xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 13ec1e3155c7e9bf037b12af186351b7fa9b9450)
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 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/eventhandler.h>
44 #include <sys/malloc.h>
45 #include <sys/libkern.h>
46 #include <sys/lock.h>
47 #include <sys/rwlock.h>
48 #include <sys/mbuf.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/errno.h>
54 #include <sys/sysctl.h>
55 #include <sys/syslog.h>
56 #include <sys/queue.h>
57 #include <sys/callout.h>
58 #include <sys/refcount.h>
59 
60 #include <net/if.h>
61 #include <net/if_types.h>
62 #include <net/if_dl.h>
63 #include <net/if_var.h>
64 #include <net/route.h>
65 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <net/if_llatbl.h>
70 #include <netinet6/in6_var.h>
71 #include <netinet6/in6_ifattach.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #include <netinet6/nd6.h>
76 #include <netinet/icmp6.h>
77 #include <netinet/ip_carp.h>
78 #include <netinet6/send.h>
79 
80 #define SDL(s) ((struct sockaddr_dl *)s)
81 
82 struct dadq;
83 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
84 static void nd6_dad_add(struct dadq *dp);
85 static void nd6_dad_del(struct dadq *dp);
86 static void nd6_dad_rele(struct dadq *);
87 static void nd6_dad_starttimer(struct dadq *, int);
88 static void nd6_dad_stoptimer(struct dadq *);
89 static void nd6_dad_timer(void *);
90 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
91 static void nd6_dad_ns_output(struct dadq *);
92 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
93 static void nd6_dad_na_input(struct ifaddr *);
94 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
95     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
96 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
97     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
98 
99 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
100 #define	V_dad_enhanced			VNET(dad_enhanced)
101 
102 SYSCTL_DECL(_net_inet6_ip6);
103 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
104     &VNET_NAME(dad_enhanced), 0,
105     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
106 
107 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
108 						   transmit DAD packet */
109 #define	V_dad_maxtry			VNET(dad_maxtry)
110 
111 /*
112  * Input a Neighbor Solicitation Message.
113  *
114  * Based on RFC 2461
115  * Based on RFC 2462 (duplicate address detection)
116  */
117 void
118 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
119 {
120 	struct ifnet *ifp;
121 	struct ip6_hdr *ip6;
122 	struct nd_neighbor_solicit *nd_ns;
123 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
124 	struct ifaddr *ifa;
125 	struct sockaddr_dl proxydl;
126 	union nd_opts ndopts;
127 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
128 	char *lladdr;
129 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
130 
131 	ifa = NULL;
132 
133 	/* RFC 6980: Nodes MUST silently ignore fragments */
134 	if(m->m_flags & M_FRAGMENTED)
135 		goto freeit;
136 
137 	ifp = m->m_pkthdr.rcvif;
138 	ip6 = mtod(m, struct ip6_hdr *);
139 	if (__predict_false(ip6->ip6_hlim != 255)) {
140 		ICMP6STAT_INC(icp6s_invlhlim);
141 		nd6log((LOG_ERR,
142 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
143 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
144 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
145 		goto bads;
146 	}
147 
148 	if (m->m_len < off + icmp6len) {
149 		m = m_pullup(m, off + icmp6len);
150 		if (m == NULL) {
151 			IP6STAT_INC(ip6s_exthdrtoolong);
152 			return;
153 		}
154 	}
155 	ip6 = mtod(m, struct ip6_hdr *);
156 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
157 
158 	saddr6 = ip6->ip6_src;
159 	daddr6 = ip6->ip6_dst;
160 	taddr6 = nd_ns->nd_ns_target;
161 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
162 		goto bad;
163 
164 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
165 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
166 		rflag = 0;
167 
168 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
169 		/* dst has to be a solicited node multicast address. */
170 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
171 		    /* don't check ifindex portion */
172 		    daddr6.s6_addr32[1] == 0 &&
173 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
174 		    daddr6.s6_addr8[12] == 0xff) {
175 			; /* good */
176 		} else {
177 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
178 			    "(wrong ip6 dst)\n"));
179 			goto bad;
180 		}
181 	} else if (!V_nd6_onlink_ns_rfc4861) {
182 		struct sockaddr_in6 src_sa6;
183 
184 		/*
185 		 * According to recent IETF discussions, it is not a good idea
186 		 * to accept a NS from an address which would not be deemed
187 		 * to be a neighbor otherwise.  This point is expected to be
188 		 * clarified in future revisions of the specification.
189 		 */
190 		bzero(&src_sa6, sizeof(src_sa6));
191 		src_sa6.sin6_family = AF_INET6;
192 		src_sa6.sin6_len = sizeof(src_sa6);
193 		src_sa6.sin6_addr = saddr6;
194 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
195 			nd6log((LOG_INFO, "nd6_ns_input: "
196 				"NS packet from non-neighbor\n"));
197 			goto bad;
198 		}
199 	}
200 
201 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
202 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
203 		goto bad;
204 	}
205 
206 	icmp6len -= sizeof(*nd_ns);
207 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
208 	if (nd6_options(&ndopts) < 0) {
209 		nd6log((LOG_INFO,
210 		    "nd6_ns_input: invalid ND option, ignored\n"));
211 		/* nd6_options have incremented stats */
212 		goto freeit;
213 	}
214 
215 	lladdr = NULL;
216 	lladdrlen = 0;
217 	if (ndopts.nd_opts_src_lladdr) {
218 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
219 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
220 	}
221 
222 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
223 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
224 		    "(link-layer address option)\n"));
225 		goto bad;
226 	}
227 
228 	/*
229 	 * Attaching target link-layer address to the NA?
230 	 * (RFC 2461 7.2.4)
231 	 *
232 	 * NS IP dst is unicast/anycast			MUST NOT add
233 	 * NS IP dst is solicited-node multicast	MUST add
234 	 *
235 	 * In implementation, we add target link-layer address by default.
236 	 * We do not add one in MUST NOT cases.
237 	 */
238 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
239 		tlladdr = 0;
240 	else
241 		tlladdr = 1;
242 
243 	/*
244 	 * Target address (taddr6) must be either:
245 	 * (1) Valid unicast/anycast address for my receiving interface,
246 	 * (2) Unicast address for which I'm offering proxy service, or
247 	 * (3) "tentative" address on which DAD is being performed.
248 	 */
249 	/* (1) and (3) check. */
250 	if (ifp->if_carp)
251 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
252 	else
253 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
254 
255 	/* (2) check. */
256 	proxy = 0;
257 	if (ifa == NULL) {
258 		struct sockaddr_dl rt_gateway;
259 		struct rt_addrinfo info;
260 		struct sockaddr_in6 dst6;
261 
262 		bzero(&dst6, sizeof(dst6));
263 		dst6.sin6_len = sizeof(struct sockaddr_in6);
264 		dst6.sin6_family = AF_INET6;
265 		dst6.sin6_addr = taddr6;
266 
267 		bzero(&rt_gateway, sizeof(rt_gateway));
268 		rt_gateway.sdl_len = sizeof(rt_gateway);
269 		bzero(&info, sizeof(info));
270 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
271 
272 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
273 		    0, 0, &info) == 0) {
274 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
275 			    rt_gateway.sdl_family == AF_LINK) {
276 				/*
277 				 * proxy NDP for single entry
278 				 */
279 				proxydl = *SDL(&rt_gateway);
280 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
281 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
282 				if (ifa)
283 					proxy = 1;
284 			}
285 		}
286 	}
287 	if (ifa == NULL) {
288 		/*
289 		 * We've got an NS packet, and we don't have that adddress
290 		 * assigned for us.  We MUST silently ignore it.
291 		 * See RFC2461 7.2.3.
292 		 */
293 		goto freeit;
294 	}
295 	myaddr6 = *IFA_IN6(ifa);
296 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
297 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
298 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
299 		goto freeit;
300 
301 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
302 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
303 		    "(if %d, NS packet %d)\n",
304 		    ip6_sprintf(ip6bufs, &taddr6),
305 		    ifp->if_addrlen, lladdrlen - 2));
306 		goto bad;
307 	}
308 
309 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
310 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
311 		    ip6_sprintf(ip6bufs, &saddr6)));
312 		goto freeit;
313 	}
314 
315 	/*
316 	 * We have neighbor solicitation packet, with target address equals to
317 	 * one of my tentative address.
318 	 *
319 	 * src addr	how to process?
320 	 * ---		---
321 	 * multicast	of course, invalid (rejected in ip6_input)
322 	 * unicast	somebody is doing address resolution -> ignore
323 	 * unspec	dup address detection
324 	 *
325 	 * The processing is defined in RFC 2462.
326 	 */
327 	if (tentative) {
328 		/*
329 		 * If source address is unspecified address, it is for
330 		 * duplicate address detection.
331 		 *
332 		 * If not, the packet is for addess resolution;
333 		 * silently ignore it.
334 		 */
335 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
336 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
337 
338 		goto freeit;
339 	}
340 
341 	/*
342 	 * If the source address is unspecified address, entries must not
343 	 * be created or updated.
344 	 * It looks that sender is performing DAD.  Output NA toward
345 	 * all-node multicast address, to tell the sender that I'm using
346 	 * the address.
347 	 * S bit ("solicited") must be zero.
348 	 */
349 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
350 		struct in6_addr in6_all;
351 
352 		in6_all = in6addr_linklocal_allnodes;
353 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
354 			goto bad;
355 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
356 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
357 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
358 		    M_GETFIB(m));
359 		goto freeit;
360 	}
361 
362 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
363 	    ND_NEIGHBOR_SOLICIT, 0);
364 
365 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
366 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
367 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
368 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
369  freeit:
370 	if (ifa != NULL)
371 		ifa_free(ifa);
372 	m_freem(m);
373 	return;
374 
375  bad:
376 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
377 		ip6_sprintf(ip6bufs, &saddr6)));
378 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
379 		ip6_sprintf(ip6bufs, &daddr6)));
380 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
381 		ip6_sprintf(ip6bufs, &taddr6)));
382  bads:
383 	ICMP6STAT_INC(icp6s_badns);
384 	if (ifa != NULL)
385 		ifa_free(ifa);
386 	m_freem(m);
387 }
388 
389 /*
390  * Output a Neighbor Solicitation Message. Caller specifies:
391  *	- ICMP6 header source IP6 address
392  *	- ND6 header target IP6 address
393  *	- ND6 header source datalink address
394  *
395  * Based on RFC 2461
396  * Based on RFC 2462 (duplicate address detection)
397  *
398  *    ln - for source address determination
399  * nonce - If non-NULL, NS is used for duplicate address detection and
400  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
401  */
402 static void
403 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
404     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
405     uint8_t *nonce, u_int fibnum)
406 {
407 	struct mbuf *m;
408 	struct m_tag *mtag;
409 	struct ip6_hdr *ip6;
410 	struct nd_neighbor_solicit *nd_ns;
411 	struct ip6_moptions im6o;
412 	int icmp6len;
413 	int maxlen;
414 
415 	NET_EPOCH_ASSERT();
416 
417 	if (IN6_IS_ADDR_MULTICAST(taddr6))
418 		return;
419 
420 	/* estimate the size of message */
421 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
422 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
423 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
424 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
425 	    __func__, max_linkhdr, maxlen, MCLBYTES));
426 
427 	if (max_linkhdr + maxlen > MHLEN)
428 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
429 	else
430 		m = m_gethdr(M_NOWAIT, MT_DATA);
431 	if (m == NULL)
432 		return;
433 	M_SETFIB(m, fibnum);
434 
435 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
436 		m->m_flags |= M_MCAST;
437 		im6o.im6o_multicast_ifp = ifp;
438 		im6o.im6o_multicast_hlim = 255;
439 		im6o.im6o_multicast_loop = 0;
440 	}
441 
442 	icmp6len = sizeof(*nd_ns);
443 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
444 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
445 
446 	/* fill neighbor solicitation packet */
447 	ip6 = mtod(m, struct ip6_hdr *);
448 	ip6->ip6_flow = 0;
449 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
450 	ip6->ip6_vfc |= IPV6_VERSION;
451 	/* ip6->ip6_plen will be set later */
452 	ip6->ip6_nxt = IPPROTO_ICMPV6;
453 	ip6->ip6_hlim = 255;
454 	if (daddr6)
455 		ip6->ip6_dst = *daddr6;
456 	else {
457 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
458 		ip6->ip6_dst.s6_addr16[1] = 0;
459 		ip6->ip6_dst.s6_addr32[1] = 0;
460 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
461 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
462 		ip6->ip6_dst.s6_addr8[12] = 0xff;
463 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
464 			goto bad;
465 	}
466 	if (nonce == NULL) {
467 		struct ifaddr *ifa = NULL;
468 
469 		/*
470 		 * RFC2461 7.2.2:
471 		 * "If the source address of the packet prompting the
472 		 * solicitation is the same as one of the addresses assigned
473 		 * to the outgoing interface, that address SHOULD be placed
474 		 * in the IP Source Address of the outgoing solicitation.
475 		 * Otherwise, any one of the addresses assigned to the
476 		 * interface should be used."
477 		 *
478 		 * We use the source address for the prompting packet
479 		 * (saddr6), if saddr6 belongs to the outgoing interface.
480 		 * Otherwise, we perform the source address selection as usual.
481 		 */
482 
483 		if (saddr6 != NULL)
484 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
485 		if (ifa != NULL) {
486 			/* ip6_src set already. */
487 			ip6->ip6_src = *saddr6;
488 			ifa_free(ifa);
489 		} else {
490 			int error;
491 			struct in6_addr dst6, src6;
492 			uint32_t scopeid;
493 
494 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
495 			error = in6_selectsrc_addr(fibnum, &dst6,
496 			    scopeid, ifp, &src6, NULL);
497 			if (error) {
498 				char ip6buf[INET6_ADDRSTRLEN];
499 				nd6log((LOG_DEBUG, "%s: source can't be "
500 				    "determined: dst=%s, error=%d\n", __func__,
501 				    ip6_sprintf(ip6buf, &dst6),
502 				    error));
503 				goto bad;
504 			}
505 			ip6->ip6_src = src6;
506 		}
507 	} else {
508 		/*
509 		 * Source address for DAD packet must always be IPv6
510 		 * unspecified address. (0::0)
511 		 * We actually don't have to 0-clear the address (we did it
512 		 * above), but we do so here explicitly to make the intention
513 		 * clearer.
514 		 */
515 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
516 	}
517 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
518 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
519 	nd_ns->nd_ns_code = 0;
520 	nd_ns->nd_ns_reserved = 0;
521 	nd_ns->nd_ns_target = *taddr6;
522 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
523 
524 	/*
525 	 * Add source link-layer address option.
526 	 *
527 	 *				spec		implementation
528 	 *				---		---
529 	 * DAD packet			MUST NOT	do not add the option
530 	 * there's no link layer address:
531 	 *				impossible	do not add the option
532 	 * there's link layer address:
533 	 *	Multicast NS		MUST add one	add the option
534 	 *	Unicast NS		SHOULD add one	add the option
535 	 */
536 	if (nonce == NULL) {
537 		struct nd_opt_hdr *nd_opt;
538 		char *mac;
539 		int optlen;
540 
541 		mac = NULL;
542 		if (ifp->if_carp)
543 			mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src);
544 		if (mac == NULL)
545 			mac = nd6_ifptomac(ifp);
546 
547 		if (mac != NULL) {
548 			nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
549 			optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
550 			/* 8 byte alignments... */
551 			optlen = (optlen + 7) & ~7;
552 			m->m_pkthdr.len += optlen;
553 			m->m_len += optlen;
554 			icmp6len += optlen;
555 			bzero(nd_opt, optlen);
556 			nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
557 			nd_opt->nd_opt_len = optlen >> 3;
558 			bcopy(mac, nd_opt + 1, ifp->if_addrlen);
559 		}
560 	}
561 	/*
562 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
563 	 * This behavior is documented as Enhanced Duplicate Address
564 	 * Detection in RFC 7527.
565 	 * net.inet6.ip6.dad_enhanced=0 disables this.
566 	 */
567 	if (V_dad_enhanced != 0 && nonce != NULL) {
568 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
569 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
570 		/* 8-byte alignment is required. */
571 		optlen = (optlen + 7) & ~7;
572 
573 		m->m_pkthdr.len += optlen;
574 		m->m_len += optlen;
575 		icmp6len += optlen;
576 		bzero((caddr_t)nd_opt, optlen);
577 		nd_opt->nd_opt_type = ND_OPT_NONCE;
578 		nd_opt->nd_opt_len = optlen >> 3;
579 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
580 	}
581 	ip6->ip6_plen = htons((u_short)icmp6len);
582 	nd_ns->nd_ns_cksum = 0;
583 	nd_ns->nd_ns_cksum =
584 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
585 
586 	if (send_sendso_input_hook != NULL) {
587 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
588 			sizeof(unsigned short), M_NOWAIT);
589 		if (mtag == NULL)
590 			goto bad;
591 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
592 		m_tag_prepend(m, mtag);
593 	}
594 
595 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
596 	    &im6o, NULL, NULL);
597 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
598 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
599 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
600 
601 	return;
602 
603   bad:
604 	m_freem(m);
605 }
606 
607 #ifndef BURN_BRIDGES
608 void
609 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
610     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
611 {
612 
613 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
614 }
615 #endif
616 /*
617  * Neighbor advertisement input handling.
618  *
619  * Based on RFC 2461
620  * Based on RFC 2462 (duplicate address detection)
621  *
622  * the following items are not implemented yet:
623  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
624  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
625  */
626 void
627 nd6_na_input(struct mbuf *m, int off, int icmp6len)
628 {
629 	struct ifnet *ifp;
630 	struct ip6_hdr *ip6;
631 	struct ifaddr *ifa;
632 	struct llentry *ln;
633 	struct mbuf *chain;
634 	struct nd_neighbor_advert *nd_na;
635 	struct in6_addr daddr6, taddr6;
636 	union nd_opts ndopts;
637 	u_char linkhdr[LLE_MAX_LINKHDR];
638 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
639 	char *lladdr;
640 	size_t linkhdrsize;
641 	int flags, is_override, is_router, is_solicited;
642 	int lladdr_off, lladdrlen, checklink;
643 	bool flush_holdchain = false;
644 
645 	NET_EPOCH_ASSERT();
646 
647 	chain = NULL;
648 	ln = NULL;
649 	checklink = 0;
650 
651 	/* RFC 6980: Nodes MUST silently ignore fragments */
652 	if(m->m_flags & M_FRAGMENTED)
653 		goto freeit;
654 
655 	ifp = m->m_pkthdr.rcvif;
656 	ip6 = mtod(m, struct ip6_hdr *);
657 	if (__predict_false(ip6->ip6_hlim != 255)) {
658 		ICMP6STAT_INC(icp6s_invlhlim);
659 		nd6log((LOG_ERR,
660 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
661 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
662 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
663 		goto bad;
664 	}
665 
666 	if (m->m_len < off + icmp6len) {
667 		m = m_pullup(m, off + icmp6len);
668 		if (m == NULL) {
669 			IP6STAT_INC(ip6s_exthdrtoolong);
670 			return;
671 		}
672 	}
673 	ip6 = mtod(m, struct ip6_hdr *);
674 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
675 
676 	flags = nd_na->nd_na_flags_reserved;
677 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
678 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
679 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
680 
681 	taddr6 = nd_na->nd_na_target;
682 	if (in6_setscope(&taddr6, ifp, NULL))
683 		goto bad;	/* XXX: impossible */
684 
685 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
686 		nd6log((LOG_ERR,
687 		    "nd6_na_input: invalid target address %s\n",
688 		    ip6_sprintf(ip6bufs, &taddr6)));
689 		goto bad;
690 	}
691 
692 	daddr6 = ip6->ip6_dst;
693 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
694 		if (is_solicited) {
695 			nd6log((LOG_ERR,
696 			    "nd6_na_input: a solicited adv is multicasted\n"));
697 			goto bad;
698 		}
699 
700 	icmp6len -= sizeof(*nd_na);
701 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
702 	if (nd6_options(&ndopts) < 0) {
703 		nd6log((LOG_INFO,
704 		    "nd6_na_input: invalid ND option, ignored\n"));
705 		/* nd6_options have incremented stats */
706 		goto freeit;
707 	}
708 
709 	lladdr = NULL;
710 	lladdrlen = 0;
711 	if (ndopts.nd_opts_tgt_lladdr) {
712 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
713 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
714 	}
715 
716 	/*
717 	 * This effectively disables the DAD check on a non-master CARP
718 	 * address.
719 	 */
720 	if (ifp->if_carp)
721 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
722 	else
723 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
724 
725 	/*
726 	 * Target address matches one of my interface address.
727 	 *
728 	 * If my address is tentative, this means that there's somebody
729 	 * already using the same address as mine.  This indicates DAD failure.
730 	 * This is defined in RFC 2462.
731 	 *
732 	 * Otherwise, process as defined in RFC 2461.
733 	 */
734 	if (ifa
735 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
736 		nd6_dad_na_input(ifa);
737 		ifa_free(ifa);
738 		goto freeit;
739 	}
740 
741 	/* Just for safety, maybe unnecessary. */
742 	if (ifa) {
743 		ifa_free(ifa);
744 		log(LOG_ERR,
745 		    "nd6_na_input: duplicate IP6 address %s\n",
746 		    ip6_sprintf(ip6bufs, &taddr6));
747 		goto freeit;
748 	}
749 
750 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
751 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
752 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
753 		    ifp->if_addrlen, lladdrlen - 2));
754 		goto bad;
755 	}
756 
757 	/*
758 	 * If no neighbor cache entry is found, NA SHOULD silently be
759 	 * discarded.
760 	 */
761 	ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
762 	if (ln == NULL) {
763 		goto freeit;
764 	}
765 
766 	/*
767 	 * Do not try to override static entry.
768 	 */
769 	if (ln->la_flags & LLE_STATIC)
770 		goto freeit;
771 
772 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
773 		/*
774 		 * If the link-layer has address, and no lladdr option came,
775 		 * discard the packet.
776 		 */
777 		if (ifp->if_addrlen && lladdr == NULL) {
778 			goto freeit;
779 		}
780 
781 		/*
782 		 * Record link-layer address, and update the state.
783 		 */
784 		if (!nd6_try_set_entry_addr(ifp, ln, lladdr))
785 			goto freeit;
786 
787 		flush_holdchain = true;
788 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
789 		if (is_solicited)
790 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
791 		else
792 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
793 		if ((ln->ln_router = is_router) != 0) {
794 			/*
795 			 * This means a router's state has changed from
796 			 * non-reachable to probably reachable, and might
797 			 * affect the status of associated prefixes..
798 			 */
799 			checklink = 1;
800 		}
801 	} else {
802 		int llchange;
803 
804 		/*
805 		 * Check if the link-layer address has changed or not.
806 		 */
807 		if (lladdr == NULL)
808 			llchange = 0;
809 		else {
810 			if (ln->la_flags & LLE_VALID) {
811 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
812 					llchange = 1;
813 				else
814 					llchange = 0;
815 			} else
816 				llchange = 1;
817 		}
818 
819 		/*
820 		 * This is VERY complex.  Look at it with care.
821 		 *
822 		 * override solicit lladdr llchange	action
823 		 *					(L: record lladdr)
824 		 *
825 		 *	0	0	n	--	(2c)
826 		 *	0	0	y	n	(2b) L
827 		 *	0	0	y	y	(1)    REACHABLE->STALE
828 		 *	0	1	n	--	(2c)   *->REACHABLE
829 		 *	0	1	y	n	(2b) L *->REACHABLE
830 		 *	0	1	y	y	(1)    REACHABLE->STALE
831 		 *	1	0	n	--	(2a)
832 		 *	1	0	y	n	(2a) L
833 		 *	1	0	y	y	(2a) L *->STALE
834 		 *	1	1	n	--	(2a)   *->REACHABLE
835 		 *	1	1	y	n	(2a) L *->REACHABLE
836 		 *	1	1	y	y	(2a) L *->REACHABLE
837 		 */
838 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
839 			/*
840 			 * If state is REACHABLE, make it STALE.
841 			 * no other updates should be done.
842 			 */
843 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
844 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
845 			goto freeit;
846 		} else if (is_override				   /* (2a) */
847 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
848 			|| lladdr == NULL) {			   /* (2c) */
849 			/*
850 			 * Update link-local address, if any.
851 			 */
852 			if (lladdr != NULL) {
853 				linkhdrsize = sizeof(linkhdr);
854 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
855 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
856 					goto freeit;
857 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
858 				    linkhdrsize, lladdr_off) == 0) {
859 					ln = NULL;
860 					goto freeit;
861 				}
862 				EVENTHANDLER_INVOKE(lle_event, ln,
863 				    LLENTRY_RESOLVED);
864 			}
865 
866 			/*
867 			 * If solicited, make the state REACHABLE.
868 			 * If not solicited and the link-layer address was
869 			 * changed, make it STALE.
870 			 */
871 			if (is_solicited)
872 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
873 			else {
874 				if (lladdr != NULL && llchange)
875 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
876 			}
877 		}
878 
879 		if (ln->ln_router && !is_router) {
880 			/*
881 			 * The peer dropped the router flag.
882 			 * Remove the sender from the Default Router List and
883 			 * update the Destination Cache entries.
884 			 */
885 			struct ifnet *nd6_ifp;
886 
887 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
888 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
889 			    (ND_IFINFO(nd6_ifp)->flags &
890 			     ND6_IFF_ACCEPT_RTADV) != 0)
891 				/*
892 				 * Even if the neighbor is not in the default
893 				 * router list, the neighbor may be used as a
894 				 * next hop for some destinations (e.g. redirect
895 				 * case). So we must call rt6_flush explicitly.
896 				 */
897 				rt6_flush(&ip6->ip6_src, ifp);
898 		}
899 		ln->ln_router = is_router;
900 	}
901         /* XXX - QL
902 	 *  Does this matter?
903 	 *  rt->rt_flags &= ~RTF_REJECT;
904 	 */
905 	ln->la_asked = 0;
906 	if (ln->la_hold != NULL)
907 		chain = nd6_grab_holdchain(ln);
908  freeit:
909 	if (ln != NULL)
910 		LLE_WUNLOCK(ln);
911 
912 	if (chain != NULL)
913 		nd6_flush_holdchain(ifp, ln, chain);
914 	if (flush_holdchain)
915 		nd6_flush_children_holdchain(ifp, ln);
916 
917 	if (checklink)
918 		pfxlist_onlink_check();
919 
920 	m_freem(m);
921 	return;
922 
923  bad:
924 	if (ln != NULL)
925 		LLE_WUNLOCK(ln);
926 
927 	ICMP6STAT_INC(icp6s_badna);
928 	m_freem(m);
929 }
930 
931 /*
932  * Neighbor advertisement output handling.
933  *
934  * Based on RFC 2461
935  *
936  * the following items are not implemented yet:
937  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
938  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
939  *
940  * tlladdr - 1 if include target link-layer address
941  * sdl0 - sockaddr_dl (= proxy NA) or NULL
942  */
943 static void
944 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
945     const struct in6_addr *taddr6, u_long flags, int tlladdr,
946     struct sockaddr *sdl0, u_int fibnum)
947 {
948 	struct mbuf *m;
949 	struct m_tag *mtag;
950 	struct ip6_hdr *ip6;
951 	struct nd_neighbor_advert *nd_na;
952 	struct ip6_moptions im6o;
953 	struct in6_addr daddr6, dst6, src6;
954 	uint32_t scopeid;
955 
956 	NET_EPOCH_ASSERT();
957 
958 	int icmp6len, maxlen, error;
959 	caddr_t mac = NULL;
960 
961 	daddr6 = *daddr6_0;	/* make a local copy for modification */
962 
963 	/* estimate the size of message */
964 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
965 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
966 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
967 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
968 	    __func__, max_linkhdr, maxlen, MCLBYTES));
969 
970 	if (max_linkhdr + maxlen > MHLEN)
971 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
972 	else
973 		m = m_gethdr(M_NOWAIT, MT_DATA);
974 	if (m == NULL)
975 		return;
976 	M_SETFIB(m, fibnum);
977 
978 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
979 		m->m_flags |= M_MCAST;
980 		im6o.im6o_multicast_ifp = ifp;
981 		im6o.im6o_multicast_hlim = 255;
982 		im6o.im6o_multicast_loop = 0;
983 	}
984 
985 	icmp6len = sizeof(*nd_na);
986 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
987 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
988 
989 	/* fill neighbor advertisement packet */
990 	ip6 = mtod(m, struct ip6_hdr *);
991 	ip6->ip6_flow = 0;
992 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
993 	ip6->ip6_vfc |= IPV6_VERSION;
994 	ip6->ip6_nxt = IPPROTO_ICMPV6;
995 	ip6->ip6_hlim = 255;
996 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
997 		/* reply to DAD */
998 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
999 		daddr6.s6_addr16[1] = 0;
1000 		daddr6.s6_addr32[1] = 0;
1001 		daddr6.s6_addr32[2] = 0;
1002 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1003 		if (in6_setscope(&daddr6, ifp, NULL))
1004 			goto bad;
1005 
1006 		flags &= ~ND_NA_FLAG_SOLICITED;
1007 	}
1008 	ip6->ip6_dst = daddr6;
1009 
1010 	/*
1011 	 * Select a source whose scope is the same as that of the dest.
1012 	 */
1013 	in6_splitscope(&daddr6, &dst6, &scopeid);
1014 	error = in6_selectsrc_addr(fibnum, &dst6,
1015 	    scopeid, ifp, &src6, NULL);
1016 	if (error) {
1017 		char ip6buf[INET6_ADDRSTRLEN];
1018 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1019 		    "determined: dst=%s, error=%d\n",
1020 		    ip6_sprintf(ip6buf, &daddr6), error));
1021 		goto bad;
1022 	}
1023 	ip6->ip6_src = src6;
1024 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1025 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1026 	nd_na->nd_na_code = 0;
1027 	nd_na->nd_na_target = *taddr6;
1028 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1029 
1030 	/*
1031 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1032 	 * see nd6_ns_input() for details.
1033 	 * Basically, if NS packet is sent to unicast/anycast addr,
1034 	 * target lladdr option SHOULD NOT be included.
1035 	 */
1036 	if (tlladdr) {
1037 		/*
1038 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1039 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1040 		 * my address) use lladdr configured for the interface.
1041 		 */
1042 		if (sdl0 == NULL) {
1043 			if (ifp->if_carp)
1044 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1045 			if (mac == NULL)
1046 				mac = nd6_ifptomac(ifp);
1047 		} else if (sdl0->sa_family == AF_LINK) {
1048 			struct sockaddr_dl *sdl;
1049 			sdl = (struct sockaddr_dl *)sdl0;
1050 			if (sdl->sdl_alen == ifp->if_addrlen)
1051 				mac = LLADDR(sdl);
1052 		}
1053 	}
1054 	if (tlladdr && mac) {
1055 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1056 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1057 
1058 		/* roundup to 8 bytes alignment! */
1059 		optlen = (optlen + 7) & ~7;
1060 
1061 		m->m_pkthdr.len += optlen;
1062 		m->m_len += optlen;
1063 		icmp6len += optlen;
1064 		bzero((caddr_t)nd_opt, optlen);
1065 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1066 		nd_opt->nd_opt_len = optlen >> 3;
1067 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1068 	} else
1069 		flags &= ~ND_NA_FLAG_OVERRIDE;
1070 
1071 	ip6->ip6_plen = htons((u_short)icmp6len);
1072 	nd_na->nd_na_flags_reserved = flags;
1073 	nd_na->nd_na_cksum = 0;
1074 	nd_na->nd_na_cksum =
1075 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1076 
1077 	if (send_sendso_input_hook != NULL) {
1078 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1079 		    sizeof(unsigned short), M_NOWAIT);
1080 		if (mtag == NULL)
1081 			goto bad;
1082 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1083 		m_tag_prepend(m, mtag);
1084 	}
1085 
1086 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1087 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1088 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1089 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1090 
1091 	return;
1092 
1093   bad:
1094 	m_freem(m);
1095 }
1096 
1097 #ifndef BURN_BRIDGES
1098 void
1099 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1100     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1101     struct sockaddr *sdl0)
1102 {
1103 
1104 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1105 	    RT_DEFAULT_FIB);
1106 }
1107 #endif
1108 
1109 caddr_t
1110 nd6_ifptomac(struct ifnet *ifp)
1111 {
1112 	switch (ifp->if_type) {
1113 	case IFT_ETHER:
1114 	case IFT_IEEE1394:
1115 	case IFT_L2VLAN:
1116 	case IFT_INFINIBAND:
1117 	case IFT_BRIDGE:
1118 		return IF_LLADDR(ifp);
1119 	default:
1120 		return NULL;
1121 	}
1122 }
1123 
1124 struct dadq {
1125 	TAILQ_ENTRY(dadq) dad_list;
1126 	struct ifaddr *dad_ifa;
1127 	int dad_count;		/* max NS to send */
1128 	int dad_ns_tcount;	/* # of trials to send NS */
1129 	int dad_ns_ocount;	/* NS sent so far */
1130 	int dad_ns_icount;
1131 	int dad_na_icount;
1132 	int dad_ns_lcount;	/* looped back NS */
1133 	int dad_loopbackprobe;	/* probing state for loopback detection */
1134 	struct callout dad_timer_ch;
1135 	struct vnet *dad_vnet;
1136 	u_int dad_refcnt;
1137 #define	ND_OPT_NONCE_LEN32 \
1138 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1139 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1140 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1141 };
1142 
1143 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1144 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1145 #define	V_dadq			VNET(dadq)
1146 #define	V_dad_rwlock		VNET(dad_rwlock)
1147 
1148 #define	DADQ_LOCKPTR()		(&V_dad_rwlock)
1149 #define	DADQ_LOCK_INIT()	rw_init(DADQ_LOCKPTR(), "nd6 DAD queue")
1150 #define	DADQ_RLOCK()		rw_rlock(DADQ_LOCKPTR())
1151 #define	DADQ_RUNLOCK()		rw_runlock(DADQ_LOCKPTR())
1152 #define	DADQ_WLOCK()		rw_wlock(DADQ_LOCKPTR())
1153 #define	DADQ_WUNLOCK()		rw_wunlock(DADQ_LOCKPTR())
1154 
1155 #define	DADQ_LOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_LOCKED);
1156 #define	DADQ_RLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_RLOCKED);
1157 #define	DADQ_WLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_WLOCKED);
1158 
1159 static void
1160 nd6_dad_add(struct dadq *dp)
1161 {
1162 	DADQ_WLOCK_ASSERT();
1163 
1164 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1165 	dp->dad_ondadq = true;
1166 }
1167 
1168 static void
1169 nd6_dad_del(struct dadq *dp)
1170 {
1171 	DADQ_WLOCK_ASSERT();
1172 
1173 	if (dp->dad_ondadq) {
1174 		/*
1175 		 * Remove dp from the dadq and release the dadq's
1176 		 * reference.
1177 		 */
1178 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1179 		dp->dad_ondadq = false;
1180 		nd6_dad_rele(dp);
1181 	}
1182 }
1183 
1184 static struct dadq *
1185 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1186 {
1187 	struct dadq *dp;
1188 
1189 	DADQ_LOCK_ASSERT();
1190 
1191 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1192 		if (dp->dad_ifa != ifa)
1193 			continue;
1194 
1195 		/*
1196 		 * Skip if the nonce matches the received one.
1197 		 * +2 in the length is required because of type and
1198 		 * length fields are included in a header.
1199 		 */
1200 		if (n != NULL &&
1201 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1202 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1203 		    ND_OPT_NONCE_LEN) == 0) {
1204 			dp->dad_ns_lcount++;
1205 			continue;
1206 		}
1207 		break;
1208 	}
1209 
1210 	return (dp);
1211 }
1212 
1213 static void
1214 nd6_dad_starttimer(struct dadq *dp, int ticks)
1215 {
1216 	DADQ_WLOCK_ASSERT();
1217 
1218 	callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp);
1219 }
1220 
1221 static void
1222 nd6_dad_stoptimer(struct dadq *dp)
1223 {
1224 	callout_drain(&dp->dad_timer_ch);
1225 }
1226 
1227 static void
1228 nd6_dad_rele(struct dadq *dp)
1229 {
1230 	if (refcount_release(&dp->dad_refcnt)) {
1231 		KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp));
1232 		ifa_free(dp->dad_ifa);
1233 		free(dp, M_IP6NDP);
1234 	}
1235 }
1236 
1237 void
1238 nd6_dad_init(void)
1239 {
1240 	DADQ_LOCK_INIT();
1241 	TAILQ_INIT(&V_dadq);
1242 }
1243 
1244 /*
1245  * Start Duplicate Address Detection (DAD) for specified interface address.
1246  */
1247 void
1248 nd6_dad_start(struct ifaddr *ifa, int delay)
1249 {
1250 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1251 	struct dadq *dp;
1252 	char ip6buf[INET6_ADDRSTRLEN];
1253 
1254 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1255 	    ("starting DAD on non-tentative address %p", ifa));
1256 
1257 	/*
1258 	 * If we don't need DAD, don't do it.
1259 	 * There are several cases:
1260 	 * - DAD is disabled globally or on the interface
1261 	 * - the interface address is anycast
1262 	 */
1263 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1264 	    V_ip6_dad_count == 0 ||
1265 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1266 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1267 		return;
1268 	}
1269 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1270 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1271 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1272 		return;
1273 
1274 	DADQ_WLOCK();
1275 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1276 		/*
1277 		 * DAD is already in progress.  Let the existing entry
1278 		 * finish it.
1279 		 */
1280 		DADQ_WUNLOCK();
1281 		return;
1282 	}
1283 
1284 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1285 	if (dp == NULL) {
1286 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1287 			"%s(%s)\n",
1288 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1289 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1290 		return;
1291 	}
1292 	callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(),
1293 	    CALLOUT_RETURNUNLOCKED);
1294 #ifdef VIMAGE
1295 	dp->dad_vnet = curvnet;
1296 #endif
1297 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1298 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1299 
1300 	/*
1301 	 * Send NS packet for DAD, ip6_dad_count times.
1302 	 * Note that we must delay the first transmission, if this is the
1303 	 * first packet to be sent from the interface after interface
1304 	 * (re)initialization.
1305 	 */
1306 	dp->dad_ifa = ifa;
1307 	ifa_ref(dp->dad_ifa);
1308 	dp->dad_count = V_ip6_dad_count;
1309 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1310 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1311 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1312 
1313 	/* Add this to the dadq and add a reference for the dadq. */
1314 	refcount_init(&dp->dad_refcnt, 1);
1315 	nd6_dad_add(dp);
1316 	nd6_dad_starttimer(dp, delay);
1317 	DADQ_WUNLOCK();
1318 }
1319 
1320 /*
1321  * terminate DAD unconditionally.  used for address removals.
1322  */
1323 void
1324 nd6_dad_stop(struct ifaddr *ifa)
1325 {
1326 	struct dadq *dp;
1327 
1328 	DADQ_WLOCK();
1329 	dp = nd6_dad_find(ifa, NULL);
1330 	if (dp == NULL) {
1331 		DADQ_WUNLOCK();
1332 		/* DAD wasn't started yet */
1333 		return;
1334 	}
1335 
1336 	/*
1337 	 * Acquire a temporary reference so that we can safely stop the callout.
1338 	 */
1339 	(void)refcount_acquire(&dp->dad_refcnt);
1340 	nd6_dad_del(dp);
1341 	DADQ_WUNLOCK();
1342 
1343 	nd6_dad_stoptimer(dp);
1344 	nd6_dad_rele(dp);
1345 }
1346 
1347 static void
1348 nd6_dad_timer(void *arg)
1349 {
1350 	struct dadq *dp = arg;
1351 	struct ifaddr *ifa = dp->dad_ifa;
1352 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1353 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1354 	char ip6buf[INET6_ADDRSTRLEN];
1355 	struct epoch_tracker et;
1356 
1357 	CURVNET_SET(dp->dad_vnet);
1358 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1359 
1360 	NET_EPOCH_ENTER(et);
1361 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1362 		/* Do not need DAD for ifdisabled interface. */
1363 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1364 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1365 		goto err;
1366 	}
1367 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1368 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1369 			"%s(%s)\n",
1370 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1371 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1372 		goto err;
1373 	}
1374 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1375 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1376 			"%s(%s)\n",
1377 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1378 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1379 		goto err;
1380 	}
1381 
1382 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1383 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1384 	    (((ifp->if_flags & IFF_UP) == 0) ||
1385 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1386 		nd6log((LOG_INFO, "%s: could not run DAD "
1387 		    "because the interface was down or not running.\n",
1388 		    if_name(ifa->ifa_ifp)));
1389 		goto err;
1390 	}
1391 
1392 	/* Need more checks? */
1393 	if (dp->dad_ns_ocount < dp->dad_count) {
1394 		/*
1395 		 * We have more NS to go.  Send NS packet for DAD.
1396 		 */
1397 		nd6_dad_starttimer(dp,
1398 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1399 		nd6_dad_ns_output(dp);
1400 		goto done;
1401 	} else {
1402 		/*
1403 		 * We have transmitted sufficient number of DAD packets.
1404 		 * See what we've got.
1405 		 */
1406 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) {
1407 			/* We've seen NS or NA, means DAD has failed. */
1408 			nd6_dad_duplicated(ifa, dp);
1409 		} else if (V_dad_enhanced != 0 &&
1410 		    dp->dad_ns_lcount > 0 &&
1411 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1412 			/*
1413 			 * Sec. 4.1 in RFC 7527 requires transmission of
1414 			 * additional probes until the loopback condition
1415 			 * becomes clear when a looped back probe is detected.
1416 			 */
1417 			log(LOG_ERR, "%s: a looped back NS message is "
1418 			    "detected during DAD for %s.  "
1419 			    "Another DAD probes are being sent.\n",
1420 			    if_name(ifa->ifa_ifp),
1421 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1422 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1423 			/*
1424 			 * Send an NS immediately and increase dad_count by
1425 			 * V_nd6_mmaxtries - 1.
1426 			 */
1427 			dp->dad_count =
1428 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1429 			nd6_dad_starttimer(dp,
1430 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1431 			nd6_dad_ns_output(dp);
1432 			goto done;
1433 		} else {
1434 			/*
1435 			 * We are done with DAD.  No NA came, no NS came.
1436 			 * No duplicate address found.  Check IFDISABLED flag
1437 			 * again in case that it is changed between the
1438 			 * beginning of this function and here.
1439 			 */
1440 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1441 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1442 
1443 			nd6log((LOG_DEBUG,
1444 			    "%s: DAD complete for %s - no duplicates found\n",
1445 			    if_name(ifa->ifa_ifp),
1446 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1447 			if (dp->dad_ns_lcount > 0)
1448 				log(LOG_ERR, "%s: DAD completed while "
1449 				    "a looped back NS message is detected "
1450 				    "during DAD for %s.\n",
1451 				    if_name(ifa->ifa_ifp),
1452 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1453 		}
1454 	}
1455 err:
1456 	nd6_dad_del(dp);
1457 	DADQ_WUNLOCK();
1458 done:
1459 	NET_EPOCH_EXIT(et);
1460 	CURVNET_RESTORE();
1461 }
1462 
1463 static void
1464 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1465 {
1466 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1467 	struct ifnet *ifp;
1468 	char ip6buf[INET6_ADDRSTRLEN];
1469 
1470 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1471 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1472 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1473 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1474 	    dp->dad_na_icount);
1475 
1476 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1477 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1478 
1479 	ifp = ifa->ifa_ifp;
1480 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1481 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1482 	log(LOG_ERR, "%s: manual intervention required\n",
1483 	    if_name(ifp));
1484 
1485 	/*
1486 	 * If the address is a link-local address formed from an interface
1487 	 * identifier based on the hardware address which is supposed to be
1488 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1489 	 * operation on the interface SHOULD be disabled.
1490 	 * [RFC 4862, Section 5.4.5]
1491 	 */
1492 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1493 		struct in6_addr in6;
1494 
1495 		/*
1496 		 * To avoid over-reaction, we only apply this logic when we are
1497 		 * very sure that hardware addresses are supposed to be unique.
1498 		 */
1499 		switch (ifp->if_type) {
1500 		case IFT_ETHER:
1501 		case IFT_ATM:
1502 		case IFT_IEEE1394:
1503 		case IFT_INFINIBAND:
1504 			in6 = ia->ia_addr.sin6_addr;
1505 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1506 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1507 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1508 				log(LOG_ERR, "%s: possible hardware address "
1509 				    "duplication detected, disable IPv6\n",
1510 				    if_name(ifp));
1511 			}
1512 			break;
1513 		}
1514 	}
1515 }
1516 
1517 /*
1518  * Transmit a neighbour solicitation for the purpose of DAD.  Returns with the
1519  * DAD queue unlocked.
1520  */
1521 static void
1522 nd6_dad_ns_output(struct dadq *dp)
1523 {
1524 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1525 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1526 	int i;
1527 
1528 	DADQ_WLOCK_ASSERT();
1529 
1530 	dp->dad_ns_tcount++;
1531 	if ((ifp->if_flags & IFF_UP) == 0) {
1532 		DADQ_WUNLOCK();
1533 		return;
1534 	}
1535 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1536 		DADQ_WUNLOCK();
1537 		return;
1538 	}
1539 
1540 	dp->dad_ns_ocount++;
1541 	if (V_dad_enhanced != 0) {
1542 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1543 			dp->dad_nonce[i] = arc4random();
1544 		/*
1545 		 * XXXHRS: Note that in the case that
1546 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1547 		 * different nonces can be looped back in an unexpected
1548 		 * order.  The current implementation recognizes only
1549 		 * the latest nonce on the sender side.  Practically it
1550 		 * should work well in almost all cases.
1551 		 */
1552 	}
1553 	DADQ_WUNLOCK();
1554 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1555 	    (uint8_t *)&dp->dad_nonce[0]);
1556 }
1557 
1558 static void
1559 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1560 {
1561 	struct dadq *dp;
1562 
1563 	if (ifa == NULL)
1564 		panic("ifa == NULL in nd6_dad_ns_input");
1565 
1566 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1567 	if (V_dad_enhanced == 0)
1568 		ndopt_nonce = NULL;
1569 	DADQ_RLOCK();
1570 	dp = nd6_dad_find(ifa, ndopt_nonce);
1571 	if (dp != NULL)
1572 		dp->dad_ns_icount++;
1573 	DADQ_RUNLOCK();
1574 }
1575 
1576 static void
1577 nd6_dad_na_input(struct ifaddr *ifa)
1578 {
1579 	struct dadq *dp;
1580 
1581 	if (ifa == NULL)
1582 		panic("ifa == NULL in nd6_dad_na_input");
1583 
1584 	DADQ_RLOCK();
1585 	dp = nd6_dad_find(ifa, NULL);
1586 	if (dp != NULL)
1587 		dp->dad_na_icount++;
1588 	DADQ_RUNLOCK();
1589 }
1590