xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 652a9748855320619e075c4e83aef2f5294412d2)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40 #include "opt_mpath.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/eventhandler.h>
45 #include <sys/malloc.h>
46 #include <sys/libkern.h>
47 #include <sys/lock.h>
48 #include <sys/rwlock.h>
49 #include <sys/mbuf.h>
50 #include <sys/socket.h>
51 #include <sys/sockio.h>
52 #include <sys/time.h>
53 #include <sys/kernel.h>
54 #include <sys/errno.h>
55 #include <sys/sysctl.h>
56 #include <sys/syslog.h>
57 #include <sys/queue.h>
58 #include <sys/callout.h>
59 #include <sys/refcount.h>
60 
61 #include <net/if.h>
62 #include <net/if_types.h>
63 #include <net/if_dl.h>
64 #include <net/if_var.h>
65 #include <net/route.h>
66 #ifdef RADIX_MPATH
67 #include <net/radix_mpath.h>
68 #endif
69 #include <net/vnet.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <net/if_llatbl.h>
74 #include <netinet6/in6_var.h>
75 #include <netinet6/in6_ifattach.h>
76 #include <netinet/ip6.h>
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/scope6_var.h>
79 #include <netinet6/nd6.h>
80 #include <netinet/icmp6.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet6/send.h>
83 
84 #define SDL(s) ((struct sockaddr_dl *)s)
85 
86 struct dadq;
87 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
88 static void nd6_dad_add(struct dadq *dp);
89 static void nd6_dad_del(struct dadq *dp);
90 static void nd6_dad_rele(struct dadq *);
91 static void nd6_dad_starttimer(struct dadq *, int, int);
92 static void nd6_dad_stoptimer(struct dadq *);
93 static void nd6_dad_timer(struct dadq *);
94 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
95 static void nd6_dad_ns_output(struct dadq *);
96 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
97 static void nd6_dad_na_input(struct ifaddr *);
98 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
99     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
100 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
101     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
102 
103 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
104 #define	V_dad_enhanced			VNET(dad_enhanced)
105 
106 SYSCTL_DECL(_net_inet6_ip6);
107 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
108     &VNET_NAME(dad_enhanced), 0,
109     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
110 
111 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
112 						   transmit DAD packet */
113 #define	V_dad_maxtry			VNET(dad_maxtry)
114 
115 /*
116  * Input a Neighbor Solicitation Message.
117  *
118  * Based on RFC 2461
119  * Based on RFC 2462 (duplicate address detection)
120  */
121 void
122 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
123 {
124 	struct ifnet *ifp;
125 	struct ip6_hdr *ip6;
126 	struct nd_neighbor_solicit *nd_ns;
127 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
128 	struct ifaddr *ifa;
129 	struct sockaddr_dl proxydl;
130 	union nd_opts ndopts;
131 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
132 	char *lladdr;
133 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
134 
135 	ifa = NULL;
136 
137 	/* RFC 6980: Nodes MUST silently ignore fragments */
138 	if(m->m_flags & M_FRAGMENTED)
139 		goto freeit;
140 
141 	ifp = m->m_pkthdr.rcvif;
142 	ip6 = mtod(m, struct ip6_hdr *);
143 	if (ip6->ip6_hlim != 255) {
144 		nd6log((LOG_ERR,
145 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
146 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
147 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
148 		goto bads;
149 	}
150 
151 	if (m->m_len < off + icmp6len) {
152 		m = m_pullup(m, off + icmp6len);
153 		if (m == NULL) {
154 			IP6STAT_INC(ip6s_exthdrtoolong);
155 			return;
156 		}
157 	}
158 	ip6 = mtod(m, struct ip6_hdr *);
159 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
160 
161 	saddr6 = ip6->ip6_src;
162 	daddr6 = ip6->ip6_dst;
163 	taddr6 = nd_ns->nd_ns_target;
164 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
165 		goto bad;
166 
167 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
168 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
169 		rflag = 0;
170 
171 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
172 		/* dst has to be a solicited node multicast address. */
173 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
174 		    /* don't check ifindex portion */
175 		    daddr6.s6_addr32[1] == 0 &&
176 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
177 		    daddr6.s6_addr8[12] == 0xff) {
178 			; /* good */
179 		} else {
180 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
181 			    "(wrong ip6 dst)\n"));
182 			goto bad;
183 		}
184 	} else if (!V_nd6_onlink_ns_rfc4861) {
185 		struct sockaddr_in6 src_sa6;
186 
187 		/*
188 		 * According to recent IETF discussions, it is not a good idea
189 		 * to accept a NS from an address which would not be deemed
190 		 * to be a neighbor otherwise.  This point is expected to be
191 		 * clarified in future revisions of the specification.
192 		 */
193 		bzero(&src_sa6, sizeof(src_sa6));
194 		src_sa6.sin6_family = AF_INET6;
195 		src_sa6.sin6_len = sizeof(src_sa6);
196 		src_sa6.sin6_addr = saddr6;
197 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
198 			nd6log((LOG_INFO, "nd6_ns_input: "
199 				"NS packet from non-neighbor\n"));
200 			goto bad;
201 		}
202 	}
203 
204 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
205 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
206 		goto bad;
207 	}
208 
209 	icmp6len -= sizeof(*nd_ns);
210 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
211 	if (nd6_options(&ndopts) < 0) {
212 		nd6log((LOG_INFO,
213 		    "nd6_ns_input: invalid ND option, ignored\n"));
214 		/* nd6_options have incremented stats */
215 		goto freeit;
216 	}
217 
218 	lladdr = NULL;
219 	lladdrlen = 0;
220 	if (ndopts.nd_opts_src_lladdr) {
221 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
222 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
223 	}
224 
225 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
226 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
227 		    "(link-layer address option)\n"));
228 		goto bad;
229 	}
230 
231 	/*
232 	 * Attaching target link-layer address to the NA?
233 	 * (RFC 2461 7.2.4)
234 	 *
235 	 * NS IP dst is unicast/anycast			MUST NOT add
236 	 * NS IP dst is solicited-node multicast	MUST add
237 	 *
238 	 * In implementation, we add target link-layer address by default.
239 	 * We do not add one in MUST NOT cases.
240 	 */
241 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
242 		tlladdr = 0;
243 	else
244 		tlladdr = 1;
245 
246 	/*
247 	 * Target address (taddr6) must be either:
248 	 * (1) Valid unicast/anycast address for my receiving interface,
249 	 * (2) Unicast address for which I'm offering proxy service, or
250 	 * (3) "tentative" address on which DAD is being performed.
251 	 */
252 	/* (1) and (3) check. */
253 	if (ifp->if_carp)
254 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
255 	else
256 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
257 
258 	/* (2) check. */
259 	proxy = 0;
260 	if (ifa == NULL) {
261 		struct sockaddr_dl rt_gateway;
262 		struct rt_addrinfo info;
263 		struct sockaddr_in6 dst6;
264 
265 		bzero(&dst6, sizeof(dst6));
266 		dst6.sin6_len = sizeof(struct sockaddr_in6);
267 		dst6.sin6_family = AF_INET6;
268 		dst6.sin6_addr = taddr6;
269 
270 		bzero(&rt_gateway, sizeof(rt_gateway));
271 		rt_gateway.sdl_len = sizeof(rt_gateway);
272 		bzero(&info, sizeof(info));
273 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
274 
275 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
276 		    0, 0, &info) == 0) {
277 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
278 			    rt_gateway.sdl_family == AF_LINK) {
279 
280 				/*
281 				 * proxy NDP for single entry
282 				 */
283 				proxydl = *SDL(&rt_gateway);
284 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
285 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
286 				if (ifa)
287 					proxy = 1;
288 			}
289 		}
290 	}
291 	if (ifa == NULL) {
292 		/*
293 		 * We've got an NS packet, and we don't have that adddress
294 		 * assigned for us.  We MUST silently ignore it.
295 		 * See RFC2461 7.2.3.
296 		 */
297 		goto freeit;
298 	}
299 	myaddr6 = *IFA_IN6(ifa);
300 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
301 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
302 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
303 		goto freeit;
304 
305 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
306 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
307 		    "(if %d, NS packet %d)\n",
308 		    ip6_sprintf(ip6bufs, &taddr6),
309 		    ifp->if_addrlen, lladdrlen - 2));
310 		goto bad;
311 	}
312 
313 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
314 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
315 		    ip6_sprintf(ip6bufs, &saddr6)));
316 		goto freeit;
317 	}
318 
319 	/*
320 	 * We have neighbor solicitation packet, with target address equals to
321 	 * one of my tentative address.
322 	 *
323 	 * src addr	how to process?
324 	 * ---		---
325 	 * multicast	of course, invalid (rejected in ip6_input)
326 	 * unicast	somebody is doing address resolution -> ignore
327 	 * unspec	dup address detection
328 	 *
329 	 * The processing is defined in RFC 2462.
330 	 */
331 	if (tentative) {
332 		/*
333 		 * If source address is unspecified address, it is for
334 		 * duplicate address detection.
335 		 *
336 		 * If not, the packet is for addess resolution;
337 		 * silently ignore it.
338 		 */
339 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
340 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
341 
342 		goto freeit;
343 	}
344 
345 	/*
346 	 * If the source address is unspecified address, entries must not
347 	 * be created or updated.
348 	 * It looks that sender is performing DAD.  Output NA toward
349 	 * all-node multicast address, to tell the sender that I'm using
350 	 * the address.
351 	 * S bit ("solicited") must be zero.
352 	 */
353 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
354 		struct in6_addr in6_all;
355 
356 		in6_all = in6addr_linklocal_allnodes;
357 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
358 			goto bad;
359 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
360 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
361 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
362 		    M_GETFIB(m));
363 		goto freeit;
364 	}
365 
366 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
367 	    ND_NEIGHBOR_SOLICIT, 0);
368 
369 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
370 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
371 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
372 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
373  freeit:
374 	if (ifa != NULL)
375 		ifa_free(ifa);
376 	m_freem(m);
377 	return;
378 
379  bad:
380 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
381 		ip6_sprintf(ip6bufs, &saddr6)));
382 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
383 		ip6_sprintf(ip6bufs, &daddr6)));
384 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
385 		ip6_sprintf(ip6bufs, &taddr6)));
386  bads:
387 	ICMP6STAT_INC(icp6s_badns);
388 	if (ifa != NULL)
389 		ifa_free(ifa);
390 	m_freem(m);
391 }
392 
393 /*
394  * Output a Neighbor Solicitation Message. Caller specifies:
395  *	- ICMP6 header source IP6 address
396  *	- ND6 header target IP6 address
397  *	- ND6 header source datalink address
398  *
399  * Based on RFC 2461
400  * Based on RFC 2462 (duplicate address detection)
401  *
402  *    ln - for source address determination
403  * nonce - If non-NULL, NS is used for duplicate address detection and
404  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
405  */
406 static void
407 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
408     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
409     uint8_t *nonce, u_int fibnum)
410 {
411 	struct mbuf *m;
412 	struct m_tag *mtag;
413 	struct ip6_hdr *ip6;
414 	struct nd_neighbor_solicit *nd_ns;
415 	struct ip6_moptions im6o;
416 	int icmp6len;
417 	int maxlen;
418 	caddr_t mac;
419 
420 	if (IN6_IS_ADDR_MULTICAST(taddr6))
421 		return;
422 
423 	/* estimate the size of message */
424 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
425 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
426 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
427 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
428 	    __func__, max_linkhdr, maxlen, MCLBYTES));
429 
430 	if (max_linkhdr + maxlen > MHLEN)
431 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
432 	else
433 		m = m_gethdr(M_NOWAIT, MT_DATA);
434 	if (m == NULL)
435 		return;
436 	M_SETFIB(m, fibnum);
437 
438 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
439 		m->m_flags |= M_MCAST;
440 		im6o.im6o_multicast_ifp = ifp;
441 		im6o.im6o_multicast_hlim = 255;
442 		im6o.im6o_multicast_loop = 0;
443 	}
444 
445 	icmp6len = sizeof(*nd_ns);
446 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
447 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
448 
449 	/* fill neighbor solicitation packet */
450 	ip6 = mtod(m, struct ip6_hdr *);
451 	ip6->ip6_flow = 0;
452 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
453 	ip6->ip6_vfc |= IPV6_VERSION;
454 	/* ip6->ip6_plen will be set later */
455 	ip6->ip6_nxt = IPPROTO_ICMPV6;
456 	ip6->ip6_hlim = 255;
457 	if (daddr6)
458 		ip6->ip6_dst = *daddr6;
459 	else {
460 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
461 		ip6->ip6_dst.s6_addr16[1] = 0;
462 		ip6->ip6_dst.s6_addr32[1] = 0;
463 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
464 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
465 		ip6->ip6_dst.s6_addr8[12] = 0xff;
466 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
467 			goto bad;
468 	}
469 	if (nonce == NULL) {
470 		struct ifaddr *ifa = NULL;
471 
472 		/*
473 		 * RFC2461 7.2.2:
474 		 * "If the source address of the packet prompting the
475 		 * solicitation is the same as one of the addresses assigned
476 		 * to the outgoing interface, that address SHOULD be placed
477 		 * in the IP Source Address of the outgoing solicitation.
478 		 * Otherwise, any one of the addresses assigned to the
479 		 * interface should be used."
480 		 *
481 		 * We use the source address for the prompting packet
482 		 * (saddr6), if saddr6 belongs to the outgoing interface.
483 		 * Otherwise, we perform the source address selection as usual.
484 		 */
485 
486 		if (saddr6 != NULL)
487 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
488 		if (ifa != NULL) {
489 			/* ip6_src set already. */
490 			ip6->ip6_src = *saddr6;
491 			ifa_free(ifa);
492 		} else {
493 			int error;
494 			struct in6_addr dst6, src6;
495 			uint32_t scopeid;
496 
497 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
498 			error = in6_selectsrc_addr(fibnum, &dst6,
499 			    scopeid, ifp, &src6, NULL);
500 			if (error) {
501 				char ip6buf[INET6_ADDRSTRLEN];
502 				nd6log((LOG_DEBUG, "%s: source can't be "
503 				    "determined: dst=%s, error=%d\n", __func__,
504 				    ip6_sprintf(ip6buf, &dst6),
505 				    error));
506 				goto bad;
507 			}
508 			ip6->ip6_src = src6;
509 		}
510 	} else {
511 		/*
512 		 * Source address for DAD packet must always be IPv6
513 		 * unspecified address. (0::0)
514 		 * We actually don't have to 0-clear the address (we did it
515 		 * above), but we do so here explicitly to make the intention
516 		 * clearer.
517 		 */
518 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
519 	}
520 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
521 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
522 	nd_ns->nd_ns_code = 0;
523 	nd_ns->nd_ns_reserved = 0;
524 	nd_ns->nd_ns_target = *taddr6;
525 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
526 
527 	/*
528 	 * Add source link-layer address option.
529 	 *
530 	 *				spec		implementation
531 	 *				---		---
532 	 * DAD packet			MUST NOT	do not add the option
533 	 * there's no link layer address:
534 	 *				impossible	do not add the option
535 	 * there's link layer address:
536 	 *	Multicast NS		MUST add one	add the option
537 	 *	Unicast NS		SHOULD add one	add the option
538 	 */
539 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
540 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
541 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
542 		/* 8 byte alignments... */
543 		optlen = (optlen + 7) & ~7;
544 
545 		m->m_pkthdr.len += optlen;
546 		m->m_len += optlen;
547 		icmp6len += optlen;
548 		bzero((caddr_t)nd_opt, optlen);
549 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
550 		nd_opt->nd_opt_len = optlen >> 3;
551 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
552 	}
553 	/*
554 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
555 	 * This behavior is documented as Enhanced Duplicate Address
556 	 * Detection in RFC 7527.
557 	 * net.inet6.ip6.dad_enhanced=0 disables this.
558 	 */
559 	if (V_dad_enhanced != 0 && nonce != NULL) {
560 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
561 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
562 		/* 8-byte alignment is required. */
563 		optlen = (optlen + 7) & ~7;
564 
565 		m->m_pkthdr.len += optlen;
566 		m->m_len += optlen;
567 		icmp6len += optlen;
568 		bzero((caddr_t)nd_opt, optlen);
569 		nd_opt->nd_opt_type = ND_OPT_NONCE;
570 		nd_opt->nd_opt_len = optlen >> 3;
571 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
572 	}
573 	ip6->ip6_plen = htons((u_short)icmp6len);
574 	nd_ns->nd_ns_cksum = 0;
575 	nd_ns->nd_ns_cksum =
576 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
577 
578 	if (send_sendso_input_hook != NULL) {
579 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
580 			sizeof(unsigned short), M_NOWAIT);
581 		if (mtag == NULL)
582 			goto bad;
583 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
584 		m_tag_prepend(m, mtag);
585 	}
586 
587 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
588 	    &im6o, NULL, NULL);
589 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
590 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
591 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
592 
593 	return;
594 
595   bad:
596 	m_freem(m);
597 }
598 
599 #ifndef BURN_BRIDGES
600 void
601 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
602     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
603 {
604 
605 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
606 }
607 #endif
608 /*
609  * Neighbor advertisement input handling.
610  *
611  * Based on RFC 2461
612  * Based on RFC 2462 (duplicate address detection)
613  *
614  * the following items are not implemented yet:
615  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
616  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
617  */
618 void
619 nd6_na_input(struct mbuf *m, int off, int icmp6len)
620 {
621 	struct ifnet *ifp;
622 	struct ip6_hdr *ip6;
623 	struct ifaddr *ifa;
624 	struct llentry *ln;
625 	struct mbuf *chain;
626 	struct nd_neighbor_advert *nd_na;
627 	struct in6_addr daddr6, taddr6;
628 	struct sockaddr_in6 sin6;
629 	union nd_opts ndopts;
630 	u_char linkhdr[LLE_MAX_LINKHDR];
631 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
632 	char *lladdr;
633 	size_t linkhdrsize;
634 	int flags, is_override, is_router, is_solicited;
635 	int lladdr_off, lladdrlen, checklink;
636 
637 	NET_EPOCH_ASSERT();
638 
639 	chain = NULL;
640 	ln = NULL;
641 	checklink = 0;
642 
643 	/* RFC 6980: Nodes MUST silently ignore fragments */
644 	if(m->m_flags & M_FRAGMENTED)
645 		goto freeit;
646 
647 	ifp = m->m_pkthdr.rcvif;
648 	ip6 = mtod(m, struct ip6_hdr *);
649 	if (ip6->ip6_hlim != 255) {
650 		nd6log((LOG_ERR,
651 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
652 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
653 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
654 		goto bad;
655 	}
656 
657 	if (m->m_len < off + icmp6len) {
658 		m = m_pullup(m, off + icmp6len);
659 		if (m == NULL) {
660 			IP6STAT_INC(ip6s_exthdrtoolong);
661 			return;
662 		}
663 	}
664 	ip6 = mtod(m, struct ip6_hdr *);
665 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
666 
667 	flags = nd_na->nd_na_flags_reserved;
668 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
669 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
670 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
671 
672 	taddr6 = nd_na->nd_na_target;
673 	if (in6_setscope(&taddr6, ifp, NULL))
674 		goto bad;	/* XXX: impossible */
675 
676 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
677 		nd6log((LOG_ERR,
678 		    "nd6_na_input: invalid target address %s\n",
679 		    ip6_sprintf(ip6bufs, &taddr6)));
680 		goto bad;
681 	}
682 
683 	daddr6 = ip6->ip6_dst;
684 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
685 		if (is_solicited) {
686 			nd6log((LOG_ERR,
687 			    "nd6_na_input: a solicited adv is multicasted\n"));
688 			goto bad;
689 		}
690 
691 	icmp6len -= sizeof(*nd_na);
692 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
693 	if (nd6_options(&ndopts) < 0) {
694 		nd6log((LOG_INFO,
695 		    "nd6_na_input: invalid ND option, ignored\n"));
696 		/* nd6_options have incremented stats */
697 		goto freeit;
698 	}
699 
700 	lladdr = NULL;
701 	lladdrlen = 0;
702 	if (ndopts.nd_opts_tgt_lladdr) {
703 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
704 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
705 	}
706 
707 	/*
708 	 * This effectively disables the DAD check on a non-master CARP
709 	 * address.
710 	 */
711 	if (ifp->if_carp)
712 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
713 	else
714 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
715 
716 	/*
717 	 * Target address matches one of my interface address.
718 	 *
719 	 * If my address is tentative, this means that there's somebody
720 	 * already using the same address as mine.  This indicates DAD failure.
721 	 * This is defined in RFC 2462.
722 	 *
723 	 * Otherwise, process as defined in RFC 2461.
724 	 */
725 	if (ifa
726 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
727 		nd6_dad_na_input(ifa);
728 		ifa_free(ifa);
729 		goto freeit;
730 	}
731 
732 	/* Just for safety, maybe unnecessary. */
733 	if (ifa) {
734 		ifa_free(ifa);
735 		log(LOG_ERR,
736 		    "nd6_na_input: duplicate IP6 address %s\n",
737 		    ip6_sprintf(ip6bufs, &taddr6));
738 		goto freeit;
739 	}
740 
741 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
742 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
743 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
744 		    ifp->if_addrlen, lladdrlen - 2));
745 		goto bad;
746 	}
747 
748 	/*
749 	 * If no neighbor cache entry is found, NA SHOULD silently be
750 	 * discarded.
751 	 */
752 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
753 	if (ln == NULL) {
754 		goto freeit;
755 	}
756 
757 	/*
758 	 * Do not try to override static entry.
759 	 */
760 	if (ln->la_flags & LLE_STATIC)
761 		goto freeit;
762 
763 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
764 		/*
765 		 * If the link-layer has address, and no lladdr option came,
766 		 * discard the packet.
767 		 */
768 		if (ifp->if_addrlen && lladdr == NULL) {
769 			goto freeit;
770 		}
771 
772 		/*
773 		 * Record link-layer address, and update the state.
774 		 */
775 		linkhdrsize = sizeof(linkhdr);
776 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
777 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
778 			return;
779 
780 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
781 		    lladdr_off) == 0) {
782 			ln = NULL;
783 			goto freeit;
784 		}
785 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
786 		if (is_solicited)
787 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
788 		else
789 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
790 		if ((ln->ln_router = is_router) != 0) {
791 			/*
792 			 * This means a router's state has changed from
793 			 * non-reachable to probably reachable, and might
794 			 * affect the status of associated prefixes..
795 			 */
796 			checklink = 1;
797 		}
798 	} else {
799 		int llchange;
800 
801 		/*
802 		 * Check if the link-layer address has changed or not.
803 		 */
804 		if (lladdr == NULL)
805 			llchange = 0;
806 		else {
807 			if (ln->la_flags & LLE_VALID) {
808 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
809 					llchange = 1;
810 				else
811 					llchange = 0;
812 			} else
813 				llchange = 1;
814 		}
815 
816 		/*
817 		 * This is VERY complex.  Look at it with care.
818 		 *
819 		 * override solicit lladdr llchange	action
820 		 *					(L: record lladdr)
821 		 *
822 		 *	0	0	n	--	(2c)
823 		 *	0	0	y	n	(2b) L
824 		 *	0	0	y	y	(1)    REACHABLE->STALE
825 		 *	0	1	n	--	(2c)   *->REACHABLE
826 		 *	0	1	y	n	(2b) L *->REACHABLE
827 		 *	0	1	y	y	(1)    REACHABLE->STALE
828 		 *	1	0	n	--	(2a)
829 		 *	1	0	y	n	(2a) L
830 		 *	1	0	y	y	(2a) L *->STALE
831 		 *	1	1	n	--	(2a)   *->REACHABLE
832 		 *	1	1	y	n	(2a) L *->REACHABLE
833 		 *	1	1	y	y	(2a) L *->REACHABLE
834 		 */
835 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
836 			/*
837 			 * If state is REACHABLE, make it STALE.
838 			 * no other updates should be done.
839 			 */
840 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
841 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
842 			goto freeit;
843 		} else if (is_override				   /* (2a) */
844 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
845 			|| lladdr == NULL) {			   /* (2c) */
846 			/*
847 			 * Update link-local address, if any.
848 			 */
849 			if (lladdr != NULL) {
850 				linkhdrsize = sizeof(linkhdr);
851 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
852 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
853 					goto freeit;
854 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
855 				    linkhdrsize, lladdr_off) == 0) {
856 					ln = NULL;
857 					goto freeit;
858 				}
859 				EVENTHANDLER_INVOKE(lle_event, ln,
860 				    LLENTRY_RESOLVED);
861 			}
862 
863 			/*
864 			 * If solicited, make the state REACHABLE.
865 			 * If not solicited and the link-layer address was
866 			 * changed, make it STALE.
867 			 */
868 			if (is_solicited)
869 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
870 			else {
871 				if (lladdr != NULL && llchange)
872 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
873 			}
874 		}
875 
876 		if (ln->ln_router && !is_router) {
877 			/*
878 			 * The peer dropped the router flag.
879 			 * Remove the sender from the Default Router List and
880 			 * update the Destination Cache entries.
881 			 */
882 			struct ifnet *nd6_ifp;
883 
884 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
885 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
886 			    (ND_IFINFO(nd6_ifp)->flags &
887 			     ND6_IFF_ACCEPT_RTADV) != 0)
888 				/*
889 				 * Even if the neighbor is not in the default
890 				 * router list, the neighbor may be used as a
891 				 * next hop for some destinations (e.g. redirect
892 				 * case). So we must call rt6_flush explicitly.
893 				 */
894 				rt6_flush(&ip6->ip6_src, ifp);
895 		}
896 		ln->ln_router = is_router;
897 	}
898         /* XXX - QL
899 	 *  Does this matter?
900 	 *  rt->rt_flags &= ~RTF_REJECT;
901 	 */
902 	ln->la_asked = 0;
903 	if (ln->la_hold != NULL) {
904 		memset(&sin6, 0, sizeof(sin6));
905 		nd6_grab_holdchain(ln, &chain, &sin6);
906 	}
907  freeit:
908 	if (ln != NULL)
909 		LLE_WUNLOCK(ln);
910 
911 	if (chain != NULL)
912 		nd6_flush_holdchain(ifp, chain, &sin6);
913 
914 	if (checklink)
915 		pfxlist_onlink_check();
916 
917 	m_freem(m);
918 	return;
919 
920  bad:
921 	if (ln != NULL)
922 		LLE_WUNLOCK(ln);
923 
924 	ICMP6STAT_INC(icp6s_badna);
925 	m_freem(m);
926 }
927 
928 /*
929  * Neighbor advertisement output handling.
930  *
931  * Based on RFC 2461
932  *
933  * the following items are not implemented yet:
934  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
935  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
936  *
937  * tlladdr - 1 if include target link-layer address
938  * sdl0 - sockaddr_dl (= proxy NA) or NULL
939  */
940 static void
941 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
942     const struct in6_addr *taddr6, u_long flags, int tlladdr,
943     struct sockaddr *sdl0, u_int fibnum)
944 {
945 	struct mbuf *m;
946 	struct m_tag *mtag;
947 	struct ip6_hdr *ip6;
948 	struct nd_neighbor_advert *nd_na;
949 	struct ip6_moptions im6o;
950 	struct in6_addr daddr6, dst6, src6;
951 	uint32_t scopeid;
952 
953 	int icmp6len, maxlen, error;
954 	caddr_t mac = NULL;
955 
956 	daddr6 = *daddr6_0;	/* make a local copy for modification */
957 
958 	/* estimate the size of message */
959 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
960 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
961 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
962 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
963 	    __func__, max_linkhdr, maxlen, MCLBYTES));
964 
965 	if (max_linkhdr + maxlen > MHLEN)
966 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
967 	else
968 		m = m_gethdr(M_NOWAIT, MT_DATA);
969 	if (m == NULL)
970 		return;
971 	M_SETFIB(m, fibnum);
972 
973 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
974 		m->m_flags |= M_MCAST;
975 		im6o.im6o_multicast_ifp = ifp;
976 		im6o.im6o_multicast_hlim = 255;
977 		im6o.im6o_multicast_loop = 0;
978 	}
979 
980 	icmp6len = sizeof(*nd_na);
981 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
982 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
983 
984 	/* fill neighbor advertisement packet */
985 	ip6 = mtod(m, struct ip6_hdr *);
986 	ip6->ip6_flow = 0;
987 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
988 	ip6->ip6_vfc |= IPV6_VERSION;
989 	ip6->ip6_nxt = IPPROTO_ICMPV6;
990 	ip6->ip6_hlim = 255;
991 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
992 		/* reply to DAD */
993 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
994 		daddr6.s6_addr16[1] = 0;
995 		daddr6.s6_addr32[1] = 0;
996 		daddr6.s6_addr32[2] = 0;
997 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
998 		if (in6_setscope(&daddr6, ifp, NULL))
999 			goto bad;
1000 
1001 		flags &= ~ND_NA_FLAG_SOLICITED;
1002 	}
1003 	ip6->ip6_dst = daddr6;
1004 
1005 	/*
1006 	 * Select a source whose scope is the same as that of the dest.
1007 	 */
1008 	in6_splitscope(&daddr6, &dst6, &scopeid);
1009 	error = in6_selectsrc_addr(fibnum, &dst6,
1010 	    scopeid, ifp, &src6, NULL);
1011 	if (error) {
1012 		char ip6buf[INET6_ADDRSTRLEN];
1013 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1014 		    "determined: dst=%s, error=%d\n",
1015 		    ip6_sprintf(ip6buf, &daddr6), error));
1016 		goto bad;
1017 	}
1018 	ip6->ip6_src = src6;
1019 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1020 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1021 	nd_na->nd_na_code = 0;
1022 	nd_na->nd_na_target = *taddr6;
1023 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1024 
1025 	/*
1026 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1027 	 * see nd6_ns_input() for details.
1028 	 * Basically, if NS packet is sent to unicast/anycast addr,
1029 	 * target lladdr option SHOULD NOT be included.
1030 	 */
1031 	if (tlladdr) {
1032 		/*
1033 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1034 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1035 		 * my address) use lladdr configured for the interface.
1036 		 */
1037 		if (sdl0 == NULL) {
1038 			if (ifp->if_carp)
1039 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1040 			if (mac == NULL)
1041 				mac = nd6_ifptomac(ifp);
1042 		} else if (sdl0->sa_family == AF_LINK) {
1043 			struct sockaddr_dl *sdl;
1044 			sdl = (struct sockaddr_dl *)sdl0;
1045 			if (sdl->sdl_alen == ifp->if_addrlen)
1046 				mac = LLADDR(sdl);
1047 		}
1048 	}
1049 	if (tlladdr && mac) {
1050 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1051 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1052 
1053 		/* roundup to 8 bytes alignment! */
1054 		optlen = (optlen + 7) & ~7;
1055 
1056 		m->m_pkthdr.len += optlen;
1057 		m->m_len += optlen;
1058 		icmp6len += optlen;
1059 		bzero((caddr_t)nd_opt, optlen);
1060 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1061 		nd_opt->nd_opt_len = optlen >> 3;
1062 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1063 	} else
1064 		flags &= ~ND_NA_FLAG_OVERRIDE;
1065 
1066 	ip6->ip6_plen = htons((u_short)icmp6len);
1067 	nd_na->nd_na_flags_reserved = flags;
1068 	nd_na->nd_na_cksum = 0;
1069 	nd_na->nd_na_cksum =
1070 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1071 
1072 	if (send_sendso_input_hook != NULL) {
1073 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1074 		    sizeof(unsigned short), M_NOWAIT);
1075 		if (mtag == NULL)
1076 			goto bad;
1077 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1078 		m_tag_prepend(m, mtag);
1079 	}
1080 
1081 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1082 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1083 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1084 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1085 
1086 	return;
1087 
1088   bad:
1089 	m_freem(m);
1090 }
1091 
1092 #ifndef BURN_BRIDGES
1093 void
1094 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1095     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1096     struct sockaddr *sdl0)
1097 {
1098 
1099 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1100 	    RT_DEFAULT_FIB);
1101 }
1102 #endif
1103 
1104 caddr_t
1105 nd6_ifptomac(struct ifnet *ifp)
1106 {
1107 	switch (ifp->if_type) {
1108 	case IFT_ETHER:
1109 	case IFT_IEEE1394:
1110 	case IFT_L2VLAN:
1111 	case IFT_INFINIBAND:
1112 	case IFT_BRIDGE:
1113 		return IF_LLADDR(ifp);
1114 	default:
1115 		return NULL;
1116 	}
1117 }
1118 
1119 struct dadq {
1120 	TAILQ_ENTRY(dadq) dad_list;
1121 	struct ifaddr *dad_ifa;
1122 	int dad_count;		/* max NS to send */
1123 	int dad_ns_tcount;	/* # of trials to send NS */
1124 	int dad_ns_ocount;	/* NS sent so far */
1125 	int dad_ns_icount;
1126 	int dad_na_icount;
1127 	int dad_ns_lcount;	/* looped back NS */
1128 	int dad_loopbackprobe;	/* probing state for loopback detection */
1129 	struct callout dad_timer_ch;
1130 	struct vnet *dad_vnet;
1131 	u_int dad_refcnt;
1132 #define	ND_OPT_NONCE_LEN32 \
1133 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1134 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1135 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1136 };
1137 
1138 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1139 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1140 #define	V_dadq			VNET(dadq)
1141 #define	V_dad_rwlock		VNET(dad_rwlock)
1142 
1143 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1144 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1145 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1146 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1147 
1148 static void
1149 nd6_dad_add(struct dadq *dp)
1150 {
1151 
1152 	DADQ_WLOCK();
1153 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1154 	dp->dad_ondadq = true;
1155 	DADQ_WUNLOCK();
1156 }
1157 
1158 static void
1159 nd6_dad_del(struct dadq *dp)
1160 {
1161 
1162 	DADQ_WLOCK();
1163 	if (dp->dad_ondadq) {
1164 		/*
1165 		 * Remove dp from the dadq and release the dadq's
1166 		 * reference.
1167 		 */
1168 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1169 		dp->dad_ondadq = false;
1170 		DADQ_WUNLOCK();
1171 		nd6_dad_rele(dp);
1172 	} else
1173 		DADQ_WUNLOCK();
1174 }
1175 
1176 static struct dadq *
1177 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1178 {
1179 	struct dadq *dp;
1180 
1181 	DADQ_RLOCK();
1182 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1183 		if (dp->dad_ifa != ifa)
1184 			continue;
1185 		/*
1186 		 * Skip if the nonce matches the received one.
1187 		 * +2 in the length is required because of type and
1188 		 * length fields are included in a header.
1189 		 */
1190 		if (n != NULL &&
1191 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1192 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1193 		        ND_OPT_NONCE_LEN) == 0) {
1194 			dp->dad_ns_lcount++;
1195 			continue;
1196 		}
1197 		refcount_acquire(&dp->dad_refcnt);
1198 		break;
1199 	}
1200 	DADQ_RUNLOCK();
1201 
1202 	return (dp);
1203 }
1204 
1205 static void
1206 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1207 {
1208 
1209 	NET_EPOCH_ASSERT();
1210 
1211 	if (send_ns != 0)
1212 		nd6_dad_ns_output(dp);
1213 	callout_reset(&dp->dad_timer_ch, ticks,
1214 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1215 }
1216 
1217 static void
1218 nd6_dad_stoptimer(struct dadq *dp)
1219 {
1220 
1221 	callout_drain(&dp->dad_timer_ch);
1222 }
1223 
1224 static void
1225 nd6_dad_rele(struct dadq *dp)
1226 {
1227 
1228 	if (refcount_release(&dp->dad_refcnt)) {
1229 		ifa_free(dp->dad_ifa);
1230 		free(dp, M_IP6NDP);
1231 	}
1232 }
1233 
1234 void
1235 nd6_dad_init(void)
1236 {
1237 
1238 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1239 	TAILQ_INIT(&V_dadq);
1240 }
1241 
1242 /*
1243  * Start Duplicate Address Detection (DAD) for specified interface address.
1244  */
1245 void
1246 nd6_dad_start(struct ifaddr *ifa, int delay)
1247 {
1248 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1249 	struct dadq *dp;
1250 	char ip6buf[INET6_ADDRSTRLEN];
1251 	struct epoch_tracker et;
1252 
1253 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1254 	    ("starting DAD on non-tentative address %p", ifa));
1255 
1256 	/*
1257 	 * If we don't need DAD, don't do it.
1258 	 * There are several cases:
1259 	 * - DAD is disabled globally or on the interface
1260 	 * - the interface address is anycast
1261 	 */
1262 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1263 	    V_ip6_dad_count == 0 ||
1264 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1265 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1266 		return;
1267 	}
1268 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1269 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1270 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1271 		return;
1272 
1273 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1274 		/*
1275 		 * DAD is already in progress.  Let the existing entry
1276 		 * finish it.
1277 		 */
1278 		nd6_dad_rele(dp);
1279 		return;
1280 	}
1281 
1282 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1283 	if (dp == NULL) {
1284 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1285 			"%s(%s)\n",
1286 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1287 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1288 		return;
1289 	}
1290 	callout_init(&dp->dad_timer_ch, 0);
1291 #ifdef VIMAGE
1292 	dp->dad_vnet = curvnet;
1293 #endif
1294 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1295 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1296 
1297 	/*
1298 	 * Send NS packet for DAD, ip6_dad_count times.
1299 	 * Note that we must delay the first transmission, if this is the
1300 	 * first packet to be sent from the interface after interface
1301 	 * (re)initialization.
1302 	 */
1303 	dp->dad_ifa = ifa;
1304 	ifa_ref(dp->dad_ifa);
1305 	dp->dad_count = V_ip6_dad_count;
1306 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1307 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1308 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1309 
1310 	/* Add this to the dadq and add a reference for the dadq. */
1311 	refcount_init(&dp->dad_refcnt, 1);
1312 	nd6_dad_add(dp);
1313 	NET_EPOCH_ENTER(et);
1314 	nd6_dad_starttimer(dp, delay, 0);
1315 	NET_EPOCH_EXIT(et);
1316 }
1317 
1318 /*
1319  * terminate DAD unconditionally.  used for address removals.
1320  */
1321 void
1322 nd6_dad_stop(struct ifaddr *ifa)
1323 {
1324 	struct dadq *dp;
1325 
1326 	dp = nd6_dad_find(ifa, NULL);
1327 	if (!dp) {
1328 		/* DAD wasn't started yet */
1329 		return;
1330 	}
1331 
1332 	nd6_dad_stoptimer(dp);
1333 	nd6_dad_del(dp);
1334 
1335 	/* Release this function's reference, acquired by nd6_dad_find(). */
1336 	nd6_dad_rele(dp);
1337 }
1338 
1339 static void
1340 nd6_dad_timer(struct dadq *dp)
1341 {
1342 	CURVNET_SET(dp->dad_vnet);
1343 	struct ifaddr *ifa = dp->dad_ifa;
1344 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1345 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1346 	char ip6buf[INET6_ADDRSTRLEN];
1347 	struct epoch_tracker et;
1348 
1349 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1350 
1351 	NET_EPOCH_ENTER(et);
1352 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1353 		/* Do not need DAD for ifdisabled interface. */
1354 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1355 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1356 		goto err;
1357 	}
1358 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1359 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1360 			"%s(%s)\n",
1361 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1362 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1363 		goto err;
1364 	}
1365 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1366 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1367 			"%s(%s)\n",
1368 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1369 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1370 		goto err;
1371 	}
1372 
1373 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1374 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1375 	    (((ifp->if_flags & IFF_UP) == 0) ||
1376 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1377 		nd6log((LOG_INFO, "%s: could not run DAD "
1378 		    "because the interface was down or not running.\n",
1379 		    if_name(ifa->ifa_ifp)));
1380 		goto err;
1381 	}
1382 
1383 	/* Need more checks? */
1384 	if (dp->dad_ns_ocount < dp->dad_count) {
1385 		/*
1386 		 * We have more NS to go.  Send NS packet for DAD.
1387 		 */
1388 		nd6_dad_starttimer(dp,
1389 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1390 		goto done;
1391 	} else {
1392 		/*
1393 		 * We have transmitted sufficient number of DAD packets.
1394 		 * See what we've got.
1395 		 */
1396 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1397 			/* We've seen NS or NA, means DAD has failed. */
1398 			nd6_dad_duplicated(ifa, dp);
1399 		else if (V_dad_enhanced != 0 &&
1400 		    dp->dad_ns_lcount > 0 &&
1401 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1402 			/*
1403 			 * Sec. 4.1 in RFC 7527 requires transmission of
1404 			 * additional probes until the loopback condition
1405 			 * becomes clear when a looped back probe is detected.
1406 			 */
1407 			log(LOG_ERR, "%s: a looped back NS message is "
1408 			    "detected during DAD for %s.  "
1409 			    "Another DAD probes are being sent.\n",
1410 			    if_name(ifa->ifa_ifp),
1411 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1412 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1413 			/*
1414 			 * Send an NS immediately and increase dad_count by
1415 			 * V_nd6_mmaxtries - 1.
1416 			 */
1417 			dp->dad_count =
1418 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1419 			nd6_dad_starttimer(dp,
1420 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1421 			    1);
1422 			goto done;
1423 		} else {
1424 			/*
1425 			 * We are done with DAD.  No NA came, no NS came.
1426 			 * No duplicate address found.  Check IFDISABLED flag
1427 			 * again in case that it is changed between the
1428 			 * beginning of this function and here.
1429 			 */
1430 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1431 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1432 
1433 			nd6log((LOG_DEBUG,
1434 			    "%s: DAD complete for %s - no duplicates found\n",
1435 			    if_name(ifa->ifa_ifp),
1436 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1437 			if (dp->dad_ns_lcount > 0)
1438 				log(LOG_ERR, "%s: DAD completed while "
1439 				    "a looped back NS message is detected "
1440 				    "during DAD for %s.\n",
1441 				    if_name(ifa->ifa_ifp),
1442 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1443 		}
1444 	}
1445 err:
1446 	nd6_dad_del(dp);
1447 done:
1448 	NET_EPOCH_EXIT(et);
1449 	CURVNET_RESTORE();
1450 }
1451 
1452 static void
1453 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1454 {
1455 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1456 	struct ifnet *ifp;
1457 	char ip6buf[INET6_ADDRSTRLEN];
1458 
1459 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1460 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1461 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1462 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1463 	    dp->dad_na_icount);
1464 
1465 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1466 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1467 
1468 	ifp = ifa->ifa_ifp;
1469 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1470 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1471 	log(LOG_ERR, "%s: manual intervention required\n",
1472 	    if_name(ifp));
1473 
1474 	/*
1475 	 * If the address is a link-local address formed from an interface
1476 	 * identifier based on the hardware address which is supposed to be
1477 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1478 	 * operation on the interface SHOULD be disabled.
1479 	 * [RFC 4862, Section 5.4.5]
1480 	 */
1481 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1482 		struct in6_addr in6;
1483 
1484 		/*
1485 		 * To avoid over-reaction, we only apply this logic when we are
1486 		 * very sure that hardware addresses are supposed to be unique.
1487 		 */
1488 		switch (ifp->if_type) {
1489 		case IFT_ETHER:
1490 		case IFT_ATM:
1491 		case IFT_IEEE1394:
1492 		case IFT_INFINIBAND:
1493 			in6 = ia->ia_addr.sin6_addr;
1494 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1495 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1496 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1497 				log(LOG_ERR, "%s: possible hardware address "
1498 				    "duplication detected, disable IPv6\n",
1499 				    if_name(ifp));
1500 			}
1501 			break;
1502 		}
1503 	}
1504 }
1505 
1506 static void
1507 nd6_dad_ns_output(struct dadq *dp)
1508 {
1509 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1510 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1511 	int i;
1512 
1513 	dp->dad_ns_tcount++;
1514 	if ((ifp->if_flags & IFF_UP) == 0) {
1515 		return;
1516 	}
1517 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1518 		return;
1519 	}
1520 
1521 	dp->dad_ns_ocount++;
1522 	if (V_dad_enhanced != 0) {
1523 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1524 			dp->dad_nonce[i] = arc4random();
1525 		/*
1526 		 * XXXHRS: Note that in the case that
1527 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1528 		 * different nonces can be looped back in an unexpected
1529 		 * order.  The current implementation recognizes only
1530 		 * the latest nonce on the sender side.  Practically it
1531 		 * should work well in almost all cases.
1532 		 */
1533 	}
1534 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1535 	    (uint8_t *)&dp->dad_nonce[0]);
1536 }
1537 
1538 static void
1539 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1540 {
1541 	struct dadq *dp;
1542 
1543 	if (ifa == NULL)
1544 		panic("ifa == NULL in nd6_dad_ns_input");
1545 
1546 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1547 	if (V_dad_enhanced == 0)
1548 		ndopt_nonce = NULL;
1549 	dp = nd6_dad_find(ifa, ndopt_nonce);
1550 	if (dp == NULL)
1551 		return;
1552 
1553 	dp->dad_ns_icount++;
1554 	nd6_dad_rele(dp);
1555 }
1556 
1557 static void
1558 nd6_dad_na_input(struct ifaddr *ifa)
1559 {
1560 	struct dadq *dp;
1561 
1562 	if (ifa == NULL)
1563 		panic("ifa == NULL in nd6_dad_na_input");
1564 
1565 	dp = nd6_dad_find(ifa, NULL);
1566 	if (dp != NULL) {
1567 		dp->dad_na_icount++;
1568 		nd6_dad_rele(dp);
1569 	}
1570 }
1571