xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 6580f5c38dd5b01aeeaed16b370f1a12423437f0)
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 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/eventhandler.h>
42 #include <sys/malloc.h>
43 #include <sys/libkern.h>
44 #include <sys/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55 #include <sys/callout.h>
56 #include <sys/refcount.h>
57 
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/if_var.h>
62 #include <net/if_private.h>
63 #include <net/route.h>
64 #include <net/vnet.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <net/if_llatbl.h>
69 #include <netinet6/in6_var.h>
70 #include <netinet6/in6_ifattach.h>
71 #include <netinet/ip6.h>
72 #include <netinet6/ip6_var.h>
73 #include <netinet6/scope6_var.h>
74 #include <netinet6/nd6.h>
75 #include <netinet/icmp6.h>
76 #include <netinet/ip_carp.h>
77 #include <netinet6/send.h>
78 
79 #define SDL(s) ((struct sockaddr_dl *)s)
80 
81 struct dadq;
82 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
83 static void nd6_dad_add(struct dadq *dp);
84 static void nd6_dad_del(struct dadq *dp);
85 static void nd6_dad_rele(struct dadq *);
86 static void nd6_dad_starttimer(struct dadq *, int);
87 static void nd6_dad_stoptimer(struct dadq *);
88 static void nd6_dad_timer(void *);
89 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
90 static void nd6_dad_ns_output(struct dadq *);
91 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
92 static void nd6_dad_na_input(struct ifaddr *);
93 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
94     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
95 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
96     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
97 
98 static struct ifaddr *nd6_proxy_fill_sdl(struct ifnet *,
99     const struct in6_addr *, struct sockaddr_dl *);
100 
101 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
102 #define	V_dad_enhanced			VNET(dad_enhanced)
103 
104 SYSCTL_DECL(_net_inet6_ip6);
105 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
106     &VNET_NAME(dad_enhanced), 0,
107     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
108 
109 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
110 						   transmit DAD packet */
111 #define	V_dad_maxtry			VNET(dad_maxtry)
112 
113 VNET_DEFINE_STATIC(int, nd6_onlink_ns_rfc4861) = 0;
114 #define	V_nd6_onlink_ns_rfc4861		VNET(nd6_onlink_ns_rfc4861)
115 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_ONLINKNSRFC4861,
116     nd6_onlink_ns_rfc4861, CTLFLAG_VNET | CTLFLAG_RW,
117     &VNET_NAME(nd6_onlink_ns_rfc4861), 0,
118     "Accept 'on-link' ICMPv6 NS messages in compliance with RFC 4861");
119 
120 /*
121  * Input a Neighbor Solicitation Message.
122  *
123  * Based on RFC 2461
124  * Based on RFC 2462 (duplicate address detection)
125  */
126 void
127 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
128 {
129 	struct ifnet *ifp;
130 	struct ip6_hdr *ip6;
131 	struct nd_neighbor_solicit *nd_ns;
132 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
133 	struct ifaddr *ifa;
134 	struct sockaddr_dl proxydl;
135 	union nd_opts ndopts;
136 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
137 	char *lladdr;
138 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
139 
140 	ifa = NULL;
141 
142 	/* RFC 6980: Nodes MUST silently ignore fragments */
143 	if(m->m_flags & M_FRAGMENTED)
144 		goto freeit;
145 
146 	ifp = m->m_pkthdr.rcvif;
147 	ip6 = mtod(m, struct ip6_hdr *);
148 	if (__predict_false(ip6->ip6_hlim != 255)) {
149 		ICMP6STAT_INC(icp6s_invlhlim);
150 		nd6log((LOG_ERR,
151 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
152 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
153 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
154 		goto bads;
155 	}
156 
157 	if (m->m_len < off + icmp6len) {
158 		m = m_pullup(m, off + icmp6len);
159 		if (m == NULL) {
160 			IP6STAT_INC(ip6s_exthdrtoolong);
161 			return;
162 		}
163 	}
164 	ip6 = mtod(m, struct ip6_hdr *);
165 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
166 
167 	saddr6 = ip6->ip6_src;
168 	daddr6 = ip6->ip6_dst;
169 	taddr6 = nd_ns->nd_ns_target;
170 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
171 		goto bad;
172 
173 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
174 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
175 		rflag = 0;
176 
177 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
178 		/* dst has to be a solicited node multicast address. */
179 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
180 		    /* don't check ifindex portion */
181 		    daddr6.s6_addr32[1] == 0 &&
182 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
183 		    daddr6.s6_addr8[12] == 0xff) {
184 			; /* good */
185 		} else {
186 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
187 			    "(wrong ip6 dst)\n"));
188 			goto bad;
189 		}
190 	} else if (!V_nd6_onlink_ns_rfc4861) {
191 		struct sockaddr_in6 src_sa6;
192 
193 		/*
194 		 * According to recent IETF discussions, it is not a good idea
195 		 * to accept a NS from an address which would not be deemed
196 		 * to be a neighbor otherwise.  This point is expected to be
197 		 * clarified in future revisions of the specification.
198 		 */
199 		bzero(&src_sa6, sizeof(src_sa6));
200 		src_sa6.sin6_family = AF_INET6;
201 		src_sa6.sin6_len = sizeof(src_sa6);
202 		src_sa6.sin6_addr = saddr6;
203 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
204 			nd6log((LOG_INFO, "nd6_ns_input: "
205 				"NS packet from non-neighbor\n"));
206 			goto bad;
207 		}
208 	}
209 
210 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
211 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
212 		goto bad;
213 	}
214 
215 	icmp6len -= sizeof(*nd_ns);
216 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
217 	if (nd6_options(&ndopts) < 0) {
218 		nd6log((LOG_INFO,
219 		    "nd6_ns_input: invalid ND option, ignored\n"));
220 		/* nd6_options have incremented stats */
221 		goto freeit;
222 	}
223 
224 	lladdr = NULL;
225 	lladdrlen = 0;
226 	if (ndopts.nd_opts_src_lladdr) {
227 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
228 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
229 	}
230 
231 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
232 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
233 		    "(link-layer address option)\n"));
234 		goto bad;
235 	}
236 
237 	/*
238 	 * Attaching target link-layer address to the NA?
239 	 * (RFC 2461 7.2.4)
240 	 *
241 	 * NS IP dst is unicast/anycast			MUST NOT add
242 	 * NS IP dst is solicited-node multicast	MUST add
243 	 *
244 	 * In implementation, we add target link-layer address by default.
245 	 * We do not add one in MUST NOT cases.
246 	 */
247 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
248 		tlladdr = 0;
249 	else
250 		tlladdr = 1;
251 
252 	/*
253 	 * Target address (taddr6) must be either:
254 	 * (1) Valid unicast/anycast address for my receiving interface,
255 	 * (2) Unicast address for which I'm offering proxy service, or
256 	 * (3) "tentative" address on which DAD is being performed.
257 	 */
258 	/* (1) and (3) check. */
259 	if (ifp->if_carp)
260 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
261 	else
262 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
263 
264 	/* (2) check. */
265 	proxy = 0;
266 	if (ifa == NULL) {
267 		if ((ifa = nd6_proxy_fill_sdl(ifp, &taddr6, &proxydl)) != NULL)
268 			proxy = 1;
269 	}
270 	if (ifa == NULL) {
271 		/*
272 		 * We've got an NS packet, and we don't have that address
273 		 * assigned for us.  We MUST silently ignore it.
274 		 * See RFC2461 7.2.3.
275 		 */
276 		goto freeit;
277 	}
278 	myaddr6 = *IFA_IN6(ifa);
279 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
280 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
281 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
282 		goto freeit;
283 
284 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
285 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
286 		    "(if %d, NS packet %d)\n",
287 		    ip6_sprintf(ip6bufs, &taddr6),
288 		    ifp->if_addrlen, lladdrlen - 2));
289 		goto bad;
290 	}
291 
292 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
293 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
294 		    ip6_sprintf(ip6bufs, &saddr6)));
295 		goto freeit;
296 	}
297 
298 	/*
299 	 * We have neighbor solicitation packet, with target address equals to
300 	 * one of my tentative address.
301 	 *
302 	 * src addr	how to process?
303 	 * ---		---
304 	 * multicast	of course, invalid (rejected in ip6_input)
305 	 * unicast	somebody is doing address resolution -> ignore
306 	 * unspec	dup address detection
307 	 *
308 	 * The processing is defined in RFC 2462.
309 	 */
310 	if (tentative) {
311 		/*
312 		 * If source address is unspecified address, it is for
313 		 * duplicate address detection.
314 		 *
315 		 * If not, the packet is for addess resolution;
316 		 * silently ignore it.
317 		 */
318 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
319 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
320 
321 		goto freeit;
322 	}
323 
324 	/*
325 	 * If the source address is unspecified address, entries must not
326 	 * be created or updated.
327 	 * It looks that sender is performing DAD.  Output NA toward
328 	 * all-node multicast address, to tell the sender that I'm using
329 	 * the address.
330 	 * S bit ("solicited") must be zero.
331 	 */
332 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
333 		struct in6_addr in6_all;
334 
335 		in6_all = in6addr_linklocal_allnodes;
336 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
337 			goto bad;
338 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
339 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
340 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
341 		    M_GETFIB(m));
342 		goto freeit;
343 	}
344 
345 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
346 	    ND_NEIGHBOR_SOLICIT, 0);
347 
348 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
349 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
350 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
351 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
352  freeit:
353 	if (ifa != NULL)
354 		ifa_free(ifa);
355 	m_freem(m);
356 	return;
357 
358  bad:
359 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
360 		ip6_sprintf(ip6bufs, &saddr6)));
361 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
362 		ip6_sprintf(ip6bufs, &daddr6)));
363 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
364 		ip6_sprintf(ip6bufs, &taddr6)));
365  bads:
366 	ICMP6STAT_INC(icp6s_badns);
367 	if (ifa != NULL)
368 		ifa_free(ifa);
369 	m_freem(m);
370 }
371 
372 static struct ifaddr *
373 nd6_proxy_fill_sdl(struct ifnet *ifp, const struct in6_addr *taddr6,
374     struct sockaddr_dl *sdl)
375 {
376 	struct ifaddr *ifa;
377 	struct llentry *ln;
378 
379 	ifa = NULL;
380 	ln = nd6_lookup(taddr6, LLE_SF(AF_INET6, 0), ifp);
381 	if (ln == NULL)
382 		return (ifa);
383 	if ((ln->la_flags & (LLE_PUB | LLE_VALID)) == (LLE_PUB | LLE_VALID)) {
384 		link_init_sdl(ifp, (struct sockaddr *)sdl, ifp->if_type);
385 		sdl->sdl_alen = ifp->if_addrlen;
386 		bcopy(ln->ll_addr, &sdl->sdl_data, ifp->if_addrlen);
387 		LLE_RUNLOCK(ln);
388 		ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
389 		    IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
390 	} else
391 		LLE_RUNLOCK(ln);
392 
393 	return (ifa);
394 }
395 
396 /*
397  * Output a Neighbor Solicitation Message. Caller specifies:
398  *	- ICMP6 header source IP6 address
399  *	- ND6 header target IP6 address
400  *	- ND6 header source datalink address
401  *
402  * Based on RFC 2461
403  * Based on RFC 2462 (duplicate address detection)
404  *
405  *    ln - for source address determination
406  * nonce - If non-NULL, NS is used for duplicate address detection and
407  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
408  */
409 static void
410 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
411     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
412     uint8_t *nonce, u_int fibnum)
413 {
414 	struct mbuf *m;
415 	struct m_tag *mtag;
416 	struct ip6_hdr *ip6;
417 	struct nd_neighbor_solicit *nd_ns;
418 	struct ip6_moptions im6o;
419 	int icmp6len;
420 	int maxlen;
421 
422 	NET_EPOCH_ASSERT();
423 
424 	if (IN6_IS_ADDR_MULTICAST(taddr6))
425 		return;
426 
427 	/* estimate the size of message */
428 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
429 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
430 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
431 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
432 	    __func__, max_linkhdr, maxlen, MCLBYTES));
433 
434 	if (max_linkhdr + maxlen > MHLEN)
435 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
436 	else
437 		m = m_gethdr(M_NOWAIT, MT_DATA);
438 	if (m == NULL)
439 		return;
440 	M_SETFIB(m, fibnum);
441 
442 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
443 		m->m_flags |= M_MCAST;
444 		im6o.im6o_multicast_ifp = ifp;
445 		im6o.im6o_multicast_hlim = 255;
446 		im6o.im6o_multicast_loop = 0;
447 	}
448 
449 	icmp6len = sizeof(*nd_ns);
450 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
451 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
452 
453 	/* fill neighbor solicitation packet */
454 	ip6 = mtod(m, struct ip6_hdr *);
455 	ip6->ip6_flow = 0;
456 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
457 	ip6->ip6_vfc |= IPV6_VERSION;
458 	/* ip6->ip6_plen will be set later */
459 	ip6->ip6_nxt = IPPROTO_ICMPV6;
460 	ip6->ip6_hlim = 255;
461 	if (daddr6)
462 		ip6->ip6_dst = *daddr6;
463 	else {
464 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
465 		ip6->ip6_dst.s6_addr16[1] = 0;
466 		ip6->ip6_dst.s6_addr32[1] = 0;
467 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
468 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
469 		ip6->ip6_dst.s6_addr8[12] = 0xff;
470 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
471 			goto bad;
472 	}
473 	if (nonce == NULL) {
474 		char ip6buf[INET6_ADDRSTRLEN];
475 		struct ifaddr *ifa = NULL;
476 
477 		/*
478 		 * RFC2461 7.2.2:
479 		 * "If the source address of the packet prompting the
480 		 * solicitation is the same as one of the addresses assigned
481 		 * to the outgoing interface, that address SHOULD be placed
482 		 * in the IP Source Address of the outgoing solicitation.
483 		 * Otherwise, any one of the addresses assigned to the
484 		 * interface should be used."
485 		 *
486 		 * We use the source address for the prompting packet
487 		 * (saddr6), if saddr6 belongs to the outgoing interface.
488 		 * Otherwise, we perform the source address selection as usual.
489 		 */
490 		if (saddr6 != NULL)
491 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
492 		if (ifa == NULL) {
493 			int error;
494 			struct in6_addr dst6, src6;
495 			uint32_t scopeid;
496 
497 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
498 			error = in6_selectsrc_addr(fibnum, &dst6,
499 			    scopeid, ifp, &src6, NULL);
500 			if (error) {
501 				nd6log((LOG_DEBUG, "%s: source can't be "
502 				    "determined: dst=%s, error=%d\n", __func__,
503 				    ip6_sprintf(ip6buf, &dst6),
504 				    error));
505 				goto bad;
506 			}
507 			ip6->ip6_src = src6;
508 		} else
509 			ip6->ip6_src = *saddr6;
510 
511 		if (ifp->if_carp != NULL) {
512 			/*
513 			 * Check that selected source address belongs to
514 			 * CARP addresses.
515 			 */
516 			if (ifa == NULL)
517 				ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
518 				    &ip6->ip6_src);
519 			/*
520 			 * Do not send NS for CARP address if we are not
521 			 * the CARP master.
522 			 */
523 			if (ifa != NULL && ifa->ifa_carp != NULL &&
524 			    !(*carp_master_p)(ifa)) {
525 				nd6log((LOG_DEBUG,
526 				    "nd6_ns_output: NS from BACKUP CARP address %s\n",
527 				    ip6_sprintf(ip6buf, &ip6->ip6_src)));
528 				ifa_free(ifa);
529 				goto bad;
530 			}
531 		}
532 		if (ifa != NULL)
533 			ifa_free(ifa);
534 	} else {
535 		/*
536 		 * Source address for DAD packet must always be IPv6
537 		 * unspecified address. (0::0)
538 		 * We actually don't have to 0-clear the address (we did it
539 		 * above), but we do so here explicitly to make the intention
540 		 * clearer.
541 		 */
542 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
543 	}
544 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
545 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
546 	nd_ns->nd_ns_code = 0;
547 	nd_ns->nd_ns_reserved = 0;
548 	nd_ns->nd_ns_target = *taddr6;
549 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
550 
551 	/*
552 	 * Add source link-layer address option.
553 	 *
554 	 *				spec		implementation
555 	 *				---		---
556 	 * DAD packet			MUST NOT	do not add the option
557 	 * there's no link layer address:
558 	 *				impossible	do not add the option
559 	 * there's link layer address:
560 	 *	Multicast NS		MUST add one	add the option
561 	 *	Unicast NS		SHOULD add one	add the option
562 	 */
563 	if (nonce == NULL) {
564 		struct nd_opt_hdr *nd_opt;
565 		char *mac;
566 		int optlen;
567 
568 		mac = NULL;
569 		if (ifp->if_carp)
570 			mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src);
571 		if (mac == NULL)
572 			mac = nd6_ifptomac(ifp);
573 
574 		if (mac != NULL) {
575 			nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
576 			optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
577 			/* 8 byte alignments... */
578 			optlen = (optlen + 7) & ~7;
579 			m->m_pkthdr.len += optlen;
580 			m->m_len += optlen;
581 			icmp6len += optlen;
582 			bzero(nd_opt, optlen);
583 			nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
584 			nd_opt->nd_opt_len = optlen >> 3;
585 			bcopy(mac, nd_opt + 1, ifp->if_addrlen);
586 		}
587 	}
588 	/*
589 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
590 	 * This behavior is documented as Enhanced Duplicate Address
591 	 * Detection in RFC 7527.
592 	 * net.inet6.ip6.dad_enhanced=0 disables this.
593 	 */
594 	if (V_dad_enhanced != 0 && nonce != NULL) {
595 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
596 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
597 		/* 8-byte alignment is required. */
598 		optlen = (optlen + 7) & ~7;
599 
600 		m->m_pkthdr.len += optlen;
601 		m->m_len += optlen;
602 		icmp6len += optlen;
603 		bzero((caddr_t)nd_opt, optlen);
604 		nd_opt->nd_opt_type = ND_OPT_NONCE;
605 		nd_opt->nd_opt_len = optlen >> 3;
606 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
607 	}
608 	ip6->ip6_plen = htons((u_short)icmp6len);
609 	nd_ns->nd_ns_cksum = 0;
610 	nd_ns->nd_ns_cksum =
611 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
612 
613 	if (send_sendso_input_hook != NULL) {
614 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
615 			sizeof(unsigned short), M_NOWAIT);
616 		if (mtag == NULL)
617 			goto bad;
618 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
619 		m_tag_prepend(m, mtag);
620 	}
621 
622 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
623 	    &im6o, NULL, NULL);
624 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
625 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
626 	ICMP6STAT_INC2(icp6s_outhist, ND_NEIGHBOR_SOLICIT);
627 
628 	return;
629 
630   bad:
631 	m_freem(m);
632 }
633 
634 #ifndef BURN_BRIDGES
635 void
636 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
637     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
638 {
639 
640 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
641 }
642 #endif
643 /*
644  * Neighbor advertisement input handling.
645  *
646  * Based on RFC 2461
647  * Based on RFC 2462 (duplicate address detection)
648  *
649  * the following items are not implemented yet:
650  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
651  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
652  */
653 void
654 nd6_na_input(struct mbuf *m, int off, int icmp6len)
655 {
656 	struct ifnet *ifp;
657 	struct ip6_hdr *ip6;
658 	struct ifaddr *ifa;
659 	struct llentry *ln;
660 	struct mbuf *chain;
661 	struct nd_neighbor_advert *nd_na;
662 	struct in6_addr daddr6, taddr6;
663 	union nd_opts ndopts;
664 	u_char linkhdr[LLE_MAX_LINKHDR];
665 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
666 	char *lladdr;
667 	size_t linkhdrsize;
668 	int flags, is_override, is_router, is_solicited;
669 	int lladdr_off, lladdrlen, checklink;
670 	bool flush_holdchain = false;
671 
672 	NET_EPOCH_ASSERT();
673 
674 	chain = NULL;
675 	ln = NULL;
676 	checklink = 0;
677 
678 	/* RFC 6980: Nodes MUST silently ignore fragments */
679 	if(m->m_flags & M_FRAGMENTED)
680 		goto freeit;
681 
682 	ifp = m->m_pkthdr.rcvif;
683 	ip6 = mtod(m, struct ip6_hdr *);
684 	if (__predict_false(ip6->ip6_hlim != 255)) {
685 		ICMP6STAT_INC(icp6s_invlhlim);
686 		nd6log((LOG_ERR,
687 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
688 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
689 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
690 		goto bad;
691 	}
692 
693 	if (m->m_len < off + icmp6len) {
694 		m = m_pullup(m, off + icmp6len);
695 		if (m == NULL) {
696 			IP6STAT_INC(ip6s_exthdrtoolong);
697 			return;
698 		}
699 	}
700 	ip6 = mtod(m, struct ip6_hdr *);
701 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
702 
703 	flags = nd_na->nd_na_flags_reserved;
704 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
705 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
706 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
707 
708 	taddr6 = nd_na->nd_na_target;
709 	if (in6_setscope(&taddr6, ifp, NULL))
710 		goto bad;	/* XXX: impossible */
711 
712 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
713 		nd6log((LOG_ERR,
714 		    "nd6_na_input: invalid target address %s\n",
715 		    ip6_sprintf(ip6bufs, &taddr6)));
716 		goto bad;
717 	}
718 
719 	daddr6 = ip6->ip6_dst;
720 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
721 		if (is_solicited) {
722 			nd6log((LOG_ERR,
723 			    "nd6_na_input: a solicited adv is multicasted\n"));
724 			goto bad;
725 		}
726 
727 	icmp6len -= sizeof(*nd_na);
728 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
729 	if (nd6_options(&ndopts) < 0) {
730 		nd6log((LOG_INFO,
731 		    "nd6_na_input: invalid ND option, ignored\n"));
732 		/* nd6_options have incremented stats */
733 		goto freeit;
734 	}
735 
736 	lladdr = NULL;
737 	lladdrlen = 0;
738 	if (ndopts.nd_opts_tgt_lladdr) {
739 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
740 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
741 	}
742 
743 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
744 	if (ifa != NULL && ifa->ifa_carp != NULL) {
745 		/*
746 		 * Silently ignore NAs for CARP addresses if we are not
747 		 * the CARP master.
748 		 */
749 		if (!(*carp_master_p)(ifa)) {
750 			nd6log((LOG_DEBUG,
751 			    "nd6_na_input: NA for BACKUP CARP address %s\n",
752 			    ip6_sprintf(ip6bufs, &taddr6)));
753 			ifa_free(ifa);
754 			goto freeit;
755 		}
756 	}
757 	/*
758 	 * Target address matches one of my interface address.
759 	 *
760 	 * If my address is tentative, this means that there's somebody
761 	 * already using the same address as mine.  This indicates DAD failure.
762 	 * This is defined in RFC 2462.
763 	 *
764 	 * Otherwise, process as defined in RFC 2461.
765 	 */
766 	if (ifa
767 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
768 		nd6_dad_na_input(ifa);
769 		ifa_free(ifa);
770 		goto freeit;
771 	}
772 
773 	/* Just for safety, maybe unnecessary. */
774 	if (ifa) {
775 		ifa_free(ifa);
776 		log(LOG_ERR,
777 		    "nd6_na_input: duplicate IP6 address %s\n",
778 		    ip6_sprintf(ip6bufs, &taddr6));
779 		goto freeit;
780 	}
781 
782 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
783 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
784 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
785 		    ifp->if_addrlen, lladdrlen - 2));
786 		goto bad;
787 	}
788 
789 	/*
790 	 * If no neighbor cache entry is found, NA SHOULD silently be
791 	 * discarded.
792 	 */
793 	ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
794 	if (ln == NULL) {
795 		goto freeit;
796 	}
797 
798 	/*
799 	 * Do not try to override static entry.
800 	 */
801 	if (ln->la_flags & LLE_STATIC)
802 		goto freeit;
803 
804 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
805 		/*
806 		 * If the link-layer has address, and no lladdr option came,
807 		 * discard the packet.
808 		 */
809 		if (ifp->if_addrlen && lladdr == NULL) {
810 			goto freeit;
811 		}
812 
813 		/*
814 		 * Record link-layer address, and update the state.
815 		 */
816 		if (!nd6_try_set_entry_addr(ifp, ln, lladdr))
817 			goto freeit;
818 
819 		flush_holdchain = true;
820 		if (is_solicited)
821 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
822 		else
823 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
824 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
825 		if ((ln->ln_router = is_router) != 0) {
826 			/*
827 			 * This means a router's state has changed from
828 			 * non-reachable to probably reachable, and might
829 			 * affect the status of associated prefixes..
830 			 */
831 			checklink = 1;
832 		}
833 	} else {
834 		int llchange;
835 
836 		/*
837 		 * Check if the link-layer address has changed or not.
838 		 */
839 		if (lladdr == NULL)
840 			llchange = 0;
841 		else {
842 			if (ln->la_flags & LLE_VALID) {
843 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
844 					llchange = 1;
845 				else
846 					llchange = 0;
847 			} else
848 				llchange = 1;
849 		}
850 
851 		/*
852 		 * This is VERY complex.  Look at it with care.
853 		 *
854 		 * override solicit lladdr llchange	action
855 		 *					(L: record lladdr)
856 		 *
857 		 *	0	0	n	--	(2c)
858 		 *	0	0	y	n	(2b) L
859 		 *	0	0	y	y	(1)    REACHABLE->STALE
860 		 *	0	1	n	--	(2c)   *->REACHABLE
861 		 *	0	1	y	n	(2b) L *->REACHABLE
862 		 *	0	1	y	y	(1)    REACHABLE->STALE
863 		 *	1	0	n	--	(2a)
864 		 *	1	0	y	n	(2a) L
865 		 *	1	0	y	y	(2a) L *->STALE
866 		 *	1	1	n	--	(2a)   *->REACHABLE
867 		 *	1	1	y	n	(2a) L *->REACHABLE
868 		 *	1	1	y	y	(2a) L *->REACHABLE
869 		 */
870 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
871 			/*
872 			 * If state is REACHABLE, make it STALE.
873 			 * no other updates should be done.
874 			 */
875 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
876 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
877 			goto freeit;
878 		} else if (is_override				   /* (2a) */
879 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
880 			|| lladdr == NULL) {			   /* (2c) */
881 			/*
882 			 * Update link-local address, if any.
883 			 */
884 			if (lladdr != NULL) {
885 				linkhdrsize = sizeof(linkhdr);
886 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
887 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
888 					goto freeit;
889 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
890 				    linkhdrsize, lladdr_off) == 0)
891 					goto freeit;
892 				EVENTHANDLER_INVOKE(lle_event, ln,
893 				    LLENTRY_RESOLVED);
894 			}
895 
896 			/*
897 			 * If solicited, make the state REACHABLE.
898 			 * If not solicited and the link-layer address was
899 			 * changed, make it STALE.
900 			 */
901 			if (is_solicited)
902 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
903 			else {
904 				if (lladdr != NULL && llchange)
905 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
906 			}
907 		}
908 
909 		if (ln->ln_router && !is_router) {
910 			/*
911 			 * The peer dropped the router flag.
912 			 * Remove the sender from the Default Router List and
913 			 * update the Destination Cache entries.
914 			 */
915 			struct ifnet *nd6_ifp;
916 
917 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
918 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
919 			    (ND_IFINFO(nd6_ifp)->flags &
920 			     ND6_IFF_ACCEPT_RTADV) != 0)
921 				/*
922 				 * Even if the neighbor is not in the default
923 				 * router list, the neighbor may be used as a
924 				 * next hop for some destinations (e.g. redirect
925 				 * case). So we must call rt6_flush explicitly.
926 				 */
927 				rt6_flush(&ip6->ip6_src, ifp);
928 		}
929 		ln->ln_router = is_router;
930 	}
931         /* XXX - QL
932 	 *  Does this matter?
933 	 *  rt->rt_flags &= ~RTF_REJECT;
934 	 */
935 	ln->la_asked = 0;
936 	if (ln->la_hold != NULL)
937 		chain = nd6_grab_holdchain(ln);
938  freeit:
939 	if (ln != NULL)
940 		LLE_WUNLOCK(ln);
941 
942 	if (chain != NULL)
943 		nd6_flush_holdchain(ifp, ln, chain);
944 	if (flush_holdchain)
945 		nd6_flush_children_holdchain(ifp, ln);
946 
947 	if (checklink)
948 		pfxlist_onlink_check();
949 
950 	m_freem(m);
951 	return;
952 
953  bad:
954 	if (ln != NULL)
955 		LLE_WUNLOCK(ln);
956 
957 	ICMP6STAT_INC(icp6s_badna);
958 	m_freem(m);
959 }
960 
961 /*
962  * Neighbor advertisement output handling.
963  *
964  * Based on RFC 2461
965  *
966  * the following items are not implemented yet:
967  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
968  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
969  *
970  * tlladdr - 1 if include target link-layer address
971  * sdl0 - sockaddr_dl (= proxy NA) or NULL
972  */
973 static void
974 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
975     const struct in6_addr *taddr6, u_long flags, int tlladdr,
976     struct sockaddr *sdl0, u_int fibnum)
977 {
978 	struct mbuf *m;
979 	struct m_tag *mtag;
980 	struct ip6_hdr *ip6;
981 	struct nd_neighbor_advert *nd_na;
982 	struct ip6_moptions im6o;
983 	struct in6_addr daddr6, dst6, src6;
984 	uint32_t scopeid;
985 
986 	NET_EPOCH_ASSERT();
987 
988 	int icmp6len, maxlen, error;
989 	caddr_t mac = NULL;
990 
991 	daddr6 = *daddr6_0;	/* make a local copy for modification */
992 
993 	/* estimate the size of message */
994 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
995 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
996 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
997 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
998 	    __func__, max_linkhdr, maxlen, MCLBYTES));
999 
1000 	if (max_linkhdr + maxlen > MHLEN)
1001 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
1002 	else
1003 		m = m_gethdr(M_NOWAIT, MT_DATA);
1004 	if (m == NULL)
1005 		return;
1006 	M_SETFIB(m, fibnum);
1007 
1008 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
1009 		m->m_flags |= M_MCAST;
1010 		im6o.im6o_multicast_ifp = ifp;
1011 		im6o.im6o_multicast_hlim = 255;
1012 		im6o.im6o_multicast_loop = 0;
1013 	}
1014 
1015 	icmp6len = sizeof(*nd_na);
1016 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1017 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
1018 
1019 	/* fill neighbor advertisement packet */
1020 	ip6 = mtod(m, struct ip6_hdr *);
1021 	ip6->ip6_flow = 0;
1022 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1023 	ip6->ip6_vfc |= IPV6_VERSION;
1024 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1025 	ip6->ip6_hlim = 255;
1026 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1027 		/* reply to DAD */
1028 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1029 		daddr6.s6_addr16[1] = 0;
1030 		daddr6.s6_addr32[1] = 0;
1031 		daddr6.s6_addr32[2] = 0;
1032 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1033 		if (in6_setscope(&daddr6, ifp, NULL))
1034 			goto bad;
1035 
1036 		flags &= ~ND_NA_FLAG_SOLICITED;
1037 	}
1038 	ip6->ip6_dst = daddr6;
1039 
1040 	/*
1041 	 * Select a source whose scope is the same as that of the dest.
1042 	 */
1043 	in6_splitscope(&daddr6, &dst6, &scopeid);
1044 	error = in6_selectsrc_addr(fibnum, &dst6,
1045 	    scopeid, ifp, &src6, NULL);
1046 	if (error) {
1047 		char ip6buf[INET6_ADDRSTRLEN];
1048 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1049 		    "determined: dst=%s, error=%d\n",
1050 		    ip6_sprintf(ip6buf, &daddr6), error));
1051 		goto bad;
1052 	}
1053 	ip6->ip6_src = src6;
1054 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1055 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1056 	nd_na->nd_na_code = 0;
1057 	nd_na->nd_na_target = *taddr6;
1058 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1059 
1060 	/*
1061 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1062 	 * see nd6_ns_input() for details.
1063 	 * Basically, if NS packet is sent to unicast/anycast addr,
1064 	 * target lladdr option SHOULD NOT be included.
1065 	 */
1066 	if (tlladdr) {
1067 		/*
1068 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1069 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1070 		 * my address) use lladdr configured for the interface.
1071 		 */
1072 		if (sdl0 == NULL) {
1073 			if (ifp->if_carp)
1074 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1075 			if (mac == NULL)
1076 				mac = nd6_ifptomac(ifp);
1077 		} else if (sdl0->sa_family == AF_LINK) {
1078 			struct sockaddr_dl *sdl;
1079 			sdl = (struct sockaddr_dl *)sdl0;
1080 			if (sdl->sdl_alen == ifp->if_addrlen)
1081 				mac = LLADDR(sdl);
1082 		}
1083 	}
1084 	if (tlladdr && mac) {
1085 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1086 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1087 
1088 		/* roundup to 8 bytes alignment! */
1089 		optlen = (optlen + 7) & ~7;
1090 
1091 		m->m_pkthdr.len += optlen;
1092 		m->m_len += optlen;
1093 		icmp6len += optlen;
1094 		bzero((caddr_t)nd_opt, optlen);
1095 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1096 		nd_opt->nd_opt_len = optlen >> 3;
1097 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1098 	} else
1099 		flags &= ~ND_NA_FLAG_OVERRIDE;
1100 
1101 	ip6->ip6_plen = htons((u_short)icmp6len);
1102 	nd_na->nd_na_flags_reserved = flags;
1103 	nd_na->nd_na_cksum = 0;
1104 	nd_na->nd_na_cksum =
1105 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1106 
1107 	if (send_sendso_input_hook != NULL) {
1108 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1109 		    sizeof(unsigned short), M_NOWAIT);
1110 		if (mtag == NULL)
1111 			goto bad;
1112 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1113 		m_tag_prepend(m, mtag);
1114 	}
1115 
1116 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1117 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1118 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1119 	ICMP6STAT_INC2(icp6s_outhist, ND_NEIGHBOR_ADVERT);
1120 
1121 	return;
1122 
1123   bad:
1124 	m_freem(m);
1125 }
1126 
1127 #ifndef BURN_BRIDGES
1128 void
1129 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1130     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1131     struct sockaddr *sdl0)
1132 {
1133 
1134 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1135 	    RT_DEFAULT_FIB);
1136 }
1137 #endif
1138 
1139 caddr_t
1140 nd6_ifptomac(struct ifnet *ifp)
1141 {
1142 	switch (ifp->if_type) {
1143 	case IFT_ETHER:
1144 	case IFT_IEEE1394:
1145 	case IFT_L2VLAN:
1146 	case IFT_INFINIBAND:
1147 	case IFT_BRIDGE:
1148 		return IF_LLADDR(ifp);
1149 	default:
1150 		return NULL;
1151 	}
1152 }
1153 
1154 struct dadq {
1155 	TAILQ_ENTRY(dadq) dad_list;
1156 	struct ifaddr *dad_ifa;
1157 	int dad_count;		/* max NS to send */
1158 	int dad_ns_tcount;	/* # of trials to send NS */
1159 	int dad_ns_ocount;	/* NS sent so far */
1160 	int dad_ns_icount;
1161 	int dad_na_icount;
1162 	int dad_ns_lcount;	/* looped back NS */
1163 	int dad_loopbackprobe;	/* probing state for loopback detection */
1164 	struct callout dad_timer_ch;
1165 	struct vnet *dad_vnet;
1166 	u_int dad_refcnt;
1167 #define	ND_OPT_NONCE_LEN32 \
1168 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1169 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1170 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1171 };
1172 
1173 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1174 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1175 #define	V_dadq			VNET(dadq)
1176 #define	V_dad_rwlock		VNET(dad_rwlock)
1177 
1178 #define	DADQ_LOCKPTR()		(&V_dad_rwlock)
1179 #define	DADQ_LOCK_INIT()	rw_init(DADQ_LOCKPTR(), "nd6 DAD queue")
1180 #define	DADQ_RLOCK()		rw_rlock(DADQ_LOCKPTR())
1181 #define	DADQ_RUNLOCK()		rw_runlock(DADQ_LOCKPTR())
1182 #define	DADQ_WLOCK()		rw_wlock(DADQ_LOCKPTR())
1183 #define	DADQ_WUNLOCK()		rw_wunlock(DADQ_LOCKPTR())
1184 
1185 #define	DADQ_LOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_LOCKED);
1186 #define	DADQ_RLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_RLOCKED);
1187 #define	DADQ_WLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_WLOCKED);
1188 
1189 static void
1190 nd6_dad_add(struct dadq *dp)
1191 {
1192 	DADQ_WLOCK_ASSERT();
1193 
1194 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1195 	dp->dad_ondadq = true;
1196 }
1197 
1198 static void
1199 nd6_dad_del(struct dadq *dp)
1200 {
1201 	DADQ_WLOCK_ASSERT();
1202 
1203 	if (dp->dad_ondadq) {
1204 		/*
1205 		 * Remove dp from the dadq and release the dadq's
1206 		 * reference.
1207 		 */
1208 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1209 		dp->dad_ondadq = false;
1210 		nd6_dad_rele(dp);
1211 	}
1212 }
1213 
1214 static struct dadq *
1215 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1216 {
1217 	struct dadq *dp;
1218 
1219 	DADQ_LOCK_ASSERT();
1220 
1221 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1222 		if (dp->dad_ifa != ifa)
1223 			continue;
1224 
1225 		/*
1226 		 * Skip if the nonce matches the received one.
1227 		 * +2 in the length is required because of type and
1228 		 * length fields are included in a header.
1229 		 */
1230 		if (n != NULL &&
1231 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1232 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1233 		    ND_OPT_NONCE_LEN) == 0) {
1234 			dp->dad_ns_lcount++;
1235 			continue;
1236 		}
1237 		break;
1238 	}
1239 
1240 	return (dp);
1241 }
1242 
1243 static void
1244 nd6_dad_starttimer(struct dadq *dp, int ticks)
1245 {
1246 	DADQ_WLOCK_ASSERT();
1247 
1248 	callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp);
1249 }
1250 
1251 static void
1252 nd6_dad_stoptimer(struct dadq *dp)
1253 {
1254 	callout_drain(&dp->dad_timer_ch);
1255 }
1256 
1257 static void
1258 nd6_dad_rele(struct dadq *dp)
1259 {
1260 	if (refcount_release(&dp->dad_refcnt)) {
1261 		KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp));
1262 		ifa_free(dp->dad_ifa);
1263 		free(dp, M_IP6NDP);
1264 	}
1265 }
1266 
1267 void
1268 nd6_dad_init(void)
1269 {
1270 	DADQ_LOCK_INIT();
1271 	TAILQ_INIT(&V_dadq);
1272 }
1273 
1274 /*
1275  * Start Duplicate Address Detection (DAD) for specified interface address.
1276  */
1277 void
1278 nd6_dad_start(struct ifaddr *ifa, int delay)
1279 {
1280 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1281 	struct dadq *dp;
1282 	char ip6buf[INET6_ADDRSTRLEN];
1283 
1284 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1285 	    ("starting DAD on non-tentative address %p", ifa));
1286 
1287 	/*
1288 	 * If we don't need DAD, don't do it.
1289 	 * There are several cases:
1290 	 * - DAD is disabled globally or on the interface
1291 	 * - the interface address is anycast
1292 	 */
1293 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1294 	    V_ip6_dad_count == 0 ||
1295 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1296 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1297 		return;
1298 	}
1299 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1300 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1301 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1302 		return;
1303 
1304 	DADQ_WLOCK();
1305 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1306 		/*
1307 		 * DAD is already in progress.  Let the existing entry
1308 		 * finish it.
1309 		 */
1310 		DADQ_WUNLOCK();
1311 		return;
1312 	}
1313 
1314 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1315 	if (dp == NULL) {
1316 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1317 			"%s(%s)\n",
1318 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1319 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1320 		return;
1321 	}
1322 	callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(),
1323 	    CALLOUT_RETURNUNLOCKED);
1324 #ifdef VIMAGE
1325 	dp->dad_vnet = curvnet;
1326 #endif
1327 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1328 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1329 
1330 	/*
1331 	 * Send NS packet for DAD, ip6_dad_count times.
1332 	 * Note that we must delay the first transmission, if this is the
1333 	 * first packet to be sent from the interface after interface
1334 	 * (re)initialization.
1335 	 */
1336 	dp->dad_ifa = ifa;
1337 	ifa_ref(dp->dad_ifa);
1338 	dp->dad_count = V_ip6_dad_count;
1339 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1340 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1341 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1342 
1343 	/* Add this to the dadq and add a reference for the dadq. */
1344 	refcount_init(&dp->dad_refcnt, 1);
1345 	nd6_dad_add(dp);
1346 	nd6_dad_starttimer(dp, delay);
1347 	DADQ_WUNLOCK();
1348 }
1349 
1350 /*
1351  * terminate DAD unconditionally.  used for address removals.
1352  */
1353 void
1354 nd6_dad_stop(struct ifaddr *ifa)
1355 {
1356 	struct dadq *dp;
1357 
1358 	DADQ_WLOCK();
1359 	dp = nd6_dad_find(ifa, NULL);
1360 	if (dp == NULL) {
1361 		DADQ_WUNLOCK();
1362 		/* DAD wasn't started yet */
1363 		return;
1364 	}
1365 
1366 	/*
1367 	 * Acquire a temporary reference so that we can safely stop the callout.
1368 	 */
1369 	(void)refcount_acquire(&dp->dad_refcnt);
1370 	nd6_dad_del(dp);
1371 	DADQ_WUNLOCK();
1372 
1373 	nd6_dad_stoptimer(dp);
1374 	nd6_dad_rele(dp);
1375 }
1376 
1377 static void
1378 nd6_dad_timer(void *arg)
1379 {
1380 	struct dadq *dp = arg;
1381 	struct ifaddr *ifa = dp->dad_ifa;
1382 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1383 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1384 	char ip6buf[INET6_ADDRSTRLEN];
1385 	struct epoch_tracker et;
1386 
1387 	CURVNET_SET(dp->dad_vnet);
1388 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1389 
1390 	NET_EPOCH_ENTER(et);
1391 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1392 		/* Do not need DAD for ifdisabled interface. */
1393 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1394 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1395 		goto err;
1396 	}
1397 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1398 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1399 			"%s(%s)\n",
1400 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1401 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1402 		goto err;
1403 	}
1404 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1405 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1406 			"%s(%s)\n",
1407 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1408 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1409 		goto err;
1410 	}
1411 
1412 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1413 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1414 	    (((ifp->if_flags & IFF_UP) == 0) ||
1415 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1416 		nd6log((LOG_INFO, "%s: could not run DAD "
1417 		    "because the interface was down or not running.\n",
1418 		    if_name(ifa->ifa_ifp)));
1419 		goto err;
1420 	}
1421 
1422 	/* Need more checks? */
1423 	if (dp->dad_ns_ocount < dp->dad_count) {
1424 		/*
1425 		 * We have more NS to go.  Send NS packet for DAD.
1426 		 */
1427 		nd6_dad_starttimer(dp,
1428 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1429 		nd6_dad_ns_output(dp);
1430 		goto done;
1431 	} else {
1432 		/*
1433 		 * We have transmitted sufficient number of DAD packets.
1434 		 * See what we've got.
1435 		 */
1436 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) {
1437 			/* We've seen NS or NA, means DAD has failed. */
1438 			nd6_dad_duplicated(ifa, dp);
1439 		} else if (V_dad_enhanced != 0 &&
1440 		    dp->dad_ns_lcount > 0 &&
1441 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1442 			/*
1443 			 * Sec. 4.1 in RFC 7527 requires transmission of
1444 			 * additional probes until the loopback condition
1445 			 * becomes clear when a looped back probe is detected.
1446 			 */
1447 			log(LOG_ERR, "%s: a looped back NS message is "
1448 			    "detected during DAD for %s.  "
1449 			    "Another DAD probes are being sent.\n",
1450 			    if_name(ifa->ifa_ifp),
1451 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1452 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1453 			/*
1454 			 * Send an NS immediately and increase dad_count by
1455 			 * V_nd6_mmaxtries - 1.
1456 			 */
1457 			dp->dad_count =
1458 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1459 			nd6_dad_starttimer(dp,
1460 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1461 			nd6_dad_ns_output(dp);
1462 			goto done;
1463 		} else {
1464 			/*
1465 			 * We are done with DAD.  No NA came, no NS came.
1466 			 * No duplicate address found.  Check IFDISABLED flag
1467 			 * again in case that it is changed between the
1468 			 * beginning of this function and here.
1469 			 */
1470 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1471 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1472 
1473 			nd6log((LOG_DEBUG,
1474 			    "%s: DAD complete for %s - no duplicates found\n",
1475 			    if_name(ifa->ifa_ifp),
1476 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1477 			if (dp->dad_ns_lcount > 0)
1478 				log(LOG_ERR, "%s: DAD completed while "
1479 				    "a looped back NS message is detected "
1480 				    "during DAD for %s.\n",
1481 				    if_name(ifa->ifa_ifp),
1482 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1483 		}
1484 	}
1485 err:
1486 	nd6_dad_del(dp);
1487 	DADQ_WUNLOCK();
1488 done:
1489 	NET_EPOCH_EXIT(et);
1490 	CURVNET_RESTORE();
1491 }
1492 
1493 static void
1494 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1495 {
1496 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1497 	struct ifnet *ifp;
1498 	char ip6buf[INET6_ADDRSTRLEN];
1499 
1500 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1501 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1502 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1503 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1504 	    dp->dad_na_icount);
1505 
1506 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1507 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1508 
1509 	ifp = ifa->ifa_ifp;
1510 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1511 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1512 	log(LOG_ERR, "%s: manual intervention required\n",
1513 	    if_name(ifp));
1514 
1515 	/*
1516 	 * If the address is a link-local address formed from an interface
1517 	 * identifier based on the hardware address which is supposed to be
1518 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1519 	 * operation on the interface SHOULD be disabled.
1520 	 * [RFC 4862, Section 5.4.5]
1521 	 */
1522 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1523 		struct in6_addr in6;
1524 
1525 		/*
1526 		 * To avoid over-reaction, we only apply this logic when we are
1527 		 * very sure that hardware addresses are supposed to be unique.
1528 		 */
1529 		switch (ifp->if_type) {
1530 		case IFT_ETHER:
1531 		case IFT_ATM:
1532 		case IFT_IEEE1394:
1533 		case IFT_INFINIBAND:
1534 			in6 = ia->ia_addr.sin6_addr;
1535 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1536 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1537 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1538 				log(LOG_ERR, "%s: possible hardware address "
1539 				    "duplication detected, disable IPv6\n",
1540 				    if_name(ifp));
1541 			}
1542 			break;
1543 		}
1544 	}
1545 }
1546 
1547 /*
1548  * Transmit a neighbour solicitation for the purpose of DAD.  Returns with the
1549  * DAD queue unlocked.
1550  */
1551 static void
1552 nd6_dad_ns_output(struct dadq *dp)
1553 {
1554 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1555 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1556 	int i;
1557 
1558 	DADQ_WLOCK_ASSERT();
1559 
1560 	dp->dad_ns_tcount++;
1561 	if ((ifp->if_flags & IFF_UP) == 0) {
1562 		DADQ_WUNLOCK();
1563 		return;
1564 	}
1565 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1566 		DADQ_WUNLOCK();
1567 		return;
1568 	}
1569 
1570 	dp->dad_ns_ocount++;
1571 	if (V_dad_enhanced != 0) {
1572 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1573 			dp->dad_nonce[i] = arc4random();
1574 		/*
1575 		 * XXXHRS: Note that in the case that
1576 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1577 		 * different nonces can be looped back in an unexpected
1578 		 * order.  The current implementation recognizes only
1579 		 * the latest nonce on the sender side.  Practically it
1580 		 * should work well in almost all cases.
1581 		 */
1582 	}
1583 	DADQ_WUNLOCK();
1584 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1585 	    (uint8_t *)&dp->dad_nonce[0]);
1586 }
1587 
1588 static void
1589 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1590 {
1591 	struct dadq *dp;
1592 
1593 	if (ifa == NULL)
1594 		panic("ifa == NULL in nd6_dad_ns_input");
1595 
1596 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1597 	if (V_dad_enhanced == 0)
1598 		ndopt_nonce = NULL;
1599 	DADQ_RLOCK();
1600 	dp = nd6_dad_find(ifa, ndopt_nonce);
1601 	if (dp != NULL)
1602 		dp->dad_ns_icount++;
1603 	DADQ_RUNLOCK();
1604 }
1605 
1606 static void
1607 nd6_dad_na_input(struct ifaddr *ifa)
1608 {
1609 	struct dadq *dp;
1610 
1611 	if (ifa == NULL)
1612 		panic("ifa == NULL in nd6_dad_na_input");
1613 
1614 	DADQ_RLOCK();
1615 	dp = nd6_dad_find(ifa, NULL);
1616 	if (dp != NULL)
1617 		dp->dad_na_icount++;
1618 	DADQ_RUNLOCK();
1619 }
1620