Home | History | Annotate | Line # | Download | only in netinet6
in6.c revision 1.193
      1 /*	$NetBSD: in6.c,v 1.193 2015/11/27 02:54:22 ozaki-r Exp $	*/
      2 /*	$KAME: in6.c,v 1.198 2001/07/18 09:12:38 itojun Exp $	*/
      3 
      4 /*
      5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      6  * All rights reserved.
      7  *
      8  * Redistribution and use in source and binary forms, with or without
      9  * modification, are permitted provided that the following conditions
     10  * are met:
     11  * 1. Redistributions of source code must retain the above copyright
     12  *    notice, this list of conditions and the following disclaimer.
     13  * 2. Redistributions in binary form must reproduce the above copyright
     14  *    notice, this list of conditions and the following disclaimer in the
     15  *    documentation and/or other materials provided with the distribution.
     16  * 3. Neither the name of the project nor the names of its contributors
     17  *    may be used to endorse or promote products derived from this software
     18  *    without specific prior written permission.
     19  *
     20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     30  * SUCH DAMAGE.
     31  */
     32 
     33 /*
     34  * Copyright (c) 1982, 1986, 1991, 1993
     35  *	The Regents of the University of California.  All rights reserved.
     36  *
     37  * Redistribution and use in source and binary forms, with or without
     38  * modification, are permitted provided that the following conditions
     39  * are met:
     40  * 1. Redistributions of source code must retain the above copyright
     41  *    notice, this list of conditions and the following disclaimer.
     42  * 2. Redistributions in binary form must reproduce the above copyright
     43  *    notice, this list of conditions and the following disclaimer in the
     44  *    documentation and/or other materials provided with the distribution.
     45  * 3. Neither the name of the University nor the names of its contributors
     46  *    may be used to endorse or promote products derived from this software
     47  *    without specific prior written permission.
     48  *
     49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     59  * SUCH DAMAGE.
     60  *
     61  *	@(#)in.c	8.2 (Berkeley) 11/15/93
     62  */
     63 
     64 #include <sys/cdefs.h>
     65 __KERNEL_RCSID(0, "$NetBSD: in6.c,v 1.193 2015/11/27 02:54:22 ozaki-r Exp $");
     66 
     67 #ifdef _KERNEL_OPT
     68 #include "opt_inet.h"
     69 #include "opt_compat_netbsd.h"
     70 #endif
     71 
     72 #include <sys/param.h>
     73 #include <sys/ioctl.h>
     74 #include <sys/errno.h>
     75 #include <sys/malloc.h>
     76 #include <sys/socket.h>
     77 #include <sys/socketvar.h>
     78 #include <sys/sockio.h>
     79 #include <sys/systm.h>
     80 #include <sys/proc.h>
     81 #include <sys/time.h>
     82 #include <sys/kernel.h>
     83 #include <sys/syslog.h>
     84 #include <sys/kauth.h>
     85 #include <sys/cprng.h>
     86 #include <sys/kmem.h>
     87 
     88 #include <net/if.h>
     89 #include <net/if_types.h>
     90 #include <net/if_llatbl.h>
     91 #include <net/if_ether.h>
     92 #include <net/if_dl.h>
     93 #include <net/pfil.h>
     94 #include <net/route.h>
     95 
     96 #include <netinet/in.h>
     97 #include <netinet/in_var.h>
     98 
     99 #include <netinet/ip6.h>
    100 #include <netinet6/ip6_var.h>
    101 #include <netinet6/nd6.h>
    102 #include <netinet6/mld6_var.h>
    103 #include <netinet6/ip6_mroute.h>
    104 #include <netinet6/in6_ifattach.h>
    105 #include <netinet6/scope6_var.h>
    106 
    107 #include <net/net_osdep.h>
    108 
    109 #ifdef COMPAT_50
    110 #include <compat/netinet6/in6_var.h>
    111 #endif
    112 
    113 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
    114 
    115 /* enable backward compatibility code for obsoleted ioctls */
    116 #define COMPAT_IN6IFIOCTL
    117 
    118 #ifdef	IN6_DEBUG
    119 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
    120 #else
    121 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
    122 #endif /* IN6_DEBUG */
    123 
    124 /*
    125  * Definitions of some constant IP6 addresses.
    126  */
    127 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
    128 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
    129 const struct in6_addr in6addr_nodelocal_allnodes =
    130 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
    131 const struct in6_addr in6addr_linklocal_allnodes =
    132 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
    133 const struct in6_addr in6addr_linklocal_allrouters =
    134 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
    135 
    136 const struct in6_addr in6mask0 = IN6MASK0;
    137 const struct in6_addr in6mask32 = IN6MASK32;
    138 const struct in6_addr in6mask64 = IN6MASK64;
    139 const struct in6_addr in6mask96 = IN6MASK96;
    140 const struct in6_addr in6mask128 = IN6MASK128;
    141 
    142 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
    143 				     0, 0, IN6ADDR_ANY_INIT, 0};
    144 
    145 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
    146 	struct ifnet *);
    147 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
    148 	const struct sockaddr_in6 *, int);
    149 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
    150 
    151 /*
    152  * Add ownaddr as loopback rtentry.  We previously add the route only if
    153  * necessary (ex. on a p2p link).  However, since we now manage addresses
    154  * separately from prefixes, we should always add the route.  We can't
    155  * rely on the cloning mechanism from the corresponding interface route
    156  * any more.
    157  */
    158 void
    159 in6_ifaddlocal(struct ifaddr *ifa)
    160 {
    161 
    162 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
    163 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
    164 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
    165 	{
    166 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
    167 		return;
    168 	}
    169 
    170 	rt_ifa_addlocal(ifa);
    171 }
    172 
    173 /*
    174  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
    175  * if it exists.
    176  */
    177 void
    178 in6_ifremlocal(struct ifaddr *ifa)
    179 {
    180 	struct in6_ifaddr *ia;
    181 	struct ifaddr *alt_ifa = NULL;
    182 	int ia_count = 0;
    183 
    184 	/*
    185 	 * Some of BSD variants do not remove cloned routes
    186 	 * from an interface direct route, when removing the direct route
    187 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
    188 	 * cloned routes, they could fail to remove the cloned routes when
    189 	 * we handle multple addresses that share a common prefix.
    190 	 * So, we should remove the route corresponding to the deleted address.
    191 	 */
    192 
    193 	/*
    194 	 * Delete the entry only if exactly one ifaddr matches the
    195 	 * address, ifa->ifa_addr.
    196 	 *
    197 	 * If more than one ifaddr matches, replace the ifaddr in
    198 	 * the routing table, rt_ifa, with a different ifaddr than
    199 	 * the one we are purging, ifa.  It is important to do
    200 	 * this, or else the routing table can accumulate dangling
    201 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
    202 	 * which will lead to crashes, later.  (More than one ifaddr
    203 	 * can match if we assign the same address to multiple---probably
    204 	 * p2p---interfaces.)
    205 	 *
    206 	 * XXX An old comment at this place said, "we should avoid
    207 	 * XXX such a configuration [i.e., interfaces with the same
    208 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
    209 	 * XXX agree, especially now that I have fixed the dangling
    210 	 * XXX ifp-pointers bug.
    211 	 */
    212 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
    213 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
    214 			continue;
    215 		if (ia->ia_ifp != ifa->ifa_ifp)
    216 			alt_ifa = &ia->ia_ifa;
    217 		if (++ia_count > 1 && alt_ifa != NULL)
    218 			break;
    219 	}
    220 
    221 	if (ia_count == 0)
    222 		return;
    223 
    224 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
    225 }
    226 
    227 int
    228 in6_mask2len(struct in6_addr *mask, u_char *lim0)
    229 {
    230 	int x = 0, y;
    231 	u_char *lim = lim0, *p;
    232 
    233 	/* ignore the scope_id part */
    234 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
    235 		lim = (u_char *)mask + sizeof(*mask);
    236 	for (p = (u_char *)mask; p < lim; x++, p++) {
    237 		if (*p != 0xff)
    238 			break;
    239 	}
    240 	y = 0;
    241 	if (p < lim) {
    242 		for (y = 0; y < NBBY; y++) {
    243 			if ((*p & (0x80 >> y)) == 0)
    244 				break;
    245 		}
    246 	}
    247 
    248 	/*
    249 	 * when the limit pointer is given, do a stricter check on the
    250 	 * remaining bits.
    251 	 */
    252 	if (p < lim) {
    253 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
    254 			return -1;
    255 		for (p = p + 1; p < lim; p++)
    256 			if (*p != 0)
    257 				return -1;
    258 	}
    259 
    260 	return x * NBBY + y;
    261 }
    262 
    263 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
    264 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
    265 
    266 static int
    267 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
    268 {
    269 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
    270 	struct	in6_ifaddr *ia = NULL;
    271 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
    272 	struct sockaddr_in6 *sa6;
    273 	int error;
    274 
    275 	switch (cmd) {
    276 	/*
    277 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
    278 	 */
    279 	case SIOCSIFADDR:
    280 	case SIOCSIFDSTADDR:
    281 	case SIOCSIFBRDADDR:
    282 	case SIOCSIFNETMASK:
    283 		return EOPNOTSUPP;
    284 	case SIOCGETSGCNT_IN6:
    285 	case SIOCGETMIFCNT_IN6:
    286 		return mrt6_ioctl(cmd, data);
    287 	case SIOCGIFADDRPREF:
    288 	case SIOCSIFADDRPREF:
    289 		if (ifp == NULL)
    290 			return EINVAL;
    291 		return ifaddrpref_ioctl(so, cmd, data, ifp);
    292 	}
    293 
    294 	if (ifp == NULL)
    295 		return EOPNOTSUPP;
    296 
    297 	switch (cmd) {
    298 	case SIOCSNDFLUSH_IN6:
    299 	case SIOCSPFXFLUSH_IN6:
    300 	case SIOCSRTRFLUSH_IN6:
    301 	case SIOCSDEFIFACE_IN6:
    302 	case SIOCSIFINFO_FLAGS:
    303 	case SIOCSIFINFO_IN6:
    304 		/* Privileged. */
    305 		/* FALLTHROUGH */
    306 	case OSIOCGIFINFO_IN6:
    307 	case SIOCGIFINFO_IN6:
    308 	case SIOCGDRLST_IN6:
    309 	case SIOCGPRLST_IN6:
    310 	case SIOCGNBRINFO_IN6:
    311 	case SIOCGDEFIFACE_IN6:
    312 		return nd6_ioctl(cmd, data, ifp);
    313 	}
    314 
    315 	switch (cmd) {
    316 	case SIOCSIFPREFIX_IN6:
    317 	case SIOCDIFPREFIX_IN6:
    318 	case SIOCAIFPREFIX_IN6:
    319 	case SIOCCIFPREFIX_IN6:
    320 	case SIOCSGIFPREFIX_IN6:
    321 	case SIOCGIFPREFIX_IN6:
    322 		log(LOG_NOTICE,
    323 		    "prefix ioctls are now invalidated. "
    324 		    "please use ifconfig.\n");
    325 		return EOPNOTSUPP;
    326 	}
    327 
    328 	switch (cmd) {
    329 	case SIOCALIFADDR:
    330 	case SIOCDLIFADDR:
    331 		/* Privileged. */
    332 		/* FALLTHROUGH */
    333 	case SIOCGLIFADDR:
    334 		return in6_lifaddr_ioctl(so, cmd, data, ifp);
    335 	}
    336 
    337 	/*
    338 	 * Find address for this interface, if it exists.
    339 	 *
    340 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
    341 	 * only, and used the first interface address as the target of other
    342 	 * operations (without checking ifra_addr).  This was because netinet
    343 	 * code/API assumed at most 1 interface address per interface.
    344 	 * Since IPv6 allows a node to assign multiple addresses
    345 	 * on a single interface, we almost always look and check the
    346 	 * presence of ifra_addr, and reject invalid ones here.
    347 	 * It also decreases duplicated code among SIOC*_IN6 operations.
    348 	 */
    349 	switch (cmd) {
    350 	case SIOCAIFADDR_IN6:
    351 #ifdef OSIOCAIFADDR_IN6
    352 	case OSIOCAIFADDR_IN6:
    353 #endif
    354 #ifdef OSIOCSIFPHYADDR_IN6
    355 	case OSIOCSIFPHYADDR_IN6:
    356 #endif
    357 	case SIOCSIFPHYADDR_IN6:
    358 		sa6 = &ifra->ifra_addr;
    359 		break;
    360 	case SIOCSIFADDR_IN6:
    361 	case SIOCGIFADDR_IN6:
    362 	case SIOCSIFDSTADDR_IN6:
    363 	case SIOCSIFNETMASK_IN6:
    364 	case SIOCGIFDSTADDR_IN6:
    365 	case SIOCGIFNETMASK_IN6:
    366 	case SIOCDIFADDR_IN6:
    367 	case SIOCGIFPSRCADDR_IN6:
    368 	case SIOCGIFPDSTADDR_IN6:
    369 	case SIOCGIFAFLAG_IN6:
    370 	case SIOCSNDFLUSH_IN6:
    371 	case SIOCSPFXFLUSH_IN6:
    372 	case SIOCSRTRFLUSH_IN6:
    373 	case SIOCGIFALIFETIME_IN6:
    374 #ifdef OSIOCGIFALIFETIME_IN6
    375 	case OSIOCGIFALIFETIME_IN6:
    376 #endif
    377 	case SIOCGIFSTAT_IN6:
    378 	case SIOCGIFSTAT_ICMP6:
    379 		sa6 = &ifr->ifr_addr;
    380 		break;
    381 	default:
    382 		sa6 = NULL;
    383 		break;
    384 	}
    385 	if (sa6 && sa6->sin6_family == AF_INET6) {
    386 		if (sa6->sin6_scope_id != 0)
    387 			error = sa6_embedscope(sa6, 0);
    388 		else
    389 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
    390 		if (error != 0)
    391 			return error;
    392 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
    393 	} else
    394 		ia = NULL;
    395 
    396 	switch (cmd) {
    397 	case SIOCSIFADDR_IN6:
    398 	case SIOCSIFDSTADDR_IN6:
    399 	case SIOCSIFNETMASK_IN6:
    400 		/*
    401 		 * Since IPv6 allows a node to assign multiple addresses
    402 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
    403 		 */
    404 		return EINVAL;
    405 
    406 	case SIOCDIFADDR_IN6:
    407 		/*
    408 		 * for IPv4, we look for existing in_ifaddr here to allow
    409 		 * "ifconfig if0 delete" to remove the first IPv4 address on
    410 		 * the interface.  For IPv6, as the spec allows multiple
    411 		 * interface address from the day one, we consider "remove the
    412 		 * first one" semantics to be not preferable.
    413 		 */
    414 		if (ia == NULL)
    415 			return EADDRNOTAVAIL;
    416 		/* FALLTHROUGH */
    417 #ifdef OSIOCAIFADDR_IN6
    418 	case OSIOCAIFADDR_IN6:
    419 #endif
    420 	case SIOCAIFADDR_IN6:
    421 		/*
    422 		 * We always require users to specify a valid IPv6 address for
    423 		 * the corresponding operation.
    424 		 */
    425 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
    426 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
    427 			return EAFNOSUPPORT;
    428 		/* Privileged. */
    429 
    430 		break;
    431 
    432 	case SIOCGIFADDR_IN6:
    433 		/* This interface is basically deprecated. use SIOCGIFCONF. */
    434 		/* FALLTHROUGH */
    435 	case SIOCGIFAFLAG_IN6:
    436 	case SIOCGIFNETMASK_IN6:
    437 	case SIOCGIFDSTADDR_IN6:
    438 	case SIOCGIFALIFETIME_IN6:
    439 #ifdef OSIOCGIFALIFETIME_IN6
    440 	case OSIOCGIFALIFETIME_IN6:
    441 #endif
    442 		/* must think again about its semantics */
    443 		if (ia == NULL)
    444 			return EADDRNOTAVAIL;
    445 		break;
    446 	}
    447 
    448 	switch (cmd) {
    449 
    450 	case SIOCGIFADDR_IN6:
    451 		ifr->ifr_addr = ia->ia_addr;
    452 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
    453 			return error;
    454 		break;
    455 
    456 	case SIOCGIFDSTADDR_IN6:
    457 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
    458 			return EINVAL;
    459 		/*
    460 		 * XXX: should we check if ifa_dstaddr is NULL and return
    461 		 * an error?
    462 		 */
    463 		ifr->ifr_dstaddr = ia->ia_dstaddr;
    464 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
    465 			return error;
    466 		break;
    467 
    468 	case SIOCGIFNETMASK_IN6:
    469 		ifr->ifr_addr = ia->ia_prefixmask;
    470 		break;
    471 
    472 	case SIOCGIFAFLAG_IN6:
    473 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
    474 		break;
    475 
    476 	case SIOCGIFSTAT_IN6:
    477 		if (ifp == NULL)
    478 			return EINVAL;
    479 		memset(&ifr->ifr_ifru.ifru_stat, 0,
    480 		    sizeof(ifr->ifr_ifru.ifru_stat));
    481 		ifr->ifr_ifru.ifru_stat =
    482 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
    483 		break;
    484 
    485 	case SIOCGIFSTAT_ICMP6:
    486 		if (ifp == NULL)
    487 			return EINVAL;
    488 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
    489 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
    490 		ifr->ifr_ifru.ifru_icmp6stat =
    491 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
    492 		break;
    493 
    494 #ifdef OSIOCGIFALIFETIME_IN6
    495 	case OSIOCGIFALIFETIME_IN6:
    496 #endif
    497 	case SIOCGIFALIFETIME_IN6:
    498 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
    499 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
    500 			time_t maxexpire;
    501 			struct in6_addrlifetime *retlt =
    502 			    &ifr->ifr_ifru.ifru_lifetime;
    503 
    504 			/*
    505 			 * XXX: adjust expiration time assuming time_t is
    506 			 * signed.
    507 			 */
    508 			maxexpire = ((time_t)~0) &
    509 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
    510 			if (ia->ia6_lifetime.ia6t_vltime <
    511 			    maxexpire - ia->ia6_updatetime) {
    512 				retlt->ia6t_expire = ia->ia6_updatetime +
    513 				    ia->ia6_lifetime.ia6t_vltime;
    514 				retlt->ia6t_expire = retlt->ia6t_expire ?
    515 				    time_mono_to_wall(retlt->ia6t_expire) :
    516 				    0;
    517 			} else
    518 				retlt->ia6t_expire = maxexpire;
    519 		}
    520 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
    521 			time_t maxexpire;
    522 			struct in6_addrlifetime *retlt =
    523 			    &ifr->ifr_ifru.ifru_lifetime;
    524 
    525 			/*
    526 			 * XXX: adjust expiration time assuming time_t is
    527 			 * signed.
    528 			 */
    529 			maxexpire = ((time_t)~0) &
    530 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
    531 			if (ia->ia6_lifetime.ia6t_pltime <
    532 			    maxexpire - ia->ia6_updatetime) {
    533 				retlt->ia6t_preferred = ia->ia6_updatetime +
    534 				    ia->ia6_lifetime.ia6t_pltime;
    535 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
    536 				    time_mono_to_wall(retlt->ia6t_preferred) :
    537 				    0;
    538 			} else
    539 				retlt->ia6t_preferred = maxexpire;
    540 		}
    541 #ifdef OSIOCFIFALIFETIME_IN6
    542 		if (cmd == OSIOCFIFALIFETIME_IN6)
    543 			in6_addrlifetime_to_in6_addrlifetime50(
    544 			    &ifr->ifru.ifru_lifetime);
    545 #endif
    546 		break;
    547 
    548 #ifdef OSIOCAIFADDR_IN6
    549 	case OSIOCAIFADDR_IN6:
    550 		in6_aliasreq50_to_in6_aliasreq(ifra);
    551 		/*FALLTHROUGH*/
    552 #endif
    553 	case SIOCAIFADDR_IN6:
    554 	{
    555 		int i;
    556 		struct nd_prefixctl prc0;
    557 		struct nd_prefix *pr;
    558 		struct in6_addrlifetime *lt;
    559 
    560 		/* reject read-only flags */
    561 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
    562 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
    563 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
    564 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
    565 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
    566 			return EINVAL;
    567 		}
    568 		/*
    569 		 * ia6t_expire and ia6t_preferred won't be used for now,
    570 		 * so just in case.
    571 		 */
    572 		lt = &ifra->ifra_lifetime;
    573 		if (lt->ia6t_expire != 0)
    574 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
    575 		if (lt->ia6t_preferred != 0)
    576 			lt->ia6t_preferred =
    577 			    time_wall_to_mono(lt->ia6t_preferred);
    578 		/*
    579 		 * first, make or update the interface address structure,
    580 		 * and link it to the list.
    581 		 */
    582 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
    583 			return error;
    584 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
    585 		    == NULL) {
    586 		    	/*
    587 			 * this can happen when the user specify the 0 valid
    588 			 * lifetime.
    589 			 */
    590 			break;
    591 		}
    592 
    593 		/*
    594 		 * then, make the prefix on-link on the interface.
    595 		 * XXX: we'd rather create the prefix before the address, but
    596 		 * we need at least one address to install the corresponding
    597 		 * interface route, so we configure the address first.
    598 		 */
    599 
    600 		/*
    601 		 * convert mask to prefix length (prefixmask has already
    602 		 * been validated in in6_update_ifa().
    603 		 */
    604 		memset(&prc0, 0, sizeof(prc0));
    605 		prc0.ndprc_ifp = ifp;
    606 		prc0.ndprc_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
    607 		    NULL);
    608 		if (prc0.ndprc_plen == 128) {
    609 			break;	/* we don't need to install a host route. */
    610 		}
    611 		prc0.ndprc_prefix = ifra->ifra_addr;
    612 		/* apply the mask for safety. */
    613 		for (i = 0; i < 4; i++) {
    614 			prc0.ndprc_prefix.sin6_addr.s6_addr32[i] &=
    615 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
    616 		}
    617 		/*
    618 		 * XXX: since we don't have an API to set prefix (not address)
    619 		 * lifetimes, we just use the same lifetimes as addresses.
    620 		 * The (temporarily) installed lifetimes can be overridden by
    621 		 * later advertised RAs (when accept_rtadv is non 0), which is
    622 		 * an intended behavior.
    623 		 */
    624 		prc0.ndprc_raf_onlink = 1; /* should be configurable? */
    625 		prc0.ndprc_raf_auto =
    626 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
    627 		prc0.ndprc_vltime = ifra->ifra_lifetime.ia6t_vltime;
    628 		prc0.ndprc_pltime = ifra->ifra_lifetime.ia6t_pltime;
    629 
    630 		/* add the prefix if not yet. */
    631 		if ((pr = nd6_prefix_lookup(&prc0)) == NULL) {
    632 			/*
    633 			 * nd6_prelist_add will install the corresponding
    634 			 * interface route.
    635 			 */
    636 			if ((error = nd6_prelist_add(&prc0, NULL, &pr)) != 0)
    637 				return error;
    638 			if (pr == NULL) {
    639 				log(LOG_ERR, "nd6_prelist_add succeeded but "
    640 				    "no prefix\n");
    641 				return EINVAL; /* XXX panic here? */
    642 			}
    643 		}
    644 
    645 		/* relate the address to the prefix */
    646 		if (ia->ia6_ndpr == NULL) {
    647 			ia->ia6_ndpr = pr;
    648 			pr->ndpr_refcnt++;
    649 
    650 			/*
    651 			 * If this is the first autoconf address from the
    652 			 * prefix, create a temporary address as well
    653 			 * (when required).
    654 			 */
    655 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
    656 			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
    657 				int e;
    658 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
    659 					log(LOG_NOTICE, "in6_control: failed "
    660 					    "to create a temporary address, "
    661 					    "errno=%d\n", e);
    662 				}
    663 			}
    664 		}
    665 
    666 		/*
    667 		 * this might affect the status of autoconfigured addresses,
    668 		 * that is, this address might make other addresses detached.
    669 		 */
    670 		pfxlist_onlink_check();
    671 
    672 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCAIFADDR_IN6,
    673 		    ifp, PFIL_IFADDR);
    674 		break;
    675 	}
    676 
    677 	case SIOCDIFADDR_IN6:
    678 	{
    679 		struct nd_prefix *pr;
    680 
    681 		/*
    682 		 * If the address being deleted is the only one that owns
    683 		 * the corresponding prefix, expire the prefix as well.
    684 		 * XXX: theoretically, we don't have to worry about such
    685 		 * relationship, since we separate the address management
    686 		 * and the prefix management.  We do this, however, to provide
    687 		 * as much backward compatibility as possible in terms of
    688 		 * the ioctl operation.
    689 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
    690 		 */
    691 		pr = ia->ia6_ndpr;
    692 		in6_purgeaddr(&ia->ia_ifa);
    693 		if (pr && pr->ndpr_refcnt == 0)
    694 			prelist_remove(pr);
    695 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCDIFADDR_IN6,
    696 		    ifp, PFIL_IFADDR);
    697 		break;
    698 	}
    699 
    700 	default:
    701 		return ENOTTY;
    702 	}
    703 
    704 	return 0;
    705 }
    706 
    707 int
    708 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
    709 {
    710 	int error, s;
    711 
    712 	switch (cmd) {
    713 	case SIOCSNDFLUSH_IN6:
    714 	case SIOCSPFXFLUSH_IN6:
    715 	case SIOCSRTRFLUSH_IN6:
    716 	case SIOCSDEFIFACE_IN6:
    717 	case SIOCSIFINFO_FLAGS:
    718 	case SIOCSIFINFO_IN6:
    719 
    720 	case SIOCALIFADDR:
    721 	case SIOCDLIFADDR:
    722 
    723 	case SIOCDIFADDR_IN6:
    724 #ifdef OSIOCAIFADDR_IN6
    725 	case OSIOCAIFADDR_IN6:
    726 #endif
    727 	case SIOCAIFADDR_IN6:
    728 		if (kauth_authorize_network(curlwp->l_cred,
    729 		    KAUTH_NETWORK_SOCKET,
    730 		    KAUTH_REQ_NETWORK_SOCKET_SETPRIV,
    731 		    so, NULL, NULL))
    732 			return EPERM;
    733 		break;
    734 	}
    735 
    736 	s = splnet();
    737 	error = in6_control1(so , cmd, data, ifp);
    738 	splx(s);
    739 	return error;
    740 }
    741 
    742 /*
    743  * Update parameters of an IPv6 interface address.
    744  * If necessary, a new entry is created and linked into address chains.
    745  * This function is separated from in6_control().
    746  * XXX: should this be performed under splnet()?
    747  */
    748 static int
    749 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
    750     struct in6_ifaddr *ia, int flags)
    751 {
    752 	int error = 0, hostIsNew = 0, plen = -1;
    753 	struct in6_ifaddr *oia;
    754 	struct sockaddr_in6 dst6;
    755 	struct in6_addrlifetime *lt;
    756 	struct in6_multi_mship *imm;
    757 	struct in6_multi *in6m_sol;
    758 	struct rtentry *rt;
    759 	int dad_delay, was_tentative;
    760 
    761 	in6m_sol = NULL;
    762 
    763 	/* Validate parameters */
    764 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
    765 		return EINVAL;
    766 
    767 	/*
    768 	 * The destination address for a p2p link must have a family
    769 	 * of AF_UNSPEC or AF_INET6.
    770 	 */
    771 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
    772 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
    773 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
    774 		return EAFNOSUPPORT;
    775 	/*
    776 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
    777 	 * does not carry fields other than sin6_len.
    778 	 */
    779 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
    780 		return EINVAL;
    781 	/*
    782 	 * Because the IPv6 address architecture is classless, we require
    783 	 * users to specify a (non 0) prefix length (mask) for a new address.
    784 	 * We also require the prefix (when specified) mask is valid, and thus
    785 	 * reject a non-consecutive mask.
    786 	 */
    787 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
    788 		return EINVAL;
    789 	if (ifra->ifra_prefixmask.sin6_len != 0) {
    790 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
    791 		    (u_char *)&ifra->ifra_prefixmask +
    792 		    ifra->ifra_prefixmask.sin6_len);
    793 		if (plen <= 0)
    794 			return EINVAL;
    795 	} else {
    796 		/*
    797 		 * In this case, ia must not be NULL.  We just use its prefix
    798 		 * length.
    799 		 */
    800 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
    801 	}
    802 	/*
    803 	 * If the destination address on a p2p interface is specified,
    804 	 * and the address is a scoped one, validate/set the scope
    805 	 * zone identifier.
    806 	 */
    807 	dst6 = ifra->ifra_dstaddr;
    808 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
    809 	    (dst6.sin6_family == AF_INET6)) {
    810 		struct in6_addr in6_tmp;
    811 		u_int32_t zoneid;
    812 
    813 		in6_tmp = dst6.sin6_addr;
    814 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
    815 			return EINVAL; /* XXX: should be impossible */
    816 
    817 		if (dst6.sin6_scope_id != 0) {
    818 			if (dst6.sin6_scope_id != zoneid)
    819 				return EINVAL;
    820 		} else		/* user omit to specify the ID. */
    821 			dst6.sin6_scope_id = zoneid;
    822 
    823 		/* convert into the internal form */
    824 		if (sa6_embedscope(&dst6, 0))
    825 			return EINVAL; /* XXX: should be impossible */
    826 	}
    827 	/*
    828 	 * The destination address can be specified only for a p2p or a
    829 	 * loopback interface.  If specified, the corresponding prefix length
    830 	 * must be 128.
    831 	 */
    832 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
    833 #ifdef FORCE_P2PPLEN
    834 		int i;
    835 #endif
    836 
    837 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
    838 			/* XXX: noisy message */
    839 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
    840 			    "be specified for a p2p or a loopback IF only\n"));
    841 			return EINVAL;
    842 		}
    843 		if (plen != 128) {
    844 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
    845 			    "be 128 when dstaddr is specified\n"));
    846 #ifdef FORCE_P2PPLEN
    847 			/*
    848 			 * To be compatible with old configurations,
    849 			 * such as ifconfig gif0 inet6 2001::1 2001::2
    850 			 * prefixlen 126, we override the specified
    851 			 * prefixmask as if the prefix length was 128.
    852 			 */
    853 			ifra->ifra_prefixmask.sin6_len =
    854 			    sizeof(struct sockaddr_in6);
    855 			for (i = 0; i < 4; i++)
    856 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
    857 				    0xffffffff;
    858 			plen = 128;
    859 #else
    860 			return EINVAL;
    861 #endif
    862 		}
    863 	}
    864 	/* lifetime consistency check */
    865 	lt = &ifra->ifra_lifetime;
    866 	if (lt->ia6t_pltime > lt->ia6t_vltime)
    867 		return EINVAL;
    868 	if (lt->ia6t_vltime == 0) {
    869 		/*
    870 		 * the following log might be noisy, but this is a typical
    871 		 * configuration mistake or a tool's bug.
    872 		 */
    873 		nd6log((LOG_INFO,
    874 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
    875 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr)));
    876 
    877 		if (ia == NULL)
    878 			return 0; /* there's nothing to do */
    879 	}
    880 
    881 	/*
    882 	 * If this is a new address, allocate a new ifaddr and link it
    883 	 * into chains.
    884 	 */
    885 	if (ia == NULL) {
    886 		hostIsNew = 1;
    887 		/*
    888 		 * When in6_update_ifa() is called in a process of a received
    889 		 * RA, it is called under an interrupt context.  So, we should
    890 		 * call malloc with M_NOWAIT.
    891 		 */
    892 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
    893 		    M_NOWAIT);
    894 		if (ia == NULL)
    895 			return ENOBUFS;
    896 		memset(ia, 0, sizeof(*ia));
    897 		LIST_INIT(&ia->ia6_memberships);
    898 		/* Initialize the address and masks, and put time stamp */
    899 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
    900 		ia->ia_addr.sin6_family = AF_INET6;
    901 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
    902 		ia->ia6_createtime = time_uptime;
    903 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
    904 			/*
    905 			 * XXX: some functions expect that ifa_dstaddr is not
    906 			 * NULL for p2p interfaces.
    907 			 */
    908 			ia->ia_ifa.ifa_dstaddr =
    909 			    (struct sockaddr *)&ia->ia_dstaddr;
    910 		} else {
    911 			ia->ia_ifa.ifa_dstaddr = NULL;
    912 		}
    913 		ia->ia_ifa.ifa_netmask =
    914 		    (struct sockaddr *)&ia->ia_prefixmask;
    915 
    916 		ia->ia_ifp = ifp;
    917 		if ((oia = in6_ifaddr) != NULL) {
    918 			for ( ; oia->ia_next; oia = oia->ia_next)
    919 				continue;
    920 			oia->ia_next = ia;
    921 		} else
    922 			in6_ifaddr = ia;
    923 		/* gain a refcnt for the link from in6_ifaddr */
    924 		ifaref(&ia->ia_ifa);
    925 
    926 		ifa_insert(ifp, &ia->ia_ifa);
    927 	}
    928 
    929 	/* update timestamp */
    930 	ia->ia6_updatetime = time_uptime;
    931 
    932 	/* set prefix mask */
    933 	if (ifra->ifra_prefixmask.sin6_len) {
    934 		/*
    935 		 * We prohibit changing the prefix length of an existing
    936 		 * address, because
    937 		 * + such an operation should be rare in IPv6, and
    938 		 * + the operation would confuse prefix management.
    939 		 */
    940 		if (ia->ia_prefixmask.sin6_len &&
    941 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
    942 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
    943 			    " existing (%s) address should not be changed\n",
    944 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
    945 			error = EINVAL;
    946 			goto unlink;
    947 		}
    948 		ia->ia_prefixmask = ifra->ifra_prefixmask;
    949 	}
    950 
    951 	/*
    952 	 * If a new destination address is specified, scrub the old one and
    953 	 * install the new destination.  Note that the interface must be
    954 	 * p2p or loopback (see the check above.)
    955 	 */
    956 	if (dst6.sin6_family == AF_INET6 &&
    957 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
    958 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
    959 		    rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0) {
    960 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
    961 			    "a route to the old destination: %s\n",
    962 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
    963 			/* proceed anyway... */
    964 		} else
    965 			ia->ia_flags &= ~IFA_ROUTE;
    966 		ia->ia_dstaddr = dst6;
    967 	}
    968 
    969 	/*
    970 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
    971 	 * to see if the address is deprecated or invalidated, but initialize
    972 	 * these members for applications.
    973 	 */
    974 	ia->ia6_lifetime = ifra->ifra_lifetime;
    975 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
    976 		ia->ia6_lifetime.ia6t_expire =
    977 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
    978 	} else
    979 		ia->ia6_lifetime.ia6t_expire = 0;
    980 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
    981 		ia->ia6_lifetime.ia6t_preferred =
    982 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
    983 	} else
    984 		ia->ia6_lifetime.ia6t_preferred = 0;
    985 
    986 	/*
    987 	 * configure address flags.
    988 	 * We need to preserve tentative state so DAD works if
    989 	 * something adds the same address before DAD finishes.
    990 	 */
    991 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
    992 	ia->ia6_flags = ifra->ifra_flags;
    993 
    994 	/*
    995 	 * Make the address tentative before joining multicast addresses,
    996 	 * so that corresponding MLD responses would not have a tentative
    997 	 * source address.
    998 	 */
    999 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
   1000 	if (ifp->if_link_state == LINK_STATE_DOWN) {
   1001 		ia->ia6_flags |= IN6_IFF_DETACHED;
   1002 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
   1003 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp))
   1004 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
   1005 
   1006 	/*
   1007 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
   1008 	 * userland, make it deprecated.
   1009 	 */
   1010 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
   1011 		ia->ia6_lifetime.ia6t_pltime = 0;
   1012 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
   1013 	}
   1014 
   1015 	/* reset the interface and routing table appropriately. */
   1016 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
   1017 		goto unlink;
   1018 	/*
   1019 	 * We are done if we have simply modified an existing address.
   1020 	 */
   1021 	if (!hostIsNew)
   1022 		return error;
   1023 
   1024 	/*
   1025 	 * Beyond this point, we should call in6_purgeaddr upon an error,
   1026 	 * not just go to unlink.
   1027 	 */
   1028 
   1029 	/* join necessary multicast groups */
   1030 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
   1031 		struct sockaddr_in6 mltaddr, mltmask;
   1032 		struct in6_addr llsol;
   1033 
   1034 		/* join solicited multicast addr for new host id */
   1035 		memset(&llsol, 0, sizeof(struct in6_addr));
   1036 		llsol.s6_addr16[0] = htons(0xff02);
   1037 		llsol.s6_addr32[1] = 0;
   1038 		llsol.s6_addr32[2] = htonl(1);
   1039 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
   1040 		llsol.s6_addr8[12] = 0xff;
   1041 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
   1042 			/* XXX: should not happen */
   1043 			log(LOG_ERR, "in6_update_ifa: "
   1044 			    "in6_setscope failed\n");
   1045 			goto cleanup;
   1046 		}
   1047 		dad_delay = 0;
   1048 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
   1049 			/*
   1050 			 * We need a random delay for DAD on the address
   1051 			 * being configured.  It also means delaying
   1052 			 * transmission of the corresponding MLD report to
   1053 			 * avoid report collision.
   1054 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
   1055 			 */
   1056 			dad_delay = cprng_fast32() %
   1057 			    (MAX_RTR_SOLICITATION_DELAY * hz);
   1058 		}
   1059 
   1060 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
   1061 		/* join solicited multicast addr for new host id */
   1062 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
   1063 		if (!imm) {
   1064 			nd6log((LOG_ERR,
   1065 			    "in6_update_ifa: addmulti "
   1066 			    "failed for %s on %s (errno=%d)\n",
   1067 			    ip6_sprintf(&llsol), if_name(ifp), error));
   1068 			goto cleanup;
   1069 		}
   1070 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
   1071 		in6m_sol = imm->i6mm_maddr;
   1072 
   1073 		sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
   1074 
   1075 		/*
   1076 		 * join link-local all-nodes address
   1077 		 */
   1078 		sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
   1079 		    0, 0, 0);
   1080 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
   1081 			goto cleanup; /* XXX: should not fail */
   1082 
   1083 		/*
   1084 		 * XXX: do we really need this automatic routes?
   1085 		 * We should probably reconsider this stuff.  Most applications
   1086 		 * actually do not need the routes, since they usually specify
   1087 		 * the outgoing interface.
   1088 		 */
   1089 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
   1090 		if (rt) {
   1091 			if (memcmp(&mltaddr.sin6_addr,
   1092 			    &satocsin6(rt_getkey(rt))->sin6_addr,
   1093 			    MLTMASK_LEN)) {
   1094 				rtfree(rt);
   1095 				rt = NULL;
   1096 			} else if (rt->rt_ifp != ifp) {
   1097 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
   1098 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
   1099 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
   1100 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
   1101 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
   1102 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
   1103 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
   1104 				rt_replace_ifa(rt, &ia->ia_ifa);
   1105 				rt->rt_ifp = ifp;
   1106 			}
   1107 		}
   1108 		if (!rt) {
   1109 			struct rt_addrinfo info;
   1110 
   1111 			memset(&info, 0, sizeof(info));
   1112 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
   1113 			info.rti_info[RTAX_GATEWAY] =
   1114 			    (struct sockaddr *)&ia->ia_addr;
   1115 			info.rti_info[RTAX_NETMASK] =
   1116 			    (struct sockaddr *)&mltmask;
   1117 			info.rti_info[RTAX_IFA] =
   1118 			    (struct sockaddr *)&ia->ia_addr;
   1119 			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
   1120 			info.rti_flags = RTF_UP | RTF_CLONING;
   1121 			error = rtrequest1(RTM_ADD, &info, NULL);
   1122 			if (error)
   1123 				goto cleanup;
   1124 		} else {
   1125 			rtfree(rt);
   1126 		}
   1127 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
   1128 		if (!imm) {
   1129 			nd6log((LOG_WARNING,
   1130 			    "in6_update_ifa: addmulti failed for "
   1131 			    "%s on %s (errno=%d)\n",
   1132 			    ip6_sprintf(&mltaddr.sin6_addr),
   1133 			    if_name(ifp), error));
   1134 			goto cleanup;
   1135 		}
   1136 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
   1137 
   1138 		/*
   1139 		 * join node information group address
   1140 		 */
   1141 		dad_delay = 0;
   1142 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
   1143 			/*
   1144 			 * The spec doesn't say anything about delay for this
   1145 			 * group, but the same logic should apply.
   1146 			 */
   1147 			dad_delay = cprng_fast32() %
   1148 			    (MAX_RTR_SOLICITATION_DELAY * hz);
   1149 		}
   1150 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
   1151 			;
   1152 		else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
   1153 		          dad_delay)) == NULL) { /* XXX jinmei */
   1154 			nd6log((LOG_WARNING, "in6_update_ifa: "
   1155 			    "addmulti failed for %s on %s (errno=%d)\n",
   1156 			    ip6_sprintf(&mltaddr.sin6_addr),
   1157 			    if_name(ifp), error));
   1158 			/* XXX not very fatal, go on... */
   1159 		} else {
   1160 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
   1161 		}
   1162 
   1163 
   1164 		/*
   1165 		 * join interface-local all-nodes address.
   1166 		 * (ff01::1%ifN, and ff01::%ifN/32)
   1167 		 */
   1168 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
   1169 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
   1170 			goto cleanup; /* XXX: should not fail */
   1171 
   1172 		/* XXX: again, do we really need the route? */
   1173 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
   1174 		if (rt) {
   1175 			/* 32bit came from "mltmask" */
   1176 			if (memcmp(&mltaddr.sin6_addr,
   1177 			    &satocsin6(rt_getkey(rt))->sin6_addr,
   1178 			    32 / NBBY)) {
   1179 				rtfree(rt);
   1180 				rt = NULL;
   1181 			} else if (rt->rt_ifp != ifp) {
   1182 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
   1183 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
   1184 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
   1185 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
   1186 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
   1187 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
   1188 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
   1189 				rt_replace_ifa(rt, &ia->ia_ifa);
   1190 				rt->rt_ifp = ifp;
   1191 			}
   1192 		}
   1193 		if (!rt) {
   1194 			struct rt_addrinfo info;
   1195 
   1196 			memset(&info, 0, sizeof(info));
   1197 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
   1198 			info.rti_info[RTAX_GATEWAY] =
   1199 			    (struct sockaddr *)&ia->ia_addr;
   1200 			info.rti_info[RTAX_NETMASK] =
   1201 			    (struct sockaddr *)&mltmask;
   1202 			info.rti_info[RTAX_IFA] =
   1203 			    (struct sockaddr *)&ia->ia_addr;
   1204 			info.rti_flags = RTF_UP | RTF_CLONING;
   1205 			error = rtrequest1(RTM_ADD, &info, NULL);
   1206 			if (error)
   1207 				goto cleanup;
   1208 #undef	MLTMASK_LEN
   1209 		} else {
   1210 			rtfree(rt);
   1211 		}
   1212 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
   1213 		if (!imm) {
   1214 			nd6log((LOG_WARNING, "in6_update_ifa: "
   1215 			    "addmulti failed for %s on %s (errno=%d)\n",
   1216 			    ip6_sprintf(&mltaddr.sin6_addr),
   1217 			    if_name(ifp), error));
   1218 			goto cleanup;
   1219 		} else {
   1220 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
   1221 		}
   1222 	}
   1223 
   1224 	/*
   1225 	 * Perform DAD, if needed.
   1226 	 * XXX It may be of use, if we can administratively
   1227 	 * disable DAD.
   1228 	 */
   1229 	if (hostIsNew && if_do_dad(ifp) &&
   1230 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
   1231 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
   1232 	{
   1233 		int mindelay, maxdelay;
   1234 
   1235 		dad_delay = 0;
   1236 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
   1237 			/*
   1238 			 * We need to impose a delay before sending an NS
   1239 			 * for DAD.  Check if we also needed a delay for the
   1240 			 * corresponding MLD message.  If we did, the delay
   1241 			 * should be larger than the MLD delay (this could be
   1242 			 * relaxed a bit, but this simple logic is at least
   1243 			 * safe).
   1244 			 */
   1245 			mindelay = 0;
   1246 			if (in6m_sol != NULL &&
   1247 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
   1248 				mindelay = in6m_sol->in6m_timer;
   1249 			}
   1250 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
   1251 			if (maxdelay - mindelay == 0)
   1252 				dad_delay = 0;
   1253 			else {
   1254 				dad_delay =
   1255 				    (cprng_fast32() % (maxdelay - mindelay)) +
   1256 				    mindelay;
   1257 			}
   1258 		}
   1259 		/* +1 ensures callout is always used */
   1260 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
   1261 	}
   1262 
   1263 	return error;
   1264 
   1265   unlink:
   1266 	/*
   1267 	 * XXX: if a change of an existing address failed, keep the entry
   1268 	 * anyway.
   1269 	 */
   1270 	if (hostIsNew)
   1271 		in6_unlink_ifa(ia, ifp);
   1272 	return error;
   1273 
   1274   cleanup:
   1275 	in6_purgeaddr(&ia->ia_ifa);
   1276 	return error;
   1277 }
   1278 
   1279 int
   1280 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
   1281     struct in6_ifaddr *ia, int flags)
   1282 {
   1283 	int rc, s;
   1284 
   1285 	s = splnet();
   1286 	rc = in6_update_ifa1(ifp, ifra, ia, flags);
   1287 	splx(s);
   1288 	return rc;
   1289 }
   1290 
   1291 void
   1292 in6_purgeaddr(struct ifaddr *ifa)
   1293 {
   1294 	struct ifnet *ifp = ifa->ifa_ifp;
   1295 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
   1296 	struct in6_multi_mship *imm;
   1297 
   1298 	/* stop DAD processing */
   1299 	nd6_dad_stop(ifa);
   1300 
   1301 	/*
   1302 	 * delete route to the destination of the address being purged.
   1303 	 * The interface must be p2p or loopback in this case.
   1304 	 */
   1305 	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
   1306 		int e;
   1307 
   1308 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
   1309 		    != 0) {
   1310 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
   1311 			    "a route to the p2p destination: %s on %s, "
   1312 			    "errno=%d\n",
   1313 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
   1314 			    e);
   1315 			/* proceed anyway... */
   1316 		} else
   1317 			ia->ia_flags &= ~IFA_ROUTE;
   1318 	}
   1319 
   1320 	/* Remove ownaddr's loopback rtentry, if it exists. */
   1321 	in6_ifremlocal(&(ia->ia_ifa));
   1322 
   1323 	/*
   1324 	 * leave from multicast groups we have joined for the interface
   1325 	 */
   1326 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
   1327 		LIST_REMOVE(imm, i6mm_chain);
   1328 		in6_leavegroup(imm);
   1329 	}
   1330 
   1331 	in6_unlink_ifa(ia, ifp);
   1332 }
   1333 
   1334 static void
   1335 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
   1336 {
   1337 	struct in6_ifaddr *oia;
   1338 	int	s = splnet();
   1339 
   1340 	ifa_remove(ifp, &ia->ia_ifa);
   1341 
   1342 	oia = ia;
   1343 	if (oia == (ia = in6_ifaddr))
   1344 		in6_ifaddr = ia->ia_next;
   1345 	else {
   1346 		while (ia->ia_next && (ia->ia_next != oia))
   1347 			ia = ia->ia_next;
   1348 		if (ia->ia_next)
   1349 			ia->ia_next = oia->ia_next;
   1350 		else {
   1351 			/* search failed */
   1352 			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
   1353 		}
   1354 	}
   1355 
   1356 	/*
   1357 	 * XXX thorpej (at) NetBSD.org -- if the interface is going
   1358 	 * XXX away, don't save the multicast entries, delete them!
   1359 	 */
   1360 	if (LIST_EMPTY(&oia->ia6_multiaddrs))
   1361 		;
   1362 	else if (oia->ia_ifa.ifa_ifp->if_output == if_nulloutput) {
   1363 		struct in6_multi *in6m, *next;
   1364 
   1365 		for (in6m = LIST_FIRST(&oia->ia6_multiaddrs); in6m != NULL;
   1366 		     in6m = next) {
   1367 			next = LIST_NEXT(in6m, in6m_entry);
   1368 			in6_delmulti(in6m);
   1369 		}
   1370 	} else
   1371 		in6_savemkludge(oia);
   1372 
   1373 	/*
   1374 	 * Release the reference to the base prefix.  There should be a
   1375 	 * positive reference.
   1376 	 */
   1377 	if (oia->ia6_ndpr == NULL) {
   1378 		nd6log((LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
   1379 		    "%p has no prefix\n", oia));
   1380 	} else {
   1381 		oia->ia6_ndpr->ndpr_refcnt--;
   1382 		oia->ia6_ndpr = NULL;
   1383 	}
   1384 
   1385 	/*
   1386 	 * Also, if the address being removed is autoconf'ed, call
   1387 	 * pfxlist_onlink_check() since the release might affect the status of
   1388 	 * other (detached) addresses.
   1389 	 */
   1390 	if ((oia->ia6_flags & IN6_IFF_AUTOCONF) != 0)
   1391 		pfxlist_onlink_check();
   1392 
   1393 	/*
   1394 	 * release another refcnt for the link from in6_ifaddr.
   1395 	 * Note that we should decrement the refcnt at least once for all *BSD.
   1396 	 */
   1397 	ifafree(&oia->ia_ifa);
   1398 
   1399 	splx(s);
   1400 }
   1401 
   1402 void
   1403 in6_purgeif(struct ifnet *ifp)
   1404 {
   1405 	if_purgeaddrs(ifp, AF_INET6, in6_purgeaddr);
   1406 
   1407 	in6_ifdetach(ifp);
   1408 }
   1409 
   1410 /*
   1411  * SIOC[GAD]LIFADDR.
   1412  *	SIOCGLIFADDR: get first address. (?)
   1413  *	SIOCGLIFADDR with IFLR_PREFIX:
   1414  *		get first address that matches the specified prefix.
   1415  *	SIOCALIFADDR: add the specified address.
   1416  *	SIOCALIFADDR with IFLR_PREFIX:
   1417  *		add the specified prefix, filling hostid part from
   1418  *		the first link-local address.  prefixlen must be <= 64.
   1419  *	SIOCDLIFADDR: delete the specified address.
   1420  *	SIOCDLIFADDR with IFLR_PREFIX:
   1421  *		delete the first address that matches the specified prefix.
   1422  * return values:
   1423  *	EINVAL on invalid parameters
   1424  *	EADDRNOTAVAIL on prefix match failed/specified address not found
   1425  *	other values may be returned from in6_ioctl()
   1426  *
   1427  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
   1428  * this is to accommodate address naming scheme other than RFC2374,
   1429  * in the future.
   1430  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
   1431  * address encoding scheme. (see figure on page 8)
   1432  */
   1433 static int
   1434 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
   1435 	struct ifnet *ifp)
   1436 {
   1437 	struct in6_ifaddr *ia;
   1438 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
   1439 	struct ifaddr *ifa;
   1440 	struct sockaddr *sa;
   1441 
   1442 	/* sanity checks */
   1443 	if (!data || !ifp) {
   1444 		panic("invalid argument to in6_lifaddr_ioctl");
   1445 		/* NOTREACHED */
   1446 	}
   1447 
   1448 	switch (cmd) {
   1449 	case SIOCGLIFADDR:
   1450 		/* address must be specified on GET with IFLR_PREFIX */
   1451 		if ((iflr->flags & IFLR_PREFIX) == 0)
   1452 			break;
   1453 		/* FALLTHROUGH */
   1454 	case SIOCALIFADDR:
   1455 	case SIOCDLIFADDR:
   1456 		/* address must be specified on ADD and DELETE */
   1457 		sa = (struct sockaddr *)&iflr->addr;
   1458 		if (sa->sa_family != AF_INET6)
   1459 			return EINVAL;
   1460 		if (sa->sa_len != sizeof(struct sockaddr_in6))
   1461 			return EINVAL;
   1462 		/* XXX need improvement */
   1463 		sa = (struct sockaddr *)&iflr->dstaddr;
   1464 		if (sa->sa_family && sa->sa_family != AF_INET6)
   1465 			return EINVAL;
   1466 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
   1467 			return EINVAL;
   1468 		break;
   1469 	default: /* shouldn't happen */
   1470 #if 0
   1471 		panic("invalid cmd to in6_lifaddr_ioctl");
   1472 		/* NOTREACHED */
   1473 #else
   1474 		return EOPNOTSUPP;
   1475 #endif
   1476 	}
   1477 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
   1478 		return EINVAL;
   1479 
   1480 	switch (cmd) {
   1481 	case SIOCALIFADDR:
   1482 	    {
   1483 		struct in6_aliasreq ifra;
   1484 		struct in6_addr *xhostid = NULL;
   1485 		int prefixlen;
   1486 
   1487 		if ((iflr->flags & IFLR_PREFIX) != 0) {
   1488 			struct sockaddr_in6 *sin6;
   1489 
   1490 			/*
   1491 			 * xhostid is to fill in the hostid part of the
   1492 			 * address.  xhostid points to the first link-local
   1493 			 * address attached to the interface.
   1494 			 */
   1495 			ia = in6ifa_ifpforlinklocal(ifp, 0);
   1496 			if (ia == NULL)
   1497 				return EADDRNOTAVAIL;
   1498 			xhostid = IFA_IN6(&ia->ia_ifa);
   1499 
   1500 		 	/* prefixlen must be <= 64. */
   1501 			if (64 < iflr->prefixlen)
   1502 				return EINVAL;
   1503 			prefixlen = iflr->prefixlen;
   1504 
   1505 			/* hostid part must be zero. */
   1506 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
   1507 			if (sin6->sin6_addr.s6_addr32[2] != 0
   1508 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
   1509 				return EINVAL;
   1510 			}
   1511 		} else
   1512 			prefixlen = iflr->prefixlen;
   1513 
   1514 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
   1515 		memset(&ifra, 0, sizeof(ifra));
   1516 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
   1517 
   1518 		memcpy(&ifra.ifra_addr, &iflr->addr,
   1519 		    ((struct sockaddr *)&iflr->addr)->sa_len);
   1520 		if (xhostid) {
   1521 			/* fill in hostid part */
   1522 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
   1523 			    xhostid->s6_addr32[2];
   1524 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
   1525 			    xhostid->s6_addr32[3];
   1526 		}
   1527 
   1528 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
   1529 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
   1530 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
   1531 			if (xhostid) {
   1532 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
   1533 				    xhostid->s6_addr32[2];
   1534 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
   1535 				    xhostid->s6_addr32[3];
   1536 			}
   1537 		}
   1538 
   1539 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
   1540 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
   1541 
   1542 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
   1543 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
   1544 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
   1545 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
   1546 	    }
   1547 	case SIOCGLIFADDR:
   1548 	case SIOCDLIFADDR:
   1549 	    {
   1550 		struct in6_addr mask, candidate, match;
   1551 		struct sockaddr_in6 *sin6;
   1552 		int cmp;
   1553 
   1554 		memset(&mask, 0, sizeof(mask));
   1555 		if (iflr->flags & IFLR_PREFIX) {
   1556 			/* lookup a prefix rather than address. */
   1557 			in6_prefixlen2mask(&mask, iflr->prefixlen);
   1558 
   1559 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
   1560 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
   1561 			match.s6_addr32[0] &= mask.s6_addr32[0];
   1562 			match.s6_addr32[1] &= mask.s6_addr32[1];
   1563 			match.s6_addr32[2] &= mask.s6_addr32[2];
   1564 			match.s6_addr32[3] &= mask.s6_addr32[3];
   1565 
   1566 			/* if you set extra bits, that's wrong */
   1567 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
   1568 				return EINVAL;
   1569 
   1570 			cmp = 1;
   1571 		} else {
   1572 			if (cmd == SIOCGLIFADDR) {
   1573 				/* on getting an address, take the 1st match */
   1574 				cmp = 0;	/* XXX */
   1575 			} else {
   1576 				/* on deleting an address, do exact match */
   1577 				in6_prefixlen2mask(&mask, 128);
   1578 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
   1579 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
   1580 
   1581 				cmp = 1;
   1582 			}
   1583 		}
   1584 
   1585 		IFADDR_FOREACH(ifa, ifp) {
   1586 			if (ifa->ifa_addr->sa_family != AF_INET6)
   1587 				continue;
   1588 			if (!cmp)
   1589 				break;
   1590 
   1591 			/*
   1592 			 * XXX: this is adhoc, but is necessary to allow
   1593 			 * a user to specify fe80::/64 (not /10) for a
   1594 			 * link-local address.
   1595 			 */
   1596 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
   1597 			in6_clearscope(&candidate);
   1598 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
   1599 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
   1600 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
   1601 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
   1602 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
   1603 				break;
   1604 		}
   1605 		if (!ifa)
   1606 			return EADDRNOTAVAIL;
   1607 		ia = ifa2ia6(ifa);
   1608 
   1609 		if (cmd == SIOCGLIFADDR) {
   1610 			int error;
   1611 
   1612 			/* fill in the if_laddrreq structure */
   1613 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
   1614 			error = sa6_recoverscope(
   1615 			    (struct sockaddr_in6 *)&iflr->addr);
   1616 			if (error != 0)
   1617 				return error;
   1618 
   1619 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
   1620 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
   1621 				    ia->ia_dstaddr.sin6_len);
   1622 				error = sa6_recoverscope(
   1623 				    (struct sockaddr_in6 *)&iflr->dstaddr);
   1624 				if (error != 0)
   1625 					return error;
   1626 			} else
   1627 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
   1628 
   1629 			iflr->prefixlen =
   1630 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
   1631 
   1632 			iflr->flags = ia->ia6_flags;	/* XXX */
   1633 
   1634 			return 0;
   1635 		} else {
   1636 			struct in6_aliasreq ifra;
   1637 
   1638 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
   1639 			memset(&ifra, 0, sizeof(ifra));
   1640 			memcpy(ifra.ifra_name, iflr->iflr_name,
   1641 			    sizeof(ifra.ifra_name));
   1642 
   1643 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
   1644 			    ia->ia_addr.sin6_len);
   1645 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
   1646 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
   1647 				    ia->ia_dstaddr.sin6_len);
   1648 			} else {
   1649 				memset(&ifra.ifra_dstaddr, 0,
   1650 				    sizeof(ifra.ifra_dstaddr));
   1651 			}
   1652 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
   1653 			    ia->ia_prefixmask.sin6_len);
   1654 
   1655 			ifra.ifra_flags = ia->ia6_flags;
   1656 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
   1657 		}
   1658 	    }
   1659 	}
   1660 
   1661 	return EOPNOTSUPP;	/* just for safety */
   1662 }
   1663 
   1664 /*
   1665  * Initialize an interface's internet6 address
   1666  * and routing table entry.
   1667  */
   1668 static int
   1669 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
   1670 	const struct sockaddr_in6 *sin6, int newhost)
   1671 {
   1672 	int	error = 0, plen, ifacount = 0;
   1673 	int	s = splnet();
   1674 	struct ifaddr *ifa;
   1675 
   1676 	/*
   1677 	 * Give the interface a chance to initialize
   1678 	 * if this is its first address,
   1679 	 * and to validate the address if necessary.
   1680 	 */
   1681 	IFADDR_FOREACH(ifa, ifp) {
   1682 		if (ifa->ifa_addr == NULL)
   1683 			continue;	/* just for safety */
   1684 		if (ifa->ifa_addr->sa_family != AF_INET6)
   1685 			continue;
   1686 		ifacount++;
   1687 	}
   1688 
   1689 	ia->ia_addr = *sin6;
   1690 
   1691 	if (ifacount <= 1 &&
   1692 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
   1693 		splx(s);
   1694 		return error;
   1695 	}
   1696 	splx(s);
   1697 
   1698 	ia->ia_ifa.ifa_metric = ifp->if_metric;
   1699 
   1700 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
   1701 
   1702 	/*
   1703 	 * Special case:
   1704 	 * If the destination address is specified for a point-to-point
   1705 	 * interface, install a route to the destination as an interface
   1706 	 * direct route.
   1707 	 */
   1708 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
   1709 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
   1710 		if ((error = rtinit(&ia->ia_ifa, RTM_ADD,
   1711 				    RTF_UP | RTF_HOST)) != 0)
   1712 			return error;
   1713 		ia->ia_flags |= IFA_ROUTE;
   1714 	}
   1715 
   1716 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
   1717 	if (newhost) {
   1718 		/* set the rtrequest function to create llinfo */
   1719 		if (ifp->if_flags & IFF_POINTOPOINT)
   1720 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
   1721 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
   1722 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
   1723 		in6_ifaddlocal(&ia->ia_ifa);
   1724 	} else {
   1725 		/* Inform the routing socket of new flags/timings */
   1726 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
   1727 	}
   1728 
   1729 	if (ifp->if_flags & IFF_MULTICAST)
   1730 		in6_restoremkludge(ia, ifp);
   1731 
   1732 	return error;
   1733 }
   1734 
   1735 static struct ifaddr *
   1736 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
   1737 {
   1738 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
   1739 		return ifa;
   1740 	return best_ifa;
   1741 }
   1742 
   1743 /*
   1744  * Find an IPv6 interface link-local address specific to an interface.
   1745  */
   1746 struct in6_ifaddr *
   1747 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
   1748 {
   1749 	struct ifaddr *best_ifa = NULL, *ifa;
   1750 
   1751 	IFADDR_FOREACH(ifa, ifp) {
   1752 		if (ifa->ifa_addr == NULL)
   1753 			continue;	/* just for safety */
   1754 		if (ifa->ifa_addr->sa_family != AF_INET6)
   1755 			continue;
   1756 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
   1757 			continue;
   1758 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
   1759 			continue;
   1760 		best_ifa = bestifa(best_ifa, ifa);
   1761 	}
   1762 
   1763 	return (struct in6_ifaddr *)best_ifa;
   1764 }
   1765 
   1766 /*
   1767  * find the internet address corresponding to a given address.
   1768  * ifaddr is returned referenced.
   1769  */
   1770 struct in6_ifaddr *
   1771 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
   1772 {
   1773 	struct in6_ifaddr *ia;
   1774 
   1775 #ifdef __FreeBSD__
   1776 	IN6_IFADDR_RLOCK();
   1777 	LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
   1778 #else
   1779 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
   1780 #endif
   1781 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
   1782 			if (zoneid != 0 &&
   1783 			    zoneid != ia->ia_addr.sin6_scope_id)
   1784 				continue;
   1785 			ifaref(&ia->ia_ifa);
   1786 			break;
   1787 		}
   1788 	}
   1789 #ifdef __FreeBSD__
   1790 	IN6_IFADDR_RUNLOCK();
   1791 #endif
   1792 	return ia;
   1793 }
   1794 
   1795 /*
   1796  * find the internet address corresponding to a given interface and address.
   1797  */
   1798 struct in6_ifaddr *
   1799 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
   1800 {
   1801 	struct ifaddr *best_ifa = NULL, *ifa;
   1802 
   1803 	IFADDR_FOREACH(ifa, ifp) {
   1804 		if (ifa->ifa_addr == NULL)
   1805 			continue;	/* just for safety */
   1806 		if (ifa->ifa_addr->sa_family != AF_INET6)
   1807 			continue;
   1808 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
   1809 			continue;
   1810 		best_ifa = bestifa(best_ifa, ifa);
   1811 	}
   1812 
   1813 	return (struct in6_ifaddr *)best_ifa;
   1814 }
   1815 
   1816 static struct in6_ifaddr *
   1817 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
   1818 {
   1819 	if (best_ia == NULL ||
   1820 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
   1821 		return ia;
   1822 	return best_ia;
   1823 }
   1824 
   1825 /*
   1826  * Convert IP6 address to printable (loggable) representation.
   1827  */
   1828 char *
   1829 ip6_sprintf(const struct in6_addr *addr)
   1830 {
   1831 	static int ip6round = 0;
   1832 	static char ip6buf[8][INET6_ADDRSTRLEN];
   1833 	char *cp = ip6buf[ip6round++ & 7];
   1834 
   1835 	in6_print(cp, INET6_ADDRSTRLEN, addr);
   1836 	return cp;
   1837 }
   1838 
   1839 /*
   1840  * Determine if an address is on a local network.
   1841  */
   1842 int
   1843 in6_localaddr(const struct in6_addr *in6)
   1844 {
   1845 	struct in6_ifaddr *ia;
   1846 
   1847 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
   1848 		return 1;
   1849 
   1850 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
   1851 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
   1852 					      &ia->ia_prefixmask.sin6_addr))
   1853 			return 1;
   1854 
   1855 	return 0;
   1856 }
   1857 
   1858 int
   1859 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
   1860 {
   1861 	struct in6_ifaddr *ia;
   1862 
   1863 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
   1864 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
   1865 		    &sa6->sin6_addr) &&
   1866 #ifdef SCOPEDROUTING
   1867 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
   1868 #endif
   1869 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
   1870 			return 1; /* true */
   1871 
   1872 		/* XXX: do we still have to go thru the rest of the list? */
   1873 	}
   1874 
   1875 	return 0;		/* false */
   1876 }
   1877 
   1878 /*
   1879  * return length of part which dst and src are equal
   1880  * hard coding...
   1881  */
   1882 int
   1883 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
   1884 {
   1885 	int match = 0;
   1886 	u_char *s = (u_char *)src, *d = (u_char *)dst;
   1887 	u_char *lim = s + 16, r;
   1888 
   1889 	while (s < lim)
   1890 		if ((r = (*d++ ^ *s++)) != 0) {
   1891 			while (r < 128) {
   1892 				match++;
   1893 				r <<= 1;
   1894 			}
   1895 			break;
   1896 		} else
   1897 			match += NBBY;
   1898 	return match;
   1899 }
   1900 
   1901 /* XXX: to be scope conscious */
   1902 int
   1903 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
   1904 {
   1905 	int bytelen, bitlen;
   1906 
   1907 	/* sanity check */
   1908 	if (len < 0 || len > 128) {
   1909 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
   1910 		    len);
   1911 		return 0;
   1912 	}
   1913 
   1914 	bytelen = len / NBBY;
   1915 	bitlen = len % NBBY;
   1916 
   1917 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
   1918 		return 0;
   1919 	if (bitlen != 0 &&
   1920 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
   1921 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
   1922 		return 0;
   1923 
   1924 	return 1;
   1925 }
   1926 
   1927 void
   1928 in6_prefixlen2mask(struct in6_addr *maskp, int len)
   1929 {
   1930 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
   1931 	int bytelen, bitlen, i;
   1932 
   1933 	/* sanity check */
   1934 	if (len < 0 || len > 128) {
   1935 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
   1936 		    len);
   1937 		return;
   1938 	}
   1939 
   1940 	memset(maskp, 0, sizeof(*maskp));
   1941 	bytelen = len / NBBY;
   1942 	bitlen = len % NBBY;
   1943 	for (i = 0; i < bytelen; i++)
   1944 		maskp->s6_addr[i] = 0xff;
   1945 	if (bitlen)
   1946 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
   1947 }
   1948 
   1949 /*
   1950  * return the best address out of the same scope. if no address was
   1951  * found, return the first valid address from designated IF.
   1952  */
   1953 struct in6_ifaddr *
   1954 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
   1955 {
   1956 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
   1957 	struct ifaddr *ifa;
   1958 	struct in6_ifaddr *best_ia = NULL, *ia;
   1959 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
   1960 
   1961 	dep[0] = dep[1] = NULL;
   1962 
   1963 	/*
   1964 	 * We first look for addresses in the same scope.
   1965 	 * If there is one, return it.
   1966 	 * If two or more, return one which matches the dst longest.
   1967 	 * If none, return one of global addresses assigned other ifs.
   1968 	 */
   1969 	IFADDR_FOREACH(ifa, ifp) {
   1970 		if (ifa->ifa_addr->sa_family != AF_INET6)
   1971 			continue;
   1972 		ia = (struct in6_ifaddr *)ifa;
   1973 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
   1974 			continue; /* XXX: is there any case to allow anycast? */
   1975 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
   1976 			continue; /* don't use this interface */
   1977 		if (ia->ia6_flags & IN6_IFF_DETACHED)
   1978 			continue;
   1979 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
   1980 			if (ip6_use_deprecated)
   1981 				dep[0] = ia;
   1982 			continue;
   1983 		}
   1984 
   1985 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
   1986 			continue;
   1987 		/*
   1988 		 * call in6_matchlen() as few as possible
   1989 		 */
   1990 		if (best_ia == NULL) {
   1991 			best_ia = ia;
   1992 			continue;
   1993 		}
   1994 		if (blen == -1)
   1995 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
   1996 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
   1997 		if (tlen > blen) {
   1998 			blen = tlen;
   1999 			best_ia = ia;
   2000 		} else if (tlen == blen)
   2001 			best_ia = bestia(best_ia, ia);
   2002 	}
   2003 	if (best_ia != NULL)
   2004 		return best_ia;
   2005 
   2006 	IFADDR_FOREACH(ifa, ifp) {
   2007 		if (ifa->ifa_addr->sa_family != AF_INET6)
   2008 			continue;
   2009 		ia = (struct in6_ifaddr *)ifa;
   2010 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
   2011 			continue; /* XXX: is there any case to allow anycast? */
   2012 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
   2013 			continue; /* don't use this interface */
   2014 		if (ia->ia6_flags & IN6_IFF_DETACHED)
   2015 			continue;
   2016 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
   2017 			if (ip6_use_deprecated)
   2018 				dep[1] = (struct in6_ifaddr *)ifa;
   2019 			continue;
   2020 		}
   2021 
   2022 		best_ia = bestia(best_ia, ia);
   2023 	}
   2024 	if (best_ia != NULL)
   2025 		return best_ia;
   2026 
   2027 	/* use the last-resort values, that are, deprecated addresses */
   2028 	if (dep[0])
   2029 		return dep[0];
   2030 	if (dep[1])
   2031 		return dep[1];
   2032 
   2033 	return NULL;
   2034 }
   2035 
   2036 /*
   2037  * perform DAD when interface becomes IFF_UP.
   2038  */
   2039 void
   2040 in6_if_link_up(struct ifnet *ifp)
   2041 {
   2042 	struct ifaddr *ifa;
   2043 	struct in6_ifaddr *ia;
   2044 
   2045 	/* Ensure it's sane to run DAD */
   2046 	if (ifp->if_link_state == LINK_STATE_DOWN)
   2047 		return;
   2048 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
   2049 		return;
   2050 
   2051 	IFADDR_FOREACH(ifa, ifp) {
   2052 		if (ifa->ifa_addr->sa_family != AF_INET6)
   2053 			continue;
   2054 		ia = (struct in6_ifaddr *)ifa;
   2055 
   2056 		/* If detached then mark as tentative */
   2057 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
   2058 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
   2059 			if (if_do_dad(ifp)) {
   2060 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
   2061 				nd6log((LOG_ERR, "in6_if_up: "
   2062 				    "%s marked tentative\n",
   2063 				    ip6_sprintf(&ia->ia_addr.sin6_addr)));
   2064 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
   2065 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   2066 		}
   2067 
   2068 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
   2069 			int rand_delay;
   2070 
   2071 			/* Clear the duplicated flag as we're starting DAD. */
   2072 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
   2073 
   2074 			/*
   2075 			 * The TENTATIVE flag was likely set by hand
   2076 			 * beforehand, implicitly indicating the need for DAD.
   2077 			 * We may be able to skip the random delay in this
   2078 			 * case, but we impose delays just in case.
   2079 			 */
   2080 			rand_delay = cprng_fast32() %
   2081 			    (MAX_RTR_SOLICITATION_DELAY * hz);
   2082 			/* +1 ensures callout is always used */
   2083 			nd6_dad_start(ifa, rand_delay + 1);
   2084 		}
   2085 	}
   2086 
   2087 	/* Restore any detached prefixes */
   2088 	pfxlist_onlink_check();
   2089 }
   2090 
   2091 void
   2092 in6_if_up(struct ifnet *ifp)
   2093 {
   2094 
   2095 	/*
   2096 	 * special cases, like 6to4, are handled in in6_ifattach
   2097 	 */
   2098 	in6_ifattach(ifp, NULL);
   2099 
   2100 	/* interface may not support link state, so bring it up also */
   2101 	in6_if_link_up(ifp);
   2102 }
   2103 
   2104 /*
   2105  * Mark all addresses as detached.
   2106  */
   2107 void
   2108 in6_if_link_down(struct ifnet *ifp)
   2109 {
   2110 	struct ifaddr *ifa;
   2111 	struct in6_ifaddr *ia;
   2112 
   2113 	/* Any prefixes on this interface should be detached as well */
   2114 	pfxlist_onlink_check();
   2115 
   2116 	IFADDR_FOREACH(ifa, ifp) {
   2117 		if (ifa->ifa_addr->sa_family != AF_INET6)
   2118 			continue;
   2119 		ia = (struct in6_ifaddr *)ifa;
   2120 
   2121 		/* Stop DAD processing */
   2122 		nd6_dad_stop(ifa);
   2123 
   2124 		/*
   2125 		 * Mark the address as detached.
   2126 		 * This satisfies RFC4862 Section 5.3, but we should apply
   2127 		 * this logic to all addresses to be a good citizen and
   2128 		 * avoid potential duplicated addresses.
   2129 		 * When the interface comes up again, detached addresses
   2130 		 * are marked tentative and DAD commences.
   2131 		 */
   2132 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
   2133 			nd6log((LOG_DEBUG, "in6_if_down: "
   2134 			    "%s marked detached\n",
   2135 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
   2136 			ia->ia6_flags |= IN6_IFF_DETACHED;
   2137 			ia->ia6_flags &=
   2138 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
   2139 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
   2140 		}
   2141 	}
   2142 }
   2143 
   2144 void
   2145 in6_if_down(struct ifnet *ifp)
   2146 {
   2147 
   2148 	in6_if_link_down(ifp);
   2149 }
   2150 
   2151 void
   2152 in6_if_link_state_change(struct ifnet *ifp, int link_state)
   2153 {
   2154 
   2155 	switch (link_state) {
   2156 	case LINK_STATE_DOWN:
   2157 		in6_if_link_down(ifp);
   2158 		break;
   2159 	case LINK_STATE_UP:
   2160 		in6_if_link_up(ifp);
   2161 		break;
   2162 	}
   2163 }
   2164 
   2165 /*
   2166  * Calculate max IPv6 MTU through all the interfaces and store it
   2167  * to in6_maxmtu.
   2168  */
   2169 void
   2170 in6_setmaxmtu(void)
   2171 {
   2172 	unsigned long maxmtu = 0;
   2173 	struct ifnet *ifp;
   2174 
   2175 	IFNET_FOREACH(ifp) {
   2176 		/* this function can be called during ifnet initialization */
   2177 		if (!ifp->if_afdata[AF_INET6])
   2178 			continue;
   2179 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
   2180 		    IN6_LINKMTU(ifp) > maxmtu)
   2181 			maxmtu = IN6_LINKMTU(ifp);
   2182 	}
   2183 	if (maxmtu)	     /* update only when maxmtu is positive */
   2184 		in6_maxmtu = maxmtu;
   2185 }
   2186 
   2187 /*
   2188  * Provide the length of interface identifiers to be used for the link attached
   2189  * to the given interface.  The length should be defined in "IPv6 over
   2190  * xxx-link" document.  Note that address architecture might also define
   2191  * the length for a particular set of address prefixes, regardless of the
   2192  * link type.  As clarified in rfc2462bis, those two definitions should be
   2193  * consistent, and those really are as of August 2004.
   2194  */
   2195 int
   2196 in6_if2idlen(struct ifnet *ifp)
   2197 {
   2198 	switch (ifp->if_type) {
   2199 	case IFT_ETHER:		/* RFC2464 */
   2200 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
   2201 	case IFT_L2VLAN:	/* ditto */
   2202 	case IFT_IEEE80211:	/* ditto */
   2203 	case IFT_FDDI:		/* RFC2467 */
   2204 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
   2205 	case IFT_PPP:		/* RFC2472 */
   2206 	case IFT_ARCNET:	/* RFC2497 */
   2207 	case IFT_FRELAY:	/* RFC2590 */
   2208 	case IFT_IEEE1394:	/* RFC3146 */
   2209 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
   2210 	case IFT_LOOP:		/* XXX: is this really correct? */
   2211 		return 64;
   2212 	default:
   2213 		/*
   2214 		 * Unknown link type:
   2215 		 * It might be controversial to use the today's common constant
   2216 		 * of 64 for these cases unconditionally.  For full compliance,
   2217 		 * we should return an error in this case.  On the other hand,
   2218 		 * if we simply miss the standard for the link type or a new
   2219 		 * standard is defined for a new link type, the IFID length
   2220 		 * is very likely to be the common constant.  As a compromise,
   2221 		 * we always use the constant, but make an explicit notice
   2222 		 * indicating the "unknown" case.
   2223 		 */
   2224 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
   2225 		return 64;
   2226 	}
   2227 }
   2228 
   2229 struct in6_llentry {
   2230 	struct llentry		base;
   2231 };
   2232 
   2233 #define	IN6_LLTBL_DEFAULT_HSIZE	32
   2234 #define	IN6_LLTBL_HASH(k, h) \
   2235 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
   2236 
   2237 /*
   2238  * Do actual deallocation of @lle.
   2239  * Called by LLE_FREE_LOCKED when number of references
   2240  * drops to zero.
   2241  */
   2242 static void
   2243 in6_lltable_destroy_lle(struct llentry *lle)
   2244 {
   2245 
   2246 	LLE_WUNLOCK(lle);
   2247 	LLE_LOCK_DESTROY(lle);
   2248 	kmem_intr_free(lle, sizeof(struct in6_llentry));
   2249 }
   2250 
   2251 static struct llentry *
   2252 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
   2253 {
   2254 	struct in6_llentry *lle;
   2255 
   2256 	lle = kmem_intr_zalloc(sizeof(struct in6_llentry), KM_NOSLEEP);
   2257 	if (lle == NULL)		/* NB: caller generates msg */
   2258 		return NULL;
   2259 
   2260 	lle->base.r_l3addr.addr6 = *addr6;
   2261 	lle->base.lle_refcnt = 1;
   2262 	lle->base.lle_free = in6_lltable_destroy_lle;
   2263 	LLE_LOCK_INIT(&lle->base);
   2264 	callout_init(&lle->base.lle_timer, CALLOUT_MPSAFE);
   2265 
   2266 	return &lle->base;
   2267 }
   2268 
   2269 static int
   2270 in6_lltable_match_prefix(const struct sockaddr *prefix,
   2271     const struct sockaddr *mask, u_int flags, struct llentry *lle)
   2272 {
   2273 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
   2274 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
   2275 
   2276 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
   2277 	    &pfx->sin6_addr, &msk->sin6_addr) &&
   2278 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
   2279 		return 1;
   2280 
   2281 	return 0;
   2282 }
   2283 
   2284 static void
   2285 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
   2286 {
   2287 	struct ifnet *ifp __diagused;
   2288 
   2289 	LLE_WLOCK_ASSERT(lle);
   2290 	KASSERT(llt != NULL);
   2291 
   2292 	/* Unlink entry from table */
   2293 	if ((lle->la_flags & LLE_LINKED) != 0) {
   2294 
   2295 		ifp = llt->llt_ifp;
   2296 		IF_AFDATA_WLOCK_ASSERT(ifp);
   2297 		lltable_unlink_entry(llt, lle);
   2298 	}
   2299 
   2300 	KASSERT(mutex_owned(softnet_lock));
   2301 	callout_halt(&lle->lle_timer, softnet_lock);
   2302 	LLE_REMREF(lle);
   2303 
   2304 	llentry_free(lle);
   2305 }
   2306 
   2307 static int
   2308 in6_lltable_rtcheck(struct ifnet *ifp,
   2309 		    u_int flags,
   2310 		    const struct sockaddr *l3addr)
   2311 {
   2312 	struct rtentry *rt;
   2313 
   2314 	KASSERTMSG(l3addr->sa_family == AF_INET6,
   2315 	    "sin_family %d", l3addr->sa_family);
   2316 
   2317 	rt = rtalloc1(l3addr, 0);
   2318 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
   2319 		struct ifaddr *ifa;
   2320 		/*
   2321 		 * Create an ND6 cache for an IPv6 neighbor
   2322 		 * that is not covered by our own prefix.
   2323 		 */
   2324 		/* XXX ifaof_ifpforaddr should take a const param */
   2325 		ifa = ifaof_ifpforaddr(l3addr, ifp);
   2326 		if (ifa != NULL) {
   2327 			ifafree(ifa);
   2328 			if (rt != NULL)
   2329 				rtfree(rt);
   2330 			return 0;
   2331 		}
   2332 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
   2333 		    ip6_sprintf(&((const struct sockaddr_in6 *)l3addr)->sin6_addr));
   2334 		if (rt != NULL)
   2335 			rtfree(rt);
   2336 		return EINVAL;
   2337 	}
   2338 	rtfree(rt);
   2339 	return 0;
   2340 }
   2341 
   2342 static inline uint32_t
   2343 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
   2344 {
   2345 
   2346 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
   2347 }
   2348 
   2349 static uint32_t
   2350 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
   2351 {
   2352 
   2353 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
   2354 }
   2355 
   2356 static void
   2357 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
   2358 {
   2359 	struct sockaddr_in6 *sin6;
   2360 
   2361 	sin6 = (struct sockaddr_in6 *)sa;
   2362 	bzero(sin6, sizeof(*sin6));
   2363 	sin6->sin6_family = AF_INET6;
   2364 	sin6->sin6_len = sizeof(*sin6);
   2365 	sin6->sin6_addr = lle->r_l3addr.addr6;
   2366 }
   2367 
   2368 static inline struct llentry *
   2369 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
   2370 {
   2371 	struct llentry *lle;
   2372 	struct llentries *lleh;
   2373 	u_int hashidx;
   2374 
   2375 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
   2376 	lleh = &llt->lle_head[hashidx];
   2377 	LIST_FOREACH(lle, lleh, lle_next) {
   2378 		if (lle->la_flags & LLE_DELETED)
   2379 			continue;
   2380 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
   2381 			break;
   2382 	}
   2383 
   2384 	return lle;
   2385 }
   2386 
   2387 static int
   2388 in6_lltable_delete(struct lltable *llt, u_int flags,
   2389 	const struct sockaddr *l3addr)
   2390 {
   2391 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
   2392 	struct llentry *lle;
   2393 
   2394 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
   2395 	KASSERTMSG(l3addr->sa_family == AF_INET6,
   2396 	    "sin_family %d", l3addr->sa_family);
   2397 
   2398 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
   2399 
   2400 	if (lle == NULL)
   2401 		return ENOENT;
   2402 
   2403 	if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
   2404 		LLE_WLOCK(lle);
   2405 		lle->la_flags |= LLE_DELETED;
   2406 #ifdef DIAGNOSTIC
   2407 		log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
   2408 #endif
   2409 		if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
   2410 			llentry_free(lle);
   2411 		else
   2412 			LLE_WUNLOCK(lle);
   2413 	}
   2414 
   2415 	return 0;
   2416 }
   2417 
   2418 static struct llentry *
   2419 in6_lltable_create(struct lltable *llt, u_int flags,
   2420 	const struct sockaddr *l3addr)
   2421 {
   2422 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
   2423 	struct ifnet *ifp = llt->llt_ifp;
   2424 	struct llentry *lle;
   2425 
   2426 	IF_AFDATA_WLOCK_ASSERT(ifp);
   2427 	KASSERTMSG(l3addr->sa_family == AF_INET6,
   2428 	    "sin_family %d", l3addr->sa_family);
   2429 
   2430 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
   2431 
   2432 	if (lle != NULL) {
   2433 		LLE_WLOCK(lle);
   2434 		return lle;
   2435 	}
   2436 
   2437 	/*
   2438 	 * A route that covers the given address must have
   2439 	 * been installed 1st because we are doing a resolution,
   2440 	 * verify this.
   2441 	 */
   2442 	if (!(flags & LLE_IFADDR) &&
   2443 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
   2444 		return NULL;
   2445 
   2446 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
   2447 	if (lle == NULL) {
   2448 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
   2449 		return NULL;
   2450 	}
   2451 	lle->la_flags = flags;
   2452 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
   2453 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
   2454 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
   2455 	}
   2456 
   2457 	lltable_link_entry(llt, lle);
   2458 	LLE_WLOCK(lle);
   2459 
   2460 	return lle;
   2461 }
   2462 
   2463 static struct llentry *
   2464 in6_lltable_lookup(struct lltable *llt, u_int flags,
   2465 	const struct sockaddr *l3addr)
   2466 {
   2467 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
   2468 	struct llentry *lle;
   2469 
   2470 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
   2471 	KASSERTMSG(l3addr->sa_family == AF_INET6,
   2472 	    "sin_family %d", l3addr->sa_family);
   2473 
   2474 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
   2475 
   2476 	if (lle == NULL)
   2477 		return NULL;
   2478 
   2479 	if (flags & LLE_EXCLUSIVE)
   2480 		LLE_WLOCK(lle);
   2481 	else
   2482 		LLE_RLOCK(lle);
   2483 	return lle;
   2484 }
   2485 
   2486 static struct lltable *
   2487 in6_lltattach(struct ifnet *ifp)
   2488 {
   2489 	struct lltable *llt;
   2490 
   2491 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
   2492 	llt->llt_af = AF_INET6;
   2493 	llt->llt_ifp = ifp;
   2494 
   2495 	llt->llt_lookup = in6_lltable_lookup;
   2496 	llt->llt_create = in6_lltable_create;
   2497 	llt->llt_delete = in6_lltable_delete;
   2498 #if notyet
   2499 	llt->llt_dump_entry = in6_lltable_dump_entry;
   2500 #endif
   2501 	llt->llt_hash = in6_lltable_hash;
   2502 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
   2503 	llt->llt_free_entry = in6_lltable_free_entry;
   2504 	llt->llt_match_prefix = in6_lltable_match_prefix;
   2505 	lltable_link(llt);
   2506 
   2507 	return llt;
   2508 }
   2509 
   2510 void *
   2511 in6_domifattach(struct ifnet *ifp)
   2512 {
   2513 	struct in6_ifextra *ext;
   2514 
   2515 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
   2516 
   2517 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
   2518 	    M_IFADDR, M_WAITOK|M_ZERO);
   2519 
   2520 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
   2521 	    M_IFADDR, M_WAITOK|M_ZERO);
   2522 
   2523 	ext->nd_ifinfo = nd6_ifattach(ifp);
   2524 	ext->scope6_id = scope6_ifattach(ifp);
   2525 	ext->nprefixes = 0;
   2526 	ext->ndefrouters = 0;
   2527 
   2528 	ext->lltable = in6_lltattach(ifp);
   2529 
   2530 	return ext;
   2531 }
   2532 
   2533 void
   2534 in6_domifdetach(struct ifnet *ifp, void *aux)
   2535 {
   2536 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
   2537 
   2538 	lltable_free(ext->lltable);
   2539 	ext->lltable = NULL;
   2540 	nd6_ifdetach(ifp, ext);
   2541 	free(ext->in6_ifstat, M_IFADDR);
   2542 	free(ext->icmp6_ifstat, M_IFADDR);
   2543 	scope6_ifdetach(ext->scope6_id);
   2544 	free(ext, M_IFADDR);
   2545 }
   2546 
   2547 /*
   2548  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
   2549  * v4 mapped addr or v4 compat addr
   2550  */
   2551 void
   2552 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
   2553 {
   2554 	memset(sin, 0, sizeof(*sin));
   2555 	sin->sin_len = sizeof(struct sockaddr_in);
   2556 	sin->sin_family = AF_INET;
   2557 	sin->sin_port = sin6->sin6_port;
   2558 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
   2559 }
   2560 
   2561 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
   2562 void
   2563 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
   2564 {
   2565 	memset(sin6, 0, sizeof(*sin6));
   2566 	sin6->sin6_len = sizeof(struct sockaddr_in6);
   2567 	sin6->sin6_family = AF_INET6;
   2568 	sin6->sin6_port = sin->sin_port;
   2569 	sin6->sin6_addr.s6_addr32[0] = 0;
   2570 	sin6->sin6_addr.s6_addr32[1] = 0;
   2571 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
   2572 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
   2573 }
   2574 
   2575 /* Convert sockaddr_in6 into sockaddr_in. */
   2576 void
   2577 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
   2578 {
   2579 	struct sockaddr_in *sin_p;
   2580 	struct sockaddr_in6 sin6;
   2581 
   2582 	/*
   2583 	 * Save original sockaddr_in6 addr and convert it
   2584 	 * to sockaddr_in.
   2585 	 */
   2586 	sin6 = *(struct sockaddr_in6 *)nam;
   2587 	sin_p = (struct sockaddr_in *)nam;
   2588 	in6_sin6_2_sin(sin_p, &sin6);
   2589 }
   2590 
   2591 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
   2592 void
   2593 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
   2594 {
   2595 	struct sockaddr_in *sin_p;
   2596 	struct sockaddr_in6 *sin6_p;
   2597 
   2598 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
   2599 	sin_p = (struct sockaddr_in *)*nam;
   2600 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
   2601 	free(*nam, M_SONAME);
   2602 	*nam = (struct sockaddr *)sin6_p;
   2603 }
   2604