xref: /freebsd/sys/netinet6/nd6_nbr.c (revision a812392203d7c4c3f0db9d8a0f3391374c49c71f)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_mpath.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/lock.h>
44 #include <sys/rwlock.h>
45 #include <sys/mbuf.h>
46 #include <sys/socket.h>
47 #include <sys/sockio.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/errno.h>
51 #include <sys/syslog.h>
52 #include <sys/queue.h>
53 #include <sys/callout.h>
54 #include <sys/refcount.h>
55 
56 #include <net/if.h>
57 #include <net/if_types.h>
58 #include <net/if_dl.h>
59 #include <net/if_var.h>
60 #include <net/route.h>
61 #ifdef RADIX_MPATH
62 #include <net/radix_mpath.h>
63 #endif
64 
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <net/if_llatbl.h>
68 #define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
69 #include <netinet6/in6_var.h>
70 #include <netinet6/in6_ifattach.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #include <netinet6/nd6.h>
75 #include <netinet/icmp6.h>
76 #include <netinet/ip_carp.h>
77 #include <netinet6/send.h>
78 
79 #define SDL(s) ((struct sockaddr_dl *)s)
80 
81 struct dadq;
82 static struct dadq *nd6_dad_find(struct ifaddr *);
83 static void nd6_dad_add(struct dadq *dp);
84 static void nd6_dad_del(struct dadq *dp);
85 static void nd6_dad_rele(struct dadq *);
86 static void nd6_dad_starttimer(struct dadq *, int);
87 static void nd6_dad_stoptimer(struct dadq *);
88 static void nd6_dad_timer(struct dadq *);
89 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
90 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
91 static void nd6_dad_ns_input(struct ifaddr *);
92 static void nd6_dad_na_input(struct ifaddr *);
93 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
94     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
95 
96 static VNET_DEFINE(int, dad_ignore_ns) = 0;	/* ignore NS in DAD
97 						   - specwise incorrect */
98 static VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to
99 						   transmit DAD packet */
100 #define	V_dad_ignore_ns			VNET(dad_ignore_ns)
101 #define	V_dad_maxtry			VNET(dad_maxtry)
102 
103 /*
104  * Input a Neighbor Solicitation Message.
105  *
106  * Based on RFC 2461
107  * Based on RFC 2462 (duplicate address detection)
108  */
109 void
110 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
111 {
112 	struct ifnet *ifp = m->m_pkthdr.rcvif;
113 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
114 	struct nd_neighbor_solicit *nd_ns;
115 	struct in6_addr saddr6 = ip6->ip6_src;
116 	struct in6_addr daddr6 = ip6->ip6_dst;
117 	struct in6_addr taddr6;
118 	struct in6_addr myaddr6;
119 	char *lladdr = NULL;
120 	struct ifaddr *ifa = NULL;
121 	int lladdrlen = 0;
122 	int anycast = 0, proxy = 0, tentative = 0;
123 	int tlladdr;
124 	int rflag;
125 	union nd_opts ndopts;
126 	struct sockaddr_dl proxydl;
127 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
128 
129 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
130 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
131 		rflag = 0;
132 #ifndef PULLDOWN_TEST
133 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
134 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
135 #else
136 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
137 	if (nd_ns == NULL) {
138 		ICMP6STAT_INC(icp6s_tooshort);
139 		return;
140 	}
141 #endif
142 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
143 	taddr6 = nd_ns->nd_ns_target;
144 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
145 		goto bad;
146 
147 	if (ip6->ip6_hlim != 255) {
148 		nd6log((LOG_ERR,
149 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
150 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
151 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
152 		goto bad;
153 	}
154 
155 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
156 		/* dst has to be a solicited node multicast address. */
157 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
158 		    /* don't check ifindex portion */
159 		    daddr6.s6_addr32[1] == 0 &&
160 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
161 		    daddr6.s6_addr8[12] == 0xff) {
162 			; /* good */
163 		} else {
164 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
165 			    "(wrong ip6 dst)\n"));
166 			goto bad;
167 		}
168 	} else if (!V_nd6_onlink_ns_rfc4861) {
169 		struct sockaddr_in6 src_sa6;
170 
171 		/*
172 		 * According to recent IETF discussions, it is not a good idea
173 		 * to accept a NS from an address which would not be deemed
174 		 * to be a neighbor otherwise.  This point is expected to be
175 		 * clarified in future revisions of the specification.
176 		 */
177 		bzero(&src_sa6, sizeof(src_sa6));
178 		src_sa6.sin6_family = AF_INET6;
179 		src_sa6.sin6_len = sizeof(src_sa6);
180 		src_sa6.sin6_addr = saddr6;
181 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
182 			nd6log((LOG_INFO, "nd6_ns_input: "
183 				"NS packet from non-neighbor\n"));
184 			goto bad;
185 		}
186 	}
187 
188 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
189 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
190 		goto bad;
191 	}
192 
193 	icmp6len -= sizeof(*nd_ns);
194 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
195 	if (nd6_options(&ndopts) < 0) {
196 		nd6log((LOG_INFO,
197 		    "nd6_ns_input: invalid ND option, ignored\n"));
198 		/* nd6_options have incremented stats */
199 		goto freeit;
200 	}
201 
202 	if (ndopts.nd_opts_src_lladdr) {
203 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
204 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
205 	}
206 
207 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
208 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
209 		    "(link-layer address option)\n"));
210 		goto bad;
211 	}
212 
213 	/*
214 	 * Attaching target link-layer address to the NA?
215 	 * (RFC 2461 7.2.4)
216 	 *
217 	 * NS IP dst is unicast/anycast			MUST NOT add
218 	 * NS IP dst is solicited-node multicast	MUST add
219 	 *
220 	 * In implementation, we add target link-layer address by default.
221 	 * We do not add one in MUST NOT cases.
222 	 */
223 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
224 		tlladdr = 0;
225 	else
226 		tlladdr = 1;
227 
228 	/*
229 	 * Target address (taddr6) must be either:
230 	 * (1) Valid unicast/anycast address for my receiving interface,
231 	 * (2) Unicast address for which I'm offering proxy service, or
232 	 * (3) "tentative" address on which DAD is being performed.
233 	 */
234 	/* (1) and (3) check. */
235 	if (ifp->if_carp)
236 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
237 	else
238 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
239 
240 	/* (2) check. */
241 	if (ifa == NULL) {
242 		struct route_in6 ro;
243 		int need_proxy;
244 
245 		bzero(&ro, sizeof(ro));
246 		ro.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
247 		ro.ro_dst.sin6_family = AF_INET6;
248 		ro.ro_dst.sin6_addr = taddr6;
249 
250 		/* Always use the default FIB. */
251 #ifdef RADIX_MPATH
252 		rtalloc_mpath_fib((struct route *)&ro, ntohl(taddr6.s6_addr32[3]),
253 		    RT_DEFAULT_FIB);
254 #else
255 		in6_rtalloc(&ro, RT_DEFAULT_FIB);
256 #endif
257 		need_proxy = (ro.ro_rt &&
258 		    (ro.ro_rt->rt_flags & RTF_ANNOUNCE) != 0 &&
259 		    ro.ro_rt->rt_gateway->sa_family == AF_LINK);
260 		if (ro.ro_rt != NULL) {
261 			if (need_proxy)
262 				proxydl = *SDL(ro.ro_rt->rt_gateway);
263 			RTFREE(ro.ro_rt);
264 		}
265 		if (need_proxy) {
266 			/*
267 			 * proxy NDP for single entry
268 			 */
269 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
270 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
271 			if (ifa)
272 				proxy = 1;
273 		}
274 	}
275 	if (ifa == NULL) {
276 		/*
277 		 * We've got an NS packet, and we don't have that adddress
278 		 * assigned for us.  We MUST silently ignore it.
279 		 * See RFC2461 7.2.3.
280 		 */
281 		goto freeit;
282 	}
283 	myaddr6 = *IFA_IN6(ifa);
284 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
285 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
286 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
287 		goto freeit;
288 
289 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
290 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
291 		    "(if %d, NS packet %d)\n",
292 		    ip6_sprintf(ip6bufs, &taddr6),
293 		    ifp->if_addrlen, lladdrlen - 2));
294 		goto bad;
295 	}
296 
297 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
298 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
299 		    ip6_sprintf(ip6bufs, &saddr6)));
300 		goto freeit;
301 	}
302 
303 	/*
304 	 * We have neighbor solicitation packet, with target address equals to
305 	 * one of my tentative address.
306 	 *
307 	 * src addr	how to process?
308 	 * ---		---
309 	 * multicast	of course, invalid (rejected in ip6_input)
310 	 * unicast	somebody is doing address resolution -> ignore
311 	 * unspec	dup address detection
312 	 *
313 	 * The processing is defined in RFC 2462.
314 	 */
315 	if (tentative) {
316 		/*
317 		 * If source address is unspecified address, it is for
318 		 * duplicate address detection.
319 		 *
320 		 * If not, the packet is for addess resolution;
321 		 * silently ignore it.
322 		 */
323 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
324 			nd6_dad_ns_input(ifa);
325 
326 		goto freeit;
327 	}
328 
329 	/*
330 	 * If the source address is unspecified address, entries must not
331 	 * be created or updated.
332 	 * It looks that sender is performing DAD.  Output NA toward
333 	 * all-node multicast address, to tell the sender that I'm using
334 	 * the address.
335 	 * S bit ("solicited") must be zero.
336 	 */
337 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
338 		struct in6_addr in6_all;
339 
340 		in6_all = in6addr_linklocal_allnodes;
341 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
342 			goto bad;
343 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
344 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
345 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
346 		    M_GETFIB(m));
347 		goto freeit;
348 	}
349 
350 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
351 	    ND_NEIGHBOR_SOLICIT, 0);
352 
353 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
354 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
355 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
356 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
357  freeit:
358 	if (ifa != NULL)
359 		ifa_free(ifa);
360 	m_freem(m);
361 	return;
362 
363  bad:
364 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
365 		ip6_sprintf(ip6bufs, &saddr6)));
366 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
367 		ip6_sprintf(ip6bufs, &daddr6)));
368 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
369 		ip6_sprintf(ip6bufs, &taddr6)));
370 	ICMP6STAT_INC(icp6s_badns);
371 	if (ifa != NULL)
372 		ifa_free(ifa);
373 	m_freem(m);
374 }
375 
376 /*
377  * Output a Neighbor Solicitation Message. Caller specifies:
378  *	- ICMP6 header source IP6 address
379  *	- ND6 header target IP6 address
380  *	- ND6 header source datalink address
381  *
382  * Based on RFC 2461
383  * Based on RFC 2462 (duplicate address detection)
384  *
385  *   ln - for source address determination
386  *  dad - duplicate address detection
387  */
388 void
389 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
390     const struct in6_addr *taddr6, struct llentry *ln, int dad)
391 {
392 	struct mbuf *m;
393 	struct m_tag *mtag;
394 	struct ip6_hdr *ip6;
395 	struct nd_neighbor_solicit *nd_ns;
396 	struct ip6_moptions im6o;
397 	int icmp6len;
398 	int maxlen;
399 	caddr_t mac;
400 	struct route_in6 ro;
401 
402 	if (IN6_IS_ADDR_MULTICAST(taddr6))
403 		return;
404 
405 	/* estimate the size of message */
406 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
407 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
408 	if (max_linkhdr + maxlen >= MCLBYTES) {
409 #ifdef DIAGNOSTIC
410 		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
411 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
412 #endif
413 		return;
414 	}
415 
416 	if (max_linkhdr + maxlen > MHLEN)
417 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
418 	else
419 		m = m_gethdr(M_NOWAIT, MT_DATA);
420 	if (m == NULL)
421 		return;
422 
423 	bzero(&ro, sizeof(ro));
424 
425 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
426 		m->m_flags |= M_MCAST;
427 		im6o.im6o_multicast_ifp = ifp;
428 		im6o.im6o_multicast_hlim = 255;
429 		im6o.im6o_multicast_loop = 0;
430 	}
431 
432 	icmp6len = sizeof(*nd_ns);
433 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
434 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
435 
436 	/* fill neighbor solicitation packet */
437 	ip6 = mtod(m, struct ip6_hdr *);
438 	ip6->ip6_flow = 0;
439 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
440 	ip6->ip6_vfc |= IPV6_VERSION;
441 	/* ip6->ip6_plen will be set later */
442 	ip6->ip6_nxt = IPPROTO_ICMPV6;
443 	ip6->ip6_hlim = 255;
444 	if (daddr6)
445 		ip6->ip6_dst = *daddr6;
446 	else {
447 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
448 		ip6->ip6_dst.s6_addr16[1] = 0;
449 		ip6->ip6_dst.s6_addr32[1] = 0;
450 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
451 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
452 		ip6->ip6_dst.s6_addr8[12] = 0xff;
453 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
454 			goto bad;
455 	}
456 	if (!dad) {
457 		struct ifaddr *ifa;
458 
459 		/*
460 		 * RFC2461 7.2.2:
461 		 * "If the source address of the packet prompting the
462 		 * solicitation is the same as one of the addresses assigned
463 		 * to the outgoing interface, that address SHOULD be placed
464 		 * in the IP Source Address of the outgoing solicitation.
465 		 * Otherwise, any one of the addresses assigned to the
466 		 * interface should be used."
467 		 *
468 		 * We use the source address for the prompting packet
469 		 * (saddr6), if:
470 		 * - saddr6 is given from the caller (by giving "ln"), and
471 		 * - saddr6 belongs to the outgoing interface.
472 		 * Otherwise, we perform the source address selection as usual.
473 		 */
474 		struct in6_addr *hsrc;
475 
476 		hsrc = NULL;
477 		if (ln != NULL) {
478 			LLE_RLOCK(ln);
479 			if (ln->la_hold != NULL) {
480 				struct ip6_hdr *hip6;		/* hold ip6 */
481 
482 				/*
483 				 * assuming every packet in la_hold has the same IP
484 				 * header
485 				 */
486 				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
487 				/* XXX pullup? */
488 				if (sizeof(*hip6) < ln->la_hold->m_len) {
489 					ip6->ip6_src = hip6->ip6_src;
490 					hsrc = &hip6->ip6_src;
491 				}
492 			}
493 			LLE_RUNLOCK(ln);
494 		}
495 		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
496 		    hsrc)) != NULL) {
497 			/* ip6_src set already. */
498 			ifa_free(ifa);
499 		} else {
500 			int error;
501 			struct sockaddr_in6 dst_sa;
502 			struct in6_addr src_in;
503 			struct ifnet *oifp;
504 
505 			bzero(&dst_sa, sizeof(dst_sa));
506 			dst_sa.sin6_family = AF_INET6;
507 			dst_sa.sin6_len = sizeof(dst_sa);
508 			dst_sa.sin6_addr = ip6->ip6_dst;
509 
510 			oifp = ifp;
511 			error = in6_selectsrc(&dst_sa, NULL,
512 			    NULL, &ro, NULL, &oifp, &src_in);
513 			if (error) {
514 				char ip6buf[INET6_ADDRSTRLEN];
515 				nd6log((LOG_DEBUG,
516 				    "nd6_ns_output: source can't be "
517 				    "determined: dst=%s, error=%d\n",
518 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
519 				    error));
520 				goto bad;
521 			}
522 			ip6->ip6_src = src_in;
523 		}
524 	} else {
525 		/*
526 		 * Source address for DAD packet must always be IPv6
527 		 * unspecified address. (0::0)
528 		 * We actually don't have to 0-clear the address (we did it
529 		 * above), but we do so here explicitly to make the intention
530 		 * clearer.
531 		 */
532 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
533 	}
534 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
535 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
536 	nd_ns->nd_ns_code = 0;
537 	nd_ns->nd_ns_reserved = 0;
538 	nd_ns->nd_ns_target = *taddr6;
539 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
540 
541 	/*
542 	 * Add source link-layer address option.
543 	 *
544 	 *				spec		implementation
545 	 *				---		---
546 	 * DAD packet			MUST NOT	do not add the option
547 	 * there's no link layer address:
548 	 *				impossible	do not add the option
549 	 * there's link layer address:
550 	 *	Multicast NS		MUST add one	add the option
551 	 *	Unicast NS		SHOULD add one	add the option
552 	 */
553 	if (!dad && (mac = nd6_ifptomac(ifp))) {
554 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
555 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
556 		/* 8 byte alignments... */
557 		optlen = (optlen + 7) & ~7;
558 
559 		m->m_pkthdr.len += optlen;
560 		m->m_len += optlen;
561 		icmp6len += optlen;
562 		bzero((caddr_t)nd_opt, optlen);
563 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
564 		nd_opt->nd_opt_len = optlen >> 3;
565 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
566 	}
567 
568 	ip6->ip6_plen = htons((u_short)icmp6len);
569 	nd_ns->nd_ns_cksum = 0;
570 	nd_ns->nd_ns_cksum =
571 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
572 
573 	if (send_sendso_input_hook != NULL) {
574 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
575 			sizeof(unsigned short), M_NOWAIT);
576 		if (mtag == NULL)
577 			goto bad;
578 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
579 		m_tag_prepend(m, mtag);
580 	}
581 
582 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
583 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
584 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
585 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
586 
587 	/* We don't cache this route. */
588 	RO_RTFREE(&ro);
589 
590 	return;
591 
592   bad:
593 	if (ro.ro_rt) {
594 		RTFREE(ro.ro_rt);
595 	}
596 	m_freem(m);
597 	return;
598 }
599 
600 /*
601  * Neighbor advertisement input handling.
602  *
603  * Based on RFC 2461
604  * Based on RFC 2462 (duplicate address detection)
605  *
606  * the following items are not implemented yet:
607  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
608  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
609  */
610 void
611 nd6_na_input(struct mbuf *m, int off, int icmp6len)
612 {
613 	struct ifnet *ifp = m->m_pkthdr.rcvif;
614 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
615 	struct nd_neighbor_advert *nd_na;
616 	struct in6_addr daddr6 = ip6->ip6_dst;
617 	struct in6_addr taddr6;
618 	int flags;
619 	int is_router;
620 	int is_solicited;
621 	int is_override;
622 	char *lladdr = NULL;
623 	int lladdrlen = 0;
624 	int checklink = 0;
625 	struct ifaddr *ifa;
626 	struct llentry *ln = NULL;
627 	union nd_opts ndopts;
628 	struct mbuf *chain = NULL;
629 	struct m_tag *mtag;
630 	struct sockaddr_in6 sin6;
631 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
632 
633 	if (ip6->ip6_hlim != 255) {
634 		nd6log((LOG_ERR,
635 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
636 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
637 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
638 		goto bad;
639 	}
640 
641 #ifndef PULLDOWN_TEST
642 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
643 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
644 #else
645 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
646 	if (nd_na == NULL) {
647 		ICMP6STAT_INC(icp6s_tooshort);
648 		return;
649 	}
650 #endif
651 
652 	flags = nd_na->nd_na_flags_reserved;
653 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
654 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
655 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
656 
657 	taddr6 = nd_na->nd_na_target;
658 	if (in6_setscope(&taddr6, ifp, NULL))
659 		goto bad;	/* XXX: impossible */
660 
661 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
662 		nd6log((LOG_ERR,
663 		    "nd6_na_input: invalid target address %s\n",
664 		    ip6_sprintf(ip6bufs, &taddr6)));
665 		goto bad;
666 	}
667 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
668 		if (is_solicited) {
669 			nd6log((LOG_ERR,
670 			    "nd6_na_input: a solicited adv is multicasted\n"));
671 			goto bad;
672 		}
673 
674 	icmp6len -= sizeof(*nd_na);
675 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
676 	if (nd6_options(&ndopts) < 0) {
677 		nd6log((LOG_INFO,
678 		    "nd6_na_input: invalid ND option, ignored\n"));
679 		/* nd6_options have incremented stats */
680 		goto freeit;
681 	}
682 
683 	if (ndopts.nd_opts_tgt_lladdr) {
684 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
685 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
686 	}
687 
688 	/*
689 	 * This effectively disables the DAD check on a non-master CARP
690 	 * address.
691 	 */
692 	if (ifp->if_carp)
693 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
694 	else
695 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
696 
697 	/*
698 	 * Target address matches one of my interface address.
699 	 *
700 	 * If my address is tentative, this means that there's somebody
701 	 * already using the same address as mine.  This indicates DAD failure.
702 	 * This is defined in RFC 2462.
703 	 *
704 	 * Otherwise, process as defined in RFC 2461.
705 	 */
706 	if (ifa
707 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
708 		ifa_free(ifa);
709 		nd6_dad_na_input(ifa);
710 		goto freeit;
711 	}
712 
713 	/* Just for safety, maybe unnecessary. */
714 	if (ifa) {
715 		ifa_free(ifa);
716 		log(LOG_ERR,
717 		    "nd6_na_input: duplicate IP6 address %s\n",
718 		    ip6_sprintf(ip6bufs, &taddr6));
719 		goto freeit;
720 	}
721 
722 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
723 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
724 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
725 		    ifp->if_addrlen, lladdrlen - 2));
726 		goto bad;
727 	}
728 
729 	/*
730 	 * If no neighbor cache entry is found, NA SHOULD silently be
731 	 * discarded.
732 	 */
733 	IF_AFDATA_RLOCK(ifp);
734 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
735 	IF_AFDATA_RUNLOCK(ifp);
736 	if (ln == NULL) {
737 		goto freeit;
738 	}
739 
740 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
741 		/*
742 		 * If the link-layer has address, and no lladdr option came,
743 		 * discard the packet.
744 		 */
745 		if (ifp->if_addrlen && lladdr == NULL) {
746 			goto freeit;
747 		}
748 
749 		/*
750 		 * Record link-layer address, and update the state.
751 		 */
752 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
753 		ln->la_flags |= LLE_VALID;
754 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
755 		if (is_solicited) {
756 			ln->ln_state = ND6_LLINFO_REACHABLE;
757 			ln->ln_byhint = 0;
758 			if (!ND6_LLINFO_PERMANENT(ln)) {
759 				nd6_llinfo_settimer_locked(ln,
760 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
761 			}
762 		} else {
763 			ln->ln_state = ND6_LLINFO_STALE;
764 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
765 		}
766 		if ((ln->ln_router = is_router) != 0) {
767 			/*
768 			 * This means a router's state has changed from
769 			 * non-reachable to probably reachable, and might
770 			 * affect the status of associated prefixes..
771 			 */
772 			checklink = 1;
773 		}
774 	} else {
775 		int llchange;
776 
777 		/*
778 		 * Check if the link-layer address has changed or not.
779 		 */
780 		if (lladdr == NULL)
781 			llchange = 0;
782 		else {
783 			if (ln->la_flags & LLE_VALID) {
784 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
785 					llchange = 1;
786 				else
787 					llchange = 0;
788 			} else
789 				llchange = 1;
790 		}
791 
792 		/*
793 		 * This is VERY complex.  Look at it with care.
794 		 *
795 		 * override solicit lladdr llchange	action
796 		 *					(L: record lladdr)
797 		 *
798 		 *	0	0	n	--	(2c)
799 		 *	0	0	y	n	(2b) L
800 		 *	0	0	y	y	(1)    REACHABLE->STALE
801 		 *	0	1	n	--	(2c)   *->REACHABLE
802 		 *	0	1	y	n	(2b) L *->REACHABLE
803 		 *	0	1	y	y	(1)    REACHABLE->STALE
804 		 *	1	0	n	--	(2a)
805 		 *	1	0	y	n	(2a) L
806 		 *	1	0	y	y	(2a) L *->STALE
807 		 *	1	1	n	--	(2a)   *->REACHABLE
808 		 *	1	1	y	n	(2a) L *->REACHABLE
809 		 *	1	1	y	y	(2a) L *->REACHABLE
810 		 */
811 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
812 			/*
813 			 * If state is REACHABLE, make it STALE.
814 			 * no other updates should be done.
815 			 */
816 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
817 				ln->ln_state = ND6_LLINFO_STALE;
818 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
819 			}
820 			goto freeit;
821 		} else if (is_override				   /* (2a) */
822 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
823 			|| lladdr == NULL) {			   /* (2c) */
824 			/*
825 			 * Update link-local address, if any.
826 			 */
827 			if (lladdr != NULL) {
828 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
829 				ln->la_flags |= LLE_VALID;
830 				EVENTHANDLER_INVOKE(lle_event, ln,
831 				    LLENTRY_RESOLVED);
832 			}
833 
834 			/*
835 			 * If solicited, make the state REACHABLE.
836 			 * If not solicited and the link-layer address was
837 			 * changed, make it STALE.
838 			 */
839 			if (is_solicited) {
840 				ln->ln_state = ND6_LLINFO_REACHABLE;
841 				ln->ln_byhint = 0;
842 				if (!ND6_LLINFO_PERMANENT(ln)) {
843 					nd6_llinfo_settimer_locked(ln,
844 					    (long)ND_IFINFO(ifp)->reachable * hz);
845 				}
846 			} else {
847 				if (lladdr != NULL && llchange) {
848 					ln->ln_state = ND6_LLINFO_STALE;
849 					nd6_llinfo_settimer_locked(ln,
850 					    (long)V_nd6_gctimer * hz);
851 				}
852 			}
853 		}
854 
855 		if (ln->ln_router && !is_router) {
856 			/*
857 			 * The peer dropped the router flag.
858 			 * Remove the sender from the Default Router List and
859 			 * update the Destination Cache entries.
860 			 */
861 			struct nd_defrouter *dr;
862 			struct in6_addr *in6;
863 
864 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
865 
866 			/*
867 			 * Lock to protect the default router list.
868 			 * XXX: this might be unnecessary, since this function
869 			 * is only called under the network software interrupt
870 			 * context.  However, we keep it just for safety.
871 			 */
872 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
873 			if (dr)
874 				defrtrlist_del(dr);
875 			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
876 			    ND6_IFF_ACCEPT_RTADV) {
877 				/*
878 				 * Even if the neighbor is not in the default
879 				 * router list, the neighbor may be used
880 				 * as a next hop for some destinations
881 				 * (e.g. redirect case). So we must
882 				 * call rt6_flush explicitly.
883 				 */
884 				rt6_flush(&ip6->ip6_src, ifp);
885 			}
886 		}
887 		ln->ln_router = is_router;
888 	}
889         /* XXX - QL
890 	 *  Does this matter?
891 	 *  rt->rt_flags &= ~RTF_REJECT;
892 	 */
893 	ln->la_asked = 0;
894 	if (ln->la_hold) {
895 		struct mbuf *m_hold, *m_hold_next;
896 
897 		/*
898 		 * reset the la_hold in advance, to explicitly
899 		 * prevent a la_hold lookup in nd6_output()
900 		 * (wouldn't happen, though...)
901 		 */
902 		for (m_hold = ln->la_hold, ln->la_hold = NULL;
903 		    m_hold; m_hold = m_hold_next) {
904 			m_hold_next = m_hold->m_nextpkt;
905 			m_hold->m_nextpkt = NULL;
906 			/*
907 			 * we assume ifp is not a loopback here, so just set
908 			 * the 2nd argument as the 1st one.
909 			 */
910 
911 			if (send_sendso_input_hook != NULL) {
912 				mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
913 				    sizeof(unsigned short), M_NOWAIT);
914 				if (mtag == NULL)
915 					goto bad;
916 				m_tag_prepend(m, mtag);
917 			}
918 
919 			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
920 		}
921 	}
922  freeit:
923 	if (ln != NULL) {
924 		if (chain)
925 			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
926 		LLE_WUNLOCK(ln);
927 
928 		if (chain)
929 			nd6_output_flush(ifp, ifp, chain, &sin6);
930 	}
931 	if (checklink)
932 		pfxlist_onlink_check();
933 
934 	m_freem(m);
935 	return;
936 
937  bad:
938 	if (ln != NULL)
939 		LLE_WUNLOCK(ln);
940 
941 	ICMP6STAT_INC(icp6s_badna);
942 	m_freem(m);
943 }
944 
945 /*
946  * Neighbor advertisement output handling.
947  *
948  * Based on RFC 2461
949  *
950  * the following items are not implemented yet:
951  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
952  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
953  *
954  * tlladdr - 1 if include target link-layer address
955  * sdl0 - sockaddr_dl (= proxy NA) or NULL
956  */
957 static void
958 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
959     const struct in6_addr *taddr6, u_long flags, int tlladdr,
960     struct sockaddr *sdl0, u_int fibnum)
961 {
962 	struct mbuf *m;
963 	struct m_tag *mtag;
964 	struct ifnet *oifp;
965 	struct ip6_hdr *ip6;
966 	struct nd_neighbor_advert *nd_na;
967 	struct ip6_moptions im6o;
968 	struct in6_addr src, daddr6;
969 	struct sockaddr_in6 dst_sa;
970 	int icmp6len, maxlen, error;
971 	caddr_t mac = NULL;
972 	struct route_in6 ro;
973 
974 	bzero(&ro, sizeof(ro));
975 
976 	daddr6 = *daddr6_0;	/* make a local copy for modification */
977 
978 	/* estimate the size of message */
979 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
980 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
981 	if (max_linkhdr + maxlen >= MCLBYTES) {
982 #ifdef DIAGNOSTIC
983 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
984 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
985 #endif
986 		return;
987 	}
988 
989 	if (max_linkhdr + maxlen > MHLEN)
990 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
991 	else
992 		m = m_gethdr(M_NOWAIT, MT_DATA);
993 	if (m == NULL)
994 		return;
995 	M_SETFIB(m, fibnum);
996 
997 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
998 		m->m_flags |= M_MCAST;
999 		im6o.im6o_multicast_ifp = ifp;
1000 		im6o.im6o_multicast_hlim = 255;
1001 		im6o.im6o_multicast_loop = 0;
1002 	}
1003 
1004 	icmp6len = sizeof(*nd_na);
1005 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1006 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
1007 
1008 	/* fill neighbor advertisement packet */
1009 	ip6 = mtod(m, struct ip6_hdr *);
1010 	ip6->ip6_flow = 0;
1011 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1012 	ip6->ip6_vfc |= IPV6_VERSION;
1013 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1014 	ip6->ip6_hlim = 255;
1015 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1016 		/* reply to DAD */
1017 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1018 		daddr6.s6_addr16[1] = 0;
1019 		daddr6.s6_addr32[1] = 0;
1020 		daddr6.s6_addr32[2] = 0;
1021 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1022 		if (in6_setscope(&daddr6, ifp, NULL))
1023 			goto bad;
1024 
1025 		flags &= ~ND_NA_FLAG_SOLICITED;
1026 	}
1027 	ip6->ip6_dst = daddr6;
1028 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1029 	dst_sa.sin6_family = AF_INET6;
1030 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1031 	dst_sa.sin6_addr = daddr6;
1032 
1033 	/*
1034 	 * Select a source whose scope is the same as that of the dest.
1035 	 */
1036 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1037 	oifp = ifp;
1038 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1039 	if (error) {
1040 		char ip6buf[INET6_ADDRSTRLEN];
1041 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1042 		    "determined: dst=%s, error=%d\n",
1043 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1044 		goto bad;
1045 	}
1046 	ip6->ip6_src = src;
1047 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1048 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1049 	nd_na->nd_na_code = 0;
1050 	nd_na->nd_na_target = *taddr6;
1051 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1052 
1053 	/*
1054 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1055 	 * see nd6_ns_input() for details.
1056 	 * Basically, if NS packet is sent to unicast/anycast addr,
1057 	 * target lladdr option SHOULD NOT be included.
1058 	 */
1059 	if (tlladdr) {
1060 		/*
1061 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1062 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1063 		 * my address) use lladdr configured for the interface.
1064 		 */
1065 		if (sdl0 == NULL) {
1066 			if (ifp->if_carp)
1067 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1068 			if (mac == NULL)
1069 				mac = nd6_ifptomac(ifp);
1070 		} else if (sdl0->sa_family == AF_LINK) {
1071 			struct sockaddr_dl *sdl;
1072 			sdl = (struct sockaddr_dl *)sdl0;
1073 			if (sdl->sdl_alen == ifp->if_addrlen)
1074 				mac = LLADDR(sdl);
1075 		}
1076 	}
1077 	if (tlladdr && mac) {
1078 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1079 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1080 
1081 		/* roundup to 8 bytes alignment! */
1082 		optlen = (optlen + 7) & ~7;
1083 
1084 		m->m_pkthdr.len += optlen;
1085 		m->m_len += optlen;
1086 		icmp6len += optlen;
1087 		bzero((caddr_t)nd_opt, optlen);
1088 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1089 		nd_opt->nd_opt_len = optlen >> 3;
1090 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1091 	} else
1092 		flags &= ~ND_NA_FLAG_OVERRIDE;
1093 
1094 	ip6->ip6_plen = htons((u_short)icmp6len);
1095 	nd_na->nd_na_flags_reserved = flags;
1096 	nd_na->nd_na_cksum = 0;
1097 	nd_na->nd_na_cksum =
1098 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1099 
1100 	if (send_sendso_input_hook != NULL) {
1101 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1102 		    sizeof(unsigned short), M_NOWAIT);
1103 		if (mtag == NULL)
1104 			goto bad;
1105 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1106 		m_tag_prepend(m, mtag);
1107 	}
1108 
1109 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1110 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1111 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1112 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1113 
1114 	/* We don't cache this route. */
1115 	RO_RTFREE(&ro);
1116 
1117 	return;
1118 
1119   bad:
1120 	if (ro.ro_rt) {
1121 		RTFREE(ro.ro_rt);
1122 	}
1123 	m_freem(m);
1124 	return;
1125 }
1126 
1127 #ifndef BURN_BRIDGES
1128 void
1129 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1130     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1131     struct sockaddr *sdl0)
1132 {
1133 
1134 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1135 	    RT_DEFAULT_FIB);
1136 }
1137 #endif
1138 
1139 caddr_t
1140 nd6_ifptomac(struct ifnet *ifp)
1141 {
1142 	switch (ifp->if_type) {
1143 	case IFT_ARCNET:
1144 	case IFT_ETHER:
1145 	case IFT_FDDI:
1146 	case IFT_IEEE1394:
1147 #ifdef IFT_L2VLAN
1148 	case IFT_L2VLAN:
1149 #endif
1150 #ifdef IFT_IEEE80211
1151 	case IFT_IEEE80211:
1152 #endif
1153 	case IFT_INFINIBAND:
1154 	case IFT_BRIDGE:
1155 	case IFT_ISO88025:
1156 		return IF_LLADDR(ifp);
1157 	default:
1158 		return NULL;
1159 	}
1160 }
1161 
1162 struct dadq {
1163 	TAILQ_ENTRY(dadq) dad_list;
1164 	struct ifaddr *dad_ifa;
1165 	int dad_count;		/* max NS to send */
1166 	int dad_ns_tcount;	/* # of trials to send NS */
1167 	int dad_ns_ocount;	/* NS sent so far */
1168 	int dad_ns_icount;
1169 	int dad_na_icount;
1170 	struct callout dad_timer_ch;
1171 	struct vnet *dad_vnet;
1172 	u_int dad_refcnt;
1173 };
1174 
1175 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1176 static VNET_DEFINE(struct rwlock, dad_rwlock);
1177 #define	V_dadq			VNET(dadq)
1178 #define	V_dad_rwlock		VNET(dad_rwlock)
1179 
1180 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1181 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1182 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1183 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1184 
1185 static void
1186 nd6_dad_add(struct dadq *dp)
1187 {
1188 
1189 	DADQ_WLOCK();
1190 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1191 	DADQ_WUNLOCK();
1192 }
1193 
1194 static void
1195 nd6_dad_del(struct dadq *dp)
1196 {
1197 
1198 	DADQ_WLOCK();
1199 	TAILQ_REMOVE(&V_dadq, dp, dad_list);
1200 	DADQ_WUNLOCK();
1201 	nd6_dad_rele(dp);
1202 }
1203 
1204 static struct dadq *
1205 nd6_dad_find(struct ifaddr *ifa)
1206 {
1207 	struct dadq *dp;
1208 
1209 	DADQ_RLOCK();
1210 	TAILQ_FOREACH(dp, &V_dadq, dad_list)
1211 		if (dp->dad_ifa == ifa) {
1212 			refcount_acquire(&dp->dad_refcnt);
1213 			break;
1214 		}
1215 	DADQ_RUNLOCK();
1216 
1217 	return (dp);
1218 }
1219 
1220 static void
1221 nd6_dad_starttimer(struct dadq *dp, int ticks)
1222 {
1223 
1224 	callout_reset(&dp->dad_timer_ch, ticks,
1225 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1226 }
1227 
1228 static void
1229 nd6_dad_stoptimer(struct dadq *dp)
1230 {
1231 
1232 	callout_drain(&dp->dad_timer_ch);
1233 }
1234 
1235 static void
1236 nd6_dad_rele(struct dadq *dp)
1237 {
1238 
1239 	if (refcount_release(&dp->dad_refcnt)) {
1240 		ifa_free(dp->dad_ifa);
1241 		free(dp, M_IP6NDP);
1242 	}
1243 }
1244 
1245 void
1246 nd6_dad_init(void)
1247 {
1248 
1249 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1250 	TAILQ_INIT(&V_dadq);
1251 }
1252 
1253 /*
1254  * Start Duplicate Address Detection (DAD) for specified interface address.
1255  */
1256 void
1257 nd6_dad_start(struct ifaddr *ifa, int delay)
1258 {
1259 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1260 	struct dadq *dp;
1261 	char ip6buf[INET6_ADDRSTRLEN];
1262 
1263 	/*
1264 	 * If we don't need DAD, don't do it.
1265 	 * There are several cases:
1266 	 * - DAD is disabled (ip6_dad_count == 0)
1267 	 * - the interface address is anycast
1268 	 */
1269 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1270 		log(LOG_DEBUG,
1271 			"nd6_dad_start: called with non-tentative address "
1272 			"%s(%s)\n",
1273 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1274 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1275 		return;
1276 	}
1277 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1278 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1279 		return;
1280 	}
1281 	if (!V_ip6_dad_count) {
1282 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1283 		return;
1284 	}
1285 	if (ifa->ifa_ifp == NULL)
1286 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1287 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1288 		return;
1289 	}
1290 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1291 		return;
1292 	if ((dp = nd6_dad_find(ifa)) != NULL) {
1293 		/* DAD already in progress */
1294 		nd6_dad_rele(dp);
1295 		return;
1296 	}
1297 
1298 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1299 	if (dp == NULL) {
1300 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1301 			"%s(%s)\n",
1302 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1303 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1304 		return;
1305 	}
1306 	callout_init(&dp->dad_timer_ch, 0);
1307 #ifdef VIMAGE
1308 	dp->dad_vnet = curvnet;
1309 #endif
1310 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1311 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1312 
1313 	/*
1314 	 * Send NS packet for DAD, ip6_dad_count times.
1315 	 * Note that we must delay the first transmission, if this is the
1316 	 * first packet to be sent from the interface after interface
1317 	 * (re)initialization.
1318 	 */
1319 	dp->dad_ifa = ifa;
1320 	ifa_ref(dp->dad_ifa);
1321 	dp->dad_count = V_ip6_dad_count;
1322 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1323 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1324 	refcount_init(&dp->dad_refcnt, 1);
1325 	nd6_dad_add(dp);
1326 	if (delay == 0) {
1327 		nd6_dad_ns_output(dp, ifa);
1328 		nd6_dad_starttimer(dp,
1329 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1330 	} else {
1331 		nd6_dad_starttimer(dp, delay);
1332 	}
1333 }
1334 
1335 /*
1336  * terminate DAD unconditionally.  used for address removals.
1337  */
1338 void
1339 nd6_dad_stop(struct ifaddr *ifa)
1340 {
1341 	struct dadq *dp;
1342 
1343 	dp = nd6_dad_find(ifa);
1344 	if (!dp) {
1345 		/* DAD wasn't started yet */
1346 		return;
1347 	}
1348 
1349 	nd6_dad_stoptimer(dp);
1350 
1351 	/*
1352 	 * The DAD queue entry may have been removed by nd6_dad_timer() while
1353 	 * we were waiting for it to stop, so re-do the lookup.
1354 	 */
1355 	nd6_dad_rele(dp);
1356 	if (nd6_dad_find(ifa) == NULL)
1357 		return;
1358 
1359 	nd6_dad_del(dp);
1360 	nd6_dad_rele(dp);
1361 }
1362 
1363 static void
1364 nd6_dad_timer(struct dadq *dp)
1365 {
1366 	CURVNET_SET(dp->dad_vnet);
1367 	struct ifaddr *ifa = dp->dad_ifa;
1368 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1369 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1370 	char ip6buf[INET6_ADDRSTRLEN];
1371 
1372 	/* Sanity check */
1373 	if (ia == NULL) {
1374 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1375 		goto err;
1376 	}
1377 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1378 		/* Do not need DAD for ifdisabled interface. */
1379 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1380 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1381 		goto err;
1382 	}
1383 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1384 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1385 			"%s(%s)\n",
1386 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1387 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1388 		goto err;
1389 	}
1390 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1391 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1392 			"%s(%s)\n",
1393 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1394 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1395 		goto err;
1396 	}
1397 
1398 	/* timeouted with IFF_{RUNNING,UP} check */
1399 	if (dp->dad_ns_tcount > V_dad_maxtry) {
1400 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1401 		    if_name(ifa->ifa_ifp)));
1402 		goto err;
1403 	}
1404 
1405 	/* Need more checks? */
1406 	if (dp->dad_ns_ocount < dp->dad_count) {
1407 		/*
1408 		 * We have more NS to go.  Send NS packet for DAD.
1409 		 */
1410 		nd6_dad_ns_output(dp, ifa);
1411 		nd6_dad_starttimer(dp,
1412 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1413 		goto done;
1414 	} else {
1415 		/*
1416 		 * We have transmitted sufficient number of DAD packets.
1417 		 * See what we've got.
1418 		 */
1419 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1420 			/* We've seen NS or NA, means DAD has failed. */
1421 			nd6_dad_duplicated(ifa, dp);
1422 		else {
1423 			/*
1424 			 * We are done with DAD.  No NA came, no NS came.
1425 			 * No duplicate address found.  Check IFDISABLED flag
1426 			 * again in case that it is changed between the
1427 			 * beginning of this function and here.
1428 			 */
1429 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1430 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1431 
1432 			nd6log((LOG_DEBUG,
1433 			    "%s: DAD complete for %s - no duplicates found\n",
1434 			    if_name(ifa->ifa_ifp),
1435 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1436 		}
1437 	}
1438 err:
1439 	nd6_dad_del(dp);
1440 done:
1441 	CURVNET_RESTORE();
1442 }
1443 
1444 static void
1445 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1446 {
1447 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1448 	struct ifnet *ifp;
1449 	char ip6buf[INET6_ADDRSTRLEN];
1450 
1451 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1452 	    "NS in/out=%d/%d, NA in=%d\n",
1453 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1454 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1455 
1456 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1457 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1458 
1459 	ifp = ifa->ifa_ifp;
1460 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1461 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1462 	log(LOG_ERR, "%s: manual intervention required\n",
1463 	    if_name(ifp));
1464 
1465 	/*
1466 	 * If the address is a link-local address formed from an interface
1467 	 * identifier based on the hardware address which is supposed to be
1468 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1469 	 * operation on the interface SHOULD be disabled.
1470 	 * [RFC 4862, Section 5.4.5]
1471 	 */
1472 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1473 		struct in6_addr in6;
1474 
1475 		/*
1476 		 * To avoid over-reaction, we only apply this logic when we are
1477 		 * very sure that hardware addresses are supposed to be unique.
1478 		 */
1479 		switch (ifp->if_type) {
1480 		case IFT_ETHER:
1481 		case IFT_FDDI:
1482 		case IFT_ATM:
1483 		case IFT_IEEE1394:
1484 #ifdef IFT_IEEE80211
1485 		case IFT_IEEE80211:
1486 #endif
1487 		case IFT_INFINIBAND:
1488 			in6 = ia->ia_addr.sin6_addr;
1489 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1490 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1491 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1492 				log(LOG_ERR, "%s: possible hardware address "
1493 				    "duplication detected, disable IPv6\n",
1494 				    if_name(ifp));
1495 			}
1496 			break;
1497 		}
1498 	}
1499 }
1500 
1501 static void
1502 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1503 {
1504 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1505 	struct ifnet *ifp = ifa->ifa_ifp;
1506 
1507 	dp->dad_ns_tcount++;
1508 	if ((ifp->if_flags & IFF_UP) == 0) {
1509 		return;
1510 	}
1511 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1512 		return;
1513 	}
1514 
1515 	dp->dad_ns_ocount++;
1516 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1517 }
1518 
1519 static void
1520 nd6_dad_ns_input(struct ifaddr *ifa)
1521 {
1522 	struct in6_ifaddr *ia;
1523 	struct ifnet *ifp;
1524 	const struct in6_addr *taddr6;
1525 	struct dadq *dp;
1526 
1527 	if (ifa == NULL)
1528 		panic("ifa == NULL in nd6_dad_ns_input");
1529 
1530 	ia = (struct in6_ifaddr *)ifa;
1531 	ifp = ifa->ifa_ifp;
1532 	taddr6 = &ia->ia_addr.sin6_addr;
1533 	dp = nd6_dad_find(ifa);
1534 	if (dp == NULL)
1535 		return;
1536 
1537 	/* Quickhack - completely ignore DAD NS packets */
1538 	if (V_dad_ignore_ns) {
1539 		char ip6buf[INET6_ADDRSTRLEN];
1540 		nd6log((LOG_INFO,
1541 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1542 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1543 		    if_name(ifa->ifa_ifp)));
1544 		return;
1545 	}
1546 
1547 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1548 
1549 	dp->dad_ns_icount++;
1550 	nd6_dad_rele(dp);
1551 }
1552 
1553 static void
1554 nd6_dad_na_input(struct ifaddr *ifa)
1555 {
1556 	struct dadq *dp;
1557 
1558 	if (ifa == NULL)
1559 		panic("ifa == NULL in nd6_dad_na_input");
1560 
1561 	dp = nd6_dad_find(ifa);
1562 	if (dp != NULL) {
1563 		dp->dad_na_icount++;
1564 		nd6_dad_rele(dp);
1565 	}
1566 }
1567