Home | History | Annotate | Line # | Download | only in netinet6
nd6_rtr.c revision 1.60
      1 /*	$NetBSD: nd6_rtr.c,v 1.60 2006/10/12 01:32:39 christos Exp $	*/
      2 /*	$KAME: nd6_rtr.c,v 1.95 2001/02/07 08:09:47 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 #include <sys/cdefs.h>
     34 __KERNEL_RCSID(0, "$NetBSD: nd6_rtr.c,v 1.60 2006/10/12 01:32:39 christos Exp $");
     35 
     36 #include <sys/param.h>
     37 #include <sys/systm.h>
     38 #include <sys/malloc.h>
     39 #include <sys/mbuf.h>
     40 #include <sys/socket.h>
     41 #include <sys/sockio.h>
     42 #include <sys/time.h>
     43 #include <sys/kernel.h>
     44 #include <sys/errno.h>
     45 #include <sys/ioctl.h>
     46 #include <sys/syslog.h>
     47 
     48 #include <net/if.h>
     49 #include <net/if_types.h>
     50 #include <net/if_dl.h>
     51 #include <net/route.h>
     52 #include <net/radix.h>
     53 
     54 #include <netinet/in.h>
     55 #include <netinet6/in6_var.h>
     56 #include <netinet6/in6_ifattach.h>
     57 #include <netinet/ip6.h>
     58 #include <netinet6/ip6_var.h>
     59 #include <netinet6/nd6.h>
     60 #include <netinet/icmp6.h>
     61 #include <netinet6/scope6_var.h>
     62 
     63 #include <net/net_osdep.h>
     64 
     65 #define SDL(s)	((struct sockaddr_dl *)s)
     66 
     67 static int rtpref __P((struct nd_defrouter *));
     68 static struct nd_defrouter *defrtrlist_update __P((struct nd_defrouter *));
     69 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
     70     struct mbuf *, int));
     71 static struct in6_ifaddr *in6_ifadd __P((struct nd_prefixctl *, int));
     72 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
     73 	struct nd_defrouter *));
     74 static void pfxrtr_add __P((struct nd_prefix *, struct nd_defrouter *));
     75 static void pfxrtr_del __P((struct nd_pfxrouter *));
     76 static struct nd_pfxrouter *find_pfxlist_reachable_router
     77 	__P((struct nd_prefix *));
     78 static void defrouter_delreq __P((struct nd_defrouter *));
     79 static void nd6_rtmsg __P((int, struct rtentry *));
     80 
     81 static int in6_init_prefix_ltimes __P((struct nd_prefix *));
     82 static void in6_init_address_ltimes __P((struct nd_prefix *ndpr,
     83 	struct in6_addrlifetime *lt6));
     84 
     85 static int rt6_deleteroute __P((struct radix_node *, void *));
     86 
     87 extern int nd6_recalc_reachtm_interval;
     88 
     89 static struct ifnet *nd6_defifp;
     90 int nd6_defifindex;
     91 
     92 int ip6_use_tempaddr = 0;
     93 
     94 int ip6_desync_factor;
     95 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
     96 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
     97 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
     98 
     99 /* RTPREF_MEDIUM has to be 0! */
    100 #define RTPREF_HIGH	1
    101 #define RTPREF_MEDIUM	0
    102 #define RTPREF_LOW	(-1)
    103 #define RTPREF_RESERVED	(-2)
    104 #define RTPREF_INVALID	(-3)	/* internal */
    105 
    106 /*
    107  * Receive Router Solicitation Message - just for routers.
    108  * Router solicitation/advertisement is mostly managed by userland program
    109  * (rtadvd) so here we have no function like nd6_ra_output().
    110  *
    111  * Based on RFC 2461
    112  */
    113 void
    114 nd6_rs_input(m, off, icmp6len)
    115 	struct	mbuf *m;
    116 	int off, icmp6len;
    117 {
    118 	struct ifnet *ifp = m->m_pkthdr.rcvif;
    119 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
    120 	struct nd_router_solicit *nd_rs;
    121 	struct in6_addr saddr6 = ip6->ip6_src;
    122 #if 0
    123 	struct in6_addr daddr6 = ip6->ip6_dst;
    124 #endif
    125 	char *lladdr = NULL;
    126 	int lladdrlen = 0;
    127 #if 0
    128 	struct sockaddr_dl *sdl = (struct sockaddr_dl *)NULL;
    129 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)NULL;
    130 	struct rtentry *rt = NULL;
    131 	int is_newentry;
    132 #endif
    133 	union nd_opts ndopts;
    134 
    135 	/* If I'm not a router, ignore it. */
    136 	if (ip6_accept_rtadv != 0 || !ip6_forwarding)
    137 		goto freeit;
    138 
    139 	/* Sanity checks */
    140 	if (ip6->ip6_hlim != 255) {
    141 		nd6log((LOG_ERR,
    142 		    "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
    143 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
    144 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
    145 		goto bad;
    146 	}
    147 
    148 	/*
    149 	 * Don't update the neighbor cache, if src = ::.
    150 	 * This indicates that the src has no IP address assigned yet.
    151 	 */
    152 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
    153 		goto freeit;
    154 
    155 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
    156 	if (nd_rs == NULL) {
    157 		icmp6stat.icp6s_tooshort++;
    158 		return;
    159 	}
    160 
    161 	icmp6len -= sizeof(*nd_rs);
    162 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
    163 	if (nd6_options(&ndopts) < 0) {
    164 		nd6log((LOG_INFO,
    165 		    "nd6_rs_input: invalid ND option, ignored\n"));
    166 		/* nd6_options have incremented stats */
    167 		goto freeit;
    168 	}
    169 
    170 	if (ndopts.nd_opts_src_lladdr) {
    171 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
    172 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
    173 	}
    174 
    175 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
    176 		nd6log((LOG_INFO,
    177 		    "nd6_rs_input: lladdrlen mismatch for %s "
    178 		    "(if %d, RS packet %d)\n",
    179 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2));
    180 		goto bad;
    181 	}
    182 
    183 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
    184 
    185  freeit:
    186 	m_freem(m);
    187 	return;
    188 
    189  bad:
    190 	icmp6stat.icp6s_badrs++;
    191 	m_freem(m);
    192 }
    193 
    194 /*
    195  * Receive Router Advertisement Message.
    196  *
    197  * Based on RFC 2461
    198  * TODO: on-link bit on prefix information
    199  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
    200  */
    201 void
    202 nd6_ra_input(m, off, icmp6len)
    203 	struct	mbuf *m;
    204 	int off, icmp6len;
    205 {
    206 	struct ifnet *ifp = m->m_pkthdr.rcvif;
    207 	struct nd_ifinfo *ndi = ND_IFINFO(ifp);
    208 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
    209 	struct nd_router_advert *nd_ra;
    210 	struct in6_addr saddr6 = ip6->ip6_src;
    211 #if 0
    212 	struct in6_addr daddr6 = ip6->ip6_dst;
    213 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
    214 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
    215 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
    216 #endif
    217 	int mcast = 0;
    218 	union nd_opts ndopts;
    219 	struct nd_defrouter *dr;
    220 
    221 	/*
    222 	 * We only accept RAs only when
    223 	 * the system-wide variable allows the acceptance, and
    224 	 * per-interface variable allows RAs on the receiving interface.
    225 	 */
    226 	if (ip6_accept_rtadv == 0)
    227 		goto freeit;
    228 	if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
    229 		goto freeit;
    230 
    231 	if (ip6->ip6_hlim != 255) {
    232 		nd6log((LOG_ERR,
    233 		    "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
    234 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
    235 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp)));
    236 		goto bad;
    237 	}
    238 
    239 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
    240 		nd6log((LOG_ERR,
    241 		    "nd6_ra_input: src %s is not link-local\n",
    242 		    ip6_sprintf(&saddr6)));
    243 		goto bad;
    244 	}
    245 
    246 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
    247 	if (nd_ra == NULL) {
    248 		icmp6stat.icp6s_tooshort++;
    249 		return;
    250 	}
    251 
    252 	icmp6len -= sizeof(*nd_ra);
    253 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
    254 	if (nd6_options(&ndopts) < 0) {
    255 		nd6log((LOG_INFO,
    256 		    "nd6_ra_input: invalid ND option, ignored\n"));
    257 		/* nd6_options have incremented stats */
    258 		goto freeit;
    259 	}
    260 
    261     {
    262 	struct nd_defrouter drtr;
    263 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
    264 
    265 	/* remember if this is a multicasted advertisement */
    266 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
    267 		mcast = 1;
    268 
    269 	memset(&drtr, 0, sizeof(drtr));
    270 	drtr.rtaddr = saddr6;
    271 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
    272 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
    273 	drtr.expire = time_second + drtr.rtlifetime;
    274 	drtr.ifp = ifp;
    275 	/* unspecified or not? (RFC 2461 6.3.4) */
    276 	if (advreachable) {
    277 		NTOHL(advreachable);
    278 		if (advreachable <= MAX_REACHABLE_TIME &&
    279 		    ndi->basereachable != advreachable) {
    280 			ndi->basereachable = advreachable;
    281 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
    282 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
    283 		}
    284 	}
    285 	if (nd_ra->nd_ra_retransmit)
    286 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
    287 	if (nd_ra->nd_ra_curhoplimit)
    288 		ndi->chlim = nd_ra->nd_ra_curhoplimit;
    289 	dr = defrtrlist_update(&drtr);
    290     }
    291 
    292 	/*
    293 	 * prefix
    294 	 */
    295 	if (ndopts.nd_opts_pi) {
    296 		struct nd_opt_hdr *pt;
    297 		struct nd_opt_prefix_info *pi = NULL;
    298 		struct nd_prefixctl pr;
    299 
    300 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
    301 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
    302 		     pt = (struct nd_opt_hdr *)((caddr_t)pt +
    303 						(pt->nd_opt_len << 3))) {
    304 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
    305 				continue;
    306 			pi = (struct nd_opt_prefix_info *)pt;
    307 
    308 			if (pi->nd_opt_pi_len != 4) {
    309 				nd6log((LOG_INFO,
    310 				    "nd6_ra_input: invalid option "
    311 				    "len %d for prefix information option, "
    312 				    "ignored\n", pi->nd_opt_pi_len));
    313 				continue;
    314 			}
    315 
    316 			if (128 < pi->nd_opt_pi_prefix_len) {
    317 				nd6log((LOG_INFO,
    318 				    "nd6_ra_input: invalid prefix "
    319 				    "len %d for prefix information option, "
    320 				    "ignored\n", pi->nd_opt_pi_prefix_len));
    321 				continue;
    322 			}
    323 
    324 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
    325 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
    326 				nd6log((LOG_INFO,
    327 				    "nd6_ra_input: invalid prefix "
    328 				    "%s, ignored\n",
    329 				    ip6_sprintf(&pi->nd_opt_pi_prefix)));
    330 				continue;
    331 			}
    332 
    333 			bzero(&pr, sizeof(pr));
    334 			pr.ndpr_prefix.sin6_family = AF_INET6;
    335 			pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
    336 			pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
    337 			pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
    338 
    339 			pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
    340 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
    341 			pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
    342 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
    343 			pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
    344 			pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
    345 			pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
    346 
    347 			(void)prelist_update(&pr, dr, m, mcast);
    348 		}
    349 	}
    350 
    351 	/*
    352 	 * MTU
    353 	 */
    354 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
    355 		u_long mtu;
    356 		u_long maxmtu;
    357 
    358 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
    359 
    360 		/* lower bound */
    361 		if (mtu < IPV6_MMTU) {
    362 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
    363 			    "mtu=%lu sent from %s, ignoring\n",
    364 			    mtu, ip6_sprintf(&ip6->ip6_src)));
    365 			goto skip;
    366 		}
    367 
    368 		/* upper bound */
    369 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
    370 		    ? ndi->maxmtu : ifp->if_mtu;
    371 		if (mtu <= maxmtu) {
    372 			int change = (ndi->linkmtu != mtu);
    373 
    374 			ndi->linkmtu = mtu;
    375 			if (change) /* in6_maxmtu may change */
    376 				in6_setmaxmtu();
    377 		} else {
    378 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
    379 			    "mtu=%lu sent from %s; "
    380 			    "exceeds maxmtu %lu, ignoring\n",
    381 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
    382 		}
    383 	}
    384 
    385  skip:
    386 
    387 	/*
    388 	 * Source link layer address
    389 	 */
    390     {
    391 	char *lladdr = NULL;
    392 	int lladdrlen = 0;
    393 
    394 	if (ndopts.nd_opts_src_lladdr) {
    395 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
    396 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
    397 	}
    398 
    399 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
    400 		nd6log((LOG_INFO,
    401 		    "nd6_ra_input: lladdrlen mismatch for %s "
    402 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
    403 		    ifp->if_addrlen, lladdrlen - 2));
    404 		goto bad;
    405 	}
    406 
    407 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
    408 
    409 	/*
    410 	 * Installing a link-layer address might change the state of the
    411 	 * router's neighbor cache, which might also affect our on-link
    412 	 * detection of adveritsed prefixes.
    413 	 */
    414 	pfxlist_onlink_check();
    415     }
    416 
    417  freeit:
    418 	m_freem(m);
    419 	return;
    420 
    421  bad:
    422 	icmp6stat.icp6s_badra++;
    423 	m_freem(m);
    424 }
    425 
    426 /*
    427  * default router list processing sub routines
    428  */
    429 
    430 /* tell the change to user processes watching the routing socket. */
    431 static void
    432 nd6_rtmsg(cmd, rt)
    433 	int cmd;
    434 	struct rtentry *rt;
    435 {
    436 	struct rt_addrinfo info;
    437 
    438 	bzero((caddr_t)&info, sizeof(info));
    439 	info.rti_info[RTAX_DST] = rt_key(rt);
    440 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
    441 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
    442 	if (rt->rt_ifp) {
    443 		info.rti_info[RTAX_IFP] =
    444 		    TAILQ_FIRST(&rt->rt_ifp->if_addrlist)->ifa_addr;
    445 		info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
    446 	}
    447 
    448 	rt_missmsg(cmd, &info, rt->rt_flags, 0);
    449 }
    450 
    451 void
    452 defrouter_addreq(new)
    453 	struct nd_defrouter *new;
    454 {
    455 	struct sockaddr_in6 def, mask, gate;
    456 	struct rtentry *newrt = NULL;
    457 	int s;
    458 	int error;
    459 
    460 	memset(&def, 0, sizeof(def));
    461 	memset(&mask, 0, sizeof(mask));
    462 	memset(&gate, 0,sizeof(gate)); /* for safety */
    463 
    464 	def.sin6_len = mask.sin6_len = gate.sin6_len =
    465 	    sizeof(struct sockaddr_in6);
    466 	def.sin6_family = mask.sin6_family = gate.sin6_family = AF_INET6;
    467 	gate.sin6_addr = new->rtaddr;
    468 #ifndef SCOPEDROUTING
    469 	gate.sin6_scope_id = 0;	/* XXX */
    470 #endif
    471 
    472 	s = splsoftnet();
    473 	error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
    474 	    (struct sockaddr *)&gate, (struct sockaddr *)&mask,
    475 	    RTF_GATEWAY, &newrt);
    476 	if (newrt) {
    477 		nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
    478 		newrt->rt_refcnt--;
    479 	}
    480 	if (error == 0)
    481 		new->installed = 1;
    482 	splx(s);
    483 	return;
    484 }
    485 
    486 struct nd_defrouter *
    487 defrouter_lookup(addr, ifp)
    488 	struct in6_addr *addr;
    489 	struct ifnet *ifp;
    490 {
    491 	struct nd_defrouter *dr;
    492 
    493 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    494 	     dr = TAILQ_NEXT(dr, dr_entry)) {
    495 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr)) {
    496 			return (dr);
    497 		}
    498 	}
    499 
    500 	return (NULL);		/* search failed */
    501 }
    502 
    503 void
    504 defrtrlist_del(dr)
    505 	struct nd_defrouter *dr;
    506 {
    507 	struct nd_defrouter *deldr = NULL;
    508 	struct nd_prefix *pr;
    509 
    510 	/*
    511 	 * Flush all the routing table entries that use the router
    512 	 * as a next hop.
    513 	 */
    514 	if (!ip6_forwarding && ip6_accept_rtadv) /* XXX: better condition? */
    515 		rt6_flush(&dr->rtaddr, dr->ifp);
    516 
    517 	if (dr->installed) {
    518 		deldr = dr;
    519 		defrouter_delreq(dr);
    520 	}
    521 	TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
    522 
    523 	/*
    524 	 * Also delete all the pointers to the router in each prefix lists.
    525 	 */
    526 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
    527 		struct nd_pfxrouter *pfxrtr;
    528 		if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
    529 			pfxrtr_del(pfxrtr);
    530 	}
    531 	pfxlist_onlink_check();
    532 
    533 	/*
    534 	 * If the router is the primary one, choose a new one.
    535 	 * Note that defrouter_select() will remove the current gateway
    536 	 * from the routing table.
    537 	 */
    538 	if (deldr)
    539 		defrouter_select();
    540 
    541 	free(dr, M_IP6NDP);
    542 }
    543 
    544 /*
    545  * Remove the default route for a given router.
    546  * This is just a subroutine function for defrouter_select(), and should
    547  * not be called from anywhere else.
    548  */
    549 static void
    550 defrouter_delreq(dr)
    551 	struct nd_defrouter *dr;
    552 {
    553 	struct sockaddr_in6 def, mask, gw;
    554 	struct rtentry *oldrt = NULL;
    555 
    556 #ifdef DIAGNOSTIC
    557 	if (dr == NULL)
    558 		panic("dr == NULL in defrouter_delreq");
    559 #endif
    560 
    561 	bzero(&def, sizeof(def));
    562 	bzero(&mask, sizeof(mask));
    563 	bzero(&gw, sizeof(gw));	/* for safety */
    564 
    565 	def.sin6_len = mask.sin6_len = gw.sin6_len =
    566 	    sizeof(struct sockaddr_in6);
    567 	def.sin6_family = mask.sin6_family = gw.sin6_family = AF_INET6;
    568 	gw.sin6_addr = dr->rtaddr;
    569 #ifndef SCOPEDROUTING
    570 	gw.sin6_scope_id = 0;	/* XXX */
    571 #endif
    572 
    573 	rtrequest(RTM_DELETE, (struct sockaddr *)&def,
    574 	    (struct sockaddr *)&gw,
    575 	    (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
    576 	if (oldrt) {
    577 		nd6_rtmsg(RTM_DELETE, oldrt);
    578 		if (oldrt->rt_refcnt <= 0) {
    579 			/*
    580 			 * XXX: borrowed from the RTM_DELETE case of
    581 			 * rtrequest().
    582 			 */
    583 			oldrt->rt_refcnt++;
    584 			rtfree(oldrt);
    585 		}
    586 	}
    587 
    588 	dr->installed = 0;
    589 }
    590 
    591 /*
    592  * remove all default routes from default router list
    593  */
    594 void
    595 defrouter_reset()
    596 {
    597 	struct nd_defrouter *dr;
    598 
    599 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    600 	     dr = TAILQ_NEXT(dr, dr_entry))
    601 		defrouter_delreq(dr);
    602 
    603 	/*
    604 	 * XXX should we also nuke any default routers in the kernel, by
    605 	 * going through them by rtalloc1()?
    606 	 */
    607 }
    608 
    609 /*
    610  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
    611  * draft-ietf-ipngwg-router-selection:
    612  * 1) Routers that are reachable or probably reachable should be preferred.
    613  *    If we have more than one (probably) reachable router, prefer ones
    614  *    with the highest router preference.
    615  * 2) When no routers on the list are known to be reachable or
    616  *    probably reachable, routers SHOULD be selected in a round-robin
    617  *    fashion, regardless of router preference values.
    618  * 3) If the Default Router List is empty, assume that all
    619  *    destinations are on-link.
    620  *
    621  * We assume nd_defrouter is sorted by router preference value.
    622  * Since the code below covers both with and without router preference cases,
    623  * we do not need to classify the cases by ifdef.
    624  *
    625  * At this moment, we do not try to install more than one default router,
    626  * even when the multipath routing is available, because we're not sure about
    627  * the benefits for stub hosts comparing to the risk of making the code
    628  * complicated and the possibility of introducing bugs.
    629  */
    630 void
    631 defrouter_select()
    632 {
    633 	int s = splsoftnet();
    634 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
    635 	struct rtentry *rt = NULL;
    636 	struct llinfo_nd6 *ln = NULL;
    637 
    638 	/*
    639 	 * This function should be called only when acting as an autoconfigured
    640 	 * host.  Although the remaining part of this function is not effective
    641 	 * if the node is not an autoconfigured host, we explicitly exclude
    642 	 * such cases here for safety.
    643 	 */
    644 	if (ip6_forwarding || !ip6_accept_rtadv) {
    645 		nd6log((LOG_WARNING,
    646 		    "defrouter_select: called unexpectedly (forwarding=%d, "
    647 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv));
    648 		splx(s);
    649 		return;
    650 	}
    651 
    652 	/*
    653 	 * Let's handle easy case (3) first:
    654 	 * If default router list is empty, there's nothing to be done.
    655 	 */
    656 	if (!TAILQ_FIRST(&nd_defrouter)) {
    657 		splx(s);
    658 		return;
    659 	}
    660 
    661 	/*
    662 	 * Search for a (probably) reachable router from the list.
    663 	 * We just pick up the first reachable one (if any), assuming that
    664 	 * the ordering rule of the list described in defrtrlist_update().
    665 	 */
    666 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    667 	     dr = TAILQ_NEXT(dr, dr_entry)) {
    668 		if (selected_dr == NULL &&
    669 		    (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) != NULL &&
    670 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) != NULL &&
    671 		    ND6_IS_LLINFO_PROBREACH(ln)) {
    672 			selected_dr = dr;
    673 		}
    674 
    675 		if (dr->installed && !installed_dr)
    676 			installed_dr = dr;
    677 		else if (dr->installed && installed_dr) {
    678 			/* this should not happen.  warn for diagnosis. */
    679 			log(LOG_ERR, "defrouter_select: more than one router"
    680 			    " is installed\n");
    681 		}
    682 	}
    683 	/*
    684 	 * If none of the default routers was found to be reachable,
    685 	 * round-robin the list regardless of preference.
    686 	 * Otherwise, if we have an installed router, check if the selected
    687 	 * (reachable) router should really be preferred to the installed one.
    688 	 * We only prefer the new router when the old one is not reachable
    689 	 * or when the new one has a really higher preference value.
    690 	 */
    691 	if (selected_dr == NULL) {
    692 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
    693 			selected_dr = TAILQ_FIRST(&nd_defrouter);
    694 		else
    695 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
    696 	} else if (installed_dr &&
    697 	    (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
    698 	    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
    699 	    ND6_IS_LLINFO_PROBREACH(ln) &&
    700 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
    701 		selected_dr = installed_dr;
    702 	}
    703 
    704 	/*
    705 	 * If the selected router is different than the installed one,
    706 	 * remove the installed router and install the selected one.
    707 	 * Note that the selected router is never NULL here.
    708 	 */
    709 	if (installed_dr != selected_dr) {
    710 		if (installed_dr)
    711 			defrouter_delreq(installed_dr);
    712 		defrouter_addreq(selected_dr);
    713 	}
    714 
    715 	splx(s);
    716 	return;
    717 }
    718 
    719 /*
    720  * for default router selection
    721  * regards router-preference field as a 2-bit signed integer
    722  */
    723 static int
    724 rtpref(struct nd_defrouter *dr)
    725 {
    726 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
    727 	case ND_RA_FLAG_RTPREF_HIGH:
    728 		return (RTPREF_HIGH);
    729 	case ND_RA_FLAG_RTPREF_MEDIUM:
    730 	case ND_RA_FLAG_RTPREF_RSV:
    731 		return (RTPREF_MEDIUM);
    732 	case ND_RA_FLAG_RTPREF_LOW:
    733 		return (RTPREF_LOW);
    734 	default:
    735 		/*
    736 		 * This case should never happen.  If it did, it would mean a
    737 		 * serious bug of kernel internal.  We thus always bark here.
    738 		 * Or, can we even panic?
    739 		 */
    740 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
    741 		return (RTPREF_INVALID);
    742 	}
    743 	/* NOTREACHED */
    744 }
    745 
    746 static struct nd_defrouter *
    747 defrtrlist_update(new)
    748 	struct nd_defrouter *new;
    749 {
    750 	struct nd_defrouter *dr, *n;
    751 	int s = splsoftnet();
    752 
    753 	if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
    754 		/* entry exists */
    755 		if (new->rtlifetime == 0) {
    756 			defrtrlist_del(dr);
    757 			dr = NULL;
    758 		} else {
    759 			int oldpref = rtpref(dr);
    760 
    761 			/* override */
    762 			dr->flags = new->flags; /* xxx flag check */
    763 			dr->rtlifetime = new->rtlifetime;
    764 			dr->expire = new->expire;
    765 
    766 			/*
    767 			 * If the preference does not change, there's no need
    768 			 * to sort the entries.
    769 			 */
    770 			if (rtpref(new) == oldpref) {
    771 				splx(s);
    772 				return (dr);
    773 			}
    774 
    775 			/*
    776 			 * preferred router may be changed, so relocate
    777 			 * this router.
    778 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
    779 			 * However, since defrtrlist_del() has many side
    780 			 * effects, we intentionally do so here.
    781 			 * defrouter_select() below will handle routing
    782 			 * changes later.
    783 			 */
    784 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
    785 			n = dr;
    786 			goto insert;
    787 		}
    788 		splx(s);
    789 		return (dr);
    790 	}
    791 
    792 	/* entry does not exist */
    793 	if (new->rtlifetime == 0) {
    794 		splx(s);
    795 		return (NULL);
    796 	}
    797 
    798 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
    799 	if (n == NULL) {
    800 		splx(s);
    801 		return (NULL);
    802 	}
    803 	bzero(n, sizeof(*n));
    804 	*n = *new;
    805 
    806 insert:
    807 	/*
    808 	 * Insert the new router in the Default Router List;
    809 	 * The Default Router List should be in the descending order
    810 	 * of router-preferece.  Routers with the same preference are
    811 	 * sorted in the arriving time order.
    812 	 */
    813 
    814 	/* insert at the end of the group */
    815 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    816 	     dr = TAILQ_NEXT(dr, dr_entry)) {
    817 		if (rtpref(n) > rtpref(dr))
    818 			break;
    819 	}
    820 	if (dr)
    821 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
    822 	else
    823 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
    824 
    825 	defrouter_select();
    826 
    827 	splx(s);
    828 
    829 	return (n);
    830 }
    831 
    832 static struct nd_pfxrouter *
    833 pfxrtr_lookup(pr, dr)
    834 	struct nd_prefix *pr;
    835 	struct nd_defrouter *dr;
    836 {
    837 	struct nd_pfxrouter *search;
    838 
    839 	for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
    840 		if (search->router == dr)
    841 			break;
    842 	}
    843 
    844 	return (search);
    845 }
    846 
    847 static void
    848 pfxrtr_add(pr, dr)
    849 	struct nd_prefix *pr;
    850 	struct nd_defrouter *dr;
    851 {
    852 	struct nd_pfxrouter *new;
    853 
    854 	new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
    855 	if (new == NULL)
    856 		return;
    857 	bzero(new, sizeof(*new));
    858 	new->router = dr;
    859 
    860 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
    861 
    862 	pfxlist_onlink_check();
    863 }
    864 
    865 static void
    866 pfxrtr_del(pfr)
    867 	struct nd_pfxrouter *pfr;
    868 {
    869 	LIST_REMOVE(pfr, pfr_entry);
    870 	free(pfr, M_IP6NDP);
    871 }
    872 
    873 struct nd_prefix *
    874 nd6_prefix_lookup(key)
    875 	struct nd_prefixctl *key;
    876 {
    877 	struct nd_prefix *search;
    878 
    879 	for (search = nd_prefix.lh_first; search; search = search->ndpr_next) {
    880 		if (key->ndpr_ifp == search->ndpr_ifp &&
    881 		    key->ndpr_plen == search->ndpr_plen &&
    882 		    in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
    883 		    &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
    884 			break;
    885 		}
    886 	}
    887 
    888 	return (search);
    889 }
    890 
    891 int
    892 nd6_prelist_add(pr, dr, newp)
    893 	struct nd_prefixctl *pr;
    894 	struct nd_prefix **newp;
    895 	struct nd_defrouter *dr;
    896 {
    897 	struct nd_prefix *new = NULL;
    898 	int i, s;
    899 	int error;
    900 
    901 	error = 0;
    902 	new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
    903 	if (new == NULL)
    904 		return ENOMEM;
    905 	bzero(new, sizeof(*new));
    906 	new->ndpr_ifp = pr->ndpr_ifp;
    907 	new->ndpr_prefix = pr->ndpr_prefix;
    908 	new->ndpr_plen = pr->ndpr_plen;
    909 	new->ndpr_vltime = pr->ndpr_vltime;
    910 	new->ndpr_pltime = pr->ndpr_pltime;
    911 	new->ndpr_flags = pr->ndpr_flags;
    912 	if ((error = in6_init_prefix_ltimes(new)) != 0) {
    913 		free(new, M_IP6NDP);
    914 		return(error);
    915 	}
    916 	new->ndpr_lastupdate = time_second;
    917 	if (newp != NULL)
    918 		*newp = new;
    919 
    920 	/* initialization */
    921 	LIST_INIT(&new->ndpr_advrtrs);
    922 	in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
    923 	/* make prefix in the canonical form */
    924 	for (i = 0; i < 4; i++)
    925 		new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
    926 		    new->ndpr_mask.s6_addr32[i];
    927 
    928 	s = splsoftnet();
    929 	/* link ndpr_entry to nd_prefix list */
    930 	LIST_INSERT_HEAD(&nd_prefix, new, ndpr_entry);
    931 	splx(s);
    932 
    933 	/* ND_OPT_PI_FLAG_ONLINK processing */
    934 	if (new->ndpr_raf_onlink) {
    935 		int e;
    936 
    937 		if ((e = nd6_prefix_onlink(new)) != 0) {
    938 			nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
    939 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
    940 			    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
    941 			    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
    942 			/* proceed anyway. XXX: is it correct? */
    943 		}
    944 	}
    945 
    946 	if (dr)
    947 		pfxrtr_add(new, dr);
    948 
    949 	return 0;
    950 }
    951 
    952 void
    953 prelist_remove(pr)
    954 	struct nd_prefix *pr;
    955 {
    956 	struct nd_pfxrouter *pfr, *next;
    957 	int e, s;
    958 
    959 	/* make sure to invalidate the prefix until it is really freed. */
    960 	pr->ndpr_vltime = 0;
    961 	pr->ndpr_pltime = 0;
    962 #if 0
    963 	/*
    964 	 * Though these flags are now meaningless, we'd rather keep the value
    965 	 * not to confuse users when executing "ndp -p".
    966 	 */
    967 	pr->ndpr_raf_onlink = 0;
    968 	pr->ndpr_raf_auto = 0;
    969 #endif
    970 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
    971 	    (e = nd6_prefix_offlink(pr)) != 0) {
    972 		nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
    973 		    "on %s, errno=%d\n",
    974 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
    975 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
    976 		/* what should we do? */
    977 	}
    978 
    979 	if (pr->ndpr_refcnt > 0)
    980 		return;		/* notice here? */
    981 
    982 	s = splsoftnet();
    983 	/* unlink ndpr_entry from nd_prefix list */
    984 	LIST_REMOVE(pr, ndpr_entry);
    985 
    986 	/* free list of routers that adversed the prefix */
    987 	for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
    988 		next = pfr->pfr_next;
    989 
    990 		free(pfr, M_IP6NDP);
    991 	}
    992 	splx(s);
    993 
    994 	free(pr, M_IP6NDP);
    995 
    996 	pfxlist_onlink_check();
    997 }
    998 
    999 static int
   1000 prelist_update(new, dr, m, mcast)
   1001 	struct nd_prefixctl *new;
   1002 	struct nd_defrouter *dr; /* may be NULL */
   1003 	struct mbuf *m;
   1004 	int mcast;
   1005 {
   1006 	struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
   1007 	struct ifaddr *ifa;
   1008 	struct ifnet *ifp = new->ndpr_ifp;
   1009 	struct nd_prefix *pr;
   1010 	int s = splsoftnet();
   1011 	int error = 0;
   1012 	int newprefix = 0;
   1013 	int auth;
   1014 	struct in6_addrlifetime lt6_tmp;
   1015 
   1016 	auth = 0;
   1017 	if (m) {
   1018 		/*
   1019 		 * Authenticity for NA consists authentication for
   1020 		 * both IP header and IP datagrams, doesn't it ?
   1021 		 */
   1022 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
   1023 		auth = (m->m_flags & M_AUTHIPHDR
   1024 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
   1025 #endif
   1026 	}
   1027 
   1028 	if ((pr = nd6_prefix_lookup(new)) != NULL) {
   1029 		/*
   1030 		 * nd6_prefix_lookup() ensures that pr and new have the same
   1031 		 * prefix on a same interface.
   1032 		 */
   1033 
   1034 		/*
   1035 		 * Update prefix information.  Note that the on-link (L) bit
   1036 		 * and the autonomous (A) bit should NOT be changed from 1
   1037 		 * to 0.
   1038 		 */
   1039 		if (new->ndpr_raf_onlink == 1)
   1040 			pr->ndpr_raf_onlink = 1;
   1041 		if (new->ndpr_raf_auto == 1)
   1042 			pr->ndpr_raf_auto = 1;
   1043 		if (new->ndpr_raf_onlink) {
   1044 			pr->ndpr_vltime = new->ndpr_vltime;
   1045 			pr->ndpr_pltime = new->ndpr_pltime;
   1046 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
   1047 			pr->ndpr_lastupdate = time_second;
   1048 		}
   1049 
   1050 		if (new->ndpr_raf_onlink &&
   1051 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
   1052 			int e;
   1053 
   1054 			if ((e = nd6_prefix_onlink(pr)) != 0) {
   1055 				nd6log((LOG_ERR,
   1056 				    "prelist_update: failed to make "
   1057 				    "the prefix %s/%d on-link on %s "
   1058 				    "(errno=%d)\n",
   1059 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1060 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
   1061 				/* proceed anyway. XXX: is it correct? */
   1062 			}
   1063 		}
   1064 
   1065 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
   1066 			pfxrtr_add(pr, dr);
   1067 	} else {
   1068 		struct nd_prefix *newpr = NULL;
   1069 
   1070 		newprefix = 1;
   1071 
   1072 		if (new->ndpr_vltime == 0)
   1073 			goto end;
   1074 		if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
   1075 			goto end;
   1076 
   1077 		error = nd6_prelist_add(new, dr, &newpr);
   1078 		if (error != 0 || newpr == NULL) {
   1079 			nd6log((LOG_NOTICE, "prelist_update: "
   1080 			    "nd6_prelist_add failed for %s/%d on %s "
   1081 			    "errno=%d, returnpr=%p\n",
   1082 			    ip6_sprintf(&new->ndpr_prefix.sin6_addr),
   1083 			    new->ndpr_plen, if_name(new->ndpr_ifp),
   1084 			    error, newpr));
   1085 			goto end; /* we should just give up in this case. */
   1086 		}
   1087 
   1088 		/*
   1089 		 * XXX: from the ND point of view, we can ignore a prefix
   1090 		 * with the on-link bit being zero.  However, we need a
   1091 		 * prefix structure for references from autoconfigured
   1092 		 * addresses.  Thus, we explicitly make sure that the prefix
   1093 		 * itself expires now.
   1094 		 */
   1095 		if (newpr->ndpr_raf_onlink == 0) {
   1096 			newpr->ndpr_vltime = 0;
   1097 			newpr->ndpr_pltime = 0;
   1098 			in6_init_prefix_ltimes(newpr);
   1099 		}
   1100 
   1101 		pr = newpr;
   1102 	}
   1103 
   1104 	/*
   1105 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
   1106 	 * Note that pr must be non NULL at this point.
   1107 	 */
   1108 
   1109 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
   1110 	if (!new->ndpr_raf_auto)
   1111 		goto end;
   1112 
   1113 	/*
   1114 	 * 5.5.3 (b). the link-local prefix should have been ignored in
   1115 	 * nd6_ra_input.
   1116 	 */
   1117 
   1118 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
   1119 	if (new->ndpr_pltime > new->ndpr_vltime) {
   1120 		error = EINVAL;	/* XXX: won't be used */
   1121 		goto end;
   1122 	}
   1123 
   1124 	/*
   1125 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
   1126 	 * an address configured by stateless autoconfiguration already in the
   1127 	 * list of addresses associated with the interface, and the Valid
   1128 	 * Lifetime is not 0, form an address.  We first check if we have
   1129 	 * a matching prefix.
   1130 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
   1131 	 * consider autoconfigured addresses while RFC2462 simply said
   1132 	 * "address".
   1133 	 */
   1134 	for (ifa = ifp->if_addrlist.tqh_first; ifa; ifa = ifa->ifa_list.tqe_next)
   1135 	{
   1136 		struct in6_ifaddr *ifa6;
   1137 		u_int32_t remaininglifetime;
   1138 
   1139 		if (ifa->ifa_addr->sa_family != AF_INET6)
   1140 			continue;
   1141 
   1142 		ifa6 = (struct in6_ifaddr *)ifa;
   1143 
   1144 		/*
   1145 		 * We only consider autoconfigured addresses as per rfc2462bis.
   1146 		 */
   1147 		if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
   1148 			continue;
   1149 
   1150 		/*
   1151 		 * Spec is not clear here, but I believe we should concentrate
   1152 		 * on unicast (i.e. not anycast) addresses.
   1153 		 * XXX: other ia6_flags? detached or duplicated?
   1154 		 */
   1155 		if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
   1156 			continue;
   1157 
   1158 		/*
   1159 		 * Ignore the address if it is not associated with a prefix
   1160 		 * or is associated with a prefix that is different from this
   1161 		 * one.  (pr is never NULL here)
   1162 		 */
   1163 		if (ifa6->ia6_ndpr != pr)
   1164 			continue;
   1165 
   1166 		if (ia6_match == NULL) /* remember the first one */
   1167 			ia6_match = ifa6;
   1168 
   1169 		/*
   1170 		 * An already autoconfigured address matched.  Now that we
   1171 		 * are sure there is at least one matched address, we can
   1172 		 * proceed to 5.5.3. (e): update the lifetimes according to the
   1173 		 * "two hours" rule and the privacy extension.
   1174 		 * We apply some clarifications in rfc2462bis:
   1175 		 * - use remaininglifetime instead of storedlifetime as a
   1176 		 *   variable name
   1177 		 * - remove the dead code in the "two-hour" rule
   1178 		 */
   1179 #define TWOHOUR		(120*60)
   1180 		lt6_tmp = ifa6->ia6_lifetime;
   1181 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
   1182 			remaininglifetime = ND6_INFINITE_LIFETIME;
   1183 		else if (time_second - ifa6->ia6_updatetime >
   1184 			 lt6_tmp.ia6t_vltime) {
   1185 			/*
   1186 			 * The case of "invalid" address.  We should usually
   1187 			 * not see this case.
   1188 			 */
   1189 			remaininglifetime = 0;
   1190 		} else
   1191 			remaininglifetime = lt6_tmp.ia6t_vltime -
   1192 			    (time_second - ifa6->ia6_updatetime);
   1193 
   1194 		/* when not updating, keep the current stored lifetime. */
   1195 		lt6_tmp.ia6t_vltime = remaininglifetime;
   1196 
   1197 		if (TWOHOUR < new->ndpr_vltime ||
   1198 		    remaininglifetime < new->ndpr_vltime) {
   1199 			lt6_tmp.ia6t_vltime = new->ndpr_vltime;
   1200 		} else if (remaininglifetime <= TWOHOUR) {
   1201 			if (auth)
   1202 				lt6_tmp.ia6t_vltime = new->ndpr_vltime;
   1203 		} else {
   1204 			/*
   1205 			 * new->ndpr_vltime <= TWOHOUR &&
   1206 			 * TWOHOUR < remaininglifetime
   1207 			 */
   1208 			lt6_tmp.ia6t_vltime = TWOHOUR;
   1209 		}
   1210 
   1211 		/* The 2 hour rule is not imposed for preferred lifetime. */
   1212 		lt6_tmp.ia6t_pltime = new->ndpr_pltime;
   1213 
   1214 		in6_init_address_ltimes(pr, &lt6_tmp);
   1215 
   1216 		/*
   1217 		 * We need to treat lifetimes for temporary addresses
   1218 		 * differently, according to
   1219 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
   1220 		 * we only update the lifetimes when they are in the maximum
   1221 		 * intervals.
   1222 		 */
   1223 		if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
   1224 			u_int32_t maxvltime, maxpltime;
   1225 
   1226 			if (ip6_temp_valid_lifetime >
   1227 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
   1228 			    ip6_desync_factor)) {
   1229 				maxvltime = ip6_temp_valid_lifetime -
   1230 				    (time_second - ifa6->ia6_createtime) -
   1231 				    ip6_desync_factor;
   1232 			} else
   1233 				maxvltime = 0;
   1234 			if (ip6_temp_preferred_lifetime >
   1235 			    (u_int32_t)((time_second - ifa6->ia6_createtime) +
   1236 			    ip6_desync_factor)) {
   1237 				maxpltime = ip6_temp_preferred_lifetime -
   1238 				    (time_second - ifa6->ia6_createtime) -
   1239 				    ip6_desync_factor;
   1240 			} else
   1241 				maxpltime = 0;
   1242 
   1243 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
   1244 			    lt6_tmp.ia6t_vltime > maxvltime) {
   1245 				lt6_tmp.ia6t_vltime = maxvltime;
   1246 			}
   1247 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
   1248 			    lt6_tmp.ia6t_pltime > maxpltime) {
   1249 				lt6_tmp.ia6t_pltime = maxpltime;
   1250 			}
   1251 		}
   1252 
   1253 		ifa6->ia6_lifetime = lt6_tmp;
   1254 		ifa6->ia6_updatetime = time_second;
   1255 	}
   1256 	if (ia6_match == NULL && new->ndpr_vltime) {
   1257 		int ifidlen;
   1258 
   1259 		/*
   1260 		 * 5.5.3 (d) (continued)
   1261 		 * No address matched and the valid lifetime is non-zero.
   1262 		 * Create a new address.
   1263 		 */
   1264 
   1265 		/*
   1266 		 * Prefix Length check:
   1267 		 * If the sum of the prefix length and interface identifier
   1268 		 * length does not equal 128 bits, the Prefix Information
   1269 		 * option MUST be ignored.  The length of the interface
   1270 		 * identifier is defined in a separate link-type specific
   1271 		 * document.
   1272 		 */
   1273 		ifidlen = in6_if2idlen(ifp);
   1274 		if (ifidlen < 0) {
   1275 			/* this should not happen, so we always log it. */
   1276 			log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
   1277 			    if_name(ifp));
   1278 			goto end;
   1279 		}
   1280 		if (ifidlen + pr->ndpr_plen != 128) {
   1281 			nd6log((LOG_INFO,
   1282 			    "prelist_update: invalid prefixlen "
   1283 			    "%d for %s, ignored\n",
   1284 			    pr->ndpr_plen, if_name(ifp)));
   1285 			goto end;
   1286 		}
   1287 
   1288 		if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
   1289 			/*
   1290 			 * note that we should use pr (not new) for reference.
   1291 			 */
   1292 			pr->ndpr_refcnt++;
   1293 			ia6->ia6_ndpr = pr;
   1294 
   1295 			/*
   1296 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
   1297 			 * When a new public address is created as described
   1298 			 * in RFC2462, also create a new temporary address.
   1299 			 *
   1300 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
   1301 			 * When an interface connects to a new link, a new
   1302 			 * randomized interface identifier should be generated
   1303 			 * immediately together with a new set of temporary
   1304 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
   1305 			 * in6_tmpifadd().
   1306 			 */
   1307 			if (ip6_use_tempaddr) {
   1308 				int e;
   1309 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
   1310 					nd6log((LOG_NOTICE, "prelist_update: "
   1311 					    "failed to create a temporary "
   1312 					    "address, errno=%d\n",
   1313 					    e));
   1314 				}
   1315 			}
   1316 
   1317 			/*
   1318 			 * A newly added address might affect the status
   1319 			 * of other addresses, so we check and update it.
   1320 			 * XXX: what if address duplication happens?
   1321 			 */
   1322 			pfxlist_onlink_check();
   1323 		} else {
   1324 			/* just set an error. do not bark here. */
   1325 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
   1326 		}
   1327 	}
   1328 
   1329  end:
   1330 	splx(s);
   1331 	return error;
   1332 }
   1333 
   1334 /*
   1335  * A supplement function used in the on-link detection below;
   1336  * detect if a given prefix has a (probably) reachable advertising router.
   1337  * XXX: lengthy function name...
   1338  */
   1339 static struct nd_pfxrouter *
   1340 find_pfxlist_reachable_router(pr)
   1341 	struct nd_prefix *pr;
   1342 {
   1343 	struct nd_pfxrouter *pfxrtr;
   1344 	struct rtentry *rt;
   1345 	struct llinfo_nd6 *ln;
   1346 
   1347 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
   1348 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
   1349 		if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0,
   1350 		    pfxrtr->router->ifp)) &&
   1351 		    (ln = (struct llinfo_nd6 *)rt->rt_llinfo) &&
   1352 		    ND6_IS_LLINFO_PROBREACH(ln))
   1353 			break;	/* found */
   1354 	}
   1355 
   1356 	return (pfxrtr);
   1357 }
   1358 
   1359 /*
   1360  * Check if each prefix in the prefix list has at least one available router
   1361  * that advertised the prefix (a router is "available" if its neighbor cache
   1362  * entry is reachable or probably reachable).
   1363  * If the check fails, the prefix may be off-link, because, for example,
   1364  * we have moved from the network but the lifetime of the prefix has not
   1365  * expired yet.  So we should not use the prefix if there is another prefix
   1366  * that has an available router.
   1367  * But, if there is no prefix that has an available router, we still regards
   1368  * all the prefixes as on-link.  This is because we can't tell if all the
   1369  * routers are simply dead or if we really moved from the network and there
   1370  * is no router around us.
   1371  */
   1372 void
   1373 pfxlist_onlink_check()
   1374 {
   1375 	struct nd_prefix *pr;
   1376 	struct in6_ifaddr *ifa;
   1377 	struct nd_defrouter *dr;
   1378 	struct nd_pfxrouter *pfxrtr = NULL;
   1379 
   1380 	/*
   1381 	 * Check if there is a prefix that has a reachable advertising
   1382 	 * router.
   1383 	 */
   1384 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
   1385 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
   1386 			break;
   1387 	}
   1388 	/*
   1389 	 * If we have no such prefix, check whether we still have a router
   1390 	 * that does not advertise any prefixes.
   1391 	 */
   1392 	if (pr == NULL) {
   1393 		for (dr = TAILQ_FIRST(&nd_defrouter); dr;
   1394 		    dr = TAILQ_NEXT(dr, dr_entry)) {
   1395 			struct nd_prefix *pr0;
   1396 
   1397 			for (pr0 = nd_prefix.lh_first; pr0;
   1398 			    pr0 = pr0->ndpr_next) {
   1399 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
   1400 					break;
   1401 			}
   1402 			if (pfxrtr)
   1403 				break;
   1404 		}
   1405 	}
   1406 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
   1407 		/*
   1408 		 * There is at least one prefix that has a reachable router,
   1409 		 * or at least a router which probably does not advertise
   1410 		 * any prefixes.  The latter would be the case when we move
   1411 		 * to a new link where we have a router that does not provide
   1412 		 * prefixes and we configure an address by hand.
   1413 		 * Detach prefixes which have no reachable advertising
   1414 		 * router, and attach other prefixes.
   1415 		 */
   1416 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
   1417 			/* XXX: a link-local prefix should never be detached */
   1418 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1419 				continue;
   1420 
   1421 			/*
   1422 			 * we aren't interested in prefixes without the L bit
   1423 			 * set.
   1424 			 */
   1425 			if (pr->ndpr_raf_onlink == 0)
   1426 				continue;
   1427 
   1428 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
   1429 			    find_pfxlist_reachable_router(pr) == NULL)
   1430 				pr->ndpr_stateflags |= NDPRF_DETACHED;
   1431 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
   1432 			    find_pfxlist_reachable_router(pr) != 0)
   1433 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
   1434 		}
   1435 	} else {
   1436 		/* there is no prefix that has a reachable router */
   1437 		for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
   1438 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1439 				continue;
   1440 
   1441 			if (pr->ndpr_raf_onlink == 0)
   1442 				continue;
   1443 
   1444 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
   1445 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
   1446 		}
   1447 	}
   1448 
   1449 	/*
   1450 	 * Remove each interface route associated with a (just) detached
   1451 	 * prefix, and reinstall the interface route for a (just) attached
   1452 	 * prefix.  Note that all attempt of reinstallation does not
   1453 	 * necessarily success, when a same prefix is shared among multiple
   1454 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
   1455 	 * so we don't have to care about them.
   1456 	 */
   1457 	for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
   1458 		int e;
   1459 
   1460 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1461 			continue;
   1462 
   1463 		if (pr->ndpr_raf_onlink == 0)
   1464 			continue;
   1465 
   1466 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
   1467 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
   1468 			if ((e = nd6_prefix_offlink(pr)) != 0) {
   1469 				nd6log((LOG_ERR,
   1470 				    "pfxlist_onlink_check: failed to "
   1471 				    "make %s/%d offlink, errno=%d\n",
   1472 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1473 				    pr->ndpr_plen, e));
   1474 			}
   1475 		}
   1476 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
   1477 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
   1478 		    pr->ndpr_raf_onlink) {
   1479 			if ((e = nd6_prefix_onlink(pr)) != 0) {
   1480 				nd6log((LOG_ERR,
   1481 				    "pfxlist_onlink_check: failed to "
   1482 				    "make %s/%d onlink, errno=%d\n",
   1483 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1484 				    pr->ndpr_plen, e));
   1485 			}
   1486 		}
   1487 	}
   1488 
   1489 	/*
   1490 	 * Changes on the prefix status might affect address status as well.
   1491 	 * Make sure that all addresses derived from an attached prefix are
   1492 	 * attached, and that all addresses derived from a detached prefix are
   1493 	 * detached.  Note, however, that a manually configured address should
   1494 	 * always be attached.
   1495 	 * The precise detection logic is same as the one for prefixes.
   1496 	 */
   1497 	for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
   1498 		if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
   1499 			continue;
   1500 
   1501 		if (ifa->ia6_ndpr == NULL) {
   1502 			/*
   1503 			 * This can happen when we first configure the address
   1504 			 * (i.e. the address exists, but the prefix does not).
   1505 			 * XXX: complicated relationships...
   1506 			 */
   1507 			continue;
   1508 		}
   1509 
   1510 		if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
   1511 			break;
   1512 	}
   1513 	if (ifa) {
   1514 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
   1515 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
   1516 				continue;
   1517 
   1518 			if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
   1519 				continue;
   1520 
   1521 			if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
   1522 				if (ifa->ia6_flags & IN6_IFF_DETACHED) {
   1523 					ifa->ia6_flags &= ~IN6_IFF_DETACHED;
   1524 					ifa->ia6_flags |= IN6_IFF_TENTATIVE;
   1525 					nd6_dad_start((struct ifaddr *)ifa,
   1526 					    0);
   1527 				}
   1528 			} else {
   1529 				if ((ifa->ia6_flags & IN6_IFF_DETACHED) == 0) {
   1530 					ifa->ia6_flags |= IN6_IFF_DETACHED;
   1531 				}
   1532 			}
   1533 		}
   1534 	}
   1535 	else {
   1536 		for (ifa = in6_ifaddr; ifa; ifa = ifa->ia_next) {
   1537 			if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
   1538 				continue;
   1539 
   1540 			if (ifa->ia6_flags & IN6_IFF_DETACHED) {
   1541 				ifa->ia6_flags &= ~IN6_IFF_DETACHED;
   1542 				ifa->ia6_flags |= IN6_IFF_TENTATIVE;
   1543 				/* Do we need a delay in this case? */
   1544 				nd6_dad_start((struct ifaddr *)ifa, 0);
   1545 			}
   1546 		}
   1547 	}
   1548 }
   1549 
   1550 int
   1551 nd6_prefix_onlink(pr)
   1552 	struct nd_prefix *pr;
   1553 {
   1554 	struct ifaddr *ifa;
   1555 	struct ifnet *ifp = pr->ndpr_ifp;
   1556 	struct sockaddr_in6 mask6;
   1557 	struct nd_prefix *opr;
   1558 	u_long rtflags;
   1559 	int error = 0;
   1560 	struct rtentry *rt = NULL;
   1561 
   1562 	/* sanity check */
   1563 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
   1564 		nd6log((LOG_ERR,
   1565 		    "nd6_prefix_onlink: %s/%d is already on-link\n",
   1566 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
   1567 		return (EEXIST);
   1568 	}
   1569 
   1570 	/*
   1571 	 * Add the interface route associated with the prefix.  Before
   1572 	 * installing the route, check if there's the same prefix on another
   1573 	 * interface, and the prefix has already installed the interface route.
   1574 	 * Although such a configuration is expected to be rare, we explicitly
   1575 	 * allow it.
   1576 	 */
   1577 	for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
   1578 		if (opr == pr)
   1579 			continue;
   1580 
   1581 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
   1582 			continue;
   1583 
   1584 		if (opr->ndpr_plen == pr->ndpr_plen &&
   1585 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
   1586 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
   1587 			return (0);
   1588 	}
   1589 
   1590 	/*
   1591 	 * We prefer link-local addresses as the associated interface address.
   1592 	 */
   1593 	/* search for a link-local addr */
   1594 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
   1595 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
   1596 	if (ifa == NULL) {
   1597 		/* XXX: freebsd does not have ifa_ifwithaf */
   1598 		for (ifa = ifp->if_addrlist.tqh_first;
   1599 		     ifa;
   1600 		     ifa = ifa->ifa_list.tqe_next)
   1601 		{
   1602 			if (ifa->ifa_addr->sa_family == AF_INET6)
   1603 				break;
   1604 		}
   1605 		/* should we care about ia6_flags? */
   1606 	}
   1607 	if (ifa == NULL) {
   1608 		/*
   1609 		 * This can still happen, when, for example, we receive an RA
   1610 		 * containing a prefix with the L bit set and the A bit clear,
   1611 		 * after removing all IPv6 addresses on the receiving
   1612 		 * interface.  This should, of course, be rare though.
   1613 		 */
   1614 		nd6log((LOG_NOTICE,
   1615 		    "nd6_prefix_onlink: failed to find any ifaddr"
   1616 		    " to add route for a prefix(%s/%d) on %s\n",
   1617 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1618 		    pr->ndpr_plen, if_name(ifp)));
   1619 		return (0);
   1620 	}
   1621 
   1622 	/*
   1623 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
   1624 	 * ifa->ifa_rtrequest = nd6_rtrequest;
   1625 	 */
   1626 	bzero(&mask6, sizeof(mask6));
   1627 	mask6.sin6_len = sizeof(mask6);
   1628 	mask6.sin6_addr = pr->ndpr_mask;
   1629 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
   1630 	rtflags = ifa->ifa_flags | RTF_UP;
   1631 	if (nd6_need_cache(ifp)) {
   1632 		/* explicitly set in case ifa_flags does not set the flag. */
   1633 		rtflags |= RTF_CLONING;
   1634 	} else {
   1635 		/*
   1636 		 * explicitly clear the cloning bit in case ifa_flags sets it.
   1637 		 */
   1638 		rtflags &= ~RTF_CLONING;
   1639 	}
   1640 	error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
   1641 	    ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
   1642 	if (error == 0) {
   1643 		if (rt != NULL) /* this should be non NULL, though */
   1644 			nd6_rtmsg(RTM_ADD, rt);
   1645 		pr->ndpr_stateflags |= NDPRF_ONLINK;
   1646 	} else {
   1647 		nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
   1648 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
   1649 		    "errno = %d\n",
   1650 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1651 		    pr->ndpr_plen, if_name(ifp),
   1652 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
   1653 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error));
   1654 	}
   1655 
   1656 	if (rt != NULL)
   1657 		rt->rt_refcnt--;
   1658 
   1659 	return (error);
   1660 }
   1661 
   1662 int
   1663 nd6_prefix_offlink(pr)
   1664 	struct nd_prefix *pr;
   1665 {
   1666 	int error = 0;
   1667 	struct ifnet *ifp = pr->ndpr_ifp;
   1668 	struct nd_prefix *opr;
   1669 	struct sockaddr_in6 sa6, mask6;
   1670 	struct rtentry *rt = NULL;
   1671 
   1672 	/* sanity check */
   1673 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
   1674 		nd6log((LOG_ERR,
   1675 		    "nd6_prefix_offlink: %s/%d is already off-link\n",
   1676 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen));
   1677 		return (EEXIST);
   1678 	}
   1679 
   1680 	bzero(&sa6, sizeof(sa6));
   1681 	sa6.sin6_family = AF_INET6;
   1682 	sa6.sin6_len = sizeof(sa6);
   1683 	bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
   1684 	    sizeof(struct in6_addr));
   1685 	bzero(&mask6, sizeof(mask6));
   1686 	mask6.sin6_family = AF_INET6;
   1687 	mask6.sin6_len = sizeof(sa6);
   1688 	bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
   1689 	error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
   1690 	    (struct sockaddr *)&mask6, 0, &rt);
   1691 	if (error == 0) {
   1692 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
   1693 
   1694 		/* report the route deletion to the routing socket. */
   1695 		if (rt != NULL)
   1696 			nd6_rtmsg(RTM_DELETE, rt);
   1697 
   1698 		/*
   1699 		 * There might be the same prefix on another interface,
   1700 		 * the prefix which could not be on-link just because we have
   1701 		 * the interface route (see comments in nd6_prefix_onlink).
   1702 		 * If there's one, try to make the prefix on-link on the
   1703 		 * interface.
   1704 		 */
   1705 		for (opr = nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
   1706 			if (opr == pr)
   1707 				continue;
   1708 
   1709 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
   1710 				continue;
   1711 
   1712 			/*
   1713 			 * KAME specific: detached prefixes should not be
   1714 			 * on-link.
   1715 			 */
   1716 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
   1717 				continue;
   1718 
   1719 			if (opr->ndpr_plen == pr->ndpr_plen &&
   1720 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
   1721 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
   1722 				int e;
   1723 
   1724 				if ((e = nd6_prefix_onlink(opr)) != 0) {
   1725 					nd6log((LOG_ERR,
   1726 					    "nd6_prefix_offlink: failed to "
   1727 					    "recover a prefix %s/%d from %s "
   1728 					    "to %s (errno = %d)\n",
   1729 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
   1730 					    opr->ndpr_plen, if_name(ifp),
   1731 					    if_name(opr->ndpr_ifp), e));
   1732 				}
   1733 			}
   1734 		}
   1735 	} else {
   1736 		/* XXX: can we still set the NDPRF_ONLINK flag? */
   1737 		nd6log((LOG_ERR,
   1738 		    "nd6_prefix_offlink: failed to delete route: "
   1739 		    "%s/%d on %s (errno = %d)\n",
   1740 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
   1741 		    error));
   1742 	}
   1743 
   1744 	if (rt != NULL) {
   1745 		if (rt->rt_refcnt <= 0) {
   1746 			/* XXX: we should free the entry ourselves. */
   1747 			rt->rt_refcnt++;
   1748 			rtfree(rt);
   1749 		}
   1750 	}
   1751 
   1752 	return (error);
   1753 }
   1754 
   1755 static struct in6_ifaddr *
   1756 in6_ifadd(pr, mcast)
   1757 	struct nd_prefixctl *pr;
   1758 	int mcast;
   1759 {
   1760 	struct ifnet *ifp = pr->ndpr_ifp;
   1761 	struct ifaddr *ifa;
   1762 	struct in6_aliasreq ifra;
   1763 	struct in6_ifaddr *ia, *ib;
   1764 	int error, plen0;
   1765 	struct in6_addr mask;
   1766 	int prefixlen = pr->ndpr_plen;
   1767 	int updateflags;
   1768 
   1769 	in6_prefixlen2mask(&mask, prefixlen);
   1770 
   1771 	/*
   1772 	 * find a link-local address (will be interface ID).
   1773 	 * Is it really mandatory? Theoretically, a global or a site-local
   1774 	 * address can be configured without a link-local address, if we
   1775 	 * have a unique interface identifier...
   1776 	 *
   1777 	 * it is not mandatory to have a link-local address, we can generate
   1778 	 * interface identifier on the fly.  we do this because:
   1779 	 * (1) it should be the easiest way to find interface identifier.
   1780 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
   1781 	 * for multiple addresses on a single interface, and possible shortcut
   1782 	 * of DAD.  we omitted DAD for this reason in the past.
   1783 	 * (3) a user can prevent autoconfiguration of global address
   1784 	 * by removing link-local address by hand (this is partly because we
   1785 	 * don't have other way to control the use of IPv6 on an interface.
   1786 	 * this has been our design choice - cf. NRL's "ifconfig auto").
   1787 	 * (4) it is easier to manage when an interface has addresses
   1788 	 * with the same interface identifier, than to have multiple addresses
   1789 	 * with different interface identifiers.
   1790 	 */
   1791 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
   1792 	if (ifa)
   1793 		ib = (struct in6_ifaddr *)ifa;
   1794 	else
   1795 		return NULL;
   1796 
   1797 #if 0 /* don't care link local addr state, and always do DAD */
   1798 	/* if link-local address is not eligible, do not autoconfigure. */
   1799 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
   1800 		printf("in6_ifadd: link-local address not ready\n");
   1801 		return NULL;
   1802 	}
   1803 #endif
   1804 
   1805 	/* prefixlen + ifidlen must be equal to 128 */
   1806 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
   1807 	if (prefixlen != plen0) {
   1808 		nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
   1809 		    "(prefix=%d ifid=%d)\n",
   1810 		    if_name(ifp), prefixlen, 128 - plen0));
   1811 		return NULL;
   1812 	}
   1813 
   1814 	/* make ifaddr */
   1815 
   1816 	memset(&ifra, 0, sizeof(ifra));
   1817 	/*
   1818 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
   1819 	 * for safety.
   1820 	 */
   1821 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
   1822 	ifra.ifra_addr.sin6_family = AF_INET6;
   1823 	ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
   1824 	/* prefix */
   1825 	ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
   1826 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
   1827 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
   1828 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
   1829 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
   1830 
   1831 	/* interface ID */
   1832 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
   1833 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
   1834 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
   1835 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
   1836 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
   1837 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
   1838 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
   1839 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
   1840 
   1841 	/* new prefix mask. */
   1842 	ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
   1843 	ifra.ifra_prefixmask.sin6_family = AF_INET6;
   1844 	bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
   1845 	    sizeof(ifra.ifra_prefixmask.sin6_addr));
   1846 
   1847 	/* lifetimes */
   1848 	ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
   1849 	ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
   1850 
   1851 	/* XXX: scope zone ID? */
   1852 
   1853 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
   1854 
   1855 	/*
   1856 	 * Make sure that we do not have this address already.  This should
   1857 	 * usually not happen, but we can still see this case, e.g., if we
   1858 	 * have manually configured the exact address to be configured.
   1859 	 */
   1860 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
   1861 		/* this should be rare enough to make an explicit log */
   1862 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
   1863 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
   1864 		return (NULL);
   1865 	}
   1866 
   1867 	/*
   1868 	 * Allocate ifaddr structure, link into chain, etc.
   1869 	 * If we are going to create a new address upon receiving a multicasted
   1870 	 * RA, we need to impose a random delay before starting DAD.
   1871 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
   1872 	 */
   1873 	updateflags = 0;
   1874 	if (mcast)
   1875 		updateflags |= IN6_IFAUPDATE_DADDELAY;
   1876 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
   1877 		nd6log((LOG_ERR,
   1878 		    "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
   1879 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
   1880 		    error));
   1881 		return (NULL);	/* ifaddr must not have been allocated. */
   1882 	}
   1883 
   1884 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
   1885 
   1886 	return (ia);		/* this is always non-NULL */
   1887 }
   1888 
   1889 int
   1890 in6_tmpifadd(ia0, forcegen, dad_delay)
   1891 	const struct in6_ifaddr *ia0; /* corresponding public address */
   1892 	int forcegen, dad_delay;
   1893 {
   1894 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
   1895 	struct in6_ifaddr *newia, *ia;
   1896 	struct in6_aliasreq ifra;
   1897 	int i, error;
   1898 	int trylimit = 3;	/* XXX: adhoc value */
   1899 	int updateflags;
   1900 	u_int32_t randid[2];
   1901 	u_int32_t vltime0, pltime0;
   1902 
   1903 	memset(&ifra, 0, sizeof(ifra));
   1904 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
   1905 	ifra.ifra_addr = ia0->ia_addr;
   1906 	/* copy prefix mask */
   1907 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
   1908 	/* clear the old IFID */
   1909 	for (i = 0; i < 4; i++) {
   1910 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
   1911 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
   1912 	}
   1913 
   1914   again:
   1915 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
   1916 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
   1917 		nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
   1918 		    "random IFID\n"));
   1919 		return (EINVAL);
   1920 	}
   1921 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
   1922 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
   1923 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
   1924 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
   1925 
   1926 	/*
   1927 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
   1928 	 * However, we may still have a chance to see collision, because
   1929 	 * there may be a time lag between generation of the ID and generation
   1930 	 * of the address.  So, we'll do one more sanity check.
   1931 	 */
   1932 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
   1933 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
   1934 		    &ifra.ifra_addr.sin6_addr)) {
   1935 			if (trylimit-- == 0) {
   1936 				/*
   1937 				 * Give up.  Something strange should have
   1938 				 * happened.
   1939 				 */
   1940 				nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
   1941 				    "find a unique random IFID\n"));
   1942 				return (EEXIST);
   1943 			}
   1944 			forcegen = 1;
   1945 			goto again;
   1946 		}
   1947 	}
   1948 
   1949 	/*
   1950 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
   1951          * public address or TEMP_VALID_LIFETIME.
   1952 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
   1953          * of the public address or TEMP_PREFERRED_LIFETIME -
   1954          * DESYNC_FACTOR.
   1955 	 */
   1956 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
   1957 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
   1958 		    (ia0->ia6_lifetime.ia6t_vltime -
   1959 		    (time_second - ia0->ia6_updatetime));
   1960 		if (vltime0 > ip6_temp_valid_lifetime)
   1961 			vltime0 = ip6_temp_valid_lifetime;
   1962 	} else
   1963 		vltime0 = ip6_temp_valid_lifetime;
   1964 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
   1965 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
   1966 		    (ia0->ia6_lifetime.ia6t_pltime -
   1967 		    (time_second - ia0->ia6_updatetime));
   1968 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
   1969 			pltime0 = ip6_temp_preferred_lifetime -
   1970 			    ip6_desync_factor;
   1971 		}
   1972 	} else
   1973 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
   1974 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
   1975 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
   1976 
   1977 	/*
   1978 	 * A temporary address is created only if this calculated Preferred
   1979 	 * Lifetime is greater than REGEN_ADVANCE time units.
   1980 	 */
   1981 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
   1982 		return (0);
   1983 
   1984 	/* XXX: scope zone ID? */
   1985 
   1986 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
   1987 
   1988 	/* allocate ifaddr structure, link into chain, etc. */
   1989 	updateflags = 0;
   1990 	if (dad_delay)
   1991 		updateflags |= IN6_IFAUPDATE_DADDELAY;
   1992 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
   1993 		return (error);
   1994 
   1995 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
   1996 	if (newia == NULL) {	/* XXX: can it happen? */
   1997 		nd6log((LOG_ERR,
   1998 		    "in6_tmpifadd: ifa update succeeded, but we got "
   1999 		    "no ifaddr\n"));
   2000 		return (EINVAL); /* XXX */
   2001 	}
   2002 	newia->ia6_ndpr = ia0->ia6_ndpr;
   2003 	newia->ia6_ndpr->ndpr_refcnt++;
   2004 
   2005 	/*
   2006 	 * A newly added address might affect the status of other addresses.
   2007 	 * XXX: when the temporary address is generated with a new public
   2008 	 * address, the onlink check is redundant.  However, it would be safe
   2009 	 * to do the check explicitly everywhere a new address is generated,
   2010 	 * and, in fact, we surely need the check when we create a new
   2011 	 * temporary address due to deprecation of an old temporary address.
   2012 	 */
   2013 	pfxlist_onlink_check();
   2014 
   2015 	return (0);
   2016 }
   2017 
   2018 static int
   2019 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
   2020 {
   2021 
   2022 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
   2023 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
   2024 		nd6log((LOG_INFO, "in6_init_prefix_ltimes: preferred lifetime"
   2025 		    "(%d) is greater than valid lifetime(%d)\n",
   2026 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime));
   2027 		return (EINVAL);
   2028 	}
   2029 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
   2030 		ndpr->ndpr_preferred = 0;
   2031 	else
   2032 		ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
   2033 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
   2034 		ndpr->ndpr_expire = 0;
   2035 	else
   2036 		ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
   2037 
   2038 	return 0;
   2039 }
   2040 
   2041 static void
   2042 in6_init_address_ltimes(struct nd_prefix *new __unused,
   2043     struct in6_addrlifetime *lt6)
   2044 {
   2045 
   2046 	/* Valid lifetime must not be updated unless explicitly specified. */
   2047 	/* init ia6t_expire */
   2048 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
   2049 		lt6->ia6t_expire = 0;
   2050 	else {
   2051 		lt6->ia6t_expire = time_second;
   2052 		lt6->ia6t_expire += lt6->ia6t_vltime;
   2053 	}
   2054 
   2055 	/* init ia6t_preferred */
   2056 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
   2057 		lt6->ia6t_preferred = 0;
   2058 	else {
   2059 		lt6->ia6t_preferred = time_second;
   2060 		lt6->ia6t_preferred += lt6->ia6t_pltime;
   2061 	}
   2062 }
   2063 
   2064 /*
   2065  * Delete all the routing table entries that use the specified gateway.
   2066  * XXX: this function causes search through all entries of routing table, so
   2067  * it shouldn't be called when acting as a router.
   2068  */
   2069 void
   2070 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp __unused)
   2071 {
   2072 	struct radix_node_head *rnh = rt_tables[AF_INET6];
   2073 	int s = splsoftnet();
   2074 
   2075 	/* We'll care only link-local addresses */
   2076 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
   2077 		splx(s);
   2078 		return;
   2079 	}
   2080 
   2081 	rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
   2082 	splx(s);
   2083 }
   2084 
   2085 static int
   2086 rt6_deleteroute(rn, arg)
   2087 	struct radix_node *rn;
   2088 	void *arg;
   2089 {
   2090 #define SIN6(s)	((struct sockaddr_in6 *)s)
   2091 	struct rtentry *rt = (struct rtentry *)rn;
   2092 	struct in6_addr *gate = (struct in6_addr *)arg;
   2093 
   2094 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
   2095 		return (0);
   2096 
   2097 	if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr))
   2098 		return (0);
   2099 
   2100 	/*
   2101 	 * Do not delete a static route.
   2102 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
   2103 	 * 'cloned' bit instead?
   2104 	 */
   2105 	if ((rt->rt_flags & RTF_STATIC) != 0)
   2106 		return (0);
   2107 
   2108 	/*
   2109 	 * We delete only host route. This means, in particular, we don't
   2110 	 * delete default route.
   2111 	 */
   2112 	if ((rt->rt_flags & RTF_HOST) == 0)
   2113 		return (0);
   2114 
   2115 	return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
   2116 	    rt_mask(rt), rt->rt_flags, 0));
   2117 #undef SIN6
   2118 }
   2119 
   2120 int
   2121 nd6_setdefaultiface(ifindex)
   2122 	int ifindex;
   2123 {
   2124 	int error = 0;
   2125 
   2126 	if (ifindex < 0 || if_indexlim <= ifindex)
   2127 		return (EINVAL);
   2128 	if (ifindex != 0 && !ifindex2ifnet[ifindex])
   2129 		return (EINVAL);
   2130 
   2131 	if (nd6_defifindex != ifindex) {
   2132 		nd6_defifindex = ifindex;
   2133 		if (nd6_defifindex > 0) {
   2134 			nd6_defifp = ifindex2ifnet[nd6_defifindex];
   2135 		} else
   2136 			nd6_defifp = NULL;
   2137 
   2138 		/*
   2139 		 * Our current implementation assumes one-to-one maping between
   2140 		 * interfaces and links, so it would be natural to use the
   2141 		 * default interface as the default link.
   2142 		 */
   2143 		scope6_setdefault(nd6_defifp);
   2144 	}
   2145 
   2146 	return (error);
   2147 }
   2148