xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 9ecd54f24fe9fa373e07c9fd7c052deb2188f545)
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 route_in6 ro;
236 		int need_proxy;
237 
238 		bzero(&ro, sizeof(ro));
239 		ro.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
240 		ro.ro_dst.sin6_family = AF_INET6;
241 		ro.ro_dst.sin6_addr = taddr6;
242 
243 		/* Always use the default FIB. */
244 #ifdef RADIX_MPATH
245 		rtalloc_mpath_fib((struct route *)&ro, RTF_ANNOUNCE,
246 		    RT_DEFAULT_FIB);
247 #else
248 		in6_rtalloc(&ro, RT_DEFAULT_FIB);
249 #endif
250 		need_proxy = (ro.ro_rt &&
251 		    (ro.ro_rt->rt_flags & RTF_ANNOUNCE) != 0 &&
252 		    ro.ro_rt->rt_gateway->sa_family == AF_LINK);
253 		if (ro.ro_rt != NULL) {
254 			if (need_proxy)
255 				proxydl = *SDL(ro.ro_rt->rt_gateway);
256 			RTFREE(ro.ro_rt);
257 		}
258 		if (need_proxy) {
259 			/*
260 			 * proxy NDP for single entry
261 			 */
262 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
263 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
264 			if (ifa)
265 				proxy = 1;
266 		}
267 	}
268 	if (ifa == NULL) {
269 		/*
270 		 * We've got an NS packet, and we don't have that adddress
271 		 * assigned for us.  We MUST silently ignore it.
272 		 * See RFC2461 7.2.3.
273 		 */
274 		goto freeit;
275 	}
276 	myaddr6 = *IFA_IN6(ifa);
277 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
278 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
279 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
280 		goto freeit;
281 
282 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
283 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
284 		    "(if %d, NS packet %d)\n",
285 		    ip6_sprintf(ip6bufs, &taddr6),
286 		    ifp->if_addrlen, lladdrlen - 2));
287 		goto bad;
288 	}
289 
290 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
291 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
292 		    ip6_sprintf(ip6bufs, &saddr6)));
293 		goto freeit;
294 	}
295 
296 	/*
297 	 * We have neighbor solicitation packet, with target address equals to
298 	 * one of my tentative address.
299 	 *
300 	 * src addr	how to process?
301 	 * ---		---
302 	 * multicast	of course, invalid (rejected in ip6_input)
303 	 * unicast	somebody is doing address resolution -> ignore
304 	 * unspec	dup address detection
305 	 *
306 	 * The processing is defined in RFC 2462.
307 	 */
308 	if (tentative) {
309 		/*
310 		 * If source address is unspecified address, it is for
311 		 * duplicate address detection.
312 		 *
313 		 * If not, the packet is for addess resolution;
314 		 * silently ignore it.
315 		 */
316 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
317 			nd6_dad_ns_input(ifa);
318 
319 		goto freeit;
320 	}
321 
322 	/*
323 	 * If the source address is unspecified address, entries must not
324 	 * be created or updated.
325 	 * It looks that sender is performing DAD.  Output NA toward
326 	 * all-node multicast address, to tell the sender that I'm using
327 	 * the address.
328 	 * S bit ("solicited") must be zero.
329 	 */
330 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
331 		struct in6_addr in6_all;
332 
333 		in6_all = in6addr_linklocal_allnodes;
334 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
335 			goto bad;
336 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
337 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
338 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
339 		    M_GETFIB(m));
340 		goto freeit;
341 	}
342 
343 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
344 	    ND_NEIGHBOR_SOLICIT, 0);
345 
346 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
347 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
348 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
349 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
350  freeit:
351 	if (ifa != NULL)
352 		ifa_free(ifa);
353 	m_freem(m);
354 	return;
355 
356  bad:
357 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
358 		ip6_sprintf(ip6bufs, &saddr6)));
359 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
360 		ip6_sprintf(ip6bufs, &daddr6)));
361 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
362 		ip6_sprintf(ip6bufs, &taddr6)));
363 	ICMP6STAT_INC(icp6s_badns);
364 	if (ifa != NULL)
365 		ifa_free(ifa);
366 	m_freem(m);
367 }
368 
369 /*
370  * Output a Neighbor Solicitation Message. Caller specifies:
371  *	- ICMP6 header source IP6 address
372  *	- ND6 header target IP6 address
373  *	- ND6 header source datalink address
374  *
375  * Based on RFC 2461
376  * Based on RFC 2462 (duplicate address detection)
377  *
378  *   ln - for source address determination
379  *  dad - duplicate address detection
380  */
381 void
382 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
383     const struct in6_addr *taddr6, struct llentry *ln, int dad)
384 {
385 	struct mbuf *m;
386 	struct m_tag *mtag;
387 	struct ip6_hdr *ip6;
388 	struct nd_neighbor_solicit *nd_ns;
389 	struct ip6_moptions im6o;
390 	int icmp6len;
391 	int maxlen;
392 	caddr_t mac;
393 	struct route_in6 ro;
394 
395 	if (IN6_IS_ADDR_MULTICAST(taddr6))
396 		return;
397 
398 	/* estimate the size of message */
399 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
400 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
401 	if (max_linkhdr + maxlen >= MCLBYTES) {
402 #ifdef DIAGNOSTIC
403 		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
404 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
405 #endif
406 		return;
407 	}
408 
409 	if (max_linkhdr + maxlen > MHLEN)
410 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
411 	else
412 		m = m_gethdr(M_NOWAIT, MT_DATA);
413 	if (m == NULL)
414 		return;
415 
416 	bzero(&ro, sizeof(ro));
417 
418 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
419 		m->m_flags |= M_MCAST;
420 		im6o.im6o_multicast_ifp = ifp;
421 		im6o.im6o_multicast_hlim = 255;
422 		im6o.im6o_multicast_loop = 0;
423 	}
424 
425 	icmp6len = sizeof(*nd_ns);
426 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
427 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
428 
429 	/* fill neighbor solicitation packet */
430 	ip6 = mtod(m, struct ip6_hdr *);
431 	ip6->ip6_flow = 0;
432 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
433 	ip6->ip6_vfc |= IPV6_VERSION;
434 	/* ip6->ip6_plen will be set later */
435 	ip6->ip6_nxt = IPPROTO_ICMPV6;
436 	ip6->ip6_hlim = 255;
437 	if (daddr6)
438 		ip6->ip6_dst = *daddr6;
439 	else {
440 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
441 		ip6->ip6_dst.s6_addr16[1] = 0;
442 		ip6->ip6_dst.s6_addr32[1] = 0;
443 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
444 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
445 		ip6->ip6_dst.s6_addr8[12] = 0xff;
446 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
447 			goto bad;
448 	}
449 	if (!dad) {
450 		struct ifaddr *ifa;
451 
452 		/*
453 		 * RFC2461 7.2.2:
454 		 * "If the source address of the packet prompting the
455 		 * solicitation is the same as one of the addresses assigned
456 		 * to the outgoing interface, that address SHOULD be placed
457 		 * in the IP Source Address of the outgoing solicitation.
458 		 * Otherwise, any one of the addresses assigned to the
459 		 * interface should be used."
460 		 *
461 		 * We use the source address for the prompting packet
462 		 * (saddr6), if:
463 		 * - saddr6 is given from the caller (by giving "ln"), and
464 		 * - saddr6 belongs to the outgoing interface.
465 		 * Otherwise, we perform the source address selection as usual.
466 		 */
467 		struct in6_addr *hsrc;
468 
469 		hsrc = NULL;
470 		if (ln != NULL) {
471 			LLE_RLOCK(ln);
472 			if (ln->la_hold != NULL) {
473 				struct ip6_hdr *hip6;		/* hold ip6 */
474 
475 				/*
476 				 * assuming every packet in la_hold has the same IP
477 				 * header
478 				 */
479 				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
480 				/* XXX pullup? */
481 				if (sizeof(*hip6) < ln->la_hold->m_len) {
482 					ip6->ip6_src = hip6->ip6_src;
483 					hsrc = &hip6->ip6_src;
484 				}
485 			}
486 			LLE_RUNLOCK(ln);
487 		}
488 		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
489 		    hsrc)) != NULL) {
490 			/* ip6_src set already. */
491 			ifa_free(ifa);
492 		} else {
493 			int error;
494 			struct sockaddr_in6 dst_sa;
495 			struct in6_addr src_in;
496 			struct ifnet *oifp;
497 
498 			bzero(&dst_sa, sizeof(dst_sa));
499 			dst_sa.sin6_family = AF_INET6;
500 			dst_sa.sin6_len = sizeof(dst_sa);
501 			dst_sa.sin6_addr = ip6->ip6_dst;
502 
503 			oifp = ifp;
504 			error = in6_selectsrc(&dst_sa, NULL,
505 			    NULL, &ro, NULL, &oifp, &src_in);
506 			if (error) {
507 				char ip6buf[INET6_ADDRSTRLEN];
508 				nd6log((LOG_DEBUG,
509 				    "nd6_ns_output: source can't be "
510 				    "determined: dst=%s, error=%d\n",
511 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
512 				    error));
513 				goto bad;
514 			}
515 			ip6->ip6_src = src_in;
516 		}
517 	} else {
518 		/*
519 		 * Source address for DAD packet must always be IPv6
520 		 * unspecified address. (0::0)
521 		 * We actually don't have to 0-clear the address (we did it
522 		 * above), but we do so here explicitly to make the intention
523 		 * clearer.
524 		 */
525 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
526 	}
527 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
528 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
529 	nd_ns->nd_ns_code = 0;
530 	nd_ns->nd_ns_reserved = 0;
531 	nd_ns->nd_ns_target = *taddr6;
532 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
533 
534 	/*
535 	 * Add source link-layer address option.
536 	 *
537 	 *				spec		implementation
538 	 *				---		---
539 	 * DAD packet			MUST NOT	do not add the option
540 	 * there's no link layer address:
541 	 *				impossible	do not add the option
542 	 * there's link layer address:
543 	 *	Multicast NS		MUST add one	add the option
544 	 *	Unicast NS		SHOULD add one	add the option
545 	 */
546 	if (!dad && (mac = nd6_ifptomac(ifp))) {
547 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
548 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
549 		/* 8 byte alignments... */
550 		optlen = (optlen + 7) & ~7;
551 
552 		m->m_pkthdr.len += optlen;
553 		m->m_len += optlen;
554 		icmp6len += optlen;
555 		bzero((caddr_t)nd_opt, optlen);
556 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
557 		nd_opt->nd_opt_len = optlen >> 3;
558 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
559 	}
560 
561 	ip6->ip6_plen = htons((u_short)icmp6len);
562 	nd_ns->nd_ns_cksum = 0;
563 	nd_ns->nd_ns_cksum =
564 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
565 
566 	if (send_sendso_input_hook != NULL) {
567 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
568 			sizeof(unsigned short), M_NOWAIT);
569 		if (mtag == NULL)
570 			goto bad;
571 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
572 		m_tag_prepend(m, mtag);
573 	}
574 
575 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
576 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
577 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
578 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
579 
580 	/* We don't cache this route. */
581 	RO_RTFREE(&ro);
582 
583 	return;
584 
585   bad:
586 	if (ro.ro_rt) {
587 		RTFREE(ro.ro_rt);
588 	}
589 	m_freem(m);
590 	return;
591 }
592 
593 /*
594  * Neighbor advertisement input handling.
595  *
596  * Based on RFC 2461
597  * Based on RFC 2462 (duplicate address detection)
598  *
599  * the following items are not implemented yet:
600  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
601  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
602  */
603 void
604 nd6_na_input(struct mbuf *m, int off, int icmp6len)
605 {
606 	struct ifnet *ifp = m->m_pkthdr.rcvif;
607 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
608 	struct nd_neighbor_advert *nd_na;
609 	struct in6_addr daddr6 = ip6->ip6_dst;
610 	struct in6_addr taddr6;
611 	int flags;
612 	int is_router;
613 	int is_solicited;
614 	int is_override;
615 	char *lladdr = NULL;
616 	int lladdrlen = 0;
617 	int checklink = 0;
618 	struct ifaddr *ifa;
619 	struct llentry *ln = NULL;
620 	union nd_opts ndopts;
621 	struct mbuf *chain = NULL;
622 	struct m_tag *mtag;
623 	struct sockaddr_in6 sin6;
624 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
625 
626 	if (ip6->ip6_hlim != 255) {
627 		nd6log((LOG_ERR,
628 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
629 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
630 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
631 		goto bad;
632 	}
633 
634 #ifndef PULLDOWN_TEST
635 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
636 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
637 #else
638 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
639 	if (nd_na == NULL) {
640 		ICMP6STAT_INC(icp6s_tooshort);
641 		return;
642 	}
643 #endif
644 
645 	flags = nd_na->nd_na_flags_reserved;
646 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
647 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
648 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
649 
650 	taddr6 = nd_na->nd_na_target;
651 	if (in6_setscope(&taddr6, ifp, NULL))
652 		goto bad;	/* XXX: impossible */
653 
654 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
655 		nd6log((LOG_ERR,
656 		    "nd6_na_input: invalid target address %s\n",
657 		    ip6_sprintf(ip6bufs, &taddr6)));
658 		goto bad;
659 	}
660 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
661 		if (is_solicited) {
662 			nd6log((LOG_ERR,
663 			    "nd6_na_input: a solicited adv is multicasted\n"));
664 			goto bad;
665 		}
666 
667 	icmp6len -= sizeof(*nd_na);
668 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
669 	if (nd6_options(&ndopts) < 0) {
670 		nd6log((LOG_INFO,
671 		    "nd6_na_input: invalid ND option, ignored\n"));
672 		/* nd6_options have incremented stats */
673 		goto freeit;
674 	}
675 
676 	if (ndopts.nd_opts_tgt_lladdr) {
677 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
678 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
679 	}
680 
681 	/*
682 	 * This effectively disables the DAD check on a non-master CARP
683 	 * address.
684 	 */
685 	if (ifp->if_carp)
686 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
687 	else
688 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
689 
690 	/*
691 	 * Target address matches one of my interface address.
692 	 *
693 	 * If my address is tentative, this means that there's somebody
694 	 * already using the same address as mine.  This indicates DAD failure.
695 	 * This is defined in RFC 2462.
696 	 *
697 	 * Otherwise, process as defined in RFC 2461.
698 	 */
699 	if (ifa
700 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
701 		ifa_free(ifa);
702 		nd6_dad_na_input(ifa);
703 		goto freeit;
704 	}
705 
706 	/* Just for safety, maybe unnecessary. */
707 	if (ifa) {
708 		ifa_free(ifa);
709 		log(LOG_ERR,
710 		    "nd6_na_input: duplicate IP6 address %s\n",
711 		    ip6_sprintf(ip6bufs, &taddr6));
712 		goto freeit;
713 	}
714 
715 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
716 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
717 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
718 		    ifp->if_addrlen, lladdrlen - 2));
719 		goto bad;
720 	}
721 
722 	/*
723 	 * If no neighbor cache entry is found, NA SHOULD silently be
724 	 * discarded.
725 	 */
726 	IF_AFDATA_RLOCK(ifp);
727 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
728 	IF_AFDATA_RUNLOCK(ifp);
729 	if (ln == NULL) {
730 		goto freeit;
731 	}
732 
733 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
734 		/*
735 		 * If the link-layer has address, and no lladdr option came,
736 		 * discard the packet.
737 		 */
738 		if (ifp->if_addrlen && lladdr == NULL) {
739 			goto freeit;
740 		}
741 
742 		/*
743 		 * Record link-layer address, and update the state.
744 		 */
745 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
746 		ln->la_flags |= LLE_VALID;
747 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
748 		if (is_solicited) {
749 			ln->ln_state = ND6_LLINFO_REACHABLE;
750 			ln->ln_byhint = 0;
751 			if (!ND6_LLINFO_PERMANENT(ln)) {
752 				nd6_llinfo_settimer_locked(ln,
753 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
754 			}
755 		} else {
756 			ln->ln_state = ND6_LLINFO_STALE;
757 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
758 		}
759 		if ((ln->ln_router = is_router) != 0) {
760 			/*
761 			 * This means a router's state has changed from
762 			 * non-reachable to probably reachable, and might
763 			 * affect the status of associated prefixes..
764 			 */
765 			checklink = 1;
766 		}
767 	} else {
768 		int llchange;
769 
770 		/*
771 		 * Check if the link-layer address has changed or not.
772 		 */
773 		if (lladdr == NULL)
774 			llchange = 0;
775 		else {
776 			if (ln->la_flags & LLE_VALID) {
777 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
778 					llchange = 1;
779 				else
780 					llchange = 0;
781 			} else
782 				llchange = 1;
783 		}
784 
785 		/*
786 		 * This is VERY complex.  Look at it with care.
787 		 *
788 		 * override solicit lladdr llchange	action
789 		 *					(L: record lladdr)
790 		 *
791 		 *	0	0	n	--	(2c)
792 		 *	0	0	y	n	(2b) L
793 		 *	0	0	y	y	(1)    REACHABLE->STALE
794 		 *	0	1	n	--	(2c)   *->REACHABLE
795 		 *	0	1	y	n	(2b) L *->REACHABLE
796 		 *	0	1	y	y	(1)    REACHABLE->STALE
797 		 *	1	0	n	--	(2a)
798 		 *	1	0	y	n	(2a) L
799 		 *	1	0	y	y	(2a) L *->STALE
800 		 *	1	1	n	--	(2a)   *->REACHABLE
801 		 *	1	1	y	n	(2a) L *->REACHABLE
802 		 *	1	1	y	y	(2a) L *->REACHABLE
803 		 */
804 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
805 			/*
806 			 * If state is REACHABLE, make it STALE.
807 			 * no other updates should be done.
808 			 */
809 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
810 				ln->ln_state = ND6_LLINFO_STALE;
811 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
812 			}
813 			goto freeit;
814 		} else if (is_override				   /* (2a) */
815 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
816 			|| lladdr == NULL) {			   /* (2c) */
817 			/*
818 			 * Update link-local address, if any.
819 			 */
820 			if (lladdr != NULL) {
821 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
822 				ln->la_flags |= LLE_VALID;
823 				EVENTHANDLER_INVOKE(lle_event, ln,
824 				    LLENTRY_RESOLVED);
825 			}
826 
827 			/*
828 			 * If solicited, make the state REACHABLE.
829 			 * If not solicited and the link-layer address was
830 			 * changed, make it STALE.
831 			 */
832 			if (is_solicited) {
833 				ln->ln_state = ND6_LLINFO_REACHABLE;
834 				ln->ln_byhint = 0;
835 				if (!ND6_LLINFO_PERMANENT(ln)) {
836 					nd6_llinfo_settimer_locked(ln,
837 					    (long)ND_IFINFO(ifp)->reachable * hz);
838 				}
839 			} else {
840 				if (lladdr != NULL && llchange) {
841 					ln->ln_state = ND6_LLINFO_STALE;
842 					nd6_llinfo_settimer_locked(ln,
843 					    (long)V_nd6_gctimer * hz);
844 				}
845 			}
846 		}
847 
848 		if (ln->ln_router && !is_router) {
849 			/*
850 			 * The peer dropped the router flag.
851 			 * Remove the sender from the Default Router List and
852 			 * update the Destination Cache entries.
853 			 */
854 			struct nd_defrouter *dr;
855 			struct in6_addr *in6;
856 
857 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
858 
859 			/*
860 			 * Lock to protect the default router list.
861 			 * XXX: this might be unnecessary, since this function
862 			 * is only called under the network software interrupt
863 			 * context.  However, we keep it just for safety.
864 			 */
865 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
866 			if (dr)
867 				defrtrlist_del(dr);
868 			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
869 			    ND6_IFF_ACCEPT_RTADV) {
870 				/*
871 				 * Even if the neighbor is not in the default
872 				 * router list, the neighbor may be used
873 				 * as a next hop for some destinations
874 				 * (e.g. redirect case). So we must
875 				 * call rt6_flush explicitly.
876 				 */
877 				rt6_flush(&ip6->ip6_src, ifp);
878 			}
879 		}
880 		ln->ln_router = is_router;
881 	}
882         /* XXX - QL
883 	 *  Does this matter?
884 	 *  rt->rt_flags &= ~RTF_REJECT;
885 	 */
886 	ln->la_asked = 0;
887 	if (ln->la_hold) {
888 		struct mbuf *m_hold, *m_hold_next;
889 
890 		/*
891 		 * reset the la_hold in advance, to explicitly
892 		 * prevent a la_hold lookup in nd6_output()
893 		 * (wouldn't happen, though...)
894 		 */
895 		for (m_hold = ln->la_hold, ln->la_hold = NULL;
896 		    m_hold; m_hold = m_hold_next) {
897 			m_hold_next = m_hold->m_nextpkt;
898 			m_hold->m_nextpkt = NULL;
899 			/*
900 			 * we assume ifp is not a loopback here, so just set
901 			 * the 2nd argument as the 1st one.
902 			 */
903 
904 			if (send_sendso_input_hook != NULL) {
905 				mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
906 				    sizeof(unsigned short), M_NOWAIT);
907 				if (mtag == NULL)
908 					goto bad;
909 				m_tag_prepend(m, mtag);
910 			}
911 
912 			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
913 		}
914 	}
915  freeit:
916 	if (ln != NULL) {
917 		if (chain)
918 			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
919 		LLE_WUNLOCK(ln);
920 
921 		if (chain)
922 			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
923 	}
924 	if (checklink)
925 		pfxlist_onlink_check();
926 
927 	m_freem(m);
928 	return;
929 
930  bad:
931 	if (ln != NULL)
932 		LLE_WUNLOCK(ln);
933 
934 	ICMP6STAT_INC(icp6s_badna);
935 	m_freem(m);
936 }
937 
938 /*
939  * Neighbor advertisement output handling.
940  *
941  * Based on RFC 2461
942  *
943  * the following items are not implemented yet:
944  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
945  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
946  *
947  * tlladdr - 1 if include target link-layer address
948  * sdl0 - sockaddr_dl (= proxy NA) or NULL
949  */
950 static void
951 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
952     const struct in6_addr *taddr6, u_long flags, int tlladdr,
953     struct sockaddr *sdl0, u_int fibnum)
954 {
955 	struct mbuf *m;
956 	struct m_tag *mtag;
957 	struct ifnet *oifp;
958 	struct ip6_hdr *ip6;
959 	struct nd_neighbor_advert *nd_na;
960 	struct ip6_moptions im6o;
961 	struct in6_addr src, daddr6;
962 	struct sockaddr_in6 dst_sa;
963 	int icmp6len, maxlen, error;
964 	caddr_t mac = NULL;
965 	struct route_in6 ro;
966 
967 	bzero(&ro, sizeof(ro));
968 
969 	daddr6 = *daddr6_0;	/* make a local copy for modification */
970 
971 	/* estimate the size of message */
972 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
973 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
974 	if (max_linkhdr + maxlen >= MCLBYTES) {
975 #ifdef DIAGNOSTIC
976 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
977 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
978 #endif
979 		return;
980 	}
981 
982 	if (max_linkhdr + maxlen > MHLEN)
983 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
984 	else
985 		m = m_gethdr(M_NOWAIT, MT_DATA);
986 	if (m == NULL)
987 		return;
988 	M_SETFIB(m, fibnum);
989 
990 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
991 		m->m_flags |= M_MCAST;
992 		im6o.im6o_multicast_ifp = ifp;
993 		im6o.im6o_multicast_hlim = 255;
994 		im6o.im6o_multicast_loop = 0;
995 	}
996 
997 	icmp6len = sizeof(*nd_na);
998 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
999 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
1000 
1001 	/* fill neighbor advertisement packet */
1002 	ip6 = mtod(m, struct ip6_hdr *);
1003 	ip6->ip6_flow = 0;
1004 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1005 	ip6->ip6_vfc |= IPV6_VERSION;
1006 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1007 	ip6->ip6_hlim = 255;
1008 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1009 		/* reply to DAD */
1010 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1011 		daddr6.s6_addr16[1] = 0;
1012 		daddr6.s6_addr32[1] = 0;
1013 		daddr6.s6_addr32[2] = 0;
1014 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1015 		if (in6_setscope(&daddr6, ifp, NULL))
1016 			goto bad;
1017 
1018 		flags &= ~ND_NA_FLAG_SOLICITED;
1019 	}
1020 	ip6->ip6_dst = daddr6;
1021 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1022 	dst_sa.sin6_family = AF_INET6;
1023 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1024 	dst_sa.sin6_addr = daddr6;
1025 
1026 	/*
1027 	 * Select a source whose scope is the same as that of the dest.
1028 	 */
1029 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1030 	oifp = ifp;
1031 	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1032 	if (error) {
1033 		char ip6buf[INET6_ADDRSTRLEN];
1034 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1035 		    "determined: dst=%s, error=%d\n",
1036 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1037 		goto bad;
1038 	}
1039 	ip6->ip6_src = src;
1040 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1041 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1042 	nd_na->nd_na_code = 0;
1043 	nd_na->nd_na_target = *taddr6;
1044 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1045 
1046 	/*
1047 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1048 	 * see nd6_ns_input() for details.
1049 	 * Basically, if NS packet is sent to unicast/anycast addr,
1050 	 * target lladdr option SHOULD NOT be included.
1051 	 */
1052 	if (tlladdr) {
1053 		/*
1054 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1055 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1056 		 * my address) use lladdr configured for the interface.
1057 		 */
1058 		if (sdl0 == NULL) {
1059 			if (ifp->if_carp)
1060 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1061 			if (mac == NULL)
1062 				mac = nd6_ifptomac(ifp);
1063 		} else if (sdl0->sa_family == AF_LINK) {
1064 			struct sockaddr_dl *sdl;
1065 			sdl = (struct sockaddr_dl *)sdl0;
1066 			if (sdl->sdl_alen == ifp->if_addrlen)
1067 				mac = LLADDR(sdl);
1068 		}
1069 	}
1070 	if (tlladdr && mac) {
1071 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1072 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1073 
1074 		/* roundup to 8 bytes alignment! */
1075 		optlen = (optlen + 7) & ~7;
1076 
1077 		m->m_pkthdr.len += optlen;
1078 		m->m_len += optlen;
1079 		icmp6len += optlen;
1080 		bzero((caddr_t)nd_opt, optlen);
1081 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1082 		nd_opt->nd_opt_len = optlen >> 3;
1083 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1084 	} else
1085 		flags &= ~ND_NA_FLAG_OVERRIDE;
1086 
1087 	ip6->ip6_plen = htons((u_short)icmp6len);
1088 	nd_na->nd_na_flags_reserved = flags;
1089 	nd_na->nd_na_cksum = 0;
1090 	nd_na->nd_na_cksum =
1091 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1092 
1093 	if (send_sendso_input_hook != NULL) {
1094 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1095 		    sizeof(unsigned short), M_NOWAIT);
1096 		if (mtag == NULL)
1097 			goto bad;
1098 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1099 		m_tag_prepend(m, mtag);
1100 	}
1101 
1102 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1103 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1104 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1105 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1106 
1107 	/* We don't cache this route. */
1108 	RO_RTFREE(&ro);
1109 
1110 	return;
1111 
1112   bad:
1113 	if (ro.ro_rt) {
1114 		RTFREE(ro.ro_rt);
1115 	}
1116 	m_freem(m);
1117 	return;
1118 }
1119 
1120 #ifndef BURN_BRIDGES
1121 void
1122 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1123     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1124     struct sockaddr *sdl0)
1125 {
1126 
1127 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1128 	    RT_DEFAULT_FIB);
1129 }
1130 #endif
1131 
1132 caddr_t
1133 nd6_ifptomac(struct ifnet *ifp)
1134 {
1135 	switch (ifp->if_type) {
1136 	case IFT_ARCNET:
1137 	case IFT_ETHER:
1138 	case IFT_FDDI:
1139 	case IFT_IEEE1394:
1140 #ifdef IFT_L2VLAN
1141 	case IFT_L2VLAN:
1142 #endif
1143 #ifdef IFT_IEEE80211
1144 	case IFT_IEEE80211:
1145 #endif
1146 	case IFT_INFINIBAND:
1147 	case IFT_BRIDGE:
1148 	case IFT_ISO88025:
1149 		return IF_LLADDR(ifp);
1150 	default:
1151 		return NULL;
1152 	}
1153 }
1154 
1155 struct dadq {
1156 	TAILQ_ENTRY(dadq) dad_list;
1157 	struct ifaddr *dad_ifa;
1158 	int dad_count;		/* max NS to send */
1159 	int dad_ns_tcount;	/* # of trials to send NS */
1160 	int dad_ns_ocount;	/* NS sent so far */
1161 	int dad_ns_icount;
1162 	int dad_na_icount;
1163 	struct callout dad_timer_ch;
1164 	struct vnet *dad_vnet;
1165 };
1166 
1167 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1168 VNET_DEFINE(int, dad_init) = 0;
1169 #define	V_dadq				VNET(dadq)
1170 #define	V_dad_init			VNET(dad_init)
1171 
1172 static struct dadq *
1173 nd6_dad_find(struct ifaddr *ifa)
1174 {
1175 	struct dadq *dp;
1176 
1177 	TAILQ_FOREACH(dp, &V_dadq, dad_list)
1178 		if (dp->dad_ifa == ifa)
1179 			return (dp);
1180 
1181 	return (NULL);
1182 }
1183 
1184 static void
1185 nd6_dad_starttimer(struct dadq *dp, int ticks)
1186 {
1187 
1188 	callout_reset(&dp->dad_timer_ch, ticks,
1189 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1190 }
1191 
1192 static void
1193 nd6_dad_stoptimer(struct dadq *dp)
1194 {
1195 
1196 	callout_stop(&dp->dad_timer_ch);
1197 }
1198 
1199 /*
1200  * Start Duplicate Address Detection (DAD) for specified interface address.
1201  */
1202 void
1203 nd6_dad_start(struct ifaddr *ifa, int delay)
1204 {
1205 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1206 	struct dadq *dp;
1207 	char ip6buf[INET6_ADDRSTRLEN];
1208 
1209 	if (!V_dad_init) {
1210 		TAILQ_INIT(&V_dadq);
1211 		V_dad_init++;
1212 	}
1213 
1214 	/*
1215 	 * If we don't need DAD, don't do it.
1216 	 * There are several cases:
1217 	 * - DAD is disabled (ip6_dad_count == 0)
1218 	 * - the interface address is anycast
1219 	 */
1220 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1221 		log(LOG_DEBUG,
1222 			"nd6_dad_start: called with non-tentative address "
1223 			"%s(%s)\n",
1224 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1225 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1226 		return;
1227 	}
1228 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1229 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1230 		return;
1231 	}
1232 	if (!V_ip6_dad_count) {
1233 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1234 		return;
1235 	}
1236 	if (ifa->ifa_ifp == NULL)
1237 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1238 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1239 		return;
1240 	}
1241 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1242 		return;
1243 	if (nd6_dad_find(ifa) != NULL) {
1244 		/* DAD already in progress */
1245 		return;
1246 	}
1247 
1248 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1249 	if (dp == NULL) {
1250 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1251 			"%s(%s)\n",
1252 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1253 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1254 		return;
1255 	}
1256 	bzero(dp, sizeof(*dp));
1257 	callout_init(&dp->dad_timer_ch, 0);
1258 #ifdef VIMAGE
1259 	dp->dad_vnet = curvnet;
1260 #endif
1261 	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1262 
1263 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1264 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1265 
1266 	/*
1267 	 * Send NS packet for DAD, ip6_dad_count times.
1268 	 * Note that we must delay the first transmission, if this is the
1269 	 * first packet to be sent from the interface after interface
1270 	 * (re)initialization.
1271 	 */
1272 	dp->dad_ifa = ifa;
1273 	ifa_ref(ifa);	/* just for safety */
1274 	dp->dad_count = V_ip6_dad_count;
1275 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1276 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1277 	if (delay == 0) {
1278 		nd6_dad_ns_output(dp, ifa);
1279 		nd6_dad_starttimer(dp,
1280 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1281 	} else {
1282 		nd6_dad_starttimer(dp, delay);
1283 	}
1284 }
1285 
1286 /*
1287  * terminate DAD unconditionally.  used for address removals.
1288  */
1289 void
1290 nd6_dad_stop(struct ifaddr *ifa)
1291 {
1292 	struct dadq *dp;
1293 
1294 	if (!V_dad_init)
1295 		return;
1296 	dp = nd6_dad_find(ifa);
1297 	if (!dp) {
1298 		/* DAD wasn't started yet */
1299 		return;
1300 	}
1301 
1302 	nd6_dad_stoptimer(dp);
1303 
1304 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1305 	free(dp, M_IP6NDP);
1306 	dp = NULL;
1307 	ifa_free(ifa);
1308 }
1309 
1310 static void
1311 nd6_dad_timer(struct dadq *dp)
1312 {
1313 	CURVNET_SET(dp->dad_vnet);
1314 	struct ifaddr *ifa = dp->dad_ifa;
1315 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1316 	char ip6buf[INET6_ADDRSTRLEN];
1317 
1318 	/* Sanity check */
1319 	if (ia == NULL) {
1320 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1321 		goto done;
1322 	}
1323 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1324 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1325 			"%s(%s)\n",
1326 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1327 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1328 		goto done;
1329 	}
1330 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1331 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1332 			"%s(%s)\n",
1333 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1334 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1335 		goto done;
1336 	}
1337 
1338 	/* timeouted with IFF_{RUNNING,UP} check */
1339 	if (dp->dad_ns_tcount > V_dad_maxtry) {
1340 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1341 		    if_name(ifa->ifa_ifp)));
1342 
1343 		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1344 		free(dp, M_IP6NDP);
1345 		dp = NULL;
1346 		ifa_free(ifa);
1347 		goto done;
1348 	}
1349 
1350 	/* Need more checks? */
1351 	if (dp->dad_ns_ocount < dp->dad_count) {
1352 		/*
1353 		 * We have more NS to go.  Send NS packet for DAD.
1354 		 */
1355 		nd6_dad_ns_output(dp, ifa);
1356 		nd6_dad_starttimer(dp,
1357 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1358 	} else {
1359 		/*
1360 		 * We have transmitted sufficient number of DAD packets.
1361 		 * See what we've got.
1362 		 */
1363 		int duplicate;
1364 
1365 		duplicate = 0;
1366 
1367 		if (dp->dad_na_icount) {
1368 			/*
1369 			 * the check is in nd6_dad_na_input(),
1370 			 * but just in case
1371 			 */
1372 			duplicate++;
1373 		}
1374 
1375 		if (dp->dad_ns_icount) {
1376 			/* We've seen NS, means DAD has failed. */
1377 			duplicate++;
1378 		}
1379 
1380 		if (duplicate) {
1381 			/* (*dp) will be freed in nd6_dad_duplicated() */
1382 			dp = NULL;
1383 			nd6_dad_duplicated(ifa);
1384 		} else {
1385 			/*
1386 			 * We are done with DAD.  No NA came, no NS came.
1387 			 * No duplicate address found.
1388 			 */
1389 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1390 
1391 			nd6log((LOG_DEBUG,
1392 			    "%s: DAD complete for %s - no duplicates found\n",
1393 			    if_name(ifa->ifa_ifp),
1394 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1395 
1396 			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1397 			free(dp, M_IP6NDP);
1398 			dp = NULL;
1399 			ifa_free(ifa);
1400 		}
1401 	}
1402 
1403 done:
1404 	CURVNET_RESTORE();
1405 }
1406 
1407 void
1408 nd6_dad_duplicated(struct ifaddr *ifa)
1409 {
1410 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1411 	struct ifnet *ifp;
1412 	struct dadq *dp;
1413 	char ip6buf[INET6_ADDRSTRLEN];
1414 
1415 	dp = nd6_dad_find(ifa);
1416 	if (dp == NULL) {
1417 		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1418 		return;
1419 	}
1420 
1421 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1422 	    "NS in/out=%d/%d, NA in=%d\n",
1423 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1424 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1425 
1426 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1427 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1428 
1429 	/* We are done with DAD, with duplicate address found. (failure) */
1430 	nd6_dad_stoptimer(dp);
1431 
1432 	ifp = ifa->ifa_ifp;
1433 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1434 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1435 	log(LOG_ERR, "%s: manual intervention required\n",
1436 	    if_name(ifp));
1437 
1438 	/*
1439 	 * If the address is a link-local address formed from an interface
1440 	 * identifier based on the hardware address which is supposed to be
1441 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1442 	 * operation on the interface SHOULD be disabled.
1443 	 * [RFC 4862, Section 5.4.5]
1444 	 */
1445 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1446 		struct in6_addr in6;
1447 
1448 		/*
1449 		 * To avoid over-reaction, we only apply this logic when we are
1450 		 * very sure that hardware addresses are supposed to be unique.
1451 		 */
1452 		switch (ifp->if_type) {
1453 		case IFT_ETHER:
1454 		case IFT_FDDI:
1455 		case IFT_ATM:
1456 		case IFT_IEEE1394:
1457 #ifdef IFT_IEEE80211
1458 		case IFT_IEEE80211:
1459 #endif
1460 		case IFT_INFINIBAND:
1461 			in6 = ia->ia_addr.sin6_addr;
1462 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1463 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1464 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1465 				log(LOG_ERR, "%s: possible hardware address "
1466 				    "duplication detected, disable IPv6\n",
1467 				    if_name(ifp));
1468 			}
1469 			break;
1470 		}
1471 	}
1472 
1473 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1474 	free(dp, M_IP6NDP);
1475 	dp = NULL;
1476 	ifa_free(ifa);
1477 }
1478 
1479 static void
1480 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1481 {
1482 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1483 	struct ifnet *ifp = ifa->ifa_ifp;
1484 
1485 	dp->dad_ns_tcount++;
1486 	if ((ifp->if_flags & IFF_UP) == 0) {
1487 		return;
1488 	}
1489 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1490 		return;
1491 	}
1492 
1493 	dp->dad_ns_ocount++;
1494 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1495 }
1496 
1497 static void
1498 nd6_dad_ns_input(struct ifaddr *ifa)
1499 {
1500 	struct in6_ifaddr *ia;
1501 	struct ifnet *ifp;
1502 	const struct in6_addr *taddr6;
1503 	struct dadq *dp;
1504 	int duplicate;
1505 
1506 	if (ifa == NULL)
1507 		panic("ifa == NULL in nd6_dad_ns_input");
1508 
1509 	ia = (struct in6_ifaddr *)ifa;
1510 	ifp = ifa->ifa_ifp;
1511 	taddr6 = &ia->ia_addr.sin6_addr;
1512 	duplicate = 0;
1513 	dp = nd6_dad_find(ifa);
1514 
1515 	/* Quickhack - completely ignore DAD NS packets */
1516 	if (V_dad_ignore_ns) {
1517 		char ip6buf[INET6_ADDRSTRLEN];
1518 		nd6log((LOG_INFO,
1519 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1520 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1521 		    if_name(ifa->ifa_ifp)));
1522 		return;
1523 	}
1524 
1525 	/*
1526 	 * if I'm yet to start DAD, someone else started using this address
1527 	 * first.  I have a duplicate and you win.
1528 	 */
1529 	if (dp == NULL || dp->dad_ns_ocount == 0)
1530 		duplicate++;
1531 
1532 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1533 
1534 	if (duplicate) {
1535 		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1536 		nd6_dad_duplicated(ifa);
1537 	} else {
1538 		/*
1539 		 * not sure if I got a duplicate.
1540 		 * increment ns count and see what happens.
1541 		 */
1542 		if (dp)
1543 			dp->dad_ns_icount++;
1544 	}
1545 }
1546 
1547 static void
1548 nd6_dad_na_input(struct ifaddr *ifa)
1549 {
1550 	struct dadq *dp;
1551 
1552 	if (ifa == NULL)
1553 		panic("ifa == NULL in nd6_dad_na_input");
1554 
1555 	dp = nd6_dad_find(ifa);
1556 	if (dp)
1557 		dp->dad_na_icount++;
1558 
1559 	/* remove the address. */
1560 	nd6_dad_duplicated(ifa);
1561 }
1562