xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 35a04710d7286aa9538917fd7f8e417dbee95b82)
1 /*	$FreeBSD$	*/
2 /*	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 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 "opt_inet.h"
34 #include "opt_inet6.h"
35 #include "opt_ipsec.h"
36 #include "opt_carp.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/socket.h>
43 #include <sys/sockio.h>
44 #include <sys/time.h>
45 #include <sys/kernel.h>
46 #include <sys/errno.h>
47 #include <sys/syslog.h>
48 #include <sys/queue.h>
49 #include <sys/callout.h>
50 
51 #include <net/if.h>
52 #include <net/if_types.h>
53 #include <net/if_dl.h>
54 #include <net/if_var.h>
55 #include <net/route.h>
56 
57 #include <netinet/in.h>
58 #include <netinet/in_var.h>
59 #include <netinet6/in6_var.h>
60 #include <netinet6/in6_ifattach.h>
61 #include <netinet/ip6.h>
62 #include <netinet6/ip6_var.h>
63 #include <netinet6/scope6_var.h>
64 #include <netinet6/nd6.h>
65 #include <netinet/icmp6.h>
66 
67 #ifdef DEV_CARP
68 #include <netinet/ip_carp.h>
69 #endif
70 
71 #define SDL(s) ((struct sockaddr_dl *)s)
72 
73 struct dadq;
74 static struct dadq *nd6_dad_find __P((struct ifaddr *));
75 static void nd6_dad_starttimer __P((struct dadq *, int));
76 static void nd6_dad_stoptimer __P((struct dadq *));
77 static void nd6_dad_timer __P((struct ifaddr *));
78 static void nd6_dad_ns_output __P((struct dadq *, struct ifaddr *));
79 static void nd6_dad_ns_input __P((struct ifaddr *));
80 static void nd6_dad_na_input __P((struct ifaddr *));
81 
82 static int dad_ignore_ns = 0;	/* ignore NS in DAD - specwise incorrect*/
83 static int dad_maxtry = 15;	/* max # of *tries* to transmit DAD packet */
84 
85 /*
86  * Input a Neighbor Solicitation Message.
87  *
88  * Based on RFC 2461
89  * Based on RFC 2462 (duplicate address detection)
90  */
91 void
92 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
93 {
94 	struct ifnet *ifp = m->m_pkthdr.rcvif;
95 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
96 	struct nd_neighbor_solicit *nd_ns;
97 	struct in6_addr saddr6 = ip6->ip6_src;
98 	struct in6_addr daddr6 = ip6->ip6_dst;
99 	struct in6_addr taddr6;
100 	struct in6_addr myaddr6;
101 	char *lladdr = NULL;
102 	struct ifaddr *ifa = NULL;
103 	int lladdrlen = 0;
104 	int anycast = 0, proxy = 0, tentative = 0;
105 	int tlladdr;
106 	union nd_opts ndopts;
107 	struct sockaddr_dl *proxydl = NULL;
108 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
109 
110 #ifndef PULLDOWN_TEST
111 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
112 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
113 #else
114 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
115 	if (nd_ns == NULL) {
116 		icmp6stat.icp6s_tooshort++;
117 		return;
118 	}
119 #endif
120 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
121 	taddr6 = nd_ns->nd_ns_target;
122 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
123 		goto bad;
124 
125 	if (ip6->ip6_hlim != 255) {
126 		nd6log((LOG_ERR,
127 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
128 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
129 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
130 		goto bad;
131 	}
132 
133 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
134 		/* dst has to be a solicited node multicast address. */
135 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
136 		    /* don't check ifindex portion */
137 		    daddr6.s6_addr32[1] == 0 &&
138 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
139 		    daddr6.s6_addr8[12] == 0xff) {
140 			; /* good */
141 		} else {
142 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
143 			    "(wrong ip6 dst)\n"));
144 			goto bad;
145 		}
146 	}
147 
148 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
149 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
150 		goto bad;
151 	}
152 
153 	icmp6len -= sizeof(*nd_ns);
154 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
155 	if (nd6_options(&ndopts) < 0) {
156 		nd6log((LOG_INFO,
157 		    "nd6_ns_input: invalid ND option, ignored\n"));
158 		/* nd6_options have incremented stats */
159 		goto freeit;
160 	}
161 
162 	if (ndopts.nd_opts_src_lladdr) {
163 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
164 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
165 	}
166 
167 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
168 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
169 		    "(link-layer address option)\n"));
170 		goto bad;
171 	}
172 
173 	/*
174 	 * Attaching target link-layer address to the NA?
175 	 * (RFC 2461 7.2.4)
176 	 *
177 	 * NS IP dst is unicast/anycast			MUST NOT add
178 	 * NS IP dst is solicited-node multicast	MUST add
179 	 *
180 	 * In implementation, we add target link-layer address by default.
181 	 * We do not add one in MUST NOT cases.
182 	 */
183 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
184 		tlladdr = 0;
185 	else
186 		tlladdr = 1;
187 
188 	/*
189 	 * Target address (taddr6) must be either:
190 	 * (1) Valid unicast/anycast address for my receiving interface,
191 	 * (2) Unicast address for which I'm offering proxy service, or
192 	 * (3) "tentative" address on which DAD is being performed.
193 	 */
194 	/* (1) and (3) check. */
195 #ifdef DEV_CARP
196 	if (ifp->if_carp)
197 		ifa = carp_iamatch6(ifp->if_carp, &taddr6);
198 	if (ifa == NULL)
199 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
200 #else
201 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
202 #endif
203 
204 	/* (2) check. */
205 	if (ifa == NULL) {
206 		struct rtentry *rt;
207 		struct sockaddr_in6 tsin6;
208 		int need_proxy;
209 
210 		bzero(&tsin6, sizeof tsin6);
211 		tsin6.sin6_len = sizeof(struct sockaddr_in6);
212 		tsin6.sin6_family = AF_INET6;
213 		tsin6.sin6_addr = taddr6;
214 
215 		rt = rtalloc1((struct sockaddr *)&tsin6, 0, 0);
216 		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
217 		    rt->rt_gateway->sa_family == AF_LINK);
218 		if (rt)
219 			rtfree(rt);
220 		if (need_proxy) {
221 			/*
222 			 * proxy NDP for single entry
223 			 */
224 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
225 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
226 			if (ifa) {
227 				proxy = 1;
228 				proxydl = SDL(rt->rt_gateway);
229 			}
230 		}
231 	}
232 	if (ifa == NULL) {
233 		/*
234 		 * We've got an NS packet, and we don't have that adddress
235 		 * assigned for us.  We MUST silently ignore it.
236 		 * See RFC2461 7.2.3.
237 		 */
238 		goto freeit;
239 	}
240 	myaddr6 = *IFA_IN6(ifa);
241 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
242 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
243 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
244 		goto freeit;
245 
246 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
247 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
248 		    "(if %d, NS packet %d)\n",
249 		    ip6_sprintf(ip6bufs, &taddr6),
250 		    ifp->if_addrlen, lladdrlen - 2));
251 		goto bad;
252 	}
253 
254 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
255 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
256 		    ip6_sprintf(ip6bufs, &saddr6)));
257 		goto freeit;
258 	}
259 
260 	/*
261 	 * We have neighbor solicitation packet, with target address equals to
262 	 * one of my tentative address.
263 	 *
264 	 * src addr	how to process?
265 	 * ---		---
266 	 * multicast	of course, invalid (rejected in ip6_input)
267 	 * unicast	somebody is doing address resolution -> ignore
268 	 * unspec	dup address detection
269 	 *
270 	 * The processing is defined in RFC 2462.
271 	 */
272 	if (tentative) {
273 		/*
274 		 * If source address is unspecified address, it is for
275 		 * duplicate address detection.
276 		 *
277 		 * If not, the packet is for addess resolution;
278 		 * silently ignore it.
279 		 */
280 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
281 			nd6_dad_ns_input(ifa);
282 
283 		goto freeit;
284 	}
285 
286 	/*
287 	 * If the source address is unspecified address, entries must not
288 	 * be created or updated.
289 	 * It looks that sender is performing DAD.  Output NA toward
290 	 * all-node multicast address, to tell the sender that I'm using
291 	 * the address.
292 	 * S bit ("solicited") must be zero.
293 	 */
294 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
295 		struct in6_addr in6_all;
296 
297 		in6_all = in6addr_linklocal_allnodes;
298 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
299 			goto bad;
300 		nd6_na_output(ifp, &in6_all, &taddr6,
301 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
302 		    (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
303 		    tlladdr, (struct sockaddr *)proxydl);
304 		goto freeit;
305 	}
306 
307 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
308 	    ND_NEIGHBOR_SOLICIT, 0);
309 
310 	nd6_na_output(ifp, &saddr6, &taddr6,
311 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
312 	    (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
313 	    tlladdr, (struct sockaddr *)proxydl);
314  freeit:
315 	m_freem(m);
316 	return;
317 
318  bad:
319 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
320 		ip6_sprintf(ip6bufs, &saddr6)));
321 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
322 		ip6_sprintf(ip6bufs, &daddr6)));
323 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
324 		ip6_sprintf(ip6bufs, &taddr6)));
325 	icmp6stat.icp6s_badns++;
326 	m_freem(m);
327 }
328 
329 /*
330  * Output a Neighbor Solicitation Message. Caller specifies:
331  *	- ICMP6 header source IP6 address
332  *	- ND6 header target IP6 address
333  *	- ND6 header source datalink address
334  *
335  * Based on RFC 2461
336  * Based on RFC 2462 (duplicate address detection)
337  *
338  *   ln - for source address determination
339  *  dad - duplicate address detection
340  */
341 void
342 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
343     const struct in6_addr *taddr6, struct llinfo_nd6 *ln, int dad)
344 {
345 	struct mbuf *m;
346 	struct ip6_hdr *ip6;
347 	struct nd_neighbor_solicit *nd_ns;
348 	struct in6_addr *src, src_in;
349 	struct ip6_moptions im6o;
350 	int icmp6len;
351 	int maxlen;
352 	caddr_t mac;
353 	struct route_in6 ro;
354 
355 	bzero(&ro, sizeof(ro));
356 
357 	if (IN6_IS_ADDR_MULTICAST(taddr6))
358 		return;
359 
360 	/* estimate the size of message */
361 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
362 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
363 	if (max_linkhdr + maxlen >= MCLBYTES) {
364 #ifdef DIAGNOSTIC
365 		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
366 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
367 #endif
368 		return;
369 	}
370 
371 	MGETHDR(m, M_DONTWAIT, MT_DATA);
372 	if (m && max_linkhdr + maxlen >= MHLEN) {
373 		MCLGET(m, M_DONTWAIT);
374 		if ((m->m_flags & M_EXT) == 0) {
375 			m_free(m);
376 			m = NULL;
377 		}
378 	}
379 	if (m == NULL)
380 		return;
381 	m->m_pkthdr.rcvif = NULL;
382 
383 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
384 		m->m_flags |= M_MCAST;
385 		im6o.im6o_multicast_ifp = ifp;
386 		im6o.im6o_multicast_hlim = 255;
387 		im6o.im6o_multicast_loop = 0;
388 	}
389 
390 	icmp6len = sizeof(*nd_ns);
391 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
392 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
393 
394 	/* fill neighbor solicitation packet */
395 	ip6 = mtod(m, struct ip6_hdr *);
396 	ip6->ip6_flow = 0;
397 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
398 	ip6->ip6_vfc |= IPV6_VERSION;
399 	/* ip6->ip6_plen will be set later */
400 	ip6->ip6_nxt = IPPROTO_ICMPV6;
401 	ip6->ip6_hlim = 255;
402 	if (daddr6)
403 		ip6->ip6_dst = *daddr6;
404 	else {
405 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
406 		ip6->ip6_dst.s6_addr16[1] = 0;
407 		ip6->ip6_dst.s6_addr32[1] = 0;
408 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
409 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
410 		ip6->ip6_dst.s6_addr8[12] = 0xff;
411 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
412 			goto bad;
413 	}
414 	if (!dad) {
415 		/*
416 		 * RFC2461 7.2.2:
417 		 * "If the source address of the packet prompting the
418 		 * solicitation is the same as one of the addresses assigned
419 		 * to the outgoing interface, that address SHOULD be placed
420 		 * in the IP Source Address of the outgoing solicitation.
421 		 * Otherwise, any one of the addresses assigned to the
422 		 * interface should be used."
423 		 *
424 		 * We use the source address for the prompting packet
425 		 * (saddr6), if:
426 		 * - saddr6 is given from the caller (by giving "ln"), and
427 		 * - saddr6 belongs to the outgoing interface.
428 		 * Otherwise, we perform the source address selection as usual.
429 		 */
430 		struct ip6_hdr *hip6;		/* hold ip6 */
431 		struct in6_addr *hsrc = NULL;
432 
433 		if (ln && ln->ln_hold) {
434 			/*
435 			 * assuming every packet in ln_hold has the same IP
436 			 * header
437 			 */
438 			hip6 = mtod(ln->ln_hold, struct ip6_hdr *);
439 			/* XXX pullup? */
440 			if (sizeof(*hip6) < ln->ln_hold->m_len)
441 				hsrc = &hip6->ip6_src;
442 			else
443 				hsrc = NULL;
444 		}
445 		if (hsrc && in6ifa_ifpwithaddr(ifp, hsrc))
446 			src = hsrc;
447 		else {
448 			int error;
449 			struct sockaddr_in6 dst_sa;
450 
451 			bzero(&dst_sa, sizeof(dst_sa));
452 			dst_sa.sin6_family = AF_INET6;
453 			dst_sa.sin6_len = sizeof(dst_sa);
454 			dst_sa.sin6_addr = ip6->ip6_dst;
455 
456 			src = in6_selectsrc(&dst_sa, NULL,
457 			    NULL, &ro, NULL, NULL, &error);
458 			if (src == NULL) {
459 				char ip6buf[INET6_ADDRSTRLEN];
460 				nd6log((LOG_DEBUG,
461 				    "nd6_ns_output: source can't be "
462 				    "determined: dst=%s, error=%d\n",
463 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
464 				    error));
465 				goto bad;
466 			}
467 		}
468 	} else {
469 		/*
470 		 * Source address for DAD packet must always be IPv6
471 		 * unspecified address. (0::0)
472 		 * We actually don't have to 0-clear the address (we did it
473 		 * above), but we do so here explicitly to make the intention
474 		 * clearer.
475 		 */
476 		bzero(&src_in, sizeof(src_in));
477 		src = &src_in;
478 	}
479 	ip6->ip6_src = *src;
480 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
481 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
482 	nd_ns->nd_ns_code = 0;
483 	nd_ns->nd_ns_reserved = 0;
484 	nd_ns->nd_ns_target = *taddr6;
485 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
486 
487 	/*
488 	 * Add source link-layer address option.
489 	 *
490 	 *				spec		implementation
491 	 *				---		---
492 	 * DAD packet			MUST NOT	do not add the option
493 	 * there's no link layer address:
494 	 *				impossible	do not add the option
495 	 * there's link layer address:
496 	 *	Multicast NS		MUST add one	add the option
497 	 *	Unicast NS		SHOULD add one	add the option
498 	 */
499 	if (!dad && (mac = nd6_ifptomac(ifp))) {
500 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
501 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
502 		/* 8 byte alignments... */
503 		optlen = (optlen + 7) & ~7;
504 
505 		m->m_pkthdr.len += optlen;
506 		m->m_len += optlen;
507 		icmp6len += optlen;
508 		bzero((caddr_t)nd_opt, optlen);
509 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
510 		nd_opt->nd_opt_len = optlen >> 3;
511 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
512 	}
513 
514 	ip6->ip6_plen = htons((u_short)icmp6len);
515 	nd_ns->nd_ns_cksum = 0;
516 	nd_ns->nd_ns_cksum =
517 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
518 
519 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
520 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
521 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
522 	icmp6stat.icp6s_outhist[ND_NEIGHBOR_SOLICIT]++;
523 
524 	if (ro.ro_rt) {		/* we don't cache this route. */
525 		RTFREE(ro.ro_rt);
526 	}
527 	return;
528 
529   bad:
530 	if (ro.ro_rt) {
531 		RTFREE(ro.ro_rt);
532 	}
533 	m_freem(m);
534 	return;
535 }
536 
537 /*
538  * Neighbor advertisement input handling.
539  *
540  * Based on RFC 2461
541  * Based on RFC 2462 (duplicate address detection)
542  *
543  * the following items are not implemented yet:
544  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
545  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
546  */
547 void
548 nd6_na_input(struct mbuf *m, int off, int icmp6len)
549 {
550 	struct ifnet *ifp = m->m_pkthdr.rcvif;
551 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
552 	struct nd_neighbor_advert *nd_na;
553 	struct in6_addr daddr6 = ip6->ip6_dst;
554 	struct in6_addr taddr6;
555 	int flags;
556 	int is_router;
557 	int is_solicited;
558 	int is_override;
559 	char *lladdr = NULL;
560 	int lladdrlen = 0;
561 	struct ifaddr *ifa;
562 	struct llinfo_nd6 *ln;
563 	struct rtentry *rt;
564 	struct sockaddr_dl *sdl;
565 	union nd_opts ndopts;
566 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
567 
568 	if (ip6->ip6_hlim != 255) {
569 		nd6log((LOG_ERR,
570 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
571 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
572 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
573 		goto bad;
574 	}
575 
576 #ifndef PULLDOWN_TEST
577 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
578 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
579 #else
580 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
581 	if (nd_na == NULL) {
582 		icmp6stat.icp6s_tooshort++;
583 		return;
584 	}
585 #endif
586 
587 	flags = nd_na->nd_na_flags_reserved;
588 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
589 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
590 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
591 
592 	taddr6 = nd_na->nd_na_target;
593 	if (in6_setscope(&taddr6, ifp, NULL))
594 		goto bad;	/* XXX: impossible */
595 
596 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
597 		nd6log((LOG_ERR,
598 		    "nd6_na_input: invalid target address %s\n",
599 		    ip6_sprintf(ip6bufs, &taddr6)));
600 		goto bad;
601 	}
602 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
603 		if (is_solicited) {
604 			nd6log((LOG_ERR,
605 			    "nd6_na_input: a solicited adv is multicasted\n"));
606 			goto bad;
607 		}
608 
609 	icmp6len -= sizeof(*nd_na);
610 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
611 	if (nd6_options(&ndopts) < 0) {
612 		nd6log((LOG_INFO,
613 		    "nd6_na_input: invalid ND option, ignored\n"));
614 		/* nd6_options have incremented stats */
615 		goto freeit;
616 	}
617 
618 	if (ndopts.nd_opts_tgt_lladdr) {
619 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
620 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
621 	}
622 
623 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
624 
625 	/*
626 	 * Target address matches one of my interface address.
627 	 *
628 	 * If my address is tentative, this means that there's somebody
629 	 * already using the same address as mine.  This indicates DAD failure.
630 	 * This is defined in RFC 2462.
631 	 *
632 	 * Otherwise, process as defined in RFC 2461.
633 	 */
634 	if (ifa
635 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
636 		nd6_dad_na_input(ifa);
637 		goto freeit;
638 	}
639 
640 	/* Just for safety, maybe unnecessary. */
641 	if (ifa) {
642 		log(LOG_ERR,
643 		    "nd6_na_input: duplicate IP6 address %s\n",
644 		    ip6_sprintf(ip6bufs, &taddr6));
645 		goto freeit;
646 	}
647 
648 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
649 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
650 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
651 		    ifp->if_addrlen, lladdrlen - 2));
652 		goto bad;
653 	}
654 
655 	/*
656 	 * If no neighbor cache entry is found, NA SHOULD silently be
657 	 * discarded.
658 	 */
659 	rt = nd6_lookup(&taddr6, 0, ifp);
660 	if ((rt == NULL) ||
661 	   ((ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) ||
662 	   ((sdl = SDL(rt->rt_gateway)) == NULL))
663 		goto freeit;
664 
665 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
666 		/*
667 		 * If the link-layer has address, and no lladdr option came,
668 		 * discard the packet.
669 		 */
670 		if (ifp->if_addrlen && lladdr == NULL)
671 			goto freeit;
672 
673 		/*
674 		 * Record link-layer address, and update the state.
675 		 */
676 		sdl->sdl_alen = ifp->if_addrlen;
677 		bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
678 		if (is_solicited) {
679 			ln->ln_state = ND6_LLINFO_REACHABLE;
680 			ln->ln_byhint = 0;
681 			if (!ND6_LLINFO_PERMANENT(ln)) {
682 				nd6_llinfo_settimer(ln,
683 				    (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
684 			}
685 		} else {
686 			ln->ln_state = ND6_LLINFO_STALE;
687 			nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
688 		}
689 		if ((ln->ln_router = is_router) != 0) {
690 			/*
691 			 * This means a router's state has changed from
692 			 * non-reachable to probably reachable, and might
693 			 * affect the status of associated prefixes..
694 			 */
695 			pfxlist_onlink_check();
696 		}
697 	} else {
698 		int llchange;
699 
700 		/*
701 		 * Check if the link-layer address has changed or not.
702 		 */
703 		if (lladdr == NULL)
704 			llchange = 0;
705 		else {
706 			if (sdl->sdl_alen) {
707 				if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
708 					llchange = 1;
709 				else
710 					llchange = 0;
711 			} else
712 				llchange = 1;
713 		}
714 
715 		/*
716 		 * This is VERY complex.  Look at it with care.
717 		 *
718 		 * override solicit lladdr llchange	action
719 		 *					(L: record lladdr)
720 		 *
721 		 *	0	0	n	--	(2c)
722 		 *	0	0	y	n	(2b) L
723 		 *	0	0	y	y	(1)    REACHABLE->STALE
724 		 *	0	1	n	--	(2c)   *->REACHABLE
725 		 *	0	1	y	n	(2b) L *->REACHABLE
726 		 *	0	1	y	y	(1)    REACHABLE->STALE
727 		 *	1	0	n	--	(2a)
728 		 *	1	0	y	n	(2a) L
729 		 *	1	0	y	y	(2a) L *->STALE
730 		 *	1	1	n	--	(2a)   *->REACHABLE
731 		 *	1	1	y	n	(2a) L *->REACHABLE
732 		 *	1	1	y	y	(2a) L *->REACHABLE
733 		 */
734 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
735 			/*
736 			 * If state is REACHABLE, make it STALE.
737 			 * no other updates should be done.
738 			 */
739 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
740 				ln->ln_state = ND6_LLINFO_STALE;
741 				nd6_llinfo_settimer(ln, (long)nd6_gctimer * hz);
742 			}
743 			goto freeit;
744 		} else if (is_override				   /* (2a) */
745 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
746 			|| lladdr == NULL) {			   /* (2c) */
747 			/*
748 			 * Update link-local address, if any.
749 			 */
750 			if (lladdr != NULL) {
751 				sdl->sdl_alen = ifp->if_addrlen;
752 				bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
753 			}
754 
755 			/*
756 			 * If solicited, make the state REACHABLE.
757 			 * If not solicited and the link-layer address was
758 			 * changed, make it STALE.
759 			 */
760 			if (is_solicited) {
761 				ln->ln_state = ND6_LLINFO_REACHABLE;
762 				ln->ln_byhint = 0;
763 				if (!ND6_LLINFO_PERMANENT(ln)) {
764 					nd6_llinfo_settimer(ln,
765 					    (long)ND_IFINFO(ifp)->reachable * hz);
766 				}
767 			} else {
768 				if (lladdr != NULL && llchange) {
769 					ln->ln_state = ND6_LLINFO_STALE;
770 					nd6_llinfo_settimer(ln,
771 					    (long)nd6_gctimer * hz);
772 				}
773 			}
774 		}
775 
776 		if (ln->ln_router && !is_router) {
777 			/*
778 			 * The peer dropped the router flag.
779 			 * Remove the sender from the Default Router List and
780 			 * update the Destination Cache entries.
781 			 */
782 			struct nd_defrouter *dr;
783 			struct in6_addr *in6;
784 			int s;
785 
786 			in6 = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
787 
788 			/*
789 			 * Lock to protect the default router list.
790 			 * XXX: this might be unnecessary, since this function
791 			 * is only called under the network software interrupt
792 			 * context.  However, we keep it just for safety.
793 			 */
794 			s = splnet();
795 			dr = defrouter_lookup(in6, ifp);
796 			if (dr)
797 				defrtrlist_del(dr);
798 			else if (!ip6_forwarding) {
799 				/*
800 				 * Even if the neighbor is not in the default
801 				 * router list, the neighbor may be used
802 				 * as a next hop for some destinations
803 				 * (e.g. redirect case). So we must
804 				 * call rt6_flush explicitly.
805 				 */
806 				rt6_flush(&ip6->ip6_src, ifp);
807 			}
808 			splx(s);
809 		}
810 		ln->ln_router = is_router;
811 	}
812 	rt->rt_flags &= ~RTF_REJECT;
813 	ln->ln_asked = 0;
814 	if (ln->ln_hold) {
815 		struct mbuf *m_hold, *m_hold_next;
816 
817 		/*
818 		 * reset the ln_hold in advance, to explicitly
819 		 * prevent a ln_hold lookup in nd6_output()
820 		 * (wouldn't happen, though...)
821 		 */
822 		for (m_hold = ln->ln_hold;
823 		    m_hold; m_hold = m_hold_next) {
824 			m_hold_next = m_hold->m_nextpkt;
825 			m_hold->m_nextpkt = NULL;
826 			/*
827 			 * we assume ifp is not a loopback here, so just set
828 			 * the 2nd argument as the 1st one.
829 			 */
830 			nd6_output(ifp, ifp, m_hold,
831 			    (struct sockaddr_in6 *)rt_key(rt), rt);
832 		}
833 		ln->ln_hold = NULL;
834 	}
835 
836  freeit:
837 	m_freem(m);
838 	return;
839 
840  bad:
841 	icmp6stat.icp6s_badna++;
842 	m_freem(m);
843 }
844 
845 /*
846  * Neighbor advertisement output handling.
847  *
848  * Based on RFC 2461
849  *
850  * the following items are not implemented yet:
851  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
852  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
853  *
854  * tlladdr - 1 if include target link-layer address
855  * sdl0 - sockaddr_dl (= proxy NA) or NULL
856  */
857 void
858 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
859     const struct in6_addr *taddr6, u_long flags, int tlladdr,
860     struct sockaddr *sdl0)
861 {
862 	struct mbuf *m;
863 	struct ip6_hdr *ip6;
864 	struct nd_neighbor_advert *nd_na;
865 	struct ip6_moptions im6o;
866 	struct in6_addr *src, daddr6;
867 	struct sockaddr_in6 dst_sa;
868 	int icmp6len, maxlen, error;
869 	caddr_t mac = NULL;
870 	struct route_in6 ro;
871 
872 	bzero(&ro, sizeof(ro));
873 
874 	daddr6 = *daddr6_0;	/* make a local copy for modification */
875 
876 	/* estimate the size of message */
877 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
878 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
879 	if (max_linkhdr + maxlen >= MCLBYTES) {
880 #ifdef DIAGNOSTIC
881 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
882 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
883 #endif
884 		return;
885 	}
886 
887 	MGETHDR(m, M_DONTWAIT, MT_DATA);
888 	if (m && max_linkhdr + maxlen >= MHLEN) {
889 		MCLGET(m, M_DONTWAIT);
890 		if ((m->m_flags & M_EXT) == 0) {
891 			m_free(m);
892 			m = NULL;
893 		}
894 	}
895 	if (m == NULL)
896 		return;
897 	m->m_pkthdr.rcvif = NULL;
898 
899 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
900 		m->m_flags |= M_MCAST;
901 		im6o.im6o_multicast_ifp = ifp;
902 		im6o.im6o_multicast_hlim = 255;
903 		im6o.im6o_multicast_loop = 0;
904 	}
905 
906 	icmp6len = sizeof(*nd_na);
907 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
908 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
909 
910 	/* fill neighbor advertisement packet */
911 	ip6 = mtod(m, struct ip6_hdr *);
912 	ip6->ip6_flow = 0;
913 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
914 	ip6->ip6_vfc |= IPV6_VERSION;
915 	ip6->ip6_nxt = IPPROTO_ICMPV6;
916 	ip6->ip6_hlim = 255;
917 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
918 		/* reply to DAD */
919 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
920 		daddr6.s6_addr16[1] = 0;
921 		daddr6.s6_addr32[1] = 0;
922 		daddr6.s6_addr32[2] = 0;
923 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
924 		if (in6_setscope(&daddr6, ifp, NULL))
925 			goto bad;
926 
927 		flags &= ~ND_NA_FLAG_SOLICITED;
928 	}
929 	ip6->ip6_dst = daddr6;
930 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
931 	dst_sa.sin6_family = AF_INET6;
932 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
933 	dst_sa.sin6_addr = daddr6;
934 
935 	/*
936 	 * Select a source whose scope is the same as that of the dest.
937 	 */
938 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
939 	src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &error);
940 	if (src == NULL) {
941 		char ip6buf[INET6_ADDRSTRLEN];
942 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
943 		    "determined: dst=%s, error=%d\n",
944 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
945 		goto bad;
946 	}
947 	ip6->ip6_src = *src;
948 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
949 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
950 	nd_na->nd_na_code = 0;
951 	nd_na->nd_na_target = *taddr6;
952 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
953 
954 	/*
955 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
956 	 * see nd6_ns_input() for details.
957 	 * Basically, if NS packet is sent to unicast/anycast addr,
958 	 * target lladdr option SHOULD NOT be included.
959 	 */
960 	if (tlladdr) {
961 		/*
962 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
963 		 * lladdr in sdl0.  If we are not proxying (sending NA for
964 		 * my address) use lladdr configured for the interface.
965 		 */
966 		if (sdl0 == NULL) {
967 #ifdef DEV_CARP
968 			if (ifp->if_carp)
969 				mac = carp_macmatch6(ifp->if_carp, m, taddr6);
970 			if (mac == NULL)
971 				mac = nd6_ifptomac(ifp);
972 #else
973 			mac = nd6_ifptomac(ifp);
974 #endif
975 		} else if (sdl0->sa_family == AF_LINK) {
976 			struct sockaddr_dl *sdl;
977 			sdl = (struct sockaddr_dl *)sdl0;
978 			if (sdl->sdl_alen == ifp->if_addrlen)
979 				mac = LLADDR(sdl);
980 		}
981 	}
982 	if (tlladdr && mac) {
983 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
984 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
985 
986 		/* roundup to 8 bytes alignment! */
987 		optlen = (optlen + 7) & ~7;
988 
989 		m->m_pkthdr.len += optlen;
990 		m->m_len += optlen;
991 		icmp6len += optlen;
992 		bzero((caddr_t)nd_opt, optlen);
993 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
994 		nd_opt->nd_opt_len = optlen >> 3;
995 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
996 	} else
997 		flags &= ~ND_NA_FLAG_OVERRIDE;
998 
999 	ip6->ip6_plen = htons((u_short)icmp6len);
1000 	nd_na->nd_na_flags_reserved = flags;
1001 	nd_na->nd_na_cksum = 0;
1002 	nd_na->nd_na_cksum =
1003 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1004 
1005 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1006 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1007 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1008 	icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1009 
1010 	if (ro.ro_rt) {		/* we don't cache this route. */
1011 		RTFREE(ro.ro_rt);
1012 	}
1013 	return;
1014 
1015   bad:
1016 	if (ro.ro_rt) {
1017 		RTFREE(ro.ro_rt);
1018 	}
1019 	m_freem(m);
1020 	return;
1021 }
1022 
1023 caddr_t
1024 nd6_ifptomac(struct ifnet *ifp)
1025 {
1026 	switch (ifp->if_type) {
1027 	case IFT_ARCNET:
1028 	case IFT_ETHER:
1029 	case IFT_FDDI:
1030 	case IFT_IEEE1394:
1031 #ifdef IFT_L2VLAN
1032 	case IFT_L2VLAN:
1033 #endif
1034 #ifdef IFT_IEEE80211
1035 	case IFT_IEEE80211:
1036 #endif
1037 #ifdef IFT_CARP
1038 	case IFT_CARP:
1039 #endif
1040 	case IFT_BRIDGE:
1041 	case IFT_ISO88025:
1042 		return IF_LLADDR(ifp);
1043 	default:
1044 		return NULL;
1045 	}
1046 }
1047 
1048 TAILQ_HEAD(dadq_head, dadq);
1049 struct dadq {
1050 	TAILQ_ENTRY(dadq) dad_list;
1051 	struct ifaddr *dad_ifa;
1052 	int dad_count;		/* max NS to send */
1053 	int dad_ns_tcount;	/* # of trials to send NS */
1054 	int dad_ns_ocount;	/* NS sent so far */
1055 	int dad_ns_icount;
1056 	int dad_na_icount;
1057 	struct callout dad_timer_ch;
1058 };
1059 
1060 static struct dadq_head dadq;
1061 static int dad_init = 0;
1062 
1063 static struct dadq *
1064 nd6_dad_find(struct ifaddr *ifa)
1065 {
1066 	struct dadq *dp;
1067 
1068 	for (dp = dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1069 		if (dp->dad_ifa == ifa)
1070 			return dp;
1071 	}
1072 	return NULL;
1073 }
1074 
1075 static void
1076 nd6_dad_starttimer(struct dadq *dp, int ticks)
1077 {
1078 
1079 	callout_reset(&dp->dad_timer_ch, ticks,
1080 	    (void (*) __P((void *)))nd6_dad_timer, (void *)dp->dad_ifa);
1081 }
1082 
1083 static void
1084 nd6_dad_stoptimer(struct dadq *dp)
1085 {
1086 
1087 	callout_stop(&dp->dad_timer_ch);
1088 }
1089 
1090 /*
1091  * Start Duplicate Address Detection (DAD) for specified interface address.
1092  */
1093 void
1094 nd6_dad_start(struct ifaddr *ifa, int delay)
1095 {
1096 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1097 	struct dadq *dp;
1098 	char ip6buf[INET6_ADDRSTRLEN];
1099 
1100 	if (!dad_init) {
1101 		TAILQ_INIT(&dadq);
1102 		dad_init++;
1103 	}
1104 
1105 	/*
1106 	 * If we don't need DAD, don't do it.
1107 	 * There are several cases:
1108 	 * - DAD is disabled (ip6_dad_count == 0)
1109 	 * - the interface address is anycast
1110 	 */
1111 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1112 		log(LOG_DEBUG,
1113 			"nd6_dad_start: called with non-tentative address "
1114 			"%s(%s)\n",
1115 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1116 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1117 		return;
1118 	}
1119 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1120 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1121 		return;
1122 	}
1123 	if (!ip6_dad_count) {
1124 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1125 		return;
1126 	}
1127 	if (ifa->ifa_ifp == NULL)
1128 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1129 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1130 		return;
1131 	}
1132 	if (nd6_dad_find(ifa) != NULL) {
1133 		/* DAD already in progress */
1134 		return;
1135 	}
1136 
1137 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1138 	if (dp == NULL) {
1139 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1140 			"%s(%s)\n",
1141 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1142 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1143 		return;
1144 	}
1145 	bzero(dp, sizeof(*dp));
1146 	callout_init(&dp->dad_timer_ch, 0);
1147 	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1148 
1149 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1150 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1151 
1152 	/*
1153 	 * Send NS packet for DAD, ip6_dad_count times.
1154 	 * Note that we must delay the first transmission, if this is the
1155 	 * first packet to be sent from the interface after interface
1156 	 * (re)initialization.
1157 	 */
1158 	dp->dad_ifa = ifa;
1159 	IFAREF(ifa);	/* just for safety */
1160 	dp->dad_count = ip6_dad_count;
1161 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1162 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1163 	if (delay == 0) {
1164 		nd6_dad_ns_output(dp, ifa);
1165 		nd6_dad_starttimer(dp,
1166 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1167 	} else {
1168 		nd6_dad_starttimer(dp, delay);
1169 	}
1170 }
1171 
1172 /*
1173  * terminate DAD unconditionally.  used for address removals.
1174  */
1175 void
1176 nd6_dad_stop(struct ifaddr *ifa)
1177 {
1178 	struct dadq *dp;
1179 
1180 	if (!dad_init)
1181 		return;
1182 	dp = nd6_dad_find(ifa);
1183 	if (!dp) {
1184 		/* DAD wasn't started yet */
1185 		return;
1186 	}
1187 
1188 	nd6_dad_stoptimer(dp);
1189 
1190 	TAILQ_REMOVE(&dadq, (struct dadq *)dp, dad_list);
1191 	free(dp, M_IP6NDP);
1192 	dp = NULL;
1193 	IFAFREE(ifa);
1194 }
1195 
1196 static void
1197 nd6_dad_timer(struct ifaddr *ifa)
1198 {
1199 	int s;
1200 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1201 	struct dadq *dp;
1202 	char ip6buf[INET6_ADDRSTRLEN];
1203 
1204 	s = splnet();		/* XXX */
1205 
1206 	/* Sanity check */
1207 	if (ia == NULL) {
1208 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1209 		goto done;
1210 	}
1211 	dp = nd6_dad_find(ifa);
1212 	if (dp == NULL) {
1213 		log(LOG_ERR, "nd6_dad_timer: DAD structure not found\n");
1214 		goto done;
1215 	}
1216 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1217 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1218 			"%s(%s)\n",
1219 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1220 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1221 		goto done;
1222 	}
1223 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1224 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1225 			"%s(%s)\n",
1226 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1227 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1228 		goto done;
1229 	}
1230 
1231 	/* timeouted with IFF_{RUNNING,UP} check */
1232 	if (dp->dad_ns_tcount > dad_maxtry) {
1233 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1234 		    if_name(ifa->ifa_ifp)));
1235 
1236 		TAILQ_REMOVE(&dadq, (struct dadq *)dp, dad_list);
1237 		free(dp, M_IP6NDP);
1238 		dp = NULL;
1239 		IFAFREE(ifa);
1240 		goto done;
1241 	}
1242 
1243 	/* Need more checks? */
1244 	if (dp->dad_ns_ocount < dp->dad_count) {
1245 		/*
1246 		 * We have more NS to go.  Send NS packet for DAD.
1247 		 */
1248 		nd6_dad_ns_output(dp, ifa);
1249 		nd6_dad_starttimer(dp,
1250 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1251 	} else {
1252 		/*
1253 		 * We have transmitted sufficient number of DAD packets.
1254 		 * See what we've got.
1255 		 */
1256 		int duplicate;
1257 
1258 		duplicate = 0;
1259 
1260 		if (dp->dad_na_icount) {
1261 			/*
1262 			 * the check is in nd6_dad_na_input(),
1263 			 * but just in case
1264 			 */
1265 			duplicate++;
1266 		}
1267 
1268 		if (dp->dad_ns_icount) {
1269 			/* We've seen NS, means DAD has failed. */
1270 			duplicate++;
1271 		}
1272 
1273 		if (duplicate) {
1274 			/* (*dp) will be freed in nd6_dad_duplicated() */
1275 			dp = NULL;
1276 			nd6_dad_duplicated(ifa);
1277 		} else {
1278 			/*
1279 			 * We are done with DAD.  No NA came, no NS came.
1280 			 * No duplicate address found.
1281 			 */
1282 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1283 
1284 			nd6log((LOG_DEBUG,
1285 			    "%s: DAD complete for %s - no duplicates found\n",
1286 			    if_name(ifa->ifa_ifp),
1287 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1288 
1289 			TAILQ_REMOVE(&dadq, (struct dadq *)dp, dad_list);
1290 			free(dp, M_IP6NDP);
1291 			dp = NULL;
1292 			IFAFREE(ifa);
1293 		}
1294 	}
1295 
1296 done:
1297 	splx(s);
1298 }
1299 
1300 void
1301 nd6_dad_duplicated(struct ifaddr *ifa)
1302 {
1303 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1304 	struct ifnet *ifp;
1305 	struct dadq *dp;
1306 	char ip6buf[INET6_ADDRSTRLEN];
1307 
1308 	dp = nd6_dad_find(ifa);
1309 	if (dp == NULL) {
1310 		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1311 		return;
1312 	}
1313 
1314 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1315 	    "NS in/out=%d/%d, NA in=%d\n",
1316 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1317 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1318 
1319 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1320 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1321 
1322 	/* We are done with DAD, with duplicate address found. (failure) */
1323 	nd6_dad_stoptimer(dp);
1324 
1325 	ifp = ifa->ifa_ifp;
1326 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1327 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1328 	log(LOG_ERR, "%s: manual intervention required\n",
1329 	    if_name(ifp));
1330 
1331 	/*
1332 	 * If the address is a link-local address formed from an interface
1333 	 * identifier based on the hardware address which is supposed to be
1334 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1335 	 * operation on the interface SHOULD be disabled.
1336 	 * [rfc2462bis-03 Section 5.4.5]
1337 	 */
1338 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1339 		struct in6_addr in6;
1340 
1341 		/*
1342 		 * To avoid over-reaction, we only apply this logic when we are
1343 		 * very sure that hardware addresses are supposed to be unique.
1344 		 */
1345 		switch (ifp->if_type) {
1346 		case IFT_ETHER:
1347 		case IFT_FDDI:
1348 		case IFT_ATM:
1349 		case IFT_IEEE1394:
1350 #ifdef IFT_IEEE80211
1351 		case IFT_IEEE80211:
1352 #endif
1353 			in6 = ia->ia_addr.sin6_addr;
1354 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1355 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1356 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1357 				log(LOG_ERR, "%s: possible hardware address "
1358 				    "duplication detected, disable IPv6\n",
1359 				    if_name(ifp));
1360 			}
1361 			break;
1362 		}
1363 	}
1364 
1365 	TAILQ_REMOVE(&dadq, (struct dadq *)dp, dad_list);
1366 	free(dp, M_IP6NDP);
1367 	dp = NULL;
1368 	IFAFREE(ifa);
1369 }
1370 
1371 static void
1372 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1373 {
1374 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1375 	struct ifnet *ifp = ifa->ifa_ifp;
1376 
1377 	dp->dad_ns_tcount++;
1378 	if ((ifp->if_flags & IFF_UP) == 0) {
1379 		return;
1380 	}
1381 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1382 		return;
1383 	}
1384 
1385 	dp->dad_ns_ocount++;
1386 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1387 }
1388 
1389 static void
1390 nd6_dad_ns_input(struct ifaddr *ifa)
1391 {
1392 	struct in6_ifaddr *ia;
1393 	struct ifnet *ifp;
1394 	const struct in6_addr *taddr6;
1395 	struct dadq *dp;
1396 	int duplicate;
1397 
1398 	if (ifa == NULL)
1399 		panic("ifa == NULL in nd6_dad_ns_input");
1400 
1401 	ia = (struct in6_ifaddr *)ifa;
1402 	ifp = ifa->ifa_ifp;
1403 	taddr6 = &ia->ia_addr.sin6_addr;
1404 	duplicate = 0;
1405 	dp = nd6_dad_find(ifa);
1406 
1407 	/* Quickhack - completely ignore DAD NS packets */
1408 	if (dad_ignore_ns) {
1409 		char ip6buf[INET6_ADDRSTRLEN];
1410 		nd6log((LOG_INFO,
1411 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1412 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1413 		    if_name(ifa->ifa_ifp)));
1414 		return;
1415 	}
1416 
1417 	/*
1418 	 * if I'm yet to start DAD, someone else started using this address
1419 	 * first.  I have a duplicate and you win.
1420 	 */
1421 	if (dp == NULL || dp->dad_ns_ocount == 0)
1422 		duplicate++;
1423 
1424 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1425 
1426 	if (duplicate) {
1427 		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1428 		nd6_dad_duplicated(ifa);
1429 	} else {
1430 		/*
1431 		 * not sure if I got a duplicate.
1432 		 * increment ns count and see what happens.
1433 		 */
1434 		if (dp)
1435 			dp->dad_ns_icount++;
1436 	}
1437 }
1438 
1439 static void
1440 nd6_dad_na_input(struct ifaddr *ifa)
1441 {
1442 	struct dadq *dp;
1443 
1444 	if (ifa == NULL)
1445 		panic("ifa == NULL in nd6_dad_na_input");
1446 
1447 	dp = nd6_dad_find(ifa);
1448 	if (dp)
1449 		dp->dad_na_icount++;
1450 
1451 	/* remove the address. */
1452 	nd6_dad_duplicated(ifa);
1453 }
1454