xref: /freebsd/sys/netinet6/nd6_nbr.c (revision c7a063741720ef81d4caa4613242579d12f1d605)
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);
88 static void nd6_dad_stoptimer(struct dadq *);
89 static void nd6_dad_timer(void *);
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 static struct ifaddr *nd6_proxy_fill_sdl(struct ifnet *,
100     const struct in6_addr *, struct sockaddr_dl *);
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;
124 	struct ip6_hdr *ip6;
125 	struct nd_neighbor_solicit *nd_ns;
126 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
127 	struct ifaddr *ifa;
128 	struct sockaddr_dl proxydl;
129 	union nd_opts ndopts;
130 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
131 	char *lladdr;
132 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
133 
134 	ifa = NULL;
135 
136 	/* RFC 6980: Nodes MUST silently ignore fragments */
137 	if(m->m_flags & M_FRAGMENTED)
138 		goto freeit;
139 
140 	ifp = m->m_pkthdr.rcvif;
141 	ip6 = mtod(m, struct ip6_hdr *);
142 	if (__predict_false(ip6->ip6_hlim != 255)) {
143 		ICMP6STAT_INC(icp6s_invlhlim);
144 		nd6log((LOG_ERR,
145 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
146 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
147 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
148 		goto bads;
149 	}
150 
151 	if (m->m_len < off + icmp6len) {
152 		m = m_pullup(m, off + icmp6len);
153 		if (m == NULL) {
154 			IP6STAT_INC(ip6s_exthdrtoolong);
155 			return;
156 		}
157 	}
158 	ip6 = mtod(m, struct ip6_hdr *);
159 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
160 
161 	saddr6 = ip6->ip6_src;
162 	daddr6 = ip6->ip6_dst;
163 	taddr6 = nd_ns->nd_ns_target;
164 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
165 		goto bad;
166 
167 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
168 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
169 		rflag = 0;
170 
171 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
172 		/* dst has to be a solicited node multicast address. */
173 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
174 		    /* don't check ifindex portion */
175 		    daddr6.s6_addr32[1] == 0 &&
176 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
177 		    daddr6.s6_addr8[12] == 0xff) {
178 			; /* good */
179 		} else {
180 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
181 			    "(wrong ip6 dst)\n"));
182 			goto bad;
183 		}
184 	} else if (!V_nd6_onlink_ns_rfc4861) {
185 		struct sockaddr_in6 src_sa6;
186 
187 		/*
188 		 * According to recent IETF discussions, it is not a good idea
189 		 * to accept a NS from an address which would not be deemed
190 		 * to be a neighbor otherwise.  This point is expected to be
191 		 * clarified in future revisions of the specification.
192 		 */
193 		bzero(&src_sa6, sizeof(src_sa6));
194 		src_sa6.sin6_family = AF_INET6;
195 		src_sa6.sin6_len = sizeof(src_sa6);
196 		src_sa6.sin6_addr = saddr6;
197 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
198 			nd6log((LOG_INFO, "nd6_ns_input: "
199 				"NS packet from non-neighbor\n"));
200 			goto bad;
201 		}
202 	}
203 
204 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
205 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
206 		goto bad;
207 	}
208 
209 	icmp6len -= sizeof(*nd_ns);
210 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
211 	if (nd6_options(&ndopts) < 0) {
212 		nd6log((LOG_INFO,
213 		    "nd6_ns_input: invalid ND option, ignored\n"));
214 		/* nd6_options have incremented stats */
215 		goto freeit;
216 	}
217 
218 	lladdr = NULL;
219 	lladdrlen = 0;
220 	if (ndopts.nd_opts_src_lladdr) {
221 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
222 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
223 	}
224 
225 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
226 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
227 		    "(link-layer address option)\n"));
228 		goto bad;
229 	}
230 
231 	/*
232 	 * Attaching target link-layer address to the NA?
233 	 * (RFC 2461 7.2.4)
234 	 *
235 	 * NS IP dst is unicast/anycast			MUST NOT add
236 	 * NS IP dst is solicited-node multicast	MUST add
237 	 *
238 	 * In implementation, we add target link-layer address by default.
239 	 * We do not add one in MUST NOT cases.
240 	 */
241 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
242 		tlladdr = 0;
243 	else
244 		tlladdr = 1;
245 
246 	/*
247 	 * Target address (taddr6) must be either:
248 	 * (1) Valid unicast/anycast address for my receiving interface,
249 	 * (2) Unicast address for which I'm offering proxy service, or
250 	 * (3) "tentative" address on which DAD is being performed.
251 	 */
252 	/* (1) and (3) check. */
253 	if (ifp->if_carp)
254 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
255 	else
256 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
257 
258 	/* (2) check. */
259 	proxy = 0;
260 	if (ifa == NULL) {
261 		if ((ifa = nd6_proxy_fill_sdl(ifp, &taddr6, &proxydl)) != NULL)
262 			proxy = 1;
263 	}
264 	if (ifa == NULL) {
265 		/*
266 		 * We've got an NS packet, and we don't have that adddress
267 		 * assigned for us.  We MUST silently ignore it.
268 		 * See RFC2461 7.2.3.
269 		 */
270 		goto freeit;
271 	}
272 	myaddr6 = *IFA_IN6(ifa);
273 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
274 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
275 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
276 		goto freeit;
277 
278 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
279 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
280 		    "(if %d, NS packet %d)\n",
281 		    ip6_sprintf(ip6bufs, &taddr6),
282 		    ifp->if_addrlen, lladdrlen - 2));
283 		goto bad;
284 	}
285 
286 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
287 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
288 		    ip6_sprintf(ip6bufs, &saddr6)));
289 		goto freeit;
290 	}
291 
292 	/*
293 	 * We have neighbor solicitation packet, with target address equals to
294 	 * one of my tentative address.
295 	 *
296 	 * src addr	how to process?
297 	 * ---		---
298 	 * multicast	of course, invalid (rejected in ip6_input)
299 	 * unicast	somebody is doing address resolution -> ignore
300 	 * unspec	dup address detection
301 	 *
302 	 * The processing is defined in RFC 2462.
303 	 */
304 	if (tentative) {
305 		/*
306 		 * If source address is unspecified address, it is for
307 		 * duplicate address detection.
308 		 *
309 		 * If not, the packet is for addess resolution;
310 		 * silently ignore it.
311 		 */
312 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
313 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
314 
315 		goto freeit;
316 	}
317 
318 	/*
319 	 * If the source address is unspecified address, entries must not
320 	 * be created or updated.
321 	 * It looks that sender is performing DAD.  Output NA toward
322 	 * all-node multicast address, to tell the sender that I'm using
323 	 * the address.
324 	 * S bit ("solicited") must be zero.
325 	 */
326 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
327 		struct in6_addr in6_all;
328 
329 		in6_all = in6addr_linklocal_allnodes;
330 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
331 			goto bad;
332 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
333 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
334 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
335 		    M_GETFIB(m));
336 		goto freeit;
337 	}
338 
339 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
340 	    ND_NEIGHBOR_SOLICIT, 0);
341 
342 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
343 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
344 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
345 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
346  freeit:
347 	if (ifa != NULL)
348 		ifa_free(ifa);
349 	m_freem(m);
350 	return;
351 
352  bad:
353 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
354 		ip6_sprintf(ip6bufs, &saddr6)));
355 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
356 		ip6_sprintf(ip6bufs, &daddr6)));
357 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
358 		ip6_sprintf(ip6bufs, &taddr6)));
359  bads:
360 	ICMP6STAT_INC(icp6s_badns);
361 	if (ifa != NULL)
362 		ifa_free(ifa);
363 	m_freem(m);
364 }
365 
366 static struct ifaddr *
367 nd6_proxy_fill_sdl(struct ifnet *ifp, const struct in6_addr *taddr6,
368     struct sockaddr_dl *sdl)
369 {
370 	struct ifaddr *ifa;
371 	struct llentry *ln;
372 
373 	ifa = NULL;
374 	ln = nd6_lookup(taddr6, LLE_SF(AF_INET6, 0), ifp);
375 	if (ln == NULL)
376 		return (ifa);
377 	if ((ln->la_flags & (LLE_PUB | LLE_VALID)) == (LLE_PUB | LLE_VALID)) {
378 		link_init_sdl(ifp, (struct sockaddr *)sdl, ifp->if_type);
379 		sdl->sdl_alen = ifp->if_addrlen;
380 		bcopy(ln->ll_addr, &sdl->sdl_data, ifp->if_addrlen);
381 		LLE_RUNLOCK(ln);
382 		ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
383 		    IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
384 	} else
385 		LLE_RUNLOCK(ln);
386 
387 	return (ifa);
388 }
389 
390 /*
391  * Output a Neighbor Solicitation Message. Caller specifies:
392  *	- ICMP6 header source IP6 address
393  *	- ND6 header target IP6 address
394  *	- ND6 header source datalink address
395  *
396  * Based on RFC 2461
397  * Based on RFC 2462 (duplicate address detection)
398  *
399  *    ln - for source address determination
400  * nonce - If non-NULL, NS is used for duplicate address detection and
401  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
402  */
403 static void
404 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
405     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
406     uint8_t *nonce, u_int fibnum)
407 {
408 	struct mbuf *m;
409 	struct m_tag *mtag;
410 	struct ip6_hdr *ip6;
411 	struct nd_neighbor_solicit *nd_ns;
412 	struct ip6_moptions im6o;
413 	int icmp6len;
414 	int maxlen;
415 
416 	NET_EPOCH_ASSERT();
417 
418 	if (IN6_IS_ADDR_MULTICAST(taddr6))
419 		return;
420 
421 	/* estimate the size of message */
422 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
423 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
424 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
425 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
426 	    __func__, max_linkhdr, maxlen, MCLBYTES));
427 
428 	if (max_linkhdr + maxlen > MHLEN)
429 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
430 	else
431 		m = m_gethdr(M_NOWAIT, MT_DATA);
432 	if (m == NULL)
433 		return;
434 	M_SETFIB(m, fibnum);
435 
436 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
437 		m->m_flags |= M_MCAST;
438 		im6o.im6o_multicast_ifp = ifp;
439 		im6o.im6o_multicast_hlim = 255;
440 		im6o.im6o_multicast_loop = 0;
441 	}
442 
443 	icmp6len = sizeof(*nd_ns);
444 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
445 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
446 
447 	/* fill neighbor solicitation packet */
448 	ip6 = mtod(m, struct ip6_hdr *);
449 	ip6->ip6_flow = 0;
450 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
451 	ip6->ip6_vfc |= IPV6_VERSION;
452 	/* ip6->ip6_plen will be set later */
453 	ip6->ip6_nxt = IPPROTO_ICMPV6;
454 	ip6->ip6_hlim = 255;
455 	if (daddr6)
456 		ip6->ip6_dst = *daddr6;
457 	else {
458 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
459 		ip6->ip6_dst.s6_addr16[1] = 0;
460 		ip6->ip6_dst.s6_addr32[1] = 0;
461 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
462 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
463 		ip6->ip6_dst.s6_addr8[12] = 0xff;
464 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
465 			goto bad;
466 	}
467 	if (nonce == NULL) {
468 		char ip6buf[INET6_ADDRSTRLEN];
469 		struct ifaddr *ifa = NULL;
470 
471 		/*
472 		 * RFC2461 7.2.2:
473 		 * "If the source address of the packet prompting the
474 		 * solicitation is the same as one of the addresses assigned
475 		 * to the outgoing interface, that address SHOULD be placed
476 		 * in the IP Source Address of the outgoing solicitation.
477 		 * Otherwise, any one of the addresses assigned to the
478 		 * interface should be used."
479 		 *
480 		 * We use the source address for the prompting packet
481 		 * (saddr6), if saddr6 belongs to the outgoing interface.
482 		 * Otherwise, we perform the source address selection as usual.
483 		 */
484 		if (saddr6 != NULL)
485 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
486 		if (ifa == NULL) {
487 			int error;
488 			struct in6_addr dst6, src6;
489 			uint32_t scopeid;
490 
491 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
492 			error = in6_selectsrc_addr(fibnum, &dst6,
493 			    scopeid, ifp, &src6, NULL);
494 			if (error) {
495 				nd6log((LOG_DEBUG, "%s: source can't be "
496 				    "determined: dst=%s, error=%d\n", __func__,
497 				    ip6_sprintf(ip6buf, &dst6),
498 				    error));
499 				goto bad;
500 			}
501 			ip6->ip6_src = src6;
502 		} else
503 			ip6->ip6_src = *saddr6;
504 
505 		if (ifp->if_carp != NULL) {
506 			/*
507 			 * Check that selected source address belongs to
508 			 * CARP addresses.
509 			 */
510 			if (ifa == NULL)
511 				ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
512 				    &ip6->ip6_src);
513 			/*
514 			 * Do not send NS for CARP address if we are not
515 			 * the CARP master.
516 			 */
517 			if (ifa != NULL && ifa->ifa_carp != NULL &&
518 			    !(*carp_master_p)(ifa)) {
519 				log(LOG_DEBUG,
520 				    "nd6_ns_output: NS from BACKUP CARP address %s\n",
521 				    ip6_sprintf(ip6buf, &ip6->ip6_src));
522 				ifa_free(ifa);
523 				goto bad;
524 			}
525 		}
526 		if (ifa != NULL)
527 			ifa_free(ifa);
528 	} else {
529 		/*
530 		 * Source address for DAD packet must always be IPv6
531 		 * unspecified address. (0::0)
532 		 * We actually don't have to 0-clear the address (we did it
533 		 * above), but we do so here explicitly to make the intention
534 		 * clearer.
535 		 */
536 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
537 	}
538 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
539 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
540 	nd_ns->nd_ns_code = 0;
541 	nd_ns->nd_ns_reserved = 0;
542 	nd_ns->nd_ns_target = *taddr6;
543 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
544 
545 	/*
546 	 * Add source link-layer address option.
547 	 *
548 	 *				spec		implementation
549 	 *				---		---
550 	 * DAD packet			MUST NOT	do not add the option
551 	 * there's no link layer address:
552 	 *				impossible	do not add the option
553 	 * there's link layer address:
554 	 *	Multicast NS		MUST add one	add the option
555 	 *	Unicast NS		SHOULD add one	add the option
556 	 */
557 	if (nonce == NULL) {
558 		struct nd_opt_hdr *nd_opt;
559 		char *mac;
560 		int optlen;
561 
562 		mac = NULL;
563 		if (ifp->if_carp)
564 			mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src);
565 		if (mac == NULL)
566 			mac = nd6_ifptomac(ifp);
567 
568 		if (mac != NULL) {
569 			nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
570 			optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
571 			/* 8 byte alignments... */
572 			optlen = (optlen + 7) & ~7;
573 			m->m_pkthdr.len += optlen;
574 			m->m_len += optlen;
575 			icmp6len += optlen;
576 			bzero(nd_opt, optlen);
577 			nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
578 			nd_opt->nd_opt_len = optlen >> 3;
579 			bcopy(mac, nd_opt + 1, ifp->if_addrlen);
580 		}
581 	}
582 	/*
583 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
584 	 * This behavior is documented as Enhanced Duplicate Address
585 	 * Detection in RFC 7527.
586 	 * net.inet6.ip6.dad_enhanced=0 disables this.
587 	 */
588 	if (V_dad_enhanced != 0 && nonce != NULL) {
589 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
590 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
591 		/* 8-byte alignment is required. */
592 		optlen = (optlen + 7) & ~7;
593 
594 		m->m_pkthdr.len += optlen;
595 		m->m_len += optlen;
596 		icmp6len += optlen;
597 		bzero((caddr_t)nd_opt, optlen);
598 		nd_opt->nd_opt_type = ND_OPT_NONCE;
599 		nd_opt->nd_opt_len = optlen >> 3;
600 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
601 	}
602 	ip6->ip6_plen = htons((u_short)icmp6len);
603 	nd_ns->nd_ns_cksum = 0;
604 	nd_ns->nd_ns_cksum =
605 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
606 
607 	if (send_sendso_input_hook != NULL) {
608 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
609 			sizeof(unsigned short), M_NOWAIT);
610 		if (mtag == NULL)
611 			goto bad;
612 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
613 		m_tag_prepend(m, mtag);
614 	}
615 
616 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
617 	    &im6o, NULL, NULL);
618 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
619 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
620 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
621 
622 	return;
623 
624   bad:
625 	m_freem(m);
626 }
627 
628 #ifndef BURN_BRIDGES
629 void
630 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
631     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
632 {
633 
634 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
635 }
636 #endif
637 /*
638  * Neighbor advertisement input handling.
639  *
640  * Based on RFC 2461
641  * Based on RFC 2462 (duplicate address detection)
642  *
643  * the following items are not implemented yet:
644  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
645  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
646  */
647 void
648 nd6_na_input(struct mbuf *m, int off, int icmp6len)
649 {
650 	struct ifnet *ifp;
651 	struct ip6_hdr *ip6;
652 	struct ifaddr *ifa;
653 	struct llentry *ln;
654 	struct mbuf *chain;
655 	struct nd_neighbor_advert *nd_na;
656 	struct in6_addr daddr6, taddr6;
657 	union nd_opts ndopts;
658 	u_char linkhdr[LLE_MAX_LINKHDR];
659 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
660 	char *lladdr;
661 	size_t linkhdrsize;
662 	int flags, is_override, is_router, is_solicited;
663 	int lladdr_off, lladdrlen, checklink;
664 	bool flush_holdchain = false;
665 
666 	NET_EPOCH_ASSERT();
667 
668 	chain = NULL;
669 	ln = NULL;
670 	checklink = 0;
671 
672 	/* RFC 6980: Nodes MUST silently ignore fragments */
673 	if(m->m_flags & M_FRAGMENTED)
674 		goto freeit;
675 
676 	ifp = m->m_pkthdr.rcvif;
677 	ip6 = mtod(m, struct ip6_hdr *);
678 	if (__predict_false(ip6->ip6_hlim != 255)) {
679 		ICMP6STAT_INC(icp6s_invlhlim);
680 		nd6log((LOG_ERR,
681 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
682 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
683 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
684 		goto bad;
685 	}
686 
687 	if (m->m_len < off + icmp6len) {
688 		m = m_pullup(m, off + icmp6len);
689 		if (m == NULL) {
690 			IP6STAT_INC(ip6s_exthdrtoolong);
691 			return;
692 		}
693 	}
694 	ip6 = mtod(m, struct ip6_hdr *);
695 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
696 
697 	flags = nd_na->nd_na_flags_reserved;
698 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
699 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
700 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
701 
702 	taddr6 = nd_na->nd_na_target;
703 	if (in6_setscope(&taddr6, ifp, NULL))
704 		goto bad;	/* XXX: impossible */
705 
706 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
707 		nd6log((LOG_ERR,
708 		    "nd6_na_input: invalid target address %s\n",
709 		    ip6_sprintf(ip6bufs, &taddr6)));
710 		goto bad;
711 	}
712 
713 	daddr6 = ip6->ip6_dst;
714 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
715 		if (is_solicited) {
716 			nd6log((LOG_ERR,
717 			    "nd6_na_input: a solicited adv is multicasted\n"));
718 			goto bad;
719 		}
720 
721 	icmp6len -= sizeof(*nd_na);
722 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
723 	if (nd6_options(&ndopts) < 0) {
724 		nd6log((LOG_INFO,
725 		    "nd6_na_input: invalid ND option, ignored\n"));
726 		/* nd6_options have incremented stats */
727 		goto freeit;
728 	}
729 
730 	lladdr = NULL;
731 	lladdrlen = 0;
732 	if (ndopts.nd_opts_tgt_lladdr) {
733 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
734 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
735 	}
736 
737 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
738 	if (ifa != NULL && ifa->ifa_carp != NULL) {
739 		/*
740 		 * Silently ignore NAs for CARP addresses if we are not
741 		 * the CARP master.
742 		 */
743 		if (!(*carp_master_p)(ifa)) {
744 			log(LOG_DEBUG,
745 			    "nd6_na_input: NA for BACKUP CARP address %s\n",
746 			    ip6_sprintf(ip6bufs, &taddr6));
747 			ifa_free(ifa);
748 			goto freeit;
749 		}
750 	}
751 	/*
752 	 * Target address matches one of my interface address.
753 	 *
754 	 * If my address is tentative, this means that there's somebody
755 	 * already using the same address as mine.  This indicates DAD failure.
756 	 * This is defined in RFC 2462.
757 	 *
758 	 * Otherwise, process as defined in RFC 2461.
759 	 */
760 	if (ifa
761 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
762 		nd6_dad_na_input(ifa);
763 		ifa_free(ifa);
764 		goto freeit;
765 	}
766 
767 	/* Just for safety, maybe unnecessary. */
768 	if (ifa) {
769 		ifa_free(ifa);
770 		log(LOG_ERR,
771 		    "nd6_na_input: duplicate IP6 address %s\n",
772 		    ip6_sprintf(ip6bufs, &taddr6));
773 		goto freeit;
774 	}
775 
776 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
777 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
778 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
779 		    ifp->if_addrlen, lladdrlen - 2));
780 		goto bad;
781 	}
782 
783 	/*
784 	 * If no neighbor cache entry is found, NA SHOULD silently be
785 	 * discarded.
786 	 */
787 	ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
788 	if (ln == NULL) {
789 		goto freeit;
790 	}
791 
792 	/*
793 	 * Do not try to override static entry.
794 	 */
795 	if (ln->la_flags & LLE_STATIC)
796 		goto freeit;
797 
798 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
799 		/*
800 		 * If the link-layer has address, and no lladdr option came,
801 		 * discard the packet.
802 		 */
803 		if (ifp->if_addrlen && lladdr == NULL) {
804 			goto freeit;
805 		}
806 
807 		/*
808 		 * Record link-layer address, and update the state.
809 		 */
810 		if (!nd6_try_set_entry_addr(ifp, ln, lladdr))
811 			goto freeit;
812 
813 		flush_holdchain = true;
814 		if (is_solicited)
815 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
816 		else
817 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
818 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
819 		if ((ln->ln_router = is_router) != 0) {
820 			/*
821 			 * This means a router's state has changed from
822 			 * non-reachable to probably reachable, and might
823 			 * affect the status of associated prefixes..
824 			 */
825 			checklink = 1;
826 		}
827 	} else {
828 		int llchange;
829 
830 		/*
831 		 * Check if the link-layer address has changed or not.
832 		 */
833 		if (lladdr == NULL)
834 			llchange = 0;
835 		else {
836 			if (ln->la_flags & LLE_VALID) {
837 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
838 					llchange = 1;
839 				else
840 					llchange = 0;
841 			} else
842 				llchange = 1;
843 		}
844 
845 		/*
846 		 * This is VERY complex.  Look at it with care.
847 		 *
848 		 * override solicit lladdr llchange	action
849 		 *					(L: record lladdr)
850 		 *
851 		 *	0	0	n	--	(2c)
852 		 *	0	0	y	n	(2b) L
853 		 *	0	0	y	y	(1)    REACHABLE->STALE
854 		 *	0	1	n	--	(2c)   *->REACHABLE
855 		 *	0	1	y	n	(2b) L *->REACHABLE
856 		 *	0	1	y	y	(1)    REACHABLE->STALE
857 		 *	1	0	n	--	(2a)
858 		 *	1	0	y	n	(2a) L
859 		 *	1	0	y	y	(2a) L *->STALE
860 		 *	1	1	n	--	(2a)   *->REACHABLE
861 		 *	1	1	y	n	(2a) L *->REACHABLE
862 		 *	1	1	y	y	(2a) L *->REACHABLE
863 		 */
864 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
865 			/*
866 			 * If state is REACHABLE, make it STALE.
867 			 * no other updates should be done.
868 			 */
869 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
870 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
871 			goto freeit;
872 		} else if (is_override				   /* (2a) */
873 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
874 			|| lladdr == NULL) {			   /* (2c) */
875 			/*
876 			 * Update link-local address, if any.
877 			 */
878 			if (lladdr != NULL) {
879 				linkhdrsize = sizeof(linkhdr);
880 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
881 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
882 					goto freeit;
883 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
884 				    linkhdrsize, lladdr_off) == 0)
885 					goto freeit;
886 				EVENTHANDLER_INVOKE(lle_event, ln,
887 				    LLENTRY_RESOLVED);
888 			}
889 
890 			/*
891 			 * If solicited, make the state REACHABLE.
892 			 * If not solicited and the link-layer address was
893 			 * changed, make it STALE.
894 			 */
895 			if (is_solicited)
896 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
897 			else {
898 				if (lladdr != NULL && llchange)
899 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
900 			}
901 		}
902 
903 		if (ln->ln_router && !is_router) {
904 			/*
905 			 * The peer dropped the router flag.
906 			 * Remove the sender from the Default Router List and
907 			 * update the Destination Cache entries.
908 			 */
909 			struct ifnet *nd6_ifp;
910 
911 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
912 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
913 			    (ND_IFINFO(nd6_ifp)->flags &
914 			     ND6_IFF_ACCEPT_RTADV) != 0)
915 				/*
916 				 * Even if the neighbor is not in the default
917 				 * router list, the neighbor may be used as a
918 				 * next hop for some destinations (e.g. redirect
919 				 * case). So we must call rt6_flush explicitly.
920 				 */
921 				rt6_flush(&ip6->ip6_src, ifp);
922 		}
923 		ln->ln_router = is_router;
924 	}
925         /* XXX - QL
926 	 *  Does this matter?
927 	 *  rt->rt_flags &= ~RTF_REJECT;
928 	 */
929 	ln->la_asked = 0;
930 	if (ln->la_hold != NULL)
931 		chain = nd6_grab_holdchain(ln);
932  freeit:
933 	if (ln != NULL)
934 		LLE_WUNLOCK(ln);
935 
936 	if (chain != NULL)
937 		nd6_flush_holdchain(ifp, ln, chain);
938 	if (flush_holdchain)
939 		nd6_flush_children_holdchain(ifp, ln);
940 
941 	if (checklink)
942 		pfxlist_onlink_check();
943 
944 	m_freem(m);
945 	return;
946 
947  bad:
948 	if (ln != NULL)
949 		LLE_WUNLOCK(ln);
950 
951 	ICMP6STAT_INC(icp6s_badna);
952 	m_freem(m);
953 }
954 
955 /*
956  * Neighbor advertisement output handling.
957  *
958  * Based on RFC 2461
959  *
960  * the following items are not implemented yet:
961  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
962  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
963  *
964  * tlladdr - 1 if include target link-layer address
965  * sdl0 - sockaddr_dl (= proxy NA) or NULL
966  */
967 static void
968 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
969     const struct in6_addr *taddr6, u_long flags, int tlladdr,
970     struct sockaddr *sdl0, u_int fibnum)
971 {
972 	struct mbuf *m;
973 	struct m_tag *mtag;
974 	struct ip6_hdr *ip6;
975 	struct nd_neighbor_advert *nd_na;
976 	struct ip6_moptions im6o;
977 	struct in6_addr daddr6, dst6, src6;
978 	uint32_t scopeid;
979 
980 	NET_EPOCH_ASSERT();
981 
982 	int icmp6len, maxlen, error;
983 	caddr_t mac = NULL;
984 
985 	daddr6 = *daddr6_0;	/* make a local copy for modification */
986 
987 	/* estimate the size of message */
988 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
989 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
990 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
991 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
992 	    __func__, max_linkhdr, maxlen, MCLBYTES));
993 
994 	if (max_linkhdr + maxlen > MHLEN)
995 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
996 	else
997 		m = m_gethdr(M_NOWAIT, MT_DATA);
998 	if (m == NULL)
999 		return;
1000 	M_SETFIB(m, fibnum);
1001 
1002 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1003 		m->m_flags |= M_MCAST;
1004 		im6o.im6o_multicast_ifp = ifp;
1005 		im6o.im6o_multicast_hlim = 255;
1006 		im6o.im6o_multicast_loop = 0;
1007 	}
1008 
1009 	icmp6len = sizeof(*nd_na);
1010 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1011 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
1012 
1013 	/* fill neighbor advertisement packet */
1014 	ip6 = mtod(m, struct ip6_hdr *);
1015 	ip6->ip6_flow = 0;
1016 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1017 	ip6->ip6_vfc |= IPV6_VERSION;
1018 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1019 	ip6->ip6_hlim = 255;
1020 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1021 		/* reply to DAD */
1022 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1023 		daddr6.s6_addr16[1] = 0;
1024 		daddr6.s6_addr32[1] = 0;
1025 		daddr6.s6_addr32[2] = 0;
1026 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1027 		if (in6_setscope(&daddr6, ifp, NULL))
1028 			goto bad;
1029 
1030 		flags &= ~ND_NA_FLAG_SOLICITED;
1031 	}
1032 	ip6->ip6_dst = daddr6;
1033 
1034 	/*
1035 	 * Select a source whose scope is the same as that of the dest.
1036 	 */
1037 	in6_splitscope(&daddr6, &dst6, &scopeid);
1038 	error = in6_selectsrc_addr(fibnum, &dst6,
1039 	    scopeid, ifp, &src6, NULL);
1040 	if (error) {
1041 		char ip6buf[INET6_ADDRSTRLEN];
1042 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1043 		    "determined: dst=%s, error=%d\n",
1044 		    ip6_sprintf(ip6buf, &daddr6), error));
1045 		goto bad;
1046 	}
1047 	ip6->ip6_src = src6;
1048 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1049 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1050 	nd_na->nd_na_code = 0;
1051 	nd_na->nd_na_target = *taddr6;
1052 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1053 
1054 	/*
1055 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1056 	 * see nd6_ns_input() for details.
1057 	 * Basically, if NS packet is sent to unicast/anycast addr,
1058 	 * target lladdr option SHOULD NOT be included.
1059 	 */
1060 	if (tlladdr) {
1061 		/*
1062 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1063 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1064 		 * my address) use lladdr configured for the interface.
1065 		 */
1066 		if (sdl0 == NULL) {
1067 			if (ifp->if_carp)
1068 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1069 			if (mac == NULL)
1070 				mac = nd6_ifptomac(ifp);
1071 		} else if (sdl0->sa_family == AF_LINK) {
1072 			struct sockaddr_dl *sdl;
1073 			sdl = (struct sockaddr_dl *)sdl0;
1074 			if (sdl->sdl_alen == ifp->if_addrlen)
1075 				mac = LLADDR(sdl);
1076 		}
1077 	}
1078 	if (tlladdr && mac) {
1079 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1080 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1081 
1082 		/* roundup to 8 bytes alignment! */
1083 		optlen = (optlen + 7) & ~7;
1084 
1085 		m->m_pkthdr.len += optlen;
1086 		m->m_len += optlen;
1087 		icmp6len += optlen;
1088 		bzero((caddr_t)nd_opt, optlen);
1089 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1090 		nd_opt->nd_opt_len = optlen >> 3;
1091 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1092 	} else
1093 		flags &= ~ND_NA_FLAG_OVERRIDE;
1094 
1095 	ip6->ip6_plen = htons((u_short)icmp6len);
1096 	nd_na->nd_na_flags_reserved = flags;
1097 	nd_na->nd_na_cksum = 0;
1098 	nd_na->nd_na_cksum =
1099 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1100 
1101 	if (send_sendso_input_hook != NULL) {
1102 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1103 		    sizeof(unsigned short), M_NOWAIT);
1104 		if (mtag == NULL)
1105 			goto bad;
1106 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1107 		m_tag_prepend(m, mtag);
1108 	}
1109 
1110 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1111 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1112 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1113 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1114 
1115 	return;
1116 
1117   bad:
1118 	m_freem(m);
1119 }
1120 
1121 #ifndef BURN_BRIDGES
1122 void
1123 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1124     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1125     struct sockaddr *sdl0)
1126 {
1127 
1128 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1129 	    RT_DEFAULT_FIB);
1130 }
1131 #endif
1132 
1133 caddr_t
1134 nd6_ifptomac(struct ifnet *ifp)
1135 {
1136 	switch (ifp->if_type) {
1137 	case IFT_ETHER:
1138 	case IFT_IEEE1394:
1139 	case IFT_L2VLAN:
1140 	case IFT_INFINIBAND:
1141 	case IFT_BRIDGE:
1142 		return IF_LLADDR(ifp);
1143 	default:
1144 		return NULL;
1145 	}
1146 }
1147 
1148 struct dadq {
1149 	TAILQ_ENTRY(dadq) dad_list;
1150 	struct ifaddr *dad_ifa;
1151 	int dad_count;		/* max NS to send */
1152 	int dad_ns_tcount;	/* # of trials to send NS */
1153 	int dad_ns_ocount;	/* NS sent so far */
1154 	int dad_ns_icount;
1155 	int dad_na_icount;
1156 	int dad_ns_lcount;	/* looped back NS */
1157 	int dad_loopbackprobe;	/* probing state for loopback detection */
1158 	struct callout dad_timer_ch;
1159 	struct vnet *dad_vnet;
1160 	u_int dad_refcnt;
1161 #define	ND_OPT_NONCE_LEN32 \
1162 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1163 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1164 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1165 };
1166 
1167 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1168 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1169 #define	V_dadq			VNET(dadq)
1170 #define	V_dad_rwlock		VNET(dad_rwlock)
1171 
1172 #define	DADQ_LOCKPTR()		(&V_dad_rwlock)
1173 #define	DADQ_LOCK_INIT()	rw_init(DADQ_LOCKPTR(), "nd6 DAD queue")
1174 #define	DADQ_RLOCK()		rw_rlock(DADQ_LOCKPTR())
1175 #define	DADQ_RUNLOCK()		rw_runlock(DADQ_LOCKPTR())
1176 #define	DADQ_WLOCK()		rw_wlock(DADQ_LOCKPTR())
1177 #define	DADQ_WUNLOCK()		rw_wunlock(DADQ_LOCKPTR())
1178 
1179 #define	DADQ_LOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_LOCKED);
1180 #define	DADQ_RLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_RLOCKED);
1181 #define	DADQ_WLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_WLOCKED);
1182 
1183 static void
1184 nd6_dad_add(struct dadq *dp)
1185 {
1186 	DADQ_WLOCK_ASSERT();
1187 
1188 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1189 	dp->dad_ondadq = true;
1190 }
1191 
1192 static void
1193 nd6_dad_del(struct dadq *dp)
1194 {
1195 	DADQ_WLOCK_ASSERT();
1196 
1197 	if (dp->dad_ondadq) {
1198 		/*
1199 		 * Remove dp from the dadq and release the dadq's
1200 		 * reference.
1201 		 */
1202 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1203 		dp->dad_ondadq = false;
1204 		nd6_dad_rele(dp);
1205 	}
1206 }
1207 
1208 static struct dadq *
1209 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1210 {
1211 	struct dadq *dp;
1212 
1213 	DADQ_LOCK_ASSERT();
1214 
1215 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1216 		if (dp->dad_ifa != ifa)
1217 			continue;
1218 
1219 		/*
1220 		 * Skip if the nonce matches the received one.
1221 		 * +2 in the length is required because of type and
1222 		 * length fields are included in a header.
1223 		 */
1224 		if (n != NULL &&
1225 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1226 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1227 		    ND_OPT_NONCE_LEN) == 0) {
1228 			dp->dad_ns_lcount++;
1229 			continue;
1230 		}
1231 		break;
1232 	}
1233 
1234 	return (dp);
1235 }
1236 
1237 static void
1238 nd6_dad_starttimer(struct dadq *dp, int ticks)
1239 {
1240 	DADQ_WLOCK_ASSERT();
1241 
1242 	callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp);
1243 }
1244 
1245 static void
1246 nd6_dad_stoptimer(struct dadq *dp)
1247 {
1248 	callout_drain(&dp->dad_timer_ch);
1249 }
1250 
1251 static void
1252 nd6_dad_rele(struct dadq *dp)
1253 {
1254 	if (refcount_release(&dp->dad_refcnt)) {
1255 		KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp));
1256 		ifa_free(dp->dad_ifa);
1257 		free(dp, M_IP6NDP);
1258 	}
1259 }
1260 
1261 void
1262 nd6_dad_init(void)
1263 {
1264 	DADQ_LOCK_INIT();
1265 	TAILQ_INIT(&V_dadq);
1266 }
1267 
1268 /*
1269  * Start Duplicate Address Detection (DAD) for specified interface address.
1270  */
1271 void
1272 nd6_dad_start(struct ifaddr *ifa, int delay)
1273 {
1274 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1275 	struct dadq *dp;
1276 	char ip6buf[INET6_ADDRSTRLEN];
1277 
1278 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1279 	    ("starting DAD on non-tentative address %p", ifa));
1280 
1281 	/*
1282 	 * If we don't need DAD, don't do it.
1283 	 * There are several cases:
1284 	 * - DAD is disabled globally or on the interface
1285 	 * - the interface address is anycast
1286 	 */
1287 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1288 	    V_ip6_dad_count == 0 ||
1289 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1290 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1291 		return;
1292 	}
1293 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1294 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1295 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1296 		return;
1297 
1298 	DADQ_WLOCK();
1299 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1300 		/*
1301 		 * DAD is already in progress.  Let the existing entry
1302 		 * finish it.
1303 		 */
1304 		DADQ_WUNLOCK();
1305 		return;
1306 	}
1307 
1308 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1309 	if (dp == NULL) {
1310 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1311 			"%s(%s)\n",
1312 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1313 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1314 		return;
1315 	}
1316 	callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(),
1317 	    CALLOUT_RETURNUNLOCKED);
1318 #ifdef VIMAGE
1319 	dp->dad_vnet = curvnet;
1320 #endif
1321 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1322 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1323 
1324 	/*
1325 	 * Send NS packet for DAD, ip6_dad_count times.
1326 	 * Note that we must delay the first transmission, if this is the
1327 	 * first packet to be sent from the interface after interface
1328 	 * (re)initialization.
1329 	 */
1330 	dp->dad_ifa = ifa;
1331 	ifa_ref(dp->dad_ifa);
1332 	dp->dad_count = V_ip6_dad_count;
1333 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1334 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1335 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1336 
1337 	/* Add this to the dadq and add a reference for the dadq. */
1338 	refcount_init(&dp->dad_refcnt, 1);
1339 	nd6_dad_add(dp);
1340 	nd6_dad_starttimer(dp, delay);
1341 	DADQ_WUNLOCK();
1342 }
1343 
1344 /*
1345  * terminate DAD unconditionally.  used for address removals.
1346  */
1347 void
1348 nd6_dad_stop(struct ifaddr *ifa)
1349 {
1350 	struct dadq *dp;
1351 
1352 	DADQ_WLOCK();
1353 	dp = nd6_dad_find(ifa, NULL);
1354 	if (dp == NULL) {
1355 		DADQ_WUNLOCK();
1356 		/* DAD wasn't started yet */
1357 		return;
1358 	}
1359 
1360 	/*
1361 	 * Acquire a temporary reference so that we can safely stop the callout.
1362 	 */
1363 	(void)refcount_acquire(&dp->dad_refcnt);
1364 	nd6_dad_del(dp);
1365 	DADQ_WUNLOCK();
1366 
1367 	nd6_dad_stoptimer(dp);
1368 	nd6_dad_rele(dp);
1369 }
1370 
1371 static void
1372 nd6_dad_timer(void *arg)
1373 {
1374 	struct dadq *dp = arg;
1375 	struct ifaddr *ifa = dp->dad_ifa;
1376 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1377 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1378 	char ip6buf[INET6_ADDRSTRLEN];
1379 	struct epoch_tracker et;
1380 
1381 	CURVNET_SET(dp->dad_vnet);
1382 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1383 
1384 	NET_EPOCH_ENTER(et);
1385 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1386 		/* Do not need DAD for ifdisabled interface. */
1387 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1388 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1389 		goto err;
1390 	}
1391 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1392 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1393 			"%s(%s)\n",
1394 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1395 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1396 		goto err;
1397 	}
1398 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1399 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1400 			"%s(%s)\n",
1401 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1402 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1403 		goto err;
1404 	}
1405 
1406 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1407 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1408 	    (((ifp->if_flags & IFF_UP) == 0) ||
1409 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1410 		nd6log((LOG_INFO, "%s: could not run DAD "
1411 		    "because the interface was down or not running.\n",
1412 		    if_name(ifa->ifa_ifp)));
1413 		goto err;
1414 	}
1415 
1416 	/* Need more checks? */
1417 	if (dp->dad_ns_ocount < dp->dad_count) {
1418 		/*
1419 		 * We have more NS to go.  Send NS packet for DAD.
1420 		 */
1421 		nd6_dad_starttimer(dp,
1422 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1423 		nd6_dad_ns_output(dp);
1424 		goto done;
1425 	} else {
1426 		/*
1427 		 * We have transmitted sufficient number of DAD packets.
1428 		 * See what we've got.
1429 		 */
1430 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) {
1431 			/* We've seen NS or NA, means DAD has failed. */
1432 			nd6_dad_duplicated(ifa, dp);
1433 		} else if (V_dad_enhanced != 0 &&
1434 		    dp->dad_ns_lcount > 0 &&
1435 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1436 			/*
1437 			 * Sec. 4.1 in RFC 7527 requires transmission of
1438 			 * additional probes until the loopback condition
1439 			 * becomes clear when a looped back probe is detected.
1440 			 */
1441 			log(LOG_ERR, "%s: a looped back NS message is "
1442 			    "detected during DAD for %s.  "
1443 			    "Another DAD probes are being sent.\n",
1444 			    if_name(ifa->ifa_ifp),
1445 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1446 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1447 			/*
1448 			 * Send an NS immediately and increase dad_count by
1449 			 * V_nd6_mmaxtries - 1.
1450 			 */
1451 			dp->dad_count =
1452 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1453 			nd6_dad_starttimer(dp,
1454 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1455 			nd6_dad_ns_output(dp);
1456 			goto done;
1457 		} else {
1458 			/*
1459 			 * We are done with DAD.  No NA came, no NS came.
1460 			 * No duplicate address found.  Check IFDISABLED flag
1461 			 * again in case that it is changed between the
1462 			 * beginning of this function and here.
1463 			 */
1464 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1465 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1466 
1467 			nd6log((LOG_DEBUG,
1468 			    "%s: DAD complete for %s - no duplicates found\n",
1469 			    if_name(ifa->ifa_ifp),
1470 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1471 			if (dp->dad_ns_lcount > 0)
1472 				log(LOG_ERR, "%s: DAD completed while "
1473 				    "a looped back NS message is detected "
1474 				    "during DAD for %s.\n",
1475 				    if_name(ifa->ifa_ifp),
1476 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1477 		}
1478 	}
1479 err:
1480 	nd6_dad_del(dp);
1481 	DADQ_WUNLOCK();
1482 done:
1483 	NET_EPOCH_EXIT(et);
1484 	CURVNET_RESTORE();
1485 }
1486 
1487 static void
1488 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1489 {
1490 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1491 	struct ifnet *ifp;
1492 	char ip6buf[INET6_ADDRSTRLEN];
1493 
1494 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1495 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1496 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1497 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1498 	    dp->dad_na_icount);
1499 
1500 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1501 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1502 
1503 	ifp = ifa->ifa_ifp;
1504 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1505 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1506 	log(LOG_ERR, "%s: manual intervention required\n",
1507 	    if_name(ifp));
1508 
1509 	/*
1510 	 * If the address is a link-local address formed from an interface
1511 	 * identifier based on the hardware address which is supposed to be
1512 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1513 	 * operation on the interface SHOULD be disabled.
1514 	 * [RFC 4862, Section 5.4.5]
1515 	 */
1516 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1517 		struct in6_addr in6;
1518 
1519 		/*
1520 		 * To avoid over-reaction, we only apply this logic when we are
1521 		 * very sure that hardware addresses are supposed to be unique.
1522 		 */
1523 		switch (ifp->if_type) {
1524 		case IFT_ETHER:
1525 		case IFT_ATM:
1526 		case IFT_IEEE1394:
1527 		case IFT_INFINIBAND:
1528 			in6 = ia->ia_addr.sin6_addr;
1529 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1530 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1531 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1532 				log(LOG_ERR, "%s: possible hardware address "
1533 				    "duplication detected, disable IPv6\n",
1534 				    if_name(ifp));
1535 			}
1536 			break;
1537 		}
1538 	}
1539 }
1540 
1541 /*
1542  * Transmit a neighbour solicitation for the purpose of DAD.  Returns with the
1543  * DAD queue unlocked.
1544  */
1545 static void
1546 nd6_dad_ns_output(struct dadq *dp)
1547 {
1548 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1549 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1550 	int i;
1551 
1552 	DADQ_WLOCK_ASSERT();
1553 
1554 	dp->dad_ns_tcount++;
1555 	if ((ifp->if_flags & IFF_UP) == 0) {
1556 		DADQ_WUNLOCK();
1557 		return;
1558 	}
1559 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1560 		DADQ_WUNLOCK();
1561 		return;
1562 	}
1563 
1564 	dp->dad_ns_ocount++;
1565 	if (V_dad_enhanced != 0) {
1566 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1567 			dp->dad_nonce[i] = arc4random();
1568 		/*
1569 		 * XXXHRS: Note that in the case that
1570 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1571 		 * different nonces can be looped back in an unexpected
1572 		 * order.  The current implementation recognizes only
1573 		 * the latest nonce on the sender side.  Practically it
1574 		 * should work well in almost all cases.
1575 		 */
1576 	}
1577 	DADQ_WUNLOCK();
1578 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1579 	    (uint8_t *)&dp->dad_nonce[0]);
1580 }
1581 
1582 static void
1583 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1584 {
1585 	struct dadq *dp;
1586 
1587 	if (ifa == NULL)
1588 		panic("ifa == NULL in nd6_dad_ns_input");
1589 
1590 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1591 	if (V_dad_enhanced == 0)
1592 		ndopt_nonce = NULL;
1593 	DADQ_RLOCK();
1594 	dp = nd6_dad_find(ifa, ndopt_nonce);
1595 	if (dp != NULL)
1596 		dp->dad_ns_icount++;
1597 	DADQ_RUNLOCK();
1598 }
1599 
1600 static void
1601 nd6_dad_na_input(struct ifaddr *ifa)
1602 {
1603 	struct dadq *dp;
1604 
1605 	if (ifa == NULL)
1606 		panic("ifa == NULL in nd6_dad_na_input");
1607 
1608 	DADQ_RLOCK();
1609 	dp = nd6_dad_find(ifa, NULL);
1610 	if (dp != NULL)
1611 		dp->dad_na_icount++;
1612 	DADQ_RUNLOCK();
1613 }
1614