xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 8a4217aacf57330755501a349d0ea662d4880386)
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/malloc.h>
45 #include <sys/libkern.h>
46 #include <sys/lock.h>
47 #include <sys/rwlock.h>
48 #include <sys/mbuf.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/errno.h>
54 #include <sys/sysctl.h>
55 #include <sys/syslog.h>
56 #include <sys/queue.h>
57 #include <sys/callout.h>
58 #include <sys/refcount.h>
59 
60 #include <net/if.h>
61 #include <net/if_types.h>
62 #include <net/if_dl.h>
63 #include <net/if_var.h>
64 #include <net/route.h>
65 #ifdef RADIX_MPATH
66 #include <net/radix_mpath.h>
67 #endif
68 #include <net/vnet.h>
69 
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
72 #include <net/if_llatbl.h>
73 #include <netinet6/in6_var.h>
74 #include <netinet6/in6_ifattach.h>
75 #include <netinet/ip6.h>
76 #include <netinet6/ip6_var.h>
77 #include <netinet6/scope6_var.h>
78 #include <netinet6/nd6.h>
79 #include <netinet/icmp6.h>
80 #include <netinet/ip_carp.h>
81 #include <netinet6/send.h>
82 
83 #define SDL(s) ((struct sockaddr_dl *)s)
84 
85 struct dadq;
86 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
87 static void nd6_dad_add(struct dadq *dp);
88 static void nd6_dad_del(struct dadq *dp);
89 static void nd6_dad_rele(struct dadq *);
90 static void nd6_dad_starttimer(struct dadq *, int, int);
91 static void nd6_dad_stoptimer(struct dadq *);
92 static void nd6_dad_timer(struct dadq *);
93 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
94 static void nd6_dad_ns_output(struct dadq *);
95 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
96 static void nd6_dad_na_input(struct ifaddr *);
97 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
98     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
99 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
100     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
101 
102 static VNET_DEFINE(int, dad_enhanced) = 1;
103 #define	V_dad_enhanced			VNET(dad_enhanced)
104 
105 SYSCTL_DECL(_net_inet6_ip6);
106 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
107     &VNET_NAME(dad_enhanced), 0,
108     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
109 
110 static VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to
111 						   transmit DAD packet */
112 #define	V_dad_maxtry			VNET(dad_maxtry)
113 
114 /*
115  * Input a Neighbor Solicitation Message.
116  *
117  * Based on RFC 2461
118  * Based on RFC 2462 (duplicate address detection)
119  */
120 void
121 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
122 {
123 	struct ifnet *ifp = m->m_pkthdr.rcvif;
124 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
125 	struct nd_neighbor_solicit *nd_ns;
126 	struct in6_addr saddr6 = ip6->ip6_src;
127 	struct in6_addr daddr6 = ip6->ip6_dst;
128 	struct in6_addr taddr6;
129 	struct in6_addr myaddr6;
130 	char *lladdr = NULL;
131 	struct ifaddr *ifa = NULL;
132 	int lladdrlen = 0;
133 	int anycast = 0, proxy = 0, tentative = 0;
134 	int tlladdr;
135 	int rflag;
136 	union nd_opts ndopts;
137 	struct sockaddr_dl proxydl;
138 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
139 
140 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
141 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
142 		rflag = 0;
143 #ifndef PULLDOWN_TEST
144 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
145 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
146 #else
147 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
148 	if (nd_ns == NULL) {
149 		ICMP6STAT_INC(icp6s_tooshort);
150 		return;
151 	}
152 #endif
153 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
154 	taddr6 = nd_ns->nd_ns_target;
155 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
156 		goto bad;
157 
158 	if (ip6->ip6_hlim != 255) {
159 		nd6log((LOG_ERR,
160 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
161 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
162 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
163 		goto bad;
164 	}
165 
166 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
167 		/* dst has to be a solicited node multicast address. */
168 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
169 		    /* don't check ifindex portion */
170 		    daddr6.s6_addr32[1] == 0 &&
171 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
172 		    daddr6.s6_addr8[12] == 0xff) {
173 			; /* good */
174 		} else {
175 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
176 			    "(wrong ip6 dst)\n"));
177 			goto bad;
178 		}
179 	} else if (!V_nd6_onlink_ns_rfc4861) {
180 		struct sockaddr_in6 src_sa6;
181 
182 		/*
183 		 * According to recent IETF discussions, it is not a good idea
184 		 * to accept a NS from an address which would not be deemed
185 		 * to be a neighbor otherwise.  This point is expected to be
186 		 * clarified in future revisions of the specification.
187 		 */
188 		bzero(&src_sa6, sizeof(src_sa6));
189 		src_sa6.sin6_family = AF_INET6;
190 		src_sa6.sin6_len = sizeof(src_sa6);
191 		src_sa6.sin6_addr = saddr6;
192 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
193 			nd6log((LOG_INFO, "nd6_ns_input: "
194 				"NS packet from non-neighbor\n"));
195 			goto bad;
196 		}
197 	}
198 
199 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
200 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
201 		goto bad;
202 	}
203 
204 	icmp6len -= sizeof(*nd_ns);
205 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
206 	if (nd6_options(&ndopts) < 0) {
207 		nd6log((LOG_INFO,
208 		    "nd6_ns_input: invalid ND option, ignored\n"));
209 		/* nd6_options have incremented stats */
210 		goto freeit;
211 	}
212 
213 	if (ndopts.nd_opts_src_lladdr) {
214 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
215 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
216 	}
217 
218 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
219 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
220 		    "(link-layer address option)\n"));
221 		goto bad;
222 	}
223 
224 	/*
225 	 * Attaching target link-layer address to the NA?
226 	 * (RFC 2461 7.2.4)
227 	 *
228 	 * NS IP dst is unicast/anycast			MUST NOT add
229 	 * NS IP dst is solicited-node multicast	MUST add
230 	 *
231 	 * In implementation, we add target link-layer address by default.
232 	 * We do not add one in MUST NOT cases.
233 	 */
234 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
235 		tlladdr = 0;
236 	else
237 		tlladdr = 1;
238 
239 	/*
240 	 * Target address (taddr6) must be either:
241 	 * (1) Valid unicast/anycast address for my receiving interface,
242 	 * (2) Unicast address for which I'm offering proxy service, or
243 	 * (3) "tentative" address on which DAD is being performed.
244 	 */
245 	/* (1) and (3) check. */
246 	if (ifp->if_carp)
247 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
248 	else
249 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
250 
251 	/* (2) check. */
252 	if (ifa == NULL) {
253 		struct sockaddr_dl rt_gateway;
254 		struct rt_addrinfo info;
255 		struct sockaddr_in6 dst6;
256 
257 		bzero(&dst6, sizeof(dst6));
258 		dst6.sin6_len = sizeof(struct sockaddr_in6);
259 		dst6.sin6_family = AF_INET6;
260 		dst6.sin6_addr = taddr6;
261 
262 		bzero(&rt_gateway, sizeof(rt_gateway));
263 		rt_gateway.sdl_len = sizeof(rt_gateway);
264 		bzero(&info, sizeof(info));
265 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
266 
267 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
268 		    0, 0, &info) == 0) {
269 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
270 			    rt_gateway.sdl_family == AF_LINK) {
271 
272 				/*
273 				 * proxy NDP for single entry
274 				 */
275 				proxydl = *SDL(&rt_gateway);
276 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
277 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
278 				if (ifa)
279 					proxy = 1;
280 			}
281 		}
282 	}
283 	if (ifa == NULL) {
284 		/*
285 		 * We've got an NS packet, and we don't have that adddress
286 		 * assigned for us.  We MUST silently ignore it.
287 		 * See RFC2461 7.2.3.
288 		 */
289 		goto freeit;
290 	}
291 	myaddr6 = *IFA_IN6(ifa);
292 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
293 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
294 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
295 		goto freeit;
296 
297 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
298 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
299 		    "(if %d, NS packet %d)\n",
300 		    ip6_sprintf(ip6bufs, &taddr6),
301 		    ifp->if_addrlen, lladdrlen - 2));
302 		goto bad;
303 	}
304 
305 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
306 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
307 		    ip6_sprintf(ip6bufs, &saddr6)));
308 		goto freeit;
309 	}
310 
311 	/*
312 	 * We have neighbor solicitation packet, with target address equals to
313 	 * one of my tentative address.
314 	 *
315 	 * src addr	how to process?
316 	 * ---		---
317 	 * multicast	of course, invalid (rejected in ip6_input)
318 	 * unicast	somebody is doing address resolution -> ignore
319 	 * unspec	dup address detection
320 	 *
321 	 * The processing is defined in RFC 2462.
322 	 */
323 	if (tentative) {
324 		/*
325 		 * If source address is unspecified address, it is for
326 		 * duplicate address detection.
327 		 *
328 		 * If not, the packet is for addess resolution;
329 		 * silently ignore it.
330 		 */
331 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
332 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
333 
334 		goto freeit;
335 	}
336 
337 	/*
338 	 * If the source address is unspecified address, entries must not
339 	 * be created or updated.
340 	 * It looks that sender is performing DAD.  Output NA toward
341 	 * all-node multicast address, to tell the sender that I'm using
342 	 * the address.
343 	 * S bit ("solicited") must be zero.
344 	 */
345 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
346 		struct in6_addr in6_all;
347 
348 		in6_all = in6addr_linklocal_allnodes;
349 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
350 			goto bad;
351 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
352 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
353 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
354 		    M_GETFIB(m));
355 		goto freeit;
356 	}
357 
358 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
359 	    ND_NEIGHBOR_SOLICIT, 0);
360 
361 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
362 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
363 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
364 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
365  freeit:
366 	if (ifa != NULL)
367 		ifa_free(ifa);
368 	m_freem(m);
369 	return;
370 
371  bad:
372 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
373 		ip6_sprintf(ip6bufs, &saddr6)));
374 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
375 		ip6_sprintf(ip6bufs, &daddr6)));
376 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
377 		ip6_sprintf(ip6bufs, &taddr6)));
378 	ICMP6STAT_INC(icp6s_badns);
379 	if (ifa != NULL)
380 		ifa_free(ifa);
381 	m_freem(m);
382 }
383 
384 /*
385  * Output a Neighbor Solicitation Message. Caller specifies:
386  *	- ICMP6 header source IP6 address
387  *	- ND6 header target IP6 address
388  *	- ND6 header source datalink address
389  *
390  * Based on RFC 2461
391  * Based on RFC 2462 (duplicate address detection)
392  *
393  *    ln - for source address determination
394  * nonce - If non-NULL, NS is used for duplicate address detection and
395  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
396  */
397 static void
398 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
399     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
400     uint8_t *nonce, u_int fibnum)
401 {
402 	struct mbuf *m;
403 	struct m_tag *mtag;
404 	struct ip6_hdr *ip6;
405 	struct nd_neighbor_solicit *nd_ns;
406 	struct ip6_moptions im6o;
407 	int icmp6len;
408 	int maxlen;
409 	caddr_t mac;
410 
411 	if (IN6_IS_ADDR_MULTICAST(taddr6))
412 		return;
413 
414 	/* estimate the size of message */
415 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
416 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
417 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
418 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
419 	    __func__, max_linkhdr, maxlen, MCLBYTES));
420 
421 	if (max_linkhdr + maxlen > MHLEN)
422 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
423 	else
424 		m = m_gethdr(M_NOWAIT, MT_DATA);
425 	if (m == NULL)
426 		return;
427 	M_SETFIB(m, fibnum);
428 
429 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
430 		m->m_flags |= M_MCAST;
431 		im6o.im6o_multicast_ifp = ifp;
432 		im6o.im6o_multicast_hlim = 255;
433 		im6o.im6o_multicast_loop = 0;
434 	}
435 
436 	icmp6len = sizeof(*nd_ns);
437 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
438 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
439 
440 	/* fill neighbor solicitation packet */
441 	ip6 = mtod(m, struct ip6_hdr *);
442 	ip6->ip6_flow = 0;
443 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
444 	ip6->ip6_vfc |= IPV6_VERSION;
445 	/* ip6->ip6_plen will be set later */
446 	ip6->ip6_nxt = IPPROTO_ICMPV6;
447 	ip6->ip6_hlim = 255;
448 	if (daddr6)
449 		ip6->ip6_dst = *daddr6;
450 	else {
451 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
452 		ip6->ip6_dst.s6_addr16[1] = 0;
453 		ip6->ip6_dst.s6_addr32[1] = 0;
454 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
455 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
456 		ip6->ip6_dst.s6_addr8[12] = 0xff;
457 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
458 			goto bad;
459 	}
460 	if (nonce == NULL) {
461 		struct ifaddr *ifa = NULL;
462 
463 		/*
464 		 * RFC2461 7.2.2:
465 		 * "If the source address of the packet prompting the
466 		 * solicitation is the same as one of the addresses assigned
467 		 * to the outgoing interface, that address SHOULD be placed
468 		 * in the IP Source Address of the outgoing solicitation.
469 		 * Otherwise, any one of the addresses assigned to the
470 		 * interface should be used."
471 		 *
472 		 * We use the source address for the prompting packet
473 		 * (saddr6), if saddr6 belongs to the outgoing interface.
474 		 * Otherwise, we perform the source address selection as usual.
475 		 */
476 
477 		if (saddr6 != NULL)
478 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
479 		if (ifa != NULL) {
480 			/* ip6_src set already. */
481 			ip6->ip6_src = *saddr6;
482 			ifa_free(ifa);
483 		} else {
484 			int error;
485 			struct in6_addr dst6, src6;
486 			uint32_t scopeid;
487 
488 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
489 			error = in6_selectsrc_addr(fibnum, &dst6,
490 			    scopeid, ifp, &src6, NULL);
491 			if (error) {
492 				char ip6buf[INET6_ADDRSTRLEN];
493 				nd6log((LOG_DEBUG, "%s: source can't be "
494 				    "determined: dst=%s, error=%d\n", __func__,
495 				    ip6_sprintf(ip6buf, &dst6),
496 				    error));
497 				goto bad;
498 			}
499 			ip6->ip6_src = src6;
500 		}
501 	} else {
502 		/*
503 		 * Source address for DAD packet must always be IPv6
504 		 * unspecified address. (0::0)
505 		 * We actually don't have to 0-clear the address (we did it
506 		 * above), but we do so here explicitly to make the intention
507 		 * clearer.
508 		 */
509 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
510 	}
511 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
512 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
513 	nd_ns->nd_ns_code = 0;
514 	nd_ns->nd_ns_reserved = 0;
515 	nd_ns->nd_ns_target = *taddr6;
516 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
517 
518 	/*
519 	 * Add source link-layer address option.
520 	 *
521 	 *				spec		implementation
522 	 *				---		---
523 	 * DAD packet			MUST NOT	do not add the option
524 	 * there's no link layer address:
525 	 *				impossible	do not add the option
526 	 * there's link layer address:
527 	 *	Multicast NS		MUST add one	add the option
528 	 *	Unicast NS		SHOULD add one	add the option
529 	 */
530 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
531 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
532 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
533 		/* 8 byte alignments... */
534 		optlen = (optlen + 7) & ~7;
535 
536 		m->m_pkthdr.len += optlen;
537 		m->m_len += optlen;
538 		icmp6len += optlen;
539 		bzero((caddr_t)nd_opt, optlen);
540 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
541 		nd_opt->nd_opt_len = optlen >> 3;
542 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
543 	}
544 	/*
545 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
546 	 * This behavior is documented as Enhanced Duplicate Address
547 	 * Detection in RFC 7527.
548 	 * net.inet6.ip6.dad_enhanced=0 disables this.
549 	 */
550 	if (V_dad_enhanced != 0 && nonce != NULL) {
551 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
552 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
553 		/* 8-byte alignment is required. */
554 		optlen = (optlen + 7) & ~7;
555 
556 		m->m_pkthdr.len += optlen;
557 		m->m_len += optlen;
558 		icmp6len += optlen;
559 		bzero((caddr_t)nd_opt, optlen);
560 		nd_opt->nd_opt_type = ND_OPT_NONCE;
561 		nd_opt->nd_opt_len = optlen >> 3;
562 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
563 	}
564 	ip6->ip6_plen = htons((u_short)icmp6len);
565 	nd_ns->nd_ns_cksum = 0;
566 	nd_ns->nd_ns_cksum =
567 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
568 
569 	if (send_sendso_input_hook != NULL) {
570 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
571 			sizeof(unsigned short), M_NOWAIT);
572 		if (mtag == NULL)
573 			goto bad;
574 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
575 		m_tag_prepend(m, mtag);
576 	}
577 
578 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
579 	    &im6o, NULL, NULL);
580 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
581 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
582 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
583 
584 	return;
585 
586   bad:
587 	m_freem(m);
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 ifnet *nd6_ifp;
862 
863 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
864 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
865 			    (ND_IFINFO(nd6_ifp)->flags &
866 			     ND6_IFF_ACCEPT_RTADV) != 0)
867 				/*
868 				 * Even if the neighbor is not in the default
869 				 * router list, the neighbor may be used as a
870 				 * next hop for some destinations (e.g. redirect
871 				 * case). So we must call rt6_flush explicitly.
872 				 */
873 				rt6_flush(&ip6->ip6_src, ifp);
874 		}
875 		ln->ln_router = is_router;
876 	}
877         /* XXX - QL
878 	 *  Does this matter?
879 	 *  rt->rt_flags &= ~RTF_REJECT;
880 	 */
881 	ln->la_asked = 0;
882 	if (ln->la_hold != NULL)
883 		nd6_grab_holdchain(ln, &chain, &sin6);
884  freeit:
885 	if (ln != NULL)
886 		LLE_WUNLOCK(ln);
887 
888 	if (chain != NULL)
889 		nd6_flush_holdchain(ifp, ifp, chain, &sin6);
890 
891 	if (checklink)
892 		pfxlist_onlink_check();
893 
894 	m_freem(m);
895 	return;
896 
897  bad:
898 	if (ln != NULL)
899 		LLE_WUNLOCK(ln);
900 
901 	ICMP6STAT_INC(icp6s_badna);
902 	m_freem(m);
903 }
904 
905 /*
906  * Neighbor advertisement output handling.
907  *
908  * Based on RFC 2461
909  *
910  * the following items are not implemented yet:
911  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
912  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
913  *
914  * tlladdr - 1 if include target link-layer address
915  * sdl0 - sockaddr_dl (= proxy NA) or NULL
916  */
917 static void
918 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
919     const struct in6_addr *taddr6, u_long flags, int tlladdr,
920     struct sockaddr *sdl0, u_int fibnum)
921 {
922 	struct mbuf *m;
923 	struct m_tag *mtag;
924 	struct ip6_hdr *ip6;
925 	struct nd_neighbor_advert *nd_na;
926 	struct ip6_moptions im6o;
927 	struct in6_addr daddr6, dst6, src6;
928 	uint32_t scopeid;
929 
930 	int icmp6len, maxlen, error;
931 	caddr_t mac = NULL;
932 
933 	daddr6 = *daddr6_0;	/* make a local copy for modification */
934 
935 	/* estimate the size of message */
936 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
937 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
938 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
939 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
940 	    __func__, max_linkhdr, maxlen, MCLBYTES));
941 
942 	if (max_linkhdr + maxlen > MHLEN)
943 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
944 	else
945 		m = m_gethdr(M_NOWAIT, MT_DATA);
946 	if (m == NULL)
947 		return;
948 	M_SETFIB(m, fibnum);
949 
950 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
951 		m->m_flags |= M_MCAST;
952 		im6o.im6o_multicast_ifp = ifp;
953 		im6o.im6o_multicast_hlim = 255;
954 		im6o.im6o_multicast_loop = 0;
955 	}
956 
957 	icmp6len = sizeof(*nd_na);
958 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
959 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
960 
961 	/* fill neighbor advertisement packet */
962 	ip6 = mtod(m, struct ip6_hdr *);
963 	ip6->ip6_flow = 0;
964 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
965 	ip6->ip6_vfc |= IPV6_VERSION;
966 	ip6->ip6_nxt = IPPROTO_ICMPV6;
967 	ip6->ip6_hlim = 255;
968 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
969 		/* reply to DAD */
970 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
971 		daddr6.s6_addr16[1] = 0;
972 		daddr6.s6_addr32[1] = 0;
973 		daddr6.s6_addr32[2] = 0;
974 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
975 		if (in6_setscope(&daddr6, ifp, NULL))
976 			goto bad;
977 
978 		flags &= ~ND_NA_FLAG_SOLICITED;
979 	}
980 	ip6->ip6_dst = daddr6;
981 
982 	/*
983 	 * Select a source whose scope is the same as that of the dest.
984 	 */
985 	in6_splitscope(&daddr6, &dst6, &scopeid);
986 	error = in6_selectsrc_addr(fibnum, &dst6,
987 	    scopeid, ifp, &src6, NULL);
988 	if (error) {
989 		char ip6buf[INET6_ADDRSTRLEN];
990 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
991 		    "determined: dst=%s, error=%d\n",
992 		    ip6_sprintf(ip6buf, &daddr6), error));
993 		goto bad;
994 	}
995 	ip6->ip6_src = src6;
996 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
997 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
998 	nd_na->nd_na_code = 0;
999 	nd_na->nd_na_target = *taddr6;
1000 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1001 
1002 	/*
1003 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1004 	 * see nd6_ns_input() for details.
1005 	 * Basically, if NS packet is sent to unicast/anycast addr,
1006 	 * target lladdr option SHOULD NOT be included.
1007 	 */
1008 	if (tlladdr) {
1009 		/*
1010 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1011 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1012 		 * my address) use lladdr configured for the interface.
1013 		 */
1014 		if (sdl0 == NULL) {
1015 			if (ifp->if_carp)
1016 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1017 			if (mac == NULL)
1018 				mac = nd6_ifptomac(ifp);
1019 		} else if (sdl0->sa_family == AF_LINK) {
1020 			struct sockaddr_dl *sdl;
1021 			sdl = (struct sockaddr_dl *)sdl0;
1022 			if (sdl->sdl_alen == ifp->if_addrlen)
1023 				mac = LLADDR(sdl);
1024 		}
1025 	}
1026 	if (tlladdr && mac) {
1027 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1028 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1029 
1030 		/* roundup to 8 bytes alignment! */
1031 		optlen = (optlen + 7) & ~7;
1032 
1033 		m->m_pkthdr.len += optlen;
1034 		m->m_len += optlen;
1035 		icmp6len += optlen;
1036 		bzero((caddr_t)nd_opt, optlen);
1037 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1038 		nd_opt->nd_opt_len = optlen >> 3;
1039 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1040 	} else
1041 		flags &= ~ND_NA_FLAG_OVERRIDE;
1042 
1043 	ip6->ip6_plen = htons((u_short)icmp6len);
1044 	nd_na->nd_na_flags_reserved = flags;
1045 	nd_na->nd_na_cksum = 0;
1046 	nd_na->nd_na_cksum =
1047 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1048 
1049 	if (send_sendso_input_hook != NULL) {
1050 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1051 		    sizeof(unsigned short), M_NOWAIT);
1052 		if (mtag == NULL)
1053 			goto bad;
1054 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1055 		m_tag_prepend(m, mtag);
1056 	}
1057 
1058 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1059 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1060 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1061 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1062 
1063 	return;
1064 
1065   bad:
1066 	m_freem(m);
1067 }
1068 
1069 #ifndef BURN_BRIDGES
1070 void
1071 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1072     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1073     struct sockaddr *sdl0)
1074 {
1075 
1076 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1077 	    RT_DEFAULT_FIB);
1078 }
1079 #endif
1080 
1081 caddr_t
1082 nd6_ifptomac(struct ifnet *ifp)
1083 {
1084 	switch (ifp->if_type) {
1085 	case IFT_ARCNET:
1086 	case IFT_ETHER:
1087 	case IFT_FDDI:
1088 	case IFT_IEEE1394:
1089 	case IFT_L2VLAN:
1090 	case IFT_INFINIBAND:
1091 	case IFT_BRIDGE:
1092 	case IFT_ISO88025:
1093 		return IF_LLADDR(ifp);
1094 	default:
1095 		return NULL;
1096 	}
1097 }
1098 
1099 struct dadq {
1100 	TAILQ_ENTRY(dadq) dad_list;
1101 	struct ifaddr *dad_ifa;
1102 	int dad_count;		/* max NS to send */
1103 	int dad_ns_tcount;	/* # of trials to send NS */
1104 	int dad_ns_ocount;	/* NS sent so far */
1105 	int dad_ns_icount;
1106 	int dad_na_icount;
1107 	int dad_ns_lcount;	/* looped back NS */
1108 	int dad_loopbackprobe;	/* probing state for loopback detection */
1109 	struct callout dad_timer_ch;
1110 	struct vnet *dad_vnet;
1111 	u_int dad_refcnt;
1112 #define	ND_OPT_NONCE_LEN32 \
1113 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1114 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1115 };
1116 
1117 static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1118 static VNET_DEFINE(struct rwlock, dad_rwlock);
1119 #define	V_dadq			VNET(dadq)
1120 #define	V_dad_rwlock		VNET(dad_rwlock)
1121 
1122 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1123 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1124 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1125 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1126 
1127 static void
1128 nd6_dad_add(struct dadq *dp)
1129 {
1130 
1131 	DADQ_WLOCK();
1132 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1133 	DADQ_WUNLOCK();
1134 }
1135 
1136 static void
1137 nd6_dad_del(struct dadq *dp)
1138 {
1139 
1140 	DADQ_WLOCK();
1141 	TAILQ_REMOVE(&V_dadq, dp, dad_list);
1142 	DADQ_WUNLOCK();
1143 	nd6_dad_rele(dp);
1144 }
1145 
1146 static struct dadq *
1147 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1148 {
1149 	struct dadq *dp;
1150 
1151 	DADQ_RLOCK();
1152 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1153 		if (dp->dad_ifa != ifa)
1154 			continue;
1155 		/*
1156 		 * Skip if the nonce matches the received one.
1157 		 * +2 in the length is required because of type and
1158 		 * length fields are included in a header.
1159 		 */
1160 		if (n != NULL &&
1161 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1162 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1163 		        ND_OPT_NONCE_LEN) == 0) {
1164 			dp->dad_ns_lcount++;
1165 			continue;
1166 		}
1167 		refcount_acquire(&dp->dad_refcnt);
1168 		break;
1169 	}
1170 	DADQ_RUNLOCK();
1171 
1172 	return (dp);
1173 }
1174 
1175 static void
1176 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1177 {
1178 
1179 	if (send_ns != 0)
1180 		nd6_dad_ns_output(dp);
1181 	callout_reset(&dp->dad_timer_ch, ticks,
1182 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1183 }
1184 
1185 static void
1186 nd6_dad_stoptimer(struct dadq *dp)
1187 {
1188 
1189 	callout_drain(&dp->dad_timer_ch);
1190 }
1191 
1192 static void
1193 nd6_dad_rele(struct dadq *dp)
1194 {
1195 
1196 	if (refcount_release(&dp->dad_refcnt)) {
1197 		ifa_free(dp->dad_ifa);
1198 		free(dp, M_IP6NDP);
1199 	}
1200 }
1201 
1202 void
1203 nd6_dad_init(void)
1204 {
1205 
1206 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1207 	TAILQ_INIT(&V_dadq);
1208 }
1209 
1210 /*
1211  * Start Duplicate Address Detection (DAD) for specified interface address.
1212  */
1213 void
1214 nd6_dad_start(struct ifaddr *ifa, int delay)
1215 {
1216 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1217 	struct dadq *dp;
1218 	char ip6buf[INET6_ADDRSTRLEN];
1219 
1220 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1221 	    ("starting DAD on non-tentative address %p", ifa));
1222 
1223 	/*
1224 	 * If we don't need DAD, don't do it.
1225 	 * There are several cases:
1226 	 * - DAD is disabled globally or on the interface
1227 	 * - the interface address is anycast
1228 	 */
1229 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1230 	    V_ip6_dad_count == 0 ||
1231 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1232 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1233 		return;
1234 	}
1235 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1236 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1237 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1238 		return;
1239 
1240 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1241 		/*
1242 		 * DAD is already in progress.  Let the existing entry
1243 		 * finish it.
1244 		 */
1245 		nd6_dad_rele(dp);
1246 		return;
1247 	}
1248 
1249 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1250 	if (dp == NULL) {
1251 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1252 			"%s(%s)\n",
1253 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1254 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1255 		return;
1256 	}
1257 	callout_init(&dp->dad_timer_ch, 0);
1258 #ifdef VIMAGE
1259 	dp->dad_vnet = curvnet;
1260 #endif
1261 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1262 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1263 
1264 	/*
1265 	 * Send NS packet for DAD, ip6_dad_count times.
1266 	 * Note that we must delay the first transmission, if this is the
1267 	 * first packet to be sent from the interface after interface
1268 	 * (re)initialization.
1269 	 */
1270 	dp->dad_ifa = ifa;
1271 	ifa_ref(dp->dad_ifa);
1272 	dp->dad_count = V_ip6_dad_count;
1273 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1274 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1275 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1276 	refcount_init(&dp->dad_refcnt, 1);
1277 	nd6_dad_add(dp);
1278 	nd6_dad_starttimer(dp, delay, 0);
1279 }
1280 
1281 /*
1282  * terminate DAD unconditionally.  used for address removals.
1283  */
1284 void
1285 nd6_dad_stop(struct ifaddr *ifa)
1286 {
1287 	struct dadq *dp;
1288 
1289 	dp = nd6_dad_find(ifa, NULL);
1290 	if (!dp) {
1291 		/* DAD wasn't started yet */
1292 		return;
1293 	}
1294 
1295 	nd6_dad_stoptimer(dp);
1296 
1297 	/*
1298 	 * The DAD queue entry may have been removed by nd6_dad_timer() while
1299 	 * we were waiting for it to stop, so re-do the lookup.
1300 	 */
1301 	nd6_dad_rele(dp);
1302 	if (nd6_dad_find(ifa, NULL) == NULL)
1303 		return;
1304 
1305 	nd6_dad_del(dp);
1306 	nd6_dad_rele(dp);
1307 }
1308 
1309 static void
1310 nd6_dad_timer(struct dadq *dp)
1311 {
1312 	CURVNET_SET(dp->dad_vnet);
1313 	struct ifaddr *ifa = dp->dad_ifa;
1314 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1315 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1316 	char ip6buf[INET6_ADDRSTRLEN];
1317 
1318 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1319 
1320 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1321 		/* Do not need DAD for ifdisabled interface. */
1322 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1323 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1324 		goto err;
1325 	}
1326 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1327 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1328 			"%s(%s)\n",
1329 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1330 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1331 		goto err;
1332 	}
1333 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1334 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1335 			"%s(%s)\n",
1336 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1337 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1338 		goto err;
1339 	}
1340 
1341 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1342 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1343 	    (((ifp->if_flags & IFF_UP) == 0) ||
1344 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1345 		nd6log((LOG_INFO, "%s: could not run DAD "
1346 		    "because the interface was down or not running.\n",
1347 		    if_name(ifa->ifa_ifp)));
1348 		goto err;
1349 	}
1350 
1351 	/* Need more checks? */
1352 	if (dp->dad_ns_ocount < dp->dad_count) {
1353 		/*
1354 		 * We have more NS to go.  Send NS packet for DAD.
1355 		 */
1356 		nd6_dad_starttimer(dp,
1357 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1358 		goto done;
1359 	} else {
1360 		/*
1361 		 * We have transmitted sufficient number of DAD packets.
1362 		 * See what we've got.
1363 		 */
1364 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1365 			/* We've seen NS or NA, means DAD has failed. */
1366 			nd6_dad_duplicated(ifa, dp);
1367 		else if (V_dad_enhanced != 0 &&
1368 		    dp->dad_ns_lcount > 0 &&
1369 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1370 			/*
1371 			 * Sec. 4.1 in RFC 7527 requires transmission of
1372 			 * additional probes until the loopback condition
1373 			 * becomes clear when a looped back probe is detected.
1374 			 */
1375 			log(LOG_ERR, "%s: a looped back NS message is "
1376 			    "detected during DAD for %s.  "
1377 			    "Another DAD probes are being sent.\n",
1378 			    if_name(ifa->ifa_ifp),
1379 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1380 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1381 			/*
1382 			 * Send an NS immediately and increase dad_count by
1383 			 * V_nd6_mmaxtries - 1.
1384 			 */
1385 			dp->dad_count =
1386 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1387 			nd6_dad_starttimer(dp,
1388 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1389 			    1);
1390 			goto done;
1391 		} else {
1392 			/*
1393 			 * We are done with DAD.  No NA came, no NS came.
1394 			 * No duplicate address found.  Check IFDISABLED flag
1395 			 * again in case that it is changed between the
1396 			 * beginning of this function and here.
1397 			 */
1398 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1399 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1400 
1401 			nd6log((LOG_DEBUG,
1402 			    "%s: DAD complete for %s - no duplicates found\n",
1403 			    if_name(ifa->ifa_ifp),
1404 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1405 			if (dp->dad_ns_lcount > 0)
1406 				log(LOG_ERR, "%s: DAD completed while "
1407 				    "a looped back NS message is detected "
1408 				    "during DAD for %s.\n",
1409 				    if_name(ifa->ifa_ifp),
1410 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1411 		}
1412 	}
1413 err:
1414 	nd6_dad_del(dp);
1415 done:
1416 	CURVNET_RESTORE();
1417 }
1418 
1419 static void
1420 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1421 {
1422 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1423 	struct ifnet *ifp;
1424 	char ip6buf[INET6_ADDRSTRLEN];
1425 
1426 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1427 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1428 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1429 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1430 	    dp->dad_na_icount);
1431 
1432 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1433 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1434 
1435 	ifp = ifa->ifa_ifp;
1436 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1437 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1438 	log(LOG_ERR, "%s: manual intervention required\n",
1439 	    if_name(ifp));
1440 
1441 	/*
1442 	 * If the address is a link-local address formed from an interface
1443 	 * identifier based on the hardware address which is supposed to be
1444 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1445 	 * operation on the interface SHOULD be disabled.
1446 	 * [RFC 4862, Section 5.4.5]
1447 	 */
1448 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1449 		struct in6_addr in6;
1450 
1451 		/*
1452 		 * To avoid over-reaction, we only apply this logic when we are
1453 		 * very sure that hardware addresses are supposed to be unique.
1454 		 */
1455 		switch (ifp->if_type) {
1456 		case IFT_ETHER:
1457 		case IFT_FDDI:
1458 		case IFT_ATM:
1459 		case IFT_IEEE1394:
1460 		case IFT_INFINIBAND:
1461 			in6 = ia->ia_addr.sin6_addr;
1462 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1463 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1464 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1465 				log(LOG_ERR, "%s: possible hardware address "
1466 				    "duplication detected, disable IPv6\n",
1467 				    if_name(ifp));
1468 			}
1469 			break;
1470 		}
1471 	}
1472 }
1473 
1474 static void
1475 nd6_dad_ns_output(struct dadq *dp)
1476 {
1477 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1478 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1479 	int i;
1480 
1481 	dp->dad_ns_tcount++;
1482 	if ((ifp->if_flags & IFF_UP) == 0) {
1483 		return;
1484 	}
1485 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1486 		return;
1487 	}
1488 
1489 	dp->dad_ns_ocount++;
1490 	if (V_dad_enhanced != 0) {
1491 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1492 			dp->dad_nonce[i] = arc4random();
1493 		/*
1494 		 * XXXHRS: Note that in the case that
1495 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1496 		 * different nonces can be looped back in an unexpected
1497 		 * order.  The current implementation recognizes only
1498 		 * the latest nonce on the sender side.  Practically it
1499 		 * should work well in almost all cases.
1500 		 */
1501 	}
1502 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1503 	    (uint8_t *)&dp->dad_nonce[0]);
1504 }
1505 
1506 static void
1507 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1508 {
1509 	struct in6_ifaddr *ia;
1510 	struct ifnet *ifp;
1511 	const struct in6_addr *taddr6;
1512 	struct dadq *dp;
1513 
1514 	if (ifa == NULL)
1515 		panic("ifa == NULL in nd6_dad_ns_input");
1516 
1517 	ia = (struct in6_ifaddr *)ifa;
1518 	ifp = ifa->ifa_ifp;
1519 	taddr6 = &ia->ia_addr.sin6_addr;
1520 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1521 	if (V_dad_enhanced == 0)
1522 		ndopt_nonce = NULL;
1523 	dp = nd6_dad_find(ifa, ndopt_nonce);
1524 	if (dp == NULL)
1525 		return;
1526 
1527 	dp->dad_ns_icount++;
1528 	nd6_dad_rele(dp);
1529 }
1530 
1531 static void
1532 nd6_dad_na_input(struct ifaddr *ifa)
1533 {
1534 	struct dadq *dp;
1535 
1536 	if (ifa == NULL)
1537 		panic("ifa == NULL in nd6_dad_na_input");
1538 
1539 	dp = nd6_dad_find(ifa, NULL);
1540 	if (dp != NULL) {
1541 		dp->dad_na_icount++;
1542 		nd6_dad_rele(dp);
1543 	}
1544 }
1545