xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 7c1b51d6dc2e165ae7333373513b080f17cf79bd)
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/libkern.h>
44 #include <sys/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55 #include <sys/callout.h>
56 #include <sys/refcount.h>
57 
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/if_var.h>
62 #include <net/route.h>
63 #ifdef RADIX_MPATH
64 #include <net/radix_mpath.h>
65 #endif
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <net/if_llatbl.h>
71 #include <netinet6/in6_var.h>
72 #include <netinet6/in6_ifattach.h>
73 #include <netinet/ip6.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/scope6_var.h>
76 #include <netinet6/nd6.h>
77 #include <netinet/icmp6.h>
78 #include <netinet/ip_carp.h>
79 #include <netinet6/send.h>
80 
81 #define SDL(s) ((struct sockaddr_dl *)s)
82 
83 struct dadq;
84 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
85 static void nd6_dad_add(struct dadq *dp);
86 static void nd6_dad_del(struct dadq *dp);
87 static void nd6_dad_rele(struct dadq *);
88 static void nd6_dad_starttimer(struct dadq *, int, int);
89 static void nd6_dad_stoptimer(struct dadq *);
90 static void nd6_dad_timer(struct dadq *);
91 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
92 static void nd6_dad_ns_output(struct dadq *);
93 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
94 static void nd6_dad_na_input(struct ifaddr *);
95 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
96     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
97 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
98     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
99 
100 static VNET_DEFINE(int, dad_enhanced) = 1;
101 #define	V_dad_enhanced			VNET(dad_enhanced)
102 
103 SYSCTL_DECL(_net_inet6_ip6);
104 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
105     &VNET_NAME(dad_enhanced), 0,
106     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
107 
108 static VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to
109 						   transmit DAD packet */
110 #define	V_dad_maxtry			VNET(dad_maxtry)
111 
112 /*
113  * Input a Neighbor Solicitation Message.
114  *
115  * Based on RFC 2461
116  * Based on RFC 2462 (duplicate address detection)
117  */
118 void
119 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
120 {
121 	struct ifnet *ifp = m->m_pkthdr.rcvif;
122 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123 	struct nd_neighbor_solicit *nd_ns;
124 	struct in6_addr saddr6 = ip6->ip6_src;
125 	struct in6_addr daddr6 = ip6->ip6_dst;
126 	struct in6_addr taddr6;
127 	struct in6_addr myaddr6;
128 	char *lladdr = NULL;
129 	struct ifaddr *ifa = NULL;
130 	int lladdrlen = 0;
131 	int anycast = 0, proxy = 0, tentative = 0;
132 	int tlladdr;
133 	int rflag;
134 	union nd_opts ndopts;
135 	struct sockaddr_dl proxydl;
136 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
137 
138 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
139 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
140 		rflag = 0;
141 #ifndef PULLDOWN_TEST
142 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
143 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
144 #else
145 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
146 	if (nd_ns == NULL) {
147 		ICMP6STAT_INC(icp6s_tooshort);
148 		return;
149 	}
150 #endif
151 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
152 	taddr6 = nd_ns->nd_ns_target;
153 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
154 		goto bad;
155 
156 	if (ip6->ip6_hlim != 255) {
157 		nd6log((LOG_ERR,
158 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
159 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
160 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
161 		goto bad;
162 	}
163 
164 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
165 		/* dst has to be a solicited node multicast address. */
166 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
167 		    /* don't check ifindex portion */
168 		    daddr6.s6_addr32[1] == 0 &&
169 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
170 		    daddr6.s6_addr8[12] == 0xff) {
171 			; /* good */
172 		} else {
173 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
174 			    "(wrong ip6 dst)\n"));
175 			goto bad;
176 		}
177 	} else if (!V_nd6_onlink_ns_rfc4861) {
178 		struct sockaddr_in6 src_sa6;
179 
180 		/*
181 		 * According to recent IETF discussions, it is not a good idea
182 		 * to accept a NS from an address which would not be deemed
183 		 * to be a neighbor otherwise.  This point is expected to be
184 		 * clarified in future revisions of the specification.
185 		 */
186 		bzero(&src_sa6, sizeof(src_sa6));
187 		src_sa6.sin6_family = AF_INET6;
188 		src_sa6.sin6_len = sizeof(src_sa6);
189 		src_sa6.sin6_addr = saddr6;
190 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
191 			nd6log((LOG_INFO, "nd6_ns_input: "
192 				"NS packet from non-neighbor\n"));
193 			goto bad;
194 		}
195 	}
196 
197 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
198 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
199 		goto bad;
200 	}
201 
202 	icmp6len -= sizeof(*nd_ns);
203 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
204 	if (nd6_options(&ndopts) < 0) {
205 		nd6log((LOG_INFO,
206 		    "nd6_ns_input: invalid ND option, ignored\n"));
207 		/* nd6_options have incremented stats */
208 		goto freeit;
209 	}
210 
211 	if (ndopts.nd_opts_src_lladdr) {
212 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
213 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
214 	}
215 
216 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
217 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
218 		    "(link-layer address option)\n"));
219 		goto bad;
220 	}
221 
222 	/*
223 	 * Attaching target link-layer address to the NA?
224 	 * (RFC 2461 7.2.4)
225 	 *
226 	 * NS IP dst is unicast/anycast			MUST NOT add
227 	 * NS IP dst is solicited-node multicast	MUST add
228 	 *
229 	 * In implementation, we add target link-layer address by default.
230 	 * We do not add one in MUST NOT cases.
231 	 */
232 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
233 		tlladdr = 0;
234 	else
235 		tlladdr = 1;
236 
237 	/*
238 	 * Target address (taddr6) must be either:
239 	 * (1) Valid unicast/anycast address for my receiving interface,
240 	 * (2) Unicast address for which I'm offering proxy service, or
241 	 * (3) "tentative" address on which DAD is being performed.
242 	 */
243 	/* (1) and (3) check. */
244 	if (ifp->if_carp)
245 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
246 	else
247 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
248 
249 	/* (2) check. */
250 	if (ifa == NULL) {
251 		struct sockaddr_dl rt_gateway;
252 		struct rt_addrinfo info;
253 		struct sockaddr_in6 dst6;
254 
255 		bzero(&dst6, sizeof(dst6));
256 		dst6.sin6_len = sizeof(struct sockaddr_in6);
257 		dst6.sin6_family = AF_INET6;
258 		dst6.sin6_addr = taddr6;
259 
260 		bzero(&rt_gateway, sizeof(rt_gateway));
261 		rt_gateway.sdl_len = sizeof(rt_gateway);
262 		bzero(&info, sizeof(info));
263 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
264 
265 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
266 		    0, 0, &info) == 0) {
267 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
268 			    rt_gateway.sdl_family == AF_LINK) {
269 
270 				/*
271 				 * proxy NDP for single entry
272 				 */
273 				proxydl = *SDL(&rt_gateway);
274 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
275 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
276 				if (ifa)
277 					proxy = 1;
278 			}
279 		}
280 	}
281 	if (ifa == NULL) {
282 		/*
283 		 * We've got an NS packet, and we don't have that adddress
284 		 * assigned for us.  We MUST silently ignore it.
285 		 * See RFC2461 7.2.3.
286 		 */
287 		goto freeit;
288 	}
289 	myaddr6 = *IFA_IN6(ifa);
290 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
291 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
292 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
293 		goto freeit;
294 
295 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
296 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
297 		    "(if %d, NS packet %d)\n",
298 		    ip6_sprintf(ip6bufs, &taddr6),
299 		    ifp->if_addrlen, lladdrlen - 2));
300 		goto bad;
301 	}
302 
303 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
304 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
305 		    ip6_sprintf(ip6bufs, &saddr6)));
306 		goto freeit;
307 	}
308 
309 	/*
310 	 * We have neighbor solicitation packet, with target address equals to
311 	 * one of my tentative address.
312 	 *
313 	 * src addr	how to process?
314 	 * ---		---
315 	 * multicast	of course, invalid (rejected in ip6_input)
316 	 * unicast	somebody is doing address resolution -> ignore
317 	 * unspec	dup address detection
318 	 *
319 	 * The processing is defined in RFC 2462.
320 	 */
321 	if (tentative) {
322 		/*
323 		 * If source address is unspecified address, it is for
324 		 * duplicate address detection.
325 		 *
326 		 * If not, the packet is for addess resolution;
327 		 * silently ignore it.
328 		 */
329 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
330 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
331 
332 		goto freeit;
333 	}
334 
335 	/*
336 	 * If the source address is unspecified address, entries must not
337 	 * be created or updated.
338 	 * It looks that sender is performing DAD.  Output NA toward
339 	 * all-node multicast address, to tell the sender that I'm using
340 	 * the address.
341 	 * S bit ("solicited") must be zero.
342 	 */
343 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
344 		struct in6_addr in6_all;
345 
346 		in6_all = in6addr_linklocal_allnodes;
347 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
348 			goto bad;
349 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
350 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
351 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
352 		    M_GETFIB(m));
353 		goto freeit;
354 	}
355 
356 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
357 	    ND_NEIGHBOR_SOLICIT, 0);
358 
359 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
360 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
361 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
362 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
363  freeit:
364 	if (ifa != NULL)
365 		ifa_free(ifa);
366 	m_freem(m);
367 	return;
368 
369  bad:
370 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
371 		ip6_sprintf(ip6bufs, &saddr6)));
372 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
373 		ip6_sprintf(ip6bufs, &daddr6)));
374 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
375 		ip6_sprintf(ip6bufs, &taddr6)));
376 	ICMP6STAT_INC(icp6s_badns);
377 	if (ifa != NULL)
378 		ifa_free(ifa);
379 	m_freem(m);
380 }
381 
382 /*
383  * Output a Neighbor Solicitation Message. Caller specifies:
384  *	- ICMP6 header source IP6 address
385  *	- ND6 header target IP6 address
386  *	- ND6 header source datalink address
387  *
388  * Based on RFC 2461
389  * Based on RFC 2462 (duplicate address detection)
390  *
391  *    ln - for source address determination
392  * nonce - If non-NULL, NS is used for duplicate address detection and
393  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
394  */
395 static void
396 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
397     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
398     uint8_t *nonce, u_int fibnum)
399 {
400 	struct mbuf *m;
401 	struct m_tag *mtag;
402 	struct ip6_hdr *ip6;
403 	struct nd_neighbor_solicit *nd_ns;
404 	struct ip6_moptions im6o;
405 	int icmp6len;
406 	int maxlen;
407 	caddr_t mac;
408 
409 	if (IN6_IS_ADDR_MULTICAST(taddr6))
410 		return;
411 
412 	/* estimate the size of message */
413 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
414 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
415 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
416 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
417 	    __func__, max_linkhdr, maxlen, MCLBYTES));
418 
419 	if (max_linkhdr + maxlen > MHLEN)
420 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
421 	else
422 		m = m_gethdr(M_NOWAIT, MT_DATA);
423 	if (m == NULL)
424 		return;
425 	M_SETFIB(m, fibnum);
426 
427 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
428 		m->m_flags |= M_MCAST;
429 		im6o.im6o_multicast_ifp = ifp;
430 		im6o.im6o_multicast_hlim = 255;
431 		im6o.im6o_multicast_loop = 0;
432 	}
433 
434 	icmp6len = sizeof(*nd_ns);
435 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
436 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
437 
438 	/* fill neighbor solicitation packet */
439 	ip6 = mtod(m, struct ip6_hdr *);
440 	ip6->ip6_flow = 0;
441 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
442 	ip6->ip6_vfc |= IPV6_VERSION;
443 	/* ip6->ip6_plen will be set later */
444 	ip6->ip6_nxt = IPPROTO_ICMPV6;
445 	ip6->ip6_hlim = 255;
446 	if (daddr6)
447 		ip6->ip6_dst = *daddr6;
448 	else {
449 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
450 		ip6->ip6_dst.s6_addr16[1] = 0;
451 		ip6->ip6_dst.s6_addr32[1] = 0;
452 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
453 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
454 		ip6->ip6_dst.s6_addr8[12] = 0xff;
455 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
456 			goto bad;
457 	}
458 	if (nonce == NULL) {
459 		struct ifaddr *ifa = NULL;
460 
461 		/*
462 		 * RFC2461 7.2.2:
463 		 * "If the source address of the packet prompting the
464 		 * solicitation is the same as one of the addresses assigned
465 		 * to the outgoing interface, that address SHOULD be placed
466 		 * in the IP Source Address of the outgoing solicitation.
467 		 * Otherwise, any one of the addresses assigned to the
468 		 * interface should be used."
469 		 *
470 		 * We use the source address for the prompting packet
471 		 * (saddr6), if saddr6 belongs to the outgoing interface.
472 		 * Otherwise, we perform the source address selection as usual.
473 		 */
474 
475 		if (saddr6 != NULL)
476 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
477 		if (ifa != NULL) {
478 			/* ip6_src set already. */
479 			ip6->ip6_src = *saddr6;
480 			ifa_free(ifa);
481 		} else {
482 			int error;
483 			struct in6_addr dst6, src6;
484 			uint32_t scopeid;
485 
486 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
487 			error = in6_selectsrc_addr(fibnum, &dst6,
488 			    scopeid, ifp, &src6, NULL);
489 			if (error) {
490 				char ip6buf[INET6_ADDRSTRLEN];
491 				nd6log((LOG_DEBUG, "%s: source can't be "
492 				    "determined: dst=%s, error=%d\n", __func__,
493 				    ip6_sprintf(ip6buf, &dst6),
494 				    error));
495 				goto bad;
496 			}
497 			ip6->ip6_src = src6;
498 		}
499 	} else {
500 		/*
501 		 * Source address for DAD packet must always be IPv6
502 		 * unspecified address. (0::0)
503 		 * We actually don't have to 0-clear the address (we did it
504 		 * above), but we do so here explicitly to make the intention
505 		 * clearer.
506 		 */
507 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
508 	}
509 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
510 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
511 	nd_ns->nd_ns_code = 0;
512 	nd_ns->nd_ns_reserved = 0;
513 	nd_ns->nd_ns_target = *taddr6;
514 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
515 
516 	/*
517 	 * Add source link-layer address option.
518 	 *
519 	 *				spec		implementation
520 	 *				---		---
521 	 * DAD packet			MUST NOT	do not add the option
522 	 * there's no link layer address:
523 	 *				impossible	do not add the option
524 	 * there's link layer address:
525 	 *	Multicast NS		MUST add one	add the option
526 	 *	Unicast NS		SHOULD add one	add the option
527 	 */
528 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
529 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
530 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
531 		/* 8 byte alignments... */
532 		optlen = (optlen + 7) & ~7;
533 
534 		m->m_pkthdr.len += optlen;
535 		m->m_len += optlen;
536 		icmp6len += optlen;
537 		bzero((caddr_t)nd_opt, optlen);
538 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
539 		nd_opt->nd_opt_len = optlen >> 3;
540 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
541 	}
542 	/*
543 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
544 	 * This behavior is documented as Enhanced Duplicate Address
545 	 * Detection in RFC 7527.
546 	 * net.inet6.ip6.dad_enhanced=0 disables this.
547 	 */
548 	if (V_dad_enhanced != 0 && nonce != NULL) {
549 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
550 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
551 		/* 8-byte alignment is required. */
552 		optlen = (optlen + 7) & ~7;
553 
554 		m->m_pkthdr.len += optlen;
555 		m->m_len += optlen;
556 		icmp6len += optlen;
557 		bzero((caddr_t)nd_opt, optlen);
558 		nd_opt->nd_opt_type = ND_OPT_NONCE;
559 		nd_opt->nd_opt_len = optlen >> 3;
560 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
561 	}
562 	ip6->ip6_plen = htons((u_short)icmp6len);
563 	nd_ns->nd_ns_cksum = 0;
564 	nd_ns->nd_ns_cksum =
565 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
566 
567 	if (send_sendso_input_hook != NULL) {
568 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
569 			sizeof(unsigned short), M_NOWAIT);
570 		if (mtag == NULL)
571 			goto bad;
572 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
573 		m_tag_prepend(m, mtag);
574 	}
575 
576 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
577 	    &im6o, NULL, NULL);
578 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
579 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
580 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
581 
582 	return;
583 
584   bad:
585 	m_freem(m);
586 }
587 
588 #ifndef BURN_BRIDGES
589 void
590 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
591     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
592 {
593 
594 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
595 }
596 #endif
597 /*
598  * Neighbor advertisement input handling.
599  *
600  * Based on RFC 2461
601  * Based on RFC 2462 (duplicate address detection)
602  *
603  * the following items are not implemented yet:
604  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
605  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
606  */
607 void
608 nd6_na_input(struct mbuf *m, int off, int icmp6len)
609 {
610 	struct ifnet *ifp = m->m_pkthdr.rcvif;
611 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
612 	struct nd_neighbor_advert *nd_na;
613 	struct in6_addr daddr6 = ip6->ip6_dst;
614 	struct in6_addr taddr6;
615 	int flags;
616 	int is_router;
617 	int is_solicited;
618 	int is_override;
619 	char *lladdr = NULL;
620 	int lladdrlen = 0;
621 	int checklink = 0;
622 	struct ifaddr *ifa;
623 	struct llentry *ln = NULL;
624 	union nd_opts ndopts;
625 	struct mbuf *chain = NULL;
626 	struct sockaddr_in6 sin6;
627 	u_char linkhdr[LLE_MAX_LINKHDR];
628 	size_t linkhdrsize;
629 	int lladdr_off;
630 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
631 
632 	if (ip6->ip6_hlim != 255) {
633 		nd6log((LOG_ERR,
634 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
635 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
636 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
637 		goto bad;
638 	}
639 
640 #ifndef PULLDOWN_TEST
641 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
642 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
643 #else
644 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
645 	if (nd_na == NULL) {
646 		ICMP6STAT_INC(icp6s_tooshort);
647 		return;
648 	}
649 #endif
650 
651 	flags = nd_na->nd_na_flags_reserved;
652 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
653 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
654 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
655 	memset(&sin6, 0, sizeof(sin6));
656 
657 	taddr6 = nd_na->nd_na_target;
658 	if (in6_setscope(&taddr6, ifp, NULL))
659 		goto bad;	/* XXX: impossible */
660 
661 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
662 		nd6log((LOG_ERR,
663 		    "nd6_na_input: invalid target address %s\n",
664 		    ip6_sprintf(ip6bufs, &taddr6)));
665 		goto bad;
666 	}
667 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
668 		if (is_solicited) {
669 			nd6log((LOG_ERR,
670 			    "nd6_na_input: a solicited adv is multicasted\n"));
671 			goto bad;
672 		}
673 
674 	icmp6len -= sizeof(*nd_na);
675 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
676 	if (nd6_options(&ndopts) < 0) {
677 		nd6log((LOG_INFO,
678 		    "nd6_na_input: invalid ND option, ignored\n"));
679 		/* nd6_options have incremented stats */
680 		goto freeit;
681 	}
682 
683 	if (ndopts.nd_opts_tgt_lladdr) {
684 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
685 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
686 	}
687 
688 	/*
689 	 * This effectively disables the DAD check on a non-master CARP
690 	 * address.
691 	 */
692 	if (ifp->if_carp)
693 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
694 	else
695 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
696 
697 	/*
698 	 * Target address matches one of my interface address.
699 	 *
700 	 * If my address is tentative, this means that there's somebody
701 	 * already using the same address as mine.  This indicates DAD failure.
702 	 * This is defined in RFC 2462.
703 	 *
704 	 * Otherwise, process as defined in RFC 2461.
705 	 */
706 	if (ifa
707 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
708 		nd6_dad_na_input(ifa);
709 		ifa_free(ifa);
710 		goto freeit;
711 	}
712 
713 	/* Just for safety, maybe unnecessary. */
714 	if (ifa) {
715 		ifa_free(ifa);
716 		log(LOG_ERR,
717 		    "nd6_na_input: duplicate IP6 address %s\n",
718 		    ip6_sprintf(ip6bufs, &taddr6));
719 		goto freeit;
720 	}
721 
722 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
723 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
724 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
725 		    ifp->if_addrlen, lladdrlen - 2));
726 		goto bad;
727 	}
728 
729 	/*
730 	 * If no neighbor cache entry is found, NA SHOULD silently be
731 	 * discarded.
732 	 */
733 	IF_AFDATA_RLOCK(ifp);
734 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
735 	IF_AFDATA_RUNLOCK(ifp);
736 	if (ln == NULL) {
737 		goto freeit;
738 	}
739 
740 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
741 		/*
742 		 * If the link-layer has address, and no lladdr option came,
743 		 * discard the packet.
744 		 */
745 		if (ifp->if_addrlen && lladdr == NULL) {
746 			goto freeit;
747 		}
748 
749 		/*
750 		 * Record link-layer address, and update the state.
751 		 */
752 		linkhdrsize = sizeof(linkhdr);
753 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
754 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
755 			return;
756 
757 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
758 		    lladdr_off) == 0) {
759 			ln = NULL;
760 			goto freeit;
761 		}
762 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
763 		if (is_solicited)
764 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
765 		else
766 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
767 		if ((ln->ln_router = is_router) != 0) {
768 			/*
769 			 * This means a router's state has changed from
770 			 * non-reachable to probably reachable, and might
771 			 * affect the status of associated prefixes..
772 			 */
773 			checklink = 1;
774 		}
775 	} else {
776 		int llchange;
777 
778 		/*
779 		 * Check if the link-layer address has changed or not.
780 		 */
781 		if (lladdr == NULL)
782 			llchange = 0;
783 		else {
784 			if (ln->la_flags & LLE_VALID) {
785 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
786 					llchange = 1;
787 				else
788 					llchange = 0;
789 			} else
790 				llchange = 1;
791 		}
792 
793 		/*
794 		 * This is VERY complex.  Look at it with care.
795 		 *
796 		 * override solicit lladdr llchange	action
797 		 *					(L: record lladdr)
798 		 *
799 		 *	0	0	n	--	(2c)
800 		 *	0	0	y	n	(2b) L
801 		 *	0	0	y	y	(1)    REACHABLE->STALE
802 		 *	0	1	n	--	(2c)   *->REACHABLE
803 		 *	0	1	y	n	(2b) L *->REACHABLE
804 		 *	0	1	y	y	(1)    REACHABLE->STALE
805 		 *	1	0	n	--	(2a)
806 		 *	1	0	y	n	(2a) L
807 		 *	1	0	y	y	(2a) L *->STALE
808 		 *	1	1	n	--	(2a)   *->REACHABLE
809 		 *	1	1	y	n	(2a) L *->REACHABLE
810 		 *	1	1	y	y	(2a) L *->REACHABLE
811 		 */
812 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
813 			/*
814 			 * If state is REACHABLE, make it STALE.
815 			 * no other updates should be done.
816 			 */
817 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
818 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
819 			goto freeit;
820 		} else if (is_override				   /* (2a) */
821 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
822 			|| lladdr == NULL) {			   /* (2c) */
823 			/*
824 			 * Update link-local address, if any.
825 			 */
826 			if (lladdr != NULL) {
827 				linkhdrsize = sizeof(linkhdr);
828 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
829 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
830 					goto freeit;
831 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
832 				    linkhdrsize, lladdr_off) == 0) {
833 					ln = NULL;
834 					goto freeit;
835 				}
836 				EVENTHANDLER_INVOKE(lle_event, ln,
837 				    LLENTRY_RESOLVED);
838 			}
839 
840 			/*
841 			 * If solicited, make the state REACHABLE.
842 			 * If not solicited and the link-layer address was
843 			 * changed, make it STALE.
844 			 */
845 			if (is_solicited)
846 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
847 			else {
848 				if (lladdr != NULL && llchange)
849 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
850 			}
851 		}
852 
853 		if (ln->ln_router && !is_router) {
854 			/*
855 			 * The peer dropped the router flag.
856 			 * Remove the sender from the Default Router List and
857 			 * update the Destination Cache entries.
858 			 */
859 			struct ifnet *nd6_ifp;
860 
861 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
862 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
863 			    (ND_IFINFO(nd6_ifp)->flags &
864 			     ND6_IFF_ACCEPT_RTADV) != 0)
865 				/*
866 				 * Even if the neighbor is not in the default
867 				 * router list, the neighbor may be used as a
868 				 * next hop for some destinations (e.g. redirect
869 				 * case). So we must call rt6_flush explicitly.
870 				 */
871 				rt6_flush(&ip6->ip6_src, ifp);
872 		}
873 		ln->ln_router = is_router;
874 	}
875         /* XXX - QL
876 	 *  Does this matter?
877 	 *  rt->rt_flags &= ~RTF_REJECT;
878 	 */
879 	ln->la_asked = 0;
880 	if (ln->la_hold != NULL)
881 		nd6_grab_holdchain(ln, &chain, &sin6);
882  freeit:
883 	if (ln != NULL)
884 		LLE_WUNLOCK(ln);
885 
886 	if (chain != NULL)
887 		nd6_flush_holdchain(ifp, ifp, chain, &sin6);
888 
889 	if (checklink)
890 		pfxlist_onlink_check();
891 
892 	m_freem(m);
893 	return;
894 
895  bad:
896 	if (ln != NULL)
897 		LLE_WUNLOCK(ln);
898 
899 	ICMP6STAT_INC(icp6s_badna);
900 	m_freem(m);
901 }
902 
903 /*
904  * Neighbor advertisement output handling.
905  *
906  * Based on RFC 2461
907  *
908  * the following items are not implemented yet:
909  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
910  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
911  *
912  * tlladdr - 1 if include target link-layer address
913  * sdl0 - sockaddr_dl (= proxy NA) or NULL
914  */
915 static void
916 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
917     const struct in6_addr *taddr6, u_long flags, int tlladdr,
918     struct sockaddr *sdl0, u_int fibnum)
919 {
920 	struct mbuf *m;
921 	struct m_tag *mtag;
922 	struct ip6_hdr *ip6;
923 	struct nd_neighbor_advert *nd_na;
924 	struct ip6_moptions im6o;
925 	struct in6_addr daddr6, dst6, src6;
926 	uint32_t scopeid;
927 
928 	int icmp6len, maxlen, error;
929 	caddr_t mac = NULL;
930 
931 	daddr6 = *daddr6_0;	/* make a local copy for modification */
932 
933 	/* estimate the size of message */
934 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
935 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
936 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
937 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
938 	    __func__, max_linkhdr, maxlen, MCLBYTES));
939 
940 	if (max_linkhdr + maxlen > MHLEN)
941 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
942 	else
943 		m = m_gethdr(M_NOWAIT, MT_DATA);
944 	if (m == NULL)
945 		return;
946 	M_SETFIB(m, fibnum);
947 
948 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
949 		m->m_flags |= M_MCAST;
950 		im6o.im6o_multicast_ifp = ifp;
951 		im6o.im6o_multicast_hlim = 255;
952 		im6o.im6o_multicast_loop = 0;
953 	}
954 
955 	icmp6len = sizeof(*nd_na);
956 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
957 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
958 
959 	/* fill neighbor advertisement packet */
960 	ip6 = mtod(m, struct ip6_hdr *);
961 	ip6->ip6_flow = 0;
962 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
963 	ip6->ip6_vfc |= IPV6_VERSION;
964 	ip6->ip6_nxt = IPPROTO_ICMPV6;
965 	ip6->ip6_hlim = 255;
966 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
967 		/* reply to DAD */
968 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
969 		daddr6.s6_addr16[1] = 0;
970 		daddr6.s6_addr32[1] = 0;
971 		daddr6.s6_addr32[2] = 0;
972 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
973 		if (in6_setscope(&daddr6, ifp, NULL))
974 			goto bad;
975 
976 		flags &= ~ND_NA_FLAG_SOLICITED;
977 	}
978 	ip6->ip6_dst = daddr6;
979 
980 	/*
981 	 * Select a source whose scope is the same as that of the dest.
982 	 */
983 	in6_splitscope(&daddr6, &dst6, &scopeid);
984 	error = in6_selectsrc_addr(fibnum, &dst6,
985 	    scopeid, ifp, &src6, NULL);
986 	if (error) {
987 		char ip6buf[INET6_ADDRSTRLEN];
988 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
989 		    "determined: dst=%s, error=%d\n",
990 		    ip6_sprintf(ip6buf, &daddr6), error));
991 		goto bad;
992 	}
993 	ip6->ip6_src = src6;
994 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
995 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
996 	nd_na->nd_na_code = 0;
997 	nd_na->nd_na_target = *taddr6;
998 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
999 
1000 	/*
1001 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1002 	 * see nd6_ns_input() for details.
1003 	 * Basically, if NS packet is sent to unicast/anycast addr,
1004 	 * target lladdr option SHOULD NOT be included.
1005 	 */
1006 	if (tlladdr) {
1007 		/*
1008 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1009 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1010 		 * my address) use lladdr configured for the interface.
1011 		 */
1012 		if (sdl0 == NULL) {
1013 			if (ifp->if_carp)
1014 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1015 			if (mac == NULL)
1016 				mac = nd6_ifptomac(ifp);
1017 		} else if (sdl0->sa_family == AF_LINK) {
1018 			struct sockaddr_dl *sdl;
1019 			sdl = (struct sockaddr_dl *)sdl0;
1020 			if (sdl->sdl_alen == ifp->if_addrlen)
1021 				mac = LLADDR(sdl);
1022 		}
1023 	}
1024 	if (tlladdr && mac) {
1025 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1026 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1027 
1028 		/* roundup to 8 bytes alignment! */
1029 		optlen = (optlen + 7) & ~7;
1030 
1031 		m->m_pkthdr.len += optlen;
1032 		m->m_len += optlen;
1033 		icmp6len += optlen;
1034 		bzero((caddr_t)nd_opt, optlen);
1035 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1036 		nd_opt->nd_opt_len = optlen >> 3;
1037 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1038 	} else
1039 		flags &= ~ND_NA_FLAG_OVERRIDE;
1040 
1041 	ip6->ip6_plen = htons((u_short)icmp6len);
1042 	nd_na->nd_na_flags_reserved = flags;
1043 	nd_na->nd_na_cksum = 0;
1044 	nd_na->nd_na_cksum =
1045 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1046 
1047 	if (send_sendso_input_hook != NULL) {
1048 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1049 		    sizeof(unsigned short), M_NOWAIT);
1050 		if (mtag == NULL)
1051 			goto bad;
1052 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1053 		m_tag_prepend(m, mtag);
1054 	}
1055 
1056 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1057 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1058 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1059 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1060 
1061 	return;
1062 
1063   bad:
1064 	m_freem(m);
1065 }
1066 
1067 #ifndef BURN_BRIDGES
1068 void
1069 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1070     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1071     struct sockaddr *sdl0)
1072 {
1073 
1074 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1075 	    RT_DEFAULT_FIB);
1076 }
1077 #endif
1078 
1079 caddr_t
1080 nd6_ifptomac(struct ifnet *ifp)
1081 {
1082 	switch (ifp->if_type) {
1083 	case IFT_ARCNET:
1084 	case IFT_ETHER:
1085 	case IFT_FDDI:
1086 	case IFT_IEEE1394:
1087 	case IFT_L2VLAN:
1088 	case IFT_INFINIBAND:
1089 	case IFT_BRIDGE:
1090 	case IFT_ISO88025:
1091 		return IF_LLADDR(ifp);
1092 	default:
1093 		return NULL;
1094 	}
1095 }
1096 
1097 struct dadq {
1098 	TAILQ_ENTRY(dadq) dad_list;
1099 	struct ifaddr *dad_ifa;
1100 	int dad_count;		/* max NS to send */
1101 	int dad_ns_tcount;	/* # of trials to send NS */
1102 	int dad_ns_ocount;	/* NS sent so far */
1103 	int dad_ns_icount;
1104 	int dad_na_icount;
1105 	int dad_ns_lcount;	/* looped back NS */
1106 	int dad_loopbackprobe;	/* probing state for loopback detection */
1107 	struct callout dad_timer_ch;
1108 	struct vnet *dad_vnet;
1109 	u_int dad_refcnt;
1110 #define	ND_OPT_NONCE_LEN32 \
1111 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1112 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1113 };
1114 
1115 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1116 static VNET_DEFINE(struct rwlock, dad_rwlock);
1117 #define	V_dadq			VNET(dadq)
1118 #define	V_dad_rwlock		VNET(dad_rwlock)
1119 
1120 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1121 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1122 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1123 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1124 
1125 static void
1126 nd6_dad_add(struct dadq *dp)
1127 {
1128 
1129 	DADQ_WLOCK();
1130 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1131 	DADQ_WUNLOCK();
1132 }
1133 
1134 static void
1135 nd6_dad_del(struct dadq *dp)
1136 {
1137 
1138 	DADQ_WLOCK();
1139 	TAILQ_REMOVE(&V_dadq, dp, dad_list);
1140 	DADQ_WUNLOCK();
1141 	nd6_dad_rele(dp);
1142 }
1143 
1144 static struct dadq *
1145 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1146 {
1147 	struct dadq *dp;
1148 
1149 	DADQ_RLOCK();
1150 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1151 		if (dp->dad_ifa != ifa)
1152 			continue;
1153 		/*
1154 		 * Skip if the nonce matches the received one.
1155 		 * +2 in the length is required because of type and
1156 		 * length fields are included in a header.
1157 		 */
1158 		if (n != NULL &&
1159 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1160 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1161 		        ND_OPT_NONCE_LEN) == 0) {
1162 			dp->dad_ns_lcount++;
1163 			continue;
1164 		}
1165 		refcount_acquire(&dp->dad_refcnt);
1166 		break;
1167 	}
1168 	DADQ_RUNLOCK();
1169 
1170 	return (dp);
1171 }
1172 
1173 static void
1174 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1175 {
1176 
1177 	if (send_ns != 0)
1178 		nd6_dad_ns_output(dp);
1179 	callout_reset(&dp->dad_timer_ch, ticks,
1180 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1181 }
1182 
1183 static void
1184 nd6_dad_stoptimer(struct dadq *dp)
1185 {
1186 
1187 	callout_drain(&dp->dad_timer_ch);
1188 }
1189 
1190 static void
1191 nd6_dad_rele(struct dadq *dp)
1192 {
1193 
1194 	if (refcount_release(&dp->dad_refcnt)) {
1195 		ifa_free(dp->dad_ifa);
1196 		free(dp, M_IP6NDP);
1197 	}
1198 }
1199 
1200 void
1201 nd6_dad_init(void)
1202 {
1203 
1204 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1205 	TAILQ_INIT(&V_dadq);
1206 }
1207 
1208 /*
1209  * Start Duplicate Address Detection (DAD) for specified interface address.
1210  */
1211 void
1212 nd6_dad_start(struct ifaddr *ifa, int delay)
1213 {
1214 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1215 	struct dadq *dp;
1216 	char ip6buf[INET6_ADDRSTRLEN];
1217 
1218 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1219 	    ("starting DAD on non-tentative address %p", ifa));
1220 
1221 	/*
1222 	 * If we don't need DAD, don't do it.
1223 	 * There are several cases:
1224 	 * - DAD is disabled globally or on the interface
1225 	 * - the interface address is anycast
1226 	 */
1227 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1228 	    V_ip6_dad_count == 0 ||
1229 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1230 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1231 		return;
1232 	}
1233 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1234 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1235 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1236 		return;
1237 
1238 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1239 		/*
1240 		 * DAD is already in progress.  Let the existing entry
1241 		 * finish it.
1242 		 */
1243 		nd6_dad_rele(dp);
1244 		return;
1245 	}
1246 
1247 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1248 	if (dp == NULL) {
1249 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1250 			"%s(%s)\n",
1251 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1252 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1253 		return;
1254 	}
1255 	callout_init(&dp->dad_timer_ch, 0);
1256 #ifdef VIMAGE
1257 	dp->dad_vnet = curvnet;
1258 #endif
1259 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1260 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1261 
1262 	/*
1263 	 * Send NS packet for DAD, ip6_dad_count times.
1264 	 * Note that we must delay the first transmission, if this is the
1265 	 * first packet to be sent from the interface after interface
1266 	 * (re)initialization.
1267 	 */
1268 	dp->dad_ifa = ifa;
1269 	ifa_ref(dp->dad_ifa);
1270 	dp->dad_count = V_ip6_dad_count;
1271 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1272 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1273 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1274 	refcount_init(&dp->dad_refcnt, 1);
1275 	nd6_dad_add(dp);
1276 	nd6_dad_starttimer(dp, delay, 0);
1277 }
1278 
1279 /*
1280  * terminate DAD unconditionally.  used for address removals.
1281  */
1282 void
1283 nd6_dad_stop(struct ifaddr *ifa)
1284 {
1285 	struct dadq *dp;
1286 
1287 	dp = nd6_dad_find(ifa, NULL);
1288 	if (!dp) {
1289 		/* DAD wasn't started yet */
1290 		return;
1291 	}
1292 
1293 	nd6_dad_stoptimer(dp);
1294 
1295 	/*
1296 	 * The DAD queue entry may have been removed by nd6_dad_timer() while
1297 	 * we were waiting for it to stop, so re-do the lookup.
1298 	 */
1299 	nd6_dad_rele(dp);
1300 	if (nd6_dad_find(ifa, NULL) == NULL)
1301 		return;
1302 
1303 	nd6_dad_del(dp);
1304 	nd6_dad_rele(dp);
1305 }
1306 
1307 static void
1308 nd6_dad_timer(struct dadq *dp)
1309 {
1310 	CURVNET_SET(dp->dad_vnet);
1311 	struct ifaddr *ifa = dp->dad_ifa;
1312 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1313 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1314 	char ip6buf[INET6_ADDRSTRLEN];
1315 
1316 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1317 
1318 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1319 		/* Do not need DAD for ifdisabled interface. */
1320 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1321 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1322 		goto err;
1323 	}
1324 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1325 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1326 			"%s(%s)\n",
1327 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1328 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1329 		goto err;
1330 	}
1331 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1332 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1333 			"%s(%s)\n",
1334 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1335 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1336 		goto err;
1337 	}
1338 
1339 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1340 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1341 	    (((ifp->if_flags & IFF_UP) == 0) ||
1342 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1343 		nd6log((LOG_INFO, "%s: could not run DAD "
1344 		    "because the interface was down or not running.\n",
1345 		    if_name(ifa->ifa_ifp)));
1346 		goto err;
1347 	}
1348 
1349 	/* Need more checks? */
1350 	if (dp->dad_ns_ocount < dp->dad_count) {
1351 		/*
1352 		 * We have more NS to go.  Send NS packet for DAD.
1353 		 */
1354 		nd6_dad_starttimer(dp,
1355 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1356 		goto done;
1357 	} else {
1358 		/*
1359 		 * We have transmitted sufficient number of DAD packets.
1360 		 * See what we've got.
1361 		 */
1362 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1363 			/* We've seen NS or NA, means DAD has failed. */
1364 			nd6_dad_duplicated(ifa, dp);
1365 		else if (V_dad_enhanced != 0 &&
1366 		    dp->dad_ns_lcount > 0 &&
1367 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1368 			/*
1369 			 * Sec. 4.1 in RFC 7527 requires transmission of
1370 			 * additional probes until the loopback condition
1371 			 * becomes clear when a looped back probe is detected.
1372 			 */
1373 			log(LOG_ERR, "%s: a looped back NS message is "
1374 			    "detected during DAD for %s.  "
1375 			    "Another DAD probes are being sent.\n",
1376 			    if_name(ifa->ifa_ifp),
1377 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1378 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1379 			/*
1380 			 * Send an NS immediately and increase dad_count by
1381 			 * V_nd6_mmaxtries - 1.
1382 			 */
1383 			dp->dad_count =
1384 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1385 			nd6_dad_starttimer(dp,
1386 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1387 			    1);
1388 			goto done;
1389 		} else {
1390 			/*
1391 			 * We are done with DAD.  No NA came, no NS came.
1392 			 * No duplicate address found.  Check IFDISABLED flag
1393 			 * again in case that it is changed between the
1394 			 * beginning of this function and here.
1395 			 */
1396 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1397 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1398 
1399 			nd6log((LOG_DEBUG,
1400 			    "%s: DAD complete for %s - no duplicates found\n",
1401 			    if_name(ifa->ifa_ifp),
1402 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1403 			if (dp->dad_ns_lcount > 0)
1404 				log(LOG_ERR, "%s: DAD completed while "
1405 				    "a looped back NS message is detected "
1406 				    "during DAD for %s.\n",
1407 				    if_name(ifa->ifa_ifp),
1408 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1409 		}
1410 	}
1411 err:
1412 	nd6_dad_del(dp);
1413 done:
1414 	CURVNET_RESTORE();
1415 }
1416 
1417 static void
1418 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1419 {
1420 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1421 	struct ifnet *ifp;
1422 	char ip6buf[INET6_ADDRSTRLEN];
1423 
1424 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1425 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1426 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1427 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1428 	    dp->dad_na_icount);
1429 
1430 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1431 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1432 
1433 	ifp = ifa->ifa_ifp;
1434 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1435 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1436 	log(LOG_ERR, "%s: manual intervention required\n",
1437 	    if_name(ifp));
1438 
1439 	/*
1440 	 * If the address is a link-local address formed from an interface
1441 	 * identifier based on the hardware address which is supposed to be
1442 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1443 	 * operation on the interface SHOULD be disabled.
1444 	 * [RFC 4862, Section 5.4.5]
1445 	 */
1446 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1447 		struct in6_addr in6;
1448 
1449 		/*
1450 		 * To avoid over-reaction, we only apply this logic when we are
1451 		 * very sure that hardware addresses are supposed to be unique.
1452 		 */
1453 		switch (ifp->if_type) {
1454 		case IFT_ETHER:
1455 		case IFT_FDDI:
1456 		case IFT_ATM:
1457 		case IFT_IEEE1394:
1458 		case IFT_INFINIBAND:
1459 			in6 = ia->ia_addr.sin6_addr;
1460 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1461 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1462 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1463 				log(LOG_ERR, "%s: possible hardware address "
1464 				    "duplication detected, disable IPv6\n",
1465 				    if_name(ifp));
1466 			}
1467 			break;
1468 		}
1469 	}
1470 }
1471 
1472 static void
1473 nd6_dad_ns_output(struct dadq *dp)
1474 {
1475 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1476 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1477 	int i;
1478 
1479 	dp->dad_ns_tcount++;
1480 	if ((ifp->if_flags & IFF_UP) == 0) {
1481 		return;
1482 	}
1483 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1484 		return;
1485 	}
1486 
1487 	dp->dad_ns_ocount++;
1488 	if (V_dad_enhanced != 0) {
1489 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1490 			dp->dad_nonce[i] = arc4random();
1491 		/*
1492 		 * XXXHRS: Note that in the case that
1493 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1494 		 * different nonces can be looped back in an unexpected
1495 		 * order.  The current implementation recognizes only
1496 		 * the latest nonce on the sender side.  Practically it
1497 		 * should work well in almost all cases.
1498 		 */
1499 	}
1500 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1501 	    (uint8_t *)&dp->dad_nonce[0]);
1502 }
1503 
1504 static void
1505 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1506 {
1507 	struct in6_ifaddr *ia;
1508 	struct ifnet *ifp;
1509 	const struct in6_addr *taddr6;
1510 	struct dadq *dp;
1511 
1512 	if (ifa == NULL)
1513 		panic("ifa == NULL in nd6_dad_ns_input");
1514 
1515 	ia = (struct in6_ifaddr *)ifa;
1516 	ifp = ifa->ifa_ifp;
1517 	taddr6 = &ia->ia_addr.sin6_addr;
1518 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1519 	if (V_dad_enhanced == 0)
1520 		ndopt_nonce = NULL;
1521 	dp = nd6_dad_find(ifa, ndopt_nonce);
1522 	if (dp == NULL)
1523 		return;
1524 
1525 	dp->dad_ns_icount++;
1526 	nd6_dad_rele(dp);
1527 }
1528 
1529 static void
1530 nd6_dad_na_input(struct ifaddr *ifa)
1531 {
1532 	struct dadq *dp;
1533 
1534 	if (ifa == NULL)
1535 		panic("ifa == NULL in nd6_dad_na_input");
1536 
1537 	dp = nd6_dad_find(ifa, NULL);
1538 	if (dp != NULL) {
1539 		dp->dad_na_icount++;
1540 		nd6_dad_rele(dp);
1541 	}
1542 }
1543