xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 6b2c1e49da284f28ec7b52f7c031474087e37104)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40 #include "opt_mpath.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/eventhandler.h>
45 #include <sys/malloc.h>
46 #include <sys/libkern.h>
47 #include <sys/lock.h>
48 #include <sys/rwlock.h>
49 #include <sys/mbuf.h>
50 #include <sys/socket.h>
51 #include <sys/sockio.h>
52 #include <sys/time.h>
53 #include <sys/kernel.h>
54 #include <sys/errno.h>
55 #include <sys/sysctl.h>
56 #include <sys/syslog.h>
57 #include <sys/queue.h>
58 #include <sys/callout.h>
59 #include <sys/refcount.h>
60 
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/if_dl.h>
64 #include <net/if_var.h>
65 #include <net/route.h>
66 #ifdef RADIX_MPATH
67 #include <net/radix_mpath.h>
68 #endif
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <net/if_llatbl.h>
74 #include <netinet6/in6_var.h>
75 #include <netinet6/in6_ifattach.h>
76 #include <netinet/ip6.h>
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/scope6_var.h>
79 #include <netinet6/nd6.h>
80 #include <netinet/icmp6.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet6/send.h>
83 
84 #define SDL(s) ((struct sockaddr_dl *)s)
85 
86 struct dadq;
87 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
88 static void nd6_dad_add(struct dadq *dp);
89 static void nd6_dad_del(struct dadq *dp);
90 static void nd6_dad_rele(struct dadq *);
91 static void nd6_dad_starttimer(struct dadq *, int, int);
92 static void nd6_dad_stoptimer(struct dadq *);
93 static void nd6_dad_timer(struct dadq *);
94 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
95 static void nd6_dad_ns_output(struct dadq *);
96 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
97 static void nd6_dad_na_input(struct ifaddr *);
98 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
99     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
100 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
101     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
102 
103 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
104 #define	V_dad_enhanced			VNET(dad_enhanced)
105 
106 SYSCTL_DECL(_net_inet6_ip6);
107 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
108     &VNET_NAME(dad_enhanced), 0,
109     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
110 
111 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
112 						   transmit DAD packet */
113 #define	V_dad_maxtry			VNET(dad_maxtry)
114 
115 /*
116  * Input a Neighbor Solicitation Message.
117  *
118  * Based on RFC 2461
119  * Based on RFC 2462 (duplicate address detection)
120  */
121 void
122 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
123 {
124 	struct ifnet *ifp;
125 	struct ip6_hdr *ip6;
126 	struct nd_neighbor_solicit *nd_ns;
127 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
128 	struct ifaddr *ifa;
129 	struct sockaddr_dl proxydl;
130 	union nd_opts ndopts;
131 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
132 	char *lladdr;
133 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
134 
135 	ifa = NULL;
136 
137 	/* RFC 6980: Nodes MUST silently ignore fragments */
138 	if(m->m_flags & M_FRAGMENTED)
139 		goto freeit;
140 
141 	ifp = m->m_pkthdr.rcvif;
142 	ip6 = mtod(m, struct ip6_hdr *);
143 	if (ip6->ip6_hlim != 255) {
144 		nd6log((LOG_ERR,
145 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
146 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
147 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
148 		goto bads;
149 	}
150 
151 	m = m_pullup(m, off + icmp6len);
152 	if (m == NULL) {
153 		IP6STAT_INC(ip6s_exthdrtoolong);
154 		return;
155 	}
156 	ip6 = mtod(m, struct ip6_hdr *);
157 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
158 
159 	saddr6 = ip6->ip6_src;
160 	daddr6 = ip6->ip6_dst;
161 	taddr6 = nd_ns->nd_ns_target;
162 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
163 		goto bad;
164 
165 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
166 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
167 		rflag = 0;
168 
169 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
170 		/* dst has to be a solicited node multicast address. */
171 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
172 		    /* don't check ifindex portion */
173 		    daddr6.s6_addr32[1] == 0 &&
174 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
175 		    daddr6.s6_addr8[12] == 0xff) {
176 			; /* good */
177 		} else {
178 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
179 			    "(wrong ip6 dst)\n"));
180 			goto bad;
181 		}
182 	} else if (!V_nd6_onlink_ns_rfc4861) {
183 		struct sockaddr_in6 src_sa6;
184 
185 		/*
186 		 * According to recent IETF discussions, it is not a good idea
187 		 * to accept a NS from an address which would not be deemed
188 		 * to be a neighbor otherwise.  This point is expected to be
189 		 * clarified in future revisions of the specification.
190 		 */
191 		bzero(&src_sa6, sizeof(src_sa6));
192 		src_sa6.sin6_family = AF_INET6;
193 		src_sa6.sin6_len = sizeof(src_sa6);
194 		src_sa6.sin6_addr = saddr6;
195 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
196 			nd6log((LOG_INFO, "nd6_ns_input: "
197 				"NS packet from non-neighbor\n"));
198 			goto bad;
199 		}
200 	}
201 
202 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
203 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
204 		goto bad;
205 	}
206 
207 	icmp6len -= sizeof(*nd_ns);
208 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
209 	if (nd6_options(&ndopts) < 0) {
210 		nd6log((LOG_INFO,
211 		    "nd6_ns_input: invalid ND option, ignored\n"));
212 		/* nd6_options have incremented stats */
213 		goto freeit;
214 	}
215 
216 	lladdr = NULL;
217 	lladdrlen = 0;
218 	if (ndopts.nd_opts_src_lladdr) {
219 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
220 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
221 	}
222 
223 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
224 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
225 		    "(link-layer address option)\n"));
226 		goto bad;
227 	}
228 
229 	/*
230 	 * Attaching target link-layer address to the NA?
231 	 * (RFC 2461 7.2.4)
232 	 *
233 	 * NS IP dst is unicast/anycast			MUST NOT add
234 	 * NS IP dst is solicited-node multicast	MUST add
235 	 *
236 	 * In implementation, we add target link-layer address by default.
237 	 * We do not add one in MUST NOT cases.
238 	 */
239 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
240 		tlladdr = 0;
241 	else
242 		tlladdr = 1;
243 
244 	/*
245 	 * Target address (taddr6) must be either:
246 	 * (1) Valid unicast/anycast address for my receiving interface,
247 	 * (2) Unicast address for which I'm offering proxy service, or
248 	 * (3) "tentative" address on which DAD is being performed.
249 	 */
250 	/* (1) and (3) check. */
251 	if (ifp->if_carp)
252 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
253 	else
254 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
255 
256 	/* (2) check. */
257 	proxy = 0;
258 	if (ifa == NULL) {
259 		struct sockaddr_dl rt_gateway;
260 		struct rt_addrinfo info;
261 		struct sockaddr_in6 dst6;
262 
263 		bzero(&dst6, sizeof(dst6));
264 		dst6.sin6_len = sizeof(struct sockaddr_in6);
265 		dst6.sin6_family = AF_INET6;
266 		dst6.sin6_addr = taddr6;
267 
268 		bzero(&rt_gateway, sizeof(rt_gateway));
269 		rt_gateway.sdl_len = sizeof(rt_gateway);
270 		bzero(&info, sizeof(info));
271 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
272 
273 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
274 		    0, 0, &info) == 0) {
275 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
276 			    rt_gateway.sdl_family == AF_LINK) {
277 
278 				/*
279 				 * proxy NDP for single entry
280 				 */
281 				proxydl = *SDL(&rt_gateway);
282 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
283 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
284 				if (ifa)
285 					proxy = 1;
286 			}
287 		}
288 	}
289 	if (ifa == NULL) {
290 		/*
291 		 * We've got an NS packet, and we don't have that adddress
292 		 * assigned for us.  We MUST silently ignore it.
293 		 * See RFC2461 7.2.3.
294 		 */
295 		goto freeit;
296 	}
297 	myaddr6 = *IFA_IN6(ifa);
298 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
299 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
300 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
301 		goto freeit;
302 
303 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
304 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
305 		    "(if %d, NS packet %d)\n",
306 		    ip6_sprintf(ip6bufs, &taddr6),
307 		    ifp->if_addrlen, lladdrlen - 2));
308 		goto bad;
309 	}
310 
311 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
312 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
313 		    ip6_sprintf(ip6bufs, &saddr6)));
314 		goto freeit;
315 	}
316 
317 	/*
318 	 * We have neighbor solicitation packet, with target address equals to
319 	 * one of my tentative address.
320 	 *
321 	 * src addr	how to process?
322 	 * ---		---
323 	 * multicast	of course, invalid (rejected in ip6_input)
324 	 * unicast	somebody is doing address resolution -> ignore
325 	 * unspec	dup address detection
326 	 *
327 	 * The processing is defined in RFC 2462.
328 	 */
329 	if (tentative) {
330 		/*
331 		 * If source address is unspecified address, it is for
332 		 * duplicate address detection.
333 		 *
334 		 * If not, the packet is for addess resolution;
335 		 * silently ignore it.
336 		 */
337 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
338 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
339 
340 		goto freeit;
341 	}
342 
343 	/*
344 	 * If the source address is unspecified address, entries must not
345 	 * be created or updated.
346 	 * It looks that sender is performing DAD.  Output NA toward
347 	 * all-node multicast address, to tell the sender that I'm using
348 	 * the address.
349 	 * S bit ("solicited") must be zero.
350 	 */
351 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
352 		struct in6_addr in6_all;
353 
354 		in6_all = in6addr_linklocal_allnodes;
355 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
356 			goto bad;
357 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
358 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
359 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
360 		    M_GETFIB(m));
361 		goto freeit;
362 	}
363 
364 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
365 	    ND_NEIGHBOR_SOLICIT, 0);
366 
367 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
368 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
369 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
370 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
371  freeit:
372 	if (ifa != NULL)
373 		ifa_free(ifa);
374 	m_freem(m);
375 	return;
376 
377  bad:
378 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
379 		ip6_sprintf(ip6bufs, &saddr6)));
380 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
381 		ip6_sprintf(ip6bufs, &daddr6)));
382 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
383 		ip6_sprintf(ip6bufs, &taddr6)));
384  bads:
385 	ICMP6STAT_INC(icp6s_badns);
386 	if (ifa != NULL)
387 		ifa_free(ifa);
388 	m_freem(m);
389 }
390 
391 /*
392  * Output a Neighbor Solicitation Message. Caller specifies:
393  *	- ICMP6 header source IP6 address
394  *	- ND6 header target IP6 address
395  *	- ND6 header source datalink address
396  *
397  * Based on RFC 2461
398  * Based on RFC 2462 (duplicate address detection)
399  *
400  *    ln - for source address determination
401  * nonce - If non-NULL, NS is used for duplicate address detection and
402  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
403  */
404 static void
405 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
406     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
407     uint8_t *nonce, u_int fibnum)
408 {
409 	struct mbuf *m;
410 	struct m_tag *mtag;
411 	struct ip6_hdr *ip6;
412 	struct nd_neighbor_solicit *nd_ns;
413 	struct ip6_moptions im6o;
414 	int icmp6len;
415 	int maxlen;
416 	caddr_t mac;
417 
418 	if (IN6_IS_ADDR_MULTICAST(taddr6))
419 		return;
420 
421 	/* estimate the size of message */
422 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
423 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
424 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
425 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
426 	    __func__, max_linkhdr, maxlen, MCLBYTES));
427 
428 	if (max_linkhdr + maxlen > MHLEN)
429 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
430 	else
431 		m = m_gethdr(M_NOWAIT, MT_DATA);
432 	if (m == NULL)
433 		return;
434 	M_SETFIB(m, fibnum);
435 
436 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
437 		m->m_flags |= M_MCAST;
438 		im6o.im6o_multicast_ifp = ifp;
439 		im6o.im6o_multicast_hlim = 255;
440 		im6o.im6o_multicast_loop = 0;
441 	}
442 
443 	icmp6len = sizeof(*nd_ns);
444 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
445 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
446 
447 	/* fill neighbor solicitation packet */
448 	ip6 = mtod(m, struct ip6_hdr *);
449 	ip6->ip6_flow = 0;
450 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
451 	ip6->ip6_vfc |= IPV6_VERSION;
452 	/* ip6->ip6_plen will be set later */
453 	ip6->ip6_nxt = IPPROTO_ICMPV6;
454 	ip6->ip6_hlim = 255;
455 	if (daddr6)
456 		ip6->ip6_dst = *daddr6;
457 	else {
458 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
459 		ip6->ip6_dst.s6_addr16[1] = 0;
460 		ip6->ip6_dst.s6_addr32[1] = 0;
461 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
462 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
463 		ip6->ip6_dst.s6_addr8[12] = 0xff;
464 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
465 			goto bad;
466 	}
467 	if (nonce == NULL) {
468 		struct ifaddr *ifa = NULL;
469 
470 		/*
471 		 * RFC2461 7.2.2:
472 		 * "If the source address of the packet prompting the
473 		 * solicitation is the same as one of the addresses assigned
474 		 * to the outgoing interface, that address SHOULD be placed
475 		 * in the IP Source Address of the outgoing solicitation.
476 		 * Otherwise, any one of the addresses assigned to the
477 		 * interface should be used."
478 		 *
479 		 * We use the source address for the prompting packet
480 		 * (saddr6), if saddr6 belongs to the outgoing interface.
481 		 * Otherwise, we perform the source address selection as usual.
482 		 */
483 
484 		if (saddr6 != NULL)
485 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
486 		if (ifa != NULL) {
487 			/* ip6_src set already. */
488 			ip6->ip6_src = *saddr6;
489 			ifa_free(ifa);
490 		} else {
491 			int error;
492 			struct in6_addr dst6, src6;
493 			uint32_t scopeid;
494 
495 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
496 			error = in6_selectsrc_addr(fibnum, &dst6,
497 			    scopeid, ifp, &src6, NULL);
498 			if (error) {
499 				char ip6buf[INET6_ADDRSTRLEN];
500 				nd6log((LOG_DEBUG, "%s: source can't be "
501 				    "determined: dst=%s, error=%d\n", __func__,
502 				    ip6_sprintf(ip6buf, &dst6),
503 				    error));
504 				goto bad;
505 			}
506 			ip6->ip6_src = src6;
507 		}
508 	} else {
509 		/*
510 		 * Source address for DAD packet must always be IPv6
511 		 * unspecified address. (0::0)
512 		 * We actually don't have to 0-clear the address (we did it
513 		 * above), but we do so here explicitly to make the intention
514 		 * clearer.
515 		 */
516 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
517 	}
518 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
519 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
520 	nd_ns->nd_ns_code = 0;
521 	nd_ns->nd_ns_reserved = 0;
522 	nd_ns->nd_ns_target = *taddr6;
523 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
524 
525 	/*
526 	 * Add source link-layer address option.
527 	 *
528 	 *				spec		implementation
529 	 *				---		---
530 	 * DAD packet			MUST NOT	do not add the option
531 	 * there's no link layer address:
532 	 *				impossible	do not add the option
533 	 * there's link layer address:
534 	 *	Multicast NS		MUST add one	add the option
535 	 *	Unicast NS		SHOULD add one	add the option
536 	 */
537 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
538 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
539 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
540 		/* 8 byte alignments... */
541 		optlen = (optlen + 7) & ~7;
542 
543 		m->m_pkthdr.len += optlen;
544 		m->m_len += optlen;
545 		icmp6len += optlen;
546 		bzero((caddr_t)nd_opt, optlen);
547 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
548 		nd_opt->nd_opt_len = optlen >> 3;
549 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
550 	}
551 	/*
552 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
553 	 * This behavior is documented as Enhanced Duplicate Address
554 	 * Detection in RFC 7527.
555 	 * net.inet6.ip6.dad_enhanced=0 disables this.
556 	 */
557 	if (V_dad_enhanced != 0 && nonce != NULL) {
558 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
559 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
560 		/* 8-byte alignment is required. */
561 		optlen = (optlen + 7) & ~7;
562 
563 		m->m_pkthdr.len += optlen;
564 		m->m_len += optlen;
565 		icmp6len += optlen;
566 		bzero((caddr_t)nd_opt, optlen);
567 		nd_opt->nd_opt_type = ND_OPT_NONCE;
568 		nd_opt->nd_opt_len = optlen >> 3;
569 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
570 	}
571 	ip6->ip6_plen = htons((u_short)icmp6len);
572 	nd_ns->nd_ns_cksum = 0;
573 	nd_ns->nd_ns_cksum =
574 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
575 
576 	if (send_sendso_input_hook != NULL) {
577 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
578 			sizeof(unsigned short), M_NOWAIT);
579 		if (mtag == NULL)
580 			goto bad;
581 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
582 		m_tag_prepend(m, mtag);
583 	}
584 
585 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
586 	    &im6o, NULL, NULL);
587 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
588 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
589 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
590 
591 	return;
592 
593   bad:
594 	m_freem(m);
595 }
596 
597 #ifndef BURN_BRIDGES
598 void
599 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
600     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
601 {
602 
603 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
604 }
605 #endif
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;
620 	struct ip6_hdr *ip6;
621 	struct ifaddr *ifa;
622 	struct llentry *ln;
623 	struct mbuf *chain;
624 	struct nd_neighbor_advert *nd_na;
625 	struct in6_addr daddr6, taddr6;
626 	struct sockaddr_in6 sin6;
627 	union nd_opts ndopts;
628 	u_char linkhdr[LLE_MAX_LINKHDR];
629 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
630 	char *lladdr;
631 	size_t linkhdrsize;
632 	int flags, is_override, is_router, is_solicited;
633 	int lladdr_off, lladdrlen, checklink;
634 
635 	NET_EPOCH_ASSERT();
636 
637 	chain = NULL;
638 	ln = NULL;
639 	checklink = 0;
640 
641 	/* RFC 6980: Nodes MUST silently ignore fragments */
642 	if(m->m_flags & M_FRAGMENTED)
643 		goto freeit;
644 
645 	ifp = m->m_pkthdr.rcvif;
646 	ip6 = mtod(m, struct ip6_hdr *);
647 	if (ip6->ip6_hlim != 255) {
648 		nd6log((LOG_ERR,
649 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
650 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
651 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
652 		goto bad;
653 	}
654 
655 	m = m_pullup(m, off + icmp6len);
656 	if (m == NULL) {
657 		IP6STAT_INC(ip6s_exthdrtoolong);
658 		return;
659 	}
660 	ip6 = mtod(m, struct ip6_hdr *);
661 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
662 
663 	flags = nd_na->nd_na_flags_reserved;
664 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
665 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
666 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
667 
668 	taddr6 = nd_na->nd_na_target;
669 	if (in6_setscope(&taddr6, ifp, NULL))
670 		goto bad;	/* XXX: impossible */
671 
672 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
673 		nd6log((LOG_ERR,
674 		    "nd6_na_input: invalid target address %s\n",
675 		    ip6_sprintf(ip6bufs, &taddr6)));
676 		goto bad;
677 	}
678 
679 	daddr6 = ip6->ip6_dst;
680 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
681 		if (is_solicited) {
682 			nd6log((LOG_ERR,
683 			    "nd6_na_input: a solicited adv is multicasted\n"));
684 			goto bad;
685 		}
686 
687 	icmp6len -= sizeof(*nd_na);
688 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
689 	if (nd6_options(&ndopts) < 0) {
690 		nd6log((LOG_INFO,
691 		    "nd6_na_input: invalid ND option, ignored\n"));
692 		/* nd6_options have incremented stats */
693 		goto freeit;
694 	}
695 
696 	lladdr = NULL;
697 	lladdrlen = 0;
698 	if (ndopts.nd_opts_tgt_lladdr) {
699 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
700 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
701 	}
702 
703 	/*
704 	 * This effectively disables the DAD check on a non-master CARP
705 	 * address.
706 	 */
707 	if (ifp->if_carp)
708 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
709 	else
710 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
711 
712 	/*
713 	 * Target address matches one of my interface address.
714 	 *
715 	 * If my address is tentative, this means that there's somebody
716 	 * already using the same address as mine.  This indicates DAD failure.
717 	 * This is defined in RFC 2462.
718 	 *
719 	 * Otherwise, process as defined in RFC 2461.
720 	 */
721 	if (ifa
722 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
723 		nd6_dad_na_input(ifa);
724 		ifa_free(ifa);
725 		goto freeit;
726 	}
727 
728 	/* Just for safety, maybe unnecessary. */
729 	if (ifa) {
730 		ifa_free(ifa);
731 		log(LOG_ERR,
732 		    "nd6_na_input: duplicate IP6 address %s\n",
733 		    ip6_sprintf(ip6bufs, &taddr6));
734 		goto freeit;
735 	}
736 
737 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
738 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
739 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
740 		    ifp->if_addrlen, lladdrlen - 2));
741 		goto bad;
742 	}
743 
744 	/*
745 	 * If no neighbor cache entry is found, NA SHOULD silently be
746 	 * discarded.
747 	 */
748 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
749 	if (ln == NULL) {
750 		goto freeit;
751 	}
752 
753 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
754 		/*
755 		 * If the link-layer has address, and no lladdr option came,
756 		 * discard the packet.
757 		 */
758 		if (ifp->if_addrlen && lladdr == NULL) {
759 			goto freeit;
760 		}
761 
762 		/*
763 		 * Record link-layer address, and update the state.
764 		 */
765 		linkhdrsize = sizeof(linkhdr);
766 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
767 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
768 			return;
769 
770 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
771 		    lladdr_off) == 0) {
772 			ln = NULL;
773 			goto freeit;
774 		}
775 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
776 		if (is_solicited)
777 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
778 		else
779 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
780 		if ((ln->ln_router = is_router) != 0) {
781 			/*
782 			 * This means a router's state has changed from
783 			 * non-reachable to probably reachable, and might
784 			 * affect the status of associated prefixes..
785 			 */
786 			checklink = 1;
787 		}
788 	} else {
789 		int llchange;
790 
791 		/*
792 		 * Check if the link-layer address has changed or not.
793 		 */
794 		if (lladdr == NULL)
795 			llchange = 0;
796 		else {
797 			if (ln->la_flags & LLE_VALID) {
798 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
799 					llchange = 1;
800 				else
801 					llchange = 0;
802 			} else
803 				llchange = 1;
804 		}
805 
806 		/*
807 		 * This is VERY complex.  Look at it with care.
808 		 *
809 		 * override solicit lladdr llchange	action
810 		 *					(L: record lladdr)
811 		 *
812 		 *	0	0	n	--	(2c)
813 		 *	0	0	y	n	(2b) L
814 		 *	0	0	y	y	(1)    REACHABLE->STALE
815 		 *	0	1	n	--	(2c)   *->REACHABLE
816 		 *	0	1	y	n	(2b) L *->REACHABLE
817 		 *	0	1	y	y	(1)    REACHABLE->STALE
818 		 *	1	0	n	--	(2a)
819 		 *	1	0	y	n	(2a) L
820 		 *	1	0	y	y	(2a) L *->STALE
821 		 *	1	1	n	--	(2a)   *->REACHABLE
822 		 *	1	1	y	n	(2a) L *->REACHABLE
823 		 *	1	1	y	y	(2a) L *->REACHABLE
824 		 */
825 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
826 			/*
827 			 * If state is REACHABLE, make it STALE.
828 			 * no other updates should be done.
829 			 */
830 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
831 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
832 			goto freeit;
833 		} else if (is_override				   /* (2a) */
834 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
835 			|| lladdr == NULL) {			   /* (2c) */
836 			/*
837 			 * Update link-local address, if any.
838 			 */
839 			if (lladdr != NULL) {
840 				linkhdrsize = sizeof(linkhdr);
841 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
842 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
843 					goto freeit;
844 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
845 				    linkhdrsize, lladdr_off) == 0) {
846 					ln = NULL;
847 					goto freeit;
848 				}
849 				EVENTHANDLER_INVOKE(lle_event, ln,
850 				    LLENTRY_RESOLVED);
851 			}
852 
853 			/*
854 			 * If solicited, make the state REACHABLE.
855 			 * If not solicited and the link-layer address was
856 			 * changed, make it STALE.
857 			 */
858 			if (is_solicited)
859 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
860 			else {
861 				if (lladdr != NULL && llchange)
862 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
863 			}
864 		}
865 
866 		if (ln->ln_router && !is_router) {
867 			/*
868 			 * The peer dropped the router flag.
869 			 * Remove the sender from the Default Router List and
870 			 * update the Destination Cache entries.
871 			 */
872 			struct ifnet *nd6_ifp;
873 
874 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
875 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
876 			    (ND_IFINFO(nd6_ifp)->flags &
877 			     ND6_IFF_ACCEPT_RTADV) != 0)
878 				/*
879 				 * Even if the neighbor is not in the default
880 				 * router list, the neighbor may be used as a
881 				 * next hop for some destinations (e.g. redirect
882 				 * case). So we must call rt6_flush explicitly.
883 				 */
884 				rt6_flush(&ip6->ip6_src, ifp);
885 		}
886 		ln->ln_router = is_router;
887 	}
888         /* XXX - QL
889 	 *  Does this matter?
890 	 *  rt->rt_flags &= ~RTF_REJECT;
891 	 */
892 	ln->la_asked = 0;
893 	if (ln->la_hold != NULL) {
894 		memset(&sin6, 0, sizeof(sin6));
895 		nd6_grab_holdchain(ln, &chain, &sin6);
896 	}
897  freeit:
898 	if (ln != NULL)
899 		LLE_WUNLOCK(ln);
900 
901 	if (chain != NULL)
902 		nd6_flush_holdchain(ifp, chain, &sin6);
903 
904 	if (checklink)
905 		pfxlist_onlink_check();
906 
907 	m_freem(m);
908 	return;
909 
910  bad:
911 	if (ln != NULL)
912 		LLE_WUNLOCK(ln);
913 
914 	ICMP6STAT_INC(icp6s_badna);
915 	m_freem(m);
916 }
917 
918 /*
919  * Neighbor advertisement output handling.
920  *
921  * Based on RFC 2461
922  *
923  * the following items are not implemented yet:
924  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
925  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
926  *
927  * tlladdr - 1 if include target link-layer address
928  * sdl0 - sockaddr_dl (= proxy NA) or NULL
929  */
930 static void
931 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
932     const struct in6_addr *taddr6, u_long flags, int tlladdr,
933     struct sockaddr *sdl0, u_int fibnum)
934 {
935 	struct mbuf *m;
936 	struct m_tag *mtag;
937 	struct ip6_hdr *ip6;
938 	struct nd_neighbor_advert *nd_na;
939 	struct ip6_moptions im6o;
940 	struct in6_addr daddr6, dst6, src6;
941 	uint32_t scopeid;
942 
943 	int icmp6len, maxlen, error;
944 	caddr_t mac = NULL;
945 
946 	daddr6 = *daddr6_0;	/* make a local copy for modification */
947 
948 	/* estimate the size of message */
949 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
950 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
951 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
952 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
953 	    __func__, max_linkhdr, maxlen, MCLBYTES));
954 
955 	if (max_linkhdr + maxlen > MHLEN)
956 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
957 	else
958 		m = m_gethdr(M_NOWAIT, MT_DATA);
959 	if (m == NULL)
960 		return;
961 	M_SETFIB(m, fibnum);
962 
963 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
964 		m->m_flags |= M_MCAST;
965 		im6o.im6o_multicast_ifp = ifp;
966 		im6o.im6o_multicast_hlim = 255;
967 		im6o.im6o_multicast_loop = 0;
968 	}
969 
970 	icmp6len = sizeof(*nd_na);
971 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
972 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
973 
974 	/* fill neighbor advertisement packet */
975 	ip6 = mtod(m, struct ip6_hdr *);
976 	ip6->ip6_flow = 0;
977 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
978 	ip6->ip6_vfc |= IPV6_VERSION;
979 	ip6->ip6_nxt = IPPROTO_ICMPV6;
980 	ip6->ip6_hlim = 255;
981 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
982 		/* reply to DAD */
983 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
984 		daddr6.s6_addr16[1] = 0;
985 		daddr6.s6_addr32[1] = 0;
986 		daddr6.s6_addr32[2] = 0;
987 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
988 		if (in6_setscope(&daddr6, ifp, NULL))
989 			goto bad;
990 
991 		flags &= ~ND_NA_FLAG_SOLICITED;
992 	}
993 	ip6->ip6_dst = daddr6;
994 
995 	/*
996 	 * Select a source whose scope is the same as that of the dest.
997 	 */
998 	in6_splitscope(&daddr6, &dst6, &scopeid);
999 	error = in6_selectsrc_addr(fibnum, &dst6,
1000 	    scopeid, ifp, &src6, NULL);
1001 	if (error) {
1002 		char ip6buf[INET6_ADDRSTRLEN];
1003 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1004 		    "determined: dst=%s, error=%d\n",
1005 		    ip6_sprintf(ip6buf, &daddr6), error));
1006 		goto bad;
1007 	}
1008 	ip6->ip6_src = src6;
1009 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1010 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1011 	nd_na->nd_na_code = 0;
1012 	nd_na->nd_na_target = *taddr6;
1013 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1014 
1015 	/*
1016 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1017 	 * see nd6_ns_input() for details.
1018 	 * Basically, if NS packet is sent to unicast/anycast addr,
1019 	 * target lladdr option SHOULD NOT be included.
1020 	 */
1021 	if (tlladdr) {
1022 		/*
1023 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1024 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1025 		 * my address) use lladdr configured for the interface.
1026 		 */
1027 		if (sdl0 == NULL) {
1028 			if (ifp->if_carp)
1029 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1030 			if (mac == NULL)
1031 				mac = nd6_ifptomac(ifp);
1032 		} else if (sdl0->sa_family == AF_LINK) {
1033 			struct sockaddr_dl *sdl;
1034 			sdl = (struct sockaddr_dl *)sdl0;
1035 			if (sdl->sdl_alen == ifp->if_addrlen)
1036 				mac = LLADDR(sdl);
1037 		}
1038 	}
1039 	if (tlladdr && mac) {
1040 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1041 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1042 
1043 		/* roundup to 8 bytes alignment! */
1044 		optlen = (optlen + 7) & ~7;
1045 
1046 		m->m_pkthdr.len += optlen;
1047 		m->m_len += optlen;
1048 		icmp6len += optlen;
1049 		bzero((caddr_t)nd_opt, optlen);
1050 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1051 		nd_opt->nd_opt_len = optlen >> 3;
1052 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1053 	} else
1054 		flags &= ~ND_NA_FLAG_OVERRIDE;
1055 
1056 	ip6->ip6_plen = htons((u_short)icmp6len);
1057 	nd_na->nd_na_flags_reserved = flags;
1058 	nd_na->nd_na_cksum = 0;
1059 	nd_na->nd_na_cksum =
1060 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1061 
1062 	if (send_sendso_input_hook != NULL) {
1063 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1064 		    sizeof(unsigned short), M_NOWAIT);
1065 		if (mtag == NULL)
1066 			goto bad;
1067 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1068 		m_tag_prepend(m, mtag);
1069 	}
1070 
1071 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1072 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1073 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1074 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1075 
1076 	return;
1077 
1078   bad:
1079 	m_freem(m);
1080 }
1081 
1082 #ifndef BURN_BRIDGES
1083 void
1084 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1085     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1086     struct sockaddr *sdl0)
1087 {
1088 
1089 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1090 	    RT_DEFAULT_FIB);
1091 }
1092 #endif
1093 
1094 caddr_t
1095 nd6_ifptomac(struct ifnet *ifp)
1096 {
1097 	switch (ifp->if_type) {
1098 	case IFT_ETHER:
1099 	case IFT_IEEE1394:
1100 	case IFT_L2VLAN:
1101 	case IFT_INFINIBAND:
1102 	case IFT_BRIDGE:
1103 		return IF_LLADDR(ifp);
1104 	default:
1105 		return NULL;
1106 	}
1107 }
1108 
1109 struct dadq {
1110 	TAILQ_ENTRY(dadq) dad_list;
1111 	struct ifaddr *dad_ifa;
1112 	int dad_count;		/* max NS to send */
1113 	int dad_ns_tcount;	/* # of trials to send NS */
1114 	int dad_ns_ocount;	/* NS sent so far */
1115 	int dad_ns_icount;
1116 	int dad_na_icount;
1117 	int dad_ns_lcount;	/* looped back NS */
1118 	int dad_loopbackprobe;	/* probing state for loopback detection */
1119 	struct callout dad_timer_ch;
1120 	struct vnet *dad_vnet;
1121 	u_int dad_refcnt;
1122 #define	ND_OPT_NONCE_LEN32 \
1123 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1124 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1125 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1126 };
1127 
1128 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1129 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1130 #define	V_dadq			VNET(dadq)
1131 #define	V_dad_rwlock		VNET(dad_rwlock)
1132 
1133 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1134 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1135 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1136 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1137 
1138 static void
1139 nd6_dad_add(struct dadq *dp)
1140 {
1141 
1142 	DADQ_WLOCK();
1143 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1144 	dp->dad_ondadq = true;
1145 	DADQ_WUNLOCK();
1146 }
1147 
1148 static void
1149 nd6_dad_del(struct dadq *dp)
1150 {
1151 
1152 	DADQ_WLOCK();
1153 	if (dp->dad_ondadq) {
1154 		/*
1155 		 * Remove dp from the dadq and release the dadq's
1156 		 * reference.
1157 		 */
1158 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1159 		dp->dad_ondadq = false;
1160 		DADQ_WUNLOCK();
1161 		nd6_dad_rele(dp);
1162 	} else
1163 		DADQ_WUNLOCK();
1164 }
1165 
1166 static struct dadq *
1167 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1168 {
1169 	struct dadq *dp;
1170 
1171 	DADQ_RLOCK();
1172 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1173 		if (dp->dad_ifa != ifa)
1174 			continue;
1175 		/*
1176 		 * Skip if the nonce matches the received one.
1177 		 * +2 in the length is required because of type and
1178 		 * length fields are included in a header.
1179 		 */
1180 		if (n != NULL &&
1181 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1182 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1183 		        ND_OPT_NONCE_LEN) == 0) {
1184 			dp->dad_ns_lcount++;
1185 			continue;
1186 		}
1187 		refcount_acquire(&dp->dad_refcnt);
1188 		break;
1189 	}
1190 	DADQ_RUNLOCK();
1191 
1192 	return (dp);
1193 }
1194 
1195 static void
1196 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1197 {
1198 
1199 	NET_EPOCH_ASSERT();
1200 
1201 	if (send_ns != 0)
1202 		nd6_dad_ns_output(dp);
1203 	callout_reset(&dp->dad_timer_ch, ticks,
1204 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1205 }
1206 
1207 static void
1208 nd6_dad_stoptimer(struct dadq *dp)
1209 {
1210 
1211 	callout_drain(&dp->dad_timer_ch);
1212 }
1213 
1214 static void
1215 nd6_dad_rele(struct dadq *dp)
1216 {
1217 
1218 	if (refcount_release(&dp->dad_refcnt)) {
1219 		ifa_free(dp->dad_ifa);
1220 		free(dp, M_IP6NDP);
1221 	}
1222 }
1223 
1224 void
1225 nd6_dad_init(void)
1226 {
1227 
1228 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1229 	TAILQ_INIT(&V_dadq);
1230 }
1231 
1232 /*
1233  * Start Duplicate Address Detection (DAD) for specified interface address.
1234  */
1235 void
1236 nd6_dad_start(struct ifaddr *ifa, int delay)
1237 {
1238 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1239 	struct dadq *dp;
1240 	char ip6buf[INET6_ADDRSTRLEN];
1241 	struct epoch_tracker et;
1242 
1243 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1244 	    ("starting DAD on non-tentative address %p", ifa));
1245 
1246 	/*
1247 	 * If we don't need DAD, don't do it.
1248 	 * There are several cases:
1249 	 * - DAD is disabled globally or on the interface
1250 	 * - the interface address is anycast
1251 	 */
1252 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1253 	    V_ip6_dad_count == 0 ||
1254 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1255 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1256 		return;
1257 	}
1258 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1259 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1260 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1261 		return;
1262 
1263 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1264 		/*
1265 		 * DAD is already in progress.  Let the existing entry
1266 		 * finish it.
1267 		 */
1268 		nd6_dad_rele(dp);
1269 		return;
1270 	}
1271 
1272 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1273 	if (dp == NULL) {
1274 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1275 			"%s(%s)\n",
1276 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1277 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1278 		return;
1279 	}
1280 	callout_init(&dp->dad_timer_ch, 0);
1281 #ifdef VIMAGE
1282 	dp->dad_vnet = curvnet;
1283 #endif
1284 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1285 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1286 
1287 	/*
1288 	 * Send NS packet for DAD, ip6_dad_count times.
1289 	 * Note that we must delay the first transmission, if this is the
1290 	 * first packet to be sent from the interface after interface
1291 	 * (re)initialization.
1292 	 */
1293 	dp->dad_ifa = ifa;
1294 	ifa_ref(dp->dad_ifa);
1295 	dp->dad_count = V_ip6_dad_count;
1296 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1297 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1298 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1299 
1300 	/* Add this to the dadq and add a reference for the dadq. */
1301 	refcount_init(&dp->dad_refcnt, 1);
1302 	nd6_dad_add(dp);
1303 	NET_EPOCH_ENTER(et);
1304 	nd6_dad_starttimer(dp, delay, 0);
1305 	NET_EPOCH_EXIT(et);
1306 }
1307 
1308 /*
1309  * terminate DAD unconditionally.  used for address removals.
1310  */
1311 void
1312 nd6_dad_stop(struct ifaddr *ifa)
1313 {
1314 	struct dadq *dp;
1315 
1316 	dp = nd6_dad_find(ifa, NULL);
1317 	if (!dp) {
1318 		/* DAD wasn't started yet */
1319 		return;
1320 	}
1321 
1322 	nd6_dad_stoptimer(dp);
1323 	nd6_dad_del(dp);
1324 
1325 	/* Release this function's reference, acquired by nd6_dad_find(). */
1326 	nd6_dad_rele(dp);
1327 }
1328 
1329 static void
1330 nd6_dad_timer(struct dadq *dp)
1331 {
1332 	CURVNET_SET(dp->dad_vnet);
1333 	struct ifaddr *ifa = dp->dad_ifa;
1334 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1335 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1336 	char ip6buf[INET6_ADDRSTRLEN];
1337 	struct epoch_tracker et;
1338 
1339 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1340 
1341 	NET_EPOCH_ENTER(et);
1342 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1343 		/* Do not need DAD for ifdisabled interface. */
1344 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1345 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1346 		goto err;
1347 	}
1348 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1349 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1350 			"%s(%s)\n",
1351 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1352 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1353 		goto err;
1354 	}
1355 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1356 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1357 			"%s(%s)\n",
1358 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1359 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1360 		goto err;
1361 	}
1362 
1363 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1364 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1365 	    (((ifp->if_flags & IFF_UP) == 0) ||
1366 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1367 		nd6log((LOG_INFO, "%s: could not run DAD "
1368 		    "because the interface was down or not running.\n",
1369 		    if_name(ifa->ifa_ifp)));
1370 		goto err;
1371 	}
1372 
1373 	/* Need more checks? */
1374 	if (dp->dad_ns_ocount < dp->dad_count) {
1375 		/*
1376 		 * We have more NS to go.  Send NS packet for DAD.
1377 		 */
1378 		nd6_dad_starttimer(dp,
1379 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1380 		goto done;
1381 	} else {
1382 		/*
1383 		 * We have transmitted sufficient number of DAD packets.
1384 		 * See what we've got.
1385 		 */
1386 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1387 			/* We've seen NS or NA, means DAD has failed. */
1388 			nd6_dad_duplicated(ifa, dp);
1389 		else if (V_dad_enhanced != 0 &&
1390 		    dp->dad_ns_lcount > 0 &&
1391 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1392 			/*
1393 			 * Sec. 4.1 in RFC 7527 requires transmission of
1394 			 * additional probes until the loopback condition
1395 			 * becomes clear when a looped back probe is detected.
1396 			 */
1397 			log(LOG_ERR, "%s: a looped back NS message is "
1398 			    "detected during DAD for %s.  "
1399 			    "Another DAD probes are being sent.\n",
1400 			    if_name(ifa->ifa_ifp),
1401 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1402 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1403 			/*
1404 			 * Send an NS immediately and increase dad_count by
1405 			 * V_nd6_mmaxtries - 1.
1406 			 */
1407 			dp->dad_count =
1408 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1409 			nd6_dad_starttimer(dp,
1410 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1411 			    1);
1412 			goto done;
1413 		} else {
1414 			/*
1415 			 * We are done with DAD.  No NA came, no NS came.
1416 			 * No duplicate address found.  Check IFDISABLED flag
1417 			 * again in case that it is changed between the
1418 			 * beginning of this function and here.
1419 			 */
1420 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1421 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1422 
1423 			nd6log((LOG_DEBUG,
1424 			    "%s: DAD complete for %s - no duplicates found\n",
1425 			    if_name(ifa->ifa_ifp),
1426 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1427 			if (dp->dad_ns_lcount > 0)
1428 				log(LOG_ERR, "%s: DAD completed while "
1429 				    "a looped back NS message is detected "
1430 				    "during DAD for %s.\n",
1431 				    if_name(ifa->ifa_ifp),
1432 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1433 		}
1434 	}
1435 err:
1436 	nd6_dad_del(dp);
1437 done:
1438 	NET_EPOCH_EXIT(et);
1439 	CURVNET_RESTORE();
1440 }
1441 
1442 static void
1443 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1444 {
1445 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1446 	struct ifnet *ifp;
1447 	char ip6buf[INET6_ADDRSTRLEN];
1448 
1449 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1450 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1451 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1452 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1453 	    dp->dad_na_icount);
1454 
1455 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1456 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1457 
1458 	ifp = ifa->ifa_ifp;
1459 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1460 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1461 	log(LOG_ERR, "%s: manual intervention required\n",
1462 	    if_name(ifp));
1463 
1464 	/*
1465 	 * If the address is a link-local address formed from an interface
1466 	 * identifier based on the hardware address which is supposed to be
1467 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1468 	 * operation on the interface SHOULD be disabled.
1469 	 * [RFC 4862, Section 5.4.5]
1470 	 */
1471 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1472 		struct in6_addr in6;
1473 
1474 		/*
1475 		 * To avoid over-reaction, we only apply this logic when we are
1476 		 * very sure that hardware addresses are supposed to be unique.
1477 		 */
1478 		switch (ifp->if_type) {
1479 		case IFT_ETHER:
1480 		case IFT_ATM:
1481 		case IFT_IEEE1394:
1482 		case IFT_INFINIBAND:
1483 			in6 = ia->ia_addr.sin6_addr;
1484 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1485 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1486 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1487 				log(LOG_ERR, "%s: possible hardware address "
1488 				    "duplication detected, disable IPv6\n",
1489 				    if_name(ifp));
1490 			}
1491 			break;
1492 		}
1493 	}
1494 }
1495 
1496 static void
1497 nd6_dad_ns_output(struct dadq *dp)
1498 {
1499 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1500 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1501 	int i;
1502 
1503 	dp->dad_ns_tcount++;
1504 	if ((ifp->if_flags & IFF_UP) == 0) {
1505 		return;
1506 	}
1507 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1508 		return;
1509 	}
1510 
1511 	dp->dad_ns_ocount++;
1512 	if (V_dad_enhanced != 0) {
1513 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1514 			dp->dad_nonce[i] = arc4random();
1515 		/*
1516 		 * XXXHRS: Note that in the case that
1517 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1518 		 * different nonces can be looped back in an unexpected
1519 		 * order.  The current implementation recognizes only
1520 		 * the latest nonce on the sender side.  Practically it
1521 		 * should work well in almost all cases.
1522 		 */
1523 	}
1524 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1525 	    (uint8_t *)&dp->dad_nonce[0]);
1526 }
1527 
1528 static void
1529 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1530 {
1531 	struct dadq *dp;
1532 
1533 	if (ifa == NULL)
1534 		panic("ifa == NULL in nd6_dad_ns_input");
1535 
1536 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1537 	if (V_dad_enhanced == 0)
1538 		ndopt_nonce = NULL;
1539 	dp = nd6_dad_find(ifa, ndopt_nonce);
1540 	if (dp == NULL)
1541 		return;
1542 
1543 	dp->dad_ns_icount++;
1544 	nd6_dad_rele(dp);
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, NULL);
1556 	if (dp != NULL) {
1557 		dp->dad_na_icount++;
1558 		nd6_dad_rele(dp);
1559 	}
1560 }
1561