in6.c revision 1021d43b569bfc8d2c5544bde2f540fa432b011f
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_inet.h"
67#include "opt_inet6.h"
68
69#include <sys/param.h>
70#include <sys/errno.h>
71#include <sys/malloc.h>
72#include <sys/socket.h>
73#include <sys/socketvar.h>
74#include <sys/sockio.h>
75#include <sys/systm.h>
76#include <sys/priv.h>
77#include <sys/proc.h>
78#include <sys/time.h>
79#include <sys/kernel.h>
80#include <sys/syslog.h>
81#include <sys/vimage.h>
82
83#include <net/if.h>
84#include <net/if_types.h>
85#include <net/route.h>
86#include <net/if_dl.h>
87
88#include <netinet/in.h>
89#include <netinet/in_var.h>
90#include <netinet/if_ether.h>
91#include <netinet/in_systm.h>
92#include <netinet/ip.h>
93#include <netinet/in_pcb.h>
94
95#include <netinet/ip6.h>
96#include <netinet6/ip6_var.h>
97#include <netinet6/nd6.h>
98#include <netinet6/mld6_var.h>
99#include <netinet6/ip6_mroute.h>
100#include <netinet6/in6_ifattach.h>
101#include <netinet6/scope6_var.h>
102#include <netinet6/in6_pcb.h>
103
104MALLOC_DEFINE(M_IP6MADDR, "in6_multi", "internet multicast address");
105
106/*
107 * Definitions of some costant IP6 addresses.
108 */
109const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
110const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
111const struct in6_addr in6addr_nodelocal_allnodes =
112	IN6ADDR_NODELOCAL_ALLNODES_INIT;
113const struct in6_addr in6addr_linklocal_allnodes =
114	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
115const struct in6_addr in6addr_linklocal_allrouters =
116	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
117
118const struct in6_addr in6mask0 = IN6MASK0;
119const struct in6_addr in6mask32 = IN6MASK32;
120const struct in6_addr in6mask64 = IN6MASK64;
121const struct in6_addr in6mask96 = IN6MASK96;
122const struct in6_addr in6mask128 = IN6MASK128;
123
124const struct sockaddr_in6 sa6_any =
125	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
126
127static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
128	struct ifnet *, struct thread *));
129static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
130	struct sockaddr_in6 *, int));
131static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
132
133struct in6_multihead in6_multihead;	/* XXX BSS initialization */
134int	(*faithprefix_p)(struct in6_addr *);
135
136/*
137 * Subroutine for in6_ifaddloop() and in6_ifremloop().
138 * This routine does actual work.
139 */
140static void
141in6_ifloop_request(int cmd, struct ifaddr *ifa)
142{
143	struct sockaddr_in6 all1_sa;
144	struct rtentry *nrt = NULL;
145	int e;
146	char ip6buf[INET6_ADDRSTRLEN];
147
148	bzero(&all1_sa, sizeof(all1_sa));
149	all1_sa.sin6_family = AF_INET6;
150	all1_sa.sin6_len = sizeof(struct sockaddr_in6);
151	all1_sa.sin6_addr = in6mask128;
152
153	/*
154	 * We specify the address itself as the gateway, and set the
155	 * RTF_LLINFO flag, so that the corresponding host route would have
156	 * the flag, and thus applications that assume traditional behavior
157	 * would be happy.  Note that we assume the caller of the function
158	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
159	 * which changes the outgoing interface to the loopback interface.
160	 */
161	e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr,
162	    (struct sockaddr *)&all1_sa, RTF_UP|RTF_HOST|RTF_LLINFO, &nrt);
163	if (e != 0) {
164		/* XXX need more descriptive message */
165
166		log(LOG_ERR, "in6_ifloop_request: "
167		    "%s operation failed for %s (errno=%d)\n",
168		    cmd == RTM_ADD ? "ADD" : "DELETE",
169		    ip6_sprintf(ip6buf,
170			    &((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr), e);
171	}
172
173	/*
174	 * Report the addition/removal of the address to the routing socket.
175	 * XXX: since we called rtinit for a p2p interface with a destination,
176	 *      we end up reporting twice in such a case.  Should we rather
177	 *      omit the second report?
178	 */
179	if (nrt) {
180		RT_LOCK(nrt);
181		/*
182		 * Make sure rt_ifa be equal to IFA, the second argument of
183		 * the function.  We need this because when we refer to
184		 * rt_ifa->ia6_flags in ip6_input, we assume that the rt_ifa
185		 * points to the address instead of the loopback address.
186		 */
187		if (cmd == RTM_ADD && ifa != nrt->rt_ifa) {
188			IFAFREE(nrt->rt_ifa);
189			IFAREF(ifa);
190			nrt->rt_ifa = ifa;
191		}
192
193		rt_newaddrmsg(cmd, ifa, e, nrt);
194		if (cmd == RTM_DELETE)
195			RTFREE_LOCKED(nrt);
196		else {
197			/* the cmd must be RTM_ADD here */
198			RT_REMREF(nrt);
199			RT_UNLOCK(nrt);
200		}
201	}
202}
203
204/*
205 * Add ownaddr as loopback rtentry.  We previously add the route only if
206 * necessary (ex. on a p2p link).  However, since we now manage addresses
207 * separately from prefixes, we should always add the route.  We can't
208 * rely on the cloning mechanism from the corresponding interface route
209 * any more.
210 */
211void
212in6_ifaddloop(struct ifaddr *ifa)
213{
214	struct rtentry *rt;
215	int need_loop;
216
217	/* If there is no loopback entry, allocate one. */
218	rt = rtalloc1(ifa->ifa_addr, 0, 0);
219	need_loop = (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
220	    (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0);
221	if (rt)
222		RTFREE_LOCKED(rt);
223	if (need_loop)
224		in6_ifloop_request(RTM_ADD, ifa);
225}
226
227/*
228 * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
229 * if it exists.
230 */
231void
232in6_ifremloop(struct ifaddr *ifa)
233{
234	struct in6_ifaddr *ia;
235	struct rtentry *rt;
236	int ia_count = 0;
237
238	/*
239	 * Some of BSD variants do not remove cloned routes
240	 * from an interface direct route, when removing the direct route
241	 * (see comments in net/net_osdep.h).  Even for variants that do remove
242	 * cloned routes, they could fail to remove the cloned routes when
243	 * we handle multple addresses that share a common prefix.
244	 * So, we should remove the route corresponding to the deleted address.
245	 */
246
247	/*
248	 * Delete the entry only if exact one ifa exists.  More than one ifa
249	 * can exist if we assign a same single address to multiple
250	 * (probably p2p) interfaces.
251	 * XXX: we should avoid such a configuration in IPv6...
252	 */
253	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
254		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
255			ia_count++;
256			if (ia_count > 1)
257				break;
258		}
259	}
260
261	if (ia_count == 1) {
262		/*
263		 * Before deleting, check if a corresponding loopbacked host
264		 * route surely exists.  With this check, we can avoid to
265		 * delete an interface direct route whose destination is same
266		 * as the address being removed.  This can happen when removing
267		 * a subnet-router anycast address on an interface attahced
268		 * to a shared medium.
269		 */
270		rt = rtalloc1(ifa->ifa_addr, 0, 0);
271		if (rt != NULL) {
272			if ((rt->rt_flags & RTF_HOST) != 0 &&
273			    (rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
274				RTFREE_LOCKED(rt);
275				in6_ifloop_request(RTM_DELETE, ifa);
276			} else
277				RT_UNLOCK(rt);
278		}
279	}
280}
281
282int
283in6_mask2len(struct in6_addr *mask, u_char *lim0)
284{
285	int x = 0, y;
286	u_char *lim = lim0, *p;
287
288	/* ignore the scope_id part */
289	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
290		lim = (u_char *)mask + sizeof(*mask);
291	for (p = (u_char *)mask; p < lim; x++, p++) {
292		if (*p != 0xff)
293			break;
294	}
295	y = 0;
296	if (p < lim) {
297		for (y = 0; y < 8; y++) {
298			if ((*p & (0x80 >> y)) == 0)
299				break;
300		}
301	}
302
303	/*
304	 * when the limit pointer is given, do a stricter check on the
305	 * remaining bits.
306	 */
307	if (p < lim) {
308		if (y != 0 && (*p & (0x00ff >> y)) != 0)
309			return (-1);
310		for (p = p + 1; p < lim; p++)
311			if (*p != 0)
312				return (-1);
313	}
314
315	return x * 8 + y;
316}
317
318#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
319#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
320
321int
322in6_control(struct socket *so, u_long cmd, caddr_t data,
323    struct ifnet *ifp, struct thread *td)
324{
325	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
326	struct	in6_ifaddr *ia = NULL;
327	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
328	struct sockaddr_in6 *sa6;
329	int error;
330
331	switch (cmd) {
332	case SIOCGETSGCNT_IN6:
333	case SIOCGETMIFCNT_IN6:
334		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
335	}
336
337	switch(cmd) {
338	case SIOCAADDRCTL_POLICY:
339	case SIOCDADDRCTL_POLICY:
340		if (td != NULL) {
341			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
342			if (error)
343				return (error);
344		}
345		return (in6_src_ioctl(cmd, data));
346	}
347
348	if (ifp == NULL)
349		return (EOPNOTSUPP);
350
351	switch (cmd) {
352	case SIOCSNDFLUSH_IN6:
353	case SIOCSPFXFLUSH_IN6:
354	case SIOCSRTRFLUSH_IN6:
355	case SIOCSDEFIFACE_IN6:
356	case SIOCSIFINFO_FLAGS:
357		if (td != NULL) {
358			error = priv_check(td, PRIV_NETINET_ND6);
359			if (error)
360				return (error);
361		}
362		/* FALLTHROUGH */
363	case OSIOCGIFINFO_IN6:
364	case SIOCGIFINFO_IN6:
365	case SIOCSIFINFO_IN6:
366	case SIOCGDRLST_IN6:
367	case SIOCGPRLST_IN6:
368	case SIOCGNBRINFO_IN6:
369	case SIOCGDEFIFACE_IN6:
370		return (nd6_ioctl(cmd, data, ifp));
371	}
372
373	switch (cmd) {
374	case SIOCSIFPREFIX_IN6:
375	case SIOCDIFPREFIX_IN6:
376	case SIOCAIFPREFIX_IN6:
377	case SIOCCIFPREFIX_IN6:
378	case SIOCSGIFPREFIX_IN6:
379	case SIOCGIFPREFIX_IN6:
380		log(LOG_NOTICE,
381		    "prefix ioctls are now invalidated. "
382		    "please use ifconfig.\n");
383		return (EOPNOTSUPP);
384	}
385
386	switch (cmd) {
387	case SIOCSSCOPE6:
388		if (td != NULL) {
389			error = priv_check(td, PRIV_NETINET_SCOPE6);
390			if (error)
391				return (error);
392		}
393		return (scope6_set(ifp,
394		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
395	case SIOCGSCOPE6:
396		return (scope6_get(ifp,
397		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
398	case SIOCGSCOPE6DEF:
399		return (scope6_get_default((struct scope6_id *)
400		    ifr->ifr_ifru.ifru_scope_id));
401	}
402
403	switch (cmd) {
404	case SIOCALIFADDR:
405		if (td != NULL) {
406			error = priv_check(td, PRIV_NET_ADDIFADDR);
407			if (error)
408				return (error);
409		}
410		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
411
412	case SIOCDLIFADDR:
413		if (td != NULL) {
414			error = priv_check(td, PRIV_NET_DELIFADDR);
415			if (error)
416				return (error);
417		}
418		/* FALLTHROUGH */
419	case SIOCGLIFADDR:
420		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
421	}
422
423	/*
424	 * Find address for this interface, if it exists.
425	 *
426	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
427	 * only, and used the first interface address as the target of other
428	 * operations (without checking ifra_addr).  This was because netinet
429	 * code/API assumed at most 1 interface address per interface.
430	 * Since IPv6 allows a node to assign multiple addresses
431	 * on a single interface, we almost always look and check the
432	 * presence of ifra_addr, and reject invalid ones here.
433	 * It also decreases duplicated code among SIOC*_IN6 operations.
434	 */
435	switch (cmd) {
436	case SIOCAIFADDR_IN6:
437	case SIOCSIFPHYADDR_IN6:
438		sa6 = &ifra->ifra_addr;
439		break;
440	case SIOCSIFADDR_IN6:
441	case SIOCGIFADDR_IN6:
442	case SIOCSIFDSTADDR_IN6:
443	case SIOCSIFNETMASK_IN6:
444	case SIOCGIFDSTADDR_IN6:
445	case SIOCGIFNETMASK_IN6:
446	case SIOCDIFADDR_IN6:
447	case SIOCGIFPSRCADDR_IN6:
448	case SIOCGIFPDSTADDR_IN6:
449	case SIOCGIFAFLAG_IN6:
450	case SIOCSNDFLUSH_IN6:
451	case SIOCSPFXFLUSH_IN6:
452	case SIOCSRTRFLUSH_IN6:
453	case SIOCGIFALIFETIME_IN6:
454	case SIOCSIFALIFETIME_IN6:
455	case SIOCGIFSTAT_IN6:
456	case SIOCGIFSTAT_ICMP6:
457		sa6 = &ifr->ifr_addr;
458		break;
459	default:
460		sa6 = NULL;
461		break;
462	}
463	if (sa6 && sa6->sin6_family == AF_INET6) {
464		int error = 0;
465
466		if (sa6->sin6_scope_id != 0)
467			error = sa6_embedscope(sa6, 0);
468		else
469			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
470		if (error != 0)
471			return (error);
472		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
473	} else
474		ia = NULL;
475
476	switch (cmd) {
477	case SIOCSIFADDR_IN6:
478	case SIOCSIFDSTADDR_IN6:
479	case SIOCSIFNETMASK_IN6:
480		/*
481		 * Since IPv6 allows a node to assign multiple addresses
482		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
483		 */
484		/* we decided to obsolete this command (20000704) */
485		return (EINVAL);
486
487	case SIOCDIFADDR_IN6:
488		/*
489		 * for IPv4, we look for existing in_ifaddr here to allow
490		 * "ifconfig if0 delete" to remove the first IPv4 address on
491		 * the interface.  For IPv6, as the spec allows multiple
492		 * interface address from the day one, we consider "remove the
493		 * first one" semantics to be not preferable.
494		 */
495		if (ia == NULL)
496			return (EADDRNOTAVAIL);
497		/* FALLTHROUGH */
498	case SIOCAIFADDR_IN6:
499		/*
500		 * We always require users to specify a valid IPv6 address for
501		 * the corresponding operation.
502		 */
503		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
504		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
505			return (EAFNOSUPPORT);
506
507		if (td != NULL) {
508			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
509			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
510			if (error)
511				return (error);
512		}
513
514		break;
515
516	case SIOCGIFADDR_IN6:
517		/* This interface is basically deprecated. use SIOCGIFCONF. */
518		/* FALLTHROUGH */
519	case SIOCGIFAFLAG_IN6:
520	case SIOCGIFNETMASK_IN6:
521	case SIOCGIFDSTADDR_IN6:
522	case SIOCGIFALIFETIME_IN6:
523		/* must think again about its semantics */
524		if (ia == NULL)
525			return (EADDRNOTAVAIL);
526		break;
527	case SIOCSIFALIFETIME_IN6:
528	    {
529		struct in6_addrlifetime *lt;
530
531		if (td != NULL) {
532			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
533			if (error)
534				return (error);
535		}
536		if (ia == NULL)
537			return (EADDRNOTAVAIL);
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_second < time_second) {
542			return EINVAL;
543		}
544		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
545		    lt->ia6t_pltime + time_second < time_second) {
546			return EINVAL;
547		}
548		break;
549	    }
550	}
551
552	switch (cmd) {
553
554	case SIOCGIFADDR_IN6:
555		ifr->ifr_addr = ia->ia_addr;
556		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
557			return (error);
558		break;
559
560	case SIOCGIFDSTADDR_IN6:
561		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
562			return (EINVAL);
563		/*
564		 * XXX: should we check if ifa_dstaddr is NULL and return
565		 * an error?
566		 */
567		ifr->ifr_dstaddr = ia->ia_dstaddr;
568		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
569			return (error);
570		break;
571
572	case SIOCGIFNETMASK_IN6:
573		ifr->ifr_addr = ia->ia_prefixmask;
574		break;
575
576	case SIOCGIFAFLAG_IN6:
577		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
578		break;
579
580	case SIOCGIFSTAT_IN6:
581		if (ifp == NULL)
582			return EINVAL;
583		bzero(&ifr->ifr_ifru.ifru_stat,
584		    sizeof(ifr->ifr_ifru.ifru_stat));
585		ifr->ifr_ifru.ifru_stat =
586		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
587		break;
588
589	case SIOCGIFSTAT_ICMP6:
590		if (ifp == NULL)
591			return EINVAL;
592		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
593		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
594		ifr->ifr_ifru.ifru_icmp6stat =
595		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
596		break;
597
598	case SIOCGIFALIFETIME_IN6:
599		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
600		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
601			time_t maxexpire;
602			struct in6_addrlifetime *retlt =
603			    &ifr->ifr_ifru.ifru_lifetime;
604
605			/*
606			 * XXX: adjust expiration time assuming time_t is
607			 * signed.
608			 */
609			maxexpire = (-1) &
610			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
611			if (ia->ia6_lifetime.ia6t_vltime <
612			    maxexpire - ia->ia6_updatetime) {
613				retlt->ia6t_expire = ia->ia6_updatetime +
614				    ia->ia6_lifetime.ia6t_vltime;
615			} else
616				retlt->ia6t_expire = maxexpire;
617		}
618		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
619			time_t maxexpire;
620			struct in6_addrlifetime *retlt =
621			    &ifr->ifr_ifru.ifru_lifetime;
622
623			/*
624			 * XXX: adjust expiration time assuming time_t is
625			 * signed.
626			 */
627			maxexpire = (-1) &
628			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
629			if (ia->ia6_lifetime.ia6t_pltime <
630			    maxexpire - ia->ia6_updatetime) {
631				retlt->ia6t_preferred = ia->ia6_updatetime +
632				    ia->ia6_lifetime.ia6t_pltime;
633			} else
634				retlt->ia6t_preferred = maxexpire;
635		}
636		break;
637
638	case SIOCSIFALIFETIME_IN6:
639		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
640		/* for sanity */
641		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
642			ia->ia6_lifetime.ia6t_expire =
643				time_second + ia->ia6_lifetime.ia6t_vltime;
644		} else
645			ia->ia6_lifetime.ia6t_expire = 0;
646		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
647			ia->ia6_lifetime.ia6t_preferred =
648				time_second + ia->ia6_lifetime.ia6t_pltime;
649		} else
650			ia->ia6_lifetime.ia6t_preferred = 0;
651		break;
652
653	case SIOCAIFADDR_IN6:
654	{
655		int i, error = 0;
656		struct nd_prefixctl pr0;
657		struct nd_prefix *pr;
658
659		/*
660		 * first, make or update the interface address structure,
661		 * and link it to the list.
662		 */
663		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
664			return (error);
665		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
666		    == NULL) {
667			/*
668			 * this can happen when the user specify the 0 valid
669			 * lifetime.
670			 */
671			break;
672		}
673
674		/*
675		 * then, make the prefix on-link on the interface.
676		 * XXX: we'd rather create the prefix before the address, but
677		 * we need at least one address to install the corresponding
678		 * interface route, so we configure the address first.
679		 */
680
681		/*
682		 * convert mask to prefix length (prefixmask has already
683		 * been validated in in6_update_ifa().
684		 */
685		bzero(&pr0, sizeof(pr0));
686		pr0.ndpr_ifp = ifp;
687		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
688		    NULL);
689		if (pr0.ndpr_plen == 128) {
690			break;	/* we don't need to install a host route. */
691		}
692		pr0.ndpr_prefix = ifra->ifra_addr;
693		/* apply the mask for safety. */
694		for (i = 0; i < 4; i++) {
695			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
696			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
697		}
698		/*
699		 * XXX: since we don't have an API to set prefix (not address)
700		 * lifetimes, we just use the same lifetimes as addresses.
701		 * The (temporarily) installed lifetimes can be overridden by
702		 * later advertised RAs (when accept_rtadv is non 0), which is
703		 * an intended behavior.
704		 */
705		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
706		pr0.ndpr_raf_auto =
707		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
708		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
709		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
710
711		/* add the prefix if not yet. */
712		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
713			/*
714			 * nd6_prelist_add will install the corresponding
715			 * interface route.
716			 */
717			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
718				return (error);
719			if (pr == NULL) {
720				log(LOG_ERR, "nd6_prelist_add succeeded but "
721				    "no prefix\n");
722				return (EINVAL); /* XXX panic here? */
723			}
724		}
725
726		/* relate the address to the prefix */
727		if (ia->ia6_ndpr == NULL) {
728			ia->ia6_ndpr = pr;
729			pr->ndpr_refcnt++;
730
731			/*
732			 * If this is the first autoconf address from the
733			 * prefix, create a temporary address as well
734			 * (when required).
735			 */
736			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
737			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
738				int e;
739				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
740					log(LOG_NOTICE, "in6_control: failed "
741					    "to create a temporary address, "
742					    "errno=%d\n", e);
743				}
744			}
745		}
746
747		/*
748		 * this might affect the status of autoconfigured addresses,
749		 * that is, this address might make other addresses detached.
750		 */
751		pfxlist_onlink_check();
752		if (error == 0 && ia)
753			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
754		break;
755	}
756
757	case SIOCDIFADDR_IN6:
758	{
759		struct nd_prefix *pr;
760
761		/*
762		 * If the address being deleted is the only one that owns
763		 * the corresponding prefix, expire the prefix as well.
764		 * XXX: theoretically, we don't have to worry about such
765		 * relationship, since we separate the address management
766		 * and the prefix management.  We do this, however, to provide
767		 * as much backward compatibility as possible in terms of
768		 * the ioctl operation.
769		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
770		 */
771		pr = ia->ia6_ndpr;
772		in6_purgeaddr(&ia->ia_ifa);
773		if (pr && pr->ndpr_refcnt == 0)
774			prelist_remove(pr);
775		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
776		break;
777	}
778
779	default:
780		if (ifp == NULL || ifp->if_ioctl == 0)
781			return (EOPNOTSUPP);
782		return ((*ifp->if_ioctl)(ifp, cmd, data));
783	}
784
785	return (0);
786}
787
788/*
789 * Update parameters of an IPv6 interface address.
790 * If necessary, a new entry is created and linked into address chains.
791 * This function is separated from in6_control().
792 * XXX: should this be performed under splnet()?
793 */
794int
795in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
796    struct in6_ifaddr *ia, int flags)
797{
798	int error = 0, hostIsNew = 0, plen = -1;
799	struct in6_ifaddr *oia;
800	struct sockaddr_in6 dst6;
801	struct in6_addrlifetime *lt;
802	struct in6_multi_mship *imm;
803	struct in6_multi *in6m_sol;
804	struct rtentry *rt;
805	int delay;
806	char ip6buf[INET6_ADDRSTRLEN];
807
808	/* Validate parameters */
809	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
810		return (EINVAL);
811
812	/*
813	 * The destination address for a p2p link must have a family
814	 * of AF_UNSPEC or AF_INET6.
815	 */
816	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
817	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
818	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
819		return (EAFNOSUPPORT);
820	/*
821	 * validate ifra_prefixmask.  don't check sin6_family, netmask
822	 * does not carry fields other than sin6_len.
823	 */
824	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
825		return (EINVAL);
826	/*
827	 * Because the IPv6 address architecture is classless, we require
828	 * users to specify a (non 0) prefix length (mask) for a new address.
829	 * We also require the prefix (when specified) mask is valid, and thus
830	 * reject a non-consecutive mask.
831	 */
832	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
833		return (EINVAL);
834	if (ifra->ifra_prefixmask.sin6_len != 0) {
835		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
836		    (u_char *)&ifra->ifra_prefixmask +
837		    ifra->ifra_prefixmask.sin6_len);
838		if (plen <= 0)
839			return (EINVAL);
840	} else {
841		/*
842		 * In this case, ia must not be NULL.  We just use its prefix
843		 * length.
844		 */
845		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
846	}
847	/*
848	 * If the destination address on a p2p interface is specified,
849	 * and the address is a scoped one, validate/set the scope
850	 * zone identifier.
851	 */
852	dst6 = ifra->ifra_dstaddr;
853	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
854	    (dst6.sin6_family == AF_INET6)) {
855		struct in6_addr in6_tmp;
856		u_int32_t zoneid;
857
858		in6_tmp = dst6.sin6_addr;
859		if (in6_setscope(&in6_tmp, ifp, &zoneid))
860			return (EINVAL); /* XXX: should be impossible */
861
862		if (dst6.sin6_scope_id != 0) {
863			if (dst6.sin6_scope_id != zoneid)
864				return (EINVAL);
865		} else		/* user omit to specify the ID. */
866			dst6.sin6_scope_id = zoneid;
867
868		/* convert into the internal form */
869		if (sa6_embedscope(&dst6, 0))
870			return (EINVAL); /* XXX: should be impossible */
871	}
872	/*
873	 * The destination address can be specified only for a p2p or a
874	 * loopback interface.  If specified, the corresponding prefix length
875	 * must be 128.
876	 */
877	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
878		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
879			/* XXX: noisy message */
880			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
881			    "be specified for a p2p or a loopback IF only\n"));
882			return (EINVAL);
883		}
884		if (plen != 128) {
885			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
886			    "be 128 when dstaddr is specified\n"));
887			return (EINVAL);
888		}
889	}
890	/* lifetime consistency check */
891	lt = &ifra->ifra_lifetime;
892	if (lt->ia6t_pltime > lt->ia6t_vltime)
893		return (EINVAL);
894	if (lt->ia6t_vltime == 0) {
895		/*
896		 * the following log might be noisy, but this is a typical
897		 * configuration mistake or a tool's bug.
898		 */
899		nd6log((LOG_INFO,
900		    "in6_update_ifa: valid lifetime is 0 for %s\n",
901		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
902
903		if (ia == NULL)
904			return (0); /* there's nothing to do */
905	}
906
907	/*
908	 * If this is a new address, allocate a new ifaddr and link it
909	 * into chains.
910	 */
911	if (ia == NULL) {
912		hostIsNew = 1;
913		/*
914		 * When in6_update_ifa() is called in a process of a received
915		 * RA, it is called under an interrupt context.  So, we should
916		 * call malloc with M_NOWAIT.
917		 */
918		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
919		    M_NOWAIT);
920		if (ia == NULL)
921			return (ENOBUFS);
922		bzero((caddr_t)ia, sizeof(*ia));
923		LIST_INIT(&ia->ia6_memberships);
924		/* Initialize the address and masks, and put time stamp */
925		IFA_LOCK_INIT(&ia->ia_ifa);
926		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
927		ia->ia_addr.sin6_family = AF_INET6;
928		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
929		ia->ia6_createtime = time_second;
930		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
931			/*
932			 * XXX: some functions expect that ifa_dstaddr is not
933			 * NULL for p2p interfaces.
934			 */
935			ia->ia_ifa.ifa_dstaddr =
936			    (struct sockaddr *)&ia->ia_dstaddr;
937		} else {
938			ia->ia_ifa.ifa_dstaddr = NULL;
939		}
940		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
941
942		ia->ia_ifp = ifp;
943		if ((oia = V_in6_ifaddr) != NULL) {
944			for ( ; oia->ia_next; oia = oia->ia_next)
945				continue;
946			oia->ia_next = ia;
947		} else
948			V_in6_ifaddr = ia;
949
950		ia->ia_ifa.ifa_refcnt = 1;
951		TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
952	}
953
954	/* update timestamp */
955	ia->ia6_updatetime = time_second;
956
957	/* set prefix mask */
958	if (ifra->ifra_prefixmask.sin6_len) {
959		/*
960		 * We prohibit changing the prefix length of an existing
961		 * address, because
962		 * + such an operation should be rare in IPv6, and
963		 * + the operation would confuse prefix management.
964		 */
965		if (ia->ia_prefixmask.sin6_len &&
966		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
967			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
968			    " existing (%s) address should not be changed\n",
969			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
970			error = EINVAL;
971			goto unlink;
972		}
973		ia->ia_prefixmask = ifra->ifra_prefixmask;
974	}
975
976	/*
977	 * If a new destination address is specified, scrub the old one and
978	 * install the new destination.  Note that the interface must be
979	 * p2p or loopback (see the check above.)
980	 */
981	if (dst6.sin6_family == AF_INET6 &&
982	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
983		int e;
984
985		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
986		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
987			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
988			    "a route to the old destination: %s\n",
989			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
990			/* proceed anyway... */
991		} else
992			ia->ia_flags &= ~IFA_ROUTE;
993		ia->ia_dstaddr = dst6;
994	}
995
996	/*
997	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
998	 * to see if the address is deprecated or invalidated, but initialize
999	 * these members for applications.
1000	 */
1001	ia->ia6_lifetime = ifra->ifra_lifetime;
1002	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1003		ia->ia6_lifetime.ia6t_expire =
1004		    time_second + ia->ia6_lifetime.ia6t_vltime;
1005	} else
1006		ia->ia6_lifetime.ia6t_expire = 0;
1007	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1008		ia->ia6_lifetime.ia6t_preferred =
1009		    time_second + ia->ia6_lifetime.ia6t_pltime;
1010	} else
1011		ia->ia6_lifetime.ia6t_preferred = 0;
1012
1013	/* reset the interface and routing table appropriately. */
1014	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1015		goto unlink;
1016
1017	/*
1018	 * configure address flags.
1019	 */
1020	ia->ia6_flags = ifra->ifra_flags;
1021	/*
1022	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1023	 * userland, make it deprecated.
1024	 */
1025	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1026		ia->ia6_lifetime.ia6t_pltime = 0;
1027		ia->ia6_lifetime.ia6t_preferred = time_second;
1028	}
1029	/*
1030	 * Make the address tentative before joining multicast addresses,
1031	 * so that corresponding MLD responses would not have a tentative
1032	 * source address.
1033	 */
1034	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1035	if (hostIsNew && in6if_do_dad(ifp))
1036		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1037
1038	/*
1039	 * We are done if we have simply modified an existing address.
1040	 */
1041	if (!hostIsNew)
1042		return (error);
1043
1044	/*
1045	 * Beyond this point, we should call in6_purgeaddr upon an error,
1046	 * not just go to unlink.
1047	 */
1048
1049	/* Join necessary multicast groups */
1050	in6m_sol = NULL;
1051	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1052		struct sockaddr_in6 mltaddr, mltmask;
1053		struct in6_addr llsol;
1054
1055		/* join solicited multicast addr for new host id */
1056		bzero(&llsol, sizeof(struct in6_addr));
1057		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
1058		llsol.s6_addr32[1] = 0;
1059		llsol.s6_addr32[2] = htonl(1);
1060		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1061		llsol.s6_addr8[12] = 0xff;
1062		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1063			/* XXX: should not happen */
1064			log(LOG_ERR, "in6_update_ifa: "
1065			    "in6_setscope failed\n");
1066			goto cleanup;
1067		}
1068		delay = 0;
1069		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1070			/*
1071			 * We need a random delay for DAD on the address
1072			 * being configured.  It also means delaying
1073			 * transmission of the corresponding MLD report to
1074			 * avoid report collision.
1075			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1076			 */
1077			delay = arc4random() %
1078			    (MAX_RTR_SOLICITATION_DELAY * hz);
1079		}
1080		imm = in6_joingroup(ifp, &llsol, &error, delay);
1081		if (imm == NULL) {
1082			nd6log((LOG_WARNING,
1083			    "in6_update_ifa: addmulti failed for "
1084			    "%s on %s (errno=%d)\n",
1085			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
1086			    error));
1087			in6_purgeaddr((struct ifaddr *)ia);
1088			return (error);
1089		}
1090		LIST_INSERT_HEAD(&ia->ia6_memberships,
1091		    imm, i6mm_chain);
1092		in6m_sol = imm->i6mm_maddr;
1093
1094		bzero(&mltmask, sizeof(mltmask));
1095		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1096		mltmask.sin6_family = AF_INET6;
1097		mltmask.sin6_addr = in6mask32;
1098#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1099
1100		/*
1101		 * join link-local all-nodes address
1102		 */
1103		bzero(&mltaddr, sizeof(mltaddr));
1104		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1105		mltaddr.sin6_family = AF_INET6;
1106		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1107		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1108		    0)
1109			goto cleanup; /* XXX: should not fail */
1110
1111		/*
1112		 * XXX: do we really need this automatic routes?
1113		 * We should probably reconsider this stuff.  Most applications
1114		 * actually do not need the routes, since they usually specify
1115		 * the outgoing interface.
1116		 */
1117		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1118		if (rt) {
1119			/* XXX: only works in !SCOPEDROUTING case. */
1120			if (memcmp(&mltaddr.sin6_addr,
1121			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1122			    MLTMASK_LEN)) {
1123				RTFREE_LOCKED(rt);
1124				rt = NULL;
1125			}
1126		}
1127		if (!rt) {
1128			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1129			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1130			    (struct sockaddr *)&ia->ia_addr,
1131			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1132			    (struct rtentry **)0);
1133			if (error)
1134				goto cleanup;
1135		} else {
1136			RTFREE_LOCKED(rt);
1137		}
1138
1139		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1140		if (!imm) {
1141			nd6log((LOG_WARNING,
1142			    "in6_update_ifa: addmulti failed for "
1143			    "%s on %s (errno=%d)\n",
1144			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1145			    if_name(ifp), error));
1146			goto cleanup;
1147		}
1148		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1149
1150		/*
1151		 * join node information group address
1152		 */
1153#define hostnamelen	strlen(V_hostname)
1154		delay = 0;
1155		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1156			/*
1157			 * The spec doesn't say anything about delay for this
1158			 * group, but the same logic should apply.
1159			 */
1160			delay = arc4random() %
1161			    (MAX_RTR_SOLICITATION_DELAY * hz);
1162		}
1163		mtx_lock(&hostname_mtx);
1164		if (in6_nigroup(ifp, V_hostname, hostnamelen,
1165		    &mltaddr.sin6_addr) == 0) {
1166			mtx_unlock(&hostname_mtx);
1167			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1168			    delay); /* XXX jinmei */
1169			if (!imm) {
1170				nd6log((LOG_WARNING, "in6_update_ifa: "
1171				    "addmulti failed for %s on %s "
1172				    "(errno=%d)\n",
1173				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1174				    if_name(ifp), error));
1175				/* XXX not very fatal, go on... */
1176			} else {
1177				LIST_INSERT_HEAD(&ia->ia6_memberships,
1178				    imm, i6mm_chain);
1179			}
1180		} else
1181			mtx_unlock(&hostname_mtx);
1182#undef hostnamelen
1183
1184		/*
1185		 * join interface-local all-nodes address.
1186		 * (ff01::1%ifN, and ff01::%ifN/32)
1187		 */
1188		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1189		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1190		    != 0)
1191			goto cleanup; /* XXX: should not fail */
1192		/* XXX: again, do we really need the route? */
1193		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1194		if (rt) {
1195			if (memcmp(&mltaddr.sin6_addr,
1196			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1197			    MLTMASK_LEN)) {
1198				RTFREE_LOCKED(rt);
1199				rt = NULL;
1200			}
1201		}
1202		if (!rt) {
1203			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1204			    (struct sockaddr *)&ia->ia_addr,
1205			    (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING,
1206			    (struct rtentry **)0);
1207			if (error)
1208				goto cleanup;
1209		} else
1210			RTFREE_LOCKED(rt);
1211
1212		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1213		if (!imm) {
1214			nd6log((LOG_WARNING, "in6_update_ifa: "
1215			    "addmulti failed for %s on %s "
1216			    "(errno=%d)\n",
1217			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1218			    if_name(ifp), error));
1219			goto cleanup;
1220		}
1221		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1222#undef	MLTMASK_LEN
1223	}
1224
1225	/*
1226	 * Perform DAD, if needed.
1227	 * XXX It may be of use, if we can administratively
1228	 * disable DAD.
1229	 */
1230	if (hostIsNew && in6if_do_dad(ifp) &&
1231	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1232	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1233	{
1234		int mindelay, maxdelay;
1235
1236		delay = 0;
1237		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1238			/*
1239			 * We need to impose a delay before sending an NS
1240			 * for DAD.  Check if we also needed a delay for the
1241			 * corresponding MLD message.  If we did, the delay
1242			 * should be larger than the MLD delay (this could be
1243			 * relaxed a bit, but this simple logic is at least
1244			 * safe).
1245			 */
1246			mindelay = 0;
1247			if (in6m_sol != NULL &&
1248			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1249				mindelay = in6m_sol->in6m_timer;
1250			}
1251			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1252			if (maxdelay - mindelay == 0)
1253				delay = 0;
1254			else {
1255				delay =
1256				    (arc4random() % (maxdelay - mindelay)) +
1257				    mindelay;
1258			}
1259		}
1260		nd6_dad_start((struct ifaddr *)ia, delay);
1261	}
1262
1263	return (error);
1264
1265  unlink:
1266	/*
1267	 * XXX: if a change of an existing address failed, keep the entry
1268	 * anyway.
1269	 */
1270	if (hostIsNew)
1271		in6_unlink_ifa(ia, ifp);
1272	return (error);
1273
1274  cleanup:
1275	in6_purgeaddr(&ia->ia_ifa);
1276	return error;
1277}
1278
1279void
1280in6_purgeaddr(struct ifaddr *ifa)
1281{
1282	struct ifnet *ifp = ifa->ifa_ifp;
1283	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1284	char ip6buf[INET6_ADDRSTRLEN];
1285	struct in6_multi_mship *imm;
1286
1287	/* stop DAD processing */
1288	nd6_dad_stop(ifa);
1289
1290	/*
1291	 * delete route to the destination of the address being purged.
1292	 * The interface must be p2p or loopback in this case.
1293	 */
1294	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1295		int e;
1296
1297		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1298		    != 0) {
1299			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1300			    "a route to the p2p destination: %s on %s, "
1301			    "errno=%d\n",
1302			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1303			    if_name(ifp), e);
1304			/* proceed anyway... */
1305		} else
1306			ia->ia_flags &= ~IFA_ROUTE;
1307	}
1308
1309	/* Remove ownaddr's loopback rtentry, if it exists. */
1310	in6_ifremloop(&(ia->ia_ifa));
1311
1312	/*
1313	 * leave from multicast groups we have joined for the interface
1314	 */
1315	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1316		LIST_REMOVE(imm, i6mm_chain);
1317		in6_leavegroup(imm);
1318	}
1319
1320	in6_unlink_ifa(ia, ifp);
1321}
1322
1323static void
1324in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1325{
1326	struct in6_ifaddr *oia;
1327	int	s = splnet();
1328
1329	TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1330
1331	oia = ia;
1332	if (oia == (ia = V_in6_ifaddr))
1333		V_in6_ifaddr = ia->ia_next;
1334	else {
1335		while (ia->ia_next && (ia->ia_next != oia))
1336			ia = ia->ia_next;
1337		if (ia->ia_next)
1338			ia->ia_next = oia->ia_next;
1339		else {
1340			/* search failed */
1341			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1342		}
1343	}
1344
1345	/*
1346	 * Release the reference to the base prefix.  There should be a
1347	 * positive reference.
1348	 */
1349	if (oia->ia6_ndpr == NULL) {
1350		nd6log((LOG_NOTICE,
1351		    "in6_unlink_ifa: autoconf'ed address "
1352		    "%p has no prefix\n", oia));
1353	} else {
1354		oia->ia6_ndpr->ndpr_refcnt--;
1355		oia->ia6_ndpr = NULL;
1356	}
1357
1358	/*
1359	 * Also, if the address being removed is autoconf'ed, call
1360	 * pfxlist_onlink_check() since the release might affect the status of
1361	 * other (detached) addresses.
1362	 */
1363	if ((oia->ia6_flags & IN6_IFF_AUTOCONF)) {
1364		pfxlist_onlink_check();
1365	}
1366
1367	/*
1368	 * release another refcnt for the link from in6_ifaddr.
1369	 * Note that we should decrement the refcnt at least once for all *BSD.
1370	 */
1371	IFAFREE(&oia->ia_ifa);
1372
1373	splx(s);
1374}
1375
1376void
1377in6_purgeif(struct ifnet *ifp)
1378{
1379	struct ifaddr *ifa, *nifa;
1380
1381	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa) {
1382		nifa = TAILQ_NEXT(ifa, ifa_list);
1383		if (ifa->ifa_addr->sa_family != AF_INET6)
1384			continue;
1385		in6_purgeaddr(ifa);
1386	}
1387
1388	in6_ifdetach(ifp);
1389}
1390
1391/*
1392 * SIOC[GAD]LIFADDR.
1393 *	SIOCGLIFADDR: get first address. (?)
1394 *	SIOCGLIFADDR with IFLR_PREFIX:
1395 *		get first address that matches the specified prefix.
1396 *	SIOCALIFADDR: add the specified address.
1397 *	SIOCALIFADDR with IFLR_PREFIX:
1398 *		add the specified prefix, filling hostid part from
1399 *		the first link-local address.  prefixlen must be <= 64.
1400 *	SIOCDLIFADDR: delete the specified address.
1401 *	SIOCDLIFADDR with IFLR_PREFIX:
1402 *		delete the first address that matches the specified prefix.
1403 * return values:
1404 *	EINVAL on invalid parameters
1405 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1406 *	other values may be returned from in6_ioctl()
1407 *
1408 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1409 * this is to accomodate address naming scheme other than RFC2374,
1410 * in the future.
1411 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1412 * address encoding scheme. (see figure on page 8)
1413 */
1414static int
1415in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1416    struct ifnet *ifp, struct thread *td)
1417{
1418	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1419	struct ifaddr *ifa;
1420	struct sockaddr *sa;
1421
1422	/* sanity checks */
1423	if (!data || !ifp) {
1424		panic("invalid argument to in6_lifaddr_ioctl");
1425		/* NOTREACHED */
1426	}
1427
1428	switch (cmd) {
1429	case SIOCGLIFADDR:
1430		/* address must be specified on GET with IFLR_PREFIX */
1431		if ((iflr->flags & IFLR_PREFIX) == 0)
1432			break;
1433		/* FALLTHROUGH */
1434	case SIOCALIFADDR:
1435	case SIOCDLIFADDR:
1436		/* address must be specified on ADD and DELETE */
1437		sa = (struct sockaddr *)&iflr->addr;
1438		if (sa->sa_family != AF_INET6)
1439			return EINVAL;
1440		if (sa->sa_len != sizeof(struct sockaddr_in6))
1441			return EINVAL;
1442		/* XXX need improvement */
1443		sa = (struct sockaddr *)&iflr->dstaddr;
1444		if (sa->sa_family && sa->sa_family != AF_INET6)
1445			return EINVAL;
1446		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1447			return EINVAL;
1448		break;
1449	default: /* shouldn't happen */
1450#if 0
1451		panic("invalid cmd to in6_lifaddr_ioctl");
1452		/* NOTREACHED */
1453#else
1454		return EOPNOTSUPP;
1455#endif
1456	}
1457	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1458		return EINVAL;
1459
1460	switch (cmd) {
1461	case SIOCALIFADDR:
1462	    {
1463		struct in6_aliasreq ifra;
1464		struct in6_addr *hostid = NULL;
1465		int prefixlen;
1466
1467		if ((iflr->flags & IFLR_PREFIX) != 0) {
1468			struct sockaddr_in6 *sin6;
1469
1470			/*
1471			 * hostid is to fill in the hostid part of the
1472			 * address.  hostid points to the first link-local
1473			 * address attached to the interface.
1474			 */
1475			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1476			if (!ifa)
1477				return EADDRNOTAVAIL;
1478			hostid = IFA_IN6(ifa);
1479
1480			/* prefixlen must be <= 64. */
1481			if (64 < iflr->prefixlen)
1482				return EINVAL;
1483			prefixlen = iflr->prefixlen;
1484
1485			/* hostid part must be zero. */
1486			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1487			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1488			    sin6->sin6_addr.s6_addr32[3] != 0) {
1489				return EINVAL;
1490			}
1491		} else
1492			prefixlen = iflr->prefixlen;
1493
1494		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1495		bzero(&ifra, sizeof(ifra));
1496		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1497
1498		bcopy(&iflr->addr, &ifra.ifra_addr,
1499		    ((struct sockaddr *)&iflr->addr)->sa_len);
1500		if (hostid) {
1501			/* fill in hostid part */
1502			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1503			    hostid->s6_addr32[2];
1504			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1505			    hostid->s6_addr32[3];
1506		}
1507
1508		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1509			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1510			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1511			if (hostid) {
1512				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1513				    hostid->s6_addr32[2];
1514				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1515				    hostid->s6_addr32[3];
1516			}
1517		}
1518
1519		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1520		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1521
1522		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1523		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1524	    }
1525	case SIOCGLIFADDR:
1526	case SIOCDLIFADDR:
1527	    {
1528		struct in6_ifaddr *ia;
1529		struct in6_addr mask, candidate, match;
1530		struct sockaddr_in6 *sin6;
1531		int cmp;
1532
1533		bzero(&mask, sizeof(mask));
1534		if (iflr->flags & IFLR_PREFIX) {
1535			/* lookup a prefix rather than address. */
1536			in6_prefixlen2mask(&mask, iflr->prefixlen);
1537
1538			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1539			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1540			match.s6_addr32[0] &= mask.s6_addr32[0];
1541			match.s6_addr32[1] &= mask.s6_addr32[1];
1542			match.s6_addr32[2] &= mask.s6_addr32[2];
1543			match.s6_addr32[3] &= mask.s6_addr32[3];
1544
1545			/* if you set extra bits, that's wrong */
1546			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1547				return EINVAL;
1548
1549			cmp = 1;
1550		} else {
1551			if (cmd == SIOCGLIFADDR) {
1552				/* on getting an address, take the 1st match */
1553				cmp = 0;	/* XXX */
1554			} else {
1555				/* on deleting an address, do exact match */
1556				in6_prefixlen2mask(&mask, 128);
1557				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1558				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1559
1560				cmp = 1;
1561			}
1562		}
1563
1564		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1565			if (ifa->ifa_addr->sa_family != AF_INET6)
1566				continue;
1567			if (!cmp)
1568				break;
1569
1570			/*
1571			 * XXX: this is adhoc, but is necessary to allow
1572			 * a user to specify fe80::/64 (not /10) for a
1573			 * link-local address.
1574			 */
1575			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1576			in6_clearscope(&candidate);
1577			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1578			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1579			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1580			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1581			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1582				break;
1583		}
1584		if (!ifa)
1585			return EADDRNOTAVAIL;
1586		ia = ifa2ia6(ifa);
1587
1588		if (cmd == SIOCGLIFADDR) {
1589			int error;
1590
1591			/* fill in the if_laddrreq structure */
1592			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1593			error = sa6_recoverscope(
1594			    (struct sockaddr_in6 *)&iflr->addr);
1595			if (error != 0)
1596				return (error);
1597
1598			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1599				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1600				    ia->ia_dstaddr.sin6_len);
1601				error = sa6_recoverscope(
1602				    (struct sockaddr_in6 *)&iflr->dstaddr);
1603				if (error != 0)
1604					return (error);
1605			} else
1606				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1607
1608			iflr->prefixlen =
1609			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1610
1611			iflr->flags = ia->ia6_flags;	/* XXX */
1612
1613			return 0;
1614		} else {
1615			struct in6_aliasreq ifra;
1616
1617			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1618			bzero(&ifra, sizeof(ifra));
1619			bcopy(iflr->iflr_name, ifra.ifra_name,
1620			    sizeof(ifra.ifra_name));
1621
1622			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1623			    ia->ia_addr.sin6_len);
1624			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1625				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1626				    ia->ia_dstaddr.sin6_len);
1627			} else {
1628				bzero(&ifra.ifra_dstaddr,
1629				    sizeof(ifra.ifra_dstaddr));
1630			}
1631			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1632			    ia->ia_prefixmask.sin6_len);
1633
1634			ifra.ifra_flags = ia->ia6_flags;
1635			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1636			    ifp, td);
1637		}
1638	    }
1639	}
1640
1641	return EOPNOTSUPP;	/* just for safety */
1642}
1643
1644/*
1645 * Initialize an interface's intetnet6 address
1646 * and routing table entry.
1647 */
1648static int
1649in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1650    struct sockaddr_in6 *sin6, int newhost)
1651{
1652	int	error = 0, plen, ifacount = 0;
1653	int	s = splimp();
1654	struct ifaddr *ifa;
1655
1656	/*
1657	 * Give the interface a chance to initialize
1658	 * if this is its first address,
1659	 * and to validate the address if necessary.
1660	 */
1661	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1662		if (ifa->ifa_addr->sa_family != AF_INET6)
1663			continue;
1664		ifacount++;
1665	}
1666
1667	ia->ia_addr = *sin6;
1668
1669	if (ifacount <= 1 && ifp->if_ioctl) {
1670		IFF_LOCKGIANT(ifp);
1671		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1672		IFF_UNLOCKGIANT(ifp);
1673		if (error) {
1674			splx(s);
1675			return (error);
1676		}
1677	}
1678	splx(s);
1679
1680	ia->ia_ifa.ifa_metric = ifp->if_metric;
1681
1682	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1683
1684	if (newhost) {
1685		/*
1686		 * set the rtrequest function to create llinfo.  It also
1687		 * adjust outgoing interface of the route for the local
1688		 * address when called via in6_ifaddloop() below.
1689		 */
1690		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1691	}
1692
1693	/*
1694	 * Special case:
1695	 * If a new destination address is specified for a point-to-point
1696	 * interface, install a route to the destination as an interface
1697	 * direct route.  In addition, if the link is expected to have neighbor
1698	 * cache entries, specify RTF_LLINFO so that a cache entry for the
1699	 * destination address will be created.
1700	 * created
1701	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1702	 * interface that share the same destination.
1703	 */
1704	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1705	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1706	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1707		int rtflags = RTF_UP | RTF_HOST;
1708		struct rtentry *rt = NULL, **rtp = NULL;
1709
1710		if (nd6_need_cache(ifp) != 0) {
1711			rtflags |= RTF_LLINFO;
1712			rtp = &rt;
1713		}
1714
1715		error = rtrequest(RTM_ADD,
1716		    (struct sockaddr *)&ia->ia_dstaddr,
1717		    (struct sockaddr *)&ia->ia_addr,
1718		    (struct sockaddr *)&ia->ia_prefixmask,
1719		    ia->ia_flags | rtflags, rtp);
1720		if (error != 0)
1721			return (error);
1722		if (rt != NULL) {
1723			struct llinfo_nd6 *ln;
1724
1725			RT_LOCK(rt);
1726			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
1727			if (ln != NULL) {
1728				/*
1729				 * Set the state to STALE because we don't
1730				 * have to perform address resolution on this
1731				 * link.
1732				 */
1733				ln->ln_state = ND6_LLINFO_STALE;
1734			}
1735			RT_REMREF(rt);
1736			RT_UNLOCK(rt);
1737		}
1738		ia->ia_flags |= IFA_ROUTE;
1739	}
1740	if (plen < 128) {
1741		/*
1742		 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1743		 */
1744		ia->ia_ifa.ifa_flags |= RTF_CLONING;
1745	}
1746
1747	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1748	if (newhost)
1749		in6_ifaddloop(&(ia->ia_ifa));
1750
1751	return (error);
1752}
1753
1754struct in6_multi_mship *
1755in6_joingroup(struct ifnet *ifp, struct in6_addr *addr,
1756    int *errorp, int delay)
1757{
1758	struct in6_multi_mship *imm;
1759
1760	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
1761	if (!imm) {
1762		*errorp = ENOBUFS;
1763		return NULL;
1764	}
1765	imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp, delay);
1766	if (!imm->i6mm_maddr) {
1767		/* *errorp is alrady set */
1768		free(imm, M_IP6MADDR);
1769		return NULL;
1770	}
1771	return imm;
1772}
1773
1774int
1775in6_leavegroup(struct in6_multi_mship *imm)
1776{
1777
1778	if (imm->i6mm_maddr)
1779		in6_delmulti(imm->i6mm_maddr);
1780	free(imm,  M_IP6MADDR);
1781	return 0;
1782}
1783
1784/*
1785 * Find an IPv6 interface link-local address specific to an interface.
1786 */
1787struct in6_ifaddr *
1788in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1789{
1790	struct ifaddr *ifa;
1791
1792	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1793		if (ifa->ifa_addr->sa_family != AF_INET6)
1794			continue;
1795		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1796			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1797			     ignoreflags) != 0)
1798				continue;
1799			break;
1800		}
1801	}
1802
1803	return ((struct in6_ifaddr *)ifa);
1804}
1805
1806
1807/*
1808 * find the internet address corresponding to a given interface and address.
1809 */
1810struct in6_ifaddr *
1811in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1812{
1813	struct ifaddr *ifa;
1814
1815	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1816		if (ifa->ifa_addr->sa_family != AF_INET6)
1817			continue;
1818		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1819			break;
1820	}
1821
1822	return ((struct in6_ifaddr *)ifa);
1823}
1824
1825/*
1826 * Convert IP6 address to printable (loggable) representation. Caller
1827 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1828 */
1829static char digits[] = "0123456789abcdef";
1830char *
1831ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1832{
1833	int i;
1834	char *cp;
1835	const u_int16_t *a = (const u_int16_t *)addr;
1836	const u_int8_t *d;
1837	int dcolon = 0, zero = 0;
1838
1839	cp = ip6buf;
1840
1841	for (i = 0; i < 8; i++) {
1842		if (dcolon == 1) {
1843			if (*a == 0) {
1844				if (i == 7)
1845					*cp++ = ':';
1846				a++;
1847				continue;
1848			} else
1849				dcolon = 2;
1850		}
1851		if (*a == 0) {
1852			if (dcolon == 0 && *(a + 1) == 0) {
1853				if (i == 0)
1854					*cp++ = ':';
1855				*cp++ = ':';
1856				dcolon = 1;
1857			} else {
1858				*cp++ = '0';
1859				*cp++ = ':';
1860			}
1861			a++;
1862			continue;
1863		}
1864		d = (const u_char *)a;
1865		/* Try to eliminate leading zeros in printout like in :0001. */
1866		zero = 1;
1867		*cp = digits[*d >> 4];
1868		if (*cp != '0') {
1869			zero = 0;
1870			cp++;
1871		}
1872		*cp = digits[*d++ & 0xf];
1873		if (zero == 0 || (*cp != '0')) {
1874			zero = 0;
1875			cp++;
1876		}
1877		*cp = digits[*d >> 4];
1878		if (zero == 0 || (*cp != '0')) {
1879			zero = 0;
1880			cp++;
1881		}
1882		*cp++ = digits[*d & 0xf];
1883		*cp++ = ':';
1884		a++;
1885	}
1886	*--cp = '\0';
1887	return (ip6buf);
1888}
1889
1890int
1891in6_localaddr(struct in6_addr *in6)
1892{
1893	struct in6_ifaddr *ia;
1894
1895	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1896		return 1;
1897
1898	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1899		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1900		    &ia->ia_prefixmask.sin6_addr)) {
1901			return 1;
1902		}
1903	}
1904
1905	return (0);
1906}
1907
1908int
1909in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1910{
1911	struct in6_ifaddr *ia;
1912
1913	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1914		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1915				       &sa6->sin6_addr) &&
1916		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1917			return (1); /* true */
1918
1919		/* XXX: do we still have to go thru the rest of the list? */
1920	}
1921
1922	return (0);		/* false */
1923}
1924
1925/*
1926 * return length of part which dst and src are equal
1927 * hard coding...
1928 */
1929int
1930in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1931{
1932	int match = 0;
1933	u_char *s = (u_char *)src, *d = (u_char *)dst;
1934	u_char *lim = s + 16, r;
1935
1936	while (s < lim)
1937		if ((r = (*d++ ^ *s++)) != 0) {
1938			while (r < 128) {
1939				match++;
1940				r <<= 1;
1941			}
1942			break;
1943		} else
1944			match += 8;
1945	return match;
1946}
1947
1948/* XXX: to be scope conscious */
1949int
1950in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1951{
1952	int bytelen, bitlen;
1953
1954	/* sanity check */
1955	if (0 > len || len > 128) {
1956		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1957		    len);
1958		return (0);
1959	}
1960
1961	bytelen = len / 8;
1962	bitlen = len % 8;
1963
1964	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1965		return (0);
1966	if (bitlen != 0 &&
1967	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1968	    p2->s6_addr[bytelen] >> (8 - bitlen))
1969		return (0);
1970
1971	return (1);
1972}
1973
1974void
1975in6_prefixlen2mask(struct in6_addr *maskp, int len)
1976{
1977	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1978	int bytelen, bitlen, i;
1979
1980	/* sanity check */
1981	if (0 > len || len > 128) {
1982		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1983		    len);
1984		return;
1985	}
1986
1987	bzero(maskp, sizeof(*maskp));
1988	bytelen = len / 8;
1989	bitlen = len % 8;
1990	for (i = 0; i < bytelen; i++)
1991		maskp->s6_addr[i] = 0xff;
1992	if (bitlen)
1993		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1994}
1995
1996/*
1997 * return the best address out of the same scope. if no address was
1998 * found, return the first valid address from designated IF.
1999 */
2000struct in6_ifaddr *
2001in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2002{
2003	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2004	struct ifaddr *ifa;
2005	struct in6_ifaddr *besta = 0;
2006	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2007
2008	dep[0] = dep[1] = NULL;
2009
2010	/*
2011	 * We first look for addresses in the same scope.
2012	 * If there is one, return it.
2013	 * If two or more, return one which matches the dst longest.
2014	 * If none, return one of global addresses assigned other ifs.
2015	 */
2016	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2017		if (ifa->ifa_addr->sa_family != AF_INET6)
2018			continue;
2019		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2020			continue; /* XXX: is there any case to allow anycast? */
2021		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2022			continue; /* don't use this interface */
2023		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2024			continue;
2025		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2026			if (V_ip6_use_deprecated)
2027				dep[0] = (struct in6_ifaddr *)ifa;
2028			continue;
2029		}
2030
2031		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2032			/*
2033			 * call in6_matchlen() as few as possible
2034			 */
2035			if (besta) {
2036				if (blen == -1)
2037					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2038				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2039				if (tlen > blen) {
2040					blen = tlen;
2041					besta = (struct in6_ifaddr *)ifa;
2042				}
2043			} else
2044				besta = (struct in6_ifaddr *)ifa;
2045		}
2046	}
2047	if (besta)
2048		return (besta);
2049
2050	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2051		if (ifa->ifa_addr->sa_family != AF_INET6)
2052			continue;
2053		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2054			continue; /* XXX: is there any case to allow anycast? */
2055		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2056			continue; /* don't use this interface */
2057		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2058			continue;
2059		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2060			if (V_ip6_use_deprecated)
2061				dep[1] = (struct in6_ifaddr *)ifa;
2062			continue;
2063		}
2064
2065		return (struct in6_ifaddr *)ifa;
2066	}
2067
2068	/* use the last-resort values, that are, deprecated addresses */
2069	if (dep[0])
2070		return dep[0];
2071	if (dep[1])
2072		return dep[1];
2073
2074	return NULL;
2075}
2076
2077/*
2078 * perform DAD when interface becomes IFF_UP.
2079 */
2080void
2081in6_if_up(struct ifnet *ifp)
2082{
2083	struct ifaddr *ifa;
2084	struct in6_ifaddr *ia;
2085
2086	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2087		if (ifa->ifa_addr->sa_family != AF_INET6)
2088			continue;
2089		ia = (struct in6_ifaddr *)ifa;
2090		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2091			/*
2092			 * The TENTATIVE flag was likely set by hand
2093			 * beforehand, implicitly indicating the need for DAD.
2094			 * We may be able to skip the random delay in this
2095			 * case, but we impose delays just in case.
2096			 */
2097			nd6_dad_start(ifa,
2098			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2099		}
2100	}
2101
2102	/*
2103	 * special cases, like 6to4, are handled in in6_ifattach
2104	 */
2105	in6_ifattach(ifp, NULL);
2106}
2107
2108int
2109in6if_do_dad(struct ifnet *ifp)
2110{
2111	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2112		return (0);
2113
2114	switch (ifp->if_type) {
2115#ifdef IFT_DUMMY
2116	case IFT_DUMMY:
2117#endif
2118	case IFT_FAITH:
2119		/*
2120		 * These interfaces do not have the IFF_LOOPBACK flag,
2121		 * but loop packets back.  We do not have to do DAD on such
2122		 * interfaces.  We should even omit it, because loop-backed
2123		 * NS would confuse the DAD procedure.
2124		 */
2125		return (0);
2126	default:
2127		/*
2128		 * Our DAD routine requires the interface up and running.
2129		 * However, some interfaces can be up before the RUNNING
2130		 * status.  Additionaly, users may try to assign addresses
2131		 * before the interface becomes up (or running).
2132		 * We simply skip DAD in such a case as a work around.
2133		 * XXX: we should rather mark "tentative" on such addresses,
2134		 * and do DAD after the interface becomes ready.
2135		 */
2136		if (!((ifp->if_flags & IFF_UP) &&
2137		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2138			return (0);
2139
2140		return (1);
2141	}
2142}
2143
2144/*
2145 * Calculate max IPv6 MTU through all the interfaces and store it
2146 * to in6_maxmtu.
2147 */
2148void
2149in6_setmaxmtu(void)
2150{
2151	unsigned long maxmtu = 0;
2152	struct ifnet *ifp;
2153
2154	IFNET_RLOCK();
2155	for (ifp = TAILQ_FIRST(&V_ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
2156		/* this function can be called during ifnet initialization */
2157		if (!ifp->if_afdata[AF_INET6])
2158			continue;
2159		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2160		    IN6_LINKMTU(ifp) > maxmtu)
2161			maxmtu = IN6_LINKMTU(ifp);
2162	}
2163	IFNET_RUNLOCK();
2164	if (maxmtu)	     /* update only when maxmtu is positive */
2165		V_in6_maxmtu = maxmtu;
2166}
2167
2168/*
2169 * Provide the length of interface identifiers to be used for the link attached
2170 * to the given interface.  The length should be defined in "IPv6 over
2171 * xxx-link" document.  Note that address architecture might also define
2172 * the length for a particular set of address prefixes, regardless of the
2173 * link type.  As clarified in rfc2462bis, those two definitions should be
2174 * consistent, and those really are as of August 2004.
2175 */
2176int
2177in6_if2idlen(struct ifnet *ifp)
2178{
2179	switch (ifp->if_type) {
2180	case IFT_ETHER:		/* RFC2464 */
2181#ifdef IFT_PROPVIRTUAL
2182	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2183#endif
2184#ifdef IFT_L2VLAN
2185	case IFT_L2VLAN:	/* ditto */
2186#endif
2187#ifdef IFT_IEEE80211
2188	case IFT_IEEE80211:	/* ditto */
2189#endif
2190#ifdef IFT_MIP
2191	case IFT_MIP:	/* ditto */
2192#endif
2193		return (64);
2194	case IFT_FDDI:		/* RFC2467 */
2195		return (64);
2196	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2197		return (64);
2198	case IFT_PPP:		/* RFC2472 */
2199		return (64);
2200	case IFT_ARCNET:	/* RFC2497 */
2201		return (64);
2202	case IFT_FRELAY:	/* RFC2590 */
2203		return (64);
2204	case IFT_IEEE1394:	/* RFC3146 */
2205		return (64);
2206	case IFT_GIF:
2207		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2208	case IFT_LOOP:
2209		return (64);	/* XXX: is this really correct? */
2210	default:
2211		/*
2212		 * Unknown link type:
2213		 * It might be controversial to use the today's common constant
2214		 * of 64 for these cases unconditionally.  For full compliance,
2215		 * we should return an error in this case.  On the other hand,
2216		 * if we simply miss the standard for the link type or a new
2217		 * standard is defined for a new link type, the IFID length
2218		 * is very likely to be the common constant.  As a compromise,
2219		 * we always use the constant, but make an explicit notice
2220		 * indicating the "unknown" case.
2221		 */
2222		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2223		return (64);
2224	}
2225}
2226
2227void *
2228in6_domifattach(struct ifnet *ifp)
2229{
2230	struct in6_ifextra *ext;
2231
2232	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2233	bzero(ext, sizeof(*ext));
2234
2235	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2236	    M_IFADDR, M_WAITOK);
2237	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2238
2239	ext->icmp6_ifstat =
2240	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2241	    M_IFADDR, M_WAITOK);
2242	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2243
2244	ext->nd_ifinfo = nd6_ifattach(ifp);
2245	ext->scope6_id = scope6_ifattach(ifp);
2246	return ext;
2247}
2248
2249void
2250in6_domifdetach(struct ifnet *ifp, void *aux)
2251{
2252	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2253
2254	scope6_ifdetach(ext->scope6_id);
2255	nd6_ifdetach(ext->nd_ifinfo);
2256	free(ext->in6_ifstat, M_IFADDR);
2257	free(ext->icmp6_ifstat, M_IFADDR);
2258	free(ext, M_IFADDR);
2259}
2260
2261/*
2262 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2263 * v4 mapped addr or v4 compat addr
2264 */
2265void
2266in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2267{
2268
2269	bzero(sin, sizeof(*sin));
2270	sin->sin_len = sizeof(struct sockaddr_in);
2271	sin->sin_family = AF_INET;
2272	sin->sin_port = sin6->sin6_port;
2273	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2274}
2275
2276/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2277void
2278in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2279{
2280	bzero(sin6, sizeof(*sin6));
2281	sin6->sin6_len = sizeof(struct sockaddr_in6);
2282	sin6->sin6_family = AF_INET6;
2283	sin6->sin6_port = sin->sin_port;
2284	sin6->sin6_addr.s6_addr32[0] = 0;
2285	sin6->sin6_addr.s6_addr32[1] = 0;
2286	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2287	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2288}
2289
2290/* Convert sockaddr_in6 into sockaddr_in. */
2291void
2292in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2293{
2294	struct sockaddr_in *sin_p;
2295	struct sockaddr_in6 sin6;
2296
2297	/*
2298	 * Save original sockaddr_in6 addr and convert it
2299	 * to sockaddr_in.
2300	 */
2301	sin6 = *(struct sockaddr_in6 *)nam;
2302	sin_p = (struct sockaddr_in *)nam;
2303	in6_sin6_2_sin(sin_p, &sin6);
2304}
2305
2306/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2307void
2308in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2309{
2310	struct sockaddr_in *sin_p;
2311	struct sockaddr_in6 *sin6_p;
2312
2313	MALLOC(sin6_p, struct sockaddr_in6 *, sizeof *sin6_p, M_SONAME,
2314	       M_WAITOK);
2315	sin_p = (struct sockaddr_in *)*nam;
2316	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2317	FREE(*nam, M_SONAME);
2318	*nam = (struct sockaddr *)sin6_p;
2319}
2320