xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 369d70ec9282cccf131ab0400b05dad13ba8116d)
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 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/eventhandler.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 #include <net/vnet.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <net/if_llatbl.h>
70 #include <netinet6/in6_var.h>
71 #include <netinet6/in6_ifattach.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #include <netinet6/nd6.h>
76 #include <netinet/icmp6.h>
77 #include <netinet/ip_carp.h>
78 #include <netinet6/send.h>
79 
80 #define SDL(s) ((struct sockaddr_dl *)s)
81 
82 struct dadq;
83 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
84 static void nd6_dad_add(struct dadq *dp);
85 static void nd6_dad_del(struct dadq *dp);
86 static void nd6_dad_rele(struct dadq *);
87 static void nd6_dad_starttimer(struct dadq *, int, int);
88 static void nd6_dad_stoptimer(struct dadq *);
89 static void nd6_dad_timer(struct dadq *);
90 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
91 static void nd6_dad_ns_output(struct dadq *);
92 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
93 static void nd6_dad_na_input(struct ifaddr *);
94 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
95     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
96 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
97     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
98 
99 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
100 #define	V_dad_enhanced			VNET(dad_enhanced)
101 
102 SYSCTL_DECL(_net_inet6_ip6);
103 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
104     &VNET_NAME(dad_enhanced), 0,
105     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
106 
107 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
108 						   transmit DAD packet */
109 #define	V_dad_maxtry			VNET(dad_maxtry)
110 
111 /*
112  * Input a Neighbor Solicitation Message.
113  *
114  * Based on RFC 2461
115  * Based on RFC 2462 (duplicate address detection)
116  */
117 void
118 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
119 {
120 	struct ifnet *ifp;
121 	struct ip6_hdr *ip6;
122 	struct nd_neighbor_solicit *nd_ns;
123 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
124 	struct ifaddr *ifa;
125 	struct sockaddr_dl proxydl;
126 	union nd_opts ndopts;
127 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
128 	char *lladdr;
129 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
130 
131 	ifa = NULL;
132 
133 	/* RFC 6980: Nodes MUST silently ignore fragments */
134 	if(m->m_flags & M_FRAGMENTED)
135 		goto freeit;
136 
137 	ifp = m->m_pkthdr.rcvif;
138 	ip6 = mtod(m, struct ip6_hdr *);
139 	if (__predict_false(ip6->ip6_hlim != 255)) {
140 		ICMP6STAT_INC(icp6s_invlhlim);
141 		nd6log((LOG_ERR,
142 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
143 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
144 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
145 		goto bads;
146 	}
147 
148 	if (m->m_len < off + icmp6len) {
149 		m = m_pullup(m, off + icmp6len);
150 		if (m == NULL) {
151 			IP6STAT_INC(ip6s_exthdrtoolong);
152 			return;
153 		}
154 	}
155 	ip6 = mtod(m, struct ip6_hdr *);
156 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
157 
158 	saddr6 = ip6->ip6_src;
159 	daddr6 = ip6->ip6_dst;
160 	taddr6 = nd_ns->nd_ns_target;
161 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
162 		goto bad;
163 
164 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
165 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
166 		rflag = 0;
167 
168 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
169 		/* dst has to be a solicited node multicast address. */
170 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
171 		    /* don't check ifindex portion */
172 		    daddr6.s6_addr32[1] == 0 &&
173 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
174 		    daddr6.s6_addr8[12] == 0xff) {
175 			; /* good */
176 		} else {
177 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
178 			    "(wrong ip6 dst)\n"));
179 			goto bad;
180 		}
181 	} else if (!V_nd6_onlink_ns_rfc4861) {
182 		struct sockaddr_in6 src_sa6;
183 
184 		/*
185 		 * According to recent IETF discussions, it is not a good idea
186 		 * to accept a NS from an address which would not be deemed
187 		 * to be a neighbor otherwise.  This point is expected to be
188 		 * clarified in future revisions of the specification.
189 		 */
190 		bzero(&src_sa6, sizeof(src_sa6));
191 		src_sa6.sin6_family = AF_INET6;
192 		src_sa6.sin6_len = sizeof(src_sa6);
193 		src_sa6.sin6_addr = saddr6;
194 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
195 			nd6log((LOG_INFO, "nd6_ns_input: "
196 				"NS packet from non-neighbor\n"));
197 			goto bad;
198 		}
199 	}
200 
201 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
202 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
203 		goto bad;
204 	}
205 
206 	icmp6len -= sizeof(*nd_ns);
207 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
208 	if (nd6_options(&ndopts) < 0) {
209 		nd6log((LOG_INFO,
210 		    "nd6_ns_input: invalid ND option, ignored\n"));
211 		/* nd6_options have incremented stats */
212 		goto freeit;
213 	}
214 
215 	lladdr = NULL;
216 	lladdrlen = 0;
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 	proxy = 0;
257 	if (ifa == NULL) {
258 		struct sockaddr_dl rt_gateway;
259 		struct rt_addrinfo info;
260 		struct sockaddr_in6 dst6;
261 
262 		bzero(&dst6, sizeof(dst6));
263 		dst6.sin6_len = sizeof(struct sockaddr_in6);
264 		dst6.sin6_family = AF_INET6;
265 		dst6.sin6_addr = taddr6;
266 
267 		bzero(&rt_gateway, sizeof(rt_gateway));
268 		rt_gateway.sdl_len = sizeof(rt_gateway);
269 		bzero(&info, sizeof(info));
270 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
271 
272 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
273 		    0, 0, &info) == 0) {
274 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
275 			    rt_gateway.sdl_family == AF_LINK) {
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  bads:
383 	ICMP6STAT_INC(icp6s_badns);
384 	if (ifa != NULL)
385 		ifa_free(ifa);
386 	m_freem(m);
387 }
388 
389 /*
390  * Output a Neighbor Solicitation Message. Caller specifies:
391  *	- ICMP6 header source IP6 address
392  *	- ND6 header target IP6 address
393  *	- ND6 header source datalink address
394  *
395  * Based on RFC 2461
396  * Based on RFC 2462 (duplicate address detection)
397  *
398  *    ln - for source address determination
399  * nonce - If non-NULL, NS is used for duplicate address detection and
400  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
401  */
402 static void
403 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
404     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
405     uint8_t *nonce, u_int fibnum)
406 {
407 	struct mbuf *m;
408 	struct m_tag *mtag;
409 	struct ip6_hdr *ip6;
410 	struct nd_neighbor_solicit *nd_ns;
411 	struct ip6_moptions im6o;
412 	int icmp6len;
413 	int maxlen;
414 	caddr_t mac;
415 
416 	if (IN6_IS_ADDR_MULTICAST(taddr6))
417 		return;
418 
419 	/* estimate the size of message */
420 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
421 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
422 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
423 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
424 	    __func__, max_linkhdr, maxlen, MCLBYTES));
425 
426 	if (max_linkhdr + maxlen > MHLEN)
427 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
428 	else
429 		m = m_gethdr(M_NOWAIT, MT_DATA);
430 	if (m == NULL)
431 		return;
432 	M_SETFIB(m, fibnum);
433 
434 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
435 		m->m_flags |= M_MCAST;
436 		im6o.im6o_multicast_ifp = ifp;
437 		im6o.im6o_multicast_hlim = 255;
438 		im6o.im6o_multicast_loop = 0;
439 	}
440 
441 	icmp6len = sizeof(*nd_ns);
442 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
443 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
444 
445 	/* fill neighbor solicitation packet */
446 	ip6 = mtod(m, struct ip6_hdr *);
447 	ip6->ip6_flow = 0;
448 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
449 	ip6->ip6_vfc |= IPV6_VERSION;
450 	/* ip6->ip6_plen will be set later */
451 	ip6->ip6_nxt = IPPROTO_ICMPV6;
452 	ip6->ip6_hlim = 255;
453 	if (daddr6)
454 		ip6->ip6_dst = *daddr6;
455 	else {
456 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
457 		ip6->ip6_dst.s6_addr16[1] = 0;
458 		ip6->ip6_dst.s6_addr32[1] = 0;
459 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
460 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
461 		ip6->ip6_dst.s6_addr8[12] = 0xff;
462 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
463 			goto bad;
464 	}
465 	if (nonce == NULL) {
466 		struct ifaddr *ifa = NULL;
467 
468 		/*
469 		 * RFC2461 7.2.2:
470 		 * "If the source address of the packet prompting the
471 		 * solicitation is the same as one of the addresses assigned
472 		 * to the outgoing interface, that address SHOULD be placed
473 		 * in the IP Source Address of the outgoing solicitation.
474 		 * Otherwise, any one of the addresses assigned to the
475 		 * interface should be used."
476 		 *
477 		 * We use the source address for the prompting packet
478 		 * (saddr6), if saddr6 belongs to the outgoing interface.
479 		 * Otherwise, we perform the source address selection as usual.
480 		 */
481 
482 		if (saddr6 != NULL)
483 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
484 		if (ifa != NULL) {
485 			/* ip6_src set already. */
486 			ip6->ip6_src = *saddr6;
487 			ifa_free(ifa);
488 		} else {
489 			int error;
490 			struct in6_addr dst6, src6;
491 			uint32_t scopeid;
492 
493 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
494 			error = in6_selectsrc_addr(fibnum, &dst6,
495 			    scopeid, ifp, &src6, NULL);
496 			if (error) {
497 				char ip6buf[INET6_ADDRSTRLEN];
498 				nd6log((LOG_DEBUG, "%s: source can't be "
499 				    "determined: dst=%s, error=%d\n", __func__,
500 				    ip6_sprintf(ip6buf, &dst6),
501 				    error));
502 				goto bad;
503 			}
504 			ip6->ip6_src = src6;
505 		}
506 	} else {
507 		/*
508 		 * Source address for DAD packet must always be IPv6
509 		 * unspecified address. (0::0)
510 		 * We actually don't have to 0-clear the address (we did it
511 		 * above), but we do so here explicitly to make the intention
512 		 * clearer.
513 		 */
514 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
515 	}
516 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
517 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
518 	nd_ns->nd_ns_code = 0;
519 	nd_ns->nd_ns_reserved = 0;
520 	nd_ns->nd_ns_target = *taddr6;
521 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
522 
523 	/*
524 	 * Add source link-layer address option.
525 	 *
526 	 *				spec		implementation
527 	 *				---		---
528 	 * DAD packet			MUST NOT	do not add the option
529 	 * there's no link layer address:
530 	 *				impossible	do not add the option
531 	 * there's link layer address:
532 	 *	Multicast NS		MUST add one	add the option
533 	 *	Unicast NS		SHOULD add one	add the option
534 	 */
535 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
536 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
537 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
538 		/* 8 byte alignments... */
539 		optlen = (optlen + 7) & ~7;
540 
541 		m->m_pkthdr.len += optlen;
542 		m->m_len += optlen;
543 		icmp6len += optlen;
544 		bzero((caddr_t)nd_opt, optlen);
545 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
546 		nd_opt->nd_opt_len = optlen >> 3;
547 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
548 	}
549 	/*
550 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
551 	 * This behavior is documented as Enhanced Duplicate Address
552 	 * Detection in RFC 7527.
553 	 * net.inet6.ip6.dad_enhanced=0 disables this.
554 	 */
555 	if (V_dad_enhanced != 0 && nonce != NULL) {
556 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
557 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
558 		/* 8-byte alignment is required. */
559 		optlen = (optlen + 7) & ~7;
560 
561 		m->m_pkthdr.len += optlen;
562 		m->m_len += optlen;
563 		icmp6len += optlen;
564 		bzero((caddr_t)nd_opt, optlen);
565 		nd_opt->nd_opt_type = ND_OPT_NONCE;
566 		nd_opt->nd_opt_len = optlen >> 3;
567 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
568 	}
569 	ip6->ip6_plen = htons((u_short)icmp6len);
570 	nd_ns->nd_ns_cksum = 0;
571 	nd_ns->nd_ns_cksum =
572 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
573 
574 	if (send_sendso_input_hook != NULL) {
575 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
576 			sizeof(unsigned short), M_NOWAIT);
577 		if (mtag == NULL)
578 			goto bad;
579 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
580 		m_tag_prepend(m, mtag);
581 	}
582 
583 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
584 	    &im6o, NULL, NULL);
585 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
586 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
587 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
588 
589 	return;
590 
591   bad:
592 	m_freem(m);
593 }
594 
595 #ifndef BURN_BRIDGES
596 void
597 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
598     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
599 {
600 
601 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
602 }
603 #endif
604 /*
605  * Neighbor advertisement input handling.
606  *
607  * Based on RFC 2461
608  * Based on RFC 2462 (duplicate address detection)
609  *
610  * the following items are not implemented yet:
611  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
612  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
613  */
614 void
615 nd6_na_input(struct mbuf *m, int off, int icmp6len)
616 {
617 	struct ifnet *ifp;
618 	struct ip6_hdr *ip6;
619 	struct ifaddr *ifa;
620 	struct llentry *ln;
621 	struct mbuf *chain;
622 	struct nd_neighbor_advert *nd_na;
623 	struct in6_addr daddr6, taddr6;
624 	struct sockaddr_in6 sin6;
625 	union nd_opts ndopts;
626 	u_char linkhdr[LLE_MAX_LINKHDR];
627 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
628 	char *lladdr;
629 	size_t linkhdrsize;
630 	int flags, is_override, is_router, is_solicited;
631 	int lladdr_off, lladdrlen, checklink;
632 
633 	NET_EPOCH_ASSERT();
634 
635 	chain = NULL;
636 	ln = NULL;
637 	checklink = 0;
638 
639 	/* RFC 6980: Nodes MUST silently ignore fragments */
640 	if(m->m_flags & M_FRAGMENTED)
641 		goto freeit;
642 
643 	ifp = m->m_pkthdr.rcvif;
644 	ip6 = mtod(m, struct ip6_hdr *);
645 	if (__predict_false(ip6->ip6_hlim != 255)) {
646 		ICMP6STAT_INC(icp6s_invlhlim);
647 		nd6log((LOG_ERR,
648 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
649 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
650 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
651 		goto bad;
652 	}
653 
654 	if (m->m_len < off + icmp6len) {
655 		m = m_pullup(m, off + icmp6len);
656 		if (m == NULL) {
657 			IP6STAT_INC(ip6s_exthdrtoolong);
658 			return;
659 		}
660 	}
661 	ip6 = mtod(m, struct ip6_hdr *);
662 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
663 
664 	flags = nd_na->nd_na_flags_reserved;
665 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
666 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
667 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
668 
669 	taddr6 = nd_na->nd_na_target;
670 	if (in6_setscope(&taddr6, ifp, NULL))
671 		goto bad;	/* XXX: impossible */
672 
673 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
674 		nd6log((LOG_ERR,
675 		    "nd6_na_input: invalid target address %s\n",
676 		    ip6_sprintf(ip6bufs, &taddr6)));
677 		goto bad;
678 	}
679 
680 	daddr6 = ip6->ip6_dst;
681 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
682 		if (is_solicited) {
683 			nd6log((LOG_ERR,
684 			    "nd6_na_input: a solicited adv is multicasted\n"));
685 			goto bad;
686 		}
687 
688 	icmp6len -= sizeof(*nd_na);
689 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
690 	if (nd6_options(&ndopts) < 0) {
691 		nd6log((LOG_INFO,
692 		    "nd6_na_input: invalid ND option, ignored\n"));
693 		/* nd6_options have incremented stats */
694 		goto freeit;
695 	}
696 
697 	lladdr = NULL;
698 	lladdrlen = 0;
699 	if (ndopts.nd_opts_tgt_lladdr) {
700 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
701 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
702 	}
703 
704 	/*
705 	 * This effectively disables the DAD check on a non-master CARP
706 	 * address.
707 	 */
708 	if (ifp->if_carp)
709 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
710 	else
711 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
712 
713 	/*
714 	 * Target address matches one of my interface address.
715 	 *
716 	 * If my address is tentative, this means that there's somebody
717 	 * already using the same address as mine.  This indicates DAD failure.
718 	 * This is defined in RFC 2462.
719 	 *
720 	 * Otherwise, process as defined in RFC 2461.
721 	 */
722 	if (ifa
723 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
724 		nd6_dad_na_input(ifa);
725 		ifa_free(ifa);
726 		goto freeit;
727 	}
728 
729 	/* Just for safety, maybe unnecessary. */
730 	if (ifa) {
731 		ifa_free(ifa);
732 		log(LOG_ERR,
733 		    "nd6_na_input: duplicate IP6 address %s\n",
734 		    ip6_sprintf(ip6bufs, &taddr6));
735 		goto freeit;
736 	}
737 
738 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
739 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
740 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
741 		    ifp->if_addrlen, lladdrlen - 2));
742 		goto bad;
743 	}
744 
745 	/*
746 	 * If no neighbor cache entry is found, NA SHOULD silently be
747 	 * discarded.
748 	 */
749 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
750 	if (ln == NULL) {
751 		goto freeit;
752 	}
753 
754 	/*
755 	 * Do not try to override static entry.
756 	 */
757 	if (ln->la_flags & LLE_STATIC)
758 		goto freeit;
759 
760 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
761 		/*
762 		 * If the link-layer has address, and no lladdr option came,
763 		 * discard the packet.
764 		 */
765 		if (ifp->if_addrlen && lladdr == NULL) {
766 			goto freeit;
767 		}
768 
769 		/*
770 		 * Record link-layer address, and update the state.
771 		 */
772 		linkhdrsize = sizeof(linkhdr);
773 		if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
774 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
775 			return;
776 
777 		if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
778 		    lladdr_off) == 0) {
779 			ln = NULL;
780 			goto freeit;
781 		}
782 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
783 		if (is_solicited)
784 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
785 		else
786 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
787 		if ((ln->ln_router = is_router) != 0) {
788 			/*
789 			 * This means a router's state has changed from
790 			 * non-reachable to probably reachable, and might
791 			 * affect the status of associated prefixes..
792 			 */
793 			checklink = 1;
794 		}
795 	} else {
796 		int llchange;
797 
798 		/*
799 		 * Check if the link-layer address has changed or not.
800 		 */
801 		if (lladdr == NULL)
802 			llchange = 0;
803 		else {
804 			if (ln->la_flags & LLE_VALID) {
805 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
806 					llchange = 1;
807 				else
808 					llchange = 0;
809 			} else
810 				llchange = 1;
811 		}
812 
813 		/*
814 		 * This is VERY complex.  Look at it with care.
815 		 *
816 		 * override solicit lladdr llchange	action
817 		 *					(L: record lladdr)
818 		 *
819 		 *	0	0	n	--	(2c)
820 		 *	0	0	y	n	(2b) L
821 		 *	0	0	y	y	(1)    REACHABLE->STALE
822 		 *	0	1	n	--	(2c)   *->REACHABLE
823 		 *	0	1	y	n	(2b) L *->REACHABLE
824 		 *	0	1	y	y	(1)    REACHABLE->STALE
825 		 *	1	0	n	--	(2a)
826 		 *	1	0	y	n	(2a) L
827 		 *	1	0	y	y	(2a) L *->STALE
828 		 *	1	1	n	--	(2a)   *->REACHABLE
829 		 *	1	1	y	n	(2a) L *->REACHABLE
830 		 *	1	1	y	y	(2a) L *->REACHABLE
831 		 */
832 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
833 			/*
834 			 * If state is REACHABLE, make it STALE.
835 			 * no other updates should be done.
836 			 */
837 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
838 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
839 			goto freeit;
840 		} else if (is_override				   /* (2a) */
841 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
842 			|| lladdr == NULL) {			   /* (2c) */
843 			/*
844 			 * Update link-local address, if any.
845 			 */
846 			if (lladdr != NULL) {
847 				linkhdrsize = sizeof(linkhdr);
848 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
849 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
850 					goto freeit;
851 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
852 				    linkhdrsize, lladdr_off) == 0) {
853 					ln = NULL;
854 					goto freeit;
855 				}
856 				EVENTHANDLER_INVOKE(lle_event, ln,
857 				    LLENTRY_RESOLVED);
858 			}
859 
860 			/*
861 			 * If solicited, make the state REACHABLE.
862 			 * If not solicited and the link-layer address was
863 			 * changed, make it STALE.
864 			 */
865 			if (is_solicited)
866 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
867 			else {
868 				if (lladdr != NULL && llchange)
869 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
870 			}
871 		}
872 
873 		if (ln->ln_router && !is_router) {
874 			/*
875 			 * The peer dropped the router flag.
876 			 * Remove the sender from the Default Router List and
877 			 * update the Destination Cache entries.
878 			 */
879 			struct ifnet *nd6_ifp;
880 
881 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
882 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
883 			    (ND_IFINFO(nd6_ifp)->flags &
884 			     ND6_IFF_ACCEPT_RTADV) != 0)
885 				/*
886 				 * Even if the neighbor is not in the default
887 				 * router list, the neighbor may be used as a
888 				 * next hop for some destinations (e.g. redirect
889 				 * case). So we must call rt6_flush explicitly.
890 				 */
891 				rt6_flush(&ip6->ip6_src, ifp);
892 		}
893 		ln->ln_router = is_router;
894 	}
895         /* XXX - QL
896 	 *  Does this matter?
897 	 *  rt->rt_flags &= ~RTF_REJECT;
898 	 */
899 	ln->la_asked = 0;
900 	if (ln->la_hold != NULL) {
901 		memset(&sin6, 0, sizeof(sin6));
902 		nd6_grab_holdchain(ln, &chain, &sin6);
903 	}
904  freeit:
905 	if (ln != NULL)
906 		LLE_WUNLOCK(ln);
907 
908 	if (chain != NULL)
909 		nd6_flush_holdchain(ifp, chain, &sin6);
910 
911 	if (checklink)
912 		pfxlist_onlink_check();
913 
914 	m_freem(m);
915 	return;
916 
917  bad:
918 	if (ln != NULL)
919 		LLE_WUNLOCK(ln);
920 
921 	ICMP6STAT_INC(icp6s_badna);
922 	m_freem(m);
923 }
924 
925 /*
926  * Neighbor advertisement output handling.
927  *
928  * Based on RFC 2461
929  *
930  * the following items are not implemented yet:
931  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
932  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
933  *
934  * tlladdr - 1 if include target link-layer address
935  * sdl0 - sockaddr_dl (= proxy NA) or NULL
936  */
937 static void
938 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
939     const struct in6_addr *taddr6, u_long flags, int tlladdr,
940     struct sockaddr *sdl0, u_int fibnum)
941 {
942 	struct mbuf *m;
943 	struct m_tag *mtag;
944 	struct ip6_hdr *ip6;
945 	struct nd_neighbor_advert *nd_na;
946 	struct ip6_moptions im6o;
947 	struct in6_addr daddr6, dst6, src6;
948 	uint32_t scopeid;
949 
950 	int icmp6len, maxlen, error;
951 	caddr_t mac = NULL;
952 
953 	daddr6 = *daddr6_0;	/* make a local copy for modification */
954 
955 	/* estimate the size of message */
956 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
957 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
958 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
959 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
960 	    __func__, max_linkhdr, maxlen, MCLBYTES));
961 
962 	if (max_linkhdr + maxlen > MHLEN)
963 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
964 	else
965 		m = m_gethdr(M_NOWAIT, MT_DATA);
966 	if (m == NULL)
967 		return;
968 	M_SETFIB(m, fibnum);
969 
970 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
971 		m->m_flags |= M_MCAST;
972 		im6o.im6o_multicast_ifp = ifp;
973 		im6o.im6o_multicast_hlim = 255;
974 		im6o.im6o_multicast_loop = 0;
975 	}
976 
977 	icmp6len = sizeof(*nd_na);
978 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
979 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
980 
981 	/* fill neighbor advertisement packet */
982 	ip6 = mtod(m, struct ip6_hdr *);
983 	ip6->ip6_flow = 0;
984 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
985 	ip6->ip6_vfc |= IPV6_VERSION;
986 	ip6->ip6_nxt = IPPROTO_ICMPV6;
987 	ip6->ip6_hlim = 255;
988 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
989 		/* reply to DAD */
990 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
991 		daddr6.s6_addr16[1] = 0;
992 		daddr6.s6_addr32[1] = 0;
993 		daddr6.s6_addr32[2] = 0;
994 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
995 		if (in6_setscope(&daddr6, ifp, NULL))
996 			goto bad;
997 
998 		flags &= ~ND_NA_FLAG_SOLICITED;
999 	}
1000 	ip6->ip6_dst = daddr6;
1001 
1002 	/*
1003 	 * Select a source whose scope is the same as that of the dest.
1004 	 */
1005 	in6_splitscope(&daddr6, &dst6, &scopeid);
1006 	error = in6_selectsrc_addr(fibnum, &dst6,
1007 	    scopeid, ifp, &src6, NULL);
1008 	if (error) {
1009 		char ip6buf[INET6_ADDRSTRLEN];
1010 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1011 		    "determined: dst=%s, error=%d\n",
1012 		    ip6_sprintf(ip6buf, &daddr6), error));
1013 		goto bad;
1014 	}
1015 	ip6->ip6_src = src6;
1016 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1017 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1018 	nd_na->nd_na_code = 0;
1019 	nd_na->nd_na_target = *taddr6;
1020 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1021 
1022 	/*
1023 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1024 	 * see nd6_ns_input() for details.
1025 	 * Basically, if NS packet is sent to unicast/anycast addr,
1026 	 * target lladdr option SHOULD NOT be included.
1027 	 */
1028 	if (tlladdr) {
1029 		/*
1030 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1031 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1032 		 * my address) use lladdr configured for the interface.
1033 		 */
1034 		if (sdl0 == NULL) {
1035 			if (ifp->if_carp)
1036 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1037 			if (mac == NULL)
1038 				mac = nd6_ifptomac(ifp);
1039 		} else if (sdl0->sa_family == AF_LINK) {
1040 			struct sockaddr_dl *sdl;
1041 			sdl = (struct sockaddr_dl *)sdl0;
1042 			if (sdl->sdl_alen == ifp->if_addrlen)
1043 				mac = LLADDR(sdl);
1044 		}
1045 	}
1046 	if (tlladdr && mac) {
1047 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1048 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1049 
1050 		/* roundup to 8 bytes alignment! */
1051 		optlen = (optlen + 7) & ~7;
1052 
1053 		m->m_pkthdr.len += optlen;
1054 		m->m_len += optlen;
1055 		icmp6len += optlen;
1056 		bzero((caddr_t)nd_opt, optlen);
1057 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1058 		nd_opt->nd_opt_len = optlen >> 3;
1059 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1060 	} else
1061 		flags &= ~ND_NA_FLAG_OVERRIDE;
1062 
1063 	ip6->ip6_plen = htons((u_short)icmp6len);
1064 	nd_na->nd_na_flags_reserved = flags;
1065 	nd_na->nd_na_cksum = 0;
1066 	nd_na->nd_na_cksum =
1067 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1068 
1069 	if (send_sendso_input_hook != NULL) {
1070 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1071 		    sizeof(unsigned short), M_NOWAIT);
1072 		if (mtag == NULL)
1073 			goto bad;
1074 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1075 		m_tag_prepend(m, mtag);
1076 	}
1077 
1078 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1079 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1080 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1081 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1082 
1083 	return;
1084 
1085   bad:
1086 	m_freem(m);
1087 }
1088 
1089 #ifndef BURN_BRIDGES
1090 void
1091 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1092     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1093     struct sockaddr *sdl0)
1094 {
1095 
1096 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1097 	    RT_DEFAULT_FIB);
1098 }
1099 #endif
1100 
1101 caddr_t
1102 nd6_ifptomac(struct ifnet *ifp)
1103 {
1104 	switch (ifp->if_type) {
1105 	case IFT_ETHER:
1106 	case IFT_IEEE1394:
1107 	case IFT_L2VLAN:
1108 	case IFT_INFINIBAND:
1109 	case IFT_BRIDGE:
1110 		return IF_LLADDR(ifp);
1111 	default:
1112 		return NULL;
1113 	}
1114 }
1115 
1116 struct dadq {
1117 	TAILQ_ENTRY(dadq) dad_list;
1118 	struct ifaddr *dad_ifa;
1119 	int dad_count;		/* max NS to send */
1120 	int dad_ns_tcount;	/* # of trials to send NS */
1121 	int dad_ns_ocount;	/* NS sent so far */
1122 	int dad_ns_icount;
1123 	int dad_na_icount;
1124 	int dad_ns_lcount;	/* looped back NS */
1125 	int dad_loopbackprobe;	/* probing state for loopback detection */
1126 	struct callout dad_timer_ch;
1127 	struct vnet *dad_vnet;
1128 	u_int dad_refcnt;
1129 #define	ND_OPT_NONCE_LEN32 \
1130 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1131 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1132 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1133 };
1134 
1135 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1136 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1137 #define	V_dadq			VNET(dadq)
1138 #define	V_dad_rwlock		VNET(dad_rwlock)
1139 
1140 #define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1141 #define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1142 #define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1143 #define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1144 
1145 static void
1146 nd6_dad_add(struct dadq *dp)
1147 {
1148 
1149 	DADQ_WLOCK();
1150 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1151 	dp->dad_ondadq = true;
1152 	DADQ_WUNLOCK();
1153 }
1154 
1155 static void
1156 nd6_dad_del(struct dadq *dp)
1157 {
1158 
1159 	DADQ_WLOCK();
1160 	if (dp->dad_ondadq) {
1161 		/*
1162 		 * Remove dp from the dadq and release the dadq's
1163 		 * reference.
1164 		 */
1165 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1166 		dp->dad_ondadq = false;
1167 		DADQ_WUNLOCK();
1168 		nd6_dad_rele(dp);
1169 	} else
1170 		DADQ_WUNLOCK();
1171 }
1172 
1173 static struct dadq *
1174 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1175 {
1176 	struct dadq *dp;
1177 
1178 	DADQ_RLOCK();
1179 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1180 		if (dp->dad_ifa != ifa)
1181 			continue;
1182 		/*
1183 		 * Skip if the nonce matches the received one.
1184 		 * +2 in the length is required because of type and
1185 		 * length fields are included in a header.
1186 		 */
1187 		if (n != NULL &&
1188 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1189 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1190 		        ND_OPT_NONCE_LEN) == 0) {
1191 			dp->dad_ns_lcount++;
1192 			continue;
1193 		}
1194 		refcount_acquire(&dp->dad_refcnt);
1195 		break;
1196 	}
1197 	DADQ_RUNLOCK();
1198 
1199 	return (dp);
1200 }
1201 
1202 static void
1203 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns)
1204 {
1205 
1206 	NET_EPOCH_ASSERT();
1207 
1208 	if (send_ns != 0)
1209 		nd6_dad_ns_output(dp);
1210 	callout_reset(&dp->dad_timer_ch, ticks,
1211 	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1212 }
1213 
1214 static void
1215 nd6_dad_stoptimer(struct dadq *dp)
1216 {
1217 
1218 	callout_drain(&dp->dad_timer_ch);
1219 }
1220 
1221 static void
1222 nd6_dad_rele(struct dadq *dp)
1223 {
1224 
1225 	if (refcount_release(&dp->dad_refcnt)) {
1226 		ifa_free(dp->dad_ifa);
1227 		free(dp, M_IP6NDP);
1228 	}
1229 }
1230 
1231 void
1232 nd6_dad_init(void)
1233 {
1234 
1235 	rw_init(&V_dad_rwlock, "nd6 DAD queue");
1236 	TAILQ_INIT(&V_dadq);
1237 }
1238 
1239 /*
1240  * Start Duplicate Address Detection (DAD) for specified interface address.
1241  */
1242 void
1243 nd6_dad_start(struct ifaddr *ifa, int delay)
1244 {
1245 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1246 	struct dadq *dp;
1247 	char ip6buf[INET6_ADDRSTRLEN];
1248 	struct epoch_tracker et;
1249 
1250 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1251 	    ("starting DAD on non-tentative address %p", ifa));
1252 
1253 	/*
1254 	 * If we don't need DAD, don't do it.
1255 	 * There are several cases:
1256 	 * - DAD is disabled globally or on the interface
1257 	 * - the interface address is anycast
1258 	 */
1259 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1260 	    V_ip6_dad_count == 0 ||
1261 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1262 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1263 		return;
1264 	}
1265 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1266 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1267 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1268 		return;
1269 
1270 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1271 		/*
1272 		 * DAD is already in progress.  Let the existing entry
1273 		 * finish it.
1274 		 */
1275 		nd6_dad_rele(dp);
1276 		return;
1277 	}
1278 
1279 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1280 	if (dp == NULL) {
1281 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1282 			"%s(%s)\n",
1283 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1284 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1285 		return;
1286 	}
1287 	callout_init(&dp->dad_timer_ch, 0);
1288 #ifdef VIMAGE
1289 	dp->dad_vnet = curvnet;
1290 #endif
1291 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1292 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1293 
1294 	/*
1295 	 * Send NS packet for DAD, ip6_dad_count times.
1296 	 * Note that we must delay the first transmission, if this is the
1297 	 * first packet to be sent from the interface after interface
1298 	 * (re)initialization.
1299 	 */
1300 	dp->dad_ifa = ifa;
1301 	ifa_ref(dp->dad_ifa);
1302 	dp->dad_count = V_ip6_dad_count;
1303 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1304 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1305 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1306 
1307 	/* Add this to the dadq and add a reference for the dadq. */
1308 	refcount_init(&dp->dad_refcnt, 1);
1309 	nd6_dad_add(dp);
1310 	NET_EPOCH_ENTER(et);
1311 	nd6_dad_starttimer(dp, delay, 0);
1312 	NET_EPOCH_EXIT(et);
1313 }
1314 
1315 /*
1316  * terminate DAD unconditionally.  used for address removals.
1317  */
1318 void
1319 nd6_dad_stop(struct ifaddr *ifa)
1320 {
1321 	struct dadq *dp;
1322 
1323 	dp = nd6_dad_find(ifa, NULL);
1324 	if (!dp) {
1325 		/* DAD wasn't started yet */
1326 		return;
1327 	}
1328 
1329 	nd6_dad_stoptimer(dp);
1330 	nd6_dad_del(dp);
1331 
1332 	/* Release this function's reference, acquired by nd6_dad_find(). */
1333 	nd6_dad_rele(dp);
1334 }
1335 
1336 static void
1337 nd6_dad_timer(struct dadq *dp)
1338 {
1339 	CURVNET_SET(dp->dad_vnet);
1340 	struct ifaddr *ifa = dp->dad_ifa;
1341 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1342 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1343 	char ip6buf[INET6_ADDRSTRLEN];
1344 	struct epoch_tracker et;
1345 
1346 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1347 
1348 	NET_EPOCH_ENTER(et);
1349 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1350 		/* Do not need DAD for ifdisabled interface. */
1351 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1352 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1353 		goto err;
1354 	}
1355 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1356 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1357 			"%s(%s)\n",
1358 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1359 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1360 		goto err;
1361 	}
1362 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1363 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1364 			"%s(%s)\n",
1365 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1366 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1367 		goto err;
1368 	}
1369 
1370 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1371 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1372 	    (((ifp->if_flags & IFF_UP) == 0) ||
1373 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1374 		nd6log((LOG_INFO, "%s: could not run DAD "
1375 		    "because the interface was down or not running.\n",
1376 		    if_name(ifa->ifa_ifp)));
1377 		goto err;
1378 	}
1379 
1380 	/* Need more checks? */
1381 	if (dp->dad_ns_ocount < dp->dad_count) {
1382 		/*
1383 		 * We have more NS to go.  Send NS packet for DAD.
1384 		 */
1385 		nd6_dad_starttimer(dp,
1386 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1);
1387 		goto done;
1388 	} else {
1389 		/*
1390 		 * We have transmitted sufficient number of DAD packets.
1391 		 * See what we've got.
1392 		 */
1393 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0)
1394 			/* We've seen NS or NA, means DAD has failed. */
1395 			nd6_dad_duplicated(ifa, dp);
1396 		else if (V_dad_enhanced != 0 &&
1397 		    dp->dad_ns_lcount > 0 &&
1398 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1399 			/*
1400 			 * Sec. 4.1 in RFC 7527 requires transmission of
1401 			 * additional probes until the loopback condition
1402 			 * becomes clear when a looped back probe is detected.
1403 			 */
1404 			log(LOG_ERR, "%s: a looped back NS message is "
1405 			    "detected during DAD for %s.  "
1406 			    "Another DAD probes are being sent.\n",
1407 			    if_name(ifa->ifa_ifp),
1408 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1409 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1410 			/*
1411 			 * Send an NS immediately and increase dad_count by
1412 			 * V_nd6_mmaxtries - 1.
1413 			 */
1414 			dp->dad_count =
1415 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1416 			nd6_dad_starttimer(dp,
1417 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000,
1418 			    1);
1419 			goto done;
1420 		} else {
1421 			/*
1422 			 * We are done with DAD.  No NA came, no NS came.
1423 			 * No duplicate address found.  Check IFDISABLED flag
1424 			 * again in case that it is changed between the
1425 			 * beginning of this function and here.
1426 			 */
1427 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1428 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1429 
1430 			nd6log((LOG_DEBUG,
1431 			    "%s: DAD complete for %s - no duplicates found\n",
1432 			    if_name(ifa->ifa_ifp),
1433 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1434 			if (dp->dad_ns_lcount > 0)
1435 				log(LOG_ERR, "%s: DAD completed while "
1436 				    "a looped back NS message is detected "
1437 				    "during DAD for %s.\n",
1438 				    if_name(ifa->ifa_ifp),
1439 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1440 		}
1441 	}
1442 err:
1443 	nd6_dad_del(dp);
1444 done:
1445 	NET_EPOCH_EXIT(et);
1446 	CURVNET_RESTORE();
1447 }
1448 
1449 static void
1450 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1451 {
1452 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1453 	struct ifnet *ifp;
1454 	char ip6buf[INET6_ADDRSTRLEN];
1455 
1456 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1457 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1458 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1459 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1460 	    dp->dad_na_icount);
1461 
1462 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1463 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1464 
1465 	ifp = ifa->ifa_ifp;
1466 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1467 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1468 	log(LOG_ERR, "%s: manual intervention required\n",
1469 	    if_name(ifp));
1470 
1471 	/*
1472 	 * If the address is a link-local address formed from an interface
1473 	 * identifier based on the hardware address which is supposed to be
1474 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1475 	 * operation on the interface SHOULD be disabled.
1476 	 * [RFC 4862, Section 5.4.5]
1477 	 */
1478 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1479 		struct in6_addr in6;
1480 
1481 		/*
1482 		 * To avoid over-reaction, we only apply this logic when we are
1483 		 * very sure that hardware addresses are supposed to be unique.
1484 		 */
1485 		switch (ifp->if_type) {
1486 		case IFT_ETHER:
1487 		case IFT_ATM:
1488 		case IFT_IEEE1394:
1489 		case IFT_INFINIBAND:
1490 			in6 = ia->ia_addr.sin6_addr;
1491 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1492 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1493 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1494 				log(LOG_ERR, "%s: possible hardware address "
1495 				    "duplication detected, disable IPv6\n",
1496 				    if_name(ifp));
1497 			}
1498 			break;
1499 		}
1500 	}
1501 }
1502 
1503 static void
1504 nd6_dad_ns_output(struct dadq *dp)
1505 {
1506 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1507 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1508 	int i;
1509 
1510 	dp->dad_ns_tcount++;
1511 	if ((ifp->if_flags & IFF_UP) == 0) {
1512 		return;
1513 	}
1514 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1515 		return;
1516 	}
1517 
1518 	dp->dad_ns_ocount++;
1519 	if (V_dad_enhanced != 0) {
1520 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1521 			dp->dad_nonce[i] = arc4random();
1522 		/*
1523 		 * XXXHRS: Note that in the case that
1524 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1525 		 * different nonces can be looped back in an unexpected
1526 		 * order.  The current implementation recognizes only
1527 		 * the latest nonce on the sender side.  Practically it
1528 		 * should work well in almost all cases.
1529 		 */
1530 	}
1531 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1532 	    (uint8_t *)&dp->dad_nonce[0]);
1533 }
1534 
1535 static void
1536 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1537 {
1538 	struct dadq *dp;
1539 
1540 	if (ifa == NULL)
1541 		panic("ifa == NULL in nd6_dad_ns_input");
1542 
1543 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1544 	if (V_dad_enhanced == 0)
1545 		ndopt_nonce = NULL;
1546 	dp = nd6_dad_find(ifa, ndopt_nonce);
1547 	if (dp == NULL)
1548 		return;
1549 
1550 	dp->dad_ns_icount++;
1551 	nd6_dad_rele(dp);
1552 }
1553 
1554 static void
1555 nd6_dad_na_input(struct ifaddr *ifa)
1556 {
1557 	struct dadq *dp;
1558 
1559 	if (ifa == NULL)
1560 		panic("ifa == NULL in nd6_dad_na_input");
1561 
1562 	dp = nd6_dad_find(ifa, NULL);
1563 	if (dp != NULL) {
1564 		dp->dad_na_icount++;
1565 		nd6_dad_rele(dp);
1566 	}
1567 }
1568