			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1446			RTFREE_LOCKED(rt);
1447		}
1448	} else {
1449		if (rt != NULL)
1450			RTFREE_LOCKED(rt);
1451	}
1452
1453	return (0);
1454}
1455
1456void
1457in6_purgeaddr(struct ifaddr *ifa)
1458{
1459	struct ifnet *ifp = ifa->ifa_ifp;
1460	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1461	int plen, error;
1462	struct ifaddr *ifa0;
1463
1464	if (ifa->ifa_carp)
1465		(*carp_detach_p)(ifa);
1466
1467	/*
1468	 * find another IPv6 address as the gateway for the
1469	 * link-local and node-local all-nodes multicast
1470	 * address routes
1471	 */
1472	IF_ADDR_RLOCK(ifp);
1473	TAILQ_FOREACH(ifa0, &ifp->if_addrhead, ifa_link) {
1474		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1475		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1476		    &ia->ia_addr.sin6_addr, sizeof(struct in6_addr)) == 0)
1477			continue;
1478		else
1479			break;
1480	}
1481	if (ifa0 != NULL)
1482		ifa_ref(ifa0);
1483	IF_ADDR_RUNLOCK(ifp);
1484
1485	/*
1486	 * Remove the loopback route to the interface address.
1487	 * The check for the current setting of "nd6_useloopback"
1488	 * is not needed.
1489	 */
1490	if (ia->ia_flags & IFA_RTSELF) {
1491		error = ifa_del_loopback_route((struct ifaddr *)ia,
1492		    (struct sockaddr *)&ia->ia_addr);
1493		if (error == 0)
1494			ia->ia_flags &= ~IFA_RTSELF;
1495	}
1496
1497	/* stop DAD processing */
1498	nd6_dad_stop(ifa);
1499
1500	/* Remove local address entry from lltable. */
1501	in6_ifremloop(ifa);
1502
1503	/* Leave multicast groups. */
1504	error = in6_purgeaddr_mc(ifp, ia, ifa0);
1505
1506	if (ifa0 != NULL)
1507		ifa_free(ifa0);
1508
1509	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1510	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1511		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1512		    (ia->ia_dstaddr.sin6_family == AF_INET6) ? RTF_HOST : 0);
1513		if (error != 0)
1514			log(LOG_INFO, "%s: err=%d, destination address delete "
1515			    "failed\n", __func__, error);
1516		ia->ia_flags &= ~IFA_ROUTE;
1517	}
1518
1519	in6_unlink_ifa(ia, ifp);
1520}
1521
1522static void
1523in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1524{
1525	int	s = splnet();
1526
1527	IF_ADDR_WLOCK(ifp);
1528	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1529	IF_ADDR_WUNLOCK(ifp);
1530	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1531
1532	/*
1533	 * Defer the release of what might be the last reference to the
1534	 * in6_ifaddr so that it can't be freed before the remainder of the
1535	 * cleanup.
1536	 */
1537	IN6_IFADDR_WLOCK();
1538	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1539	IN6_IFADDR_WUNLOCK();
1540
1541	/*
1542	 * Release the reference to the base prefix.  There should be a
1543	 * positive reference.
1544	 */
1545	if (ia->ia6_ndpr == NULL) {
1546		nd6log((LOG_NOTICE,
1547		    "in6_unlink_ifa: autoconf'ed address "
1548		    "%p has no prefix\n", ia));
1549	} else {
1550		ia->ia6_ndpr->ndpr_refcnt--;
1551		ia->ia6_ndpr = NULL;
1552	}
1553
1554	/*
1555	 * Also, if the address being removed is autoconf'ed, call
1556	 * pfxlist_onlink_check() since the release might affect the status of
1557	 * other (detached) addresses.
1558	 */
1559	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1560		pfxlist_onlink_check();
1561	}
1562	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1563	splx(s);
1564}
1565
1566void
1567in6_purgeif(struct ifnet *ifp)
1568{
1569	struct ifaddr *ifa, *nifa;
1570
1571	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1572		if (ifa->ifa_addr->sa_family != AF_INET6)
1573			continue;
1574		in6_purgeaddr(ifa);
1575	}
1576
1577	in6_ifdetach(ifp);
1578}
1579
1580/*
1581 * SIOC[GAD]LIFADDR.
1582 *	SIOCGLIFADDR: get first address. (?)
1583 *	SIOCGLIFADDR with IFLR_PREFIX:
1584 *		get first address that matches the specified prefix.
1585 *	SIOCALIFADDR: add the specified address.
1586 *	SIOCALIFADDR with IFLR_PREFIX:
1587 *		add the specified prefix, filling hostid part from
1588 *		the first link-local address.  prefixlen must be <= 64.
1589 *	SIOCDLIFADDR: delete the specified address.
1590 *	SIOCDLIFADDR with IFLR_PREFIX:
1591 *		delete the first address that matches the specified prefix.
1592 * return values:
1593 *	EINVAL on invalid parameters
1594 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1595 *	other values may be returned from in6_ioctl()
1596 *
1597 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1598 * this is to accomodate address naming scheme other than RFC2374,
1599 * in the future.
1600 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1601 * address encoding scheme. (see figure on page 8)
1602 */
1603static int
1604in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1605    struct ifnet *ifp, struct thread *td)
1606{
1607	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1608	struct ifaddr *ifa;
1609	struct sockaddr *sa;
1610
1611	/* sanity checks */
1612	if (!data || !ifp) {
1613		panic("invalid argument to in6_lifaddr_ioctl");
1614		/* NOTREACHED */
1615	}
1616
1617	switch (cmd) {
1618	case SIOCGLIFADDR:
1619		/* address must be specified on GET with IFLR_PREFIX */
1620		if ((iflr->flags & IFLR_PREFIX) == 0)
1621			break;
1622		/* FALLTHROUGH */
1623	case SIOCALIFADDR:
1624	case SIOCDLIFADDR:
1625		/* address must be specified on ADD and DELETE */
1626		sa = (struct sockaddr *)&iflr->addr;
1627		if (sa->sa_family != AF_INET6)
1628			return EINVAL;
1629		if (sa->sa_len != sizeof(struct sockaddr_in6))
1630			return EINVAL;
1631		/* XXX need improvement */
1632		sa = (struct sockaddr *)&iflr->dstaddr;
1633		if (sa->sa_family && sa->sa_family != AF_INET6)
1634			return EINVAL;
1635		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1636			return EINVAL;
1637		break;
1638	default: /* shouldn't happen */
1639#if 0
1640		panic("invalid cmd to in6_lifaddr_ioctl");
1641		/* NOTREACHED */
1642#else
1643		return EOPNOTSUPP;
1644#endif
1645	}
1646	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1647		return EINVAL;
1648
1649	switch (cmd) {
1650	case SIOCALIFADDR:
1651	    {
1652		struct in6_aliasreq ifra;
1653		struct in6_addr *hostid = NULL;
1654		int prefixlen;
1655
1656		ifa = NULL;
1657		if ((iflr->flags & IFLR_PREFIX) != 0) {
1658			struct sockaddr_in6 *sin6;
1659
1660			/*
1661			 * hostid is to fill in the hostid part of the
1662			 * address.  hostid points to the first link-local
1663			 * address attached to the interface.
1664			 */
1665			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1666			if (!ifa)
1667				return EADDRNOTAVAIL;
1668			hostid = IFA_IN6(ifa);
1669
1670			/* prefixlen must be <= 64. */
1671			if (64 < iflr->prefixlen) {
1672				if (ifa != NULL)
1673					ifa_free(ifa);
1674				return EINVAL;
1675			}
1676			prefixlen = iflr->prefixlen;
1677
1678			/* hostid part must be zero. */
1679			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1680			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1681			    sin6->sin6_addr.s6_addr32[3] != 0) {
1682				if (ifa != NULL)
1683					ifa_free(ifa);
1684				return EINVAL;
1685			}
1686		} else
1687			prefixlen = iflr->prefixlen;
1688
1689		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1690		bzero(&ifra, sizeof(ifra));
1691		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1692
1693		bcopy(&iflr->addr, &ifra.ifra_addr,
1694		    ((struct sockaddr *)&iflr->addr)->sa_len);
1695		if (hostid) {
1696			/* fill in hostid part */
1697			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1698			    hostid->s6_addr32[2];
1699			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1700			    hostid->s6_addr32[3];
1701		}
1702
1703		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1704			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1705			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1706			if (hostid) {
1707				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1708				    hostid->s6_addr32[2];
1709				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1710				    hostid->s6_addr32[3];
1711			}
1712		}
1713		if (ifa != NULL)
1714			ifa_free(ifa);
1715
1716		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1717		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1718
1719		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1720		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1721	    }
1722	case SIOCGLIFADDR:
1723	case SIOCDLIFADDR:
1724	    {
1725		struct in6_ifaddr *ia;
1726		struct in6_addr mask, candidate, match;
1727		struct sockaddr_in6 *sin6;
1728		int cmp;
1729
1730		bzero(&mask, sizeof(mask));
1731		if (iflr->flags & IFLR_PREFIX) {
1732			/* lookup a prefix rather than address. */
1733			in6_prefixlen2mask(&mask, iflr->prefixlen);
1734
1735			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1736			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1737			match.s6_addr32[0] &= mask.s6_addr32[0];
1738			match.s6_addr32[1] &= mask.s6_addr32[1];
1739			match.s6_addr32[2] &= mask.s6_addr32[2];
1740			match.s6_addr32[3] &= mask.s6_addr32[3];
1741
1742			/* if you set extra bits, that's wrong */
1743			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1744				return EINVAL;
1745
1746			cmp = 1;
1747		} else {
1748			if (cmd == SIOCGLIFADDR) {
1749				/* on getting an address, take the 1st match */
1750				cmp = 0;	/* XXX */
1751			} else {
1752				/* on deleting an address, do exact match */
1753				in6_prefixlen2mask(&mask, 128);
1754				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1755				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1756
1757				cmp = 1;
1758			}
1759		}
1760
1761		IF_ADDR_RLOCK(ifp);
1762		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1763			if (ifa->ifa_addr->sa_family != AF_INET6)
1764				continue;
1765			if (!cmp)
1766				break;
1767
1768			/*
1769			 * XXX: this is adhoc, but is necessary to allow
1770			 * a user to specify fe80::/64 (not /10) for a
1771			 * link-local address.
1772			 */
1773			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1774			in6_clearscope(&candidate);
1775			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1776			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1777			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1778			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1779			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1780				break;
1781		}
1782		if (ifa != NULL)
1783			ifa_ref(ifa);
1784		IF_ADDR_RUNLOCK(ifp);
1785		if (!ifa)
1786			return EADDRNOTAVAIL;
1787		ia = ifa2ia6(ifa);
1788
1789		if (cmd == SIOCGLIFADDR) {
1790			int error;
1791
1792			/* fill in the if_laddrreq structure */
1793			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1794			error = sa6_recoverscope(
1795			    (struct sockaddr_in6 *)&iflr->addr);
1796			if (error != 0) {
1797				ifa_free(ifa);
1798				return (error);
1799			}
1800
1801			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1802				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1803				    ia->ia_dstaddr.sin6_len);
1804				error = sa6_recoverscope(
1805				    (struct sockaddr_in6 *)&iflr->dstaddr);
1806				if (error != 0) {
1807					ifa_free(ifa);
1808					return (error);
1809				}
1810			} else
1811				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1812
1813			iflr->prefixlen =
1814			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1815
1816			iflr->flags = ia->ia6_flags;	/* XXX */
1817			ifa_free(ifa);
1818
1819			return 0;
1820		} else {
1821			struct in6_aliasreq ifra;
1822
1823			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1824			bzero(&ifra, sizeof(ifra));
1825			bcopy(iflr->iflr_name, ifra.ifra_name,
1826			    sizeof(ifra.ifra_name));
1827
1828			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1829			    ia->ia_addr.sin6_len);
1830			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1831				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1832				    ia->ia_dstaddr.sin6_len);
1833			} else {
1834				bzero(&ifra.ifra_dstaddr,
1835				    sizeof(ifra.ifra_dstaddr));
1836			}
1837			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1838			    ia->ia_prefixmask.sin6_len);
1839
1840			ifra.ifra_flags = ia->ia6_flags;
1841			ifa_free(ifa);
1842			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1843			    ifp, td);
1844		}
1845	    }
1846	}
1847
1848	return EOPNOTSUPP;	/* just for safety */
1849}
1850
1851/*
1852 * Initialize an interface's IPv6 address and routing table entry.
1853 */
1854static int
1855in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1856    struct sockaddr_in6 *sin6, int newhost)
1857{
1858	int	error = 0, plen, ifacount = 0;
1859	int	s = splimp();
1860	struct ifaddr *ifa;
1861
1862	/*
1863	 * Give the interface a chance to initialize
1864	 * if this is its first address,
1865	 * and to validate the address if necessary.
1866	 */
1867	IF_ADDR_RLOCK(ifp);
1868	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1869		if (ifa->ifa_addr->sa_family != AF_INET6)
1870			continue;
1871		ifacount++;
1872	}
1873	IF_ADDR_RUNLOCK(ifp);
1874
1875	ia->ia_addr = *sin6;
1876
1877	if (ifacount <= 1 && ifp->if_ioctl) {
1878		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1879		if (error) {
1880			splx(s);
1881			return (error);
1882		}
1883	}
1884	splx(s);
1885
1886	ia->ia_ifa.ifa_metric = ifp->if_metric;
1887
1888	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1889
1890	/*
1891	 * Special case:
1892	 * If a new destination address is specified for a point-to-point
1893	 * interface, install a route to the destination as an interface
1894	 * direct route.
1895	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1896	 * interface that share the same destination.
1897	 */
1898	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1899	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1900	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1901		int rtflags = RTF_UP | RTF_HOST;
1902		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1903		if (error)
1904			return (error);
1905		ia->ia_flags |= IFA_ROUTE;
1906		/*
1907		 * Handle the case for ::1 .
1908		 */
1909		if (ifp->if_flags & IFF_LOOPBACK)
1910			ia->ia_flags |= IFA_RTSELF;
1911	}
1912
1913	/*
1914	 * add a loopback route to self
1915	 */
1916	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1917		error = ifa_add_loopback_route((struct ifaddr *)ia,
1918		    (struct sockaddr *)&ia->ia_addr);
1919		if (error == 0)
1920			ia->ia_flags |= IFA_RTSELF;
1921	}
1922
1923	/* Add local address to lltable, if necessary (ex. on p2p link). */
1924	if (newhost)
1925		in6_ifaddloop(&(ia->ia_ifa));
1926
1927	return (error);
1928}
1929
1930/*
1931 * Find an IPv6 interface link-local address specific to an interface.
1932 * ifaddr is returned referenced.
1933 */
1934struct in6_ifaddr *
1935in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1936{
1937	struct ifaddr *ifa;
1938
1939	IF_ADDR_RLOCK(ifp);
1940	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1941		if (ifa->ifa_addr->sa_family != AF_INET6)
1942			continue;
1943		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1944			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1945			    ignoreflags) != 0)
1946				continue;
1947			ifa_ref(ifa);
1948			break;
1949		}
1950	}
1951	IF_ADDR_RUNLOCK(ifp);
1952
1953	return ((struct in6_ifaddr *)ifa);
1954}
1955
1956
1957/*
1958 * find the internet address corresponding to a given interface and address.
1959 * ifaddr is returned referenced.
1960 */
1961struct in6_ifaddr *
1962in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1963{
1964	struct ifaddr *ifa;
1965
1966	IF_ADDR_RLOCK(ifp);
1967	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1968		if (ifa->ifa_addr->sa_family != AF_INET6)
1969			continue;
1970		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1971			ifa_ref(ifa);
1972			break;
1973		}
1974	}
1975	IF_ADDR_RUNLOCK(ifp);
1976
1977	return ((struct in6_ifaddr *)ifa);
1978}
1979
1980/*
1981 * Convert IP6 address to printable (loggable) representation. Caller
1982 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1983 */
1984static char digits[] = "0123456789abcdef";
1985char *
1986ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1987{
1988	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1989	char *cp;
1990	const u_int16_t *a = (const u_int16_t *)addr;
1991	const u_int8_t *d;
1992	int dcolon = 0, zero = 0;
1993
1994	cp = ip6buf;
1995
1996	for (i = 0; i < 8; i++) {
1997		if (*(a + i) == 0) {
1998			cnt++;
1999			if (cnt == 1)
2000				idx = i;
2001		}
2002		else if (maxcnt < cnt) {
2003			maxcnt = cnt;
2004			index = idx;
2005			cnt = 0;
2006		}
2007	}
2008	if (maxcnt < cnt) {
2009		maxcnt = cnt;
2010		index = idx;
2011	}
2012
2013	for (i = 0; i < 8; i++) {
2014		if (dcolon == 1) {
2015			if (*a == 0) {
2016				if (i == 7)
2017					*cp++ = ':';
2018				a++;
2019				continue;
2020			} else
2021				dcolon = 2;
2022		}
2023		if (*a == 0) {
2024			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
2025				if (i == 0)
2026					*cp++ = ':';
2027				*cp++ = ':';
2028				dcolon = 1;
2029			} else {
2030				*cp++ = '0';
2031				*cp++ = ':';
2032			}
2033			a++;
2034			continue;
2035		}
2036		d = (const u_char *)a;
2037		/* Try to eliminate leading zeros in printout like in :0001. */
2038		zero = 1;
2039		*cp = digits[*d >> 4];
2040		if (*cp != '0') {
2041			zero = 0;
2042			cp++;
2043		}
2044		*cp = digits[*d++ & 0xf];
2045		if (zero == 0 || (*cp != '0')) {
2046			zero = 0;
2047			cp++;
2048		}
2049		*cp = digits[*d >> 4];
2050		if (zero == 0 || (*cp != '0')) {
2051			zero = 0;
2052			cp++;
2053		}
2054		*cp++ = digits[*d & 0xf];
2055		*cp++ = ':';
2056		a++;
2057	}
2058	*--cp = '\0';
2059	return (ip6buf);
2060}
2061
2062int
2063in6_localaddr(struct in6_addr *in6)
2064{
2065	struct in6_ifaddr *ia;
2066
2067	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2068		return 1;
2069
2070	IN6_IFADDR_RLOCK();
2071	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2072		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2073		    &ia->ia_prefixmask.sin6_addr)) {
2074			IN6_IFADDR_RUNLOCK();
2075			return 1;
2076		}
2077	}
2078	IN6_IFADDR_RUNLOCK();
2079
2080	return (0);
2081}
2082
2083/*
2084 * Return 1 if an internet address is for the local host and configured
2085 * on one of its interfaces.
2086 */
2087int
2088in6_localip(struct in6_addr *in6)
2089{
2090	struct in6_ifaddr *ia;
2091
2092	IN6_IFADDR_RLOCK();
2093	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2094		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2095			IN6_IFADDR_RUNLOCK();
2096			return (1);
2097		}
2098	}
2099	IN6_IFADDR_RUNLOCK();
2100	return (0);
2101}
2102
2103
2104int
2105in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2106{
2107	struct in6_ifaddr *ia;
2108
2109	IN6_IFADDR_RLOCK();
2110	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2111		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2112		    &sa6->sin6_addr) &&
2113		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
2114			IN6_IFADDR_RUNLOCK();
2115			return (1); /* true */
2116		}
2117
2118		/* XXX: do we still have to go thru the rest of the list? */
2119	}
2120	IN6_IFADDR_RUNLOCK();
2121
2122	return (0);		/* false */
2123}
2124
2125/*
2126 * return length of part which dst and src are equal
2127 * hard coding...
2128 */
2129int
2130in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2131{
2132	int match = 0;
2133	u_char *s = (u_char *)src, *d = (u_char *)dst;
2134	u_char *lim = s + 16, r;
2135
2136	while (s < lim)
2137		if ((r = (*d++ ^ *s++)) != 0) {
2138			while (r < 128) {
2139				match++;
2140				r <<= 1;
2141			}
2142			break;
2143		} else
2144			match += 8;
2145	return match;
2146}
2147
2148/* XXX: to be scope conscious */
2149int
2150in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2151{
2152	int bytelen, bitlen;
2153
2154	/* sanity check */
2155	if (0 > len || len > 128) {
2156		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2157		    len);
2158		return (0);
2159	}
2160
2161	bytelen = len / 8;
2162	bitlen = len % 8;
2163
2164	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2165		return (0);
2166	if (bitlen != 0 &&
2167	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2168	    p2->s6_addr[bytelen] >> (8 - bitlen))
2169		return (0);
2170
2171	return (1);
2172}
2173
2174void
2175in6_prefixlen2mask(struct in6_addr *maskp, int len)
2176{
2177	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2178	int bytelen, bitlen, i;
2179
2180	/* sanity check */
2181	if (0 > len || len > 128) {
2182		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2183		    len);
2184		return;
2185	}
2186
2187	bzero(maskp, sizeof(*maskp));
2188	bytelen = len / 8;
2189	bitlen = len % 8;
2190	for (i = 0; i < bytelen; i++)
2191		maskp->s6_addr[i] = 0xff;
2192	if (bitlen)
2193		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2194}
2195
2196/*
2197 * return the best address out of the same scope. if no address was
2198 * found, return the first valid address from designated IF.
2199 */
2200struct in6_ifaddr *
2201in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2202{
2203	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2204	struct ifaddr *ifa;
2205	struct in6_ifaddr *besta = 0;
2206	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2207
2208	dep[0] = dep[1] = NULL;
2209
2210	/*
2211	 * We first look for addresses in the same scope.
2212	 * If there is one, return it.
2213	 * If two or more, return one which matches the dst longest.
2214	 * If none, return one of global addresses assigned other ifs.
2215	 */
2216	IF_ADDR_RLOCK(ifp);
2217	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2218		if (ifa->ifa_addr->sa_family != AF_INET6)
2219			continue;
2220		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2221			continue; /* XXX: is there any case to allow anycast? */
2222		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2223			continue; /* don't use this interface */
2224		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2225			continue;
2226		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2227			if (V_ip6_use_deprecated)
2228				dep[0] = (struct in6_ifaddr *)ifa;
2229			continue;
2230		}
2231
2232		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2233			/*
2234			 * call in6_matchlen() as few as possible
2235			 */
2236			if (besta) {
2237				if (blen == -1)
2238					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2239				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2240				if (tlen > blen) {
2241					blen = tlen;
2242					besta = (struct in6_ifaddr *)ifa;
2243				}
2244			} else
2245				besta = (struct in6_ifaddr *)ifa;
2246		}
2247	}
2248	if (besta) {
2249		ifa_ref(&besta->ia_ifa);
2250		IF_ADDR_RUNLOCK(ifp);
2251		return (besta);
2252	}
2253
2254	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2255		if (ifa->ifa_addr->sa_family != AF_INET6)
2256			continue;
2257		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2258			continue; /* XXX: is there any case to allow anycast? */
2259		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2260			continue; /* don't use this interface */
2261		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2262			continue;
2263		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2264			if (V_ip6_use_deprecated)
2265				dep[1] = (struct in6_ifaddr *)ifa;
2266			continue;
2267		}
2268
2269		if (ifa != NULL)
2270			ifa_ref(ifa);
2271		IF_ADDR_RUNLOCK(ifp);
2272		return (struct in6_ifaddr *)ifa;
2273	}
2274
2275	/* use the last-resort values, that are, deprecated addresses */
2276	if (dep[0]) {
2277		ifa_ref((struct ifaddr *)dep[0]);
2278		IF_ADDR_RUNLOCK(ifp);
2279		return dep[0];
2280	}
2281	if (dep[1]) {
2282		ifa_ref((struct ifaddr *)dep[1]);
2283		IF_ADDR_RUNLOCK(ifp);
2284		return dep[1];
2285	}
2286
2287	IF_ADDR_RUNLOCK(ifp);
2288	return NULL;
2289}
2290
2291/*
2292 * perform DAD when interface becomes IFF_UP.
2293 */
2294void
2295in6_if_up(struct ifnet *ifp)
2296{
2297	struct ifaddr *ifa;
2298	struct in6_ifaddr *ia;
2299
2300	IF_ADDR_RLOCK(ifp);
2301	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2302		if (ifa->ifa_addr->sa_family != AF_INET6)
2303			continue;
2304		ia = (struct in6_ifaddr *)ifa;
2305		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2306			/*
2307			 * The TENTATIVE flag was likely set by hand
2308			 * beforehand, implicitly indicating the need for DAD.
2309			 * We may be able to skip the random delay in this
2310			 * case, but we impose delays just in case.
2311			 */
2312			nd6_dad_start(ifa,
2313			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2314		}
2315	}
2316	IF_ADDR_RUNLOCK(ifp);
2317
2318	/*
2319	 * special cases, like 6to4, are handled in in6_ifattach
2320	 */
2321	in6_ifattach(ifp, NULL);
2322}
2323
2324int
2325in6if_do_dad(struct ifnet *ifp)
2326{
2327	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2328		return (0);
2329
2330	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2331		return (0);
2332
2333	switch (ifp->if_type) {
2334#ifdef IFT_DUMMY
2335	case IFT_DUMMY:
2336#endif
2337	case IFT_FAITH:
2338		/*
2339		 * These interfaces do not have the IFF_LOOPBACK flag,
2340		 * but loop packets back.  We do not have to do DAD on such
2341		 * interfaces.  We should even omit it, because loop-backed
2342		 * NS would confuse the DAD procedure.
2343		 */
2344		return (0);
2345	default:
2346		/*
2347		 * Our DAD routine requires the interface up and running.
2348		 * However, some interfaces can be up before the RUNNING
2349		 * status.  Additionaly, users may try to assign addresses
2350		 * before the interface becomes up (or running).
2351		 * We simply skip DAD in such a case as a work around.
2352		 * XXX: we should rather mark "tentative" on such addresses,
2353		 * and do DAD after the interface becomes ready.
2354		 */
2355		if (!((ifp->if_flags & IFF_UP) &&
2356		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2357			return (0);
2358
2359		return (1);
2360	}
2361}
2362
2363/*
2364 * Calculate max IPv6 MTU through all the interfaces and store it
2365 * to in6_maxmtu.
2366 */
2367void
2368in6_setmaxmtu(void)
2369{
2370	unsigned long maxmtu = 0;
2371	struct ifnet *ifp;
2372
2373	IFNET_RLOCK_NOSLEEP();
2374	TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
2375		/* this function can be called during ifnet initialization */
2376		if (!ifp->if_afdata[AF_INET6])
2377			continue;
2378		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2379		    IN6_LINKMTU(ifp) > maxmtu)
2380			maxmtu = IN6_LINKMTU(ifp);
2381	}
2382	IFNET_RUNLOCK_NOSLEEP();
2383	if (maxmtu)	/* update only when maxmtu is positive */
2384		V_in6_maxmtu = maxmtu;
2385}
2386
2387/*
2388 * Provide the length of interface identifiers to be used for the link attached
2389 * to the given interface.  The length should be defined in "IPv6 over
2390 * xxx-link" document.  Note that address architecture might also define
2391 * the length for a particular set of address prefixes, regardless of the
2392 * link type.  As clarified in rfc2462bis, those two definitions should be
2393 * consistent, and those really are as of August 2004.
2394 */
2395int
2396in6_if2idlen(struct ifnet *ifp)
2397{
2398	switch (ifp->if_type) {
2399	case IFT_ETHER:		/* RFC2464 */
2400#ifdef IFT_PROPVIRTUAL
2401	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2402#endif
2403#ifdef IFT_L2VLAN
2404	case IFT_L2VLAN:	/* ditto */
2405#endif
2406#ifdef IFT_IEEE80211
2407	case IFT_IEEE80211:	/* ditto */
2408#endif
2409#ifdef IFT_MIP
2410	case IFT_MIP:	/* ditto */
2411#endif
2412	case IFT_INFINIBAND:
2413		return (64);
2414	case IFT_FDDI:		/* RFC2467 */
2415		return (64);
2416	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2417		return (64);
2418	case IFT_PPP:		/* RFC2472 */
2419		return (64);
2420	case IFT_ARCNET:	/* RFC2497 */
2421		return (64);
2422	case IFT_FRELAY:	/* RFC2590 */
2423		return (64);
2424	case IFT_IEEE1394:	/* RFC3146 */
2425		return (64);
2426	case IFT_GIF:
2427		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2428	case IFT_LOOP:
2429		return (64);	/* XXX: is this really correct? */
2430	default:
2431		/*
2432		 * Unknown link type:
2433		 * It might be controversial to use the today's common constant
2434		 * of 64 for these cases unconditionally.  For full compliance,
2435		 * we should return an error in this case.  On the other hand,
2436		 * if we simply miss the standard for the link type or a new
2437		 * standard is defined for a new link type, the IFID length
2438		 * is very likely to be the common constant.  As a compromise,
2439		 * we always use the constant, but make an explicit notice
2440		 * indicating the "unknown" case.
2441		 */
2442		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2443		return (64);
2444	}
2445}
2446
2447#include <sys/sysctl.h>
2448
2449struct in6_llentry {
2450	struct llentry		base;
2451	struct sockaddr_in6	l3_addr6;
2452};
2453
2454/*
2455 * Deletes an address from the address table.
2456 * This function is called by the timer functions
2457 * such as arptimer() and nd6_llinfo_timer(), and
2458 * the caller does the locking.
2459 */
2460static void
2461in6_lltable_free(struct lltable *llt, struct llentry *lle)
2462{
2463	LLE_WUNLOCK(lle);
2464	LLE_LOCK_DESTROY(lle);
2465	free(lle, M_LLTABLE);
2466}
2467
2468static struct llentry *
2469in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2470{
2471	struct in6_llentry *lle;
2472
2473	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2474	if (lle == NULL)		/* NB: caller generates msg */
2475		return NULL;
2476
2477	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2478	lle->base.lle_refcnt = 1;
2479	lle->base.lle_free = in6_lltable_free;
2480	LLE_LOCK_INIT(&lle->base);
2481	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2482	    CALLOUT_RETURNUNLOCKED);
2483
2484	return (&lle->base);
2485}
2486
2487static void
2488in6_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix,
2489    const struct sockaddr *mask, u_int flags)
2490{
2491	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2492	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2493	struct llentry *lle, *next;
2494	int i;
2495
2496	/*
2497	 * (flags & LLE_STATIC) means deleting all entries
2498	 * including static ND6 entries.
2499	 */
2500	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2501		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2502			if (IN6_ARE_MASKED_ADDR_EQUAL(
2503				    &((struct sockaddr_in6 *)L3_ADDR(lle))->sin6_addr,
2504				    &pfx->sin6_addr,
2505				    &msk->sin6_addr) &&
2506			    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
2507				int canceled;
2508
2509				canceled = callout_drain(&lle->la_timer);
2510				LLE_WLOCK(lle);
2511				if (canceled)
2512					LLE_REMREF(lle);
2513				llentry_free(lle);
2514			}
2515		}
2516	}
2517}
2518
2519static int
2520in6_lltable_rtcheck(struct ifnet *ifp,
2521		    u_int flags,
2522		    const struct sockaddr *l3addr)
2523{
2524	struct rtentry *rt;
2525	char ip6buf[INET6_ADDRSTRLEN];
2526
2527	KASSERT(l3addr->sa_family == AF_INET6,
2528	    ("sin_family %d", l3addr->sa_family));
2529
2530	/* Our local addresses are always only installed on the default FIB. */
2531	/* XXX rtalloc1 should take a const param */
2532	rt = in6_rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0,
2533	    RT_DEFAULT_FIB);
2534	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2535		struct ifaddr *ifa;
2536		/*
2537		 * Create an ND6 cache for an IPv6 neighbor
2538		 * that is not covered by our own prefix.
2539		 */
2540		/* XXX ifaof_ifpforaddr should take a const param */
2541		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2542		if (ifa != NULL) {
2543			ifa_free(ifa);
2544			if (rt != NULL)
2545				RTFREE_LOCKED(rt);
2546			return 0;
2547		}
2548		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2549		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2550		if (rt != NULL)
2551			RTFREE_LOCKED(rt);
2552		return EINVAL;
2553	}
2554	RTFREE_LOCKED(rt);
2555	return 0;
2556}
2557
2558static struct llentry *
2559in6_lltable_lookup(struct lltable *llt, u_int flags,
2560	const struct sockaddr *l3addr)
2561{
2562	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2563	struct ifnet *ifp = llt->llt_ifp;
2564	struct llentry *lle;
2565	struct llentries *lleh;
2566	u_int hashkey;
2567
2568	IF_AFDATA_LOCK_ASSERT(ifp);
2569	KASSERT(l3addr->sa_family == AF_INET6,
2570	    ("sin_family %d", l3addr->sa_family));
2571
2572	hashkey = sin6->sin6_addr.s6_addr32[3];
2573	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2574	LIST_FOREACH(lle, lleh, lle_next) {
2575		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2576		if (lle->la_flags & LLE_DELETED)
2577			continue;
2578		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2579		    sizeof(struct in6_addr)) == 0)
2580			break;
2581	}
2582
2583	if (lle == NULL) {
2584		if (!(flags & LLE_CREATE))
2585			return (NULL);
2586		/*
2587		 * A route that covers the given address must have
2588		 * been installed 1st because we are doing a resolution,
2589		 * verify this.
2590		 */
2591		if (!(flags & LLE_IFADDR) &&
2592		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2593			return NULL;
2594
2595		lle = in6_lltable_new(l3addr, flags);
2596		if (lle == NULL) {
2597			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2598			return NULL;
2599		}
2600		lle->la_flags = flags & ~LLE_CREATE;
2601		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2602			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2603			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2604		}
2605
2606		lle->lle_tbl  = llt;
2607		lle->lle_head = lleh;
2608		LIST_INSERT_HEAD(lleh, lle, lle_next);
2609	} else if (flags & LLE_DELETE) {
2610		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2611			LLE_WLOCK(lle);
2612			lle->la_flags = LLE_DELETED;
2613			LLE_WUNLOCK(lle);
2614#ifdef DIAGNOSTIC
2615			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2616#endif
2617		}
2618		lle = (void *)-1;
2619	}
2620	if (LLE_IS_VALID(lle)) {
2621		if (flags & LLE_EXCLUSIVE)
2622			LLE_WLOCK(lle);
2623		else
2624			LLE_RLOCK(lle);
2625	}
2626	return (lle);
2627}
2628
2629static int
2630in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2631{
2632	struct ifnet *ifp = llt->llt_ifp;
2633	struct llentry *lle;
2634	/* XXX stack use */
2635	struct {
2636		struct rt_msghdr	rtm;
2637		struct sockaddr_in6	sin6;
2638		/*
2639		 * ndp.c assumes that sdl is word aligned
2640		 */
2641#ifdef __LP64__
2642		uint32_t		pad;
2643#endif
2644		struct sockaddr_dl	sdl;
2645	} ndpc;
2646	int i, error;
2647
2648	if (ifp->if_flags & IFF_LOOPBACK)
2649		return 0;
2650
2651	LLTABLE_LOCK_ASSERT();
2652
2653	error = 0;
2654	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2655		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2656			struct sockaddr_dl *sdl;
2657
2658			/* skip deleted or invalid entries */
2659			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2660				continue;
2661			/* Skip if jailed and not a valid IP of the prison. */
2662			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2663				continue;
2664			/*
2665			 * produce a msg made of:
2666			 *  struct rt_msghdr;
2667			 *  struct sockaddr_in6 (IPv6)
2668			 *  struct sockaddr_dl;
2669			 */
2670			bzero(&ndpc, sizeof(ndpc));
2671			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2672			ndpc.rtm.rtm_version = RTM_VERSION;
2673			ndpc.rtm.rtm_type = RTM_GET;
2674			ndpc.rtm.rtm_flags = RTF_UP;
2675			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2676			ndpc.sin6.sin6_family = AF_INET6;
2677			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2678			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2679
2680			/* publish */
2681			if (lle->la_flags & LLE_PUB)
2682				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2683
2684			sdl = &ndpc.sdl;
2685			sdl->sdl_family = AF_LINK;
2686			sdl->sdl_len = sizeof(*sdl);
2687			sdl->sdl_alen = ifp->if_addrlen;
2688			sdl->sdl_index = ifp->if_index;
2689			sdl->sdl_type = ifp->if_type;
2690			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2691			ndpc.rtm.rtm_rmx.rmx_expire =
2692			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2693			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2694			if (lle->la_flags & LLE_STATIC)
2695				ndpc.rtm.rtm_flags |= RTF_STATIC;
2696			ndpc.rtm.rtm_index = ifp->if_index;
2697			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2698			if (error)
2699				break;
2700		}
2701	}
2702	return error;
2703}
2704
2705void *
2706in6_domifattach(struct ifnet *ifp)
2707{
2708	struct in6_ifextra *ext;
2709
2710	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2711	bzero(ext, sizeof(*ext));
2712
2713	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2714	    M_IFADDR, M_WAITOK);
2715	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2716
2717	ext->icmp6_ifstat =
2718	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2719	    M_IFADDR, M_WAITOK);
2720	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2721
2722	ext->nd_ifinfo = nd6_ifattach(ifp);
2723	ext->scope6_id = scope6_ifattach(ifp);
2724	ext->lltable = lltable_init(ifp, AF_INET6);
2725	if (ext->lltable != NULL) {
2726		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2727		ext->lltable->llt_lookup = in6_lltable_lookup;
2728		ext->lltable->llt_dump = in6_lltable_dump;
2729	}
2730
2731	ext->mld_ifinfo = mld_domifattach(ifp);
2732
2733	return ext;
2734}
2735
2736void
2737in6_domifdetach(struct ifnet *ifp, void *aux)
2738{
2739	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2740
2741	mld_domifdetach(ifp);
2742	scope6_ifdetach(ext->scope6_id);
2743	nd6_ifdetach(ext->nd_ifinfo);
2744	lltable_free(ext->lltable);
2745	free(ext->in6_ifstat, M_IFADDR);
2746	free(ext->icmp6_ifstat, M_IFADDR);
2747	free(ext, M_IFADDR);
2748}
2749
2750/*
2751 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2752 * v4 mapped addr or v4 compat addr
2753 */
2754void
2755in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2756{
2757
2758	bzero(sin, sizeof(*sin));
2759	sin->sin_len = sizeof(struct sockaddr_in);
2760	sin->sin_family = AF_INET;
2761	sin->sin_port = sin6->sin6_port;
2762	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2763}
2764
2765/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2766void
2767in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2768{
2769	bzero(sin6, sizeof(*sin6));
2770	sin6->sin6_len = sizeof(struct sockaddr_in6);
2771	sin6->sin6_family = AF_INET6;
2772	sin6->sin6_port = sin->sin_port;
2773	sin6->sin6_addr.s6_addr32[0] = 0;
2774	sin6->sin6_addr.s6_addr32[1] = 0;
2775	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2776	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2777}
2778
2779/* Convert sockaddr_in6 into sockaddr_in. */
2780void
2781in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2782{
2783	struct sockaddr_in *sin_p;
2784	struct sockaddr_in6 sin6;
2785
2786	/*
2787	 * Save original sockaddr_in6 addr and convert it
2788	 * to sockaddr_in.
2789	 */
2790	sin6 = *(struct sockaddr_in6 *)nam;
2791	sin_p = (struct sockaddr_in *)nam;
2792	in6_sin6_2_sin(sin_p, &sin6);
2793}
2794
2795/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2796void
2797in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2798{
2799	struct sockaddr_in *sin_p;
2800	struct sockaddr_in6 *sin6_p;
2801
2802	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2803	sin_p = (struct sockaddr_in *)*nam;
2804	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2805	free(*nam, M_SONAME);
2806	*nam = (struct sockaddr *)sin6_p;
2807}
2808