xref: /netbsd-src/sys/netinet6/nd6_nbr.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: nd6_nbr.c,v 1.156 2018/05/19 08:22:58 maxv Exp $	*/
2 /*	$KAME: nd6_nbr.c,v 1.61 2001/02/10 16:06:14 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 __KERNEL_RCSID(0, "$NetBSD: nd6_nbr.c,v 1.156 2018/05/19 08:22:58 maxv Exp $");
35 
36 #ifdef _KERNEL_OPT
37 #include "opt_inet.h"
38 #include "opt_net_mpsafe.h"
39 #endif
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kmem.h>
44 #include <sys/mbuf.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/sockio.h>
48 #include <sys/time.h>
49 #include <sys/kernel.h>
50 #include <sys/errno.h>
51 #include <sys/ioctl.h>
52 #include <sys/syslog.h>
53 #include <sys/queue.h>
54 #include <sys/callout.h>
55 #include <sys/cprng.h>
56 
57 #include <net/if.h>
58 #include <net/if_types.h>
59 #include <net/if_dl.h>
60 #include <net/route.h>
61 
62 #include <netinet/in.h>
63 #include <netinet/in_var.h>
64 #include <netinet6/in6_var.h>
65 #include <netinet6/in6_ifattach.h>
66 #include <netinet/ip6.h>
67 #include <netinet6/ip6_var.h>
68 #include <netinet6/scope6_var.h>
69 #include <netinet6/nd6.h>
70 #include <netinet/icmp6.h>
71 #include <netinet6/icmp6_private.h>
72 
73 #include "carp.h"
74 #if NCARP > 0
75 #include <netinet/ip_carp.h>
76 #endif
77 
78 struct dadq;
79 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
80 static void nd6_dad_starttimer(struct dadq *, int);
81 static void nd6_dad_destroytimer(struct dadq *);
82 static void nd6_dad_timer(struct dadq *);
83 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
84 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
85 static void nd6_dad_na_input(struct ifaddr *);
86 static void nd6_dad_duplicated(struct dadq *);
87 
88 static int dad_maxtry = 15;	/* max # of *tries* to transmit DAD packet */
89 
90 /*
91  * Input a Neighbor Solicitation Message.
92  *
93  * Based on RFC 2461
94  * Based on RFC 2462 (duplicate address detection)
95  */
96 void
97 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
98 {
99 	struct ifnet *ifp;
100 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
101 	struct nd_neighbor_solicit *nd_ns;
102 	struct in6_addr saddr6 = ip6->ip6_src;
103 	struct in6_addr daddr6 = ip6->ip6_dst;
104 	struct in6_addr taddr6;
105 	struct in6_addr myaddr6;
106 	char *lladdr = NULL;
107 	struct ifaddr *ifa = NULL;
108 	int lladdrlen = 0;
109 	int anycast = 0, proxy = 0, tentative = 0;
110 	int router = ip6_forwarding;
111 	int tlladdr;
112 	union nd_opts ndopts;
113 	const struct sockaddr_dl *proxydl = NULL;
114 	struct psref psref;
115 	struct psref psref_ia;
116 	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
117 
118 	ifp = m_get_rcvif_psref(m, &psref);
119 	if (ifp == NULL)
120 		goto freeit;
121 
122 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
123 	if (nd_ns == NULL) {
124 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
125 		m_put_rcvif_psref(ifp, &psref);
126 		return;
127 	}
128 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
129 	taddr6 = nd_ns->nd_ns_target;
130 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
131 		goto bad;
132 
133 	if (ip6->ip6_hlim != 255) {
134 		nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
135 		    ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
136 		    IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
137 		goto bad;
138 	}
139 
140 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
141 		/* dst has to be a solicited node multicast address. */
142 		/* don't check ifindex portion */
143 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
144 		    daddr6.s6_addr32[1] == 0 &&
145 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
146 		    daddr6.s6_addr8[12] == 0xff) {
147 			; /* good */
148 		} else {
149 			nd6log(LOG_INFO, "bad DAD packet (wrong ip6 dst)\n");
150 			goto bad;
151 		}
152 	} else {
153 		struct sockaddr_in6 ssin6;
154 
155 		/*
156 		 * Make sure the source address is from a neighbor's address.
157 		 */
158 		sockaddr_in6_init(&ssin6, &saddr6, 0, 0, 0);
159 		if (nd6_is_addr_neighbor(&ssin6, ifp) == 0) {
160 			nd6log(LOG_INFO,
161 			    "NS packet from non-neighbor %s on %s\n",
162 			    IN6_PRINT(ip6buf, &saddr6), if_name(ifp));
163 			goto bad;
164 		}
165 	}
166 
167 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
168 		nd6log(LOG_INFO, "bad NS target (multicast)\n");
169 		goto bad;
170 	}
171 
172 	icmp6len -= sizeof(*nd_ns);
173 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
174 	if (nd6_options(&ndopts) < 0) {
175 		nd6log(LOG_INFO, "invalid ND option, ignored\n");
176 		/* nd6_options have incremented stats */
177 		goto freeit;
178 	}
179 
180 	if (ndopts.nd_opts_src_lladdr) {
181 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
182 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
183 	}
184 
185 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
186 		nd6log(LOG_INFO,
187 		    "bad DAD packet (link-layer address option)\n");
188 		goto bad;
189 	}
190 
191 	/*
192 	 * Attaching target link-layer address to the NA?
193 	 * (RFC 2461 7.2.4)
194 	 *
195 	 * NS IP dst is multicast			MUST add
196 	 * Otherwise					MAY be omitted
197 	 *
198 	 * In this implementation, we omit the target link-layer address
199 	 * in the "MAY" case.
200 	 */
201 #if 0 /* too much! */
202 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &daddr6);
203 	if (ifa && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST))
204 		tlladdr = 0;
205 	else
206 #endif
207 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
208 		tlladdr = 0;
209 	else
210 		tlladdr = 1;
211 
212 	/*
213 	 * Target address (taddr6) must be either:
214 	 * (1) Valid unicast/anycast address for my receiving interface,
215 	 * (2) Unicast address for which I'm offering proxy service, or
216 	 * (3) "tentative" address on which DAD is being performed.
217 	 */
218 	/* (1) and (3) check. */
219 #if NCARP > 0
220 	if (ifp->if_carp && ifp->if_type != IFT_CARP) {
221 		int s = pserialize_read_enter();
222 		ifa = carp_iamatch6(ifp->if_carp, &taddr6);
223 		if (ifa != NULL)
224 			ifa_acquire(ifa, &psref_ia);
225 		pserialize_read_exit(s);
226 	} else
227 		ifa = NULL;
228 	if (!ifa)
229 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6,
230 		    &psref_ia);
231 #else
232 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6,
233 	    &psref_ia);
234 #endif
235 
236 	/* (2) check. */
237 	if (ifa == NULL) {
238 		struct rtentry *rt;
239 		struct sockaddr_in6 tsin6;
240 
241 		sockaddr_in6_init(&tsin6, &taddr6, 0, 0, 0);
242 
243 		rt = rtalloc1(sin6tosa(&tsin6), 0);
244 		if (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
245 		    rt->rt_gateway->sa_family == AF_LINK) {
246 			/*
247 			 * proxy NDP for single entry
248 			 */
249 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp,
250 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST, &psref_ia);
251 			if (ifa) {
252 				proxy = 1;
253 				proxydl = satocsdl(rt->rt_gateway);
254 				router = 0;	/* XXX */
255 			}
256 		}
257 		if (rt)
258 			rt_unref(rt);
259 	}
260 	if (ifa == NULL) {
261 		/*
262 		 * We've got an NS packet, and we don't have that address
263 		 * assigned for us.  We MUST silently ignore it.
264 		 * See RFC2461 7.2.3.
265 		 */
266 		goto freeit;
267 	}
268 	myaddr6 = *IFA_IN6(ifa);
269 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
270 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
271 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
272 		goto freeit;
273 
274 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
275 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
276 		    "(if %d, NS packet %d)\n",
277 		    IN6_PRINT(ip6buf, &taddr6),
278 		    ifp->if_addrlen, lladdrlen - 2);
279 		goto bad;
280 	}
281 
282 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
283 		nd6log(LOG_INFO, "duplicate IP6 address %s\n",
284 		    IN6_PRINT(ip6buf, &saddr6));
285 		goto freeit;
286 	}
287 
288 	/*
289 	 * We have neighbor solicitation packet, with target address equals to
290 	 * one of my tentative address.
291 	 *
292 	 * src addr	how to process?
293 	 * ---		---
294 	 * multicast	of course, invalid (rejected in ip6_input)
295 	 * unicast	somebody is doing address resolution -> ignore
296 	 * unspec	dup address detection
297 	 *
298 	 * The processing is defined in RFC 2462.
299 	 */
300 	if (tentative) {
301 		/*
302 		 * If source address is unspecified address, it is for
303 		 * duplicate address detection.
304 		 *
305 		 * If not, the packet is for addess resolution;
306 		 * silently ignore it.
307 		 */
308 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
309 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
310 		ifa_release(ifa, &psref_ia);
311 		ifa = NULL;
312 
313 		goto freeit;
314 	}
315 	ifa_release(ifa, &psref_ia);
316 	ifa = NULL;
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(ifp, &in6_all, &taddr6,
333 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
334 		    (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
335 		    tlladdr, (const struct sockaddr *)proxydl);
336 		goto freeit;
337 	}
338 
339 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_NEIGHBOR_SOLICIT, 0);
340 
341 	nd6_na_output(ifp, &saddr6, &taddr6,
342 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
343 	    (router ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
344 	    tlladdr, (const struct sockaddr *)proxydl);
345  freeit:
346 	ifa_release(ifa, &psref_ia);
347 	m_put_rcvif_psref(ifp, &psref);
348 	m_freem(m);
349 	return;
350 
351  bad:
352 	nd6log(LOG_ERR, "src=%s\n", IN6_PRINT(ip6buf, &saddr6));
353 	nd6log(LOG_ERR, "dst=%s\n", IN6_PRINT(ip6buf, &daddr6));
354 	nd6log(LOG_ERR, "tgt=%s\n", IN6_PRINT(ip6buf, &taddr6));
355 	ICMP6_STATINC(ICMP6_STAT_BADNS);
356 	ifa_release(ifa, &psref_ia);
357 	m_put_rcvif_psref(ifp, &psref);
358 	m_freem(m);
359 }
360 
361 /*
362  * Output a Neighbor Solicitation Message. Caller specifies:
363  *	- ICMP6 header source IP6 address
364  *	- ND6 header target IP6 address
365  *	- ND6 header source datalink address
366  *
367  * Based on RFC 2461
368  * Based on RFC 2462 (duplicate address detection)
369  */
370 void
371 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
372     const struct in6_addr *taddr6,
373     struct in6_addr *hsrc,
374     uint8_t *nonce		/* duplicate address detection */)
375 {
376 	struct mbuf *m;
377 	struct ip6_hdr *ip6;
378 	struct nd_neighbor_solicit *nd_ns;
379 	struct in6_addr *src, src_in;
380 	struct ip6_moptions im6o;
381 	int icmp6len;
382 	int maxlen;
383 	const void *mac;
384 	struct route ro;
385 
386 	if (IN6_IS_ADDR_MULTICAST(taddr6))
387 		return;
388 
389 	memset(&ro, 0, sizeof(ro));
390 
391 	/* estimate the size of message */
392 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
393 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
394 	KASSERTMSG(max_linkhdr + maxlen <= MCLBYTES,
395 	    "max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
396 	    max_linkhdr, maxlen, MCLBYTES);
397 
398 	MGETHDR(m, M_DONTWAIT, MT_DATA);
399 	if (m && max_linkhdr + maxlen >= MHLEN) {
400 		MCLGET(m, M_DONTWAIT);
401 		if ((m->m_flags & M_EXT) == 0) {
402 			m_free(m);
403 			m = NULL;
404 		}
405 	}
406 	if (m == NULL)
407 		return;
408 	m_reset_rcvif(m);
409 
410 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
411 		m->m_flags |= M_MCAST;
412 		im6o.im6o_multicast_if_index = if_get_index(ifp);
413 		im6o.im6o_multicast_hlim = 255;
414 		im6o.im6o_multicast_loop = 0;
415 	}
416 
417 	icmp6len = sizeof(*nd_ns);
418 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
419 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
420 
421 	/* fill neighbor solicitation packet */
422 	ip6 = mtod(m, struct ip6_hdr *);
423 	ip6->ip6_flow = 0;
424 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
425 	ip6->ip6_vfc |= IPV6_VERSION;
426 	/* ip6->ip6_plen will be set later */
427 	ip6->ip6_nxt = IPPROTO_ICMPV6;
428 	ip6->ip6_hlim = 255;
429 	if (daddr6)
430 		ip6->ip6_dst = *daddr6;
431 	else {
432 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
433 		ip6->ip6_dst.s6_addr16[1] = 0;
434 		ip6->ip6_dst.s6_addr32[1] = 0;
435 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
436 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
437 		ip6->ip6_dst.s6_addr8[12] = 0xff;
438 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
439 			goto bad;
440 	}
441 	if (nonce == NULL) {
442 		int s;
443 		/*
444 		 * RFC2461 7.2.2:
445 		 * "If the source address of the packet prompting the
446 		 * solicitation is the same as one of the addresses assigned
447 		 * to the outgoing interface, that address SHOULD be placed
448 		 * in the IP Source Address of the outgoing solicitation.
449 		 * Otherwise, any one of the addresses assigned to the
450 		 * interface should be used."
451 		 *
452 		 * We use the source address for the prompting packet
453 		 * (hsrc), if:
454 		 * - hsrc is given from the caller (by giving "ln"), and
455 		 * - hsrc belongs to the outgoing interface.
456 		 * Otherwise, we perform the source address selection as usual.
457 		 */
458 		s = pserialize_read_enter();
459 		if (hsrc && in6ifa_ifpwithaddr(ifp, hsrc)) {
460 			pserialize_read_exit(s);
461 			src = hsrc;
462 		} else {
463 			int error;
464 			struct sockaddr_in6 dst_sa;
465 
466 			pserialize_read_exit(s);
467 
468 			sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
469 
470 			error = in6_selectsrc(&dst_sa, NULL,
471 			    NULL, &ro, NULL, NULL, NULL, &src_in);
472 			if (error != 0) {
473 				char ip6buf[INET6_ADDRSTRLEN];
474 				nd6log(LOG_DEBUG, "source can't be "
475 				    "determined: dst=%s, error=%d\n",
476 				    IN6_PRINT(ip6buf, &dst_sa.sin6_addr),
477 				    error);
478 				pserialize_read_exit(s);
479 				goto bad;
480 			}
481 			src = &src_in;
482 		}
483 	} else {
484 		/*
485 		 * Source address for DAD packet must always be IPv6
486 		 * unspecified address. (0::0)
487 		 * We actually don't have to 0-clear the address (we did it
488 		 * above), but we do so here explicitly to make the intention
489 		 * clearer.
490 		 */
491 		memset(&src_in, 0, sizeof(src_in));
492 		src = &src_in;
493 	}
494 	ip6->ip6_src = *src;
495 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
496 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
497 	nd_ns->nd_ns_code = 0;
498 	nd_ns->nd_ns_reserved = 0;
499 	nd_ns->nd_ns_target = *taddr6;
500 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
501 
502 	/*
503 	 * Add source link-layer address option.
504 	 *
505 	 *				spec		implementation
506 	 *				---		---
507 	 * DAD packet			MUST NOT	do not add the option
508 	 * there's no link layer address:
509 	 *				impossible	do not add the option
510 	 * there's link layer address:
511 	 *	Multicast NS		MUST add one	add the option
512 	 *	Unicast NS		SHOULD add one	add the option
513 	 */
514 	if (nonce == NULL && (mac = nd6_ifptomac(ifp))) {
515 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
516 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
517 		/* 8 byte alignments... */
518 		optlen = (optlen + 7) & ~7;
519 
520 		m->m_pkthdr.len += optlen;
521 		m->m_len += optlen;
522 		icmp6len += optlen;
523 		memset((void *)nd_opt, 0, optlen);
524 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
525 		nd_opt->nd_opt_len = optlen >> 3;
526 		memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
527 	}
528 
529 	/* Add a nonce option (RFC 3971) to detect looped back NS messages.
530 	 * This behavior is documented in RFC 7527. */
531 	if (nonce != NULL) {
532 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
533 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
534 
535 		/* 8-byte alignment is required. */
536 		optlen = (optlen + 7) & ~7;
537 		m->m_pkthdr.len += optlen;
538 		m->m_len += optlen;
539 		icmp6len += optlen;
540 		memset(nd_opt, 0, optlen);
541 		nd_opt->nd_opt_type = ND_OPT_NONCE;
542 		nd_opt->nd_opt_len = optlen >> 3;
543 		memcpy(nd_opt + 1, nonce, ND_OPT_NONCE_LEN);
544 	}
545 
546 	ip6->ip6_plen = htons((u_int16_t)icmp6len);
547 	nd_ns->nd_ns_cksum = 0;
548 	nd_ns->nd_ns_cksum =
549 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
550 
551 	ip6_output(m, NULL, &ro, nonce != NULL ? IPV6_UNSPECSRC : 0,
552 	    &im6o, NULL, NULL);
553 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
554 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
555 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_SOLICIT);
556 
557 	rtcache_free(&ro);
558 	return;
559 
560   bad:
561 	rtcache_free(&ro);
562 	m_freem(m);
563 	return;
564 }
565 
566 /*
567  * Neighbor advertisement input handling.
568  *
569  * Based on RFC 2461
570  * Based on RFC 2462 (duplicate address detection)
571  *
572  * the following items are not implemented yet:
573  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
574  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
575  */
576 void
577 nd6_na_input(struct mbuf *m, int off, int icmp6len)
578 {
579 	struct ifnet *ifp;
580 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
581 	struct nd_neighbor_advert *nd_na;
582 	struct in6_addr saddr6 = ip6->ip6_src;
583 	struct in6_addr daddr6 = ip6->ip6_dst;
584 	struct in6_addr taddr6;
585 	int flags;
586 	int is_router;
587 	int is_solicited;
588 	int is_override;
589 	char *lladdr = NULL;
590 	int lladdrlen = 0;
591 	struct ifaddr *ifa;
592 	struct llentry *ln = NULL;
593 	union nd_opts ndopts;
594 	struct sockaddr_in6 ssin6;
595 	int rt_announce;
596 	bool checklink = false;
597 	struct psref psref;
598 	struct psref psref_ia;
599 	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
600 
601 	ifp = m_get_rcvif_psref(m, &psref);
602 	if (ifp == NULL)
603 		goto freeit;
604 
605 	if (ip6->ip6_hlim != 255) {
606 		nd6log(LOG_ERR,
607 		    "invalid hlim (%d) from %s to %s on %s\n",
608 		    ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
609 		    IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
610 		goto bad;
611 	}
612 
613 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
614 	if (nd_na == NULL) {
615 		m_put_rcvif_psref(ifp, &psref);
616 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
617 		return;
618 	}
619 
620 	flags = nd_na->nd_na_flags_reserved;
621 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
622 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
623 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
624 
625 	taddr6 = nd_na->nd_na_target;
626 	if (in6_setscope(&taddr6, ifp, NULL)) {
627 		goto bad;
628 	}
629 
630 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
631 		nd6log(LOG_ERR, "invalid target address %s\n",
632 		    IN6_PRINT(ip6buf, &taddr6));
633 		goto bad;
634 	}
635 	if (is_solicited && IN6_IS_ADDR_MULTICAST(&daddr6)) {
636 		nd6log(LOG_ERR, "a solicited adv is multicasted\n");
637 		goto bad;
638 	}
639 
640 	icmp6len -= sizeof(*nd_na);
641 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
642 	if (nd6_options(&ndopts) < 0) {
643 		nd6log(LOG_INFO, "invalid ND option, ignored\n");
644 		/* nd6_options have incremented stats */
645 		goto freeit;
646 	}
647 
648 	if (ndopts.nd_opts_tgt_lladdr) {
649 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
650 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
651 	}
652 
653 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6, &psref_ia);
654 
655 	/*
656 	 * Target address matches one of my interface address.
657 	 *
658 	 * If my address is tentative, this means that there's somebody
659 	 * already using the same address as mine.  This indicates DAD failure.
660 	 * This is defined in RFC 2462.
661 	 *
662 	 * Otherwise, process as defined in RFC 2461.
663 	 */
664 	if (ifa
665 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
666 		nd6_dad_na_input(ifa);
667 		ifa_release(ifa, &psref_ia);
668 		ifa = NULL;
669 		goto freeit;
670 	}
671 
672 	/* Just for safety, maybe unnecessary. */
673 	if (ifa) {
674 		log(LOG_ERR,
675 		    "nd6_na_input: duplicate IP6 address %s\n",
676 		    IN6_PRINT(ip6buf, &taddr6));
677 		ifa_release(ifa, &psref_ia);
678 		ifa = NULL;
679 		goto freeit;
680 	}
681 
682 	/*
683 	 * Make sure the source address is from a neighbor's address.
684 	 */
685 	sockaddr_in6_init(&ssin6, &saddr6, 0, 0, 0);
686 	if (nd6_is_addr_neighbor(&ssin6, ifp) == 0) {
687 		nd6log(LOG_INFO, "ND packet from non-neighbor %s on %s\n",
688 		    IN6_PRINT(ip6buf, &saddr6), if_name(ifp));
689 		goto bad;
690 	}
691 
692 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
693 		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
694 		    "(if %d, NA packet %d)\n", IN6_PRINT(ip6buf, &taddr6),
695 		    ifp->if_addrlen, lladdrlen - 2);
696 		goto bad;
697 	}
698 
699 	/*
700 	 * If no neighbor cache entry is found, NA SHOULD silently be
701 	 * discarded.
702 	 */
703 	ln = nd6_lookup(&taddr6, ifp, true);
704 	if (ln == NULL)
705 		goto freeit;
706 
707 	rt_announce = 0;
708 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
709 		/*
710 		 * If the link-layer has address, and no lladdr option came,
711 		 * discard the packet.
712 		 */
713 		if (ifp->if_addrlen && !lladdr)
714 			goto freeit;
715 
716 		/*
717 		 * Record link-layer address, and update the state.
718 		 */
719 		memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
720 		ln->la_flags |= LLE_VALID;
721 		rt_announce = 1;
722 		if (is_solicited) {
723 			ln->ln_state = ND6_LLINFO_REACHABLE;
724 			ln->ln_byhint = 0;
725 			if (!ND6_LLINFO_PERMANENT(ln)) {
726 				nd6_llinfo_settimer(ln,
727 				    ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
728 			}
729 		} else {
730 			ln->ln_state = ND6_LLINFO_STALE;
731 			nd6_llinfo_settimer(ln, nd6_gctimer * hz);
732 		}
733 		if ((ln->ln_router = is_router) != 0) {
734 			/*
735 			 * This means a router's state has changed from
736 			 * non-reachable to probably reachable, and might
737 			 * affect the status of associated prefixes..
738 			 */
739 			checklink = true;
740 		}
741 	} else {
742 		int llchange;
743 
744 		/*
745 		 * Check if the link-layer address has changed or not.
746 		 */
747 		if (lladdr == NULL)
748 			llchange = 0;
749 		else {
750 			if (ln->la_flags & LLE_VALID) {
751 				if (memcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
752 					llchange = rt_announce = 1;
753 				else
754 					llchange = 0;
755 			} else
756 				llchange = rt_announce = 1;
757 		}
758 
759 		/*
760 		 * This is VERY complex.  Look at it with care.
761 		 *
762 		 * override solicit lladdr llchange	action
763 		 *					(L: record lladdr)
764 		 *
765 		 *	0	0	n	--	(2c)
766 		 *	0	0	y	n	(2b) L
767 		 *	0	0	y	y	(1)    REACHABLE->STALE
768 		 *	0	1	n	--	(2c)   *->REACHABLE
769 		 *	0	1	y	n	(2b) L *->REACHABLE
770 		 *	0	1	y	y	(1)    REACHABLE->STALE
771 		 *	1	0	n	--	(2a)
772 		 *	1	0	y	n	(2a) L
773 		 *	1	0	y	y	(2a) L *->STALE
774 		 *	1	1	n	--	(2a)   *->REACHABLE
775 		 *	1	1	y	n	(2a) L *->REACHABLE
776 		 *	1	1	y	y	(2a) L *->REACHABLE
777 		 */
778 		if (!is_override && lladdr != NULL && llchange) { /* (1) */
779 			/*
780 			 * If state is REACHABLE, make it STALE.
781 			 * no other updates should be done.
782 			 */
783 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
784 				ln->ln_state = ND6_LLINFO_STALE;
785 				nd6_llinfo_settimer(ln, nd6_gctimer * hz);
786 			}
787 			goto freeit;
788 		} else if (is_override				   /* (2a) */
789 		    || (!is_override && lladdr != NULL && !llchange) /* (2b) */
790 		    || lladdr == NULL) {			   /* (2c) */
791 			/*
792 			 * Update link-local address, if any.
793 			 */
794 			if (lladdr != NULL) {
795 				memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
796 				ln->la_flags |= LLE_VALID;
797 			}
798 
799 			/*
800 			 * If solicited, make the state REACHABLE.
801 			 * If not solicited and the link-layer address was
802 			 * changed, make it STALE.
803 			 */
804 			if (is_solicited) {
805 				ln->ln_state = ND6_LLINFO_REACHABLE;
806 				ln->ln_byhint = 0;
807 				if (!ND6_LLINFO_PERMANENT(ln)) {
808 					nd6_llinfo_settimer(ln,
809 					    ND_IFINFO(ifp)->reachable * hz);
810 				}
811 			} else {
812 				if (lladdr && llchange) {
813 					ln->ln_state = ND6_LLINFO_STALE;
814 					nd6_llinfo_settimer(ln,
815 					    nd6_gctimer * hz);
816 				}
817 			}
818 		}
819 
820 		if (ln->ln_router && !is_router) {
821 			/*
822 			 * The peer dropped the router flag.
823 			 * Remove the sender from the Default Router List and
824 			 * update the Destination Cache entries.
825 			 */
826 			struct nd_defrouter *dr;
827 			const struct in6_addr *in6;
828 
829 			in6 = &ln->r_l3addr.addr6;
830 
831 			ND6_WLOCK();
832 			dr = nd6_defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
833 			if (dr)
834 				nd6_defrtrlist_del(dr, NULL);
835 			else if (!ip6_forwarding) {
836 				/*
837 				 * Even if the neighbor is not in the default
838 				 * router list, the neighbor may be used
839 				 * as a next hop for some destinations
840 				 * (e.g. redirect case). So we must
841 				 * call nd6_rt_flush explicitly.
842 				 */
843 				nd6_rt_flush(&ip6->ip6_src, ln->lle_tbl->llt_ifp);
844 			}
845 			ND6_UNLOCK();
846 		}
847 		ln->ln_router = is_router;
848 	}
849         /*
850 	 * XXX: does this matter?
851 	 * rt->rt_flags &= ~RTF_REJECT;
852 	 */
853 	ln->ln_asked = 0;
854 	nd6_llinfo_release_pkts(ln, ifp);
855 	/* FIXME */
856 #if 0
857 	if (rt_announce) /* tell user process about any new lladdr */
858 		rt_newmsg(RTM_CHANGE, rt);
859 #endif
860 
861  freeit:
862 	if (ln != NULL)
863 		LLE_WUNLOCK(ln);
864 
865 	if (checklink) {
866 		ND6_WLOCK();
867 		nd6_pfxlist_onlink_check();
868 		ND6_UNLOCK();
869 	}
870 
871 	m_put_rcvif_psref(ifp, &psref);
872 	m_freem(m);
873 	return;
874 
875  bad:
876 	if (ln != NULL)
877 		LLE_WUNLOCK(ln);
878 
879 	ICMP6_STATINC(ICMP6_STAT_BADNA);
880 	m_put_rcvif_psref(ifp, &psref);
881 	m_freem(m);
882 }
883 
884 /*
885  * Neighbor advertisement output handling.
886  *
887  * Based on RFC 2461
888  *
889  * the following items are not implemented yet:
890  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
891  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
892  */
893 void
894 nd6_na_output(
895 	struct ifnet *ifp,
896 	const struct in6_addr *daddr6_0,
897 	const struct in6_addr *taddr6,
898 	u_long flags,
899 	int tlladdr,		/* 1 if include target link-layer address */
900 	const struct sockaddr *sdl0)	/* sockaddr_dl (= proxy NA) or NULL */
901 {
902 	struct mbuf *m;
903 	struct ip6_hdr *ip6;
904 	struct nd_neighbor_advert *nd_na;
905 	struct ip6_moptions im6o;
906 	struct sockaddr *dst;
907 	union {
908 		struct sockaddr		dst;
909 		struct sockaddr_in6	dst6;
910 	} u;
911 	struct in6_addr daddr6;
912 	int icmp6len, maxlen, error;
913 	const void *mac;
914 	struct route ro;
915 
916 	mac = NULL;
917 	memset(&ro, 0, sizeof(ro));
918 
919 	daddr6 = *daddr6_0;	/* make a local copy for modification */
920 
921 	/* estimate the size of message */
922 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
923 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
924 	KASSERTMSG(max_linkhdr + maxlen <= MCLBYTES,
925 	    "max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
926 	    max_linkhdr, maxlen, MCLBYTES);
927 
928 	MGETHDR(m, M_DONTWAIT, MT_DATA);
929 	if (m && max_linkhdr + maxlen >= MHLEN) {
930 		MCLGET(m, M_DONTWAIT);
931 		if ((m->m_flags & M_EXT) == 0) {
932 			m_free(m);
933 			m = NULL;
934 		}
935 	}
936 	if (m == NULL)
937 		return;
938 	m_reset_rcvif(m);
939 
940 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
941 		m->m_flags |= M_MCAST;
942 		im6o.im6o_multicast_if_index = if_get_index(ifp);
943 		im6o.im6o_multicast_hlim = 255;
944 		im6o.im6o_multicast_loop = 0;
945 	}
946 
947 	icmp6len = sizeof(*nd_na);
948 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
949 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
950 
951 	/* fill neighbor advertisement packet */
952 	ip6 = mtod(m, struct ip6_hdr *);
953 	ip6->ip6_flow = 0;
954 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
955 	ip6->ip6_vfc |= IPV6_VERSION;
956 	ip6->ip6_nxt = IPPROTO_ICMPV6;
957 	ip6->ip6_hlim = 255;
958 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
959 		/* reply to DAD */
960 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
961 		daddr6.s6_addr16[1] = 0;
962 		daddr6.s6_addr32[1] = 0;
963 		daddr6.s6_addr32[2] = 0;
964 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
965 		if (in6_setscope(&daddr6, ifp, NULL))
966 			goto bad;
967 
968 		flags &= ~ND_NA_FLAG_SOLICITED;
969 	}
970 	ip6->ip6_dst = daddr6;
971 	sockaddr_in6_init(&u.dst6, &daddr6, 0, 0, 0);
972 	dst = &u.dst;
973 	if (rtcache_setdst(&ro, dst) != 0)
974 		goto bad;
975 
976 	/*
977 	 * Select a source whose scope is the same as that of the dest.
978 	 */
979 	error = in6_selectsrc(satosin6(dst), NULL, NULL, &ro, NULL, NULL, NULL,
980 	    &ip6->ip6_src);
981 	if (error != 0) {
982 		char ip6buf[INET6_ADDRSTRLEN];
983 		nd6log(LOG_DEBUG, "source can't be "
984 		    "determined: dst=%s, error=%d\n",
985 		    IN6_PRINT(ip6buf, &satocsin6(dst)->sin6_addr), error);
986 		goto bad;
987 	}
988 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
989 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
990 	nd_na->nd_na_code = 0;
991 	nd_na->nd_na_target = *taddr6;
992 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
993 
994 	/*
995 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
996 	 * see nd6_ns_input() for details.
997 	 * Basically, if NS packet is sent to unicast/anycast addr,
998 	 * target lladdr option SHOULD NOT be included.
999 	 */
1000 	if (tlladdr) {
1001 		/*
1002 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1003 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1004 		 * my address) use lladdr configured for the interface.
1005 		 */
1006 		if (sdl0 == NULL)
1007 			mac = nd6_ifptomac(ifp);
1008 		else if (sdl0->sa_family == AF_LINK) {
1009 			const struct sockaddr_dl *sdl;
1010 			sdl = satocsdl(sdl0);
1011 			if (sdl->sdl_alen == ifp->if_addrlen)
1012 				mac = CLLADDR(sdl);
1013 		}
1014 	}
1015 	if (tlladdr && mac) {
1016 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1017 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1018 
1019 		/* roundup to 8 bytes alignment! */
1020 		optlen = (optlen + 7) & ~7;
1021 
1022 		m->m_pkthdr.len += optlen;
1023 		m->m_len += optlen;
1024 		icmp6len += optlen;
1025 		memset((void *)nd_opt, 0, optlen);
1026 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1027 		nd_opt->nd_opt_len = optlen >> 3;
1028 		memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
1029 	} else
1030 		flags &= ~ND_NA_FLAG_OVERRIDE;
1031 
1032 	ip6->ip6_plen = htons((u_int16_t)icmp6len);
1033 	nd_na->nd_na_flags_reserved = flags;
1034 	nd_na->nd_na_cksum = 0;
1035 	nd_na->nd_na_cksum =
1036 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1037 
1038 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1039 
1040 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1041 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1042 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_ADVERT);
1043 
1044 	rtcache_free(&ro);
1045 	return;
1046 
1047   bad:
1048 	rtcache_free(&ro);
1049 	m_freem(m);
1050 	return;
1051 }
1052 
1053 const void *
1054 nd6_ifptomac(const struct ifnet *ifp)
1055 {
1056 	switch (ifp->if_type) {
1057 	case IFT_ARCNET:
1058 	case IFT_ETHER:
1059 	case IFT_FDDI:
1060 	case IFT_IEEE1394:
1061 	case IFT_PROPVIRTUAL:
1062 	case IFT_CARP:
1063 	case IFT_L2VLAN:
1064 	case IFT_IEEE80211:
1065 		return CLLADDR(ifp->if_sadl);
1066 	default:
1067 		return NULL;
1068 	}
1069 }
1070 
1071 TAILQ_HEAD(dadq_head, dadq);
1072 struct dadq {
1073 	TAILQ_ENTRY(dadq) dad_list;
1074 	struct ifaddr *dad_ifa;
1075 	int dad_count;			/* max NS to send */
1076 	int dad_ns_tcount;		/* # of trials to send NS */
1077 	int dad_ns_ocount;		/* NS sent so far */
1078 	int dad_ns_icount;
1079 	int dad_na_icount;
1080 	int dad_ns_lcount;		/* looped back NS */
1081 	struct callout dad_timer_ch;
1082 #define	ND_OPT_NONCE_STORE	3	/* dad_count should not exceed this */
1083 	/*
1084 	 * The default ip6_dad_count is 1 as specified by RFC 4862 and
1085 	 * practically must users won't exceed this.
1086 	 * A storage of 3 is defaulted to here, in-case the administrator wants
1087 	 * to match the equivalent behaviour in our ARP implementation.
1088 	 * This constraint could be removed by sending the on wire nonce as
1089 	 * hmac(key, dad_ns_ocount), but that would increase the nonce size
1090 	 * sent on the wire.
1091 	 */
1092 	uint8_t dad_nonce[ND_OPT_NONCE_STORE][ND_OPT_NONCE_LEN];
1093 };
1094 
1095 static struct dadq_head dadq;
1096 static int dad_init = 0;
1097 static kmutex_t nd6_dad_lock;
1098 
1099 static struct dadq *
1100 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1101 {
1102 	struct dadq *dp;
1103 	int i, nonce_max;
1104 
1105 	KASSERT(mutex_owned(&nd6_dad_lock));
1106 
1107 	TAILQ_FOREACH(dp, &dadq, dad_list) {
1108 		if (dp->dad_ifa != ifa)
1109 			continue;
1110 
1111 		if (nonce == NULL ||
1112 		    nonce->nd_opt_nonce_len != (ND_OPT_NONCE_LEN + 2) / 8)
1113 			break;
1114 
1115 		nonce_max = MIN(dp->dad_ns_ocount, ND_OPT_NONCE_STORE);
1116 		for (i = 0; i < nonce_max; i++) {
1117 			if (memcmp(nonce->nd_opt_nonce,
1118 			    dp->dad_nonce[i],
1119 			    ND_OPT_NONCE_LEN) == 0)
1120 				break;
1121 		}
1122 		if (i < nonce_max) {
1123 			char ip6buf[INET6_ADDRSTRLEN];
1124 
1125 			log(LOG_DEBUG,
1126 			    "%s: detected a looped back NS message for %s\n",
1127 			    ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???",
1128 			    IN6_PRINT(ip6buf, IFA_IN6(ifa)));
1129 			dp->dad_ns_lcount++;
1130 			continue;
1131 		}
1132 
1133 		break;
1134 	}
1135 	return dp;
1136 }
1137 
1138 static void
1139 nd6_dad_starttimer(struct dadq *dp, int ticks)
1140 {
1141 
1142 	callout_reset(&dp->dad_timer_ch, ticks,
1143 	    (void (*)(void *))nd6_dad_timer, dp);
1144 }
1145 
1146 static void
1147 nd6_dad_stoptimer(struct dadq *dp)
1148 {
1149 
1150 	KASSERT(mutex_owned(&nd6_dad_lock));
1151 
1152 	TAILQ_REMOVE(&dadq, dp, dad_list);
1153 	/* Tell the timer that dp is being destroyed. */
1154 	dp->dad_ifa = NULL;
1155 	callout_halt(&dp->dad_timer_ch, &nd6_dad_lock);
1156 }
1157 
1158 static void
1159 nd6_dad_destroytimer(struct dadq *dp)
1160 {
1161 
1162 	KASSERT(dp->dad_ifa == NULL);
1163 	callout_destroy(&dp->dad_timer_ch);
1164 	kmem_intr_free(dp, sizeof(*dp));
1165 }
1166 
1167 /*
1168  * Start Duplicate Address Detection (DAD) for specified interface address.
1169  *
1170  * Note that callout is used when xtick > 0 and not when xtick == 0.
1171  *
1172  * xtick: minimum delay ticks for IFF_UP event
1173  */
1174 void
1175 nd6_dad_start(struct ifaddr *ifa, int xtick)
1176 {
1177 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1178 	struct dadq *dp;
1179 	char ip6buf[INET6_ADDRSTRLEN];
1180 
1181 	if (!dad_init) {
1182 		TAILQ_INIT(&dadq);
1183 		mutex_init(&nd6_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1184 		dad_init++;
1185 	}
1186 
1187 	/*
1188 	 * If we don't need DAD, don't do it.
1189 	 * There are several cases:
1190 	 * - DAD is disabled (ip6_dad_count == 0)
1191 	 * - the interface address is anycast
1192 	 */
1193 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1194 		log(LOG_DEBUG,
1195 			"nd6_dad_start: called with non-tentative address "
1196 			"%s(%s)\n",
1197 			IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1198 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1199 		return;
1200 	}
1201 	if (ia->ia6_flags & IN6_IFF_ANYCAST || !ip6_dad_count) {
1202 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1203 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1204 		return;
1205 	}
1206 	KASSERT(ifa->ifa_ifp != NULL);
1207 	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1208 		return;
1209 
1210 	dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1211 
1212 	mutex_enter(&nd6_dad_lock);
1213 	if (nd6_dad_find(ifa, NULL) != NULL) {
1214 		mutex_exit(&nd6_dad_lock);
1215 		/* DAD already in progress */
1216 		if (dp != NULL)
1217 			kmem_intr_free(dp, sizeof(*dp));
1218 		return;
1219 	}
1220 
1221 	if (dp == NULL) {
1222 		mutex_exit(&nd6_dad_lock);
1223 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1224 			"%s(%s)\n",
1225 			IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1226 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1227 		return;
1228 	}
1229 
1230 	/*
1231 	 * Send NS packet for DAD, ip6_dad_count times.
1232 	 * Note that we must delay the first transmission, if this is the
1233 	 * first packet to be sent from the interface after interface
1234 	 * (re)initialization.
1235 	 */
1236 	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1237 	dp->dad_ifa = ifa;
1238 	ifaref(ifa);	/* just for safety */
1239 	dp->dad_count = ip6_dad_count;
1240 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1241 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1242 	dp->dad_ns_lcount = 0;
1243 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1244 
1245 	nd6log(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1246 	    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1247 
1248 	if (xtick == 0) {
1249 		nd6_dad_ns_output(dp, ifa);
1250 		nd6_dad_starttimer(dp,
1251 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1252 	} else
1253 		nd6_dad_starttimer(dp, xtick);
1254 	mutex_exit(&nd6_dad_lock);
1255 }
1256 
1257 /*
1258  * terminate DAD unconditionally.  used for address removals.
1259  */
1260 void
1261 nd6_dad_stop(struct ifaddr *ifa)
1262 {
1263 	struct dadq *dp;
1264 
1265 	if (!dad_init)
1266 		return;
1267 
1268 	mutex_enter(&nd6_dad_lock);
1269 	dp = nd6_dad_find(ifa, NULL);
1270 	if (dp == NULL) {
1271 		mutex_exit(&nd6_dad_lock);
1272 		/* DAD wasn't started yet */
1273 		return;
1274 	}
1275 
1276 	/* Prevent the timer from running anymore. */
1277 	nd6_dad_stoptimer(dp);
1278 
1279 	mutex_exit(&nd6_dad_lock);
1280 
1281 	nd6_dad_destroytimer(dp);
1282 	ifafree(ifa);
1283 }
1284 
1285 static void
1286 nd6_dad_timer(struct dadq *dp)
1287 {
1288 	struct ifaddr *ifa;
1289 	struct in6_ifaddr *ia;
1290 	int duplicate = 0;
1291 	char ip6buf[INET6_ADDRSTRLEN];
1292 	bool need_free = false;
1293 
1294 	KERNEL_LOCK_UNLESS_NET_MPSAFE();
1295 	mutex_enter(&nd6_dad_lock);
1296 
1297 	ifa = dp->dad_ifa;
1298 	if (ifa == NULL) {
1299 		/* dp is being destroyed by someone.  Do nothing. */
1300 		goto done;
1301 	}
1302 
1303 	ia = (struct in6_ifaddr *)ifa;
1304 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1305 		log(LOG_ERR, "nd6_dad_timer: called with duplicate address "
1306 			"%s(%s)\n",
1307 			IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1308 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1309 		goto done;
1310 	}
1311 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1312 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1313 			"%s(%s)\n",
1314 			IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1315 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1316 		goto done;
1317 	}
1318 
1319 	/* timeouted with IFF_{RUNNING,UP} check */
1320 	if (dp->dad_ns_tcount > dad_maxtry) {
1321 		nd6log(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1322 			if_name(ifa->ifa_ifp));
1323 
1324 		nd6_dad_stoptimer(dp);
1325 		need_free = true;
1326 		goto done;
1327 	}
1328 
1329 	/* Need more checks? */
1330 	if (dp->dad_ns_ocount < dp->dad_count) {
1331 		/*
1332 		 * We have more NS to go.  Send NS packet for DAD.
1333 		 */
1334 		nd6_dad_ns_output(dp, ifa);
1335 		nd6_dad_starttimer(dp,
1336 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1337 	} else {
1338 		/*
1339 		 * We have transmitted sufficient number of DAD packets.
1340 		 * See what we've got.
1341 		 */
1342 		if (dp->dad_na_icount) {
1343 			/*
1344 			 * the check is in nd6_dad_na_input(),
1345 			 * but just in case
1346 			 */
1347 			duplicate++;
1348 		}
1349 
1350 		if (dp->dad_ns_icount) {
1351 			/* We've seen NS, means DAD has failed. */
1352 			duplicate++;
1353 		}
1354 
1355 		if (duplicate) {
1356 			nd6_dad_duplicated(dp);
1357 			nd6_dad_stoptimer(dp);
1358 			need_free = true;
1359 		} else {
1360 			/*
1361 			 * We are done with DAD.  No NA came, no NS came.
1362 			 * No duplicate address found.
1363 			 */
1364 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1365 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1366 
1367 			nd6log(LOG_DEBUG,
1368 			    "%s: DAD complete for %s - no duplicates found\n",
1369 			    if_name(ifa->ifa_ifp),
1370 			    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1371 
1372 			nd6_dad_stoptimer(dp);
1373 			need_free = true;
1374 		}
1375 	}
1376 done:
1377 	mutex_exit(&nd6_dad_lock);
1378 
1379 	if (need_free) {
1380 		nd6_dad_destroytimer(dp);
1381 		KASSERT(ifa != NULL);
1382 		ifafree(ifa);
1383 	}
1384 
1385 	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1386 }
1387 
1388 static void
1389 nd6_dad_duplicated(struct dadq *dp)
1390 {
1391 	struct ifaddr *ifa = dp->dad_ifa;
1392 	struct in6_ifaddr *ia;
1393 	struct ifnet *ifp;
1394 	char ip6buf[INET6_ADDRSTRLEN];
1395 
1396 	KASSERT(mutex_owned(&nd6_dad_lock));
1397 	KASSERT(ifa != NULL);
1398 
1399 	ifp = ifa->ifa_ifp;
1400 	ia = (struct in6_ifaddr *)ifa;
1401 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1402 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1403 	    if_name(ifp), IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1404 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1405 	    dp->dad_na_icount);
1406 
1407 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1408 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1409 
1410 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1411 	    if_name(ifp), IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1412 	log(LOG_ERR, "%s: manual intervention required\n",
1413 	    if_name(ifp));
1414 
1415 	/* Inform the routing socket that DAD has completed */
1416 	rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
1417 
1418 	/*
1419 	 * If the address is a link-local address formed from an interface
1420 	 * identifier based on the hardware address which is supposed to be
1421 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1422 	 * operation on the interface SHOULD be disabled.
1423 	 * [rfc2462bis-03 Section 5.4.5]
1424 	 */
1425 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1426 		struct in6_addr in6;
1427 
1428 		/*
1429 		 * To avoid over-reaction, we only apply this logic when we are
1430 		 * very sure that hardware addresses are supposed to be unique.
1431 		 */
1432 		switch (ifp->if_type) {
1433 		case IFT_ETHER:
1434 		case IFT_FDDI:
1435 		case IFT_ATM:
1436 		case IFT_IEEE1394:
1437 		case IFT_IEEE80211:
1438 			in6 = ia->ia_addr.sin6_addr;
1439 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1440 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1441 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1442 				log(LOG_ERR, "%s: possible hardware address "
1443 				    "duplication detected, disable IPv6\n",
1444 				    if_name(ifp));
1445 			}
1446 			break;
1447 		}
1448 	}
1449 }
1450 
1451 static void
1452 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1453 {
1454 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1455 	struct ifnet *ifp = ifa->ifa_ifp;
1456 	uint8_t *nonce;
1457 
1458 	dp->dad_ns_tcount++;
1459 	if ((ifp->if_flags & IFF_UP) == 0) {
1460 #if 0
1461 		printf("%s: interface down?\n", if_name(ifp));
1462 #endif
1463 		return;
1464 	}
1465 	if ((ifp->if_flags & IFF_RUNNING) == 0) {
1466 #if 0
1467 		printf("%s: interface not running?\n", if_name(ifp));
1468 #endif
1469 		return;
1470 	}
1471 
1472 	dp->dad_ns_tcount = 0;
1473 	nonce = dp->dad_nonce[dp->dad_ns_ocount % ND_OPT_NONCE_STORE];
1474 	cprng_fast(nonce, ND_OPT_NONCE_LEN);
1475 	dp->dad_ns_ocount++;
1476 
1477 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, nonce);
1478 }
1479 
1480 static void
1481 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1482 {
1483 	struct dadq *dp;
1484 	int duplicate;
1485 
1486 	if (ifa == NULL)
1487 		panic("ifa == NULL in nd6_dad_ns_input");
1488 
1489 	duplicate = 0;
1490 
1491 	mutex_enter(&nd6_dad_lock);
1492 	dp = nd6_dad_find(ifa, nonce);
1493 
1494 	/*
1495 	 * if I'm yet to start DAD, someone else started using this address
1496 	 * first.  I have a duplicate and you win.
1497 	 */
1498 	if (dp == NULL || dp->dad_ns_ocount == 0)
1499 		duplicate++;
1500 
1501 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1502 
1503 	if (duplicate) {
1504 		if (dp) {
1505 			nd6_dad_duplicated(dp);
1506 			nd6_dad_stoptimer(dp);
1507 		}
1508 	} else {
1509 		/*
1510 		 * not sure if I got a duplicate.
1511 		 * increment ns count and see what happens.
1512 		 */
1513 		if (dp)
1514 			dp->dad_ns_icount++;
1515 	}
1516 	mutex_exit(&nd6_dad_lock);
1517 
1518 	if (duplicate && dp) {
1519 		nd6_dad_destroytimer(dp);
1520 		ifafree(ifa);
1521 	}
1522 }
1523 
1524 static void
1525 nd6_dad_na_input(struct ifaddr *ifa)
1526 {
1527 	struct dadq *dp;
1528 
1529 	KASSERT(ifa != NULL);
1530 
1531 	mutex_enter(&nd6_dad_lock);
1532 
1533 	dp = nd6_dad_find(ifa, NULL);
1534 	if (dp == NULL) {
1535 		mutex_exit(&nd6_dad_lock);
1536 		return;
1537 	}
1538 
1539 	dp->dad_na_icount++;
1540 
1541 	/* remove the address. */
1542 	nd6_dad_duplicated(dp);
1543 	nd6_dad_stoptimer(dp);
1544 
1545 	mutex_exit(&nd6_dad_lock);
1546 
1547 	nd6_dad_destroytimer(dp);
1548 	ifafree(ifa);
1549 }
1550