nd6_nbr.c revision 216022
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: head/sys/netinet6/nd6_nbr.c 216022 2010-11-29 00:04:08Z bz $");
34
35#include "opt_inet.h"
36#include "opt_inet6.h"
37#include "opt_ipsec.h"
38#include "opt_mpath.h"
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/malloc.h>
43#include <sys/lock.h>
44#include <sys/rwlock.h>
45#include <sys/mbuf.h>
46#include <sys/socket.h>
47#include <sys/sockio.h>
48#include <sys/time.h>
49#include <sys/kernel.h>
50#include <sys/errno.h>
51#include <sys/syslog.h>
52#include <sys/queue.h>
53#include <sys/callout.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 <net/if_llatbl.h>
67#define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
68#include <netinet6/in6_var.h>
69#include <netinet6/in6_ifattach.h>
70#include <netinet/ip6.h>
71#include <netinet6/ip6_var.h>
72#include <netinet6/scope6_var.h>
73#include <netinet6/nd6.h>
74#include <netinet/icmp6.h>
75#include <netinet/ip_carp.h>
76#include <netinet6/send.h>
77
78#define SDL(s) ((struct sockaddr_dl *)s)
79
80struct dadq;
81static struct dadq *nd6_dad_find(struct ifaddr *);
82static void nd6_dad_starttimer(struct dadq *, int);
83static void nd6_dad_stoptimer(struct dadq *);
84static void nd6_dad_timer(struct dadq *);
85static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
86static void nd6_dad_ns_input(struct ifaddr *);
87static void nd6_dad_na_input(struct ifaddr *);
88
89VNET_DEFINE(int, dad_ignore_ns) = 0;	/* ignore NS in DAD - specwise incorrect*/
90VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to transmit DAD packet */
91#define	V_dad_ignore_ns			VNET(dad_ignore_ns)
92#define	V_dad_maxtry			VNET(dad_maxtry)
93
94/*
95 * Input a Neighbor Solicitation Message.
96 *
97 * Based on RFC 2461
98 * Based on RFC 2462 (duplicate address detection)
99 */
100void
101nd6_ns_input(struct mbuf *m, int off, int icmp6len)
102{
103	struct ifnet *ifp = m->m_pkthdr.rcvif;
104	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
105	struct nd_neighbor_solicit *nd_ns;
106	struct in6_addr saddr6 = ip6->ip6_src;
107	struct in6_addr daddr6 = ip6->ip6_dst;
108	struct in6_addr taddr6;
109	struct in6_addr myaddr6;
110	char *lladdr = NULL;
111	struct ifaddr *ifa = NULL;
112	int lladdrlen = 0;
113	int anycast = 0, proxy = 0, tentative = 0;
114	int tlladdr;
115	union nd_opts ndopts;
116	struct sockaddr_dl *proxydl = NULL;
117	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
118
119#ifndef PULLDOWN_TEST
120	IP6_EXTHDR_CHECK(m, off, icmp6len,);
121	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
122#else
123	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
124	if (nd_ns == NULL) {
125		ICMP6STAT_INC(icp6s_tooshort);
126		return;
127	}
128#endif
129	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
130	taddr6 = nd_ns->nd_ns_target;
131	if (in6_setscope(&taddr6, ifp, NULL) != 0)
132		goto bad;
133
134	if (ip6->ip6_hlim != 255) {
135		nd6log((LOG_ERR,
136		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
137		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
138		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
139		goto bad;
140	}
141
142	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
143		/* dst has to be a solicited node multicast address. */
144		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
145		    /* don't check ifindex portion */
146		    daddr6.s6_addr32[1] == 0 &&
147		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
148		    daddr6.s6_addr8[12] == 0xff) {
149			; /* good */
150		} else {
151			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
152			    "(wrong ip6 dst)\n"));
153			goto bad;
154		}
155	} else if (!V_nd6_onlink_ns_rfc4861) {
156		struct sockaddr_in6 src_sa6;
157
158		/*
159		 * According to recent IETF discussions, it is not a good idea
160		 * to accept a NS from an address which would not be deemed
161		 * to be a neighbor otherwise.  This point is expected to be
162		 * clarified in future revisions of the specification.
163		 */
164		bzero(&src_sa6, sizeof(src_sa6));
165		src_sa6.sin6_family = AF_INET6;
166		src_sa6.sin6_len = sizeof(src_sa6);
167		src_sa6.sin6_addr = saddr6;
168		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
169			nd6log((LOG_INFO, "nd6_ns_input: "
170				"NS packet from non-neighbor\n"));
171			goto bad;
172		}
173	}
174
175	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
176		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
177		goto bad;
178	}
179
180	icmp6len -= sizeof(*nd_ns);
181	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
182	if (nd6_options(&ndopts) < 0) {
183		nd6log((LOG_INFO,
184		    "nd6_ns_input: invalid ND option, ignored\n"));
185		/* nd6_options have incremented stats */
186		goto freeit;
187	}
188
189	if (ndopts.nd_opts_src_lladdr) {
190		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
191		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
192	}
193
194	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
195		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
196		    "(link-layer address option)\n"));
197		goto bad;
198	}
199
200	/*
201	 * Attaching target link-layer address to the NA?
202	 * (RFC 2461 7.2.4)
203	 *
204	 * NS IP dst is unicast/anycast			MUST NOT add
205	 * NS IP dst is solicited-node multicast	MUST add
206	 *
207	 * In implementation, we add target link-layer address by default.
208	 * We do not add one in MUST NOT cases.
209	 */
210	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
211		tlladdr = 0;
212	else
213		tlladdr = 1;
214
215	/*
216	 * Target address (taddr6) must be either:
217	 * (1) Valid unicast/anycast address for my receiving interface,
218	 * (2) Unicast address for which I'm offering proxy service, or
219	 * (3) "tentative" address on which DAD is being performed.
220	 */
221	/* (1) and (3) check. */
222	if (ifp->if_carp)
223		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
224	if (ifa == NULL)
225		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
226
227	/* (2) check. */
228	if (ifa == NULL) {
229		struct rtentry *rt;
230		struct sockaddr_in6 tsin6;
231		int need_proxy;
232#ifdef RADIX_MPATH
233		struct route_in6 ro;
234#endif
235
236		bzero(&tsin6, sizeof tsin6);
237		tsin6.sin6_len = sizeof(struct sockaddr_in6);
238		tsin6.sin6_family = AF_INET6;
239		tsin6.sin6_addr = taddr6;
240
241#ifdef RADIX_MPATH
242		bzero(&ro, sizeof(ro));
243		ro.ro_dst = tsin6;
244		rtalloc_mpath((struct route *)&ro, RTF_ANNOUNCE);
245		rt = ro.ro_rt;
246#else
247		rt = rtalloc1((struct sockaddr *)&tsin6, 0, 0);
248#endif
249		need_proxy = (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
250		    rt->rt_gateway->sa_family == AF_LINK);
251		if (rt)
252			RTFREE_LOCKED(rt);
253		if (need_proxy) {
254			/*
255			 * proxy NDP for single entry
256			 */
257			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
258				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
259			if (ifa) {
260				proxy = 1;
261				proxydl = SDL(rt->rt_gateway);
262			}
263		}
264	}
265	if (ifa == NULL) {
266		/*
267		 * We've got an NS packet, and we don't have that adddress
268		 * assigned for us.  We MUST silently ignore it.
269		 * See RFC2461 7.2.3.
270		 */
271		goto freeit;
272	}
273	myaddr6 = *IFA_IN6(ifa);
274	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
275	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
276	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
277		goto freeit;
278
279	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
280		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
281		    "(if %d, NS packet %d)\n",
282		    ip6_sprintf(ip6bufs, &taddr6),
283		    ifp->if_addrlen, lladdrlen - 2));
284		goto bad;
285	}
286
287	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
288		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
289		    ip6_sprintf(ip6bufs, &saddr6)));
290		goto freeit;
291	}
292
293	/*
294	 * We have neighbor solicitation packet, with target address equals to
295	 * one of my tentative address.
296	 *
297	 * src addr	how to process?
298	 * ---		---
299	 * multicast	of course, invalid (rejected in ip6_input)
300	 * unicast	somebody is doing address resolution -> ignore
301	 * unspec	dup address detection
302	 *
303	 * The processing is defined in RFC 2462.
304	 */
305	if (tentative) {
306		/*
307		 * If source address is unspecified address, it is for
308		 * duplicate address detection.
309		 *
310		 * If not, the packet is for addess resolution;
311		 * silently ignore it.
312		 */
313		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
314			nd6_dad_ns_input(ifa);
315
316		goto freeit;
317	}
318
319	/*
320	 * If the source address is unspecified address, entries must not
321	 * be created or updated.
322	 * It looks that sender is performing DAD.  Output NA toward
323	 * all-node multicast address, to tell the sender that I'm using
324	 * the address.
325	 * S bit ("solicited") must be zero.
326	 */
327	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
328		struct in6_addr in6_all;
329
330		in6_all = in6addr_linklocal_allnodes;
331		if (in6_setscope(&in6_all, ifp, NULL) != 0)
332			goto bad;
333		nd6_na_output(ifp, &in6_all, &taddr6,
334		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
335		    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
336		    tlladdr, (struct sockaddr *)proxydl);
337		goto freeit;
338	}
339
340	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
341	    ND_NEIGHBOR_SOLICIT, 0);
342
343	nd6_na_output(ifp, &saddr6, &taddr6,
344	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
345	    (V_ip6_forwarding ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
346	    tlladdr, (struct sockaddr *)proxydl);
347 freeit:
348	if (ifa != NULL)
349		ifa_free(ifa);
350	m_freem(m);
351	return;
352
353 bad:
354	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
355		ip6_sprintf(ip6bufs, &saddr6)));
356	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
357		ip6_sprintf(ip6bufs, &daddr6)));
358	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
359		ip6_sprintf(ip6bufs, &taddr6)));
360	ICMP6STAT_INC(icp6s_badns);
361	if (ifa != NULL)
362		ifa_free(ifa);
363	m_freem(m);
364}
365
366/*
367 * Output a Neighbor Solicitation Message. Caller specifies:
368 *	- ICMP6 header source IP6 address
369 *	- ND6 header target IP6 address
370 *	- ND6 header source datalink address
371 *
372 * Based on RFC 2461
373 * Based on RFC 2462 (duplicate address detection)
374 *
375 *   ln - for source address determination
376 *  dad - duplicate address detection
377 */
378void
379nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
380    const struct in6_addr *taddr6, struct llentry *ln, int dad)
381{
382	struct mbuf *m;
383	struct m_tag *mtag;
384	struct ip6_hdr *ip6;
385	struct nd_neighbor_solicit *nd_ns;
386	struct ip6_moptions im6o;
387	int icmp6len;
388	int maxlen;
389	caddr_t mac;
390	struct route_in6 ro;
391
392	if (IN6_IS_ADDR_MULTICAST(taddr6))
393		return;
394
395	/* estimate the size of message */
396	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
397	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
398	if (max_linkhdr + maxlen >= MCLBYTES) {
399#ifdef DIAGNOSTIC
400		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
401		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
402#endif
403		return;
404	}
405
406	MGETHDR(m, M_DONTWAIT, MT_DATA);
407	if (m && max_linkhdr + maxlen >= MHLEN) {
408		MCLGET(m, M_DONTWAIT);
409		if ((m->m_flags & M_EXT) == 0) {
410			m_free(m);
411			m = NULL;
412		}
413	}
414	if (m == NULL)
415		return;
416	m->m_pkthdr.rcvif = NULL;
417
418	bzero(&ro, sizeof(ro));
419
420	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
421		m->m_flags |= M_MCAST;
422		im6o.im6o_multicast_ifp = ifp;
423		im6o.im6o_multicast_hlim = 255;
424		im6o.im6o_multicast_loop = 0;
425	}
426
427	icmp6len = sizeof(*nd_ns);
428	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
429	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
430
431	/* fill neighbor solicitation packet */
432	ip6 = mtod(m, struct ip6_hdr *);
433	ip6->ip6_flow = 0;
434	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
435	ip6->ip6_vfc |= IPV6_VERSION;
436	/* ip6->ip6_plen will be set later */
437	ip6->ip6_nxt = IPPROTO_ICMPV6;
438	ip6->ip6_hlim = 255;
439	if (daddr6)
440		ip6->ip6_dst = *daddr6;
441	else {
442		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
443		ip6->ip6_dst.s6_addr16[1] = 0;
444		ip6->ip6_dst.s6_addr32[1] = 0;
445		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
446		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
447		ip6->ip6_dst.s6_addr8[12] = 0xff;
448		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
449			goto bad;
450	}
451	if (!dad) {
452		struct ifaddr *ifa;
453
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 in6_addr *hsrc;
470
471		hsrc = NULL;
472		if (ln != NULL) {
473			LLE_RLOCK(ln);
474			if (ln->la_hold != NULL) {
475				struct ip6_hdr *hip6;		/* hold ip6 */
476
477				/*
478				 * assuming every packet in la_hold has the same IP
479				 * header
480				 */
481				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
482				/* XXX pullup? */
483				if (sizeof(*hip6) < ln->la_hold->m_len) {
484					ip6->ip6_src = hip6->ip6_src;
485					hsrc = &hip6->ip6_src;
486				}
487			}
488			LLE_RUNLOCK(ln);
489		}
490		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
491		    hsrc)) != NULL) {
492			/* ip6_src set already. */
493			ifa_free(ifa);
494		} else {
495			int error;
496			struct sockaddr_in6 dst_sa;
497			struct in6_addr src_in;
498
499			bzero(&dst_sa, sizeof(dst_sa));
500			dst_sa.sin6_family = AF_INET6;
501			dst_sa.sin6_len = sizeof(dst_sa);
502			dst_sa.sin6_addr = ip6->ip6_dst;
503
504			error = in6_selectsrc(&dst_sa, NULL,
505			    NULL, &ro, NULL, NULL, &src_in);
506			if (error) {
507				char ip6buf[INET6_ADDRSTRLEN];
508				nd6log((LOG_DEBUG,
509				    "nd6_ns_output: source can't be "
510				    "determined: dst=%s, error=%d\n",
511				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
512				    error));
513				goto bad;
514			}
515			ip6->ip6_src = src_in;
516		}
517	} else {
518		/*
519		 * Source address for DAD packet must always be IPv6
520		 * unspecified address. (0::0)
521		 * We actually don't have to 0-clear the address (we did it
522		 * above), but we do so here explicitly to make the intention
523		 * clearer.
524		 */
525		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
526	}
527	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
528	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
529	nd_ns->nd_ns_code = 0;
530	nd_ns->nd_ns_reserved = 0;
531	nd_ns->nd_ns_target = *taddr6;
532	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
533
534	/*
535	 * Add source link-layer address option.
536	 *
537	 *				spec		implementation
538	 *				---		---
539	 * DAD packet			MUST NOT	do not add the option
540	 * there's no link layer address:
541	 *				impossible	do not add the option
542	 * there's link layer address:
543	 *	Multicast NS		MUST add one	add the option
544	 *	Unicast NS		SHOULD add one	add the option
545	 */
546	if (!dad && (mac = nd6_ifptomac(ifp))) {
547		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
548		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
549		/* 8 byte alignments... */
550		optlen = (optlen + 7) & ~7;
551
552		m->m_pkthdr.len += optlen;
553		m->m_len += optlen;
554		icmp6len += optlen;
555		bzero((caddr_t)nd_opt, optlen);
556		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
557		nd_opt->nd_opt_len = optlen >> 3;
558		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
559	}
560
561	ip6->ip6_plen = htons((u_short)icmp6len);
562	nd_ns->nd_ns_cksum = 0;
563	nd_ns->nd_ns_cksum =
564	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
565
566	if (send_sendso_input_hook != NULL) {
567		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
568			sizeof(unsigned short), M_NOWAIT);
569		if (mtag == NULL)
570			goto bad;
571		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
572		m_tag_prepend(m, mtag);
573	}
574
575	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
576	icmp6_ifstat_inc(ifp, ifs6_out_msg);
577	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
578	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
579
580	if (ro.ro_rt) {		/* we don't cache this route. */
581		RTFREE(ro.ro_rt);
582	}
583	return;
584
585  bad:
586	if (ro.ro_rt) {
587		RTFREE(ro.ro_rt);
588	}
589	m_freem(m);
590	return;
591}
592
593/*
594 * Neighbor advertisement input handling.
595 *
596 * Based on RFC 2461
597 * Based on RFC 2462 (duplicate address detection)
598 *
599 * the following items are not implemented yet:
600 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
601 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
602 */
603void
604nd6_na_input(struct mbuf *m, int off, int icmp6len)
605{
606	struct ifnet *ifp = m->m_pkthdr.rcvif;
607	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
608	struct nd_neighbor_advert *nd_na;
609	struct in6_addr daddr6 = ip6->ip6_dst;
610	struct in6_addr taddr6;
611	int flags;
612	int is_router;
613	int is_solicited;
614	int is_override;
615	char *lladdr = NULL;
616	int lladdrlen = 0;
617	int checklink = 0;
618	struct ifaddr *ifa;
619	struct llentry *ln = NULL;
620	union nd_opts ndopts;
621	struct mbuf *chain = NULL;
622	struct m_tag *mtag;
623	struct sockaddr_in6 sin6;
624	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
625
626	if (ip6->ip6_hlim != 255) {
627		nd6log((LOG_ERR,
628		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
629		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
630		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
631		goto bad;
632	}
633
634#ifndef PULLDOWN_TEST
635	IP6_EXTHDR_CHECK(m, off, icmp6len,);
636	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
637#else
638	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
639	if (nd_na == NULL) {
640		ICMP6STAT_INC(icp6s_tooshort);
641		return;
642	}
643#endif
644
645	flags = nd_na->nd_na_flags_reserved;
646	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
647	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
648	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
649
650	taddr6 = nd_na->nd_na_target;
651	if (in6_setscope(&taddr6, ifp, NULL))
652		goto bad;	/* XXX: impossible */
653
654	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
655		nd6log((LOG_ERR,
656		    "nd6_na_input: invalid target address %s\n",
657		    ip6_sprintf(ip6bufs, &taddr6)));
658		goto bad;
659	}
660	if (IN6_IS_ADDR_MULTICAST(&daddr6))
661		if (is_solicited) {
662			nd6log((LOG_ERR,
663			    "nd6_na_input: a solicited adv is multicasted\n"));
664			goto bad;
665		}
666
667	icmp6len -= sizeof(*nd_na);
668	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
669	if (nd6_options(&ndopts) < 0) {
670		nd6log((LOG_INFO,
671		    "nd6_na_input: invalid ND option, ignored\n"));
672		/* nd6_options have incremented stats */
673		goto freeit;
674	}
675
676	if (ndopts.nd_opts_tgt_lladdr) {
677		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
678		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
679	}
680
681	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
682
683	/*
684	 * Target address matches one of my interface address.
685	 *
686	 * If my address is tentative, this means that there's somebody
687	 * already using the same address as mine.  This indicates DAD failure.
688	 * This is defined in RFC 2462.
689	 *
690	 * Otherwise, process as defined in RFC 2461.
691	 */
692	if (ifa
693	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
694		ifa_free(ifa);
695		nd6_dad_na_input(ifa);
696		goto freeit;
697	}
698
699	/* Just for safety, maybe unnecessary. */
700	if (ifa) {
701		ifa_free(ifa);
702		log(LOG_ERR,
703		    "nd6_na_input: duplicate IP6 address %s\n",
704		    ip6_sprintf(ip6bufs, &taddr6));
705		goto freeit;
706	}
707
708	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
709		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
710		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
711		    ifp->if_addrlen, lladdrlen - 2));
712		goto bad;
713	}
714
715	/*
716	 * If no neighbor cache entry is found, NA SHOULD silently be
717	 * discarded.
718	 */
719	IF_AFDATA_LOCK(ifp);
720	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
721	IF_AFDATA_UNLOCK(ifp);
722	if (ln == NULL) {
723		goto freeit;
724	}
725
726	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
727		/*
728		 * If the link-layer has address, and no lladdr option came,
729		 * discard the packet.
730		 */
731		if (ifp->if_addrlen && lladdr == NULL) {
732			goto freeit;
733		}
734
735		/*
736		 * Record link-layer address, and update the state.
737		 */
738		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
739		ln->la_flags |= LLE_VALID;
740		if (is_solicited) {
741			ln->ln_state = ND6_LLINFO_REACHABLE;
742			ln->ln_byhint = 0;
743			if (!ND6_LLINFO_PERMANENT(ln)) {
744				nd6_llinfo_settimer_locked(ln,
745				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
746			}
747		} else {
748			ln->ln_state = ND6_LLINFO_STALE;
749			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
750		}
751		if ((ln->ln_router = is_router) != 0) {
752			/*
753			 * This means a router's state has changed from
754			 * non-reachable to probably reachable, and might
755			 * affect the status of associated prefixes..
756			 */
757			checklink = 1;
758		}
759	} else {
760		int llchange;
761
762		/*
763		 * Check if the link-layer address has changed or not.
764		 */
765		if (lladdr == NULL)
766			llchange = 0;
767		else {
768			if (ln->la_flags & LLE_VALID) {
769				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
770					llchange = 1;
771				else
772					llchange = 0;
773			} else
774				llchange = 1;
775		}
776
777		/*
778		 * This is VERY complex.  Look at it with care.
779		 *
780		 * override solicit lladdr llchange	action
781		 *					(L: record lladdr)
782		 *
783		 *	0	0	n	--	(2c)
784		 *	0	0	y	n	(2b) L
785		 *	0	0	y	y	(1)    REACHABLE->STALE
786		 *	0	1	n	--	(2c)   *->REACHABLE
787		 *	0	1	y	n	(2b) L *->REACHABLE
788		 *	0	1	y	y	(1)    REACHABLE->STALE
789		 *	1	0	n	--	(2a)
790		 *	1	0	y	n	(2a) L
791		 *	1	0	y	y	(2a) L *->STALE
792		 *	1	1	n	--	(2a)   *->REACHABLE
793		 *	1	1	y	n	(2a) L *->REACHABLE
794		 *	1	1	y	y	(2a) L *->REACHABLE
795		 */
796		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
797			/*
798			 * If state is REACHABLE, make it STALE.
799			 * no other updates should be done.
800			 */
801			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
802				ln->ln_state = ND6_LLINFO_STALE;
803				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
804			}
805			goto freeit;
806		} else if (is_override				   /* (2a) */
807			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
808			|| lladdr == NULL) {			   /* (2c) */
809			/*
810			 * Update link-local address, if any.
811			 */
812			if (lladdr != NULL) {
813				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
814				ln->la_flags |= LLE_VALID;
815			}
816
817			/*
818			 * If solicited, make the state REACHABLE.
819			 * If not solicited and the link-layer address was
820			 * changed, make it STALE.
821			 */
822			if (is_solicited) {
823				ln->ln_state = ND6_LLINFO_REACHABLE;
824				ln->ln_byhint = 0;
825				if (!ND6_LLINFO_PERMANENT(ln)) {
826					nd6_llinfo_settimer_locked(ln,
827					    (long)ND_IFINFO(ifp)->reachable * hz);
828				}
829			} else {
830				if (lladdr != NULL && llchange) {
831					ln->ln_state = ND6_LLINFO_STALE;
832					nd6_llinfo_settimer_locked(ln,
833					    (long)V_nd6_gctimer * hz);
834				}
835			}
836		}
837
838		if (ln->ln_router && !is_router) {
839			/*
840			 * The peer dropped the router flag.
841			 * Remove the sender from the Default Router List and
842			 * update the Destination Cache entries.
843			 */
844			struct nd_defrouter *dr;
845			struct in6_addr *in6;
846
847			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
848
849			/*
850			 * Lock to protect the default router list.
851			 * XXX: this might be unnecessary, since this function
852			 * is only called under the network software interrupt
853			 * context.  However, we keep it just for safety.
854			 */
855			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
856			if (dr)
857				defrtrlist_del(dr);
858			else if (!V_ip6_forwarding) {
859				/*
860				 * Even if the neighbor is not in the default
861				 * router list, the neighbor may be used
862				 * as a next hop for some destinations
863				 * (e.g. redirect case). So we must
864				 * call rt6_flush explicitly.
865				 */
866				rt6_flush(&ip6->ip6_src, ifp);
867			}
868		}
869		ln->ln_router = is_router;
870	}
871        /* XXX - QL
872	 *  Does this matter?
873	 *  rt->rt_flags &= ~RTF_REJECT;
874	 */
875	ln->la_asked = 0;
876	if (ln->la_hold) {
877		struct mbuf *m_hold, *m_hold_next;
878
879		/*
880		 * reset the la_hold in advance, to explicitly
881		 * prevent a la_hold lookup in nd6_output()
882		 * (wouldn't happen, though...)
883		 */
884		for (m_hold = ln->la_hold, ln->la_hold = NULL;
885		    m_hold; m_hold = m_hold_next) {
886			m_hold_next = m_hold->m_nextpkt;
887			m_hold->m_nextpkt = NULL;
888			/*
889			 * we assume ifp is not a loopback here, so just set
890			 * the 2nd argument as the 1st one.
891			 */
892
893			if (send_sendso_input_hook != NULL) {
894				mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
895				    sizeof(unsigned short), M_NOWAIT);
896				if (mtag == NULL)
897					goto bad;
898				m_tag_prepend(m, mtag);
899			}
900
901			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
902		}
903	}
904 freeit:
905	if (ln != NULL) {
906		if (chain)
907			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
908		LLE_WUNLOCK(ln);
909
910		if (chain)
911			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
912	}
913	if (checklink)
914		pfxlist_onlink_check();
915
916	m_freem(m);
917	return;
918
919 bad:
920	if (ln != NULL)
921		LLE_WUNLOCK(ln);
922
923	ICMP6STAT_INC(icp6s_badna);
924	m_freem(m);
925}
926
927/*
928 * Neighbor advertisement output handling.
929 *
930 * Based on RFC 2461
931 *
932 * the following items are not implemented yet:
933 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
934 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
935 *
936 * tlladdr - 1 if include target link-layer address
937 * sdl0 - sockaddr_dl (= proxy NA) or NULL
938 */
939void
940nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
941    const struct in6_addr *taddr6, u_long flags, int tlladdr,
942    struct sockaddr *sdl0)
943{
944	struct mbuf *m;
945	struct m_tag *mtag;
946	struct ip6_hdr *ip6;
947	struct nd_neighbor_advert *nd_na;
948	struct ip6_moptions im6o;
949	struct in6_addr src, daddr6;
950	struct sockaddr_in6 dst_sa;
951	int icmp6len, maxlen, error;
952	caddr_t mac = NULL;
953	struct route_in6 ro;
954
955	bzero(&ro, sizeof(ro));
956
957	daddr6 = *daddr6_0;	/* make a local copy for modification */
958
959	/* estimate the size of message */
960	maxlen = sizeof(*ip6) + sizeof(*nd_na);
961	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
962	if (max_linkhdr + maxlen >= MCLBYTES) {
963#ifdef DIAGNOSTIC
964		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
965		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
966#endif
967		return;
968	}
969
970	MGETHDR(m, M_DONTWAIT, MT_DATA);
971	if (m && max_linkhdr + maxlen >= MHLEN) {
972		MCLGET(m, M_DONTWAIT);
973		if ((m->m_flags & M_EXT) == 0) {
974			m_free(m);
975			m = NULL;
976		}
977	}
978	if (m == NULL)
979		return;
980	m->m_pkthdr.rcvif = NULL;
981
982	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
983		m->m_flags |= M_MCAST;
984		im6o.im6o_multicast_ifp = ifp;
985		im6o.im6o_multicast_hlim = 255;
986		im6o.im6o_multicast_loop = 0;
987	}
988
989	icmp6len = sizeof(*nd_na);
990	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
991	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
992
993	/* fill neighbor advertisement packet */
994	ip6 = mtod(m, struct ip6_hdr *);
995	ip6->ip6_flow = 0;
996	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
997	ip6->ip6_vfc |= IPV6_VERSION;
998	ip6->ip6_nxt = IPPROTO_ICMPV6;
999	ip6->ip6_hlim = 255;
1000	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1001		/* reply to DAD */
1002		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1003		daddr6.s6_addr16[1] = 0;
1004		daddr6.s6_addr32[1] = 0;
1005		daddr6.s6_addr32[2] = 0;
1006		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1007		if (in6_setscope(&daddr6, ifp, NULL))
1008			goto bad;
1009
1010		flags &= ~ND_NA_FLAG_SOLICITED;
1011	}
1012	ip6->ip6_dst = daddr6;
1013	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1014	dst_sa.sin6_family = AF_INET6;
1015	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1016	dst_sa.sin6_addr = daddr6;
1017
1018	/*
1019	 * Select a source whose scope is the same as that of the dest.
1020	 */
1021	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1022	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &src);
1023	if (error) {
1024		char ip6buf[INET6_ADDRSTRLEN];
1025		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1026		    "determined: dst=%s, error=%d\n",
1027		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1028		goto bad;
1029	}
1030	ip6->ip6_src = src;
1031	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1032	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1033	nd_na->nd_na_code = 0;
1034	nd_na->nd_na_target = *taddr6;
1035	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1036
1037	/*
1038	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1039	 * see nd6_ns_input() for details.
1040	 * Basically, if NS packet is sent to unicast/anycast addr,
1041	 * target lladdr option SHOULD NOT be included.
1042	 */
1043	if (tlladdr) {
1044		/*
1045		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1046		 * lladdr in sdl0.  If we are not proxying (sending NA for
1047		 * my address) use lladdr configured for the interface.
1048		 */
1049		if (sdl0 == NULL) {
1050			if (ifp->if_carp)
1051				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1052			if (mac == NULL)
1053				mac = nd6_ifptomac(ifp);
1054		} else if (sdl0->sa_family == AF_LINK) {
1055			struct sockaddr_dl *sdl;
1056			sdl = (struct sockaddr_dl *)sdl0;
1057			if (sdl->sdl_alen == ifp->if_addrlen)
1058				mac = LLADDR(sdl);
1059		}
1060	}
1061	if (tlladdr && mac) {
1062		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1063		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1064
1065		/* roundup to 8 bytes alignment! */
1066		optlen = (optlen + 7) & ~7;
1067
1068		m->m_pkthdr.len += optlen;
1069		m->m_len += optlen;
1070		icmp6len += optlen;
1071		bzero((caddr_t)nd_opt, optlen);
1072		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1073		nd_opt->nd_opt_len = optlen >> 3;
1074		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1075	} else
1076		flags &= ~ND_NA_FLAG_OVERRIDE;
1077
1078	ip6->ip6_plen = htons((u_short)icmp6len);
1079	nd_na->nd_na_flags_reserved = flags;
1080	nd_na->nd_na_cksum = 0;
1081	nd_na->nd_na_cksum =
1082	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1083
1084	if (send_sendso_input_hook != NULL) {
1085		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1086		    sizeof(unsigned short), M_NOWAIT);
1087		if (mtag == NULL)
1088			goto bad;
1089		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1090		m_tag_prepend(m, mtag);
1091	}
1092
1093	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1094	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1095	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1096	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1097
1098	if (ro.ro_rt) {		/* we don't cache this route. */
1099		RTFREE(ro.ro_rt);
1100	}
1101	return;
1102
1103  bad:
1104	if (ro.ro_rt) {
1105		RTFREE(ro.ro_rt);
1106	}
1107	m_freem(m);
1108	return;
1109}
1110
1111caddr_t
1112nd6_ifptomac(struct ifnet *ifp)
1113{
1114	switch (ifp->if_type) {
1115	case IFT_ARCNET:
1116	case IFT_ETHER:
1117	case IFT_FDDI:
1118	case IFT_IEEE1394:
1119#ifdef IFT_L2VLAN
1120	case IFT_L2VLAN:
1121#endif
1122#ifdef IFT_IEEE80211
1123	case IFT_IEEE80211:
1124#endif
1125#ifdef IFT_CARP
1126	case IFT_CARP:
1127#endif
1128	case IFT_BRIDGE:
1129	case IFT_ISO88025:
1130		return IF_LLADDR(ifp);
1131	default:
1132		return NULL;
1133	}
1134}
1135
1136struct dadq {
1137	TAILQ_ENTRY(dadq) dad_list;
1138	struct ifaddr *dad_ifa;
1139	int dad_count;		/* max NS to send */
1140	int dad_ns_tcount;	/* # of trials to send NS */
1141	int dad_ns_ocount;	/* NS sent so far */
1142	int dad_ns_icount;
1143	int dad_na_icount;
1144	struct callout dad_timer_ch;
1145	struct vnet *dad_vnet;
1146};
1147
1148static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1149VNET_DEFINE(int, dad_init) = 0;
1150#define	V_dadq				VNET(dadq)
1151#define	V_dad_init			VNET(dad_init)
1152
1153static struct dadq *
1154nd6_dad_find(struct ifaddr *ifa)
1155{
1156	struct dadq *dp;
1157
1158	for (dp = V_dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1159		if (dp->dad_ifa == ifa)
1160			return dp;
1161	}
1162	return NULL;
1163}
1164
1165static void
1166nd6_dad_starttimer(struct dadq *dp, int ticks)
1167{
1168
1169	callout_reset(&dp->dad_timer_ch, ticks,
1170	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1171}
1172
1173static void
1174nd6_dad_stoptimer(struct dadq *dp)
1175{
1176
1177	callout_stop(&dp->dad_timer_ch);
1178}
1179
1180/*
1181 * Start Duplicate Address Detection (DAD) for specified interface address.
1182 */
1183void
1184nd6_dad_start(struct ifaddr *ifa, int delay)
1185{
1186	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1187	struct dadq *dp;
1188	char ip6buf[INET6_ADDRSTRLEN];
1189
1190	if (!V_dad_init) {
1191		TAILQ_INIT(&V_dadq);
1192		V_dad_init++;
1193	}
1194
1195	/*
1196	 * If we don't need DAD, don't do it.
1197	 * There are several cases:
1198	 * - DAD is disabled (ip6_dad_count == 0)
1199	 * - the interface address is anycast
1200	 */
1201	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1202		log(LOG_DEBUG,
1203			"nd6_dad_start: called with non-tentative address "
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	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1210		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1211		return;
1212	}
1213	if (!V_ip6_dad_count) {
1214		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1215		return;
1216	}
1217	if (ifa->ifa_ifp == NULL)
1218		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1219	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1220		return;
1221	}
1222	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1223		return;
1224	if (nd6_dad_find(ifa) != NULL) {
1225		/* DAD already in progress */
1226		return;
1227	}
1228
1229	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1230	if (dp == NULL) {
1231		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1232			"%s(%s)\n",
1233			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1234			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1235		return;
1236	}
1237	bzero(dp, sizeof(*dp));
1238	callout_init(&dp->dad_timer_ch, 0);
1239#ifdef VIMAGE
1240	dp->dad_vnet = curvnet;
1241#endif
1242	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1243
1244	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1245	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1246
1247	/*
1248	 * Send NS packet for DAD, ip6_dad_count times.
1249	 * Note that we must delay the first transmission, if this is the
1250	 * first packet to be sent from the interface after interface
1251	 * (re)initialization.
1252	 */
1253	dp->dad_ifa = ifa;
1254	ifa_ref(ifa);	/* just for safety */
1255	dp->dad_count = V_ip6_dad_count;
1256	dp->dad_ns_icount = dp->dad_na_icount = 0;
1257	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1258	if (delay == 0) {
1259		nd6_dad_ns_output(dp, ifa);
1260		nd6_dad_starttimer(dp,
1261		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1262	} else {
1263		nd6_dad_starttimer(dp, delay);
1264	}
1265}
1266
1267/*
1268 * terminate DAD unconditionally.  used for address removals.
1269 */
1270void
1271nd6_dad_stop(struct ifaddr *ifa)
1272{
1273	struct dadq *dp;
1274
1275	if (!V_dad_init)
1276		return;
1277	dp = nd6_dad_find(ifa);
1278	if (!dp) {
1279		/* DAD wasn't started yet */
1280		return;
1281	}
1282
1283	nd6_dad_stoptimer(dp);
1284
1285	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1286	free(dp, M_IP6NDP);
1287	dp = NULL;
1288	ifa_free(ifa);
1289}
1290
1291static void
1292nd6_dad_timer(struct dadq *dp)
1293{
1294	CURVNET_SET(dp->dad_vnet);
1295	int s;
1296	struct ifaddr *ifa = dp->dad_ifa;
1297	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1298	char ip6buf[INET6_ADDRSTRLEN];
1299
1300	s = splnet();		/* XXX */
1301
1302	/* Sanity check */
1303	if (ia == NULL) {
1304		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1305		goto done;
1306	}
1307	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1308		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1309			"%s(%s)\n",
1310			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1311			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1312		goto done;
1313	}
1314	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1315		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1316			"%s(%s)\n",
1317			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1318			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1319		goto done;
1320	}
1321
1322	/* timeouted with IFF_{RUNNING,UP} check */
1323	if (dp->dad_ns_tcount > V_dad_maxtry) {
1324		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1325		    if_name(ifa->ifa_ifp)));
1326
1327		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1328		free(dp, M_IP6NDP);
1329		dp = NULL;
1330		ifa_free(ifa);
1331		goto done;
1332	}
1333
1334	/* Need more checks? */
1335	if (dp->dad_ns_ocount < dp->dad_count) {
1336		/*
1337		 * We have more NS to go.  Send NS packet for DAD.
1338		 */
1339		nd6_dad_ns_output(dp, ifa);
1340		nd6_dad_starttimer(dp,
1341		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1342	} else {
1343		/*
1344		 * We have transmitted sufficient number of DAD packets.
1345		 * See what we've got.
1346		 */
1347		int duplicate;
1348
1349		duplicate = 0;
1350
1351		if (dp->dad_na_icount) {
1352			/*
1353			 * the check is in nd6_dad_na_input(),
1354			 * but just in case
1355			 */
1356			duplicate++;
1357		}
1358
1359		if (dp->dad_ns_icount) {
1360			/* We've seen NS, means DAD has failed. */
1361			duplicate++;
1362		}
1363
1364		if (duplicate) {
1365			/* (*dp) will be freed in nd6_dad_duplicated() */
1366			dp = NULL;
1367			nd6_dad_duplicated(ifa);
1368		} else {
1369			/*
1370			 * We are done with DAD.  No NA came, no NS came.
1371			 * No duplicate address found.
1372			 */
1373			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1374
1375			nd6log((LOG_DEBUG,
1376			    "%s: DAD complete for %s - no duplicates found\n",
1377			    if_name(ifa->ifa_ifp),
1378			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1379
1380			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1381			free(dp, M_IP6NDP);
1382			dp = NULL;
1383			ifa_free(ifa);
1384		}
1385	}
1386
1387done:
1388	splx(s);
1389	CURVNET_RESTORE();
1390}
1391
1392void
1393nd6_dad_duplicated(struct ifaddr *ifa)
1394{
1395	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1396	struct ifnet *ifp;
1397	struct dadq *dp;
1398	char ip6buf[INET6_ADDRSTRLEN];
1399
1400	dp = nd6_dad_find(ifa);
1401	if (dp == NULL) {
1402		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1403		return;
1404	}
1405
1406	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1407	    "NS in/out=%d/%d, NA in=%d\n",
1408	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1409	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1410
1411	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1412	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1413
1414	/* We are done with DAD, with duplicate address found. (failure) */
1415	nd6_dad_stoptimer(dp);
1416
1417	ifp = ifa->ifa_ifp;
1418	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1419	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1420	log(LOG_ERR, "%s: manual intervention required\n",
1421	    if_name(ifp));
1422
1423	/*
1424	 * If the address is a link-local address formed from an interface
1425	 * identifier based on the hardware address which is supposed to be
1426	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1427	 * operation on the interface SHOULD be disabled.
1428	 * [RFC 4862, Section 5.4.5]
1429	 */
1430	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1431		struct in6_addr in6;
1432
1433		/*
1434		 * To avoid over-reaction, we only apply this logic when we are
1435		 * very sure that hardware addresses are supposed to be unique.
1436		 */
1437		switch (ifp->if_type) {
1438		case IFT_ETHER:
1439		case IFT_FDDI:
1440		case IFT_ATM:
1441		case IFT_IEEE1394:
1442#ifdef IFT_IEEE80211
1443		case IFT_IEEE80211:
1444#endif
1445			in6 = ia->ia_addr.sin6_addr;
1446			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1447			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1448				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1449				log(LOG_ERR, "%s: possible hardware address "
1450				    "duplication detected, disable IPv6\n",
1451				    if_name(ifp));
1452			}
1453			break;
1454		}
1455	}
1456
1457	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1458	free(dp, M_IP6NDP);
1459	dp = NULL;
1460	ifa_free(ifa);
1461}
1462
1463static void
1464nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1465{
1466	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1467	struct ifnet *ifp = ifa->ifa_ifp;
1468
1469	dp->dad_ns_tcount++;
1470	if ((ifp->if_flags & IFF_UP) == 0) {
1471		return;
1472	}
1473	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1474		return;
1475	}
1476
1477	dp->dad_ns_ocount++;
1478	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1479}
1480
1481static void
1482nd6_dad_ns_input(struct ifaddr *ifa)
1483{
1484	struct in6_ifaddr *ia;
1485	struct ifnet *ifp;
1486	const struct in6_addr *taddr6;
1487	struct dadq *dp;
1488	int duplicate;
1489
1490	if (ifa == NULL)
1491		panic("ifa == NULL in nd6_dad_ns_input");
1492
1493	ia = (struct in6_ifaddr *)ifa;
1494	ifp = ifa->ifa_ifp;
1495	taddr6 = &ia->ia_addr.sin6_addr;
1496	duplicate = 0;
1497	dp = nd6_dad_find(ifa);
1498
1499	/* Quickhack - completely ignore DAD NS packets */
1500	if (V_dad_ignore_ns) {
1501		char ip6buf[INET6_ADDRSTRLEN];
1502		nd6log((LOG_INFO,
1503		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1504		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1505		    if_name(ifa->ifa_ifp)));
1506		return;
1507	}
1508
1509	/*
1510	 * if I'm yet to start DAD, someone else started using this address
1511	 * first.  I have a duplicate and you win.
1512	 */
1513	if (dp == NULL || dp->dad_ns_ocount == 0)
1514		duplicate++;
1515
1516	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1517
1518	if (duplicate) {
1519		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1520		nd6_dad_duplicated(ifa);
1521	} else {
1522		/*
1523		 * not sure if I got a duplicate.
1524		 * increment ns count and see what happens.
1525		 */
1526		if (dp)
1527			dp->dad_ns_icount++;
1528	}
1529}
1530
1531static void
1532nd6_dad_na_input(struct ifaddr *ifa)
1533{
1534	struct dadq *dp;
1535
1536	if (ifa == NULL)
1537		panic("ifa == NULL in nd6_dad_na_input");
1538
1539	dp = nd6_dad_find(ifa);
1540	if (dp)
1541		dp->dad_na_icount++;
1542
1543	/* remove the address. */
1544	nd6_dad_duplicated(ifa);
1545}
1546