xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 145992504973bd16cf3518af9ba5ce185fefa82a)
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 
55 #include <net/if.h>
56 #include <net/if_types.h>
57 #include <net/if_dl.h>
58 #include <net/if_var.h>
59 #include <net/route.h>
60 #ifdef RADIX_MPATH
61 #include <net/radix_mpath.h>
62 #endif
63 
64 #include <netinet/in.h>
65 #include <netinet/in_var.h>
66 #include <net/if_llatbl.h>
67 #define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
68 #include <netinet6/in6_var.h>
69 #include <netinet6/in6_ifattach.h>
70 #include <netinet/ip6.h>
71 #include <netinet6/ip6_var.h>
72 #include <netinet6/scope6_var.h>
73 #include <netinet6/nd6.h>
74 #include <netinet/icmp6.h>
75 #include <netinet/ip_carp.h>
76 #include <netinet6/send.h>
77 
78 #define SDL(s) ((struct sockaddr_dl *)s)
79 
80 struct dadq;
81 static struct dadq *nd6_dad_find(struct ifaddr *);
82 static void nd6_dad_starttimer(struct dadq *, int);
83 static void nd6_dad_stoptimer(struct dadq *);
84 static void nd6_dad_timer(struct dadq *);
85 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
86 static void nd6_dad_ns_input(struct ifaddr *);
87 static void nd6_dad_na_input(struct ifaddr *);
88 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
89     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
90 
91 VNET_DEFINE(int, dad_ignore_ns) = 0;	/* ignore NS in DAD - specwise incorrect*/
92 VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to transmit DAD packet */
93 #define	V_dad_ignore_ns			VNET(dad_ignore_ns)
94 #define	V_dad_maxtry			VNET(dad_maxtry)
95 
96 /*
97  * Input a Neighbor Solicitation Message.
98  *
99  * Based on RFC 2461
100  * Based on RFC 2462 (duplicate address detection)
101  */
102 void
103 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
104 {
105 	struct ifnet *ifp = m->m_pkthdr.rcvif;
106 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
107 	struct nd_neighbor_solicit *nd_ns;
108 	struct in6_addr saddr6 = ip6->ip6_src;
109 	struct in6_addr daddr6 = ip6->ip6_dst;
110 	struct in6_addr taddr6;
111 	struct in6_addr myaddr6;
112 	char *lladdr = NULL;
113 	struct ifaddr *ifa = NULL;
114 	int lladdrlen = 0;
115 	int anycast = 0, proxy = 0, tentative = 0;
116 	int tlladdr;
117 	int rflag;
118 	union nd_opts ndopts;
119 	struct sockaddr_dl proxydl;
120 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
121 
122 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
123 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
124 		rflag = 0;
125 #ifndef PULLDOWN_TEST
126 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
127 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
128 #else
129 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
130 	if (nd_ns == NULL) {
131 		ICMP6STAT_INC(icp6s_tooshort);
132 		return;
133 	}
134 #endif
135 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
136 	taddr6 = nd_ns->nd_ns_target;
137 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
138 		goto bad;
139 
140 	if (ip6->ip6_hlim != 255) {
141 		nd6log((LOG_ERR,
142 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
143 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
144 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
145 		goto bad;
146 	}
147 
148 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
149 		/* dst has to be a solicited node multicast address. */
150 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
151 		    /* don't check ifindex portion */
152 		    daddr6.s6_addr32[1] == 0 &&
153 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
154 		    daddr6.s6_addr8[12] == 0xff) {
155 			; /* good */
156 		} else {
157 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
158 			    "(wrong ip6 dst)\n"));
159 			goto bad;
160 		}
161 	} else if (!V_nd6_onlink_ns_rfc4861) {
162 		struct sockaddr_in6 src_sa6;
163 
164 		/*
165 		 * According to recent IETF discussions, it is not a good idea
166 		 * to accept a NS from an address which would not be deemed
167 		 * to be a neighbor otherwise.  This point is expected to be
168 		 * clarified in future revisions of the specification.
169 		 */
170 		bzero(&src_sa6, sizeof(src_sa6));
171 		src_sa6.sin6_family = AF_INET6;
172 		src_sa6.sin6_len = sizeof(src_sa6);
173 		src_sa6.sin6_addr = saddr6;
174 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
175 			nd6log((LOG_INFO, "nd6_ns_input: "
176 				"NS packet from non-neighbor\n"));
177 			goto bad;
178 		}
179 	}
180 
181 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
182 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
183 		goto bad;
184 	}
185 
186 	icmp6len -= sizeof(*nd_ns);
187 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
188 	if (nd6_options(&ndopts) < 0) {
189 		nd6log((LOG_INFO,
190 		    "nd6_ns_input: invalid ND option, ignored\n"));
191 		/* nd6_options have incremented stats */
192 		goto freeit;
193 	}
194 
195 	if (ndopts.nd_opts_src_lladdr) {
196 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
197 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
198 	}
199 
200 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
201 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
202 		    "(link-layer address option)\n"));
203 		goto bad;
204 	}
205 
206 	/*
207 	 * Attaching target link-layer address to the NA?
208 	 * (RFC 2461 7.2.4)
209 	 *
210 	 * NS IP dst is unicast/anycast			MUST NOT add
211 	 * NS IP dst is solicited-node multicast	MUST add
212 	 *
213 	 * In implementation, we add target link-layer address by default.
214 	 * We do not add one in MUST NOT cases.
215 	 */
216 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
217 		tlladdr = 0;
218 	else
219 		tlladdr = 1;
220 
221 	/*
222 	 * Target address (taddr6) must be either:
223 	 * (1) Valid unicast/anycast address for my receiving interface,
224 	 * (2) Unicast address for which I'm offering proxy service, or
225 	 * (3) "tentative" address on which DAD is being performed.
226 	 */
227 	/* (1) and (3) check. */
228 	if (ifp->if_carp)
229 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
230 	else
231 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
232 
233 	/* (2) check. */
234 	if (ifa == NULL) {
235 		struct rtentry *rt;
236 		struct sockaddr_in6 tsin6;
237 		int need_proxy;
238 #ifdef RADIX_MPATH
239 		struct route_in6 ro;
240 #endif
241 
242 		bzero(&tsin6, sizeof tsin6);
243 		tsin6.sin6_len = sizeof(struct sockaddr_in6);
244 		tsin6.sin6_family = AF_INET6;
245 		tsin6.sin6_addr = taddr6;
246 
247 		/* Always use the default FIB. */
248 #ifdef RADIX_MPATH
249 		bzero(&ro, sizeof(ro));
250 		ro.ro_dst = tsin6;
251 		rtalloc_mpath_fib((struct route *)&ro, RTF_ANNOUNCE,
252 		    RT_DEFAULT_FIB);
253 		rt = ro.ro_rt;
254 #else
255 		rt = in6_rtalloc1((struct sockaddr *)&tsin6, 0, 0,
256 		    RT_DEFAULT_FIB);
257 #endif
258 		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
259 		    rt->rt_gateway->sa_family == AF_LINK);
260 		if (rt != NULL) {
261 			/*
262 			 * Make a copy while we can be sure that rt_gateway
263 			 * is still stable before unlocking to avoid lock
264 			 * order problems.  proxydl will only be used if
265 			 * proxy will be set in the next block.
266 			 */
267 			if (need_proxy)
268 				proxydl = *SDL(rt->rt_gateway);
269 			RTFREE_LOCKED(rt);
270 		}
271 		if (need_proxy) {
272 			/*
273 			 * proxy NDP for single entry
274 			 */
275 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
276 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
277 			if (ifa)
278 				proxy = 1;
279 		}
280 	}
281 	if (ifa == NULL) {
282 		/*
283 		 * We've got an NS packet, and we don't have that adddress
284 		 * assigned for us.  We MUST silently ignore it.
285 		 * See RFC2461 7.2.3.
286 		 */
287 		goto freeit;
288 	}
289 	myaddr6 = *IFA_IN6(ifa);
290 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
291 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
292 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
293 		goto freeit;
294 
295 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
296 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
297 		    "(if %d, NS packet %d)\n",
298 		    ip6_sprintf(ip6bufs, &taddr6),
299 		    ifp->if_addrlen, lladdrlen - 2));
300 		goto bad;
301 	}
302 
303 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
304 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
305 		    ip6_sprintf(ip6bufs, &saddr6)));
306 		goto freeit;
307 	}
308 
309 	/*
310 	 * We have neighbor solicitation packet, with target address equals to
311 	 * one of my tentative address.
312 	 *
313 	 * src addr	how to process?
314 	 * ---		---
315 	 * multicast	of course, invalid (rejected in ip6_input)
316 	 * unicast	somebody is doing address resolution -> ignore
317 	 * unspec	dup address detection
318 	 *
319 	 * The processing is defined in RFC 2462.
320 	 */
321 	if (tentative) {
322 		/*
323 		 * If source address is unspecified address, it is for
324 		 * duplicate address detection.
325 		 *
326 		 * If not, the packet is for addess resolution;
327 		 * silently ignore it.
328 		 */
329 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
330 			nd6_dad_ns_input(ifa);
331 
332 		goto freeit;
333 	}
334 
335 	/*
336 	 * If the source address is unspecified address, entries must not
337 	 * be created or updated.
338 	 * It looks that sender is performing DAD.  Output NA toward
339 	 * all-node multicast address, to tell the sender that I'm using
340 	 * the address.
341 	 * S bit ("solicited") must be zero.
342 	 */
343 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
344 		struct in6_addr in6_all;
345 
346 		in6_all = in6addr_linklocal_allnodes;
347 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
348 			goto bad;
349 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
350 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
351 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
352 		    M_GETFIB(m));
353 		goto freeit;
354 	}
355 
356 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
357 	    ND_NEIGHBOR_SOLICIT, 0);
358 
359 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
360 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
361 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
362 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
363  freeit:
364 	if (ifa != NULL)
365 		ifa_free(ifa);
366 	m_freem(m);
367 	return;
368 
369  bad:
370 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
371 		ip6_sprintf(ip6bufs, &saddr6)));
372 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
373 		ip6_sprintf(ip6bufs, &daddr6)));
374 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
375 		ip6_sprintf(ip6bufs, &taddr6)));
376 	ICMP6STAT_INC(icp6s_badns);
377 	if (ifa != NULL)
378 		ifa_free(ifa);
379 	m_freem(m);
380 }
381 
382 /*
383  * Output a Neighbor Solicitation Message. Caller specifies:
384  *	- ICMP6 header source IP6 address
385  *	- ND6 header target IP6 address
386  *	- ND6 header source datalink address
387  *
388  * Based on RFC 2461
389  * Based on RFC 2462 (duplicate address detection)
390  *
391  *   ln - for source address determination
392  *  dad - duplicate address detection
393  */
394 void
395 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
396     const struct in6_addr *taddr6, struct llentry *ln, int dad)
397 {
398 	struct mbuf *m;
399 	struct m_tag *mtag;
400 	struct ip6_hdr *ip6;
401 	struct nd_neighbor_solicit *nd_ns;
402 	struct ip6_moptions im6o;
403 	int icmp6len;
404 	int maxlen;
405 	caddr_t mac;
406 	struct route_in6 ro;
407 
408 	if (IN6_IS_ADDR_MULTICAST(taddr6))
409 		return;
410 
411 	/* estimate the size of message */
412 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
413 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
414 	if (max_linkhdr + maxlen >= MCLBYTES) {
415 #ifdef DIAGNOSTIC
416 		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
417 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
418 #endif
419 		return;
420 	}
421 
422 	MGETHDR(m, M_DONTWAIT, MT_DATA);
423 	if (m && max_linkhdr + maxlen >= MHLEN) {
424 		MCLGET(m, M_DONTWAIT);
425 		if ((m->m_flags & M_EXT) == 0) {
426 			m_free(m);
427 			m = NULL;
428 		}
429 	}
430 	if (m == NULL)
431 		return;
432 	m->m_pkthdr.rcvif = NULL;
433 
434 	bzero(&ro, sizeof(ro));
435 
436 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
437 		m->m_flags |= M_MCAST;
438 		im6o.im6o_multicast_ifp = ifp;
439 		im6o.im6o_multicast_hlim = 255;
440 		im6o.im6o_multicast_loop = 0;
441 	}
442 
443 	icmp6len = sizeof(*nd_ns);
444 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
445 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
446 
447 	/* fill neighbor solicitation packet */
448 	ip6 = mtod(m, struct ip6_hdr *);
449 	ip6->ip6_flow = 0;
450 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
451 	ip6->ip6_vfc |= IPV6_VERSION;
452 	/* ip6->ip6_plen will be set later */
453 	ip6->ip6_nxt = IPPROTO_ICMPV6;
454 	ip6->ip6_hlim = 255;
455 	if (daddr6)
456 		ip6->ip6_dst = *daddr6;
457 	else {
458 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
459 		ip6->ip6_dst.s6_addr16[1] = 0;
460 		ip6->ip6_dst.s6_addr32[1] = 0;
461 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
462 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
463 		ip6->ip6_dst.s6_addr8[12] = 0xff;
464 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
465 			goto bad;
466 	}
467 	if (!dad) {
468 		struct ifaddr *ifa;
469 
470 		/*
471 		 * RFC2461 7.2.2:
472 		 * "If the source address of the packet prompting the
473 		 * solicitation is the same as one of the addresses assigned
474 		 * to the outgoing interface, that address SHOULD be placed
475 		 * in the IP Source Address of the outgoing solicitation.
476 		 * Otherwise, any one of the addresses assigned to the
477 		 * interface should be used."
478 		 *
479 		 * We use the source address for the prompting packet
480 		 * (saddr6), if:
481 		 * - saddr6 is given from the caller (by giving "ln"), and
482 		 * - saddr6 belongs to the outgoing interface.
483 		 * Otherwise, we perform the source address selection as usual.
484 		 */
485 		struct in6_addr *hsrc;
486 
487 		hsrc = NULL;
488 		if (ln != NULL) {
489 			LLE_RLOCK(ln);
490 			if (ln->la_hold != NULL) {
491 				struct ip6_hdr *hip6;		/* hold ip6 */
492 
493 				/*
494 				 * assuming every packet in la_hold has the same IP
495 				 * header
496 				 */
497 				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
498 				/* XXX pullup? */
499 				if (sizeof(*hip6) < ln->la_hold->m_len) {
500 					ip6->ip6_src = hip6->ip6_src;
501 					hsrc = &hip6->ip6_src;
502 				}
503 			}
504 			LLE_RUNLOCK(ln);
505 		}
506 		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
507 		    hsrc)) != NULL) {
508 			/* ip6_src set already. */
509 			ifa_free(ifa);
510 		} else {
511 			int error;
512 			struct sockaddr_in6 dst_sa;
513 			struct in6_addr src_in;
514 			struct ifnet *oifp;
515 
516 			bzero(&dst_sa, sizeof(dst_sa));
517 			dst_sa.sin6_family = AF_INET6;
518 			dst_sa.sin6_len = sizeof(dst_sa);
519 			dst_sa.sin6_addr = ip6->ip6_dst;
520 
521 			oifp = ifp;
522 			error = in6_selectsrc(&dst_sa, NULL,
523 			    NULL, &ro, NULL, &oifp, &src_in);
524 			if (error) {
525 				char ip6buf[INET6_ADDRSTRLEN];
526 				nd6log((LOG_DEBUG,
527 				    "nd6_ns_output: source can't be "
528 				    "determined: dst=%s, error=%d\n",
529 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
530 				    error));
531 				goto bad;
532 			}
533 			ip6->ip6_src = src_in;
534 		}
535 	} else {
536 		/*
537 		 * Source address for DAD packet must always be IPv6
538 		 * unspecified address. (0::0)
539 		 * We actually don't have to 0-clear the address (we did it
540 		 * above), but we do so here explicitly to make the intention
541 		 * clearer.
542 		 */
543 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
544 	}
545 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
546 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
547 	nd_ns->nd_ns_code = 0;
548 	nd_ns->nd_ns_reserved = 0;
549 	nd_ns->nd_ns_target = *taddr6;
550 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
551 
552 	/*
553 	 * Add source link-layer address option.
554 	 *
555 	 *				spec		implementation
556 	 *				---		---
557 	 * DAD packet			MUST NOT	do not add the option
558 	 * there's no link layer address:
559 	 *				impossible	do not add the option
560 	 * there's link layer address:
561 	 *	Multicast NS		MUST add one	add the option
562 	 *	Unicast NS		SHOULD add one	add the option
563 	 */
564 	if (!dad && (mac = nd6_ifptomac(ifp))) {
565 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
566 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
567 		/* 8 byte alignments... */
568 		optlen = (optlen + 7) & ~7;
569 
570 		m->m_pkthdr.len += optlen;
571 		m->m_len += optlen;
572 		icmp6len += optlen;
573 		bzero((caddr_t)nd_opt, optlen);
574 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
575 		nd_opt->nd_opt_len = optlen >> 3;
576 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
577 	}
578 
579 	ip6->ip6_plen = htons((u_short)icmp6len);
580 	nd_ns->nd_ns_cksum = 0;
581 	nd_ns->nd_ns_cksum =
582 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
583 
584 	if (send_sendso_input_hook != NULL) {
585 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
586 			sizeof(unsigned short), M_NOWAIT);
587 		if (mtag == NULL)
588 			goto bad;
589 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
590 		m_tag_prepend(m, mtag);
591 	}
592 
593 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
594 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
595 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
596 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
597 
598 	/* We don't cache this route. */
599 	RO_RTFREE(&ro);
600 
601 	return;
602 
603   bad:
604 	if (ro.ro_rt) {
605 		RTFREE(ro.ro_rt);
606 	}
607 	m_freem(m);
608 	return;
609 }
610 
611 /*
612  * Neighbor advertisement input handling.
613  *
614  * Based on RFC 2461
615  * Based on RFC 2462 (duplicate address detection)
616  *
617  * the following items are not implemented yet:
618  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
619  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
620  */
621 void
622 nd6_na_input(struct mbuf *m, int off, int icmp6len)
623 {
624 	struct ifnet *ifp = m->m_pkthdr.rcvif;
625 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
626 	struct nd_neighbor_advert *nd_na;
627 	struct in6_addr daddr6 = ip6->ip6_dst;
628 	struct in6_addr taddr6;
629 	int flags;
630 	int is_router;
631 	int is_solicited;
632 	int is_override;
633 	char *lladdr = NULL;
634 	int lladdrlen = 0;
635 	int checklink = 0;
636 	struct ifaddr *ifa;
637 	struct llentry *ln = NULL;
638 	union nd_opts ndopts;
639 	struct mbuf *chain = NULL;
640 	struct m_tag *mtag;
641 	struct sockaddr_in6 sin6;
642 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
643 
644 	if (ip6->ip6_hlim != 255) {
645 		nd6log((LOG_ERR,
646 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
647 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
648 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
649 		goto bad;
650 	}
651 
652 #ifndef PULLDOWN_TEST
653 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
654 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
655 #else
656 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
657 	if (nd_na == NULL) {
658 		ICMP6STAT_INC(icp6s_tooshort);
659 		return;
660 	}
661 #endif
662 
663 	flags = nd_na->nd_na_flags_reserved;
664 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
665 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
666 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
667 
668 	taddr6 = nd_na->nd_na_target;
669 	if (in6_setscope(&taddr6, ifp, NULL))
670 		goto bad;	/* XXX: impossible */
671 
672 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
673 		nd6log((LOG_ERR,
674 		    "nd6_na_input: invalid target address %s\n",
675 		    ip6_sprintf(ip6bufs, &taddr6)));
676 		goto bad;
677 	}
678 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
679 		if (is_solicited) {
680 			nd6log((LOG_ERR,
681 			    "nd6_na_input: a solicited adv is multicasted\n"));
682 			goto bad;
683 		}
684 
685 	icmp6len -= sizeof(*nd_na);
686 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
687 	if (nd6_options(&ndopts) < 0) {
688 		nd6log((LOG_INFO,
689 		    "nd6_na_input: invalid ND option, ignored\n"));
690 		/* nd6_options have incremented stats */
691 		goto freeit;
692 	}
693 
694 	if (ndopts.nd_opts_tgt_lladdr) {
695 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
696 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
697 	}
698 
699 	/*
700 	 * This effectively disables the DAD check on a non-master CARP
701 	 * address.
702 	 */
703 	if (ifp->if_carp)
704 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
705 	else
706 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
707 
708 	/*
709 	 * Target address matches one of my interface address.
710 	 *
711 	 * If my address is tentative, this means that there's somebody
712 	 * already using the same address as mine.  This indicates DAD failure.
713 	 * This is defined in RFC 2462.
714 	 *
715 	 * Otherwise, process as defined in RFC 2461.
716 	 */
717 	if (ifa
718 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
719 		ifa_free(ifa);
720 		nd6_dad_na_input(ifa);
721 		goto freeit;
722 	}
723 
724 	/* Just for safety, maybe unnecessary. */
725 	if (ifa) {
726 		ifa_free(ifa);
727 		log(LOG_ERR,
728 		    "nd6_na_input: duplicate IP6 address %s\n",
729 		    ip6_sprintf(ip6bufs, &taddr6));
730 		goto freeit;
731 	}
732 
733 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
734 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
735 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
736 		    ifp->if_addrlen, lladdrlen - 2));
737 		goto bad;
738 	}
739 
740 	/*
741 	 * If no neighbor cache entry is found, NA SHOULD silently be
742 	 * discarded.
743 	 */
744 	IF_AFDATA_LOCK(ifp);
745 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
746 	IF_AFDATA_UNLOCK(ifp);
747 	if (ln == NULL) {
748 		goto freeit;
749 	}
750 
751 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
752 		/*
753 		 * If the link-layer has address, and no lladdr option came,
754 		 * discard the packet.
755 		 */
756 		if (ifp->if_addrlen && lladdr == NULL) {
757 			goto freeit;
758 		}
759 
760 		/*
761 		 * Record link-layer address, and update the state.
762 		 */
763 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
764 		ln->la_flags |= LLE_VALID;
765 		if (is_solicited) {
766 			ln->ln_state = ND6_LLINFO_REACHABLE;
767 			ln->ln_byhint = 0;
768 			if (!ND6_LLINFO_PERMANENT(ln)) {
769 				nd6_llinfo_settimer_locked(ln,
770 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
771 			}
772 		} else {
773 			ln->ln_state = ND6_LLINFO_STALE;
774 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
775 		}
776 		if ((ln->ln_router = is_router) != 0) {
777 			/*
778 			 * This means a router's state has changed from
779 			 * non-reachable to probably reachable, and might
780 			 * affect the status of associated prefixes..
781 			 */
782 			checklink = 1;
783 		}
784 	} else {
785 		int llchange;
786 
787 		/*
788 		 * Check if the link-layer address has changed or not.
789 		 */
790 		if (lladdr == NULL)
791 			llchange = 0;
792 		else {
793 			if (ln->la_flags & LLE_VALID) {
794 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
795 					llchange = 1;
796 				else
797 					llchange = 0;
798 			} else
799 				llchange = 1;
800 		}
801 
802 		/*
803 		 * This is VERY complex.  Look at it with care.
804 		 *
805 		 * override solicit lladdr llchange	action
806 		 *					(L: record lladdr)
807 		 *
808 		 *	0	0	n	--	(2c)
809 		 *	0	0	y	n	(2b) L
810 		 *	0	0	y	y	(1)    REACHABLE->STALE
811 		 *	0	1	n	--	(2c)   *->REACHABLE
812 		 *	0	1	y	n	(2b) L *->REACHABLE
813 		 *	0	1	y	y	(1)    REACHABLE->STALE
814 		 *	1	0	n	--	(2a)
815 		 *	1	0	y	n	(2a) L
816 		 *	1	0	y	y	(2a) L *->STALE
817 		 *	1	1	n	--	(2a)   *->REACHABLE
818 		 *	1	1	y	n	(2a) L *->REACHABLE
819 		 *	1	1	y	y	(2a) L *->REACHABLE
820 		 */
821 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
822 			/*
823 			 * If state is REACHABLE, make it STALE.
824 			 * no other updates should be done.
825 			 */
826 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
827 				ln->ln_state = ND6_LLINFO_STALE;
828 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
829 			}
830 			goto freeit;
831 		} else if (is_override				   /* (2a) */
832 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
833 			|| lladdr == NULL) {			   /* (2c) */
834 			/*
835 			 * Update link-local address, if any.
836 			 */
837 			if (lladdr != NULL) {
838 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
839 				ln->la_flags |= LLE_VALID;
840 			}
841 
842 			/*
843 			 * If solicited, make the state REACHABLE.
844 			 * If not solicited and the link-layer address was
845 			 * changed, make it STALE.
846 			 */
847 			if (is_solicited) {
848 				ln->ln_state = ND6_LLINFO_REACHABLE;
849 				ln->ln_byhint = 0;
850 				if (!ND6_LLINFO_PERMANENT(ln)) {
851 					nd6_llinfo_settimer_locked(ln,
852 					    (long)ND_IFINFO(ifp)->reachable * hz);
853 				}
854 			} else {
855 				if (lladdr != NULL && llchange) {
856 					ln->ln_state = ND6_LLINFO_STALE;
857 					nd6_llinfo_settimer_locked(ln,
858 					    (long)V_nd6_gctimer * hz);
859 				}
860 			}
861 		}
862 
863 		if (ln->ln_router && !is_router) {
864 			/*
865 			 * The peer dropped the router flag.
866 			 * Remove the sender from the Default Router List and
867 			 * update the Destination Cache entries.
868 			 */
869 			struct nd_defrouter *dr;
870 			struct in6_addr *in6;
871 
872 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
873 
874 			/*
875 			 * Lock to protect the default router list.
876 			 * XXX: this might be unnecessary, since this function
877 			 * is only called under the network software interrupt
878 			 * context.  However, we keep it just for safety.
879 			 */
880 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
881 			if (dr)
882 				defrtrlist_del(dr);
883 			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
884 			    ND6_IFF_ACCEPT_RTADV) {
885 				/*
886 				 * Even if the neighbor is not in the default
887 				 * router list, the neighbor may be used
888 				 * as a next hop for some destinations
889 				 * (e.g. redirect case). So we must
890 				 * call rt6_flush explicitly.
891 				 */
892 				rt6_flush(&ip6->ip6_src, ifp);
893 			}
894 		}
895 		ln->ln_router = is_router;
896 	}
897         /* XXX - QL
898 	 *  Does this matter?
899 	 *  rt->rt_flags &= ~RTF_REJECT;
900 	 */
901 	ln->la_asked = 0;
902 	if (ln->la_hold) {
903 		struct mbuf *m_hold, *m_hold_next;
904 
905 		/*
906 		 * reset the la_hold in advance, to explicitly
907 		 * prevent a la_hold lookup in nd6_output()
908 		 * (wouldn't happen, though...)
909 		 */
910 		for (m_hold = ln->la_hold, ln->la_hold = NULL;
911 		    m_hold; m_hold = m_hold_next) {
912 			m_hold_next = m_hold->m_nextpkt;
913 			m_hold->m_nextpkt = NULL;
914 			/*
915 			 * we assume ifp is not a loopback here, so just set
916 			 * the 2nd argument as the 1st one.
917 			 */
918 
919 			if (send_sendso_input_hook != NULL) {
920 				mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
921 				    sizeof(unsigned short), M_NOWAIT);
922 				if (mtag == NULL)
923 					goto bad;
924 				m_tag_prepend(m, mtag);
925 			}
926 
927 			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
928 		}
929 	}
930  freeit:
931 	if (ln != NULL) {
932 		if (chain)
933 			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
934 		LLE_WUNLOCK(ln);
935 
936 		if (chain)
937 			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
938 	}
939 	if (checklink)
940 		pfxlist_onlink_check();
941 
942 	m_freem(m);
943 	return;
944 
945  bad:
946 	if (ln != NULL)
947 		LLE_WUNLOCK(ln);
948 
949 	ICMP6STAT_INC(icp6s_badna);
950 	m_freem(m);
951 }
952 
953 /*
954  * Neighbor advertisement output handling.
955  *
956  * Based on RFC 2461
957  *
958  * the following items are not implemented yet:
959  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
960  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
961  *
962  * tlladdr - 1 if include target link-layer address
963  * sdl0 - sockaddr_dl (= proxy NA) or NULL
964  */
965 static void
966 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
967     const struct in6_addr *taddr6, u_long flags, int tlladdr,
968     struct sockaddr *sdl0, u_int fibnum)
969 {
970 	struct mbuf *m;
971 	struct m_tag *mtag;
972 	struct ifnet *oifp;
973 	struct ip6_hdr *ip6;
974 	struct nd_neighbor_advert *nd_na;
975 	struct ip6_moptions im6o;
976 	struct in6_addr src, daddr6;
977 	struct sockaddr_in6 dst_sa;
978 	int icmp6len, maxlen, error;
979 	caddr_t mac = NULL;
980 	struct route_in6 ro;
981 
982 	bzero(&ro, sizeof(ro));
983 
984 	daddr6 = *daddr6_0;	/* make a local copy for modification */
985 
986 	/* estimate the size of message */
987 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
988 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
989 	if (max_linkhdr + maxlen >= MCLBYTES) {
990 #ifdef DIAGNOSTIC
991 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
992 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
993 #endif
994 		return;
995 	}
996 
997 	MGETHDR(m, M_DONTWAIT, MT_DATA);
998 	if (m && max_linkhdr + maxlen >= MHLEN) {
999 		MCLGET(m, M_DONTWAIT);
1000 		if ((m->m_flags & M_EXT) == 0) {
1001 			m_free(m);
1002 			m = NULL;
1003 		}
1004 	}
1005 	if (m == NULL)
1006 		return;
1007 	m->m_pkthdr.rcvif = NULL;
1008 	M_SETFIB(m, fibnum);
1009 
1010 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1011 		m->m_flags |= M_MCAST;
1012 		im6o.im6o_multicast_ifp = ifp;
1013 		im6o.im6o_multicast_hlim = 255;
1014 		im6o.im6o_multicast_loop = 0;
1015 	}
1016 
1017 	icmp6len = sizeof(*nd_na);
1018 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1019 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
1020 
1021 	/* fill neighbor advertisement packet */
1022 	ip6 = mtod(m, struct ip6_hdr *);
1023 	ip6->ip6_flow = 0;
1024 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1025 	ip6->ip6_vfc |= IPV6_VERSION;
1026 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1027 	ip6->ip6_hlim = 255;
1028 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1029 		/* reply to DAD */
1030 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1031 		daddr6.s6_addr16[1] = 0;
1032 		daddr6.s6_addr32[1] = 0;
1033 		daddr6.s6_addr32[2] = 0;
1034 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1035 		if (in6_setscope(&daddr6, ifp, NULL))
1036 			goto bad;
1037 
1038 		flags &= ~ND_NA_FLAG_SOLICITED;
1039 	}
1040 	ip6->ip6_dst = daddr6;
1041 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1042 	dst_sa.sin6_family = AF_INET6;
1043 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1044 	dst_sa.sin6_addr = daddr6;
1045 
1046 	/*
1047 	 * Select a source whose scope is the same as that of the dest.
1048 	 */
1049 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1050 	oifp = ifp;
1051 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1052 	if (error) {
1053 		char ip6buf[INET6_ADDRSTRLEN];
1054 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1055 		    "determined: dst=%s, error=%d\n",
1056 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1057 		goto bad;
1058 	}
1059 	ip6->ip6_src = src;
1060 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1061 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1062 	nd_na->nd_na_code = 0;
1063 	nd_na->nd_na_target = *taddr6;
1064 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1065 
1066 	/*
1067 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1068 	 * see nd6_ns_input() for details.
1069 	 * Basically, if NS packet is sent to unicast/anycast addr,
1070 	 * target lladdr option SHOULD NOT be included.
1071 	 */
1072 	if (tlladdr) {
1073 		/*
1074 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1075 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1076 		 * my address) use lladdr configured for the interface.
1077 		 */
1078 		if (sdl0 == NULL) {
1079 			if (ifp->if_carp)
1080 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1081 			if (mac == NULL)
1082 				mac = nd6_ifptomac(ifp);
1083 		} else if (sdl0->sa_family == AF_LINK) {
1084 			struct sockaddr_dl *sdl;
1085 			sdl = (struct sockaddr_dl *)sdl0;
1086 			if (sdl->sdl_alen == ifp->if_addrlen)
1087 				mac = LLADDR(sdl);
1088 		}
1089 	}
1090 	if (tlladdr && mac) {
1091 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1092 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1093 
1094 		/* roundup to 8 bytes alignment! */
1095 		optlen = (optlen + 7) & ~7;
1096 
1097 		m->m_pkthdr.len += optlen;
1098 		m->m_len += optlen;
1099 		icmp6len += optlen;
1100 		bzero((caddr_t)nd_opt, optlen);
1101 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1102 		nd_opt->nd_opt_len = optlen >> 3;
1103 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1104 	} else
1105 		flags &= ~ND_NA_FLAG_OVERRIDE;
1106 
1107 	ip6->ip6_plen = htons((u_short)icmp6len);
1108 	nd_na->nd_na_flags_reserved = flags;
1109 	nd_na->nd_na_cksum = 0;
1110 	nd_na->nd_na_cksum =
1111 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1112 
1113 	if (send_sendso_input_hook != NULL) {
1114 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1115 		    sizeof(unsigned short), M_NOWAIT);
1116 		if (mtag == NULL)
1117 			goto bad;
1118 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1119 		m_tag_prepend(m, mtag);
1120 	}
1121 
1122 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1123 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1124 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1125 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1126 
1127 	/* We don't cache this route. */
1128 	RO_RTFREE(&ro);
1129 
1130 	return;
1131 
1132   bad:
1133 	if (ro.ro_rt) {
1134 		RTFREE(ro.ro_rt);
1135 	}
1136 	m_freem(m);
1137 	return;
1138 }
1139 
1140 #ifndef BURN_BRIDGES
1141 void
1142 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1143     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1144     struct sockaddr *sdl0)
1145 {
1146 
1147 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1148 	    RT_DEFAULT_FIB);
1149 }
1150 #endif
1151 
1152 caddr_t
1153 nd6_ifptomac(struct ifnet *ifp)
1154 {
1155 	switch (ifp->if_type) {
1156 	case IFT_ARCNET:
1157 	case IFT_ETHER:
1158 	case IFT_FDDI:
1159 	case IFT_IEEE1394:
1160 #ifdef IFT_L2VLAN
1161 	case IFT_L2VLAN:
1162 #endif
1163 #ifdef IFT_IEEE80211
1164 	case IFT_IEEE80211:
1165 #endif
1166 	case IFT_INFINIBAND:
1167 	case IFT_BRIDGE:
1168 	case IFT_ISO88025:
1169 		return IF_LLADDR(ifp);
1170 	default:
1171 		return NULL;
1172 	}
1173 }
1174 
1175 struct dadq {
1176 	TAILQ_ENTRY(dadq) dad_list;
1177 	struct ifaddr *dad_ifa;
1178 	int dad_count;		/* max NS to send */
1179 	int dad_ns_tcount;	/* # of trials to send NS */
1180 	int dad_ns_ocount;	/* NS sent so far */
1181 	int dad_ns_icount;
1182 	int dad_na_icount;
1183 	struct callout dad_timer_ch;
1184 	struct vnet *dad_vnet;
1185 };
1186 
1187 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1188 VNET_DEFINE(int, dad_init) = 0;
1189 #define	V_dadq				VNET(dadq)
1190 #define	V_dad_init			VNET(dad_init)
1191 
1192 static struct dadq *
1193 nd6_dad_find(struct ifaddr *ifa)
1194 {
1195 	struct dadq *dp;
1196 
1197 	TAILQ_FOREACH(dp, &V_dadq, dad_list)
1198 		if (dp->dad_ifa == ifa)
1199 			return (dp);
1200 
1201 	return (NULL);
1202 }
1203 
1204 static void
1205 nd6_dad_starttimer(struct dadq *dp, int ticks)
1206 {
1207 
1208 	callout_reset(&dp->dad_timer_ch, ticks,
1209 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1210 }
1211 
1212 static void
1213 nd6_dad_stoptimer(struct dadq *dp)
1214 {
1215 
1216 	callout_stop(&dp->dad_timer_ch);
1217 }
1218 
1219 /*
1220  * Start Duplicate Address Detection (DAD) for specified interface address.
1221  */
1222 void
1223 nd6_dad_start(struct ifaddr *ifa, int delay)
1224 {
1225 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1226 	struct dadq *dp;
1227 	char ip6buf[INET6_ADDRSTRLEN];
1228 
1229 	if (!V_dad_init) {
1230 		TAILQ_INIT(&V_dadq);
1231 		V_dad_init++;
1232 	}
1233 
1234 	/*
1235 	 * If we don't need DAD, don't do it.
1236 	 * There are several cases:
1237 	 * - DAD is disabled (ip6_dad_count == 0)
1238 	 * - the interface address is anycast
1239 	 */
1240 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1241 		log(LOG_DEBUG,
1242 			"nd6_dad_start: called with non-tentative address "
1243 			"%s(%s)\n",
1244 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1245 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1246 		return;
1247 	}
1248 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1249 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1250 		return;
1251 	}
1252 	if (!V_ip6_dad_count) {
1253 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1254 		return;
1255 	}
1256 	if (ifa->ifa_ifp == NULL)
1257 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1258 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1259 		return;
1260 	}
1261 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1262 		return;
1263 	if (nd6_dad_find(ifa) != NULL) {
1264 		/* DAD already in progress */
1265 		return;
1266 	}
1267 
1268 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1269 	if (dp == NULL) {
1270 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1271 			"%s(%s)\n",
1272 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1273 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1274 		return;
1275 	}
1276 	bzero(dp, sizeof(*dp));
1277 	callout_init(&dp->dad_timer_ch, 0);
1278 #ifdef VIMAGE
1279 	dp->dad_vnet = curvnet;
1280 #endif
1281 	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1282 
1283 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1284 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1285 
1286 	/*
1287 	 * Send NS packet for DAD, ip6_dad_count times.
1288 	 * Note that we must delay the first transmission, if this is the
1289 	 * first packet to be sent from the interface after interface
1290 	 * (re)initialization.
1291 	 */
1292 	dp->dad_ifa = ifa;
1293 	ifa_ref(ifa);	/* just for safety */
1294 	dp->dad_count = V_ip6_dad_count;
1295 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1296 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1297 	if (delay == 0) {
1298 		nd6_dad_ns_output(dp, ifa);
1299 		nd6_dad_starttimer(dp,
1300 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1301 	} else {
1302 		nd6_dad_starttimer(dp, delay);
1303 	}
1304 }
1305 
1306 /*
1307  * terminate DAD unconditionally.  used for address removals.
1308  */
1309 void
1310 nd6_dad_stop(struct ifaddr *ifa)
1311 {
1312 	struct dadq *dp;
1313 
1314 	if (!V_dad_init)
1315 		return;
1316 	dp = nd6_dad_find(ifa);
1317 	if (!dp) {
1318 		/* DAD wasn't started yet */
1319 		return;
1320 	}
1321 
1322 	nd6_dad_stoptimer(dp);
1323 
1324 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1325 	free(dp, M_IP6NDP);
1326 	dp = NULL;
1327 	ifa_free(ifa);
1328 }
1329 
1330 static void
1331 nd6_dad_timer(struct dadq *dp)
1332 {
1333 	CURVNET_SET(dp->dad_vnet);
1334 	struct ifaddr *ifa = dp->dad_ifa;
1335 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1336 	char ip6buf[INET6_ADDRSTRLEN];
1337 
1338 	/* Sanity check */
1339 	if (ia == NULL) {
1340 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1341 		goto done;
1342 	}
1343 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1344 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1345 			"%s(%s)\n",
1346 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1347 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1348 		goto done;
1349 	}
1350 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1351 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1352 			"%s(%s)\n",
1353 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1354 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1355 		goto done;
1356 	}
1357 
1358 	/* timeouted with IFF_{RUNNING,UP} check */
1359 	if (dp->dad_ns_tcount > V_dad_maxtry) {
1360 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1361 		    if_name(ifa->ifa_ifp)));
1362 
1363 		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1364 		free(dp, M_IP6NDP);
1365 		dp = NULL;
1366 		ifa_free(ifa);
1367 		goto done;
1368 	}
1369 
1370 	/* Need more checks? */
1371 	if (dp->dad_ns_ocount < dp->dad_count) {
1372 		/*
1373 		 * We have more NS to go.  Send NS packet for DAD.
1374 		 */
1375 		nd6_dad_ns_output(dp, ifa);
1376 		nd6_dad_starttimer(dp,
1377 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1378 	} else {
1379 		/*
1380 		 * We have transmitted sufficient number of DAD packets.
1381 		 * See what we've got.
1382 		 */
1383 		int duplicate;
1384 
1385 		duplicate = 0;
1386 
1387 		if (dp->dad_na_icount) {
1388 			/*
1389 			 * the check is in nd6_dad_na_input(),
1390 			 * but just in case
1391 			 */
1392 			duplicate++;
1393 		}
1394 
1395 		if (dp->dad_ns_icount) {
1396 			/* We've seen NS, means DAD has failed. */
1397 			duplicate++;
1398 		}
1399 
1400 		if (duplicate) {
1401 			/* (*dp) will be freed in nd6_dad_duplicated() */
1402 			dp = NULL;
1403 			nd6_dad_duplicated(ifa);
1404 		} else {
1405 			/*
1406 			 * We are done with DAD.  No NA came, no NS came.
1407 			 * No duplicate address found.
1408 			 */
1409 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1410 
1411 			nd6log((LOG_DEBUG,
1412 			    "%s: DAD complete for %s - no duplicates found\n",
1413 			    if_name(ifa->ifa_ifp),
1414 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1415 
1416 			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1417 			free(dp, M_IP6NDP);
1418 			dp = NULL;
1419 			ifa_free(ifa);
1420 		}
1421 	}
1422 
1423 done:
1424 	CURVNET_RESTORE();
1425 }
1426 
1427 void
1428 nd6_dad_duplicated(struct ifaddr *ifa)
1429 {
1430 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1431 	struct ifnet *ifp;
1432 	struct dadq *dp;
1433 	char ip6buf[INET6_ADDRSTRLEN];
1434 
1435 	dp = nd6_dad_find(ifa);
1436 	if (dp == NULL) {
1437 		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1438 		return;
1439 	}
1440 
1441 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1442 	    "NS in/out=%d/%d, NA in=%d\n",
1443 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1444 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1445 
1446 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1447 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1448 
1449 	/* We are done with DAD, with duplicate address found. (failure) */
1450 	nd6_dad_stoptimer(dp);
1451 
1452 	ifp = ifa->ifa_ifp;
1453 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1454 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1455 	log(LOG_ERR, "%s: manual intervention required\n",
1456 	    if_name(ifp));
1457 
1458 	/*
1459 	 * If the address is a link-local address formed from an interface
1460 	 * identifier based on the hardware address which is supposed to be
1461 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1462 	 * operation on the interface SHOULD be disabled.
1463 	 * [RFC 4862, Section 5.4.5]
1464 	 */
1465 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1466 		struct in6_addr in6;
1467 
1468 		/*
1469 		 * To avoid over-reaction, we only apply this logic when we are
1470 		 * very sure that hardware addresses are supposed to be unique.
1471 		 */
1472 		switch (ifp->if_type) {
1473 		case IFT_ETHER:
1474 		case IFT_FDDI:
1475 		case IFT_ATM:
1476 		case IFT_IEEE1394:
1477 #ifdef IFT_IEEE80211
1478 		case IFT_IEEE80211:
1479 #endif
1480 		case IFT_INFINIBAND:
1481 			in6 = ia->ia_addr.sin6_addr;
1482 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1483 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1484 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1485 				log(LOG_ERR, "%s: possible hardware address "
1486 				    "duplication detected, disable IPv6\n",
1487 				    if_name(ifp));
1488 			}
1489 			break;
1490 		}
1491 	}
1492 
1493 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1494 	free(dp, M_IP6NDP);
1495 	dp = NULL;
1496 	ifa_free(ifa);
1497 }
1498 
1499 static void
1500 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1501 {
1502 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1503 	struct ifnet *ifp = ifa->ifa_ifp;
1504 
1505 	dp->dad_ns_tcount++;
1506 	if ((ifp->if_flags & IFF_UP) == 0) {
1507 		return;
1508 	}
1509 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1510 		return;
1511 	}
1512 
1513 	dp->dad_ns_ocount++;
1514 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1515 }
1516 
1517 static void
1518 nd6_dad_ns_input(struct ifaddr *ifa)
1519 {
1520 	struct in6_ifaddr *ia;
1521 	struct ifnet *ifp;
1522 	const struct in6_addr *taddr6;
1523 	struct dadq *dp;
1524 	int duplicate;
1525 
1526 	if (ifa == NULL)
1527 		panic("ifa == NULL in nd6_dad_ns_input");
1528 
1529 	ia = (struct in6_ifaddr *)ifa;
1530 	ifp = ifa->ifa_ifp;
1531 	taddr6 = &ia->ia_addr.sin6_addr;
1532 	duplicate = 0;
1533 	dp = nd6_dad_find(ifa);
1534 
1535 	/* Quickhack - completely ignore DAD NS packets */
1536 	if (V_dad_ignore_ns) {
1537 		char ip6buf[INET6_ADDRSTRLEN];
1538 		nd6log((LOG_INFO,
1539 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1540 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1541 		    if_name(ifa->ifa_ifp)));
1542 		return;
1543 	}
1544 
1545 	/*
1546 	 * if I'm yet to start DAD, someone else started using this address
1547 	 * first.  I have a duplicate and you win.
1548 	 */
1549 	if (dp == NULL || dp->dad_ns_ocount == 0)
1550 		duplicate++;
1551 
1552 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1553 
1554 	if (duplicate) {
1555 		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1556 		nd6_dad_duplicated(ifa);
1557 	} else {
1558 		/*
1559 		 * not sure if I got a duplicate.
1560 		 * increment ns count and see what happens.
1561 		 */
1562 		if (dp)
1563 			dp->dad_ns_icount++;
1564 	}
1565 }
1566 
1567 static void
1568 nd6_dad_na_input(struct ifaddr *ifa)
1569 {
1570 	struct dadq *dp;
1571 
1572 	if (ifa == NULL)
1573 		panic("ifa == NULL in nd6_dad_na_input");
1574 
1575 	dp = nd6_dad_find(ifa);
1576 	if (dp)
1577 		dp->dad_na_icount++;
1578 
1579 	/* remove the address. */
1580 	nd6_dad_duplicated(ifa);
1581 }
1582