in6.c revision 7db27341e03c3b01c3eeb8999e45320bea24386a
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/proc.h>
75#include <sys/time.h>
76#include <sys/kernel.h>
77#include <sys/syslog.h>
78
79#include <net/if.h>
80#include <net/if_types.h>
81#include <net/route.h>
82#include <net/if_dl.h>
83
84#include <netinet/in.h>
85#include <netinet/in_var.h>
86#include <netinet/if_ether.h>
87#include <netinet/in_systm.h>
88#include <netinet/ip.h>
89#include <netinet/in_pcb.h>
90
91#include <netinet/ip6.h>
92#include <netinet6/ip6_var.h>
93#include <netinet6/nd6.h>
94#include <netinet6/mld6_var.h>
95#include <netinet6/ip6_mroute.h>
96#include <netinet6/in6_ifattach.h>
97#include <netinet6/scope6_var.h>
98#include <netinet6/in6_pcb.h>
99
100#include <net/net_osdep.h>
101
102MALLOC_DEFINE(M_IP6MADDR, "in6_multi", "internet multicast address");
103
104/*
105 * Definitions of some costant IP6 addresses.
106 */
107const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
108const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
109const struct in6_addr in6addr_nodelocal_allnodes =
110	IN6ADDR_NODELOCAL_ALLNODES_INIT;
111const struct in6_addr in6addr_linklocal_allnodes =
112	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
113const struct in6_addr in6addr_linklocal_allrouters =
114	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
115
116const struct in6_addr in6mask0 = IN6MASK0;
117const struct in6_addr in6mask32 = IN6MASK32;
118const struct in6_addr in6mask64 = IN6MASK64;
119const struct in6_addr in6mask96 = IN6MASK96;
120const struct in6_addr in6mask128 = IN6MASK128;
121
122const struct sockaddr_in6 sa6_any =
123	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
124
125static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
126	struct ifnet *, struct thread *));
127static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
128	struct sockaddr_in6 *, int));
129static void in6_unlink_ifa __P((struct in6_ifaddr *, struct ifnet *));
130
131struct in6_multihead in6_multihead;	/* XXX BSS initialization */
132int	(*faithprefix_p)(struct in6_addr *);
133
134/*
135 * Subroutine for in6_ifaddloop() and in6_ifremloop().
136 * This routine does actual work.
137 */
138static void
139in6_ifloop_request(int cmd, struct ifaddr *ifa)
140{
141	struct sockaddr_in6 all1_sa;
142	struct rtentry *nrt = NULL;
143	int e;
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		log(LOG_ERR, "in6_ifloop_request: "
163		    "%s operation failed for %s (errno=%d)\n",
164		    cmd == RTM_ADD ? "ADD" : "DELETE",
165		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
166		    e);
167	}
168
169	/*
170	 * Report the addition/removal of the address to the routing socket.
171	 * XXX: since we called rtinit for a p2p interface with a destination,
172	 *      we end up reporting twice in such a case.  Should we rather
173	 *      omit the second report?
174	 */
175	if (nrt) {
176		RT_LOCK(nrt);
177		/*
178		 * Make sure rt_ifa be equal to IFA, the second argument of
179		 * the function.  We need this because when we refer to
180		 * rt_ifa->ia6_flags in ip6_input, we assume that the rt_ifa
181		 * points to the address instead of the loopback address.
182		 */
183		if (cmd == RTM_ADD && ifa != nrt->rt_ifa) {
184			IFAFREE(nrt->rt_ifa);
185			IFAREF(ifa);
186			nrt->rt_ifa = ifa;
187		}
188
189		rt_newaddrmsg(cmd, ifa, e, nrt);
190		if (cmd == RTM_DELETE) {
191			rtfree(nrt);
192		} else {
193			/* the cmd must be RTM_ADD here */
194			RT_REMREF(nrt);
195			RT_UNLOCK(nrt);
196		}
197	}
198}
199
200/*
201 * Add ownaddr as loopback rtentry.  We previously add the route only if
202 * necessary (ex. on a p2p link).  However, since we now manage addresses
203 * separately from prefixes, we should always add the route.  We can't
204 * rely on the cloning mechanism from the corresponding interface route
205 * any more.
206 */
207void
208in6_ifaddloop(struct ifaddr *ifa)
209{
210	struct rtentry *rt;
211	int need_loop;
212
213	/* If there is no loopback entry, allocate one. */
214	rt = rtalloc1(ifa->ifa_addr, 0, 0);
215	need_loop = (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
216	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0);
217	if (rt)
218		rtfree(rt);
219	if (need_loop)
220		in6_ifloop_request(RTM_ADD, ifa);
221}
222
223/*
224 * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
225 * if it exists.
226 */
227void
228in6_ifremloop(struct ifaddr *ifa)
229{
230	struct in6_ifaddr *ia;
231	struct rtentry *rt;
232	int ia_count = 0;
233
234	/*
235	 * Some of BSD variants do not remove cloned routes
236	 * from an interface direct route, when removing the direct route
237	 * (see comments in net/net_osdep.h).  Even for variants that do remove
238	 * cloned routes, they could fail to remove the cloned routes when
239	 * we handle multple addresses that share a common prefix.
240	 * So, we should remove the route corresponding to the deleted address.
241	 */
242
243	/*
244	 * Delete the entry only if exact one ifa exists.  More than one ifa
245	 * can exist if we assign a same single address to multiple
246	 * (probably p2p) interfaces.
247	 * XXX: we should avoid such a configuration in IPv6...
248	 */
249	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
250		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
251			ia_count++;
252			if (ia_count > 1)
253				break;
254		}
255	}
256
257	if (ia_count == 1) {
258		/*
259		 * Before deleting, check if a corresponding loopbacked host
260		 * route surely exists.  With this check, we can avoid to
261		 * delete an interface direct route whose destination is same
262		 * as the address being removed.  This can happen when removing
263		 * a subnet-router anycast address on an interface attahced
264		 * to a shared medium.
265		 */
266		rt = rtalloc1(ifa->ifa_addr, 0, 0);
267		if (rt != NULL) {
268			if ((rt->rt_flags & RTF_HOST) != 0 &&
269			    (rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
270				rtfree(rt);
271				in6_ifloop_request(RTM_DELETE, ifa);
272			} else
273				RT_UNLOCK(rt);
274		}
275	}
276}
277
278int
279in6_mask2len(mask, lim0)
280	struct in6_addr *mask;
281	u_char *lim0;
282{
283	int x = 0, y;
284	u_char *lim = lim0, *p;
285
286	/* ignore the scope_id part */
287	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
288		lim = (u_char *)mask + sizeof(*mask);
289	for (p = (u_char *)mask; p < lim; x++, p++) {
290		if (*p != 0xff)
291			break;
292	}
293	y = 0;
294	if (p < lim) {
295		for (y = 0; y < 8; y++) {
296			if ((*p & (0x80 >> y)) == 0)
297				break;
298		}
299	}
300
301	/*
302	 * when the limit pointer is given, do a stricter check on the
303	 * remaining bits.
304	 */
305	if (p < lim) {
306		if (y != 0 && (*p & (0x00ff >> y)) != 0)
307			return (-1);
308		for (p = p + 1; p < lim; p++)
309			if (*p != 0)
310				return (-1);
311	}
312
313	return x * 8 + y;
314}
315
316#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
317#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
318
319int
320in6_control(so, cmd, data, ifp, td)
321	struct	socket *so;
322	u_long cmd;
323	caddr_t	data;
324	struct ifnet *ifp;
325	struct thread *td;
326{
327	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
328	struct	in6_ifaddr *ia = NULL;
329	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
330	int error, privileged;
331	struct sockaddr_in6 *sa6;
332
333	privileged = 0;
334	if (td == NULL || !suser(td))
335		privileged++;
336
337	switch (cmd) {
338	case SIOCGETSGCNT_IN6:
339	case SIOCGETMIFCNT_IN6:
340		return (mrt6_ioctl(cmd, data));
341	}
342
343	switch(cmd) {
344	case SIOCAADDRCTL_POLICY:
345	case SIOCDADDRCTL_POLICY:
346		if (!privileged)
347			return (EPERM);
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 (!privileged)
361			return (EPERM);
362		/* FALLTHROUGH */
363	case OSIOCGIFINFO_IN6:
364	case SIOCGIFINFO_IN6:
365	case SIOCSIFINFO_IN6:
366	case SIOCGDRLST_IN6:
367	case SIOCGPRLST_IN6:
368	case SIOCGNBRINFO_IN6:
369	case SIOCGDEFIFACE_IN6:
370		return (nd6_ioctl(cmd, data, ifp));
371	}
372
373	switch (cmd) {
374	case SIOCSIFPREFIX_IN6:
375	case SIOCDIFPREFIX_IN6:
376	case SIOCAIFPREFIX_IN6:
377	case SIOCCIFPREFIX_IN6:
378	case SIOCSGIFPREFIX_IN6:
379	case SIOCGIFPREFIX_IN6:
380		log(LOG_NOTICE,
381		    "prefix ioctls are now invalidated. "
382		    "please use ifconfig.\n");
383		return (EOPNOTSUPP);
384	}
385
386	switch (cmd) {
387	case SIOCSSCOPE6:
388		if (!privileged)
389			return (EPERM);
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		if (!privileged)
404			return (EPERM);
405		/* FALLTHROUGH */
406	case SIOCGLIFADDR:
407		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
408	}
409
410	/*
411	 * Find address for this interface, if it exists.
412	 *
413	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
414	 * only, and used the first interface address as the target of other
415	 * operations (without checking ifra_addr).  This was because netinet
416	 * code/API assumed at most 1 interface address per interface.
417	 * Since IPv6 allows a node to assign multiple addresses
418	 * on a single interface, we almost always look and check the
419	 * presence of ifra_addr, and reject invalid ones here.
420	 * It also decreases duplicated code among SIOC*_IN6 operations.
421	 */
422	switch (cmd) {
423	case SIOCAIFADDR_IN6:
424	case SIOCSIFPHYADDR_IN6:
425		sa6 = &ifra->ifra_addr;
426		break;
427	case SIOCSIFADDR_IN6:
428	case SIOCGIFADDR_IN6:
429	case SIOCSIFDSTADDR_IN6:
430	case SIOCSIFNETMASK_IN6:
431	case SIOCGIFDSTADDR_IN6:
432	case SIOCGIFNETMASK_IN6:
433	case SIOCDIFADDR_IN6:
434	case SIOCGIFPSRCADDR_IN6:
435	case SIOCGIFPDSTADDR_IN6:
436	case SIOCGIFAFLAG_IN6:
437	case SIOCSNDFLUSH_IN6:
438	case SIOCSPFXFLUSH_IN6:
439	case SIOCSRTRFLUSH_IN6:
440	case SIOCGIFALIFETIME_IN6:
441	case SIOCSIFALIFETIME_IN6:
442	case SIOCGIFSTAT_IN6:
443	case SIOCGIFSTAT_ICMP6:
444		sa6 = &ifr->ifr_addr;
445		break;
446	default:
447		sa6 = NULL;
448		break;
449	}
450	if (sa6 && sa6->sin6_family == AF_INET6) {
451		int error = 0;
452
453		if (sa6->sin6_scope_id != 0)
454			error = sa6_embedscope(sa6, 0);
455		else
456			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
457		if (error != 0)
458			return (error);
459		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
460	} else
461		ia = NULL;
462
463	switch (cmd) {
464	case SIOCSIFADDR_IN6:
465	case SIOCSIFDSTADDR_IN6:
466	case SIOCSIFNETMASK_IN6:
467		/*
468		 * Since IPv6 allows a node to assign multiple addresses
469		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
470		 */
471		/* we decided to obsolete this command (20000704) */
472		return (EINVAL);
473
474	case SIOCDIFADDR_IN6:
475		/*
476		 * for IPv4, we look for existing in_ifaddr here to allow
477		 * "ifconfig if0 delete" to remove the first IPv4 address on
478		 * the interface.  For IPv6, as the spec allows multiple
479		 * interface address from the day one, we consider "remove the
480		 * first one" semantics to be not preferable.
481		 */
482		if (ia == NULL)
483			return (EADDRNOTAVAIL);
484		/* FALLTHROUGH */
485	case SIOCAIFADDR_IN6:
486		/*
487		 * We always require users to specify a valid IPv6 address for
488		 * the corresponding operation.
489		 */
490		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
491		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
492			return (EAFNOSUPPORT);
493		if (!privileged)
494			return (EPERM);
495
496		break;
497
498	case SIOCGIFADDR_IN6:
499		/* This interface is basically deprecated. use SIOCGIFCONF. */
500		/* FALLTHROUGH */
501	case SIOCGIFAFLAG_IN6:
502	case SIOCGIFNETMASK_IN6:
503	case SIOCGIFDSTADDR_IN6:
504	case SIOCGIFALIFETIME_IN6:
505		/* must think again about its semantics */
506		if (ia == NULL)
507			return (EADDRNOTAVAIL);
508		break;
509	case SIOCSIFALIFETIME_IN6:
510	    {
511		struct in6_addrlifetime *lt;
512
513		if (!privileged)
514			return (EPERM);
515		if (ia == NULL)
516			return (EADDRNOTAVAIL);
517		/* sanity for overflow - beware unsigned */
518		lt = &ifr->ifr_ifru.ifru_lifetime;
519		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
520		    lt->ia6t_vltime + time_second < time_second) {
521			return EINVAL;
522		}
523		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
524		    lt->ia6t_pltime + time_second < time_second) {
525			return EINVAL;
526		}
527		break;
528	    }
529	}
530
531	switch (cmd) {
532
533	case SIOCGIFADDR_IN6:
534		ifr->ifr_addr = ia->ia_addr;
535		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
536			return (error);
537		break;
538
539	case SIOCGIFDSTADDR_IN6:
540		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
541			return (EINVAL);
542		/*
543		 * XXX: should we check if ifa_dstaddr is NULL and return
544		 * an error?
545		 */
546		ifr->ifr_dstaddr = ia->ia_dstaddr;
547		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
548			return (error);
549		break;
550
551	case SIOCGIFNETMASK_IN6:
552		ifr->ifr_addr = ia->ia_prefixmask;
553		break;
554
555	case SIOCGIFAFLAG_IN6:
556		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
557		break;
558
559	case SIOCGIFSTAT_IN6:
560		if (ifp == NULL)
561			return EINVAL;
562		bzero(&ifr->ifr_ifru.ifru_stat,
563		    sizeof(ifr->ifr_ifru.ifru_stat));
564		ifr->ifr_ifru.ifru_stat =
565		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
566		break;
567
568	case SIOCGIFSTAT_ICMP6:
569		if (ifp == NULL)
570			return EINVAL;
571		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
572		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
573		ifr->ifr_ifru.ifru_icmp6stat =
574		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
575		break;
576
577	case SIOCGIFALIFETIME_IN6:
578		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
579		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
580			time_t maxexpire;
581			struct in6_addrlifetime *retlt =
582			    &ifr->ifr_ifru.ifru_lifetime;
583
584			/*
585			 * XXX: adjust expiration time assuming time_t is
586			 * signed.
587			 */
588			maxexpire = (-1) &
589			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
590			if (ia->ia6_lifetime.ia6t_vltime <
591			    maxexpire - ia->ia6_updatetime) {
592				retlt->ia6t_expire = ia->ia6_updatetime +
593				    ia->ia6_lifetime.ia6t_vltime;
594			} else
595				retlt->ia6t_expire = maxexpire;
596		}
597		if (ia->ia6_lifetime.ia6t_pltime != 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_pltime <
609			    maxexpire - ia->ia6_updatetime) {
610				retlt->ia6t_preferred = ia->ia6_updatetime +
611				    ia->ia6_lifetime.ia6t_pltime;
612			} else
613				retlt->ia6t_preferred = maxexpire;
614		}
615		break;
616
617	case SIOCSIFALIFETIME_IN6:
618		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
619		/* for sanity */
620		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
621			ia->ia6_lifetime.ia6t_expire =
622				time_second + ia->ia6_lifetime.ia6t_vltime;
623		} else
624			ia->ia6_lifetime.ia6t_expire = 0;
625		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
626			ia->ia6_lifetime.ia6t_preferred =
627				time_second + ia->ia6_lifetime.ia6t_pltime;
628		} else
629			ia->ia6_lifetime.ia6t_preferred = 0;
630		break;
631
632	case SIOCAIFADDR_IN6:
633	{
634		int i, error = 0;
635		struct nd_prefixctl pr0;
636		struct nd_prefix *pr;
637
638		/*
639		 * first, make or update the interface address structure,
640		 * and link it to the list.
641		 */
642		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
643			return (error);
644		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
645		    == NULL) {
646		    	/*
647			 * this can happen when the user specify the 0 valid
648			 * lifetime.
649			 */
650			break;
651		}
652
653		/*
654		 * then, make the prefix on-link on the interface.
655		 * XXX: we'd rather create the prefix before the address, but
656		 * we need at least one address to install the corresponding
657		 * interface route, so we configure the address first.
658		 */
659
660		/*
661		 * convert mask to prefix length (prefixmask has already
662		 * been validated in in6_update_ifa().
663		 */
664		bzero(&pr0, sizeof(pr0));
665		pr0.ndpr_ifp = ifp;
666		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
667		    NULL);
668		if (pr0.ndpr_plen == 128) {
669			break;	/* we don't need to install a host route. */
670		}
671		pr0.ndpr_prefix = ifra->ifra_addr;
672		/* apply the mask for safety. */
673		for (i = 0; i < 4; i++) {
674			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
675			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
676		}
677		/*
678		 * XXX: since we don't have an API to set prefix (not address)
679		 * lifetimes, we just use the same lifetimes as addresses.
680		 * The (temporarily) installed lifetimes can be overridden by
681		 * later advertised RAs (when accept_rtadv is non 0), which is
682		 * an intended behavior.
683		 */
684		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
685		pr0.ndpr_raf_auto =
686		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
687		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
688		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
689
690		/* add the prefix if not yet. */
691		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
692			/*
693			 * nd6_prelist_add will install the corresponding
694			 * interface route.
695			 */
696			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
697				return (error);
698			if (pr == NULL) {
699				log(LOG_ERR, "nd6_prelist_add succeeded but "
700				    "no prefix\n");
701				return (EINVAL); /* XXX panic here? */
702			}
703		}
704
705		/* relate the address to the prefix */
706		if (ia->ia6_ndpr == NULL) {
707			ia->ia6_ndpr = pr;
708			pr->ndpr_refcnt++;
709
710			/*
711			 * If this is the first autoconf address from the
712			 * prefix, create a temporary address as well
713			 * (when required).
714			 */
715			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
716			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
717				int e;
718				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
719					log(LOG_NOTICE, "in6_control: failed "
720					    "to create a temporary address, "
721					    "errno=%d\n", e);
722				}
723			}
724		}
725
726		/*
727		 * this might affect the status of autoconfigured addresses,
728		 * that is, this address might make other addresses detached.
729		 */
730		pfxlist_onlink_check();
731		if (error == 0 && ia)
732			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
733		break;
734	}
735
736	case SIOCDIFADDR_IN6:
737	{
738		struct nd_prefix *pr;
739
740		/*
741		 * If the address being deleted is the only one that owns
742		 * the corresponding prefix, expire the prefix as well.
743		 * XXX: theoretically, we don't have to worry about such
744		 * relationship, since we separate the address management
745		 * and the prefix management.  We do this, however, to provide
746		 * as much backward compatibility as possible in terms of
747		 * the ioctl operation.
748		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
749		 */
750		pr = ia->ia6_ndpr;
751		in6_purgeaddr(&ia->ia_ifa);
752		if (pr && pr->ndpr_refcnt == 0)
753			prelist_remove(pr);
754		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
755		break;
756	}
757
758	default:
759		if (ifp == NULL || ifp->if_ioctl == 0)
760			return (EOPNOTSUPP);
761		return ((*ifp->if_ioctl)(ifp, cmd, data));
762	}
763
764	return (0);
765}
766
767/*
768 * Update parameters of an IPv6 interface address.
769 * If necessary, a new entry is created and linked into address chains.
770 * This function is separated from in6_control().
771 * XXX: should this be performed under splnet()?
772 */
773int
774in6_update_ifa(ifp, ifra, ia, flags)
775	struct ifnet *ifp;
776	struct in6_aliasreq *ifra;
777	struct in6_ifaddr *ia;
778	int flags;
779{
780	int error = 0, hostIsNew = 0, plen = -1;
781	struct in6_ifaddr *oia;
782	struct sockaddr_in6 dst6;
783	struct in6_addrlifetime *lt;
784	struct in6_multi_mship *imm;
785	struct in6_multi *in6m_sol;
786	struct rtentry *rt;
787	int delay;
788
789	/* Validate parameters */
790	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
791		return (EINVAL);
792
793	/*
794	 * The destination address for a p2p link must have a family
795	 * of AF_UNSPEC or AF_INET6.
796	 */
797	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
798	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
799	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
800		return (EAFNOSUPPORT);
801	/*
802	 * validate ifra_prefixmask.  don't check sin6_family, netmask
803	 * does not carry fields other than sin6_len.
804	 */
805	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
806		return (EINVAL);
807	/*
808	 * Because the IPv6 address architecture is classless, we require
809	 * users to specify a (non 0) prefix length (mask) for a new address.
810	 * We also require the prefix (when specified) mask is valid, and thus
811	 * reject a non-consecutive mask.
812	 */
813	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
814		return (EINVAL);
815	if (ifra->ifra_prefixmask.sin6_len != 0) {
816		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
817		    (u_char *)&ifra->ifra_prefixmask +
818		    ifra->ifra_prefixmask.sin6_len);
819		if (plen <= 0)
820			return (EINVAL);
821	} else {
822		/*
823		 * In this case, ia must not be NULL.  We just use its prefix
824		 * length.
825		 */
826		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
827	}
828	/*
829	 * If the destination address on a p2p interface is specified,
830	 * and the address is a scoped one, validate/set the scope
831	 * zone identifier.
832	 */
833	dst6 = ifra->ifra_dstaddr;
834	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
835	    (dst6.sin6_family == AF_INET6)) {
836		struct in6_addr in6_tmp;
837		u_int32_t zoneid;
838
839		in6_tmp = dst6.sin6_addr;
840		if (in6_setscope(&in6_tmp, ifp, &zoneid))
841			return (EINVAL); /* XXX: should be impossible */
842
843		if (dst6.sin6_scope_id != 0) {
844			if (dst6.sin6_scope_id != zoneid)
845				return (EINVAL);
846		} else		/* user omit to specify the ID. */
847			dst6.sin6_scope_id = zoneid;
848
849		/* convert into the internal form */
850		if (sa6_embedscope(&dst6, 0))
851			return (EINVAL); /* XXX: should be impossible */
852	}
853	/*
854	 * The destination address can be specified only for a p2p or a
855	 * loopback interface.  If specified, the corresponding prefix length
856	 * must be 128.
857	 */
858	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
859		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
860			/* XXX: noisy message */
861			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
862			    "be specified for a p2p or a loopback IF only\n"));
863			return (EINVAL);
864		}
865		if (plen != 128) {
866			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
867			    "be 128 when dstaddr is specified\n"));
868			return (EINVAL);
869		}
870	}
871	/* lifetime consistency check */
872	lt = &ifra->ifra_lifetime;
873	if (lt->ia6t_pltime > lt->ia6t_vltime)
874		return (EINVAL);
875	if (lt->ia6t_vltime == 0) {
876		/*
877		 * the following log might be noisy, but this is a typical
878		 * configuration mistake or a tool's bug.
879		 */
880		nd6log((LOG_INFO,
881		    "in6_update_ifa: valid lifetime is 0 for %s\n",
882		    ip6_sprintf(&ifra->ifra_addr.sin6_addr)));
883
884		if (ia == NULL)
885			return (0); /* there's nothing to do */
886	}
887
888	/*
889	 * If this is a new address, allocate a new ifaddr and link it
890	 * into chains.
891	 */
892	if (ia == NULL) {
893		hostIsNew = 1;
894		/*
895		 * When in6_update_ifa() is called in a process of a received
896		 * RA, it is called under an interrupt context.  So, we should
897		 * call malloc with M_NOWAIT.
898		 */
899		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
900		    M_NOWAIT);
901		if (ia == NULL)
902			return (ENOBUFS);
903		bzero((caddr_t)ia, sizeof(*ia));
904		/* Initialize the address and masks, and put time stamp */
905		IFA_LOCK_INIT(&ia->ia_ifa);
906		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
907		ia->ia_addr.sin6_family = AF_INET6;
908		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
909		ia->ia6_createtime = time_second;
910		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
911			/*
912			 * XXX: some functions expect that ifa_dstaddr is not
913			 * NULL for p2p interfaces.
914			 */
915			ia->ia_ifa.ifa_dstaddr =
916			    (struct sockaddr *)&ia->ia_dstaddr;
917		} else {
918			ia->ia_ifa.ifa_dstaddr = NULL;
919		}
920		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
921
922		ia->ia_ifp = ifp;
923		if ((oia = in6_ifaddr) != NULL) {
924			for ( ; oia->ia_next; oia = oia->ia_next)
925				continue;
926			oia->ia_next = ia;
927		} else
928			in6_ifaddr = ia;
929
930		ia->ia_ifa.ifa_refcnt = 1;
931		TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
932	}
933
934	/* update timestamp */
935	ia->ia6_updatetime = time_second;
936
937	/* set prefix mask */
938	if (ifra->ifra_prefixmask.sin6_len) {
939		/*
940		 * We prohibit changing the prefix length of an existing
941		 * address, because
942		 * + such an operation should be rare in IPv6, and
943		 * + the operation would confuse prefix management.
944		 */
945		if (ia->ia_prefixmask.sin6_len &&
946		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
947			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
948			    " existing (%s) address should not be changed\n",
949			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
950			error = EINVAL;
951			goto unlink;
952		}
953		ia->ia_prefixmask = ifra->ifra_prefixmask;
954	}
955
956	/*
957	 * If a new destination address is specified, scrub the old one and
958	 * install the new destination.  Note that the interface must be
959	 * p2p or loopback (see the check above.)
960	 */
961	if (dst6.sin6_family == AF_INET6 &&
962	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
963		int e;
964
965		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
966		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
967			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
968			    "a route to the old destination: %s\n",
969			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
970			/* proceed anyway... */
971		} else
972			ia->ia_flags &= ~IFA_ROUTE;
973		ia->ia_dstaddr = dst6;
974	}
975
976	/*
977	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
978	 * to see if the address is deprecated or invalidated, but initialize
979	 * these members for applications.
980	 */
981	ia->ia6_lifetime = ifra->ifra_lifetime;
982	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
983		ia->ia6_lifetime.ia6t_expire =
984		    time_second + ia->ia6_lifetime.ia6t_vltime;
985	} else
986		ia->ia6_lifetime.ia6t_expire = 0;
987	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
988		ia->ia6_lifetime.ia6t_preferred =
989		    time_second + ia->ia6_lifetime.ia6t_pltime;
990	} else
991		ia->ia6_lifetime.ia6t_preferred = 0;
992
993	/* reset the interface and routing table appropriately. */
994	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
995		goto unlink;
996
997	/*
998	 * configure address flags.
999	 */
1000	ia->ia6_flags = ifra->ifra_flags;
1001	/*
1002	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1003	 * userland, make it deprecated.
1004	 */
1005	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1006		ia->ia6_lifetime.ia6t_pltime = 0;
1007		ia->ia6_lifetime.ia6t_preferred = time_second;
1008	}
1009	/*
1010	 * Make the address tentative before joining multicast addresses,
1011	 * so that corresponding MLD responses would not have a tentative
1012	 * source address.
1013	 */
1014	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1015	if (hostIsNew && in6if_do_dad(ifp))
1016		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1017
1018	/*
1019	 * We are done if we have simply modified an existing address.
1020	 */
1021	if (!hostIsNew)
1022		return (error);
1023
1024	/*
1025	 * Beyond this point, we should call in6_purgeaddr upon an error,
1026	 * not just go to unlink.
1027	 */
1028
1029	/* Join necessary multicast groups */
1030	in6m_sol = NULL;
1031	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1032		struct sockaddr_in6 mltaddr, mltmask;
1033		struct in6_addr llsol;
1034
1035		/* join solicited multicast addr for new host id */
1036		bzero(&llsol, sizeof(struct in6_addr));
1037		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1038		llsol.s6_addr32[1] = 0;
1039		llsol.s6_addr32[2] = htonl(1);
1040		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1041		llsol.s6_addr8[12] = 0xff;
1042		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1043			/* XXX: should not happen */
1044			log(LOG_ERR, "in6_update_ifa: "
1045			    "in6_setscope failed\n");
1046			goto cleanup;
1047		}
1048		delay = 0;
1049		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1050			/*
1051			 * We need a random delay for DAD on the address
1052			 * being configured.  It also means delaying
1053			 * transmission of the corresponding MLD report to
1054			 * avoid report collision.
1055			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1056			 */
1057			delay = arc4random() %
1058			    (MAX_RTR_SOLICITATION_DELAY * hz);
1059		}
1060		imm = in6_joingroup(ifp, &llsol, &error, delay);
1061		if (error != 0) {
1062			nd6log((LOG_WARNING,
1063			    "in6_update_ifa: addmulti failed for "
1064			    "%s on %s (errno=%d)\n",
1065			    ip6_sprintf(&llsol), if_name(ifp),
1066			    error));
1067			in6_purgeaddr((struct ifaddr *)ia);
1068			return (error);
1069		}
1070		in6m_sol = imm->i6mm_maddr;
1071
1072		bzero(&mltmask, sizeof(mltmask));
1073		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1074		mltmask.sin6_family = AF_INET6;
1075		mltmask.sin6_addr = in6mask32;
1076#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1077
1078		/*
1079		 * join link-local all-nodes address
1080		 */
1081		bzero(&mltaddr, sizeof(mltaddr));
1082		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1083		mltaddr.sin6_family = AF_INET6;
1084		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1085		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1086		    0)
1087			goto cleanup; /* XXX: should not fail */
1088
1089		/*
1090		 * XXX: do we really need this automatic routes?
1091		 * We should probably reconsider this stuff.  Most applications
1092		 * actually do not need the routes, since they usually specify
1093		 * the outgoing interface.
1094		 */
1095		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1096		if (rt) {
1097			if (memcmp(&mltaddr.sin6_addr,
1098			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1099			    MLTMASK_LEN)) {
1100				RTFREE_LOCKED(rt);
1101				rt = NULL;
1102			}
1103		}
1104		if (!rt) {
1105			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1106			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1107			    (struct sockaddr *)&ia->ia_addr,
1108			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1109			    (struct rtentry **)0);
1110			if (error)
1111				goto cleanup;
1112		} else
1113			RTFREE_LOCKED(rt);
1114
1115		/*
1116		 * XXX: do we really need this automatic routes?
1117		 * We should probably reconsider this stuff.  Most applications
1118		 * actually do not need the routes, since they usually specify
1119		 * the outgoing interface.
1120		 */
1121		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1122		if (rt) {
1123			/* XXX: only works in !SCOPEDROUTING case. */
1124			if (memcmp(&mltaddr.sin6_addr,
1125			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1126			    MLTMASK_LEN)) {
1127				RTFREE_LOCKED(rt);
1128				rt = NULL;
1129			}
1130		}
1131		if (!rt) {
1132			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1133			    (struct sockaddr *)&ia->ia_addr,
1134			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1135			    (struct rtentry **)0);
1136			if (error)
1137				goto cleanup;
1138		} else {
1139			RTFREE_LOCKED(rt);
1140		}
1141
1142		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1143		if (!imm) {
1144			nd6log((LOG_WARNING,
1145			    "in6_update_ifa: addmulti failed for "
1146			    "%s on %s (errno=%d)\n",
1147			    ip6_sprintf(&mltaddr.sin6_addr),
1148			    if_name(ifp), error));
1149			goto cleanup;
1150		}
1151
1152		/*
1153		 * join node information group address
1154		 */
1155#define hostnamelen	strlen(hostname)
1156		delay = 0;
1157		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1158			/*
1159			 * The spec doesn't say anything about delay for this
1160			 * group, but the same logic should apply.
1161			 */
1162			delay = arc4random() %
1163			    (MAX_RTR_SOLICITATION_DELAY * hz);
1164		}
1165		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr.sin6_addr)
1166		    == 0) {
1167			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1168			    delay); /* XXX jinmei */
1169			if (!imm) {
1170				nd6log((LOG_WARNING, "in6_update_ifa: "
1171				    "addmulti failed for %s on %s "
1172				    "(errno=%d)\n",
1173				    ip6_sprintf(&mltaddr.sin6_addr),
1174				    if_name(ifp), error));
1175				/* XXX not very fatal, go on... */
1176			}
1177		}
1178#undef hostnamelen
1179
1180		/*
1181		 * join interface-local all-nodes address.
1182		 * (ff01::1%ifN, and ff01::%ifN/32)
1183		 */
1184		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1185		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1186		    != 0)
1187			goto cleanup; /* XXX: should not fail */
1188		/* XXX: again, do we really need the route? */
1189		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1190		if (rt) {
1191			if (memcmp(&mltaddr.sin6_addr,
1192			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1193			    MLTMASK_LEN)) {
1194				RTFREE_LOCKED(rt);
1195				rt = NULL;
1196			}
1197		}
1198		if (!rt) {
1199			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1200			    (struct sockaddr *)&ia->ia_addr,
1201			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1202			    (struct rtentry **)0);
1203			if (error)
1204				goto cleanup;
1205		} else
1206			RTFREE_LOCKED(rt);
1207
1208		/* XXX: again, do we really need the route? */
1209		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1210		if (rt) {
1211			if (memcmp(&mltaddr.sin6_addr,
1212			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1213			    MLTMASK_LEN)) {
1214				RTFREE_LOCKED(rt);
1215				rt = NULL;
1216			}
1217		}
1218		if (!rt) {
1219			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1220			    (struct sockaddr *)&ia->ia_addr,
1221			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1222			    (struct rtentry **)0);
1223			if (error)
1224				goto cleanup;
1225		} else {
1226			RTFREE_LOCKED(rt);
1227		}
1228
1229		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1230		if (!imm) {
1231			nd6log((LOG_WARNING, "in6_update_ifa: "
1232			    "addmulti failed for %s on %s "
1233			    "(errno=%d)\n",
1234			    ip6_sprintf(&mltaddr.sin6_addr),
1235			    if_name(ifp), error));
1236			goto cleanup;
1237		}
1238#undef	MLTMASK_LEN
1239	}
1240
1241	/*
1242	 * Perform DAD, if needed.
1243	 * XXX It may be of use, if we can administratively
1244	 * disable DAD.
1245	 */
1246	if (hostIsNew && in6if_do_dad(ifp) &&
1247	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1248	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1249	{
1250		int mindelay, maxdelay;
1251
1252		delay = 0;
1253		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1254			/*
1255			 * We need to impose a delay before sending an NS
1256			 * for DAD.  Check if we also needed a delay for the
1257			 * corresponding MLD message.  If we did, the delay
1258			 * should be larger than the MLD delay (this could be
1259			 * relaxed a bit, but this simple logic is at least
1260			 * safe).
1261			 */
1262			mindelay = 0;
1263			if (in6m_sol != NULL &&
1264			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1265				mindelay = in6m_sol->in6m_timer;
1266			}
1267			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1268			if (maxdelay - mindelay == 0)
1269				delay = 0;
1270			else {
1271				delay =
1272				    (arc4random() % (maxdelay - mindelay)) +
1273				    mindelay;
1274			}
1275		}
1276		nd6_dad_start((struct ifaddr *)ia, delay);
1277	}
1278
1279	return (error);
1280
1281  unlink:
1282	/*
1283	 * XXX: if a change of an existing address failed, keep the entry
1284	 * anyway.
1285	 */
1286	if (hostIsNew)
1287		in6_unlink_ifa(ia, ifp);
1288	return (error);
1289
1290  cleanup:
1291	in6_purgeaddr(&ia->ia_ifa);
1292	return error;
1293}
1294
1295void
1296in6_purgeaddr(ifa)
1297	struct ifaddr *ifa;
1298{
1299	struct ifnet *ifp = ifa->ifa_ifp;
1300	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1301
1302	/* stop DAD processing */
1303	nd6_dad_stop(ifa);
1304
1305	/*
1306	 * delete route to the destination of the address being purged.
1307	 * The interface must be p2p or loopback in this case.
1308	 */
1309	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1310		int e;
1311
1312		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1313		    != 0) {
1314			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1315			    "a route to the p2p destination: %s on %s, "
1316			    "errno=%d\n",
1317			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1318			    e);
1319			/* proceed anyway... */
1320		} else
1321			ia->ia_flags &= ~IFA_ROUTE;
1322	}
1323
1324	/* Remove ownaddr's loopback rtentry, if it exists. */
1325	in6_ifremloop(&(ia->ia_ifa));
1326
1327	if (ifp->if_flags & IFF_MULTICAST) {
1328		/*
1329		 * delete solicited multicast addr for deleting host id
1330		 */
1331		struct in6_multi *in6m;
1332		struct in6_addr llsol;
1333		bzero(&llsol, sizeof(struct in6_addr));
1334		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1335		llsol.s6_addr32[1] = 0;
1336		llsol.s6_addr32[2] = htonl(1);
1337		llsol.s6_addr32[3] =
1338			ia->ia_addr.sin6_addr.s6_addr32[3];
1339		llsol.s6_addr8[12] = 0xff;
1340		(void)in6_setscope(&llsol, ifp, NULL); /* XXX proceed anyway */
1341
1342		IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1343		if (in6m)
1344			in6_delmulti(in6m);
1345	}
1346
1347	in6_unlink_ifa(ia, ifp);
1348}
1349
1350static void
1351in6_unlink_ifa(ia, ifp)
1352	struct in6_ifaddr *ia;
1353	struct ifnet *ifp;
1354{
1355	struct in6_ifaddr *oia;
1356	int	s = splnet();
1357
1358	TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1359
1360	oia = ia;
1361	if (oia == (ia = in6_ifaddr))
1362		in6_ifaddr = ia->ia_next;
1363	else {
1364		while (ia->ia_next && (ia->ia_next != oia))
1365			ia = ia->ia_next;
1366		if (ia->ia_next)
1367			ia->ia_next = oia->ia_next;
1368		else {
1369			/* search failed */
1370			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1371		}
1372	}
1373
1374	/*
1375	 * Release the reference to the base prefix.  There should be a
1376	 * positive reference.
1377	 */
1378	if (oia->ia6_ndpr == NULL) {
1379		nd6log((LOG_NOTICE,
1380		    "in6_unlink_ifa: autoconf'ed address "
1381		    "%p has no prefix\n", oia));
1382	} else {
1383		oia->ia6_ndpr->ndpr_refcnt--;
1384		oia->ia6_ndpr = NULL;
1385	}
1386
1387	/*
1388	 * Also, if the address being removed is autoconf'ed, call
1389	 * pfxlist_onlink_check() since the release might affect the status of
1390	 * other (detached) addresses.
1391	 */
1392	if ((oia->ia6_flags & IN6_IFF_AUTOCONF)) {
1393		pfxlist_onlink_check();
1394	}
1395
1396	/*
1397	 * release another refcnt for the link from in6_ifaddr.
1398	 * Note that we should decrement the refcnt at least once for all *BSD.
1399	 */
1400	IFAFREE(&oia->ia_ifa);
1401
1402	splx(s);
1403}
1404
1405void
1406in6_purgeif(ifp)
1407	struct ifnet *ifp;
1408{
1409	struct ifaddr *ifa, *nifa;
1410
1411	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa) {
1412		nifa = TAILQ_NEXT(ifa, ifa_list);
1413		if (ifa->ifa_addr->sa_family != AF_INET6)
1414			continue;
1415		in6_purgeaddr(ifa);
1416	}
1417
1418	in6_ifdetach(ifp);
1419}
1420
1421/*
1422 * SIOC[GAD]LIFADDR.
1423 *	SIOCGLIFADDR: get first address. (?)
1424 *	SIOCGLIFADDR with IFLR_PREFIX:
1425 *		get first address that matches the specified prefix.
1426 *	SIOCALIFADDR: add the specified address.
1427 *	SIOCALIFADDR with IFLR_PREFIX:
1428 *		add the specified prefix, filling hostid part from
1429 *		the first link-local address.  prefixlen must be <= 64.
1430 *	SIOCDLIFADDR: delete the specified address.
1431 *	SIOCDLIFADDR with IFLR_PREFIX:
1432 *		delete the first address that matches the specified prefix.
1433 * return values:
1434 *	EINVAL on invalid parameters
1435 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1436 *	other values may be returned from in6_ioctl()
1437 *
1438 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1439 * this is to accomodate address naming scheme other than RFC2374,
1440 * in the future.
1441 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1442 * address encoding scheme. (see figure on page 8)
1443 */
1444static int
1445in6_lifaddr_ioctl(so, cmd, data, ifp, td)
1446	struct socket *so;
1447	u_long cmd;
1448	caddr_t	data;
1449	struct ifnet *ifp;
1450	struct thread *td;
1451{
1452	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1453	struct ifaddr *ifa;
1454	struct sockaddr *sa;
1455
1456	/* sanity checks */
1457	if (!data || !ifp) {
1458		panic("invalid argument to in6_lifaddr_ioctl");
1459		/* NOTREACHED */
1460	}
1461
1462	switch (cmd) {
1463	case SIOCGLIFADDR:
1464		/* address must be specified on GET with IFLR_PREFIX */
1465		if ((iflr->flags & IFLR_PREFIX) == 0)
1466			break;
1467		/* FALLTHROUGH */
1468	case SIOCALIFADDR:
1469	case SIOCDLIFADDR:
1470		/* address must be specified on ADD and DELETE */
1471		sa = (struct sockaddr *)&iflr->addr;
1472		if (sa->sa_family != AF_INET6)
1473			return EINVAL;
1474		if (sa->sa_len != sizeof(struct sockaddr_in6))
1475			return EINVAL;
1476		/* XXX need improvement */
1477		sa = (struct sockaddr *)&iflr->dstaddr;
1478		if (sa->sa_family && sa->sa_family != AF_INET6)
1479			return EINVAL;
1480		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1481			return EINVAL;
1482		break;
1483	default: /* shouldn't happen */
1484#if 0
1485		panic("invalid cmd to in6_lifaddr_ioctl");
1486		/* NOTREACHED */
1487#else
1488		return EOPNOTSUPP;
1489#endif
1490	}
1491	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1492		return EINVAL;
1493
1494	switch (cmd) {
1495	case SIOCALIFADDR:
1496	    {
1497		struct in6_aliasreq ifra;
1498		struct in6_addr *hostid = NULL;
1499		int prefixlen;
1500
1501		if ((iflr->flags & IFLR_PREFIX) != 0) {
1502			struct sockaddr_in6 *sin6;
1503
1504			/*
1505			 * hostid is to fill in the hostid part of the
1506			 * address.  hostid points to the first link-local
1507			 * address attached to the interface.
1508			 */
1509			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1510			if (!ifa)
1511				return EADDRNOTAVAIL;
1512			hostid = IFA_IN6(ifa);
1513
1514		 	/* prefixlen must be <= 64. */
1515			if (64 < iflr->prefixlen)
1516				return EINVAL;
1517			prefixlen = iflr->prefixlen;
1518
1519			/* hostid part must be zero. */
1520			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1521			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1522			    sin6->sin6_addr.s6_addr32[3] != 0) {
1523				return EINVAL;
1524			}
1525		} else
1526			prefixlen = iflr->prefixlen;
1527
1528		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1529		bzero(&ifra, sizeof(ifra));
1530		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1531
1532		bcopy(&iflr->addr, &ifra.ifra_addr,
1533		    ((struct sockaddr *)&iflr->addr)->sa_len);
1534		if (hostid) {
1535			/* fill in hostid part */
1536			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1537			    hostid->s6_addr32[2];
1538			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1539			    hostid->s6_addr32[3];
1540		}
1541
1542		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1543			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1544			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1545			if (hostid) {
1546				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1547				    hostid->s6_addr32[2];
1548				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1549				    hostid->s6_addr32[3];
1550			}
1551		}
1552
1553		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1554		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1555
1556		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1557		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1558	    }
1559	case SIOCGLIFADDR:
1560	case SIOCDLIFADDR:
1561	    {
1562		struct in6_ifaddr *ia;
1563		struct in6_addr mask, candidate, match;
1564		struct sockaddr_in6 *sin6;
1565		int cmp;
1566
1567		bzero(&mask, sizeof(mask));
1568		if (iflr->flags & IFLR_PREFIX) {
1569			/* lookup a prefix rather than address. */
1570			in6_prefixlen2mask(&mask, iflr->prefixlen);
1571
1572			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1573			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1574			match.s6_addr32[0] &= mask.s6_addr32[0];
1575			match.s6_addr32[1] &= mask.s6_addr32[1];
1576			match.s6_addr32[2] &= mask.s6_addr32[2];
1577			match.s6_addr32[3] &= mask.s6_addr32[3];
1578
1579			/* if you set extra bits, that's wrong */
1580			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1581				return EINVAL;
1582
1583			cmp = 1;
1584		} else {
1585			if (cmd == SIOCGLIFADDR) {
1586				/* on getting an address, take the 1st match */
1587				cmp = 0;	/* XXX */
1588			} else {
1589				/* on deleting an address, do exact match */
1590				in6_prefixlen2mask(&mask, 128);
1591				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1592				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1593
1594				cmp = 1;
1595			}
1596		}
1597
1598		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1599			if (ifa->ifa_addr->sa_family != AF_INET6)
1600				continue;
1601			if (!cmp)
1602				break;
1603
1604			/*
1605			 * XXX: this is adhoc, but is necessary to allow
1606			 * a user to specify fe80::/64 (not /10) for a
1607			 * link-local address.
1608			 */
1609			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1610			in6_clearscope(&candidate);
1611			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1612			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1613			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1614			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1615			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1616				break;
1617		}
1618		if (!ifa)
1619			return EADDRNOTAVAIL;
1620		ia = ifa2ia6(ifa);
1621
1622		if (cmd == SIOCGLIFADDR) {
1623			int error;
1624
1625			/* fill in the if_laddrreq structure */
1626			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1627			error = sa6_recoverscope(
1628			    (struct sockaddr_in6 *)&iflr->addr);
1629			if (error != 0)
1630				return (error);
1631
1632			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1633				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1634				    ia->ia_dstaddr.sin6_len);
1635				error = sa6_recoverscope(
1636				    (struct sockaddr_in6 *)&iflr->dstaddr);
1637				if (error != 0)
1638					return (error);
1639			} else
1640				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1641
1642			iflr->prefixlen =
1643			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1644
1645			iflr->flags = ia->ia6_flags;	/* XXX */
1646
1647			return 0;
1648		} else {
1649			struct in6_aliasreq ifra;
1650
1651			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1652			bzero(&ifra, sizeof(ifra));
1653			bcopy(iflr->iflr_name, ifra.ifra_name,
1654			    sizeof(ifra.ifra_name));
1655
1656			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1657			    ia->ia_addr.sin6_len);
1658			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1659				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1660				    ia->ia_dstaddr.sin6_len);
1661			} else {
1662				bzero(&ifra.ifra_dstaddr,
1663				    sizeof(ifra.ifra_dstaddr));
1664			}
1665			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1666			    ia->ia_prefixmask.sin6_len);
1667
1668			ifra.ifra_flags = ia->ia6_flags;
1669			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1670			    ifp, td);
1671		}
1672	    }
1673	}
1674
1675	return EOPNOTSUPP;	/* just for safety */
1676}
1677
1678/*
1679 * Initialize an interface's intetnet6 address
1680 * and routing table entry.
1681 */
1682static int
1683in6_ifinit(ifp, ia, sin6, newhost)
1684	struct ifnet *ifp;
1685	struct in6_ifaddr *ia;
1686	struct sockaddr_in6 *sin6;
1687	int newhost;
1688{
1689	int	error = 0, plen, ifacount = 0;
1690	int	s = splimp();
1691	struct ifaddr *ifa;
1692
1693	/*
1694	 * Give the interface a chance to initialize
1695	 * if this is its first address,
1696	 * and to validate the address if necessary.
1697	 */
1698	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1699		if (ifa->ifa_addr == NULL)
1700			continue;	/* just for safety */
1701		if (ifa->ifa_addr->sa_family != AF_INET6)
1702			continue;
1703		ifacount++;
1704	}
1705
1706	ia->ia_addr = *sin6;
1707
1708	if (ifacount <= 1 && ifp->if_ioctl) {
1709		IFF_LOCKGIANT(ifp);
1710		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1711		IFF_UNLOCKGIANT(ifp);
1712		if (error) {
1713			splx(s);
1714			return (error);
1715		}
1716	}
1717	splx(s);
1718
1719	ia->ia_ifa.ifa_metric = ifp->if_metric;
1720
1721	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1722
1723	if (newhost) {
1724		/*
1725		 * set the rtrequest function to create llinfo.  It also
1726		 * adjust outgoing interface of the route for the local
1727		 * address when called via in6_ifaddloop() below.
1728		 */
1729		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1730	}
1731
1732	/*
1733	 * Special case:
1734	 * If a new destination address is specified for a point-to-point
1735	 * interface, install a route to the destination as an interface
1736	 * direct route.  In addition, if the link is expected to have neighbor
1737	 * cache entries, specify RTF_LLINFO so that a cache entry for the
1738	 * destination address will be created.
1739	 * created
1740	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1741	 * interface that share the same destination.
1742	 */
1743	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1744	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1745	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1746		int rtflags = RTF_UP | RTF_HOST;
1747		struct rtentry *rt = NULL, **rtp = NULL;
1748
1749		if (nd6_need_cache(ifp) != 0) {
1750			rtflags |= RTF_LLINFO;
1751			rtp = &rt;
1752		}
1753
1754		error = rtrequest(RTM_ADD, (struct sockaddr *)&ia->ia_dstaddr,
1755		    (struct sockaddr *)&ia->ia_addr,
1756		    (struct sockaddr *)&ia->ia_prefixmask,
1757		    ia->ia_flags | rtflags, rtp);
1758		if (error != 0)
1759			return (error);
1760		if (rt != NULL) {
1761			struct llinfo_nd6 *ln;
1762
1763			RT_LOCK(rt);
1764			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1765			if (ln != NULL) {
1766				/*
1767				 * Set the state to STALE because we don't
1768				 * have to perform address resolution on this
1769				 * link.
1770				 */
1771				ln->ln_state = ND6_LLINFO_STALE;
1772			}
1773			RT_REMREF(rt);
1774			RT_UNLOCK(rt);
1775		}
1776		ia->ia_flags |= IFA_ROUTE;
1777	}
1778	if (plen < 128) {
1779		/*
1780		 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1781		 */
1782		ia->ia_ifa.ifa_flags |= RTF_CLONING;
1783	}
1784
1785	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1786	if (newhost)
1787		in6_ifaddloop(&(ia->ia_ifa));
1788
1789	return (error);
1790}
1791
1792struct in6_multi_mship *
1793in6_joingroup(ifp, addr, errorp, delay)
1794	struct ifnet *ifp;
1795	struct in6_addr *addr;
1796	int *errorp;
1797	int delay;
1798{
1799	struct in6_multi_mship *imm;
1800
1801	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
1802	if (!imm) {
1803		*errorp = ENOBUFS;
1804		return NULL;
1805	}
1806	imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp, delay);
1807	if (!imm->i6mm_maddr) {
1808		/* *errorp is alrady set */
1809		free(imm, M_IP6MADDR);
1810		return NULL;
1811	}
1812	return imm;
1813}
1814
1815int
1816in6_leavegroup(imm)
1817	struct in6_multi_mship *imm;
1818{
1819
1820	if (imm->i6mm_maddr)
1821		in6_delmulti(imm->i6mm_maddr);
1822	free(imm,  M_IP6MADDR);
1823	return 0;
1824}
1825
1826/*
1827 * Find an IPv6 interface link-local address specific to an interface.
1828 */
1829struct in6_ifaddr *
1830in6ifa_ifpforlinklocal(ifp, ignoreflags)
1831	struct ifnet *ifp;
1832	int ignoreflags;
1833{
1834	struct ifaddr *ifa;
1835
1836	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1837		if (ifa->ifa_addr == NULL)
1838			continue;	/* just for safety */
1839		if (ifa->ifa_addr->sa_family != AF_INET6)
1840			continue;
1841		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1842			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1843			     ignoreflags) != 0)
1844				continue;
1845			break;
1846		}
1847	}
1848
1849	return ((struct in6_ifaddr *)ifa);
1850}
1851
1852
1853/*
1854 * find the internet address corresponding to a given interface and address.
1855 */
1856struct in6_ifaddr *
1857in6ifa_ifpwithaddr(ifp, addr)
1858	struct ifnet *ifp;
1859	struct in6_addr *addr;
1860{
1861	struct ifaddr *ifa;
1862
1863	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1864		if (ifa->ifa_addr == NULL)
1865			continue;	/* just for safety */
1866		if (ifa->ifa_addr->sa_family != AF_INET6)
1867			continue;
1868		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1869			break;
1870	}
1871
1872	return ((struct in6_ifaddr *)ifa);
1873}
1874
1875/*
1876 * Convert IP6 address to printable (loggable) representation.
1877 */
1878static char digits[] = "0123456789abcdef";
1879static int ip6round = 0;
1880char *
1881ip6_sprintf(addr)
1882	const struct in6_addr *addr;
1883{
1884	static char ip6buf[8][48];
1885	int i;
1886	char *cp;
1887	const u_int16_t *a = (const u_int16_t *)addr;
1888	const u_int8_t *d;
1889	int dcolon = 0;
1890
1891	ip6round = (ip6round + 1) & 7;
1892	cp = ip6buf[ip6round];
1893
1894	for (i = 0; i < 8; i++) {
1895		if (dcolon == 1) {
1896			if (*a == 0) {
1897				if (i == 7)
1898					*cp++ = ':';
1899				a++;
1900				continue;
1901			} else
1902				dcolon = 2;
1903		}
1904		if (*a == 0) {
1905			if (dcolon == 0 && *(a + 1) == 0) {
1906				if (i == 0)
1907					*cp++ = ':';
1908				*cp++ = ':';
1909				dcolon = 1;
1910			} else {
1911				*cp++ = '0';
1912				*cp++ = ':';
1913			}
1914			a++;
1915			continue;
1916		}
1917		d = (const u_char *)a;
1918		*cp++ = digits[*d >> 4];
1919		*cp++ = digits[*d++ & 0xf];
1920		*cp++ = digits[*d >> 4];
1921		*cp++ = digits[*d & 0xf];
1922		*cp++ = ':';
1923		a++;
1924	}
1925	*--cp = 0;
1926	return (ip6buf[ip6round]);
1927}
1928
1929int
1930in6_localaddr(in6)
1931	struct in6_addr *in6;
1932{
1933	struct in6_ifaddr *ia;
1934
1935	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1936		return 1;
1937
1938	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1939		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1940		    &ia->ia_prefixmask.sin6_addr)) {
1941			return 1;
1942		}
1943	}
1944
1945	return (0);
1946}
1947
1948int
1949in6_is_addr_deprecated(sa6)
1950	struct sockaddr_in6 *sa6;
1951{
1952	struct in6_ifaddr *ia;
1953
1954	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1955		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1956				       &sa6->sin6_addr) &&
1957		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1958			return (1); /* true */
1959
1960		/* XXX: do we still have to go thru the rest of the list? */
1961	}
1962
1963	return (0);		/* false */
1964}
1965
1966/*
1967 * return length of part which dst and src are equal
1968 * hard coding...
1969 */
1970int
1971in6_matchlen(src, dst)
1972struct in6_addr *src, *dst;
1973{
1974	int match = 0;
1975	u_char *s = (u_char *)src, *d = (u_char *)dst;
1976	u_char *lim = s + 16, r;
1977
1978	while (s < lim)
1979		if ((r = (*d++ ^ *s++)) != 0) {
1980			while (r < 128) {
1981				match++;
1982				r <<= 1;
1983			}
1984			break;
1985		} else
1986			match += 8;
1987	return match;
1988}
1989
1990/* XXX: to be scope conscious */
1991int
1992in6_are_prefix_equal(p1, p2, len)
1993	struct in6_addr *p1, *p2;
1994	int len;
1995{
1996	int bytelen, bitlen;
1997
1998	/* sanity check */
1999	if (0 > len || len > 128) {
2000		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2001		    len);
2002		return (0);
2003	}
2004
2005	bytelen = len / 8;
2006	bitlen = len % 8;
2007
2008	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2009		return (0);
2010	if (bitlen != 0 &&
2011	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2012	    p2->s6_addr[bytelen] >> (8 - bitlen))
2013		return (0);
2014
2015	return (1);
2016}
2017
2018void
2019in6_prefixlen2mask(maskp, len)
2020	struct in6_addr *maskp;
2021	int len;
2022{
2023	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2024	int bytelen, bitlen, i;
2025
2026	/* sanity check */
2027	if (0 > len || len > 128) {
2028		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2029		    len);
2030		return;
2031	}
2032
2033	bzero(maskp, sizeof(*maskp));
2034	bytelen = len / 8;
2035	bitlen = len % 8;
2036	for (i = 0; i < bytelen; i++)
2037		maskp->s6_addr[i] = 0xff;
2038	if (bitlen)
2039		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2040}
2041
2042/*
2043 * return the best address out of the same scope. if no address was
2044 * found, return the first valid address from designated IF.
2045 */
2046struct in6_ifaddr *
2047in6_ifawithifp(ifp, dst)
2048	struct ifnet *ifp;
2049	struct in6_addr *dst;
2050{
2051	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2052	struct ifaddr *ifa;
2053	struct in6_ifaddr *besta = 0;
2054	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2055
2056	dep[0] = dep[1] = NULL;
2057
2058	/*
2059	 * We first look for addresses in the same scope.
2060	 * If there is one, return it.
2061	 * If two or more, return one which matches the dst longest.
2062	 * If none, return one of global addresses assigned other ifs.
2063	 */
2064	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2065		if (ifa->ifa_addr->sa_family != AF_INET6)
2066			continue;
2067		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2068			continue; /* XXX: is there any case to allow anycast? */
2069		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2070			continue; /* don't use this interface */
2071		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2072			continue;
2073		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2074			if (ip6_use_deprecated)
2075				dep[0] = (struct in6_ifaddr *)ifa;
2076			continue;
2077		}
2078
2079		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2080			/*
2081			 * call in6_matchlen() as few as possible
2082			 */
2083			if (besta) {
2084				if (blen == -1)
2085					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2086				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2087				if (tlen > blen) {
2088					blen = tlen;
2089					besta = (struct in6_ifaddr *)ifa;
2090				}
2091			} else
2092				besta = (struct in6_ifaddr *)ifa;
2093		}
2094	}
2095	if (besta)
2096		return (besta);
2097
2098	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2099		if (ifa->ifa_addr->sa_family != AF_INET6)
2100			continue;
2101		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2102			continue; /* XXX: is there any case to allow anycast? */
2103		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2104			continue; /* don't use this interface */
2105		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2106			continue;
2107		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2108			if (ip6_use_deprecated)
2109				dep[1] = (struct in6_ifaddr *)ifa;
2110			continue;
2111		}
2112
2113		return (struct in6_ifaddr *)ifa;
2114	}
2115
2116	/* use the last-resort values, that are, deprecated addresses */
2117	if (dep[0])
2118		return dep[0];
2119	if (dep[1])
2120		return dep[1];
2121
2122	return NULL;
2123}
2124
2125/*
2126 * perform DAD when interface becomes IFF_UP.
2127 */
2128void
2129in6_if_up(ifp)
2130	struct ifnet *ifp;
2131{
2132	struct ifaddr *ifa;
2133	struct in6_ifaddr *ia;
2134
2135	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2136		if (ifa->ifa_addr->sa_family != AF_INET6)
2137			continue;
2138		ia = (struct in6_ifaddr *)ifa;
2139		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2140			/*
2141			 * The TENTATIVE flag was likely set by hand
2142			 * beforehand, implicitly indicating the need for DAD.
2143			 * We may be able to skip the random delay in this
2144			 * case, but we impose delays just in case.
2145			 */
2146			nd6_dad_start(ifa,
2147			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2148		}
2149	}
2150
2151	/*
2152	 * special cases, like 6to4, are handled in in6_ifattach
2153	 */
2154	in6_ifattach(ifp, NULL);
2155}
2156
2157int
2158in6if_do_dad(ifp)
2159	struct ifnet *ifp;
2160{
2161	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2162		return (0);
2163
2164	switch (ifp->if_type) {
2165#ifdef IFT_DUMMY
2166	case IFT_DUMMY:
2167#endif
2168	case IFT_FAITH:
2169		/*
2170		 * These interfaces do not have the IFF_LOOPBACK flag,
2171		 * but loop packets back.  We do not have to do DAD on such
2172		 * interfaces.  We should even omit it, because loop-backed
2173		 * NS would confuse the DAD procedure.
2174		 */
2175		return (0);
2176	default:
2177		/*
2178		 * Our DAD routine requires the interface up and running.
2179		 * However, some interfaces can be up before the RUNNING
2180		 * status.  Additionaly, users may try to assign addresses
2181		 * before the interface becomes up (or running).
2182		 * We simply skip DAD in such a case as a work around.
2183		 * XXX: we should rather mark "tentative" on such addresses,
2184		 * and do DAD after the interface becomes ready.
2185		 */
2186		if (!((ifp->if_flags & IFF_UP) &&
2187		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2188			return (0);
2189
2190		return (1);
2191	}
2192}
2193
2194/*
2195 * Calculate max IPv6 MTU through all the interfaces and store it
2196 * to in6_maxmtu.
2197 */
2198void
2199in6_setmaxmtu()
2200{
2201	unsigned long maxmtu = 0;
2202	struct ifnet *ifp;
2203
2204	IFNET_RLOCK();
2205	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
2206		/* this function can be called during ifnet initialization */
2207		if (!ifp->if_afdata[AF_INET6])
2208			continue;
2209		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2210		    IN6_LINKMTU(ifp) > maxmtu)
2211			maxmtu = IN6_LINKMTU(ifp);
2212	}
2213	IFNET_RUNLOCK();
2214	if (maxmtu)	     /* update only when maxmtu is positive */
2215		in6_maxmtu = maxmtu;
2216}
2217
2218/*
2219 * Provide the length of interface identifiers to be used for the link attached
2220 * to the given interface.  The length should be defined in "IPv6 over
2221 * xxx-link" document.  Note that address architecture might also define
2222 * the length for a particular set of address prefixes, regardless of the
2223 * link type.  As clarified in rfc2462bis, those two definitions should be
2224 * consistent, and those really are as of August 2004.
2225 */
2226int
2227in6_if2idlen(ifp)
2228	struct ifnet *ifp;
2229{
2230	switch (ifp->if_type) {
2231	case IFT_ETHER:		/* RFC2464 */
2232#ifdef IFT_PROPVIRTUAL
2233	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2234#endif
2235#ifdef IFT_L2VLAN
2236	case IFT_L2VLAN:	/* ditto */
2237#endif
2238#ifdef IFT_IEEE80211
2239	case IFT_IEEE80211:	/* ditto */
2240#endif
2241#ifdef IFT_MIP
2242	case IFT_MIP:	/* ditto */
2243#endif
2244		return (64);
2245	case IFT_FDDI:		/* RFC2467 */
2246		return (64);
2247	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2248		return (64);
2249	case IFT_PPP:		/* RFC2472 */
2250		return (64);
2251	case IFT_ARCNET:	/* RFC2497 */
2252		return (64);
2253	case IFT_FRELAY:	/* RFC2590 */
2254		return (64);
2255	case IFT_IEEE1394:	/* RFC3146 */
2256		return (64);
2257	case IFT_GIF:
2258		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2259	case IFT_LOOP:
2260		return (64);	/* XXX: is this really correct? */
2261	default:
2262		/*
2263		 * Unknown link type:
2264		 * It might be controversial to use the today's common constant
2265		 * of 64 for these cases unconditionally.  For full compliance,
2266		 * we should return an error in this case.  On the other hand,
2267		 * if we simply miss the standard for the link type or a new
2268		 * standard is defined for a new link type, the IFID length
2269		 * is very likely to be the common constant.  As a compromise,
2270		 * we always use the constant, but make an explicit notice
2271		 * indicating the "unknown" case.
2272		 */
2273		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2274		return (64);
2275	}
2276}
2277
2278void *
2279in6_domifattach(ifp)
2280	struct ifnet *ifp;
2281{
2282	struct in6_ifextra *ext;
2283
2284	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2285	bzero(ext, sizeof(*ext));
2286
2287	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2288	    M_IFADDR, M_WAITOK);
2289	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2290
2291	ext->icmp6_ifstat =
2292	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2293	    M_IFADDR, M_WAITOK);
2294	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2295
2296	ext->nd_ifinfo = nd6_ifattach(ifp);
2297	ext->scope6_id = scope6_ifattach(ifp);
2298	return ext;
2299}
2300
2301void
2302in6_domifdetach(ifp, aux)
2303	struct ifnet *ifp;
2304	void *aux;
2305{
2306	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2307
2308	scope6_ifdetach(ext->scope6_id);
2309	nd6_ifdetach(ext->nd_ifinfo);
2310	free(ext->in6_ifstat, M_IFADDR);
2311	free(ext->icmp6_ifstat, M_IFADDR);
2312	free(ext, M_IFADDR);
2313}
2314
2315/*
2316 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2317 * v4 mapped addr or v4 compat addr
2318 */
2319void
2320in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2321{
2322	bzero(sin, sizeof(*sin));
2323	sin->sin_len = sizeof(struct sockaddr_in);
2324	sin->sin_family = AF_INET;
2325	sin->sin_port = sin6->sin6_port;
2326	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2327}
2328
2329/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2330void
2331in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2332{
2333	bzero(sin6, sizeof(*sin6));
2334	sin6->sin6_len = sizeof(struct sockaddr_in6);
2335	sin6->sin6_family = AF_INET6;
2336	sin6->sin6_port = sin->sin_port;
2337	sin6->sin6_addr.s6_addr32[0] = 0;
2338	sin6->sin6_addr.s6_addr32[1] = 0;
2339	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2340	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2341}
2342
2343/* Convert sockaddr_in6 into sockaddr_in. */
2344void
2345in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2346{
2347	struct sockaddr_in *sin_p;
2348	struct sockaddr_in6 sin6;
2349
2350	/*
2351	 * Save original sockaddr_in6 addr and convert it
2352	 * to sockaddr_in.
2353	 */
2354	sin6 = *(struct sockaddr_in6 *)nam;
2355	sin_p = (struct sockaddr_in *)nam;
2356	in6_sin6_2_sin(sin_p, &sin6);
2357}
2358
2359/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2360void
2361in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2362{
2363	struct sockaddr_in *sin_p;
2364	struct sockaddr_in6 *sin6_p;
2365
2366	MALLOC(sin6_p, struct sockaddr_in6 *, sizeof *sin6_p, M_SONAME,
2367	       M_WAITOK);
2368	sin_p = (struct sockaddr_in *)*nam;
2369	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2370	FREE(*nam, M_SONAME);
2371	*nam = (struct sockaddr *)sin6_p;
2372}
2373