Home | History | Annotate | Line # | Download | only in netinet6
nd6_rtr.c revision 1.117
      1 /*	$NetBSD: nd6_rtr.c,v 1.117 2016/07/20 07:37:51 ozaki-r 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 1.117 2016/07/20 07:37:51 ozaki-r 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 #include <sys/cprng.h>
     48 
     49 #include <net/if.h>
     50 #include <net/if_types.h>
     51 #include <net/if_dl.h>
     52 
     53 #include <netinet/in.h>
     54 #include <netinet6/in6_var.h>
     55 #include <netinet6/in6_ifattach.h>
     56 #include <netinet/ip6.h>
     57 #include <netinet6/ip6_var.h>
     58 #include <netinet6/nd6.h>
     59 #include <netinet/icmp6.h>
     60 #include <netinet6/icmp6_private.h>
     61 #include <netinet6/scope6_var.h>
     62 
     63 #include <net/net_osdep.h>
     64 
     65 static int rtpref(struct nd_defrouter *);
     66 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
     67 static int prelist_update(struct nd_prefixctl *, struct nd_defrouter *,
     68     struct mbuf *, int);
     69 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *, int);
     70 static struct nd_pfxrouter *pfxrtr_lookup(struct nd_prefix *,
     71 	struct nd_defrouter *);
     72 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
     73 static void pfxrtr_del(struct nd_pfxrouter *);
     74 static struct nd_pfxrouter *find_pfxlist_reachable_router
     75 	(struct nd_prefix *);
     76 static void defrouter_delreq(struct nd_defrouter *);
     77 
     78 static int in6_init_prefix_ltimes(struct nd_prefix *);
     79 static void in6_init_address_ltimes(struct nd_prefix *,
     80 	struct in6_addrlifetime *);
     81 static void purge_detached(struct ifnet *);
     82 
     83 static int rt6_deleteroute(struct rtentry *, void *);
     84 
     85 extern int nd6_recalc_reachtm_interval;
     86 
     87 static struct ifnet *nd6_defifp;
     88 int nd6_defifindex;
     89 
     90 int ip6_use_tempaddr = 0;
     91 
     92 int ip6_desync_factor;
     93 u_int32_t ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME;
     94 u_int32_t ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME;
     95 int ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE;
     96 
     97 int nd6_numroutes = 0;
     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 static inline bool
    107 nd6_is_llinfo_probreach(struct nd_defrouter *dr)
    108 {
    109 	struct llentry *ln = NULL;
    110 
    111 	ln = nd6_lookup(&dr->rtaddr, dr->ifp, false);
    112 	if (ln == NULL)
    113 		return false;
    114 
    115 	if (!ND6_IS_LLINFO_PROBREACH(ln)) {
    116 		LLE_RUNLOCK(ln);
    117 		return false;
    118 	}
    119 
    120 	LLE_RUNLOCK(ln);
    121 	return true;
    122 }
    123 
    124 /*
    125  * Receive Router Solicitation Message - just for routers.
    126  * Router solicitation/advertisement is mostly managed by a userland program
    127  * (rtadvd) so here we have no function like nd6_ra_output().
    128  *
    129  * Based on RFC 2461
    130  */
    131 void
    132 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
    133 {
    134 	struct ifnet *ifp;
    135 	struct nd_ifinfo *ndi;
    136 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
    137 	struct nd_router_solicit *nd_rs;
    138 	struct in6_addr saddr6 = ip6->ip6_src;
    139 	char *lladdr = NULL;
    140 	int lladdrlen = 0;
    141 	union nd_opts ndopts;
    142 	struct psref psref;
    143 
    144 	ifp = m_get_rcvif_psref(m, &psref);
    145 	if (ifp == NULL)
    146 		goto freeit;
    147 
    148 	ndi = ND_IFINFO(ifp);
    149 
    150 	/* If I'm not a router, ignore it. */
    151 	if (nd6_accepts_rtadv(ndi) || !ip6_forwarding)
    152 		goto freeit;
    153 
    154 	/* Sanity checks */
    155 	if (ip6->ip6_hlim != 255) {
    156 		nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
    157 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
    158 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
    159 		goto bad;
    160 	}
    161 
    162 	/*
    163 	 * Don't update the neighbor cache, if src = ::.
    164 	 * This indicates that the src has no IP address assigned yet.
    165 	 */
    166 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
    167 		goto freeit;
    168 
    169 	IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
    170 	if (nd_rs == NULL) {
    171 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
    172 		return;
    173 	}
    174 
    175 	icmp6len -= sizeof(*nd_rs);
    176 	nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
    177 	if (nd6_options(&ndopts) < 0) {
    178 		nd6log(LOG_INFO, "invalid ND option, ignored\n");
    179 		/* nd6_options have incremented stats */
    180 		goto freeit;
    181 	}
    182 
    183 	if (ndopts.nd_opts_src_lladdr) {
    184 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
    185 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
    186 	}
    187 
    188 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
    189 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
    190 		    "(if %d, RS packet %d)\n",
    191 		    ip6_sprintf(&saddr6), ifp->if_addrlen, lladdrlen - 2);
    192 		goto bad;
    193 	}
    194 
    195 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
    196 
    197  freeit:
    198 	m_put_rcvif_psref(ifp, &psref);
    199 	m_freem(m);
    200 	return;
    201 
    202  bad:
    203 	ICMP6_STATINC(ICMP6_STAT_BADRS);
    204 	m_put_rcvif_psref(ifp, &psref);
    205 	m_freem(m);
    206 }
    207 
    208 /*
    209  * Receive Router Advertisement Message.
    210  *
    211  * Based on RFC 2461
    212  * TODO: on-link bit on prefix information
    213  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
    214  */
    215 void
    216 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
    217 {
    218 	struct ifnet *ifp;
    219 	struct nd_ifinfo *ndi;
    220 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
    221 	struct nd_router_advert *nd_ra;
    222 	struct in6_addr saddr6 = ip6->ip6_src;
    223 #if 0
    224 	struct in6_addr daddr6 = ip6->ip6_dst;
    225 	int flags; /* = nd_ra->nd_ra_flags_reserved; */
    226 	int is_managed = ((flags & ND_RA_FLAG_MANAGED) != 0);
    227 	int is_other = ((flags & ND_RA_FLAG_OTHER) != 0);
    228 #endif
    229 	int mcast = 0;
    230 	union nd_opts ndopts;
    231 	struct nd_defrouter *dr;
    232 	struct psref psref;
    233 
    234 	ifp = m_get_rcvif_psref(m, &psref);
    235 	if (ifp == NULL)
    236 		goto freeit;
    237 
    238 	ndi = ND_IFINFO(ifp);
    239 	/*
    240 	 * We only accept RAs when
    241 	 * the system-wide variable allows the acceptance, and the
    242 	 * per-interface variable allows RAs on the receiving interface.
    243 	 */
    244 	if (!nd6_accepts_rtadv(ndi))
    245 		goto freeit;
    246 
    247 	if (ip6->ip6_hlim != 255) {
    248 		nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
    249 		    ip6->ip6_hlim, ip6_sprintf(&ip6->ip6_src),
    250 		    ip6_sprintf(&ip6->ip6_dst), if_name(ifp));
    251 		goto bad;
    252 	}
    253 
    254 	if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
    255 		nd6log(LOG_ERR, "src %s is not link-local\n",
    256 		    ip6_sprintf(&saddr6));
    257 		goto bad;
    258 	}
    259 
    260 	IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
    261 	if (nd_ra == NULL) {
    262 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
    263 		m_put_rcvif_psref(ifp, &psref);
    264 		return;
    265 	}
    266 
    267 	icmp6len -= sizeof(*nd_ra);
    268 	nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
    269 	if (nd6_options(&ndopts) < 0) {
    270 		nd6log(LOG_INFO, "invalid ND option, ignored\n");
    271 		/* nd6_options have incremented stats */
    272 		goto freeit;
    273 	}
    274 
    275     {
    276 	struct nd_defrouter drtr;
    277 	u_int32_t advreachable = nd_ra->nd_ra_reachable;
    278 
    279 	/* remember if this is a multicasted advertisement */
    280 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
    281 		mcast = 1;
    282 
    283 	memset(&drtr, 0, sizeof(drtr));
    284 	drtr.rtaddr = saddr6;
    285 	drtr.flags  = nd_ra->nd_ra_flags_reserved;
    286 	drtr.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
    287 	drtr.expire = time_uptime + drtr.rtlifetime;
    288 	drtr.ifp = ifp;
    289 	/* unspecified or not? (RFC 2461 6.3.4) */
    290 	if (advreachable) {
    291 		NTOHL(advreachable);
    292 		if (advreachable <= MAX_REACHABLE_TIME &&
    293 		    ndi->basereachable != advreachable) {
    294 			ndi->basereachable = advreachable;
    295 			ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
    296 			ndi->recalctm = nd6_recalc_reachtm_interval; /* reset */
    297 		}
    298 	}
    299 	if (nd_ra->nd_ra_retransmit)
    300 		ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
    301 	if (nd_ra->nd_ra_curhoplimit) {
    302 		if (ndi->chlim < nd_ra->nd_ra_curhoplimit)
    303 			ndi->chlim = nd_ra->nd_ra_curhoplimit;
    304 		else if (ndi->chlim != nd_ra->nd_ra_curhoplimit)
    305 			log(LOG_ERR, "nd_ra_input: lower CurHopLimit sent from "
    306 			   "%s on %s (current=%d, received=%d), ignored\n",
    307 			   ip6_sprintf(&ip6->ip6_src),
    308 			   if_name(ifp), ndi->chlim, nd_ra->nd_ra_curhoplimit);
    309 	}
    310 	dr = defrtrlist_update(&drtr);
    311     }
    312 
    313 	/*
    314 	 * prefix
    315 	 */
    316 	if (ndopts.nd_opts_pi) {
    317 		struct nd_opt_hdr *pt;
    318 		struct nd_opt_prefix_info *pi = NULL;
    319 		struct nd_prefixctl prc;
    320 
    321 		for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
    322 		     pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
    323 		     pt = (struct nd_opt_hdr *)((char *)pt +
    324 						(pt->nd_opt_len << 3))) {
    325 			if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
    326 				continue;
    327 			pi = (struct nd_opt_prefix_info *)pt;
    328 
    329 			if (pi->nd_opt_pi_len != 4) {
    330 				nd6log(LOG_INFO, "invalid option "
    331 				    "len %d for prefix information option, "
    332 				    "ignored\n", pi->nd_opt_pi_len);
    333 				continue;
    334 			}
    335 
    336 			if (128 < pi->nd_opt_pi_prefix_len) {
    337 				nd6log(LOG_INFO, "invalid prefix "
    338 				    "len %d for prefix information option, "
    339 				    "ignored\n", pi->nd_opt_pi_prefix_len);
    340 				continue;
    341 			}
    342 
    343 			if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
    344 			 || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
    345 				nd6log(LOG_INFO,
    346 				    "invalid prefix %s, ignored\n",
    347 				    ip6_sprintf(&pi->nd_opt_pi_prefix));
    348 				continue;
    349 			}
    350 
    351 			memset(&prc, 0, sizeof(prc));
    352 			sockaddr_in6_init(&prc.ndprc_prefix,
    353 			    &pi->nd_opt_pi_prefix, 0, 0, 0);
    354 			prc.ndprc_ifp = ifp;
    355 
    356 			prc.ndprc_raf_onlink = (pi->nd_opt_pi_flags_reserved &
    357 			    ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
    358 			prc.ndprc_raf_auto = (pi->nd_opt_pi_flags_reserved &
    359 			    ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
    360 			prc.ndprc_plen = pi->nd_opt_pi_prefix_len;
    361 			prc.ndprc_vltime = ntohl(pi->nd_opt_pi_valid_time);
    362 			prc.ndprc_pltime = ntohl(pi->nd_opt_pi_preferred_time);
    363 
    364 			(void)prelist_update(&prc, dr, m, mcast);
    365 		}
    366 	}
    367 
    368 	/*
    369 	 * MTU
    370 	 */
    371 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
    372 		u_long mtu;
    373 		u_long maxmtu;
    374 
    375 		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
    376 
    377 		/* lower bound */
    378 		if (mtu < IPV6_MMTU) {
    379 			nd6log(LOG_INFO, "bogus mtu option "
    380 			    "mtu=%lu sent from %s, ignoring\n",
    381 			    mtu, ip6_sprintf(&ip6->ip6_src));
    382 			goto skip;
    383 		}
    384 
    385 		/* upper bound */
    386 		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
    387 		    ? ndi->maxmtu : ifp->if_mtu;
    388 		if (mtu <= maxmtu) {
    389 			int change = (ndi->linkmtu != mtu);
    390 
    391 			ndi->linkmtu = mtu;
    392 			if (change) /* in6_maxmtu may change */
    393 				in6_setmaxmtu();
    394 		} else {
    395 			nd6log(LOG_INFO,
    396 			    "bogus mtu mtu=%lu sent from %s; "
    397 			    "exceeds maxmtu %lu, ignoring\n",
    398 			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu);
    399 		}
    400 	}
    401 
    402  skip:
    403 
    404 	/*
    405 	 * Source link layer address
    406 	 */
    407     {
    408 	char *lladdr = NULL;
    409 	int lladdrlen = 0;
    410 
    411 	if (ndopts.nd_opts_src_lladdr) {
    412 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
    413 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
    414 	}
    415 
    416 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
    417 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
    418 		    "(if %d, RA packet %d)\n", ip6_sprintf(&saddr6),
    419 		    ifp->if_addrlen, lladdrlen - 2);
    420 		goto bad;
    421 	}
    422 
    423 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_ADVERT, 0);
    424 
    425 	/*
    426 	 * Installing a link-layer address might change the state of the
    427 	 * router's neighbor cache, which might also affect our on-link
    428 	 * detection of adveritsed prefixes.
    429 	 */
    430 	pfxlist_onlink_check();
    431     }
    432 
    433  freeit:
    434 	m_put_rcvif_psref(ifp, &psref);
    435 	m_freem(m);
    436 	return;
    437 
    438  bad:
    439 	ICMP6_STATINC(ICMP6_STAT_BADRA);
    440 	m_put_rcvif_psref(ifp, &psref);
    441 	m_freem(m);
    442 }
    443 
    444 /*
    445  * default router list processing sub routines
    446  */
    447 void
    448 defrouter_addreq(struct nd_defrouter *newdr)
    449 {
    450 	union {
    451 		struct sockaddr_in6 sin6;
    452 		struct sockaddr sa;
    453 	} def, mask, gate;
    454 	int s;
    455 	int error;
    456 
    457 	memset(&def, 0, sizeof(def));
    458 	memset(&mask, 0, sizeof(mask));
    459 	memset(&gate, 0,sizeof(gate)); /* for safety */
    460 
    461 	def.sin6.sin6_len = mask.sin6.sin6_len = gate.sin6.sin6_len =
    462 	    sizeof(struct sockaddr_in6);
    463 	def.sin6.sin6_family = mask.sin6.sin6_family = gate.sin6.sin6_family = AF_INET6;
    464 	gate.sin6.sin6_addr = newdr->rtaddr;
    465 #ifndef SCOPEDROUTING
    466 	gate.sin6.sin6_scope_id = 0;	/* XXX */
    467 #endif
    468 
    469 	s = splsoftnet();
    470 	error = rtrequest_newmsg(RTM_ADD, &def.sa, &gate.sa, &mask.sa,
    471 	    RTF_GATEWAY);
    472 	if (error == 0) {
    473 		nd6_numroutes++;
    474 		newdr->installed = 1;
    475 	}
    476 	splx(s);
    477 	return;
    478 }
    479 
    480 struct nd_defrouter *
    481 defrouter_lookup(const struct in6_addr *addr, struct ifnet *ifp)
    482 {
    483 	struct nd_defrouter *dr;
    484 
    485 	TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
    486 		if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
    487 			break;
    488 	}
    489 
    490 	return dr;		/* search failed */
    491 }
    492 
    493 void
    494 defrtrlist_del(struct nd_defrouter *dr, struct in6_ifextra *ext)
    495 {
    496 	struct nd_defrouter *deldr = NULL;
    497 	struct nd_prefix *pr;
    498 	struct nd_ifinfo *ndi;
    499 
    500 	if (ext == NULL)
    501 		ext = dr->ifp->if_afdata[AF_INET6];
    502 
    503 	/* detach already in progress, can not do anything */
    504 	if (ext == NULL)
    505 		return;
    506 
    507 	ndi = ext->nd_ifinfo;
    508 
    509 	/*
    510 	 * Flush all the routing table entries that use the router
    511 	 * as a next hop.
    512 	 */
    513 	/* XXX: better condition? */
    514 	if (!ip6_forwarding && nd6_accepts_rtadv(ndi))
    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 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
    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 	ext->ndefrouters--;
    542 	if (ext->ndefrouters < 0) {
    543 		log(LOG_WARNING, "defrtrlist_del: negative count on %s\n",
    544 		    dr->ifp->if_xname);
    545 	}
    546 
    547 	free(dr, M_IP6NDP);
    548 }
    549 
    550 /*
    551  * Remove the default route for a given router.
    552  * This is just a subroutine function for defrouter_select(), and should
    553  * not be called from anywhere else.
    554  */
    555 static void
    556 defrouter_delreq(struct nd_defrouter *dr)
    557 {
    558 	union {
    559 		struct sockaddr_in6 sin6;
    560 		struct sockaddr sa;
    561 	} def, mask, gw;
    562 	int error;
    563 
    564 #ifdef DIAGNOSTIC
    565 	if (dr == NULL)
    566 		panic("dr == NULL in defrouter_delreq");
    567 #endif
    568 
    569 	memset(&def, 0, sizeof(def));
    570 	memset(&mask, 0, sizeof(mask));
    571 	memset(&gw, 0, sizeof(gw));	/* for safety */
    572 
    573 	def.sin6.sin6_len = mask.sin6.sin6_len = gw.sin6.sin6_len =
    574 	    sizeof(struct sockaddr_in6);
    575 	def.sin6.sin6_family = mask.sin6.sin6_family = gw.sin6.sin6_family = AF_INET6;
    576 	gw.sin6.sin6_addr = dr->rtaddr;
    577 #ifndef SCOPEDROUTING
    578 	gw.sin6.sin6_scope_id = 0;	/* XXX */
    579 #endif
    580 
    581 	error = rtrequest_newmsg(RTM_DELETE, &def.sa, &gw.sa, &mask.sa,
    582 	    RTF_GATEWAY);
    583 	if (error == 0)
    584 		nd6_numroutes--;
    585 
    586 	dr->installed = 0;
    587 }
    588 
    589 /*
    590  * remove all default routes from default router list
    591  */
    592 void
    593 defrouter_reset(void)
    594 {
    595 	struct nd_defrouter *dr;
    596 
    597 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    598 	     dr = TAILQ_NEXT(dr, dr_entry))
    599 		defrouter_delreq(dr);
    600 
    601 	/*
    602 	 * XXX should we also nuke any default routers in the kernel, by
    603 	 * going through them by rtalloc1()?
    604 	 */
    605 }
    606 
    607 /*
    608  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
    609  * draft-ietf-ipngwg-router-selection:
    610  * 1) Routers that are reachable or probably reachable should be preferred.
    611  *    If we have more than one (probably) reachable router, prefer ones
    612  *    with the highest router preference.
    613  * 2) When no routers on the list are known to be reachable or
    614  *    probably reachable, routers SHOULD be selected in a round-robin
    615  *    fashion, regardless of router preference values.
    616  * 3) If the Default Router List is empty, assume that all
    617  *    destinations are on-link.
    618  *
    619  * We assume nd_defrouter is sorted by router preference value.
    620  * Since the code below covers both with and without router preference cases,
    621  * we do not need to classify the cases by ifdef.
    622  *
    623  * At this moment, we do not try to install more than one default router,
    624  * even when the multipath routing is available, because we're not sure about
    625  * the benefits for stub hosts comparing to the risk of making the code
    626  * complicated and the possibility of introducing bugs.
    627  */
    628 void
    629 defrouter_select(void)
    630 {
    631 	struct nd_ifinfo *ndi;
    632 	int s = splsoftnet();
    633 	struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
    634 
    635 	/*
    636 	 * This function should be called only when acting as an autoconfigured
    637 	 * host.  Although the remaining part of this function is not effective
    638 	 * if the node is not an autoconfigured host, we explicitly exclude
    639 	 * such cases here for safety.
    640 	 */
    641 	if (ip6_forwarding) {
    642 		nd6log(LOG_WARNING, "called unexpectedly (forwarding=%d, "
    643 		    "accept_rtadv=%d)\n", ip6_forwarding, ip6_accept_rtadv);
    644 		splx(s);
    645 		return;
    646 	}
    647 
    648 	/*
    649 	 * Let's handle easy case (3) first:
    650 	 * If default router list is empty, there's nothing to be done.
    651 	 */
    652 	if (!TAILQ_FIRST(&nd_defrouter)) {
    653 		splx(s);
    654 		return;
    655 	}
    656 
    657 	/*
    658 	 * Search for a (probably) reachable router from the list.
    659 	 * We just pick up the first reachable one (if any), assuming that
    660 	 * the ordering rule of the list described in defrtrlist_update().
    661 	 */
    662 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    663 	     dr = TAILQ_NEXT(dr, dr_entry)) {
    664 		ndi = ND_IFINFO(dr->ifp);
    665 		if (nd6_accepts_rtadv(ndi))
    666 			continue;
    667 
    668 		if (selected_dr == NULL &&
    669 		    nd6_is_llinfo_probreach(dr))
    670 			selected_dr = dr;
    671 
    672 		if (dr->installed && !installed_dr)
    673 			installed_dr = dr;
    674 		else if (dr->installed && installed_dr) {
    675 			/* this should not happen.  warn for diagnosis. */
    676 			log(LOG_ERR, "defrouter_select: more than one router"
    677 			    " is installed\n");
    678 		}
    679 	}
    680 	/*
    681 	 * If none of the default routers was found to be reachable,
    682 	 * round-robin the list regardless of preference.
    683 	 * Otherwise, if we have an installed router, check if the selected
    684 	 * (reachable) router should really be preferred to the installed one.
    685 	 * We only prefer the new router when the old one is not reachable
    686 	 * or when the new one has a really higher preference value.
    687 	 */
    688 	if (selected_dr == NULL) {
    689 		if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
    690 			selected_dr = TAILQ_FIRST(&nd_defrouter);
    691 		else
    692 			selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
    693 	} else if (installed_dr &&
    694 	    nd6_is_llinfo_probreach(installed_dr) &&
    695 	    rtpref(selected_dr) <= rtpref(installed_dr)) {
    696 		selected_dr = installed_dr;
    697 	}
    698 
    699 	/*
    700 	 * If the selected router is different than the installed one,
    701 	 * remove the installed router and install the selected one.
    702 	 * Note that the selected router is never NULL here.
    703 	 */
    704 	if (installed_dr != selected_dr) {
    705 		if (installed_dr)
    706 			defrouter_delreq(installed_dr);
    707 		defrouter_addreq(selected_dr);
    708 	}
    709 
    710 	splx(s);
    711 	return;
    712 }
    713 
    714 /*
    715  * for default router selection
    716  * regards router-preference field as a 2-bit signed integer
    717  */
    718 static int
    719 rtpref(struct nd_defrouter *dr)
    720 {
    721 	switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
    722 	case ND_RA_FLAG_RTPREF_HIGH:
    723 		return (RTPREF_HIGH);
    724 	case ND_RA_FLAG_RTPREF_MEDIUM:
    725 	case ND_RA_FLAG_RTPREF_RSV:
    726 		return (RTPREF_MEDIUM);
    727 	case ND_RA_FLAG_RTPREF_LOW:
    728 		return (RTPREF_LOW);
    729 	default:
    730 		/*
    731 		 * This case should never happen.  If it did, it would mean a
    732 		 * serious bug of kernel internal.  We thus always bark here.
    733 		 * Or, can we even panic?
    734 		 */
    735 		log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
    736 		return (RTPREF_INVALID);
    737 	}
    738 	/* NOTREACHED */
    739 }
    740 
    741 static struct nd_defrouter *
    742 defrtrlist_update(struct nd_defrouter *newdr)
    743 {
    744 	struct nd_defrouter *dr, *n;
    745 	struct in6_ifextra *ext = newdr->ifp->if_afdata[AF_INET6];
    746 	int s = splsoftnet();
    747 
    748 	if ((dr = defrouter_lookup(&newdr->rtaddr, newdr->ifp)) != NULL) {
    749 		/* entry exists */
    750 		if (newdr->rtlifetime == 0) {
    751 			defrtrlist_del(dr, ext);
    752 			dr = NULL;
    753 		} else {
    754 			int oldpref = rtpref(dr);
    755 
    756 			/* override */
    757 			dr->flags = newdr->flags; /* xxx flag check */
    758 			dr->rtlifetime = newdr->rtlifetime;
    759 			dr->expire = newdr->expire;
    760 
    761 			/*
    762 			 * If the preference does not change, there's no need
    763 			 * to sort the entries.
    764 			 */
    765 			if (rtpref(newdr) == oldpref) {
    766 				splx(s);
    767 				return (dr);
    768 			}
    769 
    770 			/*
    771 			 * preferred router may be changed, so relocate
    772 			 * this router.
    773 			 * XXX: calling TAILQ_REMOVE directly is a bad manner.
    774 			 * However, since defrtrlist_del() has many side
    775 			 * effects, we intentionally do so here.
    776 			 * defrouter_select() below will handle routing
    777 			 * changes later.
    778 			 */
    779 			TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
    780 			n = dr;
    781 			goto insert;
    782 		}
    783 		splx(s);
    784 		return (dr);
    785 	}
    786 
    787 	if (ip6_maxifdefrouters >= 0 &&
    788 	    ext->ndefrouters >= ip6_maxifdefrouters) {
    789 		splx(s);
    790 		return (NULL);
    791 	}
    792 
    793 	/* entry does not exist */
    794 	if (newdr->rtlifetime == 0) {
    795 		splx(s);
    796 		return (NULL);
    797 	}
    798 
    799 	if (ip6_rtadv_maxroutes <= nd6_numroutes) {
    800 		ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
    801 		splx(s);
    802 		return (NULL);
    803 	}
    804 
    805 	n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
    806 	if (n == NULL) {
    807 		splx(s);
    808 		return (NULL);
    809 	}
    810 	memset(n, 0, sizeof(*n));
    811 	*n = *newdr;
    812 
    813 insert:
    814 	/*
    815 	 * Insert the new router in the Default Router List;
    816 	 * The Default Router List should be in the descending order
    817 	 * of router-preferece.  Routers with the same preference are
    818 	 * sorted in the arriving time order.
    819 	 */
    820 
    821 	/* insert at the end of the group */
    822 	for (dr = TAILQ_FIRST(&nd_defrouter); dr;
    823 	     dr = TAILQ_NEXT(dr, dr_entry)) {
    824 		if (rtpref(n) > rtpref(dr))
    825 			break;
    826 	}
    827 	if (dr)
    828 		TAILQ_INSERT_BEFORE(dr, n, dr_entry);
    829 	else
    830 		TAILQ_INSERT_TAIL(&nd_defrouter, n, dr_entry);
    831 
    832 	defrouter_select();
    833 
    834 	ext->ndefrouters++;
    835 
    836 	splx(s);
    837 
    838 	return (n);
    839 }
    840 
    841 static struct nd_pfxrouter *
    842 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
    843 {
    844 	struct nd_pfxrouter *search;
    845 
    846 	LIST_FOREACH(search, &pr->ndpr_advrtrs, pfr_entry) {
    847 		if (search->router == dr)
    848 			break;
    849 	}
    850 
    851 	return (search);
    852 }
    853 
    854 static void
    855 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
    856 {
    857 	struct nd_pfxrouter *newpfr;
    858 
    859 	newpfr = malloc(sizeof(*newpfr), M_IP6NDP, M_NOWAIT|M_ZERO);
    860 	if (newpfr == NULL)
    861 		return;
    862 	newpfr->router = dr;
    863 
    864 	LIST_INSERT_HEAD(&pr->ndpr_advrtrs, newpfr, pfr_entry);
    865 
    866 	pfxlist_onlink_check();
    867 }
    868 
    869 static void
    870 pfxrtr_del(struct nd_pfxrouter *pfr)
    871 {
    872 	LIST_REMOVE(pfr, pfr_entry);
    873 	free(pfr, M_IP6NDP);
    874 }
    875 
    876 struct nd_prefix *
    877 nd6_prefix_lookup(struct nd_prefixctl *key)
    878 {
    879 	struct nd_prefix *search;
    880 
    881 	LIST_FOREACH(search, &nd_prefix, ndpr_entry) {
    882 		if (key->ndprc_ifp == search->ndpr_ifp &&
    883 		    key->ndprc_plen == search->ndpr_plen &&
    884 		    in6_are_prefix_equal(&key->ndprc_prefix.sin6_addr,
    885 		    &search->ndpr_prefix.sin6_addr, key->ndprc_plen)) {
    886 			break;
    887 		}
    888 	}
    889 
    890 	return (search);
    891 }
    892 
    893 static void
    894 purge_detached(struct ifnet *ifp)
    895 {
    896 	struct nd_prefix *pr, *pr_next;
    897 	struct in6_ifaddr *ia;
    898 	struct ifaddr *ifa, *ifa_next;
    899 
    900 	for (pr = nd_prefix.lh_first; pr; pr = pr_next) {
    901 		pr_next = pr->ndpr_next;
    902 
    903 		/*
    904 		 * This function is called when we need to make more room for
    905 		 * new prefixes rather than keeping old, possibly stale ones.
    906 		 * Detached prefixes would be a good candidate; if all routers
    907 		 * that advertised the prefix expired, the prefix is also
    908 		 * probably stale.
    909 		 */
    910 		if (pr->ndpr_ifp != ifp ||
    911 		    IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr) ||
    912 		    ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
    913 		    !LIST_EMPTY(&pr->ndpr_advrtrs)))
    914 			continue;
    915 
    916 		for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = ifa_next) {
    917 			ifa_next = IFADDR_READER_NEXT(ifa);
    918 			if (ifa->ifa_addr->sa_family != AF_INET6)
    919 				continue;
    920 			ia = (struct in6_ifaddr *)ifa;
    921 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) ==
    922 			    IN6_IFF_AUTOCONF && ia->ia6_ndpr == pr) {
    923 				in6_purgeaddr(ifa);
    924 			}
    925 		}
    926 		if (pr->ndpr_refcnt == 0)
    927 			prelist_remove(pr);
    928 	}
    929 }
    930 int
    931 nd6_prelist_add(struct nd_prefixctl *prc, struct nd_defrouter *dr,
    932 	struct nd_prefix **newp)
    933 {
    934 	struct nd_prefix *newpr = NULL;
    935 	int i, s;
    936 	int error;
    937 	struct in6_ifextra *ext = prc->ndprc_ifp->if_afdata[AF_INET6];
    938 
    939 	if (ip6_maxifprefixes >= 0) {
    940 		if (ext->nprefixes >= ip6_maxifprefixes / 2)
    941 			purge_detached(prc->ndprc_ifp);
    942 		if (ext->nprefixes >= ip6_maxifprefixes)
    943 			return ENOMEM;
    944 	}
    945 
    946 	error = 0;
    947 	newpr = malloc(sizeof(*newpr), M_IP6NDP, M_NOWAIT|M_ZERO);
    948 	if (newpr == NULL)
    949 		return ENOMEM;
    950 	newpr->ndpr_ifp = prc->ndprc_ifp;
    951 	newpr->ndpr_prefix = prc->ndprc_prefix;
    952 	newpr->ndpr_plen = prc->ndprc_plen;
    953 	newpr->ndpr_vltime = prc->ndprc_vltime;
    954 	newpr->ndpr_pltime = prc->ndprc_pltime;
    955 	newpr->ndpr_flags = prc->ndprc_flags;
    956 	if ((error = in6_init_prefix_ltimes(newpr)) != 0) {
    957 		free(newpr, M_IP6NDP);
    958 		return(error);
    959 	}
    960 	newpr->ndpr_lastupdate = time_uptime;
    961 	if (newp != NULL)
    962 		*newp = newpr;
    963 
    964 	/* initialization */
    965 	LIST_INIT(&newpr->ndpr_advrtrs);
    966 	in6_prefixlen2mask(&newpr->ndpr_mask, newpr->ndpr_plen);
    967 	/* make prefix in the canonical form */
    968 	for (i = 0; i < 4; i++) {
    969 		newpr->ndpr_prefix.sin6_addr.s6_addr32[i] &=
    970 		    newpr->ndpr_mask.s6_addr32[i];
    971 	}
    972 
    973 	s = splsoftnet();
    974 	/* link ndpr_entry to nd_prefix list */
    975 	LIST_INSERT_HEAD(&nd_prefix, newpr, ndpr_entry);
    976 	splx(s);
    977 
    978 	/* ND_OPT_PI_FLAG_ONLINK processing */
    979 	if (newpr->ndpr_raf_onlink) {
    980 		int e;
    981 
    982 		if ((e = nd6_prefix_onlink(newpr)) != 0) {
    983 			nd6log(LOG_ERR, "failed to make "
    984 			    "the prefix %s/%d on-link on %s (errno=%d)\n",
    985 			    ip6_sprintf(&prc->ndprc_prefix.sin6_addr),
    986 			    prc->ndprc_plen, if_name(prc->ndprc_ifp), e);
    987 			/* proceed anyway. XXX: is it correct? */
    988 		}
    989 	}
    990 
    991 	if (dr)
    992 		pfxrtr_add(newpr, dr);
    993 
    994 	ext->nprefixes++;
    995 
    996 	return 0;
    997 }
    998 
    999 void
   1000 prelist_remove(struct nd_prefix *pr)
   1001 {
   1002 	struct nd_pfxrouter *pfr, *next;
   1003 	int e, s;
   1004 	struct in6_ifextra *ext = pr->ndpr_ifp->if_afdata[AF_INET6];
   1005 
   1006 	/* make sure to invalidate the prefix until it is really freed. */
   1007 	pr->ndpr_vltime = 0;
   1008 	pr->ndpr_pltime = 0;
   1009 #if 0
   1010 	/*
   1011 	 * Though these flags are now meaningless, we'd rather keep the value
   1012 	 * not to confuse users when executing "ndp -p".
   1013 	 */
   1014 	pr->ndpr_raf_onlink = 0;
   1015 	pr->ndpr_raf_auto = 0;
   1016 #endif
   1017 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
   1018 	    (e = nd6_prefix_offlink(pr)) != 0) {
   1019 		nd6log(LOG_ERR,
   1020 		    "failed to make %s/%d offlink on %s, errno=%d\n",
   1021 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1022 		    pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
   1023 		/* what should we do? */
   1024 	}
   1025 
   1026 	if (pr->ndpr_refcnt > 0)
   1027 		return;		/* notice here? */
   1028 
   1029 	s = splsoftnet();
   1030 	/* unlink ndpr_entry from nd_prefix list */
   1031 	LIST_REMOVE(pr, ndpr_entry);
   1032 
   1033 	/* free list of routers that adversed the prefix */
   1034 	for (pfr = LIST_FIRST(&pr->ndpr_advrtrs); pfr != NULL; pfr = next) {
   1035 		next = LIST_NEXT(pfr, pfr_entry);
   1036 
   1037 		free(pfr, M_IP6NDP);
   1038 	}
   1039 
   1040 	if (ext) {
   1041 		ext->nprefixes--;
   1042 		if (ext->nprefixes < 0) {
   1043 			log(LOG_WARNING, "prelist_remove: negative count on "
   1044 			    "%s\n", pr->ndpr_ifp->if_xname);
   1045 		}
   1046 	}
   1047 	splx(s);
   1048 
   1049 	free(pr, M_IP6NDP);
   1050 
   1051 	pfxlist_onlink_check();
   1052 }
   1053 
   1054 static int
   1055 prelist_update(struct nd_prefixctl *newprc,
   1056 	struct nd_defrouter *dr, /* may be NULL */
   1057 	struct mbuf *m,
   1058 	int mcast)
   1059 {
   1060 	struct in6_ifaddr *ia6_match = NULL;
   1061 	struct ifaddr *ifa;
   1062 	struct ifnet *ifp = newprc->ndprc_ifp;
   1063 	struct nd_prefix *pr;
   1064 	int s = splsoftnet();
   1065 	int error = 0;
   1066 	int auth;
   1067 	struct in6_addrlifetime lt6_tmp;
   1068 
   1069 	auth = 0;
   1070 	if (m) {
   1071 		/*
   1072 		 * Authenticity for NA consists authentication for
   1073 		 * both IP header and IP datagrams, doesn't it ?
   1074 		 */
   1075 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
   1076 		auth = (m->m_flags & M_AUTHIPHDR
   1077 		     && m->m_flags & M_AUTHIPDGM) ? 1 : 0;
   1078 #endif
   1079 	}
   1080 
   1081 	if ((pr = nd6_prefix_lookup(newprc)) != NULL) {
   1082 		/*
   1083 		 * nd6_prefix_lookup() ensures that pr and newprc have the same
   1084 		 * prefix on a same interface.
   1085 		 */
   1086 
   1087 		/*
   1088 		 * Update prefix information.  Note that the on-link (L) bit
   1089 		 * and the autonomous (A) bit should NOT be changed from 1
   1090 		 * to 0.
   1091 		 */
   1092 		if (newprc->ndprc_raf_onlink == 1)
   1093 			pr->ndpr_raf_onlink = 1;
   1094 		if (newprc->ndprc_raf_auto == 1)
   1095 			pr->ndpr_raf_auto = 1;
   1096 		if (newprc->ndprc_raf_onlink) {
   1097 			pr->ndpr_vltime = newprc->ndprc_vltime;
   1098 			pr->ndpr_pltime = newprc->ndprc_pltime;
   1099 			(void)in6_init_prefix_ltimes(pr); /* XXX error case? */
   1100 			pr->ndpr_lastupdate = time_uptime;
   1101 		}
   1102 
   1103 		if (newprc->ndprc_raf_onlink &&
   1104 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
   1105 			int e;
   1106 
   1107 			if ((e = nd6_prefix_onlink(pr)) != 0) {
   1108 				nd6log(LOG_ERR,
   1109 				    "failed to make "
   1110 				    "the prefix %s/%d on-link on %s "
   1111 				    "(errno=%d)\n",
   1112 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1113 				    pr->ndpr_plen, if_name(pr->ndpr_ifp), e);
   1114 				/* proceed anyway. XXX: is it correct? */
   1115 			}
   1116 		}
   1117 
   1118 		if (dr && pfxrtr_lookup(pr, dr) == NULL)
   1119 			pfxrtr_add(pr, dr);
   1120 	} else {
   1121 		struct nd_prefix *newpr = NULL;
   1122 
   1123 		if (newprc->ndprc_vltime == 0)
   1124 			goto end;
   1125 		if (newprc->ndprc_raf_onlink == 0 && newprc->ndprc_raf_auto == 0)
   1126 			goto end;
   1127 
   1128 		if (ip6_rtadv_maxroutes <= nd6_numroutes) {
   1129 			ICMP6_STATINC(ICMP6_STAT_DROPPED_RAROUTE);
   1130 			goto end;
   1131 		}
   1132 
   1133 		error = nd6_prelist_add(newprc, dr, &newpr);
   1134 		if (error != 0 || newpr == NULL) {
   1135 			nd6log(LOG_NOTICE,
   1136 			    "nd6_prelist_add failed for %s/%d on %s "
   1137 			    "errno=%d, returnpr=%p\n",
   1138 			    ip6_sprintf(&newprc->ndprc_prefix.sin6_addr),
   1139 			    newprc->ndprc_plen, if_name(newprc->ndprc_ifp),
   1140 			    error, newpr);
   1141 			goto end; /* we should just give up in this case. */
   1142 		}
   1143 
   1144 		/*
   1145 		 * XXX: from the ND point of view, we can ignore a prefix
   1146 		 * with the on-link bit being zero.  However, we need a
   1147 		 * prefix structure for references from autoconfigured
   1148 		 * addresses.  Thus, we explicitly make sure that the prefix
   1149 		 * itself expires now.
   1150 		 */
   1151 		if (newpr->ndpr_raf_onlink == 0) {
   1152 			newpr->ndpr_vltime = 0;
   1153 			newpr->ndpr_pltime = 0;
   1154 			in6_init_prefix_ltimes(newpr);
   1155 		}
   1156 
   1157 		pr = newpr;
   1158 	}
   1159 
   1160 	/*
   1161 	 * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
   1162 	 * Note that pr must be non NULL at this point.
   1163 	 */
   1164 
   1165 	/* 5.5.3 (a). Ignore the prefix without the A bit set. */
   1166 	if (!newprc->ndprc_raf_auto)
   1167 		goto end;
   1168 
   1169 	/*
   1170 	 * 5.5.3 (b). the link-local prefix should have been ignored in
   1171 	 * nd6_ra_input.
   1172 	 */
   1173 
   1174 	/* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
   1175 	if (newprc->ndprc_pltime > newprc->ndprc_vltime) {
   1176 		error = EINVAL;	/* XXX: won't be used */
   1177 		goto end;
   1178 	}
   1179 
   1180 	/*
   1181 	 * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
   1182 	 * an address configured by stateless autoconfiguration already in the
   1183 	 * list of addresses associated with the interface, and the Valid
   1184 	 * Lifetime is not 0, form an address.  We first check if we have
   1185 	 * a matching prefix.
   1186 	 * Note: we apply a clarification in rfc2462bis-02 here.  We only
   1187 	 * consider autoconfigured addresses while RFC2462 simply said
   1188 	 * "address".
   1189 	 */
   1190 	IFADDR_READER_FOREACH(ifa, ifp) {
   1191 		struct in6_ifaddr *ia6;
   1192 		u_int32_t remaininglifetime;
   1193 
   1194 		if (ifa->ifa_addr->sa_family != AF_INET6)
   1195 			continue;
   1196 
   1197 		ia6 = (struct in6_ifaddr *)ifa;
   1198 
   1199 		/*
   1200 		 * We only consider autoconfigured addresses as per rfc2462bis.
   1201 		 */
   1202 		if (!(ia6->ia6_flags & IN6_IFF_AUTOCONF))
   1203 			continue;
   1204 
   1205 		/*
   1206 		 * Spec is not clear here, but I believe we should concentrate
   1207 		 * on unicast (i.e. not anycast) addresses.
   1208 		 * XXX: other ia6_flags? detached or duplicated?
   1209 		 */
   1210 		if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0)
   1211 			continue;
   1212 
   1213 		/*
   1214 		 * Ignore the address if it is not associated with a prefix
   1215 		 * or is associated with a prefix that is different from this
   1216 		 * one.  (pr is never NULL here)
   1217 		 */
   1218 		if (ia6->ia6_ndpr != pr)
   1219 			continue;
   1220 
   1221 		if (ia6_match == NULL) /* remember the first one */
   1222 			ia6_match = ia6;
   1223 
   1224 		/*
   1225 		 * An already autoconfigured address matched.  Now that we
   1226 		 * are sure there is at least one matched address, we can
   1227 		 * proceed to 5.5.3. (e): update the lifetimes according to the
   1228 		 * "two hours" rule and the privacy extension.
   1229 		 * We apply some clarifications in rfc2462bis:
   1230 		 * - use remaininglifetime instead of storedlifetime as a
   1231 		 *   variable name
   1232 		 * - remove the dead code in the "two-hour" rule
   1233 		 */
   1234 #define TWOHOUR		(120*60)
   1235 		lt6_tmp = ia6->ia6_lifetime;
   1236 		if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
   1237 			remaininglifetime = ND6_INFINITE_LIFETIME;
   1238 		else if (time_uptime - ia6->ia6_updatetime >
   1239 			 lt6_tmp.ia6t_vltime) {
   1240 			/*
   1241 			 * The case of "invalid" address.  We should usually
   1242 			 * not see this case.
   1243 			 */
   1244 			remaininglifetime = 0;
   1245 		} else
   1246 			remaininglifetime = lt6_tmp.ia6t_vltime -
   1247 			    (time_uptime - ia6->ia6_updatetime);
   1248 
   1249 		/* when not updating, keep the current stored lifetime. */
   1250 		lt6_tmp.ia6t_vltime = remaininglifetime;
   1251 
   1252 		if (TWOHOUR < newprc->ndprc_vltime ||
   1253 		    remaininglifetime < newprc->ndprc_vltime) {
   1254 			lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
   1255 		} else if (remaininglifetime <= TWOHOUR) {
   1256 			if (auth)
   1257 				lt6_tmp.ia6t_vltime = newprc->ndprc_vltime;
   1258 		} else {
   1259 			/*
   1260 			 * newprc->ndprc_vltime <= TWOHOUR &&
   1261 			 * TWOHOUR < remaininglifetime
   1262 			 */
   1263 			lt6_tmp.ia6t_vltime = TWOHOUR;
   1264 		}
   1265 
   1266 		/* The 2 hour rule is not imposed for preferred lifetime. */
   1267 		lt6_tmp.ia6t_pltime = newprc->ndprc_pltime;
   1268 
   1269 		in6_init_address_ltimes(pr, &lt6_tmp);
   1270 
   1271 		/*
   1272 		 * We need to treat lifetimes for temporary addresses
   1273 		 * differently, according to
   1274 		 * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
   1275 		 * we only update the lifetimes when they are in the maximum
   1276 		 * intervals.
   1277 		 */
   1278 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
   1279 			u_int32_t maxvltime, maxpltime;
   1280 
   1281 			if (ip6_temp_valid_lifetime >
   1282 			    (u_int32_t)((time_uptime - ia6->ia6_createtime) +
   1283 			    ip6_desync_factor)) {
   1284 				maxvltime = ip6_temp_valid_lifetime -
   1285 				    (time_uptime - ia6->ia6_createtime) -
   1286 				    ip6_desync_factor;
   1287 			} else
   1288 				maxvltime = 0;
   1289 			if (ip6_temp_preferred_lifetime >
   1290 			    (u_int32_t)((time_uptime - ia6->ia6_createtime) +
   1291 			    ip6_desync_factor)) {
   1292 				maxpltime = ip6_temp_preferred_lifetime -
   1293 				    (time_uptime - ia6->ia6_createtime) -
   1294 				    ip6_desync_factor;
   1295 			} else
   1296 				maxpltime = 0;
   1297 
   1298 			if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
   1299 			    lt6_tmp.ia6t_vltime > maxvltime) {
   1300 				lt6_tmp.ia6t_vltime = maxvltime;
   1301 			}
   1302 			if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
   1303 			    lt6_tmp.ia6t_pltime > maxpltime) {
   1304 				lt6_tmp.ia6t_pltime = maxpltime;
   1305 			}
   1306 		}
   1307 
   1308 		ia6->ia6_lifetime = lt6_tmp;
   1309 		ia6->ia6_updatetime = time_uptime;
   1310 	}
   1311 	if (ia6_match == NULL && newprc->ndprc_vltime) {
   1312 		int ifidlen;
   1313 		struct in6_ifaddr *ia6;
   1314 
   1315 		/*
   1316 		 * 5.5.3 (d) (continued)
   1317 		 * No address matched and the valid lifetime is non-zero.
   1318 		 * Create a new address.
   1319 		 */
   1320 
   1321 		/*
   1322 		 * Prefix Length check:
   1323 		 * If the sum of the prefix length and interface identifier
   1324 		 * length does not equal 128 bits, the Prefix Information
   1325 		 * option MUST be ignored.  The length of the interface
   1326 		 * identifier is defined in a separate link-type specific
   1327 		 * document.
   1328 		 */
   1329 		ifidlen = in6_if2idlen(ifp);
   1330 		if (ifidlen < 0) {
   1331 			/* this should not happen, so we always log it. */
   1332 			log(LOG_ERR, "%s: IFID undefined (%s)\n",
   1333 			    __func__, if_name(ifp));
   1334 			goto end;
   1335 		}
   1336 		if (ifidlen + pr->ndpr_plen != 128) {
   1337 			nd6log(LOG_INFO,
   1338 			    "invalid prefixlen %d for %s, ignored\n",
   1339 			    pr->ndpr_plen, if_name(ifp));
   1340 			goto end;
   1341 		}
   1342 
   1343 		if ((ia6 = in6_ifadd(newprc, mcast)) != NULL) {
   1344 			/*
   1345 			 * note that we should use pr (not newprc) for reference.
   1346 			 */
   1347 			pr->ndpr_refcnt++;
   1348 			ia6->ia6_ndpr = pr;
   1349 
   1350 			/*
   1351 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.3 (2).
   1352 			 * When a new public address is created as described
   1353 			 * in RFC2462, also create a new temporary address.
   1354 			 *
   1355 			 * draft-ietf-ipngwg-temp-addresses-v2-00 3.5.
   1356 			 * When an interface connects to a new link, a new
   1357 			 * randomized interface identifier should be generated
   1358 			 * immediately together with a new set of temporary
   1359 			 * addresses.  Thus, we specifiy 1 as the 2nd arg of
   1360 			 * in6_tmpifadd().
   1361 			 */
   1362 			if (ip6_use_tempaddr) {
   1363 				int e;
   1364 				if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
   1365 					nd6log(LOG_NOTICE,
   1366 					    "failed to create a temporary "
   1367 					    "address, errno=%d\n", e);
   1368 				}
   1369 			}
   1370 
   1371 			/*
   1372 			 * A newly added address might affect the status
   1373 			 * of other addresses, so we check and update it.
   1374 			 * XXX: what if address duplication happens?
   1375 			 */
   1376 			pfxlist_onlink_check();
   1377 		} else {
   1378 			/* just set an error. do not bark here. */
   1379 			error = EADDRNOTAVAIL; /* XXX: might be unused. */
   1380 		}
   1381 	}
   1382 
   1383  end:
   1384 	splx(s);
   1385 	return error;
   1386 }
   1387 
   1388 /*
   1389  * A supplement function used in the on-link detection below;
   1390  * detect if a given prefix has a (probably) reachable advertising router.
   1391  * XXX: lengthy function name...
   1392  */
   1393 static struct nd_pfxrouter *
   1394 find_pfxlist_reachable_router(struct nd_prefix *pr)
   1395 {
   1396 	struct nd_pfxrouter *pfxrtr;
   1397 
   1398 	for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr;
   1399 	     pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
   1400 		if (pfxrtr->router->ifp->if_flags & IFF_UP &&
   1401 		    pfxrtr->router->ifp->if_link_state != LINK_STATE_DOWN &&
   1402 		    nd6_is_llinfo_probreach(pfxrtr->router))
   1403 			break;	/* found */
   1404 	}
   1405 
   1406 	return (pfxrtr);
   1407 }
   1408 
   1409 /*
   1410  * Check if each prefix in the prefix list has at least one available router
   1411  * that advertised the prefix (a router is "available" if its neighbor cache
   1412  * entry is reachable or probably reachable).
   1413  * If the check fails, the prefix may be off-link, because, for example,
   1414  * we have moved from the network but the lifetime of the prefix has not
   1415  * expired yet.  So we should not use the prefix if there is another prefix
   1416  * that has an available router.
   1417  * But, if there is no prefix that has an available router, we still regards
   1418  * all the prefixes as on-link.  This is because we can't tell if all the
   1419  * routers are simply dead or if we really moved from the network and there
   1420  * is no router around us.
   1421  */
   1422 void
   1423 pfxlist_onlink_check(void)
   1424 {
   1425 	struct nd_prefix *pr;
   1426 	struct in6_ifaddr *ia;
   1427 	struct nd_defrouter *dr;
   1428 	struct nd_pfxrouter *pfxrtr = NULL;
   1429 
   1430 	/*
   1431 	 * Check if there is a prefix that has a reachable advertising
   1432 	 * router.
   1433 	 */
   1434 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
   1435 		if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
   1436 			break;
   1437 	}
   1438 	/*
   1439 	 * If we have no such prefix, check whether we still have a router
   1440 	 * that does not advertise any prefixes.
   1441 	 */
   1442 	if (pr == NULL) {
   1443 		TAILQ_FOREACH(dr, &nd_defrouter, dr_entry) {
   1444 			struct nd_prefix *pr0;
   1445 
   1446 			LIST_FOREACH(pr0, &nd_prefix, ndpr_entry) {
   1447 				if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
   1448 					break;
   1449 			}
   1450 			if (pfxrtr)
   1451 				break;
   1452 		}
   1453 	}
   1454 	if (pr != NULL || (TAILQ_FIRST(&nd_defrouter) && !pfxrtr)) {
   1455 		/*
   1456 		 * There is at least one prefix that has a reachable router,
   1457 		 * or at least a router which probably does not advertise
   1458 		 * any prefixes.  The latter would be the case when we move
   1459 		 * to a new link where we have a router that does not provide
   1460 		 * prefixes and we configure an address by hand.
   1461 		 * Detach prefixes which have no reachable advertising
   1462 		 * router, and attach other prefixes.
   1463 		 */
   1464 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
   1465 			/* XXX: a link-local prefix should never be detached */
   1466 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1467 				continue;
   1468 
   1469 			/*
   1470 			 * we aren't interested in prefixes without the L bit
   1471 			 * set.
   1472 			 */
   1473 			if (pr->ndpr_raf_onlink == 0)
   1474 				continue;
   1475 
   1476 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
   1477 			    find_pfxlist_reachable_router(pr) == NULL)
   1478 				pr->ndpr_stateflags |= NDPRF_DETACHED;
   1479 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
   1480 			    find_pfxlist_reachable_router(pr) != 0)
   1481 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
   1482 		}
   1483 	} else {
   1484 		/* there is no prefix that has a reachable router */
   1485 		LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
   1486 			if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1487 				continue;
   1488 
   1489 			if (pr->ndpr_raf_onlink == 0)
   1490 				continue;
   1491 
   1492 			if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
   1493 				pr->ndpr_stateflags &= ~NDPRF_DETACHED;
   1494 		}
   1495 	}
   1496 
   1497 	/*
   1498 	 * Remove each interface route associated with a (just) detached
   1499 	 * prefix, and reinstall the interface route for a (just) attached
   1500 	 * prefix.  Note that all attempt of reinstallation does not
   1501 	 * necessarily success, when a same prefix is shared among multiple
   1502 	 * interfaces.  Such cases will be handled in nd6_prefix_onlink,
   1503 	 * so we don't have to care about them.
   1504 	 */
   1505 	LIST_FOREACH(pr, &nd_prefix, ndpr_entry) {
   1506 		int e;
   1507 
   1508 		if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
   1509 			continue;
   1510 
   1511 		if (pr->ndpr_raf_onlink == 0)
   1512 			continue;
   1513 
   1514 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
   1515 		    (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
   1516 			if ((e = nd6_prefix_offlink(pr)) != 0) {
   1517 				nd6log(LOG_ERR,
   1518 				    "failed to make %s/%d offlink, errno=%d\n",
   1519 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1520 				    pr->ndpr_plen, e);
   1521 			}
   1522 		}
   1523 		if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
   1524 		    (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
   1525 		    pr->ndpr_raf_onlink) {
   1526 			if ((e = nd6_prefix_onlink(pr)) != 0) {
   1527 				nd6log(LOG_ERR,
   1528 				    "failed to make %s/%d onlink, errno=%d\n",
   1529 				    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1530 				    pr->ndpr_plen, e);
   1531 			}
   1532 		}
   1533 	}
   1534 
   1535 	/*
   1536 	 * Changes on the prefix status might affect address status as well.
   1537 	 * Make sure that all addresses derived from an attached prefix are
   1538 	 * attached, and that all addresses derived from a detached prefix are
   1539 	 * detached.  Note, however, that a manually configured address should
   1540 	 * always be attached.
   1541 	 * The precise detection logic is same as the one for prefixes.
   1542 	 */
   1543 	IN6_ADDRLIST_READER_FOREACH(ia) {
   1544 		if (!(ia->ia6_flags & IN6_IFF_AUTOCONF))
   1545 			continue;
   1546 
   1547 		if (ia->ia6_ndpr == NULL) {
   1548 			/*
   1549 			 * This can happen when we first configure the address
   1550 			 * (i.e. the address exists, but the prefix does not).
   1551 			 * XXX: complicated relationships...
   1552 			 */
   1553 			continue;
   1554 		}
   1555 
   1556 		if (find_pfxlist_reachable_router(ia->ia6_ndpr))
   1557 			break;
   1558 	}
   1559 
   1560 	if (ia) {
   1561 		IN6_ADDRLIST_READER_FOREACH(ia) {
   1562 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
   1563 				continue;
   1564 
   1565 			if (ia->ia6_ndpr == NULL) /* XXX: see above. */
   1566 				continue;
   1567 
   1568 			if (find_pfxlist_reachable_router(ia->ia6_ndpr)) {
   1569 				if (ia->ia6_flags & IN6_IFF_DETACHED) {
   1570 					ia->ia6_flags &= ~IN6_IFF_DETACHED;
   1571 					ia->ia6_flags |= IN6_IFF_TENTATIVE;
   1572 					nd6_dad_start((struct ifaddr *)ia,
   1573 					    0);
   1574 					/* We will notify the routing socket
   1575 					 * of the DAD result, so no need to
   1576 					 * here */
   1577 				}
   1578 			} else {
   1579 				if ((ia->ia6_flags & IN6_IFF_DETACHED) == 0) {
   1580 					ia->ia6_flags |= IN6_IFF_DETACHED;
   1581 					rt_newaddrmsg(RTM_NEWADDR,
   1582 					    (struct ifaddr *)ia, 0, NULL);
   1583 				}
   1584 			}
   1585 		}
   1586 	}
   1587 	else {
   1588 		IN6_ADDRLIST_READER_FOREACH(ia) {
   1589 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
   1590 				continue;
   1591 
   1592 			if (ia->ia6_flags & IN6_IFF_DETACHED) {
   1593 				ia->ia6_flags &= ~IN6_IFF_DETACHED;
   1594 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
   1595 				/* Do we need a delay in this case? */
   1596 				nd6_dad_start((struct ifaddr *)ia, 0);
   1597 			}
   1598 		}
   1599 	}
   1600 }
   1601 
   1602 int
   1603 nd6_prefix_onlink(struct nd_prefix *pr)
   1604 {
   1605 	struct ifaddr *ifa;
   1606 	struct ifnet *ifp = pr->ndpr_ifp;
   1607 	struct sockaddr_in6 mask6;
   1608 	struct nd_prefix *opr;
   1609 	u_long rtflags;
   1610 	int error = 0;
   1611 
   1612 	/* sanity check */
   1613 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
   1614 		nd6log(LOG_ERR, "%s/%d is already on-link\n",
   1615 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen);
   1616 		return (EEXIST);
   1617 	}
   1618 
   1619 	/*
   1620 	 * Add the interface route associated with the prefix.  Before
   1621 	 * installing the route, check if there's the same prefix on another
   1622 	 * interface, and the prefix has already installed the interface route.
   1623 	 * Although such a configuration is expected to be rare, we explicitly
   1624 	 * allow it.
   1625 	 */
   1626 	LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
   1627 		if (opr == pr)
   1628 			continue;
   1629 
   1630 		if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
   1631 			continue;
   1632 
   1633 		if (opr->ndpr_plen == pr->ndpr_plen &&
   1634 		    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
   1635 		    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
   1636 			return (0);
   1637 	}
   1638 
   1639 	/*
   1640 	 * We prefer link-local addresses as the associated interface address.
   1641 	 */
   1642 	/* search for a link-local addr */
   1643 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
   1644 	    IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
   1645 	if (ifa == NULL) {
   1646 		/* XXX: freebsd does not have ifa_ifwithaf */
   1647 		IFADDR_READER_FOREACH(ifa, ifp) {
   1648 			if (ifa->ifa_addr->sa_family == AF_INET6)
   1649 				break;
   1650 		}
   1651 		/* should we care about ia6_flags? */
   1652 	}
   1653 	if (ifa == NULL) {
   1654 		/*
   1655 		 * This can still happen, when, for example, we receive an RA
   1656 		 * containing a prefix with the L bit set and the A bit clear,
   1657 		 * after removing all IPv6 addresses on the receiving
   1658 		 * interface.  This should, of course, be rare though.
   1659 		 */
   1660 		nd6log(LOG_NOTICE, "failed to find any ifaddr"
   1661 		    " to add route for a prefix(%s/%d) on %s\n",
   1662 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1663 		    pr->ndpr_plen, if_name(ifp));
   1664 		return (0);
   1665 	}
   1666 
   1667 	/*
   1668 	 * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
   1669 	 * ifa->ifa_rtrequest = nd6_rtrequest;
   1670 	 */
   1671 	memset(&mask6, 0, sizeof(mask6));
   1672 	mask6.sin6_family = AF_INET6;
   1673 	mask6.sin6_len = sizeof(mask6);
   1674 	mask6.sin6_addr = pr->ndpr_mask;
   1675 	/* rtrequest() will probably set RTF_UP, but we're not sure. */
   1676 	rtflags = ifa->ifa_flags | RTF_UP;
   1677 	if (nd6_need_cache(ifp)) {
   1678 		/* explicitly set in case ifa_flags does not set the flag. */
   1679 		rtflags |= RTF_CONNECTED;
   1680 	} else {
   1681 		/*
   1682 		 * explicitly clear the cloning bit in case ifa_flags sets it.
   1683 		 */
   1684 		rtflags &= ~RTF_CONNECTED;
   1685 	}
   1686 	error = rtrequest_newmsg(RTM_ADD, sin6tosa(&pr->ndpr_prefix),
   1687 	    ifa->ifa_addr, sin6tosa(&mask6), rtflags);
   1688 	if (error == 0) {
   1689 		nd6_numroutes++;
   1690 		pr->ndpr_stateflags |= NDPRF_ONLINK;
   1691 	} else {
   1692 		nd6log(LOG_ERR, "failed to add route for a"
   1693 		    " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
   1694 		    "errno = %d\n",
   1695 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr),
   1696 		    pr->ndpr_plen, if_name(ifp),
   1697 		    ip6_sprintf(&((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
   1698 		    ip6_sprintf(&mask6.sin6_addr), rtflags, error);
   1699 	}
   1700 
   1701 	return (error);
   1702 }
   1703 
   1704 int
   1705 nd6_prefix_offlink(struct nd_prefix *pr)
   1706 {
   1707 	int error = 0;
   1708 	struct ifnet *ifp = pr->ndpr_ifp;
   1709 	struct nd_prefix *opr;
   1710 	struct sockaddr_in6 sa6, mask6;
   1711 
   1712 	/* sanity check */
   1713 	if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
   1714 		nd6log(LOG_ERR, "%s/%d is already off-link\n",
   1715 		    ip6_sprintf(&pr->ndpr_prefix.sin6_addr), pr->ndpr_plen);
   1716 		return (EEXIST);
   1717 	}
   1718 
   1719 	sockaddr_in6_init(&sa6, &pr->ndpr_prefix.sin6_addr, 0, 0, 0);
   1720 	sockaddr_in6_init(&mask6, &pr->ndpr_mask, 0, 0, 0);
   1721 	error = rtrequest_newmsg(RTM_DELETE, sin6tosa(&sa6), NULL,
   1722 	    sin6tosa(&mask6), 0);
   1723 	if (error == 0) {
   1724 		pr->ndpr_stateflags &= ~NDPRF_ONLINK;
   1725 		nd6_numroutes--;
   1726 
   1727 		/*
   1728 		 * There might be the same prefix on another interface,
   1729 		 * the prefix which could not be on-link just because we have
   1730 		 * the interface route (see comments in nd6_prefix_onlink).
   1731 		 * If there's one, try to make the prefix on-link on the
   1732 		 * interface.
   1733 		 */
   1734 		LIST_FOREACH(opr, &nd_prefix, ndpr_entry) {
   1735 			if (opr == pr)
   1736 				continue;
   1737 
   1738 			if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
   1739 				continue;
   1740 
   1741 			/*
   1742 			 * KAME specific: detached prefixes should not be
   1743 			 * on-link.
   1744 			 */
   1745 			if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
   1746 				continue;
   1747 
   1748 			if (opr->ndpr_plen == pr->ndpr_plen &&
   1749 			    in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
   1750 			    &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
   1751 				int e;
   1752 
   1753 				if ((e = nd6_prefix_onlink(opr)) != 0) {
   1754 					nd6log(LOG_ERR, "failed to "
   1755 					    "recover a prefix %s/%d from %s "
   1756 					    "to %s (errno = %d)\n",
   1757 					    ip6_sprintf(&opr->ndpr_prefix.sin6_addr),
   1758 					    opr->ndpr_plen, if_name(ifp),
   1759 					    if_name(opr->ndpr_ifp), e);
   1760 				}
   1761 			}
   1762 		}
   1763 	} else {
   1764 		/* XXX: can we still set the NDPRF_ONLINK flag? */
   1765 		nd6log(LOG_ERR, "failed to delete route: "
   1766 		    "%s/%d on %s (errno = %d)\n",
   1767 		    ip6_sprintf(&sa6.sin6_addr), pr->ndpr_plen, if_name(ifp),
   1768 		    error);
   1769 	}
   1770 
   1771 	return error;
   1772 }
   1773 
   1774 static struct in6_ifaddr *
   1775 in6_ifadd(struct nd_prefixctl *prc, int mcast)
   1776 {
   1777 	struct ifnet *ifp = prc->ndprc_ifp;
   1778 	struct ifaddr *ifa;
   1779 	struct in6_aliasreq ifra;
   1780 	struct in6_ifaddr *ia, *ib;
   1781 	int error, plen0;
   1782 	struct in6_addr mask;
   1783 	int prefixlen = prc->ndprc_plen;
   1784 	int updateflags;
   1785 
   1786 	in6_prefixlen2mask(&mask, prefixlen);
   1787 
   1788 	/*
   1789 	 * find a link-local address (will be interface ID).
   1790 	 * Is it really mandatory? Theoretically, a global or a site-local
   1791 	 * address can be configured without a link-local address, if we
   1792 	 * have a unique interface identifier...
   1793 	 *
   1794 	 * it is not mandatory to have a link-local address, we can generate
   1795 	 * interface identifier on the fly.  we do this because:
   1796 	 * (1) it should be the easiest way to find interface identifier.
   1797 	 * (2) RFC2462 5.4 suggesting the use of the same interface identifier
   1798 	 * for multiple addresses on a single interface, and possible shortcut
   1799 	 * of DAD.  we omitted DAD for this reason in the past.
   1800 	 * (3) a user can prevent autoconfiguration of global address
   1801 	 * by removing link-local address by hand (this is partly because we
   1802 	 * don't have other way to control the use of IPv6 on an interface.
   1803 	 * this has been our design choice - cf. NRL's "ifconfig auto").
   1804 	 * (4) it is easier to manage when an interface has addresses
   1805 	 * with the same interface identifier, than to have multiple addresses
   1806 	 * with different interface identifiers.
   1807 	 */
   1808 	ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
   1809 	if (ifa)
   1810 		ib = (struct in6_ifaddr *)ifa;
   1811 	else
   1812 		return NULL;
   1813 
   1814 #if 0 /* don't care link local addr state, and always do DAD */
   1815 	/* if link-local address is not eligible, do not autoconfigure. */
   1816 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY) {
   1817 		printf("in6_ifadd: link-local address not ready\n");
   1818 		return NULL;
   1819 	}
   1820 #endif
   1821 
   1822 	/* prefixlen + ifidlen must be equal to 128 */
   1823 	plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
   1824 	if (prefixlen != plen0) {
   1825 		nd6log(LOG_INFO, "wrong prefixlen for %s "
   1826 		    "(prefix=%d ifid=%d)\n",
   1827 		    if_name(ifp), prefixlen, 128 - plen0);
   1828 		return NULL;
   1829 	}
   1830 
   1831 	/* make ifaddr */
   1832 
   1833 	memset(&ifra, 0, sizeof(ifra));
   1834 	/*
   1835 	 * in6_update_ifa() does not use ifra_name, but we accurately set it
   1836 	 * for safety.
   1837 	 */
   1838 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
   1839 	sockaddr_in6_init(&ifra.ifra_addr, &prc->ndprc_prefix.sin6_addr, 0, 0, 0);
   1840 	/* prefix */
   1841 	ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
   1842 	ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
   1843 	ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
   1844 	ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
   1845 
   1846 	/* interface ID */
   1847 	ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
   1848 	    (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
   1849 	ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
   1850 	    (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
   1851 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
   1852 	    (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
   1853 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
   1854 	    (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
   1855 
   1856 	/* new prefix mask. */
   1857 	sockaddr_in6_init(&ifra.ifra_prefixmask, &mask, 0, 0, 0);
   1858 
   1859 	/* lifetimes */
   1860 	ifra.ifra_lifetime.ia6t_vltime = prc->ndprc_vltime;
   1861 	ifra.ifra_lifetime.ia6t_pltime = prc->ndprc_pltime;
   1862 
   1863 	/* XXX: scope zone ID? */
   1864 
   1865 	ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
   1866 
   1867 	/*
   1868 	 * Make sure that we do not have this address already.  This should
   1869 	 * usually not happen, but we can still see this case, e.g., if we
   1870 	 * have manually configured the exact address to be configured.
   1871 	 */
   1872 	if (in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr) != NULL) {
   1873 		/* this should be rare enough to make an explicit log */
   1874 		log(LOG_INFO, "in6_ifadd: %s is already configured\n",
   1875 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr));
   1876 		return (NULL);
   1877 	}
   1878 
   1879 	/*
   1880 	 * Allocate ifaddr structure, link into chain, etc.
   1881 	 * If we are going to create a new address upon receiving a multicasted
   1882 	 * RA, we need to impose a random delay before starting DAD.
   1883 	 * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
   1884 	 */
   1885 	updateflags = 0;
   1886 	if (mcast)
   1887 		updateflags |= IN6_IFAUPDATE_DADDELAY;
   1888 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
   1889 		nd6log(LOG_ERR, "failed to make ifaddr %s on %s (errno=%d)\n",
   1890 		    ip6_sprintf(&ifra.ifra_addr.sin6_addr), if_name(ifp),
   1891 		    error);
   1892 		return (NULL);	/* ifaddr must not have been allocated. */
   1893 	}
   1894 
   1895 	ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
   1896 
   1897 	return (ia);		/* this is always non-NULL */
   1898 }
   1899 
   1900 int
   1901 in6_tmpifadd(
   1902 	const struct in6_ifaddr *ia0, /* corresponding public address */
   1903 	int forcegen,
   1904 	int dad_delay)
   1905 {
   1906 	struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
   1907 	struct in6_ifaddr *newia, *ia;
   1908 	struct in6_aliasreq ifra;
   1909 	int i, error;
   1910 	int trylimit = 3;	/* XXX: adhoc value */
   1911 	int updateflags;
   1912 	u_int32_t randid[2];
   1913 	u_int32_t vltime0, pltime0;
   1914 	int s;
   1915 
   1916 	memset(&ifra, 0, sizeof(ifra));
   1917 	strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
   1918 	ifra.ifra_addr = ia0->ia_addr;
   1919 	/* copy prefix mask */
   1920 	ifra.ifra_prefixmask = ia0->ia_prefixmask;
   1921 	/* clear the old IFID */
   1922 	for (i = 0; i < 4; i++) {
   1923 		ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
   1924 		    ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
   1925 	}
   1926 
   1927   again:
   1928 	if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
   1929 	    (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
   1930 		nd6log(LOG_NOTICE, "failed to find a good random IFID\n");
   1931 		return (EINVAL);
   1932 	}
   1933 	ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
   1934 	    (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
   1935 	ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
   1936 	    (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
   1937 
   1938 	/*
   1939 	 * in6_get_tmpifid() quite likely provided a unique interface ID.
   1940 	 * However, we may still have a chance to see collision, because
   1941 	 * there may be a time lag between generation of the ID and generation
   1942 	 * of the address.  So, we'll do one more sanity check.
   1943 	 */
   1944 	s = pserialize_read_enter();
   1945 	IN6_ADDRLIST_READER_FOREACH(ia) {
   1946 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
   1947 		    &ifra.ifra_addr.sin6_addr)) {
   1948 			pserialize_read_exit(s);
   1949 			if (trylimit-- == 0) {
   1950 				/*
   1951 				 * Give up.  Something strange should have
   1952 				 * happened.
   1953 				 */
   1954 				nd6log(LOG_NOTICE,
   1955 				    "failed to find a unique random IFID\n");
   1956 				return (EEXIST);
   1957 			}
   1958 			forcegen = 1;
   1959 			goto again;
   1960 		}
   1961 	}
   1962 	pserialize_read_exit(s);
   1963 
   1964 	/*
   1965 	 * The Valid Lifetime is the lower of the Valid Lifetime of the
   1966          * public address or TEMP_VALID_LIFETIME.
   1967 	 * The Preferred Lifetime is the lower of the Preferred Lifetime
   1968          * of the public address or TEMP_PREFERRED_LIFETIME -
   1969          * DESYNC_FACTOR.
   1970 	 */
   1971 	if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
   1972 		vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
   1973 		    (ia0->ia6_lifetime.ia6t_vltime -
   1974 		    (time_uptime - ia0->ia6_updatetime));
   1975 		if (vltime0 > ip6_temp_valid_lifetime)
   1976 			vltime0 = ip6_temp_valid_lifetime;
   1977 	} else
   1978 		vltime0 = ip6_temp_valid_lifetime;
   1979 	if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
   1980 		pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
   1981 		    (ia0->ia6_lifetime.ia6t_pltime -
   1982 		    (time_uptime - ia0->ia6_updatetime));
   1983 		if (pltime0 > ip6_temp_preferred_lifetime - ip6_desync_factor){
   1984 			pltime0 = ip6_temp_preferred_lifetime -
   1985 			    ip6_desync_factor;
   1986 		}
   1987 	} else
   1988 		pltime0 = ip6_temp_preferred_lifetime - ip6_desync_factor;
   1989 	ifra.ifra_lifetime.ia6t_vltime = vltime0;
   1990 	ifra.ifra_lifetime.ia6t_pltime = pltime0;
   1991 
   1992 	/*
   1993 	 * A temporary address is created only if this calculated Preferred
   1994 	 * Lifetime is greater than REGEN_ADVANCE time units.
   1995 	 */
   1996 	if (ifra.ifra_lifetime.ia6t_pltime <= ip6_temp_regen_advance)
   1997 		return (0);
   1998 
   1999 	/* XXX: scope zone ID? */
   2000 
   2001 	ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
   2002 
   2003 	/* allocate ifaddr structure, link into chain, etc. */
   2004 	updateflags = 0;
   2005 	if (dad_delay)
   2006 		updateflags |= IN6_IFAUPDATE_DADDELAY;
   2007 	if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
   2008 		return (error);
   2009 
   2010 	newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
   2011 	if (newia == NULL) {	/* XXX: can it happen? */
   2012 		nd6log(LOG_ERR,
   2013 		    "ifa update succeeded, but we got no ifaddr\n");
   2014 		return (EINVAL); /* XXX */
   2015 	}
   2016 	newia->ia6_ndpr = ia0->ia6_ndpr;
   2017 	newia->ia6_ndpr->ndpr_refcnt++;
   2018 
   2019 	/*
   2020 	 * A newly added address might affect the status of other addresses.
   2021 	 * XXX: when the temporary address is generated with a new public
   2022 	 * address, the onlink check is redundant.  However, it would be safe
   2023 	 * to do the check explicitly everywhere a new address is generated,
   2024 	 * and, in fact, we surely need the check when we create a new
   2025 	 * temporary address due to deprecation of an old temporary address.
   2026 	 */
   2027 	pfxlist_onlink_check();
   2028 
   2029 	return (0);
   2030 }
   2031 
   2032 static int
   2033 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
   2034 {
   2035 
   2036 	/* check if preferred lifetime > valid lifetime.  RFC2462 5.5.3 (c) */
   2037 	if (ndpr->ndpr_pltime > ndpr->ndpr_vltime) {
   2038 		nd6log(LOG_INFO, "preferred lifetime"
   2039 		    "(%d) is greater than valid lifetime(%d)\n",
   2040 		    (u_int)ndpr->ndpr_pltime, (u_int)ndpr->ndpr_vltime);
   2041 		return (EINVAL);
   2042 	}
   2043 	if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
   2044 		ndpr->ndpr_preferred = 0;
   2045 	else
   2046 		ndpr->ndpr_preferred = time_uptime + ndpr->ndpr_pltime;
   2047 	if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
   2048 		ndpr->ndpr_expire = 0;
   2049 	else
   2050 		ndpr->ndpr_expire = time_uptime + ndpr->ndpr_vltime;
   2051 
   2052 	return 0;
   2053 }
   2054 
   2055 static void
   2056 in6_init_address_ltimes(struct nd_prefix *newpr,
   2057     struct in6_addrlifetime *lt6)
   2058 {
   2059 
   2060 	/* Valid lifetime must not be updated unless explicitly specified. */
   2061 	/* init ia6t_expire */
   2062 	if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
   2063 		lt6->ia6t_expire = 0;
   2064 	else {
   2065 		lt6->ia6t_expire = time_uptime;
   2066 		lt6->ia6t_expire += lt6->ia6t_vltime;
   2067 	}
   2068 
   2069 	/* init ia6t_preferred */
   2070 	if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
   2071 		lt6->ia6t_preferred = 0;
   2072 	else {
   2073 		lt6->ia6t_preferred = time_uptime;
   2074 		lt6->ia6t_preferred += lt6->ia6t_pltime;
   2075 	}
   2076 }
   2077 
   2078 /*
   2079  * Delete all the routing table entries that use the specified gateway.
   2080  * XXX: this function causes search through all entries of routing table, so
   2081  * it shouldn't be called when acting as a router.
   2082  */
   2083 void
   2084 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
   2085 {
   2086 	int s = splsoftnet();
   2087 
   2088 	/* We'll care only link-local addresses */
   2089 	if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
   2090 		splx(s);
   2091 		return;
   2092 	}
   2093 
   2094 	rt_walktree(AF_INET6, rt6_deleteroute, (void *)gateway);
   2095 	splx(s);
   2096 }
   2097 
   2098 static int
   2099 rt6_deleteroute(struct rtentry *rt, void *arg)
   2100 {
   2101 	struct in6_addr *gate = (struct in6_addr *)arg;
   2102 
   2103 	if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
   2104 		return (0);
   2105 
   2106 	if (!IN6_ARE_ADDR_EQUAL(gate, &satosin6(rt->rt_gateway)->sin6_addr))
   2107 		return (0);
   2108 
   2109 	/*
   2110 	 * Do not delete a static route.
   2111 	 * XXX: this seems to be a bit ad-hoc. Should we consider the
   2112 	 * 'cloned' bit instead?
   2113 	 */
   2114 	if ((rt->rt_flags & RTF_STATIC) != 0)
   2115 		return (0);
   2116 
   2117 	/*
   2118 	 * We delete only host route. This means, in particular, we don't
   2119 	 * delete default route.
   2120 	 */
   2121 	if ((rt->rt_flags & RTF_HOST) == 0)
   2122 		return (0);
   2123 
   2124 	return (rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
   2125 	    rt_mask(rt), rt->rt_flags, NULL));
   2126 }
   2127 
   2128 int
   2129 nd6_setdefaultiface(int ifindex)
   2130 {
   2131 	ifnet_t *ifp;
   2132 	int error = 0;
   2133 	int s;
   2134 
   2135 	s = pserialize_read_enter();
   2136 	ifp = if_byindex(ifindex);
   2137 	if (ifp == NULL) {
   2138 		pserialize_read_exit(s);
   2139 		return EINVAL;
   2140 	}
   2141 	if (nd6_defifindex != ifindex) {
   2142 		nd6_defifindex = ifindex;
   2143 		nd6_defifp = nd6_defifindex > 0 ? ifp : NULL;
   2144 
   2145 		/*
   2146 		 * Our current implementation assumes one-to-one maping between
   2147 		 * interfaces and links, so it would be natural to use the
   2148 		 * default interface as the default link.
   2149 		 */
   2150 		scope6_setdefault(nd6_defifp);
   2151 	}
   2152 	pserialize_read_exit(s);
   2153 
   2154 	return (error);
   2155 }
   2156