in6.c revision 271185
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 271185 2014-09-06 04:39:26Z markj $");
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	IF_AFDATA_LOCK(ifp);
159	ifa->ifa_rtrequest = nd6_rtrequest;
160	ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
161	    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
162	IF_AFDATA_UNLOCK(ifp);
163	if (ln != NULL) {
164		ln->la_expire = 0;  /* for IPv6 this means permanent */
165		ln->ln_state = ND6_LLINFO_REACHABLE;
166		/*
167		 * initialize for rtmsg generation
168		 */
169		bzero(&gateway, sizeof(gateway));
170		gateway.sdl_len = sizeof(gateway);
171		gateway.sdl_family = AF_LINK;
172		gateway.sdl_nlen = 0;
173		gateway.sdl_alen = 6;
174		memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
175		    sizeof(ln->ll_addr));
176		LLE_WUNLOCK(ln);
177	}
178
179	bzero(&rt, sizeof(rt));
180	rt.rt_gateway = (struct sockaddr *)&gateway;
181	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
182	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
183	rt_mask(&rt) = (struct sockaddr *)&mask;
184	rt_key(&rt) = (struct sockaddr *)&addr;
185	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
186	/* Announce arrival of local address to all FIBs. */
187	rt_newaddrmsg(RTM_ADD, ifa, 0, &rt);
188}
189
190void
191in6_ifremloop(struct ifaddr *ifa)
192{
193	struct sockaddr_dl gateway;
194	struct sockaddr_in6 mask, addr;
195	struct rtentry rt0;
196	struct in6_ifaddr *ia;
197	struct ifnet *ifp;
198
199	ia = ifa2ia6(ifa);
200	ifp = ifa->ifa_ifp;
201	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
202	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
203	lltable_prefix_free(AF_INET6, (struct sockaddr *)&addr,
204	            (struct sockaddr *)&mask, LLE_STATIC);
205
206	/*
207	 * initialize for rtmsg generation
208	 */
209	bzero(&gateway, sizeof(gateway));
210	gateway.sdl_len = sizeof(gateway);
211	gateway.sdl_family = AF_LINK;
212	gateway.sdl_nlen = 0;
213	gateway.sdl_alen = ifp->if_addrlen;
214	bzero(&rt0, sizeof(rt0));
215	rt0.rt_gateway = (struct sockaddr *)&gateway;
216	rt_mask(&rt0) = (struct sockaddr *)&mask;
217	rt_key(&rt0) = (struct sockaddr *)&addr;
218	rt0.rt_flags = RTF_HOST | RTF_STATIC;
219	/* Announce removal of local address to all FIBs. */
220	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
221}
222
223int
224in6_mask2len(struct in6_addr *mask, u_char *lim0)
225{
226	int x = 0, y;
227	u_char *lim = lim0, *p;
228
229	/* ignore the scope_id part */
230	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
231		lim = (u_char *)mask + sizeof(*mask);
232	for (p = (u_char *)mask; p < lim; x++, p++) {
233		if (*p != 0xff)
234			break;
235	}
236	y = 0;
237	if (p < lim) {
238		for (y = 0; y < 8; y++) {
239			if ((*p & (0x80 >> y)) == 0)
240				break;
241		}
242	}
243
244	/*
245	 * when the limit pointer is given, do a stricter check on the
246	 * remaining bits.
247	 */
248	if (p < lim) {
249		if (y != 0 && (*p & (0x00ff >> y)) != 0)
250			return (-1);
251		for (p = p + 1; p < lim; p++)
252			if (*p != 0)
253				return (-1);
254	}
255
256	return x * 8 + y;
257}
258
259#ifdef COMPAT_FREEBSD32
260struct in6_ndifreq32 {
261	char ifname[IFNAMSIZ];
262	uint32_t ifindex;
263};
264#define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
265#endif
266
267int
268in6_control(struct socket *so, u_long cmd, caddr_t data,
269    struct ifnet *ifp, struct thread *td)
270{
271	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
272	struct	in6_ifaddr *ia = NULL;
273	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
274	struct sockaddr_in6 *sa6;
275	int carp_attached = 0;
276	int error;
277	u_long ocmd = cmd;
278
279	/*
280	 * Compat to make pre-10.x ifconfig(8) operable.
281	 */
282	if (cmd == OSIOCAIFADDR_IN6)
283		cmd = SIOCAIFADDR_IN6;
284
285	switch (cmd) {
286	case SIOCGETSGCNT_IN6:
287	case SIOCGETMIFCNT_IN6:
288		/*
289		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
290		 * We cannot see how that would be needed, so do not adjust the
291		 * KPI blindly; more likely should clean up the IPv4 variant.
292		 */
293		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
294	}
295
296	switch (cmd) {
297	case SIOCAADDRCTL_POLICY:
298	case SIOCDADDRCTL_POLICY:
299		if (td != NULL) {
300			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
301			if (error)
302				return (error);
303		}
304		return (in6_src_ioctl(cmd, data));
305	}
306
307	if (ifp == NULL)
308		return (EOPNOTSUPP);
309
310	switch (cmd) {
311	case SIOCSNDFLUSH_IN6:
312	case SIOCSPFXFLUSH_IN6:
313	case SIOCSRTRFLUSH_IN6:
314	case SIOCSDEFIFACE_IN6:
315	case SIOCSIFINFO_FLAGS:
316	case SIOCSIFINFO_IN6:
317		if (td != NULL) {
318			error = priv_check(td, PRIV_NETINET_ND6);
319			if (error)
320				return (error);
321		}
322		/* FALLTHROUGH */
323	case OSIOCGIFINFO_IN6:
324	case SIOCGIFINFO_IN6:
325	case SIOCGDRLST_IN6:
326	case SIOCGPRLST_IN6:
327	case SIOCGNBRINFO_IN6:
328	case SIOCGDEFIFACE_IN6:
329		return (nd6_ioctl(cmd, data, ifp));
330
331#ifdef COMPAT_FREEBSD32
332	case SIOCGDEFIFACE32_IN6:
333		{
334			struct in6_ndifreq ndif;
335			struct in6_ndifreq32 *ndif32;
336
337			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
338			    ifp);
339			if (error)
340				return (error);
341			ndif32 = (struct in6_ndifreq32 *)data;
342			ndif32->ifindex = ndif.ifindex;
343			return (0);
344		}
345#endif
346	}
347
348	switch (cmd) {
349	case SIOCSIFPREFIX_IN6:
350	case SIOCDIFPREFIX_IN6:
351	case SIOCAIFPREFIX_IN6:
352	case SIOCCIFPREFIX_IN6:
353	case SIOCSGIFPREFIX_IN6:
354	case SIOCGIFPREFIX_IN6:
355		log(LOG_NOTICE,
356		    "prefix ioctls are now invalidated. "
357		    "please use ifconfig.\n");
358		return (EOPNOTSUPP);
359	}
360
361	switch (cmd) {
362	case SIOCSSCOPE6:
363		if (td != NULL) {
364			error = priv_check(td, PRIV_NETINET_SCOPE6);
365			if (error)
366				return (error);
367		}
368		/* FALLTHROUGH */
369	case SIOCGSCOPE6:
370	case SIOCGSCOPE6DEF:
371		return (scope6_ioctl(cmd, data, ifp));
372	}
373
374	switch (cmd) {
375	case SIOCALIFADDR:
376		if (td != NULL) {
377			error = priv_check(td, PRIV_NET_ADDIFADDR);
378			if (error)
379				return (error);
380		}
381		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
382
383	case SIOCDLIFADDR:
384		if (td != NULL) {
385			error = priv_check(td, PRIV_NET_DELIFADDR);
386			if (error)
387				return (error);
388		}
389		/* FALLTHROUGH */
390	case SIOCGLIFADDR:
391		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
392	}
393
394	/*
395	 * Find address for this interface, if it exists.
396	 *
397	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
398	 * only, and used the first interface address as the target of other
399	 * operations (without checking ifra_addr).  This was because netinet
400	 * code/API assumed at most 1 interface address per interface.
401	 * Since IPv6 allows a node to assign multiple addresses
402	 * on a single interface, we almost always look and check the
403	 * presence of ifra_addr, and reject invalid ones here.
404	 * It also decreases duplicated code among SIOC*_IN6 operations.
405	 */
406	switch (cmd) {
407	case SIOCAIFADDR_IN6:
408	case SIOCSIFPHYADDR_IN6:
409		sa6 = &ifra->ifra_addr;
410		break;
411	case SIOCSIFADDR_IN6:
412	case SIOCGIFADDR_IN6:
413	case SIOCSIFDSTADDR_IN6:
414	case SIOCSIFNETMASK_IN6:
415	case SIOCGIFDSTADDR_IN6:
416	case SIOCGIFNETMASK_IN6:
417	case SIOCDIFADDR_IN6:
418	case SIOCGIFPSRCADDR_IN6:
419	case SIOCGIFPDSTADDR_IN6:
420	case SIOCGIFAFLAG_IN6:
421	case SIOCSNDFLUSH_IN6:
422	case SIOCSPFXFLUSH_IN6:
423	case SIOCSRTRFLUSH_IN6:
424	case SIOCGIFALIFETIME_IN6:
425	case SIOCSIFALIFETIME_IN6:
426	case SIOCGIFSTAT_IN6:
427	case SIOCGIFSTAT_ICMP6:
428		sa6 = &ifr->ifr_addr;
429		break;
430	case SIOCSIFADDR:
431	case SIOCSIFBRDADDR:
432	case SIOCSIFDSTADDR:
433	case SIOCSIFNETMASK:
434		/*
435		 * Although we should pass any non-INET6 ioctl requests
436		 * down to driver, we filter some legacy INET requests.
437		 * Drivers trust SIOCSIFADDR et al to come from an already
438		 * privileged layer, and do not perform any credentials
439		 * checks or input validation.
440		 */
441		return (EINVAL);
442	default:
443		sa6 = NULL;
444		break;
445	}
446	if (sa6 && sa6->sin6_family == AF_INET6) {
447		if (sa6->sin6_scope_id != 0)
448			error = sa6_embedscope(sa6, 0);
449		else
450			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
451		if (error != 0)
452			return (error);
453		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
454		    &sa6->sin6_addr)) != 0)
455			return (error);
456		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
457	} else
458		ia = NULL;
459
460	switch (cmd) {
461	case SIOCSIFADDR_IN6:
462	case SIOCSIFDSTADDR_IN6:
463	case SIOCSIFNETMASK_IN6:
464		/*
465		 * Since IPv6 allows a node to assign multiple addresses
466		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
467		 */
468		/* we decided to obsolete this command (20000704) */
469		error = EINVAL;
470		goto out;
471
472	case SIOCDIFADDR_IN6:
473		/*
474		 * for IPv4, we look for existing in_ifaddr here to allow
475		 * "ifconfig if0 delete" to remove the first IPv4 address on
476		 * the interface.  For IPv6, as the spec allows multiple
477		 * interface address from the day one, we consider "remove the
478		 * first one" semantics to be not preferable.
479		 */
480		if (ia == NULL) {
481			error = EADDRNOTAVAIL;
482			goto out;
483		}
484		/* FALLTHROUGH */
485	case SIOCAIFADDR_IN6:
486		/*
487		 * We always require users to specify a valid IPv6 address for
488		 * the corresponding operation.
489		 */
490		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
491		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
492			error = EAFNOSUPPORT;
493			goto out;
494		}
495
496		if (td != NULL) {
497			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
498			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
499			if (error)
500				goto out;
501		}
502		/* FALLTHROUGH */
503	case SIOCGIFSTAT_IN6:
504	case SIOCGIFSTAT_ICMP6:
505		if (ifp->if_afdata[AF_INET6] == NULL) {
506			error = EPFNOSUPPORT;
507			goto out;
508		}
509		break;
510
511	case SIOCGIFADDR_IN6:
512		/* This interface is basically deprecated. use SIOCGIFCONF. */
513		/* FALLTHROUGH */
514	case SIOCGIFAFLAG_IN6:
515	case SIOCGIFNETMASK_IN6:
516	case SIOCGIFDSTADDR_IN6:
517	case SIOCGIFALIFETIME_IN6:
518		/* must think again about its semantics */
519		if (ia == NULL) {
520			error = EADDRNOTAVAIL;
521			goto out;
522		}
523		break;
524
525	case SIOCSIFALIFETIME_IN6:
526	    {
527		struct in6_addrlifetime *lt;
528
529		if (td != NULL) {
530			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
531			if (error)
532				goto out;
533		}
534		if (ia == NULL) {
535			error = EADDRNOTAVAIL;
536			goto out;
537		}
538		/* sanity for overflow - beware unsigned */
539		lt = &ifr->ifr_ifru.ifru_lifetime;
540		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
541		    lt->ia6t_vltime + time_uptime < time_uptime) {
542			error = EINVAL;
543			goto out;
544		}
545		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
546		    lt->ia6t_pltime + time_uptime < time_uptime) {
547			error = EINVAL;
548			goto out;
549		}
550		break;
551	    }
552	}
553
554	switch (cmd) {
555	case SIOCGIFADDR_IN6:
556		ifr->ifr_addr = ia->ia_addr;
557		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
558			goto out;
559		break;
560
561	case SIOCGIFDSTADDR_IN6:
562		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
563			error = EINVAL;
564			goto out;
565		}
566		/*
567		 * XXX: should we check if ifa_dstaddr is NULL and return
568		 * an error?
569		 */
570		ifr->ifr_dstaddr = ia->ia_dstaddr;
571		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
572			goto out;
573		break;
574
575	case SIOCGIFNETMASK_IN6:
576		ifr->ifr_addr = ia->ia_prefixmask;
577		break;
578
579	case SIOCGIFAFLAG_IN6:
580		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
581		break;
582
583	case SIOCGIFSTAT_IN6:
584		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
585		    ifp->if_afdata[AF_INET6])->in6_ifstat,
586		    &ifr->ifr_ifru.ifru_stat,
587		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
588		break;
589
590	case SIOCGIFSTAT_ICMP6:
591		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
592		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
593		    &ifr->ifr_ifru.ifru_icmp6stat,
594		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
595		break;
596
597	case SIOCGIFALIFETIME_IN6:
598		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
599		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
600			time_t maxexpire;
601			struct in6_addrlifetime *retlt =
602			    &ifr->ifr_ifru.ifru_lifetime;
603
604			/*
605			 * XXX: adjust expiration time assuming time_t is
606			 * signed.
607			 */
608			maxexpire = (-1) &
609			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
610			if (ia->ia6_lifetime.ia6t_vltime <
611			    maxexpire - ia->ia6_updatetime) {
612				retlt->ia6t_expire = ia->ia6_updatetime +
613				    ia->ia6_lifetime.ia6t_vltime;
614			} else
615				retlt->ia6t_expire = maxexpire;
616		}
617		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
618			time_t maxexpire;
619			struct in6_addrlifetime *retlt =
620			    &ifr->ifr_ifru.ifru_lifetime;
621
622			/*
623			 * XXX: adjust expiration time assuming time_t is
624			 * signed.
625			 */
626			maxexpire = (-1) &
627			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
628			if (ia->ia6_lifetime.ia6t_pltime <
629			    maxexpire - ia->ia6_updatetime) {
630				retlt->ia6t_preferred = ia->ia6_updatetime +
631				    ia->ia6_lifetime.ia6t_pltime;
632			} else
633				retlt->ia6t_preferred = maxexpire;
634		}
635		break;
636
637	case SIOCSIFALIFETIME_IN6:
638		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
639		/* for sanity */
640		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
641			ia->ia6_lifetime.ia6t_expire =
642				time_uptime + ia->ia6_lifetime.ia6t_vltime;
643		} else
644			ia->ia6_lifetime.ia6t_expire = 0;
645		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
646			ia->ia6_lifetime.ia6t_preferred =
647				time_uptime + ia->ia6_lifetime.ia6t_pltime;
648		} else
649			ia->ia6_lifetime.ia6t_preferred = 0;
650		break;
651
652	case SIOCAIFADDR_IN6:
653	{
654		int i;
655		struct nd_prefixctl pr0;
656		struct nd_prefix *pr;
657
658		/*
659		 * first, make or update the interface address structure,
660		 * and link it to the list.
661		 */
662		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
663			goto out;
664		if (ia != NULL)
665			ifa_free(&ia->ia_ifa);
666		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
667		    == NULL) {
668			/*
669			 * this can happen when the user specify the 0 valid
670			 * lifetime.
671			 */
672			break;
673		}
674
675		if (cmd == ocmd && ifra->ifra_vhid > 0) {
676			if (carp_attach_p != NULL)
677				error = (*carp_attach_p)(&ia->ia_ifa,
678				    ifra->ifra_vhid);
679			else
680				error = EPROTONOSUPPORT;
681			if (error)
682				goto out;
683			else
684				carp_attached = 1;
685		}
686
687		/*
688		 * then, make the prefix on-link on the interface.
689		 * XXX: we'd rather create the prefix before the address, but
690		 * we need at least one address to install the corresponding
691		 * interface route, so we configure the address first.
692		 */
693
694		/*
695		 * convert mask to prefix length (prefixmask has already
696		 * been validated in in6_update_ifa().
697		 */
698		bzero(&pr0, sizeof(pr0));
699		pr0.ndpr_ifp = ifp;
700		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
701		    NULL);
702		if (pr0.ndpr_plen == 128) {
703			break;	/* we don't need to install a host route. */
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			if (pr == NULL) {
736				if (carp_attached)
737					(*carp_detach_p)(&ia->ia_ifa);
738				log(LOG_ERR, "nd6_prelist_add succeeded but "
739				    "no prefix\n");
740				error = EINVAL;
741				goto out;
742			}
743		}
744
745		/* relate the address to the prefix */
746		if (ia->ia6_ndpr == NULL) {
747			ia->ia6_ndpr = pr;
748			pr->ndpr_refcnt++;
749
750			/*
751			 * If this is the first autoconf address from the
752			 * prefix, create a temporary address as well
753			 * (when required).
754			 */
755			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
756			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
757				int e;
758				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
759					log(LOG_NOTICE, "in6_control: failed "
760					    "to create a temporary address, "
761					    "errno=%d\n", e);
762				}
763			}
764		}
765
766		/*
767		 * this might affect the status of autoconfigured addresses,
768		 * that is, this address might make other addresses detached.
769		 */
770		pfxlist_onlink_check();
771		if (error == 0 && ia) {
772			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
773				/*
774				 * Try to clear the flag when a new
775				 * IPv6 address is added onto an
776				 * IFDISABLED interface and it
777				 * succeeds.
778				 */
779				struct in6_ndireq nd;
780
781				memset(&nd, 0, sizeof(nd));
782				nd.ndi.flags = ND_IFINFO(ifp)->flags;
783				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
784				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
785				    (caddr_t)&nd, ifp) < 0)
786					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
787					    "SIOCSIFINFO_FLAGS for -ifdisabled "
788					    "failed.");
789				/*
790				 * Ignore failure of clearing the flag
791				 * intentionally.  The failure means
792				 * address duplication was detected.
793				 */
794			}
795			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
796		}
797		break;
798	}
799
800	case SIOCDIFADDR_IN6:
801	{
802		struct nd_prefix *pr;
803
804		/*
805		 * If the address being deleted is the only one that owns
806		 * the corresponding prefix, expire the prefix as well.
807		 * XXX: theoretically, we don't have to worry about such
808		 * relationship, since we separate the address management
809		 * and the prefix management.  We do this, however, to provide
810		 * as much backward compatibility as possible in terms of
811		 * the ioctl operation.
812		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
813		 */
814		pr = ia->ia6_ndpr;
815		in6_purgeaddr(&ia->ia_ifa);
816		if (pr && pr->ndpr_refcnt == 0)
817			prelist_remove(pr);
818		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
819		break;
820	}
821
822	default:
823		if (ifp->if_ioctl == NULL) {
824			error = EOPNOTSUPP;
825			goto out;
826		}
827		error = (*ifp->if_ioctl)(ifp, cmd, data);
828		goto out;
829	}
830
831	error = 0;
832out:
833	if (ia != NULL)
834		ifa_free(&ia->ia_ifa);
835	return (error);
836}
837
838
839/*
840 * Join necessary multicast groups.  Factored out from in6_update_ifa().
841 * This entire work should only be done once, for the default FIB.
842 */
843static int
844in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
845    struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
846{
847	char ip6buf[INET6_ADDRSTRLEN];
848	struct sockaddr_in6 mltaddr, mltmask;
849	struct in6_addr llsol;
850	struct in6_multi_mship *imm;
851	struct rtentry *rt;
852	int delay, error;
853
854	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
855
856	/* Join solicited multicast addr for new host id. */
857	bzero(&llsol, sizeof(struct in6_addr));
858	llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
859	llsol.s6_addr32[1] = 0;
860	llsol.s6_addr32[2] = htonl(1);
861	llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
862	llsol.s6_addr8[12] = 0xff;
863	if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
864		/* XXX: should not happen */
865		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
866		goto cleanup;
867	}
868	delay = 0;
869	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
870		/*
871		 * We need a random delay for DAD on the address being
872		 * configured.  It also means delaying transmission of the
873		 * corresponding MLD report to avoid report collision.
874		 * [RFC 4861, Section 6.3.7]
875		 */
876		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
877	}
878	imm = in6_joingroup(ifp, &llsol, &error, delay);
879	if (imm == NULL) {
880		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
881		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &llsol),
882		    if_name(ifp), error));
883		goto cleanup;
884	}
885	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
886	*in6m_sol = imm->i6mm_maddr;
887
888	bzero(&mltmask, sizeof(mltmask));
889	mltmask.sin6_len = sizeof(struct sockaddr_in6);
890	mltmask.sin6_family = AF_INET6;
891	mltmask.sin6_addr = in6mask32;
892#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
893
894	/*
895	 * Join link-local all-nodes address.
896	 */
897	bzero(&mltaddr, sizeof(mltaddr));
898	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
899	mltaddr.sin6_family = AF_INET6;
900	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
901	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
902		goto cleanup; /* XXX: should not fail */
903
904	/*
905	 * XXX: do we really need this automatic routes?  We should probably
906	 * reconsider this stuff.  Most applications actually do not need the
907	 * routes, since they usually specify the outgoing interface.
908	 */
909	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
910	if (rt != NULL) {
911		/* XXX: only works in !SCOPEDROUTING case. */
912		if (memcmp(&mltaddr.sin6_addr,
913		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
914		    MLTMASK_LEN)) {
915			RTFREE_LOCKED(rt);
916			rt = NULL;
917		}
918	}
919	if (rt == NULL) {
920		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
921		    (struct sockaddr *)&ia->ia_addr,
922		    (struct sockaddr *)&mltmask, RTF_UP,
923		    (struct rtentry **)0, RT_DEFAULT_FIB);
924		if (error)
925			goto cleanup;
926	} else
927		RTFREE_LOCKED(rt);
928
929	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
930	if (imm == NULL) {
931		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
932		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
933		    &mltaddr.sin6_addr), if_name(ifp), error));
934		goto cleanup;
935	}
936	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
937
938	/*
939	 * Join node information group address.
940	 */
941	delay = 0;
942	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
943		/*
944		 * The spec does not say anything about delay for this group,
945		 * but the same logic should apply.
946		 */
947		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
948	}
949	if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
950		/* XXX jinmei */
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	if (V_icmp6_nodeinfo_oldmcprefix &&
961	     in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
962		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
963		if (imm == NULL)
964			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
965			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
966			    &mltaddr.sin6_addr), if_name(ifp), error));
967			/* XXX not very fatal, go on... */
968		else
969			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
970	}
971
972	/*
973	 * Join interface-local all-nodes address.
974	 * (ff01::1%ifN, and ff01::%ifN/32)
975	 */
976	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
977	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
978		goto cleanup; /* XXX: should not fail */
979	/* XXX: again, do we really need the route? */
980	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
981	if (rt != NULL) {
982		if (memcmp(&mltaddr.sin6_addr,
983		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
984		    MLTMASK_LEN)) {
985			RTFREE_LOCKED(rt);
986			rt = NULL;
987		}
988	}
989	if (rt == NULL) {
990		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
991		    (struct sockaddr *)&ia->ia_addr,
992		    (struct sockaddr *)&mltmask, RTF_UP,
993		    (struct rtentry **)0, RT_DEFAULT_FIB);
994		if (error)
995			goto cleanup;
996	} else
997		RTFREE_LOCKED(rt);
998
999	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1000	if (imm == NULL) {
1001		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
1002		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
1003		    &mltaddr.sin6_addr), if_name(ifp), error));
1004		goto cleanup;
1005	}
1006	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1007#undef	MLTMASK_LEN
1008
1009cleanup:
1010	return (error);
1011}
1012
1013/*
1014 * Update parameters of an IPv6 interface address.
1015 * If necessary, a new entry is created and linked into address chains.
1016 * This function is separated from in6_control().
1017 */
1018int
1019in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1020    struct in6_ifaddr *ia, int flags)
1021{
1022	int error = 0, hostIsNew = 0, plen = -1;
1023	struct sockaddr_in6 dst6;
1024	struct in6_addrlifetime *lt;
1025	struct in6_multi *in6m_sol;
1026	int delay;
1027	char ip6buf[INET6_ADDRSTRLEN];
1028
1029	/* Validate parameters */
1030	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1031		return (EINVAL);
1032
1033	/*
1034	 * The destination address for a p2p link must have a family
1035	 * of AF_UNSPEC or AF_INET6.
1036	 */
1037	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1038	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1039	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1040		return (EAFNOSUPPORT);
1041	/*
1042	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1043	 * does not carry fields other than sin6_len.
1044	 */
1045	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1046		return (EINVAL);
1047	/*
1048	 * Because the IPv6 address architecture is classless, we require
1049	 * users to specify a (non 0) prefix length (mask) for a new address.
1050	 * We also require the prefix (when specified) mask is valid, and thus
1051	 * reject a non-consecutive mask.
1052	 */
1053	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1054		return (EINVAL);
1055	if (ifra->ifra_prefixmask.sin6_len != 0) {
1056		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1057		    (u_char *)&ifra->ifra_prefixmask +
1058		    ifra->ifra_prefixmask.sin6_len);
1059		if (plen <= 0)
1060			return (EINVAL);
1061	} else {
1062		/*
1063		 * In this case, ia must not be NULL.  We just use its prefix
1064		 * length.
1065		 */
1066		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1067	}
1068	/*
1069	 * If the destination address on a p2p interface is specified,
1070	 * and the address is a scoped one, validate/set the scope
1071	 * zone identifier.
1072	 */
1073	dst6 = ifra->ifra_dstaddr;
1074	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1075	    (dst6.sin6_family == AF_INET6)) {
1076		struct in6_addr in6_tmp;
1077		u_int32_t zoneid;
1078
1079		in6_tmp = dst6.sin6_addr;
1080		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1081			return (EINVAL); /* XXX: should be impossible */
1082
1083		if (dst6.sin6_scope_id != 0) {
1084			if (dst6.sin6_scope_id != zoneid)
1085				return (EINVAL);
1086		} else		/* user omit to specify the ID. */
1087			dst6.sin6_scope_id = zoneid;
1088
1089		/* convert into the internal form */
1090		if (sa6_embedscope(&dst6, 0))
1091			return (EINVAL); /* XXX: should be impossible */
1092	}
1093	/*
1094	 * The destination address can be specified only for a p2p or a
1095	 * loopback interface.  If specified, the corresponding prefix length
1096	 * must be 128.
1097	 */
1098	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1099		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1100			/* XXX: noisy message */
1101			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1102			    "be specified for a p2p or a loopback IF only\n"));
1103			return (EINVAL);
1104		}
1105		if (plen != 128) {
1106			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1107			    "be 128 when dstaddr is specified\n"));
1108			return (EINVAL);
1109		}
1110	}
1111	/* lifetime consistency check */
1112	lt = &ifra->ifra_lifetime;
1113	if (lt->ia6t_pltime > lt->ia6t_vltime)
1114		return (EINVAL);
1115	if (lt->ia6t_vltime == 0) {
1116		/*
1117		 * the following log might be noisy, but this is a typical
1118		 * configuration mistake or a tool's bug.
1119		 */
1120		nd6log((LOG_INFO,
1121		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1122		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1123
1124		if (ia == NULL)
1125			return (0); /* there's nothing to do */
1126	}
1127
1128	/*
1129	 * If this is a new address, allocate a new ifaddr and link it
1130	 * into chains.
1131	 */
1132	if (ia == NULL) {
1133		hostIsNew = 1;
1134		/*
1135		 * When in6_update_ifa() is called in a process of a received
1136		 * RA, it is called under an interrupt context.  So, we should
1137		 * call malloc with M_NOWAIT.
1138		 */
1139		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
1140		    M_NOWAIT);
1141		if (ia == NULL)
1142			return (ENOBUFS);
1143		bzero((caddr_t)ia, sizeof(*ia));
1144		ifa_init(&ia->ia_ifa);
1145		LIST_INIT(&ia->ia6_memberships);
1146		/* Initialize the address and masks, and put time stamp */
1147		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1148		ia->ia_addr.sin6_family = AF_INET6;
1149		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1150		ia->ia6_createtime = time_uptime;
1151		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1152			/*
1153			 * XXX: some functions expect that ifa_dstaddr is not
1154			 * NULL for p2p interfaces.
1155			 */
1156			ia->ia_ifa.ifa_dstaddr =
1157			    (struct sockaddr *)&ia->ia_dstaddr;
1158		} else {
1159			ia->ia_ifa.ifa_dstaddr = NULL;
1160		}
1161		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1162		ia->ia_ifp = ifp;
1163		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1164		IF_ADDR_WLOCK(ifp);
1165		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1166		IF_ADDR_WUNLOCK(ifp);
1167
1168		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1169		IN6_IFADDR_WLOCK();
1170		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1171		LIST_INSERT_HEAD(IN6ADDR_HASH(&ifra->ifra_addr.sin6_addr),
1172		    ia, ia6_hash);
1173		IN6_IFADDR_WUNLOCK();
1174	}
1175
1176	/* update timestamp */
1177	ia->ia6_updatetime = time_uptime;
1178
1179	/* set prefix mask */
1180	if (ifra->ifra_prefixmask.sin6_len) {
1181		/*
1182		 * We prohibit changing the prefix length of an existing
1183		 * address, because
1184		 * + such an operation should be rare in IPv6, and
1185		 * + the operation would confuse prefix management.
1186		 */
1187		if (ia->ia_prefixmask.sin6_len &&
1188		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1189			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1190			    " existing (%s) address should not be changed\n",
1191			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1192			error = EINVAL;
1193			goto unlink;
1194		}
1195		ia->ia_prefixmask = ifra->ifra_prefixmask;
1196		ia->ia_prefixmask.sin6_family = AF_INET6;
1197	}
1198
1199	/*
1200	 * If a new destination address is specified, scrub the old one and
1201	 * install the new destination.  Note that the interface must be
1202	 * p2p or loopback (see the check above.)
1203	 */
1204	if (dst6.sin6_family == AF_INET6 &&
1205	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1206		int e;
1207
1208		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1209		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1210			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1211			    "a route to the old destination: %s\n",
1212			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1213			/* proceed anyway... */
1214		} else
1215			ia->ia_flags &= ~IFA_ROUTE;
1216		ia->ia_dstaddr = dst6;
1217	}
1218
1219	/*
1220	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1221	 * to see if the address is deprecated or invalidated, but initialize
1222	 * these members for applications.
1223	 */
1224	ia->ia6_lifetime = ifra->ifra_lifetime;
1225	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1226		ia->ia6_lifetime.ia6t_expire =
1227		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1228	} else
1229		ia->ia6_lifetime.ia6t_expire = 0;
1230	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1231		ia->ia6_lifetime.ia6t_preferred =
1232		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1233	} else
1234		ia->ia6_lifetime.ia6t_preferred = 0;
1235
1236	/* reset the interface and routing table appropriately. */
1237	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1238		goto unlink;
1239
1240	/*
1241	 * configure address flags.
1242	 */
1243	ia->ia6_flags = ifra->ifra_flags;
1244	/*
1245	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1246	 * userland, make it deprecated.
1247	 */
1248	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1249		ia->ia6_lifetime.ia6t_pltime = 0;
1250		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1251	}
1252	/*
1253	 * Make the address tentative before joining multicast addresses,
1254	 * so that corresponding MLD responses would not have a tentative
1255	 * source address.
1256	 */
1257	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1258	if (hostIsNew && in6if_do_dad(ifp))
1259		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1260
1261	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
1262	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1263		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1264
1265	/*
1266	 * We are done if we have simply modified an existing address.
1267	 */
1268	if (!hostIsNew)
1269		return (error);
1270
1271	/*
1272	 * Beyond this point, we should call in6_purgeaddr upon an error,
1273	 * not just go to unlink.
1274	 */
1275
1276	/* Join necessary multicast groups. */
1277	in6m_sol = NULL;
1278	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1279		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1280		if (error)
1281			goto cleanup;
1282	}
1283
1284	/*
1285	 * Perform DAD, if needed.
1286	 * XXX It may be of use, if we can administratively disable DAD.
1287	 */
1288	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1289	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1290	{
1291		int mindelay, maxdelay;
1292
1293		delay = 0;
1294		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1295			/*
1296			 * We need to impose a delay before sending an NS
1297			 * for DAD.  Check if we also needed a delay for the
1298			 * corresponding MLD message.  If we did, the delay
1299			 * should be larger than the MLD delay (this could be
1300			 * relaxed a bit, but this simple logic is at least
1301			 * safe).
1302			 * XXX: Break data hiding guidelines and look at
1303			 * state for the solicited multicast group.
1304			 */
1305			mindelay = 0;
1306			if (in6m_sol != NULL &&
1307			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1308				mindelay = in6m_sol->in6m_timer;
1309			}
1310			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1311			if (maxdelay - mindelay == 0)
1312				delay = 0;
1313			else {
1314				delay =
1315				    (arc4random() % (maxdelay - mindelay)) +
1316				    mindelay;
1317			}
1318		}
1319		nd6_dad_start((struct ifaddr *)ia, delay);
1320	}
1321
1322	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1323	ifa_free(&ia->ia_ifa);
1324	return (error);
1325
1326  unlink:
1327	/*
1328	 * XXX: if a change of an existing address failed, keep the entry
1329	 * anyway.
1330	 */
1331	if (hostIsNew) {
1332		in6_unlink_ifa(ia, ifp);
1333		ifa_free(&ia->ia_ifa);
1334	}
1335	return (error);
1336
1337  cleanup:
1338	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1339	ifa_free(&ia->ia_ifa);
1340	in6_purgeaddr(&ia->ia_ifa);
1341	return error;
1342}
1343
1344/*
1345 * Leave multicast groups.  Factored out from in6_purgeaddr().
1346 * This entire work should only be done once, for the default FIB.
1347 */
1348static int
1349in6_purgeaddr_mc(struct ifnet *ifp, struct in6_ifaddr *ia, struct ifaddr *ifa0)
1350{
1351	struct sockaddr_in6 mltaddr, mltmask;
1352	struct in6_multi_mship *imm;
1353	struct rtentry *rt;
1354	struct sockaddr_in6 sin6;
1355	int error;
1356
1357	/*
1358	 * Leave from multicast groups we have joined for the interface.
1359	 */
1360	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1361		LIST_REMOVE(imm, i6mm_chain);
1362		in6_leavegroup(imm);
1363	}
1364
1365	/*
1366	 * Remove the link-local all-nodes address.
1367	 */
1368	bzero(&mltmask, sizeof(mltmask));
1369	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1370	mltmask.sin6_family = AF_INET6;
1371	mltmask.sin6_addr = in6mask32;
1372
1373	bzero(&mltaddr, sizeof(mltaddr));
1374	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1375	mltaddr.sin6_family = AF_INET6;
1376	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1377
1378	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1379		return (error);
1380
1381	/*
1382	 * As for the mltaddr above, proactively prepare the sin6 to avoid
1383	 * rtentry un- and re-locking.
1384	 */
1385	if (ifa0 != NULL) {
1386		bzero(&sin6, sizeof(sin6));
1387		sin6.sin6_len = sizeof(sin6);
1388		sin6.sin6_family = AF_INET6;
1389		memcpy(&sin6.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1390		    sizeof(sin6.sin6_addr));
1391		error = in6_setscope(&sin6.sin6_addr, ifa0->ifa_ifp, NULL);
1392		if (error != 0)
1393			return (error);
1394	}
1395
1396	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1397	if (rt != NULL && rt->rt_gateway != NULL &&
1398	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1399		    &ia->ia_addr.sin6_addr,
1400		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1401		/*
1402		 * If no more IPv6 address exists on this interface then
1403		 * remove the multicast address route.
1404		 */
1405		if (ifa0 == NULL) {
1406			memcpy(&mltaddr.sin6_addr,
1407			    &satosin6(rt_key(rt))->sin6_addr,
1408			    sizeof(mltaddr.sin6_addr));
1409			RTFREE_LOCKED(rt);
1410			error = in6_rtrequest(RTM_DELETE,
1411			    (struct sockaddr *)&mltaddr,
1412			    (struct sockaddr *)&ia->ia_addr,
1413			    (struct sockaddr *)&mltmask, RTF_UP,
1414			    (struct rtentry **)0, RT_DEFAULT_FIB);
1415			if (error)
1416				log(LOG_INFO, "%s: link-local all-nodes "
1417				    "multicast address deletion error\n",
1418				    __func__);
1419		} else {
1420			/*
1421			 * Replace the gateway of the route.
1422			 */
1423			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1424			RTFREE_LOCKED(rt);
1425		}
1426	} else {
1427		if (rt != NULL)
1428			RTFREE_LOCKED(rt);
1429	}
1430
1431	/*
1432	 * Remove the node-local all-nodes address.
1433	 */
1434	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1435	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1436		return (error);
1437
1438	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1439	if (rt != NULL && rt->rt_gateway != NULL &&
1440	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1441		    &ia->ia_addr.sin6_addr,
1442		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1443		/*
1444		 * If no more IPv6 address exists on this interface then
1445		 * remove the multicast address route.
1446		 */
1447		if (ifa0 == NULL) {
1448			memcpy(&mltaddr.sin6_addr,
1449			    &satosin6(rt_key(rt))->sin6_addr,
1450			    sizeof(mltaddr.sin6_addr));
1451
1452			RTFREE_LOCKED(rt);
1453			error = in6_rtrequest(RTM_DELETE,
1454			    (struct sockaddr *)&mltaddr,
1455			    (struct sockaddr *)&ia->ia_addr,
1456			    (struct sockaddr *)&mltmask, RTF_UP,
1457			    (struct rtentry **)0, RT_DEFAULT_FIB);
1458			if (error)
1459				log(LOG_INFO, "%s: node-local all-nodes"
1460				    "multicast address deletion error\n",
1461				    __func__);
1462		} else {
1463			/*
1464			 * Replace the gateway of the route.
1465			 */
1466			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1467			RTFREE_LOCKED(rt);
1468		}
1469	} else {
1470		if (rt != NULL)
1471			RTFREE_LOCKED(rt);
1472	}
1473
1474	return (0);
1475}
1476
1477void
1478in6_purgeaddr(struct ifaddr *ifa)
1479{
1480	struct ifnet *ifp = ifa->ifa_ifp;
1481	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1482	int plen, error;
1483	struct ifaddr *ifa0;
1484
1485	if (ifa->ifa_carp)
1486		(*carp_detach_p)(ifa);
1487
1488	/*
1489	 * find another IPv6 address as the gateway for the
1490	 * link-local and node-local all-nodes multicast
1491	 * address routes
1492	 */
1493	IF_ADDR_RLOCK(ifp);
1494	TAILQ_FOREACH(ifa0, &ifp->if_addrhead, ifa_link) {
1495		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1496		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1497		    &ia->ia_addr.sin6_addr, sizeof(struct in6_addr)) == 0)
1498			continue;
1499		else
1500			break;
1501	}
1502	if (ifa0 != NULL)
1503		ifa_ref(ifa0);
1504	IF_ADDR_RUNLOCK(ifp);
1505
1506	/*
1507	 * Remove the loopback route to the interface address.
1508	 * The check for the current setting of "nd6_useloopback"
1509	 * is not needed.
1510	 */
1511	if (ia->ia_flags & IFA_RTSELF) {
1512		error = ifa_del_loopback_route((struct ifaddr *)ia,
1513		    (struct sockaddr *)&ia->ia_addr);
1514		if (error == 0)
1515			ia->ia_flags &= ~IFA_RTSELF;
1516	}
1517
1518	/* stop DAD processing */
1519	nd6_dad_stop(ifa);
1520
1521	/* Remove local address entry from lltable. */
1522	in6_ifremloop(ifa);
1523
1524	/* Leave multicast groups. */
1525	error = in6_purgeaddr_mc(ifp, ia, ifa0);
1526
1527	if (ifa0 != NULL)
1528		ifa_free(ifa0);
1529
1530	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1531	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1532		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1533		    (ia->ia_dstaddr.sin6_family == AF_INET6) ? RTF_HOST : 0);
1534		if (error != 0)
1535			log(LOG_INFO, "%s: err=%d, destination address delete "
1536			    "failed\n", __func__, error);
1537		ia->ia_flags &= ~IFA_ROUTE;
1538	}
1539
1540	in6_unlink_ifa(ia, ifp);
1541}
1542
1543static void
1544in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1545{
1546
1547	IF_ADDR_WLOCK(ifp);
1548	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1549	IF_ADDR_WUNLOCK(ifp);
1550	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1551
1552	/*
1553	 * Defer the release of what might be the last reference to the
1554	 * in6_ifaddr so that it can't be freed before the remainder of the
1555	 * cleanup.
1556	 */
1557	IN6_IFADDR_WLOCK();
1558	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1559	LIST_REMOVE(ia, ia6_hash);
1560	IN6_IFADDR_WUNLOCK();
1561
1562	/*
1563	 * Release the reference to the base prefix.  There should be a
1564	 * positive reference.
1565	 */
1566	if (ia->ia6_ndpr == NULL) {
1567		nd6log((LOG_NOTICE,
1568		    "in6_unlink_ifa: autoconf'ed address "
1569		    "%p has no prefix\n", ia));
1570	} else {
1571		ia->ia6_ndpr->ndpr_refcnt--;
1572		ia->ia6_ndpr = NULL;
1573	}
1574
1575	/*
1576	 * Also, if the address being removed is autoconf'ed, call
1577	 * pfxlist_onlink_check() since the release might affect the status of
1578	 * other (detached) addresses.
1579	 */
1580	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1581		pfxlist_onlink_check();
1582	}
1583	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1584}
1585
1586void
1587in6_purgeif(struct ifnet *ifp)
1588{
1589	struct ifaddr *ifa, *nifa;
1590
1591	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1592		if (ifa->ifa_addr->sa_family != AF_INET6)
1593			continue;
1594		in6_purgeaddr(ifa);
1595	}
1596
1597	in6_ifdetach(ifp);
1598}
1599
1600/*
1601 * SIOC[GAD]LIFADDR.
1602 *	SIOCGLIFADDR: get first address. (?)
1603 *	SIOCGLIFADDR with IFLR_PREFIX:
1604 *		get first address that matches the specified prefix.
1605 *	SIOCALIFADDR: add the specified address.
1606 *	SIOCALIFADDR with IFLR_PREFIX:
1607 *		add the specified prefix, filling hostid part from
1608 *		the first link-local address.  prefixlen must be <= 64.
1609 *	SIOCDLIFADDR: delete the specified address.
1610 *	SIOCDLIFADDR with IFLR_PREFIX:
1611 *		delete the first address that matches the specified prefix.
1612 * return values:
1613 *	EINVAL on invalid parameters
1614 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1615 *	other values may be returned from in6_ioctl()
1616 *
1617 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1618 * this is to accomodate address naming scheme other than RFC2374,
1619 * in the future.
1620 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1621 * address encoding scheme. (see figure on page 8)
1622 */
1623static int
1624in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1625    struct ifnet *ifp, struct thread *td)
1626{
1627	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1628	struct ifaddr *ifa;
1629	struct sockaddr *sa;
1630
1631	/* sanity checks */
1632	if (!data || !ifp) {
1633		panic("invalid argument to in6_lifaddr_ioctl");
1634		/* NOTREACHED */
1635	}
1636
1637	switch (cmd) {
1638	case SIOCGLIFADDR:
1639		/* address must be specified on GET with IFLR_PREFIX */
1640		if ((iflr->flags & IFLR_PREFIX) == 0)
1641			break;
1642		/* FALLTHROUGH */
1643	case SIOCALIFADDR:
1644	case SIOCDLIFADDR:
1645		/* address must be specified on ADD and DELETE */
1646		sa = (struct sockaddr *)&iflr->addr;
1647		if (sa->sa_family != AF_INET6)
1648			return EINVAL;
1649		if (sa->sa_len != sizeof(struct sockaddr_in6))
1650			return EINVAL;
1651		/* XXX need improvement */
1652		sa = (struct sockaddr *)&iflr->dstaddr;
1653		if (sa->sa_family && sa->sa_family != AF_INET6)
1654			return EINVAL;
1655		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1656			return EINVAL;
1657		break;
1658	default: /* shouldn't happen */
1659#if 0
1660		panic("invalid cmd to in6_lifaddr_ioctl");
1661		/* NOTREACHED */
1662#else
1663		return EOPNOTSUPP;
1664#endif
1665	}
1666	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1667		return EINVAL;
1668
1669	switch (cmd) {
1670	case SIOCALIFADDR:
1671	    {
1672		struct in6_aliasreq ifra;
1673		struct in6_addr *hostid = NULL;
1674		int prefixlen;
1675
1676		ifa = NULL;
1677		if ((iflr->flags & IFLR_PREFIX) != 0) {
1678			struct sockaddr_in6 *sin6;
1679
1680			/*
1681			 * hostid is to fill in the hostid part of the
1682			 * address.  hostid points to the first link-local
1683			 * address attached to the interface.
1684			 */
1685			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1686			if (!ifa)
1687				return EADDRNOTAVAIL;
1688			hostid = IFA_IN6(ifa);
1689
1690			/* prefixlen must be <= 64. */
1691			if (64 < iflr->prefixlen) {
1692				if (ifa != NULL)
1693					ifa_free(ifa);
1694				return EINVAL;
1695			}
1696			prefixlen = iflr->prefixlen;
1697
1698			/* hostid part must be zero. */
1699			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1700			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1701			    sin6->sin6_addr.s6_addr32[3] != 0) {
1702				if (ifa != NULL)
1703					ifa_free(ifa);
1704				return EINVAL;
1705			}
1706		} else
1707			prefixlen = iflr->prefixlen;
1708
1709		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1710		bzero(&ifra, sizeof(ifra));
1711		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1712
1713		bcopy(&iflr->addr, &ifra.ifra_addr,
1714		    ((struct sockaddr *)&iflr->addr)->sa_len);
1715		if (hostid) {
1716			/* fill in hostid part */
1717			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1718			    hostid->s6_addr32[2];
1719			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1720			    hostid->s6_addr32[3];
1721		}
1722
1723		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1724			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1725			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1726			if (hostid) {
1727				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1728				    hostid->s6_addr32[2];
1729				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1730				    hostid->s6_addr32[3];
1731			}
1732		}
1733		if (ifa != NULL)
1734			ifa_free(ifa);
1735
1736		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1737		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1738
1739		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1740		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1741	    }
1742	case SIOCGLIFADDR:
1743	case SIOCDLIFADDR:
1744	    {
1745		struct in6_ifaddr *ia;
1746		struct in6_addr mask, candidate, match;
1747		struct sockaddr_in6 *sin6;
1748		int cmp;
1749
1750		bzero(&mask, sizeof(mask));
1751		if (iflr->flags & IFLR_PREFIX) {
1752			/* lookup a prefix rather than address. */
1753			in6_prefixlen2mask(&mask, iflr->prefixlen);
1754
1755			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1756			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1757			match.s6_addr32[0] &= mask.s6_addr32[0];
1758			match.s6_addr32[1] &= mask.s6_addr32[1];
1759			match.s6_addr32[2] &= mask.s6_addr32[2];
1760			match.s6_addr32[3] &= mask.s6_addr32[3];
1761
1762			/* if you set extra bits, that's wrong */
1763			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1764				return EINVAL;
1765
1766			cmp = 1;
1767		} else {
1768			if (cmd == SIOCGLIFADDR) {
1769				/* on getting an address, take the 1st match */
1770				cmp = 0;	/* XXX */
1771			} else {
1772				/* on deleting an address, do exact match */
1773				in6_prefixlen2mask(&mask, 128);
1774				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1775				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1776
1777				cmp = 1;
1778			}
1779		}
1780
1781		IF_ADDR_RLOCK(ifp);
1782		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1783			if (ifa->ifa_addr->sa_family != AF_INET6)
1784				continue;
1785			if (!cmp)
1786				break;
1787
1788			/*
1789			 * XXX: this is adhoc, but is necessary to allow
1790			 * a user to specify fe80::/64 (not /10) for a
1791			 * link-local address.
1792			 */
1793			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1794			in6_clearscope(&candidate);
1795			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1796			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1797			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1798			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1799			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1800				break;
1801		}
1802		if (ifa != NULL)
1803			ifa_ref(ifa);
1804		IF_ADDR_RUNLOCK(ifp);
1805		if (!ifa)
1806			return EADDRNOTAVAIL;
1807		ia = ifa2ia6(ifa);
1808
1809		if (cmd == SIOCGLIFADDR) {
1810			int error;
1811
1812			/* fill in the if_laddrreq structure */
1813			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1814			error = sa6_recoverscope(
1815			    (struct sockaddr_in6 *)&iflr->addr);
1816			if (error != 0) {
1817				ifa_free(ifa);
1818				return (error);
1819			}
1820
1821			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1822				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1823				    ia->ia_dstaddr.sin6_len);
1824				error = sa6_recoverscope(
1825				    (struct sockaddr_in6 *)&iflr->dstaddr);
1826				if (error != 0) {
1827					ifa_free(ifa);
1828					return (error);
1829				}
1830			} else
1831				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1832
1833			iflr->prefixlen =
1834			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1835
1836			iflr->flags = ia->ia6_flags;	/* XXX */
1837			ifa_free(ifa);
1838
1839			return 0;
1840		} else {
1841			struct in6_aliasreq ifra;
1842
1843			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1844			bzero(&ifra, sizeof(ifra));
1845			bcopy(iflr->iflr_name, ifra.ifra_name,
1846			    sizeof(ifra.ifra_name));
1847
1848			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1849			    ia->ia_addr.sin6_len);
1850			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1851				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1852				    ia->ia_dstaddr.sin6_len);
1853			} else {
1854				bzero(&ifra.ifra_dstaddr,
1855				    sizeof(ifra.ifra_dstaddr));
1856			}
1857			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1858			    ia->ia_prefixmask.sin6_len);
1859
1860			ifra.ifra_flags = ia->ia6_flags;
1861			ifa_free(ifa);
1862			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1863			    ifp, td);
1864		}
1865	    }
1866	}
1867
1868	return EOPNOTSUPP;	/* just for safety */
1869}
1870
1871/*
1872 * Initialize an interface's IPv6 address and routing table entry.
1873 */
1874static int
1875in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1876    struct sockaddr_in6 *sin6, int newhost)
1877{
1878	int	error = 0, plen, ifacount = 0;
1879	struct ifaddr *ifa;
1880
1881	/*
1882	 * Give the interface a chance to initialize
1883	 * if this is its first address,
1884	 * and to validate the address if necessary.
1885	 */
1886	IF_ADDR_RLOCK(ifp);
1887	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1888		if (ifa->ifa_addr->sa_family != AF_INET6)
1889			continue;
1890		ifacount++;
1891	}
1892	IF_ADDR_RUNLOCK(ifp);
1893
1894	ia->ia_addr = *sin6;
1895
1896	if (ifacount <= 1 && ifp->if_ioctl) {
1897		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1898		if (error)
1899			return (error);
1900	}
1901
1902	ia->ia_ifa.ifa_metric = ifp->if_metric;
1903
1904	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1905
1906	/*
1907	 * Special case:
1908	 * If a new destination address is specified for a point-to-point
1909	 * interface, install a route to the destination as an interface
1910	 * direct route.
1911	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1912	 * interface that share the same destination.
1913	 */
1914	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1915	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1916	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1917		int rtflags = RTF_UP | RTF_HOST;
1918		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1919		if (error)
1920			return (error);
1921		ia->ia_flags |= IFA_ROUTE;
1922		/*
1923		 * Handle the case for ::1 .
1924		 */
1925		if (ifp->if_flags & IFF_LOOPBACK)
1926			ia->ia_flags |= IFA_RTSELF;
1927	}
1928
1929	/*
1930	 * add a loopback route to self
1931	 */
1932	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1933		error = ifa_add_loopback_route((struct ifaddr *)ia,
1934		    (struct sockaddr *)&ia->ia_addr);
1935		if (error == 0)
1936			ia->ia_flags |= IFA_RTSELF;
1937	}
1938
1939	/* Add local address to lltable, if necessary (ex. on p2p link). */
1940	if (newhost)
1941		in6_ifaddloop(&(ia->ia_ifa));
1942
1943	return (error);
1944}
1945
1946/*
1947 * Find an IPv6 interface link-local address specific to an interface.
1948 * ifaddr is returned referenced.
1949 */
1950struct in6_ifaddr *
1951in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1952{
1953	struct ifaddr *ifa;
1954
1955	IF_ADDR_RLOCK(ifp);
1956	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1957		if (ifa->ifa_addr->sa_family != AF_INET6)
1958			continue;
1959		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1960			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1961			    ignoreflags) != 0)
1962				continue;
1963			ifa_ref(ifa);
1964			break;
1965		}
1966	}
1967	IF_ADDR_RUNLOCK(ifp);
1968
1969	return ((struct in6_ifaddr *)ifa);
1970}
1971
1972
1973/*
1974 * find the internet address corresponding to a given interface and address.
1975 * ifaddr is returned referenced.
1976 */
1977struct in6_ifaddr *
1978in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1979{
1980	struct ifaddr *ifa;
1981
1982	IF_ADDR_RLOCK(ifp);
1983	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1984		if (ifa->ifa_addr->sa_family != AF_INET6)
1985			continue;
1986		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1987			ifa_ref(ifa);
1988			break;
1989		}
1990	}
1991	IF_ADDR_RUNLOCK(ifp);
1992
1993	return ((struct in6_ifaddr *)ifa);
1994}
1995
1996/*
1997 * Find a link-local scoped address on ifp and return it if any.
1998 */
1999struct in6_ifaddr *
2000in6ifa_llaonifp(struct ifnet *ifp)
2001{
2002	struct sockaddr_in6 *sin6;
2003	struct ifaddr *ifa;
2004
2005	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2006		return (NULL);
2007	if_addr_rlock(ifp);
2008	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2009		if (ifa->ifa_addr->sa_family != AF_INET6)
2010			continue;
2011		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
2012		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
2013		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
2014		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
2015			break;
2016	}
2017	if_addr_runlock(ifp);
2018
2019	return ((struct in6_ifaddr *)ifa);
2020}
2021
2022/*
2023 * Convert IP6 address to printable (loggable) representation. Caller
2024 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
2025 */
2026static char digits[] = "0123456789abcdef";
2027char *
2028ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
2029{
2030	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
2031	char *cp;
2032	const u_int16_t *a = (const u_int16_t *)addr;
2033	const u_int8_t *d;
2034	int dcolon = 0, zero = 0;
2035
2036	cp = ip6buf;
2037
2038	for (i = 0; i < 8; i++) {
2039		if (*(a + i) == 0) {
2040			cnt++;
2041			if (cnt == 1)
2042				idx = i;
2043		}
2044		else if (maxcnt < cnt) {
2045			maxcnt = cnt;
2046			index = idx;
2047			cnt = 0;
2048		}
2049	}
2050	if (maxcnt < cnt) {
2051		maxcnt = cnt;
2052		index = idx;
2053	}
2054
2055	for (i = 0; i < 8; i++) {
2056		if (dcolon == 1) {
2057			if (*a == 0) {
2058				if (i == 7)
2059					*cp++ = ':';
2060				a++;
2061				continue;
2062			} else
2063				dcolon = 2;
2064		}
2065		if (*a == 0) {
2066			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
2067				if (i == 0)
2068					*cp++ = ':';
2069				*cp++ = ':';
2070				dcolon = 1;
2071			} else {
2072				*cp++ = '0';
2073				*cp++ = ':';
2074			}
2075			a++;
2076			continue;
2077		}
2078		d = (const u_char *)a;
2079		/* Try to eliminate leading zeros in printout like in :0001. */
2080		zero = 1;
2081		*cp = digits[*d >> 4];
2082		if (*cp != '0') {
2083			zero = 0;
2084			cp++;
2085		}
2086		*cp = digits[*d++ & 0xf];
2087		if (zero == 0 || (*cp != '0')) {
2088			zero = 0;
2089			cp++;
2090		}
2091		*cp = digits[*d >> 4];
2092		if (zero == 0 || (*cp != '0')) {
2093			zero = 0;
2094			cp++;
2095		}
2096		*cp++ = digits[*d & 0xf];
2097		*cp++ = ':';
2098		a++;
2099	}
2100	*--cp = '\0';
2101	return (ip6buf);
2102}
2103
2104int
2105in6_localaddr(struct in6_addr *in6)
2106{
2107	struct in6_ifaddr *ia;
2108
2109	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2110		return 1;
2111
2112	IN6_IFADDR_RLOCK();
2113	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2114		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2115		    &ia->ia_prefixmask.sin6_addr)) {
2116			IN6_IFADDR_RUNLOCK();
2117			return 1;
2118		}
2119	}
2120	IN6_IFADDR_RUNLOCK();
2121
2122	return (0);
2123}
2124
2125/*
2126 * Return 1 if an internet address is for the local host and configured
2127 * on one of its interfaces.
2128 */
2129int
2130in6_localip(struct in6_addr *in6)
2131{
2132	struct in6_ifaddr *ia;
2133
2134	IN6_IFADDR_RLOCK();
2135	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
2136		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2137			IN6_IFADDR_RUNLOCK();
2138			return (1);
2139		}
2140	}
2141	IN6_IFADDR_RUNLOCK();
2142	return (0);
2143}
2144
2145int
2146in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2147{
2148	struct in6_ifaddr *ia;
2149
2150	IN6_IFADDR_RLOCK();
2151	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
2152		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
2153			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2154				IN6_IFADDR_RUNLOCK();
2155				return (1); /* true */
2156			}
2157			break;
2158		}
2159	}
2160	IN6_IFADDR_RUNLOCK();
2161
2162	return (0);		/* false */
2163}
2164
2165/*
2166 * return length of part which dst and src are equal
2167 * hard coding...
2168 */
2169int
2170in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2171{
2172	int match = 0;
2173	u_char *s = (u_char *)src, *d = (u_char *)dst;
2174	u_char *lim = s + 16, r;
2175
2176	while (s < lim)
2177		if ((r = (*d++ ^ *s++)) != 0) {
2178			while (r < 128) {
2179				match++;
2180				r <<= 1;
2181			}
2182			break;
2183		} else
2184			match += 8;
2185	return match;
2186}
2187
2188/* XXX: to be scope conscious */
2189int
2190in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2191{
2192	int bytelen, bitlen;
2193
2194	/* sanity check */
2195	if (0 > len || len > 128) {
2196		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2197		    len);
2198		return (0);
2199	}
2200
2201	bytelen = len / 8;
2202	bitlen = len % 8;
2203
2204	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2205		return (0);
2206	if (bitlen != 0 &&
2207	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2208	    p2->s6_addr[bytelen] >> (8 - bitlen))
2209		return (0);
2210
2211	return (1);
2212}
2213
2214void
2215in6_prefixlen2mask(struct in6_addr *maskp, int len)
2216{
2217	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2218	int bytelen, bitlen, i;
2219
2220	/* sanity check */
2221	if (0 > len || len > 128) {
2222		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2223		    len);
2224		return;
2225	}
2226
2227	bzero(maskp, sizeof(*maskp));
2228	bytelen = len / 8;
2229	bitlen = len % 8;
2230	for (i = 0; i < bytelen; i++)
2231		maskp->s6_addr[i] = 0xff;
2232	if (bitlen)
2233		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2234}
2235
2236/*
2237 * return the best address out of the same scope. if no address was
2238 * found, return the first valid address from designated IF.
2239 */
2240struct in6_ifaddr *
2241in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2242{
2243	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2244	struct ifaddr *ifa;
2245	struct in6_ifaddr *besta = 0;
2246	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2247
2248	dep[0] = dep[1] = NULL;
2249
2250	/*
2251	 * We first look for addresses in the same scope.
2252	 * If there is one, return it.
2253	 * If two or more, return one which matches the dst longest.
2254	 * If none, return one of global addresses assigned other ifs.
2255	 */
2256	IF_ADDR_RLOCK(ifp);
2257	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2258		if (ifa->ifa_addr->sa_family != AF_INET6)
2259			continue;
2260		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2261			continue; /* XXX: is there any case to allow anycast? */
2262		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2263			continue; /* don't use this interface */
2264		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2265			continue;
2266		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2267			if (V_ip6_use_deprecated)
2268				dep[0] = (struct in6_ifaddr *)ifa;
2269			continue;
2270		}
2271
2272		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2273			/*
2274			 * call in6_matchlen() as few as possible
2275			 */
2276			if (besta) {
2277				if (blen == -1)
2278					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2279				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2280				if (tlen > blen) {
2281					blen = tlen;
2282					besta = (struct in6_ifaddr *)ifa;
2283				}
2284			} else
2285				besta = (struct in6_ifaddr *)ifa;
2286		}
2287	}
2288	if (besta) {
2289		ifa_ref(&besta->ia_ifa);
2290		IF_ADDR_RUNLOCK(ifp);
2291		return (besta);
2292	}
2293
2294	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2295		if (ifa->ifa_addr->sa_family != AF_INET6)
2296			continue;
2297		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2298			continue; /* XXX: is there any case to allow anycast? */
2299		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2300			continue; /* don't use this interface */
2301		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2302			continue;
2303		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2304			if (V_ip6_use_deprecated)
2305				dep[1] = (struct in6_ifaddr *)ifa;
2306			continue;
2307		}
2308
2309		if (ifa != NULL)
2310			ifa_ref(ifa);
2311		IF_ADDR_RUNLOCK(ifp);
2312		return (struct in6_ifaddr *)ifa;
2313	}
2314
2315	/* use the last-resort values, that are, deprecated addresses */
2316	if (dep[0]) {
2317		ifa_ref((struct ifaddr *)dep[0]);
2318		IF_ADDR_RUNLOCK(ifp);
2319		return dep[0];
2320	}
2321	if (dep[1]) {
2322		ifa_ref((struct ifaddr *)dep[1]);
2323		IF_ADDR_RUNLOCK(ifp);
2324		return dep[1];
2325	}
2326
2327	IF_ADDR_RUNLOCK(ifp);
2328	return NULL;
2329}
2330
2331/*
2332 * perform DAD when interface becomes IFF_UP.
2333 */
2334void
2335in6_if_up(struct ifnet *ifp)
2336{
2337	struct ifaddr *ifa;
2338	struct in6_ifaddr *ia;
2339
2340	IF_ADDR_RLOCK(ifp);
2341	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2342		if (ifa->ifa_addr->sa_family != AF_INET6)
2343			continue;
2344		ia = (struct in6_ifaddr *)ifa;
2345		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2346			/*
2347			 * The TENTATIVE flag was likely set by hand
2348			 * beforehand, implicitly indicating the need for DAD.
2349			 * We may be able to skip the random delay in this
2350			 * case, but we impose delays just in case.
2351			 */
2352			nd6_dad_start(ifa,
2353			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2354		}
2355	}
2356	IF_ADDR_RUNLOCK(ifp);
2357
2358	/*
2359	 * special cases, like 6to4, are handled in in6_ifattach
2360	 */
2361	in6_ifattach(ifp, NULL);
2362}
2363
2364int
2365in6if_do_dad(struct ifnet *ifp)
2366{
2367	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2368		return (0);
2369
2370	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2371		return (0);
2372
2373	switch (ifp->if_type) {
2374#ifdef IFT_DUMMY
2375	case IFT_DUMMY:
2376#endif
2377	case IFT_FAITH:
2378		/*
2379		 * These interfaces do not have the IFF_LOOPBACK flag,
2380		 * but loop packets back.  We do not have to do DAD on such
2381		 * interfaces.  We should even omit it, because loop-backed
2382		 * NS would confuse the DAD procedure.
2383		 */
2384		return (0);
2385	default:
2386		/*
2387		 * Our DAD routine requires the interface up and running.
2388		 * However, some interfaces can be up before the RUNNING
2389		 * status.  Additionaly, users may try to assign addresses
2390		 * before the interface becomes up (or running).
2391		 * We simply skip DAD in such a case as a work around.
2392		 * XXX: we should rather mark "tentative" on such addresses,
2393		 * and do DAD after the interface becomes ready.
2394		 */
2395		if (!((ifp->if_flags & IFF_UP) &&
2396		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2397			return (0);
2398
2399		return (1);
2400	}
2401}
2402
2403/*
2404 * Calculate max IPv6 MTU through all the interfaces and store it
2405 * to in6_maxmtu.
2406 */
2407void
2408in6_setmaxmtu(void)
2409{
2410	unsigned long maxmtu = 0;
2411	struct ifnet *ifp;
2412
2413	IFNET_RLOCK_NOSLEEP();
2414	TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
2415		/* this function can be called during ifnet initialization */
2416		if (!ifp->if_afdata[AF_INET6])
2417			continue;
2418		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2419		    IN6_LINKMTU(ifp) > maxmtu)
2420			maxmtu = IN6_LINKMTU(ifp);
2421	}
2422	IFNET_RUNLOCK_NOSLEEP();
2423	if (maxmtu)	/* update only when maxmtu is positive */
2424		V_in6_maxmtu = maxmtu;
2425}
2426
2427/*
2428 * Provide the length of interface identifiers to be used for the link attached
2429 * to the given interface.  The length should be defined in "IPv6 over
2430 * xxx-link" document.  Note that address architecture might also define
2431 * the length for a particular set of address prefixes, regardless of the
2432 * link type.  As clarified in rfc2462bis, those two definitions should be
2433 * consistent, and those really are as of August 2004.
2434 */
2435int
2436in6_if2idlen(struct ifnet *ifp)
2437{
2438	switch (ifp->if_type) {
2439	case IFT_ETHER:		/* RFC2464 */
2440#ifdef IFT_PROPVIRTUAL
2441	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2442#endif
2443#ifdef IFT_L2VLAN
2444	case IFT_L2VLAN:	/* ditto */
2445#endif
2446#ifdef IFT_IEEE80211
2447	case IFT_IEEE80211:	/* ditto */
2448#endif
2449#ifdef IFT_MIP
2450	case IFT_MIP:	/* ditto */
2451#endif
2452	case IFT_INFINIBAND:
2453		return (64);
2454	case IFT_FDDI:		/* RFC2467 */
2455		return (64);
2456	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2457		return (64);
2458	case IFT_PPP:		/* RFC2472 */
2459		return (64);
2460	case IFT_ARCNET:	/* RFC2497 */
2461		return (64);
2462	case IFT_FRELAY:	/* RFC2590 */
2463		return (64);
2464	case IFT_IEEE1394:	/* RFC3146 */
2465		return (64);
2466	case IFT_GIF:
2467		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2468	case IFT_LOOP:
2469		return (64);	/* XXX: is this really correct? */
2470	default:
2471		/*
2472		 * Unknown link type:
2473		 * It might be controversial to use the today's common constant
2474		 * of 64 for these cases unconditionally.  For full compliance,
2475		 * we should return an error in this case.  On the other hand,
2476		 * if we simply miss the standard for the link type or a new
2477		 * standard is defined for a new link type, the IFID length
2478		 * is very likely to be the common constant.  As a compromise,
2479		 * we always use the constant, but make an explicit notice
2480		 * indicating the "unknown" case.
2481		 */
2482		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2483		return (64);
2484	}
2485}
2486
2487#include <sys/sysctl.h>
2488
2489struct in6_llentry {
2490	struct llentry		base;
2491	struct sockaddr_in6	l3_addr6;
2492};
2493
2494/*
2495 * Deletes an address from the address table.
2496 * This function is called by the timer functions
2497 * such as arptimer() and nd6_llinfo_timer(), and
2498 * the caller does the locking.
2499 */
2500static void
2501in6_lltable_free(struct lltable *llt, struct llentry *lle)
2502{
2503	LLE_WUNLOCK(lle);
2504	LLE_LOCK_DESTROY(lle);
2505	free(lle, M_LLTABLE);
2506}
2507
2508static struct llentry *
2509in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2510{
2511	struct in6_llentry *lle;
2512
2513	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2514	if (lle == NULL)		/* NB: caller generates msg */
2515		return NULL;
2516
2517	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2518	lle->base.lle_refcnt = 1;
2519	lle->base.lle_free = in6_lltable_free;
2520	LLE_LOCK_INIT(&lle->base);
2521	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2522	    CALLOUT_RETURNUNLOCKED);
2523
2524	return (&lle->base);
2525}
2526
2527static void
2528in6_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix,
2529    const struct sockaddr *mask, u_int flags)
2530{
2531	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2532	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2533	struct llentry *lle, *next;
2534	int i;
2535
2536	/*
2537	 * (flags & LLE_STATIC) means deleting all entries
2538	 * including static ND6 entries.
2539	 */
2540	IF_AFDATA_WLOCK(llt->llt_ifp);
2541	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2542		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2543			if (IN6_ARE_MASKED_ADDR_EQUAL(
2544			    &satosin6(L3_ADDR(lle))->sin6_addr,
2545			    &pfx->sin6_addr, &msk->sin6_addr) &&
2546			    ((flags & LLE_STATIC) ||
2547			    !(lle->la_flags & LLE_STATIC))) {
2548				LLE_WLOCK(lle);
2549				if (callout_stop(&lle->la_timer))
2550					LLE_REMREF(lle);
2551				llentry_free(lle);
2552			}
2553		}
2554	}
2555	IF_AFDATA_WUNLOCK(llt->llt_ifp);
2556}
2557
2558static int
2559in6_lltable_rtcheck(struct ifnet *ifp,
2560		    u_int flags,
2561		    const struct sockaddr *l3addr)
2562{
2563	struct rtentry *rt;
2564	char ip6buf[INET6_ADDRSTRLEN];
2565
2566	KASSERT(l3addr->sa_family == AF_INET6,
2567	    ("sin_family %d", l3addr->sa_family));
2568
2569	/* Our local addresses are always only installed on the default FIB. */
2570	/* XXX rtalloc1 should take a const param */
2571	rt = in6_rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0,
2572	    RT_DEFAULT_FIB);
2573	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2574		struct ifaddr *ifa;
2575		/*
2576		 * Create an ND6 cache for an IPv6 neighbor
2577		 * that is not covered by our own prefix.
2578		 */
2579		/* XXX ifaof_ifpforaddr should take a const param */
2580		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2581		if (ifa != NULL) {
2582			ifa_free(ifa);
2583			if (rt != NULL)
2584				RTFREE_LOCKED(rt);
2585			return 0;
2586		}
2587		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2588		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2589		if (rt != NULL)
2590			RTFREE_LOCKED(rt);
2591		return EINVAL;
2592	}
2593	RTFREE_LOCKED(rt);
2594	return 0;
2595}
2596
2597static struct llentry *
2598in6_lltable_lookup(struct lltable *llt, u_int flags,
2599	const struct sockaddr *l3addr)
2600{
2601	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2602	struct ifnet *ifp = llt->llt_ifp;
2603	struct llentry *lle;
2604	struct llentries *lleh;
2605	u_int hashkey;
2606
2607	IF_AFDATA_LOCK_ASSERT(ifp);
2608	KASSERT(l3addr->sa_family == AF_INET6,
2609	    ("sin_family %d", l3addr->sa_family));
2610
2611	hashkey = sin6->sin6_addr.s6_addr32[3];
2612	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2613	LIST_FOREACH(lle, lleh, lle_next) {
2614		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2615		if (lle->la_flags & LLE_DELETED)
2616			continue;
2617		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2618		    sizeof(struct in6_addr)) == 0)
2619			break;
2620	}
2621
2622	if (lle == NULL) {
2623		if (!(flags & LLE_CREATE))
2624			return (NULL);
2625		IF_AFDATA_WLOCK_ASSERT(ifp);
2626		/*
2627		 * A route that covers the given address must have
2628		 * been installed 1st because we are doing a resolution,
2629		 * verify this.
2630		 */
2631		if (!(flags & LLE_IFADDR) &&
2632		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2633			return NULL;
2634
2635		lle = in6_lltable_new(l3addr, flags);
2636		if (lle == NULL) {
2637			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2638			return NULL;
2639		}
2640		lle->la_flags = flags & ~LLE_CREATE;
2641		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2642			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2643			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2644		}
2645
2646		lle->lle_tbl  = llt;
2647		lle->lle_head = lleh;
2648		lle->la_flags |= LLE_LINKED;
2649		LIST_INSERT_HEAD(lleh, lle, lle_next);
2650	} else if (flags & LLE_DELETE) {
2651		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2652			LLE_WLOCK(lle);
2653			lle->la_flags |= LLE_DELETED;
2654#ifdef DIAGNOSTIC
2655			log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2656#endif
2657			if ((lle->la_flags &
2658			    (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2659				llentry_free(lle);
2660			else
2661				LLE_WUNLOCK(lle);
2662		}
2663		lle = (void *)-1;
2664	}
2665	if (LLE_IS_VALID(lle)) {
2666		if (flags & LLE_EXCLUSIVE)
2667			LLE_WLOCK(lle);
2668		else
2669			LLE_RLOCK(lle);
2670	}
2671	return (lle);
2672}
2673
2674static int
2675in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2676{
2677	struct ifnet *ifp = llt->llt_ifp;
2678	struct llentry *lle;
2679	/* XXX stack use */
2680	struct {
2681		struct rt_msghdr	rtm;
2682		struct sockaddr_in6	sin6;
2683		/*
2684		 * ndp.c assumes that sdl is word aligned
2685		 */
2686#ifdef __LP64__
2687		uint32_t		pad;
2688#endif
2689		struct sockaddr_dl	sdl;
2690	} ndpc;
2691	int i, error;
2692
2693	if (ifp->if_flags & IFF_LOOPBACK)
2694		return 0;
2695
2696	LLTABLE_LOCK_ASSERT();
2697
2698	error = 0;
2699	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2700		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2701			struct sockaddr_dl *sdl;
2702
2703			/* skip deleted or invalid entries */
2704			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2705				continue;
2706			/* Skip if jailed and not a valid IP of the prison. */
2707			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2708				continue;
2709			/*
2710			 * produce a msg made of:
2711			 *  struct rt_msghdr;
2712			 *  struct sockaddr_in6 (IPv6)
2713			 *  struct sockaddr_dl;
2714			 */
2715			bzero(&ndpc, sizeof(ndpc));
2716			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2717			ndpc.rtm.rtm_version = RTM_VERSION;
2718			ndpc.rtm.rtm_type = RTM_GET;
2719			ndpc.rtm.rtm_flags = RTF_UP;
2720			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2721			ndpc.sin6.sin6_family = AF_INET6;
2722			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2723			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2724			if (V_deembed_scopeid)
2725				sa6_recoverscope(&ndpc.sin6);
2726
2727			/* publish */
2728			if (lle->la_flags & LLE_PUB)
2729				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2730
2731			sdl = &ndpc.sdl;
2732			sdl->sdl_family = AF_LINK;
2733			sdl->sdl_len = sizeof(*sdl);
2734			sdl->sdl_alen = ifp->if_addrlen;
2735			sdl->sdl_index = ifp->if_index;
2736			sdl->sdl_type = ifp->if_type;
2737			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2738			ndpc.rtm.rtm_rmx.rmx_expire =
2739			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2740			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2741			if (lle->la_flags & LLE_STATIC)
2742				ndpc.rtm.rtm_flags |= RTF_STATIC;
2743			ndpc.rtm.rtm_index = ifp->if_index;
2744			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2745			if (error)
2746				break;
2747		}
2748	}
2749	return error;
2750}
2751
2752void *
2753in6_domifattach(struct ifnet *ifp)
2754{
2755	struct in6_ifextra *ext;
2756
2757	/* There are not IPv6-capable interfaces. */
2758	switch (ifp->if_type) {
2759	case IFT_PFLOG:
2760	case IFT_PFSYNC:
2761	case IFT_USB:
2762		return (NULL);
2763	}
2764	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2765	bzero(ext, sizeof(*ext));
2766
2767	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2768	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2769	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2770	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2771
2772	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2773	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2774	    M_WAITOK);
2775	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2776	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2777
2778	ext->nd_ifinfo = nd6_ifattach(ifp);
2779	ext->scope6_id = scope6_ifattach(ifp);
2780	ext->lltable = lltable_init(ifp, AF_INET6);
2781	if (ext->lltable != NULL) {
2782		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2783		ext->lltable->llt_lookup = in6_lltable_lookup;
2784		ext->lltable->llt_dump = in6_lltable_dump;
2785	}
2786
2787	ext->mld_ifinfo = mld_domifattach(ifp);
2788
2789	return ext;
2790}
2791
2792void
2793in6_domifdetach(struct ifnet *ifp, void *aux)
2794{
2795	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2796
2797	mld_domifdetach(ifp);
2798	scope6_ifdetach(ext->scope6_id);
2799	nd6_ifdetach(ext->nd_ifinfo);
2800	lltable_free(ext->lltable);
2801	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2802	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2803	free(ext->in6_ifstat, M_IFADDR);
2804	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2805	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2806	free(ext->icmp6_ifstat, M_IFADDR);
2807	free(ext, M_IFADDR);
2808}
2809
2810/*
2811 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2812 * v4 mapped addr or v4 compat addr
2813 */
2814void
2815in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2816{
2817
2818	bzero(sin, sizeof(*sin));
2819	sin->sin_len = sizeof(struct sockaddr_in);
2820	sin->sin_family = AF_INET;
2821	sin->sin_port = sin6->sin6_port;
2822	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2823}
2824
2825/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2826void
2827in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2828{
2829	bzero(sin6, sizeof(*sin6));
2830	sin6->sin6_len = sizeof(struct sockaddr_in6);
2831	sin6->sin6_family = AF_INET6;
2832	sin6->sin6_port = sin->sin_port;
2833	sin6->sin6_addr.s6_addr32[0] = 0;
2834	sin6->sin6_addr.s6_addr32[1] = 0;
2835	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2836	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2837}
2838
2839/* Convert sockaddr_in6 into sockaddr_in. */
2840void
2841in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2842{
2843	struct sockaddr_in *sin_p;
2844	struct sockaddr_in6 sin6;
2845
2846	/*
2847	 * Save original sockaddr_in6 addr and convert it
2848	 * to sockaddr_in.
2849	 */
2850	sin6 = *(struct sockaddr_in6 *)nam;
2851	sin_p = (struct sockaddr_in *)nam;
2852	in6_sin6_2_sin(sin_p, &sin6);
2853}
2854
2855/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2856void
2857in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2858{
2859	struct sockaddr_in *sin_p;
2860	struct sockaddr_in6 *sin6_p;
2861
2862	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2863	sin_p = (struct sockaddr_in *)*nam;
2864	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2865	free(*nam, M_SONAME);
2866	*nam = (struct sockaddr *)sin6_p;
2867}
2868