1/*	$NetBSD: nd6_nbr.c,v 1.183 2023/03/29 13:01:44 kardel Exp $	*/
2/*	$KAME: nd6_nbr.c,v 1.61 2001/02/10 16:06:14 jinmei Exp $	*/
3
4/*
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__KERNEL_RCSID(0, "$NetBSD: nd6_nbr.c,v 1.183 2023/03/29 13:01:44 kardel Exp $");
35
36#ifdef _KERNEL_OPT
37#include "opt_inet.h"
38#include "opt_net_mpsafe.h"
39#endif
40
41#include <sys/param.h>
42#include <sys/systm.h>
43#include <sys/kmem.h>
44#include <sys/mbuf.h>
45#include <sys/socket.h>
46#include <sys/socketvar.h>
47#include <sys/sockio.h>
48#include <sys/time.h>
49#include <sys/kernel.h>
50#include <sys/errno.h>
51#include <sys/ioctl.h>
52#include <sys/syslog.h>
53#include <sys/queue.h>
54#include <sys/callout.h>
55#include <sys/cprng.h>
56
57#include <net/if.h>
58#include <net/if_types.h>
59#include <net/if_dl.h>
60#include <net/if_llatbl.h>
61#include <net/nd.h>
62#include <net/route.h>
63
64#include <netinet/in.h>
65#include <netinet/in_var.h>
66#include <netinet6/in6_var.h>
67#include <netinet6/in6_ifattach.h>
68#include <netinet/ip6.h>
69#include <netinet6/ip6_var.h>
70#include <netinet6/scope6_var.h>
71#include <netinet6/nd6.h>
72#include <netinet/icmp6.h>
73#include <netinet6/icmp6_private.h>
74
75#include "carp.h"
76#if NCARP > 0
77#include <netinet/ip_carp.h>
78#endif
79
80struct dadq;
81static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *, bool *);
82static bool nd6_dad_ownnonce(struct ifaddr *, struct nd_opt_nonce *nonce);
83static void nd6_dad_starttimer(struct dadq *, int);
84static void nd6_dad_destroytimer(struct dadq *);
85static void nd6_dad_timer(struct dadq *);
86static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
87static void nd6_dad_input(struct ifaddr *, struct nd_opt_nonce *,
88    const struct sockaddr_dl *);
89static void nd6_dad_duplicated(struct ifaddr *, struct dadq *,
90    const struct sockaddr_dl *);
91
92static int dad_maxtry = 15;	/* max # of *tries* to transmit DAD packet */
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, *ifpc;
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 router = ip6_forwarding;
115	int tlladdr;
116	union nd_opts ndopts;
117	const struct sockaddr_dl *proxydl = NULL;
118	struct psref psref;
119	struct psref psref_c;
120	struct psref psref_ia;
121	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
122
123	ifp = ifpc = m_get_rcvif_psref(m, &psref);
124	if (ifp == NULL)
125		goto freeit;
126
127	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
128	if (nd_ns == NULL) {
129		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
130		m_put_rcvif_psref(ifp, &psref);
131		return;
132	}
133	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
134	taddr6 = nd_ns->nd_ns_target;
135	if (in6_setscope(&taddr6, ifp, NULL) != 0)
136		goto bad;
137
138	if (ip6->ip6_hlim != 255) {
139		nd6log(LOG_ERR, "invalid hlim (%d) from %s to %s on %s\n",
140		    ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
141		    IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
142		goto bad;
143	}
144
145	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
146		/* dst has to be a solicited node multicast address. */
147		/* don't check ifindex portion */
148		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
149		    daddr6.s6_addr32[1] == 0 &&
150		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
151		    daddr6.s6_addr8[12] == 0xff) {
152			; /* good */
153		} else {
154			nd6log(LOG_INFO, "bad DAD packet (wrong ip6 dst)\n");
155			goto bad;
156		}
157	} else {
158		struct sockaddr_in6 ssin6;
159
160		/*
161		 * Make sure the source address is from a neighbor's address.
162		 */
163		sockaddr_in6_init(&ssin6, &saddr6, 0, 0, 0);
164		if (nd6_is_addr_neighbor(&ssin6, ifp) == 0) {
165			nd6log(LOG_INFO,
166			    "NS packet from non-neighbor %s on %s\n",
167			    IN6_PRINT(ip6buf, &saddr6), if_name(ifp));
168			goto bad;
169		}
170	}
171
172	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
173		nd6log(LOG_INFO, "bad NS target (multicast)\n");
174		goto bad;
175	}
176
177	icmp6len -= sizeof(*nd_ns);
178	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
179	if (nd6_options(&ndopts) < 0) {
180		nd6log(LOG_INFO, "invalid ND option, ignored\n");
181		/* nd6_options have incremented stats */
182		goto freeit;
183	}
184
185	if (ndopts.nd_opts_src_lladdr) {
186		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
187		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
188	}
189
190	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
191		nd6log(LOG_INFO,
192		    "bad DAD packet (link-layer address option)\n");
193		goto bad;
194	}
195
196	/*
197	 * Attaching target link-layer address to the NA?
198	 * (RFC 2461 7.2.4)
199	 *
200	 * NS IP dst is multicast			MUST add
201	 * Otherwise					MAY be omitted
202	 *
203	 * In this implementation, we omit the target link-layer address
204	 * in the "MAY" case.
205	 */
206#if 0 /* too much! */
207	ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &daddr6);
208	if (ifa && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST))
209		tlladdr = 0;
210	else
211#endif
212	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
213		tlladdr = 0;
214	else
215		tlladdr = 1;
216
217	/*
218	 * Target address (taddr6) must be either:
219	 * (1) Valid unicast/anycast address for my receiving interface,
220	 * (2) Unicast address for which I'm offering proxy service, or
221	 * (3) "tentative" address on which DAD is being performed.
222	 */
223	/* (1) and (3) check. */
224#if NCARP > 0
225	if (ifp->if_carp && ifp->if_type != IFT_CARP) {
226		int s = pserialize_read_enter();
227		ifa = carp_iamatch6(ifp->if_carp, &taddr6);
228		if (ifa != NULL) {
229			ifa_acquire(ifa, &psref_ia);
230			if (ifa->ifa_ifp && ifa->ifa_ifp != ifp) {
231				ifpc = ifa->ifa_ifp;
232				if_acquire(ifpc, &psref_c);
233			}
234		}
235
236
237		pserialize_read_exit(s);
238	} else
239		ifa = NULL;
240	if (!ifa)
241		ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6,
242		    &psref_ia);
243#else
244	ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6,
245	    &psref_ia);
246#endif
247
248	/* (2) check. */
249	if (ifa == NULL) {
250		struct rtentry *rt;
251		struct sockaddr_in6 tsin6;
252
253		sockaddr_in6_init(&tsin6, &taddr6, 0, 0, 0);
254
255		rt = rtalloc1(sin6tosa(&tsin6), 0);
256		if (rt && (rt->rt_flags & RTF_ANNOUNCE) != 0 &&
257		    rt->rt_gateway->sa_family == AF_LINK) {
258			/*
259			 * proxy NDP for single entry
260			 */
261			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal_psref(ifp,
262				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST, &psref_ia);
263			if (ifa) {
264				proxy = 1;
265				proxydl = satocsdl(rt->rt_gateway);
266				router = 0;	/* XXX */
267			}
268		}
269		if (rt)
270			rt_unref(rt);
271	}
272	if (ifa == NULL) {
273		/*
274		 * We've got an NS packet, and we don't have that address
275		 * assigned for us.  We MUST silently ignore it.
276		 * See RFC2461 7.2.3.
277		 */
278		goto freeit;
279	}
280	myaddr6 = *IFA_IN6(ifa);
281	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
282	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
283	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
284		goto freeit;
285
286	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
287		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
288		    "(if %d, NS packet %d)\n",
289		    IN6_PRINT(ip6buf, &taddr6),
290		    ifp->if_addrlen, lladdrlen - 2);
291		goto bad;
292	}
293
294	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
295		nd6log(LOG_INFO, "duplicate IP6 address %s\n",
296		    IN6_PRINT(ip6buf, &saddr6));
297		goto freeit;
298	}
299
300	/*
301	 * We have neighbor solicitation packet, with target address equals to
302	 * one of my tentative address.
303	 *
304	 * src addr	how to process?
305	 * ---		---
306	 * multicast	of course, invalid (rejected in ip6_input)
307	 * unicast	somebody is doing address resolution -> ignore
308	 * unspec	dup address detection
309	 *
310	 * The processing is defined in RFC 2462.
311	 */
312	if (tentative) {
313		/*
314		 * If source address is unspecified address, it is for
315		 * duplicate address detection.
316		 *
317		 * If not, the packet is for address resolution;
318		 * silently ignore it.
319		 */
320		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
321			struct sockaddr_dl sdl, *sdlp;
322
323			if (lladdr != NULL)
324				sdlp = sockaddr_dl_init(&sdl, sizeof(sdl),
325				    ifp->if_index, ifp->if_type,
326				    NULL, 0, lladdr, lladdrlen);
327			else
328				sdlp = NULL;
329			nd6_dad_input(ifa, ndopts.nd_opts_nonce, sdlp);
330		}
331		goto freeit;
332	}
333
334	/*
335	 * It looks that sender is performing DAD.
336	 * Check that the nonce is not being used by the same address
337	 * on another interface.
338	 */
339	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6) && ndopts.nd_opts_nonce != NULL) {
340		if (nd6_dad_ownnonce(ifa, ndopts.nd_opts_nonce))
341			goto freeit;
342	}
343
344	ifa_release(ifa, &psref_ia);
345	ifa = NULL;
346
347	/*
348	 * If the source address is unspecified address, entries must not
349	 * be created or updated.
350	 * It looks that sender is performing DAD.  Output NA toward
351	 * all-node multicast address, to tell the sender that I'm using
352	 * the address.
353	 * S bit ("solicited") must be zero.
354	 */
355	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
356		struct in6_addr in6_all;
357
358		in6_all = in6addr_linklocal_allnodes;
359		if (in6_setscope(&in6_all, ifp, NULL) != 0)
360			goto bad;
361		nd6_na_output(ifpc, &in6_all, &taddr6,
362		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
363		    (ip6_forwarding ? ND_NA_FLAG_ROUTER : 0),
364		    tlladdr, (const struct sockaddr *)proxydl);
365		goto freeit;
366	}
367
368	nd6_cache_lladdr(ifpc, &saddr6, lladdr, lladdrlen, ND_NEIGHBOR_SOLICIT, 0);
369
370	nd6_na_output(ifp, &saddr6, &taddr6,
371	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
372	    (router ? ND_NA_FLAG_ROUTER : 0) | ND_NA_FLAG_SOLICITED,
373	    tlladdr, (const struct sockaddr *)proxydl);
374 freeit:
375	ifa_release(ifa, &psref_ia);
376	m_put_rcvif_psref(ifp, &psref);
377	if (ifp != ifpc)
378		if_put(ifpc, &psref_c);
379
380	m_freem(m);
381	return;
382
383 bad:
384	nd6log(LOG_ERR, "src=%s\n", IN6_PRINT(ip6buf, &saddr6));
385	nd6log(LOG_ERR, "dst=%s\n", IN6_PRINT(ip6buf, &daddr6));
386	nd6log(LOG_ERR, "tgt=%s\n", IN6_PRINT(ip6buf, &taddr6));
387	ICMP6_STATINC(ICMP6_STAT_BADNS);
388	ifa_release(ifa, &psref_ia);
389	m_put_rcvif_psref(ifp, &psref);
390	m_freem(m);
391}
392
393/*
394 * Output a Neighbor Solicitation Message. Caller specifies:
395 *	- ICMP6 header source IP6 address
396 *	- ND6 header target IP6 address
397 *	- ND6 header source datalink address
398 *
399 * Based on RFC 2461
400 * Based on RFC 2462 (duplicate address detection)
401 */
402void
403nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
404    const struct in6_addr *taddr6,
405    const struct in6_addr *hsrc,
406    const uint8_t *nonce	/* duplicate address detection */)
407{
408	struct mbuf *m;
409	struct ip6_hdr *ip6;
410	struct nd_neighbor_solicit *nd_ns;
411	const struct in6_addr *src;
412	struct in6_addr src_in;
413	struct ip6_moptions im6o;
414	int icmp6len;
415	int maxlen;
416	const void *mac;
417	struct route ro;
418
419	if (IN6_IS_ADDR_MULTICAST(taddr6))
420		return;
421
422	memset(&ro, 0, sizeof(ro));
423
424	/* estimate the size of message */
425	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
426	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
427	KASSERTMSG(max_linkhdr + maxlen <= MCLBYTES,
428	    "max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
429	    max_linkhdr, maxlen, MCLBYTES);
430
431	MGETHDR(m, M_DONTWAIT, MT_DATA);
432	if (m && max_linkhdr + maxlen >= MHLEN) {
433		MCLGET(m, M_DONTWAIT);
434		if ((m->m_flags & M_EXT) == 0) {
435			m_free(m);
436			m = NULL;
437		}
438	}
439	if (m == NULL)
440		return;
441	m_reset_rcvif(m);
442
443	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
444		m->m_flags |= M_MCAST;
445		im6o.im6o_multicast_if_index = if_get_index(ifp);
446		im6o.im6o_multicast_hlim = 255;
447		im6o.im6o_multicast_loop = 0;
448	}
449
450	icmp6len = sizeof(*nd_ns);
451	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
452	m->m_data += max_linkhdr;	/* or m_align() equivalent? */
453
454	/* fill neighbor solicitation packet */
455	ip6 = mtod(m, struct ip6_hdr *);
456	ip6->ip6_flow = 0;
457	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
458	ip6->ip6_vfc |= IPV6_VERSION;
459	/* ip6->ip6_plen will be set later */
460	ip6->ip6_nxt = IPPROTO_ICMPV6;
461	ip6->ip6_hlim = 255;
462	if (daddr6)
463		ip6->ip6_dst = *daddr6;
464	else {
465		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
466		ip6->ip6_dst.s6_addr16[1] = 0;
467		ip6->ip6_dst.s6_addr32[1] = 0;
468		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
469		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
470		ip6->ip6_dst.s6_addr8[12] = 0xff;
471		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
472			goto bad;
473	}
474	if (nonce == NULL) {
475		int s;
476		/*
477		 * RFC2461 7.2.2:
478		 * "If the source address of the packet prompting the
479		 * solicitation is the same as one of the addresses assigned
480		 * to the outgoing interface, that address SHOULD be placed
481		 * in the IP Source Address of the outgoing solicitation.
482		 * Otherwise, any one of the addresses assigned to the
483		 * interface should be used."
484		 *
485		 * We use the source address for the prompting packet
486		 * (hsrc), if:
487		 * - hsrc is given from the caller (by giving "ln"), and
488		 * - hsrc belongs to the outgoing interface.
489		 * Otherwise, we perform the source address selection as usual.
490		 */
491		s = pserialize_read_enter();
492		if (hsrc && in6ifa_ifpwithaddr(ifp, hsrc)) {
493			pserialize_read_exit(s);
494			src = hsrc;
495		} else {
496			int error;
497			struct sockaddr_in6 dst_sa;
498
499			pserialize_read_exit(s);
500
501			sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
502
503			error = in6_selectsrc(&dst_sa, NULL,
504			    NULL, &ro, NULL, NULL, NULL, &src_in);
505			if (error != 0) {
506				char ip6buf[INET6_ADDRSTRLEN];
507				nd6log(LOG_DEBUG, "source can't be "
508				    "determined: dst=%s, error=%d\n",
509				    IN6_PRINT(ip6buf, &dst_sa.sin6_addr),
510				    error);
511				goto bad;
512			}
513			src = &src_in;
514		}
515	} else {
516		/*
517		 * Source address for DAD packet must always be IPv6
518		 * unspecified address. (0::0)
519		 * We actually don't have to 0-clear the address (we did it
520		 * above), but we do so here explicitly to make the intention
521		 * clearer.
522		 */
523		memset(&src_in, 0, sizeof(src_in));
524		src = &src_in;
525	}
526	ip6->ip6_src = *src;
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 (nonce == NULL && (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		memset((void *)nd_opt, 0, optlen);
556		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
557		nd_opt->nd_opt_len = optlen >> 3;
558		memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
559	}
560
561	/* Add a nonce option (RFC 3971) to detect looped back NS messages.
562	 * This behavior is documented in RFC 7527. */
563	if (nonce != NULL) {
564		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
565		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
566
567		/* 8-byte alignment is required. */
568		optlen = (optlen + 7) & ~7;
569		m->m_pkthdr.len += optlen;
570		m->m_len += optlen;
571		icmp6len += optlen;
572		memset(nd_opt, 0, optlen);
573		nd_opt->nd_opt_type = ND_OPT_NONCE;
574		nd_opt->nd_opt_len = optlen >> 3;
575		memcpy(nd_opt + 1, nonce, ND_OPT_NONCE_LEN);
576	}
577
578	ip6->ip6_plen = htons((u_int16_t)icmp6len);
579	nd_ns->nd_ns_cksum = 0;
580	nd_ns->nd_ns_cksum =
581	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
582
583	ip6_output(m, NULL, &ro, nonce != NULL ? IPV6_UNSPECSRC : 0,
584	    &im6o, NULL, NULL);
585	icmp6_ifstat_inc(ifp, ifs6_out_msg);
586	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
587	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_SOLICIT);
588
589	rtcache_free(&ro);
590	return;
591
592  bad:
593	rtcache_free(&ro);
594	m_freem(m);
595	return;
596}
597
598/*
599 * Neighbor advertisement input handling.
600 *
601 * Based on RFC 2461
602 * Based on RFC 2462 (duplicate address detection)
603 *
604 * the following items are not implemented yet:
605 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
606 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
607 */
608void
609nd6_na_input(struct mbuf *m, int off, int icmp6len)
610{
611	struct ifnet *ifp;
612	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
613	struct nd_neighbor_advert *nd_na;
614	struct in6_addr saddr6 = ip6->ip6_src;
615	struct in6_addr daddr6 = ip6->ip6_dst;
616	struct in6_addr taddr6;
617	int flags;
618	int is_router;
619	int is_solicited;
620	int is_override;
621	int rt_cmd;
622	char *lladdr = NULL;
623	int lladdrlen = 0;
624	struct ifaddr *ifa;
625	struct llentry *ln = NULL;
626	union nd_opts ndopts;
627	struct sockaddr_in6 ssin6;
628	struct psref psref;
629	struct psref psref_ia;
630	char ip6buf[INET6_ADDRSTRLEN], ip6buf2[INET6_ADDRSTRLEN];
631
632	ifp = m_get_rcvif_psref(m, &psref);
633	if (ifp == NULL)
634		goto freeit;
635
636	if (ip6->ip6_hlim != 255) {
637		nd6log(LOG_ERR,
638		    "invalid hlim (%d) from %s to %s on %s\n",
639		    ip6->ip6_hlim, IN6_PRINT(ip6buf, &ip6->ip6_src),
640		    IN6_PRINT(ip6buf2, &ip6->ip6_dst), if_name(ifp));
641		goto bad;
642	}
643
644	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
645	if (nd_na == NULL) {
646		m_put_rcvif_psref(ifp, &psref);
647		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
648		return;
649	}
650
651	flags = nd_na->nd_na_flags_reserved;
652	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
653	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
654	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
655
656	taddr6 = nd_na->nd_na_target;
657	if (in6_setscope(&taddr6, ifp, NULL)) {
658		goto bad;
659	}
660
661	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
662		nd6log(LOG_ERR, "invalid target address %s\n",
663		    IN6_PRINT(ip6buf, &taddr6));
664		goto bad;
665	}
666	if (is_solicited && IN6_IS_ADDR_MULTICAST(&daddr6)) {
667		nd6log(LOG_ERR, "a solicited adv is multicasted\n");
668		goto bad;
669	}
670
671	icmp6len -= sizeof(*nd_na);
672	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
673	if (nd6_options(&ndopts) < 0) {
674		nd6log(LOG_INFO, "invalid ND option, ignored\n");
675		/* nd6_options have incremented stats */
676		goto freeit;
677	}
678
679	if (ndopts.nd_opts_tgt_lladdr != NULL) {
680		struct ifnet *ifp_ll;
681		struct psref psref_ll;
682
683		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
684		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
685
686		if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
687			nd6log(LOG_INFO, "lladdrlen mismatch for %s "
688			    "(if %d, NA packet %d)\n", IN6_PRINT(ip6buf, &taddr6),
689			    ifp->if_addrlen, lladdrlen - 2);
690			goto bad;
691		}
692
693		ifp_ll = if_get_bylla(lladdr, ifp->if_addrlen, &psref_ll);
694		if (ifp_ll != NULL) {
695			/* it's from me, ignore it. */
696			if_put(ifp_ll, &psref_ll);
697			goto freeit;
698		}
699	}
700
701	ifa = (struct ifaddr *)in6ifa_ifpwithaddr_psref(ifp, &taddr6, &psref_ia);
702
703	/*
704	 * Target address matches one of my interface address.
705	 *
706	 * If my address is tentative, this means that there's somebody
707	 * already using the same address as mine.  This indicates DAD failure.
708	 * This is defined in RFC 2462.
709	 *
710	 * Otherwise, process as defined in RFC 2461.
711	 */
712	if (ifa) {
713		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE) {
714			struct sockaddr_dl sdl, *sdlp;
715
716			if (lladdr != NULL)
717				sdlp = sockaddr_dl_init(&sdl, sizeof(sdl),
718				    ifp->if_index, ifp->if_type,
719				    NULL, 0, lladdr, lladdrlen);
720			else
721				sdlp = NULL;
722			nd6_dad_input(ifa, NULL, sdlp);
723		} else
724			log(LOG_ERR,
725			    "nd6_na_input: duplicate IP6 address %s\n",
726			    IN6_PRINT(ip6buf, &taddr6));
727		ifa_release(ifa, &psref_ia);
728		ifa = NULL;
729		goto freeit;
730	}
731
732	/*
733	 * Make sure the source address is from a neighbor's address.
734	 */
735	sockaddr_in6_init(&ssin6, &saddr6, 0, 0, 0);
736	if (nd6_is_addr_neighbor(&ssin6, ifp) == 0) {
737		nd6log(LOG_INFO, "ND packet from non-neighbor %s on %s\n",
738		    IN6_PRINT(ip6buf, &saddr6), if_name(ifp));
739		goto bad;
740	}
741
742	/*
743	 * If no neighbor cache entry is found, NA SHOULD silently be
744	 * discarded.
745	 */
746	ln = nd6_lookup(&taddr6, ifp, true);
747	if (ln == NULL)
748		goto freeit;
749
750	rt_cmd = 0;
751	if (ln->ln_state <= ND_LLINFO_INCOMPLETE) {
752		/*
753		 * If the link-layer has address, and no lladdr option came,
754		 * discard the packet.
755		 */
756		if (ifp->if_addrlen && !lladdr)
757			goto freeit;
758
759		/*
760		 * Record link-layer address, and update the state.
761		 */
762		memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
763		ln->la_flags |= LLE_VALID;
764		rt_cmd = RTM_ADD;
765		if (is_solicited) {
766			ln->ln_state = ND_LLINFO_REACHABLE;
767			ln->ln_byhint = 0;
768			if (!ND_IS_LLINFO_PERMANENT(ln))
769				nd_set_timer(ln, ND_TIMER_REACHABLE);
770		} else {
771			ln->ln_state = ND_LLINFO_STALE;
772			nd_set_timer(ln, ND_TIMER_GC);
773		}
774	} else {
775		bool llchange;
776
777		/*
778		 * Check if the link-layer address has changed or not.
779		 */
780		if (lladdr == NULL)
781			llchange = false;
782		else {
783			if (ln->la_flags & LLE_VALID) {
784				if (memcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
785					llchange = true;
786				else
787					llchange = false;
788			} else
789				llchange = true;
790		}
791		if (llchange)
792			rt_cmd = RTM_CHANGE;
793
794		/*
795		 * This is VERY complex.  Look at it with care.
796		 *
797		 * override solicit lladdr llchange	action
798		 *					(L: record lladdr)
799		 *
800		 *	0	0	n	--	(2c)
801		 *	0	0	y	n	(2b) L
802		 *	0	0	y	y	(1)    REACHABLE->STALE
803		 *	0	1	n	--	(2c)   *->REACHABLE
804		 *	0	1	y	n	(2b) L *->REACHABLE
805		 *	0	1	y	y	(1)    REACHABLE->STALE
806		 *	1	0	n	--	(2a)
807		 *	1	0	y	n	(2a) L
808		 *	1	0	y	y	(2a) L *->STALE
809		 *	1	1	n	--	(2a)   *->REACHABLE
810		 *	1	1	y	n	(2a) L *->REACHABLE
811		 *	1	1	y	y	(2a) L *->REACHABLE
812		 */
813		if (!is_override && lladdr != NULL && llchange) { /* (1) */
814			/*
815			 * If state is REACHABLE, make it STALE.
816			 * no other updates should be done.
817			 */
818			if (ln->ln_state == ND_LLINFO_REACHABLE) {
819				ln->ln_state = ND_LLINFO_STALE;
820				nd_set_timer(ln, ND_TIMER_GC);
821			}
822			goto freeit;
823		} else if (is_override				   /* (2a) */
824		    || (!is_override && lladdr != NULL && !llchange) /* (2b) */
825		    || lladdr == NULL) {			   /* (2c) */
826			/*
827			 * Update link-local address, if any.
828			 */
829			if (lladdr != NULL) {
830				memcpy(&ln->ll_addr, lladdr, ifp->if_addrlen);
831				ln->la_flags |= LLE_VALID;
832			}
833
834			/*
835			 * If solicited, make the state REACHABLE.
836			 * If not solicited and the link-layer address was
837			 * changed, make it STALE.
838			 */
839			if (is_solicited) {
840				ln->ln_state = ND_LLINFO_REACHABLE;
841				ln->ln_byhint = 0;
842				if (!ND_IS_LLINFO_PERMANENT(ln))
843					nd_set_timer(ln, ND_TIMER_REACHABLE);
844			} else {
845				if (lladdr && llchange) {
846					ln->ln_state = ND_LLINFO_STALE;
847					nd_set_timer(ln, ND_TIMER_GC);
848				}
849			}
850		}
851		ln->ln_router = is_router;
852	}
853        /*
854	 * XXX: does this matter?
855	 * rt->rt_flags &= ~RTF_REJECT;
856	 */
857	ln->ln_asked = 0;
858	nd6_llinfo_release_pkts(ln, ifp);
859
860	if (rt_cmd != 0) {
861		struct sockaddr_in6 sin6;
862
863		sockaddr_in6_init(&sin6, &ln->r_l3addr.addr6, 0, 0, 0);
864		rt_clonedmsg(rt_cmd, sin6tosa(&ssin6), sin6tosa(&sin6),
865		    (char *)&ln->ll_addr, ln->lle_tbl->llt_ifp);
866	}
867
868 freeit:
869	if (ln != NULL)
870		LLE_WUNLOCK(ln);
871
872	m_put_rcvif_psref(ifp, &psref);
873	m_freem(m);
874	return;
875
876 bad:
877	if (ln != NULL)
878		LLE_WUNLOCK(ln);
879
880	ICMP6_STATINC(ICMP6_STAT_BADNA);
881	m_put_rcvif_psref(ifp, &psref);
882	m_freem(m);
883}
884
885/*
886 * Neighbor advertisement output handling.
887 *
888 * Based on RFC 2461
889 *
890 * the following items are not implemented yet:
891 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
892 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
893 */
894void
895nd6_na_output(
896	struct ifnet *ifp,
897	const struct in6_addr *daddr6_0,
898	const struct in6_addr *taddr6,
899	u_long flags,
900	int tlladdr,		/* 1 if include target link-layer address */
901	const struct sockaddr *sdl0)	/* sockaddr_dl (= proxy NA) or NULL */
902{
903	struct mbuf *m;
904	struct ip6_hdr *ip6;
905	struct nd_neighbor_advert *nd_na;
906	struct ip6_moptions im6o;
907	struct sockaddr *dst;
908	union {
909		struct sockaddr		dst;
910		struct sockaddr_in6	dst6;
911	} u;
912	struct in6_addr daddr6;
913	int icmp6len, maxlen, error;
914	const void *mac;
915	struct route ro;
916
917	mac = NULL;
918	memset(&ro, 0, sizeof(ro));
919
920	daddr6 = *daddr6_0;	/* make a local copy for modification */
921
922	/* estimate the size of message */
923	maxlen = sizeof(*ip6) + sizeof(*nd_na);
924	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
925	KASSERTMSG(max_linkhdr + maxlen <= MCLBYTES,
926	    "max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
927	    max_linkhdr, maxlen, MCLBYTES);
928
929	MGETHDR(m, M_DONTWAIT, MT_DATA);
930	if (m && max_linkhdr + maxlen >= MHLEN) {
931		MCLGET(m, M_DONTWAIT);
932		if ((m->m_flags & M_EXT) == 0) {
933			m_free(m);
934			m = NULL;
935		}
936	}
937	if (m == NULL)
938		return;
939	m_reset_rcvif(m);
940
941	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
942		m->m_flags |= M_MCAST;
943		im6o.im6o_multicast_if_index = if_get_index(ifp);
944		im6o.im6o_multicast_hlim = 255;
945		im6o.im6o_multicast_loop = 0;
946	}
947
948	icmp6len = sizeof(*nd_na);
949	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
950	m->m_data += max_linkhdr;	/* or m_align() equivalent? */
951
952	/* fill neighbor advertisement packet */
953	ip6 = mtod(m, struct ip6_hdr *);
954	ip6->ip6_flow = 0;
955	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
956	ip6->ip6_vfc |= IPV6_VERSION;
957	ip6->ip6_nxt = IPPROTO_ICMPV6;
958	ip6->ip6_hlim = 255;
959	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
960		/* reply to DAD */
961		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
962		daddr6.s6_addr16[1] = 0;
963		daddr6.s6_addr32[1] = 0;
964		daddr6.s6_addr32[2] = 0;
965		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
966		if (in6_setscope(&daddr6, ifp, NULL))
967			goto bad;
968
969		flags &= ~ND_NA_FLAG_SOLICITED;
970	}
971	ip6->ip6_dst = daddr6;
972	sockaddr_in6_init(&u.dst6, &daddr6, 0, 0, 0);
973	dst = &u.dst;
974	if (rtcache_setdst(&ro, dst) != 0)
975		goto bad;
976
977	/*
978	 * Select a source whose scope is the same as that of the dest.
979	 */
980	error = in6_selectsrc(satosin6(dst), NULL, NULL, &ro, NULL, NULL, NULL,
981	    &ip6->ip6_src);
982	if (error != 0) {
983		char ip6buf[INET6_ADDRSTRLEN];
984		nd6log(LOG_DEBUG, "source can't be "
985		    "determined: dst=%s, error=%d\n",
986		    IN6_PRINT(ip6buf, &satocsin6(dst)->sin6_addr), error);
987		goto bad;
988	}
989	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
990	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
991	nd_na->nd_na_code = 0;
992	nd_na->nd_na_target = *taddr6;
993	in6_clearscope(&nd_na->nd_na_target); /* XXX */
994
995	/*
996	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
997	 * see nd6_ns_input() for details.
998	 * Basically, if NS packet is sent to unicast/anycast addr,
999	 * target lladdr option SHOULD NOT be included.
1000	 */
1001	if (tlladdr) {
1002		/*
1003		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1004		 * lladdr in sdl0.  If we are not proxying (sending NA for
1005		 * my address) use lladdr configured for the interface.
1006		 */
1007		if (sdl0 == NULL)
1008			mac = nd6_ifptomac(ifp);
1009		else if (sdl0->sa_family == AF_LINK) {
1010			const struct sockaddr_dl *sdl;
1011			sdl = satocsdl(sdl0);
1012			if (sdl->sdl_alen == ifp->if_addrlen)
1013				mac = CLLADDR(sdl);
1014		}
1015	}
1016	if (tlladdr && mac) {
1017		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1018		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1019
1020		/* roundup to 8 bytes alignment! */
1021		optlen = (optlen + 7) & ~7;
1022
1023		m->m_pkthdr.len += optlen;
1024		m->m_len += optlen;
1025		icmp6len += optlen;
1026		memset((void *)nd_opt, 0, optlen);
1027		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1028		nd_opt->nd_opt_len = optlen >> 3;
1029		memcpy((void *)(nd_opt + 1), mac, ifp->if_addrlen);
1030	} else
1031		flags &= ~ND_NA_FLAG_OVERRIDE;
1032
1033	ip6->ip6_plen = htons((u_int16_t)icmp6len);
1034	nd_na->nd_na_flags_reserved = flags;
1035	nd_na->nd_na_cksum = 0;
1036	nd_na->nd_na_cksum =
1037	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1038
1039	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1040
1041	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1042	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1043	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_NEIGHBOR_ADVERT);
1044
1045	rtcache_free(&ro);
1046	return;
1047
1048  bad:
1049	rtcache_free(&ro);
1050	m_freem(m);
1051	return;
1052}
1053
1054const void *
1055nd6_ifptomac(const struct ifnet *ifp)
1056{
1057	switch (ifp->if_type) {
1058	case IFT_ARCNET:
1059	case IFT_ETHER:
1060	case IFT_IEEE1394:
1061	case IFT_PROPVIRTUAL:
1062	case IFT_CARP:
1063	case IFT_L2VLAN:
1064	case IFT_IEEE80211:
1065		return CLLADDR(ifp->if_sadl);
1066	default:
1067		return NULL;
1068	}
1069}
1070
1071TAILQ_HEAD(dadq_head, dadq);
1072struct dadq {
1073	TAILQ_ENTRY(dadq) dad_list;
1074	struct ifaddr *dad_ifa;
1075	int dad_count;			/* max NS to send */
1076	int dad_ns_tcount;		/* # of trials to send NS */
1077	int dad_ns_ocount;		/* NS sent so far */
1078	int dad_ns_lcount;		/* looped back NS */
1079	struct callout dad_timer_ch;
1080#define	ND_OPT_NONCE_STORE	3	/* dad_count should not exceed this */
1081	/*
1082	 * The default ip6_dad_count is 1 as specified by RFC 4862 and
1083	 * practically must users won't exceed this.
1084	 * A storage of 3 is defaulted to here, in-case the administrator wants
1085	 * to match the equivalent behaviour in our ARP implementation.
1086	 * This constraint could be removed by sending the on wire nonce as
1087	 * hmac(key, dad_ns_ocount), but that would increase the nonce size
1088	 * sent on the wire.
1089	 */
1090	uint8_t dad_nonce[ND_OPT_NONCE_STORE][ND_OPT_NONCE_LEN];
1091};
1092
1093static struct dadq_head dadq;
1094static kmutex_t nd6_dad_lock;
1095
1096void
1097nd6_nbr_init(void)
1098{
1099
1100	TAILQ_INIT(&dadq);
1101	mutex_init(&nd6_dad_lock, MUTEX_DEFAULT, IPL_NONE);
1102}
1103
1104static struct dadq *
1105nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *nonce, bool *found_nonce)
1106{
1107	struct in6_addr *myaddr6, *dadaddr6;
1108	bool match_ifa;
1109	struct dadq *dp;
1110	int i, nonce_max;
1111
1112	KASSERT(mutex_owned(&nd6_dad_lock));
1113	KASSERT(ifa != NULL);
1114
1115	myaddr6 = IFA_IN6(ifa);
1116	if (nonce != NULL &&
1117	    nonce->nd_opt_nonce_len != (ND_OPT_NONCE_LEN + 2) / 8)
1118		nonce = NULL;
1119	match_ifa = nonce == NULL || found_nonce == NULL || *found_nonce == false;
1120	if (found_nonce != NULL)
1121		*found_nonce = false;
1122
1123	TAILQ_FOREACH(dp, &dadq, dad_list) {
1124		if (match_ifa) {
1125			if (dp->dad_ifa != ifa)
1126				continue;
1127		} else {
1128			dadaddr6 = IFA_IN6(dp->dad_ifa);
1129			if (!IN6_ARE_ADDR_EQUAL(myaddr6, dadaddr6))
1130				continue;
1131		}
1132
1133		if (nonce == NULL)
1134			break;
1135
1136		nonce_max = MIN(dp->dad_ns_ocount, ND_OPT_NONCE_STORE);
1137		for (i = 0; i < nonce_max; i++) {
1138			if (memcmp(nonce->nd_opt_nonce,
1139			    dp->dad_nonce[i],
1140			    ND_OPT_NONCE_LEN) == 0)
1141				break;
1142		}
1143		if (i < nonce_max) {
1144			char ip6buf[INET6_ADDRSTRLEN];
1145
1146			*found_nonce = true;
1147			log(LOG_DEBUG,
1148			    "%s: detected a looped back NS message for %s\n",
1149			    if_name(ifa->ifa_ifp), IN6_PRINT(ip6buf, myaddr6));
1150			dp->dad_ns_lcount++;
1151			continue;
1152		}
1153
1154		break;
1155	}
1156	return dp;
1157}
1158
1159static bool
1160nd6_dad_ownnonce(struct ifaddr *ifa, struct nd_opt_nonce *nonce)
1161{
1162	bool found_nonce = true;
1163
1164	mutex_enter(&nd6_dad_lock);
1165	nd6_dad_find(ifa, nonce, &found_nonce);
1166	mutex_exit(&nd6_dad_lock);
1167
1168	return found_nonce;
1169}
1170
1171static void
1172nd6_dad_starttimer(struct dadq *dp, int ticks)
1173{
1174
1175	callout_reset(&dp->dad_timer_ch, ticks,
1176	    (void (*)(void *))nd6_dad_timer, dp);
1177}
1178
1179static void
1180nd6_dad_stoptimer(struct dadq *dp)
1181{
1182
1183	KASSERT(mutex_owned(&nd6_dad_lock));
1184
1185	TAILQ_REMOVE(&dadq, dp, dad_list);
1186	/* Tell the timer that dp is being destroyed. */
1187	dp->dad_ifa = NULL;
1188	callout_halt(&dp->dad_timer_ch, &nd6_dad_lock);
1189}
1190
1191static void
1192nd6_dad_destroytimer(struct dadq *dp)
1193{
1194
1195	KASSERT(dp->dad_ifa == NULL);
1196	callout_destroy(&dp->dad_timer_ch);
1197	kmem_intr_free(dp, sizeof(*dp));
1198}
1199
1200/*
1201 * Start Duplicate Address Detection (DAD) for specified interface address.
1202 *
1203 * Note that callout is used when xtick > 0 and not when xtick == 0.
1204 *
1205 * xtick: minimum delay ticks for IFF_UP event
1206 */
1207void
1208nd6_dad_start(struct ifaddr *ifa, int xtick)
1209{
1210	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1211	struct dadq *dp;
1212	char ip6buf[INET6_ADDRSTRLEN];
1213
1214	/*
1215	 * If we don't need DAD, don't do it.
1216	 * There are several cases:
1217	 * - DAD is disabled
1218	 * - the interface address is anycast
1219	 */
1220	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1221		log(LOG_DEBUG,
1222			"nd6_dad_start: called with non-tentative address "
1223			"%s(%s)\n",
1224			IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1225			if_name(ifa->ifa_ifp));
1226		return;
1227	}
1228	if (ia->ia6_flags & IN6_IFF_ANYCAST || !ip6_dad_enabled()) {
1229		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1230		rt_addrmsg(RTM_NEWADDR, ifa);
1231		return;
1232	}
1233	if (!(ifa->ifa_ifp->if_flags & IFF_UP))
1234		return;
1235
1236	dp = kmem_intr_alloc(sizeof(*dp), KM_NOSLEEP);
1237
1238	mutex_enter(&nd6_dad_lock);
1239	if (nd6_dad_find(ifa, NULL, NULL) != NULL) {
1240		mutex_exit(&nd6_dad_lock);
1241		/* DAD already in progress */
1242		if (dp != NULL)
1243			kmem_intr_free(dp, sizeof(*dp));
1244		return;
1245	}
1246
1247	if (dp == NULL) {
1248		mutex_exit(&nd6_dad_lock);
1249		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1250			"%s(%s)\n",
1251			IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1252			if_name(ifa->ifa_ifp));
1253		return;
1254	}
1255
1256	/*
1257	 * Send NS packet for DAD, ip6_dad_count times.
1258	 * Note that we must delay the first transmission, if this is the
1259	 * first packet to be sent from the interface after interface
1260	 * (re)initialization.
1261	 */
1262	callout_init(&dp->dad_timer_ch, CALLOUT_MPSAFE);
1263	dp->dad_ifa = ifa;
1264	ifaref(ifa);	/* just for safety */
1265	dp->dad_count = ip6_dad_count;
1266	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1267	dp->dad_ns_lcount = 0;
1268	TAILQ_INSERT_TAIL(&dadq, (struct dadq *)dp, dad_list);
1269
1270	nd6log(LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1271	    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1272
1273	if (xtick == 0) {
1274		nd6_dad_ns_output(dp, ifa);
1275		nd6_dad_starttimer(dp,
1276		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1277	} else
1278		nd6_dad_starttimer(dp, xtick);
1279	mutex_exit(&nd6_dad_lock);
1280}
1281
1282/*
1283 * terminate DAD unconditionally.  used for address removals.
1284 */
1285void
1286nd6_dad_stop(struct ifaddr *ifa)
1287{
1288	struct dadq *dp;
1289
1290	mutex_enter(&nd6_dad_lock);
1291	dp = nd6_dad_find(ifa, NULL, NULL);
1292	if (dp == NULL) {
1293		mutex_exit(&nd6_dad_lock);
1294		/* DAD wasn't started yet */
1295		return;
1296	}
1297
1298	/* Prevent the timer from running anymore. */
1299	nd6_dad_stoptimer(dp);
1300
1301	mutex_exit(&nd6_dad_lock);
1302
1303	nd6_dad_destroytimer(dp);
1304	ifafree(ifa);
1305}
1306
1307static void
1308nd6_dad_timer(struct dadq *dp)
1309{
1310	struct ifaddr *ifa;
1311	struct in6_ifaddr *ia;
1312	char ip6buf[INET6_ADDRSTRLEN];
1313	bool need_free = false;
1314
1315	KERNEL_LOCK_UNLESS_NET_MPSAFE();
1316	mutex_enter(&nd6_dad_lock);
1317
1318	ifa = dp->dad_ifa;
1319	if (ifa == NULL) {
1320		/* dp is being destroyed by someone.  Do nothing. */
1321		goto done;
1322	}
1323
1324	ia = (struct in6_ifaddr *)ifa;
1325	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1326		log(LOG_ERR, "nd6_dad_timer: called with duplicate address "
1327			"%s(%s)\n",
1328			IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1329			if_name(ifa->ifa_ifp));
1330		goto done;
1331	}
1332	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1333		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1334			"%s(%s)\n",
1335			IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr),
1336			if_name(ifa->ifa_ifp));
1337		goto done;
1338	}
1339
1340	/* timeouted with IFF_{RUNNING,UP} check */
1341	if (dp->dad_ns_tcount > dad_maxtry) {
1342		nd6log(LOG_INFO, "%s: could not run DAD, driver problem?\n",
1343			if_name(ifa->ifa_ifp));
1344
1345		nd6_dad_stoptimer(dp);
1346		need_free = true;
1347		goto done;
1348	}
1349
1350	/* Need more checks? */
1351	if (dp->dad_ns_ocount < dp->dad_count) {
1352		/*
1353		 * We have more NS to go.  Send NS packet for DAD.
1354		 */
1355		nd6_dad_ns_output(dp, ifa);
1356		nd6_dad_starttimer(dp,
1357		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1358	} else {
1359		/*
1360		 * We are done with DAD.  No NA came, no NS came.
1361		 * No duplicate address found.
1362		 */
1363		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1364		rt_addrmsg(RTM_NEWADDR, ifa);
1365
1366		nd6log(LOG_DEBUG,
1367		    "%s: DAD complete for %s - no duplicates found\n",
1368		    if_name(ifa->ifa_ifp),
1369		    IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr));
1370
1371		nd6_dad_stoptimer(dp);
1372		need_free = true;
1373	}
1374done:
1375	mutex_exit(&nd6_dad_lock);
1376
1377	if (need_free) {
1378		nd6_dad_destroytimer(dp);
1379		KASSERT(ifa != NULL);
1380		ifafree(ifa);
1381	}
1382
1383	KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1384}
1385
1386static void
1387nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp,
1388    const struct sockaddr_dl *from)
1389{
1390	struct in6_ifaddr *ia;
1391	struct ifnet *ifp;
1392	char ip6buf[INET6_ADDRSTRLEN], llabuf[LLA_ADDRSTRLEN], *llastr;
1393
1394	KASSERT(mutex_owned(&nd6_dad_lock));
1395	KASSERT(ifa != NULL);
1396
1397	ifp = ifa->ifa_ifp;
1398	ia = (struct in6_ifaddr *)ifa;
1399
1400	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1401	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1402
1403	if (__predict_false(from == NULL))
1404		llastr = NULL;
1405	else
1406		llastr = lla_snprintf(llabuf, sizeof(llabuf),
1407		    CLLADDR(from), from->sdl_alen);
1408
1409	log(LOG_ERR, "%s: DAD duplicate address %s from %s\n",
1410	    if_name(ifp), IN6_PRINT(ip6buf, &ia->ia_addr.sin6_addr), llastr);
1411
1412	/* Inform the routing socket that DAD has completed */
1413	rt_addrmsg_src(RTM_NEWADDR, ifa, (const struct sockaddr *)from);
1414
1415	/*
1416	 * If the address is a link-local address formed from an interface
1417	 * identifier based on the hardware address which is supposed to be
1418	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1419	 * operation on the interface SHOULD be disabled.
1420	 * [rfc2462bis-03 Section 5.4.5]
1421	 */
1422	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1423		struct in6_addr in6;
1424
1425		/*
1426		 * To avoid over-reaction, we only apply this logic when we are
1427		 * very sure that hardware addresses are supposed to be unique.
1428		 */
1429		switch (ifp->if_type) {
1430		case IFT_ETHER:
1431		case IFT_ATM:
1432		case IFT_IEEE1394:
1433		case IFT_IEEE80211:
1434			in6 = ia->ia_addr.sin6_addr;
1435			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1436			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1437				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1438				log(LOG_ERR, "%s: possible hardware address "
1439				    "duplication detected, disable IPv6\n",
1440				    if_name(ifp));
1441			}
1442			break;
1443		}
1444	}
1445}
1446
1447static void
1448nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1449{
1450	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1451	struct ifnet *ifp = ifa->ifa_ifp;
1452	uint8_t *nonce;
1453
1454	dp->dad_ns_tcount++;
1455	if ((ifp->if_flags & IFF_UP) == 0) {
1456#if 0
1457		printf("%s: interface down?\n", if_name(ifp));
1458#endif
1459		return;
1460	}
1461	if ((ifp->if_flags & IFF_RUNNING) == 0) {
1462#if 0
1463		printf("%s: interface not running?\n", if_name(ifp));
1464#endif
1465		return;
1466	}
1467
1468	dp->dad_ns_tcount = 0;
1469	nonce = dp->dad_nonce[dp->dad_ns_ocount % ND_OPT_NONCE_STORE];
1470	cprng_fast(nonce, ND_OPT_NONCE_LEN);
1471	dp->dad_ns_ocount++;
1472
1473	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, nonce);
1474}
1475
1476static void
1477nd6_dad_input(struct ifaddr *ifa, struct nd_opt_nonce *nonce,
1478    const struct sockaddr_dl *from)
1479{
1480	struct dadq *dp;
1481	bool found_nonce = false;
1482
1483	KASSERT(ifa != NULL);
1484
1485	mutex_enter(&nd6_dad_lock);
1486	dp = nd6_dad_find(ifa, nonce, &found_nonce);
1487	if (!found_nonce) {
1488		nd6_dad_duplicated(ifa, dp, from);
1489		if (dp != NULL)
1490			nd6_dad_stoptimer(dp);
1491	}
1492	mutex_exit(&nd6_dad_lock);
1493	if (dp != NULL) {
1494		nd6_dad_destroytimer(dp);
1495		ifafree(ifa);
1496	}
1497}
1498