in6.c revision 238476
1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $
30 */
31
32/*-
33 * Copyright (c) 1982, 1986, 1991, 1993
34 *	The Regents of the University of California.  All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 *    notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 *    notice, this list of conditions and the following disclaimer in the
43 *    documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 *    may be used to endorse or promote products derived from this software
46 *    without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 *	@(#)in.c	8.2 (Berkeley) 11/15/93
61 */
62
63#include <sys/cdefs.h>
64__FBSDID("$FreeBSD: stable/9/sys/netinet6/in6.c 238476 2012-07-15 11:13:09Z bz $");
65
66#include "opt_compat.h"
67#include "opt_inet.h"
68#include "opt_inet6.h"
69
70#include <sys/param.h>
71#include <sys/errno.h>
72#include <sys/jail.h>
73#include <sys/malloc.h>
74#include <sys/socket.h>
75#include <sys/socketvar.h>
76#include <sys/sockio.h>
77#include <sys/systm.h>
78#include <sys/priv.h>
79#include <sys/proc.h>
80#include <sys/time.h>
81#include <sys/kernel.h>
82#include <sys/syslog.h>
83
84#include <net/if.h>
85#include <net/if_var.h>
86#include <net/if_types.h>
87#include <net/route.h>
88#include <net/if_dl.h>
89#include <net/vnet.h>
90
91#include <netinet/in.h>
92#include <netinet/in_var.h>
93#include <net/if_llatbl.h>
94#include <netinet/if_ether.h>
95#include <netinet/in_systm.h>
96#include <netinet/ip.h>
97#include <netinet/in_pcb.h>
98
99#include <netinet/ip6.h>
100#include <netinet6/ip6_var.h>
101#include <netinet6/nd6.h>
102#include <netinet6/mld6_var.h>
103#include <netinet6/ip6_mroute.h>
104#include <netinet6/in6_ifattach.h>
105#include <netinet6/scope6_var.h>
106#include <netinet6/in6_pcb.h>
107
108/*
109 * Definitions of some costant IP6 addresses.
110 */
111const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
112const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
113const struct in6_addr in6addr_nodelocal_allnodes =
114	IN6ADDR_NODELOCAL_ALLNODES_INIT;
115const struct in6_addr in6addr_linklocal_allnodes =
116	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
117const struct in6_addr in6addr_linklocal_allrouters =
118	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
119const struct in6_addr in6addr_linklocal_allv2routers =
120	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
121
122const struct in6_addr in6mask0 = IN6MASK0;
123const struct in6_addr in6mask32 = IN6MASK32;
124const struct in6_addr in6mask64 = IN6MASK64;
125const struct in6_addr in6mask96 = IN6MASK96;
126const struct in6_addr in6mask128 = IN6MASK128;
127
128const struct sockaddr_in6 sa6_any =
129	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
130
131static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
132	struct ifnet *, struct thread *));
133static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
134	struct sockaddr_in6 *, int));
135static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
136
137int	(*faithprefix_p)(struct in6_addr *);
138
139#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
140#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
141
142void
143in6_ifaddloop(struct ifaddr *ifa)
144{
145	struct sockaddr_dl gateway;
146	struct sockaddr_in6 mask, addr;
147	struct rtentry rt;
148	struct in6_ifaddr *ia;
149	struct ifnet *ifp;
150	struct llentry *ln;
151
152	ia = ifa2ia6(ifa);
153	ifp = ifa->ifa_ifp;
154	IF_AFDATA_LOCK(ifp);
155	ifa->ifa_rtrequest = nd6_rtrequest;
156	ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
157	    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
158	IF_AFDATA_UNLOCK(ifp);
159	if (ln != NULL) {
160		ln->la_expire = 0;  /* for IPv6 this means permanent */
161		ln->ln_state = ND6_LLINFO_REACHABLE;
162		/*
163		 * initialize for rtmsg generation
164		 */
165		bzero(&gateway, sizeof(gateway));
166		gateway.sdl_len = sizeof(gateway);
167		gateway.sdl_family = AF_LINK;
168		gateway.sdl_nlen = 0;
169		gateway.sdl_alen = 6;
170		memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
171		    sizeof(ln->ll_addr));
172		LLE_WUNLOCK(ln);
173	}
174
175	bzero(&rt, sizeof(rt));
176	rt.rt_gateway = (struct sockaddr *)&gateway;
177	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
178	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
179	rt_mask(&rt) = (struct sockaddr *)&mask;
180	rt_key(&rt) = (struct sockaddr *)&addr;
181	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
182	/* Announce arrival of local address to all FIBs. */
183	rt_newaddrmsg(RTM_ADD, ifa, 0, &rt);
184}
185
186void
187in6_ifremloop(struct ifaddr *ifa)
188{
189	struct sockaddr_dl gateway;
190	struct sockaddr_in6 mask, addr;
191	struct rtentry rt0;
192	struct in6_ifaddr *ia;
193	struct ifnet *ifp;
194
195	ia = ifa2ia6(ifa);
196	ifp = ifa->ifa_ifp;
197	IF_AFDATA_LOCK(ifp);
198	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
199	    (struct sockaddr *)&ia->ia_addr);
200	IF_AFDATA_UNLOCK(ifp);
201
202	/*
203	 * initialize for rtmsg generation
204	 */
205	bzero(&gateway, sizeof(gateway));
206	gateway.sdl_len = sizeof(gateway);
207	gateway.sdl_family = AF_LINK;
208	gateway.sdl_nlen = 0;
209	gateway.sdl_alen = ifp->if_addrlen;
210	bzero(&rt0, sizeof(rt0));
211	rt0.rt_gateway = (struct sockaddr *)&gateway;
212	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
213	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
214	rt_mask(&rt0) = (struct sockaddr *)&mask;
215	rt_key(&rt0) = (struct sockaddr *)&addr;
216	rt0.rt_flags = RTF_HOST | RTF_STATIC;
217	/* Announce removal of local address to all FIBs. */
218	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
219}
220
221int
222in6_mask2len(struct in6_addr *mask, u_char *lim0)
223{
224	int x = 0, y;
225	u_char *lim = lim0, *p;
226
227	/* ignore the scope_id part */
228	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
229		lim = (u_char *)mask + sizeof(*mask);
230	for (p = (u_char *)mask; p < lim; x++, p++) {
231		if (*p != 0xff)
232			break;
233	}
234	y = 0;
235	if (p < lim) {
236		for (y = 0; y < 8; y++) {
237			if ((*p & (0x80 >> y)) == 0)
238				break;
239		}
240	}
241
242	/*
243	 * when the limit pointer is given, do a stricter check on the
244	 * remaining bits.
245	 */
246	if (p < lim) {
247		if (y != 0 && (*p & (0x00ff >> y)) != 0)
248			return (-1);
249		for (p = p + 1; p < lim; p++)
250			if (*p != 0)
251				return (-1);
252	}
253
254	return x * 8 + y;
255}
256
257#ifdef COMPAT_FREEBSD32
258struct in6_ndifreq32 {
259        char ifname[IFNAMSIZ];
260        uint32_t ifindex;
261};
262#define	SIOCGDEFIFACE32_IN6     _IOWR('i', 86, struct in6_ndifreq32)
263#endif
264
265int
266in6_control(struct socket *so, u_long cmd, caddr_t data,
267    struct ifnet *ifp, struct thread *td)
268{
269	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
270	struct	in6_ifaddr *ia = NULL;
271	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
272	struct sockaddr_in6 *sa6;
273	int error;
274
275	switch (cmd) {
276	case SIOCGETSGCNT_IN6:
277	case SIOCGETMIFCNT_IN6:
278		/*
279		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
280		 * We cannot see how that would be needed, so do not adjust the
281		 * KPI blindly; more likely should clean up the IPv4 variant.
282		 */
283		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
284	}
285
286	switch(cmd) {
287	case SIOCAADDRCTL_POLICY:
288	case SIOCDADDRCTL_POLICY:
289		if (td != NULL) {
290			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
291			if (error)
292				return (error);
293		}
294		return (in6_src_ioctl(cmd, data));
295	}
296
297	if (ifp == NULL)
298		return (EOPNOTSUPP);
299
300	switch (cmd) {
301	case SIOCSNDFLUSH_IN6:
302	case SIOCSPFXFLUSH_IN6:
303	case SIOCSRTRFLUSH_IN6:
304	case SIOCSDEFIFACE_IN6:
305	case SIOCSIFINFO_FLAGS:
306	case SIOCSIFINFO_IN6:
307		if (td != NULL) {
308			error = priv_check(td, PRIV_NETINET_ND6);
309			if (error)
310				return (error);
311		}
312		/* FALLTHROUGH */
313	case OSIOCGIFINFO_IN6:
314	case SIOCGIFINFO_IN6:
315	case SIOCGDRLST_IN6:
316	case SIOCGPRLST_IN6:
317	case SIOCGNBRINFO_IN6:
318	case SIOCGDEFIFACE_IN6:
319		return (nd6_ioctl(cmd, data, ifp));
320
321#ifdef COMPAT_FREEBSD32
322	case SIOCGDEFIFACE32_IN6:
323		{
324			struct in6_ndifreq ndif;
325			struct in6_ndifreq32 *ndif32;
326
327			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
328			    ifp);
329			if (error)
330				return (error);
331			ndif32 = (struct in6_ndifreq32 *)data;
332			ndif32->ifindex = ndif.ifindex;
333			return (0);
334		}
335#endif
336	}
337
338	switch (cmd) {
339	case SIOCSIFPREFIX_IN6:
340	case SIOCDIFPREFIX_IN6:
341	case SIOCAIFPREFIX_IN6:
342	case SIOCCIFPREFIX_IN6:
343	case SIOCSGIFPREFIX_IN6:
344	case SIOCGIFPREFIX_IN6:
345		log(LOG_NOTICE,
346		    "prefix ioctls are now invalidated. "
347		    "please use ifconfig.\n");
348		return (EOPNOTSUPP);
349	}
350
351	switch (cmd) {
352	case SIOCSSCOPE6:
353		if (td != NULL) {
354			error = priv_check(td, PRIV_NETINET_SCOPE6);
355			if (error)
356				return (error);
357		}
358		return (scope6_set(ifp,
359		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
360	case SIOCGSCOPE6:
361		return (scope6_get(ifp,
362		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
363	case SIOCGSCOPE6DEF:
364		return (scope6_get_default((struct scope6_id *)
365		    ifr->ifr_ifru.ifru_scope_id));
366	}
367
368	switch (cmd) {
369	case SIOCALIFADDR:
370		if (td != NULL) {
371			error = priv_check(td, PRIV_NET_ADDIFADDR);
372			if (error)
373				return (error);
374		}
375		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
376
377	case SIOCDLIFADDR:
378		if (td != NULL) {
379			error = priv_check(td, PRIV_NET_DELIFADDR);
380			if (error)
381				return (error);
382		}
383		/* FALLTHROUGH */
384	case SIOCGLIFADDR:
385		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
386	}
387
388	/*
389	 * Find address for this interface, if it exists.
390	 *
391	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
392	 * only, and used the first interface address as the target of other
393	 * operations (without checking ifra_addr).  This was because netinet
394	 * code/API assumed at most 1 interface address per interface.
395	 * Since IPv6 allows a node to assign multiple addresses
396	 * on a single interface, we almost always look and check the
397	 * presence of ifra_addr, and reject invalid ones here.
398	 * It also decreases duplicated code among SIOC*_IN6 operations.
399	 */
400	switch (cmd) {
401	case SIOCAIFADDR_IN6:
402	case SIOCSIFPHYADDR_IN6:
403		sa6 = &ifra->ifra_addr;
404		break;
405	case SIOCSIFADDR_IN6:
406	case SIOCGIFADDR_IN6:
407	case SIOCSIFDSTADDR_IN6:
408	case SIOCSIFNETMASK_IN6:
409	case SIOCGIFDSTADDR_IN6:
410	case SIOCGIFNETMASK_IN6:
411	case SIOCDIFADDR_IN6:
412	case SIOCGIFPSRCADDR_IN6:
413	case SIOCGIFPDSTADDR_IN6:
414	case SIOCGIFAFLAG_IN6:
415	case SIOCSNDFLUSH_IN6:
416	case SIOCSPFXFLUSH_IN6:
417	case SIOCSRTRFLUSH_IN6:
418	case SIOCGIFALIFETIME_IN6:
419	case SIOCSIFALIFETIME_IN6:
420	case SIOCGIFSTAT_IN6:
421	case SIOCGIFSTAT_ICMP6:
422		sa6 = &ifr->ifr_addr;
423		break;
424	default:
425		sa6 = NULL;
426		break;
427	}
428	if (sa6 && sa6->sin6_family == AF_INET6) {
429		if (sa6->sin6_scope_id != 0)
430			error = sa6_embedscope(sa6, 0);
431		else
432			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
433		if (error != 0)
434			return (error);
435		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
436		    &sa6->sin6_addr)) != 0)
437			return (error);
438		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
439	} else
440		ia = NULL;
441
442	switch (cmd) {
443	case SIOCSIFADDR_IN6:
444	case SIOCSIFDSTADDR_IN6:
445	case SIOCSIFNETMASK_IN6:
446		/*
447		 * Since IPv6 allows a node to assign multiple addresses
448		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
449		 */
450		/* we decided to obsolete this command (20000704) */
451		error = EINVAL;
452		goto out;
453
454	case SIOCDIFADDR_IN6:
455		/*
456		 * for IPv4, we look for existing in_ifaddr here to allow
457		 * "ifconfig if0 delete" to remove the first IPv4 address on
458		 * the interface.  For IPv6, as the spec allows multiple
459		 * interface address from the day one, we consider "remove the
460		 * first one" semantics to be not preferable.
461		 */
462		if (ia == NULL) {
463			error = EADDRNOTAVAIL;
464			goto out;
465		}
466		/* FALLTHROUGH */
467	case SIOCAIFADDR_IN6:
468		/*
469		 * We always require users to specify a valid IPv6 address for
470		 * the corresponding operation.
471		 */
472		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
473		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
474			error = EAFNOSUPPORT;
475			goto out;
476		}
477
478		if (td != NULL) {
479			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
480			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
481			if (error)
482				goto out;
483		}
484		break;
485
486	case SIOCGIFADDR_IN6:
487		/* This interface is basically deprecated. use SIOCGIFCONF. */
488		/* FALLTHROUGH */
489	case SIOCGIFAFLAG_IN6:
490	case SIOCGIFNETMASK_IN6:
491	case SIOCGIFDSTADDR_IN6:
492	case SIOCGIFALIFETIME_IN6:
493		/* must think again about its semantics */
494		if (ia == NULL) {
495			error = EADDRNOTAVAIL;
496			goto out;
497		}
498		break;
499
500	case SIOCSIFALIFETIME_IN6:
501	    {
502		struct in6_addrlifetime *lt;
503
504		if (td != NULL) {
505			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
506			if (error)
507				goto out;
508		}
509		if (ia == NULL) {
510			error = EADDRNOTAVAIL;
511			goto out;
512		}
513		/* sanity for overflow - beware unsigned */
514		lt = &ifr->ifr_ifru.ifru_lifetime;
515		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
516		    lt->ia6t_vltime + time_second < time_second) {
517			error = EINVAL;
518			goto out;
519		}
520		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
521		    lt->ia6t_pltime + time_second < time_second) {
522			error = EINVAL;
523			goto out;
524		}
525		break;
526	    }
527	}
528
529	switch (cmd) {
530	case SIOCGIFADDR_IN6:
531		ifr->ifr_addr = ia->ia_addr;
532		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
533			goto out;
534		break;
535
536	case SIOCGIFDSTADDR_IN6:
537		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
538			error = EINVAL;
539			goto out;
540		}
541		/*
542		 * XXX: should we check if ifa_dstaddr is NULL and return
543		 * an error?
544		 */
545		ifr->ifr_dstaddr = ia->ia_dstaddr;
546		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
547			goto out;
548		break;
549
550	case SIOCGIFNETMASK_IN6:
551		ifr->ifr_addr = ia->ia_prefixmask;
552		break;
553
554	case SIOCGIFAFLAG_IN6:
555		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
556		break;
557
558	case SIOCGIFSTAT_IN6:
559		if (ifp == NULL) {
560			error = EINVAL;
561			goto out;
562		}
563		bzero(&ifr->ifr_ifru.ifru_stat,
564		    sizeof(ifr->ifr_ifru.ifru_stat));
565		ifr->ifr_ifru.ifru_stat =
566		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
567		break;
568
569	case SIOCGIFSTAT_ICMP6:
570		if (ifp == NULL) {
571			error = EINVAL;
572			goto out;
573		}
574		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
575		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
576		ifr->ifr_ifru.ifru_icmp6stat =
577		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
578		break;
579
580	case SIOCGIFALIFETIME_IN6:
581		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
582		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
583			time_t maxexpire;
584			struct in6_addrlifetime *retlt =
585			    &ifr->ifr_ifru.ifru_lifetime;
586
587			/*
588			 * XXX: adjust expiration time assuming time_t is
589			 * signed.
590			 */
591			maxexpire = (-1) &
592			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
593			if (ia->ia6_lifetime.ia6t_vltime <
594			    maxexpire - ia->ia6_updatetime) {
595				retlt->ia6t_expire = ia->ia6_updatetime +
596				    ia->ia6_lifetime.ia6t_vltime;
597			} else
598				retlt->ia6t_expire = maxexpire;
599		}
600		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
601			time_t maxexpire;
602			struct in6_addrlifetime *retlt =
603			    &ifr->ifr_ifru.ifru_lifetime;
604
605			/*
606			 * XXX: adjust expiration time assuming time_t is
607			 * signed.
608			 */
609			maxexpire = (-1) &
610			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
611			if (ia->ia6_lifetime.ia6t_pltime <
612			    maxexpire - ia->ia6_updatetime) {
613				retlt->ia6t_preferred = ia->ia6_updatetime +
614				    ia->ia6_lifetime.ia6t_pltime;
615			} else
616				retlt->ia6t_preferred = maxexpire;
617		}
618		break;
619
620	case SIOCSIFALIFETIME_IN6:
621		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
622		/* for sanity */
623		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
624			ia->ia6_lifetime.ia6t_expire =
625				time_second + ia->ia6_lifetime.ia6t_vltime;
626		} else
627			ia->ia6_lifetime.ia6t_expire = 0;
628		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
629			ia->ia6_lifetime.ia6t_preferred =
630				time_second + ia->ia6_lifetime.ia6t_pltime;
631		} else
632			ia->ia6_lifetime.ia6t_preferred = 0;
633		break;
634
635	case SIOCAIFADDR_IN6:
636	{
637		int i;
638		struct nd_prefixctl pr0;
639		struct nd_prefix *pr;
640
641		/*
642		 * first, make or update the interface address structure,
643		 * and link it to the list.
644		 */
645		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
646			goto out;
647		if (ia != NULL)
648			ifa_free(&ia->ia_ifa);
649		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
650		    == NULL) {
651			/*
652			 * this can happen when the user specify the 0 valid
653			 * lifetime.
654			 */
655			break;
656		}
657
658		/*
659		 * then, make the prefix on-link on the interface.
660		 * XXX: we'd rather create the prefix before the address, but
661		 * we need at least one address to install the corresponding
662		 * interface route, so we configure the address first.
663		 */
664
665		/*
666		 * convert mask to prefix length (prefixmask has already
667		 * been validated in in6_update_ifa().
668		 */
669		bzero(&pr0, sizeof(pr0));
670		pr0.ndpr_ifp = ifp;
671		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
672		    NULL);
673		if (pr0.ndpr_plen == 128) {
674			break;	/* we don't need to install a host route. */
675		}
676		pr0.ndpr_prefix = ifra->ifra_addr;
677		/* apply the mask for safety. */
678		for (i = 0; i < 4; i++) {
679			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
680			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
681		}
682		/*
683		 * XXX: since we don't have an API to set prefix (not address)
684		 * lifetimes, we just use the same lifetimes as addresses.
685		 * The (temporarily) installed lifetimes can be overridden by
686		 * later advertised RAs (when accept_rtadv is non 0), which is
687		 * an intended behavior.
688		 */
689		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
690		pr0.ndpr_raf_auto =
691		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
692		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
693		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
694
695		/* add the prefix if not yet. */
696		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
697			/*
698			 * nd6_prelist_add will install the corresponding
699			 * interface route.
700			 */
701			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
702				goto out;
703			if (pr == NULL) {
704				log(LOG_ERR, "nd6_prelist_add succeeded but "
705				    "no prefix\n");
706				error = EINVAL;
707				goto out;
708			}
709		}
710
711		/* relate the address to the prefix */
712		if (ia->ia6_ndpr == NULL) {
713			ia->ia6_ndpr = pr;
714			pr->ndpr_refcnt++;
715
716			/*
717			 * If this is the first autoconf address from the
718			 * prefix, create a temporary address as well
719			 * (when required).
720			 */
721			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
722			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
723				int e;
724				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
725					log(LOG_NOTICE, "in6_control: failed "
726					    "to create a temporary address, "
727					    "errno=%d\n", e);
728				}
729			}
730		}
731
732		/*
733		 * this might affect the status of autoconfigured addresses,
734		 * that is, this address might make other addresses detached.
735		 */
736		pfxlist_onlink_check();
737		if (error == 0 && ia) {
738			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
739				/*
740				 * Try to clear the flag when a new
741				 * IPv6 address is added onto an
742				 * IFDISABLED interface and it
743				 * succeeds.
744				 */
745				struct in6_ndireq nd;
746
747				memset(&nd, 0, sizeof(nd));
748				nd.ndi.flags = ND_IFINFO(ifp)->flags;
749				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
750				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
751				    (caddr_t)&nd, ifp) < 0)
752					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
753					    "SIOCSIFINFO_FLAGS for -ifdisabled "
754					    "failed.");
755				/*
756				 * Ignore failure of clearing the flag
757				 * intentionally.  The failure means
758				 * address duplication was detected.
759				 */
760			}
761			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
762		}
763		break;
764	}
765
766	case SIOCDIFADDR_IN6:
767	{
768		struct nd_prefix *pr;
769
770		/*
771		 * If the address being deleted is the only one that owns
772		 * the corresponding prefix, expire the prefix as well.
773		 * XXX: theoretically, we don't have to worry about such
774		 * relationship, since we separate the address management
775		 * and the prefix management.  We do this, however, to provide
776		 * as much backward compatibility as possible in terms of
777		 * the ioctl operation.
778		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
779		 */
780		pr = ia->ia6_ndpr;
781		in6_purgeaddr(&ia->ia_ifa);
782		if (pr && pr->ndpr_refcnt == 0)
783			prelist_remove(pr);
784		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
785		break;
786	}
787
788	default:
789		if (ifp == NULL || ifp->if_ioctl == 0) {
790			error = EOPNOTSUPP;
791			goto out;
792		}
793		error = (*ifp->if_ioctl)(ifp, cmd, data);
794		goto out;
795	}
796
797	error = 0;
798out:
799	if (ia != NULL)
800		ifa_free(&ia->ia_ifa);
801	return (error);
802}
803
804
805/*
806 * Join necessary multicast groups.  Factored out from in6_update_ifa().
807 * This entire work should only be done once, for the default FIB.
808 */
809static int
810in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
811    struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
812{
813	char ip6buf[INET6_ADDRSTRLEN];
814	struct sockaddr_in6 mltaddr, mltmask;
815	struct in6_addr llsol;
816	struct in6_multi_mship *imm;
817	struct rtentry *rt;
818	int delay, error;
819
820	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
821
822	/* Join solicited multicast addr for new host id. */
823	bzero(&llsol, sizeof(struct in6_addr));
824	llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
825	llsol.s6_addr32[1] = 0;
826	llsol.s6_addr32[2] = htonl(1);
827	llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
828	llsol.s6_addr8[12] = 0xff;
829	if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
830		/* XXX: should not happen */
831		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
832		goto cleanup;
833	}
834	delay = 0;
835	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
836		/*
837		 * We need a random delay for DAD on the address being
838		 * configured.  It also means delaying transmission of the
839		 * corresponding MLD report to avoid report collision.
840		 * [RFC 4861, Section 6.3.7]
841		 */
842		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
843	}
844	imm = in6_joingroup(ifp, &llsol, &error, delay);
845	if (imm == NULL) {
846		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
847		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &llsol),
848		    if_name(ifp), error));
849		goto cleanup;
850	}
851	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
852	*in6m_sol = imm->i6mm_maddr;
853
854	bzero(&mltmask, sizeof(mltmask));
855	mltmask.sin6_len = sizeof(struct sockaddr_in6);
856	mltmask.sin6_family = AF_INET6;
857	mltmask.sin6_addr = in6mask32;
858#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
859
860	/*
861	 * Join link-local all-nodes address.
862	 */
863	bzero(&mltaddr, sizeof(mltaddr));
864	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
865	mltaddr.sin6_family = AF_INET6;
866	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
867	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
868		goto cleanup; /* XXX: should not fail */
869
870	/*
871	 * XXX: do we really need this automatic routes?  We should probably
872	 * reconsider this stuff.  Most applications actually do not need the
873	 * routes, since they usually specify the outgoing interface.
874	 */
875	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
876	if (rt != NULL) {
877		/* XXX: only works in !SCOPEDROUTING case. */
878		if (memcmp(&mltaddr.sin6_addr,
879		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
880		    MLTMASK_LEN)) {
881			RTFREE_LOCKED(rt);
882			rt = NULL;
883		}
884	}
885	if (rt == NULL) {
886		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
887		    (struct sockaddr *)&ia->ia_addr,
888		    (struct sockaddr *)&mltmask, RTF_UP,
889		    (struct rtentry **)0, RT_DEFAULT_FIB);
890		if (error)
891			goto cleanup;
892	} else
893		RTFREE_LOCKED(rt);
894
895	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
896	if (imm == NULL) {
897		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
898		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
899		    &mltaddr.sin6_addr), if_name(ifp), error));
900		goto cleanup;
901	}
902	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
903
904	/*
905	 * Join node information group address.
906	 */
907	delay = 0;
908	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
909		/*
910		 * The spec does not say anything about delay for this group,
911		 * but the same logic should apply.
912		 */
913		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
914	}
915	if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
916		/* XXX jinmei */
917		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
918		if (imm == NULL)
919			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
920			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
921			    &mltaddr.sin6_addr), if_name(ifp), error));
922			/* XXX not very fatal, go on... */
923		else
924			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
925	}
926
927	/*
928	 * Join interface-local all-nodes address.
929	 * (ff01::1%ifN, and ff01::%ifN/32)
930	 */
931	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
932	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
933		goto cleanup; /* XXX: should not fail */
934	/* XXX: again, do we really need the route? */
935	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
936	if (rt != NULL) {
937		if (memcmp(&mltaddr.sin6_addr,
938		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
939		    MLTMASK_LEN)) {
940			RTFREE_LOCKED(rt);
941			rt = NULL;
942		}
943	}
944	if (rt == NULL) {
945		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
946		    (struct sockaddr *)&ia->ia_addr,
947		    (struct sockaddr *)&mltmask, RTF_UP,
948		    (struct rtentry **)0, RT_DEFAULT_FIB);
949		if (error)
950			goto cleanup;
951	} else
952		RTFREE_LOCKED(rt);
953
954	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
955	if (imm == NULL) {
956		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
957		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
958		    &mltaddr.sin6_addr), if_name(ifp), error));
959		goto cleanup;
960	}
961	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
962#undef	MLTMASK_LEN
963
964cleanup:
965	return (error);
966}
967
968/*
969 * Update parameters of an IPv6 interface address.
970 * If necessary, a new entry is created and linked into address chains.
971 * This function is separated from in6_control().
972 * XXX: should this be performed under splnet()?
973 */
974int
975in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
976    struct in6_ifaddr *ia, int flags)
977{
978	int error = 0, hostIsNew = 0, plen = -1;
979	struct sockaddr_in6 dst6;
980	struct in6_addrlifetime *lt;
981	struct in6_multi *in6m_sol;
982	int delay;
983	char ip6buf[INET6_ADDRSTRLEN];
984
985	/* Validate parameters */
986	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
987		return (EINVAL);
988
989	/*
990	 * The destination address for a p2p link must have a family
991	 * of AF_UNSPEC or AF_INET6.
992	 */
993	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
994	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
995	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
996		return (EAFNOSUPPORT);
997	/*
998	 * validate ifra_prefixmask.  don't check sin6_family, netmask
999	 * does not carry fields other than sin6_len.
1000	 */
1001	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1002		return (EINVAL);
1003	/*
1004	 * Because the IPv6 address architecture is classless, we require
1005	 * users to specify a (non 0) prefix length (mask) for a new address.
1006	 * We also require the prefix (when specified) mask is valid, and thus
1007	 * reject a non-consecutive mask.
1008	 */
1009	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1010		return (EINVAL);
1011	if (ifra->ifra_prefixmask.sin6_len != 0) {
1012		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1013		    (u_char *)&ifra->ifra_prefixmask +
1014		    ifra->ifra_prefixmask.sin6_len);
1015		if (plen <= 0)
1016			return (EINVAL);
1017	} else {
1018		/*
1019		 * In this case, ia must not be NULL.  We just use its prefix
1020		 * length.
1021		 */
1022		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1023	}
1024	/*
1025	 * If the destination address on a p2p interface is specified,
1026	 * and the address is a scoped one, validate/set the scope
1027	 * zone identifier.
1028	 */
1029	dst6 = ifra->ifra_dstaddr;
1030	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1031	    (dst6.sin6_family == AF_INET6)) {
1032		struct in6_addr in6_tmp;
1033		u_int32_t zoneid;
1034
1035		in6_tmp = dst6.sin6_addr;
1036		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1037			return (EINVAL); /* XXX: should be impossible */
1038
1039		if (dst6.sin6_scope_id != 0) {
1040			if (dst6.sin6_scope_id != zoneid)
1041				return (EINVAL);
1042		} else		/* user omit to specify the ID. */
1043			dst6.sin6_scope_id = zoneid;
1044
1045		/* convert into the internal form */
1046		if (sa6_embedscope(&dst6, 0))
1047			return (EINVAL); /* XXX: should be impossible */
1048	}
1049	/*
1050	 * The destination address can be specified only for a p2p or a
1051	 * loopback interface.  If specified, the corresponding prefix length
1052	 * must be 128.
1053	 */
1054	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1055		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1056			/* XXX: noisy message */
1057			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1058			    "be specified for a p2p or a loopback IF only\n"));
1059			return (EINVAL);
1060		}
1061		if (plen != 128) {
1062			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1063			    "be 128 when dstaddr is specified\n"));
1064			return (EINVAL);
1065		}
1066	}
1067	/* lifetime consistency check */
1068	lt = &ifra->ifra_lifetime;
1069	if (lt->ia6t_pltime > lt->ia6t_vltime)
1070		return (EINVAL);
1071	if (lt->ia6t_vltime == 0) {
1072		/*
1073		 * the following log might be noisy, but this is a typical
1074		 * configuration mistake or a tool's bug.
1075		 */
1076		nd6log((LOG_INFO,
1077		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1078		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1079
1080		if (ia == NULL)
1081			return (0); /* there's nothing to do */
1082	}
1083
1084	/*
1085	 * If this is a new address, allocate a new ifaddr and link it
1086	 * into chains.
1087	 */
1088	if (ia == NULL) {
1089		hostIsNew = 1;
1090		/*
1091		 * When in6_update_ifa() is called in a process of a received
1092		 * RA, it is called under an interrupt context.  So, we should
1093		 * call malloc with M_NOWAIT.
1094		 */
1095		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
1096		    M_NOWAIT);
1097		if (ia == NULL)
1098			return (ENOBUFS);
1099		bzero((caddr_t)ia, sizeof(*ia));
1100		ifa_init(&ia->ia_ifa);
1101		LIST_INIT(&ia->ia6_memberships);
1102		/* Initialize the address and masks, and put time stamp */
1103		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1104		ia->ia_addr.sin6_family = AF_INET6;
1105		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1106		ia->ia6_createtime = time_second;
1107		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1108			/*
1109			 * XXX: some functions expect that ifa_dstaddr is not
1110			 * NULL for p2p interfaces.
1111			 */
1112			ia->ia_ifa.ifa_dstaddr =
1113			    (struct sockaddr *)&ia->ia_dstaddr;
1114		} else {
1115			ia->ia_ifa.ifa_dstaddr = NULL;
1116		}
1117		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1118		ia->ia_ifp = ifp;
1119		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1120		IF_ADDR_WLOCK(ifp);
1121		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1122		IF_ADDR_WUNLOCK(ifp);
1123
1124		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1125		IN6_IFADDR_WLOCK();
1126		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1127		IN6_IFADDR_WUNLOCK();
1128	}
1129
1130	/* update timestamp */
1131	ia->ia6_updatetime = time_second;
1132
1133	/* set prefix mask */
1134	if (ifra->ifra_prefixmask.sin6_len) {
1135		/*
1136		 * We prohibit changing the prefix length of an existing
1137		 * address, because
1138		 * + such an operation should be rare in IPv6, and
1139		 * + the operation would confuse prefix management.
1140		 */
1141		if (ia->ia_prefixmask.sin6_len &&
1142		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1143			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1144			    " existing (%s) address should not be changed\n",
1145			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1146			error = EINVAL;
1147			goto unlink;
1148		}
1149		ia->ia_prefixmask = ifra->ifra_prefixmask;
1150	}
1151
1152	/*
1153	 * If a new destination address is specified, scrub the old one and
1154	 * install the new destination.  Note that the interface must be
1155	 * p2p or loopback (see the check above.)
1156	 */
1157	if (dst6.sin6_family == AF_INET6 &&
1158	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1159		int e;
1160
1161		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1162		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1163			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1164			    "a route to the old destination: %s\n",
1165			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1166			/* proceed anyway... */
1167		} else
1168			ia->ia_flags &= ~IFA_ROUTE;
1169		ia->ia_dstaddr = dst6;
1170	}
1171
1172	/*
1173	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1174	 * to see if the address is deprecated or invalidated, but initialize
1175	 * these members for applications.
1176	 */
1177	ia->ia6_lifetime = ifra->ifra_lifetime;
1178	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1179		ia->ia6_lifetime.ia6t_expire =
1180		    time_second + ia->ia6_lifetime.ia6t_vltime;
1181	} else
1182		ia->ia6_lifetime.ia6t_expire = 0;
1183	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1184		ia->ia6_lifetime.ia6t_preferred =
1185		    time_second + ia->ia6_lifetime.ia6t_pltime;
1186	} else
1187		ia->ia6_lifetime.ia6t_preferred = 0;
1188
1189	/* reset the interface and routing table appropriately. */
1190	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1191		goto unlink;
1192
1193	/*
1194	 * configure address flags.
1195	 */
1196	ia->ia6_flags = ifra->ifra_flags;
1197	/*
1198	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1199	 * userland, make it deprecated.
1200	 */
1201	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1202		ia->ia6_lifetime.ia6t_pltime = 0;
1203		ia->ia6_lifetime.ia6t_preferred = time_second;
1204	}
1205	/*
1206	 * Make the address tentative before joining multicast addresses,
1207	 * so that corresponding MLD responses would not have a tentative
1208	 * source address.
1209	 */
1210	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1211	if (hostIsNew && in6if_do_dad(ifp))
1212		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1213
1214	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
1215	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1216		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1217
1218	/*
1219	 * We are done if we have simply modified an existing address.
1220	 */
1221	if (!hostIsNew)
1222		return (error);
1223
1224	/*
1225	 * Beyond this point, we should call in6_purgeaddr upon an error,
1226	 * not just go to unlink.
1227	 */
1228
1229	/* Join necessary multicast groups. */
1230	in6m_sol = NULL;
1231	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1232		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1233		if (error)
1234			goto cleanup;
1235	}
1236
1237	/*
1238	 * Perform DAD, if needed.
1239	 * XXX It may be of use, if we can administratively disable DAD.
1240	 */
1241	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1242	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1243	{
1244		int mindelay, maxdelay;
1245
1246		delay = 0;
1247		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1248			/*
1249			 * We need to impose a delay before sending an NS
1250			 * for DAD.  Check if we also needed a delay for the
1251			 * corresponding MLD message.  If we did, the delay
1252			 * should be larger than the MLD delay (this could be
1253			 * relaxed a bit, but this simple logic is at least
1254			 * safe).
1255			 * XXX: Break data hiding guidelines and look at
1256			 * state for the solicited multicast group.
1257			 */
1258			mindelay = 0;
1259			if (in6m_sol != NULL &&
1260			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1261				mindelay = in6m_sol->in6m_timer;
1262			}
1263			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1264			if (maxdelay - mindelay == 0)
1265				delay = 0;
1266			else {
1267				delay =
1268				    (arc4random() % (maxdelay - mindelay)) +
1269				    mindelay;
1270			}
1271		}
1272		nd6_dad_start((struct ifaddr *)ia, delay);
1273	}
1274
1275	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1276	ifa_free(&ia->ia_ifa);
1277	return (error);
1278
1279  unlink:
1280	/*
1281	 * XXX: if a change of an existing address failed, keep the entry
1282	 * anyway.
1283	 */
1284	if (hostIsNew) {
1285		in6_unlink_ifa(ia, ifp);
1286		ifa_free(&ia->ia_ifa);
1287	}
1288	return (error);
1289
1290  cleanup:
1291	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1292	ifa_free(&ia->ia_ifa);
1293	in6_purgeaddr(&ia->ia_ifa);
1294	return error;
1295}
1296
1297/*
1298 * Leave multicast groups.  Factored out from in6_purgeaddr().
1299 * This entire work should only be done once, for the default FIB.
1300 */
1301static int
1302in6_purgeaddr_mc(struct ifnet *ifp, struct in6_ifaddr *ia, struct ifaddr *ifa0)
1303{
1304	struct sockaddr_in6 mltaddr, mltmask;
1305	struct in6_multi_mship *imm;
1306	struct rtentry *rt;
1307	struct sockaddr_in6 sin6;
1308	int error;
1309
1310	/*
1311	 * Leave from multicast groups we have joined for the interface.
1312	 */
1313	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1314		LIST_REMOVE(imm, i6mm_chain);
1315		in6_leavegroup(imm);
1316	}
1317
1318	/*
1319	 * Remove the link-local all-nodes address.
1320	 */
1321	bzero(&mltmask, sizeof(mltmask));
1322	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1323	mltmask.sin6_family = AF_INET6;
1324	mltmask.sin6_addr = in6mask32;
1325
1326	bzero(&mltaddr, sizeof(mltaddr));
1327	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1328	mltaddr.sin6_family = AF_INET6;
1329	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1330
1331	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1332		return (error);
1333
1334	/*
1335	 * As for the mltaddr above, proactively prepare the sin6 to avoid
1336	 * rtentry un- and re-locking.
1337	 */
1338	if (ifa0 != NULL) {
1339		bzero(&sin6, sizeof(sin6));
1340		sin6.sin6_len = sizeof(sin6);
1341		sin6.sin6_family = AF_INET6;
1342		memcpy(&sin6.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1343		    sizeof(sin6.sin6_addr));
1344		error = in6_setscope(&sin6.sin6_addr, ifa0->ifa_ifp, NULL);
1345		if (error != 0)
1346			return (error);
1347	}
1348
1349	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1350	if (rt != NULL && rt->rt_gateway != NULL &&
1351	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1352		    &ia->ia_addr.sin6_addr,
1353		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1354		/*
1355		 * If no more IPv6 address exists on this interface then
1356		 * remove the multicast address route.
1357		 */
1358		if (ifa0 == NULL) {
1359			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1360			       sizeof(mltaddr.sin6_addr));
1361			RTFREE_LOCKED(rt);
1362			error = in6_rtrequest(RTM_DELETE,
1363			    (struct sockaddr *)&mltaddr,
1364			    (struct sockaddr *)&ia->ia_addr,
1365			    (struct sockaddr *)&mltmask, RTF_UP,
1366			    (struct rtentry **)0, RT_DEFAULT_FIB);
1367			if (error)
1368				log(LOG_INFO, "%s: link-local all-nodes "
1369				    "multicast address deletion error\n",
1370				    __func__);
1371		} else {
1372			/*
1373			 * Replace the gateway of the route.
1374			 */
1375			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1376			RTFREE_LOCKED(rt);
1377		}
1378	} else {
1379		if (rt != NULL)
1380			RTFREE_LOCKED(rt);
1381	}
1382
1383	/*
1384	 * Remove the node-local all-nodes address.
1385	 */
1386	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1387	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1388		return (error);
1389
1390	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1391	if (rt != NULL && rt->rt_gateway != NULL &&
1392	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1393		    &ia->ia_addr.sin6_addr,
1394		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1395		/*
1396		 * If no more IPv6 address exists on this interface then
1397		 * remove the multicast address route.
1398		 */
1399		if (ifa0 == NULL) {
1400			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1401			       sizeof(mltaddr.sin6_addr));
1402
1403			RTFREE_LOCKED(rt);
1404			error = in6_rtrequest(RTM_DELETE,
1405			    (struct sockaddr *)&mltaddr,
1406			    (struct sockaddr *)&ia->ia_addr,
1407			    (struct sockaddr *)&mltmask, RTF_UP,
1408			    (struct rtentry **)0, RT_DEFAULT_FIB);
1409			if (error)
1410				log(LOG_INFO, "%s: node-local all-nodes"
1411				    "multicast address deletion error\n",
1412				    __func__);
1413		} else {
1414			/*
1415			 * Replace the gateway of the route.
1416			 */
1417			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1418			RTFREE_LOCKED(rt);
1419		}
1420	} else {
1421		if (rt != NULL)
1422			RTFREE_LOCKED(rt);
1423	}
1424
1425	return (0);
1426}
1427
1428void
1429in6_purgeaddr(struct ifaddr *ifa)
1430{
1431	struct ifnet *ifp = ifa->ifa_ifp;
1432	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1433	int plen, error;
1434	struct ifaddr *ifa0;
1435
1436	/*
1437	 * find another IPv6 address as the gateway for the
1438	 * link-local and node-local all-nodes multicast
1439	 * address routes
1440	 */
1441	IF_ADDR_RLOCK(ifp);
1442	TAILQ_FOREACH(ifa0, &ifp->if_addrhead, ifa_link) {
1443		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1444		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1445			   &ia->ia_addr.sin6_addr,
1446			   sizeof(struct in6_addr)) == 0)
1447			continue;
1448		else
1449			break;
1450	}
1451	if (ifa0 != NULL)
1452		ifa_ref(ifa0);
1453	IF_ADDR_RUNLOCK(ifp);
1454
1455	/*
1456	 * Remove the loopback route to the interface address.
1457	 * The check for the current setting of "nd6_useloopback"
1458	 * is not needed.
1459	 */
1460	if (ia->ia_flags & IFA_RTSELF) {
1461		error = ifa_del_loopback_route((struct ifaddr *)ia,
1462				       (struct sockaddr *)&ia->ia_addr);
1463		if (error == 0)
1464			ia->ia_flags &= ~IFA_RTSELF;
1465	}
1466
1467	/* stop DAD processing */
1468	nd6_dad_stop(ifa);
1469
1470	/* Remove local address entry from lltable. */
1471	in6_ifremloop(ifa);
1472
1473	/* Leave multicast groups. */
1474	error = in6_purgeaddr_mc(ifp, ia, ifa0);
1475
1476	if (ifa0 != NULL)
1477		ifa_free(ifa0);
1478
1479	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1480	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1481		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1482		    (ia->ia_dstaddr.sin6_family == AF_INET6) ? RTF_HOST : 0);
1483		if (error != 0)
1484			log(LOG_INFO, "%s: err=%d, destination address delete "
1485			    "failed\n", __func__, error);
1486		ia->ia_flags &= ~IFA_ROUTE;
1487	}
1488
1489	in6_unlink_ifa(ia, ifp);
1490}
1491
1492static void
1493in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1494{
1495	int	s = splnet();
1496
1497	IF_ADDR_WLOCK(ifp);
1498	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1499	IF_ADDR_WUNLOCK(ifp);
1500	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1501
1502	/*
1503	 * Defer the release of what might be the last reference to the
1504	 * in6_ifaddr so that it can't be freed before the remainder of the
1505	 * cleanup.
1506	 */
1507	IN6_IFADDR_WLOCK();
1508	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1509	IN6_IFADDR_WUNLOCK();
1510
1511	/*
1512	 * Release the reference to the base prefix.  There should be a
1513	 * positive reference.
1514	 */
1515	if (ia->ia6_ndpr == NULL) {
1516		nd6log((LOG_NOTICE,
1517		    "in6_unlink_ifa: autoconf'ed address "
1518		    "%p has no prefix\n", ia));
1519	} else {
1520		ia->ia6_ndpr->ndpr_refcnt--;
1521		ia->ia6_ndpr = NULL;
1522	}
1523
1524	/*
1525	 * Also, if the address being removed is autoconf'ed, call
1526	 * pfxlist_onlink_check() since the release might affect the status of
1527	 * other (detached) addresses.
1528	 */
1529	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1530		pfxlist_onlink_check();
1531	}
1532	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1533	splx(s);
1534}
1535
1536void
1537in6_purgeif(struct ifnet *ifp)
1538{
1539	struct ifaddr *ifa, *nifa;
1540
1541	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1542		if (ifa->ifa_addr->sa_family != AF_INET6)
1543			continue;
1544		in6_purgeaddr(ifa);
1545	}
1546
1547	in6_ifdetach(ifp);
1548}
1549
1550/*
1551 * SIOC[GAD]LIFADDR.
1552 *	SIOCGLIFADDR: get first address. (?)
1553 *	SIOCGLIFADDR with IFLR_PREFIX:
1554 *		get first address that matches the specified prefix.
1555 *	SIOCALIFADDR: add the specified address.
1556 *	SIOCALIFADDR with IFLR_PREFIX:
1557 *		add the specified prefix, filling hostid part from
1558 *		the first link-local address.  prefixlen must be <= 64.
1559 *	SIOCDLIFADDR: delete the specified address.
1560 *	SIOCDLIFADDR with IFLR_PREFIX:
1561 *		delete the first address that matches the specified prefix.
1562 * return values:
1563 *	EINVAL on invalid parameters
1564 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1565 *	other values may be returned from in6_ioctl()
1566 *
1567 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1568 * this is to accomodate address naming scheme other than RFC2374,
1569 * in the future.
1570 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1571 * address encoding scheme. (see figure on page 8)
1572 */
1573static int
1574in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1575    struct ifnet *ifp, struct thread *td)
1576{
1577	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1578	struct ifaddr *ifa;
1579	struct sockaddr *sa;
1580
1581	/* sanity checks */
1582	if (!data || !ifp) {
1583		panic("invalid argument to in6_lifaddr_ioctl");
1584		/* NOTREACHED */
1585	}
1586
1587	switch (cmd) {
1588	case SIOCGLIFADDR:
1589		/* address must be specified on GET with IFLR_PREFIX */
1590		if ((iflr->flags & IFLR_PREFIX) == 0)
1591			break;
1592		/* FALLTHROUGH */
1593	case SIOCALIFADDR:
1594	case SIOCDLIFADDR:
1595		/* address must be specified on ADD and DELETE */
1596		sa = (struct sockaddr *)&iflr->addr;
1597		if (sa->sa_family != AF_INET6)
1598			return EINVAL;
1599		if (sa->sa_len != sizeof(struct sockaddr_in6))
1600			return EINVAL;
1601		/* XXX need improvement */
1602		sa = (struct sockaddr *)&iflr->dstaddr;
1603		if (sa->sa_family && sa->sa_family != AF_INET6)
1604			return EINVAL;
1605		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1606			return EINVAL;
1607		break;
1608	default: /* shouldn't happen */
1609#if 0
1610		panic("invalid cmd to in6_lifaddr_ioctl");
1611		/* NOTREACHED */
1612#else
1613		return EOPNOTSUPP;
1614#endif
1615	}
1616	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1617		return EINVAL;
1618
1619	switch (cmd) {
1620	case SIOCALIFADDR:
1621	    {
1622		struct in6_aliasreq ifra;
1623		struct in6_addr *hostid = NULL;
1624		int prefixlen;
1625
1626		ifa = NULL;
1627		if ((iflr->flags & IFLR_PREFIX) != 0) {
1628			struct sockaddr_in6 *sin6;
1629
1630			/*
1631			 * hostid is to fill in the hostid part of the
1632			 * address.  hostid points to the first link-local
1633			 * address attached to the interface.
1634			 */
1635			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1636			if (!ifa)
1637				return EADDRNOTAVAIL;
1638			hostid = IFA_IN6(ifa);
1639
1640			/* prefixlen must be <= 64. */
1641			if (64 < iflr->prefixlen) {
1642				if (ifa != NULL)
1643					ifa_free(ifa);
1644				return EINVAL;
1645			}
1646			prefixlen = iflr->prefixlen;
1647
1648			/* hostid part must be zero. */
1649			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1650			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1651			    sin6->sin6_addr.s6_addr32[3] != 0) {
1652				if (ifa != NULL)
1653					ifa_free(ifa);
1654				return EINVAL;
1655			}
1656		} else
1657			prefixlen = iflr->prefixlen;
1658
1659		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1660		bzero(&ifra, sizeof(ifra));
1661		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1662
1663		bcopy(&iflr->addr, &ifra.ifra_addr,
1664		    ((struct sockaddr *)&iflr->addr)->sa_len);
1665		if (hostid) {
1666			/* fill in hostid part */
1667			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1668			    hostid->s6_addr32[2];
1669			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1670			    hostid->s6_addr32[3];
1671		}
1672
1673		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1674			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1675			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1676			if (hostid) {
1677				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1678				    hostid->s6_addr32[2];
1679				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1680				    hostid->s6_addr32[3];
1681			}
1682		}
1683		if (ifa != NULL)
1684			ifa_free(ifa);
1685
1686		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1687		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1688
1689		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1690		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1691	    }
1692	case SIOCGLIFADDR:
1693	case SIOCDLIFADDR:
1694	    {
1695		struct in6_ifaddr *ia;
1696		struct in6_addr mask, candidate, match;
1697		struct sockaddr_in6 *sin6;
1698		int cmp;
1699
1700		bzero(&mask, sizeof(mask));
1701		if (iflr->flags & IFLR_PREFIX) {
1702			/* lookup a prefix rather than address. */
1703			in6_prefixlen2mask(&mask, iflr->prefixlen);
1704
1705			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1706			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1707			match.s6_addr32[0] &= mask.s6_addr32[0];
1708			match.s6_addr32[1] &= mask.s6_addr32[1];
1709			match.s6_addr32[2] &= mask.s6_addr32[2];
1710			match.s6_addr32[3] &= mask.s6_addr32[3];
1711
1712			/* if you set extra bits, that's wrong */
1713			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1714				return EINVAL;
1715
1716			cmp = 1;
1717		} else {
1718			if (cmd == SIOCGLIFADDR) {
1719				/* on getting an address, take the 1st match */
1720				cmp = 0;	/* XXX */
1721			} else {
1722				/* on deleting an address, do exact match */
1723				in6_prefixlen2mask(&mask, 128);
1724				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1725				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1726
1727				cmp = 1;
1728			}
1729		}
1730
1731		IF_ADDR_RLOCK(ifp);
1732		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1733			if (ifa->ifa_addr->sa_family != AF_INET6)
1734				continue;
1735			if (!cmp)
1736				break;
1737
1738			/*
1739			 * XXX: this is adhoc, but is necessary to allow
1740			 * a user to specify fe80::/64 (not /10) for a
1741			 * link-local address.
1742			 */
1743			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1744			in6_clearscope(&candidate);
1745			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1746			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1747			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1748			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1749			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1750				break;
1751		}
1752		if (ifa != NULL)
1753			ifa_ref(ifa);
1754		IF_ADDR_RUNLOCK(ifp);
1755		if (!ifa)
1756			return EADDRNOTAVAIL;
1757		ia = ifa2ia6(ifa);
1758
1759		if (cmd == SIOCGLIFADDR) {
1760			int error;
1761
1762			/* fill in the if_laddrreq structure */
1763			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1764			error = sa6_recoverscope(
1765			    (struct sockaddr_in6 *)&iflr->addr);
1766			if (error != 0) {
1767				ifa_free(ifa);
1768				return (error);
1769			}
1770
1771			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1772				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1773				    ia->ia_dstaddr.sin6_len);
1774				error = sa6_recoverscope(
1775				    (struct sockaddr_in6 *)&iflr->dstaddr);
1776				if (error != 0) {
1777					ifa_free(ifa);
1778					return (error);
1779				}
1780			} else
1781				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1782
1783			iflr->prefixlen =
1784			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1785
1786			iflr->flags = ia->ia6_flags;	/* XXX */
1787			ifa_free(ifa);
1788
1789			return 0;
1790		} else {
1791			struct in6_aliasreq ifra;
1792
1793			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1794			bzero(&ifra, sizeof(ifra));
1795			bcopy(iflr->iflr_name, ifra.ifra_name,
1796			    sizeof(ifra.ifra_name));
1797
1798			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1799			    ia->ia_addr.sin6_len);
1800			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1801				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1802				    ia->ia_dstaddr.sin6_len);
1803			} else {
1804				bzero(&ifra.ifra_dstaddr,
1805				    sizeof(ifra.ifra_dstaddr));
1806			}
1807			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1808			    ia->ia_prefixmask.sin6_len);
1809
1810			ifra.ifra_flags = ia->ia6_flags;
1811			ifa_free(ifa);
1812			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1813			    ifp, td);
1814		}
1815	    }
1816	}
1817
1818	return EOPNOTSUPP;	/* just for safety */
1819}
1820
1821/*
1822 * Initialize an interface's IPv6 address and routing table entry.
1823 */
1824static int
1825in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1826    struct sockaddr_in6 *sin6, int newhost)
1827{
1828	int	error = 0, plen, ifacount = 0;
1829	int	s = splimp();
1830	struct ifaddr *ifa;
1831
1832	/*
1833	 * Give the interface a chance to initialize
1834	 * if this is its first address,
1835	 * and to validate the address if necessary.
1836	 */
1837	IF_ADDR_RLOCK(ifp);
1838	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1839		if (ifa->ifa_addr->sa_family != AF_INET6)
1840			continue;
1841		ifacount++;
1842	}
1843	IF_ADDR_RUNLOCK(ifp);
1844
1845	ia->ia_addr = *sin6;
1846
1847	if (ifacount <= 1 && ifp->if_ioctl) {
1848		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1849		if (error) {
1850			splx(s);
1851			return (error);
1852		}
1853	}
1854	splx(s);
1855
1856	ia->ia_ifa.ifa_metric = ifp->if_metric;
1857
1858	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1859
1860	/*
1861	 * Special case:
1862	 * If a new destination address is specified for a point-to-point
1863	 * interface, install a route to the destination as an interface
1864	 * direct route.
1865	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1866	 * interface that share the same destination.
1867	 */
1868	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1869	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1870	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1871		int rtflags = RTF_UP | RTF_HOST;
1872		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1873		if (error)
1874			return (error);
1875		ia->ia_flags |= IFA_ROUTE;
1876		/*
1877		 * Handle the case for ::1 .
1878		 */
1879		if (ifp->if_flags & IFF_LOOPBACK)
1880			ia->ia_flags |= IFA_RTSELF;
1881	}
1882
1883	/*
1884	 * add a loopback route to self
1885	 */
1886	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1887		error = ifa_add_loopback_route((struct ifaddr *)ia,
1888				       (struct sockaddr *)&ia->ia_addr);
1889		if (error == 0)
1890			ia->ia_flags |= IFA_RTSELF;
1891	}
1892
1893	/* Add local address to lltable, if necessary (ex. on p2p link). */
1894	if (newhost)
1895		in6_ifaddloop(&(ia->ia_ifa));
1896
1897	return (error);
1898}
1899
1900/*
1901 * Find an IPv6 interface link-local address specific to an interface.
1902 * ifaddr is returned referenced.
1903 */
1904struct in6_ifaddr *
1905in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1906{
1907	struct ifaddr *ifa;
1908
1909	IF_ADDR_RLOCK(ifp);
1910	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1911		if (ifa->ifa_addr->sa_family != AF_INET6)
1912			continue;
1913		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1914			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1915			     ignoreflags) != 0)
1916				continue;
1917			ifa_ref(ifa);
1918			break;
1919		}
1920	}
1921	IF_ADDR_RUNLOCK(ifp);
1922
1923	return ((struct in6_ifaddr *)ifa);
1924}
1925
1926
1927/*
1928 * find the internet address corresponding to a given interface and address.
1929 * ifaddr is returned referenced.
1930 */
1931struct in6_ifaddr *
1932in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1933{
1934	struct ifaddr *ifa;
1935
1936	IF_ADDR_RLOCK(ifp);
1937	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1938		if (ifa->ifa_addr->sa_family != AF_INET6)
1939			continue;
1940		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1941			ifa_ref(ifa);
1942			break;
1943		}
1944	}
1945	IF_ADDR_RUNLOCK(ifp);
1946
1947	return ((struct in6_ifaddr *)ifa);
1948}
1949
1950/*
1951 * Convert IP6 address to printable (loggable) representation. Caller
1952 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1953 */
1954static char digits[] = "0123456789abcdef";
1955char *
1956ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1957{
1958	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1959	char *cp;
1960	const u_int16_t *a = (const u_int16_t *)addr;
1961	const u_int8_t *d;
1962	int dcolon = 0, zero = 0;
1963
1964	cp = ip6buf;
1965
1966	for (i = 0; i < 8; i++) {
1967		if (*(a + i) == 0) {
1968			cnt++;
1969			if (cnt == 1)
1970				idx = i;
1971		}
1972		else if (maxcnt < cnt) {
1973			maxcnt = cnt;
1974			index = idx;
1975			cnt = 0;
1976		}
1977	}
1978	if (maxcnt < cnt) {
1979		maxcnt = cnt;
1980		index = idx;
1981	}
1982
1983	for (i = 0; i < 8; i++) {
1984		if (dcolon == 1) {
1985			if (*a == 0) {
1986				if (i == 7)
1987					*cp++ = ':';
1988				a++;
1989				continue;
1990			} else
1991				dcolon = 2;
1992		}
1993		if (*a == 0) {
1994			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1995				if (i == 0)
1996					*cp++ = ':';
1997				*cp++ = ':';
1998				dcolon = 1;
1999			} else {
2000				*cp++ = '0';
2001				*cp++ = ':';
2002			}
2003			a++;
2004			continue;
2005		}
2006		d = (const u_char *)a;
2007		/* Try to eliminate leading zeros in printout like in :0001. */
2008		zero = 1;
2009		*cp = digits[*d >> 4];
2010		if (*cp != '0') {
2011			zero = 0;
2012			cp++;
2013		}
2014		*cp = digits[*d++ & 0xf];
2015		if (zero == 0 || (*cp != '0')) {
2016			zero = 0;
2017			cp++;
2018		}
2019		*cp = digits[*d >> 4];
2020		if (zero == 0 || (*cp != '0')) {
2021			zero = 0;
2022			cp++;
2023		}
2024		*cp++ = digits[*d & 0xf];
2025		*cp++ = ':';
2026		a++;
2027	}
2028	*--cp = '\0';
2029	return (ip6buf);
2030}
2031
2032int
2033in6_localaddr(struct in6_addr *in6)
2034{
2035	struct in6_ifaddr *ia;
2036
2037	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2038		return 1;
2039
2040	IN6_IFADDR_RLOCK();
2041	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2042		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2043		    &ia->ia_prefixmask.sin6_addr)) {
2044			IN6_IFADDR_RUNLOCK();
2045			return 1;
2046		}
2047	}
2048	IN6_IFADDR_RUNLOCK();
2049
2050	return (0);
2051}
2052
2053/*
2054 * Return 1 if an internet address is for the local host and configured
2055 * on one of its interfaces.
2056 */
2057int
2058in6_localip(struct in6_addr *in6)
2059{
2060	struct in6_ifaddr *ia;
2061
2062	IN6_IFADDR_RLOCK();
2063	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2064		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2065			IN6_IFADDR_RUNLOCK();
2066			return (1);
2067		}
2068	}
2069	IN6_IFADDR_RUNLOCK();
2070	return (0);
2071}
2072
2073
2074int
2075in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2076{
2077	struct in6_ifaddr *ia;
2078
2079	IN6_IFADDR_RLOCK();
2080	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2081		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2082				       &sa6->sin6_addr) &&
2083		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
2084			IN6_IFADDR_RUNLOCK();
2085			return (1); /* true */
2086		}
2087
2088		/* XXX: do we still have to go thru the rest of the list? */
2089	}
2090	IN6_IFADDR_RUNLOCK();
2091
2092	return (0);		/* false */
2093}
2094
2095/*
2096 * return length of part which dst and src are equal
2097 * hard coding...
2098 */
2099int
2100in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2101{
2102	int match = 0;
2103	u_char *s = (u_char *)src, *d = (u_char *)dst;
2104	u_char *lim = s + 16, r;
2105
2106	while (s < lim)
2107		if ((r = (*d++ ^ *s++)) != 0) {
2108			while (r < 128) {
2109				match++;
2110				r <<= 1;
2111			}
2112			break;
2113		} else
2114			match += 8;
2115	return match;
2116}
2117
2118/* XXX: to be scope conscious */
2119int
2120in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2121{
2122	int bytelen, bitlen;
2123
2124	/* sanity check */
2125	if (0 > len || len > 128) {
2126		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2127		    len);
2128		return (0);
2129	}
2130
2131	bytelen = len / 8;
2132	bitlen = len % 8;
2133
2134	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2135		return (0);
2136	if (bitlen != 0 &&
2137	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2138	    p2->s6_addr[bytelen] >> (8 - bitlen))
2139		return (0);
2140
2141	return (1);
2142}
2143
2144void
2145in6_prefixlen2mask(struct in6_addr *maskp, int len)
2146{
2147	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2148	int bytelen, bitlen, i;
2149
2150	/* sanity check */
2151	if (0 > len || len > 128) {
2152		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2153		    len);
2154		return;
2155	}
2156
2157	bzero(maskp, sizeof(*maskp));
2158	bytelen = len / 8;
2159	bitlen = len % 8;
2160	for (i = 0; i < bytelen; i++)
2161		maskp->s6_addr[i] = 0xff;
2162	if (bitlen)
2163		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2164}
2165
2166/*
2167 * return the best address out of the same scope. if no address was
2168 * found, return the first valid address from designated IF.
2169 */
2170struct in6_ifaddr *
2171in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2172{
2173	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2174	struct ifaddr *ifa;
2175	struct in6_ifaddr *besta = 0;
2176	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2177
2178	dep[0] = dep[1] = NULL;
2179
2180	/*
2181	 * We first look for addresses in the same scope.
2182	 * If there is one, return it.
2183	 * If two or more, return one which matches the dst longest.
2184	 * If none, return one of global addresses assigned other ifs.
2185	 */
2186	IF_ADDR_RLOCK(ifp);
2187	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2188		if (ifa->ifa_addr->sa_family != AF_INET6)
2189			continue;
2190		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2191			continue; /* XXX: is there any case to allow anycast? */
2192		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2193			continue; /* don't use this interface */
2194		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2195			continue;
2196		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2197			if (V_ip6_use_deprecated)
2198				dep[0] = (struct in6_ifaddr *)ifa;
2199			continue;
2200		}
2201
2202		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2203			/*
2204			 * call in6_matchlen() as few as possible
2205			 */
2206			if (besta) {
2207				if (blen == -1)
2208					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2209				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2210				if (tlen > blen) {
2211					blen = tlen;
2212					besta = (struct in6_ifaddr *)ifa;
2213				}
2214			} else
2215				besta = (struct in6_ifaddr *)ifa;
2216		}
2217	}
2218	if (besta) {
2219		ifa_ref(&besta->ia_ifa);
2220		IF_ADDR_RUNLOCK(ifp);
2221		return (besta);
2222	}
2223
2224	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2225		if (ifa->ifa_addr->sa_family != AF_INET6)
2226			continue;
2227		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2228			continue; /* XXX: is there any case to allow anycast? */
2229		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2230			continue; /* don't use this interface */
2231		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2232			continue;
2233		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2234			if (V_ip6_use_deprecated)
2235				dep[1] = (struct in6_ifaddr *)ifa;
2236			continue;
2237		}
2238
2239		if (ifa != NULL)
2240			ifa_ref(ifa);
2241		IF_ADDR_RUNLOCK(ifp);
2242		return (struct in6_ifaddr *)ifa;
2243	}
2244
2245	/* use the last-resort values, that are, deprecated addresses */
2246	if (dep[0]) {
2247		ifa_ref((struct ifaddr *)dep[0]);
2248		IF_ADDR_RUNLOCK(ifp);
2249		return dep[0];
2250	}
2251	if (dep[1]) {
2252		ifa_ref((struct ifaddr *)dep[1]);
2253		IF_ADDR_RUNLOCK(ifp);
2254		return dep[1];
2255	}
2256
2257	IF_ADDR_RUNLOCK(ifp);
2258	return NULL;
2259}
2260
2261/*
2262 * perform DAD when interface becomes IFF_UP.
2263 */
2264void
2265in6_if_up(struct ifnet *ifp)
2266{
2267	struct ifaddr *ifa;
2268	struct in6_ifaddr *ia;
2269
2270	IF_ADDR_RLOCK(ifp);
2271	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2272		if (ifa->ifa_addr->sa_family != AF_INET6)
2273			continue;
2274		ia = (struct in6_ifaddr *)ifa;
2275		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2276			/*
2277			 * The TENTATIVE flag was likely set by hand
2278			 * beforehand, implicitly indicating the need for DAD.
2279			 * We may be able to skip the random delay in this
2280			 * case, but we impose delays just in case.
2281			 */
2282			nd6_dad_start(ifa,
2283			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2284		}
2285	}
2286	IF_ADDR_RUNLOCK(ifp);
2287
2288	/*
2289	 * special cases, like 6to4, are handled in in6_ifattach
2290	 */
2291	in6_ifattach(ifp, NULL);
2292}
2293
2294int
2295in6if_do_dad(struct ifnet *ifp)
2296{
2297	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2298		return (0);
2299
2300	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2301		return (0);
2302
2303	switch (ifp->if_type) {
2304#ifdef IFT_DUMMY
2305	case IFT_DUMMY:
2306#endif
2307	case IFT_FAITH:
2308		/*
2309		 * These interfaces do not have the IFF_LOOPBACK flag,
2310		 * but loop packets back.  We do not have to do DAD on such
2311		 * interfaces.  We should even omit it, because loop-backed
2312		 * NS would confuse the DAD procedure.
2313		 */
2314		return (0);
2315	default:
2316		/*
2317		 * Our DAD routine requires the interface up and running.
2318		 * However, some interfaces can be up before the RUNNING
2319		 * status.  Additionaly, users may try to assign addresses
2320		 * before the interface becomes up (or running).
2321		 * We simply skip DAD in such a case as a work around.
2322		 * XXX: we should rather mark "tentative" on such addresses,
2323		 * and do DAD after the interface becomes ready.
2324		 */
2325		if (!((ifp->if_flags & IFF_UP) &&
2326		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2327			return (0);
2328
2329		return (1);
2330	}
2331}
2332
2333/*
2334 * Calculate max IPv6 MTU through all the interfaces and store it
2335 * to in6_maxmtu.
2336 */
2337void
2338in6_setmaxmtu(void)
2339{
2340	unsigned long maxmtu = 0;
2341	struct ifnet *ifp;
2342
2343	IFNET_RLOCK_NOSLEEP();
2344	TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
2345		/* this function can be called during ifnet initialization */
2346		if (!ifp->if_afdata[AF_INET6])
2347			continue;
2348		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2349		    IN6_LINKMTU(ifp) > maxmtu)
2350			maxmtu = IN6_LINKMTU(ifp);
2351	}
2352	IFNET_RUNLOCK_NOSLEEP();
2353	if (maxmtu)	     /* update only when maxmtu is positive */
2354		V_in6_maxmtu = maxmtu;
2355}
2356
2357/*
2358 * Provide the length of interface identifiers to be used for the link attached
2359 * to the given interface.  The length should be defined in "IPv6 over
2360 * xxx-link" document.  Note that address architecture might also define
2361 * the length for a particular set of address prefixes, regardless of the
2362 * link type.  As clarified in rfc2462bis, those two definitions should be
2363 * consistent, and those really are as of August 2004.
2364 */
2365int
2366in6_if2idlen(struct ifnet *ifp)
2367{
2368	switch (ifp->if_type) {
2369	case IFT_ETHER:		/* RFC2464 */
2370#ifdef IFT_PROPVIRTUAL
2371	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2372#endif
2373#ifdef IFT_L2VLAN
2374	case IFT_L2VLAN:	/* ditto */
2375#endif
2376#ifdef IFT_IEEE80211
2377	case IFT_IEEE80211:	/* ditto */
2378#endif
2379#ifdef IFT_MIP
2380	case IFT_MIP:	/* ditto */
2381#endif
2382	case IFT_INFINIBAND:
2383		return (64);
2384	case IFT_FDDI:		/* RFC2467 */
2385		return (64);
2386	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2387		return (64);
2388	case IFT_PPP:		/* RFC2472 */
2389		return (64);
2390	case IFT_ARCNET:	/* RFC2497 */
2391		return (64);
2392	case IFT_FRELAY:	/* RFC2590 */
2393		return (64);
2394	case IFT_IEEE1394:	/* RFC3146 */
2395		return (64);
2396	case IFT_GIF:
2397		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2398	case IFT_LOOP:
2399		return (64);	/* XXX: is this really correct? */
2400	default:
2401		/*
2402		 * Unknown link type:
2403		 * It might be controversial to use the today's common constant
2404		 * of 64 for these cases unconditionally.  For full compliance,
2405		 * we should return an error in this case.  On the other hand,
2406		 * if we simply miss the standard for the link type or a new
2407		 * standard is defined for a new link type, the IFID length
2408		 * is very likely to be the common constant.  As a compromise,
2409		 * we always use the constant, but make an explicit notice
2410		 * indicating the "unknown" case.
2411		 */
2412		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2413		return (64);
2414	}
2415}
2416
2417#include <sys/sysctl.h>
2418
2419struct in6_llentry {
2420	struct llentry		base;
2421	struct sockaddr_in6	l3_addr6;
2422};
2423
2424static struct llentry *
2425in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2426{
2427	struct in6_llentry *lle;
2428
2429	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE,
2430	    M_DONTWAIT | M_ZERO);
2431	if (lle == NULL)		/* NB: caller generates msg */
2432		return NULL;
2433
2434	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2435	lle->base.lle_refcnt = 1;
2436	LLE_LOCK_INIT(&lle->base);
2437	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2438	    CALLOUT_RETURNUNLOCKED);
2439
2440	return &lle->base;
2441}
2442
2443/*
2444 * Deletes an address from the address table.
2445 * This function is called by the timer functions
2446 * such as arptimer() and nd6_llinfo_timer(), and
2447 * the caller does the locking.
2448 */
2449static void
2450in6_lltable_free(struct lltable *llt, struct llentry *lle)
2451{
2452	LLE_WUNLOCK(lle);
2453	LLE_LOCK_DESTROY(lle);
2454	free(lle, M_LLTABLE);
2455}
2456
2457static void
2458in6_lltable_prefix_free(struct lltable *llt,
2459			const struct sockaddr *prefix,
2460			const struct sockaddr *mask,
2461			u_int flags)
2462{
2463	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2464	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2465	struct llentry *lle, *next;
2466	register int i;
2467
2468	/*
2469	 * (flags & LLE_STATIC) means deleting all entries
2470	 * including static ND6 entries
2471	 */
2472	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
2473		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2474			if (IN6_ARE_MASKED_ADDR_EQUAL(
2475				    &((struct sockaddr_in6 *)L3_ADDR(lle))->sin6_addr,
2476				    &pfx->sin6_addr,
2477				    &msk->sin6_addr) &&
2478			    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
2479				int canceled;
2480
2481				canceled = callout_drain(&lle->la_timer);
2482				LLE_WLOCK(lle);
2483				if (canceled)
2484					LLE_REMREF(lle);
2485				llentry_free(lle);
2486			}
2487		}
2488	}
2489}
2490
2491static int
2492in6_lltable_rtcheck(struct ifnet *ifp,
2493		    u_int flags,
2494		    const struct sockaddr *l3addr)
2495{
2496	struct rtentry *rt;
2497	char ip6buf[INET6_ADDRSTRLEN];
2498
2499	KASSERT(l3addr->sa_family == AF_INET6,
2500	    ("sin_family %d", l3addr->sa_family));
2501
2502	/* Our local addresses are always only installed on the default FIB. */
2503	/* XXX rtalloc1 should take a const param */
2504	rt = in6_rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0,
2505	    RT_DEFAULT_FIB);
2506	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2507		struct ifaddr *ifa;
2508		/*
2509		 * Create an ND6 cache for an IPv6 neighbor
2510		 * that is not covered by our own prefix.
2511		 */
2512		/* XXX ifaof_ifpforaddr should take a const param */
2513		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2514		if (ifa != NULL) {
2515			ifa_free(ifa);
2516			if (rt != NULL)
2517				RTFREE_LOCKED(rt);
2518			return 0;
2519		}
2520		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2521		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2522		if (rt != NULL)
2523			RTFREE_LOCKED(rt);
2524		return EINVAL;
2525	}
2526	RTFREE_LOCKED(rt);
2527	return 0;
2528}
2529
2530static struct llentry *
2531in6_lltable_lookup(struct lltable *llt, u_int flags,
2532	const struct sockaddr *l3addr)
2533{
2534	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2535	struct ifnet *ifp = llt->llt_ifp;
2536	struct llentry *lle;
2537	struct llentries *lleh;
2538	u_int hashkey;
2539
2540	IF_AFDATA_LOCK_ASSERT(ifp);
2541	KASSERT(l3addr->sa_family == AF_INET6,
2542	    ("sin_family %d", l3addr->sa_family));
2543
2544	hashkey = sin6->sin6_addr.s6_addr32[3];
2545	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2546	LIST_FOREACH(lle, lleh, lle_next) {
2547		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2548		if (lle->la_flags & LLE_DELETED)
2549			continue;
2550		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2551			 sizeof(struct in6_addr)) == 0)
2552			break;
2553	}
2554
2555	if (lle == NULL) {
2556		if (!(flags & LLE_CREATE))
2557			return (NULL);
2558		/*
2559		 * A route that covers the given address must have
2560		 * been installed 1st because we are doing a resolution,
2561		 * verify this.
2562		 */
2563		if (!(flags & LLE_IFADDR) &&
2564		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2565			return NULL;
2566
2567		lle = in6_lltable_new(l3addr, flags);
2568		if (lle == NULL) {
2569			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2570			return NULL;
2571		}
2572		lle->la_flags = flags & ~LLE_CREATE;
2573		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2574			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2575			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2576		}
2577
2578		lle->lle_tbl  = llt;
2579		lle->lle_head = lleh;
2580		LIST_INSERT_HEAD(lleh, lle, lle_next);
2581	} else if (flags & LLE_DELETE) {
2582		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2583			LLE_WLOCK(lle);
2584			lle->la_flags = LLE_DELETED;
2585			LLE_WUNLOCK(lle);
2586#ifdef DIAGNOSTIC
2587			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2588#endif
2589		}
2590		lle = (void *)-1;
2591	}
2592	if (LLE_IS_VALID(lle)) {
2593		if (flags & LLE_EXCLUSIVE)
2594			LLE_WLOCK(lle);
2595		else
2596			LLE_RLOCK(lle);
2597	}
2598	return (lle);
2599}
2600
2601static int
2602in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2603{
2604	struct ifnet *ifp = llt->llt_ifp;
2605	struct llentry *lle;
2606	/* XXX stack use */
2607	struct {
2608		struct rt_msghdr	rtm;
2609		struct sockaddr_in6	sin6;
2610		/*
2611		 * ndp.c assumes that sdl is word aligned
2612		 */
2613#ifdef __LP64__
2614		uint32_t		pad;
2615#endif
2616		struct sockaddr_dl	sdl;
2617	} ndpc;
2618	int i, error;
2619
2620	if (ifp->if_flags & IFF_LOOPBACK)
2621		return 0;
2622
2623	LLTABLE_LOCK_ASSERT();
2624
2625	error = 0;
2626	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2627		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2628			struct sockaddr_dl *sdl;
2629
2630			/* skip deleted or invalid entries */
2631			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2632				continue;
2633			/* Skip if jailed and not a valid IP of the prison. */
2634			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2635				continue;
2636			/*
2637			 * produce a msg made of:
2638			 *  struct rt_msghdr;
2639			 *  struct sockaddr_in6 (IPv6)
2640			 *  struct sockaddr_dl;
2641			 */
2642			bzero(&ndpc, sizeof(ndpc));
2643			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2644			ndpc.rtm.rtm_version = RTM_VERSION;
2645			ndpc.rtm.rtm_type = RTM_GET;
2646			ndpc.rtm.rtm_flags = RTF_UP;
2647			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2648			ndpc.sin6.sin6_family = AF_INET6;
2649			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2650			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2651
2652			/* publish */
2653			if (lle->la_flags & LLE_PUB)
2654				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2655
2656			sdl = &ndpc.sdl;
2657			sdl->sdl_family = AF_LINK;
2658			sdl->sdl_len = sizeof(*sdl);
2659			sdl->sdl_alen = ifp->if_addrlen;
2660			sdl->sdl_index = ifp->if_index;
2661			sdl->sdl_type = ifp->if_type;
2662			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2663			ndpc.rtm.rtm_rmx.rmx_expire =
2664			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2665			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2666			if (lle->la_flags & LLE_STATIC)
2667				ndpc.rtm.rtm_flags |= RTF_STATIC;
2668			ndpc.rtm.rtm_index = ifp->if_index;
2669			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2670			if (error)
2671				break;
2672		}
2673	}
2674	return error;
2675}
2676
2677void *
2678in6_domifattach(struct ifnet *ifp)
2679{
2680	struct in6_ifextra *ext;
2681
2682	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2683	bzero(ext, sizeof(*ext));
2684
2685	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2686	    M_IFADDR, M_WAITOK);
2687	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2688
2689	ext->icmp6_ifstat =
2690	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2691	    M_IFADDR, M_WAITOK);
2692	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2693
2694	ext->nd_ifinfo = nd6_ifattach(ifp);
2695	ext->scope6_id = scope6_ifattach(ifp);
2696	ext->lltable = lltable_init(ifp, AF_INET6);
2697	if (ext->lltable != NULL) {
2698		ext->lltable->llt_free = in6_lltable_free;
2699		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2700		ext->lltable->llt_lookup = in6_lltable_lookup;
2701		ext->lltable->llt_dump = in6_lltable_dump;
2702	}
2703
2704	ext->mld_ifinfo = mld_domifattach(ifp);
2705
2706	return ext;
2707}
2708
2709void
2710in6_domifdetach(struct ifnet *ifp, void *aux)
2711{
2712	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2713
2714	mld_domifdetach(ifp);
2715	scope6_ifdetach(ext->scope6_id);
2716	nd6_ifdetach(ext->nd_ifinfo);
2717	lltable_free(ext->lltable);
2718	free(ext->in6_ifstat, M_IFADDR);
2719	free(ext->icmp6_ifstat, M_IFADDR);
2720	free(ext, M_IFADDR);
2721}
2722
2723/*
2724 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2725 * v4 mapped addr or v4 compat addr
2726 */
2727void
2728in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2729{
2730
2731	bzero(sin, sizeof(*sin));
2732	sin->sin_len = sizeof(struct sockaddr_in);
2733	sin->sin_family = AF_INET;
2734	sin->sin_port = sin6->sin6_port;
2735	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2736}
2737
2738/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2739void
2740in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2741{
2742	bzero(sin6, sizeof(*sin6));
2743	sin6->sin6_len = sizeof(struct sockaddr_in6);
2744	sin6->sin6_family = AF_INET6;
2745	sin6->sin6_port = sin->sin_port;
2746	sin6->sin6_addr.s6_addr32[0] = 0;
2747	sin6->sin6_addr.s6_addr32[1] = 0;
2748	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2749	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2750}
2751
2752/* Convert sockaddr_in6 into sockaddr_in. */
2753void
2754in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2755{
2756	struct sockaddr_in *sin_p;
2757	struct sockaddr_in6 sin6;
2758
2759	/*
2760	 * Save original sockaddr_in6 addr and convert it
2761	 * to sockaddr_in.
2762	 */
2763	sin6 = *(struct sockaddr_in6 *)nam;
2764	sin_p = (struct sockaddr_in *)nam;
2765	in6_sin6_2_sin(sin_p, &sin6);
2766}
2767
2768/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2769void
2770in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2771{
2772	struct sockaddr_in *sin_p;
2773	struct sockaddr_in6 *sin6_p;
2774
2775	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
2776	       M_WAITOK);
2777	sin_p = (struct sockaddr_in *)*nam;
2778	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2779	free(*nam, M_SONAME);
2780	*nam = (struct sockaddr *)sin6_p;
2781}
2782