xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 1f4bcc459a76b7aa664f3fd557684cd0ba6da352)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_mpath.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/malloc.h>
43 #include <sys/libkern.h>
44 #include <sys/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55 #include <sys/callout.h>
56 #include <sys/refcount.h>
57 
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/if_var.h>
62 #include <net/route.h>
63 #ifdef RADIX_MPATH
64 #include <net/radix_mpath.h>
65 #endif
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <net/if_llatbl.h>
71 #include <netinet6/in6_var.h>
72 #include <netinet6/in6_ifattach.h>
73 #include <netinet/ip6.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/scope6_var.h>
76 #include <netinet6/nd6.h>
77 #include <netinet/icmp6.h>
78 #include <netinet/ip_carp.h>
79 #include <netinet6/send.h>
80 
81 #define SDL(s) ((struct sockaddr_dl *)s)
82 
83 struct dadq;
84 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
85 static void nd6_dad_add(struct dadq *dp);
86 static void nd6_dad_del(struct dadq *dp);
87 static void nd6_dad_rele(struct dadq *);
88 static void nd6_dad_starttimer(struct dadq *, int, int);
89 static void nd6_dad_stoptimer(struct dadq *);
90 static void nd6_dad_timer(struct dadq *);
91 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
92 static void nd6_dad_ns_output(struct dadq *);
93 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
94 static void nd6_dad_na_input(struct ifaddr *);
95 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
96     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
97 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
98     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
99 
100 static VNET_DEFINE(int, dad_enhanced) = 1;
101 #define	V_dad_enhanced			VNET(dad_enhanced)
102 
103 SYSCTL_DECL(_net_inet6_ip6);
104 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
105     &VNET_NAME(dad_enhanced), 0,
106     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
107 
108 static VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to
109 						   transmit DAD packet */
110 #define	V_dad_maxtry			VNET(dad_maxtry)
111 
112 /*
113  * Input a Neighbor Solicitation Message.
114  *
115  * Based on RFC 2461
116  * Based on RFC 2462 (duplicate address detection)
117  */
118 void
119 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
120 {
121 	struct ifnet *ifp = m->m_pkthdr.rcvif;
122 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
123 	struct nd_neighbor_solicit *nd_ns;
124 	struct in6_addr saddr6 = ip6->ip6_src;
125 	struct in6_addr daddr6 = ip6->ip6_dst;
126 	struct in6_addr taddr6;
127 	struct in6_addr myaddr6;
128 	char *lladdr = NULL;
129 	struct ifaddr *ifa = NULL;
130 	int lladdrlen = 0;
131 	int anycast = 0, proxy = 0, tentative = 0;
132 	int tlladdr;
133 	int rflag;
134 	union nd_opts ndopts;
135 	struct sockaddr_dl proxydl;
136 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
137 
138 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
139 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
140 		rflag = 0;
141 #ifndef PULLDOWN_TEST
142 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
143 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
144 #else
145 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
146 	if (nd_ns == NULL) {
147 		ICMP6STAT_INC(icp6s_tooshort);
148 		return;
149 	}
150 #endif
151 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
152 	taddr6 = nd_ns->nd_ns_target;
153 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
154 		goto bad;
155 
156 	if (ip6->ip6_hlim != 255) {
157 		nd6log((LOG_ERR,
158 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
159 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
160 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
161 		goto bad;
162 	}
163 
164 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
165 		/* dst has to be a solicited node multicast address. */
166 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
167 		    /* don't check ifindex portion */
168 		    daddr6.s6_addr32[1] == 0 &&
169 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
170 		    daddr6.s6_addr8[12] == 0xff) {
171 			; /* good */
172 		} else {
173 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
174 			    "(wrong ip6 dst)\n"));
175 			goto bad;
176 		}
177 	} else if (!V_nd6_onlink_ns_rfc4861) {
178 		struct sockaddr_in6 src_sa6;
179 
180 		/*
181 		 * According to recent IETF discussions, it is not a good idea
182 		 * to accept a NS from an address which would not be deemed
183 		 * to be a neighbor otherwise.  This point is expected to be
184 		 * clarified in future revisions of the specification.
185 		 */
186 		bzero(&src_sa6, sizeof(src_sa6));
187 		src_sa6.sin6_family = AF_INET6;
188 		src_sa6.sin6_len = sizeof(src_sa6);
189 		src_sa6.sin6_addr = saddr6;
190 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
191 			nd6log((LOG_INFO, "nd6_ns_input: "
192 				"NS packet from non-neighbor\n"));
193 			goto bad;
194 		}
195 	}
196 
197 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
198 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
199 		goto bad;
200 	}
201 
202 	icmp6len -= sizeof(*nd_ns);
203 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
204 	if (nd6_options(&ndopts) < 0) {
205 		nd6log((LOG_INFO,
206 		    "nd6_ns_input: invalid ND option, ignored\n"));
207 		/* nd6_options have incremented stats */
208 		goto freeit;
209 	}
210 
211 	if (ndopts.nd_opts_src_lladdr) {
212 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
213 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
214 	}
215 
216 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
217 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
218 		    "(link-layer address option)\n"));
219 		goto bad;
220 	}
221 
222 	/*
223 	 * Attaching target link-layer address to the NA?
224 	 * (RFC 2461 7.2.4)
225 	 *
226 	 * NS IP dst is unicast/anycast			MUST NOT add
227 	 * NS IP dst is solicited-node multicast	MUST add
228 	 *
229 	 * In implementation, we add target link-layer address by default.
230 	 * We do not add one in MUST NOT cases.
231 	 */
232 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
233 		tlladdr = 0;
234 	else
235 		tlladdr = 1;
236 
237 	/*
238 	 * Target address (taddr6) must be either:
239 	 * (1) Valid unicast/anycast address for my receiving interface,
240 	 * (2) Unicast address for which I'm offering proxy service, or
241 	 * (3) "tentative" address on which DAD is being performed.
242 	 */
243 	/* (1) and (3) check. */
244 	if (ifp->if_carp)
245 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
246 	else
247 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
248 
249 	/* (2) check. */
250 	if (ifa == NULL) {
251 		struct sockaddr_dl rt_gateway;
252 		struct rt_addrinfo info;
253 		struct sockaddr_in6 dst6;
254 
255 		bzero(&dst6, sizeof(dst6));
256 		dst6.sin6_len = sizeof(struct sockaddr_in6);
257 		dst6.sin6_family = AF_INET6;
258 		dst6.sin6_addr = taddr6;
259 
260 		bzero(&rt_gateway, sizeof(rt_gateway));
261 		rt_gateway.sdl_len = sizeof(rt_gateway);
262 		bzero(&info, sizeof(info));
263 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
264 
265 		/* Always use the default FIB. */
266 		if (rib_lookup_info(RT_DEFAULT_FIB, (struct sockaddr *)&dst6,
267 		    0, 0, &info) == 0) {
268 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
269 			    rt_gateway.sdl_family == AF_LINK) {
270 
271 				/*
272 				 * proxy NDP for single entry
273 				 */
274 				proxydl = *SDL(&rt_gateway);
275 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
276 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
277 				if (ifa)
278 					proxy = 1;
279 			}
280 		}
281 	}
282 	if (ifa == NULL) {
283 		/*
284 		 * We've got an NS packet, and we don't have that adddress
285 		 * assigned for us.  We MUST silently ignore it.
286 		 * See RFC2461 7.2.3.
287 		 */
288 		goto freeit;
289 	}
290 	myaddr6 = *IFA_IN6(ifa);
291 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
292 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
293 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
294 		goto freeit;
295 
296 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
297 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
298 		    "(if %d, NS packet %d)\n",
299 		    ip6_sprintf(ip6bufs, &taddr6),
300 		    ifp->if_addrlen, lladdrlen - 2));
301 		goto bad;
302 	}
303 
304 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
305 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
306 		    ip6_sprintf(ip6bufs, &saddr6)));
307 		goto freeit;
308 	}
309 
310 	/*
311 	 * We have neighbor solicitation packet, with target address equals to
312 	 * one of my tentative address.
313 	 *
314 	 * src addr	how to process?
315 	 * ---		---
316 	 * multicast	of course, invalid (rejected in ip6_input)
317 	 * unicast	somebody is doing address resolution -> ignore
318 	 * unspec	dup address detection
319 	 *
320 	 * The processing is defined in RFC 2462.
321 	 */
322 	if (tentative) {
323 		/*
324 		 * If source address is unspecified address, it is for
325 		 * duplicate address detection.
326 		 *
327 		 * If not, the packet is for addess resolution;
328 		 * silently ignore it.
329 		 */
330 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
331 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
332 
333 		goto freeit;
334 	}
335 
336 	/*
337 	 * If the source address is unspecified address, entries must not
338 	 * be created or updated.
339 	 * It looks that sender is performing DAD.  Output NA toward
340 	 * all-node multicast address, to tell the sender that I'm using
341 	 * the address.
342 	 * S bit ("solicited") must be zero.
343 	 */
344 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
345 		struct in6_addr in6_all;
346 
347 		in6_all = in6addr_linklocal_allnodes;
348 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
349 			goto bad;
350 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
351 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
352 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
353 		    M_GETFIB(m));
354 		goto freeit;
355 	}
356 
357 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
358 	    ND_NEIGHBOR_SOLICIT, 0);
359 
360 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
361 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
362 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
363 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
364  freeit:
365 	if (ifa != NULL)
366 		ifa_free(ifa);
367 	m_freem(m);
368 	return;
369 
370  bad:
371 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
372 		ip6_sprintf(ip6bufs, &saddr6)));
373 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
374 		ip6_sprintf(ip6bufs, &daddr6)));
375 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
376 		ip6_sprintf(ip6bufs, &taddr6)));
377 	ICMP6STAT_INC(icp6s_badns);
378 	if (ifa != NULL)
379 		ifa_free(ifa);
380 	m_freem(m);
381 }
382 
383 /*
384  * Output a Neighbor Solicitation Message. Caller specifies:
385  *	- ICMP6 header source IP6 address
386  *	- ND6 header target IP6 address
387  *	- ND6 header source datalink address
388  *
389  * Based on RFC 2461
390  * Based on RFC 2462 (duplicate address detection)
391  *
392  *    ln - for source address determination
393  * nonce - If non-NULL, NS is used for duplicate address detection and
394  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
395  */
396 static void
397 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
398     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
399     uint8_t *nonce, u_int fibnum)
400 {
401 	struct mbuf *m;
402 	struct m_tag *mtag;
403 	struct ip6_hdr *ip6;
404 	struct nd_neighbor_solicit *nd_ns;
405 	struct ip6_moptions im6o;
406 	int icmp6len;
407 	int maxlen;
408 	caddr_t mac;
409 
410 	if (IN6_IS_ADDR_MULTICAST(taddr6))
411 		return;
412 
413 	/* estimate the size of message */
414 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
415 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
416 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
417 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
418 	    __func__, max_linkhdr, maxlen, MCLBYTES));
419 
420 	if (max_linkhdr + maxlen > MHLEN)
421 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
422 	else
423 		m = m_gethdr(M_NOWAIT, MT_DATA);
424 	if (m == NULL)
425 		return;
426 	M_SETFIB(m, fibnum);
427 
428 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
429 		m->m_flags |= M_MCAST;
430 		im6o.im6o_multicast_ifp = ifp;
431 		im6o.im6o_multicast_hlim = 255;
432 		im6o.im6o_multicast_loop = 0;
433 	}
434 
435 	icmp6len = sizeof(*nd_ns);
436 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
437 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
438 
439 	/* fill neighbor solicitation packet */
440 	ip6 = mtod(m, struct ip6_hdr *);
441 	ip6->ip6_flow = 0;
442 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
443 	ip6->ip6_vfc |= IPV6_VERSION;
444 	/* ip6->ip6_plen will be set later */
445 	ip6->ip6_nxt = IPPROTO_ICMPV6;
446 	ip6->ip6_hlim = 255;
447 	if (daddr6)
448 		ip6->ip6_dst = *daddr6;
449 	else {
450 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
451 		ip6->ip6_dst.s6_addr16[1] = 0;
452 		ip6->ip6_dst.s6_addr32[1] = 0;
453 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
454 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
455 		ip6->ip6_dst.s6_addr8[12] = 0xff;
456 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
457 			goto bad;
458 	}
459 	if (nonce == NULL) {
460 		struct ifaddr *ifa = NULL;
461 
462 		/*
463 		 * RFC2461 7.2.2:
464 		 * "If the source address of the packet prompting the
465 		 * solicitation is the same as one of the addresses assigned
466 		 * to the outgoing interface, that address SHOULD be placed
467 		 * in the IP Source Address of the outgoing solicitation.
468 		 * Otherwise, any one of the addresses assigned to the
469 		 * interface should be used."
470 		 *
471 		 * We use the source address for the prompting packet
472 		 * (saddr6), if saddr6 belongs to the outgoing interface.
473 		 * Otherwise, we perform the source address selection as usual.
474 		 */
475 
476 		if (saddr6 != NULL)
477 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
478 		if (ifa != NULL) {
479 			/* ip6_src set already. */
480 			ip6->ip6_src = *saddr6;
481 			ifa_free(ifa);
482 		} else {
483 			int error;
484 			struct in6_addr dst6, src6;
485 			uint32_t scopeid;
486 
487 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
488 			error = in6_selectsrc_addr(RT_DEFAULT_FIB, &dst6,
489 			    scopeid, ifp, &src6, NULL);
490 			if (error) {
491 				char ip6buf[INET6_ADDRSTRLEN];
492 				nd6log((LOG_DEBUG, "%s: source can't be "
493 				    "determined: dst=%s, error=%d\n", __func__,
494 				    ip6_sprintf(ip6buf, &dst6),
495 				    error));
496 				goto bad;
497 			}
498 			ip6->ip6_src = src6;
499 		}
500 	} else {
501 		/*
502 		 * Source address for DAD packet must always be IPv6
503 		 * unspecified address. (0::0)
504 		 * We actually don't have to 0-clear the address (we did it
505 		 * above), but we do so here explicitly to make the intention
506 		 * clearer.
507 		 */
508 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
509 	}
510 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
511 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
512 	nd_ns->nd_ns_code = 0;
513 	nd_ns->nd_ns_reserved = 0;
514 	nd_ns->nd_ns_target = *taddr6;
515 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
516 
517 	/*
518 	 * Add source link-layer address option.
519 	 *
520 	 *				spec		implementation
521 	 *				---		---
522 	 * DAD packet			MUST NOT	do not add the option
523 	 * there's no link layer address:
524 	 *				impossible	do not add the option
525 	 * there's link layer address:
526 	 *	Multicast NS		MUST add one	add the option
527 	 *	Unicast NS		SHOULD add one	add the option
528 	 */
529 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
530 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
531 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
532 		/* 8 byte alignments... */
533 		optlen = (optlen + 7) & ~7;
534 
535 		m->m_pkthdr.len += optlen;
536 		m->m_len += optlen;
537 		icmp6len += optlen;
538 		bzero((caddr_t)nd_opt, optlen);
539 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
540 		nd_opt->nd_opt_len = optlen >> 3;
541 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
542 	}
543 	/*
544 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
545 	 * This behavior is documented as Enhanced Duplicate Address
546 	 * Detection in RFC 7527.
547 	 * net.inet6.ip6.dad_enhanced=0 disables this.
548 	 */
549 	if (V_dad_enhanced != 0 && nonce != NULL) {
550 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
551 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
552 		/* 8-byte alignment is required. */
553 		optlen = (optlen + 7) & ~7;
554 
555 		m->m_pkthdr.len += optlen;
556 		m->m_len += optlen;
557 		icmp6len += optlen;
558 		bzero((caddr_t)nd_opt, optlen);
559 		nd_opt->nd_opt_type = ND_OPT_NONCE;
560 		nd_opt->nd_opt_len = optlen >> 3;
561 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
562 	}
563 	ip6->ip6_plen = htons((u_short)icmp6len);
564 	nd_ns->nd_ns_cksum = 0;
565 	nd_ns->nd_ns_cksum =
566 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
567 
568 	if (send_sendso_input_hook != NULL) {
569 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
570 			sizeof(unsigned short), M_NOWAIT);
571 		if (mtag == NULL)
572 			goto bad;
573 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
574 		m_tag_prepend(m, mtag);
575 	}
576 
577 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
578 	    &im6o, NULL, NULL);
579 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
580 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
581 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
582 
583 	return;
584 
585   bad:
586 	m_freem(m);
587 	return;
588 }
589 
590 #ifndef BURN_BRIDGES
591 void
592 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
593     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
594 {
595 
596 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
597 }
598 #endif
599 /*
600  * Neighbor advertisement input handling.
601  *
602  * Based on RFC 2461
603  * Based on RFC 2462 (duplicate address detection)
604  *
605  * the following items are not implemented yet:
606  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
607  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
608  */
609 void
610 nd6_na_input(struct mbuf *m, int off, int icmp6len)
611 {
612 	struct ifnet *ifp = m->m_pkthdr.rcvif;
613 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
614 	struct nd_neighbor_advert *nd_na;
615 	struct in6_addr daddr6 = ip6->ip6_dst;
616 	struct in6_addr taddr6;
617 	int flags;
618 	int is_router;
619 	int is_solicited;
620 	int is_override;
621 	char *lladdr = NULL;
622 	int lladdrlen = 0;
623 	int checklink = 0;
624 	struct ifaddr *ifa;
625 	struct llentry *ln = NULL;
626 	union nd_opts ndopts;
627 	struct mbuf *chain = NULL;
628 	struct sockaddr_in6 sin6;
629 	u_char linkhdr[LLE_MAX_LINKHDR];
630 	size_t linkhdrsize;
631 	int lladdr_off;
632 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
633 
634 	if (ip6->ip6_hlim != 255) {
635 		nd6log((LOG_ERR,
636 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
637 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
638 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
639 		goto bad;
640 	}
641 
642 #ifndef PULLDOWN_TEST
643 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
644 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
645 #else
646 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
647 	if (nd_na == NULL) {
648 		ICMP6STAT_INC(icp6s_tooshort);
649 		return;
650 	}
651 #endif
652 
653 	flags = nd_na->nd_na_flags_reserved;
654 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
655 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
656 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
657 	memset(&sin6, 0, sizeof(sin6));
658 
659 	taddr6 = nd_na->nd_na_target;
660 	if (in6_setscope(&taddr6, ifp, NULL))
661 		goto bad;	/* XXX: impossible */
662 
663 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
664 		nd6log((LOG_ERR,
665 		    "nd6_na_input: invalid target address %s\n",
666 		    ip6_sprintf(ip6bufs, &taddr6)));
667 		goto bad;
668 	}
669 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
670 		if (is_solicited) {
671 			nd6log((LOG_ERR,
672 			    "nd6_na_input: a solicited adv is multicasted\n"));
673 			goto bad;
674 		}
675 
676 	icmp6len -= sizeof(*nd_na);
677 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
678 	if (nd6_options(&ndopts) < 0) {
679 		nd6log((LOG_INFO,
680 		    "nd6_na_input: invalid ND option, ignored\n"));
681 		/* nd6_options have incremented stats */
682 		goto freeit;
683 	}
684 
685 	if (ndopts.nd_opts_tgt_lladdr) {
686 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
687 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
688 	}
689 
690 	/*
691 	 * This effectively disables the DAD check on a non-master CARP
692 	 * address.
693 	 */
694 	if (ifp->if_carp)
695 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
696 	else
697 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
698 
699 	/*
700 	 * Target address matches one of my interface address.
701 	 *
702 	 * If my address is tentative, this means that there's somebody
703 	 * already using the same address as mine.  This indicates DAD failure.
704 	 * This is defined in RFC 2462.
705 	 *
706 	 * Otherwise, process as defined in RFC 2461.
707 	 */
708 	if (ifa
709 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
710 		nd6_dad_na_input(ifa);
711 		ifa_free(ifa);
712 		goto freeit;
713 	}
714 
715 	/* Just for safety, maybe unnecessary. */
716 	if (ifa) {
717 		ifa_free(ifa);
718 		log(LOG_ERR,
719 		    "nd6_na_input: duplicate IP6 address %s\n",
720 		    ip6_sprintf(ip6bufs, &taddr6));
721 		goto freeit;
722 	}
723 
724 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
725 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
726 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
727 		    ifp->if_addrlen, lladdrlen - 2));
728 		goto bad;
729 	}
730 
731 	/*
732 	 * If no neighbor cache entry is found, NA SHOULD silently be
733 	 * discarded.
734 	 */
735 	IF_AFDATA_RLOCK(ifp);
736 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
737 	IF_AFDATA_RUNLOCK(ifp);
738 	if (ln == NULL) {
739 		goto freeit;
740 	}
741 
742 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
743 		/*
744 		 * If the link-layer has address, and no lladdr option came,
745 		 * discard the packet.
746 		 */
747 		if (ifp->if_addrlen && lladdr == NULL) {
748 			goto freeit;
749 		}
750 
751 		/*
752 		 * Record link-layer address, and update the state.
753 		 */
754 		linkhdrsize = sizeof(linkhdr);
755 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
756 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
757 			return;
758 
759 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
760 		    lladdr_off) == 0) {
761 			ln = NULL;
762 			goto freeit;
763 		}
764 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
765 		if (is_solicited)
766 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
767 		else
768 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
769 		if ((ln->ln_router = is_router) != 0) {
770 			/*
771 			 * This means a router's state has changed from
772 			 * non-reachable to probably reachable, and might
773 			 * affect the status of associated prefixes..
774 			 */
775 			checklink = 1;
776 		}
777 	} else {
778 		int llchange;
779 
780 		/*
781 		 * Check if the link-layer address has changed or not.
782 		 */
783 		if (lladdr == NULL)
784 			llchange = 0;
785 		else {
786 			if (ln->la_flags & LLE_VALID) {
787 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
788 					llchange = 1;
789 				else
790 					llchange = 0;
791 			} else
792 				llchange = 1;
793 		}
794 
795 		/*
796 		 * This is VERY complex.  Look at it with care.
797 		 *
798 		 * override solicit lladdr llchange	action
799 		 *					(L: record lladdr)
800 		 *
801 		 *	0	0	n	--	(2c)
802 		 *	0	0	y	n	(2b) L
803 		 *	0	0	y	y	(1)    REACHABLE->STALE
804 		 *	0	1	n	--	(2c)   *->REACHABLE
805 		 *	0	1	y	n	(2b) L *->REACHABLE
806 		 *	0	1	y	y	(1)    REACHABLE->STALE
807 		 *	1	0	n	--	(2a)
808 		 *	1	0	y	n	(2a) L
809 		 *	1	0	y	y	(2a) L *->STALE
810 		 *	1	1	n	--	(2a)   *->REACHABLE
811 		 *	1	1	y	n	(2a) L *->REACHABLE
812 		 *	1	1	y	y	(2a) L *->REACHABLE
813 		 */
814 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
815 			/*
816 			 * If state is REACHABLE, make it STALE.
817 			 * no other updates should be done.
818 			 */
819 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
820 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
821 			goto freeit;
822 		} else if (is_override				   /* (2a) */
823 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
824 			|| lladdr == NULL) {			   /* (2c) */
825 			/*
826 			 * Update link-local address, if any.
827 			 */
828 			if (lladdr != NULL) {
829 				linkhdrsize = sizeof(linkhdr);
830 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
831 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
832 					goto freeit;
833 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
834 				    linkhdrsize, lladdr_off) == 0) {
835 					ln = NULL;
836 					goto freeit;
837 				}
838 				EVENTHANDLER_INVOKE(lle_event, ln,
839 				    LLENTRY_RESOLVED);
840 			}
841 
842 			/*
843 			 * If solicited, make the state REACHABLE.
844 			 * If not solicited and the link-layer address was
845 			 * changed, make it STALE.
846 			 */
847 			if (is_solicited)
848 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
849 			else {
850 				if (lladdr != NULL && llchange)
851 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
852 			}
853 		}
854 
855 		if (ln->ln_router && !is_router) {
856 			/*
857 			 * The peer dropped the router flag.
858 			 * Remove the sender from the Default Router List and
859 			 * update the Destination Cache entries.
860 			 */
861 			struct nd_defrouter *dr;
862 			struct in6_addr *in6;
863 			struct ifnet *nd6_ifp;
864 
865 			in6 = &ln->r_l3addr.addr6;
866 
867 			/*
868 			 * Lock to protect the default router list.
869 			 * XXX: this might be unnecessary, since this function
870 			 * is only called under the network software interrupt
871 			 * context.  However, we keep it just for safety.
872 			 */
873 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
874 			dr = defrouter_lookup(in6, nd6_ifp);
875 			if (dr)
876 				defrtrlist_del(dr);
877 			else if (ND_IFINFO(nd6_ifp)->flags &
878 			    ND6_IFF_ACCEPT_RTADV) {
879 				/*
880 				 * Even if the neighbor is not in the default
881 				 * router list, the neighbor may be used
882 				 * as a next hop for some destinations
883 				 * (e.g. redirect case). So we must
884 				 * call rt6_flush explicitly.
885 				 */
886 				rt6_flush(&ip6->ip6_src, ifp);
887 			}
888 		}
889 		ln->ln_router = is_router;
890 	}
891         /* XXX - QL
892 	 *  Does this matter?
893 	 *  rt->rt_flags &= ~RTF_REJECT;
894 	 */
895 	ln->la_asked = 0;
896 	if (ln->la_hold != NULL)
897 		nd6_grab_holdchain(ln, &chain, &sin6);
898  freeit:
899 	if (ln != NULL)
900 		LLE_WUNLOCK(ln);
901 
902 	if (chain != NULL)
903 		nd6_flush_holdchain(ifp, ifp, chain, &sin6);
904 
905 	if (checklink)
906 		pfxlist_onlink_check();
907 
908 	m_freem(m);
909 	return;
910 
911  bad:
912 	if (ln != NULL)
913 		LLE_WUNLOCK(ln);
914 
915 	ICMP6STAT_INC(icp6s_badna);
916 	m_freem(m);
917 }
918 
919 /*
920  * Neighbor advertisement output handling.
921  *
922  * Based on RFC 2461
923  *
924  * the following items are not implemented yet:
925  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
926  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
927  *
928  * tlladdr - 1 if include target link-layer address
929  * sdl0 - sockaddr_dl (= proxy NA) or NULL
930  */
931 static void
932 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
933     const struct in6_addr *taddr6, u_long flags, int tlladdr,
934     struct sockaddr *sdl0, u_int fibnum)
935 {
936 	struct mbuf *m;
937 	struct m_tag *mtag;
938 	struct ip6_hdr *ip6;
939 	struct nd_neighbor_advert *nd_na;
940 	struct ip6_moptions im6o;
941 	struct in6_addr daddr6, dst6, src6;
942 	uint32_t scopeid;
943 
944 	int icmp6len, maxlen, error;
945 	caddr_t mac = NULL;
946 
947 	daddr6 = *daddr6_0;	/* make a local copy for modification */
948 
949 	/* estimate the size of message */
950 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
951 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
952 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
953 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
954 	    __func__, max_linkhdr, maxlen, MCLBYTES));
955 
956 	if (max_linkhdr + maxlen > MHLEN)
957 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
958 	else
959 		m = m_gethdr(M_NOWAIT, MT_DATA);
960 	if (m == NULL)
961 		return;
962 	M_SETFIB(m, fibnum);
963 
964 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
965 		m->m_flags |= M_MCAST;
966 		im6o.im6o_multicast_ifp = ifp;
967 		im6o.im6o_multicast_hlim = 255;
968 		im6o.im6o_multicast_loop = 0;
969 	}
970 
971 	icmp6len = sizeof(*nd_na);
972 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
973 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
974 
975 	/* fill neighbor advertisement packet */
976 	ip6 = mtod(m, struct ip6_hdr *);
977 	ip6->ip6_flow = 0;
978 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
979 	ip6->ip6_vfc |= IPV6_VERSION;
980 	ip6->ip6_nxt = IPPROTO_ICMPV6;
981 	ip6->ip6_hlim = 255;
982 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
983 		/* reply to DAD */
984 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
985 		daddr6.s6_addr16[1] = 0;
986 		daddr6.s6_addr32[1] = 0;
987 		daddr6.s6_addr32[2] = 0;
988 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
989 		if (in6_setscope(&daddr6, ifp, NULL))
990 			goto bad;
991 
992 		flags &= ~ND_NA_FLAG_SOLICITED;
993 	}
994 	ip6->ip6_dst = daddr6;
995 
996 	/*
997 	 * Select a source whose scope is the same as that of the dest.
998 	 */
999 	in6_splitscope(&daddr6, &dst6, &scopeid);
1000 	error = in6_selectsrc_addr(RT_DEFAULT_FIB, &dst6,
1001 	    scopeid, ifp, &src6, NULL);
1002 	if (error) {
1003 		char ip6buf[INET6_ADDRSTRLEN];
1004 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1005 		    "determined: dst=%s, error=%d\n",
1006 		    ip6_sprintf(ip6buf, &daddr6), error));
1007 		goto bad;
1008 	}
1009 	ip6->ip6_src = src6;
1010 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1011 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1012 	nd_na->nd_na_code = 0;
1013 	nd_na->nd_na_target = *taddr6;
1014 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1015 
1016 	/*
1017 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1018 	 * see nd6_ns_input() for details.
1019 	 * Basically, if NS packet is sent to unicast/anycast addr,
1020 	 * target lladdr option SHOULD NOT be included.
1021 	 */
1022 	if (tlladdr) {
1023 		/*
1024 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1025 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1026 		 * my address) use lladdr configured for the interface.
1027 		 */
1028 		if (sdl0 == NULL) {
1029 			if (ifp->if_carp)
1030 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1031 			if (mac == NULL)
1032 				mac = nd6_ifptomac(ifp);
1033 		} else if (sdl0->sa_family == AF_LINK) {
1034 			struct sockaddr_dl *sdl;
1035 			sdl = (struct sockaddr_dl *)sdl0;
1036 			if (sdl->sdl_alen == ifp->if_addrlen)
1037 				mac = LLADDR(sdl);
1038 		}
1039 	}
1040 	if (tlladdr && mac) {
1041 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1042 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1043 
1044 		/* roundup to 8 bytes alignment! */
1045 		optlen = (optlen + 7) & ~7;
1046 
1047 		m->m_pkthdr.len += optlen;
1048 		m->m_len += optlen;
1049 		icmp6len += optlen;
1050 		bzero((caddr_t)nd_opt, optlen);
1051 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1052 		nd_opt->nd_opt_len = optlen >> 3;
1053 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1054 	} else
1055 		flags &= ~ND_NA_FLAG_OVERRIDE;
1056 
1057 	ip6->ip6_plen = htons((u_short)icmp6len);
1058 	nd_na->nd_na_flags_reserved = flags;
1059 	nd_na->nd_na_cksum = 0;
1060 	nd_na->nd_na_cksum =
1061 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1062 
1063 	if (send_sendso_input_hook != NULL) {
1064 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1065 		    sizeof(unsigned short), M_NOWAIT);
1066 		if (mtag == NULL)
1067 			goto bad;
1068 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1069 		m_tag_prepend(m, mtag);
1070 	}
1071 
1072 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1073 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1074 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1075 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1076 
1077 	return;
1078 
1079   bad:
1080 	m_freem(m);
1081 	return;
1082 }
1083 
1084 #ifndef BURN_BRIDGES
1085 void
1086 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1087     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1088     struct sockaddr *sdl0)
1089 {
1090 
1091 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1092 	    RT_DEFAULT_FIB);
1093 }
1094 #endif
1095 
1096 caddr_t
1097 nd6_ifptomac(struct ifnet *ifp)
1098 {
1099 	switch (ifp->if_type) {
1100 	case IFT_ARCNET:
1101 	case IFT_ETHER:
1102 	case IFT_FDDI:
1103 	case IFT_IEEE1394:
1104 	case IFT_L2VLAN:
1105 	case IFT_IEEE80211:
1106 	case IFT_INFINIBAND:
1107 	case IFT_BRIDGE:
1108 	case IFT_ISO88025:
1109 		return IF_LLADDR(ifp);
1110 	default:
1111 		return NULL;
1112 	}
1113 }
1114 
1115 struct dadq {
1116 	TAILQ_ENTRY(dadq) dad_list;
1117 	struct ifaddr *dad_ifa;
1118 	int dad_count;		/* max NS to send */
1119 	int dad_ns_tcount;	/* # of trials to send NS */
1120 	int dad_ns_ocount;	/* NS sent so far */
1121 	int dad_ns_icount;
1122 	int dad_na_icount;
1123 	int dad_ns_lcount;	/* looped back NS */
1124 	int dad_loopbackprobe;	/* probing state for loopback detection */
1125 	struct callout dad_timer_ch;
1126 	struct vnet *dad_vnet;
1127 	u_int dad_refcnt;
1128 #define	ND_OPT_NONCE_LEN32 \
1129 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1130 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1131 };
1132 
1133 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1134 static VNET_DEFINE(struct rwlock, dad_rwlock);
1135 #define	V_dadq			VNET(dadq)
1136 #define	V_dad_rwlock		VNET(dad_rwlock)
1137 
1138 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1139 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1140 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1141 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1142 
1143 static void
1144 nd6_dad_add(struct dadq *dp)
1145 {
1146 
1147 	DADQ_WLOCK();
1148 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1149 	DADQ_WUNLOCK();
1150 }
1151 
1152 static void
1153 nd6_dad_del(struct dadq *dp)
1154 {
1155 
1156 	DADQ_WLOCK();
1157 	TAILQ_REMOVE(&V_dadq, dp, dad_list);
1158 	DADQ_WUNLOCK();
1159 	nd6_dad_rele(dp);
1160 }
1161 
1162 static struct dadq *
1163 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1164 {
1165 	struct dadq *dp;
1166 
1167 	DADQ_RLOCK();
1168 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1169 		if (dp->dad_ifa != ifa)
1170 			continue;
1171 		/*
1172 		 * Skip if the nonce matches the received one.
1173 		 * +2 in the length is required because of type and
1174 		 * length fields are included in a header.
1175 		 */
1176 		if (n != NULL &&
1177 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1178 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1179 		        ND_OPT_NONCE_LEN) == 0) {
1180 			dp->dad_ns_lcount++;
1181 			continue;
1182 		}
1183 		refcount_acquire(&dp->dad_refcnt);
1184 		break;
1185 	}
1186 	DADQ_RUNLOCK();
1187 
1188 	return (dp);
1189 }
1190 
1191 static void
1192 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1193 {
1194 
1195 	if (send_ns != 0)
1196 		nd6_dad_ns_output(dp);
1197 	callout_reset(&dp->dad_timer_ch, ticks,
1198 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1199 }
1200 
1201 static void
1202 nd6_dad_stoptimer(struct dadq *dp)
1203 {
1204 
1205 	callout_drain(&dp->dad_timer_ch);
1206 }
1207 
1208 static void
1209 nd6_dad_rele(struct dadq *dp)
1210 {
1211 
1212 	if (refcount_release(&dp->dad_refcnt)) {
1213 		ifa_free(dp->dad_ifa);
1214 		free(dp, M_IP6NDP);
1215 	}
1216 }
1217 
1218 void
1219 nd6_dad_init(void)
1220 {
1221 
1222 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1223 	TAILQ_INIT(&V_dadq);
1224 }
1225 
1226 /*
1227  * Start Duplicate Address Detection (DAD) for specified interface address.
1228  */
1229 void
1230 nd6_dad_start(struct ifaddr *ifa, int delay)
1231 {
1232 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1233 	struct dadq *dp;
1234 	char ip6buf[INET6_ADDRSTRLEN];
1235 	int send_ns;
1236 
1237 	/*
1238 	 * If we don't need DAD, don't do it.
1239 	 * There are several cases:
1240 	 * - DAD is disabled (ip6_dad_count == 0)
1241 	 * - the interface address is anycast
1242 	 */
1243 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1244 		log(LOG_DEBUG,
1245 			"nd6_dad_start: called with non-tentative address "
1246 			"%s(%s)\n",
1247 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1248 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1249 		return;
1250 	}
1251 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1252 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1253 		return;
1254 	}
1255 	if (!V_ip6_dad_count) {
1256 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1257 		return;
1258 	}
1259 	if (ifa->ifa_ifp == NULL)
1260 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1261 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) {
1262 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1263 		return;
1264 	}
1265 	if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1266 	    !(ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) ||
1267 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)) {
1268 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1269 		return;
1270 	}
1271 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1272 		/*
1273 		 * DAD already in progress.  Let the existing entry
1274 		 * to finish it.
1275 		 */
1276 		return;
1277 	}
1278 
1279 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1280 	if (dp == NULL) {
1281 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1282 			"%s(%s)\n",
1283 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1284 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1285 		return;
1286 	}
1287 	callout_init(&dp->dad_timer_ch, 0);
1288 #ifdef VIMAGE
1289 	dp->dad_vnet = curvnet;
1290 #endif
1291 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1292 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1293 
1294 	/*
1295 	 * Send NS packet for DAD, ip6_dad_count times.
1296 	 * Note that we must delay the first transmission, if this is the
1297 	 * first packet to be sent from the interface after interface
1298 	 * (re)initialization.
1299 	 */
1300 	dp->dad_ifa = ifa;
1301 	ifa_ref(dp->dad_ifa);
1302 	dp->dad_count = V_ip6_dad_count;
1303 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1304 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1305 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1306 	refcount_init(&dp->dad_refcnt, 1);
1307 	nd6_dad_add(dp);
1308 	send_ns = 0;
1309 	if (delay == 0) {
1310 		send_ns = 1;
1311 		delay = (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000;
1312 	}
1313 	nd6_dad_starttimer(dp, delay, send_ns);
1314 }
1315 
1316 /*
1317  * terminate DAD unconditionally.  used for address removals.
1318  */
1319 void
1320 nd6_dad_stop(struct ifaddr *ifa)
1321 {
1322 	struct dadq *dp;
1323 
1324 	dp = nd6_dad_find(ifa, NULL);
1325 	if (!dp) {
1326 		/* DAD wasn't started yet */
1327 		return;
1328 	}
1329 
1330 	nd6_dad_stoptimer(dp);
1331 
1332 	/*
1333 	 * The DAD queue entry may have been removed by nd6_dad_timer() while
1334 	 * we were waiting for it to stop, so re-do the lookup.
1335 	 */
1336 	nd6_dad_rele(dp);
1337 	if (nd6_dad_find(ifa, NULL) == NULL)
1338 		return;
1339 
1340 	nd6_dad_del(dp);
1341 	nd6_dad_rele(dp);
1342 }
1343 
1344 static void
1345 nd6_dad_timer(struct dadq *dp)
1346 {
1347 	CURVNET_SET(dp->dad_vnet);
1348 	struct ifaddr *ifa = dp->dad_ifa;
1349 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1350 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1351 	char ip6buf[INET6_ADDRSTRLEN];
1352 
1353 	/* Sanity check */
1354 	if (ia == NULL) {
1355 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1356 		goto err;
1357 	}
1358 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1359 		/* Do not need DAD for ifdisabled interface. */
1360 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1361 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1362 		goto err;
1363 	}
1364 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1365 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1366 			"%s(%s)\n",
1367 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1368 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1369 		goto err;
1370 	}
1371 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1372 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1373 			"%s(%s)\n",
1374 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1375 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1376 		goto err;
1377 	}
1378 
1379 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1380 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1381 	    (((ifp->if_flags & IFF_UP) == 0) ||
1382 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1383 		nd6log((LOG_INFO, "%s: could not run DAD "
1384 		    "because the interface was down or not running.\n",
1385 		    if_name(ifa->ifa_ifp)));
1386 		goto err;
1387 	}
1388 
1389 	/* Need more checks? */
1390 	if (dp->dad_ns_ocount < dp->dad_count) {
1391 		/*
1392 		 * We have more NS to go.  Send NS packet for DAD.
1393 		 */
1394 		nd6_dad_starttimer(dp,
1395 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1396 		goto done;
1397 	} else {
1398 		/*
1399 		 * We have transmitted sufficient number of DAD packets.
1400 		 * See what we've got.
1401 		 */
1402 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1403 			/* We've seen NS or NA, means DAD has failed. */
1404 			nd6_dad_duplicated(ifa, dp);
1405 		else if (V_dad_enhanced != 0 &&
1406 		    dp->dad_ns_lcount > 0 &&
1407 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1408 			/*
1409 			 * Sec. 4.1 in RFC 7527 requires transmission of
1410 			 * additional probes until the loopback condition
1411 			 * becomes clear when a looped back probe is detected.
1412 			 */
1413 			log(LOG_ERR, "%s: a looped back NS message is "
1414 			    "detected during DAD for %s.  "
1415 			    "Another DAD probes are being sent.\n",
1416 			    if_name(ifa->ifa_ifp),
1417 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1418 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1419 			/*
1420 			 * Send an NS immediately and increase dad_count by
1421 			 * V_nd6_mmaxtries - 1.
1422 			 */
1423 			dp->dad_count =
1424 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1425 			nd6_dad_starttimer(dp,
1426 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1427 			    1);
1428 			goto done;
1429 		} else {
1430 			/*
1431 			 * We are done with DAD.  No NA came, no NS came.
1432 			 * No duplicate address found.  Check IFDISABLED flag
1433 			 * again in case that it is changed between the
1434 			 * beginning of this function and here.
1435 			 */
1436 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1437 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1438 
1439 			nd6log((LOG_DEBUG,
1440 			    "%s: DAD complete for %s - no duplicates found\n",
1441 			    if_name(ifa->ifa_ifp),
1442 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1443 			if (dp->dad_ns_lcount > 0)
1444 				log(LOG_ERR, "%s: DAD completed while "
1445 				    "a looped back NS message is detected "
1446 				    "during DAD for %s.\n",
1447 				    if_name(ifa->ifa_ifp),
1448 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1449 		}
1450 	}
1451 err:
1452 	nd6_dad_del(dp);
1453 done:
1454 	CURVNET_RESTORE();
1455 }
1456 
1457 static void
1458 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1459 {
1460 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1461 	struct ifnet *ifp;
1462 	char ip6buf[INET6_ADDRSTRLEN];
1463 
1464 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1465 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1466 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1467 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1468 	    dp->dad_na_icount);
1469 
1470 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1471 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1472 
1473 	ifp = ifa->ifa_ifp;
1474 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1475 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1476 	log(LOG_ERR, "%s: manual intervention required\n",
1477 	    if_name(ifp));
1478 
1479 	/*
1480 	 * If the address is a link-local address formed from an interface
1481 	 * identifier based on the hardware address which is supposed to be
1482 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1483 	 * operation on the interface SHOULD be disabled.
1484 	 * [RFC 4862, Section 5.4.5]
1485 	 */
1486 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1487 		struct in6_addr in6;
1488 
1489 		/*
1490 		 * To avoid over-reaction, we only apply this logic when we are
1491 		 * very sure that hardware addresses are supposed to be unique.
1492 		 */
1493 		switch (ifp->if_type) {
1494 		case IFT_ETHER:
1495 		case IFT_FDDI:
1496 		case IFT_ATM:
1497 		case IFT_IEEE1394:
1498 		case IFT_IEEE80211:
1499 		case IFT_INFINIBAND:
1500 			in6 = ia->ia_addr.sin6_addr;
1501 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1502 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1503 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1504 				log(LOG_ERR, "%s: possible hardware address "
1505 				    "duplication detected, disable IPv6\n",
1506 				    if_name(ifp));
1507 			}
1508 			break;
1509 		}
1510 	}
1511 }
1512 
1513 static void
1514 nd6_dad_ns_output(struct dadq *dp)
1515 {
1516 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1517 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1518 	int i;
1519 
1520 	dp->dad_ns_tcount++;
1521 	if ((ifp->if_flags & IFF_UP) == 0) {
1522 		return;
1523 	}
1524 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1525 		return;
1526 	}
1527 
1528 	dp->dad_ns_ocount++;
1529 	if (V_dad_enhanced != 0) {
1530 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1531 			dp->dad_nonce[i] = arc4random();
1532 		/*
1533 		 * XXXHRS: Note that in the case that
1534 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1535 		 * different nonces can be looped back in an unexpected
1536 		 * order.  The current implementation recognizes only
1537 		 * the latest nonce on the sender side.  Practically it
1538 		 * should work well in almost all cases.
1539 		 */
1540 	}
1541 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1542 	    (uint8_t *)&dp->dad_nonce[0]);
1543 }
1544 
1545 static void
1546 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1547 {
1548 	struct in6_ifaddr *ia;
1549 	struct ifnet *ifp;
1550 	const struct in6_addr *taddr6;
1551 	struct dadq *dp;
1552 
1553 	if (ifa == NULL)
1554 		panic("ifa == NULL in nd6_dad_ns_input");
1555 
1556 	ia = (struct in6_ifaddr *)ifa;
1557 	ifp = ifa->ifa_ifp;
1558 	taddr6 = &ia->ia_addr.sin6_addr;
1559 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1560 	if (V_dad_enhanced == 0)
1561 		ndopt_nonce = NULL;
1562 	dp = nd6_dad_find(ifa, ndopt_nonce);
1563 	if (dp == NULL)
1564 		return;
1565 
1566 	dp->dad_ns_icount++;
1567 	nd6_dad_rele(dp);
1568 }
1569 
1570 static void
1571 nd6_dad_na_input(struct ifaddr *ifa)
1572 {
1573 	struct dadq *dp;
1574 
1575 	if (ifa == NULL)
1576 		panic("ifa == NULL in nd6_dad_na_input");
1577 
1578 	dp = nd6_dad_find(ifa, NULL);
1579 	if (dp != NULL) {
1580 		dp->dad_na_icount++;
1581 		nd6_dad_rele(dp);
1582 	}
1583 }
1584