in6.c revision 2ea10fad249fb6f33048faaa674b5782fd5feaae
1/*	$FreeBSD$	*/
2/*	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $	*/
3
4/*-
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33/*-
34 * Copyright (c) 1982, 1986, 1991, 1993
35 *	The Regents of the University of California.  All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 *    notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 *    notice, this list of conditions and the following disclaimer in the
44 *    documentation and/or other materials provided with the distribution.
45 * 4. Neither the name of the University nor the names of its contributors
46 *    may be used to endorse or promote products derived from this software
47 *    without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 *	@(#)in.c	8.2 (Berkeley) 11/15/93
62 */
63
64#include "opt_inet.h"
65#include "opt_inet6.h"
66
67#include <sys/param.h>
68#include <sys/errno.h>
69#include <sys/malloc.h>
70#include <sys/socket.h>
71#include <sys/socketvar.h>
72#include <sys/sockio.h>
73#include <sys/systm.h>
74#include <sys/priv.h>
75#include <sys/proc.h>
76#include <sys/time.h>
77#include <sys/kernel.h>
78#include <sys/syslog.h>
79
80#include <net/if.h>
81#include <net/if_types.h>
82#include <net/route.h>
83#include <net/if_dl.h>
84
85#include <netinet/in.h>
86#include <netinet/in_var.h>
87#include <netinet/if_ether.h>
88#include <netinet/in_systm.h>
89#include <netinet/ip.h>
90#include <netinet/in_pcb.h>
91
92#include <netinet/ip6.h>
93#include <netinet6/ip6_var.h>
94#include <netinet6/nd6.h>
95#include <netinet6/mld6_var.h>
96#include <netinet6/ip6_mroute.h>
97#include <netinet6/in6_ifattach.h>
98#include <netinet6/scope6_var.h>
99#include <netinet6/in6_pcb.h>
100
101MALLOC_DEFINE(M_IP6MADDR, "in6_multi", "internet multicast address");
102
103/*
104 * Definitions of some costant IP6 addresses.
105 */
106const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
107const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
108const struct in6_addr in6addr_nodelocal_allnodes =
109	IN6ADDR_NODELOCAL_ALLNODES_INIT;
110const struct in6_addr in6addr_linklocal_allnodes =
111	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
112const struct in6_addr in6addr_linklocal_allrouters =
113	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
114
115const struct in6_addr in6mask0 = IN6MASK0;
116const struct in6_addr in6mask32 = IN6MASK32;
117const struct in6_addr in6mask64 = IN6MASK64;
118const struct in6_addr in6mask96 = IN6MASK96;
119const struct in6_addr in6mask128 = IN6MASK128;
120
121const struct sockaddr_in6 sa6_any =
122	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
123
124static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
125	struct ifnet *, struct thread *));
126static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
127	struct sockaddr_in6 *, int));
128static void in6_unlink_ifa __P((struct in6_ifaddr *, struct ifnet *));
129
130struct in6_multihead in6_multihead;	/* XXX BSS initialization */
131int	(*faithprefix_p)(struct in6_addr *);
132
133/*
134 * Subroutine for in6_ifaddloop() and in6_ifremloop().
135 * This routine does actual work.
136 */
137static void
138in6_ifloop_request(int cmd, struct ifaddr *ifa)
139{
140	struct sockaddr_in6 all1_sa;
141	struct rtentry *nrt = NULL;
142	int e;
143	char ip6buf[INET6_ADDRSTRLEN];
144
145	bzero(&all1_sa, sizeof(all1_sa));
146	all1_sa.sin6_family = AF_INET6;
147	all1_sa.sin6_len = sizeof(struct sockaddr_in6);
148	all1_sa.sin6_addr = in6mask128;
149
150	/*
151	 * We specify the address itself as the gateway, and set the
152	 * RTF_LLINFO flag, so that the corresponding host route would have
153	 * the flag, and thus applications that assume traditional behavior
154	 * would be happy.  Note that we assume the caller of the function
155	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
156	 * which changes the outgoing interface to the loopback interface.
157	 */
158	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr,
159	    (struct sockaddr *)&all1_sa, RTF_UP|RTF_HOST|RTF_LLINFO, &nrt);
160	if (e != 0) {
161		/* XXX need more descriptive message */
162
163		log(LOG_ERR, "in6_ifloop_request: "
164		    "%s operation failed for %s (errno=%d)\n",
165		    cmd == RTM_ADD ? "ADD" : "DELETE",
166		    ip6_sprintf(ip6buf,
167			    &((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr), e);
168	}
169
170	/*
171	 * Report the addition/removal of the address to the routing socket.
172	 * XXX: since we called rtinit for a p2p interface with a destination,
173	 *      we end up reporting twice in such a case.  Should we rather
174	 *      omit the second report?
175	 */
176	if (nrt) {
177		RT_LOCK(nrt);
178		/*
179		 * Make sure rt_ifa be equal to IFA, the second argument of
180		 * the function.  We need this because when we refer to
181		 * rt_ifa->ia6_flags in ip6_input, we assume that the rt_ifa
182		 * points to the address instead of the loopback address.
183		 */
184		if (cmd == RTM_ADD && ifa != nrt->rt_ifa) {
185			IFAFREE(nrt->rt_ifa);
186			IFAREF(ifa);
187			nrt->rt_ifa = ifa;
188		}
189
190		rt_newaddrmsg(cmd, ifa, e, nrt);
191		if (cmd == RTM_DELETE) {
192			rtfree(nrt);
193		} else {
194			/* the cmd must be RTM_ADD here */
195			RT_REMREF(nrt);
196			RT_UNLOCK(nrt);
197		}
198	}
199}
200
201/*
202 * Add ownaddr as loopback rtentry.  We previously add the route only if
203 * necessary (ex. on a p2p link).  However, since we now manage addresses
204 * separately from prefixes, we should always add the route.  We can't
205 * rely on the cloning mechanism from the corresponding interface route
206 * any more.
207 */
208void
209in6_ifaddloop(struct ifaddr *ifa)
210{
211	struct rtentry *rt;
212	int need_loop;
213
214	/* If there is no loopback entry, allocate one. */
215	rt = rtalloc1(ifa->ifa_addr, 0, 0);
216	need_loop = (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
217	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0);
218	if (rt)
219		rtfree(rt);
220	if (need_loop)
221		in6_ifloop_request(RTM_ADD, ifa);
222}
223
224/*
225 * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
226 * if it exists.
227 */
228void
229in6_ifremloop(struct ifaddr *ifa)
230{
231	struct in6_ifaddr *ia;
232	struct rtentry *rt;
233	int ia_count = 0;
234
235	/*
236	 * Some of BSD variants do not remove cloned routes
237	 * from an interface direct route, when removing the direct route
238	 * (see comments in net/net_osdep.h).  Even for variants that do remove
239	 * cloned routes, they could fail to remove the cloned routes when
240	 * we handle multple addresses that share a common prefix.
241	 * So, we should remove the route corresponding to the deleted address.
242	 */
243
244	/*
245	 * Delete the entry only if exact one ifa exists.  More than one ifa
246	 * can exist if we assign a same single address to multiple
247	 * (probably p2p) interfaces.
248	 * XXX: we should avoid such a configuration in IPv6...
249	 */
250	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
251		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
252			ia_count++;
253			if (ia_count > 1)
254				break;
255		}
256	}
257
258	if (ia_count == 1) {
259		/*
260		 * Before deleting, check if a corresponding loopbacked host
261		 * route surely exists.  With this check, we can avoid to
262		 * delete an interface direct route whose destination is same
263		 * as the address being removed.  This can happen when removing
264		 * a subnet-router anycast address on an interface attahced
265		 * to a shared medium.
266		 */
267		rt = rtalloc1(ifa->ifa_addr, 0, 0);
268		if (rt != NULL) {
269			if ((rt->rt_flags & RTF_HOST) != 0 &&
270			    (rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
271				rtfree(rt);
272				in6_ifloop_request(RTM_DELETE, ifa);
273			} else
274				RT_UNLOCK(rt);
275		}
276	}
277}
278
279int
280in6_mask2len(mask, lim0)
281	struct in6_addr *mask;
282	u_char *lim0;
283{
284	int x = 0, y;
285	u_char *lim = lim0, *p;
286
287	/* ignore the scope_id part */
288	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
289		lim = (u_char *)mask + sizeof(*mask);
290	for (p = (u_char *)mask; p < lim; x++, p++) {
291		if (*p != 0xff)
292			break;
293	}
294	y = 0;
295	if (p < lim) {
296		for (y = 0; y < 8; y++) {
297			if ((*p & (0x80 >> y)) == 0)
298				break;
299		}
300	}
301
302	/*
303	 * when the limit pointer is given, do a stricter check on the
304	 * remaining bits.
305	 */
306	if (p < lim) {
307		if (y != 0 && (*p & (0x00ff >> y)) != 0)
308			return (-1);
309		for (p = p + 1; p < lim; p++)
310			if (*p != 0)
311				return (-1);
312	}
313
314	return x * 8 + y;
315}
316
317#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
318#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
319
320int
321in6_control(so, cmd, data, ifp, td)
322	struct	socket *so;
323	u_long cmd;
324	caddr_t	data;
325	struct ifnet *ifp;
326	struct thread *td;
327{
328	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
329	struct	in6_ifaddr *ia = NULL;
330	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
331	struct sockaddr_in6 *sa6;
332	int error;
333
334	switch (cmd) {
335	case SIOCGETSGCNT_IN6:
336	case SIOCGETMIFCNT_IN6:
337		return (mrt6_ioctl(cmd, data));
338	}
339
340	switch(cmd) {
341	case SIOCAADDRCTL_POLICY:
342	case SIOCDADDRCTL_POLICY:
343		if (td != NULL) {
344			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
345			if (error)
346				return (error);
347		}
348		return (in6_src_ioctl(cmd, data));
349	}
350
351	if (ifp == NULL)
352		return (EOPNOTSUPP);
353
354	switch (cmd) {
355	case SIOCSNDFLUSH_IN6:
356	case SIOCSPFXFLUSH_IN6:
357	case SIOCSRTRFLUSH_IN6:
358	case SIOCSDEFIFACE_IN6:
359	case SIOCSIFINFO_FLAGS:
360		if (td != NULL) {
361			error = priv_check(td, PRIV_NETINET_ND6);
362			if (error)
363				return (error);
364		}
365		/* FALLTHROUGH */
366	case OSIOCGIFINFO_IN6:
367	case SIOCGIFINFO_IN6:
368	case SIOCSIFINFO_IN6:
369	case SIOCGDRLST_IN6:
370	case SIOCGPRLST_IN6:
371	case SIOCGNBRINFO_IN6:
372	case SIOCGDEFIFACE_IN6:
373		return (nd6_ioctl(cmd, data, ifp));
374	}
375
376	switch (cmd) {
377	case SIOCSIFPREFIX_IN6:
378	case SIOCDIFPREFIX_IN6:
379	case SIOCAIFPREFIX_IN6:
380	case SIOCCIFPREFIX_IN6:
381	case SIOCSGIFPREFIX_IN6:
382	case SIOCGIFPREFIX_IN6:
383		log(LOG_NOTICE,
384		    "prefix ioctls are now invalidated. "
385		    "please use ifconfig.\n");
386		return (EOPNOTSUPP);
387	}
388
389	switch (cmd) {
390	case SIOCSSCOPE6:
391		if (td != NULL) {
392			error = priv_check(td, PRIV_NETINET_SCOPE6);
393			if (error)
394				return (error);
395		}
396		return (scope6_set(ifp,
397		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
398	case SIOCGSCOPE6:
399		return (scope6_get(ifp,
400		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
401	case SIOCGSCOPE6DEF:
402		return (scope6_get_default((struct scope6_id *)
403		    ifr->ifr_ifru.ifru_scope_id));
404	}
405
406	switch (cmd) {
407	case SIOCALIFADDR:
408	case SIOCDLIFADDR:
409		/*
410		 * XXXRW: Is this checked at another layer?  What priv to use
411		 * here?
412		 */
413		if (td != NULL) {
414			error = suser(td);
415			if (error)
416				return (error);
417		}
418		/* FALLTHROUGH */
419	case SIOCGLIFADDR:
420		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
421	}
422
423	/*
424	 * Find address for this interface, if it exists.
425	 *
426	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
427	 * only, and used the first interface address as the target of other
428	 * operations (without checking ifra_addr).  This was because netinet
429	 * code/API assumed at most 1 interface address per interface.
430	 * Since IPv6 allows a node to assign multiple addresses
431	 * on a single interface, we almost always look and check the
432	 * presence of ifra_addr, and reject invalid ones here.
433	 * It also decreases duplicated code among SIOC*_IN6 operations.
434	 */
435	switch (cmd) {
436	case SIOCAIFADDR_IN6:
437	case SIOCSIFPHYADDR_IN6:
438		sa6 = &ifra->ifra_addr;
439		break;
440	case SIOCSIFADDR_IN6:
441	case SIOCGIFADDR_IN6:
442	case SIOCSIFDSTADDR_IN6:
443	case SIOCSIFNETMASK_IN6:
444	case SIOCGIFDSTADDR_IN6:
445	case SIOCGIFNETMASK_IN6:
446	case SIOCDIFADDR_IN6:
447	case SIOCGIFPSRCADDR_IN6:
448	case SIOCGIFPDSTADDR_IN6:
449	case SIOCGIFAFLAG_IN6:
450	case SIOCSNDFLUSH_IN6:
451	case SIOCSPFXFLUSH_IN6:
452	case SIOCSRTRFLUSH_IN6:
453	case SIOCGIFALIFETIME_IN6:
454	case SIOCSIFALIFETIME_IN6:
455	case SIOCGIFSTAT_IN6:
456	case SIOCGIFSTAT_ICMP6:
457		sa6 = &ifr->ifr_addr;
458		break;
459	default:
460		sa6 = NULL;
461		break;
462	}
463	if (sa6 && sa6->sin6_family == AF_INET6) {
464		int error = 0;
465
466		if (sa6->sin6_scope_id != 0)
467			error = sa6_embedscope(sa6, 0);
468		else
469			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
470		if (error != 0)
471			return (error);
472		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
473	} else
474		ia = NULL;
475
476	switch (cmd) {
477	case SIOCSIFADDR_IN6:
478	case SIOCSIFDSTADDR_IN6:
479	case SIOCSIFNETMASK_IN6:
480		/*
481		 * Since IPv6 allows a node to assign multiple addresses
482		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
483		 */
484		/* we decided to obsolete this command (20000704) */
485		return (EINVAL);
486
487	case SIOCDIFADDR_IN6:
488		/*
489		 * for IPv4, we look for existing in_ifaddr here to allow
490		 * "ifconfig if0 delete" to remove the first IPv4 address on
491		 * the interface.  For IPv6, as the spec allows multiple
492		 * interface address from the day one, we consider "remove the
493		 * first one" semantics to be not preferable.
494		 */
495		if (ia == NULL)
496			return (EADDRNOTAVAIL);
497		/* FALLTHROUGH */
498	case SIOCAIFADDR_IN6:
499		/*
500		 * We always require users to specify a valid IPv6 address for
501		 * the corresponding operation.
502		 */
503		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
504		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
505			return (EAFNOSUPPORT);
506
507		/*
508		 * XXXRW: Is this checked at another layer?  What priv to use
509		 * here?
510		 */
511		if (td != NULL) {
512			error = suser(td);
513			if (error)
514				return (error);
515		}
516
517		break;
518
519	case SIOCGIFADDR_IN6:
520		/* This interface is basically deprecated. use SIOCGIFCONF. */
521		/* FALLTHROUGH */
522	case SIOCGIFAFLAG_IN6:
523	case SIOCGIFNETMASK_IN6:
524	case SIOCGIFDSTADDR_IN6:
525	case SIOCGIFALIFETIME_IN6:
526		/* must think again about its semantics */
527		if (ia == NULL)
528			return (EADDRNOTAVAIL);
529		break;
530	case SIOCSIFALIFETIME_IN6:
531	    {
532		struct in6_addrlifetime *lt;
533
534		if (td != NULL) {
535			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
536			if (error)
537				return (error);
538		}
539		if (ia == NULL)
540			return (EADDRNOTAVAIL);
541		/* sanity for overflow - beware unsigned */
542		lt = &ifr->ifr_ifru.ifru_lifetime;
543		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
544		    lt->ia6t_vltime + time_second < time_second) {
545			return EINVAL;
546		}
547		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
548		    lt->ia6t_pltime + time_second < time_second) {
549			return EINVAL;
550		}
551		break;
552	    }
553	}
554
555	switch (cmd) {
556
557	case SIOCGIFADDR_IN6:
558		ifr->ifr_addr = ia->ia_addr;
559		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
560			return (error);
561		break;
562
563	case SIOCGIFDSTADDR_IN6:
564		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
565			return (EINVAL);
566		/*
567		 * XXX: should we check if ifa_dstaddr is NULL and return
568		 * an error?
569		 */
570		ifr->ifr_dstaddr = ia->ia_dstaddr;
571		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
572			return (error);
573		break;
574
575	case SIOCGIFNETMASK_IN6:
576		ifr->ifr_addr = ia->ia_prefixmask;
577		break;
578
579	case SIOCGIFAFLAG_IN6:
580		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
581		break;
582
583	case SIOCGIFSTAT_IN6:
584		if (ifp == NULL)
585			return EINVAL;
586		bzero(&ifr->ifr_ifru.ifru_stat,
587		    sizeof(ifr->ifr_ifru.ifru_stat));
588		ifr->ifr_ifru.ifru_stat =
589		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
590		break;
591
592	case SIOCGIFSTAT_ICMP6:
593		if (ifp == NULL)
594			return EINVAL;
595		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
596		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
597		ifr->ifr_ifru.ifru_icmp6stat =
598		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
599		break;
600
601	case SIOCGIFALIFETIME_IN6:
602		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
603		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
604			time_t maxexpire;
605			struct in6_addrlifetime *retlt =
606			    &ifr->ifr_ifru.ifru_lifetime;
607
608			/*
609			 * XXX: adjust expiration time assuming time_t is
610			 * signed.
611			 */
612			maxexpire = (-1) &
613			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
614			if (ia->ia6_lifetime.ia6t_vltime <
615			    maxexpire - ia->ia6_updatetime) {
616				retlt->ia6t_expire = ia->ia6_updatetime +
617				    ia->ia6_lifetime.ia6t_vltime;
618			} else
619				retlt->ia6t_expire = maxexpire;
620		}
621		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
622			time_t maxexpire;
623			struct in6_addrlifetime *retlt =
624			    &ifr->ifr_ifru.ifru_lifetime;
625
626			/*
627			 * XXX: adjust expiration time assuming time_t is
628			 * signed.
629			 */
630			maxexpire = (-1) &
631			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
632			if (ia->ia6_lifetime.ia6t_pltime <
633			    maxexpire - ia->ia6_updatetime) {
634				retlt->ia6t_preferred = ia->ia6_updatetime +
635				    ia->ia6_lifetime.ia6t_pltime;
636			} else
637				retlt->ia6t_preferred = maxexpire;
638		}
639		break;
640
641	case SIOCSIFALIFETIME_IN6:
642		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
643		/* for sanity */
644		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
645			ia->ia6_lifetime.ia6t_expire =
646				time_second + ia->ia6_lifetime.ia6t_vltime;
647		} else
648			ia->ia6_lifetime.ia6t_expire = 0;
649		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
650			ia->ia6_lifetime.ia6t_preferred =
651				time_second + ia->ia6_lifetime.ia6t_pltime;
652		} else
653			ia->ia6_lifetime.ia6t_preferred = 0;
654		break;
655
656	case SIOCAIFADDR_IN6:
657	{
658		int i, error = 0;
659		struct nd_prefixctl pr0;
660		struct nd_prefix *pr;
661
662		/*
663		 * first, make or update the interface address structure,
664		 * and link it to the list.
665		 */
666		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
667			return (error);
668		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
669		    == NULL) {
670		    	/*
671			 * this can happen when the user specify the 0 valid
672			 * lifetime.
673			 */
674			break;
675		}
676
677		/*
678		 * then, make the prefix on-link on the interface.
679		 * XXX: we'd rather create the prefix before the address, but
680		 * we need at least one address to install the corresponding
681		 * interface route, so we configure the address first.
682		 */
683
684		/*
685		 * convert mask to prefix length (prefixmask has already
686		 * been validated in in6_update_ifa().
687		 */
688		bzero(&pr0, sizeof(pr0));
689		pr0.ndpr_ifp = ifp;
690		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
691		    NULL);
692		if (pr0.ndpr_plen == 128) {
693			break;	/* we don't need to install a host route. */
694		}
695		pr0.ndpr_prefix = ifra->ifra_addr;
696		/* apply the mask for safety. */
697		for (i = 0; i < 4; i++) {
698			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
699			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
700		}
701		/*
702		 * XXX: since we don't have an API to set prefix (not address)
703		 * lifetimes, we just use the same lifetimes as addresses.
704		 * The (temporarily) installed lifetimes can be overridden by
705		 * later advertised RAs (when accept_rtadv is non 0), which is
706		 * an intended behavior.
707		 */
708		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
709		pr0.ndpr_raf_auto =
710		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
711		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
712		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
713
714		/* add the prefix if not yet. */
715		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
716			/*
717			 * nd6_prelist_add will install the corresponding
718			 * interface route.
719			 */
720			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
721				return (error);
722			if (pr == NULL) {
723				log(LOG_ERR, "nd6_prelist_add succeeded but "
724				    "no prefix\n");
725				return (EINVAL); /* XXX panic here? */
726			}
727		}
728
729		/* relate the address to the prefix */
730		if (ia->ia6_ndpr == NULL) {
731			ia->ia6_ndpr = pr;
732			pr->ndpr_refcnt++;
733
734			/*
735			 * If this is the first autoconf address from the
736			 * prefix, create a temporary address as well
737			 * (when required).
738			 */
739			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
740			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
741				int e;
742				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
743					log(LOG_NOTICE, "in6_control: failed "
744					    "to create a temporary address, "
745					    "errno=%d\n", e);
746				}
747			}
748		}
749
750		/*
751		 * this might affect the status of autoconfigured addresses,
752		 * that is, this address might make other addresses detached.
753		 */
754		pfxlist_onlink_check();
755		if (error == 0 && ia)
756			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
757		break;
758	}
759
760	case SIOCDIFADDR_IN6:
761	{
762		struct nd_prefix *pr;
763
764		/*
765		 * If the address being deleted is the only one that owns
766		 * the corresponding prefix, expire the prefix as well.
767		 * XXX: theoretically, we don't have to worry about such
768		 * relationship, since we separate the address management
769		 * and the prefix management.  We do this, however, to provide
770		 * as much backward compatibility as possible in terms of
771		 * the ioctl operation.
772		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
773		 */
774		pr = ia->ia6_ndpr;
775		in6_purgeaddr(&ia->ia_ifa);
776		if (pr && pr->ndpr_refcnt == 0)
777			prelist_remove(pr);
778		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
779		break;
780	}
781
782	default:
783		if (ifp == NULL || ifp->if_ioctl == 0)
784			return (EOPNOTSUPP);
785		return ((*ifp->if_ioctl)(ifp, cmd, data));
786	}
787
788	return (0);
789}
790
791/*
792 * Update parameters of an IPv6 interface address.
793 * If necessary, a new entry is created and linked into address chains.
794 * This function is separated from in6_control().
795 * XXX: should this be performed under splnet()?
796 */
797int
798in6_update_ifa(ifp, ifra, ia, flags)
799	struct ifnet *ifp;
800	struct in6_aliasreq *ifra;
801	struct in6_ifaddr *ia;
802	int flags;
803{
804	int error = 0, hostIsNew = 0, plen = -1;
805	struct in6_ifaddr *oia;
806	struct sockaddr_in6 dst6;
807	struct in6_addrlifetime *lt;
808	struct in6_multi_mship *imm;
809	struct in6_multi *in6m_sol;
810	struct rtentry *rt;
811	int delay;
812	char ip6buf[INET6_ADDRSTRLEN];
813
814	/* Validate parameters */
815	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
816		return (EINVAL);
817
818	/*
819	 * The destination address for a p2p link must have a family
820	 * of AF_UNSPEC or AF_INET6.
821	 */
822	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
823	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
824	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
825		return (EAFNOSUPPORT);
826	/*
827	 * validate ifra_prefixmask.  don't check sin6_family, netmask
828	 * does not carry fields other than sin6_len.
829	 */
830	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
831		return (EINVAL);
832	/*
833	 * Because the IPv6 address architecture is classless, we require
834	 * users to specify a (non 0) prefix length (mask) for a new address.
835	 * We also require the prefix (when specified) mask is valid, and thus
836	 * reject a non-consecutive mask.
837	 */
838	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
839		return (EINVAL);
840	if (ifra->ifra_prefixmask.sin6_len != 0) {
841		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
842		    (u_char *)&ifra->ifra_prefixmask +
843		    ifra->ifra_prefixmask.sin6_len);
844		if (plen <= 0)
845			return (EINVAL);
846	} else {
847		/*
848		 * In this case, ia must not be NULL.  We just use its prefix
849		 * length.
850		 */
851		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
852	}
853	/*
854	 * If the destination address on a p2p interface is specified,
855	 * and the address is a scoped one, validate/set the scope
856	 * zone identifier.
857	 */
858	dst6 = ifra->ifra_dstaddr;
859	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
860	    (dst6.sin6_family == AF_INET6)) {
861		struct in6_addr in6_tmp;
862		u_int32_t zoneid;
863
864		in6_tmp = dst6.sin6_addr;
865		if (in6_setscope(&in6_tmp, ifp, &zoneid))
866			return (EINVAL); /* XXX: should be impossible */
867
868		if (dst6.sin6_scope_id != 0) {
869			if (dst6.sin6_scope_id != zoneid)
870				return (EINVAL);
871		} else		/* user omit to specify the ID. */
872			dst6.sin6_scope_id = zoneid;
873
874		/* convert into the internal form */
875		if (sa6_embedscope(&dst6, 0))
876			return (EINVAL); /* XXX: should be impossible */
877	}
878	/*
879	 * The destination address can be specified only for a p2p or a
880	 * loopback interface.  If specified, the corresponding prefix length
881	 * must be 128.
882	 */
883	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
884		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
885			/* XXX: noisy message */
886			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
887			    "be specified for a p2p or a loopback IF only\n"));
888			return (EINVAL);
889		}
890		if (plen != 128) {
891			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
892			    "be 128 when dstaddr is specified\n"));
893			return (EINVAL);
894		}
895	}
896	/* lifetime consistency check */
897	lt = &ifra->ifra_lifetime;
898	if (lt->ia6t_pltime > lt->ia6t_vltime)
899		return (EINVAL);
900	if (lt->ia6t_vltime == 0) {
901		/*
902		 * the following log might be noisy, but this is a typical
903		 * configuration mistake or a tool's bug.
904		 */
905		nd6log((LOG_INFO,
906		    "in6_update_ifa: valid lifetime is 0 for %s\n",
907		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
908
909		if (ia == NULL)
910			return (0); /* there's nothing to do */
911	}
912
913	/*
914	 * If this is a new address, allocate a new ifaddr and link it
915	 * into chains.
916	 */
917	if (ia == NULL) {
918		hostIsNew = 1;
919		/*
920		 * When in6_update_ifa() is called in a process of a received
921		 * RA, it is called under an interrupt context.  So, we should
922		 * call malloc with M_NOWAIT.
923		 */
924		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
925		    M_NOWAIT);
926		if (ia == NULL)
927			return (ENOBUFS);
928		bzero((caddr_t)ia, sizeof(*ia));
929		/* Initialize the address and masks, and put time stamp */
930		IFA_LOCK_INIT(&ia->ia_ifa);
931		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
932		ia->ia_addr.sin6_family = AF_INET6;
933		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
934		ia->ia6_createtime = time_second;
935		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
936			/*
937			 * XXX: some functions expect that ifa_dstaddr is not
938			 * NULL for p2p interfaces.
939			 */
940			ia->ia_ifa.ifa_dstaddr =
941			    (struct sockaddr *)&ia->ia_dstaddr;
942		} else {
943			ia->ia_ifa.ifa_dstaddr = NULL;
944		}
945		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
946
947		ia->ia_ifp = ifp;
948		if ((oia = in6_ifaddr) != NULL) {
949			for ( ; oia->ia_next; oia = oia->ia_next)
950				continue;
951			oia->ia_next = ia;
952		} else
953			in6_ifaddr = ia;
954
955		ia->ia_ifa.ifa_refcnt = 1;
956		TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
957	}
958
959	/* update timestamp */
960	ia->ia6_updatetime = time_second;
961
962	/* set prefix mask */
963	if (ifra->ifra_prefixmask.sin6_len) {
964		/*
965		 * We prohibit changing the prefix length of an existing
966		 * address, because
967		 * + such an operation should be rare in IPv6, and
968		 * + the operation would confuse prefix management.
969		 */
970		if (ia->ia_prefixmask.sin6_len &&
971		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
972			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
973			    " existing (%s) address should not be changed\n",
974			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
975			error = EINVAL;
976			goto unlink;
977		}
978		ia->ia_prefixmask = ifra->ifra_prefixmask;
979	}
980
981	/*
982	 * If a new destination address is specified, scrub the old one and
983	 * install the new destination.  Note that the interface must be
984	 * p2p or loopback (see the check above.)
985	 */
986	if (dst6.sin6_family == AF_INET6 &&
987	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
988		int e;
989
990		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
991		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
992			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
993			    "a route to the old destination: %s\n",
994			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
995			/* proceed anyway... */
996		} else
997			ia->ia_flags &= ~IFA_ROUTE;
998		ia->ia_dstaddr = dst6;
999	}
1000
1001	/*
1002	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1003	 * to see if the address is deprecated or invalidated, but initialize
1004	 * these members for applications.
1005	 */
1006	ia->ia6_lifetime = ifra->ifra_lifetime;
1007	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1008		ia->ia6_lifetime.ia6t_expire =
1009		    time_second + ia->ia6_lifetime.ia6t_vltime;
1010	} else
1011		ia->ia6_lifetime.ia6t_expire = 0;
1012	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1013		ia->ia6_lifetime.ia6t_preferred =
1014		    time_second + ia->ia6_lifetime.ia6t_pltime;
1015	} else
1016		ia->ia6_lifetime.ia6t_preferred = 0;
1017
1018	/* reset the interface and routing table appropriately. */
1019	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1020		goto unlink;
1021
1022	/*
1023	 * configure address flags.
1024	 */
1025	ia->ia6_flags = ifra->ifra_flags;
1026	/*
1027	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1028	 * userland, make it deprecated.
1029	 */
1030	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1031		ia->ia6_lifetime.ia6t_pltime = 0;
1032		ia->ia6_lifetime.ia6t_preferred = time_second;
1033	}
1034	/*
1035	 * Make the address tentative before joining multicast addresses,
1036	 * so that corresponding MLD responses would not have a tentative
1037	 * source address.
1038	 */
1039	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1040	if (hostIsNew && in6if_do_dad(ifp))
1041		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1042
1043	/*
1044	 * We are done if we have simply modified an existing address.
1045	 */
1046	if (!hostIsNew)
1047		return (error);
1048
1049	/*
1050	 * Beyond this point, we should call in6_purgeaddr upon an error,
1051	 * not just go to unlink.
1052	 */
1053
1054	/* Join necessary multicast groups */
1055	in6m_sol = NULL;
1056	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1057		struct sockaddr_in6 mltaddr, mltmask;
1058		struct in6_addr llsol;
1059
1060		/* join solicited multicast addr for new host id */
1061		bzero(&llsol, sizeof(struct in6_addr));
1062		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1063		llsol.s6_addr32[1] = 0;
1064		llsol.s6_addr32[2] = htonl(1);
1065		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1066		llsol.s6_addr8[12] = 0xff;
1067		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1068			/* XXX: should not happen */
1069			log(LOG_ERR, "in6_update_ifa: "
1070			    "in6_setscope failed\n");
1071			goto cleanup;
1072		}
1073		delay = 0;
1074		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1075			/*
1076			 * We need a random delay for DAD on the address
1077			 * being configured.  It also means delaying
1078			 * transmission of the corresponding MLD report to
1079			 * avoid report collision.
1080			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1081			 */
1082			delay = arc4random() %
1083			    (MAX_RTR_SOLICITATION_DELAY * hz);
1084		}
1085		imm = in6_joingroup(ifp, &llsol, &error, delay);
1086		if (error != 0) {
1087			nd6log((LOG_WARNING,
1088			    "in6_update_ifa: addmulti failed for "
1089			    "%s on %s (errno=%d)\n",
1090			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
1091			    error));
1092			in6_purgeaddr((struct ifaddr *)ia);
1093			return (error);
1094		}
1095		in6m_sol = imm->i6mm_maddr;
1096
1097		bzero(&mltmask, sizeof(mltmask));
1098		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1099		mltmask.sin6_family = AF_INET6;
1100		mltmask.sin6_addr = in6mask32;
1101#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1102
1103		/*
1104		 * join link-local all-nodes address
1105		 */
1106		bzero(&mltaddr, sizeof(mltaddr));
1107		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1108		mltaddr.sin6_family = AF_INET6;
1109		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1110		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1111		    0)
1112			goto cleanup; /* XXX: should not fail */
1113
1114		/*
1115		 * XXX: do we really need this automatic routes?
1116		 * We should probably reconsider this stuff.  Most applications
1117		 * actually do not need the routes, since they usually specify
1118		 * the outgoing interface.
1119		 */
1120		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1121		if (rt) {
1122			if (memcmp(&mltaddr.sin6_addr,
1123			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1124			    MLTMASK_LEN)) {
1125				RTFREE_LOCKED(rt);
1126				rt = NULL;
1127			}
1128		}
1129		if (!rt) {
1130			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1131			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1132			    (struct sockaddr *)&ia->ia_addr,
1133			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1134			    (struct rtentry **)0);
1135			if (error)
1136				goto cleanup;
1137		} else
1138			RTFREE_LOCKED(rt);
1139
1140		/*
1141		 * XXX: do we really need this automatic routes?
1142		 * We should probably reconsider this stuff.  Most applications
1143		 * actually do not need the routes, since they usually specify
1144		 * the outgoing interface.
1145		 */
1146		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1147		if (rt) {
1148			/* XXX: only works in !SCOPEDROUTING case. */
1149			if (memcmp(&mltaddr.sin6_addr,
1150			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1151			    MLTMASK_LEN)) {
1152				RTFREE_LOCKED(rt);
1153				rt = NULL;
1154			}
1155		}
1156		if (!rt) {
1157			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1158			    (struct sockaddr *)&ia->ia_addr,
1159			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1160			    (struct rtentry **)0);
1161			if (error)
1162				goto cleanup;
1163		} else {
1164			RTFREE_LOCKED(rt);
1165		}
1166
1167		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1168		if (!imm) {
1169			nd6log((LOG_WARNING,
1170			    "in6_update_ifa: addmulti failed for "
1171			    "%s on %s (errno=%d)\n",
1172			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1173			    if_name(ifp), error));
1174			goto cleanup;
1175		}
1176
1177		/*
1178		 * join node information group address
1179		 */
1180#define hostnamelen	strlen(hostname)
1181		delay = 0;
1182		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1183			/*
1184			 * The spec doesn't say anything about delay for this
1185			 * group, but the same logic should apply.
1186			 */
1187			delay = arc4random() %
1188			    (MAX_RTR_SOLICITATION_DELAY * hz);
1189		}
1190		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr.sin6_addr)
1191		    == 0) {
1192			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1193			    delay); /* XXX jinmei */
1194			if (!imm) {
1195				nd6log((LOG_WARNING, "in6_update_ifa: "
1196				    "addmulti failed for %s on %s "
1197				    "(errno=%d)\n",
1198				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1199				    if_name(ifp), error));
1200				/* XXX not very fatal, go on... */
1201			}
1202		}
1203#undef hostnamelen
1204
1205		/*
1206		 * join interface-local all-nodes address.
1207		 * (ff01::1%ifN, and ff01::%ifN/32)
1208		 */
1209		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1210		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1211		    != 0)
1212			goto cleanup; /* XXX: should not fail */
1213		/* XXX: again, do we really need the route? */
1214		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1215		if (rt) {
1216			if (memcmp(&mltaddr.sin6_addr,
1217			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1218			    MLTMASK_LEN)) {
1219				RTFREE_LOCKED(rt);
1220				rt = NULL;
1221			}
1222		}
1223		if (!rt) {
1224			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1225			    (struct sockaddr *)&ia->ia_addr,
1226			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1227			    (struct rtentry **)0);
1228			if (error)
1229				goto cleanup;
1230		} else
1231			RTFREE_LOCKED(rt);
1232
1233		/* XXX: again, do we really need the route? */
1234		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1235		if (rt) {
1236			if (memcmp(&mltaddr.sin6_addr,
1237			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1238			    MLTMASK_LEN)) {
1239				RTFREE_LOCKED(rt);
1240				rt = NULL;
1241			}
1242		}
1243		if (!rt) {
1244			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1245			    (struct sockaddr *)&ia->ia_addr,
1246			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1247			    (struct rtentry **)0);
1248			if (error)
1249				goto cleanup;
1250		} else {
1251			RTFREE_LOCKED(rt);
1252		}
1253
1254		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1255		if (!imm) {
1256			nd6log((LOG_WARNING, "in6_update_ifa: "
1257			    "addmulti failed for %s on %s "
1258			    "(errno=%d)\n",
1259			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1260			    if_name(ifp), error));
1261			goto cleanup;
1262		}
1263#undef	MLTMASK_LEN
1264	}
1265
1266	/*
1267	 * Perform DAD, if needed.
1268	 * XXX It may be of use, if we can administratively
1269	 * disable DAD.
1270	 */
1271	if (hostIsNew && in6if_do_dad(ifp) &&
1272	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1273	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1274	{
1275		int mindelay, maxdelay;
1276
1277		delay = 0;
1278		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1279			/*
1280			 * We need to impose a delay before sending an NS
1281			 * for DAD.  Check if we also needed a delay for the
1282			 * corresponding MLD message.  If we did, the delay
1283			 * should be larger than the MLD delay (this could be
1284			 * relaxed a bit, but this simple logic is at least
1285			 * safe).
1286			 */
1287			mindelay = 0;
1288			if (in6m_sol != NULL &&
1289			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1290				mindelay = in6m_sol->in6m_timer;
1291			}
1292			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1293			if (maxdelay - mindelay == 0)
1294				delay = 0;
1295			else {
1296				delay =
1297				    (arc4random() % (maxdelay - mindelay)) +
1298				    mindelay;
1299			}
1300		}
1301		nd6_dad_start((struct ifaddr *)ia, delay);
1302	}
1303
1304	return (error);
1305
1306  unlink:
1307	/*
1308	 * XXX: if a change of an existing address failed, keep the entry
1309	 * anyway.
1310	 */
1311	if (hostIsNew)
1312		in6_unlink_ifa(ia, ifp);
1313	return (error);
1314
1315  cleanup:
1316	in6_purgeaddr(&ia->ia_ifa);
1317	return error;
1318}
1319
1320void
1321in6_purgeaddr(ifa)
1322	struct ifaddr *ifa;
1323{
1324	struct ifnet *ifp = ifa->ifa_ifp;
1325	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1326	char ip6buf[INET6_ADDRSTRLEN];
1327
1328	/* stop DAD processing */
1329	nd6_dad_stop(ifa);
1330
1331	/*
1332	 * delete route to the destination of the address being purged.
1333	 * The interface must be p2p or loopback in this case.
1334	 */
1335	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1336		int e;
1337
1338		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1339		    != 0) {
1340			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1341			    "a route to the p2p destination: %s on %s, "
1342			    "errno=%d\n",
1343			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1344			    if_name(ifp), e);
1345			/* proceed anyway... */
1346		} else
1347			ia->ia_flags &= ~IFA_ROUTE;
1348	}
1349
1350	/* Remove ownaddr's loopback rtentry, if it exists. */
1351	in6_ifremloop(&(ia->ia_ifa));
1352
1353	if (ifp->if_flags & IFF_MULTICAST) {
1354		/*
1355		 * delete solicited multicast addr for deleting host id
1356		 */
1357		struct in6_multi *in6m;
1358		struct in6_addr llsol;
1359		bzero(&llsol, sizeof(struct in6_addr));
1360		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1361		llsol.s6_addr32[1] = 0;
1362		llsol.s6_addr32[2] = htonl(1);
1363		llsol.s6_addr32[3] =
1364			ia->ia_addr.sin6_addr.s6_addr32[3];
1365		llsol.s6_addr8[12] = 0xff;
1366		(void)in6_setscope(&llsol, ifp, NULL); /* XXX proceed anyway */
1367
1368		IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1369		if (in6m)
1370			in6_delmulti(in6m);
1371	}
1372
1373	in6_unlink_ifa(ia, ifp);
1374}
1375
1376static void
1377in6_unlink_ifa(ia, ifp)
1378	struct in6_ifaddr *ia;
1379	struct ifnet *ifp;
1380{
1381	struct in6_ifaddr *oia;
1382	int	s = splnet();
1383
1384	TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1385
1386	oia = ia;
1387	if (oia == (ia = in6_ifaddr))
1388		in6_ifaddr = ia->ia_next;
1389	else {
1390		while (ia->ia_next && (ia->ia_next != oia))
1391			ia = ia->ia_next;
1392		if (ia->ia_next)
1393			ia->ia_next = oia->ia_next;
1394		else {
1395			/* search failed */
1396			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1397		}
1398	}
1399
1400	/*
1401	 * Release the reference to the base prefix.  There should be a
1402	 * positive reference.
1403	 */
1404	if (oia->ia6_ndpr == NULL) {
1405		nd6log((LOG_NOTICE,
1406		    "in6_unlink_ifa: autoconf'ed address "
1407		    "%p has no prefix\n", oia));
1408	} else {
1409		oia->ia6_ndpr->ndpr_refcnt--;
1410		oia->ia6_ndpr = NULL;
1411	}
1412
1413	/*
1414	 * Also, if the address being removed is autoconf'ed, call
1415	 * pfxlist_onlink_check() since the release might affect the status of
1416	 * other (detached) addresses.
1417	 */
1418	if ((oia->ia6_flags & IN6_IFF_AUTOCONF)) {
1419		pfxlist_onlink_check();
1420	}
1421
1422	/*
1423	 * release another refcnt for the link from in6_ifaddr.
1424	 * Note that we should decrement the refcnt at least once for all *BSD.
1425	 */
1426	IFAFREE(&oia->ia_ifa);
1427
1428	splx(s);
1429}
1430
1431void
1432in6_purgeif(ifp)
1433	struct ifnet *ifp;
1434{
1435	struct ifaddr *ifa, *nifa;
1436
1437	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa) {
1438		nifa = TAILQ_NEXT(ifa, ifa_list);
1439		if (ifa->ifa_addr->sa_family != AF_INET6)
1440			continue;
1441		in6_purgeaddr(ifa);
1442	}
1443
1444	in6_ifdetach(ifp);
1445}
1446
1447/*
1448 * SIOC[GAD]LIFADDR.
1449 *	SIOCGLIFADDR: get first address. (?)
1450 *	SIOCGLIFADDR with IFLR_PREFIX:
1451 *		get first address that matches the specified prefix.
1452 *	SIOCALIFADDR: add the specified address.
1453 *	SIOCALIFADDR with IFLR_PREFIX:
1454 *		add the specified prefix, filling hostid part from
1455 *		the first link-local address.  prefixlen must be <= 64.
1456 *	SIOCDLIFADDR: delete the specified address.
1457 *	SIOCDLIFADDR with IFLR_PREFIX:
1458 *		delete the first address that matches the specified prefix.
1459 * return values:
1460 *	EINVAL on invalid parameters
1461 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1462 *	other values may be returned from in6_ioctl()
1463 *
1464 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1465 * this is to accomodate address naming scheme other than RFC2374,
1466 * in the future.
1467 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1468 * address encoding scheme. (see figure on page 8)
1469 */
1470static int
1471in6_lifaddr_ioctl(so, cmd, data, ifp, td)
1472	struct socket *so;
1473	u_long cmd;
1474	caddr_t	data;
1475	struct ifnet *ifp;
1476	struct thread *td;
1477{
1478	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1479	struct ifaddr *ifa;
1480	struct sockaddr *sa;
1481
1482	/* sanity checks */
1483	if (!data || !ifp) {
1484		panic("invalid argument to in6_lifaddr_ioctl");
1485		/* NOTREACHED */
1486	}
1487
1488	switch (cmd) {
1489	case SIOCGLIFADDR:
1490		/* address must be specified on GET with IFLR_PREFIX */
1491		if ((iflr->flags & IFLR_PREFIX) == 0)
1492			break;
1493		/* FALLTHROUGH */
1494	case SIOCALIFADDR:
1495	case SIOCDLIFADDR:
1496		/* address must be specified on ADD and DELETE */
1497		sa = (struct sockaddr *)&iflr->addr;
1498		if (sa->sa_family != AF_INET6)
1499			return EINVAL;
1500		if (sa->sa_len != sizeof(struct sockaddr_in6))
1501			return EINVAL;
1502		/* XXX need improvement */
1503		sa = (struct sockaddr *)&iflr->dstaddr;
1504		if (sa->sa_family && sa->sa_family != AF_INET6)
1505			return EINVAL;
1506		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1507			return EINVAL;
1508		break;
1509	default: /* shouldn't happen */
1510#if 0
1511		panic("invalid cmd to in6_lifaddr_ioctl");
1512		/* NOTREACHED */
1513#else
1514		return EOPNOTSUPP;
1515#endif
1516	}
1517	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1518		return EINVAL;
1519
1520	switch (cmd) {
1521	case SIOCALIFADDR:
1522	    {
1523		struct in6_aliasreq ifra;
1524		struct in6_addr *hostid = NULL;
1525		int prefixlen;
1526
1527		if ((iflr->flags & IFLR_PREFIX) != 0) {
1528			struct sockaddr_in6 *sin6;
1529
1530			/*
1531			 * hostid is to fill in the hostid part of the
1532			 * address.  hostid points to the first link-local
1533			 * address attached to the interface.
1534			 */
1535			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1536			if (!ifa)
1537				return EADDRNOTAVAIL;
1538			hostid = IFA_IN6(ifa);
1539
1540		 	/* prefixlen must be <= 64. */
1541			if (64 < iflr->prefixlen)
1542				return EINVAL;
1543			prefixlen = iflr->prefixlen;
1544
1545			/* hostid part must be zero. */
1546			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1547			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1548			    sin6->sin6_addr.s6_addr32[3] != 0) {
1549				return EINVAL;
1550			}
1551		} else
1552			prefixlen = iflr->prefixlen;
1553
1554		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1555		bzero(&ifra, sizeof(ifra));
1556		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1557
1558		bcopy(&iflr->addr, &ifra.ifra_addr,
1559		    ((struct sockaddr *)&iflr->addr)->sa_len);
1560		if (hostid) {
1561			/* fill in hostid part */
1562			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1563			    hostid->s6_addr32[2];
1564			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1565			    hostid->s6_addr32[3];
1566		}
1567
1568		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1569			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1570			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1571			if (hostid) {
1572				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1573				    hostid->s6_addr32[2];
1574				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1575				    hostid->s6_addr32[3];
1576			}
1577		}
1578
1579		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1580		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1581
1582		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1583		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1584	    }
1585	case SIOCGLIFADDR:
1586	case SIOCDLIFADDR:
1587	    {
1588		struct in6_ifaddr *ia;
1589		struct in6_addr mask, candidate, match;
1590		struct sockaddr_in6 *sin6;
1591		int cmp;
1592
1593		bzero(&mask, sizeof(mask));
1594		if (iflr->flags & IFLR_PREFIX) {
1595			/* lookup a prefix rather than address. */
1596			in6_prefixlen2mask(&mask, iflr->prefixlen);
1597
1598			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1599			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1600			match.s6_addr32[0] &= mask.s6_addr32[0];
1601			match.s6_addr32[1] &= mask.s6_addr32[1];
1602			match.s6_addr32[2] &= mask.s6_addr32[2];
1603			match.s6_addr32[3] &= mask.s6_addr32[3];
1604
1605			/* if you set extra bits, that's wrong */
1606			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1607				return EINVAL;
1608
1609			cmp = 1;
1610		} else {
1611			if (cmd == SIOCGLIFADDR) {
1612				/* on getting an address, take the 1st match */
1613				cmp = 0;	/* XXX */
1614			} else {
1615				/* on deleting an address, do exact match */
1616				in6_prefixlen2mask(&mask, 128);
1617				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1618				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1619
1620				cmp = 1;
1621			}
1622		}
1623
1624		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1625			if (ifa->ifa_addr->sa_family != AF_INET6)
1626				continue;
1627			if (!cmp)
1628				break;
1629
1630			/*
1631			 * XXX: this is adhoc, but is necessary to allow
1632			 * a user to specify fe80::/64 (not /10) for a
1633			 * link-local address.
1634			 */
1635			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1636			in6_clearscope(&candidate);
1637			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1638			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1639			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1640			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1641			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1642				break;
1643		}
1644		if (!ifa)
1645			return EADDRNOTAVAIL;
1646		ia = ifa2ia6(ifa);
1647
1648		if (cmd == SIOCGLIFADDR) {
1649			int error;
1650
1651			/* fill in the if_laddrreq structure */
1652			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1653			error = sa6_recoverscope(
1654			    (struct sockaddr_in6 *)&iflr->addr);
1655			if (error != 0)
1656				return (error);
1657
1658			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1659				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1660				    ia->ia_dstaddr.sin6_len);
1661				error = sa6_recoverscope(
1662				    (struct sockaddr_in6 *)&iflr->dstaddr);
1663				if (error != 0)
1664					return (error);
1665			} else
1666				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1667
1668			iflr->prefixlen =
1669			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1670
1671			iflr->flags = ia->ia6_flags;	/* XXX */
1672
1673			return 0;
1674		} else {
1675			struct in6_aliasreq ifra;
1676
1677			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1678			bzero(&ifra, sizeof(ifra));
1679			bcopy(iflr->iflr_name, ifra.ifra_name,
1680			    sizeof(ifra.ifra_name));
1681
1682			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1683			    ia->ia_addr.sin6_len);
1684			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1685				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1686				    ia->ia_dstaddr.sin6_len);
1687			} else {
1688				bzero(&ifra.ifra_dstaddr,
1689				    sizeof(ifra.ifra_dstaddr));
1690			}
1691			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1692			    ia->ia_prefixmask.sin6_len);
1693
1694			ifra.ifra_flags = ia->ia6_flags;
1695			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1696			    ifp, td);
1697		}
1698	    }
1699	}
1700
1701	return EOPNOTSUPP;	/* just for safety */
1702}
1703
1704/*
1705 * Initialize an interface's intetnet6 address
1706 * and routing table entry.
1707 */
1708static int
1709in6_ifinit(ifp, ia, sin6, newhost)
1710	struct ifnet *ifp;
1711	struct in6_ifaddr *ia;
1712	struct sockaddr_in6 *sin6;
1713	int newhost;
1714{
1715	int	error = 0, plen, ifacount = 0;
1716	int	s = splimp();
1717	struct ifaddr *ifa;
1718
1719	/*
1720	 * Give the interface a chance to initialize
1721	 * if this is its first address,
1722	 * and to validate the address if necessary.
1723	 */
1724	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1725		if (ifa->ifa_addr->sa_family != AF_INET6)
1726			continue;
1727		ifacount++;
1728	}
1729
1730	ia->ia_addr = *sin6;
1731
1732	if (ifacount <= 1 && ifp->if_ioctl) {
1733		IFF_LOCKGIANT(ifp);
1734		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1735		IFF_UNLOCKGIANT(ifp);
1736		if (error) {
1737			splx(s);
1738			return (error);
1739		}
1740	}
1741	splx(s);
1742
1743	ia->ia_ifa.ifa_metric = ifp->if_metric;
1744
1745	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1746
1747	if (newhost) {
1748		/*
1749		 * set the rtrequest function to create llinfo.  It also
1750		 * adjust outgoing interface of the route for the local
1751		 * address when called via in6_ifaddloop() below.
1752		 */
1753		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1754	}
1755
1756	/*
1757	 * Special case:
1758	 * If a new destination address is specified for a point-to-point
1759	 * interface, install a route to the destination as an interface
1760	 * direct route.  In addition, if the link is expected to have neighbor
1761	 * cache entries, specify RTF_LLINFO so that a cache entry for the
1762	 * destination address will be created.
1763	 * created
1764	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1765	 * interface that share the same destination.
1766	 */
1767	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1768	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1769	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1770		int rtflags = RTF_UP | RTF_HOST;
1771		struct rtentry *rt = NULL, **rtp = NULL;
1772
1773		if (nd6_need_cache(ifp) != 0) {
1774			rtflags |= RTF_LLINFO;
1775			rtp = &rt;
1776		}
1777
1778		error = rtrequest(RTM_ADD, (struct sockaddr *)&ia->ia_dstaddr,
1779		    (struct sockaddr *)&ia->ia_addr,
1780		    (struct sockaddr *)&ia->ia_prefixmask,
1781		    ia->ia_flags | rtflags, rtp);
1782		if (error != 0)
1783			return (error);
1784		if (rt != NULL) {
1785			struct llinfo_nd6 *ln;
1786
1787			RT_LOCK(rt);
1788			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1789			if (ln != NULL) {
1790				/*
1791				 * Set the state to STALE because we don't
1792				 * have to perform address resolution on this
1793				 * link.
1794				 */
1795				ln->ln_state = ND6_LLINFO_STALE;
1796			}
1797			RT_REMREF(rt);
1798			RT_UNLOCK(rt);
1799		}
1800		ia->ia_flags |= IFA_ROUTE;
1801	}
1802	if (plen < 128) {
1803		/*
1804		 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1805		 */
1806		ia->ia_ifa.ifa_flags |= RTF_CLONING;
1807	}
1808
1809	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1810	if (newhost)
1811		in6_ifaddloop(&(ia->ia_ifa));
1812
1813	return (error);
1814}
1815
1816struct in6_multi_mship *
1817in6_joingroup(ifp, addr, errorp, delay)
1818	struct ifnet *ifp;
1819	struct in6_addr *addr;
1820	int *errorp;
1821	int delay;
1822{
1823	struct in6_multi_mship *imm;
1824
1825	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
1826	if (!imm) {
1827		*errorp = ENOBUFS;
1828		return NULL;
1829	}
1830	imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp, delay);
1831	if (!imm->i6mm_maddr) {
1832		/* *errorp is alrady set */
1833		free(imm, M_IP6MADDR);
1834		return NULL;
1835	}
1836	return imm;
1837}
1838
1839int
1840in6_leavegroup(imm)
1841	struct in6_multi_mship *imm;
1842{
1843
1844	if (imm->i6mm_maddr)
1845		in6_delmulti(imm->i6mm_maddr);
1846	free(imm,  M_IP6MADDR);
1847	return 0;
1848}
1849
1850/*
1851 * Find an IPv6 interface link-local address specific to an interface.
1852 */
1853struct in6_ifaddr *
1854in6ifa_ifpforlinklocal(ifp, ignoreflags)
1855	struct ifnet *ifp;
1856	int ignoreflags;
1857{
1858	struct ifaddr *ifa;
1859
1860	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1861		if (ifa->ifa_addr->sa_family != AF_INET6)
1862			continue;
1863		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1864			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1865			     ignoreflags) != 0)
1866				continue;
1867			break;
1868		}
1869	}
1870
1871	return ((struct in6_ifaddr *)ifa);
1872}
1873
1874
1875/*
1876 * find the internet address corresponding to a given interface and address.
1877 */
1878struct in6_ifaddr *
1879in6ifa_ifpwithaddr(ifp, addr)
1880	struct ifnet *ifp;
1881	struct in6_addr *addr;
1882{
1883	struct ifaddr *ifa;
1884
1885	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1886		if (ifa->ifa_addr->sa_family != AF_INET6)
1887			continue;
1888		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1889			break;
1890	}
1891
1892	return ((struct in6_ifaddr *)ifa);
1893}
1894
1895/*
1896 * Convert IP6 address to printable (loggable) representation. Caller
1897 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1898 */
1899static char digits[] = "0123456789abcdef";
1900char *
1901ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1902{
1903	int i;
1904	char *cp;
1905	const u_int16_t *a = (const u_int16_t *)addr;
1906	const u_int8_t *d;
1907	int dcolon = 0, zero = 0;
1908
1909	cp = ip6buf;
1910
1911	for (i = 0; i < 8; i++) {
1912		if (dcolon == 1) {
1913			if (*a == 0) {
1914				if (i == 7)
1915					*cp++ = ':';
1916				a++;
1917				continue;
1918			} else
1919				dcolon = 2;
1920		}
1921		if (*a == 0) {
1922			if (dcolon == 0 && *(a + 1) == 0) {
1923				if (i == 0)
1924					*cp++ = ':';
1925				*cp++ = ':';
1926				dcolon = 1;
1927			} else {
1928				*cp++ = '0';
1929				*cp++ = ':';
1930			}
1931			a++;
1932			continue;
1933		}
1934		d = (const u_char *)a;
1935		/* Try to eliminate leading zeros in printout like in :0001. */
1936		zero = 1;
1937		*cp = digits[*d >> 4];
1938		if (*cp != '0') {
1939			zero = 0;
1940			cp++;
1941		}
1942		*cp = digits[*d++ & 0xf];
1943		if (zero == 0 || (*cp != '0')) {
1944			zero = 0;
1945			cp++;
1946		}
1947		*cp = digits[*d >> 4];
1948		if (zero == 0 || (*cp != '0')) {
1949			zero = 0;
1950			cp++;
1951		}
1952		*cp++ = digits[*d & 0xf];
1953		*cp++ = ':';
1954		a++;
1955	}
1956	*--cp = '\0';
1957	return (ip6buf);
1958}
1959
1960int
1961in6_localaddr(in6)
1962	struct in6_addr *in6;
1963{
1964	struct in6_ifaddr *ia;
1965
1966	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1967		return 1;
1968
1969	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1970		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1971		    &ia->ia_prefixmask.sin6_addr)) {
1972			return 1;
1973		}
1974	}
1975
1976	return (0);
1977}
1978
1979int
1980in6_is_addr_deprecated(sa6)
1981	struct sockaddr_in6 *sa6;
1982{
1983	struct in6_ifaddr *ia;
1984
1985	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1986		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1987				       &sa6->sin6_addr) &&
1988		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1989			return (1); /* true */
1990
1991		/* XXX: do we still have to go thru the rest of the list? */
1992	}
1993
1994	return (0);		/* false */
1995}
1996
1997/*
1998 * return length of part which dst and src are equal
1999 * hard coding...
2000 */
2001int
2002in6_matchlen(src, dst)
2003struct in6_addr *src, *dst;
2004{
2005	int match = 0;
2006	u_char *s = (u_char *)src, *d = (u_char *)dst;
2007	u_char *lim = s + 16, r;
2008
2009	while (s < lim)
2010		if ((r = (*d++ ^ *s++)) != 0) {
2011			while (r < 128) {
2012				match++;
2013				r <<= 1;
2014			}
2015			break;
2016		} else
2017			match += 8;
2018	return match;
2019}
2020
2021/* XXX: to be scope conscious */
2022int
2023in6_are_prefix_equal(p1, p2, len)
2024	struct in6_addr *p1, *p2;
2025	int len;
2026{
2027	int bytelen, bitlen;
2028
2029	/* sanity check */
2030	if (0 > len || len > 128) {
2031		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2032		    len);
2033		return (0);
2034	}
2035
2036	bytelen = len / 8;
2037	bitlen = len % 8;
2038
2039	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2040		return (0);
2041	if (bitlen != 0 &&
2042	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2043	    p2->s6_addr[bytelen] >> (8 - bitlen))
2044		return (0);
2045
2046	return (1);
2047}
2048
2049void
2050in6_prefixlen2mask(maskp, len)
2051	struct in6_addr *maskp;
2052	int len;
2053{
2054	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2055	int bytelen, bitlen, i;
2056
2057	/* sanity check */
2058	if (0 > len || len > 128) {
2059		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2060		    len);
2061		return;
2062	}
2063
2064	bzero(maskp, sizeof(*maskp));
2065	bytelen = len / 8;
2066	bitlen = len % 8;
2067	for (i = 0; i < bytelen; i++)
2068		maskp->s6_addr[i] = 0xff;
2069	if (bitlen)
2070		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2071}
2072
2073/*
2074 * return the best address out of the same scope. if no address was
2075 * found, return the first valid address from designated IF.
2076 */
2077struct in6_ifaddr *
2078in6_ifawithifp(ifp, dst)
2079	struct ifnet *ifp;
2080	struct in6_addr *dst;
2081{
2082	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2083	struct ifaddr *ifa;
2084	struct in6_ifaddr *besta = 0;
2085	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2086
2087	dep[0] = dep[1] = NULL;
2088
2089	/*
2090	 * We first look for addresses in the same scope.
2091	 * If there is one, return it.
2092	 * If two or more, return one which matches the dst longest.
2093	 * If none, return one of global addresses assigned other ifs.
2094	 */
2095	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2096		if (ifa->ifa_addr->sa_family != AF_INET6)
2097			continue;
2098		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2099			continue; /* XXX: is there any case to allow anycast? */
2100		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2101			continue; /* don't use this interface */
2102		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2103			continue;
2104		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2105			if (ip6_use_deprecated)
2106				dep[0] = (struct in6_ifaddr *)ifa;
2107			continue;
2108		}
2109
2110		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2111			/*
2112			 * call in6_matchlen() as few as possible
2113			 */
2114			if (besta) {
2115				if (blen == -1)
2116					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2117				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2118				if (tlen > blen) {
2119					blen = tlen;
2120					besta = (struct in6_ifaddr *)ifa;
2121				}
2122			} else
2123				besta = (struct in6_ifaddr *)ifa;
2124		}
2125	}
2126	if (besta)
2127		return (besta);
2128
2129	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2130		if (ifa->ifa_addr->sa_family != AF_INET6)
2131			continue;
2132		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2133			continue; /* XXX: is there any case to allow anycast? */
2134		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2135			continue; /* don't use this interface */
2136		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2137			continue;
2138		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2139			if (ip6_use_deprecated)
2140				dep[1] = (struct in6_ifaddr *)ifa;
2141			continue;
2142		}
2143
2144		return (struct in6_ifaddr *)ifa;
2145	}
2146
2147	/* use the last-resort values, that are, deprecated addresses */
2148	if (dep[0])
2149		return dep[0];
2150	if (dep[1])
2151		return dep[1];
2152
2153	return NULL;
2154}
2155
2156/*
2157 * perform DAD when interface becomes IFF_UP.
2158 */
2159void
2160in6_if_up(ifp)
2161	struct ifnet *ifp;
2162{
2163	struct ifaddr *ifa;
2164	struct in6_ifaddr *ia;
2165
2166	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2167		if (ifa->ifa_addr->sa_family != AF_INET6)
2168			continue;
2169		ia = (struct in6_ifaddr *)ifa;
2170		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2171			/*
2172			 * The TENTATIVE flag was likely set by hand
2173			 * beforehand, implicitly indicating the need for DAD.
2174			 * We may be able to skip the random delay in this
2175			 * case, but we impose delays just in case.
2176			 */
2177			nd6_dad_start(ifa,
2178			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2179		}
2180	}
2181
2182	/*
2183	 * special cases, like 6to4, are handled in in6_ifattach
2184	 */
2185	in6_ifattach(ifp, NULL);
2186}
2187
2188int
2189in6if_do_dad(ifp)
2190	struct ifnet *ifp;
2191{
2192	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2193		return (0);
2194
2195	switch (ifp->if_type) {
2196#ifdef IFT_DUMMY
2197	case IFT_DUMMY:
2198#endif
2199	case IFT_FAITH:
2200		/*
2201		 * These interfaces do not have the IFF_LOOPBACK flag,
2202		 * but loop packets back.  We do not have to do DAD on such
2203		 * interfaces.  We should even omit it, because loop-backed
2204		 * NS would confuse the DAD procedure.
2205		 */
2206		return (0);
2207	default:
2208		/*
2209		 * Our DAD routine requires the interface up and running.
2210		 * However, some interfaces can be up before the RUNNING
2211		 * status.  Additionaly, users may try to assign addresses
2212		 * before the interface becomes up (or running).
2213		 * We simply skip DAD in such a case as a work around.
2214		 * XXX: we should rather mark "tentative" on such addresses,
2215		 * and do DAD after the interface becomes ready.
2216		 */
2217		if (!((ifp->if_flags & IFF_UP) &&
2218		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2219			return (0);
2220
2221		return (1);
2222	}
2223}
2224
2225/*
2226 * Calculate max IPv6 MTU through all the interfaces and store it
2227 * to in6_maxmtu.
2228 */
2229void
2230in6_setmaxmtu()
2231{
2232	unsigned long maxmtu = 0;
2233	struct ifnet *ifp;
2234
2235	IFNET_RLOCK();
2236	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
2237		/* this function can be called during ifnet initialization */
2238		if (!ifp->if_afdata[AF_INET6])
2239			continue;
2240		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2241		    IN6_LINKMTU(ifp) > maxmtu)
2242			maxmtu = IN6_LINKMTU(ifp);
2243	}
2244	IFNET_RUNLOCK();
2245	if (maxmtu)	     /* update only when maxmtu is positive */
2246		in6_maxmtu = maxmtu;
2247}
2248
2249/*
2250 * Provide the length of interface identifiers to be used for the link attached
2251 * to the given interface.  The length should be defined in "IPv6 over
2252 * xxx-link" document.  Note that address architecture might also define
2253 * the length for a particular set of address prefixes, regardless of the
2254 * link type.  As clarified in rfc2462bis, those two definitions should be
2255 * consistent, and those really are as of August 2004.
2256 */
2257int
2258in6_if2idlen(ifp)
2259	struct ifnet *ifp;
2260{
2261	switch (ifp->if_type) {
2262	case IFT_ETHER:		/* RFC2464 */
2263#ifdef IFT_PROPVIRTUAL
2264	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2265#endif
2266#ifdef IFT_L2VLAN
2267	case IFT_L2VLAN:	/* ditto */
2268#endif
2269#ifdef IFT_IEEE80211
2270	case IFT_IEEE80211:	/* ditto */
2271#endif
2272#ifdef IFT_MIP
2273	case IFT_MIP:	/* ditto */
2274#endif
2275		return (64);
2276	case IFT_FDDI:		/* RFC2467 */
2277		return (64);
2278	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2279		return (64);
2280	case IFT_PPP:		/* RFC2472 */
2281		return (64);
2282	case IFT_ARCNET:	/* RFC2497 */
2283		return (64);
2284	case IFT_FRELAY:	/* RFC2590 */
2285		return (64);
2286	case IFT_IEEE1394:	/* RFC3146 */
2287		return (64);
2288	case IFT_GIF:
2289		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2290	case IFT_LOOP:
2291		return (64);	/* XXX: is this really correct? */
2292	default:
2293		/*
2294		 * Unknown link type:
2295		 * It might be controversial to use the today's common constant
2296		 * of 64 for these cases unconditionally.  For full compliance,
2297		 * we should return an error in this case.  On the other hand,
2298		 * if we simply miss the standard for the link type or a new
2299		 * standard is defined for a new link type, the IFID length
2300		 * is very likely to be the common constant.  As a compromise,
2301		 * we always use the constant, but make an explicit notice
2302		 * indicating the "unknown" case.
2303		 */
2304		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2305		return (64);
2306	}
2307}
2308
2309void *
2310in6_domifattach(ifp)
2311	struct ifnet *ifp;
2312{
2313	struct in6_ifextra *ext;
2314
2315	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2316	bzero(ext, sizeof(*ext));
2317
2318	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2319	    M_IFADDR, M_WAITOK);
2320	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2321
2322	ext->icmp6_ifstat =
2323	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2324	    M_IFADDR, M_WAITOK);
2325	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2326
2327	ext->nd_ifinfo = nd6_ifattach(ifp);
2328	ext->scope6_id = scope6_ifattach(ifp);
2329	return ext;
2330}
2331
2332void
2333in6_domifdetach(ifp, aux)
2334	struct ifnet *ifp;
2335	void *aux;
2336{
2337	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2338
2339	scope6_ifdetach(ext->scope6_id);
2340	nd6_ifdetach(ext->nd_ifinfo);
2341	free(ext->in6_ifstat, M_IFADDR);
2342	free(ext->icmp6_ifstat, M_IFADDR);
2343	free(ext, M_IFADDR);
2344}
2345
2346/*
2347 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2348 * v4 mapped addr or v4 compat addr
2349 */
2350void
2351in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2352{
2353	bzero(sin, sizeof(*sin));
2354	sin->sin_len = sizeof(struct sockaddr_in);
2355	sin->sin_family = AF_INET;
2356	sin->sin_port = sin6->sin6_port;
2357	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2358}
2359
2360/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2361void
2362in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2363{
2364	bzero(sin6, sizeof(*sin6));
2365	sin6->sin6_len = sizeof(struct sockaddr_in6);
2366	sin6->sin6_family = AF_INET6;
2367	sin6->sin6_port = sin->sin_port;
2368	sin6->sin6_addr.s6_addr32[0] = 0;
2369	sin6->sin6_addr.s6_addr32[1] = 0;
2370	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2371	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2372}
2373
2374/* Convert sockaddr_in6 into sockaddr_in. */
2375void
2376in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2377{
2378	struct sockaddr_in *sin_p;
2379	struct sockaddr_in6 sin6;
2380
2381	/*
2382	 * Save original sockaddr_in6 addr and convert it
2383	 * to sockaddr_in.
2384	 */
2385	sin6 = *(struct sockaddr_in6 *)nam;
2386	sin_p = (struct sockaddr_in *)nam;
2387	in6_sin6_2_sin(sin_p, &sin6);
2388}
2389
2390/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2391void
2392in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2393{
2394	struct sockaddr_in *sin_p;
2395	struct sockaddr_in6 *sin6_p;
2396
2397	MALLOC(sin6_p, struct sockaddr_in6 *, sizeof *sin6_p, M_SONAME,
2398	       M_WAITOK);
2399	sin_p = (struct sockaddr_in *)*nam;
2400	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2401	FREE(*nam, M_SONAME);
2402	*nam = (struct sockaddr *)sin6_p;
2403}
2404