xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 907b59d76938e654f0d040a888e8dfca3de1e222)
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 }
588 
589 #ifndef BURN_BRIDGES
590 void
591 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
592     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
593 {
594 
595 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
596 }
597 #endif
598 /*
599  * Neighbor advertisement input handling.
600  *
601  * Based on RFC 2461
602  * Based on RFC 2462 (duplicate address detection)
603  *
604  * the following items are not implemented yet:
605  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
606  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
607  */
608 void
609 nd6_na_input(struct mbuf *m, int off, int icmp6len)
610 {
611 	struct ifnet *ifp = m->m_pkthdr.rcvif;
612 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
613 	struct nd_neighbor_advert *nd_na;
614 	struct in6_addr daddr6 = ip6->ip6_dst;
615 	struct in6_addr taddr6;
616 	int flags;
617 	int is_router;
618 	int is_solicited;
619 	int is_override;
620 	char *lladdr = NULL;
621 	int lladdrlen = 0;
622 	int checklink = 0;
623 	struct ifaddr *ifa;
624 	struct llentry *ln = NULL;
625 	union nd_opts ndopts;
626 	struct mbuf *chain = NULL;
627 	struct sockaddr_in6 sin6;
628 	u_char linkhdr[LLE_MAX_LINKHDR];
629 	size_t linkhdrsize;
630 	int lladdr_off;
631 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
632 
633 	if (ip6->ip6_hlim != 255) {
634 		nd6log((LOG_ERR,
635 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
636 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
637 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
638 		goto bad;
639 	}
640 
641 #ifndef PULLDOWN_TEST
642 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
643 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
644 #else
645 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
646 	if (nd_na == NULL) {
647 		ICMP6STAT_INC(icp6s_tooshort);
648 		return;
649 	}
650 #endif
651 
652 	flags = nd_na->nd_na_flags_reserved;
653 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
654 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
655 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
656 	memset(&sin6, 0, sizeof(sin6));
657 
658 	taddr6 = nd_na->nd_na_target;
659 	if (in6_setscope(&taddr6, ifp, NULL))
660 		goto bad;	/* XXX: impossible */
661 
662 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
663 		nd6log((LOG_ERR,
664 		    "nd6_na_input: invalid target address %s\n",
665 		    ip6_sprintf(ip6bufs, &taddr6)));
666 		goto bad;
667 	}
668 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
669 		if (is_solicited) {
670 			nd6log((LOG_ERR,
671 			    "nd6_na_input: a solicited adv is multicasted\n"));
672 			goto bad;
673 		}
674 
675 	icmp6len -= sizeof(*nd_na);
676 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
677 	if (nd6_options(&ndopts) < 0) {
678 		nd6log((LOG_INFO,
679 		    "nd6_na_input: invalid ND option, ignored\n"));
680 		/* nd6_options have incremented stats */
681 		goto freeit;
682 	}
683 
684 	if (ndopts.nd_opts_tgt_lladdr) {
685 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
686 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
687 	}
688 
689 	/*
690 	 * This effectively disables the DAD check on a non-master CARP
691 	 * address.
692 	 */
693 	if (ifp->if_carp)
694 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
695 	else
696 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
697 
698 	/*
699 	 * Target address matches one of my interface address.
700 	 *
701 	 * If my address is tentative, this means that there's somebody
702 	 * already using the same address as mine.  This indicates DAD failure.
703 	 * This is defined in RFC 2462.
704 	 *
705 	 * Otherwise, process as defined in RFC 2461.
706 	 */
707 	if (ifa
708 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
709 		nd6_dad_na_input(ifa);
710 		ifa_free(ifa);
711 		goto freeit;
712 	}
713 
714 	/* Just for safety, maybe unnecessary. */
715 	if (ifa) {
716 		ifa_free(ifa);
717 		log(LOG_ERR,
718 		    "nd6_na_input: duplicate IP6 address %s\n",
719 		    ip6_sprintf(ip6bufs, &taddr6));
720 		goto freeit;
721 	}
722 
723 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
724 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
725 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
726 		    ifp->if_addrlen, lladdrlen - 2));
727 		goto bad;
728 	}
729 
730 	/*
731 	 * If no neighbor cache entry is found, NA SHOULD silently be
732 	 * discarded.
733 	 */
734 	IF_AFDATA_RLOCK(ifp);
735 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
736 	IF_AFDATA_RUNLOCK(ifp);
737 	if (ln == NULL) {
738 		goto freeit;
739 	}
740 
741 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
742 		/*
743 		 * If the link-layer has address, and no lladdr option came,
744 		 * discard the packet.
745 		 */
746 		if (ifp->if_addrlen && lladdr == NULL) {
747 			goto freeit;
748 		}
749 
750 		/*
751 		 * Record link-layer address, and update the state.
752 		 */
753 		linkhdrsize = sizeof(linkhdr);
754 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
755 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
756 			return;
757 
758 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
759 		    lladdr_off) == 0) {
760 			ln = NULL;
761 			goto freeit;
762 		}
763 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
764 		if (is_solicited)
765 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
766 		else
767 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
768 		if ((ln->ln_router = is_router) != 0) {
769 			/*
770 			 * This means a router's state has changed from
771 			 * non-reachable to probably reachable, and might
772 			 * affect the status of associated prefixes..
773 			 */
774 			checklink = 1;
775 		}
776 	} else {
777 		int llchange;
778 
779 		/*
780 		 * Check if the link-layer address has changed or not.
781 		 */
782 		if (lladdr == NULL)
783 			llchange = 0;
784 		else {
785 			if (ln->la_flags & LLE_VALID) {
786 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
787 					llchange = 1;
788 				else
789 					llchange = 0;
790 			} else
791 				llchange = 1;
792 		}
793 
794 		/*
795 		 * This is VERY complex.  Look at it with care.
796 		 *
797 		 * override solicit lladdr llchange	action
798 		 *					(L: record lladdr)
799 		 *
800 		 *	0	0	n	--	(2c)
801 		 *	0	0	y	n	(2b) L
802 		 *	0	0	y	y	(1)    REACHABLE->STALE
803 		 *	0	1	n	--	(2c)   *->REACHABLE
804 		 *	0	1	y	n	(2b) L *->REACHABLE
805 		 *	0	1	y	y	(1)    REACHABLE->STALE
806 		 *	1	0	n	--	(2a)
807 		 *	1	0	y	n	(2a) L
808 		 *	1	0	y	y	(2a) L *->STALE
809 		 *	1	1	n	--	(2a)   *->REACHABLE
810 		 *	1	1	y	n	(2a) L *->REACHABLE
811 		 *	1	1	y	y	(2a) L *->REACHABLE
812 		 */
813 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
814 			/*
815 			 * If state is REACHABLE, make it STALE.
816 			 * no other updates should be done.
817 			 */
818 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
819 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
820 			goto freeit;
821 		} else if (is_override				   /* (2a) */
822 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
823 			|| lladdr == NULL) {			   /* (2c) */
824 			/*
825 			 * Update link-local address, if any.
826 			 */
827 			if (lladdr != NULL) {
828 				linkhdrsize = sizeof(linkhdr);
829 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
830 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
831 					goto freeit;
832 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
833 				    linkhdrsize, lladdr_off) == 0) {
834 					ln = NULL;
835 					goto freeit;
836 				}
837 				EVENTHANDLER_INVOKE(lle_event, ln,
838 				    LLENTRY_RESOLVED);
839 			}
840 
841 			/*
842 			 * If solicited, make the state REACHABLE.
843 			 * If not solicited and the link-layer address was
844 			 * changed, make it STALE.
845 			 */
846 			if (is_solicited)
847 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
848 			else {
849 				if (lladdr != NULL && llchange)
850 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
851 			}
852 		}
853 
854 		if (ln->ln_router && !is_router) {
855 			/*
856 			 * The peer dropped the router flag.
857 			 * Remove the sender from the Default Router List and
858 			 * update the Destination Cache entries.
859 			 */
860 			struct ifnet *nd6_ifp;
861 
862 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
863 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
864 			    (ND_IFINFO(nd6_ifp)->flags &
865 			     ND6_IFF_ACCEPT_RTADV) != 0)
866 				/*
867 				 * Even if the neighbor is not in the default
868 				 * router list, the neighbor may be used as a
869 				 * next hop for some destinations (e.g. redirect
870 				 * case). So we must call rt6_flush explicitly.
871 				 */
872 				rt6_flush(&ip6->ip6_src, ifp);
873 		}
874 		ln->ln_router = is_router;
875 	}
876         /* XXX - QL
877 	 *  Does this matter?
878 	 *  rt->rt_flags &= ~RTF_REJECT;
879 	 */
880 	ln->la_asked = 0;
881 	if (ln->la_hold != NULL)
882 		nd6_grab_holdchain(ln, &chain, &sin6);
883  freeit:
884 	if (ln != NULL)
885 		LLE_WUNLOCK(ln);
886 
887 	if (chain != NULL)
888 		nd6_flush_holdchain(ifp, ifp, chain, &sin6);
889 
890 	if (checklink)
891 		pfxlist_onlink_check();
892 
893 	m_freem(m);
894 	return;
895 
896  bad:
897 	if (ln != NULL)
898 		LLE_WUNLOCK(ln);
899 
900 	ICMP6STAT_INC(icp6s_badna);
901 	m_freem(m);
902 }
903 
904 /*
905  * Neighbor advertisement output handling.
906  *
907  * Based on RFC 2461
908  *
909  * the following items are not implemented yet:
910  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
911  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
912  *
913  * tlladdr - 1 if include target link-layer address
914  * sdl0 - sockaddr_dl (= proxy NA) or NULL
915  */
916 static void
917 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
918     const struct in6_addr *taddr6, u_long flags, int tlladdr,
919     struct sockaddr *sdl0, u_int fibnum)
920 {
921 	struct mbuf *m;
922 	struct m_tag *mtag;
923 	struct ip6_hdr *ip6;
924 	struct nd_neighbor_advert *nd_na;
925 	struct ip6_moptions im6o;
926 	struct in6_addr daddr6, dst6, src6;
927 	uint32_t scopeid;
928 
929 	int icmp6len, maxlen, error;
930 	caddr_t mac = NULL;
931 
932 	daddr6 = *daddr6_0;	/* make a local copy for modification */
933 
934 	/* estimate the size of message */
935 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
936 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
937 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
938 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
939 	    __func__, max_linkhdr, maxlen, MCLBYTES));
940 
941 	if (max_linkhdr + maxlen > MHLEN)
942 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
943 	else
944 		m = m_gethdr(M_NOWAIT, MT_DATA);
945 	if (m == NULL)
946 		return;
947 	M_SETFIB(m, fibnum);
948 
949 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
950 		m->m_flags |= M_MCAST;
951 		im6o.im6o_multicast_ifp = ifp;
952 		im6o.im6o_multicast_hlim = 255;
953 		im6o.im6o_multicast_loop = 0;
954 	}
955 
956 	icmp6len = sizeof(*nd_na);
957 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
958 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
959 
960 	/* fill neighbor advertisement packet */
961 	ip6 = mtod(m, struct ip6_hdr *);
962 	ip6->ip6_flow = 0;
963 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
964 	ip6->ip6_vfc |= IPV6_VERSION;
965 	ip6->ip6_nxt = IPPROTO_ICMPV6;
966 	ip6->ip6_hlim = 255;
967 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
968 		/* reply to DAD */
969 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
970 		daddr6.s6_addr16[1] = 0;
971 		daddr6.s6_addr32[1] = 0;
972 		daddr6.s6_addr32[2] = 0;
973 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
974 		if (in6_setscope(&daddr6, ifp, NULL))
975 			goto bad;
976 
977 		flags &= ~ND_NA_FLAG_SOLICITED;
978 	}
979 	ip6->ip6_dst = daddr6;
980 
981 	/*
982 	 * Select a source whose scope is the same as that of the dest.
983 	 */
984 	in6_splitscope(&daddr6, &dst6, &scopeid);
985 	error = in6_selectsrc_addr(RT_DEFAULT_FIB, &dst6,
986 	    scopeid, ifp, &src6, NULL);
987 	if (error) {
988 		char ip6buf[INET6_ADDRSTRLEN];
989 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
990 		    "determined: dst=%s, error=%d\n",
991 		    ip6_sprintf(ip6buf, &daddr6), error));
992 		goto bad;
993 	}
994 	ip6->ip6_src = src6;
995 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
996 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
997 	nd_na->nd_na_code = 0;
998 	nd_na->nd_na_target = *taddr6;
999 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1000 
1001 	/*
1002 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1003 	 * see nd6_ns_input() for details.
1004 	 * Basically, if NS packet is sent to unicast/anycast addr,
1005 	 * target lladdr option SHOULD NOT be included.
1006 	 */
1007 	if (tlladdr) {
1008 		/*
1009 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1010 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1011 		 * my address) use lladdr configured for the interface.
1012 		 */
1013 		if (sdl0 == NULL) {
1014 			if (ifp->if_carp)
1015 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1016 			if (mac == NULL)
1017 				mac = nd6_ifptomac(ifp);
1018 		} else if (sdl0->sa_family == AF_LINK) {
1019 			struct sockaddr_dl *sdl;
1020 			sdl = (struct sockaddr_dl *)sdl0;
1021 			if (sdl->sdl_alen == ifp->if_addrlen)
1022 				mac = LLADDR(sdl);
1023 		}
1024 	}
1025 	if (tlladdr && mac) {
1026 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1027 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1028 
1029 		/* roundup to 8 bytes alignment! */
1030 		optlen = (optlen + 7) & ~7;
1031 
1032 		m->m_pkthdr.len += optlen;
1033 		m->m_len += optlen;
1034 		icmp6len += optlen;
1035 		bzero((caddr_t)nd_opt, optlen);
1036 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1037 		nd_opt->nd_opt_len = optlen >> 3;
1038 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1039 	} else
1040 		flags &= ~ND_NA_FLAG_OVERRIDE;
1041 
1042 	ip6->ip6_plen = htons((u_short)icmp6len);
1043 	nd_na->nd_na_flags_reserved = flags;
1044 	nd_na->nd_na_cksum = 0;
1045 	nd_na->nd_na_cksum =
1046 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1047 
1048 	if (send_sendso_input_hook != NULL) {
1049 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1050 		    sizeof(unsigned short), M_NOWAIT);
1051 		if (mtag == NULL)
1052 			goto bad;
1053 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1054 		m_tag_prepend(m, mtag);
1055 	}
1056 
1057 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1058 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1059 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1060 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1061 
1062 	return;
1063 
1064   bad:
1065 	m_freem(m);
1066 }
1067 
1068 #ifndef BURN_BRIDGES
1069 void
1070 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1071     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1072     struct sockaddr *sdl0)
1073 {
1074 
1075 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1076 	    RT_DEFAULT_FIB);
1077 }
1078 #endif
1079 
1080 caddr_t
1081 nd6_ifptomac(struct ifnet *ifp)
1082 {
1083 	switch (ifp->if_type) {
1084 	case IFT_ARCNET:
1085 	case IFT_ETHER:
1086 	case IFT_FDDI:
1087 	case IFT_IEEE1394:
1088 	case IFT_L2VLAN:
1089 	case IFT_IEEE80211:
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 	/*
1221 	 * If we don't need DAD, don't do it.
1222 	 * There are several cases:
1223 	 * - DAD is disabled (ip6_dad_count == 0)
1224 	 * - the interface address is anycast
1225 	 */
1226 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1227 		log(LOG_DEBUG,
1228 			"nd6_dad_start: called with non-tentative address "
1229 			"%s(%s)\n",
1230 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1231 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1232 		return;
1233 	}
1234 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1235 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1236 		return;
1237 	}
1238 	if (!V_ip6_dad_count) {
1239 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1240 		return;
1241 	}
1242 	if (ifa->ifa_ifp == NULL)
1243 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1244 	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) {
1245 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1246 		return;
1247 	}
1248 	if (!(ifa->ifa_ifp->if_flags & IFF_UP) ||
1249 	    !(ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) ||
1250 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)) {
1251 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1252 		return;
1253 	}
1254 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1255 		/*
1256 		 * DAD is already in progress.  Let the existing entry
1257 		 * finish it.
1258 		 */
1259 		nd6_dad_rele(dp);
1260 		return;
1261 	}
1262 
1263 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1264 	if (dp == NULL) {
1265 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1266 			"%s(%s)\n",
1267 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1268 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1269 		return;
1270 	}
1271 	callout_init(&dp->dad_timer_ch, 0);
1272 #ifdef VIMAGE
1273 	dp->dad_vnet = curvnet;
1274 #endif
1275 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1276 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1277 
1278 	/*
1279 	 * Send NS packet for DAD, ip6_dad_count times.
1280 	 * Note that we must delay the first transmission, if this is the
1281 	 * first packet to be sent from the interface after interface
1282 	 * (re)initialization.
1283 	 */
1284 	dp->dad_ifa = ifa;
1285 	ifa_ref(dp->dad_ifa);
1286 	dp->dad_count = V_ip6_dad_count;
1287 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1288 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1289 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1290 	refcount_init(&dp->dad_refcnt, 1);
1291 	nd6_dad_add(dp);
1292 	nd6_dad_starttimer(dp, delay, 0);
1293 }
1294 
1295 /*
1296  * terminate DAD unconditionally.  used for address removals.
1297  */
1298 void
1299 nd6_dad_stop(struct ifaddr *ifa)
1300 {
1301 	struct dadq *dp;
1302 
1303 	dp = nd6_dad_find(ifa, NULL);
1304 	if (!dp) {
1305 		/* DAD wasn't started yet */
1306 		return;
1307 	}
1308 
1309 	nd6_dad_stoptimer(dp);
1310 
1311 	/*
1312 	 * The DAD queue entry may have been removed by nd6_dad_timer() while
1313 	 * we were waiting for it to stop, so re-do the lookup.
1314 	 */
1315 	nd6_dad_rele(dp);
1316 	if (nd6_dad_find(ifa, NULL) == NULL)
1317 		return;
1318 
1319 	nd6_dad_del(dp);
1320 	nd6_dad_rele(dp);
1321 }
1322 
1323 static void
1324 nd6_dad_timer(struct dadq *dp)
1325 {
1326 	CURVNET_SET(dp->dad_vnet);
1327 	struct ifaddr *ifa = dp->dad_ifa;
1328 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1329 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1330 	char ip6buf[INET6_ADDRSTRLEN];
1331 
1332 	/* Sanity check */
1333 	if (ia == NULL) {
1334 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1335 		goto err;
1336 	}
1337 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1338 		/* Do not need DAD for ifdisabled interface. */
1339 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1340 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1341 		goto err;
1342 	}
1343 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1344 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1345 			"%s(%s)\n",
1346 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1347 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1348 		goto err;
1349 	}
1350 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1351 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1352 			"%s(%s)\n",
1353 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1354 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1355 		goto err;
1356 	}
1357 
1358 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1359 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1360 	    (((ifp->if_flags & IFF_UP) == 0) ||
1361 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1362 		nd6log((LOG_INFO, "%s: could not run DAD "
1363 		    "because the interface was down or not running.\n",
1364 		    if_name(ifa->ifa_ifp)));
1365 		goto err;
1366 	}
1367 
1368 	/* Need more checks? */
1369 	if (dp->dad_ns_ocount < dp->dad_count) {
1370 		/*
1371 		 * We have more NS to go.  Send NS packet for DAD.
1372 		 */
1373 		nd6_dad_starttimer(dp,
1374 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1375 		goto done;
1376 	} else {
1377 		/*
1378 		 * We have transmitted sufficient number of DAD packets.
1379 		 * See what we've got.
1380 		 */
1381 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1382 			/* We've seen NS or NA, means DAD has failed. */
1383 			nd6_dad_duplicated(ifa, dp);
1384 		else if (V_dad_enhanced != 0 &&
1385 		    dp->dad_ns_lcount > 0 &&
1386 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1387 			/*
1388 			 * Sec. 4.1 in RFC 7527 requires transmission of
1389 			 * additional probes until the loopback condition
1390 			 * becomes clear when a looped back probe is detected.
1391 			 */
1392 			log(LOG_ERR, "%s: a looped back NS message is "
1393 			    "detected during DAD for %s.  "
1394 			    "Another DAD probes are being sent.\n",
1395 			    if_name(ifa->ifa_ifp),
1396 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1397 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1398 			/*
1399 			 * Send an NS immediately and increase dad_count by
1400 			 * V_nd6_mmaxtries - 1.
1401 			 */
1402 			dp->dad_count =
1403 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1404 			nd6_dad_starttimer(dp,
1405 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1406 			    1);
1407 			goto done;
1408 		} else {
1409 			/*
1410 			 * We are done with DAD.  No NA came, no NS came.
1411 			 * No duplicate address found.  Check IFDISABLED flag
1412 			 * again in case that it is changed between the
1413 			 * beginning of this function and here.
1414 			 */
1415 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1416 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1417 
1418 			nd6log((LOG_DEBUG,
1419 			    "%s: DAD complete for %s - no duplicates found\n",
1420 			    if_name(ifa->ifa_ifp),
1421 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1422 			if (dp->dad_ns_lcount > 0)
1423 				log(LOG_ERR, "%s: DAD completed while "
1424 				    "a looped back NS message is detected "
1425 				    "during DAD for %s.\n",
1426 				    if_name(ifa->ifa_ifp),
1427 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1428 		}
1429 	}
1430 err:
1431 	nd6_dad_del(dp);
1432 done:
1433 	CURVNET_RESTORE();
1434 }
1435 
1436 static void
1437 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1438 {
1439 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1440 	struct ifnet *ifp;
1441 	char ip6buf[INET6_ADDRSTRLEN];
1442 
1443 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1444 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1445 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1446 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1447 	    dp->dad_na_icount);
1448 
1449 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1450 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1451 
1452 	ifp = ifa->ifa_ifp;
1453 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1454 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1455 	log(LOG_ERR, "%s: manual intervention required\n",
1456 	    if_name(ifp));
1457 
1458 	/*
1459 	 * If the address is a link-local address formed from an interface
1460 	 * identifier based on the hardware address which is supposed to be
1461 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1462 	 * operation on the interface SHOULD be disabled.
1463 	 * [RFC 4862, Section 5.4.5]
1464 	 */
1465 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1466 		struct in6_addr in6;
1467 
1468 		/*
1469 		 * To avoid over-reaction, we only apply this logic when we are
1470 		 * very sure that hardware addresses are supposed to be unique.
1471 		 */
1472 		switch (ifp->if_type) {
1473 		case IFT_ETHER:
1474 		case IFT_FDDI:
1475 		case IFT_ATM:
1476 		case IFT_IEEE1394:
1477 		case IFT_IEEE80211:
1478 		case IFT_INFINIBAND:
1479 			in6 = ia->ia_addr.sin6_addr;
1480 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1481 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1482 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1483 				log(LOG_ERR, "%s: possible hardware address "
1484 				    "duplication detected, disable IPv6\n",
1485 				    if_name(ifp));
1486 			}
1487 			break;
1488 		}
1489 	}
1490 }
1491 
1492 static void
1493 nd6_dad_ns_output(struct dadq *dp)
1494 {
1495 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1496 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1497 	int i;
1498 
1499 	dp->dad_ns_tcount++;
1500 	if ((ifp->if_flags & IFF_UP) == 0) {
1501 		return;
1502 	}
1503 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1504 		return;
1505 	}
1506 
1507 	dp->dad_ns_ocount++;
1508 	if (V_dad_enhanced != 0) {
1509 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1510 			dp->dad_nonce[i] = arc4random();
1511 		/*
1512 		 * XXXHRS: Note that in the case that
1513 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1514 		 * different nonces can be looped back in an unexpected
1515 		 * order.  The current implementation recognizes only
1516 		 * the latest nonce on the sender side.  Practically it
1517 		 * should work well in almost all cases.
1518 		 */
1519 	}
1520 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1521 	    (uint8_t *)&dp->dad_nonce[0]);
1522 }
1523 
1524 static void
1525 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1526 {
1527 	struct in6_ifaddr *ia;
1528 	struct ifnet *ifp;
1529 	const struct in6_addr *taddr6;
1530 	struct dadq *dp;
1531 
1532 	if (ifa == NULL)
1533 		panic("ifa == NULL in nd6_dad_ns_input");
1534 
1535 	ia = (struct in6_ifaddr *)ifa;
1536 	ifp = ifa->ifa_ifp;
1537 	taddr6 = &ia->ia_addr.sin6_addr;
1538 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1539 	if (V_dad_enhanced == 0)
1540 		ndopt_nonce = NULL;
1541 	dp = nd6_dad_find(ifa, ndopt_nonce);
1542 	if (dp == NULL)
1543 		return;
1544 
1545 	dp->dad_ns_icount++;
1546 	nd6_dad_rele(dp);
1547 }
1548 
1549 static void
1550 nd6_dad_na_input(struct ifaddr *ifa)
1551 {
1552 	struct dadq *dp;
1553 
1554 	if (ifa == NULL)
1555 		panic("ifa == NULL in nd6_dad_na_input");
1556 
1557 	dp = nd6_dad_find(ifa, NULL);
1558 	if (dp != NULL) {
1559 		dp->dad_na_icount++;
1560 		nd6_dad_rele(dp);
1561 	}
1562 }
1563