in6.c revision 194760
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: head/sys/netinet6/in6.c 194760 2009-06-23 20:19:09Z rwatson $");
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/jail.h>
72#include <sys/malloc.h>
73#include <sys/socket.h>
74#include <sys/socketvar.h>
75#include <sys/sockio.h>
76#include <sys/systm.h>
77#include <sys/priv.h>
78#include <sys/proc.h>
79#include <sys/time.h>
80#include <sys/kernel.h>
81#include <sys/syslog.h>
82#include <sys/vimage.h>
83
84#include <net/if.h>
85#include <net/if_types.h>
86#include <net/route.h>
87#include <net/if_dl.h>
88#include <net/vnet.h>
89
90#include <netinet/in.h>
91#include <netinet/in_var.h>
92#include <net/if_llatbl.h>
93#include <netinet/if_ether.h>
94#include <netinet/in_systm.h>
95#include <netinet/ip.h>
96#include <netinet/in_pcb.h>
97
98#include <netinet/ip6.h>
99#include <netinet6/ip6_var.h>
100#include <netinet6/nd6.h>
101#include <netinet6/mld6_var.h>
102#include <netinet6/ip6_mroute.h>
103#include <netinet6/in6_ifattach.h>
104#include <netinet6/scope6_var.h>
105#include <netinet6/in6_pcb.h>
106#include <netinet6/vinet6.h>
107
108/*
109 * Definitions of some costant IP6 addresses.
110 */
111const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
112const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
113const struct in6_addr in6addr_nodelocal_allnodes =
114	IN6ADDR_NODELOCAL_ALLNODES_INIT;
115const struct in6_addr in6addr_linklocal_allnodes =
116	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
117const struct in6_addr in6addr_linklocal_allrouters =
118	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
119const struct in6_addr in6addr_linklocal_allv2routers =
120	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
121
122const struct in6_addr in6mask0 = IN6MASK0;
123const struct in6_addr in6mask32 = IN6MASK32;
124const struct in6_addr in6mask64 = IN6MASK64;
125const struct in6_addr in6mask96 = IN6MASK96;
126const struct in6_addr in6mask128 = IN6MASK128;
127
128const struct sockaddr_in6 sa6_any =
129	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
130
131static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
132	struct ifnet *, struct thread *));
133static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
134	struct sockaddr_in6 *, int));
135static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
136
137int	(*faithprefix_p)(struct in6_addr *);
138
139
140
141int
142in6_mask2len(struct in6_addr *mask, u_char *lim0)
143{
144	int x = 0, y;
145	u_char *lim = lim0, *p;
146
147	/* ignore the scope_id part */
148	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
149		lim = (u_char *)mask + sizeof(*mask);
150	for (p = (u_char *)mask; p < lim; x++, p++) {
151		if (*p != 0xff)
152			break;
153	}
154	y = 0;
155	if (p < lim) {
156		for (y = 0; y < 8; y++) {
157			if ((*p & (0x80 >> y)) == 0)
158				break;
159		}
160	}
161
162	/*
163	 * when the limit pointer is given, do a stricter check on the
164	 * remaining bits.
165	 */
166	if (p < lim) {
167		if (y != 0 && (*p & (0x00ff >> y)) != 0)
168			return (-1);
169		for (p = p + 1; p < lim; p++)
170			if (*p != 0)
171				return (-1);
172	}
173
174	return x * 8 + y;
175}
176
177#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
178#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
179
180int
181in6_control(struct socket *so, u_long cmd, caddr_t data,
182    struct ifnet *ifp, struct thread *td)
183{
184	INIT_VNET_INET6(curvnet);
185	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
186	struct	in6_ifaddr *ia = NULL;
187	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
188	struct sockaddr_in6 *sa6;
189	int error;
190
191	switch (cmd) {
192	case SIOCGETSGCNT_IN6:
193	case SIOCGETMIFCNT_IN6:
194		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
195	}
196
197	switch(cmd) {
198	case SIOCAADDRCTL_POLICY:
199	case SIOCDADDRCTL_POLICY:
200		if (td != NULL) {
201			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
202			if (error)
203				return (error);
204		}
205		return (in6_src_ioctl(cmd, data));
206	}
207
208	if (ifp == NULL)
209		return (EOPNOTSUPP);
210
211	switch (cmd) {
212	case SIOCSNDFLUSH_IN6:
213	case SIOCSPFXFLUSH_IN6:
214	case SIOCSRTRFLUSH_IN6:
215	case SIOCSDEFIFACE_IN6:
216	case SIOCSIFINFO_FLAGS:
217	case SIOCSIFINFO_IN6:
218		if (td != NULL) {
219			error = priv_check(td, PRIV_NETINET_ND6);
220			if (error)
221				return (error);
222		}
223		/* FALLTHROUGH */
224	case OSIOCGIFINFO_IN6:
225	case SIOCGIFINFO_IN6:
226	case SIOCGDRLST_IN6:
227	case SIOCGPRLST_IN6:
228	case SIOCGNBRINFO_IN6:
229	case SIOCGDEFIFACE_IN6:
230		return (nd6_ioctl(cmd, data, ifp));
231	}
232
233	switch (cmd) {
234	case SIOCSIFPREFIX_IN6:
235	case SIOCDIFPREFIX_IN6:
236	case SIOCAIFPREFIX_IN6:
237	case SIOCCIFPREFIX_IN6:
238	case SIOCSGIFPREFIX_IN6:
239	case SIOCGIFPREFIX_IN6:
240		log(LOG_NOTICE,
241		    "prefix ioctls are now invalidated. "
242		    "please use ifconfig.\n");
243		return (EOPNOTSUPP);
244	}
245
246	switch (cmd) {
247	case SIOCSSCOPE6:
248		if (td != NULL) {
249			error = priv_check(td, PRIV_NETINET_SCOPE6);
250			if (error)
251				return (error);
252		}
253		return (scope6_set(ifp,
254		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
255	case SIOCGSCOPE6:
256		return (scope6_get(ifp,
257		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
258	case SIOCGSCOPE6DEF:
259		return (scope6_get_default((struct scope6_id *)
260		    ifr->ifr_ifru.ifru_scope_id));
261	}
262
263	switch (cmd) {
264	case SIOCALIFADDR:
265		if (td != NULL) {
266			error = priv_check(td, PRIV_NET_ADDIFADDR);
267			if (error)
268				return (error);
269		}
270		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
271
272	case SIOCDLIFADDR:
273		if (td != NULL) {
274			error = priv_check(td, PRIV_NET_DELIFADDR);
275			if (error)
276				return (error);
277		}
278		/* FALLTHROUGH */
279	case SIOCGLIFADDR:
280		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
281	}
282
283	/*
284	 * Find address for this interface, if it exists.
285	 *
286	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
287	 * only, and used the first interface address as the target of other
288	 * operations (without checking ifra_addr).  This was because netinet
289	 * code/API assumed at most 1 interface address per interface.
290	 * Since IPv6 allows a node to assign multiple addresses
291	 * on a single interface, we almost always look and check the
292	 * presence of ifra_addr, and reject invalid ones here.
293	 * It also decreases duplicated code among SIOC*_IN6 operations.
294	 */
295	switch (cmd) {
296	case SIOCAIFADDR_IN6:
297	case SIOCSIFPHYADDR_IN6:
298		sa6 = &ifra->ifra_addr;
299		break;
300	case SIOCSIFADDR_IN6:
301	case SIOCGIFADDR_IN6:
302	case SIOCSIFDSTADDR_IN6:
303	case SIOCSIFNETMASK_IN6:
304	case SIOCGIFDSTADDR_IN6:
305	case SIOCGIFNETMASK_IN6:
306	case SIOCDIFADDR_IN6:
307	case SIOCGIFPSRCADDR_IN6:
308	case SIOCGIFPDSTADDR_IN6:
309	case SIOCGIFAFLAG_IN6:
310	case SIOCSNDFLUSH_IN6:
311	case SIOCSPFXFLUSH_IN6:
312	case SIOCSRTRFLUSH_IN6:
313	case SIOCGIFALIFETIME_IN6:
314	case SIOCSIFALIFETIME_IN6:
315	case SIOCGIFSTAT_IN6:
316	case SIOCGIFSTAT_ICMP6:
317		sa6 = &ifr->ifr_addr;
318		break;
319	default:
320		sa6 = NULL;
321		break;
322	}
323	if (sa6 && sa6->sin6_family == AF_INET6) {
324		if (sa6->sin6_scope_id != 0)
325			error = sa6_embedscope(sa6, 0);
326		else
327			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
328		if (error != 0)
329			return (error);
330		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
331		    &sa6->sin6_addr)) != 0)
332			return (error);
333		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
334	} else
335		ia = NULL;
336
337	switch (cmd) {
338	case SIOCSIFADDR_IN6:
339	case SIOCSIFDSTADDR_IN6:
340	case SIOCSIFNETMASK_IN6:
341		/*
342		 * Since IPv6 allows a node to assign multiple addresses
343		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
344		 */
345		/* we decided to obsolete this command (20000704) */
346		error = EINVAL;
347		goto out;
348
349	case SIOCDIFADDR_IN6:
350		/*
351		 * for IPv4, we look for existing in_ifaddr here to allow
352		 * "ifconfig if0 delete" to remove the first IPv4 address on
353		 * the interface.  For IPv6, as the spec allows multiple
354		 * interface address from the day one, we consider "remove the
355		 * first one" semantics to be not preferable.
356		 */
357		if (ia == NULL) {
358			error = EADDRNOTAVAIL;
359			goto out;
360		}
361		/* FALLTHROUGH */
362	case SIOCAIFADDR_IN6:
363		/*
364		 * We always require users to specify a valid IPv6 address for
365		 * the corresponding operation.
366		 */
367		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
368		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
369			error = EAFNOSUPPORT;
370			goto out;
371		}
372
373		if (td != NULL) {
374			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
375			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
376			if (error)
377				goto out;
378		}
379		break;
380
381	case SIOCGIFADDR_IN6:
382		/* This interface is basically deprecated. use SIOCGIFCONF. */
383		/* FALLTHROUGH */
384	case SIOCGIFAFLAG_IN6:
385	case SIOCGIFNETMASK_IN6:
386	case SIOCGIFDSTADDR_IN6:
387	case SIOCGIFALIFETIME_IN6:
388		/* must think again about its semantics */
389		if (ia == NULL) {
390			error = EADDRNOTAVAIL;
391			goto out;
392		}
393		break;
394
395	case SIOCSIFALIFETIME_IN6:
396	    {
397		struct in6_addrlifetime *lt;
398
399		if (td != NULL) {
400			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
401			if (error)
402				goto out;
403		}
404		if (ia == NULL) {
405			error = EADDRNOTAVAIL;
406			goto out;
407		}
408		/* sanity for overflow - beware unsigned */
409		lt = &ifr->ifr_ifru.ifru_lifetime;
410		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
411		    lt->ia6t_vltime + time_second < time_second) {
412			error = EINVAL;
413			goto out;
414		}
415		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
416		    lt->ia6t_pltime + time_second < time_second) {
417			error = EINVAL;
418			goto out;
419		}
420		break;
421	    }
422	}
423
424	switch (cmd) {
425	case SIOCGIFADDR_IN6:
426		ifr->ifr_addr = ia->ia_addr;
427		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
428			goto out;
429		break;
430
431	case SIOCGIFDSTADDR_IN6:
432		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
433			error = EINVAL;
434			goto out;
435		}
436		/*
437		 * XXX: should we check if ifa_dstaddr is NULL and return
438		 * an error?
439		 */
440		ifr->ifr_dstaddr = ia->ia_dstaddr;
441		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
442			goto out;
443		break;
444
445	case SIOCGIFNETMASK_IN6:
446		ifr->ifr_addr = ia->ia_prefixmask;
447		break;
448
449	case SIOCGIFAFLAG_IN6:
450		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
451		break;
452
453	case SIOCGIFSTAT_IN6:
454		if (ifp == NULL) {
455			error = EINVAL;
456			goto out;
457		}
458		bzero(&ifr->ifr_ifru.ifru_stat,
459		    sizeof(ifr->ifr_ifru.ifru_stat));
460		ifr->ifr_ifru.ifru_stat =
461		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
462		break;
463
464	case SIOCGIFSTAT_ICMP6:
465		if (ifp == NULL) {
466			error = EINVAL;
467			goto out;
468		}
469		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
470		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
471		ifr->ifr_ifru.ifru_icmp6stat =
472		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
473		break;
474
475	case SIOCGIFALIFETIME_IN6:
476		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
477		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
478			time_t maxexpire;
479			struct in6_addrlifetime *retlt =
480			    &ifr->ifr_ifru.ifru_lifetime;
481
482			/*
483			 * XXX: adjust expiration time assuming time_t is
484			 * signed.
485			 */
486			maxexpire = (-1) &
487			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
488			if (ia->ia6_lifetime.ia6t_vltime <
489			    maxexpire - ia->ia6_updatetime) {
490				retlt->ia6t_expire = ia->ia6_updatetime +
491				    ia->ia6_lifetime.ia6t_vltime;
492			} else
493				retlt->ia6t_expire = maxexpire;
494		}
495		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
496			time_t maxexpire;
497			struct in6_addrlifetime *retlt =
498			    &ifr->ifr_ifru.ifru_lifetime;
499
500			/*
501			 * XXX: adjust expiration time assuming time_t is
502			 * signed.
503			 */
504			maxexpire = (-1) &
505			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
506			if (ia->ia6_lifetime.ia6t_pltime <
507			    maxexpire - ia->ia6_updatetime) {
508				retlt->ia6t_preferred = ia->ia6_updatetime +
509				    ia->ia6_lifetime.ia6t_pltime;
510			} else
511				retlt->ia6t_preferred = maxexpire;
512		}
513		break;
514
515	case SIOCSIFALIFETIME_IN6:
516		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
517		/* for sanity */
518		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
519			ia->ia6_lifetime.ia6t_expire =
520				time_second + ia->ia6_lifetime.ia6t_vltime;
521		} else
522			ia->ia6_lifetime.ia6t_expire = 0;
523		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
524			ia->ia6_lifetime.ia6t_preferred =
525				time_second + ia->ia6_lifetime.ia6t_pltime;
526		} else
527			ia->ia6_lifetime.ia6t_preferred = 0;
528		break;
529
530	case SIOCAIFADDR_IN6:
531	{
532		int i;
533		struct nd_prefixctl pr0;
534		struct nd_prefix *pr;
535
536		/*
537		 * first, make or update the interface address structure,
538		 * and link it to the list.
539		 */
540		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
541			goto out;
542		if (ia != NULL)
543			ifa_free(&ia->ia_ifa);
544		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
545		    == NULL) {
546			/*
547			 * this can happen when the user specify the 0 valid
548			 * lifetime.
549			 */
550			break;
551		}
552
553		/*
554		 * then, make the prefix on-link on the interface.
555		 * XXX: we'd rather create the prefix before the address, but
556		 * we need at least one address to install the corresponding
557		 * interface route, so we configure the address first.
558		 */
559
560		/*
561		 * convert mask to prefix length (prefixmask has already
562		 * been validated in in6_update_ifa().
563		 */
564		bzero(&pr0, sizeof(pr0));
565		pr0.ndpr_ifp = ifp;
566		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
567		    NULL);
568		if (pr0.ndpr_plen == 128) {
569			break;	/* we don't need to install a host route. */
570		}
571		pr0.ndpr_prefix = ifra->ifra_addr;
572		/* apply the mask for safety. */
573		for (i = 0; i < 4; i++) {
574			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
575			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
576		}
577		/*
578		 * XXX: since we don't have an API to set prefix (not address)
579		 * lifetimes, we just use the same lifetimes as addresses.
580		 * The (temporarily) installed lifetimes can be overridden by
581		 * later advertised RAs (when accept_rtadv is non 0), which is
582		 * an intended behavior.
583		 */
584		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
585		pr0.ndpr_raf_auto =
586		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
587		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
588		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
589
590		/* add the prefix if not yet. */
591		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
592			/*
593			 * nd6_prelist_add will install the corresponding
594			 * interface route.
595			 */
596			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
597				goto out;
598			if (pr == NULL) {
599				log(LOG_ERR, "nd6_prelist_add succeeded but "
600				    "no prefix\n");
601				error = EINVAL;
602				goto out;
603			}
604		}
605
606		/* relate the address to the prefix */
607		if (ia->ia6_ndpr == NULL) {
608			ia->ia6_ndpr = pr;
609			pr->ndpr_refcnt++;
610
611			/*
612			 * If this is the first autoconf address from the
613			 * prefix, create a temporary address as well
614			 * (when required).
615			 */
616			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
617			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
618				int e;
619				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
620					log(LOG_NOTICE, "in6_control: failed "
621					    "to create a temporary address, "
622					    "errno=%d\n", e);
623				}
624			}
625		}
626
627		/*
628		 * this might affect the status of autoconfigured addresses,
629		 * that is, this address might make other addresses detached.
630		 */
631		pfxlist_onlink_check();
632		if (error == 0 && ia)
633			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
634		break;
635	}
636
637	case SIOCDIFADDR_IN6:
638	{
639		struct nd_prefix *pr;
640
641		/*
642		 * If the address being deleted is the only one that owns
643		 * the corresponding prefix, expire the prefix as well.
644		 * XXX: theoretically, we don't have to worry about such
645		 * relationship, since we separate the address management
646		 * and the prefix management.  We do this, however, to provide
647		 * as much backward compatibility as possible in terms of
648		 * the ioctl operation.
649		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
650		 */
651		pr = ia->ia6_ndpr;
652		in6_purgeaddr(&ia->ia_ifa);
653		if (pr && pr->ndpr_refcnt == 0)
654			prelist_remove(pr);
655		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
656		break;
657	}
658
659	default:
660		if (ifp == NULL || ifp->if_ioctl == 0) {
661			error = EOPNOTSUPP;
662			goto out;
663		}
664		error = (*ifp->if_ioctl)(ifp, cmd, data);
665		goto out;
666	}
667
668	error = 0;
669out:
670	if (ia != NULL)
671		ifa_free(&ia->ia_ifa);
672	return (error);
673}
674
675/*
676 * Update parameters of an IPv6 interface address.
677 * If necessary, a new entry is created and linked into address chains.
678 * This function is separated from in6_control().
679 * XXX: should this be performed under splnet()?
680 */
681int
682in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
683    struct in6_ifaddr *ia, int flags)
684{
685	INIT_VNET_INET6(ifp->if_vnet);
686	int error = 0, hostIsNew = 0, plen = -1;
687	struct in6_ifaddr *oia;
688	struct sockaddr_in6 dst6;
689	struct in6_addrlifetime *lt;
690	struct in6_multi_mship *imm;
691	struct in6_multi *in6m_sol;
692	struct rtentry *rt;
693	int delay;
694	char ip6buf[INET6_ADDRSTRLEN];
695
696	/* Validate parameters */
697	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
698		return (EINVAL);
699
700	/*
701	 * The destination address for a p2p link must have a family
702	 * of AF_UNSPEC or AF_INET6.
703	 */
704	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
705	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
706	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
707		return (EAFNOSUPPORT);
708	/*
709	 * validate ifra_prefixmask.  don't check sin6_family, netmask
710	 * does not carry fields other than sin6_len.
711	 */
712	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
713		return (EINVAL);
714	/*
715	 * Because the IPv6 address architecture is classless, we require
716	 * users to specify a (non 0) prefix length (mask) for a new address.
717	 * We also require the prefix (when specified) mask is valid, and thus
718	 * reject a non-consecutive mask.
719	 */
720	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
721		return (EINVAL);
722	if (ifra->ifra_prefixmask.sin6_len != 0) {
723		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
724		    (u_char *)&ifra->ifra_prefixmask +
725		    ifra->ifra_prefixmask.sin6_len);
726		if (plen <= 0)
727			return (EINVAL);
728	} else {
729		/*
730		 * In this case, ia must not be NULL.  We just use its prefix
731		 * length.
732		 */
733		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
734	}
735	/*
736	 * If the destination address on a p2p interface is specified,
737	 * and the address is a scoped one, validate/set the scope
738	 * zone identifier.
739	 */
740	dst6 = ifra->ifra_dstaddr;
741	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
742	    (dst6.sin6_family == AF_INET6)) {
743		struct in6_addr in6_tmp;
744		u_int32_t zoneid;
745
746		in6_tmp = dst6.sin6_addr;
747		if (in6_setscope(&in6_tmp, ifp, &zoneid))
748			return (EINVAL); /* XXX: should be impossible */
749
750		if (dst6.sin6_scope_id != 0) {
751			if (dst6.sin6_scope_id != zoneid)
752				return (EINVAL);
753		} else		/* user omit to specify the ID. */
754			dst6.sin6_scope_id = zoneid;
755
756		/* convert into the internal form */
757		if (sa6_embedscope(&dst6, 0))
758			return (EINVAL); /* XXX: should be impossible */
759	}
760	/*
761	 * The destination address can be specified only for a p2p or a
762	 * loopback interface.  If specified, the corresponding prefix length
763	 * must be 128.
764	 */
765	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
766		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
767			/* XXX: noisy message */
768			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
769			    "be specified for a p2p or a loopback IF only\n"));
770			return (EINVAL);
771		}
772		if (plen != 128) {
773			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
774			    "be 128 when dstaddr is specified\n"));
775			return (EINVAL);
776		}
777	}
778	/* lifetime consistency check */
779	lt = &ifra->ifra_lifetime;
780	if (lt->ia6t_pltime > lt->ia6t_vltime)
781		return (EINVAL);
782	if (lt->ia6t_vltime == 0) {
783		/*
784		 * the following log might be noisy, but this is a typical
785		 * configuration mistake or a tool's bug.
786		 */
787		nd6log((LOG_INFO,
788		    "in6_update_ifa: valid lifetime is 0 for %s\n",
789		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
790
791		if (ia == NULL)
792			return (0); /* there's nothing to do */
793	}
794
795	/*
796	 * If this is a new address, allocate a new ifaddr and link it
797	 * into chains.
798	 */
799	if (ia == NULL) {
800		hostIsNew = 1;
801		/*
802		 * When in6_update_ifa() is called in a process of a received
803		 * RA, it is called under an interrupt context.  So, we should
804		 * call malloc with M_NOWAIT.
805		 */
806		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
807		    M_NOWAIT);
808		if (ia == NULL)
809			return (ENOBUFS);
810		bzero((caddr_t)ia, sizeof(*ia));
811		ifa_init(&ia->ia_ifa);
812		LIST_INIT(&ia->ia6_memberships);
813		/* Initialize the address and masks, and put time stamp */
814		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
815		ia->ia_addr.sin6_family = AF_INET6;
816		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
817		ia->ia6_createtime = time_second;
818		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
819			/*
820			 * XXX: some functions expect that ifa_dstaddr is not
821			 * NULL for p2p interfaces.
822			 */
823			ia->ia_ifa.ifa_dstaddr =
824			    (struct sockaddr *)&ia->ia_dstaddr;
825		} else {
826			ia->ia_ifa.ifa_dstaddr = NULL;
827		}
828		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
829
830		ia->ia_ifp = ifp;
831		if ((oia = V_in6_ifaddr) != NULL) {
832			for ( ; oia->ia_next; oia = oia->ia_next)
833				continue;
834			oia->ia_next = ia;
835		} else
836			V_in6_ifaddr = ia;
837
838		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
839		IF_ADDR_LOCK(ifp);
840		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
841		IF_ADDR_UNLOCK(ifp);
842	}
843
844	/* update timestamp */
845	ia->ia6_updatetime = time_second;
846
847	/* set prefix mask */
848	if (ifra->ifra_prefixmask.sin6_len) {
849		/*
850		 * We prohibit changing the prefix length of an existing
851		 * address, because
852		 * + such an operation should be rare in IPv6, and
853		 * + the operation would confuse prefix management.
854		 */
855		if (ia->ia_prefixmask.sin6_len &&
856		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
857			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
858			    " existing (%s) address should not be changed\n",
859			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
860			error = EINVAL;
861			goto unlink;
862		}
863		ia->ia_prefixmask = ifra->ifra_prefixmask;
864	}
865
866	/*
867	 * If a new destination address is specified, scrub the old one and
868	 * install the new destination.  Note that the interface must be
869	 * p2p or loopback (see the check above.)
870	 */
871	if (dst6.sin6_family == AF_INET6 &&
872	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
873		int e;
874
875		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
876		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
877			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
878			    "a route to the old destination: %s\n",
879			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
880			/* proceed anyway... */
881		} else
882			ia->ia_flags &= ~IFA_ROUTE;
883		ia->ia_dstaddr = dst6;
884	}
885
886	/*
887	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
888	 * to see if the address is deprecated or invalidated, but initialize
889	 * these members for applications.
890	 */
891	ia->ia6_lifetime = ifra->ifra_lifetime;
892	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
893		ia->ia6_lifetime.ia6t_expire =
894		    time_second + ia->ia6_lifetime.ia6t_vltime;
895	} else
896		ia->ia6_lifetime.ia6t_expire = 0;
897	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
898		ia->ia6_lifetime.ia6t_preferred =
899		    time_second + ia->ia6_lifetime.ia6t_pltime;
900	} else
901		ia->ia6_lifetime.ia6t_preferred = 0;
902
903	/* reset the interface and routing table appropriately. */
904	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
905		goto unlink;
906
907	/*
908	 * configure address flags.
909	 */
910	ia->ia6_flags = ifra->ifra_flags;
911	/*
912	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
913	 * userland, make it deprecated.
914	 */
915	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
916		ia->ia6_lifetime.ia6t_pltime = 0;
917		ia->ia6_lifetime.ia6t_preferred = time_second;
918	}
919	/*
920	 * Make the address tentative before joining multicast addresses,
921	 * so that corresponding MLD responses would not have a tentative
922	 * source address.
923	 */
924	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
925	if (hostIsNew && in6if_do_dad(ifp))
926		ia->ia6_flags |= IN6_IFF_TENTATIVE;
927
928	/*
929	 * We are done if we have simply modified an existing address.
930	 */
931	if (!hostIsNew)
932		return (error);
933
934	/*
935	 * Beyond this point, we should call in6_purgeaddr upon an error,
936	 * not just go to unlink.
937	 */
938
939	/* Join necessary multicast groups */
940	in6m_sol = NULL;
941	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
942		struct sockaddr_in6 mltaddr, mltmask;
943		struct in6_addr llsol;
944
945		/* join solicited multicast addr for new host id */
946		bzero(&llsol, sizeof(struct in6_addr));
947		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
948		llsol.s6_addr32[1] = 0;
949		llsol.s6_addr32[2] = htonl(1);
950		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
951		llsol.s6_addr8[12] = 0xff;
952		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
953			/* XXX: should not happen */
954			log(LOG_ERR, "in6_update_ifa: "
955			    "in6_setscope failed\n");
956			goto cleanup;
957		}
958		delay = 0;
959		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
960			/*
961			 * We need a random delay for DAD on the address
962			 * being configured.  It also means delaying
963			 * transmission of the corresponding MLD report to
964			 * avoid report collision.
965			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
966			 */
967			delay = arc4random() %
968			    (MAX_RTR_SOLICITATION_DELAY * hz);
969		}
970		imm = in6_joingroup(ifp, &llsol, &error, delay);
971		if (imm == NULL) {
972			nd6log((LOG_WARNING,
973			    "in6_update_ifa: addmulti failed for "
974			    "%s on %s (errno=%d)\n",
975			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
976			    error));
977			in6_purgeaddr((struct ifaddr *)ia);
978			return (error);
979		}
980		LIST_INSERT_HEAD(&ia->ia6_memberships,
981		    imm, i6mm_chain);
982		in6m_sol = imm->i6mm_maddr;
983
984		bzero(&mltmask, sizeof(mltmask));
985		mltmask.sin6_len = sizeof(struct sockaddr_in6);
986		mltmask.sin6_family = AF_INET6;
987		mltmask.sin6_addr = in6mask32;
988#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
989
990		/*
991		 * join link-local all-nodes address
992		 */
993		bzero(&mltaddr, sizeof(mltaddr));
994		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
995		mltaddr.sin6_family = AF_INET6;
996		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
997		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
998		    0)
999			goto cleanup; /* XXX: should not fail */
1000
1001		/*
1002		 * XXX: do we really need this automatic routes?
1003		 * We should probably reconsider this stuff.  Most applications
1004		 * actually do not need the routes, since they usually specify
1005		 * the outgoing interface.
1006		 */
1007		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1008		if (rt) {
1009			/* XXX: only works in !SCOPEDROUTING case. */
1010			if (memcmp(&mltaddr.sin6_addr,
1011			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1012			    MLTMASK_LEN)) {
1013				RTFREE_LOCKED(rt);
1014				rt = NULL;
1015			}
1016		}
1017		if (!rt) {
1018			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1019			    (struct sockaddr *)&ia->ia_addr,
1020			    (struct sockaddr *)&mltmask, RTF_UP,
1021			    (struct rtentry **)0);
1022			if (error)
1023				goto cleanup;
1024		} else {
1025			RTFREE_LOCKED(rt);
1026		}
1027
1028		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1029		if (!imm) {
1030			nd6log((LOG_WARNING,
1031			    "in6_update_ifa: addmulti failed for "
1032			    "%s on %s (errno=%d)\n",
1033			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1034			    if_name(ifp), error));
1035			goto cleanup;
1036		}
1037		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1038
1039		/*
1040		 * join node information group address
1041		 */
1042		delay = 0;
1043		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1044			/*
1045			 * The spec doesn't say anything about delay for this
1046			 * group, but the same logic should apply.
1047			 */
1048			delay = arc4random() %
1049			    (MAX_RTR_SOLICITATION_DELAY * hz);
1050		}
1051		if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
1052			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1053			    delay); /* XXX jinmei */
1054			if (!imm) {
1055				nd6log((LOG_WARNING, "in6_update_ifa: "
1056				    "addmulti failed for %s on %s "
1057				    "(errno=%d)\n",
1058				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1059				    if_name(ifp), error));
1060				/* XXX not very fatal, go on... */
1061			} else {
1062				LIST_INSERT_HEAD(&ia->ia6_memberships,
1063				    imm, i6mm_chain);
1064			}
1065		}
1066
1067		/*
1068		 * join interface-local all-nodes address.
1069		 * (ff01::1%ifN, and ff01::%ifN/32)
1070		 */
1071		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1072		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1073		    != 0)
1074			goto cleanup; /* XXX: should not fail */
1075		/* XXX: again, do we really need the route? */
1076		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1077		if (rt) {
1078			if (memcmp(&mltaddr.sin6_addr,
1079			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1080			    MLTMASK_LEN)) {
1081				RTFREE_LOCKED(rt);
1082				rt = NULL;
1083			}
1084		}
1085		if (!rt) {
1086			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1087			    (struct sockaddr *)&ia->ia_addr,
1088			    (struct sockaddr *)&mltmask, RTF_UP,
1089			    (struct rtentry **)0);
1090			if (error)
1091				goto cleanup;
1092		} else
1093			RTFREE_LOCKED(rt);
1094
1095		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1096		if (!imm) {
1097			nd6log((LOG_WARNING, "in6_update_ifa: "
1098			    "addmulti failed for %s on %s "
1099			    "(errno=%d)\n",
1100			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1101			    if_name(ifp), error));
1102			goto cleanup;
1103		}
1104		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1105#undef	MLTMASK_LEN
1106	}
1107
1108	/*
1109	 * Perform DAD, if needed.
1110	 * XXX It may be of use, if we can administratively
1111	 * disable DAD.
1112	 */
1113	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1114	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1115	{
1116		int mindelay, maxdelay;
1117
1118		delay = 0;
1119		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1120			/*
1121			 * We need to impose a delay before sending an NS
1122			 * for DAD.  Check if we also needed a delay for the
1123			 * corresponding MLD message.  If we did, the delay
1124			 * should be larger than the MLD delay (this could be
1125			 * relaxed a bit, but this simple logic is at least
1126			 * safe).
1127			 * XXX: Break data hiding guidelines and look at
1128			 * state for the solicited multicast group.
1129			 */
1130			mindelay = 0;
1131			if (in6m_sol != NULL &&
1132			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1133				mindelay = in6m_sol->in6m_timer;
1134			}
1135			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1136			if (maxdelay - mindelay == 0)
1137				delay = 0;
1138			else {
1139				delay =
1140				    (arc4random() % (maxdelay - mindelay)) +
1141				    mindelay;
1142			}
1143		}
1144		nd6_dad_start((struct ifaddr *)ia, delay);
1145	}
1146
1147	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1148	ifa_free(&ia->ia_ifa);
1149	return (error);
1150
1151  unlink:
1152	/*
1153	 * XXX: if a change of an existing address failed, keep the entry
1154	 * anyway.
1155	 */
1156	if (hostIsNew) {
1157		ifa_free(&ia->ia_ifa);
1158		in6_unlink_ifa(ia, ifp);
1159	}
1160	return (error);
1161
1162  cleanup:
1163	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1164	ifa_free(&ia->ia_ifa);
1165	in6_purgeaddr(&ia->ia_ifa);
1166	return error;
1167}
1168
1169void
1170in6_purgeaddr(struct ifaddr *ifa)
1171{
1172	struct ifnet *ifp = ifa->ifa_ifp;
1173	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1174	struct in6_multi_mship *imm;
1175	struct sockaddr_in6 mltaddr, mltmask;
1176	struct rtentry rt0;
1177	struct sockaddr_dl gateway;
1178	struct sockaddr_in6 mask, addr;
1179	int plen, error;
1180	struct rtentry *rt;
1181	struct ifaddr *ifa0, *nifa;
1182
1183	/*
1184	 * find another IPv6 address as the gateway for the
1185	 * link-local and node-local all-nodes multicast
1186	 * address routes
1187	 */
1188	IF_ADDR_LOCK(ifp);
1189	TAILQ_FOREACH_SAFE(ifa0, &ifp->if_addrhead, ifa_link, nifa) {
1190		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1191		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1192			   &ia->ia_addr.sin6_addr,
1193			   sizeof(struct in6_addr)) == 0)
1194			continue;
1195		else
1196			break;
1197	}
1198	if (ifa0 != NULL)
1199		ifa_ref(ifa0);
1200	IF_ADDR_UNLOCK(ifp);
1201
1202	/* stop DAD processing */
1203	nd6_dad_stop(ifa);
1204
1205	IF_AFDATA_LOCK(ifp);
1206	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
1207	    (struct sockaddr *)&ia->ia_addr);
1208	IF_AFDATA_UNLOCK(ifp);
1209
1210	/*
1211	 * initialize for rtmsg generation
1212	 */
1213	bzero(&gateway, sizeof(gateway));
1214	gateway.sdl_len = sizeof(gateway);
1215	gateway.sdl_family = AF_LINK;
1216	gateway.sdl_nlen = 0;
1217	gateway.sdl_alen = ifp->if_addrlen;
1218	/* */
1219	bzero(&rt0, sizeof(rt0));
1220	rt0.rt_gateway = (struct sockaddr *)&gateway;
1221	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
1222	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
1223	rt_mask(&rt0) = (struct sockaddr *)&mask;
1224	rt_key(&rt0) = (struct sockaddr *)&addr;
1225	rt0.rt_flags = RTF_HOST | RTF_STATIC;
1226	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
1227
1228	/*
1229	 * leave from multicast groups we have joined for the interface
1230	 */
1231	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1232		LIST_REMOVE(imm, i6mm_chain);
1233		in6_leavegroup(imm);
1234	}
1235
1236	/*
1237	 * remove the link-local all-nodes address
1238	 */
1239	bzero(&mltmask, sizeof(mltmask));
1240	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1241	mltmask.sin6_family = AF_INET6;
1242	mltmask.sin6_addr = in6mask32;
1243
1244	bzero(&mltaddr, sizeof(mltaddr));
1245	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1246	mltaddr.sin6_family = AF_INET6;
1247	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1248
1249	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1250	    0)
1251		goto cleanup;
1252
1253	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1254	if (rt != NULL && rt->rt_gateway != NULL &&
1255	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1256		    &ia->ia_addr.sin6_addr,
1257		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1258		/*
1259		 * if no more IPv6 address exists on this interface
1260		 * then remove the multicast address route
1261		 */
1262		if (ifa0 == NULL) {
1263			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1264			       sizeof(mltaddr.sin6_addr));
1265			RTFREE_LOCKED(rt);
1266			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1267					  (struct sockaddr *)&ia->ia_addr,
1268					  (struct sockaddr *)&mltmask, RTF_UP,
1269					  (struct rtentry **)0);
1270			if (error)
1271				log(LOG_INFO, "in6_purgeaddr: link-local all-nodes"
1272				    "multicast address deletion error\n");
1273		} else {
1274			/*
1275			 * replace the gateway of the route
1276			 */
1277			struct sockaddr_in6 sa;
1278
1279			bzero(&sa, sizeof(sa));
1280			sa.sin6_len = sizeof(struct sockaddr_in6);
1281			sa.sin6_family = AF_INET6;
1282			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1283			       sizeof(sa.sin6_addr));
1284			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1285			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1286			RTFREE_LOCKED(rt);
1287		}
1288	} else {
1289		if (rt != NULL)
1290			RTFREE_LOCKED(rt);
1291	}
1292
1293	/*
1294	 * remove the node-local all-nodes address
1295	 */
1296	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1297	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1298	    0)
1299		goto cleanup;
1300
1301	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1302	if (rt != NULL && rt->rt_gateway != NULL &&
1303	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1304		    &ia->ia_addr.sin6_addr,
1305		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1306		/*
1307		 * if no more IPv6 address exists on this interface
1308		 * then remove the multicast address route
1309		 */
1310		if (ifa0 == NULL) {
1311			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1312			       sizeof(mltaddr.sin6_addr));
1313
1314			RTFREE_LOCKED(rt);
1315			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1316					  (struct sockaddr *)&ia->ia_addr,
1317					  (struct sockaddr *)&mltmask, RTF_UP,
1318					  (struct rtentry **)0);
1319
1320			if (error)
1321				log(LOG_INFO, "in6_purgeaddr: node-local all-nodes"
1322				    "multicast address deletion error\n");
1323		} else {
1324			/*
1325			 * replace the gateway of the route
1326			 */
1327			struct sockaddr_in6 sa;
1328
1329			bzero(&sa, sizeof(sa));
1330			sa.sin6_len = sizeof(struct sockaddr_in6);
1331			sa.sin6_family = AF_INET6;
1332			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1333			       sizeof(sa.sin6_addr));
1334			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1335			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1336			RTFREE_LOCKED(rt);
1337		}
1338	} else {
1339		if (rt != NULL)
1340			RTFREE_LOCKED(rt);
1341	}
1342
1343cleanup:
1344
1345	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1346	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1347		int error;
1348		struct sockaddr *dstaddr;
1349
1350		/*
1351		 * use the interface address if configuring an
1352		 * interface address with a /128 prefix len
1353		 */
1354		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1355			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1356		else
1357			dstaddr = (struct sockaddr *)&ia->ia_addr;
1358
1359		error = rtrequest(RTM_DELETE,
1360		    (struct sockaddr *)dstaddr,
1361		    (struct sockaddr *)&ia->ia_addr,
1362		    (struct sockaddr *)&ia->ia_prefixmask,
1363		    ia->ia_flags | RTF_HOST, NULL);
1364		if (error != 0)
1365			return;
1366		ia->ia_flags &= ~IFA_ROUTE;
1367	}
1368	if (ifa0 != NULL)
1369		ifa_free(ifa0);
1370
1371	in6_unlink_ifa(ia, ifp);
1372}
1373
1374static void
1375in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1376{
1377	INIT_VNET_INET6(ifp->if_vnet);
1378	struct in6_ifaddr *oia;
1379	int	s = splnet();
1380
1381	IF_ADDR_LOCK(ifp);
1382	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1383	IF_ADDR_UNLOCK(ifp);
1384	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1385
1386	oia = ia;
1387	if (oia == (ia = V_in6_ifaddr))
1388		V_in6_ifaddr = ia->ia_next;
1389	else {
1390		while (ia->ia_next && (ia->ia_next != oia))
1391			ia = ia->ia_next;
1392		if (ia->ia_next)
1393			ia->ia_next = oia->ia_next;
1394		else {
1395			/* search failed */
1396			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1397		}
1398	}
1399
1400	/*
1401	 * Release the reference to the base prefix.  There should be a
1402	 * positive reference.
1403	 */
1404	if (oia->ia6_ndpr == NULL) {
1405		nd6log((LOG_NOTICE,
1406		    "in6_unlink_ifa: autoconf'ed address "
1407		    "%p has no prefix\n", oia));
1408	} else {
1409		oia->ia6_ndpr->ndpr_refcnt--;
1410		oia->ia6_ndpr = NULL;
1411	}
1412
1413	/*
1414	 * Also, if the address being removed is autoconf'ed, call
1415	 * pfxlist_onlink_check() since the release might affect the status of
1416	 * other (detached) addresses.
1417	 */
1418	if ((oia->ia6_flags & IN6_IFF_AUTOCONF)) {
1419		pfxlist_onlink_check();
1420	}
1421
1422	/*
1423	 * release another refcnt for the link from in6_ifaddr.
1424	 * Note that we should decrement the refcnt at least once for all *BSD.
1425	 */
1426	ifa_free(&oia->ia_ifa);
1427
1428	splx(s);
1429}
1430
1431void
1432in6_purgeif(struct ifnet *ifp)
1433{
1434	struct ifaddr *ifa, *nifa;
1435
1436	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1437		if (ifa->ifa_addr->sa_family != AF_INET6)
1438			continue;
1439		in6_purgeaddr(ifa);
1440	}
1441
1442	in6_ifdetach(ifp);
1443}
1444
1445/*
1446 * SIOC[GAD]LIFADDR.
1447 *	SIOCGLIFADDR: get first address. (?)
1448 *	SIOCGLIFADDR with IFLR_PREFIX:
1449 *		get first address that matches the specified prefix.
1450 *	SIOCALIFADDR: add the specified address.
1451 *	SIOCALIFADDR with IFLR_PREFIX:
1452 *		add the specified prefix, filling hostid part from
1453 *		the first link-local address.  prefixlen must be <= 64.
1454 *	SIOCDLIFADDR: delete the specified address.
1455 *	SIOCDLIFADDR with IFLR_PREFIX:
1456 *		delete the first address that matches the specified prefix.
1457 * return values:
1458 *	EINVAL on invalid parameters
1459 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1460 *	other values may be returned from in6_ioctl()
1461 *
1462 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1463 * this is to accomodate address naming scheme other than RFC2374,
1464 * in the future.
1465 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1466 * address encoding scheme. (see figure on page 8)
1467 */
1468static int
1469in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1470    struct ifnet *ifp, struct thread *td)
1471{
1472	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1473	struct ifaddr *ifa;
1474	struct sockaddr *sa;
1475
1476	/* sanity checks */
1477	if (!data || !ifp) {
1478		panic("invalid argument to in6_lifaddr_ioctl");
1479		/* NOTREACHED */
1480	}
1481
1482	switch (cmd) {
1483	case SIOCGLIFADDR:
1484		/* address must be specified on GET with IFLR_PREFIX */
1485		if ((iflr->flags & IFLR_PREFIX) == 0)
1486			break;
1487		/* FALLTHROUGH */
1488	case SIOCALIFADDR:
1489	case SIOCDLIFADDR:
1490		/* address must be specified on ADD and DELETE */
1491		sa = (struct sockaddr *)&iflr->addr;
1492		if (sa->sa_family != AF_INET6)
1493			return EINVAL;
1494		if (sa->sa_len != sizeof(struct sockaddr_in6))
1495			return EINVAL;
1496		/* XXX need improvement */
1497		sa = (struct sockaddr *)&iflr->dstaddr;
1498		if (sa->sa_family && sa->sa_family != AF_INET6)
1499			return EINVAL;
1500		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1501			return EINVAL;
1502		break;
1503	default: /* shouldn't happen */
1504#if 0
1505		panic("invalid cmd to in6_lifaddr_ioctl");
1506		/* NOTREACHED */
1507#else
1508		return EOPNOTSUPP;
1509#endif
1510	}
1511	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1512		return EINVAL;
1513
1514	switch (cmd) {
1515	case SIOCALIFADDR:
1516	    {
1517		struct in6_aliasreq ifra;
1518		struct in6_addr *hostid = NULL;
1519		int prefixlen;
1520
1521		ifa = NULL;
1522		if ((iflr->flags & IFLR_PREFIX) != 0) {
1523			struct sockaddr_in6 *sin6;
1524
1525			/*
1526			 * hostid is to fill in the hostid part of the
1527			 * address.  hostid points to the first link-local
1528			 * address attached to the interface.
1529			 */
1530			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1531			if (!ifa)
1532				return EADDRNOTAVAIL;
1533			hostid = IFA_IN6(ifa);
1534
1535			/* prefixlen must be <= 64. */
1536			if (64 < iflr->prefixlen)
1537				return EINVAL;
1538			prefixlen = iflr->prefixlen;
1539
1540			/* hostid part must be zero. */
1541			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1542			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1543			    sin6->sin6_addr.s6_addr32[3] != 0) {
1544				return EINVAL;
1545			}
1546		} else
1547			prefixlen = iflr->prefixlen;
1548
1549		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1550		bzero(&ifra, sizeof(ifra));
1551		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1552
1553		bcopy(&iflr->addr, &ifra.ifra_addr,
1554		    ((struct sockaddr *)&iflr->addr)->sa_len);
1555		if (hostid) {
1556			/* fill in hostid part */
1557			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1558			    hostid->s6_addr32[2];
1559			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1560			    hostid->s6_addr32[3];
1561		}
1562
1563		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1564			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1565			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1566			if (hostid) {
1567				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1568				    hostid->s6_addr32[2];
1569				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1570				    hostid->s6_addr32[3];
1571			}
1572		}
1573		if (ifa != NULL)
1574			ifa_free(ifa);
1575
1576		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1577		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1578
1579		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1580		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1581	    }
1582	case SIOCGLIFADDR:
1583	case SIOCDLIFADDR:
1584	    {
1585		struct in6_ifaddr *ia;
1586		struct in6_addr mask, candidate, match;
1587		struct sockaddr_in6 *sin6;
1588		int cmp;
1589
1590		bzero(&mask, sizeof(mask));
1591		if (iflr->flags & IFLR_PREFIX) {
1592			/* lookup a prefix rather than address. */
1593			in6_prefixlen2mask(&mask, iflr->prefixlen);
1594
1595			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1596			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1597			match.s6_addr32[0] &= mask.s6_addr32[0];
1598			match.s6_addr32[1] &= mask.s6_addr32[1];
1599			match.s6_addr32[2] &= mask.s6_addr32[2];
1600			match.s6_addr32[3] &= mask.s6_addr32[3];
1601
1602			/* if you set extra bits, that's wrong */
1603			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1604				return EINVAL;
1605
1606			cmp = 1;
1607		} else {
1608			if (cmd == SIOCGLIFADDR) {
1609				/* on getting an address, take the 1st match */
1610				cmp = 0;	/* XXX */
1611			} else {
1612				/* on deleting an address, do exact match */
1613				in6_prefixlen2mask(&mask, 128);
1614				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1615				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1616
1617				cmp = 1;
1618			}
1619		}
1620
1621		IF_ADDR_LOCK(ifp);
1622		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1623			if (ifa->ifa_addr->sa_family != AF_INET6)
1624				continue;
1625			if (!cmp)
1626				break;
1627
1628			/*
1629			 * XXX: this is adhoc, but is necessary to allow
1630			 * a user to specify fe80::/64 (not /10) for a
1631			 * link-local address.
1632			 */
1633			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1634			in6_clearscope(&candidate);
1635			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1636			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1637			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1638			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1639			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1640				break;
1641		}
1642		IF_ADDR_UNLOCK(ifp);
1643		if (!ifa)
1644			return EADDRNOTAVAIL;
1645		ia = ifa2ia6(ifa);
1646
1647		if (cmd == SIOCGLIFADDR) {
1648			int error;
1649
1650			/* fill in the if_laddrreq structure */
1651			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1652			error = sa6_recoverscope(
1653			    (struct sockaddr_in6 *)&iflr->addr);
1654			if (error != 0)
1655				return (error);
1656
1657			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1658				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1659				    ia->ia_dstaddr.sin6_len);
1660				error = sa6_recoverscope(
1661				    (struct sockaddr_in6 *)&iflr->dstaddr);
1662				if (error != 0)
1663					return (error);
1664			} else
1665				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1666
1667			iflr->prefixlen =
1668			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1669
1670			iflr->flags = ia->ia6_flags;	/* XXX */
1671
1672			return 0;
1673		} else {
1674			struct in6_aliasreq ifra;
1675
1676			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1677			bzero(&ifra, sizeof(ifra));
1678			bcopy(iflr->iflr_name, ifra.ifra_name,
1679			    sizeof(ifra.ifra_name));
1680
1681			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1682			    ia->ia_addr.sin6_len);
1683			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1684				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1685				    ia->ia_dstaddr.sin6_len);
1686			} else {
1687				bzero(&ifra.ifra_dstaddr,
1688				    sizeof(ifra.ifra_dstaddr));
1689			}
1690			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1691			    ia->ia_prefixmask.sin6_len);
1692
1693			ifra.ifra_flags = ia->ia6_flags;
1694			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1695			    ifp, td);
1696		}
1697	    }
1698	}
1699
1700	return EOPNOTSUPP;	/* just for safety */
1701}
1702
1703/*
1704 * Initialize an interface's intetnet6 address
1705 * and routing table entry.
1706 */
1707static int
1708in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1709    struct sockaddr_in6 *sin6, int newhost)
1710{
1711	int	error = 0, plen, ifacount = 0;
1712	int	s = splimp();
1713	struct ifaddr *ifa;
1714
1715	/*
1716	 * Give the interface a chance to initialize
1717	 * if this is its first address,
1718	 * and to validate the address if necessary.
1719	 */
1720	IF_ADDR_LOCK(ifp);
1721	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1722		if (ifa->ifa_addr->sa_family != AF_INET6)
1723			continue;
1724		ifacount++;
1725	}
1726	IF_ADDR_UNLOCK(ifp);
1727
1728	ia->ia_addr = *sin6;
1729
1730	if (ifacount <= 1 && ifp->if_ioctl) {
1731		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1732		if (error) {
1733			splx(s);
1734			return (error);
1735		}
1736	}
1737	splx(s);
1738
1739	ia->ia_ifa.ifa_metric = ifp->if_metric;
1740
1741	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1742
1743	/*
1744	 * Special case:
1745	 * If a new destination address is specified for a point-to-point
1746	 * interface, install a route to the destination as an interface
1747	 * direct route.
1748	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1749	 * interface that share the same destination.
1750	 */
1751	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1752	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128) {
1753		struct sockaddr *dstaddr;
1754		int rtflags = RTF_UP | RTF_HOST;
1755
1756		/*
1757		 * use the interface address if configuring an
1758		 * interface address with a /128 prefix len
1759		 */
1760		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1761			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1762		else
1763			dstaddr = (struct sockaddr *)&ia->ia_addr;
1764
1765		error = rtrequest(RTM_ADD,
1766		    (struct sockaddr *)dstaddr,
1767		    (struct sockaddr *)&ia->ia_addr,
1768		    (struct sockaddr *)&ia->ia_prefixmask,
1769		    ia->ia_flags | rtflags, NULL);
1770		if (error != 0)
1771			return (error);
1772		ia->ia_flags |= IFA_ROUTE;
1773	}
1774
1775	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1776	if (newhost) {
1777		struct llentry *ln;
1778		struct rtentry rt;
1779		struct sockaddr_dl gateway;
1780		struct sockaddr_in6 mask, addr;
1781
1782		IF_AFDATA_LOCK(ifp);
1783		ia->ia_ifa.ifa_rtrequest = NULL;
1784
1785		/* XXX QL
1786		 * we need to report rt_newaddrmsg
1787		 */
1788		ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR | LLE_EXCLUSIVE),
1789		    (struct sockaddr *)&ia->ia_addr);
1790		IF_AFDATA_UNLOCK(ifp);
1791		if (ln != NULL) {
1792			ln->la_expire = 0;  /* for IPv6 this means permanent */
1793			ln->ln_state = ND6_LLINFO_REACHABLE;
1794			/*
1795			 * initialize for rtmsg generation
1796			 */
1797			bzero(&gateway, sizeof(gateway));
1798			gateway.sdl_len = sizeof(gateway);
1799			gateway.sdl_family = AF_LINK;
1800			gateway.sdl_nlen = 0;
1801			gateway.sdl_alen = 6;
1802			memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned, sizeof(ln->ll_addr));
1803			/* */
1804			LLE_WUNLOCK(ln);
1805		}
1806
1807		bzero(&rt, sizeof(rt));
1808		rt.rt_gateway = (struct sockaddr *)&gateway;
1809		memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
1810		memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
1811		rt_mask(&rt) = (struct sockaddr *)&mask;
1812		rt_key(&rt) = (struct sockaddr *)&addr;
1813		rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
1814		rt_newaddrmsg(RTM_ADD, &ia->ia_ifa, 0, &rt);
1815	}
1816
1817	return (error);
1818}
1819
1820/*
1821 * Find an IPv6 interface link-local address specific to an interface.
1822 * ifaddr is returned referenced.
1823 */
1824struct in6_ifaddr *
1825in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1826{
1827	struct ifaddr *ifa;
1828
1829	IF_ADDR_LOCK(ifp);
1830	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1831		if (ifa->ifa_addr->sa_family != AF_INET6)
1832			continue;
1833		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1834			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1835			     ignoreflags) != 0)
1836				continue;
1837			ifa_ref(ifa);
1838			break;
1839		}
1840	}
1841	IF_ADDR_UNLOCK(ifp);
1842
1843	return ((struct in6_ifaddr *)ifa);
1844}
1845
1846
1847/*
1848 * find the internet address corresponding to a given interface and address.
1849 * ifaddr is returned referenced.
1850 */
1851struct in6_ifaddr *
1852in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1853{
1854	struct ifaddr *ifa;
1855
1856	IF_ADDR_LOCK(ifp);
1857	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1858		if (ifa->ifa_addr->sa_family != AF_INET6)
1859			continue;
1860		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1861			ifa_ref(ifa);
1862			break;
1863		}
1864	}
1865	IF_ADDR_UNLOCK(ifp);
1866
1867	return ((struct in6_ifaddr *)ifa);
1868}
1869
1870/*
1871 * Convert IP6 address to printable (loggable) representation. Caller
1872 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1873 */
1874static char digits[] = "0123456789abcdef";
1875char *
1876ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1877{
1878	int i;
1879	char *cp;
1880	const u_int16_t *a = (const u_int16_t *)addr;
1881	const u_int8_t *d;
1882	int dcolon = 0, zero = 0;
1883
1884	cp = ip6buf;
1885
1886	for (i = 0; i < 8; i++) {
1887		if (dcolon == 1) {
1888			if (*a == 0) {
1889				if (i == 7)
1890					*cp++ = ':';
1891				a++;
1892				continue;
1893			} else
1894				dcolon = 2;
1895		}
1896		if (*a == 0) {
1897			if (dcolon == 0 && *(a + 1) == 0) {
1898				if (i == 0)
1899					*cp++ = ':';
1900				*cp++ = ':';
1901				dcolon = 1;
1902			} else {
1903				*cp++ = '0';
1904				*cp++ = ':';
1905			}
1906			a++;
1907			continue;
1908		}
1909		d = (const u_char *)a;
1910		/* Try to eliminate leading zeros in printout like in :0001. */
1911		zero = 1;
1912		*cp = digits[*d >> 4];
1913		if (*cp != '0') {
1914			zero = 0;
1915			cp++;
1916		}
1917		*cp = digits[*d++ & 0xf];
1918		if (zero == 0 || (*cp != '0')) {
1919			zero = 0;
1920			cp++;
1921		}
1922		*cp = digits[*d >> 4];
1923		if (zero == 0 || (*cp != '0')) {
1924			zero = 0;
1925			cp++;
1926		}
1927		*cp++ = digits[*d & 0xf];
1928		*cp++ = ':';
1929		a++;
1930	}
1931	*--cp = '\0';
1932	return (ip6buf);
1933}
1934
1935int
1936in6_localaddr(struct in6_addr *in6)
1937{
1938	INIT_VNET_INET6(curvnet);
1939	struct in6_ifaddr *ia;
1940
1941	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1942		return 1;
1943
1944	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1945		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1946		    &ia->ia_prefixmask.sin6_addr)) {
1947			return 1;
1948		}
1949	}
1950
1951	return (0);
1952}
1953
1954int
1955in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1956{
1957	INIT_VNET_INET6(curvnet);
1958	struct in6_ifaddr *ia;
1959
1960	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1961		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1962				       &sa6->sin6_addr) &&
1963		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1964			return (1); /* true */
1965
1966		/* XXX: do we still have to go thru the rest of the list? */
1967	}
1968
1969	return (0);		/* false */
1970}
1971
1972/*
1973 * return length of part which dst and src are equal
1974 * hard coding...
1975 */
1976int
1977in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1978{
1979	int match = 0;
1980	u_char *s = (u_char *)src, *d = (u_char *)dst;
1981	u_char *lim = s + 16, r;
1982
1983	while (s < lim)
1984		if ((r = (*d++ ^ *s++)) != 0) {
1985			while (r < 128) {
1986				match++;
1987				r <<= 1;
1988			}
1989			break;
1990		} else
1991			match += 8;
1992	return match;
1993}
1994
1995/* XXX: to be scope conscious */
1996int
1997in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1998{
1999	int bytelen, bitlen;
2000
2001	/* sanity check */
2002	if (0 > len || len > 128) {
2003		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2004		    len);
2005		return (0);
2006	}
2007
2008	bytelen = len / 8;
2009	bitlen = len % 8;
2010
2011	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2012		return (0);
2013	if (bitlen != 0 &&
2014	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2015	    p2->s6_addr[bytelen] >> (8 - bitlen))
2016		return (0);
2017
2018	return (1);
2019}
2020
2021void
2022in6_prefixlen2mask(struct in6_addr *maskp, int len)
2023{
2024	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2025	int bytelen, bitlen, i;
2026
2027	/* sanity check */
2028	if (0 > len || len > 128) {
2029		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2030		    len);
2031		return;
2032	}
2033
2034	bzero(maskp, sizeof(*maskp));
2035	bytelen = len / 8;
2036	bitlen = len % 8;
2037	for (i = 0; i < bytelen; i++)
2038		maskp->s6_addr[i] = 0xff;
2039	if (bitlen)
2040		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2041}
2042
2043/*
2044 * return the best address out of the same scope. if no address was
2045 * found, return the first valid address from designated IF.
2046 */
2047struct in6_ifaddr *
2048in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2049{
2050	INIT_VNET_INET6(curvnet);
2051	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2052	struct ifaddr *ifa;
2053	struct in6_ifaddr *besta = 0;
2054	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2055
2056	dep[0] = dep[1] = NULL;
2057
2058	/*
2059	 * We first look for addresses in the same scope.
2060	 * If there is one, return it.
2061	 * If two or more, return one which matches the dst longest.
2062	 * If none, return one of global addresses assigned other ifs.
2063	 */
2064	IF_ADDR_LOCK(ifp);
2065	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2066		if (ifa->ifa_addr->sa_family != AF_INET6)
2067			continue;
2068		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2069			continue; /* XXX: is there any case to allow anycast? */
2070		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2071			continue; /* don't use this interface */
2072		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2073			continue;
2074		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2075			if (V_ip6_use_deprecated)
2076				dep[0] = (struct in6_ifaddr *)ifa;
2077			continue;
2078		}
2079
2080		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2081			/*
2082			 * call in6_matchlen() as few as possible
2083			 */
2084			if (besta) {
2085				if (blen == -1)
2086					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2087				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2088				if (tlen > blen) {
2089					blen = tlen;
2090					besta = (struct in6_ifaddr *)ifa;
2091				}
2092			} else
2093				besta = (struct in6_ifaddr *)ifa;
2094		}
2095	}
2096	if (besta) {
2097		ifa_ref(&besta->ia_ifa);
2098		IF_ADDR_UNLOCK(ifp);
2099		return (besta);
2100	}
2101
2102	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2103		if (ifa->ifa_addr->sa_family != AF_INET6)
2104			continue;
2105		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2106			continue; /* XXX: is there any case to allow anycast? */
2107		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2108			continue; /* don't use this interface */
2109		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2110			continue;
2111		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2112			if (V_ip6_use_deprecated)
2113				dep[1] = (struct in6_ifaddr *)ifa;
2114			continue;
2115		}
2116
2117		if (ifa != NULL)
2118			ifa_ref(ifa);
2119		IF_ADDR_UNLOCK(ifp);
2120		return (struct in6_ifaddr *)ifa;
2121	}
2122	IF_ADDR_UNLOCK(ifp);
2123
2124	/* use the last-resort values, that are, deprecated addresses */
2125	if (dep[0])
2126		return dep[0];
2127	if (dep[1])
2128		return dep[1];
2129
2130	return NULL;
2131}
2132
2133/*
2134 * perform DAD when interface becomes IFF_UP.
2135 */
2136void
2137in6_if_up(struct ifnet *ifp)
2138{
2139	struct ifaddr *ifa;
2140	struct in6_ifaddr *ia;
2141
2142	IF_ADDR_LOCK(ifp);
2143	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2144		if (ifa->ifa_addr->sa_family != AF_INET6)
2145			continue;
2146		ia = (struct in6_ifaddr *)ifa;
2147		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2148			/*
2149			 * The TENTATIVE flag was likely set by hand
2150			 * beforehand, implicitly indicating the need for DAD.
2151			 * We may be able to skip the random delay in this
2152			 * case, but we impose delays just in case.
2153			 */
2154			nd6_dad_start(ifa,
2155			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2156		}
2157	}
2158	IF_ADDR_UNLOCK(ifp);
2159
2160	/*
2161	 * special cases, like 6to4, are handled in in6_ifattach
2162	 */
2163	in6_ifattach(ifp, NULL);
2164}
2165
2166int
2167in6if_do_dad(struct ifnet *ifp)
2168{
2169	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2170		return (0);
2171
2172	switch (ifp->if_type) {
2173#ifdef IFT_DUMMY
2174	case IFT_DUMMY:
2175#endif
2176	case IFT_FAITH:
2177		/*
2178		 * These interfaces do not have the IFF_LOOPBACK flag,
2179		 * but loop packets back.  We do not have to do DAD on such
2180		 * interfaces.  We should even omit it, because loop-backed
2181		 * NS would confuse the DAD procedure.
2182		 */
2183		return (0);
2184	default:
2185		/*
2186		 * Our DAD routine requires the interface up and running.
2187		 * However, some interfaces can be up before the RUNNING
2188		 * status.  Additionaly, users may try to assign addresses
2189		 * before the interface becomes up (or running).
2190		 * We simply skip DAD in such a case as a work around.
2191		 * XXX: we should rather mark "tentative" on such addresses,
2192		 * and do DAD after the interface becomes ready.
2193		 */
2194		if (!((ifp->if_flags & IFF_UP) &&
2195		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2196			return (0);
2197
2198		return (1);
2199	}
2200}
2201
2202/*
2203 * Calculate max IPv6 MTU through all the interfaces and store it
2204 * to in6_maxmtu.
2205 */
2206void
2207in6_setmaxmtu(void)
2208{
2209	INIT_VNET_NET(curvnet);
2210	INIT_VNET_INET6(curvnet);
2211	unsigned long maxmtu = 0;
2212	struct ifnet *ifp;
2213
2214	IFNET_RLOCK();
2215	for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
2216	    ifp = TAILQ_NEXT(ifp, if_list)) {
2217		/* this function can be called during ifnet initialization */
2218		if (!ifp->if_afdata[AF_INET6])
2219			continue;
2220		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2221		    IN6_LINKMTU(ifp) > maxmtu)
2222			maxmtu = IN6_LINKMTU(ifp);
2223	}
2224	IFNET_RUNLOCK();
2225	if (maxmtu)	     /* update only when maxmtu is positive */
2226		V_in6_maxmtu = maxmtu;
2227}
2228
2229/*
2230 * Provide the length of interface identifiers to be used for the link attached
2231 * to the given interface.  The length should be defined in "IPv6 over
2232 * xxx-link" document.  Note that address architecture might also define
2233 * the length for a particular set of address prefixes, regardless of the
2234 * link type.  As clarified in rfc2462bis, those two definitions should be
2235 * consistent, and those really are as of August 2004.
2236 */
2237int
2238in6_if2idlen(struct ifnet *ifp)
2239{
2240	switch (ifp->if_type) {
2241	case IFT_ETHER:		/* RFC2464 */
2242#ifdef IFT_PROPVIRTUAL
2243	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2244#endif
2245#ifdef IFT_L2VLAN
2246	case IFT_L2VLAN:	/* ditto */
2247#endif
2248#ifdef IFT_IEEE80211
2249	case IFT_IEEE80211:	/* ditto */
2250#endif
2251#ifdef IFT_MIP
2252	case IFT_MIP:	/* ditto */
2253#endif
2254		return (64);
2255	case IFT_FDDI:		/* RFC2467 */
2256		return (64);
2257	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2258		return (64);
2259	case IFT_PPP:		/* RFC2472 */
2260		return (64);
2261	case IFT_ARCNET:	/* RFC2497 */
2262		return (64);
2263	case IFT_FRELAY:	/* RFC2590 */
2264		return (64);
2265	case IFT_IEEE1394:	/* RFC3146 */
2266		return (64);
2267	case IFT_GIF:
2268		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2269	case IFT_LOOP:
2270		return (64);	/* XXX: is this really correct? */
2271	default:
2272		/*
2273		 * Unknown link type:
2274		 * It might be controversial to use the today's common constant
2275		 * of 64 for these cases unconditionally.  For full compliance,
2276		 * we should return an error in this case.  On the other hand,
2277		 * if we simply miss the standard for the link type or a new
2278		 * standard is defined for a new link type, the IFID length
2279		 * is very likely to be the common constant.  As a compromise,
2280		 * we always use the constant, but make an explicit notice
2281		 * indicating the "unknown" case.
2282		 */
2283		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2284		return (64);
2285	}
2286}
2287
2288#include <sys/sysctl.h>
2289
2290struct in6_llentry {
2291	struct llentry		base;
2292	struct sockaddr_in6	l3_addr6;
2293};
2294
2295static struct llentry *
2296in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2297{
2298	struct in6_llentry *lle;
2299
2300	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE,
2301	    M_DONTWAIT | M_ZERO);
2302	if (lle == NULL)		/* NB: caller generates msg */
2303		return NULL;
2304
2305	callout_init(&lle->base.ln_timer_ch, CALLOUT_MPSAFE);
2306	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2307	lle->base.lle_refcnt = 1;
2308	LLE_LOCK_INIT(&lle->base);
2309	return &lle->base;
2310}
2311
2312/*
2313 * Deletes an address from the address table.
2314 * This function is called by the timer functions
2315 * such as arptimer() and nd6_llinfo_timer(), and
2316 * the caller does the locking.
2317 */
2318static void
2319in6_lltable_free(struct lltable *llt, struct llentry *lle)
2320{
2321	LLE_WUNLOCK(lle);
2322	LLE_LOCK_DESTROY(lle);
2323	free(lle, M_LLTABLE);
2324}
2325
2326static void
2327in6_lltable_prefix_free(struct lltable *llt,
2328			const struct sockaddr *prefix,
2329			const struct sockaddr *mask)
2330{
2331	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2332	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2333	struct llentry *lle, *next;
2334	register int i;
2335
2336	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
2337		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2338			if (IN6_ARE_MASKED_ADDR_EQUAL(
2339				    &((struct sockaddr_in6 *)L3_ADDR(lle))->sin6_addr,
2340				    &pfx->sin6_addr,
2341				    &msk->sin6_addr)) {
2342				callout_drain(&lle->la_timer);
2343				LLE_WLOCK(lle);
2344				llentry_free(lle);
2345			}
2346		}
2347	}
2348}
2349
2350static int
2351in6_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr)
2352{
2353	struct rtentry *rt;
2354	char ip6buf[INET6_ADDRSTRLEN];
2355
2356	KASSERT(l3addr->sa_family == AF_INET6,
2357	    ("sin_family %d", l3addr->sa_family));
2358
2359	/* XXX rtalloc1 should take a const param */
2360	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
2361	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2362		struct ifaddr *ifa;
2363		/*
2364		 * Create an ND6 cache for an IPv6 neighbor
2365		 * that is not covered by our own prefix.
2366		 */
2367		/* XXX ifaof_ifpforaddr should take a const param */
2368		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2369		if (ifa != NULL) {
2370			ifa_free(ifa);
2371			if (rt != NULL)
2372				RTFREE_LOCKED(rt);
2373			return 0;
2374		}
2375		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2376		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2377		if (rt != NULL)
2378			RTFREE_LOCKED(rt);
2379		return EINVAL;
2380	}
2381	RTFREE_LOCKED(rt);
2382	return 0;
2383}
2384
2385static struct llentry *
2386in6_lltable_lookup(struct lltable *llt, u_int flags,
2387	const struct sockaddr *l3addr)
2388{
2389	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2390	struct ifnet *ifp = llt->llt_ifp;
2391	struct llentry *lle;
2392	struct llentries *lleh;
2393	u_int hashkey;
2394
2395	IF_AFDATA_LOCK_ASSERT(ifp);
2396	KASSERT(l3addr->sa_family == AF_INET6,
2397	    ("sin_family %d", l3addr->sa_family));
2398
2399	hashkey = sin6->sin6_addr.s6_addr32[3];
2400	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2401	LIST_FOREACH(lle, lleh, lle_next) {
2402		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2403		if (lle->la_flags & LLE_DELETED)
2404			continue;
2405		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2406			 sizeof(struct in6_addr)) == 0)
2407			break;
2408	}
2409
2410	if (lle == NULL) {
2411		if (!(flags & LLE_CREATE))
2412			return (NULL);
2413		/*
2414		 * A route that covers the given address must have
2415		 * been installed 1st because we are doing a resolution,
2416		 * verify this.
2417		 */
2418		if (!(flags & LLE_IFADDR) &&
2419		    in6_lltable_rtcheck(ifp, l3addr) != 0)
2420			return NULL;
2421
2422		lle = in6_lltable_new(l3addr, flags);
2423		if (lle == NULL) {
2424			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2425			return NULL;
2426		}
2427		lle->la_flags = flags & ~LLE_CREATE;
2428		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2429			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2430			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2431		}
2432
2433		lle->lle_tbl  = llt;
2434		lle->lle_head = lleh;
2435		LIST_INSERT_HEAD(lleh, lle, lle_next);
2436	} else if (flags & LLE_DELETE) {
2437		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2438			LLE_WLOCK(lle);
2439			lle->la_flags = LLE_DELETED;
2440			LLE_WUNLOCK(lle);
2441#ifdef DIAGNOSTICS
2442			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2443#endif
2444		}
2445		lle = (void *)-1;
2446	}
2447	if (LLE_IS_VALID(lle)) {
2448		if (flags & LLE_EXCLUSIVE)
2449			LLE_WLOCK(lle);
2450		else
2451			LLE_RLOCK(lle);
2452	}
2453	return (lle);
2454}
2455
2456static int
2457in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2458{
2459	struct ifnet *ifp = llt->llt_ifp;
2460	struct llentry *lle;
2461	/* XXX stack use */
2462	struct {
2463		struct rt_msghdr	rtm;
2464		struct sockaddr_in6	sin6;
2465		/*
2466		 * ndp.c assumes that sdl is word aligned
2467		 */
2468#ifdef __LP64__
2469		uint32_t		pad;
2470#endif
2471		struct sockaddr_dl	sdl;
2472	} ndpc;
2473	int i, error;
2474
2475	/* XXXXX
2476	 * current IFNET_RLOCK() is mapped to IFNET_WLOCK()
2477	 * so it is okay to use this ASSERT, change it when
2478	 * IFNET lock is finalized
2479	 */
2480	IFNET_WLOCK_ASSERT();
2481
2482	error = 0;
2483	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2484		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2485			struct sockaddr_dl *sdl;
2486
2487			/* skip deleted or invalid entries */
2488			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2489				continue;
2490			/* Skip if jailed and not a valid IP of the prison. */
2491			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2492				continue;
2493			/*
2494			 * produce a msg made of:
2495			 *  struct rt_msghdr;
2496			 *  struct sockaddr_in6 (IPv6)
2497			 *  struct sockaddr_dl;
2498			 */
2499			bzero(&ndpc, sizeof(ndpc));
2500			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2501			ndpc.rtm.rtm_version = RTM_VERSION;
2502			ndpc.rtm.rtm_type = RTM_GET;
2503			ndpc.rtm.rtm_flags = RTF_UP;
2504			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2505			ndpc.sin6.sin6_family = AF_INET6;
2506			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2507			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2508
2509			/* publish */
2510			if (lle->la_flags & LLE_PUB)
2511				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2512
2513			sdl = &ndpc.sdl;
2514			sdl->sdl_family = AF_LINK;
2515			sdl->sdl_len = sizeof(*sdl);
2516			sdl->sdl_alen = ifp->if_addrlen;
2517			sdl->sdl_index = ifp->if_index;
2518			sdl->sdl_type = ifp->if_type;
2519			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2520			ndpc.rtm.rtm_rmx.rmx_expire =
2521			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2522			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2523			if (lle->la_flags & LLE_STATIC)
2524				ndpc.rtm.rtm_flags |= RTF_STATIC;
2525			ndpc.rtm.rtm_index = ifp->if_index;
2526			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2527			if (error)
2528				break;
2529		}
2530	}
2531	return error;
2532}
2533
2534void *
2535in6_domifattach(struct ifnet *ifp)
2536{
2537	struct in6_ifextra *ext;
2538
2539	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2540	bzero(ext, sizeof(*ext));
2541
2542	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2543	    M_IFADDR, M_WAITOK);
2544	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2545
2546	ext->icmp6_ifstat =
2547	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2548	    M_IFADDR, M_WAITOK);
2549	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2550
2551	ext->nd_ifinfo = nd6_ifattach(ifp);
2552	ext->scope6_id = scope6_ifattach(ifp);
2553	ext->lltable = lltable_init(ifp, AF_INET6);
2554	if (ext->lltable != NULL) {
2555		ext->lltable->llt_new = in6_lltable_new;
2556		ext->lltable->llt_free = in6_lltable_free;
2557		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2558		ext->lltable->llt_rtcheck = in6_lltable_rtcheck;
2559		ext->lltable->llt_lookup = in6_lltable_lookup;
2560		ext->lltable->llt_dump = in6_lltable_dump;
2561	}
2562
2563	ext->mld_ifinfo = mld_domifattach(ifp);
2564
2565	return ext;
2566}
2567
2568void
2569in6_domifdetach(struct ifnet *ifp, void *aux)
2570{
2571	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2572
2573	mld_domifdetach(ifp);
2574	scope6_ifdetach(ext->scope6_id);
2575	nd6_ifdetach(ext->nd_ifinfo);
2576	lltable_free(ext->lltable);
2577	free(ext->in6_ifstat, M_IFADDR);
2578	free(ext->icmp6_ifstat, M_IFADDR);
2579	free(ext, M_IFADDR);
2580}
2581
2582/*
2583 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2584 * v4 mapped addr or v4 compat addr
2585 */
2586void
2587in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2588{
2589
2590	bzero(sin, sizeof(*sin));
2591	sin->sin_len = sizeof(struct sockaddr_in);
2592	sin->sin_family = AF_INET;
2593	sin->sin_port = sin6->sin6_port;
2594	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2595}
2596
2597/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2598void
2599in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2600{
2601	bzero(sin6, sizeof(*sin6));
2602	sin6->sin6_len = sizeof(struct sockaddr_in6);
2603	sin6->sin6_family = AF_INET6;
2604	sin6->sin6_port = sin->sin_port;
2605	sin6->sin6_addr.s6_addr32[0] = 0;
2606	sin6->sin6_addr.s6_addr32[1] = 0;
2607	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2608	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2609}
2610
2611/* Convert sockaddr_in6 into sockaddr_in. */
2612void
2613in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2614{
2615	struct sockaddr_in *sin_p;
2616	struct sockaddr_in6 sin6;
2617
2618	/*
2619	 * Save original sockaddr_in6 addr and convert it
2620	 * to sockaddr_in.
2621	 */
2622	sin6 = *(struct sockaddr_in6 *)nam;
2623	sin_p = (struct sockaddr_in *)nam;
2624	in6_sin6_2_sin(sin_p, &sin6);
2625}
2626
2627/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2628void
2629in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2630{
2631	struct sockaddr_in *sin_p;
2632	struct sockaddr_in6 *sin6_p;
2633
2634	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
2635	       M_WAITOK);
2636	sin_p = (struct sockaddr_in *)*nam;
2637	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2638	free(*nam, M_SONAME);
2639	*nam = (struct sockaddr *)sin6_p;
2640}
2641