in6.c revision 728e6ff16e41c425792b3b9c07acdff4489b5f59
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("$FreeBSD$");
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
109/*
110 * Definitions of some costant IP6 addresses.
111 */
112const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
113const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
114const struct in6_addr in6addr_nodelocal_allnodes =
115	IN6ADDR_NODELOCAL_ALLNODES_INIT;
116const struct in6_addr in6addr_linklocal_allnodes =
117	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
118const struct in6_addr in6addr_linklocal_allrouters =
119	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
120const struct in6_addr in6addr_linklocal_allv2routers =
121	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
122
123const struct in6_addr in6mask0 = IN6MASK0;
124const struct in6_addr in6mask32 = IN6MASK32;
125const struct in6_addr in6mask64 = IN6MASK64;
126const struct in6_addr in6mask96 = IN6MASK96;
127const struct in6_addr in6mask128 = IN6MASK128;
128
129const struct sockaddr_in6 sa6_any =
130	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
131
132static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
133	struct ifnet *, struct thread *));
134static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
135	struct sockaddr_in6 *, int));
136static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
137
138int	(*faithprefix_p)(struct in6_addr *);
139
140#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
141#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
142
143void
144in6_ifaddloop(struct ifaddr *ifa)
145{
146	struct sockaddr_dl gateway;
147	struct sockaddr_in6 mask, addr;
148	struct rtentry rt;
149	struct in6_ifaddr *ia;
150	struct ifnet *ifp;
151	struct llentry *ln;
152
153	ia = ifa2ia6(ifa);
154	ifp = ifa->ifa_ifp;
155	IF_AFDATA_LOCK(ifp);
156	ifa->ifa_rtrequest = nd6_rtrequest;
157	ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
158	    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
159	IF_AFDATA_UNLOCK(ifp);
160	if (ln != NULL) {
161		ln->la_expire = 0;  /* for IPv6 this means permanent */
162		ln->ln_state = ND6_LLINFO_REACHABLE;
163		/*
164		 * initialize for rtmsg generation
165		 */
166		bzero(&gateway, sizeof(gateway));
167		gateway.sdl_len = sizeof(gateway);
168		gateway.sdl_family = AF_LINK;
169		gateway.sdl_nlen = 0;
170		gateway.sdl_alen = 6;
171		memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
172		    sizeof(ln->ll_addr));
173		LLE_WUNLOCK(ln);
174	}
175
176	bzero(&rt, sizeof(rt));
177	rt.rt_gateway = (struct sockaddr *)&gateway;
178	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
179	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
180	rt_mask(&rt) = (struct sockaddr *)&mask;
181	rt_key(&rt) = (struct sockaddr *)&addr;
182	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
183	rt_newaddrmsg(RTM_ADD, ifa, 0, &rt);
184}
185
186void
187in6_ifremloop(struct ifaddr *ifa)
188{
189	struct sockaddr_dl gateway;
190	struct sockaddr_in6 mask, addr;
191	struct rtentry rt0;
192	struct in6_ifaddr *ia;
193	struct ifnet *ifp;
194
195	ia = ifa2ia6(ifa);
196	ifp = ifa->ifa_ifp;
197	IF_AFDATA_LOCK(ifp);
198	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
199	    (struct sockaddr *)&ia->ia_addr);
200	IF_AFDATA_UNLOCK(ifp);
201
202	/*
203	 * initialize for rtmsg generation
204	 */
205	bzero(&gateway, sizeof(gateway));
206	gateway.sdl_len = sizeof(gateway);
207	gateway.sdl_family = AF_LINK;
208	gateway.sdl_nlen = 0;
209	gateway.sdl_alen = ifp->if_addrlen;
210	bzero(&rt0, sizeof(rt0));
211	rt0.rt_gateway = (struct sockaddr *)&gateway;
212	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
213	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
214	rt_mask(&rt0) = (struct sockaddr *)&mask;
215	rt_key(&rt0) = (struct sockaddr *)&addr;
216	rt0.rt_flags = RTF_HOST | RTF_STATIC;
217	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
218}
219
220int
221in6_mask2len(struct in6_addr *mask, u_char *lim0)
222{
223	int x = 0, y;
224	u_char *lim = lim0, *p;
225
226	/* ignore the scope_id part */
227	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
228		lim = (u_char *)mask + sizeof(*mask);
229	for (p = (u_char *)mask; p < lim; x++, p++) {
230		if (*p != 0xff)
231			break;
232	}
233	y = 0;
234	if (p < lim) {
235		for (y = 0; y < 8; y++) {
236			if ((*p & (0x80 >> y)) == 0)
237				break;
238		}
239	}
240
241	/*
242	 * when the limit pointer is given, do a stricter check on the
243	 * remaining bits.
244	 */
245	if (p < lim) {
246		if (y != 0 && (*p & (0x00ff >> y)) != 0)
247			return (-1);
248		for (p = p + 1; p < lim; p++)
249			if (*p != 0)
250				return (-1);
251	}
252
253	return x * 8 + y;
254}
255
256#ifdef COMPAT_FREEBSD32
257struct in6_ndifreq32 {
258        char ifname[IFNAMSIZ];
259        uint32_t ifindex;
260};
261#define	SIOCGDEFIFACE32_IN6     _IOWR('i', 86, struct in6_ndifreq32)
262#endif
263
264int
265in6_control(struct socket *so, u_long cmd, caddr_t data,
266    struct ifnet *ifp, struct thread *td)
267{
268	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
269	struct	in6_ifaddr *ia = NULL;
270	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
271	struct sockaddr_in6 *sa6;
272	int carp_attached = 0;
273	int error;
274	u_long ocmd = cmd;
275
276	/*
277	 * Compat to make pre-10.x ifconfig(8) operable.
278	 */
279	if (cmd == OSIOCAIFADDR_IN6)
280		cmd = SIOCAIFADDR_IN6;
281
282	switch (cmd) {
283	case SIOCGETSGCNT_IN6:
284	case SIOCGETMIFCNT_IN6:
285		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
286	}
287
288	switch(cmd) {
289	case SIOCAADDRCTL_POLICY:
290	case SIOCDADDRCTL_POLICY:
291		if (td != NULL) {
292			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
293			if (error)
294				return (error);
295		}
296		return (in6_src_ioctl(cmd, data));
297	}
298
299	if (ifp == NULL)
300		return (EOPNOTSUPP);
301
302	switch (cmd) {
303	case SIOCSNDFLUSH_IN6:
304	case SIOCSPFXFLUSH_IN6:
305	case SIOCSRTRFLUSH_IN6:
306	case SIOCSDEFIFACE_IN6:
307	case SIOCSIFINFO_FLAGS:
308	case SIOCSIFINFO_IN6:
309		if (td != NULL) {
310			error = priv_check(td, PRIV_NETINET_ND6);
311			if (error)
312				return (error);
313		}
314		/* FALLTHROUGH */
315	case OSIOCGIFINFO_IN6:
316	case SIOCGIFINFO_IN6:
317	case SIOCGDRLST_IN6:
318	case SIOCGPRLST_IN6:
319	case SIOCGNBRINFO_IN6:
320	case SIOCGDEFIFACE_IN6:
321		return (nd6_ioctl(cmd, data, ifp));
322
323#ifdef COMPAT_FREEBSD32
324	case SIOCGDEFIFACE32_IN6:
325		{
326			struct in6_ndifreq ndif;
327			struct in6_ndifreq32 *ndif32;
328
329			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
330			    ifp);
331			if (error)
332				return (error);
333			ndif32 = (struct in6_ndifreq32 *)data;
334			ndif32->ifindex = ndif.ifindex;
335			return (0);
336		}
337#endif
338	}
339
340	switch (cmd) {
341	case SIOCSIFPREFIX_IN6:
342	case SIOCDIFPREFIX_IN6:
343	case SIOCAIFPREFIX_IN6:
344	case SIOCCIFPREFIX_IN6:
345	case SIOCSGIFPREFIX_IN6:
346	case SIOCGIFPREFIX_IN6:
347		log(LOG_NOTICE,
348		    "prefix ioctls are now invalidated. "
349		    "please use ifconfig.\n");
350		return (EOPNOTSUPP);
351	}
352
353	switch (cmd) {
354	case SIOCSSCOPE6:
355		if (td != NULL) {
356			error = priv_check(td, PRIV_NETINET_SCOPE6);
357			if (error)
358				return (error);
359		}
360		return (scope6_set(ifp,
361		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
362	case SIOCGSCOPE6:
363		return (scope6_get(ifp,
364		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
365	case SIOCGSCOPE6DEF:
366		return (scope6_get_default((struct scope6_id *)
367		    ifr->ifr_ifru.ifru_scope_id));
368	}
369
370	switch (cmd) {
371	case SIOCALIFADDR:
372		if (td != NULL) {
373			error = priv_check(td, PRIV_NET_ADDIFADDR);
374			if (error)
375				return (error);
376		}
377		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
378
379	case SIOCDLIFADDR:
380		if (td != NULL) {
381			error = priv_check(td, PRIV_NET_DELIFADDR);
382			if (error)
383				return (error);
384		}
385		/* FALLTHROUGH */
386	case SIOCGLIFADDR:
387		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
388	}
389
390	/*
391	 * Find address for this interface, if it exists.
392	 *
393	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
394	 * only, and used the first interface address as the target of other
395	 * operations (without checking ifra_addr).  This was because netinet
396	 * code/API assumed at most 1 interface address per interface.
397	 * Since IPv6 allows a node to assign multiple addresses
398	 * on a single interface, we almost always look and check the
399	 * presence of ifra_addr, and reject invalid ones here.
400	 * It also decreases duplicated code among SIOC*_IN6 operations.
401	 */
402	switch (cmd) {
403	case SIOCAIFADDR_IN6:
404	case SIOCSIFPHYADDR_IN6:
405		sa6 = &ifra->ifra_addr;
406		break;
407	case SIOCSIFADDR_IN6:
408	case SIOCGIFADDR_IN6:
409	case SIOCSIFDSTADDR_IN6:
410	case SIOCSIFNETMASK_IN6:
411	case SIOCGIFDSTADDR_IN6:
412	case SIOCGIFNETMASK_IN6:
413	case SIOCDIFADDR_IN6:
414	case SIOCGIFPSRCADDR_IN6:
415	case SIOCGIFPDSTADDR_IN6:
416	case SIOCGIFAFLAG_IN6:
417	case SIOCSNDFLUSH_IN6:
418	case SIOCSPFXFLUSH_IN6:
419	case SIOCSRTRFLUSH_IN6:
420	case SIOCGIFALIFETIME_IN6:
421	case SIOCSIFALIFETIME_IN6:
422	case SIOCGIFSTAT_IN6:
423	case SIOCGIFSTAT_ICMP6:
424		sa6 = &ifr->ifr_addr;
425		break;
426	default:
427		sa6 = NULL;
428		break;
429	}
430	if (sa6 && sa6->sin6_family == AF_INET6) {
431		if (sa6->sin6_scope_id != 0)
432			error = sa6_embedscope(sa6, 0);
433		else
434			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
435		if (error != 0)
436			return (error);
437		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
438		    &sa6->sin6_addr)) != 0)
439			return (error);
440		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
441	} else
442		ia = NULL;
443
444	switch (cmd) {
445	case SIOCSIFADDR_IN6:
446	case SIOCSIFDSTADDR_IN6:
447	case SIOCSIFNETMASK_IN6:
448		/*
449		 * Since IPv6 allows a node to assign multiple addresses
450		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
451		 */
452		/* we decided to obsolete this command (20000704) */
453		error = EINVAL;
454		goto out;
455
456	case SIOCDIFADDR_IN6:
457		/*
458		 * for IPv4, we look for existing in_ifaddr here to allow
459		 * "ifconfig if0 delete" to remove the first IPv4 address on
460		 * the interface.  For IPv6, as the spec allows multiple
461		 * interface address from the day one, we consider "remove the
462		 * first one" semantics to be not preferable.
463		 */
464		if (ia == NULL) {
465			error = EADDRNOTAVAIL;
466			goto out;
467		}
468		/* FALLTHROUGH */
469	case SIOCAIFADDR_IN6:
470		/*
471		 * We always require users to specify a valid IPv6 address for
472		 * the corresponding operation.
473		 */
474		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
475		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
476			error = EAFNOSUPPORT;
477			goto out;
478		}
479
480		if (td != NULL) {
481			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
482			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
483			if (error)
484				goto out;
485		}
486		break;
487
488	case SIOCGIFADDR_IN6:
489		/* This interface is basically deprecated. use SIOCGIFCONF. */
490		/* FALLTHROUGH */
491	case SIOCGIFAFLAG_IN6:
492	case SIOCGIFNETMASK_IN6:
493	case SIOCGIFDSTADDR_IN6:
494	case SIOCGIFALIFETIME_IN6:
495		/* must think again about its semantics */
496		if (ia == NULL) {
497			error = EADDRNOTAVAIL;
498			goto out;
499		}
500		break;
501
502	case SIOCSIFALIFETIME_IN6:
503	    {
504		struct in6_addrlifetime *lt;
505
506		if (td != NULL) {
507			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
508			if (error)
509				goto out;
510		}
511		if (ia == NULL) {
512			error = EADDRNOTAVAIL;
513			goto out;
514		}
515		/* sanity for overflow - beware unsigned */
516		lt = &ifr->ifr_ifru.ifru_lifetime;
517		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
518		    lt->ia6t_vltime + time_second < time_second) {
519			error = EINVAL;
520			goto out;
521		}
522		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
523		    lt->ia6t_pltime + time_second < time_second) {
524			error = EINVAL;
525			goto out;
526		}
527		break;
528	    }
529	}
530
531	switch (cmd) {
532	case SIOCGIFADDR_IN6:
533		ifr->ifr_addr = ia->ia_addr;
534		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
535			goto out;
536		break;
537
538	case SIOCGIFDSTADDR_IN6:
539		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
540			error = EINVAL;
541			goto out;
542		}
543		/*
544		 * XXX: should we check if ifa_dstaddr is NULL and return
545		 * an error?
546		 */
547		ifr->ifr_dstaddr = ia->ia_dstaddr;
548		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
549			goto out;
550		break;
551
552	case SIOCGIFNETMASK_IN6:
553		ifr->ifr_addr = ia->ia_prefixmask;
554		break;
555
556	case SIOCGIFAFLAG_IN6:
557		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
558		break;
559
560	case SIOCGIFSTAT_IN6:
561		if (ifp == NULL) {
562			error = EINVAL;
563			goto out;
564		}
565		bzero(&ifr->ifr_ifru.ifru_stat,
566		    sizeof(ifr->ifr_ifru.ifru_stat));
567		ifr->ifr_ifru.ifru_stat =
568		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
569		break;
570
571	case SIOCGIFSTAT_ICMP6:
572		if (ifp == NULL) {
573			error = EINVAL;
574			goto out;
575		}
576		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
577		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
578		ifr->ifr_ifru.ifru_icmp6stat =
579		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
580		break;
581
582	case SIOCGIFALIFETIME_IN6:
583		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
584		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
585			time_t maxexpire;
586			struct in6_addrlifetime *retlt =
587			    &ifr->ifr_ifru.ifru_lifetime;
588
589			/*
590			 * XXX: adjust expiration time assuming time_t is
591			 * signed.
592			 */
593			maxexpire = (-1) &
594			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
595			if (ia->ia6_lifetime.ia6t_vltime <
596			    maxexpire - ia->ia6_updatetime) {
597				retlt->ia6t_expire = ia->ia6_updatetime +
598				    ia->ia6_lifetime.ia6t_vltime;
599			} else
600				retlt->ia6t_expire = maxexpire;
601		}
602		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
603			time_t maxexpire;
604			struct in6_addrlifetime *retlt =
605			    &ifr->ifr_ifru.ifru_lifetime;
606
607			/*
608			 * XXX: adjust expiration time assuming time_t is
609			 * signed.
610			 */
611			maxexpire = (-1) &
612			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
613			if (ia->ia6_lifetime.ia6t_pltime <
614			    maxexpire - ia->ia6_updatetime) {
615				retlt->ia6t_preferred = ia->ia6_updatetime +
616				    ia->ia6_lifetime.ia6t_pltime;
617			} else
618				retlt->ia6t_preferred = maxexpire;
619		}
620		break;
621
622	case SIOCSIFALIFETIME_IN6:
623		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
624		/* for sanity */
625		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
626			ia->ia6_lifetime.ia6t_expire =
627				time_second + ia->ia6_lifetime.ia6t_vltime;
628		} else
629			ia->ia6_lifetime.ia6t_expire = 0;
630		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
631			ia->ia6_lifetime.ia6t_preferred =
632				time_second + ia->ia6_lifetime.ia6t_pltime;
633		} else
634			ia->ia6_lifetime.ia6t_preferred = 0;
635		break;
636
637	case SIOCAIFADDR_IN6:
638	{
639		int i;
640		struct nd_prefixctl pr0;
641		struct nd_prefix *pr;
642
643		/*
644		 * first, make or update the interface address structure,
645		 * and link it to the list.
646		 */
647		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
648			goto out;
649		if (ia != NULL)
650			ifa_free(&ia->ia_ifa);
651		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
652		    == NULL) {
653			/*
654			 * this can happen when the user specify the 0 valid
655			 * lifetime.
656			 */
657			break;
658		}
659
660		if (cmd == ocmd && ifra->ifra_vhid > 0) {
661			if (carp_attach_p != NULL)
662				error = (*carp_attach_p)(&ia->ia_ifa,
663				    ifra->ifra_vhid);
664			else
665				error = EPROTONOSUPPORT;
666			if (error)
667				goto out;
668			else
669				carp_attached = 1;
670		}
671
672		/*
673		 * then, make the prefix on-link on the interface.
674		 * XXX: we'd rather create the prefix before the address, but
675		 * we need at least one address to install the corresponding
676		 * interface route, so we configure the address first.
677		 */
678
679		/*
680		 * convert mask to prefix length (prefixmask has already
681		 * been validated in in6_update_ifa().
682		 */
683		bzero(&pr0, sizeof(pr0));
684		pr0.ndpr_ifp = ifp;
685		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
686		    NULL);
687		if (pr0.ndpr_plen == 128) {
688			break;	/* we don't need to install a host route. */
689		}
690		pr0.ndpr_prefix = ifra->ifra_addr;
691		/* apply the mask for safety. */
692		for (i = 0; i < 4; i++) {
693			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
694			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
695		}
696		/*
697		 * XXX: since we don't have an API to set prefix (not address)
698		 * lifetimes, we just use the same lifetimes as addresses.
699		 * The (temporarily) installed lifetimes can be overridden by
700		 * later advertised RAs (when accept_rtadv is non 0), which is
701		 * an intended behavior.
702		 */
703		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
704		pr0.ndpr_raf_auto =
705		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
706		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
707		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
708
709		/* add the prefix if not yet. */
710		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
711			/*
712			 * nd6_prelist_add will install the corresponding
713			 * interface route.
714			 */
715			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
716				if (carp_attached)
717					(*carp_detach_p)(&ia->ia_ifa);
718				goto out;
719			}
720			if (pr == NULL) {
721				if (carp_attached)
722					(*carp_detach_p)(&ia->ia_ifa);
723				log(LOG_ERR, "nd6_prelist_add succeeded but "
724				    "no prefix\n");
725				error = EINVAL;
726				goto out;
727			}
728		}
729
730		/* relate the address to the prefix */
731		if (ia->ia6_ndpr == NULL) {
732			ia->ia6_ndpr = pr;
733			pr->ndpr_refcnt++;
734
735			/*
736			 * If this is the first autoconf address from the
737			 * prefix, create a temporary address as well
738			 * (when required).
739			 */
740			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
741			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
742				int e;
743				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
744					log(LOG_NOTICE, "in6_control: failed "
745					    "to create a temporary address, "
746					    "errno=%d\n", e);
747				}
748			}
749		}
750
751		/*
752		 * this might affect the status of autoconfigured addresses,
753		 * that is, this address might make other addresses detached.
754		 */
755		pfxlist_onlink_check();
756		if (error == 0 && ia) {
757			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
758				/*
759				 * Try to clear the flag when a new
760				 * IPv6 address is added onto an
761				 * IFDISABLED interface and it
762				 * succeeds.
763				 */
764				struct in6_ndireq nd;
765
766				memset(&nd, 0, sizeof(nd));
767				nd.ndi.flags = ND_IFINFO(ifp)->flags;
768				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
769				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
770				    (caddr_t)&nd, ifp) < 0)
771					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
772					    "SIOCSIFINFO_FLAGS for -ifdisabled "
773					    "failed.");
774				/*
775				 * Ignore failure of clearing the flag
776				 * intentionally.  The failure means
777				 * address duplication was detected.
778				 */
779			}
780			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
781		}
782		break;
783	}
784
785	case SIOCDIFADDR_IN6:
786	{
787		struct nd_prefix *pr;
788
789		/*
790		 * If the address being deleted is the only one that owns
791		 * the corresponding prefix, expire the prefix as well.
792		 * XXX: theoretically, we don't have to worry about such
793		 * relationship, since we separate the address management
794		 * and the prefix management.  We do this, however, to provide
795		 * as much backward compatibility as possible in terms of
796		 * the ioctl operation.
797		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
798		 */
799		pr = ia->ia6_ndpr;
800		in6_purgeaddr(&ia->ia_ifa);
801		if (pr && pr->ndpr_refcnt == 0)
802			prelist_remove(pr);
803		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
804		break;
805	}
806
807	default:
808		if (ifp == NULL || ifp->if_ioctl == 0) {
809			error = EOPNOTSUPP;
810			goto out;
811		}
812		error = (*ifp->if_ioctl)(ifp, cmd, data);
813		goto out;
814	}
815
816	error = 0;
817out:
818	if (ia != NULL)
819		ifa_free(&ia->ia_ifa);
820	return (error);
821}
822
823/*
824 * Update parameters of an IPv6 interface address.
825 * If necessary, a new entry is created and linked into address chains.
826 * This function is separated from in6_control().
827 * XXX: should this be performed under splnet()?
828 */
829int
830in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
831    struct in6_ifaddr *ia, int flags)
832{
833	int error = 0, hostIsNew = 0, plen = -1;
834	struct sockaddr_in6 dst6;
835	struct in6_addrlifetime *lt;
836	struct in6_multi_mship *imm;
837	struct in6_multi *in6m_sol;
838	struct rtentry *rt;
839	int delay;
840	char ip6buf[INET6_ADDRSTRLEN];
841
842	/* Validate parameters */
843	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
844		return (EINVAL);
845
846	/*
847	 * The destination address for a p2p link must have a family
848	 * of AF_UNSPEC or AF_INET6.
849	 */
850	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
851	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
852	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
853		return (EAFNOSUPPORT);
854	/*
855	 * validate ifra_prefixmask.  don't check sin6_family, netmask
856	 * does not carry fields other than sin6_len.
857	 */
858	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
859		return (EINVAL);
860	/*
861	 * Because the IPv6 address architecture is classless, we require
862	 * users to specify a (non 0) prefix length (mask) for a new address.
863	 * We also require the prefix (when specified) mask is valid, and thus
864	 * reject a non-consecutive mask.
865	 */
866	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
867		return (EINVAL);
868	if (ifra->ifra_prefixmask.sin6_len != 0) {
869		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
870		    (u_char *)&ifra->ifra_prefixmask +
871		    ifra->ifra_prefixmask.sin6_len);
872		if (plen <= 0)
873			return (EINVAL);
874	} else {
875		/*
876		 * In this case, ia must not be NULL.  We just use its prefix
877		 * length.
878		 */
879		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
880	}
881	/*
882	 * If the destination address on a p2p interface is specified,
883	 * and the address is a scoped one, validate/set the scope
884	 * zone identifier.
885	 */
886	dst6 = ifra->ifra_dstaddr;
887	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
888	    (dst6.sin6_family == AF_INET6)) {
889		struct in6_addr in6_tmp;
890		u_int32_t zoneid;
891
892		in6_tmp = dst6.sin6_addr;
893		if (in6_setscope(&in6_tmp, ifp, &zoneid))
894			return (EINVAL); /* XXX: should be impossible */
895
896		if (dst6.sin6_scope_id != 0) {
897			if (dst6.sin6_scope_id != zoneid)
898				return (EINVAL);
899		} else		/* user omit to specify the ID. */
900			dst6.sin6_scope_id = zoneid;
901
902		/* convert into the internal form */
903		if (sa6_embedscope(&dst6, 0))
904			return (EINVAL); /* XXX: should be impossible */
905	}
906	/*
907	 * The destination address can be specified only for a p2p or a
908	 * loopback interface.  If specified, the corresponding prefix length
909	 * must be 128.
910	 */
911	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
912		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
913			/* XXX: noisy message */
914			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
915			    "be specified for a p2p or a loopback IF only\n"));
916			return (EINVAL);
917		}
918		if (plen != 128) {
919			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
920			    "be 128 when dstaddr is specified\n"));
921			return (EINVAL);
922		}
923	}
924	/* lifetime consistency check */
925	lt = &ifra->ifra_lifetime;
926	if (lt->ia6t_pltime > lt->ia6t_vltime)
927		return (EINVAL);
928	if (lt->ia6t_vltime == 0) {
929		/*
930		 * the following log might be noisy, but this is a typical
931		 * configuration mistake or a tool's bug.
932		 */
933		nd6log((LOG_INFO,
934		    "in6_update_ifa: valid lifetime is 0 for %s\n",
935		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
936
937		if (ia == NULL)
938			return (0); /* there's nothing to do */
939	}
940
941	/*
942	 * If this is a new address, allocate a new ifaddr and link it
943	 * into chains.
944	 */
945	if (ia == NULL) {
946		hostIsNew = 1;
947		/*
948		 * When in6_update_ifa() is called in a process of a received
949		 * RA, it is called under an interrupt context.  So, we should
950		 * call malloc with M_NOWAIT.
951		 */
952		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
953		    M_NOWAIT);
954		if (ia == NULL)
955			return (ENOBUFS);
956		bzero((caddr_t)ia, sizeof(*ia));
957		ifa_init(&ia->ia_ifa);
958		LIST_INIT(&ia->ia6_memberships);
959		/* Initialize the address and masks, and put time stamp */
960		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
961		ia->ia_addr.sin6_family = AF_INET6;
962		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
963		ia->ia6_createtime = time_second;
964		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
965			/*
966			 * XXX: some functions expect that ifa_dstaddr is not
967			 * NULL for p2p interfaces.
968			 */
969			ia->ia_ifa.ifa_dstaddr =
970			    (struct sockaddr *)&ia->ia_dstaddr;
971		} else {
972			ia->ia_ifa.ifa_dstaddr = NULL;
973		}
974		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
975		ia->ia_ifp = ifp;
976		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
977		IF_ADDR_WLOCK(ifp);
978		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
979		IF_ADDR_WUNLOCK(ifp);
980
981		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
982		IN6_IFADDR_WLOCK();
983		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
984		IN6_IFADDR_WUNLOCK();
985	}
986
987	/* update timestamp */
988	ia->ia6_updatetime = time_second;
989
990	/* set prefix mask */
991	if (ifra->ifra_prefixmask.sin6_len) {
992		/*
993		 * We prohibit changing the prefix length of an existing
994		 * address, because
995		 * + such an operation should be rare in IPv6, and
996		 * + the operation would confuse prefix management.
997		 */
998		if (ia->ia_prefixmask.sin6_len &&
999		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1000			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1001			    " existing (%s) address should not be changed\n",
1002			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1003			error = EINVAL;
1004			goto unlink;
1005		}
1006		ia->ia_prefixmask = ifra->ifra_prefixmask;
1007	}
1008
1009	/*
1010	 * If a new destination address is specified, scrub the old one and
1011	 * install the new destination.  Note that the interface must be
1012	 * p2p or loopback (see the check above.)
1013	 */
1014	if (dst6.sin6_family == AF_INET6 &&
1015	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1016		int e;
1017
1018		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1019		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1020			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1021			    "a route to the old destination: %s\n",
1022			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1023			/* proceed anyway... */
1024		} else
1025			ia->ia_flags &= ~IFA_ROUTE;
1026		ia->ia_dstaddr = dst6;
1027	}
1028
1029	/*
1030	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1031	 * to see if the address is deprecated or invalidated, but initialize
1032	 * these members for applications.
1033	 */
1034	ia->ia6_lifetime = ifra->ifra_lifetime;
1035	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1036		ia->ia6_lifetime.ia6t_expire =
1037		    time_second + ia->ia6_lifetime.ia6t_vltime;
1038	} else
1039		ia->ia6_lifetime.ia6t_expire = 0;
1040	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1041		ia->ia6_lifetime.ia6t_preferred =
1042		    time_second + ia->ia6_lifetime.ia6t_pltime;
1043	} else
1044		ia->ia6_lifetime.ia6t_preferred = 0;
1045
1046	/* reset the interface and routing table appropriately. */
1047	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1048		goto unlink;
1049
1050	/*
1051	 * configure address flags.
1052	 */
1053	ia->ia6_flags = ifra->ifra_flags;
1054	/*
1055	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1056	 * userland, make it deprecated.
1057	 */
1058	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1059		ia->ia6_lifetime.ia6t_pltime = 0;
1060		ia->ia6_lifetime.ia6t_preferred = time_second;
1061	}
1062	/*
1063	 * Make the address tentative before joining multicast addresses,
1064	 * so that corresponding MLD responses would not have a tentative
1065	 * source address.
1066	 */
1067	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1068	if (hostIsNew && in6if_do_dad(ifp))
1069		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1070
1071	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
1072	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1073		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1074
1075	/*
1076	 * We are done if we have simply modified an existing address.
1077	 */
1078	if (!hostIsNew)
1079		return (error);
1080
1081	/*
1082	 * Beyond this point, we should call in6_purgeaddr upon an error,
1083	 * not just go to unlink.
1084	 */
1085
1086	/* Join necessary multicast groups */
1087	in6m_sol = NULL;
1088	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1089		struct sockaddr_in6 mltaddr, mltmask;
1090		struct in6_addr llsol;
1091
1092		/* join solicited multicast addr for new host id */
1093		bzero(&llsol, sizeof(struct in6_addr));
1094		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1095		llsol.s6_addr32[1] = 0;
1096		llsol.s6_addr32[2] = htonl(1);
1097		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1098		llsol.s6_addr8[12] = 0xff;
1099		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1100			/* XXX: should not happen */
1101			log(LOG_ERR, "in6_update_ifa: "
1102			    "in6_setscope failed\n");
1103			goto cleanup;
1104		}
1105		delay = 0;
1106		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1107			/*
1108			 * We need a random delay for DAD on the address
1109			 * being configured.  It also means delaying
1110			 * transmission of the corresponding MLD report to
1111			 * avoid report collision.
1112			 * [RFC 4861, Section 6.3.7]
1113			 */
1114			delay = arc4random() %
1115			    (MAX_RTR_SOLICITATION_DELAY * hz);
1116		}
1117		imm = in6_joingroup(ifp, &llsol, &error, delay);
1118		if (imm == NULL) {
1119			nd6log((LOG_WARNING,
1120			    "in6_update_ifa: addmulti failed for "
1121			    "%s on %s (errno=%d)\n",
1122			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
1123			    error));
1124			goto cleanup;
1125		}
1126		LIST_INSERT_HEAD(&ia->ia6_memberships,
1127		    imm, i6mm_chain);
1128		in6m_sol = imm->i6mm_maddr;
1129
1130		bzero(&mltmask, sizeof(mltmask));
1131		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1132		mltmask.sin6_family = AF_INET6;
1133		mltmask.sin6_addr = in6mask32;
1134#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1135
1136		/*
1137		 * join link-local all-nodes address
1138		 */
1139		bzero(&mltaddr, sizeof(mltaddr));
1140		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1141		mltaddr.sin6_family = AF_INET6;
1142		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1143		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1144		    0)
1145			goto cleanup; /* XXX: should not fail */
1146
1147		/*
1148		 * XXX: do we really need this automatic routes?
1149		 * We should probably reconsider this stuff.  Most applications
1150		 * actually do not need the routes, since they usually specify
1151		 * the outgoing interface.
1152		 */
1153		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1154		if (rt) {
1155			/* XXX: only works in !SCOPEDROUTING case. */
1156			if (memcmp(&mltaddr.sin6_addr,
1157			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1158			    MLTMASK_LEN)) {
1159				RTFREE_LOCKED(rt);
1160				rt = NULL;
1161			}
1162		}
1163		if (!rt) {
1164			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1165			    (struct sockaddr *)&ia->ia_addr,
1166			    (struct sockaddr *)&mltmask, RTF_UP,
1167			    (struct rtentry **)0);
1168			if (error)
1169				goto cleanup;
1170		} else {
1171			RTFREE_LOCKED(rt);
1172		}
1173
1174		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1175		if (!imm) {
1176			nd6log((LOG_WARNING,
1177			    "in6_update_ifa: addmulti failed for "
1178			    "%s on %s (errno=%d)\n",
1179			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1180			    if_name(ifp), error));
1181			goto cleanup;
1182		}
1183		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1184
1185		/*
1186		 * join node information group address
1187		 */
1188		delay = 0;
1189		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1190			/*
1191			 * The spec doesn't say anything about delay for this
1192			 * group, but the same logic should apply.
1193			 */
1194			delay = arc4random() %
1195			    (MAX_RTR_SOLICITATION_DELAY * hz);
1196		}
1197		if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
1198			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1199			    delay); /* XXX jinmei */
1200			if (!imm) {
1201				nd6log((LOG_WARNING, "in6_update_ifa: "
1202				    "addmulti failed for %s on %s "
1203				    "(errno=%d)\n",
1204				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1205				    if_name(ifp), error));
1206				/* XXX not very fatal, go on... */
1207			} else {
1208				LIST_INSERT_HEAD(&ia->ia6_memberships,
1209				    imm, i6mm_chain);
1210			}
1211		}
1212
1213		/*
1214		 * join interface-local all-nodes address.
1215		 * (ff01::1%ifN, and ff01::%ifN/32)
1216		 */
1217		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1218		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1219		    != 0)
1220			goto cleanup; /* XXX: should not fail */
1221		/* XXX: again, do we really need the route? */
1222		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1223		if (rt) {
1224			if (memcmp(&mltaddr.sin6_addr,
1225			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1226			    MLTMASK_LEN)) {
1227				RTFREE_LOCKED(rt);
1228				rt = NULL;
1229			}
1230		}
1231		if (!rt) {
1232			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1233			    (struct sockaddr *)&ia->ia_addr,
1234			    (struct sockaddr *)&mltmask, RTF_UP,
1235			    (struct rtentry **)0);
1236			if (error)
1237				goto cleanup;
1238		} else
1239			RTFREE_LOCKED(rt);
1240
1241		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1242		if (!imm) {
1243			nd6log((LOG_WARNING, "in6_update_ifa: "
1244			    "addmulti failed for %s on %s "
1245			    "(errno=%d)\n",
1246			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1247			    if_name(ifp), error));
1248			goto cleanup;
1249		}
1250		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1251#undef	MLTMASK_LEN
1252	}
1253
1254	/*
1255	 * Perform DAD, if needed.
1256	 * XXX It may be of use, if we can administratively
1257	 * disable DAD.
1258	 */
1259	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1260	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1261	{
1262		int mindelay, maxdelay;
1263
1264		delay = 0;
1265		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1266			/*
1267			 * We need to impose a delay before sending an NS
1268			 * for DAD.  Check if we also needed a delay for the
1269			 * corresponding MLD message.  If we did, the delay
1270			 * should be larger than the MLD delay (this could be
1271			 * relaxed a bit, but this simple logic is at least
1272			 * safe).
1273			 * XXX: Break data hiding guidelines and look at
1274			 * state for the solicited multicast group.
1275			 */
1276			mindelay = 0;
1277			if (in6m_sol != NULL &&
1278			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1279				mindelay = in6m_sol->in6m_timer;
1280			}
1281			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1282			if (maxdelay - mindelay == 0)
1283				delay = 0;
1284			else {
1285				delay =
1286				    (arc4random() % (maxdelay - mindelay)) +
1287				    mindelay;
1288			}
1289		}
1290		nd6_dad_start((struct ifaddr *)ia, delay);
1291	}
1292
1293	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1294	ifa_free(&ia->ia_ifa);
1295	return (error);
1296
1297  unlink:
1298	/*
1299	 * XXX: if a change of an existing address failed, keep the entry
1300	 * anyway.
1301	 */
1302	if (hostIsNew) {
1303		in6_unlink_ifa(ia, ifp);
1304		ifa_free(&ia->ia_ifa);
1305	}
1306	return (error);
1307
1308  cleanup:
1309	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1310	ifa_free(&ia->ia_ifa);
1311	in6_purgeaddr(&ia->ia_ifa);
1312	return error;
1313}
1314
1315void
1316in6_purgeaddr(struct ifaddr *ifa)
1317{
1318	struct ifnet *ifp = ifa->ifa_ifp;
1319	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1320	struct in6_multi_mship *imm;
1321	struct sockaddr_in6 mltaddr, mltmask;
1322	int plen, error;
1323	struct rtentry *rt;
1324	struct ifaddr *ifa0;
1325
1326	if (ifa->ifa_carp)
1327		(*carp_detach_p)(ifa);
1328
1329	/*
1330	 * find another IPv6 address as the gateway for the
1331	 * link-local and node-local all-nodes multicast
1332	 * address routes
1333	 */
1334	IF_ADDR_RLOCK(ifp);
1335	TAILQ_FOREACH(ifa0, &ifp->if_addrhead, ifa_link) {
1336		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1337		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1338			   &ia->ia_addr.sin6_addr,
1339			   sizeof(struct in6_addr)) == 0)
1340			continue;
1341		else
1342			break;
1343	}
1344	if (ifa0 != NULL)
1345		ifa_ref(ifa0);
1346	IF_ADDR_RUNLOCK(ifp);
1347
1348	/*
1349	 * Remove the loopback route to the interface address.
1350	 * The check for the current setting of "nd6_useloopback"
1351	 * is not needed.
1352	 */
1353	if (ia->ia_flags & IFA_RTSELF) {
1354		error = ifa_del_loopback_route((struct ifaddr *)ia,
1355				       (struct sockaddr *)&ia->ia_addr);
1356		if (error == 0)
1357			ia->ia_flags &= ~IFA_RTSELF;
1358	}
1359
1360	/* stop DAD processing */
1361	nd6_dad_stop(ifa);
1362
1363	in6_ifremloop(ifa);
1364
1365	/*
1366	 * leave from multicast groups we have joined for the interface
1367	 */
1368	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1369		LIST_REMOVE(imm, i6mm_chain);
1370		in6_leavegroup(imm);
1371	}
1372
1373	/*
1374	 * remove the link-local all-nodes address
1375	 */
1376	bzero(&mltmask, sizeof(mltmask));
1377	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1378	mltmask.sin6_family = AF_INET6;
1379	mltmask.sin6_addr = in6mask32;
1380
1381	bzero(&mltaddr, sizeof(mltaddr));
1382	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1383	mltaddr.sin6_family = AF_INET6;
1384	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1385
1386	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1387		goto cleanup;
1388
1389	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1390	if (rt != NULL && rt->rt_gateway != NULL &&
1391	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1392		    &ia->ia_addr.sin6_addr,
1393		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1394		/*
1395		 * if no more IPv6 address exists on this interface
1396		 * then remove the multicast address route
1397		 */
1398		if (ifa0 == NULL) {
1399			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1400			       sizeof(mltaddr.sin6_addr));
1401			RTFREE_LOCKED(rt);
1402			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1403					  (struct sockaddr *)&ia->ia_addr,
1404					  (struct sockaddr *)&mltmask, RTF_UP,
1405					  (struct rtentry **)0);
1406			if (error)
1407				log(LOG_INFO, "in6_purgeaddr: link-local all-nodes"
1408				    "multicast address deletion error\n");
1409		} else {
1410			/*
1411			 * replace the gateway of the route
1412			 */
1413			struct sockaddr_in6 sa;
1414
1415			bzero(&sa, sizeof(sa));
1416			sa.sin6_len = sizeof(struct sockaddr_in6);
1417			sa.sin6_family = AF_INET6;
1418			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1419			       sizeof(sa.sin6_addr));
1420			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1421			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1422			RTFREE_LOCKED(rt);
1423		}
1424	} else {
1425		if (rt != NULL)
1426			RTFREE_LOCKED(rt);
1427	}
1428
1429	/*
1430	 * remove the node-local all-nodes address
1431	 */
1432	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1433	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1434	    0)
1435		goto cleanup;
1436
1437	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1438	if (rt != NULL && rt->rt_gateway != NULL &&
1439	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1440		    &ia->ia_addr.sin6_addr,
1441		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1442		/*
1443		 * if no more IPv6 address exists on this interface
1444		 * then remove the multicast address route
1445		 */
1446		if (ifa0 == NULL) {
1447			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1448			       sizeof(mltaddr.sin6_addr));
1449
1450			RTFREE_LOCKED(rt);
1451			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1452					  (struct sockaddr *)&ia->ia_addr,
1453					  (struct sockaddr *)&mltmask, RTF_UP,
1454					  (struct rtentry **)0);
1455
1456			if (error)
1457				log(LOG_INFO, "in6_purgeaddr: node-local all-nodes"
1458				    "multicast address deletion error\n");
1459		} else {
1460			/*
1461			 * replace the gateway of the route
1462			 */
1463			struct sockaddr_in6 sa;
1464
1465			bzero(&sa, sizeof(sa));
1466			sa.sin6_len = sizeof(struct sockaddr_in6);
1467			sa.sin6_family = AF_INET6;
1468			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1469			       sizeof(sa.sin6_addr));
1470			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1471			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1472			RTFREE_LOCKED(rt);
1473		}
1474	} else {
1475		if (rt != NULL)
1476			RTFREE_LOCKED(rt);
1477	}
1478
1479cleanup:
1480
1481	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1482	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1483		int error;
1484		struct sockaddr *dstaddr;
1485
1486		/*
1487		 * use the interface address if configuring an
1488		 * interface address with a /128 prefix len
1489		 */
1490		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1491			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1492		else
1493			dstaddr = (struct sockaddr *)&ia->ia_addr;
1494
1495		error = rtrequest(RTM_DELETE,
1496		    (struct sockaddr *)dstaddr,
1497		    (struct sockaddr *)&ia->ia_addr,
1498		    (struct sockaddr *)&ia->ia_prefixmask,
1499		    ia->ia_flags | RTF_HOST, NULL);
1500		if (error != 0)
1501			return;
1502		ia->ia_flags &= ~IFA_ROUTE;
1503	}
1504	if (ifa0 != NULL)
1505		ifa_free(ifa0);
1506
1507	in6_unlink_ifa(ia, ifp);
1508}
1509
1510static void
1511in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1512{
1513	int	s = splnet();
1514
1515	IF_ADDR_WLOCK(ifp);
1516	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1517	IF_ADDR_WUNLOCK(ifp);
1518	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1519
1520	/*
1521	 * Defer the release of what might be the last reference to the
1522	 * in6_ifaddr so that it can't be freed before the remainder of the
1523	 * cleanup.
1524	 */
1525	IN6_IFADDR_WLOCK();
1526	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1527	IN6_IFADDR_WUNLOCK();
1528
1529	/*
1530	 * Release the reference to the base prefix.  There should be a
1531	 * positive reference.
1532	 */
1533	if (ia->ia6_ndpr == NULL) {
1534		nd6log((LOG_NOTICE,
1535		    "in6_unlink_ifa: autoconf'ed address "
1536		    "%p has no prefix\n", ia));
1537	} else {
1538		ia->ia6_ndpr->ndpr_refcnt--;
1539		ia->ia6_ndpr = NULL;
1540	}
1541
1542	/*
1543	 * Also, if the address being removed is autoconf'ed, call
1544	 * pfxlist_onlink_check() since the release might affect the status of
1545	 * other (detached) addresses.
1546	 */
1547	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1548		pfxlist_onlink_check();
1549	}
1550	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1551	splx(s);
1552}
1553
1554void
1555in6_purgeif(struct ifnet *ifp)
1556{
1557	struct ifaddr *ifa, *nifa;
1558
1559	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1560		if (ifa->ifa_addr->sa_family != AF_INET6)
1561			continue;
1562		in6_purgeaddr(ifa);
1563	}
1564
1565	in6_ifdetach(ifp);
1566}
1567
1568/*
1569 * SIOC[GAD]LIFADDR.
1570 *	SIOCGLIFADDR: get first address. (?)
1571 *	SIOCGLIFADDR with IFLR_PREFIX:
1572 *		get first address that matches the specified prefix.
1573 *	SIOCALIFADDR: add the specified address.
1574 *	SIOCALIFADDR with IFLR_PREFIX:
1575 *		add the specified prefix, filling hostid part from
1576 *		the first link-local address.  prefixlen must be <= 64.
1577 *	SIOCDLIFADDR: delete the specified address.
1578 *	SIOCDLIFADDR with IFLR_PREFIX:
1579 *		delete the first address that matches the specified prefix.
1580 * return values:
1581 *	EINVAL on invalid parameters
1582 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1583 *	other values may be returned from in6_ioctl()
1584 *
1585 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1586 * this is to accomodate address naming scheme other than RFC2374,
1587 * in the future.
1588 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1589 * address encoding scheme. (see figure on page 8)
1590 */
1591static int
1592in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1593    struct ifnet *ifp, struct thread *td)
1594{
1595	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1596	struct ifaddr *ifa;
1597	struct sockaddr *sa;
1598
1599	/* sanity checks */
1600	if (!data || !ifp) {
1601		panic("invalid argument to in6_lifaddr_ioctl");
1602		/* NOTREACHED */
1603	}
1604
1605	switch (cmd) {
1606	case SIOCGLIFADDR:
1607		/* address must be specified on GET with IFLR_PREFIX */
1608		if ((iflr->flags & IFLR_PREFIX) == 0)
1609			break;
1610		/* FALLTHROUGH */
1611	case SIOCALIFADDR:
1612	case SIOCDLIFADDR:
1613		/* address must be specified on ADD and DELETE */
1614		sa = (struct sockaddr *)&iflr->addr;
1615		if (sa->sa_family != AF_INET6)
1616			return EINVAL;
1617		if (sa->sa_len != sizeof(struct sockaddr_in6))
1618			return EINVAL;
1619		/* XXX need improvement */
1620		sa = (struct sockaddr *)&iflr->dstaddr;
1621		if (sa->sa_family && sa->sa_family != AF_INET6)
1622			return EINVAL;
1623		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1624			return EINVAL;
1625		break;
1626	default: /* shouldn't happen */
1627#if 0
1628		panic("invalid cmd to in6_lifaddr_ioctl");
1629		/* NOTREACHED */
1630#else
1631		return EOPNOTSUPP;
1632#endif
1633	}
1634	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1635		return EINVAL;
1636
1637	switch (cmd) {
1638	case SIOCALIFADDR:
1639	    {
1640		struct in6_aliasreq ifra;
1641		struct in6_addr *hostid = NULL;
1642		int prefixlen;
1643
1644		ifa = NULL;
1645		if ((iflr->flags & IFLR_PREFIX) != 0) {
1646			struct sockaddr_in6 *sin6;
1647
1648			/*
1649			 * hostid is to fill in the hostid part of the
1650			 * address.  hostid points to the first link-local
1651			 * address attached to the interface.
1652			 */
1653			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1654			if (!ifa)
1655				return EADDRNOTAVAIL;
1656			hostid = IFA_IN6(ifa);
1657
1658			/* prefixlen must be <= 64. */
1659			if (64 < iflr->prefixlen)
1660				return EINVAL;
1661			prefixlen = iflr->prefixlen;
1662
1663			/* hostid part must be zero. */
1664			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1665			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1666			    sin6->sin6_addr.s6_addr32[3] != 0) {
1667				return EINVAL;
1668			}
1669		} else
1670			prefixlen = iflr->prefixlen;
1671
1672		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1673		bzero(&ifra, sizeof(ifra));
1674		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1675
1676		bcopy(&iflr->addr, &ifra.ifra_addr,
1677		    ((struct sockaddr *)&iflr->addr)->sa_len);
1678		if (hostid) {
1679			/* fill in hostid part */
1680			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1681			    hostid->s6_addr32[2];
1682			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1683			    hostid->s6_addr32[3];
1684		}
1685
1686		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1687			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1688			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1689			if (hostid) {
1690				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1691				    hostid->s6_addr32[2];
1692				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1693				    hostid->s6_addr32[3];
1694			}
1695		}
1696		if (ifa != NULL)
1697			ifa_free(ifa);
1698
1699		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1700		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1701
1702		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1703		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1704	    }
1705	case SIOCGLIFADDR:
1706	case SIOCDLIFADDR:
1707	    {
1708		struct in6_ifaddr *ia;
1709		struct in6_addr mask, candidate, match;
1710		struct sockaddr_in6 *sin6;
1711		int cmp;
1712
1713		bzero(&mask, sizeof(mask));
1714		if (iflr->flags & IFLR_PREFIX) {
1715			/* lookup a prefix rather than address. */
1716			in6_prefixlen2mask(&mask, iflr->prefixlen);
1717
1718			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1719			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1720			match.s6_addr32[0] &= mask.s6_addr32[0];
1721			match.s6_addr32[1] &= mask.s6_addr32[1];
1722			match.s6_addr32[2] &= mask.s6_addr32[2];
1723			match.s6_addr32[3] &= mask.s6_addr32[3];
1724
1725			/* if you set extra bits, that's wrong */
1726			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1727				return EINVAL;
1728
1729			cmp = 1;
1730		} else {
1731			if (cmd == SIOCGLIFADDR) {
1732				/* on getting an address, take the 1st match */
1733				cmp = 0;	/* XXX */
1734			} else {
1735				/* on deleting an address, do exact match */
1736				in6_prefixlen2mask(&mask, 128);
1737				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1738				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1739
1740				cmp = 1;
1741			}
1742		}
1743
1744		IF_ADDR_RLOCK(ifp);
1745		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1746			if (ifa->ifa_addr->sa_family != AF_INET6)
1747				continue;
1748			if (!cmp)
1749				break;
1750
1751			/*
1752			 * XXX: this is adhoc, but is necessary to allow
1753			 * a user to specify fe80::/64 (not /10) for a
1754			 * link-local address.
1755			 */
1756			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1757			in6_clearscope(&candidate);
1758			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1759			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1760			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1761			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1762			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1763				break;
1764		}
1765		if (ifa != NULL)
1766			ifa_ref(ifa);
1767		IF_ADDR_RUNLOCK(ifp);
1768		if (!ifa)
1769			return EADDRNOTAVAIL;
1770		ia = ifa2ia6(ifa);
1771
1772		if (cmd == SIOCGLIFADDR) {
1773			int error;
1774
1775			/* fill in the if_laddrreq structure */
1776			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1777			error = sa6_recoverscope(
1778			    (struct sockaddr_in6 *)&iflr->addr);
1779			if (error != 0) {
1780				ifa_free(ifa);
1781				return (error);
1782			}
1783
1784			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1785				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1786				    ia->ia_dstaddr.sin6_len);
1787				error = sa6_recoverscope(
1788				    (struct sockaddr_in6 *)&iflr->dstaddr);
1789				if (error != 0) {
1790					ifa_free(ifa);
1791					return (error);
1792				}
1793			} else
1794				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1795
1796			iflr->prefixlen =
1797			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1798
1799			iflr->flags = ia->ia6_flags;	/* XXX */
1800			ifa_free(ifa);
1801
1802			return 0;
1803		} else {
1804			struct in6_aliasreq ifra;
1805
1806			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1807			bzero(&ifra, sizeof(ifra));
1808			bcopy(iflr->iflr_name, ifra.ifra_name,
1809			    sizeof(ifra.ifra_name));
1810
1811			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1812			    ia->ia_addr.sin6_len);
1813			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1814				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1815				    ia->ia_dstaddr.sin6_len);
1816			} else {
1817				bzero(&ifra.ifra_dstaddr,
1818				    sizeof(ifra.ifra_dstaddr));
1819			}
1820			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1821			    ia->ia_prefixmask.sin6_len);
1822
1823			ifra.ifra_flags = ia->ia6_flags;
1824			ifa_free(ifa);
1825			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1826			    ifp, td);
1827		}
1828	    }
1829	}
1830
1831	return EOPNOTSUPP;	/* just for safety */
1832}
1833
1834/*
1835 * Initialize an interface's intetnet6 address
1836 * and routing table entry.
1837 */
1838static int
1839in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1840    struct sockaddr_in6 *sin6, int newhost)
1841{
1842	int	error = 0, plen, ifacount = 0;
1843	int	s = splimp();
1844	struct ifaddr *ifa;
1845
1846	/*
1847	 * Give the interface a chance to initialize
1848	 * if this is its first address,
1849	 * and to validate the address if necessary.
1850	 */
1851	IF_ADDR_RLOCK(ifp);
1852	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1853		if (ifa->ifa_addr->sa_family != AF_INET6)
1854			continue;
1855		ifacount++;
1856	}
1857	IF_ADDR_RUNLOCK(ifp);
1858
1859	ia->ia_addr = *sin6;
1860
1861	if (ifacount <= 1 && ifp->if_ioctl) {
1862		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1863		if (error) {
1864			splx(s);
1865			return (error);
1866		}
1867	}
1868	splx(s);
1869
1870	ia->ia_ifa.ifa_metric = ifp->if_metric;
1871
1872	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1873
1874	/*
1875	 * Special case:
1876	 * If a new destination address is specified for a point-to-point
1877	 * interface, install a route to the destination as an interface
1878	 * direct route.
1879	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1880	 * interface that share the same destination.
1881	 */
1882	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1883	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1884	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1885		int rtflags = RTF_UP | RTF_HOST;
1886
1887		error = rtrequest(RTM_ADD,
1888		    (struct sockaddr *)&ia->ia_dstaddr,
1889		    (struct sockaddr *)&ia->ia_addr,
1890		    (struct sockaddr *)&ia->ia_prefixmask,
1891		    ia->ia_flags | rtflags, NULL);
1892		if (error != 0)
1893			return (error);
1894		ia->ia_flags |= IFA_ROUTE;
1895		/*
1896		 * Handle the case for ::1 .
1897		 */
1898		if (ifp->if_flags & IFF_LOOPBACK)
1899			ia->ia_flags |= IFA_RTSELF;
1900	}
1901
1902	/*
1903	 * add a loopback route to self
1904	 */
1905	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1906		error = ifa_add_loopback_route((struct ifaddr *)ia,
1907				       (struct sockaddr *)&ia->ia_addr);
1908		if (error == 0)
1909			ia->ia_flags |= IFA_RTSELF;
1910	}
1911
1912	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1913	if (newhost)
1914		in6_ifaddloop(&(ia->ia_ifa));
1915
1916	return (error);
1917}
1918
1919/*
1920 * Find an IPv6 interface link-local address specific to an interface.
1921 * ifaddr is returned referenced.
1922 */
1923struct in6_ifaddr *
1924in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1925{
1926	struct ifaddr *ifa;
1927
1928	IF_ADDR_RLOCK(ifp);
1929	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1930		if (ifa->ifa_addr->sa_family != AF_INET6)
1931			continue;
1932		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1933			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1934			     ignoreflags) != 0)
1935				continue;
1936			ifa_ref(ifa);
1937			break;
1938		}
1939	}
1940	IF_ADDR_RUNLOCK(ifp);
1941
1942	return ((struct in6_ifaddr *)ifa);
1943}
1944
1945
1946/*
1947 * find the internet address corresponding to a given interface and address.
1948 * ifaddr is returned referenced.
1949 */
1950struct in6_ifaddr *
1951in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
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_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1960			ifa_ref(ifa);
1961			break;
1962		}
1963	}
1964	IF_ADDR_RUNLOCK(ifp);
1965
1966	return ((struct in6_ifaddr *)ifa);
1967}
1968
1969/*
1970 * Convert IP6 address to printable (loggable) representation. Caller
1971 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1972 */
1973static char digits[] = "0123456789abcdef";
1974char *
1975ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1976{
1977	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1978	char *cp;
1979	const u_int16_t *a = (const u_int16_t *)addr;
1980	const u_int8_t *d;
1981	int dcolon = 0, zero = 0;
1982
1983	cp = ip6buf;
1984
1985	for (i = 0; i < 8; i++) {
1986		if (*(a + i) == 0) {
1987			cnt++;
1988			if (cnt == 1)
1989				idx = i;
1990		}
1991		else if (maxcnt < cnt) {
1992			maxcnt = cnt;
1993			index = idx;
1994			cnt = 0;
1995		}
1996	}
1997	if (maxcnt < cnt) {
1998		maxcnt = cnt;
1999		index = idx;
2000	}
2001
2002	for (i = 0; i < 8; i++) {
2003		if (dcolon == 1) {
2004			if (*a == 0) {
2005				if (i == 7)
2006					*cp++ = ':';
2007				a++;
2008				continue;
2009			} else
2010				dcolon = 2;
2011		}
2012		if (*a == 0) {
2013			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
2014				if (i == 0)
2015					*cp++ = ':';
2016				*cp++ = ':';
2017				dcolon = 1;
2018			} else {
2019				*cp++ = '0';
2020				*cp++ = ':';
2021			}
2022			a++;
2023			continue;
2024		}
2025		d = (const u_char *)a;
2026		/* Try to eliminate leading zeros in printout like in :0001. */
2027		zero = 1;
2028		*cp = digits[*d >> 4];
2029		if (*cp != '0') {
2030			zero = 0;
2031			cp++;
2032		}
2033		*cp = digits[*d++ & 0xf];
2034		if (zero == 0 || (*cp != '0')) {
2035			zero = 0;
2036			cp++;
2037		}
2038		*cp = digits[*d >> 4];
2039		if (zero == 0 || (*cp != '0')) {
2040			zero = 0;
2041			cp++;
2042		}
2043		*cp++ = digits[*d & 0xf];
2044		*cp++ = ':';
2045		a++;
2046	}
2047	*--cp = '\0';
2048	return (ip6buf);
2049}
2050
2051int
2052in6_localaddr(struct in6_addr *in6)
2053{
2054	struct in6_ifaddr *ia;
2055
2056	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2057		return 1;
2058
2059	IN6_IFADDR_RLOCK();
2060	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2061		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2062		    &ia->ia_prefixmask.sin6_addr)) {
2063			IN6_IFADDR_RUNLOCK();
2064			return 1;
2065		}
2066	}
2067	IN6_IFADDR_RUNLOCK();
2068
2069	return (0);
2070}
2071
2072/*
2073 * Return 1 if an internet address is for the local host and configured
2074 * on one of its interfaces.
2075 */
2076int
2077in6_localip(struct in6_addr *in6)
2078{
2079	struct in6_ifaddr *ia;
2080
2081	IN6_IFADDR_RLOCK();
2082	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2083		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2084			IN6_IFADDR_RUNLOCK();
2085			return (1);
2086		}
2087	}
2088	IN6_IFADDR_RUNLOCK();
2089	return (0);
2090}
2091
2092
2093int
2094in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2095{
2096	struct in6_ifaddr *ia;
2097
2098	IN6_IFADDR_RLOCK();
2099	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2100		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2101				       &sa6->sin6_addr) &&
2102		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
2103			IN6_IFADDR_RUNLOCK();
2104			return (1); /* true */
2105		}
2106
2107		/* XXX: do we still have to go thru the rest of the list? */
2108	}
2109	IN6_IFADDR_RUNLOCK();
2110
2111	return (0);		/* false */
2112}
2113
2114/*
2115 * return length of part which dst and src are equal
2116 * hard coding...
2117 */
2118int
2119in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2120{
2121	int match = 0;
2122	u_char *s = (u_char *)src, *d = (u_char *)dst;
2123	u_char *lim = s + 16, r;
2124
2125	while (s < lim)
2126		if ((r = (*d++ ^ *s++)) != 0) {
2127			while (r < 128) {
2128				match++;
2129				r <<= 1;
2130			}
2131			break;
2132		} else
2133			match += 8;
2134	return match;
2135}
2136
2137/* XXX: to be scope conscious */
2138int
2139in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2140{
2141	int bytelen, bitlen;
2142
2143	/* sanity check */
2144	if (0 > len || len > 128) {
2145		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2146		    len);
2147		return (0);
2148	}
2149
2150	bytelen = len / 8;
2151	bitlen = len % 8;
2152
2153	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2154		return (0);
2155	if (bitlen != 0 &&
2156	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2157	    p2->s6_addr[bytelen] >> (8 - bitlen))
2158		return (0);
2159
2160	return (1);
2161}
2162
2163void
2164in6_prefixlen2mask(struct in6_addr *maskp, int len)
2165{
2166	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2167	int bytelen, bitlen, i;
2168
2169	/* sanity check */
2170	if (0 > len || len > 128) {
2171		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2172		    len);
2173		return;
2174	}
2175
2176	bzero(maskp, sizeof(*maskp));
2177	bytelen = len / 8;
2178	bitlen = len % 8;
2179	for (i = 0; i < bytelen; i++)
2180		maskp->s6_addr[i] = 0xff;
2181	if (bitlen)
2182		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2183}
2184
2185/*
2186 * return the best address out of the same scope. if no address was
2187 * found, return the first valid address from designated IF.
2188 */
2189struct in6_ifaddr *
2190in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2191{
2192	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2193	struct ifaddr *ifa;
2194	struct in6_ifaddr *besta = 0;
2195	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2196
2197	dep[0] = dep[1] = NULL;
2198
2199	/*
2200	 * We first look for addresses in the same scope.
2201	 * If there is one, return it.
2202	 * If two or more, return one which matches the dst longest.
2203	 * If none, return one of global addresses assigned other ifs.
2204	 */
2205	IF_ADDR_RLOCK(ifp);
2206	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2207		if (ifa->ifa_addr->sa_family != AF_INET6)
2208			continue;
2209		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2210			continue; /* XXX: is there any case to allow anycast? */
2211		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2212			continue; /* don't use this interface */
2213		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2214			continue;
2215		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2216			if (V_ip6_use_deprecated)
2217				dep[0] = (struct in6_ifaddr *)ifa;
2218			continue;
2219		}
2220
2221		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2222			/*
2223			 * call in6_matchlen() as few as possible
2224			 */
2225			if (besta) {
2226				if (blen == -1)
2227					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2228				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2229				if (tlen > blen) {
2230					blen = tlen;
2231					besta = (struct in6_ifaddr *)ifa;
2232				}
2233			} else
2234				besta = (struct in6_ifaddr *)ifa;
2235		}
2236	}
2237	if (besta) {
2238		ifa_ref(&besta->ia_ifa);
2239		IF_ADDR_RUNLOCK(ifp);
2240		return (besta);
2241	}
2242
2243	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2244		if (ifa->ifa_addr->sa_family != AF_INET6)
2245			continue;
2246		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2247			continue; /* XXX: is there any case to allow anycast? */
2248		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2249			continue; /* don't use this interface */
2250		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2251			continue;
2252		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2253			if (V_ip6_use_deprecated)
2254				dep[1] = (struct in6_ifaddr *)ifa;
2255			continue;
2256		}
2257
2258		if (ifa != NULL)
2259			ifa_ref(ifa);
2260		IF_ADDR_RUNLOCK(ifp);
2261		return (struct in6_ifaddr *)ifa;
2262	}
2263	IF_ADDR_RUNLOCK(ifp);
2264
2265	/* use the last-resort values, that are, deprecated addresses */
2266	if (dep[0])
2267		return dep[0];
2268	if (dep[1])
2269		return dep[1];
2270
2271	return NULL;
2272}
2273
2274/*
2275 * perform DAD when interface becomes IFF_UP.
2276 */
2277void
2278in6_if_up(struct ifnet *ifp)
2279{
2280	struct ifaddr *ifa;
2281	struct in6_ifaddr *ia;
2282
2283	IF_ADDR_RLOCK(ifp);
2284	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2285		if (ifa->ifa_addr->sa_family != AF_INET6)
2286			continue;
2287		ia = (struct in6_ifaddr *)ifa;
2288		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2289			/*
2290			 * The TENTATIVE flag was likely set by hand
2291			 * beforehand, implicitly indicating the need for DAD.
2292			 * We may be able to skip the random delay in this
2293			 * case, but we impose delays just in case.
2294			 */
2295			nd6_dad_start(ifa,
2296			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2297		}
2298	}
2299	IF_ADDR_RUNLOCK(ifp);
2300
2301	/*
2302	 * special cases, like 6to4, are handled in in6_ifattach
2303	 */
2304	in6_ifattach(ifp, NULL);
2305}
2306
2307int
2308in6if_do_dad(struct ifnet *ifp)
2309{
2310	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2311		return (0);
2312
2313	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2314		return (0);
2315
2316	switch (ifp->if_type) {
2317#ifdef IFT_DUMMY
2318	case IFT_DUMMY:
2319#endif
2320	case IFT_FAITH:
2321		/*
2322		 * These interfaces do not have the IFF_LOOPBACK flag,
2323		 * but loop packets back.  We do not have to do DAD on such
2324		 * interfaces.  We should even omit it, because loop-backed
2325		 * NS would confuse the DAD procedure.
2326		 */
2327		return (0);
2328	default:
2329		/*
2330		 * Our DAD routine requires the interface up and running.
2331		 * However, some interfaces can be up before the RUNNING
2332		 * status.  Additionaly, users may try to assign addresses
2333		 * before the interface becomes up (or running).
2334		 * We simply skip DAD in such a case as a work around.
2335		 * XXX: we should rather mark "tentative" on such addresses,
2336		 * and do DAD after the interface becomes ready.
2337		 */
2338		if (!((ifp->if_flags & IFF_UP) &&
2339		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2340			return (0);
2341
2342		return (1);
2343	}
2344}
2345
2346/*
2347 * Calculate max IPv6 MTU through all the interfaces and store it
2348 * to in6_maxmtu.
2349 */
2350void
2351in6_setmaxmtu(void)
2352{
2353	unsigned long maxmtu = 0;
2354	struct ifnet *ifp;
2355
2356	IFNET_RLOCK_NOSLEEP();
2357	TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
2358		/* this function can be called during ifnet initialization */
2359		if (!ifp->if_afdata[AF_INET6])
2360			continue;
2361		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2362		    IN6_LINKMTU(ifp) > maxmtu)
2363			maxmtu = IN6_LINKMTU(ifp);
2364	}
2365	IFNET_RUNLOCK_NOSLEEP();
2366	if (maxmtu)	     /* update only when maxmtu is positive */
2367		V_in6_maxmtu = maxmtu;
2368}
2369
2370/*
2371 * Provide the length of interface identifiers to be used for the link attached
2372 * to the given interface.  The length should be defined in "IPv6 over
2373 * xxx-link" document.  Note that address architecture might also define
2374 * the length for a particular set of address prefixes, regardless of the
2375 * link type.  As clarified in rfc2462bis, those two definitions should be
2376 * consistent, and those really are as of August 2004.
2377 */
2378int
2379in6_if2idlen(struct ifnet *ifp)
2380{
2381	switch (ifp->if_type) {
2382	case IFT_ETHER:		/* RFC2464 */
2383#ifdef IFT_PROPVIRTUAL
2384	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2385#endif
2386#ifdef IFT_L2VLAN
2387	case IFT_L2VLAN:	/* ditto */
2388#endif
2389#ifdef IFT_IEEE80211
2390	case IFT_IEEE80211:	/* ditto */
2391#endif
2392#ifdef IFT_MIP
2393	case IFT_MIP:	/* ditto */
2394#endif
2395	case IFT_INFINIBAND:
2396		return (64);
2397	case IFT_FDDI:		/* RFC2467 */
2398		return (64);
2399	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2400		return (64);
2401	case IFT_PPP:		/* RFC2472 */
2402		return (64);
2403	case IFT_ARCNET:	/* RFC2497 */
2404		return (64);
2405	case IFT_FRELAY:	/* RFC2590 */
2406		return (64);
2407	case IFT_IEEE1394:	/* RFC3146 */
2408		return (64);
2409	case IFT_GIF:
2410		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2411	case IFT_LOOP:
2412		return (64);	/* XXX: is this really correct? */
2413	default:
2414		/*
2415		 * Unknown link type:
2416		 * It might be controversial to use the today's common constant
2417		 * of 64 for these cases unconditionally.  For full compliance,
2418		 * we should return an error in this case.  On the other hand,
2419		 * if we simply miss the standard for the link type or a new
2420		 * standard is defined for a new link type, the IFID length
2421		 * is very likely to be the common constant.  As a compromise,
2422		 * we always use the constant, but make an explicit notice
2423		 * indicating the "unknown" case.
2424		 */
2425		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2426		return (64);
2427	}
2428}
2429
2430#include <sys/sysctl.h>
2431
2432struct in6_llentry {
2433	struct llentry		base;
2434	struct sockaddr_in6	l3_addr6;
2435};
2436
2437static struct llentry *
2438in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2439{
2440	struct in6_llentry *lle;
2441
2442	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE,
2443	    M_DONTWAIT | M_ZERO);
2444	if (lle == NULL)		/* NB: caller generates msg */
2445		return NULL;
2446
2447	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2448	lle->base.lle_refcnt = 1;
2449	LLE_LOCK_INIT(&lle->base);
2450	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2451	    CALLOUT_RETURNUNLOCKED);
2452
2453	return &lle->base;
2454}
2455
2456/*
2457 * Deletes an address from the address table.
2458 * This function is called by the timer functions
2459 * such as arptimer() and nd6_llinfo_timer(), and
2460 * the caller does the locking.
2461 */
2462static void
2463in6_lltable_free(struct lltable *llt, struct llentry *lle)
2464{
2465	LLE_WUNLOCK(lle);
2466	LLE_LOCK_DESTROY(lle);
2467	free(lle, M_LLTABLE);
2468}
2469
2470static void
2471in6_lltable_prefix_free(struct lltable *llt,
2472			const struct sockaddr *prefix,
2473			const struct sockaddr *mask,
2474			u_int flags)
2475{
2476	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2477	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2478	struct llentry *lle, *next;
2479	register int i;
2480
2481	/*
2482	 * (flags & LLE_STATIC) means deleting all entries
2483	 * including static ND6 entries
2484	 */
2485	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
2486		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2487			if (IN6_ARE_MASKED_ADDR_EQUAL(
2488				    &((struct sockaddr_in6 *)L3_ADDR(lle))->sin6_addr,
2489				    &pfx->sin6_addr,
2490				    &msk->sin6_addr) &&
2491			    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
2492				int canceled;
2493
2494				canceled = callout_drain(&lle->la_timer);
2495				LLE_WLOCK(lle);
2496				if (canceled)
2497					LLE_REMREF(lle);
2498				llentry_free(lle);
2499			}
2500		}
2501	}
2502}
2503
2504static int
2505in6_lltable_rtcheck(struct ifnet *ifp,
2506		    u_int flags,
2507		    const struct sockaddr *l3addr)
2508{
2509	struct rtentry *rt;
2510	char ip6buf[INET6_ADDRSTRLEN];
2511
2512	KASSERT(l3addr->sa_family == AF_INET6,
2513	    ("sin_family %d", l3addr->sa_family));
2514
2515	/* XXX rtalloc1 should take a const param */
2516	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
2517	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2518		struct ifaddr *ifa;
2519		/*
2520		 * Create an ND6 cache for an IPv6 neighbor
2521		 * that is not covered by our own prefix.
2522		 */
2523		/* XXX ifaof_ifpforaddr should take a const param */
2524		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2525		if (ifa != NULL) {
2526			ifa_free(ifa);
2527			if (rt != NULL)
2528				RTFREE_LOCKED(rt);
2529			return 0;
2530		}
2531		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2532		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2533		if (rt != NULL)
2534			RTFREE_LOCKED(rt);
2535		return EINVAL;
2536	}
2537	RTFREE_LOCKED(rt);
2538	return 0;
2539}
2540
2541static struct llentry *
2542in6_lltable_lookup(struct lltable *llt, u_int flags,
2543	const struct sockaddr *l3addr)
2544{
2545	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2546	struct ifnet *ifp = llt->llt_ifp;
2547	struct llentry *lle;
2548	struct llentries *lleh;
2549	u_int hashkey;
2550
2551	IF_AFDATA_LOCK_ASSERT(ifp);
2552	KASSERT(l3addr->sa_family == AF_INET6,
2553	    ("sin_family %d", l3addr->sa_family));
2554
2555	hashkey = sin6->sin6_addr.s6_addr32[3];
2556	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2557	LIST_FOREACH(lle, lleh, lle_next) {
2558		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2559		if (lle->la_flags & LLE_DELETED)
2560			continue;
2561		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2562			 sizeof(struct in6_addr)) == 0)
2563			break;
2564	}
2565
2566	if (lle == NULL) {
2567		if (!(flags & LLE_CREATE))
2568			return (NULL);
2569		/*
2570		 * A route that covers the given address must have
2571		 * been installed 1st because we are doing a resolution,
2572		 * verify this.
2573		 */
2574		if (!(flags & LLE_IFADDR) &&
2575		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2576			return NULL;
2577
2578		lle = in6_lltable_new(l3addr, flags);
2579		if (lle == NULL) {
2580			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2581			return NULL;
2582		}
2583		lle->la_flags = flags & ~LLE_CREATE;
2584		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2585			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2586			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2587		}
2588
2589		lle->lle_tbl  = llt;
2590		lle->lle_head = lleh;
2591		LIST_INSERT_HEAD(lleh, lle, lle_next);
2592	} else if (flags & LLE_DELETE) {
2593		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2594			LLE_WLOCK(lle);
2595			lle->la_flags = LLE_DELETED;
2596			LLE_WUNLOCK(lle);
2597#ifdef DIAGNOSTIC
2598			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2599#endif
2600		}
2601		lle = (void *)-1;
2602	}
2603	if (LLE_IS_VALID(lle)) {
2604		if (flags & LLE_EXCLUSIVE)
2605			LLE_WLOCK(lle);
2606		else
2607			LLE_RLOCK(lle);
2608	}
2609	return (lle);
2610}
2611
2612static int
2613in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2614{
2615	struct ifnet *ifp = llt->llt_ifp;
2616	struct llentry *lle;
2617	/* XXX stack use */
2618	struct {
2619		struct rt_msghdr	rtm;
2620		struct sockaddr_in6	sin6;
2621		/*
2622		 * ndp.c assumes that sdl is word aligned
2623		 */
2624#ifdef __LP64__
2625		uint32_t		pad;
2626#endif
2627		struct sockaddr_dl	sdl;
2628	} ndpc;
2629	int i, error;
2630
2631	if (ifp->if_flags & IFF_LOOPBACK)
2632		return 0;
2633
2634	LLTABLE_LOCK_ASSERT();
2635
2636	error = 0;
2637	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2638		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2639			struct sockaddr_dl *sdl;
2640
2641			/* skip deleted or invalid entries */
2642			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2643				continue;
2644			/* Skip if jailed and not a valid IP of the prison. */
2645			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2646				continue;
2647			/*
2648			 * produce a msg made of:
2649			 *  struct rt_msghdr;
2650			 *  struct sockaddr_in6 (IPv6)
2651			 *  struct sockaddr_dl;
2652			 */
2653			bzero(&ndpc, sizeof(ndpc));
2654			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2655			ndpc.rtm.rtm_version = RTM_VERSION;
2656			ndpc.rtm.rtm_type = RTM_GET;
2657			ndpc.rtm.rtm_flags = RTF_UP;
2658			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2659			ndpc.sin6.sin6_family = AF_INET6;
2660			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2661			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2662
2663			/* publish */
2664			if (lle->la_flags & LLE_PUB)
2665				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2666
2667			sdl = &ndpc.sdl;
2668			sdl->sdl_family = AF_LINK;
2669			sdl->sdl_len = sizeof(*sdl);
2670			sdl->sdl_alen = ifp->if_addrlen;
2671			sdl->sdl_index = ifp->if_index;
2672			sdl->sdl_type = ifp->if_type;
2673			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2674			ndpc.rtm.rtm_rmx.rmx_expire =
2675			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2676			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2677			if (lle->la_flags & LLE_STATIC)
2678				ndpc.rtm.rtm_flags |= RTF_STATIC;
2679			ndpc.rtm.rtm_index = ifp->if_index;
2680			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2681			if (error)
2682				break;
2683		}
2684	}
2685	return error;
2686}
2687
2688void *
2689in6_domifattach(struct ifnet *ifp)
2690{
2691	struct in6_ifextra *ext;
2692
2693	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2694	bzero(ext, sizeof(*ext));
2695
2696	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2697	    M_IFADDR, M_WAITOK);
2698	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2699
2700	ext->icmp6_ifstat =
2701	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2702	    M_IFADDR, M_WAITOK);
2703	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2704
2705	ext->nd_ifinfo = nd6_ifattach(ifp);
2706	ext->scope6_id = scope6_ifattach(ifp);
2707	ext->lltable = lltable_init(ifp, AF_INET6);
2708	if (ext->lltable != NULL) {
2709		ext->lltable->llt_free = in6_lltable_free;
2710		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2711		ext->lltable->llt_lookup = in6_lltable_lookup;
2712		ext->lltable->llt_dump = in6_lltable_dump;
2713	}
2714
2715	ext->mld_ifinfo = mld_domifattach(ifp);
2716
2717	return ext;
2718}
2719
2720void
2721in6_domifdetach(struct ifnet *ifp, void *aux)
2722{
2723	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2724
2725	mld_domifdetach(ifp);
2726	scope6_ifdetach(ext->scope6_id);
2727	nd6_ifdetach(ext->nd_ifinfo);
2728	lltable_free(ext->lltable);
2729	free(ext->in6_ifstat, M_IFADDR);
2730	free(ext->icmp6_ifstat, M_IFADDR);
2731	free(ext, M_IFADDR);
2732}
2733
2734/*
2735 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2736 * v4 mapped addr or v4 compat addr
2737 */
2738void
2739in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2740{
2741
2742	bzero(sin, sizeof(*sin));
2743	sin->sin_len = sizeof(struct sockaddr_in);
2744	sin->sin_family = AF_INET;
2745	sin->sin_port = sin6->sin6_port;
2746	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2747}
2748
2749/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2750void
2751in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2752{
2753	bzero(sin6, sizeof(*sin6));
2754	sin6->sin6_len = sizeof(struct sockaddr_in6);
2755	sin6->sin6_family = AF_INET6;
2756	sin6->sin6_port = sin->sin_port;
2757	sin6->sin6_addr.s6_addr32[0] = 0;
2758	sin6->sin6_addr.s6_addr32[1] = 0;
2759	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2760	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2761}
2762
2763/* Convert sockaddr_in6 into sockaddr_in. */
2764void
2765in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2766{
2767	struct sockaddr_in *sin_p;
2768	struct sockaddr_in6 sin6;
2769
2770	/*
2771	 * Save original sockaddr_in6 addr and convert it
2772	 * to sockaddr_in.
2773	 */
2774	sin6 = *(struct sockaddr_in6 *)nam;
2775	sin_p = (struct sockaddr_in *)nam;
2776	in6_sin6_2_sin(sin_p, &sin6);
2777}
2778
2779/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2780void
2781in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2782{
2783	struct sockaddr_in *sin_p;
2784	struct sockaddr_in6 *sin6_p;
2785
2786	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
2787	       M_WAITOK);
2788	sin_p = (struct sockaddr_in *)*nam;
2789	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2790	free(*nam, M_SONAME);
2791	*nam = (struct sockaddr *)sin6_p;
2792}
2793