in6.c revision 751dff36104851edee18dc55bd333b085a912d68
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/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
108MALLOC_DEFINE(M_IP6MADDR, "in6_multi", "internet multicast address");
109
110/*
111 * Definitions of some costant IP6 addresses.
112 */
113const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
114const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
115const struct in6_addr in6addr_nodelocal_allnodes =
116	IN6ADDR_NODELOCAL_ALLNODES_INIT;
117const struct in6_addr in6addr_linklocal_allnodes =
118	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
119const struct in6_addr in6addr_linklocal_allrouters =
120	IN6ADDR_LINKLOCAL_ALLROUTERS_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
137struct in6_multihead in6_multihead;	/* XXX BSS initialization */
138int	(*faithprefix_p)(struct in6_addr *);
139
140
141
142int
143in6_mask2len(struct in6_addr *mask, u_char *lim0)
144{
145	int x = 0, y;
146	u_char *lim = lim0, *p;
147
148	/* ignore the scope_id part */
149	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
150		lim = (u_char *)mask + sizeof(*mask);
151	for (p = (u_char *)mask; p < lim; x++, p++) {
152		if (*p != 0xff)
153			break;
154	}
155	y = 0;
156	if (p < lim) {
157		for (y = 0; y < 8; y++) {
158			if ((*p & (0x80 >> y)) == 0)
159				break;
160		}
161	}
162
163	/*
164	 * when the limit pointer is given, do a stricter check on the
165	 * remaining bits.
166	 */
167	if (p < lim) {
168		if (y != 0 && (*p & (0x00ff >> y)) != 0)
169			return (-1);
170		for (p = p + 1; p < lim; p++)
171			if (*p != 0)
172				return (-1);
173	}
174
175	return x * 8 + y;
176}
177
178#define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
179#define ia62ifa(ia6)	(&((ia6)->ia_ifa))
180
181int
182in6_control(struct socket *so, u_long cmd, caddr_t data,
183    struct ifnet *ifp, struct thread *td)
184{
185	INIT_VNET_INET6(curvnet);
186	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
187	struct	in6_ifaddr *ia = NULL;
188	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
189	struct sockaddr_in6 *sa6;
190	int error;
191
192	switch (cmd) {
193	case SIOCGETSGCNT_IN6:
194	case SIOCGETMIFCNT_IN6:
195		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
196	}
197
198	switch(cmd) {
199	case SIOCAADDRCTL_POLICY:
200	case SIOCDADDRCTL_POLICY:
201		if (td != NULL) {
202			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
203			if (error)
204				return (error);
205		}
206		return (in6_src_ioctl(cmd, data));
207	}
208
209	if (ifp == NULL)
210		return (EOPNOTSUPP);
211
212	switch (cmd) {
213	case SIOCSNDFLUSH_IN6:
214	case SIOCSPFXFLUSH_IN6:
215	case SIOCSRTRFLUSH_IN6:
216	case SIOCSDEFIFACE_IN6:
217	case SIOCSIFINFO_FLAGS:
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 SIOCSIFINFO_IN6:
227	case SIOCGDRLST_IN6:
228	case SIOCGPRLST_IN6:
229	case SIOCGNBRINFO_IN6:
230	case SIOCGDEFIFACE_IN6:
231		return (nd6_ioctl(cmd, data, ifp));
232	}
233
234	switch (cmd) {
235	case SIOCSIFPREFIX_IN6:
236	case SIOCDIFPREFIX_IN6:
237	case SIOCAIFPREFIX_IN6:
238	case SIOCCIFPREFIX_IN6:
239	case SIOCSGIFPREFIX_IN6:
240	case SIOCGIFPREFIX_IN6:
241		log(LOG_NOTICE,
242		    "prefix ioctls are now invalidated. "
243		    "please use ifconfig.\n");
244		return (EOPNOTSUPP);
245	}
246
247	switch (cmd) {
248	case SIOCSSCOPE6:
249		if (td != NULL) {
250			error = priv_check(td, PRIV_NETINET_SCOPE6);
251			if (error)
252				return (error);
253		}
254		return (scope6_set(ifp,
255		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
256	case SIOCGSCOPE6:
257		return (scope6_get(ifp,
258		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
259	case SIOCGSCOPE6DEF:
260		return (scope6_get_default((struct scope6_id *)
261		    ifr->ifr_ifru.ifru_scope_id));
262	}
263
264	switch (cmd) {
265	case SIOCALIFADDR:
266		if (td != NULL) {
267			error = priv_check(td, PRIV_NET_ADDIFADDR);
268			if (error)
269				return (error);
270		}
271		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
272
273	case SIOCDLIFADDR:
274		if (td != NULL) {
275			error = priv_check(td, PRIV_NET_DELIFADDR);
276			if (error)
277				return (error);
278		}
279		/* FALLTHROUGH */
280	case SIOCGLIFADDR:
281		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
282	}
283
284	/*
285	 * Find address for this interface, if it exists.
286	 *
287	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
288	 * only, and used the first interface address as the target of other
289	 * operations (without checking ifra_addr).  This was because netinet
290	 * code/API assumed at most 1 interface address per interface.
291	 * Since IPv6 allows a node to assign multiple addresses
292	 * on a single interface, we almost always look and check the
293	 * presence of ifra_addr, and reject invalid ones here.
294	 * It also decreases duplicated code among SIOC*_IN6 operations.
295	 */
296	switch (cmd) {
297	case SIOCAIFADDR_IN6:
298	case SIOCSIFPHYADDR_IN6:
299		sa6 = &ifra->ifra_addr;
300		break;
301	case SIOCSIFADDR_IN6:
302	case SIOCGIFADDR_IN6:
303	case SIOCSIFDSTADDR_IN6:
304	case SIOCSIFNETMASK_IN6:
305	case SIOCGIFDSTADDR_IN6:
306	case SIOCGIFNETMASK_IN6:
307	case SIOCDIFADDR_IN6:
308	case SIOCGIFPSRCADDR_IN6:
309	case SIOCGIFPDSTADDR_IN6:
310	case SIOCGIFAFLAG_IN6:
311	case SIOCSNDFLUSH_IN6:
312	case SIOCSPFXFLUSH_IN6:
313	case SIOCSRTRFLUSH_IN6:
314	case SIOCGIFALIFETIME_IN6:
315	case SIOCSIFALIFETIME_IN6:
316	case SIOCGIFSTAT_IN6:
317	case SIOCGIFSTAT_ICMP6:
318		sa6 = &ifr->ifr_addr;
319		break;
320	default:
321		sa6 = NULL;
322		break;
323	}
324	if (sa6 && sa6->sin6_family == AF_INET6) {
325		int error = 0;
326
327		if (sa6->sin6_scope_id != 0)
328			error = sa6_embedscope(sa6, 0);
329		else
330			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
331		if (error != 0)
332			return (error);
333		if (td != NULL && !prison_check_ip6(td->td_ucred,
334		    &sa6->sin6_addr))
335			return (EADDRNOTAVAIL);
336		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
337	} else
338		ia = NULL;
339
340	switch (cmd) {
341	case SIOCSIFADDR_IN6:
342	case SIOCSIFDSTADDR_IN6:
343	case SIOCSIFNETMASK_IN6:
344		/*
345		 * Since IPv6 allows a node to assign multiple addresses
346		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
347		 */
348		/* we decided to obsolete this command (20000704) */
349		return (EINVAL);
350
351	case SIOCDIFADDR_IN6:
352		/*
353		 * for IPv4, we look for existing in_ifaddr here to allow
354		 * "ifconfig if0 delete" to remove the first IPv4 address on
355		 * the interface.  For IPv6, as the spec allows multiple
356		 * interface address from the day one, we consider "remove the
357		 * first one" semantics to be not preferable.
358		 */
359		if (ia == NULL)
360			return (EADDRNOTAVAIL);
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			return (EAFNOSUPPORT);
370
371		if (td != NULL) {
372			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
373			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
374			if (error)
375				return (error);
376		}
377
378		break;
379
380	case SIOCGIFADDR_IN6:
381		/* This interface is basically deprecated. use SIOCGIFCONF. */
382		/* FALLTHROUGH */
383	case SIOCGIFAFLAG_IN6:
384	case SIOCGIFNETMASK_IN6:
385	case SIOCGIFDSTADDR_IN6:
386	case SIOCGIFALIFETIME_IN6:
387		/* must think again about its semantics */
388		if (ia == NULL)
389			return (EADDRNOTAVAIL);
390		break;
391	case SIOCSIFALIFETIME_IN6:
392	    {
393		struct in6_addrlifetime *lt;
394
395		if (td != NULL) {
396			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
397			if (error)
398				return (error);
399		}
400		if (ia == NULL)
401			return (EADDRNOTAVAIL);
402		/* sanity for overflow - beware unsigned */
403		lt = &ifr->ifr_ifru.ifru_lifetime;
404		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
405		    lt->ia6t_vltime + time_second < time_second) {
406			return EINVAL;
407		}
408		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
409		    lt->ia6t_pltime + time_second < time_second) {
410			return EINVAL;
411		}
412		break;
413	    }
414	}
415
416	switch (cmd) {
417
418	case SIOCGIFADDR_IN6:
419		ifr->ifr_addr = ia->ia_addr;
420		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
421			return (error);
422		break;
423
424	case SIOCGIFDSTADDR_IN6:
425		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
426			return (EINVAL);
427		/*
428		 * XXX: should we check if ifa_dstaddr is NULL and return
429		 * an error?
430		 */
431		ifr->ifr_dstaddr = ia->ia_dstaddr;
432		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
433			return (error);
434		break;
435
436	case SIOCGIFNETMASK_IN6:
437		ifr->ifr_addr = ia->ia_prefixmask;
438		break;
439
440	case SIOCGIFAFLAG_IN6:
441		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
442		break;
443
444	case SIOCGIFSTAT_IN6:
445		if (ifp == NULL)
446			return EINVAL;
447		bzero(&ifr->ifr_ifru.ifru_stat,
448		    sizeof(ifr->ifr_ifru.ifru_stat));
449		ifr->ifr_ifru.ifru_stat =
450		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
451		break;
452
453	case SIOCGIFSTAT_ICMP6:
454		if (ifp == NULL)
455			return EINVAL;
456		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
457		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
458		ifr->ifr_ifru.ifru_icmp6stat =
459		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
460		break;
461
462	case SIOCGIFALIFETIME_IN6:
463		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
464		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
465			time_t maxexpire;
466			struct in6_addrlifetime *retlt =
467			    &ifr->ifr_ifru.ifru_lifetime;
468
469			/*
470			 * XXX: adjust expiration time assuming time_t is
471			 * signed.
472			 */
473			maxexpire = (-1) &
474			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
475			if (ia->ia6_lifetime.ia6t_vltime <
476			    maxexpire - ia->ia6_updatetime) {
477				retlt->ia6t_expire = ia->ia6_updatetime +
478				    ia->ia6_lifetime.ia6t_vltime;
479			} else
480				retlt->ia6t_expire = maxexpire;
481		}
482		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
483			time_t maxexpire;
484			struct in6_addrlifetime *retlt =
485			    &ifr->ifr_ifru.ifru_lifetime;
486
487			/*
488			 * XXX: adjust expiration time assuming time_t is
489			 * signed.
490			 */
491			maxexpire = (-1) &
492			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
493			if (ia->ia6_lifetime.ia6t_pltime <
494			    maxexpire - ia->ia6_updatetime) {
495				retlt->ia6t_preferred = ia->ia6_updatetime +
496				    ia->ia6_lifetime.ia6t_pltime;
497			} else
498				retlt->ia6t_preferred = maxexpire;
499		}
500		break;
501
502	case SIOCSIFALIFETIME_IN6:
503		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
504		/* for sanity */
505		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
506			ia->ia6_lifetime.ia6t_expire =
507				time_second + ia->ia6_lifetime.ia6t_vltime;
508		} else
509			ia->ia6_lifetime.ia6t_expire = 0;
510		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
511			ia->ia6_lifetime.ia6t_preferred =
512				time_second + ia->ia6_lifetime.ia6t_pltime;
513		} else
514			ia->ia6_lifetime.ia6t_preferred = 0;
515		break;
516
517	case SIOCAIFADDR_IN6:
518	{
519		int i, error = 0;
520		struct nd_prefixctl pr0;
521		struct nd_prefix *pr;
522
523		/*
524		 * first, make or update the interface address structure,
525		 * and link it to the list.
526		 */
527		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
528			return (error);
529		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
530		    == NULL) {
531			/*
532			 * this can happen when the user specify the 0 valid
533			 * lifetime.
534			 */
535			break;
536		}
537
538		/*
539		 * then, make the prefix on-link on the interface.
540		 * XXX: we'd rather create the prefix before the address, but
541		 * we need at least one address to install the corresponding
542		 * interface route, so we configure the address first.
543		 */
544
545		/*
546		 * convert mask to prefix length (prefixmask has already
547		 * been validated in in6_update_ifa().
548		 */
549		bzero(&pr0, sizeof(pr0));
550		pr0.ndpr_ifp = ifp;
551		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
552		    NULL);
553		if (pr0.ndpr_plen == 128) {
554			break;	/* we don't need to install a host route. */
555		}
556		pr0.ndpr_prefix = ifra->ifra_addr;
557		/* apply the mask for safety. */
558		for (i = 0; i < 4; i++) {
559			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
560			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
561		}
562		/*
563		 * XXX: since we don't have an API to set prefix (not address)
564		 * lifetimes, we just use the same lifetimes as addresses.
565		 * The (temporarily) installed lifetimes can be overridden by
566		 * later advertised RAs (when accept_rtadv is non 0), which is
567		 * an intended behavior.
568		 */
569		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
570		pr0.ndpr_raf_auto =
571		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
572		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
573		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
574
575		/* add the prefix if not yet. */
576		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
577			/*
578			 * nd6_prelist_add will install the corresponding
579			 * interface route.
580			 */
581			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
582				return (error);
583			if (pr == NULL) {
584				log(LOG_ERR, "nd6_prelist_add succeeded but "
585				    "no prefix\n");
586				return (EINVAL); /* XXX panic here? */
587			}
588		}
589
590		/* relate the address to the prefix */
591		if (ia->ia6_ndpr == NULL) {
592			ia->ia6_ndpr = pr;
593			pr->ndpr_refcnt++;
594
595			/*
596			 * If this is the first autoconf address from the
597			 * prefix, create a temporary address as well
598			 * (when required).
599			 */
600			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
601			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
602				int e;
603				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
604					log(LOG_NOTICE, "in6_control: failed "
605					    "to create a temporary address, "
606					    "errno=%d\n", e);
607				}
608			}
609		}
610
611		/*
612		 * this might affect the status of autoconfigured addresses,
613		 * that is, this address might make other addresses detached.
614		 */
615		pfxlist_onlink_check();
616		if (error == 0 && ia)
617			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
618		break;
619	}
620
621	case SIOCDIFADDR_IN6:
622	{
623		struct nd_prefix *pr;
624
625		/*
626		 * If the address being deleted is the only one that owns
627		 * the corresponding prefix, expire the prefix as well.
628		 * XXX: theoretically, we don't have to worry about such
629		 * relationship, since we separate the address management
630		 * and the prefix management.  We do this, however, to provide
631		 * as much backward compatibility as possible in terms of
632		 * the ioctl operation.
633		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
634		 */
635		pr = ia->ia6_ndpr;
636		in6_purgeaddr(&ia->ia_ifa);
637		if (pr && pr->ndpr_refcnt == 0)
638			prelist_remove(pr);
639		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
640		break;
641	}
642
643	default:
644		if (ifp == NULL || ifp->if_ioctl == 0)
645			return (EOPNOTSUPP);
646		return ((*ifp->if_ioctl)(ifp, cmd, data));
647	}
648
649	return (0);
650}
651
652/*
653 * Update parameters of an IPv6 interface address.
654 * If necessary, a new entry is created and linked into address chains.
655 * This function is separated from in6_control().
656 * XXX: should this be performed under splnet()?
657 */
658int
659in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
660    struct in6_ifaddr *ia, int flags)
661{
662	INIT_VNET_INET6(ifp->if_vnet);
663	INIT_VPROCG(TD_TO_VPROCG(curthread)); /* XXX V_hostname needs this */
664	int error = 0, hostIsNew = 0, plen = -1;
665	struct in6_ifaddr *oia;
666	struct sockaddr_in6 dst6;
667	struct in6_addrlifetime *lt;
668	struct in6_multi_mship *imm;
669	struct in6_multi *in6m_sol;
670	struct rtentry *rt;
671	int delay;
672	char ip6buf[INET6_ADDRSTRLEN];
673
674	/* Validate parameters */
675	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
676		return (EINVAL);
677
678	/*
679	 * The destination address for a p2p link must have a family
680	 * of AF_UNSPEC or AF_INET6.
681	 */
682	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
683	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
684	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
685		return (EAFNOSUPPORT);
686	/*
687	 * validate ifra_prefixmask.  don't check sin6_family, netmask
688	 * does not carry fields other than sin6_len.
689	 */
690	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
691		return (EINVAL);
692	/*
693	 * Because the IPv6 address architecture is classless, we require
694	 * users to specify a (non 0) prefix length (mask) for a new address.
695	 * We also require the prefix (when specified) mask is valid, and thus
696	 * reject a non-consecutive mask.
697	 */
698	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
699		return (EINVAL);
700	if (ifra->ifra_prefixmask.sin6_len != 0) {
701		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
702		    (u_char *)&ifra->ifra_prefixmask +
703		    ifra->ifra_prefixmask.sin6_len);
704		if (plen <= 0)
705			return (EINVAL);
706	} else {
707		/*
708		 * In this case, ia must not be NULL.  We just use its prefix
709		 * length.
710		 */
711		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
712	}
713	/*
714	 * If the destination address on a p2p interface is specified,
715	 * and the address is a scoped one, validate/set the scope
716	 * zone identifier.
717	 */
718	dst6 = ifra->ifra_dstaddr;
719	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
720	    (dst6.sin6_family == AF_INET6)) {
721		struct in6_addr in6_tmp;
722		u_int32_t zoneid;
723
724		in6_tmp = dst6.sin6_addr;
725		if (in6_setscope(&in6_tmp, ifp, &zoneid))
726			return (EINVAL); /* XXX: should be impossible */
727
728		if (dst6.sin6_scope_id != 0) {
729			if (dst6.sin6_scope_id != zoneid)
730				return (EINVAL);
731		} else		/* user omit to specify the ID. */
732			dst6.sin6_scope_id = zoneid;
733
734		/* convert into the internal form */
735		if (sa6_embedscope(&dst6, 0))
736			return (EINVAL); /* XXX: should be impossible */
737	}
738	/*
739	 * The destination address can be specified only for a p2p or a
740	 * loopback interface.  If specified, the corresponding prefix length
741	 * must be 128.
742	 */
743	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
744		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
745			/* XXX: noisy message */
746			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
747			    "be specified for a p2p or a loopback IF only\n"));
748			return (EINVAL);
749		}
750		if (plen != 128) {
751			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
752			    "be 128 when dstaddr is specified\n"));
753			return (EINVAL);
754		}
755	}
756	/* lifetime consistency check */
757	lt = &ifra->ifra_lifetime;
758	if (lt->ia6t_pltime > lt->ia6t_vltime)
759		return (EINVAL);
760	if (lt->ia6t_vltime == 0) {
761		/*
762		 * the following log might be noisy, but this is a typical
763		 * configuration mistake or a tool's bug.
764		 */
765		nd6log((LOG_INFO,
766		    "in6_update_ifa: valid lifetime is 0 for %s\n",
767		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
768
769		if (ia == NULL)
770			return (0); /* there's nothing to do */
771	}
772
773	/*
774	 * If this is a new address, allocate a new ifaddr and link it
775	 * into chains.
776	 */
777	if (ia == NULL) {
778		hostIsNew = 1;
779		/*
780		 * When in6_update_ifa() is called in a process of a received
781		 * RA, it is called under an interrupt context.  So, we should
782		 * call malloc with M_NOWAIT.
783		 */
784		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
785		    M_NOWAIT);
786		if (ia == NULL)
787			return (ENOBUFS);
788		bzero((caddr_t)ia, sizeof(*ia));
789		LIST_INIT(&ia->ia6_memberships);
790		/* Initialize the address and masks, and put time stamp */
791		IFA_LOCK_INIT(&ia->ia_ifa);
792		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
793		ia->ia_addr.sin6_family = AF_INET6;
794		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
795		ia->ia6_createtime = time_second;
796		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
797			/*
798			 * XXX: some functions expect that ifa_dstaddr is not
799			 * NULL for p2p interfaces.
800			 */
801			ia->ia_ifa.ifa_dstaddr =
802			    (struct sockaddr *)&ia->ia_dstaddr;
803		} else {
804			ia->ia_ifa.ifa_dstaddr = NULL;
805		}
806		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
807
808		ia->ia_ifp = ifp;
809		if ((oia = V_in6_ifaddr) != NULL) {
810			for ( ; oia->ia_next; oia = oia->ia_next)
811				continue;
812			oia->ia_next = ia;
813		} else
814			V_in6_ifaddr = ia;
815
816		ia->ia_ifa.ifa_refcnt = 1;
817		TAILQ_INSERT_TAIL(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
818	}
819
820	/* update timestamp */
821	ia->ia6_updatetime = time_second;
822
823	/* set prefix mask */
824	if (ifra->ifra_prefixmask.sin6_len) {
825		/*
826		 * We prohibit changing the prefix length of an existing
827		 * address, because
828		 * + such an operation should be rare in IPv6, and
829		 * + the operation would confuse prefix management.
830		 */
831		if (ia->ia_prefixmask.sin6_len &&
832		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
833			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
834			    " existing (%s) address should not be changed\n",
835			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
836			error = EINVAL;
837			goto unlink;
838		}
839		ia->ia_prefixmask = ifra->ifra_prefixmask;
840	}
841
842	/*
843	 * If a new destination address is specified, scrub the old one and
844	 * install the new destination.  Note that the interface must be
845	 * p2p or loopback (see the check above.)
846	 */
847	if (dst6.sin6_family == AF_INET6 &&
848	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
849		int e;
850
851		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
852		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
853			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
854			    "a route to the old destination: %s\n",
855			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
856			/* proceed anyway... */
857		} else
858			ia->ia_flags &= ~IFA_ROUTE;
859		ia->ia_dstaddr = dst6;
860	}
861
862	/*
863	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
864	 * to see if the address is deprecated or invalidated, but initialize
865	 * these members for applications.
866	 */
867	ia->ia6_lifetime = ifra->ifra_lifetime;
868	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
869		ia->ia6_lifetime.ia6t_expire =
870		    time_second + ia->ia6_lifetime.ia6t_vltime;
871	} else
872		ia->ia6_lifetime.ia6t_expire = 0;
873	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
874		ia->ia6_lifetime.ia6t_preferred =
875		    time_second + ia->ia6_lifetime.ia6t_pltime;
876	} else
877		ia->ia6_lifetime.ia6t_preferred = 0;
878
879	/* reset the interface and routing table appropriately. */
880	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
881		goto unlink;
882
883	/*
884	 * configure address flags.
885	 */
886	ia->ia6_flags = ifra->ifra_flags;
887	/*
888	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
889	 * userland, make it deprecated.
890	 */
891	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
892		ia->ia6_lifetime.ia6t_pltime = 0;
893		ia->ia6_lifetime.ia6t_preferred = time_second;
894	}
895	/*
896	 * Make the address tentative before joining multicast addresses,
897	 * so that corresponding MLD responses would not have a tentative
898	 * source address.
899	 */
900	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
901	if (hostIsNew && in6if_do_dad(ifp))
902		ia->ia6_flags |= IN6_IFF_TENTATIVE;
903
904	/*
905	 * We are done if we have simply modified an existing address.
906	 */
907	if (!hostIsNew)
908		return (error);
909
910	/*
911	 * Beyond this point, we should call in6_purgeaddr upon an error,
912	 * not just go to unlink.
913	 */
914
915	/* Join necessary multicast groups */
916	in6m_sol = NULL;
917	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
918		struct sockaddr_in6 mltaddr, mltmask;
919		struct in6_addr llsol;
920
921		/* join solicited multicast addr for new host id */
922		bzero(&llsol, sizeof(struct in6_addr));
923		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
924		llsol.s6_addr32[1] = 0;
925		llsol.s6_addr32[2] = htonl(1);
926		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
927		llsol.s6_addr8[12] = 0xff;
928		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
929			/* XXX: should not happen */
930			log(LOG_ERR, "in6_update_ifa: "
931			    "in6_setscope failed\n");
932			goto cleanup;
933		}
934		delay = 0;
935		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
936			/*
937			 * We need a random delay for DAD on the address
938			 * being configured.  It also means delaying
939			 * transmission of the corresponding MLD report to
940			 * avoid report collision.
941			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
942			 */
943			delay = arc4random() %
944			    (MAX_RTR_SOLICITATION_DELAY * hz);
945		}
946		imm = in6_joingroup(ifp, &llsol, &error, delay);
947		if (imm == NULL) {
948			nd6log((LOG_WARNING,
949			    "in6_update_ifa: addmulti failed for "
950			    "%s on %s (errno=%d)\n",
951			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
952			    error));
953			in6_purgeaddr((struct ifaddr *)ia);
954			return (error);
955		}
956		LIST_INSERT_HEAD(&ia->ia6_memberships,
957		    imm, i6mm_chain);
958		in6m_sol = imm->i6mm_maddr;
959
960		bzero(&mltmask, sizeof(mltmask));
961		mltmask.sin6_len = sizeof(struct sockaddr_in6);
962		mltmask.sin6_family = AF_INET6;
963		mltmask.sin6_addr = in6mask32;
964#define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
965
966		/*
967		 * join link-local all-nodes address
968		 */
969		bzero(&mltaddr, sizeof(mltaddr));
970		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
971		mltaddr.sin6_family = AF_INET6;
972		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
973		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
974		    0)
975			goto cleanup; /* XXX: should not fail */
976
977		/*
978		 * XXX: do we really need this automatic routes?
979		 * We should probably reconsider this stuff.  Most applications
980		 * actually do not need the routes, since they usually specify
981		 * the outgoing interface.
982		 */
983		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
984		if (rt) {
985			/* XXX: only works in !SCOPEDROUTING case. */
986			if (memcmp(&mltaddr.sin6_addr,
987			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
988			    MLTMASK_LEN)) {
989				RTFREE_LOCKED(rt);
990				rt = NULL;
991			}
992		}
993		if (!rt) {
994			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
995			    (struct sockaddr *)&ia->ia_addr,
996			    (struct sockaddr *)&mltmask, RTF_UP,
997			    (struct rtentry **)0);
998			if (error)
999				goto cleanup;
1000		} else {
1001			RTFREE_LOCKED(rt);
1002		}
1003
1004		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1005		if (!imm) {
1006			nd6log((LOG_WARNING,
1007			    "in6_update_ifa: addmulti failed for "
1008			    "%s on %s (errno=%d)\n",
1009			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1010			    if_name(ifp), error));
1011			goto cleanup;
1012		}
1013		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1014
1015		/*
1016		 * join node information group address
1017		 */
1018#define hostnamelen	strlen(V_hostname)
1019		delay = 0;
1020		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1021			/*
1022			 * The spec doesn't say anything about delay for this
1023			 * group, but the same logic should apply.
1024			 */
1025			delay = arc4random() %
1026			    (MAX_RTR_SOLICITATION_DELAY * hz);
1027		}
1028		mtx_lock(&hostname_mtx);
1029		if (in6_nigroup(ifp, V_hostname, hostnamelen,
1030		    &mltaddr.sin6_addr) == 0) {
1031			mtx_unlock(&hostname_mtx);
1032			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1033			    delay); /* XXX jinmei */
1034			if (!imm) {
1035				nd6log((LOG_WARNING, "in6_update_ifa: "
1036				    "addmulti failed for %s on %s "
1037				    "(errno=%d)\n",
1038				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1039				    if_name(ifp), error));
1040				/* XXX not very fatal, go on... */
1041			} else {
1042				LIST_INSERT_HEAD(&ia->ia6_memberships,
1043				    imm, i6mm_chain);
1044			}
1045		} else
1046			mtx_unlock(&hostname_mtx);
1047#undef hostnamelen
1048
1049		/*
1050		 * join interface-local all-nodes address.
1051		 * (ff01::1%ifN, and ff01::%ifN/32)
1052		 */
1053		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1054		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1055		    != 0)
1056			goto cleanup; /* XXX: should not fail */
1057		/* XXX: again, do we really need the route? */
1058		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1059		if (rt) {
1060			if (memcmp(&mltaddr.sin6_addr,
1061			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1062			    MLTMASK_LEN)) {
1063				RTFREE_LOCKED(rt);
1064				rt = NULL;
1065			}
1066		}
1067		if (!rt) {
1068			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1069			    (struct sockaddr *)&ia->ia_addr,
1070			    (struct sockaddr *)&mltmask, RTF_UP,
1071			    (struct rtentry **)0);
1072			if (error)
1073				goto cleanup;
1074		} else
1075			RTFREE_LOCKED(rt);
1076
1077		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1078		if (!imm) {
1079			nd6log((LOG_WARNING, "in6_update_ifa: "
1080			    "addmulti failed for %s on %s "
1081			    "(errno=%d)\n",
1082			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1083			    if_name(ifp), error));
1084			goto cleanup;
1085		}
1086		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1087#undef	MLTMASK_LEN
1088	}
1089
1090	/*
1091	 * Perform DAD, if needed.
1092	 * XXX It may be of use, if we can administratively
1093	 * disable DAD.
1094	 */
1095	if (hostIsNew && in6if_do_dad(ifp) &&
1096	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1097	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1098	{
1099		int mindelay, maxdelay;
1100
1101		delay = 0;
1102		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1103			/*
1104			 * We need to impose a delay before sending an NS
1105			 * for DAD.  Check if we also needed a delay for the
1106			 * corresponding MLD message.  If we did, the delay
1107			 * should be larger than the MLD delay (this could be
1108			 * relaxed a bit, but this simple logic is at least
1109			 * safe).
1110			 */
1111			mindelay = 0;
1112			if (in6m_sol != NULL &&
1113			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1114				mindelay = in6m_sol->in6m_timer;
1115			}
1116			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1117			if (maxdelay - mindelay == 0)
1118				delay = 0;
1119			else {
1120				delay =
1121				    (arc4random() % (maxdelay - mindelay)) +
1122				    mindelay;
1123			}
1124		}
1125		nd6_dad_start((struct ifaddr *)ia, delay);
1126	}
1127
1128	return (error);
1129
1130  unlink:
1131	/*
1132	 * XXX: if a change of an existing address failed, keep the entry
1133	 * anyway.
1134	 */
1135	if (hostIsNew)
1136		in6_unlink_ifa(ia, ifp);
1137	return (error);
1138
1139  cleanup:
1140	in6_purgeaddr(&ia->ia_ifa);
1141	return error;
1142}
1143
1144void
1145in6_purgeaddr(struct ifaddr *ifa)
1146{
1147	struct ifnet *ifp = ifa->ifa_ifp;
1148	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1149	struct in6_multi_mship *imm;
1150
1151	/* stop DAD processing */
1152	nd6_dad_stop(ifa);
1153
1154	IF_AFDATA_LOCK(ifp);
1155	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
1156	    (struct sockaddr *)&ia->ia_addr);
1157	IF_AFDATA_UNLOCK(ifp);
1158
1159	/*
1160	 * leave from multicast groups we have joined for the interface
1161	 */
1162	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1163		LIST_REMOVE(imm, i6mm_chain);
1164		in6_leavegroup(imm);
1165	}
1166
1167	in6_unlink_ifa(ia, ifp);
1168}
1169
1170static void
1171in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1172{
1173	INIT_VNET_INET6(ifp->if_vnet);
1174	struct in6_ifaddr *oia;
1175	int	s = splnet();
1176
1177	TAILQ_REMOVE(&ifp->if_addrlist, &ia->ia_ifa, ifa_list);
1178
1179	oia = ia;
1180	if (oia == (ia = V_in6_ifaddr))
1181		V_in6_ifaddr = ia->ia_next;
1182	else {
1183		while (ia->ia_next && (ia->ia_next != oia))
1184			ia = ia->ia_next;
1185		if (ia->ia_next)
1186			ia->ia_next = oia->ia_next;
1187		else {
1188			/* search failed */
1189			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1190		}
1191	}
1192
1193	/*
1194	 * Release the reference to the base prefix.  There should be a
1195	 * positive reference.
1196	 */
1197	if (oia->ia6_ndpr == NULL) {
1198		nd6log((LOG_NOTICE,
1199		    "in6_unlink_ifa: autoconf'ed address "
1200		    "%p has no prefix\n", oia));
1201	} else {
1202		oia->ia6_ndpr->ndpr_refcnt--;
1203		oia->ia6_ndpr = NULL;
1204	}
1205
1206	/*
1207	 * Also, if the address being removed is autoconf'ed, call
1208	 * pfxlist_onlink_check() since the release might affect the status of
1209	 * other (detached) addresses.
1210	 */
1211	if ((oia->ia6_flags & IN6_IFF_AUTOCONF)) {
1212		pfxlist_onlink_check();
1213	}
1214
1215	/*
1216	 * release another refcnt for the link from in6_ifaddr.
1217	 * Note that we should decrement the refcnt at least once for all *BSD.
1218	 */
1219	IFAFREE(&oia->ia_ifa);
1220
1221	splx(s);
1222}
1223
1224void
1225in6_purgeif(struct ifnet *ifp)
1226{
1227	struct ifaddr *ifa, *nifa;
1228
1229	for (ifa = TAILQ_FIRST(&ifp->if_addrlist); ifa != NULL; ifa = nifa) {
1230		nifa = TAILQ_NEXT(ifa, ifa_list);
1231		if (ifa->ifa_addr->sa_family != AF_INET6)
1232			continue;
1233		in6_purgeaddr(ifa);
1234	}
1235
1236	in6_ifdetach(ifp);
1237}
1238
1239/*
1240 * SIOC[GAD]LIFADDR.
1241 *	SIOCGLIFADDR: get first address. (?)
1242 *	SIOCGLIFADDR with IFLR_PREFIX:
1243 *		get first address that matches the specified prefix.
1244 *	SIOCALIFADDR: add the specified address.
1245 *	SIOCALIFADDR with IFLR_PREFIX:
1246 *		add the specified prefix, filling hostid part from
1247 *		the first link-local address.  prefixlen must be <= 64.
1248 *	SIOCDLIFADDR: delete the specified address.
1249 *	SIOCDLIFADDR with IFLR_PREFIX:
1250 *		delete the first address that matches the specified prefix.
1251 * return values:
1252 *	EINVAL on invalid parameters
1253 *	EADDRNOTAVAIL on prefix match failed/specified address not found
1254 *	other values may be returned from in6_ioctl()
1255 *
1256 * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1257 * this is to accomodate address naming scheme other than RFC2374,
1258 * in the future.
1259 * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1260 * address encoding scheme. (see figure on page 8)
1261 */
1262static int
1263in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1264    struct ifnet *ifp, struct thread *td)
1265{
1266	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1267	struct ifaddr *ifa;
1268	struct sockaddr *sa;
1269
1270	/* sanity checks */
1271	if (!data || !ifp) {
1272		panic("invalid argument to in6_lifaddr_ioctl");
1273		/* NOTREACHED */
1274	}
1275
1276	switch (cmd) {
1277	case SIOCGLIFADDR:
1278		/* address must be specified on GET with IFLR_PREFIX */
1279		if ((iflr->flags & IFLR_PREFIX) == 0)
1280			break;
1281		/* FALLTHROUGH */
1282	case SIOCALIFADDR:
1283	case SIOCDLIFADDR:
1284		/* address must be specified on ADD and DELETE */
1285		sa = (struct sockaddr *)&iflr->addr;
1286		if (sa->sa_family != AF_INET6)
1287			return EINVAL;
1288		if (sa->sa_len != sizeof(struct sockaddr_in6))
1289			return EINVAL;
1290		/* XXX need improvement */
1291		sa = (struct sockaddr *)&iflr->dstaddr;
1292		if (sa->sa_family && sa->sa_family != AF_INET6)
1293			return EINVAL;
1294		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1295			return EINVAL;
1296		break;
1297	default: /* shouldn't happen */
1298#if 0
1299		panic("invalid cmd to in6_lifaddr_ioctl");
1300		/* NOTREACHED */
1301#else
1302		return EOPNOTSUPP;
1303#endif
1304	}
1305	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1306		return EINVAL;
1307
1308	switch (cmd) {
1309	case SIOCALIFADDR:
1310	    {
1311		struct in6_aliasreq ifra;
1312		struct in6_addr *hostid = NULL;
1313		int prefixlen;
1314
1315		if ((iflr->flags & IFLR_PREFIX) != 0) {
1316			struct sockaddr_in6 *sin6;
1317
1318			/*
1319			 * hostid is to fill in the hostid part of the
1320			 * address.  hostid points to the first link-local
1321			 * address attached to the interface.
1322			 */
1323			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1324			if (!ifa)
1325				return EADDRNOTAVAIL;
1326			hostid = IFA_IN6(ifa);
1327
1328			/* prefixlen must be <= 64. */
1329			if (64 < iflr->prefixlen)
1330				return EINVAL;
1331			prefixlen = iflr->prefixlen;
1332
1333			/* hostid part must be zero. */
1334			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1335			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1336			    sin6->sin6_addr.s6_addr32[3] != 0) {
1337				return EINVAL;
1338			}
1339		} else
1340			prefixlen = iflr->prefixlen;
1341
1342		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1343		bzero(&ifra, sizeof(ifra));
1344		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1345
1346		bcopy(&iflr->addr, &ifra.ifra_addr,
1347		    ((struct sockaddr *)&iflr->addr)->sa_len);
1348		if (hostid) {
1349			/* fill in hostid part */
1350			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1351			    hostid->s6_addr32[2];
1352			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1353			    hostid->s6_addr32[3];
1354		}
1355
1356		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1357			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1358			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1359			if (hostid) {
1360				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1361				    hostid->s6_addr32[2];
1362				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1363				    hostid->s6_addr32[3];
1364			}
1365		}
1366
1367		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1368		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1369
1370		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1371		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1372	    }
1373	case SIOCGLIFADDR:
1374	case SIOCDLIFADDR:
1375	    {
1376		struct in6_ifaddr *ia;
1377		struct in6_addr mask, candidate, match;
1378		struct sockaddr_in6 *sin6;
1379		int cmp;
1380
1381		bzero(&mask, sizeof(mask));
1382		if (iflr->flags & IFLR_PREFIX) {
1383			/* lookup a prefix rather than address. */
1384			in6_prefixlen2mask(&mask, iflr->prefixlen);
1385
1386			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1387			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1388			match.s6_addr32[0] &= mask.s6_addr32[0];
1389			match.s6_addr32[1] &= mask.s6_addr32[1];
1390			match.s6_addr32[2] &= mask.s6_addr32[2];
1391			match.s6_addr32[3] &= mask.s6_addr32[3];
1392
1393			/* if you set extra bits, that's wrong */
1394			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1395				return EINVAL;
1396
1397			cmp = 1;
1398		} else {
1399			if (cmd == SIOCGLIFADDR) {
1400				/* on getting an address, take the 1st match */
1401				cmp = 0;	/* XXX */
1402			} else {
1403				/* on deleting an address, do exact match */
1404				in6_prefixlen2mask(&mask, 128);
1405				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1406				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1407
1408				cmp = 1;
1409			}
1410		}
1411
1412		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1413			if (ifa->ifa_addr->sa_family != AF_INET6)
1414				continue;
1415			if (!cmp)
1416				break;
1417
1418			/*
1419			 * XXX: this is adhoc, but is necessary to allow
1420			 * a user to specify fe80::/64 (not /10) for a
1421			 * link-local address.
1422			 */
1423			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1424			in6_clearscope(&candidate);
1425			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1426			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1427			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1428			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1429			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1430				break;
1431		}
1432		if (!ifa)
1433			return EADDRNOTAVAIL;
1434		ia = ifa2ia6(ifa);
1435
1436		if (cmd == SIOCGLIFADDR) {
1437			int error;
1438
1439			/* fill in the if_laddrreq structure */
1440			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1441			error = sa6_recoverscope(
1442			    (struct sockaddr_in6 *)&iflr->addr);
1443			if (error != 0)
1444				return (error);
1445
1446			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1447				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1448				    ia->ia_dstaddr.sin6_len);
1449				error = sa6_recoverscope(
1450				    (struct sockaddr_in6 *)&iflr->dstaddr);
1451				if (error != 0)
1452					return (error);
1453			} else
1454				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1455
1456			iflr->prefixlen =
1457			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1458
1459			iflr->flags = ia->ia6_flags;	/* XXX */
1460
1461			return 0;
1462		} else {
1463			struct in6_aliasreq ifra;
1464
1465			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1466			bzero(&ifra, sizeof(ifra));
1467			bcopy(iflr->iflr_name, ifra.ifra_name,
1468			    sizeof(ifra.ifra_name));
1469
1470			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1471			    ia->ia_addr.sin6_len);
1472			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1473				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1474				    ia->ia_dstaddr.sin6_len);
1475			} else {
1476				bzero(&ifra.ifra_dstaddr,
1477				    sizeof(ifra.ifra_dstaddr));
1478			}
1479			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1480			    ia->ia_prefixmask.sin6_len);
1481
1482			ifra.ifra_flags = ia->ia6_flags;
1483			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1484			    ifp, td);
1485		}
1486	    }
1487	}
1488
1489	return EOPNOTSUPP;	/* just for safety */
1490}
1491
1492/*
1493 * Initialize an interface's intetnet6 address
1494 * and routing table entry.
1495 */
1496static int
1497in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1498    struct sockaddr_in6 *sin6, int newhost)
1499{
1500	int	error = 0, plen, ifacount = 0;
1501	int	s = splimp();
1502	struct ifaddr *ifa;
1503
1504	/*
1505	 * Give the interface a chance to initialize
1506	 * if this is its first address,
1507	 * and to validate the address if necessary.
1508	 */
1509	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1510		if (ifa->ifa_addr->sa_family != AF_INET6)
1511			continue;
1512		ifacount++;
1513	}
1514
1515	ia->ia_addr = *sin6;
1516
1517	if (ifacount <= 1 && ifp->if_ioctl) {
1518		IFF_LOCKGIANT(ifp);
1519		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1520		IFF_UNLOCKGIANT(ifp);
1521		if (error) {
1522			splx(s);
1523			return (error);
1524		}
1525	}
1526	splx(s);
1527
1528	ia->ia_ifa.ifa_metric = ifp->if_metric;
1529
1530	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1531
1532	/*
1533	 * Special case:
1534	 * If a new destination address is specified for a point-to-point
1535	 * interface, install a route to the destination as an interface
1536	 * direct route.
1537	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1538	 * interface that share the same destination.
1539	 */
1540	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1541	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128) {
1542		struct sockaddr *dstaddr;
1543		int rtflags = RTF_UP | RTF_HOST;
1544
1545		/*
1546		 * use the interface address if configuring an
1547		 * interface address with a /128 prefix len
1548		 */
1549		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1550			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1551		else
1552			dstaddr = (struct sockaddr *)&ia->ia_addr;
1553
1554		error = rtrequest(RTM_ADD,
1555		    (struct sockaddr *)dstaddr,
1556		    (struct sockaddr *)&ia->ia_addr,
1557		    (struct sockaddr *)&ia->ia_prefixmask,
1558		    ia->ia_flags | rtflags, NULL);
1559		if (error != 0)
1560			return (error);
1561		ia->ia_flags |= IFA_ROUTE;
1562	}
1563
1564	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1565	if (newhost) {
1566		struct llentry *ln;
1567
1568		IF_AFDATA_LOCK(ifp);
1569		ia->ia_ifa.ifa_rtrequest = NULL;
1570
1571		/* XXX QL
1572		 * we need to report rt_newaddrmsg
1573		 */
1574		ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR | LLE_EXCLUSIVE),
1575		    (struct sockaddr *)&ia->ia_addr);
1576		IF_AFDATA_UNLOCK(ifp);
1577		if (ln != NULL) {
1578			ln->la_expire = 0;  /* for IPv6 this means permanent */
1579			ln->ln_state = ND6_LLINFO_REACHABLE;
1580			LLE_WUNLOCK(ln);
1581		}
1582	}
1583
1584	return (error);
1585}
1586
1587struct in6_multi_mship *
1588in6_joingroup(struct ifnet *ifp, struct in6_addr *addr,
1589    int *errorp, int delay)
1590{
1591	struct in6_multi_mship *imm;
1592
1593	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
1594	if (!imm) {
1595		*errorp = ENOBUFS;
1596		return NULL;
1597	}
1598	imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp, delay);
1599	if (!imm->i6mm_maddr) {
1600		/* *errorp is alrady set */
1601		free(imm, M_IP6MADDR);
1602		return NULL;
1603	}
1604	return imm;
1605}
1606
1607int
1608in6_leavegroup(struct in6_multi_mship *imm)
1609{
1610
1611	if (imm->i6mm_maddr)
1612		in6_delmulti(imm->i6mm_maddr);
1613	free(imm,  M_IP6MADDR);
1614	return 0;
1615}
1616
1617/*
1618 * Find an IPv6 interface link-local address specific to an interface.
1619 */
1620struct in6_ifaddr *
1621in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1622{
1623	struct ifaddr *ifa;
1624
1625	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1626		if (ifa->ifa_addr->sa_family != AF_INET6)
1627			continue;
1628		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1629			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1630			     ignoreflags) != 0)
1631				continue;
1632			break;
1633		}
1634	}
1635
1636	return ((struct in6_ifaddr *)ifa);
1637}
1638
1639
1640/*
1641 * find the internet address corresponding to a given interface and address.
1642 */
1643struct in6_ifaddr *
1644in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1645{
1646	struct ifaddr *ifa;
1647
1648	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1649		if (ifa->ifa_addr->sa_family != AF_INET6)
1650			continue;
1651		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1652			break;
1653	}
1654
1655	return ((struct in6_ifaddr *)ifa);
1656}
1657
1658/*
1659 * Convert IP6 address to printable (loggable) representation. Caller
1660 * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1661 */
1662static char digits[] = "0123456789abcdef";
1663char *
1664ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1665{
1666	int i;
1667	char *cp;
1668	const u_int16_t *a = (const u_int16_t *)addr;
1669	const u_int8_t *d;
1670	int dcolon = 0, zero = 0;
1671
1672	cp = ip6buf;
1673
1674	for (i = 0; i < 8; i++) {
1675		if (dcolon == 1) {
1676			if (*a == 0) {
1677				if (i == 7)
1678					*cp++ = ':';
1679				a++;
1680				continue;
1681			} else
1682				dcolon = 2;
1683		}
1684		if (*a == 0) {
1685			if (dcolon == 0 && *(a + 1) == 0) {
1686				if (i == 0)
1687					*cp++ = ':';
1688				*cp++ = ':';
1689				dcolon = 1;
1690			} else {
1691				*cp++ = '0';
1692				*cp++ = ':';
1693			}
1694			a++;
1695			continue;
1696		}
1697		d = (const u_char *)a;
1698		/* Try to eliminate leading zeros in printout like in :0001. */
1699		zero = 1;
1700		*cp = digits[*d >> 4];
1701		if (*cp != '0') {
1702			zero = 0;
1703			cp++;
1704		}
1705		*cp = digits[*d++ & 0xf];
1706		if (zero == 0 || (*cp != '0')) {
1707			zero = 0;
1708			cp++;
1709		}
1710		*cp = digits[*d >> 4];
1711		if (zero == 0 || (*cp != '0')) {
1712			zero = 0;
1713			cp++;
1714		}
1715		*cp++ = digits[*d & 0xf];
1716		*cp++ = ':';
1717		a++;
1718	}
1719	*--cp = '\0';
1720	return (ip6buf);
1721}
1722
1723int
1724in6_localaddr(struct in6_addr *in6)
1725{
1726	INIT_VNET_INET6(curvnet);
1727	struct in6_ifaddr *ia;
1728
1729	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1730		return 1;
1731
1732	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1733		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1734		    &ia->ia_prefixmask.sin6_addr)) {
1735			return 1;
1736		}
1737	}
1738
1739	return (0);
1740}
1741
1742int
1743in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1744{
1745	INIT_VNET_INET6(curvnet);
1746	struct in6_ifaddr *ia;
1747
1748	for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1749		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1750				       &sa6->sin6_addr) &&
1751		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1752			return (1); /* true */
1753
1754		/* XXX: do we still have to go thru the rest of the list? */
1755	}
1756
1757	return (0);		/* false */
1758}
1759
1760/*
1761 * return length of part which dst and src are equal
1762 * hard coding...
1763 */
1764int
1765in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1766{
1767	int match = 0;
1768	u_char *s = (u_char *)src, *d = (u_char *)dst;
1769	u_char *lim = s + 16, r;
1770
1771	while (s < lim)
1772		if ((r = (*d++ ^ *s++)) != 0) {
1773			while (r < 128) {
1774				match++;
1775				r <<= 1;
1776			}
1777			break;
1778		} else
1779			match += 8;
1780	return match;
1781}
1782
1783/* XXX: to be scope conscious */
1784int
1785in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1786{
1787	int bytelen, bitlen;
1788
1789	/* sanity check */
1790	if (0 > len || len > 128) {
1791		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1792		    len);
1793		return (0);
1794	}
1795
1796	bytelen = len / 8;
1797	bitlen = len % 8;
1798
1799	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1800		return (0);
1801	if (bitlen != 0 &&
1802	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1803	    p2->s6_addr[bytelen] >> (8 - bitlen))
1804		return (0);
1805
1806	return (1);
1807}
1808
1809void
1810in6_prefixlen2mask(struct in6_addr *maskp, int len)
1811{
1812	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1813	int bytelen, bitlen, i;
1814
1815	/* sanity check */
1816	if (0 > len || len > 128) {
1817		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1818		    len);
1819		return;
1820	}
1821
1822	bzero(maskp, sizeof(*maskp));
1823	bytelen = len / 8;
1824	bitlen = len % 8;
1825	for (i = 0; i < bytelen; i++)
1826		maskp->s6_addr[i] = 0xff;
1827	if (bitlen)
1828		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1829}
1830
1831/*
1832 * return the best address out of the same scope. if no address was
1833 * found, return the first valid address from designated IF.
1834 */
1835struct in6_ifaddr *
1836in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1837{
1838	INIT_VNET_INET6(curvnet);
1839	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1840	struct ifaddr *ifa;
1841	struct in6_ifaddr *besta = 0;
1842	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1843
1844	dep[0] = dep[1] = NULL;
1845
1846	/*
1847	 * We first look for addresses in the same scope.
1848	 * If there is one, return it.
1849	 * If two or more, return one which matches the dst longest.
1850	 * If none, return one of global addresses assigned other ifs.
1851	 */
1852	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1853		if (ifa->ifa_addr->sa_family != AF_INET6)
1854			continue;
1855		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1856			continue; /* XXX: is there any case to allow anycast? */
1857		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1858			continue; /* don't use this interface */
1859		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1860			continue;
1861		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1862			if (V_ip6_use_deprecated)
1863				dep[0] = (struct in6_ifaddr *)ifa;
1864			continue;
1865		}
1866
1867		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1868			/*
1869			 * call in6_matchlen() as few as possible
1870			 */
1871			if (besta) {
1872				if (blen == -1)
1873					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1874				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1875				if (tlen > blen) {
1876					blen = tlen;
1877					besta = (struct in6_ifaddr *)ifa;
1878				}
1879			} else
1880				besta = (struct in6_ifaddr *)ifa;
1881		}
1882	}
1883	if (besta)
1884		return (besta);
1885
1886	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1887		if (ifa->ifa_addr->sa_family != AF_INET6)
1888			continue;
1889		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1890			continue; /* XXX: is there any case to allow anycast? */
1891		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1892			continue; /* don't use this interface */
1893		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1894			continue;
1895		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1896			if (V_ip6_use_deprecated)
1897				dep[1] = (struct in6_ifaddr *)ifa;
1898			continue;
1899		}
1900
1901		return (struct in6_ifaddr *)ifa;
1902	}
1903
1904	/* use the last-resort values, that are, deprecated addresses */
1905	if (dep[0])
1906		return dep[0];
1907	if (dep[1])
1908		return dep[1];
1909
1910	return NULL;
1911}
1912
1913/*
1914 * perform DAD when interface becomes IFF_UP.
1915 */
1916void
1917in6_if_up(struct ifnet *ifp)
1918{
1919	struct ifaddr *ifa;
1920	struct in6_ifaddr *ia;
1921
1922	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1923		if (ifa->ifa_addr->sa_family != AF_INET6)
1924			continue;
1925		ia = (struct in6_ifaddr *)ifa;
1926		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1927			/*
1928			 * The TENTATIVE flag was likely set by hand
1929			 * beforehand, implicitly indicating the need for DAD.
1930			 * We may be able to skip the random delay in this
1931			 * case, but we impose delays just in case.
1932			 */
1933			nd6_dad_start(ifa,
1934			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1935		}
1936	}
1937
1938	/*
1939	 * special cases, like 6to4, are handled in in6_ifattach
1940	 */
1941	in6_ifattach(ifp, NULL);
1942}
1943
1944int
1945in6if_do_dad(struct ifnet *ifp)
1946{
1947	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1948		return (0);
1949
1950	switch (ifp->if_type) {
1951#ifdef IFT_DUMMY
1952	case IFT_DUMMY:
1953#endif
1954	case IFT_FAITH:
1955		/*
1956		 * These interfaces do not have the IFF_LOOPBACK flag,
1957		 * but loop packets back.  We do not have to do DAD on such
1958		 * interfaces.  We should even omit it, because loop-backed
1959		 * NS would confuse the DAD procedure.
1960		 */
1961		return (0);
1962	default:
1963		/*
1964		 * Our DAD routine requires the interface up and running.
1965		 * However, some interfaces can be up before the RUNNING
1966		 * status.  Additionaly, users may try to assign addresses
1967		 * before the interface becomes up (or running).
1968		 * We simply skip DAD in such a case as a work around.
1969		 * XXX: we should rather mark "tentative" on such addresses,
1970		 * and do DAD after the interface becomes ready.
1971		 */
1972		if (!((ifp->if_flags & IFF_UP) &&
1973		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
1974			return (0);
1975
1976		return (1);
1977	}
1978}
1979
1980/*
1981 * Calculate max IPv6 MTU through all the interfaces and store it
1982 * to in6_maxmtu.
1983 */
1984void
1985in6_setmaxmtu(void)
1986{
1987	INIT_VNET_NET(curvnet);
1988	INIT_VNET_INET6(curvnet);
1989	unsigned long maxmtu = 0;
1990	struct ifnet *ifp;
1991
1992	IFNET_RLOCK();
1993	for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
1994	    ifp = TAILQ_NEXT(ifp, if_list)) {
1995		/* this function can be called during ifnet initialization */
1996		if (!ifp->if_afdata[AF_INET6])
1997			continue;
1998		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1999		    IN6_LINKMTU(ifp) > maxmtu)
2000			maxmtu = IN6_LINKMTU(ifp);
2001	}
2002	IFNET_RUNLOCK();
2003	if (maxmtu)	     /* update only when maxmtu is positive */
2004		V_in6_maxmtu = maxmtu;
2005}
2006
2007/*
2008 * Provide the length of interface identifiers to be used for the link attached
2009 * to the given interface.  The length should be defined in "IPv6 over
2010 * xxx-link" document.  Note that address architecture might also define
2011 * the length for a particular set of address prefixes, regardless of the
2012 * link type.  As clarified in rfc2462bis, those two definitions should be
2013 * consistent, and those really are as of August 2004.
2014 */
2015int
2016in6_if2idlen(struct ifnet *ifp)
2017{
2018	switch (ifp->if_type) {
2019	case IFT_ETHER:		/* RFC2464 */
2020#ifdef IFT_PROPVIRTUAL
2021	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2022#endif
2023#ifdef IFT_L2VLAN
2024	case IFT_L2VLAN:	/* ditto */
2025#endif
2026#ifdef IFT_IEEE80211
2027	case IFT_IEEE80211:	/* ditto */
2028#endif
2029#ifdef IFT_MIP
2030	case IFT_MIP:	/* ditto */
2031#endif
2032		return (64);
2033	case IFT_FDDI:		/* RFC2467 */
2034		return (64);
2035	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2036		return (64);
2037	case IFT_PPP:		/* RFC2472 */
2038		return (64);
2039	case IFT_ARCNET:	/* RFC2497 */
2040		return (64);
2041	case IFT_FRELAY:	/* RFC2590 */
2042		return (64);
2043	case IFT_IEEE1394:	/* RFC3146 */
2044		return (64);
2045	case IFT_GIF:
2046		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2047	case IFT_LOOP:
2048		return (64);	/* XXX: is this really correct? */
2049	default:
2050		/*
2051		 * Unknown link type:
2052		 * It might be controversial to use the today's common constant
2053		 * of 64 for these cases unconditionally.  For full compliance,
2054		 * we should return an error in this case.  On the other hand,
2055		 * if we simply miss the standard for the link type or a new
2056		 * standard is defined for a new link type, the IFID length
2057		 * is very likely to be the common constant.  As a compromise,
2058		 * we always use the constant, but make an explicit notice
2059		 * indicating the "unknown" case.
2060		 */
2061		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2062		return (64);
2063	}
2064}
2065
2066#include <sys/sysctl.h>
2067
2068struct in6_llentry {
2069	struct llentry		base;
2070	struct sockaddr_in6	l3_addr6;
2071};
2072
2073static struct llentry *
2074in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2075{
2076	struct in6_llentry *lle;
2077
2078	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE,
2079	    M_DONTWAIT | M_ZERO);
2080	if (lle == NULL)		/* NB: caller generates msg */
2081		return NULL;
2082
2083	callout_init(&lle->base.ln_timer_ch, CALLOUT_MPSAFE);
2084	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2085	lle->base.lle_refcnt = 1;
2086	LLE_LOCK_INIT(&lle->base);
2087	return &lle->base;
2088}
2089
2090/*
2091 * Deletes an address from the address table.
2092 * This function is called by the timer functions
2093 * such as arptimer() and nd6_llinfo_timer(), and
2094 * the caller does the locking.
2095 */
2096static void
2097in6_lltable_free(struct lltable *llt, struct llentry *lle)
2098{
2099	LLE_WUNLOCK(lle);
2100	LLE_LOCK_DESTROY(lle);
2101	free(lle, M_LLTABLE);
2102}
2103
2104static int
2105in6_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr)
2106{
2107	struct rtentry *rt;
2108	char ip6buf[INET6_ADDRSTRLEN];
2109
2110	KASSERT(l3addr->sa_family == AF_INET6,
2111	    ("sin_family %d", l3addr->sa_family));
2112
2113	/* XXX rtalloc1 should take a const param */
2114	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
2115	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2116		struct ifaddr *ifa;
2117		/*
2118		 * Create an ND6 cache for an IPv6 neighbor
2119		 * that is not covered by our own prefix.
2120		 */
2121		/* XXX ifaof_ifpforaddr should take a const param */
2122		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2123		if (ifa != NULL) {
2124			if (rt != NULL)
2125				rtfree(rt);
2126			return 0;
2127		}
2128		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2129		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2130		if (rt != NULL)
2131			rtfree(rt);
2132		return EINVAL;
2133	}
2134	rtfree(rt);
2135	return 0;
2136}
2137
2138static struct llentry *
2139in6_lltable_lookup(struct lltable *llt, u_int flags,
2140	const struct sockaddr *l3addr)
2141{
2142	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2143	struct ifnet *ifp = llt->llt_ifp;
2144	struct llentry *lle;
2145	struct llentries *lleh;
2146	u_int hashkey;
2147
2148	IF_AFDATA_LOCK_ASSERT(ifp);
2149	KASSERT(l3addr->sa_family == AF_INET6,
2150	    ("sin_family %d", l3addr->sa_family));
2151
2152	hashkey = sin6->sin6_addr.s6_addr32[3];
2153	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2154	LIST_FOREACH(lle, lleh, lle_next) {
2155		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2156		if (lle->la_flags & LLE_DELETED)
2157			continue;
2158		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2159			 sizeof(struct in6_addr)) == 0)
2160			break;
2161	}
2162
2163	if (lle == NULL) {
2164		if (!(flags & LLE_CREATE))
2165			return (NULL);
2166		/*
2167		 * A route that covers the given address must have
2168		 * been installed 1st because we are doing a resolution,
2169		 * verify this.
2170		 */
2171		if (!(flags & LLE_IFADDR) &&
2172		    in6_lltable_rtcheck(ifp, l3addr) != 0)
2173			return NULL;
2174
2175		lle = in6_lltable_new(l3addr, flags);
2176		if (lle == NULL) {
2177			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2178			return NULL;
2179		}
2180		lle->la_flags = flags & ~LLE_CREATE;
2181		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2182			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2183			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2184		}
2185
2186		lle->lle_tbl  = llt;
2187		lle->lle_head = lleh;
2188		LIST_INSERT_HEAD(lleh, lle, lle_next);
2189	} else if (flags & LLE_DELETE) {
2190		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2191			LLE_WLOCK(lle);
2192			lle->la_flags = LLE_DELETED;
2193			LLE_WUNLOCK(lle);
2194#ifdef DIAGNOSTICS
2195			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2196#endif
2197		}
2198		lle = (void *)-1;
2199	}
2200	if (LLE_IS_VALID(lle)) {
2201		if (flags & LLE_EXCLUSIVE)
2202			LLE_WLOCK(lle);
2203		else
2204			LLE_RLOCK(lle);
2205	}
2206	return (lle);
2207}
2208
2209static int
2210in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2211{
2212	struct ifnet *ifp = llt->llt_ifp;
2213	struct llentry *lle;
2214	/* XXX stack use */
2215	struct {
2216		struct rt_msghdr	rtm;
2217		struct sockaddr_in6	sin6;
2218		/*
2219		 * ndp.c assumes that sdl is word aligned
2220		 */
2221#ifdef __LP64__
2222		uint32_t		pad;
2223#endif
2224		struct sockaddr_dl	sdl;
2225	} ndpc;
2226	int i, error;
2227
2228	/* XXXXX
2229	 * current IFNET_RLOCK() is mapped to IFNET_WLOCK()
2230	 * so it is okay to use this ASSERT, change it when
2231	 * IFNET lock is finalized
2232	 */
2233	IFNET_WLOCK_ASSERT();
2234
2235	error = 0;
2236	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2237		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2238			struct sockaddr_dl *sdl;
2239
2240			/* skip deleted or invalid entries */
2241			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2242				continue;
2243			/* Skip if jailed and not a valid IP of the prison. */
2244			if (jailed(wr->td->td_ucred) &&
2245			    !prison_if(wr->td->td_ucred, L3_ADDR(lle)))
2246				continue;
2247			/*
2248			 * produce a msg made of:
2249			 *  struct rt_msghdr;
2250			 *  struct sockaddr_in6 (IPv6)
2251			 *  struct sockaddr_dl;
2252			 */
2253			bzero(&ndpc, sizeof(ndpc));
2254			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2255			ndpc.rtm.rtm_version = RTM_VERSION;
2256			ndpc.rtm.rtm_type = RTM_GET;
2257			ndpc.rtm.rtm_flags = RTF_UP;
2258			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2259			ndpc.sin6.sin6_family = AF_INET6;
2260			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2261			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2262
2263			/* publish */
2264			if (lle->la_flags & LLE_PUB)
2265				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2266
2267			sdl = &ndpc.sdl;
2268			sdl->sdl_family = AF_LINK;
2269			sdl->sdl_len = sizeof(*sdl);
2270			sdl->sdl_alen = ifp->if_addrlen;
2271			sdl->sdl_index = ifp->if_index;
2272			sdl->sdl_type = ifp->if_type;
2273			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2274			ndpc.rtm.rtm_rmx.rmx_expire =
2275			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2276			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2277			if (lle->la_flags & LLE_STATIC)
2278				ndpc.rtm.rtm_flags |= RTF_STATIC;
2279			ndpc.rtm.rtm_index = ifp->if_index;
2280			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2281			if (error)
2282				break;
2283		}
2284	}
2285	return error;
2286}
2287
2288void *
2289in6_domifattach(struct ifnet *ifp)
2290{
2291	struct in6_ifextra *ext;
2292
2293	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2294	bzero(ext, sizeof(*ext));
2295
2296	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2297	    M_IFADDR, M_WAITOK);
2298	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2299
2300	ext->icmp6_ifstat =
2301	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2302	    M_IFADDR, M_WAITOK);
2303	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2304
2305	ext->nd_ifinfo = nd6_ifattach(ifp);
2306	ext->scope6_id = scope6_ifattach(ifp);
2307	ext->lltable = lltable_init(ifp, AF_INET6);
2308	if (ext->lltable != NULL) {
2309		ext->lltable->llt_new = in6_lltable_new;
2310		ext->lltable->llt_free = in6_lltable_free;
2311		ext->lltable->llt_rtcheck = in6_lltable_rtcheck;
2312		ext->lltable->llt_lookup = in6_lltable_lookup;
2313		ext->lltable->llt_dump = in6_lltable_dump;
2314	}
2315	return ext;
2316}
2317
2318void
2319in6_domifdetach(struct ifnet *ifp, void *aux)
2320{
2321	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2322
2323	scope6_ifdetach(ext->scope6_id);
2324	nd6_ifdetach(ext->nd_ifinfo);
2325	lltable_free(ext->lltable);
2326	free(ext->in6_ifstat, M_IFADDR);
2327	free(ext->icmp6_ifstat, M_IFADDR);
2328	free(ext, M_IFADDR);
2329}
2330
2331/*
2332 * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2333 * v4 mapped addr or v4 compat addr
2334 */
2335void
2336in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2337{
2338
2339	bzero(sin, sizeof(*sin));
2340	sin->sin_len = sizeof(struct sockaddr_in);
2341	sin->sin_family = AF_INET;
2342	sin->sin_port = sin6->sin6_port;
2343	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2344}
2345
2346/* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2347void
2348in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2349{
2350	bzero(sin6, sizeof(*sin6));
2351	sin6->sin6_len = sizeof(struct sockaddr_in6);
2352	sin6->sin6_family = AF_INET6;
2353	sin6->sin6_port = sin->sin_port;
2354	sin6->sin6_addr.s6_addr32[0] = 0;
2355	sin6->sin6_addr.s6_addr32[1] = 0;
2356	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2357	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2358}
2359
2360/* Convert sockaddr_in6 into sockaddr_in. */
2361void
2362in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2363{
2364	struct sockaddr_in *sin_p;
2365	struct sockaddr_in6 sin6;
2366
2367	/*
2368	 * Save original sockaddr_in6 addr and convert it
2369	 * to sockaddr_in.
2370	 */
2371	sin6 = *(struct sockaddr_in6 *)nam;
2372	sin_p = (struct sockaddr_in *)nam;
2373	in6_sin6_2_sin(sin_p, &sin6);
2374}
2375
2376/* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2377void
2378in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2379{
2380	struct sockaddr_in *sin_p;
2381	struct sockaddr_in6 *sin6_p;
2382
2383	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
2384	       M_WAITOK);
2385	sin_p = (struct sockaddr_in *)*nam;
2386	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2387	free(*nam, M_SONAME);
2388	*nam = (struct sockaddr *)sin6_p;
2389}
2390