in6.c revision e6f8b0995d59e493018009921005c7f50759dc53
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_LOCKED(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_LOCKED(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_LOCKED(rt);
272				in6_ifloop_request(RTM_DELETE, ifa);
273			} else
274				RT_UNLOCK(rt);
275		}
276	}
277}
278
279int
280in6_mask2len(struct in6_addr *mask, u_char *lim0)
281{
282	int x = 0, y;
283	u_char *lim = lim0, *p;
284
285	/* ignore the scope_id part */
286	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
287		lim = (u_char *)mask + sizeof(*mask);
288	for (p = (u_char *)mask; p < lim; x++, p++) {
289		if (*p != 0xff)
290			break;
291	}
292	y = 0;
293	if (p < lim) {
294		for (y = 0; y < 8; y++) {
295			if ((*p & (0x80 >> y)) == 0)
296				break;
297		}
298	}
299
300	/*
301	 * when the limit pointer is given, do a stricter check on the
302	 * remaining bits.
303	 */
304	if (p < lim) {
305		if (y != 0 && (*p & (0x00ff >> y)) != 0)
306			return (-1);
307		for (p = p + 1; p < lim; p++)
308			if (*p != 0)
309				return (-1);
310	}
311
312	return x * 8 + y;
313}
314
315#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
316#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
317
318int
319in6_control(struct socket *so, u_long cmd, caddr_t data,
320    struct ifnet *ifp, struct thread *td)
321{
322	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
323	struct	in6_ifaddr *ia = NULL;
324	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
325	struct sockaddr_in6 *sa6;
326	int error;
327
328	switch (cmd) {
329	case SIOCGETSGCNT_IN6:
330	case SIOCGETMIFCNT_IN6:
331		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
332	}
333
334	switch(cmd) {
335	case SIOCAADDRCTL_POLICY:
336	case SIOCDADDRCTL_POLICY:
337		if (td != NULL) {
338			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
339			if (error)
340				return (error);
341		}
342		return (in6_src_ioctl(cmd, data));
343	}
344
345	if (ifp == NULL)
346		return (EOPNOTSUPP);
347
348	switch (cmd) {
349	case SIOCSNDFLUSH_IN6:
350	case SIOCSPFXFLUSH_IN6:
351	case SIOCSRTRFLUSH_IN6:
352	case SIOCSDEFIFACE_IN6:
353	case SIOCSIFINFO_FLAGS:
354		if (td != NULL) {
355			error = priv_check(td, PRIV_NETINET_ND6);
356			if (error)
357				return (error);
358		}
359		/* FALLTHROUGH */
360	case OSIOCGIFINFO_IN6:
361	case SIOCGIFINFO_IN6:
362	case SIOCSIFINFO_IN6:
363	case SIOCGDRLST_IN6:
364	case SIOCGPRLST_IN6:
365	case SIOCGNBRINFO_IN6:
366	case SIOCGDEFIFACE_IN6:
367		return (nd6_ioctl(cmd, data, ifp));
368	}
369
370	switch (cmd) {
371	case SIOCSIFPREFIX_IN6:
372	case SIOCDIFPREFIX_IN6:
373	case SIOCAIFPREFIX_IN6:
374	case SIOCCIFPREFIX_IN6:
375	case SIOCSGIFPREFIX_IN6:
376	case SIOCGIFPREFIX_IN6:
377		log(LOG_NOTICE,
378		    "prefix ioctls are now invalidated. "
379		    "please use ifconfig.\n");
380		return (EOPNOTSUPP);
381	}
382
383	switch (cmd) {
384	case SIOCSSCOPE6:
385		if (td != NULL) {
386			error = priv_check(td, PRIV_NETINET_SCOPE6);
387			if (error)
388				return (error);
389		}
390		return (scope6_set(ifp,
391		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
392	case SIOCGSCOPE6:
393		return (scope6_get(ifp,
394		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
395	case SIOCGSCOPE6DEF:
396		return (scope6_get_default((struct scope6_id *)
397		    ifr->ifr_ifru.ifru_scope_id));
398	}
399
400	switch (cmd) {
401	case SIOCALIFADDR:
402	case SIOCDLIFADDR:
403		/*
404		 * XXXRW: Is this checked at another layer?  What priv to use
405		 * here?
406		 */
407		if (td != NULL) {
408			error = suser(td);
409			if (error)
410				return (error);
411		}
412		/* FALLTHROUGH */
413	case SIOCGLIFADDR:
414		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
415	}
416
417	/*
418	 * Find address for this interface, if it exists.
419	 *
420	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
421	 * only, and used the first interface address as the target of other
422	 * operations (without checking ifra_addr).  This was because netinet
423	 * code/API assumed at most 1 interface address per interface.
424	 * Since IPv6 allows a node to assign multiple addresses
425	 * on a single interface, we almost always look and check the
426	 * presence of ifra_addr, and reject invalid ones here.
427	 * It also decreases duplicated code among SIOC*_IN6 operations.
428	 */
429	switch (cmd) {
430	case SIOCAIFADDR_IN6:
431	case SIOCSIFPHYADDR_IN6:
432		sa6 = &ifra->ifra_addr;
433		break;
434	case SIOCSIFADDR_IN6:
435	case SIOCGIFADDR_IN6:
436	case SIOCSIFDSTADDR_IN6:
437	case SIOCSIFNETMASK_IN6:
438	case SIOCGIFDSTADDR_IN6:
439	case SIOCGIFNETMASK_IN6:
440	case SIOCDIFADDR_IN6:
441	case SIOCGIFPSRCADDR_IN6:
442	case SIOCGIFPDSTADDR_IN6:
443	case SIOCGIFAFLAG_IN6:
444	case SIOCSNDFLUSH_IN6:
445	case SIOCSPFXFLUSH_IN6:
446	case SIOCSRTRFLUSH_IN6:
447	case SIOCGIFALIFETIME_IN6:
448	case SIOCSIFALIFETIME_IN6:
449	case SIOCGIFSTAT_IN6:
450	case SIOCGIFSTAT_ICMP6:
451		sa6 = &ifr->ifr_addr;
452		break;
453	default:
454		sa6 = NULL;
455		break;
456	}
457	if (sa6 && sa6->sin6_family == AF_INET6) {
458		int error = 0;
459
460		if (sa6->sin6_scope_id != 0)
461			error = sa6_embedscope(sa6, 0);
462		else
463			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
464		if (error != 0)
465			return (error);
466		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
467	} else
468		ia = NULL;
469
470	switch (cmd) {
471	case SIOCSIFADDR_IN6:
472	case SIOCSIFDSTADDR_IN6:
473	case SIOCSIFNETMASK_IN6:
474		/*
475		 * Since IPv6 allows a node to assign multiple addresses
476		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
477		 */
478		/* we decided to obsolete this command (20000704) */
479		return (EINVAL);
480
481	case SIOCDIFADDR_IN6:
482		/*
483		 * for IPv4, we look for existing in_ifaddr here to allow
484		 * "ifconfig if0 delete" to remove the first IPv4 address on
485		 * the interface.  For IPv6, as the spec allows multiple
486		 * interface address from the day one, we consider "remove the
487		 * first one" semantics to be not preferable.
488		 */
489		if (ia == NULL)
490			return (EADDRNOTAVAIL);
491		/* FALLTHROUGH */
492	case SIOCAIFADDR_IN6:
493		/*
494		 * We always require users to specify a valid IPv6 address for
495		 * the corresponding operation.
496		 */
497		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
498		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
499			return (EAFNOSUPPORT);
500
501		/*
502		 * XXXRW: Is this checked at another layer?  What priv to use
503		 * here?
504		 */
505		if (td != NULL) {
506			error = suser(td);
507			if (error)
508				return (error);
509		}
510
511		break;
512
513	case SIOCGIFADDR_IN6:
514		/* This interface is basically deprecated. use SIOCGIFCONF. */
515		/* FALLTHROUGH */
516	case SIOCGIFAFLAG_IN6:
517	case SIOCGIFNETMASK_IN6:
518	case SIOCGIFDSTADDR_IN6:
519	case SIOCGIFALIFETIME_IN6:
520		/* must think again about its semantics */
521		if (ia == NULL)
522			return (EADDRNOTAVAIL);
523		break;
524	case SIOCSIFALIFETIME_IN6:
525	    {
526		struct in6_addrlifetime *lt;
527
528		if (td != NULL) {
529			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
530			if (error)
531				return (error);
532		}
533		if (ia == NULL)
534			return (EADDRNOTAVAIL);
535		/* sanity for overflow - beware unsigned */
536		lt = &ifr->ifr_ifru.ifru_lifetime;
537		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
538		    lt->ia6t_vltime + time_second < time_second) {
539			return EINVAL;
540		}
541		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
542		    lt->ia6t_pltime + time_second < time_second) {
543			return EINVAL;
544		}
545		break;
546	    }
547	}
548
549	switch (cmd) {
550
551	case SIOCGIFADDR_IN6:
552		ifr->ifr_addr = ia->ia_addr;
553		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
554			return (error);
555		break;
556
557	case SIOCGIFDSTADDR_IN6:
558		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
559			return (EINVAL);
560		/*
561		 * XXX: should we check if ifa_dstaddr is NULL and return
562		 * an error?
563		 */
564		ifr->ifr_dstaddr = ia->ia_dstaddr;
565		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
566			return (error);
567		break;
568
569	case SIOCGIFNETMASK_IN6:
570		ifr->ifr_addr = ia->ia_prefixmask;
571		break;
572
573	case SIOCGIFAFLAG_IN6:
574		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
575		break;
576
577	case SIOCGIFSTAT_IN6:
578		if (ifp == NULL)
579			return EINVAL;
580		bzero(&ifr->ifr_ifru.ifru_stat,
581		    sizeof(ifr->ifr_ifru.ifru_stat));
582		ifr->ifr_ifru.ifru_stat =
583		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
584		break;
585
586	case SIOCGIFSTAT_ICMP6:
587		if (ifp == NULL)
588			return EINVAL;
589		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
590		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
591		ifr->ifr_ifru.ifru_icmp6stat =
592		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
593		break;
594
595	case SIOCGIFALIFETIME_IN6:
596		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
597		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
598			time_t maxexpire;
599			struct in6_addrlifetime *retlt =
600			    &ifr->ifr_ifru.ifru_lifetime;
601
602			/*
603			 * XXX: adjust expiration time assuming time_t is
604			 * signed.
605			 */
606			maxexpire = (-1) &
607			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
608			if (ia->ia6_lifetime.ia6t_vltime <
609			    maxexpire - ia->ia6_updatetime) {
610				retlt->ia6t_expire = ia->ia6_updatetime +
611				    ia->ia6_lifetime.ia6t_vltime;
612			} else
613				retlt->ia6t_expire = maxexpire;
614		}
615		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
616			time_t maxexpire;
617			struct in6_addrlifetime *retlt =
618			    &ifr->ifr_ifru.ifru_lifetime;
619
620			/*
621			 * XXX: adjust expiration time assuming time_t is
622			 * signed.
623			 */
624			maxexpire = (-1) &
625			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
626			if (ia->ia6_lifetime.ia6t_pltime <
627			    maxexpire - ia->ia6_updatetime) {
628				retlt->ia6t_preferred = ia->ia6_updatetime +
629				    ia->ia6_lifetime.ia6t_pltime;
630			} else
631				retlt->ia6t_preferred = maxexpire;
632		}
633		break;
634
635	case SIOCSIFALIFETIME_IN6:
636		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
637		/* for sanity */
638		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
639			ia->ia6_lifetime.ia6t_expire =
640				time_second + ia->ia6_lifetime.ia6t_vltime;
641		} else
642			ia->ia6_lifetime.ia6t_expire = 0;
643		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
644			ia->ia6_lifetime.ia6t_preferred =
645				time_second + ia->ia6_lifetime.ia6t_pltime;
646		} else
647			ia->ia6_lifetime.ia6t_preferred = 0;
648		break;
649
650	case SIOCAIFADDR_IN6:
651	{
652		int i, error = 0;
653		struct nd_prefixctl pr0;
654		struct nd_prefix *pr;
655
656		/*
657		 * first, make or update the interface address structure,
658		 * and link it to the list.
659		 */
660		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
661			return (error);
662		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
663		    == NULL) {
664		    	/*
665			 * this can happen when the user specify the 0 valid
666			 * lifetime.
667			 */
668			break;
669		}
670
671		/*
672		 * then, make the prefix on-link on the interface.
673		 * XXX: we'd rather create the prefix before the address, but
674		 * we need at least one address to install the corresponding
675		 * interface route, so we configure the address first.
676		 */
677
678		/*
679		 * convert mask to prefix length (prefixmask has already
680		 * been validated in in6_update_ifa().
681		 */
682		bzero(&pr0, sizeof(pr0));
683		pr0.ndpr_ifp = ifp;
684		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
685		    NULL);
686		if (pr0.ndpr_plen == 128) {
687			break;	/* we don't need to install a host route. */
688		}
689		pr0.ndpr_prefix = ifra->ifra_addr;
690		/* apply the mask for safety. */
691		for (i = 0; i < 4; i++) {
692			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
693			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
694		}
695		/*
696		 * XXX: since we don't have an API to set prefix (not address)
697		 * lifetimes, we just use the same lifetimes as addresses.
698		 * The (temporarily) installed lifetimes can be overridden by
699		 * later advertised RAs (when accept_rtadv is non 0), which is
700		 * an intended behavior.
701		 */
702		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
703		pr0.ndpr_raf_auto =
704		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
705		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
706		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
707
708		/* add the prefix if not yet. */
709		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
710			/*
711			 * nd6_prelist_add will install the corresponding
712			 * interface route.
713			 */
714			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
715				return (error);
716			if (pr == NULL) {
717				log(LOG_ERR, "nd6_prelist_add succeeded but "
718				    "no prefix\n");
719				return (EINVAL); /* XXX panic here? */
720			}
721		}
722
723		/* relate the address to the prefix */
724		if (ia->ia6_ndpr == NULL) {
725			ia->ia6_ndpr = pr;
726			pr->ndpr_refcnt++;
727
728			/*
729			 * If this is the first autoconf address from the
730			 * prefix, create a temporary address as well
731			 * (when required).
732			 */
733			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
734			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
735				int e;
736				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
737					log(LOG_NOTICE, "in6_control: failed "
738					    "to create a temporary address, "
739					    "errno=%d\n", e);
740				}
741			}
742		}
743
744		/*
745		 * this might affect the status of autoconfigured addresses,
746		 * that is, this address might make other addresses detached.
747		 */
748		pfxlist_onlink_check();
749		if (error == 0 && ia)
750			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
751		break;
752	}
753
754	case SIOCDIFADDR_IN6:
755	{
756		struct nd_prefix *pr;
757
758		/*
759		 * If the address being deleted is the only one that owns
760		 * the corresponding prefix, expire the prefix as well.
761		 * XXX: theoretically, we don't have to worry about such
762		 * relationship, since we separate the address management
763		 * and the prefix management.  We do this, however, to provide
764		 * as much backward compatibility as possible in terms of
765		 * the ioctl operation.
766		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
767		 */
768		pr = ia->ia6_ndpr;
769		in6_purgeaddr(&ia->ia_ifa);
770		if (pr && pr->ndpr_refcnt == 0)
771			prelist_remove(pr);
772		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
773		break;
774	}
775
776	default:
777		if (ifp == NULL || ifp->if_ioctl == 0)
778			return (EOPNOTSUPP);
779		return ((*ifp->if_ioctl)(ifp, cmd, data));
780	}
781
782	return (0);
783}
784
785/*
786 * Update parameters of an IPv6 interface address.
787 * If necessary, a new entry is created and linked into address chains.
788 * This function is separated from in6_control().
789 * XXX: should this be performed under splnet()?
790 */
791int
792in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
793    struct in6_ifaddr *ia, int flags)
794{
795	int error = 0, hostIsNew = 0, plen = -1;
796	struct in6_ifaddr *oia;
797	struct sockaddr_in6 dst6;
798	struct in6_addrlifetime *lt;
799	struct in6_multi_mship *imm;
800	struct in6_multi *in6m_sol;
801	struct rtentry *rt;
802	int delay;
803	char ip6buf[INET6_ADDRSTRLEN];
804
805	/* Validate parameters */
806	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
807		return (EINVAL);
808
809	/*
810	 * The destination address for a p2p link must have a family
811	 * of AF_UNSPEC or AF_INET6.
812	 */
813	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
814	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
815	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
816		return (EAFNOSUPPORT);
817	/*
818	 * validate ifra_prefixmask.  don't check sin6_family, netmask
819	 * does not carry fields other than sin6_len.
820	 */
821	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
822		return (EINVAL);
823	/*
824	 * Because the IPv6 address architecture is classless, we require
825	 * users to specify a (non 0) prefix length (mask) for a new address.
826	 * We also require the prefix (when specified) mask is valid, and thus
827	 * reject a non-consecutive mask.
828	 */
829	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
830		return (EINVAL);
831	if (ifra->ifra_prefixmask.sin6_len != 0) {
832		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
833		    (u_char *)&ifra->ifra_prefixmask +
834		    ifra->ifra_prefixmask.sin6_len);
835		if (plen <= 0)
836			return (EINVAL);
837	} else {
838		/*
839		 * In this case, ia must not be NULL.  We just use its prefix
840		 * length.
841		 */
842		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
843	}
844	/*
845	 * If the destination address on a p2p interface is specified,
846	 * and the address is a scoped one, validate/set the scope
847	 * zone identifier.
848	 */
849	dst6 = ifra->ifra_dstaddr;
850	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
851	    (dst6.sin6_family == AF_INET6)) {
852		struct in6_addr in6_tmp;
853		u_int32_t zoneid;
854
855		in6_tmp = dst6.sin6_addr;
856		if (in6_setscope(&in6_tmp, ifp, &zoneid))
857			return (EINVAL); /* XXX: should be impossible */
858
859		if (dst6.sin6_scope_id != 0) {
860			if (dst6.sin6_scope_id != zoneid)
861				return (EINVAL);
862		} else		/* user omit to specify the ID. */
863			dst6.sin6_scope_id = zoneid;
864
865		/* convert into the internal form */
866		if (sa6_embedscope(&dst6, 0))
867			return (EINVAL); /* XXX: should be impossible */
868	}
869	/*
870	 * The destination address can be specified only for a p2p or a
871	 * loopback interface.  If specified, the corresponding prefix length
872	 * must be 128.
873	 */
874	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
875		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
876			/* XXX: noisy message */
877			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
878			    "be specified for a p2p or a loopback IF only\n"));
879			return (EINVAL);
880		}
881		if (plen != 128) {
882			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
883			    "be 128 when dstaddr is specified\n"));
884			return (EINVAL);
885		}
886	}
887	/* lifetime consistency check */
888	lt = &ifra->ifra_lifetime;
889	if (lt->ia6t_pltime > lt->ia6t_vltime)
890		return (EINVAL);
891	if (lt->ia6t_vltime == 0) {
892		/*
893		 * the following log might be noisy, but this is a typical
894		 * configuration mistake or a tool's bug.
895		 */
896		nd6log((LOG_INFO,
897		    "in6_update_ifa: valid lifetime is 0 for %s\n",
898		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
899
900		if (ia == NULL)
901			return (0); /* there's nothing to do */
902	}
903
904	/*
905	 * If this is a new address, allocate a new ifaddr and link it
906	 * into chains.
907	 */
908	if (ia == NULL) {
909		hostIsNew = 1;
910		/*
911		 * When in6_update_ifa() is called in a process of a received
912		 * RA, it is called under an interrupt context.  So, we should
913		 * call malloc with M_NOWAIT.
914		 */
915		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
916		    M_NOWAIT);
917		if (ia == NULL)
918			return (ENOBUFS);
919		bzero((caddr_t)ia, sizeof(*ia));
920		LIST_INIT(&ia->ia6_memberships);
921		/* Initialize the address and masks, and put time stamp */
922		IFA_LOCK_INIT(&ia->ia_ifa);
923		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
924		ia->ia_addr.sin6_family = AF_INET6;
925		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
926		ia->ia6_createtime = time_second;
927		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
928			/*
929			 * XXX: some functions expect that ifa_dstaddr is not
930			 * NULL for p2p interfaces.
931			 */
932			ia->ia_ifa.ifa_dstaddr =
933			    (struct sockaddr *)&ia->ia_dstaddr;
934		} else {
935			ia->ia_ifa.ifa_dstaddr = NULL;
936		}
937		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
938
939		ia->ia_ifp = ifp;
940		if ((oia = in6_ifaddr) != NULL) {
941			for ( ; oia->ia_next; oia = oia->ia_next)
942				continue;
943			oia->ia_next = ia;
944		} else
945			in6_ifaddr = ia;
946
947		ia->ia_ifa.ifa_refcnt = 1;
948		TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
949	}
950
951	/* update timestamp */
952	ia->ia6_updatetime = time_second;
953
954	/* set prefix mask */
955	if (ifra->ifra_prefixmask.sin6_len) {
956		/*
957		 * We prohibit changing the prefix length of an existing
958		 * address, because
959		 * + such an operation should be rare in IPv6, and
960		 * + the operation would confuse prefix management.
961		 */
962		if (ia->ia_prefixmask.sin6_len &&
963		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
964			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
965			    " existing (%s) address should not be changed\n",
966			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
967			error = EINVAL;
968			goto unlink;
969		}
970		ia->ia_prefixmask = ifra->ifra_prefixmask;
971	}
972
973	/*
974	 * If a new destination address is specified, scrub the old one and
975	 * install the new destination.  Note that the interface must be
976	 * p2p or loopback (see the check above.)
977	 */
978	if (dst6.sin6_family == AF_INET6 &&
979	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
980		int e;
981
982		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
983		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
984			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
985			    "a route to the old destination: %s\n",
986			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
987			/* proceed anyway... */
988		} else
989			ia->ia_flags &= ~IFA_ROUTE;
990		ia->ia_dstaddr = dst6;
991	}
992
993	/*
994	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
995	 * to see if the address is deprecated or invalidated, but initialize
996	 * these members for applications.
997	 */
998	ia->ia6_lifetime = ifra->ifra_lifetime;
999	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1000		ia->ia6_lifetime.ia6t_expire =
1001		    time_second + ia->ia6_lifetime.ia6t_vltime;
1002	} else
1003		ia->ia6_lifetime.ia6t_expire = 0;
1004	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1005		ia->ia6_lifetime.ia6t_preferred =
1006		    time_second + ia->ia6_lifetime.ia6t_pltime;
1007	} else
1008		ia->ia6_lifetime.ia6t_preferred = 0;
1009
1010	/* reset the interface and routing table appropriately. */
1011	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1012		goto unlink;
1013
1014	/*
1015	 * configure address flags.
1016	 */
1017	ia->ia6_flags = ifra->ifra_flags;
1018	/*
1019	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1020	 * userland, make it deprecated.
1021	 */
1022	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1023		ia->ia6_lifetime.ia6t_pltime = 0;
1024		ia->ia6_lifetime.ia6t_preferred = time_second;
1025	}
1026	/*
1027	 * Make the address tentative before joining multicast addresses,
1028	 * so that corresponding MLD responses would not have a tentative
1029	 * source address.
1030	 */
1031	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1032	if (hostIsNew && in6if_do_dad(ifp))
1033		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1034
1035	/*
1036	 * We are done if we have simply modified an existing address.
1037	 */
1038	if (!hostIsNew)
1039		return (error);
1040
1041	/*
1042	 * Beyond this point, we should call in6_purgeaddr upon an error,
1043	 * not just go to unlink.
1044	 */
1045
1046	/* Join necessary multicast groups */
1047	in6m_sol = NULL;
1048	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1049		struct sockaddr_in6 mltaddr, mltmask;
1050		struct in6_addr llsol;
1051
1052		/* join solicited multicast addr for new host id */
1053		bzero(&llsol, sizeof(struct in6_addr));
1054		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1055		llsol.s6_addr32[1] = 0;
1056		llsol.s6_addr32[2] = htonl(1);
1057		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1058		llsol.s6_addr8[12] = 0xff;
1059		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1060			/* XXX: should not happen */
1061			log(LOG_ERR, "in6_update_ifa: "
1062			    "in6_setscope failed\n");
1063			goto cleanup;
1064		}
1065		delay = 0;
1066		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1067			/*
1068			 * We need a random delay for DAD on the address
1069			 * being configured.  It also means delaying
1070			 * transmission of the corresponding MLD report to
1071			 * avoid report collision.
1072			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1073			 */
1074			delay = arc4random() %
1075			    (MAX_RTR_SOLICITATION_DELAY * hz);
1076		}
1077		imm = in6_joingroup(ifp, &llsol, &error, delay);
1078		if (imm == NULL) {
1079			nd6log((LOG_WARNING,
1080			    "in6_update_ifa: addmulti failed for "
1081			    "%s on %s (errno=%d)\n",
1082			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
1083			    error));
1084			in6_purgeaddr((struct ifaddr *)ia);
1085			return (error);
1086		}
1087		LIST_INSERT_HEAD(&ia->ia6_memberships,
1088		    imm, i6mm_chain);
1089		in6m_sol = imm->i6mm_maddr;
1090
1091		bzero(&mltmask, sizeof(mltmask));
1092		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1093		mltmask.sin6_family = AF_INET6;
1094		mltmask.sin6_addr = in6mask32;
1095#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1096
1097		/*
1098		 * join link-local all-nodes address
1099		 */
1100		bzero(&mltaddr, sizeof(mltaddr));
1101		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1102		mltaddr.sin6_family = AF_INET6;
1103		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1104		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1105		    0)
1106			goto cleanup; /* XXX: should not fail */
1107
1108		/*
1109		 * XXX: do we really need this automatic routes?
1110		 * We should probably reconsider this stuff.  Most applications
1111		 * actually do not need the routes, since they usually specify
1112		 * the outgoing interface.
1113		 */
1114		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1115		if (rt) {
1116			if (memcmp(&mltaddr.sin6_addr,
1117			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1118			    MLTMASK_LEN)) {
1119				RTFREE_LOCKED(rt);
1120				rt = NULL;
1121			}
1122		}
1123		if (!rt) {
1124			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1125			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1126			    (struct sockaddr *)&ia->ia_addr,
1127			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1128			    (struct rtentry **)0);
1129			if (error)
1130				goto cleanup;
1131		} else
1132			RTFREE_LOCKED(rt);
1133
1134		/*
1135		 * XXX: do we really need this automatic routes?
1136		 * We should probably reconsider this stuff.  Most applications
1137		 * actually do not need the routes, since they usually specify
1138		 * the outgoing interface.
1139		 */
1140		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1141		if (rt) {
1142			/* XXX: only works in !SCOPEDROUTING case. */
1143			if (memcmp(&mltaddr.sin6_addr,
1144			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1145			    MLTMASK_LEN)) {
1146				RTFREE_LOCKED(rt);
1147				rt = NULL;
1148			}
1149		}
1150		if (!rt) {
1151			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1152			    (struct sockaddr *)&ia->ia_addr,
1153			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1154			    (struct rtentry **)0);
1155			if (error)
1156				goto cleanup;
1157		} else {
1158			RTFREE_LOCKED(rt);
1159		}
1160
1161		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1162		if (!imm) {
1163			nd6log((LOG_WARNING,
1164			    "in6_update_ifa: addmulti failed for "
1165			    "%s on %s (errno=%d)\n",
1166			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1167			    if_name(ifp), error));
1168			goto cleanup;
1169		}
1170		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1171
1172		/*
1173		 * join node information group address
1174		 */
1175#define hostnamelen	strlen(hostname)
1176		delay = 0;
1177		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1178			/*
1179			 * The spec doesn't say anything about delay for this
1180			 * group, but the same logic should apply.
1181			 */
1182			delay = arc4random() %
1183			    (MAX_RTR_SOLICITATION_DELAY * hz);
1184		}
1185		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr.sin6_addr)
1186		    == 0) {
1187			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1188			    delay); /* XXX jinmei */
1189			if (!imm) {
1190				nd6log((LOG_WARNING, "in6_update_ifa: "
1191				    "addmulti failed for %s on %s "
1192				    "(errno=%d)\n",
1193				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1194				    if_name(ifp), error));
1195				/* XXX not very fatal, go on... */
1196			} else {
1197				LIST_INSERT_HEAD(&ia->ia6_memberships,
1198				    imm, i6mm_chain);
1199			}
1200		}
1201#undef hostnamelen
1202
1203		/*
1204		 * join interface-local all-nodes address.
1205		 * (ff01::1%ifN, and ff01::%ifN/32)
1206		 */
1207		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1208		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1209		    != 0)
1210			goto cleanup; /* XXX: should not fail */
1211		/* XXX: again, do we really need the route? */
1212		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1213		if (rt) {
1214			if (memcmp(&mltaddr.sin6_addr,
1215			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1216			    MLTMASK_LEN)) {
1217				RTFREE_LOCKED(rt);
1218				rt = NULL;
1219			}
1220		}
1221		if (!rt) {
1222			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1223			    (struct sockaddr *)&ia->ia_addr,
1224			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1225			    (struct rtentry **)0);
1226			if (error)
1227				goto cleanup;
1228		} else
1229			RTFREE_LOCKED(rt);
1230
1231		/* XXX: again, do we really need the route? */
1232		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1233		if (rt) {
1234			if (memcmp(&mltaddr.sin6_addr,
1235			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1236			    MLTMASK_LEN)) {
1237				RTFREE_LOCKED(rt);
1238				rt = NULL;
1239			}
1240		}
1241		if (!rt) {
1242			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1243			    (struct sockaddr *)&ia->ia_addr,
1244			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1245			    (struct rtentry **)0);
1246			if (error)
1247				goto cleanup;
1248		} else {
1249			RTFREE_LOCKED(rt);
1250		}
1251
1252		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1253		if (!imm) {
1254			nd6log((LOG_WARNING, "in6_update_ifa: "
1255			    "addmulti failed for %s on %s "
1256			    "(errno=%d)\n",
1257			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1258			    if_name(ifp), error));
1259			goto cleanup;
1260		}
1261		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1262#undef	MLTMASK_LEN
1263	}
1264
1265	/*
1266	 * Perform DAD, if needed.
1267	 * XXX It may be of use, if we can administratively
1268	 * disable DAD.
1269	 */
1270	if (hostIsNew && in6if_do_dad(ifp) &&
1271	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1272	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1273	{
1274		int mindelay, maxdelay;
1275
1276		delay = 0;
1277		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1278			/*
1279			 * We need to impose a delay before sending an NS
1280			 * for DAD.  Check if we also needed a delay for the
1281			 * corresponding MLD message.  If we did, the delay
1282			 * should be larger than the MLD delay (this could be
1283			 * relaxed a bit, but this simple logic is at least
1284			 * safe).
1285			 */
1286			mindelay = 0;
1287			if (in6m_sol != NULL &&
1288			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1289				mindelay = in6m_sol->in6m_timer;
1290			}
1291			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1292			if (maxdelay - mindelay == 0)
1293				delay = 0;
1294			else {
1295				delay =
1296				    (arc4random() % (maxdelay - mindelay)) +
1297				    mindelay;
1298			}
1299		}
1300		nd6_dad_start((struct ifaddr *)ia, delay);
1301	}
1302
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	return (error);
1313
1314  cleanup:
1315	in6_purgeaddr(&ia->ia_ifa);
1316	return error;
1317}
1318
1319void
1320in6_purgeaddr(struct ifaddr *ifa)
1321{
1322	struct ifnet *ifp = ifa->ifa_ifp;
1323	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1324	char ip6buf[INET6_ADDRSTRLEN];
1325	struct in6_multi_mship *imm;
1326
1327	/* stop DAD processing */
1328	nd6_dad_stop(ifa);
1329
1330	/*
1331	 * delete route to the destination of the address being purged.
1332	 * The interface must be p2p or loopback in this case.
1333	 */
1334	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1335		int e;
1336
1337		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1338		    != 0) {
1339			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1340			    "a route to the p2p destination: %s on %s, "
1341			    "errno=%d\n",
1342			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1343			    if_name(ifp), e);
1344			/* proceed anyway... */
1345		} else
1346			ia->ia_flags &= ~IFA_ROUTE;
1347	}
1348
1349	/* Remove ownaddr's loopback rtentry, if it exists. */
1350	in6_ifremloop(&(ia->ia_ifa));
1351
1352	/*
1353	 * leave from multicast groups we have joined for the interface
1354	 */
1355	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1356		LIST_REMOVE(imm, i6mm_chain);
1357		in6_leavegroup(imm);
1358	}
1359
1360	in6_unlink_ifa(ia, ifp);
1361}
1362
1363static void
1364in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1365{
1366	struct in6_ifaddr *oia;
1367	int	s = splnet();
1368
1369	TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1370
1371	oia = ia;
1372	if (oia == (ia = in6_ifaddr))
1373		in6_ifaddr = ia->ia_next;
1374	else {
1375		while (ia->ia_next && (ia->ia_next != oia))
1376			ia = ia->ia_next;
1377		if (ia->ia_next)
1378			ia->ia_next = oia->ia_next;
1379		else {
1380			/* search failed */
1381			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1382		}
1383	}
1384
1385	/*
1386	 * Release the reference to the base prefix.  There should be a
1387	 * positive reference.
1388	 */
1389	if (oia->ia6_ndpr == NULL) {
1390		nd6log((LOG_NOTICE,
1391		    "in6_unlink_ifa: autoconf'ed address "
1392		    "%p has no prefix\n", oia));
1393	} else {
1394		oia->ia6_ndpr->ndpr_refcnt--;
1395		oia->ia6_ndpr = NULL;
1396	}
1397
1398	/*
1399	 * Also, if the address being removed is autoconf'ed, call
1400	 * pfxlist_onlink_check() since the release might affect the status of
1401	 * other (detached) addresses.
1402	 */
1403	if ((oia->ia6_flags & IN6_IFF_AUTOCONF)) {
1404		pfxlist_onlink_check();
1405	}
1406
1407	/*
1408	 * release another refcnt for the link from in6_ifaddr.
1409	 * Note that we should decrement the refcnt at least once for all *BSD.
1410	 */
1411	IFAFREE(&oia->ia_ifa);
1412
1413	splx(s);
1414}
1415
1416void
1417in6_purgeif(struct ifnet *ifp)
1418{
1419	struct ifaddr *ifa, *nifa;
1420
1421	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa) {
1422		nifa = TAILQ_NEXT(ifa, ifa_list);
1423		if (ifa->ifa_addr->sa_family != AF_INET6)
1424			continue;
1425		in6_purgeaddr(ifa);
1426	}
1427
1428	in6_ifdetach(ifp);
1429}
1430
1431/*
1432 * SIOC[GAD]LIFADDR.
1433 *	SIOCGLIFADDR: get first address. (?)
1434 *	SIOCGLIFADDR with IFLR_PREFIX:
1435 *		get first address that matches the specified prefix.
1436 *	SIOCALIFADDR: add the specified address.
1437 *	SIOCALIFADDR with IFLR_PREFIX:
1438 *		add the specified prefix, filling hostid part from
1439 *		the first link-local address.  prefixlen must be <= 64.
1440 *	SIOCDLIFADDR: delete the specified address.
1441 *	SIOCDLIFADDR with IFLR_PREFIX:
1442 *		delete the first address that matches the specified prefix.
1443 * return values:
1444 *	EINVAL on invalid parameters
1445 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1446 *	other values may be returned from in6_ioctl()
1447 *
1448 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1449 * this is to accomodate address naming scheme other than RFC2374,
1450 * in the future.
1451 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1452 * address encoding scheme. (see figure on page 8)
1453 */
1454static int
1455in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1456    struct ifnet *ifp, struct thread *td)
1457{
1458	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1459	struct ifaddr *ifa;
1460	struct sockaddr *sa;
1461
1462	/* sanity checks */
1463	if (!data || !ifp) {
1464		panic("invalid argument to in6_lifaddr_ioctl");
1465		/* NOTREACHED */
1466	}
1467
1468	switch (cmd) {
1469	case SIOCGLIFADDR:
1470		/* address must be specified on GET with IFLR_PREFIX */
1471		if ((iflr->flags & IFLR_PREFIX) == 0)
1472			break;
1473		/* FALLTHROUGH */
1474	case SIOCALIFADDR:
1475	case SIOCDLIFADDR:
1476		/* address must be specified on ADD and DELETE */
1477		sa = (struct sockaddr *)&iflr->addr;
1478		if (sa->sa_family != AF_INET6)
1479			return EINVAL;
1480		if (sa->sa_len != sizeof(struct sockaddr_in6))
1481			return EINVAL;
1482		/* XXX need improvement */
1483		sa = (struct sockaddr *)&iflr->dstaddr;
1484		if (sa->sa_family && sa->sa_family != AF_INET6)
1485			return EINVAL;
1486		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1487			return EINVAL;
1488		break;
1489	default: /* shouldn't happen */
1490#if 0
1491		panic("invalid cmd to in6_lifaddr_ioctl");
1492		/* NOTREACHED */
1493#else
1494		return EOPNOTSUPP;
1495#endif
1496	}
1497	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1498		return EINVAL;
1499
1500	switch (cmd) {
1501	case SIOCALIFADDR:
1502	    {
1503		struct in6_aliasreq ifra;
1504		struct in6_addr *hostid = NULL;
1505		int prefixlen;
1506
1507		if ((iflr->flags & IFLR_PREFIX) != 0) {
1508			struct sockaddr_in6 *sin6;
1509
1510			/*
1511			 * hostid is to fill in the hostid part of the
1512			 * address.  hostid points to the first link-local
1513			 * address attached to the interface.
1514			 */
1515			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1516			if (!ifa)
1517				return EADDRNOTAVAIL;
1518			hostid = IFA_IN6(ifa);
1519
1520		 	/* prefixlen must be <= 64. */
1521			if (64 < iflr->prefixlen)
1522				return EINVAL;
1523			prefixlen = iflr->prefixlen;
1524
1525			/* hostid part must be zero. */
1526			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1527			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1528			    sin6->sin6_addr.s6_addr32[3] != 0) {
1529				return EINVAL;
1530			}
1531		} else
1532			prefixlen = iflr->prefixlen;
1533
1534		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1535		bzero(&ifra, sizeof(ifra));
1536		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1537
1538		bcopy(&iflr->addr, &ifra.ifra_addr,
1539		    ((struct sockaddr *)&iflr->addr)->sa_len);
1540		if (hostid) {
1541			/* fill in hostid part */
1542			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1543			    hostid->s6_addr32[2];
1544			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1545			    hostid->s6_addr32[3];
1546		}
1547
1548		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1549			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1550			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1551			if (hostid) {
1552				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1553				    hostid->s6_addr32[2];
1554				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1555				    hostid->s6_addr32[3];
1556			}
1557		}
1558
1559		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1560		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1561
1562		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1563		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1564	    }
1565	case SIOCGLIFADDR:
1566	case SIOCDLIFADDR:
1567	    {
1568		struct in6_ifaddr *ia;
1569		struct in6_addr mask, candidate, match;
1570		struct sockaddr_in6 *sin6;
1571		int cmp;
1572
1573		bzero(&mask, sizeof(mask));
1574		if (iflr->flags & IFLR_PREFIX) {
1575			/* lookup a prefix rather than address. */
1576			in6_prefixlen2mask(&mask, iflr->prefixlen);
1577
1578			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1579			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1580			match.s6_addr32[0] &= mask.s6_addr32[0];
1581			match.s6_addr32[1] &= mask.s6_addr32[1];
1582			match.s6_addr32[2] &= mask.s6_addr32[2];
1583			match.s6_addr32[3] &= mask.s6_addr32[3];
1584
1585			/* if you set extra bits, that's wrong */
1586			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1587				return EINVAL;
1588
1589			cmp = 1;
1590		} else {
1591			if (cmd == SIOCGLIFADDR) {
1592				/* on getting an address, take the 1st match */
1593				cmp = 0;	/* XXX */
1594			} else {
1595				/* on deleting an address, do exact match */
1596				in6_prefixlen2mask(&mask, 128);
1597				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1598				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1599
1600				cmp = 1;
1601			}
1602		}
1603
1604		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1605			if (ifa->ifa_addr->sa_family != AF_INET6)
1606				continue;
1607			if (!cmp)
1608				break;
1609
1610			/*
1611			 * XXX: this is adhoc, but is necessary to allow
1612			 * a user to specify fe80::/64 (not /10) for a
1613			 * link-local address.
1614			 */
1615			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1616			in6_clearscope(&candidate);
1617			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1618			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1619			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1620			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1621			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1622				break;
1623		}
1624		if (!ifa)
1625			return EADDRNOTAVAIL;
1626		ia = ifa2ia6(ifa);
1627
1628		if (cmd == SIOCGLIFADDR) {
1629			int error;
1630
1631			/* fill in the if_laddrreq structure */
1632			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1633			error = sa6_recoverscope(
1634			    (struct sockaddr_in6 *)&iflr->addr);
1635			if (error != 0)
1636				return (error);
1637
1638			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1639				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1640				    ia->ia_dstaddr.sin6_len);
1641				error = sa6_recoverscope(
1642				    (struct sockaddr_in6 *)&iflr->dstaddr);
1643				if (error != 0)
1644					return (error);
1645			} else
1646				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1647
1648			iflr->prefixlen =
1649			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1650
1651			iflr->flags = ia->ia6_flags;	/* XXX */
1652
1653			return 0;
1654		} else {
1655			struct in6_aliasreq ifra;
1656
1657			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1658			bzero(&ifra, sizeof(ifra));
1659			bcopy(iflr->iflr_name, ifra.ifra_name,
1660			    sizeof(ifra.ifra_name));
1661
1662			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1663			    ia->ia_addr.sin6_len);
1664			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1665				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1666				    ia->ia_dstaddr.sin6_len);
1667			} else {
1668				bzero(&ifra.ifra_dstaddr,
1669				    sizeof(ifra.ifra_dstaddr));
1670			}
1671			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1672			    ia->ia_prefixmask.sin6_len);
1673
1674			ifra.ifra_flags = ia->ia6_flags;
1675			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1676			    ifp, td);
1677		}
1678	    }
1679	}
1680
1681	return EOPNOTSUPP;	/* just for safety */
1682}
1683
1684/*
1685 * Initialize an interface's intetnet6 address
1686 * and routing table entry.
1687 */
1688static int
1689in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1690    struct sockaddr_in6 *sin6, int newhost)
1691{
1692	int	error = 0, plen, ifacount = 0;
1693	int	s = splimp();
1694	struct ifaddr *ifa;
1695
1696	/*
1697	 * Give the interface a chance to initialize
1698	 * if this is its first address,
1699	 * and to validate the address if necessary.
1700	 */
1701	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1702		if (ifa->ifa_addr->sa_family != AF_INET6)
1703			continue;
1704		ifacount++;
1705	}
1706
1707	ia->ia_addr = *sin6;
1708
1709	if (ifacount <= 1 && ifp->if_ioctl) {
1710		IFF_LOCKGIANT(ifp);
1711		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1712		IFF_UNLOCKGIANT(ifp);
1713		if (error) {
1714			splx(s);
1715			return (error);
1716		}
1717	}
1718	splx(s);
1719
1720	ia->ia_ifa.ifa_metric = ifp->if_metric;
1721
1722	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1723
1724	if (newhost) {
1725		/*
1726		 * set the rtrequest function to create llinfo.  It also
1727		 * adjust outgoing interface of the route for the local
1728		 * address when called via in6_ifaddloop() below.
1729		 */
1730		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1731	}
1732
1733	/*
1734	 * Special case:
1735	 * If a new destination address is specified for a point-to-point
1736	 * interface, install a route to the destination as an interface
1737	 * direct route.  In addition, if the link is expected to have neighbor
1738	 * cache entries, specify RTF_LLINFO so that a cache entry for the
1739	 * destination address will be created.
1740	 * created
1741	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1742	 * interface that share the same destination.
1743	 */
1744	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1745	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1746	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1747		int rtflags = RTF_UP | RTF_HOST;
1748		struct rtentry *rt = NULL, **rtp = NULL;
1749
1750		if (nd6_need_cache(ifp) != 0) {
1751			rtflags |= RTF_LLINFO;
1752			rtp = &rt;
1753		}
1754
1755		error = rtrequest(RTM_ADD, (struct sockaddr *)&ia->ia_dstaddr,
1756		    (struct sockaddr *)&ia->ia_addr,
1757		    (struct sockaddr *)&ia->ia_prefixmask,
1758		    ia->ia_flags | rtflags, rtp);
1759		if (error != 0)
1760			return (error);
1761		if (rt != NULL) {
1762			struct llinfo_nd6 *ln;
1763
1764			RT_LOCK(rt);
1765			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1766			if (ln != NULL) {
1767				/*
1768				 * Set the state to STALE because we don't
1769				 * have to perform address resolution on this
1770				 * link.
1771				 */
1772				ln->ln_state = ND6_LLINFO_STALE;
1773			}
1774			RT_REMREF(rt);
1775			RT_UNLOCK(rt);
1776		}
1777		ia->ia_flags |= IFA_ROUTE;
1778	}
1779	if (plen < 128) {
1780		/*
1781		 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1782		 */
1783		ia->ia_ifa.ifa_flags |= RTF_CLONING;
1784	}
1785
1786	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1787	if (newhost)
1788		in6_ifaddloop(&(ia->ia_ifa));
1789
1790	return (error);
1791}
1792
1793struct in6_multi_mship *
1794in6_joingroup(struct ifnet *ifp, struct in6_addr *addr,
1795    int *errorp, int delay)
1796{
1797	struct in6_multi_mship *imm;
1798
1799	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
1800	if (!imm) {
1801		*errorp = ENOBUFS;
1802		return NULL;
1803	}
1804	imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp, delay);
1805	if (!imm->i6mm_maddr) {
1806		/* *errorp is alrady set */
1807		free(imm, M_IP6MADDR);
1808		return NULL;
1809	}
1810	return imm;
1811}
1812
1813int
1814in6_leavegroup(struct in6_multi_mship *imm)
1815{
1816
1817	if (imm->i6mm_maddr)
1818		in6_delmulti(imm->i6mm_maddr);
1819	free(imm,  M_IP6MADDR);
1820	return 0;
1821}
1822
1823/*
1824 * Find an IPv6 interface link-local address specific to an interface.
1825 */
1826struct in6_ifaddr *
1827in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1828{
1829	struct ifaddr *ifa;
1830
1831	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1832		if (ifa->ifa_addr->sa_family != AF_INET6)
1833			continue;
1834		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1835			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1836			     ignoreflags) != 0)
1837				continue;
1838			break;
1839		}
1840	}
1841
1842	return ((struct in6_ifaddr *)ifa);
1843}
1844
1845
1846/*
1847 * find the internet address corresponding to a given interface and address.
1848 */
1849struct in6_ifaddr *
1850in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1851{
1852	struct ifaddr *ifa;
1853
1854	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1855		if (ifa->ifa_addr->sa_family != AF_INET6)
1856			continue;
1857		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1858			break;
1859	}
1860
1861	return ((struct in6_ifaddr *)ifa);
1862}
1863
1864/*
1865 * Convert IP6 address to printable (loggable) representation. Caller
1866 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1867 */
1868static char digits[] = "0123456789abcdef";
1869char *
1870ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1871{
1872	int i;
1873	char *cp;
1874	const u_int16_t *a = (const u_int16_t *)addr;
1875	const u_int8_t *d;
1876	int dcolon = 0, zero = 0;
1877
1878	cp = ip6buf;
1879
1880	for (i = 0; i < 8; i++) {
1881		if (dcolon == 1) {
1882			if (*a == 0) {
1883				if (i == 7)
1884					*cp++ = ':';
1885				a++;
1886				continue;
1887			} else
1888				dcolon = 2;
1889		}
1890		if (*a == 0) {
1891			if (dcolon == 0 && *(a + 1) == 0) {
1892				if (i == 0)
1893					*cp++ = ':';
1894				*cp++ = ':';
1895				dcolon = 1;
1896			} else {
1897				*cp++ = '0';
1898				*cp++ = ':';
1899			}
1900			a++;
1901			continue;
1902		}
1903		d = (const u_char *)a;
1904		/* Try to eliminate leading zeros in printout like in :0001. */
1905		zero = 1;
1906		*cp = digits[*d >> 4];
1907		if (*cp != '0') {
1908			zero = 0;
1909			cp++;
1910		}
1911		*cp = digits[*d++ & 0xf];
1912		if (zero == 0 || (*cp != '0')) {
1913			zero = 0;
1914			cp++;
1915		}
1916		*cp = digits[*d >> 4];
1917		if (zero == 0 || (*cp != '0')) {
1918			zero = 0;
1919			cp++;
1920		}
1921		*cp++ = digits[*d & 0xf];
1922		*cp++ = ':';
1923		a++;
1924	}
1925	*--cp = '\0';
1926	return (ip6buf);
1927}
1928
1929int
1930in6_localaddr(struct in6_addr *in6)
1931{
1932	struct in6_ifaddr *ia;
1933
1934	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1935		return 1;
1936
1937	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1938		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1939		    &ia->ia_prefixmask.sin6_addr)) {
1940			return 1;
1941		}
1942	}
1943
1944	return (0);
1945}
1946
1947int
1948in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1949{
1950	struct in6_ifaddr *ia;
1951
1952	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1953		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1954				       &sa6->sin6_addr) &&
1955		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1956			return (1); /* true */
1957
1958		/* XXX: do we still have to go thru the rest of the list? */
1959	}
1960
1961	return (0);		/* false */
1962}
1963
1964/*
1965 * return length of part which dst and src are equal
1966 * hard coding...
1967 */
1968int
1969in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1970{
1971	int match = 0;
1972	u_char *s = (u_char *)src, *d = (u_char *)dst;
1973	u_char *lim = s + 16, r;
1974
1975	while (s < lim)
1976		if ((r = (*d++ ^ *s++)) != 0) {
1977			while (r < 128) {
1978				match++;
1979				r <<= 1;
1980			}
1981			break;
1982		} else
1983			match += 8;
1984	return match;
1985}
1986
1987/* XXX: to be scope conscious */
1988int
1989in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1990{
1991	int bytelen, bitlen;
1992
1993	/* sanity check */
1994	if (0 > len || len > 128) {
1995		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1996		    len);
1997		return (0);
1998	}
1999
2000	bytelen = len / 8;
2001	bitlen = len % 8;
2002
2003	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2004		return (0);
2005	if (bitlen != 0 &&
2006	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2007	    p2->s6_addr[bytelen] >> (8 - bitlen))
2008		return (0);
2009
2010	return (1);
2011}
2012
2013void
2014in6_prefixlen2mask(struct in6_addr *maskp, int len)
2015{
2016	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2017	int bytelen, bitlen, i;
2018
2019	/* sanity check */
2020	if (0 > len || len > 128) {
2021		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2022		    len);
2023		return;
2024	}
2025
2026	bzero(maskp, sizeof(*maskp));
2027	bytelen = len / 8;
2028	bitlen = len % 8;
2029	for (i = 0; i < bytelen; i++)
2030		maskp->s6_addr[i] = 0xff;
2031	if (bitlen)
2032		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2033}
2034
2035/*
2036 * return the best address out of the same scope. if no address was
2037 * found, return the first valid address from designated IF.
2038 */
2039struct in6_ifaddr *
2040in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2041{
2042	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2043	struct ifaddr *ifa;
2044	struct in6_ifaddr *besta = 0;
2045	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2046
2047	dep[0] = dep[1] = NULL;
2048
2049	/*
2050	 * We first look for addresses in the same scope.
2051	 * If there is one, return it.
2052	 * If two or more, return one which matches the dst longest.
2053	 * If none, return one of global addresses assigned other ifs.
2054	 */
2055	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2056		if (ifa->ifa_addr->sa_family != AF_INET6)
2057			continue;
2058		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2059			continue; /* XXX: is there any case to allow anycast? */
2060		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2061			continue; /* don't use this interface */
2062		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2063			continue;
2064		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2065			if (ip6_use_deprecated)
2066				dep[0] = (struct in6_ifaddr *)ifa;
2067			continue;
2068		}
2069
2070		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2071			/*
2072			 * call in6_matchlen() as few as possible
2073			 */
2074			if (besta) {
2075				if (blen == -1)
2076					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2077				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2078				if (tlen > blen) {
2079					blen = tlen;
2080					besta = (struct in6_ifaddr *)ifa;
2081				}
2082			} else
2083				besta = (struct in6_ifaddr *)ifa;
2084		}
2085	}
2086	if (besta)
2087		return (besta);
2088
2089	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2090		if (ifa->ifa_addr->sa_family != AF_INET6)
2091			continue;
2092		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2093			continue; /* XXX: is there any case to allow anycast? */
2094		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2095			continue; /* don't use this interface */
2096		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2097			continue;
2098		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2099			if (ip6_use_deprecated)
2100				dep[1] = (struct in6_ifaddr *)ifa;
2101			continue;
2102		}
2103
2104		return (struct in6_ifaddr *)ifa;
2105	}
2106
2107	/* use the last-resort values, that are, deprecated addresses */
2108	if (dep[0])
2109		return dep[0];
2110	if (dep[1])
2111		return dep[1];
2112
2113	return NULL;
2114}
2115
2116/*
2117 * perform DAD when interface becomes IFF_UP.
2118 */
2119void
2120in6_if_up(struct ifnet *ifp)
2121{
2122	struct ifaddr *ifa;
2123	struct in6_ifaddr *ia;
2124
2125	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2126		if (ifa->ifa_addr->sa_family != AF_INET6)
2127			continue;
2128		ia = (struct in6_ifaddr *)ifa;
2129		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2130			/*
2131			 * The TENTATIVE flag was likely set by hand
2132			 * beforehand, implicitly indicating the need for DAD.
2133			 * We may be able to skip the random delay in this
2134			 * case, but we impose delays just in case.
2135			 */
2136			nd6_dad_start(ifa,
2137			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2138		}
2139	}
2140
2141	/*
2142	 * special cases, like 6to4, are handled in in6_ifattach
2143	 */
2144	in6_ifattach(ifp, NULL);
2145}
2146
2147int
2148in6if_do_dad(struct ifnet *ifp)
2149{
2150	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2151		return (0);
2152
2153	switch (ifp->if_type) {
2154#ifdef IFT_DUMMY
2155	case IFT_DUMMY:
2156#endif
2157	case IFT_FAITH:
2158		/*
2159		 * These interfaces do not have the IFF_LOOPBACK flag,
2160		 * but loop packets back.  We do not have to do DAD on such
2161		 * interfaces.  We should even omit it, because loop-backed
2162		 * NS would confuse the DAD procedure.
2163		 */
2164		return (0);
2165	default:
2166		/*
2167		 * Our DAD routine requires the interface up and running.
2168		 * However, some interfaces can be up before the RUNNING
2169		 * status.  Additionaly, users may try to assign addresses
2170		 * before the interface becomes up (or running).
2171		 * We simply skip DAD in such a case as a work around.
2172		 * XXX: we should rather mark "tentative" on such addresses,
2173		 * and do DAD after the interface becomes ready.
2174		 */
2175		if (!((ifp->if_flags & IFF_UP) &&
2176		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2177			return (0);
2178
2179		return (1);
2180	}
2181}
2182
2183/*
2184 * Calculate max IPv6 MTU through all the interfaces and store it
2185 * to in6_maxmtu.
2186 */
2187void
2188in6_setmaxmtu(void)
2189{
2190	unsigned long maxmtu = 0;
2191	struct ifnet *ifp;
2192
2193	IFNET_RLOCK();
2194	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
2195		/* this function can be called during ifnet initialization */
2196		if (!ifp->if_afdata[AF_INET6])
2197			continue;
2198		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2199		    IN6_LINKMTU(ifp) > maxmtu)
2200			maxmtu = IN6_LINKMTU(ifp);
2201	}
2202	IFNET_RUNLOCK();
2203	if (maxmtu)	     /* update only when maxmtu is positive */
2204		in6_maxmtu = maxmtu;
2205}
2206
2207/*
2208 * Provide the length of interface identifiers to be used for the link attached
2209 * to the given interface.  The length should be defined in "IPv6 over
2210 * xxx-link" document.  Note that address architecture might also define
2211 * the length for a particular set of address prefixes, regardless of the
2212 * link type.  As clarified in rfc2462bis, those two definitions should be
2213 * consistent, and those really are as of August 2004.
2214 */
2215int
2216in6_if2idlen(struct ifnet *ifp)
2217{
2218	switch (ifp->if_type) {
2219	case IFT_ETHER:		/* RFC2464 */
2220#ifdef IFT_PROPVIRTUAL
2221	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2222#endif
2223#ifdef IFT_L2VLAN
2224	case IFT_L2VLAN:	/* ditto */
2225#endif
2226#ifdef IFT_IEEE80211
2227	case IFT_IEEE80211:	/* ditto */
2228#endif
2229#ifdef IFT_MIP
2230	case IFT_MIP:	/* ditto */
2231#endif
2232		return (64);
2233	case IFT_FDDI:		/* RFC2467 */
2234		return (64);
2235	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2236		return (64);
2237	case IFT_PPP:		/* RFC2472 */
2238		return (64);
2239	case IFT_ARCNET:	/* RFC2497 */
2240		return (64);
2241	case IFT_FRELAY:	/* RFC2590 */
2242		return (64);
2243	case IFT_IEEE1394:	/* RFC3146 */
2244		return (64);
2245	case IFT_GIF:
2246		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2247	case IFT_LOOP:
2248		return (64);	/* XXX: is this really correct? */
2249	default:
2250		/*
2251		 * Unknown link type:
2252		 * It might be controversial to use the today's common constant
2253		 * of 64 for these cases unconditionally.  For full compliance,
2254		 * we should return an error in this case.  On the other hand,
2255		 * if we simply miss the standard for the link type or a new
2256		 * standard is defined for a new link type, the IFID length
2257		 * is very likely to be the common constant.  As a compromise,
2258		 * we always use the constant, but make an explicit notice
2259		 * indicating the "unknown" case.
2260		 */
2261		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2262		return (64);
2263	}
2264}
2265
2266void *
2267in6_domifattach(struct ifnet *ifp)
2268{
2269	struct in6_ifextra *ext;
2270
2271	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2272	bzero(ext, sizeof(*ext));
2273
2274	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2275	    M_IFADDR, M_WAITOK);
2276	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2277
2278	ext->icmp6_ifstat =
2279	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2280	    M_IFADDR, M_WAITOK);
2281	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2282
2283	ext->nd_ifinfo = nd6_ifattach(ifp);
2284	ext->scope6_id = scope6_ifattach(ifp);
2285	return ext;
2286}
2287
2288void
2289in6_domifdetach(struct ifnet *ifp, void *aux)
2290{
2291	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2292
2293	scope6_ifdetach(ext->scope6_id);
2294	nd6_ifdetach(ext->nd_ifinfo);
2295	free(ext->in6_ifstat, M_IFADDR);
2296	free(ext->icmp6_ifstat, M_IFADDR);
2297	free(ext, M_IFADDR);
2298}
2299
2300/*
2301 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2302 * v4 mapped addr or v4 compat addr
2303 */
2304void
2305in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2306{
2307
2308	bzero(sin, sizeof(*sin));
2309	sin->sin_len = sizeof(struct sockaddr_in);
2310	sin->sin_family = AF_INET;
2311	sin->sin_port = sin6->sin6_port;
2312	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2313}
2314
2315/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2316void
2317in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2318{
2319	bzero(sin6, sizeof(*sin6));
2320	sin6->sin6_len = sizeof(struct sockaddr_in6);
2321	sin6->sin6_family = AF_INET6;
2322	sin6->sin6_port = sin->sin_port;
2323	sin6->sin6_addr.s6_addr32[0] = 0;
2324	sin6->sin6_addr.s6_addr32[1] = 0;
2325	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2326	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2327}
2328
2329/* Convert sockaddr_in6 into sockaddr_in. */
2330void
2331in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2332{
2333	struct sockaddr_in *sin_p;
2334	struct sockaddr_in6 sin6;
2335
2336	/*
2337	 * Save original sockaddr_in6 addr and convert it
2338	 * to sockaddr_in.
2339	 */
2340	sin6 = *(struct sockaddr_in6 *)nam;
2341	sin_p = (struct sockaddr_in *)nam;
2342	in6_sin6_2_sin(sin_p, &sin6);
2343}
2344
2345/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2346void
2347in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2348{
2349	struct sockaddr_in *sin_p;
2350	struct sockaddr_in6 *sin6_p;
2351
2352	MALLOC(sin6_p, struct sockaddr_in6 *, sizeof *sin6_p, M_SONAME,
2353	       M_WAITOK);
2354	sin_p = (struct sockaddr_in *)*nam;
2355	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2356	FREE(*nam, M_SONAME);
2357	*nam = (struct sockaddr *)sin6_p;
2358}
2359