Home | History | Annotate | Line # | Download | only in netinet6
nd6.c revision 1.8
      1 /*	$NetBSD: nd6.c,v 1.8 1999/07/31 18:41:17 itojun Exp $	*/
      2 
      3 /*
      4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      5  * All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the project nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  */
     31 
     32 /*
     33  * XXX
     34  * KAME 970409 note:
     35  * BSD/OS version heavily modifies this code, related to llinfo.
     36  * Since we don't have BSD/OS version of net/route.c in our hand,
     37  * I left the code mostly as it was in 970310.  -- itojun
     38  */
     39 
     40 #include <sys/param.h>
     41 #include <sys/systm.h>
     42 #include <sys/malloc.h>
     43 #include <sys/mbuf.h>
     44 #include <sys/socket.h>
     45 #include <sys/sockio.h>
     46 #include <sys/time.h>
     47 #include <sys/kernel.h>
     48 #include <sys/errno.h>
     49 #if !defined(__FreeBSD__) || __FreeBSD__ < 3
     50 #include <sys/ioctl.h>
     51 #endif
     52 #include <sys/syslog.h>
     53 #include <sys/queue.h>
     54 
     55 #include <net/if.h>
     56 #include <net/if_dl.h>
     57 #include <net/if_types.h>
     58 #include <net/if_atm.h>
     59 #include <net/route.h>
     60 
     61 #include <netinet/in.h>
     62 #ifndef __NetBSD__
     63 #include <netinet/if_ether.h>
     64 #ifdef __FreeBSD__
     65 #include <netinet/if_fddi.h>
     66 #endif
     67 #ifdef __bsdi__
     68 #include <net/if_fddi.h>
     69 #endif
     70 #else /* __NetBSD__ */
     71 #include <net/if_ether.h>
     72 #include <netinet/if_inarp.h>
     73 #include <net/if_fddi.h>
     74 #endif /* __NetBSD__ */
     75 #include <netinet6/in6_var.h>
     76 #include <netinet6/ip6.h>
     77 #include <netinet6/ip6_var.h>
     78 #include <netinet6/nd6.h>
     79 #include <netinet6/icmp6.h>
     80 
     81 #include "loop.h"
     82 #ifdef __NetBSD__
     83 extern struct ifnet loif[NLOOP];
     84 #endif
     85 
     86 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
     87 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
     88 
     89 #if !defined(__FreeBSD__) || __FreeBSD__ < 3
     90 #define time_second time.tv_sec
     91 #endif
     92 
     93 #define SIN6(s) ((struct sockaddr_in6 *)s)
     94 #define SDL(s) ((struct sockaddr_dl *)s)
     95 
     96 /* timer values */
     97 int	nd6_prune	= 1;	/* walk list every 1 seconds */
     98 int	nd6_delay	= 5;	/* delay first probe time 5 second */
     99 int	nd6_umaxtries	= 3;	/* maximum unicast query */
    100 int	nd6_mmaxtries	= 3;	/* maximum multicast query */
    101 int	nd6_useloopback = 1;	/* use loopback interface for local traffic */
    102 int	nd6_proxyall	= 0;	/* enable Proxy Neighbor Advertisement */
    103 
    104 /* for debugging? */
    105 static int nd6_inuse, nd6_allocated;
    106 
    107 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
    108 struct nd_ifinfo *nd_ifinfo;
    109 struct nd_drhead nd_defrouter = { 0 };
    110 struct nd_prhead nd_prefix = { 0 };
    111 
    112 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
    113 #if 0
    114 extern	int ip6_forwarding;
    115 #endif
    116 
    117 static void nd6_slowtimo __P((void *));
    118 
    119 void
    120 nd6_init()
    121 {
    122 	static int nd6_init_done = 0;
    123 
    124 	if (nd6_init_done) {
    125 		log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
    126 		return;
    127 	}
    128 	nd6_init_done = 1;
    129 
    130 	/* start timer */
    131 	timeout(nd6_slowtimo, (caddr_t)0, ND6_SLOWTIMER_INTERVAL * hz);
    132 }
    133 
    134 void
    135 nd6_ifattach(ifp)
    136 	struct ifnet *ifp;
    137 {
    138 	static size_t if_indexlim = 8;
    139 
    140 	/*
    141 	 * We have some arrays that should be indexed by if_index.
    142 	 * since if_index will grow dynamically, they should grow too.
    143 	 */
    144 	if (nd_ifinfo == NULL || if_index >= if_indexlim) {
    145 		size_t n;
    146 		caddr_t q;
    147 
    148 		while (if_index >= if_indexlim)
    149 			if_indexlim <<= 1;
    150 
    151 		/* grow nd_ifinfo */
    152 		n = if_indexlim * sizeof(struct nd_ifinfo);
    153 		q = (caddr_t)malloc(n, M_IP6NDP, M_WAITOK);
    154 		bzero(q, n);
    155 		if (nd_ifinfo) {
    156 			bcopy((caddr_t)nd_ifinfo, q, n/2);
    157 			free((caddr_t)nd_ifinfo, M_IP6NDP);
    158 		}
    159 		nd_ifinfo = (struct nd_ifinfo *)q;
    160 	}
    161 
    162 #define ND nd_ifinfo[ifp->if_index]
    163 	ND.linkmtu = ifindex2ifnet[ifp->if_index]->if_mtu;
    164 	ND.chlim = IPV6_DEFHLIM;
    165 	ND.basereachable = REACHABLE_TIME;
    166 	ND.reachable = ND_COMPUTE_RTIME(ND.basereachable);
    167 	ND.retrans = RETRANS_TIMER;
    168 	ND.receivedra = 0;
    169 	nd6_setmtu(ifp);
    170 #undef ND
    171 }
    172 
    173 /*
    174  * Reset ND level link MTU. This function is called when the physical MTU
    175  * changes, which means we might have to adjust the ND level MTU.
    176  */
    177 void
    178 nd6_setmtu(ifp)
    179 	struct ifnet *ifp;
    180 {
    181 #define MIN(a,b) ((a) < (b) ? (a) : (b))
    182 	struct nd_ifinfo *ndi = &nd_ifinfo[ifp->if_index];
    183 	u_long oldmaxmtu = ndi->maxmtu;
    184 	u_long oldlinkmtu = ndi->linkmtu;
    185 
    186 	switch(ifp->if_type) {
    187 	 case IFT_ETHER:
    188 		 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
    189 		 break;
    190 #if defined(__FreeBSD__) || defined(__bsdi__)
    191 	 case IFT_FDDI:
    192 		 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu);
    193 		 break;
    194 #endif
    195 	 case IFT_ATM:
    196 		 ndi->maxmtu = MIN(ATMMTU, ifp->if_mtu);
    197 		 break;
    198 	 default:
    199 		 ndi->maxmtu = ifp->if_mtu;
    200 		 break;
    201 	}
    202 
    203 	if (oldmaxmtu != ndi->maxmtu) {
    204 		/*
    205 		 * If the ND level MTU is not set yet, or if the maxmtu
    206 		 * is reset to a smaller value than the ND level MTU,
    207 		 * also reset the ND level MTU.
    208 		 */
    209 		if (ndi->linkmtu == 0 ||
    210 		    ndi->maxmtu < ndi->linkmtu) {
    211 			ndi->linkmtu = ndi->maxmtu;
    212 			/* also adjust in6_maxmtu if necessary. */
    213 			if (oldlinkmtu == 0) {
    214 				/*
    215 				 * XXX: the case analysis is grotty, but
    216 				 * it is not efficient to call in6_setmaxmtu()
    217 				 * here when we are during the initialization
    218 				 * procedure.
    219 				 */
    220 				if (in6_maxmtu < ndi->linkmtu)
    221 					in6_maxmtu = ndi->linkmtu;
    222 			}
    223 			else
    224 				in6_setmaxmtu();
    225 		}
    226 	}
    227 #undef MIN
    228 }
    229 
    230 void
    231 nd6_option_init(opt, icmp6len, ndopts)
    232 	void *opt;
    233 	int icmp6len;
    234 	union nd_opts *ndopts;
    235 {
    236 	bzero(ndopts, sizeof(*ndopts));
    237 	ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
    238 	ndopts->nd_opts_last
    239 		= (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
    240 
    241 	if (icmp6len == 0) {
    242 		ndopts->nd_opts_done = 1;
    243 		ndopts->nd_opts_search = NULL;
    244 	}
    245 }
    246 
    247 /*
    248  * Take one ND option.
    249  */
    250 struct nd_opt_hdr *
    251 nd6_option(ndopts)
    252 	union nd_opts *ndopts;
    253 {
    254 	struct nd_opt_hdr *nd_opt;
    255 	int olen;
    256 
    257 	if (!ndopts)
    258 		panic("ndopts == NULL in nd6_option\n");
    259 	if (!ndopts->nd_opts_last)
    260 		panic("uninitialized ndopts in nd6_option\n");
    261 	if (!ndopts->nd_opts_search)
    262 		return NULL;
    263 	if (ndopts->nd_opts_done)
    264 		return NULL;
    265 
    266 	nd_opt = ndopts->nd_opts_search;
    267 
    268 	olen = nd_opt->nd_opt_len << 3;
    269 	if (olen == 0) {
    270 		/*
    271 		 * Message validation requires that all included
    272 		 * options have a length that is greater than zero.
    273 		 */
    274 		bzero(ndopts, sizeof(*ndopts));
    275 		return NULL;
    276 	}
    277 
    278 	ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
    279 	if (!(ndopts->nd_opts_search < ndopts->nd_opts_last)) {
    280 		ndopts->nd_opts_done = 1;
    281 		ndopts->nd_opts_search = NULL;
    282 	}
    283 	return nd_opt;
    284 }
    285 
    286 /*
    287  * Parse multiple ND options.
    288  * This function is much easier to use, for ND routines that do not need
    289  * multiple options of the same type.
    290  */
    291 int
    292 nd6_options(ndopts)
    293 	union nd_opts *ndopts;
    294 {
    295 	struct nd_opt_hdr *nd_opt;
    296 	int i = 0;
    297 
    298 	if (!ndopts)
    299 		panic("ndopts == NULL in nd6_options\n");
    300 	if (!ndopts->nd_opts_last)
    301 		panic("uninitialized ndopts in nd6_options\n");
    302 	if (!ndopts->nd_opts_search)
    303 		return 0;
    304 
    305 	while (1) {
    306 		nd_opt = nd6_option(ndopts);
    307 		if (!nd_opt && !ndopts->nd_opts_last) {
    308 			/*
    309 			 * Message validation requires that all included
    310 			 * options have a length that is greater than zero.
    311 			 */
    312 			bzero(ndopts, sizeof(*ndopts));
    313 			return -1;
    314 		}
    315 
    316 		if (!nd_opt)
    317 			goto skip1;
    318 
    319 		switch (nd_opt->nd_opt_type) {
    320 		case ND_OPT_SOURCE_LINKADDR:
    321 		case ND_OPT_TARGET_LINKADDR:
    322 		case ND_OPT_MTU:
    323 		case ND_OPT_REDIRECTED_HEADER:
    324 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
    325 				printf("duplicated ND6 option found "
    326 					"(type=%d)\n", nd_opt->nd_opt_type);
    327 				/* XXX bark? */
    328 			} else {
    329 				ndopts->nd_opt_array[nd_opt->nd_opt_type]
    330 					= nd_opt;
    331 			}
    332 			break;
    333 		case ND_OPT_PREFIX_INFORMATION:
    334 			if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
    335 				ndopts->nd_opt_array[nd_opt->nd_opt_type]
    336 					= nd_opt;
    337 			}
    338 			ndopts->nd_opts_pi_end =
    339 				(struct nd_opt_prefix_info *)nd_opt;
    340 			break;
    341 		default:
    342 			/*
    343 			 * Unknown options must be silently ignored,
    344 			 * to accomodate future extension to the protocol.
    345 			 */
    346 			log(LOG_INFO,
    347 			    "nd6_options: unsupported option %d - "
    348 			    "option ignored\n", nd_opt->nd_opt_type);
    349 		}
    350 
    351 skip1:
    352 		i++;
    353 		if (i > 10) {
    354 			printf("too many loop in nd opt\n");
    355 			break;
    356 		}
    357 
    358 		if (ndopts->nd_opts_done)
    359 			break;
    360 	}
    361 
    362 	return 0;
    363 }
    364 
    365 /*
    366  * ND6 timer routine to expire default route list and prefix list
    367  */
    368 void
    369 nd6_timer(ignored_arg)
    370 	void	*ignored_arg;
    371 {
    372 	int s;
    373 	register struct llinfo_nd6 *ln;
    374 	register struct nd_defrouter *dr;
    375 	register struct nd_prefix *pr;
    376 
    377 	s = splsoftnet();
    378 	timeout(nd6_timer, (caddr_t)0, nd6_prune * hz);
    379 
    380 	ln = llinfo_nd6.ln_next;
    381 	/* XXX BSD/OS separates this code -- itojun */
    382 	while (ln && ln != &llinfo_nd6) {
    383 		struct rtentry *rt;
    384 		struct ifnet *ifp;
    385 		struct sockaddr_in6 *dst;
    386 		struct llinfo_nd6 *next = ln->ln_next;
    387 
    388 		if ((rt = ln->ln_rt) == NULL) {
    389 			ln = next;
    390 			continue;
    391 		}
    392 		if ((ifp = rt->rt_ifp) == NULL) {
    393 			ln = next;
    394 			continue;
    395 		}
    396 		dst = (struct sockaddr_in6 *)rt_key(rt);
    397 
    398 		if (ln->ln_expire > time_second) {
    399 			ln = next;
    400 			continue;
    401 		}
    402 
    403 		/* sanity check */
    404 		if (!rt)
    405 			panic("rt=0 in nd6_timer(ln=%p)\n", ln);
    406 		if (!dst)
    407 			panic("dst=0 in nd6_timer(ln=%p)\n", ln);
    408 
    409 		switch (ln->ln_state) {
    410 		case ND6_LLINFO_INCOMPLETE:
    411 			if (ln->ln_asked < nd6_mmaxtries) {
    412 				ln->ln_asked++;
    413 				ln->ln_expire = time_second +
    414 					nd_ifinfo[ifp->if_index].retrans / 1000;
    415 				nd6_ns_output(ifp, NULL, &dst->sin6_addr,
    416 					ln, 0);
    417 			} else {
    418 				struct mbuf *m = ln->ln_hold;
    419 				if (m) {
    420 					if (rt->rt_ifp) {
    421 						/*
    422 						 * Fake rcvif to make ICMP error
    423 						 * more helpful in diagnosing
    424 						 * for the receiver.
    425 						 * XXX: should we consider
    426 						 * older rcvif?
    427 						 */
    428 						m->m_pkthdr.rcvif = rt->rt_ifp;
    429 					}
    430 					icmp6_error(m, ICMP6_DST_UNREACH,
    431 						    ICMP6_DST_UNREACH_ADDR, 0);
    432 					ln->ln_hold = NULL;
    433 				}
    434 				nd6_free(rt);
    435 			}
    436 			break;
    437 		case ND6_LLINFO_REACHABLE:
    438 			if (ln->ln_expire) {
    439 				ln->ln_state = ND6_LLINFO_STALE;
    440 			}
    441 			break;
    442 		/*
    443 		 * ND6_LLINFO_STALE state requires nothing for timer
    444 		 * routine.
    445 		 */
    446 		case ND6_LLINFO_DELAY:
    447 			ln->ln_asked = 1;
    448 			ln->ln_state = ND6_LLINFO_PROBE;
    449 			ln->ln_expire = time_second +
    450 				nd_ifinfo[ifp->if_index].retrans / 1000;
    451 			nd6_ns_output(ifp, &dst->sin6_addr, &dst->sin6_addr,
    452 				ln, 0);
    453 			break;
    454 
    455 		case ND6_LLINFO_PROBE:
    456 			if (ln->ln_asked < nd6_umaxtries) {
    457 				ln->ln_asked++;
    458 				ln->ln_expire = time_second +
    459 					nd_ifinfo[ifp->if_index].retrans / 1000;
    460 				nd6_ns_output(ifp, &dst->sin6_addr,
    461 					       &dst->sin6_addr, ln, 0);
    462 			} else {
    463 				nd6_free(rt);
    464 			}
    465 			break;
    466 		case ND6_LLINFO_WAITDELETE:
    467 			nd6_free(rt);
    468 			break;
    469 		}
    470 		ln = next;
    471 	}
    472 
    473 	/* expire */
    474 	dr = nd_defrouter.lh_first;
    475 	while (dr) {
    476 		if (dr->expire && dr->expire < time_second) {
    477 			struct nd_defrouter *t;
    478 			t = dr->dr_next;
    479 			defrtrlist_del(dr);
    480 			dr = t;
    481 		} else
    482 			dr = dr->dr_next;
    483 	}
    484 	pr = nd_prefix.lh_first;
    485 	while (pr) {
    486 		struct in6_ifaddr *ia6;
    487 		struct in6_addrlifetime *lt6;
    488 
    489 		if (IN6_IS_ADDR_UNSPECIFIED(&pr->ndpr_addr))
    490 			ia6 = NULL;
    491 		else
    492 			ia6 = in6ifa_ifpwithaddr(pr->ndpr_ifp, &pr->ndpr_addr);
    493 
    494 		if (ia6) {
    495 			/* check address lifetime */
    496 			lt6 = &ia6->ia6_lifetime;
    497 			if (lt6->ia6t_preferred && lt6->ia6t_preferred < time_second)
    498 				ia6->ia6_flags |= IN6_IFF_DEPRECATED;
    499 			if (lt6->ia6t_expire && lt6->ia6t_expire < time_second) {
    500 				if (!IN6_IS_ADDR_UNSPECIFIED(&pr->ndpr_addr))
    501 					in6_ifdel(pr->ndpr_ifp, &pr->ndpr_addr);
    502 				/* xxx ND_OPT_PI_FLAG_ONLINK processing */
    503 			}
    504 		}
    505 
    506 		/*
    507 		 * check prefix lifetime.
    508 		 * since pltime is just for autoconf, pltime processing for
    509 		 * prefix is not necessary.
    510 		 *
    511 		 * we offset expire time by NDPR_KEEP_EXPIRE, so that we
    512 		 * can use the old prefix information to validate the
    513 		 * next prefix information to come.  See prelist_update()
    514 		 * for actual validation.
    515 		 */
    516 		if (pr->ndpr_expire
    517 		 && pr->ndpr_expire + NDPR_KEEP_EXPIRED < time_second) {
    518 			struct nd_prefix *t;
    519 			t = pr->ndpr_next;
    520 
    521 			/*
    522 			 * address expiration and prefix expiration are
    523 			 * separate.  NEVER perform in6_ifdel here.
    524 			 */
    525 
    526 			prelist_remove(pr);
    527 			pr = t;
    528 		} else
    529 			pr = pr->ndpr_next;
    530 	}
    531 	splx(s);
    532 }
    533 
    534 static struct sockaddr_in6 all1_sa = {
    535 	sizeof(struct sockaddr_in6), AF_INET6, 0, 0,
    536 	{{{0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff}}}, 0};
    537 
    538 struct rtentry *
    539 nd6_lookup(addr6, create, ifp)
    540 	struct in6_addr *addr6;
    541 	int create;
    542 	struct ifnet *ifp;
    543 {
    544 	struct rtentry *rt;
    545 	struct sockaddr_in6 sin6;
    546 
    547 	bzero(&sin6, sizeof(sin6));
    548 	sin6.sin6_len = sizeof(struct sockaddr_in6);
    549 	sin6.sin6_family = AF_INET6;
    550 	sin6.sin6_addr = *addr6;
    551 	rt = rtalloc1((struct sockaddr *)&sin6, create
    552 #ifdef __FreeBSD__
    553 		      , 0UL
    554 #endif /*__FreeBSD__*/
    555 		      );
    556 	if (rt && (rt->rt_flags & RTF_LLINFO) == 0) {
    557 		/*
    558 		 * This is the case for the default route.
    559 		 * If we want to create a neighbor cache for the address, we
    560 		 * should free the route for the destination and allocate an
    561 		 * interface route.
    562 		 */
    563 		if (create) {
    564 			RTFREE(rt);
    565 			rt = 0;
    566 		}
    567 	}
    568 	if (!rt) {
    569 		if (create && ifp) {
    570 			/*
    571 			 * If no route is available and create is set,
    572 			 * we allocate a host route for the destination
    573 			 * and treat it like an interface route.
    574 			 * This hack is necessary for a neighbor which can't
    575 			 * be covered by our own prefix.
    576 			 */
    577 			struct ifaddr *ifa =
    578 				ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
    579 			if (ifa == NULL)
    580 				return(NULL);
    581 
    582 			/*
    583 			 * Create a new route. RTF_LLINFO is necessary
    584 			 * to create a Neighbor Cache entry for the
    585 			 * destination in nd6_rtrequest which will be
    586 			 * called in rtequest via ifa->ifa_rtrequest.
    587 			 */
    588 			if (rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
    589 				      ifa->ifa_addr,
    590 				      (struct sockaddr *)&all1_sa,
    591 				      (ifa->ifa_flags |
    592 				       RTF_HOST | RTF_LLINFO) & ~RTF_CLONING,
    593 				      &rt))
    594 				log(LOG_ERR,
    595 				    "nd6_lookup: failed to add route for a "
    596 				    "neighbor(%s)\n", ip6_sprintf(addr6));
    597 			if (rt == NULL)
    598 				return(NULL);
    599 			if (rt->rt_llinfo) {
    600 				struct llinfo_nd6 *ln =
    601 					(struct llinfo_nd6 *)rt->rt_llinfo;
    602 				ln->ln_state = ND6_LLINFO_NOSTATE;
    603 			}
    604 		}
    605 		else
    606 			return(NULL);
    607 	}
    608 	rt->rt_refcnt--;
    609 	if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
    610 	   rt->rt_gateway->sa_family != AF_LINK) {
    611 		if (create) {
    612 			log(LOG_DEBUG, "nd6_lookup: failed to lookup %s\n",
    613 			    ip6_sprintf(addr6));
    614 			/* xxx more logs... kazu */
    615 		}
    616 		return(0);
    617 	}
    618 	return(rt);
    619 }
    620 
    621 /*
    622  * Detect if a given IPv6 address identifies a neighbor on a given link.
    623  * XXX: should take care of the destination of a p2p link?
    624  */
    625 int
    626 nd6_is_addr_neighbor(addr, ifp)
    627 	struct in6_addr *addr;
    628 	struct ifnet *ifp;
    629 {
    630 	register struct ifaddr *ifa;
    631 	int i;
    632 
    633 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
    634 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
    635 
    636 	/* A link-local address is always a neighbor. */
    637 	if (IN6_IS_ADDR_LINKLOCAL(addr))
    638 		return(1);
    639 
    640 	/*
    641 	 * If the address matches one of our addresses,
    642 	 * it should be a neighbor.
    643 	 */
    644 #ifdef __bsdi__
    645 	for (ifa = ifp->if_addrlist; ifa; ifa = ifa->ifa_next)
    646 #else
    647 	for (ifa = ifp->if_addrlist.tqh_first;
    648 	     ifa;
    649 	     ifa = ifa->ifa_list.tqe_next)
    650 #endif
    651 	{
    652 		if (ifa->ifa_addr->sa_family != AF_INET6)
    653 			next: continue;
    654 
    655 		for (i = 0; i < 4; i++) {
    656 			if ((IFADDR6(ifa).s6_addr32[i] ^ addr->s6_addr32[i]) &
    657 			    IFMASK6(ifa).s6_addr32[i])
    658 				goto next;
    659 		}
    660 		return(1);
    661 	}
    662 
    663 	/*
    664 	 * Even if the address matches none of our addresses, it might be
    665 	 * in the neighbor cache.
    666 	 */
    667 	if (nd6_lookup(addr, 0, ifp))
    668 		return(1);
    669 
    670 	return(0);
    671 #undef IFADDR6
    672 #undef IFMASK6
    673 }
    674 
    675 /*
    676  * Free an nd6 llinfo entry.
    677  */
    678 void
    679 nd6_free(rt)
    680 	struct rtentry *rt;
    681 {
    682 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
    683 	struct sockaddr_dl *sdl;
    684 
    685 	if (ln->ln_router) {
    686 		/* remove from default router list */
    687 		struct nd_defrouter *dr;
    688 		struct in6_addr *in6;
    689 		int s;
    690 		in6 = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
    691 
    692 		s = splsoftnet();
    693 		dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->
    694 				      sin6_addr,
    695 				      rt->rt_ifp);
    696 		if (dr)
    697 			defrtrlist_del(dr);
    698 		else if (!ip6_forwarding && ip6_accept_rtadv) {
    699 			/*
    700 			 * rt6_flush must be called in any case.
    701 			 * see the comment in nd6_na_input().
    702 			 */
    703 			rt6_flush(in6, rt->rt_ifp);
    704 		}
    705 		splx(s);
    706 	}
    707 
    708 	if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) &&
    709 	   sdl->sdl_family == AF_LINK) {
    710 		sdl->sdl_alen = 0;
    711 		ln->ln_state = ND6_LLINFO_WAITDELETE;
    712 		ln->ln_asked = 0;
    713 		rt->rt_flags &= ~RTF_REJECT;
    714 		return;
    715 	}
    716 	rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt),
    717 		  0, (struct rtentry **)0);
    718 }
    719 
    720 /*
    721  * Upper-layer reachability hint for Neighbor Unreachability Detection.
    722  *
    723  * XXX cost-effective metods?
    724  */
    725 void
    726 nd6_nud_hint(rt, dst6)
    727 	struct rtentry *rt;
    728 	struct in6_addr *dst6;
    729 {
    730 	struct llinfo_nd6 *ln;
    731 
    732 	/*
    733 	 * If the caller specified "rt", use that.  Otherwise, resolve the
    734 	 * routing table by supplied "dst6".
    735 	 */
    736 	if (!rt) {
    737 		if (!dst6)
    738 			return;
    739 		if (!(rt = nd6_lookup(dst6, 0, NULL)))
    740 			return;
    741 	}
    742 
    743 	if ((rt->rt_flags & RTF_GATEWAY)
    744 	 || (rt->rt_flags & RTF_LLINFO) == 0
    745 	 || !rt->rt_llinfo
    746 	 || !rt->rt_gateway
    747 	 || rt->rt_gateway->sa_family != AF_LINK) {
    748 		/* This is not a host route. */
    749 		return;
    750 	}
    751 
    752 	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
    753 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE)
    754 		return;
    755 
    756 	ln->ln_state = ND6_LLINFO_REACHABLE;
    757 	if (ln->ln_expire)
    758 		ln->ln_expire = time_second +
    759 			nd_ifinfo[rt->rt_ifp->if_index].reachable;
    760 }
    761 
    762 /*
    763  * Resolve an IP6 address into an ethernet address. If success,
    764  * desten is filled in. If there is no entry in ndptab,
    765  * set one up and multicast a solicitation for the IP6 address.
    766  * Hold onto this mbuf and resend it once the address
    767  * is finally resolved. A return value of 1 indicates
    768  * that desten has been filled in and the packet should be sent
    769  * normally; a 0 return indicates that the packet has been
    770  * taken over here, either now or for later transmission.
    771  */
    772 int
    773 nd6_resolve(ifp, rt, m, dst, desten)
    774 	struct ifnet *ifp;
    775 	struct rtentry *rt;
    776 	struct mbuf *m;
    777 	struct sockaddr *dst;
    778 	u_char *desten;
    779 {
    780 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)NULL;
    781 	struct sockaddr_dl *sdl;
    782 
    783 	if (m->m_flags & M_MCAST) {
    784 		switch (ifp->if_type) {
    785 		case IFT_ETHER:
    786 		case IFT_FDDI:
    787 			ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
    788 						 desten);
    789 			return(1);
    790 			break;
    791 		default:
    792 			return(0);
    793 		}
    794 	}
    795 	if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
    796 		ln = (struct llinfo_nd6 *)rt->rt_llinfo;
    797 	else {
    798 		if ((rt = nd6_lookup(&(SIN6(dst)->sin6_addr), 1, ifp)) != NULL)
    799 			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
    800 	}
    801 	if (!ln || !rt) {
    802 		log(LOG_DEBUG, "nd6_resolve: can't allocate llinfo for %s\n",
    803 			ip6_sprintf(&(SIN6(dst)->sin6_addr)));
    804 		m_freem(m);
    805 		return(0);
    806 	}
    807 	sdl = SDL(rt->rt_gateway);
    808 	/*
    809 	 * Ckeck the address family and length is valid, the address
    810 	 * is resolved; otherwise, try to resolve.
    811 	 */
    812 	if (ln->ln_state >= ND6_LLINFO_REACHABLE
    813 	   && sdl->sdl_family == AF_LINK
    814 	   && sdl->sdl_alen != 0) {
    815 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
    816 		if (ln->ln_state == ND6_LLINFO_STALE) {
    817 			ln->ln_asked = 0;
    818 			ln->ln_state = ND6_LLINFO_DELAY;
    819 			ln->ln_expire = time_second + nd6_delay;
    820 		}
    821 		return(1);
    822 	}
    823 	/*
    824 	 * There is an ndp entry, but no ethernet address
    825 	 * response yet. Replace the held mbuf with this
    826 	 * latest one.
    827 	 *
    828 	 * XXX Does the code conform to rate-limiting rule?
    829 	 * (RFC 2461 7.2.2)
    830 	 */
    831 	if (ln->ln_state == ND6_LLINFO_WAITDELETE ||
    832 	    ln->ln_state == ND6_LLINFO_NOSTATE)
    833 		ln->ln_state = ND6_LLINFO_INCOMPLETE;
    834 	if (ln->ln_hold)
    835 		m_freem(ln->ln_hold);
    836 	ln->ln_hold = m;
    837 	if (ln->ln_expire) {
    838 		rt->rt_flags &= ~RTF_REJECT;
    839 		if (ln->ln_asked < nd6_mmaxtries &&
    840 		    ln->ln_expire < time_second) {
    841 			ln->ln_asked++;
    842 			ln->ln_expire = time_second +
    843 				nd_ifinfo[ifp->if_index].retrans / 1000;
    844 			nd6_ns_output(ifp, NULL, &(SIN6(dst)->sin6_addr),
    845 				ln, 0);
    846 		}
    847 	}
    848 	return(0);
    849 }
    850 
    851 void
    852 nd6_rtrequest(req, rt, sa)
    853 	int	req;
    854 	struct rtentry *rt;
    855 	struct sockaddr *sa; /* xxx unused */
    856 {
    857 	struct sockaddr *gate = rt->rt_gateway;
    858 	struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
    859 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
    860 	struct ifnet *ifp = rt->rt_ifp;
    861 	struct ifaddr *ifa;
    862 
    863 	if (rt->rt_flags & RTF_GATEWAY)
    864 		return;
    865 
    866 	switch (req) {
    867 	case RTM_ADD:
    868 		/*
    869 		 * There is no backward compatibility :)
    870 		 *
    871 		 * if ((rt->rt_flags & RTF_HOST) == 0 &&
    872 		 *     SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
    873 		 *	   rt->rt_flags |= RTF_CLONING;
    874 		 */
    875 		if (rt->rt_flags & RTF_CLONING || rt->rt_flags & RTF_LLINFO) {
    876 			/*
    877 			 * Case 1: This route should come from
    878 			 * a route to interface. RTF_LLINFO flag is set
    879 			 * for a host route whose destination should be
    880 			 * treated as on-link.
    881 			 */
    882 			rt_setgate(rt, rt_key(rt),
    883 				   (struct sockaddr *)&null_sdl);
    884 			gate = rt->rt_gateway;
    885 			SDL(gate)->sdl_type = ifp->if_type;
    886 			SDL(gate)->sdl_index = ifp->if_index;
    887 			if (ln)
    888 				ln->ln_expire = time_second;
    889 #if 1
    890 			if (ln && ln->ln_expire == 0) {
    891 				/* cludge for desktops */
    892 #if 0
    893 				printf("nd6_request: time.tv_sec is zero; "
    894 				       "treat it as 1\n");
    895 #endif
    896 				ln->ln_expire = 1;
    897 			}
    898 #endif
    899 			if (rt->rt_flags & RTF_CLONING)
    900 				break;
    901 		}
    902 		/* Announce a new entry if requested. */
    903 		if (rt->rt_flags & RTF_ANNOUNCE)
    904 			nd6_na_output(ifp,
    905 				      &SIN6(rt_key(rt))->sin6_addr,
    906 				      &SIN6(rt_key(rt))->sin6_addr,
    907 				      ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
    908 				      1);
    909 		/* FALLTHROUGH */
    910 	case RTM_RESOLVE:
    911 		if (gate->sa_family != AF_LINK ||
    912 		   gate->sa_len < sizeof(null_sdl)) {
    913 			log(LOG_DEBUG, "nd6_rtrequest: bad gateway value\n");
    914 			break;
    915 		}
    916 		SDL(gate)->sdl_type = ifp->if_type;
    917 		SDL(gate)->sdl_index = ifp->if_index;
    918 		if (ln != 0)
    919 			break;	/* This happens on a route change */
    920 		/*
    921 		 * Case 2: This route may come from cloning, or a manual route
    922 		 * add with a LL address.
    923 		 */
    924 		R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
    925 		rt->rt_llinfo = (caddr_t)ln;
    926 		if (!ln) {
    927 			log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
    928 			break;
    929 		}
    930 		nd6_inuse++;
    931 		nd6_allocated++;
    932 		Bzero(ln, sizeof(*ln));
    933 		ln->ln_rt = rt;
    934 		/* this is required for "ndp" command. - shin */
    935 		if (req == RTM_ADD) {
    936 		        /*
    937 			 * gate should have some valid AF_LINK entry,
    938 			 * and ln->ln_expire should have some lifetime
    939 			 * which is specified by ndp command.
    940 			 */
    941 			ln->ln_state = ND6_LLINFO_REACHABLE;
    942 		} else {
    943 		        /*
    944 			 * When req == RTM_RESOLVE, rt is created and
    945 			 * initialized in rtrequest(), so rt_expire is 0.
    946 			 */
    947 			ln->ln_state = ND6_LLINFO_INCOMPLETE;
    948 			ln->ln_expire = time_second;
    949 		}
    950 		rt->rt_flags |= RTF_LLINFO;
    951 #if 0
    952 		insque(ln, &llinfo_nd6);
    953 #else
    954 		ln->ln_next = llinfo_nd6.ln_next;
    955 		llinfo_nd6.ln_next = ln;
    956 		ln->ln_prev = &llinfo_nd6;
    957 		ln->ln_next->ln_prev = ln;
    958 #endif
    959 
    960 		/*
    961 		 * check if rt_key(rt) is one of my address assigned
    962 		 * to the interface.
    963 		 */
    964 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
    965 					  &SIN6(rt_key(rt))->sin6_addr);
    966 		if (ifa) {
    967 			caddr_t macp = nd6_ifptomac(ifp);
    968 			ln->ln_expire = 0;
    969 			ln->ln_state = ND6_LLINFO_REACHABLE;
    970 			if (macp) {
    971 				Bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
    972 				SDL(gate)->sdl_alen = ifp->if_addrlen;
    973 			}
    974 			if (nd6_useloopback) {
    975 #ifdef __bsdi__
    976 				extern struct ifnet loif;
    977 				rt->rt_ifp = &loif;	/*XXX*/
    978 #endif /*__bsdi__*/
    979 #if defined(__FreeBSD__) || defined(__NetBSD__)
    980 				rt->rt_ifp = &loif[0];	/*XXX*/
    981 #endif
    982 				/*
    983 				 * Make sure rt_ifa be equal to the ifaddr
    984 				 * corresponding to the address.
    985 				 * We need this because when we refer
    986 				 * rt_ifa->ia6_flags in ip6_input, we assume
    987 				 * that the rt_ifa points to the address instead
    988 				 * of the loopback address.
    989 				 */
    990 				if (ifa != rt->rt_ifa) {
    991 					rt->rt_ifa->ifa_refcnt--;
    992 					ifa->ifa_refcnt++;
    993 					rt->rt_ifa = ifa;
    994 				}
    995 			}
    996 		}
    997 		break;
    998 
    999 	case RTM_DELETE:
   1000 		if (!ln)
   1001 			break;
   1002 		nd6_inuse--;
   1003 #if 0
   1004 		remque(ln);
   1005 #else
   1006 		ln->ln_next->ln_prev = ln->ln_prev;
   1007 		ln->ln_prev->ln_next = ln->ln_next;
   1008 		ln->ln_prev = NULL;
   1009 #endif
   1010 		rt->rt_llinfo = 0;
   1011 		rt->rt_flags &= ~RTF_LLINFO;
   1012 		if (ln->ln_hold)
   1013 			m_freem(ln->ln_hold);
   1014 		Free((caddr_t)ln);
   1015 	}
   1016 }
   1017 
   1018 void
   1019 nd6_p2p_rtrequest(req, rt, sa)
   1020 	int	req;
   1021 	struct rtentry *rt;
   1022 	struct sockaddr *sa; /* xxx unused */
   1023 {
   1024 	struct sockaddr *gate = rt->rt_gateway;
   1025 	static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
   1026 	struct ifnet *ifp = rt->rt_ifp;
   1027 	struct ifaddr *ifa;
   1028 
   1029 	if (rt->rt_flags & RTF_GATEWAY)
   1030 		return;
   1031 
   1032 	switch (req) {
   1033 	case RTM_ADD:
   1034 		/*
   1035 		 * There is no backward compatibility :)
   1036 		 *
   1037 		 * if ((rt->rt_flags & RTF_HOST) == 0 &&
   1038 		 *     SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
   1039 		 *	   rt->rt_flags |= RTF_CLONING;
   1040 		 */
   1041 		if (rt->rt_flags & RTF_CLONING) {
   1042 			/*
   1043 			 * Case 1: This route should come from
   1044 			 * a route to interface.
   1045 			 */
   1046 			rt_setgate(rt, rt_key(rt),
   1047 				   (struct sockaddr *)&null_sdl);
   1048 			gate = rt->rt_gateway;
   1049 			SDL(gate)->sdl_type = ifp->if_type;
   1050 			SDL(gate)->sdl_index = ifp->if_index;
   1051 			break;
   1052 		}
   1053 		/* Announce a new entry if rqquested. */
   1054 		if (rt->rt_flags & RTF_ANNOUNCE)
   1055 			nd6_na_output(ifp,
   1056 				      &SIN6(rt_key(rt))->sin6_addr,
   1057 				      &SIN6(rt_key(rt))->sin6_addr,
   1058 				      ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
   1059 				      1);
   1060 		/* FALLTHROUGH */
   1061 	case RTM_RESOLVE:
   1062 		/*
   1063 		 * check if rt_key(rt) is one of my address assigned
   1064 		 * to the interface.
   1065 		 */
   1066  		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
   1067 					  &SIN6(rt_key(rt))->sin6_addr);
   1068 		if (ifa) {
   1069 			if (nd6_useloopback) {
   1070 #ifdef __bsdi__
   1071 				extern struct ifnet loif;
   1072 				rt->rt_ifp = &loif;	/*XXX*/
   1073 #endif /*__bsdi__*/
   1074 #if defined(__FreeBSD__) || defined(__NetBSD__)
   1075 				rt->rt_ifp = &loif[0];	/*XXX*/
   1076 #endif
   1077 			}
   1078 		}
   1079 		break;
   1080 	}
   1081 }
   1082 
   1083 int
   1084 nd6_ioctl(cmd, data, ifp)
   1085 	u_long cmd;
   1086 	caddr_t	data;
   1087 	struct ifnet *ifp;
   1088 {
   1089 	struct in6_drlist *drl = (struct in6_drlist *)data;
   1090 	struct in6_prlist *prl = (struct in6_prlist *)data;
   1091 	struct in6_ndireq *ndi = (struct in6_ndireq *)data;
   1092 	struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
   1093 	struct nd_defrouter *dr, any;
   1094 	struct nd_prefix *pr;
   1095 	struct rtentry *rt;
   1096 	int i = 0, error = 0;
   1097 	int s;
   1098 
   1099 	switch (cmd) {
   1100 	case SIOCGDRLST_IN6:
   1101 		bzero(drl, sizeof(*drl));
   1102 		s = splsoftnet();
   1103 		dr = nd_defrouter.lh_first;
   1104 		while (dr && i < DRLSTSIZ) {
   1105 			drl->defrouter[i].rtaddr = dr->rtaddr;
   1106 			if (IN6_IS_ADDR_LINKLOCAL(&drl->defrouter[i].rtaddr)) {
   1107 				/* XXX: need to this hack for KAME stack */
   1108 				drl->defrouter[i].rtaddr.s6_addr16[1] = 0;
   1109 			}
   1110 			else
   1111 				log(LOG_ERR,
   1112 				    "default router list contains a "
   1113 				    "non-linklocal address(%s)\n",
   1114 				    ip6_sprintf(&drl->defrouter[i].rtaddr));
   1115 
   1116 			drl->defrouter[i].flags = dr->flags;
   1117 			drl->defrouter[i].rtlifetime = dr->rtlifetime;
   1118 			drl->defrouter[i].expire = dr->expire;
   1119 			drl->defrouter[i].if_index = dr->ifp->if_index;
   1120 			i++;
   1121 			dr = dr->dr_next;
   1122 		}
   1123 		splx(s);
   1124 		break;
   1125 	case SIOCGPRLST_IN6:
   1126 		bzero(prl, sizeof(*prl));
   1127 		s = splsoftnet();
   1128 		pr = nd_prefix.lh_first;
   1129 		while (pr && i < PRLSTSIZ) {
   1130 			struct nd_pfxrouter *pfr;
   1131 			int j;
   1132 
   1133 			prl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
   1134 			prl->prefix[i].raflags = pr->ndpr_raf;
   1135 			prl->prefix[i].prefixlen = pr->ndpr_plen;
   1136 			prl->prefix[i].vltime = pr->ndpr_vltime;
   1137 			prl->prefix[i].pltime = pr->ndpr_pltime;
   1138 			prl->prefix[i].if_index = pr->ndpr_ifp->if_index;
   1139 			prl->prefix[i].expire = pr->ndpr_expire;
   1140 
   1141 			pfr = pr->ndpr_advrtrs.lh_first;
   1142 			j = 0;
   1143 			while(pfr) {
   1144 				if (j < DRLSTSIZ) {
   1145 #define RTRADDR prl->prefix[i].advrtr[j]
   1146 					RTRADDR = pfr->router->rtaddr;
   1147 					if (IN6_IS_ADDR_LINKLOCAL(&RTRADDR)) {
   1148 						/* XXX: hack for KAME */
   1149 						RTRADDR.s6_addr16[1] = 0;
   1150 					}
   1151 					else
   1152 						log(LOG_ERR,
   1153 						    "a router(%s) advertises "
   1154 						    "a prefix with "
   1155 						    "non-link local address\n",
   1156 						    ip6_sprintf(&RTRADDR));
   1157 #undef RTRADDR
   1158 				}
   1159 				j++;
   1160 				pfr = pfr->pfr_next;
   1161 			}
   1162 			prl->prefix[i].advrtrs = j;
   1163 
   1164 			i++;
   1165 			pr = pr->ndpr_next;
   1166 		}
   1167 		splx(s);
   1168 		break;
   1169 	case SIOCGIFINFO_IN6:
   1170 		ndi->ndi = nd_ifinfo[ifp->if_index];
   1171 		break;
   1172 	case SIOCSNDFLUSH_IN6:
   1173 		/* flush default router list */
   1174 		/*
   1175 		 * xxx sumikawa: should not delete route if default
   1176 		 * route equals to the top of default router list
   1177 		 */
   1178 		bzero(&any, sizeof(any));
   1179 		defrouter_delreq(&any, 0);
   1180 		/* xxx sumikawa: flush prefix list */
   1181 		break;
   1182 	case SIOCSPFXFLUSH_IN6:
   1183 	    {
   1184 		/* flush all the prefix advertised by routers */
   1185 		struct nd_prefix *pr, *next;
   1186 
   1187 		s = splsoftnet();
   1188 		for (pr = nd_prefix.lh_first; pr; pr = next) {
   1189 			next = pr->ndpr_next;
   1190 			if (!IN6_IS_ADDR_UNSPECIFIED(&pr->ndpr_addr))
   1191 				in6_ifdel(pr->ndpr_ifp, &pr->ndpr_addr);
   1192 			prelist_remove(pr);
   1193 		}
   1194 		splx(s);
   1195 		break;
   1196 	    }
   1197 	case SIOCSRTRFLUSH_IN6:
   1198 	    {
   1199 		/* flush all the default routers */
   1200 		struct nd_defrouter *dr, *next;
   1201 
   1202 		s = splsoftnet();
   1203 		if ((dr = nd_defrouter.lh_first) != NULL) {
   1204 			/*
   1205 			 * The first entry of the list may be stored in
   1206 			 * the routing table, so we'll delete it later.
   1207 			 */
   1208 			for (dr = dr->dr_next; dr; dr = next) {
   1209 				next = dr->dr_next;
   1210 				defrtrlist_del(dr);
   1211 			}
   1212 			defrtrlist_del(nd_defrouter.lh_first);
   1213 		}
   1214 		splx(s);
   1215 		break;
   1216 	    }
   1217 	case SIOCGNBRINFO_IN6:
   1218 	    {
   1219 		  struct llinfo_nd6 *ln;
   1220 
   1221 		  s = splsoftnet();
   1222 		  if ((rt = nd6_lookup(&nbi->addr, 0, ifp)) == NULL) {
   1223 			  error = EINVAL;
   1224 			  break;
   1225 		  }
   1226 		  ln = (struct llinfo_nd6 *)rt->rt_llinfo;
   1227 		  nbi->state = ln->ln_state;
   1228 		  nbi->asked = ln->ln_asked;
   1229 		  nbi->isrouter = ln->ln_router;
   1230 		  nbi->expire = ln->ln_expire;
   1231 		  splx(s);
   1232 
   1233 		  break;
   1234 	    }
   1235 	}
   1236 	return(error);
   1237 }
   1238 
   1239 /*
   1240  * Create neighbor cache entry and cache link-layer address,
   1241  * on reception of inbound ND6 packets. (RS/RA/NS/redirect)
   1242  */
   1243 struct rtentry *
   1244 nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code)
   1245 	struct ifnet *ifp;
   1246 	struct in6_addr *from;
   1247 	char *lladdr;
   1248 	int lladdrlen;
   1249 	int type;	/* ICMP6 type */
   1250 	int code;	/* type dependent information */
   1251 {
   1252 	struct rtentry *rt = NULL;
   1253 	struct llinfo_nd6 *ln = NULL;
   1254 	int is_newentry;
   1255 	struct sockaddr_dl *sdl = NULL;
   1256 	int do_update;
   1257 	int olladdr;
   1258 	int llchange;
   1259 	int newstate = 0;
   1260 
   1261 	if (!ifp)
   1262 		panic("ifp == NULL in nd6_cache_lladdr");
   1263 	if (!from)
   1264 		panic("from == NULL in nd6_cache_lladdr");
   1265 
   1266 	/* nothing must be updated for unspecified address */
   1267 	if (IN6_IS_ADDR_UNSPECIFIED(from))
   1268 		return NULL;
   1269 
   1270 	/*
   1271 	 * Validation about ifp->if_addrlen and lladdrlen must be done in
   1272 	 * the caller.
   1273 	 *
   1274 	 * XXX If the link does not have link-layer adderss, what should
   1275 	 * we do? (ifp->if_addrlen == 0)
   1276 	 * Spec says nothing in sections for RA, RS and NA.  There's small
   1277 	 * description on it in NS section (RFC 2461 7.2.3).
   1278 	 */
   1279 
   1280 	rt = nd6_lookup(from, 0, ifp);
   1281 	if (!rt) {
   1282 #if 0
   1283 		/* nothing must be done if there's no lladdr */
   1284 		if (!lladdr || !lladdrlen)
   1285 			return NULL;
   1286 #endif
   1287 
   1288 		rt = nd6_lookup(from, 1, ifp);
   1289 		is_newentry = 1;
   1290 	} else
   1291 		is_newentry = 0;
   1292 
   1293 	if (!rt)
   1294 		return NULL;
   1295 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
   1296 fail:
   1297 		nd6_free(rt);
   1298 		return NULL;
   1299 	}
   1300 	ln = (struct llinfo_nd6 *)rt->rt_llinfo;
   1301 	if (!ln)
   1302 		goto fail;
   1303 	if (!rt->rt_gateway)
   1304 		goto fail;
   1305 	if (rt->rt_gateway->sa_family != AF_LINK)
   1306 		goto fail;
   1307 	sdl = SDL(rt->rt_gateway);
   1308 
   1309 	olladdr = (sdl->sdl_alen) ? 1 : 0;
   1310 	if (olladdr && lladdr) {
   1311 		if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
   1312 			llchange = 1;
   1313 		else
   1314 			llchange = 0;
   1315 	} else
   1316 		llchange = 0;
   1317 
   1318 	/*
   1319 	 * newentry olladdr  lladdr  llchange	(*=record)
   1320 	 *	0	n	n	--	(1)
   1321 	 *	0	y	n	--	(2)
   1322 	 *	0	n	y	--	(3) * STALE
   1323 	 *	0	y	y	n	(4) *
   1324 	 *	0	y	y	y	(5) * STALE
   1325 	 *	1	--	n	--	(6)   NOSTATE(= PASSIVE)
   1326 	 *	1	--	y	--	(7) * STALE
   1327 	 */
   1328 
   1329 	if (lladdr) {		/*(3-5) and (7)*/
   1330 		/*
   1331 		 * Record source link-layer address
   1332 		 * XXX is it dependent to ifp->if_type?
   1333 		 */
   1334 		sdl->sdl_alen = ifp->if_addrlen;
   1335 		bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
   1336 	}
   1337 
   1338 	if (!is_newentry) {
   1339 		if ((!olladdr && lladdr)		/*(3)*/
   1340 		 || (olladdr && lladdr && llchange)) {	/*(5)*/
   1341 			do_update = 1;
   1342 			newstate = ND6_LLINFO_STALE;
   1343 		} else					/*(1-2,4)*/
   1344 			do_update = 0;
   1345 	} else {
   1346 		do_update = 1;
   1347 		if (!lladdr)				/*(6)*/
   1348 			newstate = ND6_LLINFO_NOSTATE;
   1349 		else					/*(7)*/
   1350 			newstate = ND6_LLINFO_STALE;
   1351 	}
   1352 
   1353 	if (do_update) {
   1354 		/*
   1355 		 * Update the state of the neighbor cache.
   1356 		 */
   1357 		ln->ln_state = newstate;
   1358 
   1359 		if (ln->ln_state == ND6_LLINFO_STALE) {
   1360 			rt->rt_flags &= ~RTF_REJECT;
   1361 			if (ln->ln_hold) {
   1362 				(*ifp->if_output)(ifp, ln->ln_hold,
   1363 						  rt_key(rt), rt);
   1364 				ln->ln_hold = 0;
   1365 			}
   1366 		} else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
   1367 			/* probe right away */
   1368 			ln->ln_expire = time_second;
   1369 		}
   1370 	}
   1371 
   1372 	/*
   1373 	 * ICMP6 type dependent behavior.
   1374 	 *
   1375 	 * NS: clear IsRouter if new entry
   1376 	 * RS: clear IsRouter
   1377 	 * RA: set IsRouter if there's lladdr
   1378 	 * redir: clear IsRouter if new entry
   1379 	 *
   1380 	 * RA case, (1):
   1381 	 * The spec says that we must set IsRouter in the following cases:
   1382 	 * - If lladdr exist, set IsRouter.  This means (1-5).
   1383 	 * - If it is old entry (!newentry), set IsRouter.  This means (7).
   1384 	 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
   1385 	 * A quetion arises for (1) case.  (1) case has no lladdr in the
   1386 	 * neighbor cache, this is similar to (6).
   1387 	 * This case is rare but we figured that we MUST NOT set IsRouter.
   1388 	 *
   1389 	 * newentry olladdr  lladdr  llchange	    NS  RS  RA	redir
   1390 	 *							D R
   1391 	 *	0	n	n	--	(1)	c   ?     s
   1392 	 *	0	y	n	--	(2)	c   s     s
   1393 	 *	0	n	y	--	(3)	c   s     s
   1394 	 *	0	y	y	n	(4)	c   s     s
   1395 	 *	0	y	y	y	(5)	c   s     s
   1396 	 *	1	--	n	--	(6) c	c 	c s
   1397 	 *	1	--	y	--	(7) c	c   s	c s
   1398 	 *
   1399 	 *					(c=clear s=set)
   1400 	 */
   1401 	switch (type & 0xff) {
   1402 	case ND_NEIGHBOR_SOLICIT:
   1403 		/*
   1404 		 * New entry must have is_router flag cleared.
   1405 		 */
   1406 		if (is_newentry)	/*(6-7)*/
   1407 			ln->ln_router = 0;
   1408 		break;
   1409 	case ND_REDIRECT:
   1410 		/*
   1411 		 * If the icmp is a redirect to a better router, always set the
   1412 		 * is_router flag. Otherwise, if the entry is newly created,
   1413 		 * clear the flag. [RFC 2461, sec 8.3]
   1414 		 *
   1415 		 */
   1416 		if (code == ND_REDIRECT_ROUTER)
   1417 			ln->ln_router = 1;
   1418 		else if (is_newentry) /*(6-7)*/
   1419 			ln->ln_router = 0;
   1420 		break;
   1421 	case ND_ROUTER_SOLICIT:
   1422 		/*
   1423 		 * is_router flag must always be cleared.
   1424 		 */
   1425 		ln->ln_router = 0;
   1426 		break;
   1427 	case ND_ROUTER_ADVERT:
   1428 		/*
   1429 		 * Mark an entry with lladdr as a router.
   1430 		 */
   1431 		if ((!is_newentry && (olladdr || lladdr))	/*(2-5)*/
   1432 		 || (is_newentry && lladdr)) {			/*(7)*/
   1433 			ln->ln_router = 1;
   1434 		}
   1435 		break;
   1436 	}
   1437 
   1438 	return rt;
   1439 }
   1440 
   1441 static void
   1442 nd6_slowtimo(ignored_arg)
   1443     void *ignored_arg;
   1444 {
   1445 	int s = splsoftnet();
   1446 	register int i;
   1447 	register struct nd_ifinfo *nd6if;
   1448 
   1449 	timeout(nd6_slowtimo, (caddr_t)0, ND6_SLOWTIMER_INTERVAL * hz);
   1450 	for (i = 1; i < if_index + 1; i++) {
   1451 		nd6if = &nd_ifinfo[i];
   1452 		if (nd6if->basereachable && /* already initialized */
   1453 		    (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
   1454 			/*
   1455 			 * Since reachable time rarely changes by router
   1456 			 * advertisements, we SHOULD insure that a new random
   1457 			 * value gets recomputed at least once every few hours.
   1458 			 * (RFC 2461, 6.3.4)
   1459 			 */
   1460 			nd6if->recalctm = nd6_recalc_reachtm_interval;
   1461 			nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
   1462 		}
   1463 	}
   1464 	splx(s);
   1465 }
   1466 
   1467 #ifdef NEWIP6OUTPUT
   1468 /* for experimental */
   1469 #define senderr(e) { error = (e); goto bad;}
   1470 int
   1471 nd6_output(ifp, m0, dst, rt0)
   1472 	register struct ifnet *ifp;
   1473 	struct mbuf *m0;
   1474 	struct sockaddr_in6 *dst;
   1475 	struct rtentry *rt0;
   1476 {
   1477 	register struct mbuf *m = m0;
   1478 	register struct rtentry *rt = rt0;
   1479 	struct llinfo_nd6 *ln = NULL;
   1480 	int error = 0;
   1481 
   1482 	if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
   1483 		goto sendpkt;
   1484 
   1485 	/*
   1486 	 * XXX: we currently do not make neighbor cache on any interface
   1487 	 * other than Ethernet and FDDI.
   1488 	 */
   1489 	if (ifp->if_type != IFT_ETHER && ifp->if_type != IFT_FDDI)
   1490 		goto sendpkt;
   1491 
   1492 	/*
   1493 	 * next hop determination. This routine is derived from ether_outpout.
   1494 	 */
   1495 	if (rt) {
   1496 		if ((rt->rt_flags & RTF_UP) == 0) {
   1497 #ifdef __FreeBSD__
   1498 			if ((rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL)) !=
   1499 				NULL) {
   1500 #else
   1501 			if ((rt0 = rt = rtalloc1((struct sockaddr *)dst, 1)) !=
   1502 				NULL) {
   1503 #endif
   1504 				rt->rt_refcnt--;
   1505 				if (rt->rt_ifp != ifp)
   1506 					return nd6_output(ifp, m0, dst, rt); /* XXX: loop care? */
   1507 			} else
   1508 				senderr(EHOSTUNREACH);
   1509 		}
   1510 		if (rt->rt_flags & RTF_GATEWAY) {
   1511 			if (rt->rt_gwroute == 0)
   1512 				goto lookup;
   1513 			if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
   1514 				rtfree(rt); rt = rt0;
   1515 #ifdef __FreeBSD__
   1516 			lookup: rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1, 0UL);
   1517 #else
   1518 			lookup: rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1);
   1519 #endif
   1520 				if ((rt = rt->rt_gwroute) == 0)
   1521 					senderr(EHOSTUNREACH);
   1522 #ifdef __bsdi__
   1523 				/* the "G" test below also prevents rt == rt0 */
   1524 				if ((rt->rt_flags & RTF_GATEWAY) ||
   1525 				    (rt->rt_ifp != ifp)) {
   1526 					rt->rt_refcnt--;
   1527 					rt0->rt_gwroute = 0;
   1528 					senderr(EHOSTUNREACH);
   1529 				}
   1530 #endif
   1531 			}
   1532 		}
   1533 		if (rt->rt_flags & RTF_REJECT)
   1534 			senderr(rt == rt0 ? EHOSTDOWN : EHOSTUNREACH);
   1535 	}
   1536 
   1537 	/*
   1538 	 * Address resolution or Neighbor Unreachability Detection
   1539 	 * for the next hop.
   1540 	 * At this point, the destination of the packet must be a unicast
   1541 	 * or an anycast address(i.e. not a multicast).
   1542 	 */
   1543 
   1544 	/* Look up the neighbor cache for the nexthop */
   1545 	if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
   1546 		ln = (struct llinfo_nd6 *)rt->rt_llinfo;
   1547 	else {
   1548 		if ((rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
   1549 			ln = (struct llinfo_nd6 *)rt->rt_llinfo;
   1550 	}
   1551 	if (!ln || !rt) {
   1552 		log(LOG_DEBUG, "nd6_output: can't allocate llinfo for %s "
   1553 		    "(ln=%p, rt=%p)\n",
   1554 		    ip6_sprintf(&dst->sin6_addr), ln, rt);
   1555 		senderr(EIO);	/* XXX: good error? */
   1556 	}
   1557 
   1558 
   1559 	/*
   1560 	 * The first time we send a packet to a neighbor whose entry is
   1561 	 * STALE, we have to change the state to DELAY and a sets a timer to
   1562 	 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
   1563 	 * neighbor unreachability detection on expiration.
   1564 	 * (RFC 2461 7.3.3)
   1565 	 */
   1566 	if (ln->ln_state == ND6_LLINFO_STALE) {
   1567 		ln->ln_asked = 0;
   1568 		ln->ln_state = ND6_LLINFO_DELAY;
   1569 		ln->ln_expire = time_second + nd6_delay;
   1570 	}
   1571 
   1572 	/*
   1573 	 * If the neighbor cache entry has a state other than INCOMPLETE
   1574 	 * (i.e. its link-layer address is already reloved), just
   1575 	 * send the packet.
   1576 	 */
   1577 	if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
   1578 		goto sendpkt;
   1579 
   1580 	/*
   1581 	 * There is a neighbor cache entry, but no ethernet address
   1582 	 * response yet. Replace the held mbuf (if any) with this
   1583 	 * latest one.
   1584 	 *
   1585 	 * XXX Does the code conform to rate-limiting rule?
   1586 	 * (RFC 2461 7.2.2)
   1587 	 */
   1588 	if (ln->ln_state == ND6_LLINFO_WAITDELETE ||
   1589 	    ln->ln_state == ND6_LLINFO_NOSTATE)
   1590 		ln->ln_state = ND6_LLINFO_INCOMPLETE;
   1591 	if (ln->ln_hold)
   1592 		m_freem(ln->ln_hold);
   1593 	ln->ln_hold = m;
   1594 	if (ln->ln_expire) {
   1595 		rt->rt_flags &= ~RTF_REJECT;
   1596 		if (ln->ln_asked < nd6_mmaxtries &&
   1597 		    ln->ln_expire < time_second) {
   1598 			ln->ln_asked++;
   1599 			ln->ln_expire = time_second +
   1600 				nd_ifinfo[ifp->if_index].retrans / 1000;
   1601 			nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
   1602 		}
   1603 	}
   1604 	return(0);
   1605 
   1606   sendpkt:
   1607 	return((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
   1608 
   1609   bad:
   1610 	if (m)
   1611 		m_freem(m);
   1612 	return (error);
   1613 }
   1614 #undef senderr
   1615 
   1616 int
   1617 nd6_storelladdr(ifp, rt, m, dst, desten)
   1618 	struct ifnet *ifp;
   1619 	struct rtentry *rt;
   1620 	struct mbuf *m;
   1621 	struct sockaddr *dst;
   1622 	u_char *desten;
   1623 {
   1624 	struct sockaddr_dl *sdl;
   1625 
   1626 	if (m->m_flags & M_MCAST) {
   1627 		switch (ifp->if_type) {
   1628 		case IFT_ETHER:
   1629 		case IFT_FDDI:
   1630 			ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
   1631 						 desten);
   1632 			return(1);
   1633 			break;
   1634 		default:
   1635 			return(0);
   1636 		}
   1637 	}
   1638 
   1639 	if (rt == NULL ||
   1640 	    rt->rt_gateway->sa_family != AF_LINK) {
   1641 		printf("nd6_storelladdr: something odd happens\n");
   1642 		return(0);
   1643 	}
   1644 	sdl = SDL(rt->rt_gateway);
   1645 	if (sdl->sdl_alen != 0)
   1646 		bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
   1647 
   1648 	return(1);
   1649 }
   1650 #endif /* NEWIP6OUTPUT */
   1651