in6.c revision 303458
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/10/sys/netinet6/in6.c 303458 2016-07-28 20:08:01Z sbruno $");
65
66#include "opt_compat.h"
67#include "opt_inet.h"
68#include "opt_inet6.h"
69
70#include <sys/param.h>
71#include <sys/errno.h>
72#include <sys/jail.h>
73#include <sys/malloc.h>
74#include <sys/socket.h>
75#include <sys/socketvar.h>
76#include <sys/sockio.h>
77#include <sys/systm.h>
78#include <sys/priv.h>
79#include <sys/proc.h>
80#include <sys/time.h>
81#include <sys/kernel.h>
82#include <sys/syslog.h>
83
84#include <net/if.h>
85#include <net/if_var.h>
86#include <net/if_types.h>
87#include <net/route.h>
88#include <net/if_dl.h>
89#include <net/vnet.h>
90
91#include <netinet/in.h>
92#include <netinet/in_var.h>
93#include <net/if_llatbl.h>
94#include <netinet/if_ether.h>
95#include <netinet/in_systm.h>
96#include <netinet/ip.h>
97#include <netinet/in_pcb.h>
98#include <netinet/ip_carp.h>
99
100#include <netinet/ip6.h>
101#include <netinet6/ip6_var.h>
102#include <netinet6/nd6.h>
103#include <netinet6/mld6_var.h>
104#include <netinet6/ip6_mroute.h>
105#include <netinet6/in6_ifattach.h>
106#include <netinet6/scope6_var.h>
107#include <netinet6/in6_pcb.h>
108
109VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
110#define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
111
112/*
113 * Definitions of some costant IP6 addresses.
114 */
115const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
116const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
117const struct in6_addr in6addr_nodelocal_allnodes =
118	IN6ADDR_NODELOCAL_ALLNODES_INIT;
119const struct in6_addr in6addr_linklocal_allnodes =
120	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
121const struct in6_addr in6addr_linklocal_allrouters =
122	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
123const struct in6_addr in6addr_linklocal_allv2routers =
124	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
125
126const struct in6_addr in6mask0 = IN6MASK0;
127const struct in6_addr in6mask32 = IN6MASK32;
128const struct in6_addr in6mask64 = IN6MASK64;
129const struct in6_addr in6mask96 = IN6MASK96;
130const struct in6_addr in6mask128 = IN6MASK128;
131
132const struct sockaddr_in6 sa6_any =
133	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
134
135static int in6_lifaddr_ioctl(struct socket *, u_long, caddr_t,
136	struct ifnet *, struct thread *);
137static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
138	struct sockaddr_in6 *, int);
139static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
140
141int	(*faithprefix_p)(struct in6_addr *);
142
143#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
144#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
145
146void
147in6_ifaddloop(struct ifaddr *ifa)
148{
149	struct sockaddr_dl gateway;
150	struct sockaddr_in6 mask, addr;
151	struct rtentry rt;
152	struct in6_ifaddr *ia;
153	struct ifnet *ifp;
154	struct llentry *ln;
155
156	ia = ifa2ia6(ifa);
157	ifp = ifa->ifa_ifp;
158	/*
159	 * initialize for rtmsg generation
160	 */
161	bzero(&gateway, sizeof(gateway));
162	gateway.sdl_len = sizeof(gateway);
163	gateway.sdl_family = AF_LINK;
164	if (nd6_need_cache(ifp) != 0) {
165		IF_AFDATA_LOCK(ifp);
166		ifa->ifa_rtrequest = nd6_rtrequest;
167		ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
168		    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
169		IF_AFDATA_UNLOCK(ifp);
170		if (ln != NULL) {
171			ln->la_expire = 0;  /* for IPv6 this means permanent */
172			ln->ln_state = ND6_LLINFO_REACHABLE;
173
174			gateway.sdl_alen = 6;
175			memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
176			    sizeof(ln->ll_addr));
177			LLE_WUNLOCK(ln);
178		}
179	}
180	bzero(&rt, sizeof(rt));
181	rt.rt_gateway = (struct sockaddr *)&gateway;
182	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
183	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
184	rt_mask(&rt) = (struct sockaddr *)&mask;
185	rt_key(&rt) = (struct sockaddr *)&addr;
186	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
187	/* Announce arrival of local address to all FIBs. */
188	rt_newaddrmsg(RTM_ADD, ifa, 0, &rt);
189}
190
191void
192in6_ifremloop(struct ifaddr *ifa)
193{
194	struct sockaddr_dl gateway;
195	struct sockaddr_in6 mask, addr;
196	struct rtentry rt0;
197	struct in6_ifaddr *ia;
198	struct ifnet *ifp;
199
200	ia = ifa2ia6(ifa);
201	ifp = ifa->ifa_ifp;
202	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
203	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
204	lltable_prefix_free(AF_INET6, (struct sockaddr *)&addr,
205	            (struct sockaddr *)&mask, LLE_STATIC);
206
207	/*
208	 * initialize for rtmsg generation
209	 */
210	bzero(&gateway, sizeof(gateway));
211	gateway.sdl_len = sizeof(gateway);
212	gateway.sdl_family = AF_LINK;
213	gateway.sdl_nlen = 0;
214	gateway.sdl_alen = ifp->if_addrlen;
215	bzero(&rt0, sizeof(rt0));
216	rt0.rt_gateway = (struct sockaddr *)&gateway;
217	rt_mask(&rt0) = (struct sockaddr *)&mask;
218	rt_key(&rt0) = (struct sockaddr *)&addr;
219	rt0.rt_flags = RTF_HOST | RTF_STATIC;
220	/* Announce removal of local address to all FIBs. */
221	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
222}
223
224int
225in6_mask2len(struct in6_addr *mask, u_char *lim0)
226{
227	int x = 0, y;
228	u_char *lim = lim0, *p;
229
230	/* ignore the scope_id part */
231	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
232		lim = (u_char *)mask + sizeof(*mask);
233	for (p = (u_char *)mask; p < lim; x++, p++) {
234		if (*p != 0xff)
235			break;
236	}
237	y = 0;
238	if (p < lim) {
239		for (y = 0; y < 8; y++) {
240			if ((*p & (0x80 >> y)) == 0)
241				break;
242		}
243	}
244
245	/*
246	 * when the limit pointer is given, do a stricter check on the
247	 * remaining bits.
248	 */
249	if (p < lim) {
250		if (y != 0 && (*p & (0x00ff >> y)) != 0)
251			return (-1);
252		for (p = p + 1; p < lim; p++)
253			if (*p != 0)
254				return (-1);
255	}
256
257	return x * 8 + y;
258}
259
260#ifdef COMPAT_FREEBSD32
261struct in6_ndifreq32 {
262	char ifname[IFNAMSIZ];
263	uint32_t ifindex;
264};
265#define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
266#endif
267
268int
269in6_control(struct socket *so, u_long cmd, caddr_t data,
270    struct ifnet *ifp, struct thread *td)
271{
272	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
273	struct	in6_ifaddr *ia = NULL;
274	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
275	struct sockaddr_in6 *sa6;
276	int carp_attached = 0;
277	int error;
278	u_long ocmd = cmd;
279
280	/*
281	 * Compat to make pre-10.x ifconfig(8) operable.
282	 */
283	if (cmd == OSIOCAIFADDR_IN6)
284		cmd = SIOCAIFADDR_IN6;
285
286	switch (cmd) {
287	case SIOCGETSGCNT_IN6:
288	case SIOCGETMIFCNT_IN6:
289		/*
290		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
291		 * We cannot see how that would be needed, so do not adjust the
292		 * KPI blindly; more likely should clean up the IPv4 variant.
293		 */
294		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
295	}
296
297	switch (cmd) {
298	case SIOCAADDRCTL_POLICY:
299	case SIOCDADDRCTL_POLICY:
300		if (td != NULL) {
301			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
302			if (error)
303				return (error);
304		}
305		return (in6_src_ioctl(cmd, data));
306	}
307
308	if (ifp == NULL)
309		return (EOPNOTSUPP);
310
311	switch (cmd) {
312	case SIOCSNDFLUSH_IN6:
313	case SIOCSPFXFLUSH_IN6:
314	case SIOCSRTRFLUSH_IN6:
315	case SIOCSDEFIFACE_IN6:
316	case SIOCSIFINFO_FLAGS:
317	case SIOCSIFINFO_IN6:
318		if (td != NULL) {
319			error = priv_check(td, PRIV_NETINET_ND6);
320			if (error)
321				return (error);
322		}
323		/* FALLTHROUGH */
324	case OSIOCGIFINFO_IN6:
325	case SIOCGIFINFO_IN6:
326	case SIOCGNBRINFO_IN6:
327	case SIOCGDEFIFACE_IN6:
328		return (nd6_ioctl(cmd, data, ifp));
329
330#ifdef COMPAT_FREEBSD32
331	case SIOCGDEFIFACE32_IN6:
332		{
333			struct in6_ndifreq ndif;
334			struct in6_ndifreq32 *ndif32;
335
336			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
337			    ifp);
338			if (error)
339				return (error);
340			ndif32 = (struct in6_ndifreq32 *)data;
341			ndif32->ifindex = ndif.ifindex;
342			return (0);
343		}
344#endif
345	}
346
347	switch (cmd) {
348	case SIOCSIFPREFIX_IN6:
349	case SIOCDIFPREFIX_IN6:
350	case SIOCAIFPREFIX_IN6:
351	case SIOCCIFPREFIX_IN6:
352	case SIOCSGIFPREFIX_IN6:
353	case SIOCGIFPREFIX_IN6:
354		log(LOG_NOTICE,
355		    "prefix ioctls are now invalidated. "
356		    "please use ifconfig.\n");
357		return (EOPNOTSUPP);
358	}
359
360	switch (cmd) {
361	case SIOCSSCOPE6:
362		if (td != NULL) {
363			error = priv_check(td, PRIV_NETINET_SCOPE6);
364			if (error)
365				return (error);
366		}
367		/* FALLTHROUGH */
368	case SIOCGSCOPE6:
369	case SIOCGSCOPE6DEF:
370		return (scope6_ioctl(cmd, data, ifp));
371	}
372
373	switch (cmd) {
374	case SIOCALIFADDR:
375		if (td != NULL) {
376			error = priv_check(td, PRIV_NET_ADDIFADDR);
377			if (error)
378				return (error);
379		}
380		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
381
382	case SIOCDLIFADDR:
383		if (td != NULL) {
384			error = priv_check(td, PRIV_NET_DELIFADDR);
385			if (error)
386				return (error);
387		}
388		/* FALLTHROUGH */
389	case SIOCGLIFADDR:
390		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
391	}
392
393	/*
394	 * Find address for this interface, if it exists.
395	 *
396	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
397	 * only, and used the first interface address as the target of other
398	 * operations (without checking ifra_addr).  This was because netinet
399	 * code/API assumed at most 1 interface address per interface.
400	 * Since IPv6 allows a node to assign multiple addresses
401	 * on a single interface, we almost always look and check the
402	 * presence of ifra_addr, and reject invalid ones here.
403	 * It also decreases duplicated code among SIOC*_IN6 operations.
404	 */
405	switch (cmd) {
406	case SIOCAIFADDR_IN6:
407	case SIOCSIFPHYADDR_IN6:
408		sa6 = &ifra->ifra_addr;
409		break;
410	case SIOCSIFADDR_IN6:
411	case SIOCGIFADDR_IN6:
412	case SIOCSIFDSTADDR_IN6:
413	case SIOCSIFNETMASK_IN6:
414	case SIOCGIFDSTADDR_IN6:
415	case SIOCGIFNETMASK_IN6:
416	case SIOCDIFADDR_IN6:
417	case SIOCGIFPSRCADDR_IN6:
418	case SIOCGIFPDSTADDR_IN6:
419	case SIOCGIFAFLAG_IN6:
420	case SIOCSNDFLUSH_IN6:
421	case SIOCSPFXFLUSH_IN6:
422	case SIOCSRTRFLUSH_IN6:
423	case SIOCGIFALIFETIME_IN6:
424	case SIOCSIFALIFETIME_IN6:
425	case SIOCGIFSTAT_IN6:
426	case SIOCGIFSTAT_ICMP6:
427		sa6 = &ifr->ifr_addr;
428		break;
429	case SIOCSIFADDR:
430	case SIOCSIFBRDADDR:
431	case SIOCSIFDSTADDR:
432	case SIOCSIFNETMASK:
433		/*
434		 * Although we should pass any non-INET6 ioctl requests
435		 * down to driver, we filter some legacy INET requests.
436		 * Drivers trust SIOCSIFADDR et al to come from an already
437		 * privileged layer, and do not perform any credentials
438		 * checks or input validation.
439		 */
440		return (EINVAL);
441	default:
442		sa6 = NULL;
443		break;
444	}
445	if (sa6 && sa6->sin6_family == AF_INET6) {
446		if (sa6->sin6_scope_id != 0)
447			error = sa6_embedscope(sa6, 0);
448		else
449			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
450		if (error != 0)
451			return (error);
452		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
453		    &sa6->sin6_addr)) != 0)
454			return (error);
455		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
456	} else
457		ia = NULL;
458
459	switch (cmd) {
460	case SIOCSIFADDR_IN6:
461	case SIOCSIFDSTADDR_IN6:
462	case SIOCSIFNETMASK_IN6:
463		/*
464		 * Since IPv6 allows a node to assign multiple addresses
465		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
466		 */
467		/* we decided to obsolete this command (20000704) */
468		error = EINVAL;
469		goto out;
470
471	case SIOCDIFADDR_IN6:
472		/*
473		 * for IPv4, we look for existing in_ifaddr here to allow
474		 * "ifconfig if0 delete" to remove the first IPv4 address on
475		 * the interface.  For IPv6, as the spec allows multiple
476		 * interface address from the day one, we consider "remove the
477		 * first one" semantics to be not preferable.
478		 */
479		if (ia == NULL) {
480			error = EADDRNOTAVAIL;
481			goto out;
482		}
483		/* FALLTHROUGH */
484	case SIOCAIFADDR_IN6:
485		/*
486		 * We always require users to specify a valid IPv6 address for
487		 * the corresponding operation.
488		 */
489		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
490		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
491			error = EAFNOSUPPORT;
492			goto out;
493		}
494
495		if (td != NULL) {
496			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
497			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
498			if (error)
499				goto out;
500		}
501		/* FALLTHROUGH */
502	case SIOCGIFSTAT_IN6:
503	case SIOCGIFSTAT_ICMP6:
504		if (ifp->if_afdata[AF_INET6] == NULL) {
505			error = EPFNOSUPPORT;
506			goto out;
507		}
508		break;
509
510	case SIOCGIFADDR_IN6:
511		/* This interface is basically deprecated. use SIOCGIFCONF. */
512		/* FALLTHROUGH */
513	case SIOCGIFAFLAG_IN6:
514	case SIOCGIFNETMASK_IN6:
515	case SIOCGIFDSTADDR_IN6:
516	case SIOCGIFALIFETIME_IN6:
517		/* must think again about its semantics */
518		if (ia == NULL) {
519			error = EADDRNOTAVAIL;
520			goto out;
521		}
522		break;
523
524	case SIOCSIFALIFETIME_IN6:
525	    {
526		struct in6_addrlifetime *lt;
527
528		if (td != NULL) {
529			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
530			if (error)
531				goto out;
532		}
533		if (ia == NULL) {
534			error = EADDRNOTAVAIL;
535			goto out;
536		}
537		/* sanity for overflow - beware unsigned */
538		lt = &ifr->ifr_ifru.ifru_lifetime;
539		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
540		    lt->ia6t_vltime + time_uptime < time_uptime) {
541			error = EINVAL;
542			goto out;
543		}
544		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
545		    lt->ia6t_pltime + time_uptime < time_uptime) {
546			error = EINVAL;
547			goto out;
548		}
549		break;
550	    }
551	}
552
553	switch (cmd) {
554	case SIOCGIFADDR_IN6:
555		ifr->ifr_addr = ia->ia_addr;
556		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
557			goto out;
558		break;
559
560	case SIOCGIFDSTADDR_IN6:
561		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
562			error = EINVAL;
563			goto out;
564		}
565		/*
566		 * XXX: should we check if ifa_dstaddr is NULL and return
567		 * an error?
568		 */
569		ifr->ifr_dstaddr = ia->ia_dstaddr;
570		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
571			goto out;
572		break;
573
574	case SIOCGIFNETMASK_IN6:
575		ifr->ifr_addr = ia->ia_prefixmask;
576		break;
577
578	case SIOCGIFAFLAG_IN6:
579		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
580		break;
581
582	case SIOCGIFSTAT_IN6:
583		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
584		    ifp->if_afdata[AF_INET6])->in6_ifstat,
585		    &ifr->ifr_ifru.ifru_stat,
586		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
587		break;
588
589	case SIOCGIFSTAT_ICMP6:
590		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
591		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
592		    &ifr->ifr_ifru.ifru_icmp6stat,
593		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
594		break;
595
596	case SIOCGIFALIFETIME_IN6:
597		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
598		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
599			time_t maxexpire;
600			struct in6_addrlifetime *retlt =
601			    &ifr->ifr_ifru.ifru_lifetime;
602
603			/*
604			 * XXX: adjust expiration time assuming time_t is
605			 * signed.
606			 */
607			maxexpire = (-1) &
608			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
609			if (ia->ia6_lifetime.ia6t_vltime <
610			    maxexpire - ia->ia6_updatetime) {
611				retlt->ia6t_expire = ia->ia6_updatetime +
612				    ia->ia6_lifetime.ia6t_vltime;
613			} else
614				retlt->ia6t_expire = maxexpire;
615		}
616		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
617			time_t maxexpire;
618			struct in6_addrlifetime *retlt =
619			    &ifr->ifr_ifru.ifru_lifetime;
620
621			/*
622			 * XXX: adjust expiration time assuming time_t is
623			 * signed.
624			 */
625			maxexpire = (-1) &
626			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
627			if (ia->ia6_lifetime.ia6t_pltime <
628			    maxexpire - ia->ia6_updatetime) {
629				retlt->ia6t_preferred = ia->ia6_updatetime +
630				    ia->ia6_lifetime.ia6t_pltime;
631			} else
632				retlt->ia6t_preferred = maxexpire;
633		}
634		break;
635
636	case SIOCSIFALIFETIME_IN6:
637		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
638		/* for sanity */
639		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
640			ia->ia6_lifetime.ia6t_expire =
641				time_uptime + ia->ia6_lifetime.ia6t_vltime;
642		} else
643			ia->ia6_lifetime.ia6t_expire = 0;
644		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
645			ia->ia6_lifetime.ia6t_preferred =
646				time_uptime + ia->ia6_lifetime.ia6t_pltime;
647		} else
648			ia->ia6_lifetime.ia6t_preferred = 0;
649		break;
650
651	case SIOCAIFADDR_IN6:
652	{
653		int i;
654		struct nd_prefixctl pr0;
655		struct nd_prefix *pr;
656
657		/*
658		 * first, make or update the interface address structure,
659		 * and link it to the list.
660		 */
661		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
662			goto out;
663		if (ia != NULL)
664			ifa_free(&ia->ia_ifa);
665		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
666		    == NULL) {
667			/*
668			 * this can happen when the user specify the 0 valid
669			 * lifetime.
670			 */
671			break;
672		}
673
674		if (cmd == ocmd && ifra->ifra_vhid > 0) {
675			if (carp_attach_p != NULL)
676				error = (*carp_attach_p)(&ia->ia_ifa,
677				    ifra->ifra_vhid);
678			else
679				error = EPROTONOSUPPORT;
680			if (error)
681				goto out;
682			else
683				carp_attached = 1;
684		}
685
686		/*
687		 * then, make the prefix on-link on the interface.
688		 * XXX: we'd rather create the prefix before the address, but
689		 * we need at least one address to install the corresponding
690		 * interface route, so we configure the address first.
691		 */
692
693		/*
694		 * convert mask to prefix length (prefixmask has already
695		 * been validated in in6_update_ifa().
696		 */
697		bzero(&pr0, sizeof(pr0));
698		pr0.ndpr_ifp = ifp;
699		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
700		    NULL);
701		if (pr0.ndpr_plen == 128) {
702			/* we don't need to install a host route. */
703			goto aifaddr_out;
704		}
705		pr0.ndpr_prefix = ifra->ifra_addr;
706		/* apply the mask for safety. */
707		for (i = 0; i < 4; i++) {
708			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
709			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
710		}
711		/*
712		 * XXX: since we don't have an API to set prefix (not address)
713		 * lifetimes, we just use the same lifetimes as addresses.
714		 * The (temporarily) installed lifetimes can be overridden by
715		 * later advertised RAs (when accept_rtadv is non 0), which is
716		 * an intended behavior.
717		 */
718		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
719		pr0.ndpr_raf_auto =
720		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
721		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
722		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
723
724		/* add the prefix if not yet. */
725		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
726			/*
727			 * nd6_prelist_add will install the corresponding
728			 * interface route.
729			 */
730			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
731				if (carp_attached)
732					(*carp_detach_p)(&ia->ia_ifa);
733				goto out;
734			}
735		}
736
737		/* relate the address to the prefix */
738		if (ia->ia6_ndpr == NULL) {
739			ia->ia6_ndpr = pr;
740			pr->ndpr_refcnt++;
741
742			/*
743			 * If this is the first autoconf address from the
744			 * prefix, create a temporary address as well
745			 * (when required).
746			 */
747			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
748			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
749				int e;
750				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
751					log(LOG_NOTICE, "in6_control: failed "
752					    "to create a temporary address, "
753					    "errno=%d\n", e);
754				}
755			}
756		}
757
758		/*
759		 * this might affect the status of autoconfigured addresses,
760		 * that is, this address might make other addresses detached.
761		 */
762		pfxlist_onlink_check();
763aifaddr_out:
764		if (error != 0 || ia == NULL)
765			break;
766		/*
767		 * Try to clear the flag when a new IPv6 address is added
768		 * onto an IFDISABLED interface and it succeeds.
769		 */
770		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
771			struct in6_ndireq nd;
772
773			memset(&nd, 0, sizeof(nd));
774			nd.ndi.flags = ND_IFINFO(ifp)->flags;
775			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
776			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
777				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
778				    "SIOCSIFINFO_FLAGS for -ifdisabled "
779				    "failed.");
780			/*
781			 * Ignore failure of clearing the flag intentionally.
782			 * The failure means address duplication was detected.
783			 */
784		}
785		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
786		break;
787	}
788
789	case SIOCDIFADDR_IN6:
790	{
791		struct nd_prefix *pr;
792
793		/*
794		 * If the address being deleted is the only one that owns
795		 * the corresponding prefix, expire the prefix as well.
796		 * XXX: theoretically, we don't have to worry about such
797		 * relationship, since we separate the address management
798		 * and the prefix management.  We do this, however, to provide
799		 * as much backward compatibility as possible in terms of
800		 * the ioctl operation.
801		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
802		 */
803		pr = ia->ia6_ndpr;
804		in6_purgeaddr(&ia->ia_ifa);
805		if (pr && pr->ndpr_refcnt == 0)
806			prelist_remove(pr);
807		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
808		break;
809	}
810
811	default:
812		if (ifp->if_ioctl == NULL) {
813			error = EOPNOTSUPP;
814			goto out;
815		}
816		error = (*ifp->if_ioctl)(ifp, cmd, data);
817		goto out;
818	}
819
820	error = 0;
821out:
822	if (ia != NULL)
823		ifa_free(&ia->ia_ifa);
824	return (error);
825}
826
827
828/*
829 * Join necessary multicast groups.  Factored out from in6_update_ifa().
830 * This entire work should only be done once, for the default FIB.
831 */
832static int
833in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
834    struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
835{
836	char ip6buf[INET6_ADDRSTRLEN];
837	struct sockaddr_in6 mltaddr, mltmask;
838	struct in6_addr llsol;
839	struct in6_multi_mship *imm;
840	struct rtentry *rt;
841	int delay, error;
842
843	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
844
845	/* Join solicited multicast addr for new host id. */
846	bzero(&llsol, sizeof(struct in6_addr));
847	llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
848	llsol.s6_addr32[1] = 0;
849	llsol.s6_addr32[2] = htonl(1);
850	llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
851	llsol.s6_addr8[12] = 0xff;
852	if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
853		/* XXX: should not happen */
854		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
855		goto cleanup;
856	}
857	delay = 0;
858	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
859		/*
860		 * We need a random delay for DAD on the address being
861		 * configured.  It also means delaying transmission of the
862		 * corresponding MLD report to avoid report collision.
863		 * [RFC 4861, Section 6.3.7]
864		 */
865		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
866	}
867	imm = in6_joingroup(ifp, &llsol, &error, delay);
868	if (imm == NULL) {
869		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
870		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &llsol),
871		    if_name(ifp), error));
872		goto cleanup;
873	}
874	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
875	*in6m_sol = imm->i6mm_maddr;
876
877	bzero(&mltmask, sizeof(mltmask));
878	mltmask.sin6_len = sizeof(struct sockaddr_in6);
879	mltmask.sin6_family = AF_INET6;
880	mltmask.sin6_addr = in6mask32;
881#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
882
883	/*
884	 * Join link-local all-nodes address.
885	 */
886	bzero(&mltaddr, sizeof(mltaddr));
887	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
888	mltaddr.sin6_family = AF_INET6;
889	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
890	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
891		goto cleanup; /* XXX: should not fail */
892
893	/*
894	 * XXX: do we really need this automatic routes?  We should probably
895	 * reconsider this stuff.  Most applications actually do not need the
896	 * routes, since they usually specify the outgoing interface.
897	 */
898	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
899	if (rt != NULL) {
900		/* XXX: only works in !SCOPEDROUTING case. */
901		if (memcmp(&mltaddr.sin6_addr,
902		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
903		    MLTMASK_LEN)) {
904			RTFREE_LOCKED(rt);
905			rt = NULL;
906		}
907	}
908	if (rt == NULL) {
909		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
910		    (struct sockaddr *)&ia->ia_addr,
911		    (struct sockaddr *)&mltmask, RTF_UP,
912		    (struct rtentry **)0, RT_DEFAULT_FIB);
913		if (error)
914			goto cleanup;
915	} else
916		RTFREE_LOCKED(rt);
917
918	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
919	if (imm == NULL) {
920		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
921		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
922		    &mltaddr.sin6_addr), if_name(ifp), error));
923		goto cleanup;
924	}
925	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
926
927	/*
928	 * Join node information group address.
929	 */
930	delay = 0;
931	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
932		/*
933		 * The spec does not say anything about delay for this group,
934		 * but the same logic should apply.
935		 */
936		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
937	}
938	if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
939		/* XXX jinmei */
940		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
941		if (imm == NULL)
942			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
943			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
944			    &mltaddr.sin6_addr), if_name(ifp), error));
945			/* XXX not very fatal, go on... */
946		else
947			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
948	}
949	if (V_icmp6_nodeinfo_oldmcprefix &&
950	     in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
951		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
952		if (imm == NULL)
953			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
954			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
955			    &mltaddr.sin6_addr), if_name(ifp), error));
956			/* XXX not very fatal, go on... */
957		else
958			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
959	}
960
961	/*
962	 * Join interface-local all-nodes address.
963	 * (ff01::1%ifN, and ff01::%ifN/32)
964	 */
965	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
966	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
967		goto cleanup; /* XXX: should not fail */
968	/* XXX: again, do we really need the route? */
969	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
970	if (rt != NULL) {
971		if (memcmp(&mltaddr.sin6_addr,
972		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
973		    MLTMASK_LEN)) {
974			RTFREE_LOCKED(rt);
975			rt = NULL;
976		}
977	}
978	if (rt == NULL) {
979		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
980		    (struct sockaddr *)&ia->ia_addr,
981		    (struct sockaddr *)&mltmask, RTF_UP,
982		    (struct rtentry **)0, RT_DEFAULT_FIB);
983		if (error)
984			goto cleanup;
985	} else
986		RTFREE_LOCKED(rt);
987
988	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
989	if (imm == NULL) {
990		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
991		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
992		    &mltaddr.sin6_addr), if_name(ifp), error));
993		goto cleanup;
994	}
995	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
996#undef	MLTMASK_LEN
997
998cleanup:
999	return (error);
1000}
1001
1002/*
1003 * Update parameters of an IPv6 interface address.
1004 * If necessary, a new entry is created and linked into address chains.
1005 * This function is separated from in6_control().
1006 */
1007int
1008in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1009    struct in6_ifaddr *ia, int flags)
1010{
1011	int error = 0, hostIsNew = 0, plen = -1;
1012	struct sockaddr_in6 dst6;
1013	struct in6_addrlifetime *lt;
1014	struct in6_multi *in6m_sol;
1015	int delay;
1016	char ip6buf[INET6_ADDRSTRLEN];
1017
1018	/* Validate parameters */
1019	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1020		return (EINVAL);
1021
1022	/*
1023	 * The destination address for a p2p link must have a family
1024	 * of AF_UNSPEC or AF_INET6.
1025	 */
1026	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1027	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1028	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1029		return (EAFNOSUPPORT);
1030	/*
1031	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1032	 * does not carry fields other than sin6_len.
1033	 */
1034	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1035		return (EINVAL);
1036	/*
1037	 * Because the IPv6 address architecture is classless, we require
1038	 * users to specify a (non 0) prefix length (mask) for a new address.
1039	 * We also require the prefix (when specified) mask is valid, and thus
1040	 * reject a non-consecutive mask.
1041	 */
1042	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1043		return (EINVAL);
1044	if (ifra->ifra_prefixmask.sin6_len != 0) {
1045		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1046		    (u_char *)&ifra->ifra_prefixmask +
1047		    ifra->ifra_prefixmask.sin6_len);
1048		if (plen <= 0)
1049			return (EINVAL);
1050	} else {
1051		/*
1052		 * In this case, ia must not be NULL.  We just use its prefix
1053		 * length.
1054		 */
1055		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1056	}
1057	/*
1058	 * If the destination address on a p2p interface is specified,
1059	 * and the address is a scoped one, validate/set the scope
1060	 * zone identifier.
1061	 */
1062	dst6 = ifra->ifra_dstaddr;
1063	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1064	    (dst6.sin6_family == AF_INET6)) {
1065		struct in6_addr in6_tmp;
1066		u_int32_t zoneid;
1067
1068		in6_tmp = dst6.sin6_addr;
1069		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1070			return (EINVAL); /* XXX: should be impossible */
1071
1072		if (dst6.sin6_scope_id != 0) {
1073			if (dst6.sin6_scope_id != zoneid)
1074				return (EINVAL);
1075		} else		/* user omit to specify the ID. */
1076			dst6.sin6_scope_id = zoneid;
1077
1078		/* convert into the internal form */
1079		if (sa6_embedscope(&dst6, 0))
1080			return (EINVAL); /* XXX: should be impossible */
1081	}
1082	/*
1083	 * The destination address can be specified only for a p2p or a
1084	 * loopback interface.  If specified, the corresponding prefix length
1085	 * must be 128.
1086	 */
1087	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1088		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1089			/* XXX: noisy message */
1090			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1091			    "be specified for a p2p or a loopback IF only\n"));
1092			return (EINVAL);
1093		}
1094		if (plen != 128) {
1095			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1096			    "be 128 when dstaddr is specified\n"));
1097			return (EINVAL);
1098		}
1099	}
1100	/* lifetime consistency check */
1101	lt = &ifra->ifra_lifetime;
1102	if (lt->ia6t_pltime > lt->ia6t_vltime)
1103		return (EINVAL);
1104	if (lt->ia6t_vltime == 0) {
1105		/*
1106		 * the following log might be noisy, but this is a typical
1107		 * configuration mistake or a tool's bug.
1108		 */
1109		nd6log((LOG_INFO,
1110		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1111		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1112
1113		if (ia == NULL)
1114			return (0); /* there's nothing to do */
1115	}
1116
1117	/*
1118	 * If this is a new address, allocate a new ifaddr and link it
1119	 * into chains.
1120	 */
1121	if (ia == NULL) {
1122		hostIsNew = 1;
1123		/*
1124		 * When in6_update_ifa() is called in a process of a received
1125		 * RA, it is called under an interrupt context.  So, we should
1126		 * call malloc with M_NOWAIT.
1127		 */
1128		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
1129		    M_NOWAIT);
1130		if (ia == NULL)
1131			return (ENOBUFS);
1132		bzero((caddr_t)ia, sizeof(*ia));
1133		ifa_init(&ia->ia_ifa);
1134		LIST_INIT(&ia->ia6_memberships);
1135		/* Initialize the address and masks, and put time stamp */
1136		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1137		ia->ia_addr.sin6_family = AF_INET6;
1138		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1139		ia->ia6_createtime = time_uptime;
1140		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1141			/*
1142			 * XXX: some functions expect that ifa_dstaddr is not
1143			 * NULL for p2p interfaces.
1144			 */
1145			ia->ia_ifa.ifa_dstaddr =
1146			    (struct sockaddr *)&ia->ia_dstaddr;
1147		} else {
1148			ia->ia_ifa.ifa_dstaddr = NULL;
1149		}
1150		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1151		ia->ia_ifp = ifp;
1152		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1153		IF_ADDR_WLOCK(ifp);
1154		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1155		IF_ADDR_WUNLOCK(ifp);
1156
1157		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1158		IN6_IFADDR_WLOCK();
1159		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1160		LIST_INSERT_HEAD(IN6ADDR_HASH(&ifra->ifra_addr.sin6_addr),
1161		    ia, ia6_hash);
1162		IN6_IFADDR_WUNLOCK();
1163	}
1164
1165	/* update timestamp */
1166	ia->ia6_updatetime = time_uptime;
1167
1168	/* set prefix mask */
1169	if (ifra->ifra_prefixmask.sin6_len) {
1170		/*
1171		 * We prohibit changing the prefix length of an existing
1172		 * address, because
1173		 * + such an operation should be rare in IPv6, and
1174		 * + the operation would confuse prefix management.
1175		 */
1176		if (ia->ia_prefixmask.sin6_len &&
1177		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1178			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1179			    " existing (%s) address should not be changed\n",
1180			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1181			error = EINVAL;
1182			goto unlink;
1183		}
1184		ia->ia_prefixmask = ifra->ifra_prefixmask;
1185		ia->ia_prefixmask.sin6_family = AF_INET6;
1186	}
1187
1188	/*
1189	 * If a new destination address is specified, scrub the old one and
1190	 * install the new destination.  Note that the interface must be
1191	 * p2p or loopback (see the check above.)
1192	 */
1193	if (dst6.sin6_family == AF_INET6 &&
1194	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1195		int e;
1196
1197		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1198		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1199			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1200			    "a route to the old destination: %s\n",
1201			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1202			/* proceed anyway... */
1203		} else
1204			ia->ia_flags &= ~IFA_ROUTE;
1205		ia->ia_dstaddr = dst6;
1206	}
1207
1208	/*
1209	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1210	 * to see if the address is deprecated or invalidated, but initialize
1211	 * these members for applications.
1212	 */
1213	ia->ia6_lifetime = ifra->ifra_lifetime;
1214	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1215		ia->ia6_lifetime.ia6t_expire =
1216		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1217	} else
1218		ia->ia6_lifetime.ia6t_expire = 0;
1219	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1220		ia->ia6_lifetime.ia6t_preferred =
1221		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1222	} else
1223		ia->ia6_lifetime.ia6t_preferred = 0;
1224
1225	/* reset the interface and routing table appropriately. */
1226	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1227		goto unlink;
1228
1229	/*
1230	 * configure address flags.
1231	 */
1232	ia->ia6_flags = ifra->ifra_flags;
1233	/*
1234	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1235	 * userland, make it deprecated.
1236	 */
1237	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1238		ia->ia6_lifetime.ia6t_pltime = 0;
1239		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1240	}
1241	/*
1242	 * Make the address tentative before joining multicast addresses,
1243	 * so that corresponding MLD responses would not have a tentative
1244	 * source address.
1245	 */
1246	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1247
1248	/*
1249	 * DAD should be performed for an new address or addresses on
1250	 * an interface with ND6_IFF_IFDISABLED.
1251	 */
1252	if (in6if_do_dad(ifp) &&
1253	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1254		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1255
1256	/*
1257	 * We are done if we have simply modified an existing address.
1258	 */
1259	if (!hostIsNew)
1260		return (error);
1261
1262	/*
1263	 * Beyond this point, we should call in6_purgeaddr upon an error,
1264	 * not just go to unlink.
1265	 */
1266
1267	/* Join necessary multicast groups. */
1268	in6m_sol = NULL;
1269	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1270		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1271		if (error)
1272			goto cleanup;
1273	}
1274
1275	/* Perform DAD, if the address is TENTATIVE. */
1276	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1277		int mindelay, maxdelay;
1278
1279		delay = 0;
1280		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1281			/*
1282			 * We need to impose a delay before sending an NS
1283			 * for DAD.  Check if we also needed a delay for the
1284			 * corresponding MLD message.  If we did, the delay
1285			 * should be larger than the MLD delay (this could be
1286			 * relaxed a bit, but this simple logic is at least
1287			 * safe).
1288			 * XXX: Break data hiding guidelines and look at
1289			 * state for the solicited multicast group.
1290			 */
1291			mindelay = 0;
1292			if (in6m_sol != NULL &&
1293			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1294				mindelay = in6m_sol->in6m_timer;
1295			}
1296			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1297			if (maxdelay - mindelay == 0)
1298				delay = 0;
1299			else {
1300				delay =
1301				    (arc4random() % (maxdelay - mindelay)) +
1302				    mindelay;
1303			}
1304		}
1305		nd6_dad_start((struct ifaddr *)ia, delay);
1306	}
1307
1308	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1309	ifa_free(&ia->ia_ifa);
1310	return (error);
1311
1312  unlink:
1313	/*
1314	 * XXX: if a change of an existing address failed, keep the entry
1315	 * anyway.
1316	 */
1317	if (hostIsNew) {
1318		in6_unlink_ifa(ia, ifp);
1319		ifa_free(&ia->ia_ifa);
1320	}
1321	return (error);
1322
1323  cleanup:
1324	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1325	ifa_free(&ia->ia_ifa);
1326	in6_purgeaddr(&ia->ia_ifa);
1327	return error;
1328}
1329
1330/*
1331 * Leave multicast groups.  Factored out from in6_purgeaddr().
1332 * This entire work should only be done once, for the default FIB.
1333 */
1334static int
1335in6_purgeaddr_mc(struct ifnet *ifp, struct in6_ifaddr *ia, struct ifaddr *ifa0)
1336{
1337	struct sockaddr_in6 mltaddr, mltmask;
1338	struct in6_multi_mship *imm;
1339	struct rtentry *rt;
1340	struct sockaddr_in6 sin6;
1341	int error;
1342
1343	/*
1344	 * Leave from multicast groups we have joined for the interface.
1345	 */
1346	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1347		LIST_REMOVE(imm, i6mm_chain);
1348		in6_leavegroup(imm);
1349	}
1350
1351	/*
1352	 * Remove the link-local all-nodes address.
1353	 */
1354	bzero(&mltmask, sizeof(mltmask));
1355	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1356	mltmask.sin6_family = AF_INET6;
1357	mltmask.sin6_addr = in6mask32;
1358
1359	bzero(&mltaddr, sizeof(mltaddr));
1360	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1361	mltaddr.sin6_family = AF_INET6;
1362	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1363
1364	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1365		return (error);
1366
1367	/*
1368	 * As for the mltaddr above, proactively prepare the sin6 to avoid
1369	 * rtentry un- and re-locking.
1370	 */
1371	if (ifa0 != NULL) {
1372		bzero(&sin6, sizeof(sin6));
1373		sin6.sin6_len = sizeof(sin6);
1374		sin6.sin6_family = AF_INET6;
1375		memcpy(&sin6.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1376		    sizeof(sin6.sin6_addr));
1377		error = in6_setscope(&sin6.sin6_addr, ifa0->ifa_ifp, NULL);
1378		if (error != 0)
1379			return (error);
1380	}
1381
1382	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1383	if (rt != NULL && rt->rt_gateway != NULL &&
1384	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1385		    &ia->ia_addr.sin6_addr,
1386		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1387		/*
1388		 * If no more IPv6 address exists on this interface then
1389		 * remove the multicast address route.
1390		 */
1391		if (ifa0 == NULL) {
1392			memcpy(&mltaddr.sin6_addr,
1393			    &satosin6(rt_key(rt))->sin6_addr,
1394			    sizeof(mltaddr.sin6_addr));
1395			RTFREE_LOCKED(rt);
1396			error = in6_rtrequest(RTM_DELETE,
1397			    (struct sockaddr *)&mltaddr,
1398			    (struct sockaddr *)&ia->ia_addr,
1399			    (struct sockaddr *)&mltmask, RTF_UP,
1400			    (struct rtentry **)0, RT_DEFAULT_FIB);
1401			if (error)
1402				log(LOG_INFO, "%s: link-local all-nodes "
1403				    "multicast address deletion error\n",
1404				    __func__);
1405		} else {
1406			/*
1407			 * Replace the gateway of the route.
1408			 */
1409			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1410			RTFREE_LOCKED(rt);
1411		}
1412	} else {
1413		if (rt != NULL)
1414			RTFREE_LOCKED(rt);
1415	}
1416
1417	/*
1418	 * Remove the node-local all-nodes address.
1419	 */
1420	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1421	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1422		return (error);
1423
1424	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1425	if (rt != NULL && rt->rt_gateway != NULL &&
1426	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1427		    &ia->ia_addr.sin6_addr,
1428		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1429		/*
1430		 * If no more IPv6 address exists on this interface then
1431		 * remove the multicast address route.
1432		 */
1433		if (ifa0 == NULL) {
1434			memcpy(&mltaddr.sin6_addr,
1435			    &satosin6(rt_key(rt))->sin6_addr,
1436			    sizeof(mltaddr.sin6_addr));
1437
1438			RTFREE_LOCKED(rt);
1439			error = in6_rtrequest(RTM_DELETE,
1440			    (struct sockaddr *)&mltaddr,
1441			    (struct sockaddr *)&ia->ia_addr,
1442			    (struct sockaddr *)&mltmask, RTF_UP,
1443			    (struct rtentry **)0, RT_DEFAULT_FIB);
1444			if (error)
1445				log(LOG_INFO, "%s: node-local all-nodes"
1446				    "multicast address deletion error\n",
1447				    __func__);
1448		} else {
1449			/*
1450			 * Replace the gateway of the route.
1451			 */
1452			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1453			RTFREE_LOCKED(rt);
1454		}
1455	} else {
1456		if (rt != NULL)
1457			RTFREE_LOCKED(rt);
1458	}
1459
1460	return (0);
1461}
1462
1463void
1464in6_purgeaddr(struct ifaddr *ifa)
1465{
1466	struct ifnet *ifp = ifa->ifa_ifp;
1467	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1468	int plen, error;
1469	struct ifaddr *ifa0;
1470
1471	if (ifa->ifa_carp)
1472		(*carp_detach_p)(ifa);
1473
1474	/*
1475	 * find another IPv6 address as the gateway for the
1476	 * link-local and node-local all-nodes multicast
1477	 * address routes
1478	 */
1479	IF_ADDR_RLOCK(ifp);
1480	TAILQ_FOREACH(ifa0, &ifp->if_addrhead, ifa_link) {
1481		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1482		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1483		    &ia->ia_addr.sin6_addr, sizeof(struct in6_addr)) == 0)
1484			continue;
1485		else
1486			break;
1487	}
1488	if (ifa0 != NULL)
1489		ifa_ref(ifa0);
1490	IF_ADDR_RUNLOCK(ifp);
1491
1492	/*
1493	 * Remove the loopback route to the interface address.
1494	 * The check for the current setting of "nd6_useloopback"
1495	 * is not needed.
1496	 */
1497	if (ia->ia_flags & IFA_RTSELF) {
1498		error = ifa_del_loopback_route((struct ifaddr *)ia,
1499		    (struct sockaddr *)&ia->ia_addr);
1500		if (error == 0)
1501			ia->ia_flags &= ~IFA_RTSELF;
1502	}
1503
1504	/* stop DAD processing */
1505	nd6_dad_stop(ifa);
1506
1507	/* Remove local address entry from lltable. */
1508	in6_ifremloop(ifa);
1509
1510	/* Leave multicast groups. */
1511	error = in6_purgeaddr_mc(ifp, ia, ifa0);
1512
1513	if (ifa0 != NULL)
1514		ifa_free(ifa0);
1515
1516	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1517	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1518		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1519		    (ia->ia_dstaddr.sin6_family == AF_INET6) ? RTF_HOST : 0);
1520		if (error != 0)
1521			log(LOG_INFO, "%s: err=%d, destination address delete "
1522			    "failed\n", __func__, error);
1523		ia->ia_flags &= ~IFA_ROUTE;
1524	}
1525
1526	in6_unlink_ifa(ia, ifp);
1527}
1528
1529static void
1530in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1531{
1532	char ip6buf[INET6_ADDRSTRLEN];
1533
1534	IF_ADDR_WLOCK(ifp);
1535	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1536	IF_ADDR_WUNLOCK(ifp);
1537	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1538
1539	/*
1540	 * Defer the release of what might be the last reference to the
1541	 * in6_ifaddr so that it can't be freed before the remainder of the
1542	 * cleanup.
1543	 */
1544	IN6_IFADDR_WLOCK();
1545	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1546	LIST_REMOVE(ia, ia6_hash);
1547	IN6_IFADDR_WUNLOCK();
1548
1549	/*
1550	 * Release the reference to the base prefix.  There should be a
1551	 * positive reference.
1552	 */
1553	if (ia->ia6_ndpr == NULL) {
1554		nd6log((LOG_NOTICE,
1555		    "in6_unlink_ifa: autoconf'ed address "
1556		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1557	} else {
1558		ia->ia6_ndpr->ndpr_refcnt--;
1559		ia->ia6_ndpr = NULL;
1560	}
1561
1562	/*
1563	 * Also, if the address being removed is autoconf'ed, call
1564	 * pfxlist_onlink_check() since the release might affect the status of
1565	 * other (detached) addresses.
1566	 */
1567	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1568		pfxlist_onlink_check();
1569	}
1570	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1571}
1572
1573void
1574in6_purgeif(struct ifnet *ifp)
1575{
1576	struct ifaddr *ifa, *nifa;
1577
1578	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1579		if (ifa->ifa_addr->sa_family != AF_INET6)
1580			continue;
1581		in6_purgeaddr(ifa);
1582	}
1583
1584	in6_ifdetach(ifp);
1585}
1586
1587/*
1588 * SIOC[GAD]LIFADDR.
1589 *	SIOCGLIFADDR: get first address. (?)
1590 *	SIOCGLIFADDR with IFLR_PREFIX:
1591 *		get first address that matches the specified prefix.
1592 *	SIOCALIFADDR: add the specified address.
1593 *	SIOCALIFADDR with IFLR_PREFIX:
1594 *		add the specified prefix, filling hostid part from
1595 *		the first link-local address.  prefixlen must be <= 64.
1596 *	SIOCDLIFADDR: delete the specified address.
1597 *	SIOCDLIFADDR with IFLR_PREFIX:
1598 *		delete the first address that matches the specified prefix.
1599 * return values:
1600 *	EINVAL on invalid parameters
1601 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1602 *	other values may be returned from in6_ioctl()
1603 *
1604 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1605 * this is to accomodate address naming scheme other than RFC2374,
1606 * in the future.
1607 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1608 * address encoding scheme. (see figure on page 8)
1609 * Notifies other subsystems about address change/arrival:
1610 * 1) Notifies device handler on the first IPv6 address assignment
1611 * 2) Handle routing table changes for P2P links and route
1612 * 3) Handle routing table changes for address host route
1613 */
1614static int
1615in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1616    struct ifnet *ifp, struct thread *td)
1617{
1618	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1619	struct ifaddr *ifa;
1620	struct sockaddr *sa;
1621
1622	/* sanity checks */
1623	if (!data || !ifp) {
1624		panic("invalid argument to in6_lifaddr_ioctl");
1625		/* NOTREACHED */
1626	}
1627
1628	switch (cmd) {
1629	case SIOCGLIFADDR:
1630		/* address must be specified on GET with IFLR_PREFIX */
1631		if ((iflr->flags & IFLR_PREFIX) == 0)
1632			break;
1633		/* FALLTHROUGH */
1634	case SIOCALIFADDR:
1635	case SIOCDLIFADDR:
1636		/* address must be specified on ADD and DELETE */
1637		sa = (struct sockaddr *)&iflr->addr;
1638		if (sa->sa_family != AF_INET6)
1639			return EINVAL;
1640		if (sa->sa_len != sizeof(struct sockaddr_in6))
1641			return EINVAL;
1642		/* XXX need improvement */
1643		sa = (struct sockaddr *)&iflr->dstaddr;
1644		if (sa->sa_family && sa->sa_family != AF_INET6)
1645			return EINVAL;
1646		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1647			return EINVAL;
1648		break;
1649	default: /* shouldn't happen */
1650#if 0
1651		panic("invalid cmd to in6_lifaddr_ioctl");
1652		/* NOTREACHED */
1653#else
1654		return EOPNOTSUPP;
1655#endif
1656	}
1657	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1658		return EINVAL;
1659
1660	switch (cmd) {
1661	case SIOCALIFADDR:
1662	    {
1663		struct in6_aliasreq ifra;
1664		struct in6_addr *hostid = NULL;
1665		int prefixlen;
1666
1667		ifa = NULL;
1668		if ((iflr->flags & IFLR_PREFIX) != 0) {
1669			struct sockaddr_in6 *sin6;
1670
1671			/*
1672			 * hostid is to fill in the hostid part of the
1673			 * address.  hostid points to the first link-local
1674			 * address attached to the interface.
1675			 */
1676			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1677			if (!ifa)
1678				return EADDRNOTAVAIL;
1679			hostid = IFA_IN6(ifa);
1680
1681			/* prefixlen must be <= 64. */
1682			if (64 < iflr->prefixlen) {
1683				if (ifa != NULL)
1684					ifa_free(ifa);
1685				return EINVAL;
1686			}
1687			prefixlen = iflr->prefixlen;
1688
1689			/* hostid part must be zero. */
1690			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1691			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1692			    sin6->sin6_addr.s6_addr32[3] != 0) {
1693				if (ifa != NULL)
1694					ifa_free(ifa);
1695				return EINVAL;
1696			}
1697		} else
1698			prefixlen = iflr->prefixlen;
1699
1700		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1701		bzero(&ifra, sizeof(ifra));
1702		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1703
1704		bcopy(&iflr->addr, &ifra.ifra_addr,
1705		    ((struct sockaddr *)&iflr->addr)->sa_len);
1706		if (hostid) {
1707			/* fill in hostid part */
1708			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1709			    hostid->s6_addr32[2];
1710			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1711			    hostid->s6_addr32[3];
1712		}
1713
1714		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1715			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1716			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1717			if (hostid) {
1718				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1719				    hostid->s6_addr32[2];
1720				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1721				    hostid->s6_addr32[3];
1722			}
1723		}
1724		if (ifa != NULL)
1725			ifa_free(ifa);
1726
1727		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1728		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1729
1730		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1731		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1732	    }
1733	case SIOCGLIFADDR:
1734	case SIOCDLIFADDR:
1735	    {
1736		struct in6_ifaddr *ia;
1737		struct in6_addr mask, candidate, match;
1738		struct sockaddr_in6 *sin6;
1739		int cmp;
1740
1741		bzero(&mask, sizeof(mask));
1742		if (iflr->flags & IFLR_PREFIX) {
1743			/* lookup a prefix rather than address. */
1744			in6_prefixlen2mask(&mask, iflr->prefixlen);
1745
1746			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1747			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1748			match.s6_addr32[0] &= mask.s6_addr32[0];
1749			match.s6_addr32[1] &= mask.s6_addr32[1];
1750			match.s6_addr32[2] &= mask.s6_addr32[2];
1751			match.s6_addr32[3] &= mask.s6_addr32[3];
1752
1753			/* if you set extra bits, that's wrong */
1754			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1755				return EINVAL;
1756
1757			cmp = 1;
1758		} else {
1759			if (cmd == SIOCGLIFADDR) {
1760				/* on getting an address, take the 1st match */
1761				cmp = 0;	/* XXX */
1762			} else {
1763				/* on deleting an address, do exact match */
1764				in6_prefixlen2mask(&mask, 128);
1765				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1766				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1767
1768				cmp = 1;
1769			}
1770		}
1771
1772		IF_ADDR_RLOCK(ifp);
1773		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1774			if (ifa->ifa_addr->sa_family != AF_INET6)
1775				continue;
1776			if (!cmp)
1777				break;
1778
1779			/*
1780			 * XXX: this is adhoc, but is necessary to allow
1781			 * a user to specify fe80::/64 (not /10) for a
1782			 * link-local address.
1783			 */
1784			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1785			in6_clearscope(&candidate);
1786			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1787			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1788			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1789			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1790			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1791				break;
1792		}
1793		if (ifa != NULL)
1794			ifa_ref(ifa);
1795		IF_ADDR_RUNLOCK(ifp);
1796		if (!ifa)
1797			return EADDRNOTAVAIL;
1798		ia = ifa2ia6(ifa);
1799
1800		if (cmd == SIOCGLIFADDR) {
1801			int error;
1802
1803			/* fill in the if_laddrreq structure */
1804			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1805			error = sa6_recoverscope(
1806			    (struct sockaddr_in6 *)&iflr->addr);
1807			if (error != 0) {
1808				ifa_free(ifa);
1809				return (error);
1810			}
1811
1812			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1813				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1814				    ia->ia_dstaddr.sin6_len);
1815				error = sa6_recoverscope(
1816				    (struct sockaddr_in6 *)&iflr->dstaddr);
1817				if (error != 0) {
1818					ifa_free(ifa);
1819					return (error);
1820				}
1821			} else
1822				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1823
1824			iflr->prefixlen =
1825			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1826
1827			iflr->flags = ia->ia6_flags;	/* XXX */
1828			ifa_free(ifa);
1829
1830			return 0;
1831		} else {
1832			struct in6_aliasreq ifra;
1833
1834			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1835			bzero(&ifra, sizeof(ifra));
1836			bcopy(iflr->iflr_name, ifra.ifra_name,
1837			    sizeof(ifra.ifra_name));
1838
1839			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1840			    ia->ia_addr.sin6_len);
1841			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1842				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1843				    ia->ia_dstaddr.sin6_len);
1844			} else {
1845				bzero(&ifra.ifra_dstaddr,
1846				    sizeof(ifra.ifra_dstaddr));
1847			}
1848			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1849			    ia->ia_prefixmask.sin6_len);
1850
1851			ifra.ifra_flags = ia->ia6_flags;
1852			ifa_free(ifa);
1853			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1854			    ifp, td);
1855		}
1856	    }
1857	}
1858
1859	return EOPNOTSUPP;	/* just for safety */
1860}
1861
1862/*
1863 * Initialize an interface's IPv6 address and routing table entry.
1864 */
1865static int
1866in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1867    struct sockaddr_in6 *sin6, int newhost)
1868{
1869	int	error = 0, plen, ifacount = 0;
1870	struct ifaddr *ifa;
1871
1872	/*
1873	 * Give the interface a chance to initialize
1874	 * if this is its first address,
1875	 * and to validate the address if necessary.
1876	 */
1877	IF_ADDR_RLOCK(ifp);
1878	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1879		if (ifa->ifa_addr->sa_family != AF_INET6)
1880			continue;
1881		ifacount++;
1882	}
1883	IF_ADDR_RUNLOCK(ifp);
1884
1885	ia->ia_addr = *sin6;
1886
1887	if (ifacount <= 1 && ifp->if_ioctl) {
1888		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1889		if (error)
1890			return (error);
1891	}
1892
1893	ia->ia_ifa.ifa_metric = ifp->if_metric;
1894
1895	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1896
1897	/*
1898	 * Special case:
1899	 * If a new destination address is specified for a point-to-point
1900	 * interface, install a route to the destination as an interface
1901	 * direct route.
1902	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1903	 * interface that share the same destination.
1904	 */
1905	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1906	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1907	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1908		int rtflags = RTF_UP | RTF_HOST;
1909		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1910		if (error)
1911			return (error);
1912		ia->ia_flags |= IFA_ROUTE;
1913		/*
1914		 * Handle the case for ::1 .
1915		 */
1916		if (ifp->if_flags & IFF_LOOPBACK)
1917			ia->ia_flags |= IFA_RTSELF;
1918	}
1919
1920	/*
1921	 * add a loopback route to self
1922	 */
1923	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1924		error = ifa_add_loopback_route((struct ifaddr *)ia,
1925		    (struct sockaddr *)&ia->ia_addr);
1926		if (error == 0)
1927			ia->ia_flags |= IFA_RTSELF;
1928	}
1929
1930	/* Add local address to lltable, if necessary (ex. on p2p link). */
1931	if (newhost)
1932		in6_ifaddloop(&(ia->ia_ifa));
1933
1934	return (error);
1935}
1936
1937/*
1938 * Find an IPv6 interface link-local address specific to an interface.
1939 * ifaddr is returned referenced.
1940 */
1941struct in6_ifaddr *
1942in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1943{
1944	struct ifaddr *ifa;
1945
1946	IF_ADDR_RLOCK(ifp);
1947	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1948		if (ifa->ifa_addr->sa_family != AF_INET6)
1949			continue;
1950		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1951			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1952			    ignoreflags) != 0)
1953				continue;
1954			ifa_ref(ifa);
1955			break;
1956		}
1957	}
1958	IF_ADDR_RUNLOCK(ifp);
1959
1960	return ((struct in6_ifaddr *)ifa);
1961}
1962
1963
1964/*
1965 * find the internet address corresponding to a given interface and address.
1966 * ifaddr is returned referenced.
1967 */
1968struct in6_ifaddr *
1969in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1970{
1971	struct ifaddr *ifa;
1972
1973	IF_ADDR_RLOCK(ifp);
1974	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1975		if (ifa->ifa_addr->sa_family != AF_INET6)
1976			continue;
1977		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1978			ifa_ref(ifa);
1979			break;
1980		}
1981	}
1982	IF_ADDR_RUNLOCK(ifp);
1983
1984	return ((struct in6_ifaddr *)ifa);
1985}
1986
1987/*
1988 * Find a link-local scoped address on ifp and return it if any.
1989 */
1990struct in6_ifaddr *
1991in6ifa_llaonifp(struct ifnet *ifp)
1992{
1993	struct sockaddr_in6 *sin6;
1994	struct ifaddr *ifa;
1995
1996	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1997		return (NULL);
1998	IF_ADDR_RLOCK(ifp);
1999	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2000		if (ifa->ifa_addr->sa_family != AF_INET6)
2001			continue;
2002		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
2003		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
2004		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
2005		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
2006			break;
2007	}
2008	IF_ADDR_RUNLOCK(ifp);
2009
2010	return ((struct in6_ifaddr *)ifa);
2011}
2012
2013/*
2014 * Convert IP6 address to printable (loggable) representation. Caller
2015 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
2016 */
2017static char digits[] = "0123456789abcdef";
2018char *
2019ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
2020{
2021	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
2022	char *cp;
2023	const u_int16_t *a = (const u_int16_t *)addr;
2024	const u_int8_t *d;
2025	int dcolon = 0, zero = 0;
2026
2027	cp = ip6buf;
2028
2029	for (i = 0; i < 8; i++) {
2030		if (*(a + i) == 0) {
2031			cnt++;
2032			if (cnt == 1)
2033				idx = i;
2034		}
2035		else if (maxcnt < cnt) {
2036			maxcnt = cnt;
2037			index = idx;
2038			cnt = 0;
2039		}
2040	}
2041	if (maxcnt < cnt) {
2042		maxcnt = cnt;
2043		index = idx;
2044	}
2045
2046	for (i = 0; i < 8; i++) {
2047		if (dcolon == 1) {
2048			if (*a == 0) {
2049				if (i == 7)
2050					*cp++ = ':';
2051				a++;
2052				continue;
2053			} else
2054				dcolon = 2;
2055		}
2056		if (*a == 0) {
2057			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
2058				if (i == 0)
2059					*cp++ = ':';
2060				*cp++ = ':';
2061				dcolon = 1;
2062			} else {
2063				*cp++ = '0';
2064				*cp++ = ':';
2065			}
2066			a++;
2067			continue;
2068		}
2069		d = (const u_char *)a;
2070		/* Try to eliminate leading zeros in printout like in :0001. */
2071		zero = 1;
2072		*cp = digits[*d >> 4];
2073		if (*cp != '0') {
2074			zero = 0;
2075			cp++;
2076		}
2077		*cp = digits[*d++ & 0xf];
2078		if (zero == 0 || (*cp != '0')) {
2079			zero = 0;
2080			cp++;
2081		}
2082		*cp = digits[*d >> 4];
2083		if (zero == 0 || (*cp != '0')) {
2084			zero = 0;
2085			cp++;
2086		}
2087		*cp++ = digits[*d & 0xf];
2088		*cp++ = ':';
2089		a++;
2090	}
2091	*--cp = '\0';
2092	return (ip6buf);
2093}
2094
2095int
2096in6_localaddr(struct in6_addr *in6)
2097{
2098	struct in6_ifaddr *ia;
2099
2100	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2101		return 1;
2102
2103	IN6_IFADDR_RLOCK();
2104	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2105		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2106		    &ia->ia_prefixmask.sin6_addr)) {
2107			IN6_IFADDR_RUNLOCK();
2108			return 1;
2109		}
2110	}
2111	IN6_IFADDR_RUNLOCK();
2112
2113	return (0);
2114}
2115
2116/*
2117 * Return 1 if an internet address is for the local host and configured
2118 * on one of its interfaces.
2119 */
2120int
2121in6_localip(struct in6_addr *in6)
2122{
2123	struct in6_ifaddr *ia;
2124
2125	IN6_IFADDR_RLOCK();
2126	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
2127		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2128			IN6_IFADDR_RUNLOCK();
2129			return (1);
2130		}
2131	}
2132	IN6_IFADDR_RUNLOCK();
2133	return (0);
2134}
2135
2136int
2137in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2138{
2139	struct in6_ifaddr *ia;
2140
2141	IN6_IFADDR_RLOCK();
2142	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
2143		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
2144			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2145				IN6_IFADDR_RUNLOCK();
2146				return (1); /* true */
2147			}
2148			break;
2149		}
2150	}
2151	IN6_IFADDR_RUNLOCK();
2152
2153	return (0);		/* false */
2154}
2155
2156/*
2157 * return length of part which dst and src are equal
2158 * hard coding...
2159 */
2160int
2161in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2162{
2163	int match = 0;
2164	u_char *s = (u_char *)src, *d = (u_char *)dst;
2165	u_char *lim = s + 16, r;
2166
2167	while (s < lim)
2168		if ((r = (*d++ ^ *s++)) != 0) {
2169			while (r < 128) {
2170				match++;
2171				r <<= 1;
2172			}
2173			break;
2174		} else
2175			match += 8;
2176	return match;
2177}
2178
2179/* XXX: to be scope conscious */
2180int
2181in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2182{
2183	int bytelen, bitlen;
2184
2185	/* sanity check */
2186	if (0 > len || len > 128) {
2187		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2188		    len);
2189		return (0);
2190	}
2191
2192	bytelen = len / 8;
2193	bitlen = len % 8;
2194
2195	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2196		return (0);
2197	if (bitlen != 0 &&
2198	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2199	    p2->s6_addr[bytelen] >> (8 - bitlen))
2200		return (0);
2201
2202	return (1);
2203}
2204
2205void
2206in6_prefixlen2mask(struct in6_addr *maskp, int len)
2207{
2208	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2209	int bytelen, bitlen, i;
2210
2211	/* sanity check */
2212	if (0 > len || len > 128) {
2213		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2214		    len);
2215		return;
2216	}
2217
2218	bzero(maskp, sizeof(*maskp));
2219	bytelen = len / 8;
2220	bitlen = len % 8;
2221	for (i = 0; i < bytelen; i++)
2222		maskp->s6_addr[i] = 0xff;
2223	if (bitlen)
2224		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2225}
2226
2227/*
2228 * return the best address out of the same scope. if no address was
2229 * found, return the first valid address from designated IF.
2230 */
2231struct in6_ifaddr *
2232in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2233{
2234	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2235	struct ifaddr *ifa;
2236	struct in6_ifaddr *besta = 0;
2237	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2238
2239	dep[0] = dep[1] = NULL;
2240
2241	/*
2242	 * We first look for addresses in the same scope.
2243	 * If there is one, return it.
2244	 * If two or more, return one which matches the dst longest.
2245	 * If none, return one of global addresses assigned other ifs.
2246	 */
2247	IF_ADDR_RLOCK(ifp);
2248	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2249		if (ifa->ifa_addr->sa_family != AF_INET6)
2250			continue;
2251		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2252			continue; /* XXX: is there any case to allow anycast? */
2253		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2254			continue; /* don't use this interface */
2255		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2256			continue;
2257		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2258			if (V_ip6_use_deprecated)
2259				dep[0] = (struct in6_ifaddr *)ifa;
2260			continue;
2261		}
2262
2263		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2264			/*
2265			 * call in6_matchlen() as few as possible
2266			 */
2267			if (besta) {
2268				if (blen == -1)
2269					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2270				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2271				if (tlen > blen) {
2272					blen = tlen;
2273					besta = (struct in6_ifaddr *)ifa;
2274				}
2275			} else
2276				besta = (struct in6_ifaddr *)ifa;
2277		}
2278	}
2279	if (besta) {
2280		ifa_ref(&besta->ia_ifa);
2281		IF_ADDR_RUNLOCK(ifp);
2282		return (besta);
2283	}
2284
2285	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2286		if (ifa->ifa_addr->sa_family != AF_INET6)
2287			continue;
2288		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2289			continue; /* XXX: is there any case to allow anycast? */
2290		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2291			continue; /* don't use this interface */
2292		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2293			continue;
2294		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2295			if (V_ip6_use_deprecated)
2296				dep[1] = (struct in6_ifaddr *)ifa;
2297			continue;
2298		}
2299
2300		if (ifa != NULL)
2301			ifa_ref(ifa);
2302		IF_ADDR_RUNLOCK(ifp);
2303		return (struct in6_ifaddr *)ifa;
2304	}
2305
2306	/* use the last-resort values, that are, deprecated addresses */
2307	if (dep[0]) {
2308		ifa_ref((struct ifaddr *)dep[0]);
2309		IF_ADDR_RUNLOCK(ifp);
2310		return dep[0];
2311	}
2312	if (dep[1]) {
2313		ifa_ref((struct ifaddr *)dep[1]);
2314		IF_ADDR_RUNLOCK(ifp);
2315		return dep[1];
2316	}
2317
2318	IF_ADDR_RUNLOCK(ifp);
2319	return NULL;
2320}
2321
2322/*
2323 * perform DAD when interface becomes IFF_UP.
2324 */
2325void
2326in6_if_up(struct ifnet *ifp)
2327{
2328	struct ifaddr *ifa;
2329	struct in6_ifaddr *ia;
2330
2331	IF_ADDR_RLOCK(ifp);
2332	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2333		if (ifa->ifa_addr->sa_family != AF_INET6)
2334			continue;
2335		ia = (struct in6_ifaddr *)ifa;
2336		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2337			/*
2338			 * The TENTATIVE flag was likely set by hand
2339			 * beforehand, implicitly indicating the need for DAD.
2340			 * We may be able to skip the random delay in this
2341			 * case, but we impose delays just in case.
2342			 */
2343			nd6_dad_start(ifa,
2344			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2345		}
2346	}
2347	IF_ADDR_RUNLOCK(ifp);
2348
2349	/*
2350	 * special cases, like 6to4, are handled in in6_ifattach
2351	 */
2352	in6_ifattach(ifp, NULL);
2353}
2354
2355int
2356in6if_do_dad(struct ifnet *ifp)
2357{
2358	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2359		return (0);
2360
2361	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) ||
2362	    (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD))
2363		return (0);
2364
2365	switch (ifp->if_type) {
2366#ifdef IFT_DUMMY
2367	case IFT_DUMMY:
2368#endif
2369	case IFT_FAITH:
2370		/*
2371		 * These interfaces do not have the IFF_LOOPBACK flag,
2372		 * but loop packets back.  We do not have to do DAD on such
2373		 * interfaces.  We should even omit it, because loop-backed
2374		 * NS would confuse the DAD procedure.
2375		 */
2376		return (0);
2377	default:
2378		/*
2379		 * Our DAD routine requires the interface up and running.
2380		 * However, some interfaces can be up before the RUNNING
2381		 * status.  Additionaly, users may try to assign addresses
2382		 * before the interface becomes up (or running).
2383		 * This function returns EAGAIN in that case.
2384		 * The caller should mark "tentative" on the address instead of
2385		 * performing DAD immediately.
2386		 */
2387		if (!((ifp->if_flags & IFF_UP) &&
2388		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2389			return (EAGAIN);
2390
2391		return (1);
2392	}
2393}
2394
2395/*
2396 * Calculate max IPv6 MTU through all the interfaces and store it
2397 * to in6_maxmtu.
2398 */
2399void
2400in6_setmaxmtu(void)
2401{
2402	unsigned long maxmtu = 0;
2403	struct ifnet *ifp;
2404
2405	IFNET_RLOCK_NOSLEEP();
2406	TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
2407		/* this function can be called during ifnet initialization */
2408		if (!ifp->if_afdata[AF_INET6])
2409			continue;
2410		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2411		    IN6_LINKMTU(ifp) > maxmtu)
2412			maxmtu = IN6_LINKMTU(ifp);
2413	}
2414	IFNET_RUNLOCK_NOSLEEP();
2415	if (maxmtu)	/* update only when maxmtu is positive */
2416		V_in6_maxmtu = maxmtu;
2417}
2418
2419/*
2420 * Provide the length of interface identifiers to be used for the link attached
2421 * to the given interface.  The length should be defined in "IPv6 over
2422 * xxx-link" document.  Note that address architecture might also define
2423 * the length for a particular set of address prefixes, regardless of the
2424 * link type.  As clarified in rfc2462bis, those two definitions should be
2425 * consistent, and those really are as of August 2004.
2426 */
2427int
2428in6_if2idlen(struct ifnet *ifp)
2429{
2430	switch (ifp->if_type) {
2431	case IFT_ETHER:		/* RFC2464 */
2432#ifdef IFT_PROPVIRTUAL
2433	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2434#endif
2435#ifdef IFT_L2VLAN
2436	case IFT_L2VLAN:	/* ditto */
2437#endif
2438#ifdef IFT_IEEE80211
2439	case IFT_IEEE80211:	/* ditto */
2440#endif
2441#ifdef IFT_MIP
2442	case IFT_MIP:	/* ditto */
2443#endif
2444	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2445	case IFT_INFINIBAND:
2446		return (64);
2447	case IFT_FDDI:		/* RFC2467 */
2448		return (64);
2449	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2450		return (64);
2451	case IFT_PPP:		/* RFC2472 */
2452		return (64);
2453	case IFT_ARCNET:	/* RFC2497 */
2454		return (64);
2455	case IFT_FRELAY:	/* RFC2590 */
2456		return (64);
2457	case IFT_IEEE1394:	/* RFC3146 */
2458		return (64);
2459	case IFT_GIF:
2460		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2461	case IFT_LOOP:
2462		return (64);	/* XXX: is this really correct? */
2463	default:
2464		/*
2465		 * Unknown link type:
2466		 * It might be controversial to use the today's common constant
2467		 * of 64 for these cases unconditionally.  For full compliance,
2468		 * we should return an error in this case.  On the other hand,
2469		 * if we simply miss the standard for the link type or a new
2470		 * standard is defined for a new link type, the IFID length
2471		 * is very likely to be the common constant.  As a compromise,
2472		 * we always use the constant, but make an explicit notice
2473		 * indicating the "unknown" case.
2474		 */
2475		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2476		return (64);
2477	}
2478}
2479
2480#include <sys/sysctl.h>
2481
2482struct in6_llentry {
2483	struct llentry		base;
2484	struct sockaddr_in6	l3_addr6;
2485};
2486
2487/*
2488 * Deletes an address from the address table.
2489 * This function is called by the timer functions
2490 * such as arptimer() and nd6_llinfo_timer(), and
2491 * the caller does the locking.
2492 */
2493static void
2494in6_lltable_free(struct lltable *llt, struct llentry *lle)
2495{
2496	LLE_WUNLOCK(lle);
2497	LLE_LOCK_DESTROY(lle);
2498	free(lle, M_LLTABLE);
2499}
2500
2501static struct llentry *
2502in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2503{
2504	struct in6_llentry *lle;
2505
2506	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2507	if (lle == NULL)		/* NB: caller generates msg */
2508		return NULL;
2509
2510	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2511	lle->base.lle_refcnt = 1;
2512	lle->base.lle_free = in6_lltable_free;
2513	LLE_LOCK_INIT(&lle->base);
2514	callout_init(&lle->base.ln_timer_ch, 1);
2515
2516	return (&lle->base);
2517}
2518
2519static void
2520in6_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix,
2521    const struct sockaddr *mask, u_int flags)
2522{
2523	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2524	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2525	struct llentry *lle, *next;
2526	int i;
2527
2528	/*
2529	 * (flags & LLE_STATIC) means deleting all entries
2530	 * including static ND6 entries.
2531	 */
2532	IF_AFDATA_WLOCK(llt->llt_ifp);
2533	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2534		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2535			if (IN6_ARE_MASKED_ADDR_EQUAL(
2536			    &satosin6(L3_ADDR(lle))->sin6_addr,
2537			    &pfx->sin6_addr, &msk->sin6_addr) &&
2538			    ((flags & LLE_STATIC) ||
2539			    !(lle->la_flags & LLE_STATIC))) {
2540				LLE_WLOCK(lle);
2541				if (callout_stop(&lle->la_timer))
2542					LLE_REMREF(lle);
2543				llentry_free(lle);
2544			}
2545		}
2546	}
2547	IF_AFDATA_WUNLOCK(llt->llt_ifp);
2548}
2549
2550static int
2551in6_lltable_rtcheck(struct ifnet *ifp,
2552		    u_int flags,
2553		    const struct sockaddr *l3addr)
2554{
2555	struct rtentry *rt;
2556	char ip6buf[INET6_ADDRSTRLEN];
2557
2558	KASSERT(l3addr->sa_family == AF_INET6,
2559	    ("sin_family %d", l3addr->sa_family));
2560
2561	/* Our local addresses are always only installed on the default FIB. */
2562	/* XXX rtalloc1 should take a const param */
2563	rt = in6_rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0,
2564	    RT_DEFAULT_FIB);
2565	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2566		struct ifaddr *ifa;
2567		/*
2568		 * Create an ND6 cache for an IPv6 neighbor
2569		 * that is not covered by our own prefix.
2570		 */
2571		/* XXX ifaof_ifpforaddr should take a const param */
2572		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2573		if (ifa != NULL) {
2574			ifa_free(ifa);
2575			if (rt != NULL)
2576				RTFREE_LOCKED(rt);
2577			return 0;
2578		}
2579		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2580		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2581		if (rt != NULL)
2582			RTFREE_LOCKED(rt);
2583		return EINVAL;
2584	}
2585	RTFREE_LOCKED(rt);
2586	return 0;
2587}
2588
2589static struct llentry *
2590in6_lltable_lookup(struct lltable *llt, u_int flags,
2591	const struct sockaddr *l3addr)
2592{
2593	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2594	struct ifnet *ifp = llt->llt_ifp;
2595	struct llentry *lle;
2596	struct llentries *lleh;
2597	u_int hashkey;
2598
2599	IF_AFDATA_LOCK_ASSERT(ifp);
2600	KASSERT(l3addr->sa_family == AF_INET6,
2601	    ("sin_family %d", l3addr->sa_family));
2602
2603	hashkey = sin6->sin6_addr.s6_addr32[3];
2604	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2605	LIST_FOREACH(lle, lleh, lle_next) {
2606		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2607		if (lle->la_flags & LLE_DELETED)
2608			continue;
2609		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2610		    sizeof(struct in6_addr)) == 0)
2611			break;
2612	}
2613
2614	if (lle == NULL) {
2615		if (!(flags & LLE_CREATE))
2616			return (NULL);
2617		IF_AFDATA_WLOCK_ASSERT(ifp);
2618		/*
2619		 * A route that covers the given address must have
2620		 * been installed 1st because we are doing a resolution,
2621		 * verify this.
2622		 */
2623		if (!(flags & LLE_IFADDR) &&
2624		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2625			return NULL;
2626
2627		lle = in6_lltable_new(l3addr, flags);
2628		if (lle == NULL) {
2629			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2630			return NULL;
2631		}
2632		lle->la_flags = flags & ~LLE_CREATE;
2633		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2634			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2635			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2636		}
2637
2638		lle->lle_tbl  = llt;
2639		lle->lle_head = lleh;
2640		lle->la_flags |= LLE_LINKED;
2641		LIST_INSERT_HEAD(lleh, lle, lle_next);
2642	} else if (flags & LLE_DELETE) {
2643		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2644			LLE_WLOCK(lle);
2645			lle->la_flags |= LLE_DELETED;
2646			EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2647#ifdef DIAGNOSTIC
2648			log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2649#endif
2650			if ((lle->la_flags &
2651			    (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2652				llentry_free(lle);
2653			else
2654				LLE_WUNLOCK(lle);
2655		}
2656		lle = (void *)-1;
2657	}
2658	if (LLE_IS_VALID(lle)) {
2659		if (flags & LLE_EXCLUSIVE)
2660			LLE_WLOCK(lle);
2661		else
2662			LLE_RLOCK(lle);
2663	}
2664	return (lle);
2665}
2666
2667static int
2668in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2669{
2670	struct ifnet *ifp = llt->llt_ifp;
2671	struct llentry *lle;
2672	/* XXX stack use */
2673	struct {
2674		struct rt_msghdr	rtm;
2675		struct sockaddr_in6	sin6;
2676		/*
2677		 * ndp.c assumes that sdl is word aligned
2678		 */
2679#ifdef __LP64__
2680		uint32_t		pad;
2681#endif
2682		struct sockaddr_dl	sdl;
2683	} ndpc;
2684	int i, error;
2685
2686	if (ifp->if_flags & IFF_LOOPBACK)
2687		return 0;
2688
2689	LLTABLE_LOCK_ASSERT();
2690
2691	error = 0;
2692	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2693		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2694			struct sockaddr_dl *sdl;
2695
2696			/* skip deleted or invalid entries */
2697			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2698				continue;
2699			/* Skip if jailed and not a valid IP of the prison. */
2700			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2701				continue;
2702			/*
2703			 * produce a msg made of:
2704			 *  struct rt_msghdr;
2705			 *  struct sockaddr_in6 (IPv6)
2706			 *  struct sockaddr_dl;
2707			 */
2708			bzero(&ndpc, sizeof(ndpc));
2709			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2710			ndpc.rtm.rtm_version = RTM_VERSION;
2711			ndpc.rtm.rtm_type = RTM_GET;
2712			ndpc.rtm.rtm_flags = RTF_UP;
2713			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2714			ndpc.sin6.sin6_family = AF_INET6;
2715			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2716			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2717			if (V_deembed_scopeid)
2718				sa6_recoverscope(&ndpc.sin6);
2719
2720			/* publish */
2721			if (lle->la_flags & LLE_PUB)
2722				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2723
2724			sdl = &ndpc.sdl;
2725			sdl->sdl_family = AF_LINK;
2726			sdl->sdl_len = sizeof(*sdl);
2727			sdl->sdl_alen = ifp->if_addrlen;
2728			sdl->sdl_index = ifp->if_index;
2729			sdl->sdl_type = ifp->if_type;
2730			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2731			ndpc.rtm.rtm_rmx.rmx_expire =
2732			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2733			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2734			if (lle->la_flags & LLE_STATIC)
2735				ndpc.rtm.rtm_flags |= RTF_STATIC;
2736			ndpc.rtm.rtm_index = ifp->if_index;
2737			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2738			if (error)
2739				break;
2740		}
2741	}
2742	return error;
2743}
2744
2745void *
2746in6_domifattach(struct ifnet *ifp)
2747{
2748	struct in6_ifextra *ext;
2749
2750	/* There are not IPv6-capable interfaces. */
2751	switch (ifp->if_type) {
2752	case IFT_PFLOG:
2753	case IFT_PFSYNC:
2754	case IFT_USB:
2755		return (NULL);
2756	}
2757	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2758	bzero(ext, sizeof(*ext));
2759
2760	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2761	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2762	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2763	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2764
2765	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2766	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2767	    M_WAITOK);
2768	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2769	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2770
2771	ext->nd_ifinfo = nd6_ifattach(ifp);
2772	ext->scope6_id = scope6_ifattach(ifp);
2773	ext->lltable = lltable_init(ifp, AF_INET6);
2774	if (ext->lltable != NULL) {
2775		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2776		ext->lltable->llt_lookup = in6_lltable_lookup;
2777		ext->lltable->llt_dump = in6_lltable_dump;
2778	}
2779
2780	ext->mld_ifinfo = mld_domifattach(ifp);
2781
2782	return ext;
2783}
2784
2785void
2786in6_domifdetach(struct ifnet *ifp, void *aux)
2787{
2788	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2789
2790	mld_domifdetach(ifp);
2791	scope6_ifdetach(ext->scope6_id);
2792	nd6_ifdetach(ext->nd_ifinfo);
2793	lltable_free(ext->lltable);
2794	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2795	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2796	free(ext->in6_ifstat, M_IFADDR);
2797	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2798	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2799	free(ext->icmp6_ifstat, M_IFADDR);
2800	free(ext, M_IFADDR);
2801}
2802
2803/*
2804 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2805 * v4 mapped addr or v4 compat addr
2806 */
2807void
2808in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2809{
2810
2811	bzero(sin, sizeof(*sin));
2812	sin->sin_len = sizeof(struct sockaddr_in);
2813	sin->sin_family = AF_INET;
2814	sin->sin_port = sin6->sin6_port;
2815	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2816}
2817
2818/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2819void
2820in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2821{
2822	bzero(sin6, sizeof(*sin6));
2823	sin6->sin6_len = sizeof(struct sockaddr_in6);
2824	sin6->sin6_family = AF_INET6;
2825	sin6->sin6_port = sin->sin_port;
2826	sin6->sin6_addr.s6_addr32[0] = 0;
2827	sin6->sin6_addr.s6_addr32[1] = 0;
2828	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2829	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2830}
2831
2832/* Convert sockaddr_in6 into sockaddr_in. */
2833void
2834in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2835{
2836	struct sockaddr_in *sin_p;
2837	struct sockaddr_in6 sin6;
2838
2839	/*
2840	 * Save original sockaddr_in6 addr and convert it
2841	 * to sockaddr_in.
2842	 */
2843	sin6 = *(struct sockaddr_in6 *)nam;
2844	sin_p = (struct sockaddr_in *)nam;
2845	in6_sin6_2_sin(sin_p, &sin6);
2846}
2847
2848/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2849void
2850in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2851{
2852	struct sockaddr_in *sin_p;
2853	struct sockaddr_in6 *sin6_p;
2854
2855	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2856	sin_p = (struct sockaddr_in *)*nam;
2857	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2858	free(*nam, M_SONAME);
2859	*nam = (struct sockaddr *)sin6_p;
2860}
2861