xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 3231251399f9e60fc7175b6352eb5291e9296fe8)
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 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(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 	/* RFC 6980: Nodes MUST silently ignore fragments */
141 	if(m->m_flags & M_FRAGMENTED)
142 		goto freeit;
143 
144 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
145 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
146 		rflag = 0;
147 #ifndef PULLDOWN_TEST
148 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
149 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
150 #else
151 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
152 	if (nd_ns == NULL) {
153 		ICMP6STAT_INC(icp6s_tooshort);
154 		return;
155 	}
156 #endif
157 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
158 	taddr6 = nd_ns->nd_ns_target;
159 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
160 		goto bad;
161 
162 	if (ip6->ip6_hlim != 255) {
163 		nd6log((LOG_ERR,
164 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
165 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
166 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
167 		goto bad;
168 	}
169 
170 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
171 		/* dst has to be a solicited node multicast address. */
172 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
173 		    /* don't check ifindex portion */
174 		    daddr6.s6_addr32[1] == 0 &&
175 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
176 		    daddr6.s6_addr8[12] == 0xff) {
177 			; /* good */
178 		} else {
179 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
180 			    "(wrong ip6 dst)\n"));
181 			goto bad;
182 		}
183 	} else if (!V_nd6_onlink_ns_rfc4861) {
184 		struct sockaddr_in6 src_sa6;
185 
186 		/*
187 		 * According to recent IETF discussions, it is not a good idea
188 		 * to accept a NS from an address which would not be deemed
189 		 * to be a neighbor otherwise.  This point is expected to be
190 		 * clarified in future revisions of the specification.
191 		 */
192 		bzero(&src_sa6, sizeof(src_sa6));
193 		src_sa6.sin6_family = AF_INET6;
194 		src_sa6.sin6_len = sizeof(src_sa6);
195 		src_sa6.sin6_addr = saddr6;
196 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
197 			nd6log((LOG_INFO, "nd6_ns_input: "
198 				"NS packet from non-neighbor\n"));
199 			goto bad;
200 		}
201 	}
202 
203 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
204 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
205 		goto bad;
206 	}
207 
208 	icmp6len -= sizeof(*nd_ns);
209 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
210 	if (nd6_options(&ndopts) < 0) {
211 		nd6log((LOG_INFO,
212 		    "nd6_ns_input: invalid ND option, ignored\n"));
213 		/* nd6_options have incremented stats */
214 		goto freeit;
215 	}
216 
217 	if (ndopts.nd_opts_src_lladdr) {
218 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
219 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
220 	}
221 
222 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
223 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
224 		    "(link-layer address option)\n"));
225 		goto bad;
226 	}
227 
228 	/*
229 	 * Attaching target link-layer address to the NA?
230 	 * (RFC 2461 7.2.4)
231 	 *
232 	 * NS IP dst is unicast/anycast			MUST NOT add
233 	 * NS IP dst is solicited-node multicast	MUST add
234 	 *
235 	 * In implementation, we add target link-layer address by default.
236 	 * We do not add one in MUST NOT cases.
237 	 */
238 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
239 		tlladdr = 0;
240 	else
241 		tlladdr = 1;
242 
243 	/*
244 	 * Target address (taddr6) must be either:
245 	 * (1) Valid unicast/anycast address for my receiving interface,
246 	 * (2) Unicast address for which I'm offering proxy service, or
247 	 * (3) "tentative" address on which DAD is being performed.
248 	 */
249 	/* (1) and (3) check. */
250 	if (ifp->if_carp)
251 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
252 	else
253 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
254 
255 	/* (2) check. */
256 	if (ifa == NULL) {
257 		struct sockaddr_dl rt_gateway;
258 		struct rt_addrinfo info;
259 		struct sockaddr_in6 dst6;
260 
261 		bzero(&dst6, sizeof(dst6));
262 		dst6.sin6_len = sizeof(struct sockaddr_in6);
263 		dst6.sin6_family = AF_INET6;
264 		dst6.sin6_addr = taddr6;
265 
266 		bzero(&rt_gateway, sizeof(rt_gateway));
267 		rt_gateway.sdl_len = sizeof(rt_gateway);
268 		bzero(&info, sizeof(info));
269 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
270 
271 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
272 		    0, 0, &info) == 0) {
273 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
274 			    rt_gateway.sdl_family == AF_LINK) {
275 
276 				/*
277 				 * proxy NDP for single entry
278 				 */
279 				proxydl = *SDL(&rt_gateway);
280 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
281 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
282 				if (ifa)
283 					proxy = 1;
284 			}
285 		}
286 	}
287 	if (ifa == NULL) {
288 		/*
289 		 * We've got an NS packet, and we don't have that adddress
290 		 * assigned for us.  We MUST silently ignore it.
291 		 * See RFC2461 7.2.3.
292 		 */
293 		goto freeit;
294 	}
295 	myaddr6 = *IFA_IN6(ifa);
296 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
297 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
298 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
299 		goto freeit;
300 
301 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
302 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
303 		    "(if %d, NS packet %d)\n",
304 		    ip6_sprintf(ip6bufs, &taddr6),
305 		    ifp->if_addrlen, lladdrlen - 2));
306 		goto bad;
307 	}
308 
309 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
310 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
311 		    ip6_sprintf(ip6bufs, &saddr6)));
312 		goto freeit;
313 	}
314 
315 	/*
316 	 * We have neighbor solicitation packet, with target address equals to
317 	 * one of my tentative address.
318 	 *
319 	 * src addr	how to process?
320 	 * ---		---
321 	 * multicast	of course, invalid (rejected in ip6_input)
322 	 * unicast	somebody is doing address resolution -> ignore
323 	 * unspec	dup address detection
324 	 *
325 	 * The processing is defined in RFC 2462.
326 	 */
327 	if (tentative) {
328 		/*
329 		 * If source address is unspecified address, it is for
330 		 * duplicate address detection.
331 		 *
332 		 * If not, the packet is for addess resolution;
333 		 * silently ignore it.
334 		 */
335 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
336 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
337 
338 		goto freeit;
339 	}
340 
341 	/*
342 	 * If the source address is unspecified address, entries must not
343 	 * be created or updated.
344 	 * It looks that sender is performing DAD.  Output NA toward
345 	 * all-node multicast address, to tell the sender that I'm using
346 	 * the address.
347 	 * S bit ("solicited") must be zero.
348 	 */
349 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
350 		struct in6_addr in6_all;
351 
352 		in6_all = in6addr_linklocal_allnodes;
353 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
354 			goto bad;
355 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
356 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
357 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
358 		    M_GETFIB(m));
359 		goto freeit;
360 	}
361 
362 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
363 	    ND_NEIGHBOR_SOLICIT, 0);
364 
365 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
366 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
367 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
368 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
369  freeit:
370 	if (ifa != NULL)
371 		ifa_free(ifa);
372 	m_freem(m);
373 	return;
374 
375  bad:
376 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
377 		ip6_sprintf(ip6bufs, &saddr6)));
378 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
379 		ip6_sprintf(ip6bufs, &daddr6)));
380 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
381 		ip6_sprintf(ip6bufs, &taddr6)));
382 	ICMP6STAT_INC(icp6s_badns);
383 	if (ifa != NULL)
384 		ifa_free(ifa);
385 	m_freem(m);
386 }
387 
388 /*
389  * Output a Neighbor Solicitation Message. Caller specifies:
390  *	- ICMP6 header source IP6 address
391  *	- ND6 header target IP6 address
392  *	- ND6 header source datalink address
393  *
394  * Based on RFC 2461
395  * Based on RFC 2462 (duplicate address detection)
396  *
397  *    ln - for source address determination
398  * nonce - If non-NULL, NS is used for duplicate address detection and
399  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
400  */
401 static void
402 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
403     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
404     uint8_t *nonce, u_int fibnum)
405 {
406 	struct mbuf *m;
407 	struct m_tag *mtag;
408 	struct ip6_hdr *ip6;
409 	struct nd_neighbor_solicit *nd_ns;
410 	struct ip6_moptions im6o;
411 	int icmp6len;
412 	int maxlen;
413 	caddr_t mac;
414 
415 	if (IN6_IS_ADDR_MULTICAST(taddr6))
416 		return;
417 
418 	/* estimate the size of message */
419 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
420 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
421 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
422 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
423 	    __func__, max_linkhdr, maxlen, MCLBYTES));
424 
425 	if (max_linkhdr + maxlen > MHLEN)
426 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
427 	else
428 		m = m_gethdr(M_NOWAIT, MT_DATA);
429 	if (m == NULL)
430 		return;
431 	M_SETFIB(m, fibnum);
432 
433 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
434 		m->m_flags |= M_MCAST;
435 		im6o.im6o_multicast_ifp = ifp;
436 		im6o.im6o_multicast_hlim = 255;
437 		im6o.im6o_multicast_loop = 0;
438 	}
439 
440 	icmp6len = sizeof(*nd_ns);
441 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
442 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
443 
444 	/* fill neighbor solicitation packet */
445 	ip6 = mtod(m, struct ip6_hdr *);
446 	ip6->ip6_flow = 0;
447 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
448 	ip6->ip6_vfc |= IPV6_VERSION;
449 	/* ip6->ip6_plen will be set later */
450 	ip6->ip6_nxt = IPPROTO_ICMPV6;
451 	ip6->ip6_hlim = 255;
452 	if (daddr6)
453 		ip6->ip6_dst = *daddr6;
454 	else {
455 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
456 		ip6->ip6_dst.s6_addr16[1] = 0;
457 		ip6->ip6_dst.s6_addr32[1] = 0;
458 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
459 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
460 		ip6->ip6_dst.s6_addr8[12] = 0xff;
461 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
462 			goto bad;
463 	}
464 	if (nonce == NULL) {
465 		struct ifaddr *ifa = NULL;
466 
467 		/*
468 		 * RFC2461 7.2.2:
469 		 * "If the source address of the packet prompting the
470 		 * solicitation is the same as one of the addresses assigned
471 		 * to the outgoing interface, that address SHOULD be placed
472 		 * in the IP Source Address of the outgoing solicitation.
473 		 * Otherwise, any one of the addresses assigned to the
474 		 * interface should be used."
475 		 *
476 		 * We use the source address for the prompting packet
477 		 * (saddr6), if saddr6 belongs to the outgoing interface.
478 		 * Otherwise, we perform the source address selection as usual.
479 		 */
480 
481 		if (saddr6 != NULL)
482 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
483 		if (ifa != NULL) {
484 			/* ip6_src set already. */
485 			ip6->ip6_src = *saddr6;
486 			ifa_free(ifa);
487 		} else {
488 			int error;
489 			struct in6_addr dst6, src6;
490 			uint32_t scopeid;
491 
492 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
493 			error = in6_selectsrc_addr(fibnum, &dst6,
494 			    scopeid, ifp, &src6, NULL);
495 			if (error) {
496 				char ip6buf[INET6_ADDRSTRLEN];
497 				nd6log((LOG_DEBUG, "%s: source can't be "
498 				    "determined: dst=%s, error=%d\n", __func__,
499 				    ip6_sprintf(ip6buf, &dst6),
500 				    error));
501 				goto bad;
502 			}
503 			ip6->ip6_src = src6;
504 		}
505 	} else {
506 		/*
507 		 * Source address for DAD packet must always be IPv6
508 		 * unspecified address. (0::0)
509 		 * We actually don't have to 0-clear the address (we did it
510 		 * above), but we do so here explicitly to make the intention
511 		 * clearer.
512 		 */
513 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
514 	}
515 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
516 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
517 	nd_ns->nd_ns_code = 0;
518 	nd_ns->nd_ns_reserved = 0;
519 	nd_ns->nd_ns_target = *taddr6;
520 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
521 
522 	/*
523 	 * Add source link-layer address option.
524 	 *
525 	 *				spec		implementation
526 	 *				---		---
527 	 * DAD packet			MUST NOT	do not add the option
528 	 * there's no link layer address:
529 	 *				impossible	do not add the option
530 	 * there's link layer address:
531 	 *	Multicast NS		MUST add one	add the option
532 	 *	Unicast NS		SHOULD add one	add the option
533 	 */
534 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
535 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
536 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
537 		/* 8 byte alignments... */
538 		optlen = (optlen + 7) & ~7;
539 
540 		m->m_pkthdr.len += optlen;
541 		m->m_len += optlen;
542 		icmp6len += optlen;
543 		bzero((caddr_t)nd_opt, optlen);
544 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
545 		nd_opt->nd_opt_len = optlen >> 3;
546 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
547 	}
548 	/*
549 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
550 	 * This behavior is documented as Enhanced Duplicate Address
551 	 * Detection in RFC 7527.
552 	 * net.inet6.ip6.dad_enhanced=0 disables this.
553 	 */
554 	if (V_dad_enhanced != 0 && nonce != NULL) {
555 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
556 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
557 		/* 8-byte alignment is required. */
558 		optlen = (optlen + 7) & ~7;
559 
560 		m->m_pkthdr.len += optlen;
561 		m->m_len += optlen;
562 		icmp6len += optlen;
563 		bzero((caddr_t)nd_opt, optlen);
564 		nd_opt->nd_opt_type = ND_OPT_NONCE;
565 		nd_opt->nd_opt_len = optlen >> 3;
566 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
567 	}
568 	ip6->ip6_plen = htons((u_short)icmp6len);
569 	nd_ns->nd_ns_cksum = 0;
570 	nd_ns->nd_ns_cksum =
571 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
572 
573 	if (send_sendso_input_hook != NULL) {
574 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
575 			sizeof(unsigned short), M_NOWAIT);
576 		if (mtag == NULL)
577 			goto bad;
578 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
579 		m_tag_prepend(m, mtag);
580 	}
581 
582 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
583 	    &im6o, NULL, NULL);
584 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
585 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
586 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
587 
588 	return;
589 
590   bad:
591 	m_freem(m);
592 }
593 
594 #ifndef BURN_BRIDGES
595 void
596 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
597     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
598 {
599 
600 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
601 }
602 #endif
603 /*
604  * Neighbor advertisement input handling.
605  *
606  * Based on RFC 2461
607  * Based on RFC 2462 (duplicate address detection)
608  *
609  * the following items are not implemented yet:
610  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
611  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
612  */
613 void
614 nd6_na_input(struct mbuf *m, int off, int icmp6len)
615 {
616 	struct epoch_tracker et;
617 	struct ifnet *ifp = m->m_pkthdr.rcvif;
618 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
619 	struct nd_neighbor_advert *nd_na;
620 	struct in6_addr daddr6 = ip6->ip6_dst;
621 	struct in6_addr taddr6;
622 	int flags;
623 	int is_router;
624 	int is_solicited;
625 	int is_override;
626 	char *lladdr = NULL;
627 	int lladdrlen = 0;
628 	int checklink = 0;
629 	struct ifaddr *ifa;
630 	struct llentry *ln = NULL;
631 	union nd_opts ndopts;
632 	struct mbuf *chain = NULL;
633 	struct sockaddr_in6 sin6;
634 	u_char linkhdr[LLE_MAX_LINKHDR];
635 	size_t linkhdrsize;
636 	int lladdr_off;
637 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
638 
639 	/* RFC 6980: Nodes MUST silently ignore fragments */
640 	if(m->m_flags & M_FRAGMENTED)
641 		goto freeit;
642 
643 	if (ip6->ip6_hlim != 255) {
644 		nd6log((LOG_ERR,
645 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
646 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
647 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
648 		goto bad;
649 	}
650 
651 #ifndef PULLDOWN_TEST
652 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
653 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
654 #else
655 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
656 	if (nd_na == NULL) {
657 		ICMP6STAT_INC(icp6s_tooshort);
658 		return;
659 	}
660 #endif
661 
662 	flags = nd_na->nd_na_flags_reserved;
663 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
664 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
665 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
666 	memset(&sin6, 0, sizeof(sin6));
667 
668 	taddr6 = nd_na->nd_na_target;
669 	if (in6_setscope(&taddr6, ifp, NULL))
670 		goto bad;	/* XXX: impossible */
671 
672 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
673 		nd6log((LOG_ERR,
674 		    "nd6_na_input: invalid target address %s\n",
675 		    ip6_sprintf(ip6bufs, &taddr6)));
676 		goto bad;
677 	}
678 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
679 		if (is_solicited) {
680 			nd6log((LOG_ERR,
681 			    "nd6_na_input: a solicited adv is multicasted\n"));
682 			goto bad;
683 		}
684 
685 	icmp6len -= sizeof(*nd_na);
686 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
687 	if (nd6_options(&ndopts) < 0) {
688 		nd6log((LOG_INFO,
689 		    "nd6_na_input: invalid ND option, ignored\n"));
690 		/* nd6_options have incremented stats */
691 		goto freeit;
692 	}
693 
694 	if (ndopts.nd_opts_tgt_lladdr) {
695 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
696 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
697 	}
698 
699 	/*
700 	 * This effectively disables the DAD check on a non-master CARP
701 	 * address.
702 	 */
703 	if (ifp->if_carp)
704 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
705 	else
706 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
707 
708 	/*
709 	 * Target address matches one of my interface address.
710 	 *
711 	 * If my address is tentative, this means that there's somebody
712 	 * already using the same address as mine.  This indicates DAD failure.
713 	 * This is defined in RFC 2462.
714 	 *
715 	 * Otherwise, process as defined in RFC 2461.
716 	 */
717 	if (ifa
718 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
719 		nd6_dad_na_input(ifa);
720 		ifa_free(ifa);
721 		goto freeit;
722 	}
723 
724 	/* Just for safety, maybe unnecessary. */
725 	if (ifa) {
726 		ifa_free(ifa);
727 		log(LOG_ERR,
728 		    "nd6_na_input: duplicate IP6 address %s\n",
729 		    ip6_sprintf(ip6bufs, &taddr6));
730 		goto freeit;
731 	}
732 
733 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
734 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
735 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
736 		    ifp->if_addrlen, lladdrlen - 2));
737 		goto bad;
738 	}
739 
740 	/*
741 	 * If no neighbor cache entry is found, NA SHOULD silently be
742 	 * discarded.
743 	 */
744 	NET_EPOCH_ENTER(et);
745 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
746 	NET_EPOCH_EXIT(et);
747 	if (ln == NULL) {
748 		goto freeit;
749 	}
750 
751 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
752 		/*
753 		 * If the link-layer has address, and no lladdr option came,
754 		 * discard the packet.
755 		 */
756 		if (ifp->if_addrlen && lladdr == NULL) {
757 			goto freeit;
758 		}
759 
760 		/*
761 		 * Record link-layer address, and update the state.
762 		 */
763 		linkhdrsize = sizeof(linkhdr);
764 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
765 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
766 			return;
767 
768 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
769 		    lladdr_off) == 0) {
770 			ln = NULL;
771 			goto freeit;
772 		}
773 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
774 		if (is_solicited)
775 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
776 		else
777 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
778 		if ((ln->ln_router = is_router) != 0) {
779 			/*
780 			 * This means a router's state has changed from
781 			 * non-reachable to probably reachable, and might
782 			 * affect the status of associated prefixes..
783 			 */
784 			checklink = 1;
785 		}
786 	} else {
787 		int llchange;
788 
789 		/*
790 		 * Check if the link-layer address has changed or not.
791 		 */
792 		if (lladdr == NULL)
793 			llchange = 0;
794 		else {
795 			if (ln->la_flags & LLE_VALID) {
796 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
797 					llchange = 1;
798 				else
799 					llchange = 0;
800 			} else
801 				llchange = 1;
802 		}
803 
804 		/*
805 		 * This is VERY complex.  Look at it with care.
806 		 *
807 		 * override solicit lladdr llchange	action
808 		 *					(L: record lladdr)
809 		 *
810 		 *	0	0	n	--	(2c)
811 		 *	0	0	y	n	(2b) L
812 		 *	0	0	y	y	(1)    REACHABLE->STALE
813 		 *	0	1	n	--	(2c)   *->REACHABLE
814 		 *	0	1	y	n	(2b) L *->REACHABLE
815 		 *	0	1	y	y	(1)    REACHABLE->STALE
816 		 *	1	0	n	--	(2a)
817 		 *	1	0	y	n	(2a) L
818 		 *	1	0	y	y	(2a) L *->STALE
819 		 *	1	1	n	--	(2a)   *->REACHABLE
820 		 *	1	1	y	n	(2a) L *->REACHABLE
821 		 *	1	1	y	y	(2a) L *->REACHABLE
822 		 */
823 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
824 			/*
825 			 * If state is REACHABLE, make it STALE.
826 			 * no other updates should be done.
827 			 */
828 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
829 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
830 			goto freeit;
831 		} else if (is_override				   /* (2a) */
832 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
833 			|| lladdr == NULL) {			   /* (2c) */
834 			/*
835 			 * Update link-local address, if any.
836 			 */
837 			if (lladdr != NULL) {
838 				linkhdrsize = sizeof(linkhdr);
839 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
840 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
841 					goto freeit;
842 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
843 				    linkhdrsize, lladdr_off) == 0) {
844 					ln = NULL;
845 					goto freeit;
846 				}
847 				EVENTHANDLER_INVOKE(lle_event, ln,
848 				    LLENTRY_RESOLVED);
849 			}
850 
851 			/*
852 			 * If solicited, make the state REACHABLE.
853 			 * If not solicited and the link-layer address was
854 			 * changed, make it STALE.
855 			 */
856 			if (is_solicited)
857 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
858 			else {
859 				if (lladdr != NULL && llchange)
860 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
861 			}
862 		}
863 
864 		if (ln->ln_router && !is_router) {
865 			/*
866 			 * The peer dropped the router flag.
867 			 * Remove the sender from the Default Router List and
868 			 * update the Destination Cache entries.
869 			 */
870 			struct ifnet *nd6_ifp;
871 
872 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
873 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
874 			    (ND_IFINFO(nd6_ifp)->flags &
875 			     ND6_IFF_ACCEPT_RTADV) != 0)
876 				/*
877 				 * Even if the neighbor is not in the default
878 				 * router list, the neighbor may be used as a
879 				 * next hop for some destinations (e.g. redirect
880 				 * case). So we must call rt6_flush explicitly.
881 				 */
882 				rt6_flush(&ip6->ip6_src, ifp);
883 		}
884 		ln->ln_router = is_router;
885 	}
886         /* XXX - QL
887 	 *  Does this matter?
888 	 *  rt->rt_flags &= ~RTF_REJECT;
889 	 */
890 	ln->la_asked = 0;
891 	if (ln->la_hold != NULL)
892 		nd6_grab_holdchain(ln, &chain, &sin6);
893  freeit:
894 	if (ln != NULL)
895 		LLE_WUNLOCK(ln);
896 
897 	if (chain != NULL)
898 		nd6_flush_holdchain(ifp, chain, &sin6);
899 
900 	if (checklink)
901 		pfxlist_onlink_check();
902 
903 	m_freem(m);
904 	return;
905 
906  bad:
907 	if (ln != NULL)
908 		LLE_WUNLOCK(ln);
909 
910 	ICMP6STAT_INC(icp6s_badna);
911 	m_freem(m);
912 }
913 
914 /*
915  * Neighbor advertisement output handling.
916  *
917  * Based on RFC 2461
918  *
919  * the following items are not implemented yet:
920  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
921  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
922  *
923  * tlladdr - 1 if include target link-layer address
924  * sdl0 - sockaddr_dl (= proxy NA) or NULL
925  */
926 static void
927 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
928     const struct in6_addr *taddr6, u_long flags, int tlladdr,
929     struct sockaddr *sdl0, u_int fibnum)
930 {
931 	struct mbuf *m;
932 	struct m_tag *mtag;
933 	struct ip6_hdr *ip6;
934 	struct nd_neighbor_advert *nd_na;
935 	struct ip6_moptions im6o;
936 	struct in6_addr daddr6, dst6, src6;
937 	uint32_t scopeid;
938 
939 	int icmp6len, maxlen, error;
940 	caddr_t mac = NULL;
941 
942 	daddr6 = *daddr6_0;	/* make a local copy for modification */
943 
944 	/* estimate the size of message */
945 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
946 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
947 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
948 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
949 	    __func__, max_linkhdr, maxlen, MCLBYTES));
950 
951 	if (max_linkhdr + maxlen > MHLEN)
952 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
953 	else
954 		m = m_gethdr(M_NOWAIT, MT_DATA);
955 	if (m == NULL)
956 		return;
957 	M_SETFIB(m, fibnum);
958 
959 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
960 		m->m_flags |= M_MCAST;
961 		im6o.im6o_multicast_ifp = ifp;
962 		im6o.im6o_multicast_hlim = 255;
963 		im6o.im6o_multicast_loop = 0;
964 	}
965 
966 	icmp6len = sizeof(*nd_na);
967 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
968 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
969 
970 	/* fill neighbor advertisement packet */
971 	ip6 = mtod(m, struct ip6_hdr *);
972 	ip6->ip6_flow = 0;
973 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
974 	ip6->ip6_vfc |= IPV6_VERSION;
975 	ip6->ip6_nxt = IPPROTO_ICMPV6;
976 	ip6->ip6_hlim = 255;
977 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
978 		/* reply to DAD */
979 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
980 		daddr6.s6_addr16[1] = 0;
981 		daddr6.s6_addr32[1] = 0;
982 		daddr6.s6_addr32[2] = 0;
983 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
984 		if (in6_setscope(&daddr6, ifp, NULL))
985 			goto bad;
986 
987 		flags &= ~ND_NA_FLAG_SOLICITED;
988 	}
989 	ip6->ip6_dst = daddr6;
990 
991 	/*
992 	 * Select a source whose scope is the same as that of the dest.
993 	 */
994 	in6_splitscope(&daddr6, &dst6, &scopeid);
995 	error = in6_selectsrc_addr(fibnum, &dst6,
996 	    scopeid, ifp, &src6, NULL);
997 	if (error) {
998 		char ip6buf[INET6_ADDRSTRLEN];
999 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1000 		    "determined: dst=%s, error=%d\n",
1001 		    ip6_sprintf(ip6buf, &daddr6), error));
1002 		goto bad;
1003 	}
1004 	ip6->ip6_src = src6;
1005 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1006 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1007 	nd_na->nd_na_code = 0;
1008 	nd_na->nd_na_target = *taddr6;
1009 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1010 
1011 	/*
1012 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1013 	 * see nd6_ns_input() for details.
1014 	 * Basically, if NS packet is sent to unicast/anycast addr,
1015 	 * target lladdr option SHOULD NOT be included.
1016 	 */
1017 	if (tlladdr) {
1018 		/*
1019 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1020 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1021 		 * my address) use lladdr configured for the interface.
1022 		 */
1023 		if (sdl0 == NULL) {
1024 			if (ifp->if_carp)
1025 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1026 			if (mac == NULL)
1027 				mac = nd6_ifptomac(ifp);
1028 		} else if (sdl0->sa_family == AF_LINK) {
1029 			struct sockaddr_dl *sdl;
1030 			sdl = (struct sockaddr_dl *)sdl0;
1031 			if (sdl->sdl_alen == ifp->if_addrlen)
1032 				mac = LLADDR(sdl);
1033 		}
1034 	}
1035 	if (tlladdr && mac) {
1036 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1037 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1038 
1039 		/* roundup to 8 bytes alignment! */
1040 		optlen = (optlen + 7) & ~7;
1041 
1042 		m->m_pkthdr.len += optlen;
1043 		m->m_len += optlen;
1044 		icmp6len += optlen;
1045 		bzero((caddr_t)nd_opt, optlen);
1046 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1047 		nd_opt->nd_opt_len = optlen >> 3;
1048 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1049 	} else
1050 		flags &= ~ND_NA_FLAG_OVERRIDE;
1051 
1052 	ip6->ip6_plen = htons((u_short)icmp6len);
1053 	nd_na->nd_na_flags_reserved = flags;
1054 	nd_na->nd_na_cksum = 0;
1055 	nd_na->nd_na_cksum =
1056 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1057 
1058 	if (send_sendso_input_hook != NULL) {
1059 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1060 		    sizeof(unsigned short), M_NOWAIT);
1061 		if (mtag == NULL)
1062 			goto bad;
1063 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1064 		m_tag_prepend(m, mtag);
1065 	}
1066 
1067 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1068 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1069 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1070 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1071 
1072 	return;
1073 
1074   bad:
1075 	m_freem(m);
1076 }
1077 
1078 #ifndef BURN_BRIDGES
1079 void
1080 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1081     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1082     struct sockaddr *sdl0)
1083 {
1084 
1085 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1086 	    RT_DEFAULT_FIB);
1087 }
1088 #endif
1089 
1090 caddr_t
1091 nd6_ifptomac(struct ifnet *ifp)
1092 {
1093 	switch (ifp->if_type) {
1094 	case IFT_ETHER:
1095 	case IFT_IEEE1394:
1096 	case IFT_L2VLAN:
1097 	case IFT_INFINIBAND:
1098 	case IFT_BRIDGE:
1099 		return IF_LLADDR(ifp);
1100 	default:
1101 		return NULL;
1102 	}
1103 }
1104 
1105 struct dadq {
1106 	TAILQ_ENTRY(dadq) dad_list;
1107 	struct ifaddr *dad_ifa;
1108 	int dad_count;		/* max NS to send */
1109 	int dad_ns_tcount;	/* # of trials to send NS */
1110 	int dad_ns_ocount;	/* NS sent so far */
1111 	int dad_ns_icount;
1112 	int dad_na_icount;
1113 	int dad_ns_lcount;	/* looped back NS */
1114 	int dad_loopbackprobe;	/* probing state for loopback detection */
1115 	struct callout dad_timer_ch;
1116 	struct vnet *dad_vnet;
1117 	u_int dad_refcnt;
1118 #define	ND_OPT_NONCE_LEN32 \
1119 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1120 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1121 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1122 };
1123 
1124 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1125 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1126 #define	V_dadq			VNET(dadq)
1127 #define	V_dad_rwlock		VNET(dad_rwlock)
1128 
1129 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1130 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1131 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1132 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1133 
1134 static void
1135 nd6_dad_add(struct dadq *dp)
1136 {
1137 
1138 	DADQ_WLOCK();
1139 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1140 	dp->dad_ondadq = true;
1141 	DADQ_WUNLOCK();
1142 }
1143 
1144 static void
1145 nd6_dad_del(struct dadq *dp)
1146 {
1147 
1148 	DADQ_WLOCK();
1149 	if (dp->dad_ondadq) {
1150 		/*
1151 		 * Remove dp from the dadq and release the dadq's
1152 		 * reference.
1153 		 */
1154 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1155 		dp->dad_ondadq = false;
1156 		DADQ_WUNLOCK();
1157 		nd6_dad_rele(dp);
1158 	} else
1159 		DADQ_WUNLOCK();
1160 }
1161 
1162 static struct dadq *
1163 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1164 {
1165 	struct dadq *dp;
1166 
1167 	DADQ_RLOCK();
1168 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1169 		if (dp->dad_ifa != ifa)
1170 			continue;
1171 		/*
1172 		 * Skip if the nonce matches the received one.
1173 		 * +2 in the length is required because of type and
1174 		 * length fields are included in a header.
1175 		 */
1176 		if (n != NULL &&
1177 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1178 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1179 		        ND_OPT_NONCE_LEN) == 0) {
1180 			dp->dad_ns_lcount++;
1181 			continue;
1182 		}
1183 		refcount_acquire(&dp->dad_refcnt);
1184 		break;
1185 	}
1186 	DADQ_RUNLOCK();
1187 
1188 	return (dp);
1189 }
1190 
1191 static void
1192 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1193 {
1194 
1195 	if (send_ns != 0)
1196 		nd6_dad_ns_output(dp);
1197 	callout_reset(&dp->dad_timer_ch, ticks,
1198 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1199 }
1200 
1201 static void
1202 nd6_dad_stoptimer(struct dadq *dp)
1203 {
1204 
1205 	callout_drain(&dp->dad_timer_ch);
1206 }
1207 
1208 static void
1209 nd6_dad_rele(struct dadq *dp)
1210 {
1211 
1212 	if (refcount_release(&dp->dad_refcnt)) {
1213 		ifa_free(dp->dad_ifa);
1214 		free(dp, M_IP6NDP);
1215 	}
1216 }
1217 
1218 void
1219 nd6_dad_init(void)
1220 {
1221 
1222 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1223 	TAILQ_INIT(&V_dadq);
1224 }
1225 
1226 /*
1227  * Start Duplicate Address Detection (DAD) for specified interface address.
1228  */
1229 void
1230 nd6_dad_start(struct ifaddr *ifa, int delay)
1231 {
1232 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1233 	struct dadq *dp;
1234 	char ip6buf[INET6_ADDRSTRLEN];
1235 
1236 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1237 	    ("starting DAD on non-tentative address %p", ifa));
1238 
1239 	/*
1240 	 * If we don't need DAD, don't do it.
1241 	 * There are several cases:
1242 	 * - DAD is disabled globally or on the interface
1243 	 * - the interface address is anycast
1244 	 */
1245 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1246 	    V_ip6_dad_count == 0 ||
1247 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1248 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1249 		return;
1250 	}
1251 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1252 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1253 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1254 		return;
1255 
1256 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1257 		/*
1258 		 * DAD is already in progress.  Let the existing entry
1259 		 * finish it.
1260 		 */
1261 		nd6_dad_rele(dp);
1262 		return;
1263 	}
1264 
1265 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1266 	if (dp == NULL) {
1267 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1268 			"%s(%s)\n",
1269 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1270 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1271 		return;
1272 	}
1273 	callout_init(&dp->dad_timer_ch, 0);
1274 #ifdef VIMAGE
1275 	dp->dad_vnet = curvnet;
1276 #endif
1277 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1278 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1279 
1280 	/*
1281 	 * Send NS packet for DAD, ip6_dad_count times.
1282 	 * Note that we must delay the first transmission, if this is the
1283 	 * first packet to be sent from the interface after interface
1284 	 * (re)initialization.
1285 	 */
1286 	dp->dad_ifa = ifa;
1287 	ifa_ref(dp->dad_ifa);
1288 	dp->dad_count = V_ip6_dad_count;
1289 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1290 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1291 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1292 
1293 	/* Add this to the dadq and add a reference for the dadq. */
1294 	refcount_init(&dp->dad_refcnt, 1);
1295 	nd6_dad_add(dp);
1296 	nd6_dad_starttimer(dp, delay, 0);
1297 }
1298 
1299 /*
1300  * terminate DAD unconditionally.  used for address removals.
1301  */
1302 void
1303 nd6_dad_stop(struct ifaddr *ifa)
1304 {
1305 	struct dadq *dp;
1306 
1307 	dp = nd6_dad_find(ifa, NULL);
1308 	if (!dp) {
1309 		/* DAD wasn't started yet */
1310 		return;
1311 	}
1312 
1313 	nd6_dad_stoptimer(dp);
1314 	nd6_dad_del(dp);
1315 
1316 	/* Release this function's reference, acquired by nd6_dad_find(). */
1317 	nd6_dad_rele(dp);
1318 }
1319 
1320 static void
1321 nd6_dad_timer(struct dadq *dp)
1322 {
1323 	CURVNET_SET(dp->dad_vnet);
1324 	struct ifaddr *ifa = dp->dad_ifa;
1325 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1326 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1327 	char ip6buf[INET6_ADDRSTRLEN];
1328 
1329 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1330 
1331 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1332 		/* Do not need DAD for ifdisabled interface. */
1333 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1334 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1335 		goto err;
1336 	}
1337 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1338 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1339 			"%s(%s)\n",
1340 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1341 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1342 		goto err;
1343 	}
1344 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1345 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1346 			"%s(%s)\n",
1347 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1348 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1349 		goto err;
1350 	}
1351 
1352 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1353 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1354 	    (((ifp->if_flags & IFF_UP) == 0) ||
1355 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1356 		nd6log((LOG_INFO, "%s: could not run DAD "
1357 		    "because the interface was down or not running.\n",
1358 		    if_name(ifa->ifa_ifp)));
1359 		goto err;
1360 	}
1361 
1362 	/* Need more checks? */
1363 	if (dp->dad_ns_ocount < dp->dad_count) {
1364 		/*
1365 		 * We have more NS to go.  Send NS packet for DAD.
1366 		 */
1367 		nd6_dad_starttimer(dp,
1368 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1369 		goto done;
1370 	} else {
1371 		/*
1372 		 * We have transmitted sufficient number of DAD packets.
1373 		 * See what we've got.
1374 		 */
1375 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1376 			/* We've seen NS or NA, means DAD has failed. */
1377 			nd6_dad_duplicated(ifa, dp);
1378 		else if (V_dad_enhanced != 0 &&
1379 		    dp->dad_ns_lcount > 0 &&
1380 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1381 			/*
1382 			 * Sec. 4.1 in RFC 7527 requires transmission of
1383 			 * additional probes until the loopback condition
1384 			 * becomes clear when a looped back probe is detected.
1385 			 */
1386 			log(LOG_ERR, "%s: a looped back NS message is "
1387 			    "detected during DAD for %s.  "
1388 			    "Another DAD probes are being sent.\n",
1389 			    if_name(ifa->ifa_ifp),
1390 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1391 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1392 			/*
1393 			 * Send an NS immediately and increase dad_count by
1394 			 * V_nd6_mmaxtries - 1.
1395 			 */
1396 			dp->dad_count =
1397 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1398 			nd6_dad_starttimer(dp,
1399 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1400 			    1);
1401 			goto done;
1402 		} else {
1403 			/*
1404 			 * We are done with DAD.  No NA came, no NS came.
1405 			 * No duplicate address found.  Check IFDISABLED flag
1406 			 * again in case that it is changed between the
1407 			 * beginning of this function and here.
1408 			 */
1409 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1410 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1411 
1412 			nd6log((LOG_DEBUG,
1413 			    "%s: DAD complete for %s - no duplicates found\n",
1414 			    if_name(ifa->ifa_ifp),
1415 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1416 			if (dp->dad_ns_lcount > 0)
1417 				log(LOG_ERR, "%s: DAD completed while "
1418 				    "a looped back NS message is detected "
1419 				    "during DAD for %s.\n",
1420 				    if_name(ifa->ifa_ifp),
1421 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1422 		}
1423 	}
1424 err:
1425 	nd6_dad_del(dp);
1426 done:
1427 	CURVNET_RESTORE();
1428 }
1429 
1430 static void
1431 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1432 {
1433 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1434 	struct ifnet *ifp;
1435 	char ip6buf[INET6_ADDRSTRLEN];
1436 
1437 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1438 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1439 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1440 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1441 	    dp->dad_na_icount);
1442 
1443 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1444 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1445 
1446 	ifp = ifa->ifa_ifp;
1447 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1448 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1449 	log(LOG_ERR, "%s: manual intervention required\n",
1450 	    if_name(ifp));
1451 
1452 	/*
1453 	 * If the address is a link-local address formed from an interface
1454 	 * identifier based on the hardware address which is supposed to be
1455 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1456 	 * operation on the interface SHOULD be disabled.
1457 	 * [RFC 4862, Section 5.4.5]
1458 	 */
1459 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1460 		struct in6_addr in6;
1461 
1462 		/*
1463 		 * To avoid over-reaction, we only apply this logic when we are
1464 		 * very sure that hardware addresses are supposed to be unique.
1465 		 */
1466 		switch (ifp->if_type) {
1467 		case IFT_ETHER:
1468 		case IFT_ATM:
1469 		case IFT_IEEE1394:
1470 		case IFT_INFINIBAND:
1471 			in6 = ia->ia_addr.sin6_addr;
1472 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1473 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1474 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1475 				log(LOG_ERR, "%s: possible hardware address "
1476 				    "duplication detected, disable IPv6\n",
1477 				    if_name(ifp));
1478 			}
1479 			break;
1480 		}
1481 	}
1482 }
1483 
1484 static void
1485 nd6_dad_ns_output(struct dadq *dp)
1486 {
1487 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1488 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1489 	int i;
1490 
1491 	dp->dad_ns_tcount++;
1492 	if ((ifp->if_flags & IFF_UP) == 0) {
1493 		return;
1494 	}
1495 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1496 		return;
1497 	}
1498 
1499 	dp->dad_ns_ocount++;
1500 	if (V_dad_enhanced != 0) {
1501 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1502 			dp->dad_nonce[i] = arc4random();
1503 		/*
1504 		 * XXXHRS: Note that in the case that
1505 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1506 		 * different nonces can be looped back in an unexpected
1507 		 * order.  The current implementation recognizes only
1508 		 * the latest nonce on the sender side.  Practically it
1509 		 * should work well in almost all cases.
1510 		 */
1511 	}
1512 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1513 	    (uint8_t *)&dp->dad_nonce[0]);
1514 }
1515 
1516 static void
1517 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1518 {
1519 	struct dadq *dp;
1520 
1521 	if (ifa == NULL)
1522 		panic("ifa == NULL in nd6_dad_ns_input");
1523 
1524 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1525 	if (V_dad_enhanced == 0)
1526 		ndopt_nonce = NULL;
1527 	dp = nd6_dad_find(ifa, ndopt_nonce);
1528 	if (dp == NULL)
1529 		return;
1530 
1531 	dp->dad_ns_icount++;
1532 	nd6_dad_rele(dp);
1533 }
1534 
1535 static void
1536 nd6_dad_na_input(struct ifaddr *ifa)
1537 {
1538 	struct dadq *dp;
1539 
1540 	if (ifa == NULL)
1541 		panic("ifa == NULL in nd6_dad_na_input");
1542 
1543 	dp = nd6_dad_find(ifa, NULL);
1544 	if (dp != NULL) {
1545 		dp->dad_na_icount++;
1546 		nd6_dad_rele(dp);
1547 	}
1548 }
1549