nd6_nbr.c revision 194760
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 194760 2009-06-23 20:19:09Z rwatson $");
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 dadq *);
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		ICMP6STAT_INC(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_LOCKED(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	if (ifa != NULL)
359		ifa_free(ifa);
360	m_freem(m);
361	return;
362
363 bad:
364	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
365		ip6_sprintf(ip6bufs, &saddr6)));
366	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
367		ip6_sprintf(ip6bufs, &daddr6)));
368	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
369		ip6_sprintf(ip6bufs, &taddr6)));
370	ICMP6STAT_INC(icp6s_badns);
371	if (ifa != NULL)
372		ifa_free(ifa);
373	m_freem(m);
374}
375
376/*
377 * Output a Neighbor Solicitation Message. Caller specifies:
378 *	- ICMP6 header source IP6 address
379 *	- ND6 header target IP6 address
380 *	- ND6 header source datalink address
381 *
382 * Based on RFC 2461
383 * Based on RFC 2462 (duplicate address detection)
384 *
385 *   ln - for source address determination
386 *  dad - duplicate address detection
387 */
388void
389nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
390    const struct in6_addr *taddr6, struct llentry *ln, int dad)
391{
392	INIT_VNET_INET6(ifp->if_vnet);
393	struct mbuf *m;
394	struct ip6_hdr *ip6;
395	struct nd_neighbor_solicit *nd_ns;
396	struct in6_addr *src, src_in;
397	struct ip6_moptions im6o;
398	int icmp6len;
399	int maxlen;
400	caddr_t mac;
401	struct route_in6 ro;
402
403	bzero(&ro, sizeof(ro));
404
405	if (IN6_IS_ADDR_MULTICAST(taddr6))
406		return;
407
408	/* estimate the size of message */
409	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
410	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
411	if (max_linkhdr + maxlen >= MCLBYTES) {
412#ifdef DIAGNOSTIC
413		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
414		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
415#endif
416		return;
417	}
418
419	MGETHDR(m, M_DONTWAIT, MT_DATA);
420	if (m && max_linkhdr + maxlen >= MHLEN) {
421		MCLGET(m, M_DONTWAIT);
422		if ((m->m_flags & M_EXT) == 0) {
423			m_free(m);
424			m = NULL;
425		}
426	}
427	if (m == NULL)
428		return;
429	m->m_pkthdr.rcvif = NULL;
430
431	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
432		m->m_flags |= M_MCAST;
433		im6o.im6o_multicast_ifp = ifp;
434		im6o.im6o_multicast_hlim = 255;
435		im6o.im6o_multicast_loop = 0;
436	}
437
438	icmp6len = sizeof(*nd_ns);
439	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
440	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
441
442	/* fill neighbor solicitation packet */
443	ip6 = mtod(m, struct ip6_hdr *);
444	ip6->ip6_flow = 0;
445	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
446	ip6->ip6_vfc |= IPV6_VERSION;
447	/* ip6->ip6_plen will be set later */
448	ip6->ip6_nxt = IPPROTO_ICMPV6;
449	ip6->ip6_hlim = 255;
450	if (daddr6)
451		ip6->ip6_dst = *daddr6;
452	else {
453		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
454		ip6->ip6_dst.s6_addr16[1] = 0;
455		ip6->ip6_dst.s6_addr32[1] = 0;
456		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
457		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
458		ip6->ip6_dst.s6_addr8[12] = 0xff;
459		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
460			goto bad;
461	}
462	if (!dad) {
463		struct ifaddr *ifa;
464
465		/*
466		 * RFC2461 7.2.2:
467		 * "If the source address of the packet prompting the
468		 * solicitation is the same as one of the addresses assigned
469		 * to the outgoing interface, that address SHOULD be placed
470		 * in the IP Source Address of the outgoing solicitation.
471		 * Otherwise, any one of the addresses assigned to the
472		 * interface should be used."
473		 *
474		 * We use the source address for the prompting packet
475		 * (saddr6), if:
476		 * - saddr6 is given from the caller (by giving "ln"), and
477		 * - saddr6 belongs to the outgoing interface.
478		 * Otherwise, we perform the source address selection as usual.
479		 */
480		struct ip6_hdr *hip6;		/* hold ip6 */
481		struct in6_addr *hsrc = NULL;
482
483		if ((ln != NULL) && ln->la_hold) {
484			/*
485			 * assuming every packet in la_hold has the same IP
486			 * header
487			 */
488			hip6 = mtod(ln->la_hold, struct ip6_hdr *);
489			/* XXX pullup? */
490			if (sizeof(*hip6) < ln->la_hold->m_len)
491				hsrc = &hip6->ip6_src;
492			else
493				hsrc = NULL;
494		}
495		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
496		    hsrc)) != NULL) {
497			src = hsrc;
498			ifa_free(ifa);
499		} else {
500			int error;
501			struct sockaddr_in6 dst_sa;
502
503			bzero(&dst_sa, sizeof(dst_sa));
504			dst_sa.sin6_family = AF_INET6;
505			dst_sa.sin6_len = sizeof(dst_sa);
506			dst_sa.sin6_addr = ip6->ip6_dst;
507
508			src = in6_selectsrc(&dst_sa, NULL,
509			    NULL, &ro, NULL, NULL, &error);
510			if (src == NULL) {
511				char ip6buf[INET6_ADDRSTRLEN];
512				nd6log((LOG_DEBUG,
513				    "nd6_ns_output: source can't be "
514				    "determined: dst=%s, error=%d\n",
515				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
516				    error));
517				goto bad;
518			}
519		}
520	} else {
521		/*
522		 * Source address for DAD packet must always be IPv6
523		 * unspecified address. (0::0)
524		 * We actually don't have to 0-clear the address (we did it
525		 * above), but we do so here explicitly to make the intention
526		 * clearer.
527		 */
528		bzero(&src_in, sizeof(src_in));
529		src = &src_in;
530	}
531	ip6->ip6_src = *src;
532	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
533	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
534	nd_ns->nd_ns_code = 0;
535	nd_ns->nd_ns_reserved = 0;
536	nd_ns->nd_ns_target = *taddr6;
537	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
538
539	/*
540	 * Add source link-layer address option.
541	 *
542	 *				spec		implementation
543	 *				---		---
544	 * DAD packet			MUST NOT	do not add the option
545	 * there's no link layer address:
546	 *				impossible	do not add the option
547	 * there's link layer address:
548	 *	Multicast NS		MUST add one	add the option
549	 *	Unicast NS		SHOULD add one	add the option
550	 */
551	if (!dad && (mac = nd6_ifptomac(ifp))) {
552		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
553		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
554		/* 8 byte alignments... */
555		optlen = (optlen + 7) & ~7;
556
557		m->m_pkthdr.len += optlen;
558		m->m_len += optlen;
559		icmp6len += optlen;
560		bzero((caddr_t)nd_opt, optlen);
561		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
562		nd_opt->nd_opt_len = optlen >> 3;
563		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
564	}
565
566	ip6->ip6_plen = htons((u_short)icmp6len);
567	nd_ns->nd_ns_cksum = 0;
568	nd_ns->nd_ns_cksum =
569	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
570
571	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
572	icmp6_ifstat_inc(ifp, ifs6_out_msg);
573	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
574	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
575
576	if (ro.ro_rt) {		/* we don't cache this route. */
577		RTFREE(ro.ro_rt);
578	}
579	return;
580
581  bad:
582	if (ro.ro_rt) {
583		RTFREE(ro.ro_rt);
584	}
585	m_freem(m);
586	return;
587}
588
589/*
590 * Neighbor advertisement input handling.
591 *
592 * Based on RFC 2461
593 * Based on RFC 2462 (duplicate address detection)
594 *
595 * the following items are not implemented yet:
596 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
597 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
598 */
599void
600nd6_na_input(struct mbuf *m, int off, int icmp6len)
601{
602	INIT_VNET_INET6(curvnet);
603	struct ifnet *ifp = m->m_pkthdr.rcvif;
604	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
605	struct nd_neighbor_advert *nd_na;
606	struct in6_addr daddr6 = ip6->ip6_dst;
607	struct in6_addr taddr6;
608	int flags;
609	int is_router;
610	int is_solicited;
611	int is_override;
612	char *lladdr = NULL;
613	int lladdrlen = 0;
614	int checklink = 0;
615	struct ifaddr *ifa;
616	struct llentry *ln = NULL;
617	union nd_opts ndopts;
618	struct mbuf *chain = NULL;
619	struct sockaddr_in6 sin6;
620	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
621
622	if (ip6->ip6_hlim != 255) {
623		nd6log((LOG_ERR,
624		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
625		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
626		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
627		goto bad;
628	}
629
630#ifndef PULLDOWN_TEST
631	IP6_EXTHDR_CHECK(m, off, icmp6len,);
632	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
633#else
634	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
635	if (nd_na == NULL) {
636		ICMP6STAT_INC(icp6s_tooshort);
637		return;
638	}
639#endif
640
641	flags = nd_na->nd_na_flags_reserved;
642	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
643	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
644	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
645
646	taddr6 = nd_na->nd_na_target;
647	if (in6_setscope(&taddr6, ifp, NULL))
648		goto bad;	/* XXX: impossible */
649
650	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
651		nd6log((LOG_ERR,
652		    "nd6_na_input: invalid target address %s\n",
653		    ip6_sprintf(ip6bufs, &taddr6)));
654		goto bad;
655	}
656	if (IN6_IS_ADDR_MULTICAST(&daddr6))
657		if (is_solicited) {
658			nd6log((LOG_ERR,
659			    "nd6_na_input: a solicited adv is multicasted\n"));
660			goto bad;
661		}
662
663	icmp6len -= sizeof(*nd_na);
664	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
665	if (nd6_options(&ndopts) < 0) {
666		nd6log((LOG_INFO,
667		    "nd6_na_input: invalid ND option, ignored\n"));
668		/* nd6_options have incremented stats */
669		goto freeit;
670	}
671
672	if (ndopts.nd_opts_tgt_lladdr) {
673		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
674		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
675	}
676
677	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
678
679	/*
680	 * Target address matches one of my interface address.
681	 *
682	 * If my address is tentative, this means that there's somebody
683	 * already using the same address as mine.  This indicates DAD failure.
684	 * This is defined in RFC 2462.
685	 *
686	 * Otherwise, process as defined in RFC 2461.
687	 */
688	if (ifa
689	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
690		ifa_free(ifa);
691		nd6_dad_na_input(ifa);
692		goto freeit;
693	}
694
695	/* Just for safety, maybe unnecessary. */
696	if (ifa) {
697		ifa_free(ifa);
698		log(LOG_ERR,
699		    "nd6_na_input: duplicate IP6 address %s\n",
700		    ip6_sprintf(ip6bufs, &taddr6));
701		goto freeit;
702	}
703
704	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
705		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
706		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
707		    ifp->if_addrlen, lladdrlen - 2));
708		goto bad;
709	}
710
711	/*
712	 * If no neighbor cache entry is found, NA SHOULD silently be
713	 * discarded.
714	 */
715	IF_AFDATA_LOCK(ifp);
716	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
717	IF_AFDATA_UNLOCK(ifp);
718	if (ln == NULL) {
719		goto freeit;
720	}
721
722	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
723		/*
724		 * If the link-layer has address, and no lladdr option came,
725		 * discard the packet.
726		 */
727		if (ifp->if_addrlen && lladdr == NULL) {
728			goto freeit;
729		}
730
731		/*
732		 * Record link-layer address, and update the state.
733		 */
734		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
735		ln->la_flags |= LLE_VALID;
736		if (is_solicited) {
737			ln->ln_state = ND6_LLINFO_REACHABLE;
738			ln->ln_byhint = 0;
739			if (!ND6_LLINFO_PERMANENT(ln)) {
740				nd6_llinfo_settimer_locked(ln,
741				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
742			}
743		} else {
744			ln->ln_state = ND6_LLINFO_STALE;
745			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
746		}
747		if ((ln->ln_router = is_router) != 0) {
748			/*
749			 * This means a router's state has changed from
750			 * non-reachable to probably reachable, and might
751			 * affect the status of associated prefixes..
752			 */
753			checklink = 1;
754		}
755	} else {
756		int llchange;
757
758		/*
759		 * Check if the link-layer address has changed or not.
760		 */
761		if (lladdr == NULL)
762			llchange = 0;
763		else {
764			if (ln->la_flags & LLE_VALID) {
765				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
766					llchange = 1;
767				else
768					llchange = 0;
769			} else
770				llchange = 1;
771		}
772
773		/*
774		 * This is VERY complex.  Look at it with care.
775		 *
776		 * override solicit lladdr llchange	action
777		 *					(L: record lladdr)
778		 *
779		 *	0	0	n	--	(2c)
780		 *	0	0	y	n	(2b) L
781		 *	0	0	y	y	(1)    REACHABLE->STALE
782		 *	0	1	n	--	(2c)   *->REACHABLE
783		 *	0	1	y	n	(2b) L *->REACHABLE
784		 *	0	1	y	y	(1)    REACHABLE->STALE
785		 *	1	0	n	--	(2a)
786		 *	1	0	y	n	(2a) L
787		 *	1	0	y	y	(2a) L *->STALE
788		 *	1	1	n	--	(2a)   *->REACHABLE
789		 *	1	1	y	n	(2a) L *->REACHABLE
790		 *	1	1	y	y	(2a) L *->REACHABLE
791		 */
792		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
793			/*
794			 * If state is REACHABLE, make it STALE.
795			 * no other updates should be done.
796			 */
797			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
798				ln->ln_state = ND6_LLINFO_STALE;
799				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
800			}
801			goto freeit;
802		} else if (is_override				   /* (2a) */
803			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
804			|| lladdr == NULL) {			   /* (2c) */
805			/*
806			 * Update link-local address, if any.
807			 */
808			if (lladdr != NULL) {
809				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
810				ln->la_flags |= LLE_VALID;
811			}
812
813			/*
814			 * If solicited, make the state REACHABLE.
815			 * If not solicited and the link-layer address was
816			 * changed, make it STALE.
817			 */
818			if (is_solicited) {
819				ln->ln_state = ND6_LLINFO_REACHABLE;
820				ln->ln_byhint = 0;
821				if (!ND6_LLINFO_PERMANENT(ln)) {
822					nd6_llinfo_settimer_locked(ln,
823					    (long)ND_IFINFO(ifp)->reachable * hz);
824				}
825			} else {
826				if (lladdr != NULL && llchange) {
827					ln->ln_state = ND6_LLINFO_STALE;
828					nd6_llinfo_settimer_locked(ln,
829					    (long)V_nd6_gctimer * hz);
830				}
831			}
832		}
833
834		if (ln->ln_router && !is_router) {
835			/*
836			 * The peer dropped the router flag.
837			 * Remove the sender from the Default Router List and
838			 * update the Destination Cache entries.
839			 */
840			struct nd_defrouter *dr;
841			struct in6_addr *in6;
842
843			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
844
845			/*
846			 * Lock to protect the default router list.
847			 * XXX: this might be unnecessary, since this function
848			 * is only called under the network software interrupt
849			 * context.  However, we keep it just for safety.
850			 */
851			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
852			if (dr)
853				defrtrlist_del(dr);
854			else if (!V_ip6_forwarding) {
855				/*
856				 * Even if the neighbor is not in the default
857				 * router list, the neighbor may be used
858				 * as a next hop for some destinations
859				 * (e.g. redirect case). So we must
860				 * call rt6_flush explicitly.
861				 */
862				rt6_flush(&ip6->ip6_src, ifp);
863			}
864		}
865		ln->ln_router = is_router;
866	}
867        /* XXX - QL
868	 *  Does this matter?
869	 *  rt->rt_flags &= ~RTF_REJECT;
870	 */
871	ln->la_asked = 0;
872	if (ln->la_hold) {
873		struct mbuf *m_hold, *m_hold_next;
874
875		/*
876		 * reset the la_hold in advance, to explicitly
877		 * prevent a la_hold lookup in nd6_output()
878		 * (wouldn't happen, though...)
879		 */
880		for (m_hold = ln->la_hold, ln->la_hold = NULL;
881		    m_hold; m_hold = m_hold_next) {
882			m_hold_next = m_hold->m_nextpkt;
883			m_hold->m_nextpkt = NULL;
884			/*
885			 * we assume ifp is not a loopback here, so just set
886			 * the 2nd argument as the 1st one.
887			 */
888			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
889		}
890	}
891 freeit:
892	if (ln != NULL) {
893		if (chain)
894			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
895		LLE_WUNLOCK(ln);
896
897		if (chain)
898			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
899	}
900	if (checklink)
901		pfxlist_onlink_check();
902
903	m_freem(m);
904	return;
905
906 bad:
907	if (ln != NULL)
908		LLE_WUNLOCK(ln);
909
910	ICMP6STAT_INC(icp6s_badna);
911	m_freem(m);
912}
913
914/*
915 * Neighbor advertisement output handling.
916 *
917 * Based on RFC 2461
918 *
919 * the following items are not implemented yet:
920 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
921 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
922 *
923 * tlladdr - 1 if include target link-layer address
924 * sdl0 - sockaddr_dl (= proxy NA) or NULL
925 */
926void
927nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
928    const struct in6_addr *taddr6, u_long flags, int tlladdr,
929    struct sockaddr *sdl0)
930{
931	INIT_VNET_INET6(ifp->if_vnet);
932	struct mbuf *m;
933	struct ip6_hdr *ip6;
934	struct nd_neighbor_advert *nd_na;
935	struct ip6_moptions im6o;
936	struct in6_addr *src, daddr6;
937	struct sockaddr_in6 dst_sa;
938	int icmp6len, maxlen, error;
939	caddr_t mac = NULL;
940	struct route_in6 ro;
941
942	bzero(&ro, sizeof(ro));
943
944	daddr6 = *daddr6_0;	/* make a local copy for modification */
945
946	/* estimate the size of message */
947	maxlen = sizeof(*ip6) + sizeof(*nd_na);
948	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
949	if (max_linkhdr + maxlen >= MCLBYTES) {
950#ifdef DIAGNOSTIC
951		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
952		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
953#endif
954		return;
955	}
956
957	MGETHDR(m, M_DONTWAIT, MT_DATA);
958	if (m && max_linkhdr + maxlen >= MHLEN) {
959		MCLGET(m, M_DONTWAIT);
960		if ((m->m_flags & M_EXT) == 0) {
961			m_free(m);
962			m = NULL;
963		}
964	}
965	if (m == NULL)
966		return;
967	m->m_pkthdr.rcvif = NULL;
968
969	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
970		m->m_flags |= M_MCAST;
971		im6o.im6o_multicast_ifp = ifp;
972		im6o.im6o_multicast_hlim = 255;
973		im6o.im6o_multicast_loop = 0;
974	}
975
976	icmp6len = sizeof(*nd_na);
977	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
978	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
979
980	/* fill neighbor advertisement packet */
981	ip6 = mtod(m, struct ip6_hdr *);
982	ip6->ip6_flow = 0;
983	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
984	ip6->ip6_vfc |= IPV6_VERSION;
985	ip6->ip6_nxt = IPPROTO_ICMPV6;
986	ip6->ip6_hlim = 255;
987	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
988		/* reply to DAD */
989		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
990		daddr6.s6_addr16[1] = 0;
991		daddr6.s6_addr32[1] = 0;
992		daddr6.s6_addr32[2] = 0;
993		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
994		if (in6_setscope(&daddr6, ifp, NULL))
995			goto bad;
996
997		flags &= ~ND_NA_FLAG_SOLICITED;
998	}
999	ip6->ip6_dst = daddr6;
1000	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1001	dst_sa.sin6_family = AF_INET6;
1002	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1003	dst_sa.sin6_addr = daddr6;
1004
1005	/*
1006	 * Select a source whose scope is the same as that of the dest.
1007	 */
1008	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1009	src = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, NULL, &error);
1010	if (src == NULL) {
1011		char ip6buf[INET6_ADDRSTRLEN];
1012		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1013		    "determined: dst=%s, error=%d\n",
1014		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1015		goto bad;
1016	}
1017	ip6->ip6_src = *src;
1018	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1019	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1020	nd_na->nd_na_code = 0;
1021	nd_na->nd_na_target = *taddr6;
1022	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1023
1024	/*
1025	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1026	 * see nd6_ns_input() for details.
1027	 * Basically, if NS packet is sent to unicast/anycast addr,
1028	 * target lladdr option SHOULD NOT be included.
1029	 */
1030	if (tlladdr) {
1031		/*
1032		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1033		 * lladdr in sdl0.  If we are not proxying (sending NA for
1034		 * my address) use lladdr configured for the interface.
1035		 */
1036		if (sdl0 == NULL) {
1037#ifdef DEV_CARP
1038			if (ifp->if_carp)
1039				mac = carp_macmatch6(ifp->if_carp, m, taddr6);
1040			if (mac == NULL)
1041				mac = nd6_ifptomac(ifp);
1042#else
1043			mac = nd6_ifptomac(ifp);
1044#endif
1045		} else if (sdl0->sa_family == AF_LINK) {
1046			struct sockaddr_dl *sdl;
1047			sdl = (struct sockaddr_dl *)sdl0;
1048			if (sdl->sdl_alen == ifp->if_addrlen)
1049				mac = LLADDR(sdl);
1050		}
1051	}
1052	if (tlladdr && mac) {
1053		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1054		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1055
1056		/* roundup to 8 bytes alignment! */
1057		optlen = (optlen + 7) & ~7;
1058
1059		m->m_pkthdr.len += optlen;
1060		m->m_len += optlen;
1061		icmp6len += optlen;
1062		bzero((caddr_t)nd_opt, optlen);
1063		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1064		nd_opt->nd_opt_len = optlen >> 3;
1065		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1066	} else
1067		flags &= ~ND_NA_FLAG_OVERRIDE;
1068
1069	ip6->ip6_plen = htons((u_short)icmp6len);
1070	nd_na->nd_na_flags_reserved = flags;
1071	nd_na->nd_na_cksum = 0;
1072	nd_na->nd_na_cksum =
1073	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1074
1075	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1076	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1077	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1078	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1079
1080	if (ro.ro_rt) {		/* we don't cache this route. */
1081		RTFREE(ro.ro_rt);
1082	}
1083	return;
1084
1085  bad:
1086	if (ro.ro_rt) {
1087		RTFREE(ro.ro_rt);
1088	}
1089	m_freem(m);
1090	return;
1091}
1092
1093caddr_t
1094nd6_ifptomac(struct ifnet *ifp)
1095{
1096	switch (ifp->if_type) {
1097	case IFT_ARCNET:
1098	case IFT_ETHER:
1099	case IFT_FDDI:
1100	case IFT_IEEE1394:
1101#ifdef IFT_L2VLAN
1102	case IFT_L2VLAN:
1103#endif
1104#ifdef IFT_IEEE80211
1105	case IFT_IEEE80211:
1106#endif
1107#ifdef IFT_CARP
1108	case IFT_CARP:
1109#endif
1110	case IFT_BRIDGE:
1111	case IFT_ISO88025:
1112		return IF_LLADDR(ifp);
1113	default:
1114		return NULL;
1115	}
1116}
1117
1118struct dadq {
1119	TAILQ_ENTRY(dadq) dad_list;
1120	struct ifaddr *dad_ifa;
1121	int dad_count;		/* max NS to send */
1122	int dad_ns_tcount;	/* # of trials to send NS */
1123	int dad_ns_ocount;	/* NS sent so far */
1124	int dad_ns_icount;
1125	int dad_na_icount;
1126	struct callout dad_timer_ch;
1127	struct vnet *dad_vnet;
1128};
1129
1130#ifdef VIMAGE_GLOBALS
1131static TAILQ_HEAD(, dadq) dadq;
1132int dad_init;
1133#endif
1134
1135static struct dadq *
1136nd6_dad_find(struct ifaddr *ifa)
1137{
1138	INIT_VNET_INET6(curvnet);
1139	struct dadq *dp;
1140
1141	for (dp = V_dadq.tqh_first; dp; dp = dp->dad_list.tqe_next) {
1142		if (dp->dad_ifa == ifa)
1143			return dp;
1144	}
1145	return NULL;
1146}
1147
1148static void
1149nd6_dad_starttimer(struct dadq *dp, int ticks)
1150{
1151
1152	callout_reset(&dp->dad_timer_ch, ticks,
1153	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1154}
1155
1156static void
1157nd6_dad_stoptimer(struct dadq *dp)
1158{
1159
1160	callout_stop(&dp->dad_timer_ch);
1161}
1162
1163/*
1164 * Start Duplicate Address Detection (DAD) for specified interface address.
1165 */
1166void
1167nd6_dad_start(struct ifaddr *ifa, int delay)
1168{
1169	INIT_VNET_INET6(curvnet);
1170	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1171	struct dadq *dp;
1172	char ip6buf[INET6_ADDRSTRLEN];
1173
1174	if (!V_dad_init) {
1175		TAILQ_INIT(&V_dadq);
1176		V_dad_init++;
1177	}
1178
1179	/*
1180	 * If we don't need DAD, don't do it.
1181	 * There are several cases:
1182	 * - DAD is disabled (ip6_dad_count == 0)
1183	 * - the interface address is anycast
1184	 */
1185	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1186		log(LOG_DEBUG,
1187			"nd6_dad_start: called with non-tentative address "
1188			"%s(%s)\n",
1189			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1190			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1191		return;
1192	}
1193	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1194		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1195		return;
1196	}
1197	if (!V_ip6_dad_count) {
1198		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1199		return;
1200	}
1201	if (ifa->ifa_ifp == NULL)
1202		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1203	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1204		return;
1205	}
1206	if (nd6_dad_find(ifa) != NULL) {
1207		/* DAD already in progress */
1208		return;
1209	}
1210
1211	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1212	if (dp == NULL) {
1213		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1214			"%s(%s)\n",
1215			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1216			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1217		return;
1218	}
1219	bzero(dp, sizeof(*dp));
1220	callout_init(&dp->dad_timer_ch, 0);
1221#ifdef VIMAGE
1222	dp->dad_vnet = curvnet;
1223#endif
1224	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1225
1226	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1227	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1228
1229	/*
1230	 * Send NS packet for DAD, ip6_dad_count times.
1231	 * Note that we must delay the first transmission, if this is the
1232	 * first packet to be sent from the interface after interface
1233	 * (re)initialization.
1234	 */
1235	dp->dad_ifa = ifa;
1236	ifa_ref(ifa);	/* just for safety */
1237	dp->dad_count = V_ip6_dad_count;
1238	dp->dad_ns_icount = dp->dad_na_icount = 0;
1239	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1240	if (delay == 0) {
1241		nd6_dad_ns_output(dp, ifa);
1242		nd6_dad_starttimer(dp,
1243		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1244	} else {
1245		nd6_dad_starttimer(dp, delay);
1246	}
1247}
1248
1249/*
1250 * terminate DAD unconditionally.  used for address removals.
1251 */
1252void
1253nd6_dad_stop(struct ifaddr *ifa)
1254{
1255	INIT_VNET_INET6(curvnet);
1256	struct dadq *dp;
1257
1258	if (!V_dad_init)
1259		return;
1260	dp = nd6_dad_find(ifa);
1261	if (!dp) {
1262		/* DAD wasn't started yet */
1263		return;
1264	}
1265
1266	nd6_dad_stoptimer(dp);
1267
1268	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1269	free(dp, M_IP6NDP);
1270	dp = NULL;
1271	ifa_free(ifa);
1272}
1273
1274static void
1275nd6_dad_timer(struct dadq *dp)
1276{
1277	CURVNET_SET(dp->dad_vnet);
1278	INIT_VNET_INET6(curvnet);
1279	int s;
1280	struct ifaddr *ifa = dp->dad_ifa;
1281	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1282	char ip6buf[INET6_ADDRSTRLEN];
1283
1284	s = splnet();		/* XXX */
1285
1286	/* Sanity check */
1287	if (ia == NULL) {
1288		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1289		goto done;
1290	}
1291	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1292		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1293			"%s(%s)\n",
1294			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1295			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1296		goto done;
1297	}
1298	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1299		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1300			"%s(%s)\n",
1301			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1302			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1303		goto done;
1304	}
1305
1306	/* timeouted with IFF_{RUNNING,UP} check */
1307	if (dp->dad_ns_tcount > V_dad_maxtry) {
1308		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1309		    if_name(ifa->ifa_ifp)));
1310
1311		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1312		free(dp, M_IP6NDP);
1313		dp = NULL;
1314		ifa_free(ifa);
1315		goto done;
1316	}
1317
1318	/* Need more checks? */
1319	if (dp->dad_ns_ocount < dp->dad_count) {
1320		/*
1321		 * We have more NS to go.  Send NS packet for DAD.
1322		 */
1323		nd6_dad_ns_output(dp, ifa);
1324		nd6_dad_starttimer(dp,
1325		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1326	} else {
1327		/*
1328		 * We have transmitted sufficient number of DAD packets.
1329		 * See what we've got.
1330		 */
1331		int duplicate;
1332
1333		duplicate = 0;
1334
1335		if (dp->dad_na_icount) {
1336			/*
1337			 * the check is in nd6_dad_na_input(),
1338			 * but just in case
1339			 */
1340			duplicate++;
1341		}
1342
1343		if (dp->dad_ns_icount) {
1344			/* We've seen NS, means DAD has failed. */
1345			duplicate++;
1346		}
1347
1348		if (duplicate) {
1349			/* (*dp) will be freed in nd6_dad_duplicated() */
1350			dp = NULL;
1351			nd6_dad_duplicated(ifa);
1352		} else {
1353			/*
1354			 * We are done with DAD.  No NA came, no NS came.
1355			 * No duplicate address found.
1356			 */
1357			ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1358
1359			nd6log((LOG_DEBUG,
1360			    "%s: DAD complete for %s - no duplicates found\n",
1361			    if_name(ifa->ifa_ifp),
1362			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1363
1364			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1365			free(dp, M_IP6NDP);
1366			dp = NULL;
1367			ifa_free(ifa);
1368		}
1369	}
1370
1371done:
1372	splx(s);
1373	CURVNET_RESTORE();
1374}
1375
1376void
1377nd6_dad_duplicated(struct ifaddr *ifa)
1378{
1379	INIT_VNET_INET6(curvnet);
1380	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1381	struct ifnet *ifp;
1382	struct dadq *dp;
1383	char ip6buf[INET6_ADDRSTRLEN];
1384
1385	dp = nd6_dad_find(ifa);
1386	if (dp == NULL) {
1387		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1388		return;
1389	}
1390
1391	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1392	    "NS in/out=%d/%d, NA in=%d\n",
1393	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1394	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1395
1396	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1397	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1398
1399	/* We are done with DAD, with duplicate address found. (failure) */
1400	nd6_dad_stoptimer(dp);
1401
1402	ifp = ifa->ifa_ifp;
1403	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1404	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1405	log(LOG_ERR, "%s: manual intervention required\n",
1406	    if_name(ifp));
1407
1408	/*
1409	 * If the address is a link-local address formed from an interface
1410	 * identifier based on the hardware address which is supposed to be
1411	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1412	 * operation on the interface SHOULD be disabled.
1413	 * [rfc2462bis-03 Section 5.4.5]
1414	 */
1415	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1416		struct in6_addr in6;
1417
1418		/*
1419		 * To avoid over-reaction, we only apply this logic when we are
1420		 * very sure that hardware addresses are supposed to be unique.
1421		 */
1422		switch (ifp->if_type) {
1423		case IFT_ETHER:
1424		case IFT_FDDI:
1425		case IFT_ATM:
1426		case IFT_IEEE1394:
1427#ifdef IFT_IEEE80211
1428		case IFT_IEEE80211:
1429#endif
1430			in6 = ia->ia_addr.sin6_addr;
1431			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1432			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1433				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1434				log(LOG_ERR, "%s: possible hardware address "
1435				    "duplication detected, disable IPv6\n",
1436				    if_name(ifp));
1437			}
1438			break;
1439		}
1440	}
1441
1442	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1443	free(dp, M_IP6NDP);
1444	dp = NULL;
1445	ifa_free(ifa);
1446}
1447
1448static void
1449nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1450{
1451	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1452	struct ifnet *ifp = ifa->ifa_ifp;
1453
1454	dp->dad_ns_tcount++;
1455	if ((ifp->if_flags & IFF_UP) == 0) {
1456		return;
1457	}
1458	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1459		return;
1460	}
1461
1462	dp->dad_ns_ocount++;
1463	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1464}
1465
1466static void
1467nd6_dad_ns_input(struct ifaddr *ifa)
1468{
1469	INIT_VNET_INET6(curvnet);
1470	struct in6_ifaddr *ia;
1471	struct ifnet *ifp;
1472	const struct in6_addr *taddr6;
1473	struct dadq *dp;
1474	int duplicate;
1475
1476	if (ifa == NULL)
1477		panic("ifa == NULL in nd6_dad_ns_input");
1478
1479	ia = (struct in6_ifaddr *)ifa;
1480	ifp = ifa->ifa_ifp;
1481	taddr6 = &ia->ia_addr.sin6_addr;
1482	duplicate = 0;
1483	dp = nd6_dad_find(ifa);
1484
1485	/* Quickhack - completely ignore DAD NS packets */
1486	if (V_dad_ignore_ns) {
1487		char ip6buf[INET6_ADDRSTRLEN];
1488		nd6log((LOG_INFO,
1489		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1490		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1491		    if_name(ifa->ifa_ifp)));
1492		return;
1493	}
1494
1495	/*
1496	 * if I'm yet to start DAD, someone else started using this address
1497	 * first.  I have a duplicate and you win.
1498	 */
1499	if (dp == NULL || dp->dad_ns_ocount == 0)
1500		duplicate++;
1501
1502	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1503
1504	if (duplicate) {
1505		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1506		nd6_dad_duplicated(ifa);
1507	} else {
1508		/*
1509		 * not sure if I got a duplicate.
1510		 * increment ns count and see what happens.
1511		 */
1512		if (dp)
1513			dp->dad_ns_icount++;
1514	}
1515}
1516
1517static void
1518nd6_dad_na_input(struct ifaddr *ifa)
1519{
1520	struct dadq *dp;
1521
1522	if (ifa == NULL)
1523		panic("ifa == NULL in nd6_dad_na_input");
1524
1525	dp = nd6_dad_find(ifa);
1526	if (dp)
1527		dp->dad_na_icount++;
1528
1529	/* remove the address. */
1530	nd6_dad_duplicated(ifa);
1531}
1532