xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 5686c6c38a3e1cc78804eaf5f880bda23dcf592f)
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 	if (max_linkhdr + maxlen > MHLEN)
423 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
424 	else
425 		m = m_gethdr(M_NOWAIT, MT_DATA);
426 	if (m == NULL)
427 		return;
428 
429 	bzero(&ro, sizeof(ro));
430 
431 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
432 		m->m_flags |= M_MCAST;
433 		im6o.im6o_multicast_ifp = ifp;
434 		im6o.im6o_multicast_hlim = 255;
435 		im6o.im6o_multicast_loop = 0;
436 	}
437 
438 	icmp6len = sizeof(*nd_ns);
439 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
440 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
441 
442 	/* fill neighbor solicitation packet */
443 	ip6 = mtod(m, struct ip6_hdr *);
444 	ip6->ip6_flow = 0;
445 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
446 	ip6->ip6_vfc |= IPV6_VERSION;
447 	/* ip6->ip6_plen will be set later */
448 	ip6->ip6_nxt = IPPROTO_ICMPV6;
449 	ip6->ip6_hlim = 255;
450 	if (daddr6)
451 		ip6->ip6_dst = *daddr6;
452 	else {
453 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
454 		ip6->ip6_dst.s6_addr16[1] = 0;
455 		ip6->ip6_dst.s6_addr32[1] = 0;
456 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
457 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
458 		ip6->ip6_dst.s6_addr8[12] = 0xff;
459 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
460 			goto bad;
461 	}
462 	if (!dad) {
463 		struct ifaddr *ifa;
464 
465 		/*
466 		 * RFC2461 7.2.2:
467 		 * "If the source address of the packet prompting the
468 		 * solicitation is the same as one of the addresses assigned
469 		 * to the outgoing interface, that address SHOULD be placed
470 		 * in the IP Source Address of the outgoing solicitation.
471 		 * Otherwise, any one of the addresses assigned to the
472 		 * interface should be used."
473 		 *
474 		 * We use the source address for the prompting packet
475 		 * (saddr6), if:
476 		 * - saddr6 is given from the caller (by giving "ln"), and
477 		 * - saddr6 belongs to the outgoing interface.
478 		 * Otherwise, we perform the source address selection as usual.
479 		 */
480 		struct in6_addr *hsrc;
481 
482 		hsrc = NULL;
483 		if (ln != NULL) {
484 			LLE_RLOCK(ln);
485 			if (ln->la_hold != NULL) {
486 				struct ip6_hdr *hip6;		/* hold ip6 */
487 
488 				/*
489 				 * assuming every packet in la_hold has the same IP
490 				 * header
491 				 */
492 				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
493 				/* XXX pullup? */
494 				if (sizeof(*hip6) < ln->la_hold->m_len) {
495 					ip6->ip6_src = hip6->ip6_src;
496 					hsrc = &hip6->ip6_src;
497 				}
498 			}
499 			LLE_RUNLOCK(ln);
500 		}
501 		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
502 		    hsrc)) != NULL) {
503 			/* ip6_src set already. */
504 			ifa_free(ifa);
505 		} else {
506 			int error;
507 			struct sockaddr_in6 dst_sa;
508 			struct in6_addr src_in;
509 			struct ifnet *oifp;
510 
511 			bzero(&dst_sa, sizeof(dst_sa));
512 			dst_sa.sin6_family = AF_INET6;
513 			dst_sa.sin6_len = sizeof(dst_sa);
514 			dst_sa.sin6_addr = ip6->ip6_dst;
515 
516 			oifp = ifp;
517 			error = in6_selectsrc(&dst_sa, NULL,
518 			    NULL, &ro, NULL, &oifp, &src_in);
519 			if (error) {
520 				char ip6buf[INET6_ADDRSTRLEN];
521 				nd6log((LOG_DEBUG,
522 				    "nd6_ns_output: source can't be "
523 				    "determined: dst=%s, error=%d\n",
524 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
525 				    error));
526 				goto bad;
527 			}
528 			ip6->ip6_src = src_in;
529 		}
530 	} else {
531 		/*
532 		 * Source address for DAD packet must always be IPv6
533 		 * unspecified address. (0::0)
534 		 * We actually don't have to 0-clear the address (we did it
535 		 * above), but we do so here explicitly to make the intention
536 		 * clearer.
537 		 */
538 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
539 	}
540 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
541 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
542 	nd_ns->nd_ns_code = 0;
543 	nd_ns->nd_ns_reserved = 0;
544 	nd_ns->nd_ns_target = *taddr6;
545 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
546 
547 	/*
548 	 * Add source link-layer address option.
549 	 *
550 	 *				spec		implementation
551 	 *				---		---
552 	 * DAD packet			MUST NOT	do not add the option
553 	 * there's no link layer address:
554 	 *				impossible	do not add the option
555 	 * there's link layer address:
556 	 *	Multicast NS		MUST add one	add the option
557 	 *	Unicast NS		SHOULD add one	add the option
558 	 */
559 	if (!dad && (mac = nd6_ifptomac(ifp))) {
560 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
561 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
562 		/* 8 byte alignments... */
563 		optlen = (optlen + 7) & ~7;
564 
565 		m->m_pkthdr.len += optlen;
566 		m->m_len += optlen;
567 		icmp6len += optlen;
568 		bzero((caddr_t)nd_opt, optlen);
569 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
570 		nd_opt->nd_opt_len = optlen >> 3;
571 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
572 	}
573 
574 	ip6->ip6_plen = htons((u_short)icmp6len);
575 	nd_ns->nd_ns_cksum = 0;
576 	nd_ns->nd_ns_cksum =
577 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
578 
579 	if (send_sendso_input_hook != NULL) {
580 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
581 			sizeof(unsigned short), M_NOWAIT);
582 		if (mtag == NULL)
583 			goto bad;
584 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
585 		m_tag_prepend(m, mtag);
586 	}
587 
588 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
589 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
590 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
591 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
592 
593 	/* We don't cache this route. */
594 	RO_RTFREE(&ro);
595 
596 	return;
597 
598   bad:
599 	if (ro.ro_rt) {
600 		RTFREE(ro.ro_rt);
601 	}
602 	m_freem(m);
603 	return;
604 }
605 
606 /*
607  * Neighbor advertisement input handling.
608  *
609  * Based on RFC 2461
610  * Based on RFC 2462 (duplicate address detection)
611  *
612  * the following items are not implemented yet:
613  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
614  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
615  */
616 void
617 nd6_na_input(struct mbuf *m, int off, int icmp6len)
618 {
619 	struct ifnet *ifp = m->m_pkthdr.rcvif;
620 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
621 	struct nd_neighbor_advert *nd_na;
622 	struct in6_addr daddr6 = ip6->ip6_dst;
623 	struct in6_addr taddr6;
624 	int flags;
625 	int is_router;
626 	int is_solicited;
627 	int is_override;
628 	char *lladdr = NULL;
629 	int lladdrlen = 0;
630 	int checklink = 0;
631 	struct ifaddr *ifa;
632 	struct llentry *ln = NULL;
633 	union nd_opts ndopts;
634 	struct mbuf *chain = NULL;
635 	struct m_tag *mtag;
636 	struct sockaddr_in6 sin6;
637 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
638 
639 	if (ip6->ip6_hlim != 255) {
640 		nd6log((LOG_ERR,
641 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
642 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
643 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
644 		goto bad;
645 	}
646 
647 #ifndef PULLDOWN_TEST
648 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
649 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
650 #else
651 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
652 	if (nd_na == NULL) {
653 		ICMP6STAT_INC(icp6s_tooshort);
654 		return;
655 	}
656 #endif
657 
658 	flags = nd_na->nd_na_flags_reserved;
659 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
660 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
661 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
662 
663 	taddr6 = nd_na->nd_na_target;
664 	if (in6_setscope(&taddr6, ifp, NULL))
665 		goto bad;	/* XXX: impossible */
666 
667 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
668 		nd6log((LOG_ERR,
669 		    "nd6_na_input: invalid target address %s\n",
670 		    ip6_sprintf(ip6bufs, &taddr6)));
671 		goto bad;
672 	}
673 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
674 		if (is_solicited) {
675 			nd6log((LOG_ERR,
676 			    "nd6_na_input: a solicited adv is multicasted\n"));
677 			goto bad;
678 		}
679 
680 	icmp6len -= sizeof(*nd_na);
681 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
682 	if (nd6_options(&ndopts) < 0) {
683 		nd6log((LOG_INFO,
684 		    "nd6_na_input: invalid ND option, ignored\n"));
685 		/* nd6_options have incremented stats */
686 		goto freeit;
687 	}
688 
689 	if (ndopts.nd_opts_tgt_lladdr) {
690 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
691 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
692 	}
693 
694 	/*
695 	 * This effectively disables the DAD check on a non-master CARP
696 	 * address.
697 	 */
698 	if (ifp->if_carp)
699 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
700 	else
701 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
702 
703 	/*
704 	 * Target address matches one of my interface address.
705 	 *
706 	 * If my address is tentative, this means that there's somebody
707 	 * already using the same address as mine.  This indicates DAD failure.
708 	 * This is defined in RFC 2462.
709 	 *
710 	 * Otherwise, process as defined in RFC 2461.
711 	 */
712 	if (ifa
713 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
714 		ifa_free(ifa);
715 		nd6_dad_na_input(ifa);
716 		goto freeit;
717 	}
718 
719 	/* Just for safety, maybe unnecessary. */
720 	if (ifa) {
721 		ifa_free(ifa);
722 		log(LOG_ERR,
723 		    "nd6_na_input: duplicate IP6 address %s\n",
724 		    ip6_sprintf(ip6bufs, &taddr6));
725 		goto freeit;
726 	}
727 
728 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
729 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
730 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
731 		    ifp->if_addrlen, lladdrlen - 2));
732 		goto bad;
733 	}
734 
735 	/*
736 	 * If no neighbor cache entry is found, NA SHOULD silently be
737 	 * discarded.
738 	 */
739 	IF_AFDATA_LOCK(ifp);
740 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
741 	IF_AFDATA_UNLOCK(ifp);
742 	if (ln == NULL) {
743 		goto freeit;
744 	}
745 
746 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
747 		/*
748 		 * If the link-layer has address, and no lladdr option came,
749 		 * discard the packet.
750 		 */
751 		if (ifp->if_addrlen && lladdr == NULL) {
752 			goto freeit;
753 		}
754 
755 		/*
756 		 * Record link-layer address, and update the state.
757 		 */
758 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
759 		ln->la_flags |= LLE_VALID;
760 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
761 		if (is_solicited) {
762 			ln->ln_state = ND6_LLINFO_REACHABLE;
763 			ln->ln_byhint = 0;
764 			if (!ND6_LLINFO_PERMANENT(ln)) {
765 				nd6_llinfo_settimer_locked(ln,
766 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
767 			}
768 		} else {
769 			ln->ln_state = ND6_LLINFO_STALE;
770 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
771 		}
772 		if ((ln->ln_router = is_router) != 0) {
773 			/*
774 			 * This means a router's state has changed from
775 			 * non-reachable to probably reachable, and might
776 			 * affect the status of associated prefixes..
777 			 */
778 			checklink = 1;
779 		}
780 	} else {
781 		int llchange;
782 
783 		/*
784 		 * Check if the link-layer address has changed or not.
785 		 */
786 		if (lladdr == NULL)
787 			llchange = 0;
788 		else {
789 			if (ln->la_flags & LLE_VALID) {
790 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
791 					llchange = 1;
792 				else
793 					llchange = 0;
794 			} else
795 				llchange = 1;
796 		}
797 
798 		/*
799 		 * This is VERY complex.  Look at it with care.
800 		 *
801 		 * override solicit lladdr llchange	action
802 		 *					(L: record lladdr)
803 		 *
804 		 *	0	0	n	--	(2c)
805 		 *	0	0	y	n	(2b) L
806 		 *	0	0	y	y	(1)    REACHABLE->STALE
807 		 *	0	1	n	--	(2c)   *->REACHABLE
808 		 *	0	1	y	n	(2b) L *->REACHABLE
809 		 *	0	1	y	y	(1)    REACHABLE->STALE
810 		 *	1	0	n	--	(2a)
811 		 *	1	0	y	n	(2a) L
812 		 *	1	0	y	y	(2a) L *->STALE
813 		 *	1	1	n	--	(2a)   *->REACHABLE
814 		 *	1	1	y	n	(2a) L *->REACHABLE
815 		 *	1	1	y	y	(2a) L *->REACHABLE
816 		 */
817 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
818 			/*
819 			 * If state is REACHABLE, make it STALE.
820 			 * no other updates should be done.
821 			 */
822 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
823 				ln->ln_state = ND6_LLINFO_STALE;
824 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
825 			}
826 			goto freeit;
827 		} else if (is_override				   /* (2a) */
828 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
829 			|| lladdr == NULL) {			   /* (2c) */
830 			/*
831 			 * Update link-local address, if any.
832 			 */
833 			if (lladdr != NULL) {
834 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
835 				ln->la_flags |= LLE_VALID;
836 				EVENTHANDLER_INVOKE(lle_event, ln,
837 				    LLENTRY_RESOLVED);
838 			}
839 
840 			/*
841 			 * If solicited, make the state REACHABLE.
842 			 * If not solicited and the link-layer address was
843 			 * changed, make it STALE.
844 			 */
845 			if (is_solicited) {
846 				ln->ln_state = ND6_LLINFO_REACHABLE;
847 				ln->ln_byhint = 0;
848 				if (!ND6_LLINFO_PERMANENT(ln)) {
849 					nd6_llinfo_settimer_locked(ln,
850 					    (long)ND_IFINFO(ifp)->reachable * hz);
851 				}
852 			} else {
853 				if (lladdr != NULL && llchange) {
854 					ln->ln_state = ND6_LLINFO_STALE;
855 					nd6_llinfo_settimer_locked(ln,
856 					    (long)V_nd6_gctimer * hz);
857 				}
858 			}
859 		}
860 
861 		if (ln->ln_router && !is_router) {
862 			/*
863 			 * The peer dropped the router flag.
864 			 * Remove the sender from the Default Router List and
865 			 * update the Destination Cache entries.
866 			 */
867 			struct nd_defrouter *dr;
868 			struct in6_addr *in6;
869 
870 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
871 
872 			/*
873 			 * Lock to protect the default router list.
874 			 * XXX: this might be unnecessary, since this function
875 			 * is only called under the network software interrupt
876 			 * context.  However, we keep it just for safety.
877 			 */
878 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
879 			if (dr)
880 				defrtrlist_del(dr);
881 			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
882 			    ND6_IFF_ACCEPT_RTADV) {
883 				/*
884 				 * Even if the neighbor is not in the default
885 				 * router list, the neighbor may be used
886 				 * as a next hop for some destinations
887 				 * (e.g. redirect case). So we must
888 				 * call rt6_flush explicitly.
889 				 */
890 				rt6_flush(&ip6->ip6_src, ifp);
891 			}
892 		}
893 		ln->ln_router = is_router;
894 	}
895         /* XXX - QL
896 	 *  Does this matter?
897 	 *  rt->rt_flags &= ~RTF_REJECT;
898 	 */
899 	ln->la_asked = 0;
900 	if (ln->la_hold) {
901 		struct mbuf *m_hold, *m_hold_next;
902 
903 		/*
904 		 * reset the la_hold in advance, to explicitly
905 		 * prevent a la_hold lookup in nd6_output()
906 		 * (wouldn't happen, though...)
907 		 */
908 		for (m_hold = ln->la_hold, ln->la_hold = NULL;
909 		    m_hold; m_hold = m_hold_next) {
910 			m_hold_next = m_hold->m_nextpkt;
911 			m_hold->m_nextpkt = NULL;
912 			/*
913 			 * we assume ifp is not a loopback here, so just set
914 			 * the 2nd argument as the 1st one.
915 			 */
916 
917 			if (send_sendso_input_hook != NULL) {
918 				mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
919 				    sizeof(unsigned short), M_NOWAIT);
920 				if (mtag == NULL)
921 					goto bad;
922 				m_tag_prepend(m, mtag);
923 			}
924 
925 			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
926 		}
927 	}
928  freeit:
929 	if (ln != NULL) {
930 		if (chain)
931 			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
932 		LLE_WUNLOCK(ln);
933 
934 		if (chain)
935 			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
936 	}
937 	if (checklink)
938 		pfxlist_onlink_check();
939 
940 	m_freem(m);
941 	return;
942 
943  bad:
944 	if (ln != NULL)
945 		LLE_WUNLOCK(ln);
946 
947 	ICMP6STAT_INC(icp6s_badna);
948 	m_freem(m);
949 }
950 
951 /*
952  * Neighbor advertisement output handling.
953  *
954  * Based on RFC 2461
955  *
956  * the following items are not implemented yet:
957  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
958  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
959  *
960  * tlladdr - 1 if include target link-layer address
961  * sdl0 - sockaddr_dl (= proxy NA) or NULL
962  */
963 static void
964 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
965     const struct in6_addr *taddr6, u_long flags, int tlladdr,
966     struct sockaddr *sdl0, u_int fibnum)
967 {
968 	struct mbuf *m;
969 	struct m_tag *mtag;
970 	struct ifnet *oifp;
971 	struct ip6_hdr *ip6;
972 	struct nd_neighbor_advert *nd_na;
973 	struct ip6_moptions im6o;
974 	struct in6_addr src, daddr6;
975 	struct sockaddr_in6 dst_sa;
976 	int icmp6len, maxlen, error;
977 	caddr_t mac = NULL;
978 	struct route_in6 ro;
979 
980 	bzero(&ro, sizeof(ro));
981 
982 	daddr6 = *daddr6_0;	/* make a local copy for modification */
983 
984 	/* estimate the size of message */
985 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
986 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
987 	if (max_linkhdr + maxlen >= MCLBYTES) {
988 #ifdef DIAGNOSTIC
989 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
990 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
991 #endif
992 		return;
993 	}
994 
995 	if (max_linkhdr + maxlen > MHLEN)
996 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
997 	else
998 		m = m_gethdr(M_NOWAIT, MT_DATA);
999 	if (m == NULL)
1000 		return;
1001 	M_SETFIB(m, fibnum);
1002 
1003 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1004 		m->m_flags |= M_MCAST;
1005 		im6o.im6o_multicast_ifp = ifp;
1006 		im6o.im6o_multicast_hlim = 255;
1007 		im6o.im6o_multicast_loop = 0;
1008 	}
1009 
1010 	icmp6len = sizeof(*nd_na);
1011 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1012 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
1013 
1014 	/* fill neighbor advertisement packet */
1015 	ip6 = mtod(m, struct ip6_hdr *);
1016 	ip6->ip6_flow = 0;
1017 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1018 	ip6->ip6_vfc |= IPV6_VERSION;
1019 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1020 	ip6->ip6_hlim = 255;
1021 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1022 		/* reply to DAD */
1023 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1024 		daddr6.s6_addr16[1] = 0;
1025 		daddr6.s6_addr32[1] = 0;
1026 		daddr6.s6_addr32[2] = 0;
1027 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1028 		if (in6_setscope(&daddr6, ifp, NULL))
1029 			goto bad;
1030 
1031 		flags &= ~ND_NA_FLAG_SOLICITED;
1032 	}
1033 	ip6->ip6_dst = daddr6;
1034 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1035 	dst_sa.sin6_family = AF_INET6;
1036 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1037 	dst_sa.sin6_addr = daddr6;
1038 
1039 	/*
1040 	 * Select a source whose scope is the same as that of the dest.
1041 	 */
1042 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1043 	oifp = ifp;
1044 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1045 	if (error) {
1046 		char ip6buf[INET6_ADDRSTRLEN];
1047 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1048 		    "determined: dst=%s, error=%d\n",
1049 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1050 		goto bad;
1051 	}
1052 	ip6->ip6_src = src;
1053 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1054 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1055 	nd_na->nd_na_code = 0;
1056 	nd_na->nd_na_target = *taddr6;
1057 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1058 
1059 	/*
1060 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1061 	 * see nd6_ns_input() for details.
1062 	 * Basically, if NS packet is sent to unicast/anycast addr,
1063 	 * target lladdr option SHOULD NOT be included.
1064 	 */
1065 	if (tlladdr) {
1066 		/*
1067 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1068 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1069 		 * my address) use lladdr configured for the interface.
1070 		 */
1071 		if (sdl0 == NULL) {
1072 			if (ifp->if_carp)
1073 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1074 			if (mac == NULL)
1075 				mac = nd6_ifptomac(ifp);
1076 		} else if (sdl0->sa_family == AF_LINK) {
1077 			struct sockaddr_dl *sdl;
1078 			sdl = (struct sockaddr_dl *)sdl0;
1079 			if (sdl->sdl_alen == ifp->if_addrlen)
1080 				mac = LLADDR(sdl);
1081 		}
1082 	}
1083 	if (tlladdr && mac) {
1084 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1085 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1086 
1087 		/* roundup to 8 bytes alignment! */
1088 		optlen = (optlen + 7) & ~7;
1089 
1090 		m->m_pkthdr.len += optlen;
1091 		m->m_len += optlen;
1092 		icmp6len += optlen;
1093 		bzero((caddr_t)nd_opt, optlen);
1094 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1095 		nd_opt->nd_opt_len = optlen >> 3;
1096 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1097 	} else
1098 		flags &= ~ND_NA_FLAG_OVERRIDE;
1099 
1100 	ip6->ip6_plen = htons((u_short)icmp6len);
1101 	nd_na->nd_na_flags_reserved = flags;
1102 	nd_na->nd_na_cksum = 0;
1103 	nd_na->nd_na_cksum =
1104 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1105 
1106 	if (send_sendso_input_hook != NULL) {
1107 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1108 		    sizeof(unsigned short), M_NOWAIT);
1109 		if (mtag == NULL)
1110 			goto bad;
1111 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1112 		m_tag_prepend(m, mtag);
1113 	}
1114 
1115 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1116 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1117 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1118 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1119 
1120 	/* We don't cache this route. */
1121 	RO_RTFREE(&ro);
1122 
1123 	return;
1124 
1125   bad:
1126 	if (ro.ro_rt) {
1127 		RTFREE(ro.ro_rt);
1128 	}
1129 	m_freem(m);
1130 	return;
1131 }
1132 
1133 #ifndef BURN_BRIDGES
1134 void
1135 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1136     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1137     struct sockaddr *sdl0)
1138 {
1139 
1140 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1141 	    RT_DEFAULT_FIB);
1142 }
1143 #endif
1144 
1145 caddr_t
1146 nd6_ifptomac(struct ifnet *ifp)
1147 {
1148 	switch (ifp->if_type) {
1149 	case IFT_ARCNET:
1150 	case IFT_ETHER:
1151 	case IFT_FDDI:
1152 	case IFT_IEEE1394:
1153 #ifdef IFT_L2VLAN
1154 	case IFT_L2VLAN:
1155 #endif
1156 #ifdef IFT_IEEE80211
1157 	case IFT_IEEE80211:
1158 #endif
1159 	case IFT_INFINIBAND:
1160 	case IFT_BRIDGE:
1161 	case IFT_ISO88025:
1162 		return IF_LLADDR(ifp);
1163 	default:
1164 		return NULL;
1165 	}
1166 }
1167 
1168 struct dadq {
1169 	TAILQ_ENTRY(dadq) dad_list;
1170 	struct ifaddr *dad_ifa;
1171 	int dad_count;		/* max NS to send */
1172 	int dad_ns_tcount;	/* # of trials to send NS */
1173 	int dad_ns_ocount;	/* NS sent so far */
1174 	int dad_ns_icount;
1175 	int dad_na_icount;
1176 	struct callout dad_timer_ch;
1177 	struct vnet *dad_vnet;
1178 };
1179 
1180 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1181 VNET_DEFINE(int, dad_init) = 0;
1182 #define	V_dadq				VNET(dadq)
1183 #define	V_dad_init			VNET(dad_init)
1184 
1185 static struct dadq *
1186 nd6_dad_find(struct ifaddr *ifa)
1187 {
1188 	struct dadq *dp;
1189 
1190 	TAILQ_FOREACH(dp, &V_dadq, dad_list)
1191 		if (dp->dad_ifa == ifa)
1192 			return (dp);
1193 
1194 	return (NULL);
1195 }
1196 
1197 static void
1198 nd6_dad_starttimer(struct dadq *dp, int ticks)
1199 {
1200 
1201 	callout_reset(&dp->dad_timer_ch, ticks,
1202 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1203 }
1204 
1205 static void
1206 nd6_dad_stoptimer(struct dadq *dp)
1207 {
1208 
1209 	callout_stop(&dp->dad_timer_ch);
1210 }
1211 
1212 /*
1213  * Start Duplicate Address Detection (DAD) for specified interface address.
1214  */
1215 void
1216 nd6_dad_start(struct ifaddr *ifa, int delay)
1217 {
1218 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1219 	struct dadq *dp;
1220 	char ip6buf[INET6_ADDRSTRLEN];
1221 
1222 	if (!V_dad_init) {
1223 		TAILQ_INIT(&V_dadq);
1224 		V_dad_init++;
1225 	}
1226 
1227 	/*
1228 	 * If we don't need DAD, don't do it.
1229 	 * There are several cases:
1230 	 * - DAD is disabled (ip6_dad_count == 0)
1231 	 * - the interface address is anycast
1232 	 */
1233 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1234 		log(LOG_DEBUG,
1235 			"nd6_dad_start: called with non-tentative address "
1236 			"%s(%s)\n",
1237 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1238 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1239 		return;
1240 	}
1241 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1242 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1243 		return;
1244 	}
1245 	if (!V_ip6_dad_count) {
1246 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1247 		return;
1248 	}
1249 	if (ifa->ifa_ifp == NULL)
1250 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1251 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1252 		return;
1253 	}
1254 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1255 		return;
1256 	if (nd6_dad_find(ifa) != NULL) {
1257 		/* DAD already in progress */
1258 		return;
1259 	}
1260 
1261 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1262 	if (dp == NULL) {
1263 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1264 			"%s(%s)\n",
1265 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1266 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1267 		return;
1268 	}
1269 	bzero(dp, sizeof(*dp));
1270 	callout_init(&dp->dad_timer_ch, 0);
1271 #ifdef VIMAGE
1272 	dp->dad_vnet = curvnet;
1273 #endif
1274 	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1275 
1276 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1277 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1278 
1279 	/*
1280 	 * Send NS packet for DAD, ip6_dad_count times.
1281 	 * Note that we must delay the first transmission, if this is the
1282 	 * first packet to be sent from the interface after interface
1283 	 * (re)initialization.
1284 	 */
1285 	dp->dad_ifa = ifa;
1286 	ifa_ref(ifa);	/* just for safety */
1287 	dp->dad_count = V_ip6_dad_count;
1288 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1289 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1290 	if (delay == 0) {
1291 		nd6_dad_ns_output(dp, ifa);
1292 		nd6_dad_starttimer(dp,
1293 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1294 	} else {
1295 		nd6_dad_starttimer(dp, delay);
1296 	}
1297 }
1298 
1299 /*
1300  * terminate DAD unconditionally.  used for address removals.
1301  */
1302 void
1303 nd6_dad_stop(struct ifaddr *ifa)
1304 {
1305 	struct dadq *dp;
1306 
1307 	if (!V_dad_init)
1308 		return;
1309 	dp = nd6_dad_find(ifa);
1310 	if (!dp) {
1311 		/* DAD wasn't started yet */
1312 		return;
1313 	}
1314 
1315 	nd6_dad_stoptimer(dp);
1316 
1317 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1318 	free(dp, M_IP6NDP);
1319 	dp = NULL;
1320 	ifa_free(ifa);
1321 }
1322 
1323 static void
1324 nd6_dad_timer(struct dadq *dp)
1325 {
1326 	CURVNET_SET(dp->dad_vnet);
1327 	struct ifaddr *ifa = dp->dad_ifa;
1328 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1329 	char ip6buf[INET6_ADDRSTRLEN];
1330 
1331 	/* Sanity check */
1332 	if (ia == NULL) {
1333 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1334 		goto done;
1335 	}
1336 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1337 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1338 			"%s(%s)\n",
1339 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1340 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1341 		goto done;
1342 	}
1343 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1344 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative 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 
1351 	/* timeouted with IFF_{RUNNING,UP} check */
1352 	if (dp->dad_ns_tcount > V_dad_maxtry) {
1353 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1354 		    if_name(ifa->ifa_ifp)));
1355 
1356 		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1357 		free(dp, M_IP6NDP);
1358 		dp = NULL;
1359 		ifa_free(ifa);
1360 		goto done;
1361 	}
1362 
1363 	/* Need more checks? */
1364 	if (dp->dad_ns_ocount < dp->dad_count) {
1365 		/*
1366 		 * We have more NS to go.  Send NS packet for DAD.
1367 		 */
1368 		nd6_dad_ns_output(dp, ifa);
1369 		nd6_dad_starttimer(dp,
1370 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1371 	} else {
1372 		/*
1373 		 * We have transmitted sufficient number of DAD packets.
1374 		 * See what we've got.
1375 		 */
1376 		int duplicate;
1377 
1378 		duplicate = 0;
1379 
1380 		if (dp->dad_na_icount) {
1381 			/*
1382 			 * the check is in nd6_dad_na_input(),
1383 			 * but just in case
1384 			 */
1385 			duplicate++;
1386 		}
1387 
1388 		if (dp->dad_ns_icount) {
1389 			/* We've seen NS, means DAD has failed. */
1390 			duplicate++;
1391 		}
1392 
1393 		if (duplicate) {
1394 			/* (*dp) will be freed in nd6_dad_duplicated() */
1395 			dp = NULL;
1396 			nd6_dad_duplicated(ifa);
1397 		} else {
1398 			/*
1399 			 * We are done with DAD.  No NA came, no NS came.
1400 			 * No duplicate address found.
1401 			 */
1402 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1403 
1404 			nd6log((LOG_DEBUG,
1405 			    "%s: DAD complete for %s - no duplicates found\n",
1406 			    if_name(ifa->ifa_ifp),
1407 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1408 
1409 			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1410 			free(dp, M_IP6NDP);
1411 			dp = NULL;
1412 			ifa_free(ifa);
1413 		}
1414 	}
1415 
1416 done:
1417 	CURVNET_RESTORE();
1418 }
1419 
1420 void
1421 nd6_dad_duplicated(struct ifaddr *ifa)
1422 {
1423 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1424 	struct ifnet *ifp;
1425 	struct dadq *dp;
1426 	char ip6buf[INET6_ADDRSTRLEN];
1427 
1428 	dp = nd6_dad_find(ifa);
1429 	if (dp == NULL) {
1430 		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1431 		return;
1432 	}
1433 
1434 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1435 	    "NS in/out=%d/%d, NA in=%d\n",
1436 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1437 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1438 
1439 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1440 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1441 
1442 	/* We are done with DAD, with duplicate address found. (failure) */
1443 	nd6_dad_stoptimer(dp);
1444 
1445 	ifp = ifa->ifa_ifp;
1446 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1447 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1448 	log(LOG_ERR, "%s: manual intervention required\n",
1449 	    if_name(ifp));
1450 
1451 	/*
1452 	 * If the address is a link-local address formed from an interface
1453 	 * identifier based on the hardware address which is supposed to be
1454 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1455 	 * operation on the interface SHOULD be disabled.
1456 	 * [RFC 4862, Section 5.4.5]
1457 	 */
1458 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1459 		struct in6_addr in6;
1460 
1461 		/*
1462 		 * To avoid over-reaction, we only apply this logic when we are
1463 		 * very sure that hardware addresses are supposed to be unique.
1464 		 */
1465 		switch (ifp->if_type) {
1466 		case IFT_ETHER:
1467 		case IFT_FDDI:
1468 		case IFT_ATM:
1469 		case IFT_IEEE1394:
1470 #ifdef IFT_IEEE80211
1471 		case IFT_IEEE80211:
1472 #endif
1473 		case IFT_INFINIBAND:
1474 			in6 = ia->ia_addr.sin6_addr;
1475 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1476 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1477 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1478 				log(LOG_ERR, "%s: possible hardware address "
1479 				    "duplication detected, disable IPv6\n",
1480 				    if_name(ifp));
1481 			}
1482 			break;
1483 		}
1484 	}
1485 
1486 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1487 	free(dp, M_IP6NDP);
1488 	dp = NULL;
1489 	ifa_free(ifa);
1490 }
1491 
1492 static void
1493 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1494 {
1495 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1496 	struct ifnet *ifp = ifa->ifa_ifp;
1497 
1498 	dp->dad_ns_tcount++;
1499 	if ((ifp->if_flags & IFF_UP) == 0) {
1500 		return;
1501 	}
1502 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1503 		return;
1504 	}
1505 
1506 	dp->dad_ns_ocount++;
1507 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1508 }
1509 
1510 static void
1511 nd6_dad_ns_input(struct ifaddr *ifa)
1512 {
1513 	struct in6_ifaddr *ia;
1514 	struct ifnet *ifp;
1515 	const struct in6_addr *taddr6;
1516 	struct dadq *dp;
1517 	int duplicate;
1518 
1519 	if (ifa == NULL)
1520 		panic("ifa == NULL in nd6_dad_ns_input");
1521 
1522 	ia = (struct in6_ifaddr *)ifa;
1523 	ifp = ifa->ifa_ifp;
1524 	taddr6 = &ia->ia_addr.sin6_addr;
1525 	duplicate = 0;
1526 	dp = nd6_dad_find(ifa);
1527 
1528 	/* Quickhack - completely ignore DAD NS packets */
1529 	if (V_dad_ignore_ns) {
1530 		char ip6buf[INET6_ADDRSTRLEN];
1531 		nd6log((LOG_INFO,
1532 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1533 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1534 		    if_name(ifa->ifa_ifp)));
1535 		return;
1536 	}
1537 
1538 	/*
1539 	 * if I'm yet to start DAD, someone else started using this address
1540 	 * first.  I have a duplicate and you win.
1541 	 */
1542 	if (dp == NULL || dp->dad_ns_ocount == 0)
1543 		duplicate++;
1544 
1545 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1546 
1547 	if (duplicate) {
1548 		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1549 		nd6_dad_duplicated(ifa);
1550 	} else {
1551 		/*
1552 		 * not sure if I got a duplicate.
1553 		 * increment ns count and see what happens.
1554 		 */
1555 		if (dp)
1556 			dp->dad_ns_icount++;
1557 	}
1558 }
1559 
1560 static void
1561 nd6_dad_na_input(struct ifaddr *ifa)
1562 {
1563 	struct dadq *dp;
1564 
1565 	if (ifa == NULL)
1566 		panic("ifa == NULL in nd6_dad_na_input");
1567 
1568 	dp = nd6_dad_find(ifa);
1569 	if (dp)
1570 		dp->dad_na_icount++;
1571 
1572 	/* remove the address. */
1573 	nd6_dad_duplicated(ifa);
1574 }
1575