xref: /freebsd/sys/netinet6/nd6_nbr.c (revision 9162f64b58d01ec01481d60b6cdc06ffd8e8c7fc)
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/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/syslog.h>
53 #include <sys/queue.h>
54 #include <sys/callout.h>
55 #include <sys/vimage.h>
56 
57 #include <net/if.h>
58 #include <net/if_types.h>
59 #include <net/if_dl.h>
60 #include <net/if_var.h>
61 #include <net/route.h>
62 #ifdef RADIX_MPATH
63 #include <net/radix_mpath.h>
64 #endif
65 
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <net/if_llatbl.h>
69 #define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
70 #include <netinet6/in6_var.h>
71 #include <netinet6/in6_ifattach.h>
72 #include <netinet/ip6.h>
73 #include <netinet6/ip6_var.h>
74 #include <netinet6/scope6_var.h>
75 #include <netinet6/nd6.h>
76 #include <netinet/icmp6.h>
77 #include <netinet6/vinet6.h>
78 
79 #ifdef DEV_CARP
80 #include <netinet/ip_carp.h>
81 #endif
82 
83 #define SDL(s) ((struct sockaddr_dl *)s)
84 
85 struct dadq;
86 static struct dadq *nd6_dad_find(struct ifaddr *);
87 static void nd6_dad_starttimer(struct dadq *, int);
88 static void nd6_dad_stoptimer(struct dadq *);
89 static void nd6_dad_timer(struct ifaddr *);
90 static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
91 static void nd6_dad_ns_input(struct ifaddr *);
92 static void nd6_dad_na_input(struct ifaddr *);
93 
94 #ifdef VIMAGE_GLOBALS
95 int dad_ignore_ns;
96 int dad_maxtry;
97 #endif
98 
99 /*
100  * Input a Neighbor Solicitation Message.
101  *
102  * Based on RFC 2461
103  * Based on RFC 2462 (duplicate address detection)
104  */
105 void
106 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
107 {
108 	INIT_VNET_INET6(curvnet);
109 	struct ifnet *ifp = m->m_pkthdr.rcvif;
110 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
111 	struct nd_neighbor_solicit *nd_ns;
112 	struct in6_addr saddr6 = ip6->ip6_src;
113 	struct in6_addr daddr6 = ip6->ip6_dst;
114 	struct in6_addr taddr6;
115 	struct in6_addr myaddr6;
116 	char *lladdr = NULL;
117 	struct ifaddr *ifa = NULL;
118 	int lladdrlen = 0;
119 	int anycast = 0, proxy = 0, tentative = 0;
120 	int tlladdr;
121 	union nd_opts ndopts;
122 	struct sockaddr_dl *proxydl = NULL;
123 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
124 
125 #ifndef PULLDOWN_TEST
126 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
127 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
128 #else
129 	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
130 	if (nd_ns == NULL) {
131 		V_icmp6stat.icp6s_tooshort++;
132 		return;
133 	}
134 #endif
135 	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
136 	taddr6 = nd_ns->nd_ns_target;
137 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
138 		goto bad;
139 
140 	if (ip6->ip6_hlim != 255) {
141 		nd6log((LOG_ERR,
142 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
143 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
144 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
145 		goto bad;
146 	}
147 
148 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
149 		/* dst has to be a solicited node multicast address. */
150 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
151 		    /* don't check ifindex portion */
152 		    daddr6.s6_addr32[1] == 0 &&
153 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
154 		    daddr6.s6_addr8[12] == 0xff) {
155 			; /* good */
156 		} else {
157 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
158 			    "(wrong ip6 dst)\n"));
159 			goto bad;
160 		}
161 	} else if (!V_nd6_onlink_ns_rfc4861) {
162 		struct sockaddr_in6 src_sa6;
163 
164 		/*
165 		 * According to recent IETF discussions, it is not a good idea
166 		 * to accept a NS from an address which would not be deemed
167 		 * to be a neighbor otherwise.  This point is expected to be
168 		 * clarified in future revisions of the specification.
169 		 */
170 		bzero(&src_sa6, sizeof(src_sa6));
171 		src_sa6.sin6_family = AF_INET6;
172 		src_sa6.sin6_len = sizeof(src_sa6);
173 		src_sa6.sin6_addr = saddr6;
174 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
175 			nd6log((LOG_INFO, "nd6_ns_input: "
176 				"NS packet from non-neighbor\n"));
177 			goto bad;
178 		}
179 	}
180 
181 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
182 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
183 		goto bad;
184 	}
185 
186 	icmp6len -= sizeof(*nd_ns);
187 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
188 	if (nd6_options(&ndopts) < 0) {
189 		nd6log((LOG_INFO,
190 		    "nd6_ns_input: invalid ND option, ignored\n"));
191 		/* nd6_options have incremented stats */
192 		goto freeit;
193 	}
194 
195 	if (ndopts.nd_opts_src_lladdr) {
196 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
197 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
198 	}
199 
200 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
201 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
202 		    "(link-layer address option)\n"));
203 		goto bad;
204 	}
205 
206 	/*
207 	 * Attaching target link-layer address to the NA?
208 	 * (RFC 2461 7.2.4)
209 	 *
210 	 * NS IP dst is unicast/anycast			MUST NOT add
211 	 * NS IP dst is solicited-node multicast	MUST add
212 	 *
213 	 * In implementation, we add target link-layer address by default.
214 	 * We do not add one in MUST NOT cases.
215 	 */
216 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
217 		tlladdr = 0;
218 	else
219 		tlladdr = 1;
220 
221 	/*
222 	 * Target address (taddr6) must be either:
223 	 * (1) Valid unicast/anycast address for my receiving interface,
224 	 * (2) Unicast address for which I'm offering proxy service, or
225 	 * (3) "tentative" address on which DAD is being performed.
226 	 */
227 	/* (1) and (3) check. */
228 #ifdef DEV_CARP
229 	if (ifp->if_carp)
230 		ifa = carp_iamatch6(ifp->if_carp, &taddr6);
231 	if (ifa == NULL)
232 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
233 #else
234 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
235 #endif
236 
237 	/* (2) check. */
238 	if (ifa == NULL) {
239 		struct rtentry *rt;
240 		struct sockaddr_in6 tsin6;
241 		int need_proxy;
242 #ifdef RADIX_MPATH
243 		struct route_in6 ro;
244 #endif
245 
246 		bzero(&tsin6, sizeof tsin6);
247 		tsin6.sin6_len = sizeof(struct sockaddr_in6);
248 		tsin6.sin6_family = AF_INET6;
249 		tsin6.sin6_addr = taddr6;
250 
251 #ifdef RADIX_MPATH
252 		bzero(&ro, sizeof(ro));
253 		ro.ro_dst = tsin6;
254 		rtalloc_mpath((struct route *)&ro, RTF_ANNOUNCE);
255 		rt = ro.ro_rt;
256 #else
257 		rt = rtalloc1((struct sockaddr *)&tsin6, 0, 0);
258 #endif
259 		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
260 		    rt->rt_gateway->sa_family == AF_LINK);
261 		if (rt)
262 			rtfree(rt);
263 		if (need_proxy) {
264 			/*
265 			 * proxy NDP for single entry
266 			 */
267 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
268 				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
269 			if (ifa) {
270 				proxy = 1;
271 				proxydl = SDL(rt->rt_gateway);
272 			}
273 		}
274 	}
275 	if (ifa == NULL) {
276 		/*
277 		 * We've got an NS packet, and we don't have that adddress
278 		 * assigned for us.  We MUST silently ignore it.
279 		 * See RFC2461 7.2.3.
280 		 */
281 		goto freeit;
282 	}
283 	myaddr6 = *IFA_IN6(ifa);
284 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
285 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
286 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
287 		goto freeit;
288 
289 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
290 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
291 		    "(if %d, NS packet %d)\n",
292 		    ip6_sprintf(ip6bufs, &taddr6),
293 		    ifp->if_addrlen, lladdrlen - 2));
294 		goto bad;
295 	}
296 
297 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
298 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
299 		    ip6_sprintf(ip6bufs, &saddr6)));
300 		goto freeit;
301 	}
302 
303 	/*
304 	 * We have neighbor solicitation packet, with target address equals to
305 	 * one of my tentative address.
306 	 *
307 	 * src addr	how to process?
308 	 * ---		---
309 	 * multicast	of course, invalid (rejected in ip6_input)
310 	 * unicast	somebody is doing address resolution -> ignore
311 	 * unspec	dup address detection
312 	 *
313 	 * The processing is defined in RFC 2462.
314 	 */
315 	if (tentative) {
316 		/*
317 		 * If source address is unspecified address, it is for
318 		 * duplicate address detection.
319 		 *
320 		 * If not, the packet is for addess resolution;
321 		 * silently ignore it.
322 		 */
323 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
324 			nd6_dad_ns_input(ifa);
325 
326 		goto freeit;
327 	}
328 
329 	/*
330 	 * If the source address is unspecified address, entries must not
331 	 * be created or updated.
332 	 * It looks that sender is performing DAD.  Output NA toward
333 	 * all-node multicast address, to tell the sender that I'm using
334 	 * the address.
335 	 * S bit ("solicited") must be zero.
336 	 */
337 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
338 		struct in6_addr in6_all;
339 
340 		in6_all = in6addr_linklocal_allnodes;
341 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
342 			goto bad;
343 		nd6_na_output(ifp, &in6_all, &taddr6,
344 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
345 		    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
346 		    tlladdr, (struct sockaddr *)proxydl);
347 		goto freeit;
348 	}
349 
350 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
351 	    ND_NEIGHBOR_SOLICIT, 0);
352 
353 	nd6_na_output(ifp, &saddr6, &taddr6,
354 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
355 	    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
356 	    tlladdr, (struct sockaddr *)proxydl);
357  freeit:
358 	m_freem(m);
359 	return;
360 
361  bad:
362 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
363 		ip6_sprintf(ip6bufs, &saddr6)));
364 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
365 		ip6_sprintf(ip6bufs, &daddr6)));
366 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
367 		ip6_sprintf(ip6bufs, &taddr6)));
368 	V_icmp6stat.icp6s_badns++;
369 	m_freem(m);
370 }
371 
372 /*
373  * Output a Neighbor Solicitation Message. Caller specifies:
374  *	- ICMP6 header source IP6 address
375  *	- ND6 header target IP6 address
376  *	- ND6 header source datalink address
377  *
378  * Based on RFC 2461
379  * Based on RFC 2462 (duplicate address detection)
380  *
381  *   ln - for source address determination
382  *  dad - duplicate address detection
383  */
384 void
385 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
386     const struct in6_addr *taddr6, struct llentry *ln, int dad)
387 {
388 	INIT_VNET_INET6(ifp->if_vnet);
389 	struct mbuf *m;
390 	struct ip6_hdr *ip6;
391 	struct nd_neighbor_solicit *nd_ns;
392 	struct in6_addr *src, src_in;
393 	struct ip6_moptions im6o;
394 	int icmp6len;
395 	int maxlen;
396 	caddr_t mac;
397 	struct route_in6 ro;
398 
399 	bzero(&ro, sizeof(ro));
400 
401 	if (IN6_IS_ADDR_MULTICAST(taddr6))
402 		return;
403 
404 	/* estimate the size of message */
405 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
406 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
407 	if (max_linkhdr + maxlen >= MCLBYTES) {
408 #ifdef DIAGNOSTIC
409 		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
410 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
411 #endif
412 		return;
413 	}
414 
415 	MGETHDR(m, M_DONTWAIT, MT_DATA);
416 	if (m && max_linkhdr + maxlen >= MHLEN) {
417 		MCLGET(m, M_DONTWAIT);
418 		if ((m->m_flags & M_EXT) == 0) {
419 			m_free(m);
420 			m = NULL;
421 		}
422 	}
423 	if (m == NULL)
424 		return;
425 	m->m_pkthdr.rcvif = NULL;
426 
427 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
428 		m->m_flags |= M_MCAST;
429 		im6o.im6o_multicast_ifp = ifp;
430 		im6o.im6o_multicast_hlim = 255;
431 		im6o.im6o_multicast_loop = 0;
432 	}
433 
434 	icmp6len = sizeof(*nd_ns);
435 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
436 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
437 
438 	/* fill neighbor solicitation packet */
439 	ip6 = mtod(m, struct ip6_hdr *);
440 	ip6->ip6_flow = 0;
441 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
442 	ip6->ip6_vfc |= IPV6_VERSION;
443 	/* ip6->ip6_plen will be set later */
444 	ip6->ip6_nxt = IPPROTO_ICMPV6;
445 	ip6->ip6_hlim = 255;
446 	if (daddr6)
447 		ip6->ip6_dst = *daddr6;
448 	else {
449 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
450 		ip6->ip6_dst.s6_addr16[1] = 0;
451 		ip6->ip6_dst.s6_addr32[1] = 0;
452 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
453 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
454 		ip6->ip6_dst.s6_addr8[12] = 0xff;
455 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
456 			goto bad;
457 	}
458 	if (!dad) {
459 		/*
460 		 * RFC2461 7.2.2:
461 		 * "If the source address of the packet prompting the
462 		 * solicitation is the same as one of the addresses assigned
463 		 * to the outgoing interface, that address SHOULD be placed
464 		 * in the IP Source Address of the outgoing solicitation.
465 		 * Otherwise, any one of the addresses assigned to the
466 		 * interface should be used."
467 		 *
468 		 * We use the source address for the prompting packet
469 		 * (saddr6), if:
470 		 * - saddr6 is given from the caller (by giving "ln"), and
471 		 * - saddr6 belongs to the outgoing interface.
472 		 * Otherwise, we perform the source address selection as usual.
473 		 */
474 		struct ip6_hdr *hip6;		/* hold ip6 */
475 		struct in6_addr *hsrc = NULL;
476 
477 		if ((ln != NULL) && ln->la_hold) {
478 			/*
479 			 * assuming every packet in la_hold has the same IP
480 			 * header
481 			 */
482 			hip6 = mtod(ln->la_hold, struct ip6_hdr *);
483 			/* XXX pullup? */
484 			if (sizeof(*hip6) < ln->la_hold->m_len)
485 				hsrc = &hip6->ip6_src;
486 			else
487 				hsrc = NULL;
488 		}
489 		if (hsrc && in6ifa_ifpwithaddr(ifp, hsrc))
490 			src = hsrc;
491 		else {
492 			int error;
493 			struct sockaddr_in6 dst_sa;
494 
495 			bzero(&dst_sa, sizeof(dst_sa));
496 			dst_sa.sin6_family = AF_INET6;
497 			dst_sa.sin6_len = sizeof(dst_sa);
498 			dst_sa.sin6_addr = ip6->ip6_dst;
499 
500 			src = in6_selectsrc(&dst_sa, NULL,
501 			    NULL, &ro, NULL, NULL, &error);
502 			if (src == NULL) {
503 				char ip6buf[INET6_ADDRSTRLEN];
504 				nd6log((LOG_DEBUG,
505 				    "nd6_ns_output: source can't be "
506 				    "determined: dst=%s, error=%d\n",
507 				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
508 				    error));
509 				goto bad;
510 			}
511 		}
512 	} else {
513 		/*
514 		 * Source address for DAD packet must always be IPv6
515 		 * unspecified address. (0::0)
516 		 * We actually don't have to 0-clear the address (we did it
517 		 * above), but we do so here explicitly to make the intention
518 		 * clearer.
519 		 */
520 		bzero(&src_in, sizeof(src_in));
521 		src = &src_in;
522 	}
523 	ip6->ip6_src = *src;
524 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
525 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
526 	nd_ns->nd_ns_code = 0;
527 	nd_ns->nd_ns_reserved = 0;
528 	nd_ns->nd_ns_target = *taddr6;
529 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
530 
531 	/*
532 	 * Add source link-layer address option.
533 	 *
534 	 *				spec		implementation
535 	 *				---		---
536 	 * DAD packet			MUST NOT	do not add the option
537 	 * there's no link layer address:
538 	 *				impossible	do not add the option
539 	 * there's link layer address:
540 	 *	Multicast NS		MUST add one	add the option
541 	 *	Unicast NS		SHOULD add one	add the option
542 	 */
543 	if (!dad && (mac = nd6_ifptomac(ifp))) {
544 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
545 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
546 		/* 8 byte alignments... */
547 		optlen = (optlen + 7) & ~7;
548 
549 		m->m_pkthdr.len += optlen;
550 		m->m_len += optlen;
551 		icmp6len += optlen;
552 		bzero((caddr_t)nd_opt, optlen);
553 		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
554 		nd_opt->nd_opt_len = optlen >> 3;
555 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
556 	}
557 
558 	ip6->ip6_plen = htons((u_short)icmp6len);
559 	nd_ns->nd_ns_cksum = 0;
560 	nd_ns->nd_ns_cksum =
561 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
562 
563 	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
564 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
565 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
566 	V_icmp6stat.icp6s_outhist[ND_NEIGHBOR_SOLICIT]++;
567 
568 	if (ro.ro_rt) {		/* we don't cache this route. */
569 		RTFREE(ro.ro_rt);
570 	}
571 	return;
572 
573   bad:
574 	if (ro.ro_rt) {
575 		RTFREE(ro.ro_rt);
576 	}
577 	m_freem(m);
578 	return;
579 }
580 
581 /*
582  * Neighbor advertisement input handling.
583  *
584  * Based on RFC 2461
585  * Based on RFC 2462 (duplicate address detection)
586  *
587  * the following items are not implemented yet:
588  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
589  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
590  */
591 void
592 nd6_na_input(struct mbuf *m, int off, int icmp6len)
593 {
594 	INIT_VNET_INET6(curvnet);
595 	struct ifnet *ifp = m->m_pkthdr.rcvif;
596 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
597 	struct nd_neighbor_advert *nd_na;
598 	struct in6_addr daddr6 = ip6->ip6_dst;
599 	struct in6_addr taddr6;
600 	int flags;
601 	int is_router;
602 	int is_solicited;
603 	int is_override;
604 	char *lladdr = NULL;
605 	int lladdrlen = 0;
606 	int checklink = 0;
607 	struct ifaddr *ifa;
608 	struct llentry *ln = NULL;
609 	union nd_opts ndopts;
610 	struct mbuf *chain = NULL;
611 	struct sockaddr_in6 sin6;
612 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
613 
614 	if (ip6->ip6_hlim != 255) {
615 		nd6log((LOG_ERR,
616 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
617 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
618 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
619 		goto bad;
620 	}
621 
622 #ifndef PULLDOWN_TEST
623 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
624 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
625 #else
626 	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
627 	if (nd_na == NULL) {
628 		V_icmp6stat.icp6s_tooshort++;
629 		return;
630 	}
631 #endif
632 
633 	flags = nd_na->nd_na_flags_reserved;
634 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
635 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
636 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
637 
638 	taddr6 = nd_na->nd_na_target;
639 	if (in6_setscope(&taddr6, ifp, NULL))
640 		goto bad;	/* XXX: impossible */
641 
642 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
643 		nd6log((LOG_ERR,
644 		    "nd6_na_input: invalid target address %s\n",
645 		    ip6_sprintf(ip6bufs, &taddr6)));
646 		goto bad;
647 	}
648 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
649 		if (is_solicited) {
650 			nd6log((LOG_ERR,
651 			    "nd6_na_input: a solicited adv is multicasted\n"));
652 			goto bad;
653 		}
654 
655 	icmp6len -= sizeof(*nd_na);
656 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
657 	if (nd6_options(&ndopts) < 0) {
658 		nd6log((LOG_INFO,
659 		    "nd6_na_input: invalid ND option, ignored\n"));
660 		/* nd6_options have incremented stats */
661 		goto freeit;
662 	}
663 
664 	if (ndopts.nd_opts_tgt_lladdr) {
665 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
666 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
667 	}
668 
669 	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
670 
671 	/*
672 	 * Target address matches one of my interface address.
673 	 *
674 	 * If my address is tentative, this means that there's somebody
675 	 * already using the same address as mine.  This indicates DAD failure.
676 	 * This is defined in RFC 2462.
677 	 *
678 	 * Otherwise, process as defined in RFC 2461.
679 	 */
680 	if (ifa
681 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
682 		nd6_dad_na_input(ifa);
683 		goto freeit;
684 	}
685 
686 	/* Just for safety, maybe unnecessary. */
687 	if (ifa) {
688 		log(LOG_ERR,
689 		    "nd6_na_input: duplicate IP6 address %s\n",
690 		    ip6_sprintf(ip6bufs, &taddr6));
691 		goto freeit;
692 	}
693 
694 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
695 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
696 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
697 		    ifp->if_addrlen, lladdrlen - 2));
698 		goto bad;
699 	}
700 
701 	/*
702 	 * If no neighbor cache entry is found, NA SHOULD silently be
703 	 * discarded.
704 	 */
705 	IF_AFDATA_LOCK(ifp);
706 	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
707 	IF_AFDATA_UNLOCK(ifp);
708 	if (ln == NULL) {
709 		goto freeit;
710 	}
711 
712 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
713 		/*
714 		 * If the link-layer has address, and no lladdr option came,
715 		 * discard the packet.
716 		 */
717 		if (ifp->if_addrlen && lladdr == NULL) {
718 			goto freeit;
719 		}
720 
721 		/*
722 		 * Record link-layer address, and update the state.
723 		 */
724 		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
725 		ln->la_flags |= LLE_VALID;
726 		if (is_solicited) {
727 			ln->ln_state = ND6_LLINFO_REACHABLE;
728 			ln->ln_byhint = 0;
729 			if (!ND6_LLINFO_PERMANENT(ln)) {
730 				nd6_llinfo_settimer_locked(ln,
731 				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
732 			}
733 		} else {
734 			ln->ln_state = ND6_LLINFO_STALE;
735 			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
736 		}
737 		if ((ln->ln_router = is_router) != 0) {
738 			/*
739 			 * This means a router's state has changed from
740 			 * non-reachable to probably reachable, and might
741 			 * affect the status of associated prefixes..
742 			 */
743 			checklink = 1;
744 		}
745 	} else {
746 		int llchange;
747 
748 		/*
749 		 * Check if the link-layer address has changed or not.
750 		 */
751 		if (lladdr == NULL)
752 			llchange = 0;
753 		else {
754 			if (ln->la_flags & LLE_VALID) {
755 				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
756 					llchange = 1;
757 				else
758 					llchange = 0;
759 			} else
760 				llchange = 1;
761 		}
762 
763 		/*
764 		 * This is VERY complex.  Look at it with care.
765 		 *
766 		 * override solicit lladdr llchange	action
767 		 *					(L: record lladdr)
768 		 *
769 		 *	0	0	n	--	(2c)
770 		 *	0	0	y	n	(2b) L
771 		 *	0	0	y	y	(1)    REACHABLE->STALE
772 		 *	0	1	n	--	(2c)   *->REACHABLE
773 		 *	0	1	y	n	(2b) L *->REACHABLE
774 		 *	0	1	y	y	(1)    REACHABLE->STALE
775 		 *	1	0	n	--	(2a)
776 		 *	1	0	y	n	(2a) L
777 		 *	1	0	y	y	(2a) L *->STALE
778 		 *	1	1	n	--	(2a)   *->REACHABLE
779 		 *	1	1	y	n	(2a) L *->REACHABLE
780 		 *	1	1	y	y	(2a) L *->REACHABLE
781 		 */
782 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
783 			/*
784 			 * If state is REACHABLE, make it STALE.
785 			 * no other updates should be done.
786 			 */
787 			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
788 				ln->ln_state = ND6_LLINFO_STALE;
789 				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
790 			}
791 			goto freeit;
792 		} else if (is_override				   /* (2a) */
793 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
794 			|| lladdr == NULL) {			   /* (2c) */
795 			/*
796 			 * Update link-local address, if any.
797 			 */
798 			if (lladdr != NULL) {
799 				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
800 				ln->la_flags |= LLE_VALID;
801 			}
802 
803 			/*
804 			 * If solicited, make the state REACHABLE.
805 			 * If not solicited and the link-layer address was
806 			 * changed, make it STALE.
807 			 */
808 			if (is_solicited) {
809 				ln->ln_state = ND6_LLINFO_REACHABLE;
810 				ln->ln_byhint = 0;
811 				if (!ND6_LLINFO_PERMANENT(ln)) {
812 					nd6_llinfo_settimer_locked(ln,
813 					    (long)ND_IFINFO(ifp)->reachable * hz);
814 				}
815 			} else {
816 				if (lladdr != NULL && llchange) {
817 					ln->ln_state = ND6_LLINFO_STALE;
818 					nd6_llinfo_settimer_locked(ln,
819 					    (long)V_nd6_gctimer * hz);
820 				}
821 			}
822 		}
823 
824 		if (ln->ln_router && !is_router) {
825 			/*
826 			 * The peer dropped the router flag.
827 			 * Remove the sender from the Default Router List and
828 			 * update the Destination Cache entries.
829 			 */
830 			struct nd_defrouter *dr;
831 			struct in6_addr *in6;
832 
833 			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
834 
835 			/*
836 			 * Lock to protect the default router list.
837 			 * XXX: this might be unnecessary, since this function
838 			 * is only called under the network software interrupt
839 			 * context.  However, we keep it just for safety.
840 			 */
841 			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
842 			if (dr)
843 				defrtrlist_del(dr);
844 			else if (!V_ip6_forwarding) {
845 				/*
846 				 * Even if the neighbor is not in the default
847 				 * router list, the neighbor may be used
848 				 * as a next hop for some destinations
849 				 * (e.g. redirect case). So we must
850 				 * call rt6_flush explicitly.
851 				 */
852 				rt6_flush(&ip6->ip6_src, ifp);
853 			}
854 		}
855 		ln->ln_router = is_router;
856 	}
857         /* XXX - QL
858 	 *  Does this matter?
859 	 *  rt->rt_flags &= ~RTF_REJECT;
860 	 */
861 	ln->la_asked = 0;
862 	if (ln->la_hold) {
863 		struct mbuf *m_hold, *m_hold_next;
864 
865 		/*
866 		 * reset the la_hold in advance, to explicitly
867 		 * prevent a la_hold lookup in nd6_output()
868 		 * (wouldn't happen, though...)
869 		 */
870 		for (m_hold = ln->la_hold, ln->la_hold = NULL;
871 		    m_hold; m_hold = m_hold_next) {
872 			m_hold_next = m_hold->m_nextpkt;
873 			m_hold->m_nextpkt = NULL;
874 			/*
875 			 * we assume ifp is not a loopback here, so just set
876 			 * the 2nd argument as the 1st one.
877 			 */
878 			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
879 		}
880 	}
881  freeit:
882 	if (ln != NULL) {
883 		if (chain)
884 			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
885 		LLE_WUNLOCK(ln);
886 
887 		if (chain)
888 			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
889 	}
890 	if (checklink)
891 		pfxlist_onlink_check();
892 
893 	m_freem(m);
894 	return;
895 
896  bad:
897 	if (ln != NULL)
898 		LLE_WUNLOCK(ln);
899 
900 	V_icmp6stat.icp6s_badna++;
901 	m_freem(m);
902 }
903 
904 /*
905  * Neighbor advertisement output handling.
906  *
907  * Based on RFC 2461
908  *
909  * the following items are not implemented yet:
910  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
911  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
912  *
913  * tlladdr - 1 if include target link-layer address
914  * sdl0 - sockaddr_dl (= proxy NA) or NULL
915  */
916 void
917 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
918     const struct in6_addr *taddr6, u_long flags, int tlladdr,
919     struct sockaddr *sdl0)
920 {
921 	INIT_VNET_INET6(ifp->if_vnet);
922 	struct mbuf *m;
923 	struct ip6_hdr *ip6;
924 	struct nd_neighbor_advert *nd_na;
925 	struct ip6_moptions im6o;
926 	struct in6_addr *src, daddr6;
927 	struct sockaddr_in6 dst_sa;
928 	int icmp6len, maxlen, error;
929 	caddr_t mac = NULL;
930 	struct route_in6 ro;
931 
932 	bzero(&ro, sizeof(ro));
933 
934 	daddr6 = *daddr6_0;	/* make a local copy for modification */
935 
936 	/* estimate the size of message */
937 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
938 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
939 	if (max_linkhdr + maxlen >= MCLBYTES) {
940 #ifdef DIAGNOSTIC
941 		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
942 		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
943 #endif
944 		return;
945 	}
946 
947 	MGETHDR(m, M_DONTWAIT, MT_DATA);
948 	if (m && max_linkhdr + maxlen >= MHLEN) {
949 		MCLGET(m, M_DONTWAIT);
950 		if ((m->m_flags & M_EXT) == 0) {
951 			m_free(m);
952 			m = NULL;
953 		}
954 	}
955 	if (m == NULL)
956 		return;
957 	m->m_pkthdr.rcvif = NULL;
958 
959 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
960 		m->m_flags |= M_MCAST;
961 		im6o.im6o_multicast_ifp = ifp;
962 		im6o.im6o_multicast_hlim = 255;
963 		im6o.im6o_multicast_loop = 0;
964 	}
965 
966 	icmp6len = sizeof(*nd_na);
967 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
968 	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
969 
970 	/* fill neighbor advertisement packet */
971 	ip6 = mtod(m, struct ip6_hdr *);
972 	ip6->ip6_flow = 0;
973 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
974 	ip6->ip6_vfc |= IPV6_VERSION;
975 	ip6->ip6_nxt = IPPROTO_ICMPV6;
976 	ip6->ip6_hlim = 255;
977 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
978 		/* reply to DAD */
979 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
980 		daddr6.s6_addr16[1] = 0;
981 		daddr6.s6_addr32[1] = 0;
982 		daddr6.s6_addr32[2] = 0;
983 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
984 		if (in6_setscope(&daddr6, ifp, NULL))
985 			goto bad;
986 
987 		flags &= ~ND_NA_FLAG_SOLICITED;
988 	}
989 	ip6->ip6_dst = daddr6;
990 	bzero(&dst_sa, sizeof(struct sockaddr_in6));
991 	dst_sa.sin6_family = AF_INET6;
992 	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
993 	dst_sa.sin6_addr = daddr6;
994 
995 	/*
996 	 * Select a source whose scope is the same as that of the dest.
997 	 */
998 	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
999 	src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &error);
1000 	if (src == NULL) {
1001 		char ip6buf[INET6_ADDRSTRLEN];
1002 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1003 		    "determined: dst=%s, error=%d\n",
1004 		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1005 		goto bad;
1006 	}
1007 	ip6->ip6_src = *src;
1008 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1009 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1010 	nd_na->nd_na_code = 0;
1011 	nd_na->nd_na_target = *taddr6;
1012 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1013 
1014 	/*
1015 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1016 	 * see nd6_ns_input() for details.
1017 	 * Basically, if NS packet is sent to unicast/anycast addr,
1018 	 * target lladdr option SHOULD NOT be included.
1019 	 */
1020 	if (tlladdr) {
1021 		/*
1022 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1023 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1024 		 * my address) use lladdr configured for the interface.
1025 		 */
1026 		if (sdl0 == NULL) {
1027 #ifdef DEV_CARP
1028 			if (ifp->if_carp)
1029 				mac = carp_macmatch6(ifp->if_carp, m, taddr6);
1030 			if (mac == NULL)
1031 				mac = nd6_ifptomac(ifp);
1032 #else
1033 			mac = nd6_ifptomac(ifp);
1034 #endif
1035 		} else if (sdl0->sa_family == AF_LINK) {
1036 			struct sockaddr_dl *sdl;
1037 			sdl = (struct sockaddr_dl *)sdl0;
1038 			if (sdl->sdl_alen == ifp->if_addrlen)
1039 				mac = LLADDR(sdl);
1040 		}
1041 	}
1042 	if (tlladdr && mac) {
1043 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1044 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1045 
1046 		/* roundup to 8 bytes alignment! */
1047 		optlen = (optlen + 7) & ~7;
1048 
1049 		m->m_pkthdr.len += optlen;
1050 		m->m_len += optlen;
1051 		icmp6len += optlen;
1052 		bzero((caddr_t)nd_opt, optlen);
1053 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1054 		nd_opt->nd_opt_len = optlen >> 3;
1055 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1056 	} else
1057 		flags &= ~ND_NA_FLAG_OVERRIDE;
1058 
1059 	ip6->ip6_plen = htons((u_short)icmp6len);
1060 	nd_na->nd_na_flags_reserved = flags;
1061 	nd_na->nd_na_cksum = 0;
1062 	nd_na->nd_na_cksum =
1063 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1064 
1065 	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1066 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1067 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1068 	V_icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1069 
1070 	if (ro.ro_rt) {		/* we don't cache this route. */
1071 		RTFREE(ro.ro_rt);
1072 	}
1073 	return;
1074 
1075   bad:
1076 	if (ro.ro_rt) {
1077 		RTFREE(ro.ro_rt);
1078 	}
1079 	m_freem(m);
1080 	return;
1081 }
1082 
1083 caddr_t
1084 nd6_ifptomac(struct ifnet *ifp)
1085 {
1086 	switch (ifp->if_type) {
1087 	case IFT_ARCNET:
1088 	case IFT_ETHER:
1089 	case IFT_FDDI:
1090 	case IFT_IEEE1394:
1091 #ifdef IFT_L2VLAN
1092 	case IFT_L2VLAN:
1093 #endif
1094 #ifdef IFT_IEEE80211
1095 	case IFT_IEEE80211:
1096 #endif
1097 #ifdef IFT_CARP
1098 	case IFT_CARP:
1099 #endif
1100 	case IFT_BRIDGE:
1101 	case IFT_ISO88025:
1102 		return IF_LLADDR(ifp);
1103 	default:
1104 		return NULL;
1105 	}
1106 }
1107 
1108 TAILQ_HEAD(dadq_head, dadq);
1109 struct dadq {
1110 	TAILQ_ENTRY(dadq) dad_list;
1111 	struct ifaddr *dad_ifa;
1112 	int dad_count;		/* max NS to send */
1113 	int dad_ns_tcount;	/* # of trials to send NS */
1114 	int dad_ns_ocount;	/* NS sent so far */
1115 	int dad_ns_icount;
1116 	int dad_na_icount;
1117 	struct callout dad_timer_ch;
1118 };
1119 
1120 #ifdef VIMAGE_GLOBALS
1121 static struct dadq_head dadq;
1122 int dad_init;
1123 #endif
1124 
1125 static struct dadq *
1126 nd6_dad_find(struct ifaddr *ifa)
1127 {
1128 	INIT_VNET_INET6(curvnet);
1129 	struct dadq *dp;
1130 
1131 	for (dp = V_dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1132 		if (dp->dad_ifa == ifa)
1133 			return dp;
1134 	}
1135 	return NULL;
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, (void *)dp->dad_ifa);
1144 }
1145 
1146 static void
1147 nd6_dad_stoptimer(struct dadq *dp)
1148 {
1149 
1150 	callout_stop(&dp->dad_timer_ch);
1151 }
1152 
1153 /*
1154  * Start Duplicate Address Detection (DAD) for specified interface address.
1155  */
1156 void
1157 nd6_dad_start(struct ifaddr *ifa, int delay)
1158 {
1159 	INIT_VNET_INET6(curvnet);
1160 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1161 	struct dadq *dp;
1162 	char ip6buf[INET6_ADDRSTRLEN];
1163 
1164 	if (!V_dad_init) {
1165 		TAILQ_INIT(&V_dadq);
1166 		V_dad_init++;
1167 	}
1168 
1169 	/*
1170 	 * If we don't need DAD, don't do it.
1171 	 * There are several cases:
1172 	 * - DAD is disabled (ip6_dad_count == 0)
1173 	 * - the interface address is anycast
1174 	 */
1175 	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1176 		log(LOG_DEBUG,
1177 			"nd6_dad_start: called with non-tentative address "
1178 			"%s(%s)\n",
1179 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1180 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1181 		return;
1182 	}
1183 	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1184 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1185 		return;
1186 	}
1187 	if (!V_ip6_dad_count) {
1188 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1189 		return;
1190 	}
1191 	if (ifa->ifa_ifp == NULL)
1192 		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1193 	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1194 		return;
1195 	}
1196 	if (nd6_dad_find(ifa) != NULL) {
1197 		/* DAD already in progress */
1198 		return;
1199 	}
1200 
1201 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1202 	if (dp == NULL) {
1203 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1204 			"%s(%s)\n",
1205 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1206 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1207 		return;
1208 	}
1209 	bzero(dp, sizeof(*dp));
1210 	callout_init(&dp->dad_timer_ch, 0);
1211 	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1212 
1213 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1214 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1215 
1216 	/*
1217 	 * Send NS packet for DAD, ip6_dad_count times.
1218 	 * Note that we must delay the first transmission, if this is the
1219 	 * first packet to be sent from the interface after interface
1220 	 * (re)initialization.
1221 	 */
1222 	dp->dad_ifa = ifa;
1223 	IFAREF(ifa);	/* just for safety */
1224 	dp->dad_count = V_ip6_dad_count;
1225 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1226 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1227 	if (delay == 0) {
1228 		nd6_dad_ns_output(dp, ifa);
1229 		nd6_dad_starttimer(dp,
1230 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1231 	} else {
1232 		nd6_dad_starttimer(dp, delay);
1233 	}
1234 }
1235 
1236 /*
1237  * terminate DAD unconditionally.  used for address removals.
1238  */
1239 void
1240 nd6_dad_stop(struct ifaddr *ifa)
1241 {
1242 	INIT_VNET_INET6(curvnet);
1243 	struct dadq *dp;
1244 
1245 	if (!V_dad_init)
1246 		return;
1247 	dp = nd6_dad_find(ifa);
1248 	if (!dp) {
1249 		/* DAD wasn't started yet */
1250 		return;
1251 	}
1252 
1253 	nd6_dad_stoptimer(dp);
1254 
1255 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1256 	free(dp, M_IP6NDP);
1257 	dp = NULL;
1258 	IFAFREE(ifa);
1259 }
1260 
1261 static void
1262 nd6_dad_timer(struct ifaddr *ifa)
1263 {
1264 	CURVNET_SET(dp->dad_vnet);
1265 	INIT_VNET_INET6(curvnet);
1266 	int s;
1267 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1268 	struct dadq *dp;
1269 	char ip6buf[INET6_ADDRSTRLEN];
1270 
1271 	s = splnet();		/* XXX */
1272 
1273 	/* Sanity check */
1274 	if (ia == NULL) {
1275 		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1276 		goto done;
1277 	}
1278 	dp = nd6_dad_find(ifa);
1279 	if (dp == NULL) {
1280 		log(LOG_ERR, "nd6_dad_timer: DAD structure not found\n");
1281 		goto done;
1282 	}
1283 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1284 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1285 			"%s(%s)\n",
1286 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1287 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1288 		goto done;
1289 	}
1290 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1291 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1292 			"%s(%s)\n",
1293 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1294 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1295 		goto done;
1296 	}
1297 
1298 	/* timeouted with IFF_{RUNNING,UP} check */
1299 	if (dp->dad_ns_tcount > V_dad_maxtry) {
1300 		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1301 		    if_name(ifa->ifa_ifp)));
1302 
1303 		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1304 		free(dp, M_IP6NDP);
1305 		dp = NULL;
1306 		IFAFREE(ifa);
1307 		goto done;
1308 	}
1309 
1310 	/* Need more checks? */
1311 	if (dp->dad_ns_ocount < dp->dad_count) {
1312 		/*
1313 		 * We have more NS to go.  Send NS packet for DAD.
1314 		 */
1315 		nd6_dad_ns_output(dp, ifa);
1316 		nd6_dad_starttimer(dp,
1317 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1318 	} else {
1319 		/*
1320 		 * We have transmitted sufficient number of DAD packets.
1321 		 * See what we've got.
1322 		 */
1323 		int duplicate;
1324 
1325 		duplicate = 0;
1326 
1327 		if (dp->dad_na_icount) {
1328 			/*
1329 			 * the check is in nd6_dad_na_input(),
1330 			 * but just in case
1331 			 */
1332 			duplicate++;
1333 		}
1334 
1335 		if (dp->dad_ns_icount) {
1336 			/* We've seen NS, means DAD has failed. */
1337 			duplicate++;
1338 		}
1339 
1340 		if (duplicate) {
1341 			/* (*dp) will be freed in nd6_dad_duplicated() */
1342 			dp = NULL;
1343 			nd6_dad_duplicated(ifa);
1344 		} else {
1345 			/*
1346 			 * We are done with DAD.  No NA came, no NS came.
1347 			 * No duplicate address found.
1348 			 */
1349 			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1350 
1351 			nd6log((LOG_DEBUG,
1352 			    "%s: DAD complete for %s - no duplicates found\n",
1353 			    if_name(ifa->ifa_ifp),
1354 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1355 
1356 			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1357 			free(dp, M_IP6NDP);
1358 			dp = NULL;
1359 			IFAFREE(ifa);
1360 		}
1361 	}
1362 
1363 done:
1364 	splx(s);
1365 	CURVNET_RESTORE();
1366 }
1367 
1368 void
1369 nd6_dad_duplicated(struct ifaddr *ifa)
1370 {
1371 	INIT_VNET_INET6(curvnet);
1372 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1373 	struct ifnet *ifp;
1374 	struct dadq *dp;
1375 	char ip6buf[INET6_ADDRSTRLEN];
1376 
1377 	dp = nd6_dad_find(ifa);
1378 	if (dp == NULL) {
1379 		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1380 		return;
1381 	}
1382 
1383 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1384 	    "NS in/out=%d/%d, NA in=%d\n",
1385 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1386 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1387 
1388 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1389 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1390 
1391 	/* We are done with DAD, with duplicate address found. (failure) */
1392 	nd6_dad_stoptimer(dp);
1393 
1394 	ifp = ifa->ifa_ifp;
1395 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1396 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1397 	log(LOG_ERR, "%s: manual intervention required\n",
1398 	    if_name(ifp));
1399 
1400 	/*
1401 	 * If the address is a link-local address formed from an interface
1402 	 * identifier based on the hardware address which is supposed to be
1403 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1404 	 * operation on the interface SHOULD be disabled.
1405 	 * [rfc2462bis-03 Section 5.4.5]
1406 	 */
1407 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1408 		struct in6_addr in6;
1409 
1410 		/*
1411 		 * To avoid over-reaction, we only apply this logic when we are
1412 		 * very sure that hardware addresses are supposed to be unique.
1413 		 */
1414 		switch (ifp->if_type) {
1415 		case IFT_ETHER:
1416 		case IFT_FDDI:
1417 		case IFT_ATM:
1418 		case IFT_IEEE1394:
1419 #ifdef IFT_IEEE80211
1420 		case IFT_IEEE80211:
1421 #endif
1422 			in6 = ia->ia_addr.sin6_addr;
1423 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1424 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1425 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1426 				log(LOG_ERR, "%s: possible hardware address "
1427 				    "duplication detected, disable IPv6\n",
1428 				    if_name(ifp));
1429 			}
1430 			break;
1431 		}
1432 	}
1433 
1434 	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1435 	free(dp, M_IP6NDP);
1436 	dp = NULL;
1437 	IFAFREE(ifa);
1438 }
1439 
1440 static void
1441 nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1442 {
1443 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1444 	struct ifnet *ifp = ifa->ifa_ifp;
1445 
1446 	dp->dad_ns_tcount++;
1447 	if ((ifp->if_flags & IFF_UP) == 0) {
1448 		return;
1449 	}
1450 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1451 		return;
1452 	}
1453 
1454 	dp->dad_ns_ocount++;
1455 	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1456 }
1457 
1458 static void
1459 nd6_dad_ns_input(struct ifaddr *ifa)
1460 {
1461 	INIT_VNET_INET6(curvnet);
1462 	struct in6_ifaddr *ia;
1463 	struct ifnet *ifp;
1464 	const struct in6_addr *taddr6;
1465 	struct dadq *dp;
1466 	int duplicate;
1467 
1468 	if (ifa == NULL)
1469 		panic("ifa == NULL in nd6_dad_ns_input");
1470 
1471 	ia = (struct in6_ifaddr *)ifa;
1472 	ifp = ifa->ifa_ifp;
1473 	taddr6 = &ia->ia_addr.sin6_addr;
1474 	duplicate = 0;
1475 	dp = nd6_dad_find(ifa);
1476 
1477 	/* Quickhack - completely ignore DAD NS packets */
1478 	if (V_dad_ignore_ns) {
1479 		char ip6buf[INET6_ADDRSTRLEN];
1480 		nd6log((LOG_INFO,
1481 		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1482 		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1483 		    if_name(ifa->ifa_ifp)));
1484 		return;
1485 	}
1486 
1487 	/*
1488 	 * if I'm yet to start DAD, someone else started using this address
1489 	 * first.  I have a duplicate and you win.
1490 	 */
1491 	if (dp == NULL || dp->dad_ns_ocount == 0)
1492 		duplicate++;
1493 
1494 	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1495 
1496 	if (duplicate) {
1497 		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1498 		nd6_dad_duplicated(ifa);
1499 	} else {
1500 		/*
1501 		 * not sure if I got a duplicate.
1502 		 * increment ns count and see what happens.
1503 		 */
1504 		if (dp)
1505 			dp->dad_ns_icount++;
1506 	}
1507 }
1508 
1509 static void
1510 nd6_dad_na_input(struct ifaddr *ifa)
1511 {
1512 	struct dadq *dp;
1513 
1514 	if (ifa == NULL)
1515 		panic("ifa == NULL in nd6_dad_na_input");
1516 
1517 	dp = nd6_dad_find(ifa);
1518 	if (dp)
1519 		dp->dad_na_icount++;
1520 
1521 	/* remove the address. */
1522 	nd6_dad_duplicated(ifa);
1523 }
1524