xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 33644623554bb0fc57ed3c7d874193a498679b22)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 #include "opt_carp.h"
39 #include "opt_mpath.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/socket.h>
46 #include <sys/sockio.h>
47 #include <sys/time.h>
48 #include <sys/kernel.h>
49 #include <sys/errno.h>
50 #include <sys/syslog.h>
51 #include <sys/queue.h>
52 #include <sys/callout.h>
53 #include <sys/vimage.h>
54 
55 #include <net/if.h>
56 #include <net/if_types.h>
57 #include <net/if_dl.h>
58 #include <net/if_var.h>
59 #include <net/route.h>
60 #ifdef RADIX_MPATH
61 #include <net/radix_mpath.h>
62 #endif
63 
64 #include <netinet/in.h>
65 #include <netinet/in_var.h>
66 #include <netinet6/in6_var.h>
67 #include <netinet6/in6_ifattach.h>
68 #include <netinet/ip6.h>
69 #include <netinet6/ip6_var.h>
70 #include <netinet6/scope6_var.h>
71 #include <netinet6/nd6.h>
72 #include <netinet/icmp6.h>
73 
74 #ifdef DEV_CARP
75 #include <netinet/ip_carp.h>
76 #endif
77 
78 #define SDL(s) ((struct sockaddr_dl *)s)
79 
80 struct dadq;
81 static struct dadq *nd6_dad_find(struct ifaddr *);
82 static void nd6_dad_starttimer(struct dadq *, int);
83 static void nd6_dad_stoptimer(struct dadq *);
84 static void nd6_dad_timer(struct ifaddr *);
85 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
86 static void nd6_dad_ns_input(struct ifaddr *);
87 static void nd6_dad_na_input(struct ifaddr *);
88 
89 #ifdef VIMAGE_GLOBALS
90 int dad_ignore_ns;
91 int dad_maxtry;
92 #endif
93 
94 /*
95  * Input a Neighbor Solicitation Message.
96  *
97  * Based on RFC 2461
98  * Based on RFC 2462 (duplicate address detection)
99  */
100 void
101 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
102 {
103 	INIT_VNET_INET6(curvnet);
104 	struct ifnet *ifp = m->m_pkthdr.rcvif;
105 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
106 	struct nd_neighbor_solicit *nd_ns;
107 	struct in6_addr saddr6 = ip6->ip6_src;
108 	struct in6_addr daddr6 = ip6->ip6_dst;
109 	struct in6_addr taddr6;
110 	struct in6_addr myaddr6;
111 	char *lladdr = NULL;
112 	struct ifaddr *ifa = NULL;
113 	int lladdrlen = 0;
114 	int anycast = 0, proxy = 0, tentative = 0;
115 	int tlladdr;
116 	union nd_opts ndopts;
117 	struct sockaddr_dl *proxydl = NULL;
118 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
119 
120 #ifndef PULLDOWN_TEST
121 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
122 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
123 #else
124 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
125 	if (nd_ns == NULL) {
126 		V_icmp6stat.icp6s_tooshort++;
127 		return;
128 	}
129 #endif
130 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
131 	taddr6 = nd_ns->nd_ns_target;
132 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
133 		goto bad;
134 
135 	if (ip6->ip6_hlim != 255) {
136 		nd6log((LOG_ERR,
137 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
138 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
139 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
140 		goto bad;
141 	}
142 
143 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
144 		/* dst has to be a solicited node multicast address. */
145 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
146 		    /* don't check ifindex portion */
147 		    daddr6.s6_addr32[1] == 0 &&
148 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
149 		    daddr6.s6_addr8[12] == 0xff) {
150 			; /* good */
151 		} else {
152 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
153 			    "(wrong ip6 dst)\n"));
154 			goto bad;
155 		}
156 	} else if (!V_nd6_onlink_ns_rfc4861) {
157 		struct sockaddr_in6 src_sa6;
158 
159 		/*
160 		 * According to recent IETF discussions, it is not a good idea
161 		 * to accept a NS from an address which would not be deemed
162 		 * to be a neighbor otherwise.  This point is expected to be
163 		 * clarified in future revisions of the specification.
164 		 */
165 		bzero(&src_sa6, sizeof(src_sa6));
166 		src_sa6.sin6_family = AF_INET6;
167 		src_sa6.sin6_len = sizeof(src_sa6);
168 		src_sa6.sin6_addr = saddr6;
169 		if (!nd6_is_addr_neighbor(&src_sa6, ifp)) {
170 			nd6log((LOG_INFO, "nd6_ns_input: "
171 				"NS packet from non-neighbor\n"));
172 			goto bad;
173 		}
174 	}
175 
176 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
177 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
178 		goto bad;
179 	}
180 
181 	icmp6len -= sizeof(*nd_ns);
182 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
183 	if (nd6_options(&ndopts) < 0) {
184 		nd6log((LOG_INFO,
185 		    "nd6_ns_input: invalid ND option, ignored\n"));
186 		/* nd6_options have incremented stats */
187 		goto freeit;
188 	}
189 
190 	if (ndopts.nd_opts_src_lladdr) {
191 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
192 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
193 	}
194 
195 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
196 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
197 		    "(link-layer address option)\n"));
198 		goto bad;
199 	}
200 
201 	/*
202 	 * Attaching target link-layer address to the NA?
203 	 * (RFC 2461 7.2.4)
204 	 *
205 	 * NS IP dst is unicast/anycast			MUST NOT add
206 	 * NS IP dst is solicited-node multicast	MUST add
207 	 *
208 	 * In implementation, we add target link-layer address by default.
209 	 * We do not add one in MUST NOT cases.
210 	 */
211 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
212 		tlladdr = 0;
213 	else
214 		tlladdr = 1;
215 
216 	/*
217 	 * Target address (taddr6) must be either:
218 	 * (1) Valid unicast/anycast address for my receiving interface,
219 	 * (2) Unicast address for which I'm offering proxy service, or
220 	 * (3) "tentative" address on which DAD is being performed.
221 	 */
222 	/* (1) and (3) check. */
223 #ifdef DEV_CARP
224 	if (ifp->if_carp)
225 		ifa = carp_iamatch6(ifp->if_carp, &taddr6);
226 	if (ifa == NULL)
227 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
228 #else
229 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
230 #endif
231 
232 	/* (2) check. */
233 	if (ifa == NULL) {
234 		struct rtentry *rt;
235 		struct sockaddr_in6 tsin6;
236 		int need_proxy;
237 #ifdef RADIX_MPATH
238 		struct route_in6 ro;
239 #endif
240 
241 		bzero(&tsin6, sizeof tsin6);
242 		tsin6.sin6_len = sizeof(struct sockaddr_in6);
243 		tsin6.sin6_family = AF_INET6;
244 		tsin6.sin6_addr = taddr6;
245 
246 #ifdef RADIX_MPATH
247 		bzero(&ro, sizeof(ro));
248 		ro.ro_dst = tsin6;
249 		rtalloc_mpath((struct route *)&ro, RTF_ANNOUNCE);
250 		rt = ro.ro_rt;
251 #else
252 		rt = rtalloc1((struct sockaddr *)&tsin6, 0, 0);
253 #endif
254 		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
255 		    rt->rt_gateway->sa_family == AF_LINK);
256 		if (rt)
257 			rtfree(rt);
258 		if (need_proxy) {
259 			/*
260 			 * proxy NDP for single entry
261 			 */
262 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
263 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
264 			if (ifa) {
265 				proxy = 1;
266 				proxydl = SDL(rt->rt_gateway);
267 			}
268 		}
269 	}
270 	if (ifa == NULL) {
271 		/*
272 		 * We've got an NS packet, and we don't have that adddress
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);
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(ifp, &in6_all, &taddr6,
339 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
340 		    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
341 		    tlladdr, (struct sockaddr *)proxydl);
342 		goto freeit;
343 	}
344 
345 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
346 	    ND_NEIGHBOR_SOLICIT, 0);
347 
348 	nd6_na_output(ifp, &saddr6, &taddr6,
349 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
350 	    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
351 	    tlladdr, (struct sockaddr *)proxydl);
352  freeit:
353 	m_freem(m);
354 	return;
355 
356  bad:
357 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
358 		ip6_sprintf(ip6bufs, &saddr6)));
359 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
360 		ip6_sprintf(ip6bufs, &daddr6)));
361 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
362 		ip6_sprintf(ip6bufs, &taddr6)));
363 	V_icmp6stat.icp6s_badns++;
364 	m_freem(m);
365 }
366 
367 /*
368  * Output a Neighbor Solicitation Message. Caller specifies:
369  *	- ICMP6 header source IP6 address
370  *	- ND6 header target IP6 address
371  *	- ND6 header source datalink address
372  *
373  * Based on RFC 2461
374  * Based on RFC 2462 (duplicate address detection)
375  *
376  *   ln - for source address determination
377  *  dad - duplicate address detection
378  */
379 void
380 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
381     const struct in6_addr *taddr6, struct llinfo_nd6 *ln, int dad)
382 {
383 	INIT_VNET_INET6(ifp->if_vnet);
384 	struct mbuf *m;
385 	struct ip6_hdr *ip6;
386 	struct nd_neighbor_solicit *nd_ns;
387 	struct in6_addr *src, src_in;
388 	struct ip6_moptions im6o;
389 	int icmp6len;
390 	int maxlen;
391 	caddr_t mac;
392 	struct route_in6 ro;
393 
394 	bzero(&ro, sizeof(ro));
395 
396 	if (IN6_IS_ADDR_MULTICAST(taddr6))
397 		return;
398 
399 	/* estimate the size of message */
400 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
401 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
402 	if (max_linkhdr + maxlen >= MCLBYTES) {
403 #ifdef DIAGNOSTIC
404 		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
405 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
406 #endif
407 		return;
408 	}
409 
410 	MGETHDR(m, M_DONTWAIT, MT_DATA);
411 	if (m && max_linkhdr + maxlen >= MHLEN) {
412 		MCLGET(m, M_DONTWAIT);
413 		if ((m->m_flags & M_EXT) == 0) {
414 			m_free(m);
415 			m = NULL;
416 		}
417 	}
418 	if (m == NULL)
419 		return;
420 	m->m_pkthdr.rcvif = NULL;
421 
422 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
423 		m->m_flags |= M_MCAST;
424 		im6o.im6o_multicast_ifp = ifp;
425 		im6o.im6o_multicast_hlim = 255;
426 		im6o.im6o_multicast_loop = 0;
427 	}
428 
429 	icmp6len = sizeof(*nd_ns);
430 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
431 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
432 
433 	/* fill neighbor solicitation packet */
434 	ip6 = mtod(m, struct ip6_hdr *);
435 	ip6->ip6_flow = 0;
436 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
437 	ip6->ip6_vfc |= IPV6_VERSION;
438 	/* ip6->ip6_plen will be set later */
439 	ip6->ip6_nxt = IPPROTO_ICMPV6;
440 	ip6->ip6_hlim = 255;
441 	if (daddr6)
442 		ip6->ip6_dst = *daddr6;
443 	else {
444 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
445 		ip6->ip6_dst.s6_addr16[1] = 0;
446 		ip6->ip6_dst.s6_addr32[1] = 0;
447 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
448 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
449 		ip6->ip6_dst.s6_addr8[12] = 0xff;
450 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
451 			goto bad;
452 	}
453 	if (!dad) {
454 		/*
455 		 * RFC2461 7.2.2:
456 		 * "If the source address of the packet prompting the
457 		 * solicitation is the same as one of the addresses assigned
458 		 * to the outgoing interface, that address SHOULD be placed
459 		 * in the IP Source Address of the outgoing solicitation.
460 		 * Otherwise, any one of the addresses assigned to the
461 		 * interface should be used."
462 		 *
463 		 * We use the source address for the prompting packet
464 		 * (saddr6), if:
465 		 * - saddr6 is given from the caller (by giving "ln"), and
466 		 * - saddr6 belongs to the outgoing interface.
467 		 * Otherwise, we perform the source address selection as usual.
468 		 */
469 		struct ip6_hdr *hip6;		/* hold ip6 */
470 		struct in6_addr *hsrc = NULL;
471 
472 		if (ln && ln->ln_hold) {
473 			/*
474 			 * assuming every packet in ln_hold has the same IP
475 			 * header
476 			 */
477 			hip6 = mtod(ln->ln_hold, struct ip6_hdr *);
478 			/* XXX pullup? */
479 			if (sizeof(*hip6) < ln->ln_hold->m_len)
480 				hsrc = &hip6->ip6_src;
481 			else
482 				hsrc = NULL;
483 		}
484 		if (hsrc && in6ifa_ifpwithaddr(ifp, hsrc))
485 			src = hsrc;
486 		else {
487 			int error;
488 			struct sockaddr_in6 dst_sa;
489 
490 			bzero(&dst_sa, sizeof(dst_sa));
491 			dst_sa.sin6_family = AF_INET6;
492 			dst_sa.sin6_len = sizeof(dst_sa);
493 			dst_sa.sin6_addr = ip6->ip6_dst;
494 
495 			src = in6_selectsrc(&dst_sa, NULL,
496 			    NULL, &ro, NULL, NULL, &error);
497 			if (src == NULL) {
498 				char ip6buf[INET6_ADDRSTRLEN];
499 				nd6log((LOG_DEBUG,
500 				    "nd6_ns_output: source can't be "
501 				    "determined: dst=%s, error=%d\n",
502 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
503 				    error));
504 				goto bad;
505 			}
506 		}
507 	} else {
508 		/*
509 		 * Source address for DAD packet must always be IPv6
510 		 * unspecified address. (0::0)
511 		 * We actually don't have to 0-clear the address (we did it
512 		 * above), but we do so here explicitly to make the intention
513 		 * clearer.
514 		 */
515 		bzero(&src_in, sizeof(src_in));
516 		src = &src_in;
517 	}
518 	ip6->ip6_src = *src;
519 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
520 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
521 	nd_ns->nd_ns_code = 0;
522 	nd_ns->nd_ns_reserved = 0;
523 	nd_ns->nd_ns_target = *taddr6;
524 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
525 
526 	/*
527 	 * Add source link-layer address option.
528 	 *
529 	 *				spec		implementation
530 	 *				---		---
531 	 * DAD packet			MUST NOT	do not add the option
532 	 * there's no link layer address:
533 	 *				impossible	do not add the option
534 	 * there's link layer address:
535 	 *	Multicast NS		MUST add one	add the option
536 	 *	Unicast NS		SHOULD add one	add the option
537 	 */
538 	if (!dad && (mac = nd6_ifptomac(ifp))) {
539 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
540 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
541 		/* 8 byte alignments... */
542 		optlen = (optlen + 7) & ~7;
543 
544 		m->m_pkthdr.len += optlen;
545 		m->m_len += optlen;
546 		icmp6len += optlen;
547 		bzero((caddr_t)nd_opt, optlen);
548 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
549 		nd_opt->nd_opt_len = optlen >> 3;
550 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
551 	}
552 
553 	ip6->ip6_plen = htons((u_short)icmp6len);
554 	nd_ns->nd_ns_cksum = 0;
555 	nd_ns->nd_ns_cksum =
556 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
557 
558 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
559 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
560 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
561 	V_icmp6stat.icp6s_outhist[ND_NEIGHBOR_SOLICIT]++;
562 
563 	if (ro.ro_rt) {		/* we don't cache this route. */
564 		RTFREE(ro.ro_rt);
565 	}
566 	return;
567 
568   bad:
569 	if (ro.ro_rt) {
570 		RTFREE(ro.ro_rt);
571 	}
572 	m_freem(m);
573 	return;
574 }
575 
576 /*
577  * Neighbor advertisement input handling.
578  *
579  * Based on RFC 2461
580  * Based on RFC 2462 (duplicate address detection)
581  *
582  * the following items are not implemented yet:
583  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
584  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
585  */
586 void
587 nd6_na_input(struct mbuf *m, int off, int icmp6len)
588 {
589 	INIT_VNET_INET6(curvnet);
590 	struct ifnet *ifp = m->m_pkthdr.rcvif;
591 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
592 	struct nd_neighbor_advert *nd_na;
593 	struct in6_addr daddr6 = ip6->ip6_dst;
594 	struct in6_addr taddr6;
595 	int flags;
596 	int is_router;
597 	int is_solicited;
598 	int is_override;
599 	char *lladdr = NULL;
600 	int lladdrlen = 0;
601 	struct ifaddr *ifa;
602 	struct llinfo_nd6 *ln;
603 	struct rtentry *rt;
604 	struct sockaddr_dl *sdl;
605 	union nd_opts ndopts;
606 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
607 
608 	if (ip6->ip6_hlim != 255) {
609 		nd6log((LOG_ERR,
610 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
611 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
612 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
613 		goto bad;
614 	}
615 
616 #ifndef PULLDOWN_TEST
617 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
618 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
619 #else
620 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
621 	if (nd_na == NULL) {
622 		V_icmp6stat.icp6s_tooshort++;
623 		return;
624 	}
625 #endif
626 
627 	flags = nd_na->nd_na_flags_reserved;
628 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
629 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
630 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
631 
632 	taddr6 = nd_na->nd_na_target;
633 	if (in6_setscope(&taddr6, ifp, NULL))
634 		goto bad;	/* XXX: impossible */
635 
636 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
637 		nd6log((LOG_ERR,
638 		    "nd6_na_input: invalid target address %s\n",
639 		    ip6_sprintf(ip6bufs, &taddr6)));
640 		goto bad;
641 	}
642 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
643 		if (is_solicited) {
644 			nd6log((LOG_ERR,
645 			    "nd6_na_input: a solicited adv is multicasted\n"));
646 			goto bad;
647 		}
648 
649 	icmp6len -= sizeof(*nd_na);
650 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
651 	if (nd6_options(&ndopts) < 0) {
652 		nd6log((LOG_INFO,
653 		    "nd6_na_input: invalid ND option, ignored\n"));
654 		/* nd6_options have incremented stats */
655 		goto freeit;
656 	}
657 
658 	if (ndopts.nd_opts_tgt_lladdr) {
659 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
660 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
661 	}
662 
663 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
664 
665 	/*
666 	 * Target address matches one of my interface address.
667 	 *
668 	 * If my address is tentative, this means that there's somebody
669 	 * already using the same address as mine.  This indicates DAD failure.
670 	 * This is defined in RFC 2462.
671 	 *
672 	 * Otherwise, process as defined in RFC 2461.
673 	 */
674 	if (ifa
675 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
676 		nd6_dad_na_input(ifa);
677 		goto freeit;
678 	}
679 
680 	/* Just for safety, maybe unnecessary. */
681 	if (ifa) {
682 		log(LOG_ERR,
683 		    "nd6_na_input: duplicate IP6 address %s\n",
684 		    ip6_sprintf(ip6bufs, &taddr6));
685 		goto freeit;
686 	}
687 
688 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
689 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
690 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
691 		    ifp->if_addrlen, lladdrlen - 2));
692 		goto bad;
693 	}
694 
695 	/*
696 	 * If no neighbor cache entry is found, NA SHOULD silently be
697 	 * discarded.
698 	 */
699 	rt = nd6_lookup(&taddr6, 0, ifp);
700 	if ((rt == NULL) ||
701 	   ((ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) ||
702 	   ((sdl = SDL(rt->rt_gateway)) == NULL))
703 		goto freeit;
704 
705 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
706 		/*
707 		 * If the link-layer has address, and no lladdr option came,
708 		 * discard the packet.
709 		 */
710 		if (ifp->if_addrlen && lladdr == NULL)
711 			goto freeit;
712 
713 		/*
714 		 * Record link-layer address, and update the state.
715 		 */
716 		sdl->sdl_alen = ifp->if_addrlen;
717 		bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
718 		if (is_solicited) {
719 			ln->ln_state = ND6_LLINFO_REACHABLE;
720 			ln->ln_byhint = 0;
721 			if (!ND6_LLINFO_PERMANENT(ln)) {
722 				nd6_llinfo_settimer(ln,
723 				    (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
724 			}
725 		} else {
726 			ln->ln_state = ND6_LLINFO_STALE;
727 			nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
728 		}
729 		if ((ln->ln_router = is_router) != 0) {
730 			/*
731 			 * This means a router's state has changed from
732 			 * non-reachable to probably reachable, and might
733 			 * affect the status of associated prefixes..
734 			 */
735 			pfxlist_onlink_check();
736 		}
737 	} else {
738 		int llchange;
739 
740 		/*
741 		 * Check if the link-layer address has changed or not.
742 		 */
743 		if (lladdr == NULL)
744 			llchange = 0;
745 		else {
746 			if (sdl->sdl_alen) {
747 				if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
748 					llchange = 1;
749 				else
750 					llchange = 0;
751 			} else
752 				llchange = 1;
753 		}
754 
755 		/*
756 		 * This is VERY complex.  Look at it with care.
757 		 *
758 		 * override solicit lladdr llchange	action
759 		 *					(L: record lladdr)
760 		 *
761 		 *	0	0	n	--	(2c)
762 		 *	0	0	y	n	(2b) L
763 		 *	0	0	y	y	(1)    REACHABLE->STALE
764 		 *	0	1	n	--	(2c)   *->REACHABLE
765 		 *	0	1	y	n	(2b) L *->REACHABLE
766 		 *	0	1	y	y	(1)    REACHABLE->STALE
767 		 *	1	0	n	--	(2a)
768 		 *	1	0	y	n	(2a) L
769 		 *	1	0	y	y	(2a) L *->STALE
770 		 *	1	1	n	--	(2a)   *->REACHABLE
771 		 *	1	1	y	n	(2a) L *->REACHABLE
772 		 *	1	1	y	y	(2a) L *->REACHABLE
773 		 */
774 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
775 			/*
776 			 * If state is REACHABLE, make it STALE.
777 			 * no other updates should be done.
778 			 */
779 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
780 				ln->ln_state = ND6_LLINFO_STALE;
781 				nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz);
782 			}
783 			goto freeit;
784 		} else if (is_override				   /* (2a) */
785 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
786 			|| lladdr == NULL) {			   /* (2c) */
787 			/*
788 			 * Update link-local address, if any.
789 			 */
790 			if (lladdr != NULL) {
791 				sdl->sdl_alen = ifp->if_addrlen;
792 				bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
793 			}
794 
795 			/*
796 			 * If solicited, make the state REACHABLE.
797 			 * If not solicited and the link-layer address was
798 			 * changed, make it STALE.
799 			 */
800 			if (is_solicited) {
801 				ln->ln_state = ND6_LLINFO_REACHABLE;
802 				ln->ln_byhint = 0;
803 				if (!ND6_LLINFO_PERMANENT(ln)) {
804 					nd6_llinfo_settimer(ln,
805 					    (long)ND_IFINFO(ifp)->reachable * hz);
806 				}
807 			} else {
808 				if (lladdr != NULL && llchange) {
809 					ln->ln_state = ND6_LLINFO_STALE;
810 					nd6_llinfo_settimer(ln,
811 					    (long)V_nd6_gctimer * hz);
812 				}
813 			}
814 		}
815 
816 		if (ln->ln_router && !is_router) {
817 			/*
818 			 * The peer dropped the router flag.
819 			 * Remove the sender from the Default Router List and
820 			 * update the Destination Cache entries.
821 			 */
822 			struct nd_defrouter *dr;
823 			struct in6_addr *in6;
824 			int s;
825 
826 			in6 = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
827 
828 			/*
829 			 * Lock to protect the default router list.
830 			 * XXX: this might be unnecessary, since this function
831 			 * is only called under the network software interrupt
832 			 * context.  However, we keep it just for safety.
833 			 */
834 			s = splnet();
835 			dr = defrouter_lookup(in6, ifp);
836 			if (dr)
837 				defrtrlist_del(dr);
838 			else if (!V_ip6_forwarding) {
839 				/*
840 				 * Even if the neighbor is not in the default
841 				 * router list, the neighbor may be used
842 				 * as a next hop for some destinations
843 				 * (e.g. redirect case). So we must
844 				 * call rt6_flush explicitly.
845 				 */
846 				rt6_flush(&ip6->ip6_src, ifp);
847 			}
848 			splx(s);
849 		}
850 		ln->ln_router = is_router;
851 	}
852 	rt->rt_flags &= ~RTF_REJECT;
853 	ln->ln_asked = 0;
854 	if (ln->ln_hold) {
855 		struct mbuf *m_hold, *m_hold_next;
856 
857 		/*
858 		 * reset the ln_hold in advance, to explicitly
859 		 * prevent a ln_hold lookup in nd6_output()
860 		 * (wouldn't happen, though...)
861 		 */
862 		for (m_hold = ln->ln_hold;
863 		    m_hold; m_hold = m_hold_next) {
864 			m_hold_next = m_hold->m_nextpkt;
865 			m_hold->m_nextpkt = NULL;
866 			/*
867 			 * we assume ifp is not a loopback here, so just set
868 			 * the 2nd argument as the 1st one.
869 			 */
870 			nd6_output(ifp, ifp, m_hold,
871 			    (struct sockaddr_in6 *)rt_key(rt), rt);
872 		}
873 		ln->ln_hold = NULL;
874 	}
875 
876  freeit:
877 	m_freem(m);
878 	return;
879 
880  bad:
881 	V_icmp6stat.icp6s_badna++;
882 	m_freem(m);
883 }
884 
885 /*
886  * Neighbor advertisement output handling.
887  *
888  * Based on RFC 2461
889  *
890  * the following items are not implemented yet:
891  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
892  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
893  *
894  * tlladdr - 1 if include target link-layer address
895  * sdl0 - sockaddr_dl (= proxy NA) or NULL
896  */
897 void
898 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
899     const struct in6_addr *taddr6, u_long flags, int tlladdr,
900     struct sockaddr *sdl0)
901 {
902 	INIT_VNET_INET6(ifp->if_vnet);
903 	struct mbuf *m;
904 	struct ip6_hdr *ip6;
905 	struct nd_neighbor_advert *nd_na;
906 	struct ip6_moptions im6o;
907 	struct in6_addr *src, daddr6;
908 	struct sockaddr_in6 dst_sa;
909 	int icmp6len, maxlen, error;
910 	caddr_t mac = NULL;
911 	struct route_in6 ro;
912 
913 	bzero(&ro, sizeof(ro));
914 
915 	daddr6 = *daddr6_0;	/* make a local copy for modification */
916 
917 	/* estimate the size of message */
918 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
919 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
920 	if (max_linkhdr + maxlen >= MCLBYTES) {
921 #ifdef DIAGNOSTIC
922 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
923 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
924 #endif
925 		return;
926 	}
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->m_pkthdr.rcvif = NULL;
939 
940 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
941 		m->m_flags |= M_MCAST;
942 		im6o.im6o_multicast_ifp = 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 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
972 	dst_sa.sin6_family = AF_INET6;
973 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
974 	dst_sa.sin6_addr = daddr6;
975 
976 	/*
977 	 * Select a source whose scope is the same as that of the dest.
978 	 */
979 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
980 	src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &error);
981 	if (src == NULL) {
982 		char ip6buf[INET6_ADDRSTRLEN];
983 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
984 		    "determined: dst=%s, error=%d\n",
985 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
986 		goto bad;
987 	}
988 	ip6->ip6_src = *src;
989 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
990 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
991 	nd_na->nd_na_code = 0;
992 	nd_na->nd_na_target = *taddr6;
993 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
994 
995 	/*
996 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
997 	 * see nd6_ns_input() for details.
998 	 * Basically, if NS packet is sent to unicast/anycast addr,
999 	 * target lladdr option SHOULD NOT be included.
1000 	 */
1001 	if (tlladdr) {
1002 		/*
1003 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1004 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1005 		 * my address) use lladdr configured for the interface.
1006 		 */
1007 		if (sdl0 == NULL) {
1008 #ifdef DEV_CARP
1009 			if (ifp->if_carp)
1010 				mac = carp_macmatch6(ifp->if_carp, m, taddr6);
1011 			if (mac == NULL)
1012 				mac = nd6_ifptomac(ifp);
1013 #else
1014 			mac = nd6_ifptomac(ifp);
1015 #endif
1016 		} else if (sdl0->sa_family == AF_LINK) {
1017 			struct sockaddr_dl *sdl;
1018 			sdl = (struct sockaddr_dl *)sdl0;
1019 			if (sdl->sdl_alen == ifp->if_addrlen)
1020 				mac = LLADDR(sdl);
1021 		}
1022 	}
1023 	if (tlladdr && mac) {
1024 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1025 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1026 
1027 		/* roundup to 8 bytes alignment! */
1028 		optlen = (optlen + 7) & ~7;
1029 
1030 		m->m_pkthdr.len += optlen;
1031 		m->m_len += optlen;
1032 		icmp6len += optlen;
1033 		bzero((caddr_t)nd_opt, optlen);
1034 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1035 		nd_opt->nd_opt_len = optlen >> 3;
1036 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1037 	} else
1038 		flags &= ~ND_NA_FLAG_OVERRIDE;
1039 
1040 	ip6->ip6_plen = htons((u_short)icmp6len);
1041 	nd_na->nd_na_flags_reserved = flags;
1042 	nd_na->nd_na_cksum = 0;
1043 	nd_na->nd_na_cksum =
1044 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1045 
1046 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1047 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1048 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1049 	V_icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1050 
1051 	if (ro.ro_rt) {		/* we don't cache this route. */
1052 		RTFREE(ro.ro_rt);
1053 	}
1054 	return;
1055 
1056   bad:
1057 	if (ro.ro_rt) {
1058 		RTFREE(ro.ro_rt);
1059 	}
1060 	m_freem(m);
1061 	return;
1062 }
1063 
1064 caddr_t
1065 nd6_ifptomac(struct ifnet *ifp)
1066 {
1067 	switch (ifp->if_type) {
1068 	case IFT_ARCNET:
1069 	case IFT_ETHER:
1070 	case IFT_FDDI:
1071 	case IFT_IEEE1394:
1072 #ifdef IFT_L2VLAN
1073 	case IFT_L2VLAN:
1074 #endif
1075 #ifdef IFT_IEEE80211
1076 	case IFT_IEEE80211:
1077 #endif
1078 #ifdef IFT_CARP
1079 	case IFT_CARP:
1080 #endif
1081 	case IFT_BRIDGE:
1082 	case IFT_ISO88025:
1083 		return IF_LLADDR(ifp);
1084 	default:
1085 		return NULL;
1086 	}
1087 }
1088 
1089 TAILQ_HEAD(dadq_head, dadq);
1090 struct dadq {
1091 	TAILQ_ENTRY(dadq) dad_list;
1092 	struct ifaddr *dad_ifa;
1093 	int dad_count;		/* max NS to send */
1094 	int dad_ns_tcount;	/* # of trials to send NS */
1095 	int dad_ns_ocount;	/* NS sent so far */
1096 	int dad_ns_icount;
1097 	int dad_na_icount;
1098 	struct callout dad_timer_ch;
1099 };
1100 
1101 #ifdef VIMAGE_GLOBALS
1102 static struct dadq_head dadq;
1103 int dad_init;
1104 #endif
1105 
1106 static struct dadq *
1107 nd6_dad_find(struct ifaddr *ifa)
1108 {
1109 	INIT_VNET_INET6(curvnet);
1110 	struct dadq *dp;
1111 
1112 	for (dp = V_dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1113 		if (dp->dad_ifa == ifa)
1114 			return dp;
1115 	}
1116 	return NULL;
1117 }
1118 
1119 static void
1120 nd6_dad_starttimer(struct dadq *dp, int ticks)
1121 {
1122 
1123 	callout_reset(&dp->dad_timer_ch, ticks,
1124 	    (void (*)(void *))nd6_dad_timer, (void *)dp->dad_ifa);
1125 }
1126 
1127 static void
1128 nd6_dad_stoptimer(struct dadq *dp)
1129 {
1130 
1131 	callout_stop(&dp->dad_timer_ch);
1132 }
1133 
1134 /*
1135  * Start Duplicate Address Detection (DAD) for specified interface address.
1136  */
1137 void
1138 nd6_dad_start(struct ifaddr *ifa, int delay)
1139 {
1140 	INIT_VNET_INET6(curvnet);
1141 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1142 	struct dadq *dp;
1143 	char ip6buf[INET6_ADDRSTRLEN];
1144 
1145 	if (!V_dad_init) {
1146 		TAILQ_INIT(&V_dadq);
1147 		V_dad_init++;
1148 	}
1149 
1150 	/*
1151 	 * If we don't need DAD, don't do it.
1152 	 * There are several cases:
1153 	 * - DAD is disabled (ip6_dad_count == 0)
1154 	 * - the interface address is anycast
1155 	 */
1156 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1157 		log(LOG_DEBUG,
1158 			"nd6_dad_start: called with non-tentative address "
1159 			"%s(%s)\n",
1160 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1161 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1162 		return;
1163 	}
1164 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1165 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1166 		return;
1167 	}
1168 	if (!V_ip6_dad_count) {
1169 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1170 		return;
1171 	}
1172 	if (ifa->ifa_ifp == NULL)
1173 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1174 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1175 		return;
1176 	}
1177 	if (nd6_dad_find(ifa) != NULL) {
1178 		/* DAD already in progress */
1179 		return;
1180 	}
1181 
1182 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1183 	if (dp == NULL) {
1184 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1185 			"%s(%s)\n",
1186 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1187 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1188 		return;
1189 	}
1190 	bzero(dp, sizeof(*dp));
1191 	callout_init(&dp->dad_timer_ch, 0);
1192 	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1193 
1194 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1195 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1196 
1197 	/*
1198 	 * Send NS packet for DAD, ip6_dad_count times.
1199 	 * Note that we must delay the first transmission, if this is the
1200 	 * first packet to be sent from the interface after interface
1201 	 * (re)initialization.
1202 	 */
1203 	dp->dad_ifa = ifa;
1204 	IFAREF(ifa);	/* just for safety */
1205 	dp->dad_count = V_ip6_dad_count;
1206 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1207 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1208 	if (delay == 0) {
1209 		nd6_dad_ns_output(dp, ifa);
1210 		nd6_dad_starttimer(dp,
1211 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1212 	} else {
1213 		nd6_dad_starttimer(dp, delay);
1214 	}
1215 }
1216 
1217 /*
1218  * terminate DAD unconditionally.  used for address removals.
1219  */
1220 void
1221 nd6_dad_stop(struct ifaddr *ifa)
1222 {
1223 	INIT_VNET_INET6(curvnet);
1224 	struct dadq *dp;
1225 
1226 	if (!V_dad_init)
1227 		return;
1228 	dp = nd6_dad_find(ifa);
1229 	if (!dp) {
1230 		/* DAD wasn't started yet */
1231 		return;
1232 	}
1233 
1234 	nd6_dad_stoptimer(dp);
1235 
1236 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1237 	free(dp, M_IP6NDP);
1238 	dp = NULL;
1239 	IFAFREE(ifa);
1240 }
1241 
1242 static void
1243 nd6_dad_timer(struct ifaddr *ifa)
1244 {
1245 	CURVNET_SET(dp->dad_vnet);
1246 	INIT_VNET_INET6(curvnet);
1247 	int s;
1248 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1249 	struct dadq *dp;
1250 	char ip6buf[INET6_ADDRSTRLEN];
1251 
1252 	s = splnet();		/* XXX */
1253 
1254 	/* Sanity check */
1255 	if (ia == NULL) {
1256 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1257 		goto done;
1258 	}
1259 	dp = nd6_dad_find(ifa);
1260 	if (dp == NULL) {
1261 		log(LOG_ERR, "nd6_dad_timer: DAD structure not found\n");
1262 		goto done;
1263 	}
1264 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1265 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1266 			"%s(%s)\n",
1267 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1268 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1269 		goto done;
1270 	}
1271 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1272 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1273 			"%s(%s)\n",
1274 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1275 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1276 		goto done;
1277 	}
1278 
1279 	/* timeouted with IFF_{RUNNING,UP} check */
1280 	if (dp->dad_ns_tcount > V_dad_maxtry) {
1281 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1282 		    if_name(ifa->ifa_ifp)));
1283 
1284 		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1285 		free(dp, M_IP6NDP);
1286 		dp = NULL;
1287 		IFAFREE(ifa);
1288 		goto done;
1289 	}
1290 
1291 	/* Need more checks? */
1292 	if (dp->dad_ns_ocount < dp->dad_count) {
1293 		/*
1294 		 * We have more NS to go.  Send NS packet for DAD.
1295 		 */
1296 		nd6_dad_ns_output(dp, ifa);
1297 		nd6_dad_starttimer(dp,
1298 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1299 	} else {
1300 		/*
1301 		 * We have transmitted sufficient number of DAD packets.
1302 		 * See what we've got.
1303 		 */
1304 		int duplicate;
1305 
1306 		duplicate = 0;
1307 
1308 		if (dp->dad_na_icount) {
1309 			/*
1310 			 * the check is in nd6_dad_na_input(),
1311 			 * but just in case
1312 			 */
1313 			duplicate++;
1314 		}
1315 
1316 		if (dp->dad_ns_icount) {
1317 			/* We've seen NS, means DAD has failed. */
1318 			duplicate++;
1319 		}
1320 
1321 		if (duplicate) {
1322 			/* (*dp) will be freed in nd6_dad_duplicated() */
1323 			dp = NULL;
1324 			nd6_dad_duplicated(ifa);
1325 		} else {
1326 			/*
1327 			 * We are done with DAD.  No NA came, no NS came.
1328 			 * No duplicate address found.
1329 			 */
1330 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1331 
1332 			nd6log((LOG_DEBUG,
1333 			    "%s: DAD complete for %s - no duplicates found\n",
1334 			    if_name(ifa->ifa_ifp),
1335 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1336 
1337 			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1338 			free(dp, M_IP6NDP);
1339 			dp = NULL;
1340 			IFAFREE(ifa);
1341 		}
1342 	}
1343 
1344 done:
1345 	splx(s);
1346 	CURVNET_RESTORE();
1347 }
1348 
1349 void
1350 nd6_dad_duplicated(struct ifaddr *ifa)
1351 {
1352 	INIT_VNET_INET6(curvnet);
1353 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1354 	struct ifnet *ifp;
1355 	struct dadq *dp;
1356 	char ip6buf[INET6_ADDRSTRLEN];
1357 
1358 	dp = nd6_dad_find(ifa);
1359 	if (dp == NULL) {
1360 		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1361 		return;
1362 	}
1363 
1364 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1365 	    "NS in/out=%d/%d, NA in=%d\n",
1366 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1367 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1368 
1369 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1370 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1371 
1372 	/* We are done with DAD, with duplicate address found. (failure) */
1373 	nd6_dad_stoptimer(dp);
1374 
1375 	ifp = ifa->ifa_ifp;
1376 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1377 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1378 	log(LOG_ERR, "%s: manual intervention required\n",
1379 	    if_name(ifp));
1380 
1381 	/*
1382 	 * If the address is a link-local address formed from an interface
1383 	 * identifier based on the hardware address which is supposed to be
1384 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1385 	 * operation on the interface SHOULD be disabled.
1386 	 * [rfc2462bis-03 Section 5.4.5]
1387 	 */
1388 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1389 		struct in6_addr in6;
1390 
1391 		/*
1392 		 * To avoid over-reaction, we only apply this logic when we are
1393 		 * very sure that hardware addresses are supposed to be unique.
1394 		 */
1395 		switch (ifp->if_type) {
1396 		case IFT_ETHER:
1397 		case IFT_FDDI:
1398 		case IFT_ATM:
1399 		case IFT_IEEE1394:
1400 #ifdef IFT_IEEE80211
1401 		case IFT_IEEE80211:
1402 #endif
1403 			in6 = ia->ia_addr.sin6_addr;
1404 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1405 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1406 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1407 				log(LOG_ERR, "%s: possible hardware address "
1408 				    "duplication detected, disable IPv6\n",
1409 				    if_name(ifp));
1410 			}
1411 			break;
1412 		}
1413 	}
1414 
1415 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1416 	free(dp, M_IP6NDP);
1417 	dp = NULL;
1418 	IFAFREE(ifa);
1419 }
1420 
1421 static void
1422 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1423 {
1424 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1425 	struct ifnet *ifp = ifa->ifa_ifp;
1426 
1427 	dp->dad_ns_tcount++;
1428 	if ((ifp->if_flags & IFF_UP) == 0) {
1429 		return;
1430 	}
1431 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1432 		return;
1433 	}
1434 
1435 	dp->dad_ns_ocount++;
1436 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1437 }
1438 
1439 static void
1440 nd6_dad_ns_input(struct ifaddr *ifa)
1441 {
1442 	INIT_VNET_INET6(curvnet);
1443 	struct in6_ifaddr *ia;
1444 	struct ifnet *ifp;
1445 	const struct in6_addr *taddr6;
1446 	struct dadq *dp;
1447 	int duplicate;
1448 
1449 	if (ifa == NULL)
1450 		panic("ifa == NULL in nd6_dad_ns_input");
1451 
1452 	ia = (struct in6_ifaddr *)ifa;
1453 	ifp = ifa->ifa_ifp;
1454 	taddr6 = &ia->ia_addr.sin6_addr;
1455 	duplicate = 0;
1456 	dp = nd6_dad_find(ifa);
1457 
1458 	/* Quickhack - completely ignore DAD NS packets */
1459 	if (V_dad_ignore_ns) {
1460 		char ip6buf[INET6_ADDRSTRLEN];
1461 		nd6log((LOG_INFO,
1462 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1463 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1464 		    if_name(ifa->ifa_ifp)));
1465 		return;
1466 	}
1467 
1468 	/*
1469 	 * if I'm yet to start DAD, someone else started using this address
1470 	 * first.  I have a duplicate and you win.
1471 	 */
1472 	if (dp == NULL || dp->dad_ns_ocount == 0)
1473 		duplicate++;
1474 
1475 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1476 
1477 	if (duplicate) {
1478 		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1479 		nd6_dad_duplicated(ifa);
1480 	} else {
1481 		/*
1482 		 * not sure if I got a duplicate.
1483 		 * increment ns count and see what happens.
1484 		 */
1485 		if (dp)
1486 			dp->dad_ns_icount++;
1487 	}
1488 }
1489 
1490 static void
1491 nd6_dad_na_input(struct ifaddr *ifa)
1492 {
1493 	struct dadq *dp;
1494 
1495 	if (ifa == NULL)
1496 		panic("ifa == NULL in nd6_dad_na_input");
1497 
1498 	dp = nd6_dad_find(ifa);
1499 	if (dp)
1500 		dp->dad_na_icount++;
1501 
1502 	/* remove the address. */
1503 	nd6_dad_duplicated(ifa);
1504 }
1505