nd6_nbr.c revision 186119
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 186119 2008-12-15 06:10:57Z qingli $");
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
85struct dadq;
86static struct dadq *nd6_dad_find(struct ifaddr *);
87static void nd6_dad_starttimer(struct dadq *, int);
88static void nd6_dad_stoptimer(struct dadq *);
89static void nd6_dad_timer(struct ifaddr *);
90static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
91static void nd6_dad_ns_input(struct ifaddr *);
92static void nd6_dad_na_input(struct ifaddr *);
93
94#ifdef VIMAGE_GLOBALS
95int dad_ignore_ns;
96int 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 */
105void
106nd6_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 */
384void
385nd6_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 && 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 */
591void
592nd6_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	struct ifaddr *ifa;
607	struct llentry *ln = NULL;
608	union nd_opts ndopts;
609	struct mbuf *chain = NULL;
610	struct sockaddr_in6 sin6;
611	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
612
613	if (ip6->ip6_hlim != 255) {
614		nd6log((LOG_ERR,
615		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
616		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
617		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
618		goto bad;
619	}
620
621#ifndef PULLDOWN_TEST
622	IP6_EXTHDR_CHECK(m, off, icmp6len,);
623	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
624#else
625	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
626	if (nd_na == NULL) {
627		V_icmp6stat.icp6s_tooshort++;
628		return;
629	}
630#endif
631
632	flags = nd_na->nd_na_flags_reserved;
633	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
634	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
635	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
636
637	taddr6 = nd_na->nd_na_target;
638	if (in6_setscope(&taddr6, ifp, NULL))
639		goto bad;	/* XXX: impossible */
640
641	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
642		nd6log((LOG_ERR,
643		    "nd6_na_input: invalid target address %s\n",
644		    ip6_sprintf(ip6bufs, &taddr6)));
645		goto bad;
646	}
647	if (IN6_IS_ADDR_MULTICAST(&daddr6))
648		if (is_solicited) {
649			nd6log((LOG_ERR,
650			    "nd6_na_input: a solicited adv is multicasted\n"));
651			goto bad;
652		}
653
654	icmp6len -= sizeof(*nd_na);
655	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
656	if (nd6_options(&ndopts) < 0) {
657		nd6log((LOG_INFO,
658		    "nd6_na_input: invalid ND option, ignored\n"));
659		/* nd6_options have incremented stats */
660		goto freeit;
661	}
662
663	if (ndopts.nd_opts_tgt_lladdr) {
664		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
665		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
666	}
667
668	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
669
670	/*
671	 * Target address matches one of my interface address.
672	 *
673	 * If my address is tentative, this means that there's somebody
674	 * already using the same address as mine.  This indicates DAD failure.
675	 * This is defined in RFC 2462.
676	 *
677	 * Otherwise, process as defined in RFC 2461.
678	 */
679	if (ifa
680	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
681		nd6_dad_na_input(ifa);
682		goto freeit;
683	}
684
685	/* Just for safety, maybe unnecessary. */
686	if (ifa) {
687		log(LOG_ERR,
688		    "nd6_na_input: duplicate IP6 address %s\n",
689		    ip6_sprintf(ip6bufs, &taddr6));
690		goto freeit;
691	}
692
693	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
694		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
695		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
696		    ifp->if_addrlen, lladdrlen - 2));
697		goto bad;
698	}
699
700	/*
701	 * If no neighbor cache entry is found, NA SHOULD silently be
702	 * discarded.
703	 */
704	IF_AFDATA_LOCK(ifp);
705	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
706	IF_AFDATA_UNLOCK(ifp);
707	if (ln == NULL) {
708		goto freeit;
709	}
710
711	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
712		/*
713		 * If the link-layer has address, and no lladdr option came,
714		 * discard the packet.
715		 */
716		if (ifp->if_addrlen && lladdr == NULL) {
717			goto freeit;
718		}
719
720		/*
721		 * Record link-layer address, and update the state.
722		 */
723		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
724		ln->la_flags |= LLE_VALID;
725		if (is_solicited) {
726			ln->ln_state = ND6_LLINFO_REACHABLE;
727			ln->ln_byhint = 0;
728			if (!ND6_LLINFO_PERMANENT(ln)) {
729				nd6_llinfo_settimer_locked(ln,
730				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
731			}
732		} else {
733			ln->ln_state = ND6_LLINFO_STALE;
734			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
735		}
736		if ((ln->ln_router = is_router) != 0) {
737			/*
738			 * This means a router's state has changed from
739			 * non-reachable to probably reachable, and might
740			 * affect the status of associated prefixes..
741			 */
742			pfxlist_onlink_check();
743		}
744	} else {
745		int llchange;
746
747		/*
748		 * Check if the link-layer address has changed or not.
749		 */
750		if (lladdr == NULL)
751			llchange = 0;
752		else {
753			if (ln->la_flags & LLE_VALID) {
754				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
755					llchange = 1;
756				else
757					llchange = 0;
758			} else
759				llchange = 1;
760		}
761
762		/*
763		 * This is VERY complex.  Look at it with care.
764		 *
765		 * override solicit lladdr llchange	action
766		 *					(L: record lladdr)
767		 *
768		 *	0	0	n	--	(2c)
769		 *	0	0	y	n	(2b) L
770		 *	0	0	y	y	(1)    REACHABLE->STALE
771		 *	0	1	n	--	(2c)   *->REACHABLE
772		 *	0	1	y	n	(2b) L *->REACHABLE
773		 *	0	1	y	y	(1)    REACHABLE->STALE
774		 *	1	0	n	--	(2a)
775		 *	1	0	y	n	(2a) L
776		 *	1	0	y	y	(2a) L *->STALE
777		 *	1	1	n	--	(2a)   *->REACHABLE
778		 *	1	1	y	n	(2a) L *->REACHABLE
779		 *	1	1	y	y	(2a) L *->REACHABLE
780		 */
781		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
782			/*
783			 * If state is REACHABLE, make it STALE.
784			 * no other updates should be done.
785			 */
786			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
787				ln->ln_state = ND6_LLINFO_STALE;
788				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
789			}
790			goto freeit;
791		} else if (is_override				   /* (2a) */
792			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
793			|| lladdr == NULL) {			   /* (2c) */
794			/*
795			 * Update link-local address, if any.
796			 */
797			if (lladdr != NULL) {
798				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
799				ln->la_flags |= LLE_VALID;
800			}
801
802			/*
803			 * If solicited, make the state REACHABLE.
804			 * If not solicited and the link-layer address was
805			 * changed, make it STALE.
806			 */
807			if (is_solicited) {
808				ln->ln_state = ND6_LLINFO_REACHABLE;
809				ln->ln_byhint = 0;
810				if (!ND6_LLINFO_PERMANENT(ln)) {
811					nd6_llinfo_settimer_locked(ln,
812					    (long)ND_IFINFO(ifp)->reachable * hz);
813				}
814			} else {
815				if (lladdr != NULL && llchange) {
816					ln->ln_state = ND6_LLINFO_STALE;
817					nd6_llinfo_settimer_locked(ln,
818					    (long)V_nd6_gctimer * hz);
819				}
820			}
821		}
822
823		if (ln->ln_router && !is_router) {
824			/*
825			 * The peer dropped the router flag.
826			 * Remove the sender from the Default Router List and
827			 * update the Destination Cache entries.
828			 */
829			struct nd_defrouter *dr;
830			struct in6_addr *in6;
831
832			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
833
834			/*
835			 * Lock to protect the default router list.
836			 * XXX: this might be unnecessary, since this function
837			 * is only called under the network software interrupt
838			 * context.  However, we keep it just for safety.
839			 */
840			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
841			if (dr)
842				defrtrlist_del(dr);
843			else if (!V_ip6_forwarding) {
844				/*
845				 * Even if the neighbor is not in the default
846				 * router list, the neighbor may be used
847				 * as a next hop for some destinations
848				 * (e.g. redirect case). So we must
849				 * call rt6_flush explicitly.
850				 */
851				rt6_flush(&ip6->ip6_src, ifp);
852			}
853		}
854		ln->ln_router = is_router;
855	}
856        /* XXX - QL
857	 *  Does this matter?
858	 *  rt->rt_flags &= ~RTF_REJECT;
859	 */
860	ln->la_asked = 0;
861	if (ln->la_hold) {
862		struct mbuf *m_hold, *m_hold_next;
863
864		/*
865		 * reset the la_hold in advance, to explicitly
866		 * prevent a la_hold lookup in nd6_output()
867		 * (wouldn't happen, though...)
868		 */
869		for (m_hold = ln->la_hold, ln->la_hold = NULL;
870		    m_hold; m_hold = m_hold_next) {
871			m_hold_next = m_hold->m_nextpkt;
872			m_hold->m_nextpkt = NULL;
873			/*
874			 * we assume ifp is not a loopback here, so just set
875			 * the 2nd argument as the 1st one.
876			 */
877			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
878		}
879	}
880 freeit:
881	if (ln) {
882		if (chain)
883			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
884		LLE_WUNLOCK(ln);
885
886		if (chain)
887			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
888	}
889	m_freem(m);
890	return;
891
892 bad:
893	if (ln)
894		LLE_WUNLOCK(ln);
895
896	V_icmp6stat.icp6s_badna++;
897	m_freem(m);
898}
899
900/*
901 * Neighbor advertisement output handling.
902 *
903 * Based on RFC 2461
904 *
905 * the following items are not implemented yet:
906 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
907 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
908 *
909 * tlladdr - 1 if include target link-layer address
910 * sdl0 - sockaddr_dl (= proxy NA) or NULL
911 */
912void
913nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
914    const struct in6_addr *taddr6, u_long flags, int tlladdr,
915    struct sockaddr *sdl0)
916{
917	INIT_VNET_INET6(ifp->if_vnet);
918	struct mbuf *m;
919	struct ip6_hdr *ip6;
920	struct nd_neighbor_advert *nd_na;
921	struct ip6_moptions im6o;
922	struct in6_addr *src, daddr6;
923	struct sockaddr_in6 dst_sa;
924	int icmp6len, maxlen, error;
925	caddr_t mac = NULL;
926	struct route_in6 ro;
927
928	bzero(&ro, sizeof(ro));
929
930	daddr6 = *daddr6_0;	/* make a local copy for modification */
931
932	/* estimate the size of message */
933	maxlen = sizeof(*ip6) + sizeof(*nd_na);
934	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
935	if (max_linkhdr + maxlen >= MCLBYTES) {
936#ifdef DIAGNOSTIC
937		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
938		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
939#endif
940		return;
941	}
942
943	MGETHDR(m, M_DONTWAIT, MT_DATA);
944	if (m && max_linkhdr + maxlen >= MHLEN) {
945		MCLGET(m, M_DONTWAIT);
946		if ((m->m_flags & M_EXT) == 0) {
947			m_free(m);
948			m = NULL;
949		}
950	}
951	if (m == NULL)
952		return;
953	m->m_pkthdr.rcvif = NULL;
954
955	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
956		m->m_flags |= M_MCAST;
957		im6o.im6o_multicast_ifp = ifp;
958		im6o.im6o_multicast_hlim = 255;
959		im6o.im6o_multicast_loop = 0;
960	}
961
962	icmp6len = sizeof(*nd_na);
963	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
964	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
965
966	/* fill neighbor advertisement packet */
967	ip6 = mtod(m, struct ip6_hdr *);
968	ip6->ip6_flow = 0;
969	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
970	ip6->ip6_vfc |= IPV6_VERSION;
971	ip6->ip6_nxt = IPPROTO_ICMPV6;
972	ip6->ip6_hlim = 255;
973	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
974		/* reply to DAD */
975		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
976		daddr6.s6_addr16[1] = 0;
977		daddr6.s6_addr32[1] = 0;
978		daddr6.s6_addr32[2] = 0;
979		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
980		if (in6_setscope(&daddr6, ifp, NULL))
981			goto bad;
982
983		flags &= ~ND_NA_FLAG_SOLICITED;
984	}
985	ip6->ip6_dst = daddr6;
986	bzero(&dst_sa, sizeof(struct sockaddr_in6));
987	dst_sa.sin6_family = AF_INET6;
988	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
989	dst_sa.sin6_addr = daddr6;
990
991	/*
992	 * Select a source whose scope is the same as that of the dest.
993	 */
994	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
995	src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &error);
996	if (src == NULL) {
997		char ip6buf[INET6_ADDRSTRLEN];
998		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
999		    "determined: dst=%s, error=%d\n",
1000		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1001		goto bad;
1002	}
1003	ip6->ip6_src = *src;
1004	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1005	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1006	nd_na->nd_na_code = 0;
1007	nd_na->nd_na_target = *taddr6;
1008	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1009
1010	/*
1011	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1012	 * see nd6_ns_input() for details.
1013	 * Basically, if NS packet is sent to unicast/anycast addr,
1014	 * target lladdr option SHOULD NOT be included.
1015	 */
1016	if (tlladdr) {
1017		/*
1018		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1019		 * lladdr in sdl0.  If we are not proxying (sending NA for
1020		 * my address) use lladdr configured for the interface.
1021		 */
1022		if (sdl0 == NULL) {
1023#ifdef DEV_CARP
1024			if (ifp->if_carp)
1025				mac = carp_macmatch6(ifp->if_carp, m, taddr6);
1026			if (mac == NULL)
1027				mac = nd6_ifptomac(ifp);
1028#else
1029			mac = nd6_ifptomac(ifp);
1030#endif
1031		} else if (sdl0->sa_family == AF_LINK) {
1032			struct sockaddr_dl *sdl;
1033			sdl = (struct sockaddr_dl *)sdl0;
1034			if (sdl->sdl_alen == ifp->if_addrlen)
1035				mac = LLADDR(sdl);
1036		}
1037	}
1038	if (tlladdr && mac) {
1039		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1040		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1041
1042		/* roundup to 8 bytes alignment! */
1043		optlen = (optlen + 7) & ~7;
1044
1045		m->m_pkthdr.len += optlen;
1046		m->m_len += optlen;
1047		icmp6len += optlen;
1048		bzero((caddr_t)nd_opt, optlen);
1049		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1050		nd_opt->nd_opt_len = optlen >> 3;
1051		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1052	} else
1053		flags &= ~ND_NA_FLAG_OVERRIDE;
1054
1055	ip6->ip6_plen = htons((u_short)icmp6len);
1056	nd_na->nd_na_flags_reserved = flags;
1057	nd_na->nd_na_cksum = 0;
1058	nd_na->nd_na_cksum =
1059	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1060
1061	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1062	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1063	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1064	V_icmp6stat.icp6s_outhist[ND_NEIGHBOR_ADVERT]++;
1065
1066	if (ro.ro_rt) {		/* we don't cache this route. */
1067		RTFREE(ro.ro_rt);
1068	}
1069	return;
1070
1071  bad:
1072	if (ro.ro_rt) {
1073		RTFREE(ro.ro_rt);
1074	}
1075	m_freem(m);
1076	return;
1077}
1078
1079caddr_t
1080nd6_ifptomac(struct ifnet *ifp)
1081{
1082	switch (ifp->if_type) {
1083	case IFT_ARCNET:
1084	case IFT_ETHER:
1085	case IFT_FDDI:
1086	case IFT_IEEE1394:
1087#ifdef IFT_L2VLAN
1088	case IFT_L2VLAN:
1089#endif
1090#ifdef IFT_IEEE80211
1091	case IFT_IEEE80211:
1092#endif
1093#ifdef IFT_CARP
1094	case IFT_CARP:
1095#endif
1096	case IFT_BRIDGE:
1097	case IFT_ISO88025:
1098		return IF_LLADDR(ifp);
1099	default:
1100		return NULL;
1101	}
1102}
1103
1104TAILQ_HEAD(dadq_head, dadq);
1105struct dadq {
1106	TAILQ_ENTRY(dadq) dad_list;
1107	struct ifaddr *dad_ifa;
1108	int dad_count;		/* max NS to send */
1109	int dad_ns_tcount;	/* # of trials to send NS */
1110	int dad_ns_ocount;	/* NS sent so far */
1111	int dad_ns_icount;
1112	int dad_na_icount;
1113	struct callout dad_timer_ch;
1114};
1115
1116#ifdef VIMAGE_GLOBALS
1117static struct dadq_head dadq;
1118int dad_init;
1119#endif
1120
1121static struct dadq *
1122nd6_dad_find(struct ifaddr *ifa)
1123{
1124	INIT_VNET_INET6(curvnet);
1125	struct dadq *dp;
1126
1127	for (dp = V_dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1128		if (dp->dad_ifa == ifa)
1129			return dp;
1130	}
1131	return NULL;
1132}
1133
1134static void
1135nd6_dad_starttimer(struct dadq *dp, int ticks)
1136{
1137
1138	callout_reset(&dp->dad_timer_ch, ticks,
1139	    (void (*)(void *))nd6_dad_timer, (void *)dp->dad_ifa);
1140}
1141
1142static void
1143nd6_dad_stoptimer(struct dadq *dp)
1144{
1145
1146	callout_stop(&dp->dad_timer_ch);
1147}
1148
1149/*
1150 * Start Duplicate Address Detection (DAD) for specified interface address.
1151 */
1152void
1153nd6_dad_start(struct ifaddr *ifa, int delay)
1154{
1155	INIT_VNET_INET6(curvnet);
1156	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1157	struct dadq *dp;
1158	char ip6buf[INET6_ADDRSTRLEN];
1159
1160	if (!V_dad_init) {
1161		TAILQ_INIT(&V_dadq);
1162		V_dad_init++;
1163	}
1164
1165	/*
1166	 * If we don't need DAD, don't do it.
1167	 * There are several cases:
1168	 * - DAD is disabled (ip6_dad_count == 0)
1169	 * - the interface address is anycast
1170	 */
1171	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1172		log(LOG_DEBUG,
1173			"nd6_dad_start: called with non-tentative address "
1174			"%s(%s)\n",
1175			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1176			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1177		return;
1178	}
1179	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1180		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1181		return;
1182	}
1183	if (!V_ip6_dad_count) {
1184		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1185		return;
1186	}
1187	if (ifa->ifa_ifp == NULL)
1188		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1189	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1190		return;
1191	}
1192	if (nd6_dad_find(ifa) != NULL) {
1193		/* DAD already in progress */
1194		return;
1195	}
1196
1197	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1198	if (dp == NULL) {
1199		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1200			"%s(%s)\n",
1201			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1202			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1203		return;
1204	}
1205	bzero(dp, sizeof(*dp));
1206	callout_init(&dp->dad_timer_ch, 0);
1207	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1208
1209	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1210	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1211
1212	/*
1213	 * Send NS packet for DAD, ip6_dad_count times.
1214	 * Note that we must delay the first transmission, if this is the
1215	 * first packet to be sent from the interface after interface
1216	 * (re)initialization.
1217	 */
1218	dp->dad_ifa = ifa;
1219	IFAREF(ifa);	/* just for safety */
1220	dp->dad_count = V_ip6_dad_count;
1221	dp->dad_ns_icount = dp->dad_na_icount = 0;
1222	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1223	if (delay == 0) {
1224		nd6_dad_ns_output(dp, ifa);
1225		nd6_dad_starttimer(dp,
1226		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1227	} else {
1228		nd6_dad_starttimer(dp, delay);
1229	}
1230}
1231
1232/*
1233 * terminate DAD unconditionally.  used for address removals.
1234 */
1235void
1236nd6_dad_stop(struct ifaddr *ifa)
1237{
1238	INIT_VNET_INET6(curvnet);
1239	struct dadq *dp;
1240
1241	if (!V_dad_init)
1242		return;
1243	dp = nd6_dad_find(ifa);
1244	if (!dp) {
1245		/* DAD wasn't started yet */
1246		return;
1247	}
1248
1249	nd6_dad_stoptimer(dp);
1250
1251	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1252	free(dp, M_IP6NDP);
1253	dp = NULL;
1254	IFAFREE(ifa);
1255}
1256
1257static void
1258nd6_dad_timer(struct ifaddr *ifa)
1259{
1260	CURVNET_SET(dp->dad_vnet);
1261	INIT_VNET_INET6(curvnet);
1262	int s;
1263	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1264	struct dadq *dp;
1265	char ip6buf[INET6_ADDRSTRLEN];
1266
1267	s = splnet();		/* XXX */
1268
1269	/* Sanity check */
1270	if (ia == NULL) {
1271		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1272		goto done;
1273	}
1274	dp = nd6_dad_find(ifa);
1275	if (dp == NULL) {
1276		log(LOG_ERR, "nd6_dad_timer: DAD structure not found\n");
1277		goto done;
1278	}
1279	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1280		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1281			"%s(%s)\n",
1282			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1283			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1284		goto done;
1285	}
1286	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1287		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1288			"%s(%s)\n",
1289			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1290			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1291		goto done;
1292	}
1293
1294	/* timeouted with IFF_{RUNNING,UP} check */
1295	if (dp->dad_ns_tcount > V_dad_maxtry) {
1296		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1297		    if_name(ifa->ifa_ifp)));
1298
1299		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1300		free(dp, M_IP6NDP);
1301		dp = NULL;
1302		IFAFREE(ifa);
1303		goto done;
1304	}
1305
1306	/* Need more checks? */
1307	if (dp->dad_ns_ocount < dp->dad_count) {
1308		/*
1309		 * We have more NS to go.  Send NS packet for DAD.
1310		 */
1311		nd6_dad_ns_output(dp, ifa);
1312		nd6_dad_starttimer(dp,
1313		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1314	} else {
1315		/*
1316		 * We have transmitted sufficient number of DAD packets.
1317		 * See what we've got.
1318		 */
1319		int duplicate;
1320
1321		duplicate = 0;
1322
1323		if (dp->dad_na_icount) {
1324			/*
1325			 * the check is in nd6_dad_na_input(),
1326			 * but just in case
1327			 */
1328			duplicate++;
1329		}
1330
1331		if (dp->dad_ns_icount) {
1332			/* We've seen NS, means DAD has failed. */
1333			duplicate++;
1334		}
1335
1336		if (duplicate) {
1337			/* (*dp) will be freed in nd6_dad_duplicated() */
1338			dp = NULL;
1339			nd6_dad_duplicated(ifa);
1340		} else {
1341			/*
1342			 * We are done with DAD.  No NA came, no NS came.
1343			 * No duplicate address found.
1344			 */
1345			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1346
1347			nd6log((LOG_DEBUG,
1348			    "%s: DAD complete for %s - no duplicates found\n",
1349			    if_name(ifa->ifa_ifp),
1350			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1351
1352			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1353			free(dp, M_IP6NDP);
1354			dp = NULL;
1355			IFAFREE(ifa);
1356		}
1357	}
1358
1359done:
1360	splx(s);
1361	CURVNET_RESTORE();
1362}
1363
1364void
1365nd6_dad_duplicated(struct ifaddr *ifa)
1366{
1367	INIT_VNET_INET6(curvnet);
1368	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1369	struct ifnet *ifp;
1370	struct dadq *dp;
1371	char ip6buf[INET6_ADDRSTRLEN];
1372
1373	dp = nd6_dad_find(ifa);
1374	if (dp == NULL) {
1375		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1376		return;
1377	}
1378
1379	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1380	    "NS in/out=%d/%d, NA in=%d\n",
1381	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1382	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1383
1384	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1385	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1386
1387	/* We are done with DAD, with duplicate address found. (failure) */
1388	nd6_dad_stoptimer(dp);
1389
1390	ifp = ifa->ifa_ifp;
1391	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1392	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1393	log(LOG_ERR, "%s: manual intervention required\n",
1394	    if_name(ifp));
1395
1396	/*
1397	 * If the address is a link-local address formed from an interface
1398	 * identifier based on the hardware address which is supposed to be
1399	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1400	 * operation on the interface SHOULD be disabled.
1401	 * [rfc2462bis-03 Section 5.4.5]
1402	 */
1403	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1404		struct in6_addr in6;
1405
1406		/*
1407		 * To avoid over-reaction, we only apply this logic when we are
1408		 * very sure that hardware addresses are supposed to be unique.
1409		 */
1410		switch (ifp->if_type) {
1411		case IFT_ETHER:
1412		case IFT_FDDI:
1413		case IFT_ATM:
1414		case IFT_IEEE1394:
1415#ifdef IFT_IEEE80211
1416		case IFT_IEEE80211:
1417#endif
1418			in6 = ia->ia_addr.sin6_addr;
1419			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1420			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1421				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1422				log(LOG_ERR, "%s: possible hardware address "
1423				    "duplication detected, disable IPv6\n",
1424				    if_name(ifp));
1425			}
1426			break;
1427		}
1428	}
1429
1430	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1431	free(dp, M_IP6NDP);
1432	dp = NULL;
1433	IFAFREE(ifa);
1434}
1435
1436static void
1437nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1438{
1439	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1440	struct ifnet *ifp = ifa->ifa_ifp;
1441
1442	dp->dad_ns_tcount++;
1443	if ((ifp->if_flags & IFF_UP) == 0) {
1444		return;
1445	}
1446	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1447		return;
1448	}
1449
1450	dp->dad_ns_ocount++;
1451	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1452}
1453
1454static void
1455nd6_dad_ns_input(struct ifaddr *ifa)
1456{
1457	INIT_VNET_INET6(curvnet);
1458	struct in6_ifaddr *ia;
1459	struct ifnet *ifp;
1460	const struct in6_addr *taddr6;
1461	struct dadq *dp;
1462	int duplicate;
1463
1464	if (ifa == NULL)
1465		panic("ifa == NULL in nd6_dad_ns_input");
1466
1467	ia = (struct in6_ifaddr *)ifa;
1468	ifp = ifa->ifa_ifp;
1469	taddr6 = &ia->ia_addr.sin6_addr;
1470	duplicate = 0;
1471	dp = nd6_dad_find(ifa);
1472
1473	/* Quickhack - completely ignore DAD NS packets */
1474	if (V_dad_ignore_ns) {
1475		char ip6buf[INET6_ADDRSTRLEN];
1476		nd6log((LOG_INFO,
1477		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1478		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1479		    if_name(ifa->ifa_ifp)));
1480		return;
1481	}
1482
1483	/*
1484	 * if I'm yet to start DAD, someone else started using this address
1485	 * first.  I have a duplicate and you win.
1486	 */
1487	if (dp == NULL || dp->dad_ns_ocount == 0)
1488		duplicate++;
1489
1490	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1491
1492	if (duplicate) {
1493		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1494		nd6_dad_duplicated(ifa);
1495	} else {
1496		/*
1497		 * not sure if I got a duplicate.
1498		 * increment ns count and see what happens.
1499		 */
1500		if (dp)
1501			dp->dad_ns_icount++;
1502	}
1503}
1504
1505static void
1506nd6_dad_na_input(struct ifaddr *ifa)
1507{
1508	struct dadq *dp;
1509
1510	if (ifa == NULL)
1511		panic("ifa == NULL in nd6_dad_na_input");
1512
1513	dp = nd6_dad_find(ifa);
1514	if (dp)
1515		dp->dad_na_icount++;
1516
1517	/* remove the address. */
1518	nd6_dad_duplicated(ifa);
1519}
1520