nd6_nbr.c revision 273188
1193833Sjhb/*-
2193833Sjhb * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3193833Sjhb * All rights reserved.
4193833Sjhb *
5193833Sjhb * Redistribution and use in source and binary forms, with or without
6193833Sjhb * modification, are permitted provided that the following conditions
7193833Sjhb * are met:
8193833Sjhb * 1. Redistributions of source code must retain the above copyright
9193833Sjhb *    notice, this list of conditions and the following disclaimer.
10193833Sjhb * 2. Redistributions in binary form must reproduce the above copyright
11193833Sjhb *    notice, this list of conditions and the following disclaimer in the
12193833Sjhb *    documentation and/or other materials provided with the distribution.
13193833Sjhb * 3. Neither the name of the project nor the names of its contributors
14193833Sjhb *    may be used to endorse or promote products derived from this software
15193833Sjhb *    without specific prior written permission.
16193833Sjhb *
17193833Sjhb * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18193833Sjhb * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19193833Sjhb * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20193833Sjhb * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21193833Sjhb * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22193833Sjhb * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23193833Sjhb * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24193833Sjhb * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25193833Sjhb * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26193833Sjhb * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27193833Sjhb * SUCH DAMAGE.
28193833Sjhb *
29193833Sjhb *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
30193833Sjhb */
31209031Suqs
32193833Sjhb#include <sys/cdefs.h>
33193833Sjhb__FBSDID("$FreeBSD: releng/10.1/sys/netinet6/nd6_nbr.c 273188 2014-10-16 22:00:24Z hrs $");
34193833Sjhb
35193833Sjhb#include "opt_inet.h"
36193833Sjhb#include "opt_inet6.h"
37193833Sjhb#include "opt_ipsec.h"
38193833Sjhb#include "opt_mpath.h"
39193833Sjhb
40193833Sjhb#include <sys/param.h>
41193833Sjhb#include <sys/systm.h>
42193833Sjhb#include <sys/malloc.h>
43193833Sjhb#include <sys/lock.h>
44193833Sjhb#include <sys/rwlock.h>
45193833Sjhb#include <sys/mbuf.h>
46193833Sjhb#include <sys/socket.h>
47193833Sjhb#include <sys/sockio.h>
48193833Sjhb#include <sys/time.h>
49193833Sjhb#include <sys/kernel.h>
50193833Sjhb#include <sys/errno.h>
51193833Sjhb#include <sys/syslog.h>
52193833Sjhb#include <sys/queue.h>
53193833Sjhb#include <sys/callout.h>
54193833Sjhb
55193833Sjhb#include <net/if.h>
56193833Sjhb#include <net/if_types.h>
57193833Sjhb#include <net/if_dl.h>
58#include <net/if_var.h>
59#include <net/route.h>
60#ifdef RADIX_MPATH
61#include <net/radix_mpath.h>
62#endif
63
64#include <netinet/in.h>
65#include <netinet/in_var.h>
66#include <net/if_llatbl.h>
67#define	L3_ADDR_SIN6(le)	((struct sockaddr_in6 *) L3_ADDR(le))
68#include <netinet6/in6_var.h>
69#include <netinet6/in6_ifattach.h>
70#include <netinet/ip6.h>
71#include <netinet6/ip6_var.h>
72#include <netinet6/scope6_var.h>
73#include <netinet6/nd6.h>
74#include <netinet/icmp6.h>
75#include <netinet/ip_carp.h>
76#include <netinet6/send.h>
77
78#define SDL(s) ((struct sockaddr_dl *)s)
79
80struct dadq;
81static struct dadq *nd6_dad_find(struct ifaddr *);
82static void nd6_dad_starttimer(struct dadq *, int);
83static void nd6_dad_stoptimer(struct dadq *);
84static void nd6_dad_timer(struct dadq *);
85static void nd6_dad_ns_output(struct dadq *, struct ifaddr *);
86static void nd6_dad_ns_input(struct ifaddr *);
87static void nd6_dad_na_input(struct ifaddr *);
88static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
89    const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
90
91static VNET_DEFINE(int, dad_ignore_ns) = 0;	/* ignore NS in DAD
92						   - specwise incorrect */
93static VNET_DEFINE(int, dad_maxtry) = 15;	/* max # of *tries* to
94						   transmit DAD packet */
95#define	V_dad_ignore_ns			VNET(dad_ignore_ns)
96#define	V_dad_maxtry			VNET(dad_maxtry)
97
98/*
99 * Input a Neighbor Solicitation Message.
100 *
101 * Based on RFC 2461
102 * Based on RFC 2462 (duplicate address detection)
103 */
104void
105nd6_ns_input(struct mbuf *m, int off, int icmp6len)
106{
107	struct ifnet *ifp = m->m_pkthdr.rcvif;
108	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
109	struct nd_neighbor_solicit *nd_ns;
110	struct in6_addr saddr6 = ip6->ip6_src;
111	struct in6_addr daddr6 = ip6->ip6_dst;
112	struct in6_addr taddr6;
113	struct in6_addr myaddr6;
114	char *lladdr = NULL;
115	struct ifaddr *ifa = NULL;
116	int lladdrlen = 0;
117	int anycast = 0, proxy = 0, tentative = 0;
118	int tlladdr;
119	int rflag;
120	union nd_opts ndopts;
121	struct sockaddr_dl proxydl;
122	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
123
124	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
125	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
126		rflag = 0;
127#ifndef PULLDOWN_TEST
128	IP6_EXTHDR_CHECK(m, off, icmp6len,);
129	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
130#else
131	IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len);
132	if (nd_ns == NULL) {
133		ICMP6STAT_INC(icp6s_tooshort);
134		return;
135	}
136#endif
137	ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */
138	taddr6 = nd_ns->nd_ns_target;
139	if (in6_setscope(&taddr6, ifp, NULL) != 0)
140		goto bad;
141
142	if (ip6->ip6_hlim != 255) {
143		nd6log((LOG_ERR,
144		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
145		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
146		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
147		goto bad;
148	}
149
150	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
151		/* dst has to be a solicited node multicast address. */
152		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
153		    /* don't check ifindex portion */
154		    daddr6.s6_addr32[1] == 0 &&
155		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
156		    daddr6.s6_addr8[12] == 0xff) {
157			; /* good */
158		} else {
159			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
160			    "(wrong ip6 dst)\n"));
161			goto bad;
162		}
163	} else if (!V_nd6_onlink_ns_rfc4861) {
164		struct sockaddr_in6 src_sa6;
165
166		/*
167		 * According to recent IETF discussions, it is not a good idea
168		 * to accept a NS from an address which would not be deemed
169		 * to be a neighbor otherwise.  This point is expected to be
170		 * clarified in future revisions of the specification.
171		 */
172		bzero(&src_sa6, sizeof(src_sa6));
173		src_sa6.sin6_family = AF_INET6;
174		src_sa6.sin6_len = sizeof(src_sa6);
175		src_sa6.sin6_addr = saddr6;
176		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
177			nd6log((LOG_INFO, "nd6_ns_input: "
178				"NS packet from non-neighbor\n"));
179			goto bad;
180		}
181	}
182
183	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
184		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
185		goto bad;
186	}
187
188	icmp6len -= sizeof(*nd_ns);
189	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
190	if (nd6_options(&ndopts) < 0) {
191		nd6log((LOG_INFO,
192		    "nd6_ns_input: invalid ND option, ignored\n"));
193		/* nd6_options have incremented stats */
194		goto freeit;
195	}
196
197	if (ndopts.nd_opts_src_lladdr) {
198		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
199		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
200	}
201
202	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
203		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
204		    "(link-layer address option)\n"));
205		goto bad;
206	}
207
208	/*
209	 * Attaching target link-layer address to the NA?
210	 * (RFC 2461 7.2.4)
211	 *
212	 * NS IP dst is unicast/anycast			MUST NOT add
213	 * NS IP dst is solicited-node multicast	MUST add
214	 *
215	 * In implementation, we add target link-layer address by default.
216	 * We do not add one in MUST NOT cases.
217	 */
218	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
219		tlladdr = 0;
220	else
221		tlladdr = 1;
222
223	/*
224	 * Target address (taddr6) must be either:
225	 * (1) Valid unicast/anycast address for my receiving interface,
226	 * (2) Unicast address for which I'm offering proxy service, or
227	 * (3) "tentative" address on which DAD is being performed.
228	 */
229	/* (1) and (3) check. */
230	if (ifp->if_carp)
231		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
232	else
233		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
234
235	/* (2) check. */
236	if (ifa == NULL) {
237		struct route_in6 ro;
238		int need_proxy;
239
240		bzero(&ro, sizeof(ro));
241		ro.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
242		ro.ro_dst.sin6_family = AF_INET6;
243		ro.ro_dst.sin6_addr = taddr6;
244
245		/* Always use the default FIB. */
246#ifdef RADIX_MPATH
247		rtalloc_mpath_fib((struct route *)&ro, RTF_ANNOUNCE,
248		    RT_DEFAULT_FIB);
249#else
250		in6_rtalloc(&ro, RT_DEFAULT_FIB);
251#endif
252		need_proxy = (ro.ro_rt &&
253		    (ro.ro_rt->rt_flags & RTF_ANNOUNCE) != 0 &&
254		    ro.ro_rt->rt_gateway->sa_family == AF_LINK);
255		if (ro.ro_rt != NULL) {
256			if (need_proxy)
257				proxydl = *SDL(ro.ro_rt->rt_gateway);
258			RTFREE(ro.ro_rt);
259		}
260		if (need_proxy) {
261			/*
262			 * proxy NDP for single entry
263			 */
264			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
265				IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
266			if (ifa)
267				proxy = 1;
268		}
269	}
270	if (ifa == NULL) {
271		/*
272		 * We've got an NS packet, and we don't have that adddress
273		 * assigned for us.  We MUST silently ignore it.
274		 * See RFC2461 7.2.3.
275		 */
276		goto freeit;
277	}
278	myaddr6 = *IFA_IN6(ifa);
279	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
280	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
281	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
282		goto freeit;
283
284	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
285		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
286		    "(if %d, NS packet %d)\n",
287		    ip6_sprintf(ip6bufs, &taddr6),
288		    ifp->if_addrlen, lladdrlen - 2));
289		goto bad;
290	}
291
292	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
293		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
294		    ip6_sprintf(ip6bufs, &saddr6)));
295		goto freeit;
296	}
297
298	/*
299	 * We have neighbor solicitation packet, with target address equals to
300	 * one of my tentative address.
301	 *
302	 * src addr	how to process?
303	 * ---		---
304	 * multicast	of course, invalid (rejected in ip6_input)
305	 * unicast	somebody is doing address resolution -> ignore
306	 * unspec	dup address detection
307	 *
308	 * The processing is defined in RFC 2462.
309	 */
310	if (tentative) {
311		/*
312		 * If source address is unspecified address, it is for
313		 * duplicate address detection.
314		 *
315		 * If not, the packet is for addess resolution;
316		 * silently ignore it.
317		 */
318		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
319			nd6_dad_ns_input(ifa);
320
321		goto freeit;
322	}
323
324	/*
325	 * If the source address is unspecified address, entries must not
326	 * be created or updated.
327	 * It looks that sender is performing DAD.  Output NA toward
328	 * all-node multicast address, to tell the sender that I'm using
329	 * the address.
330	 * S bit ("solicited") must be zero.
331	 */
332	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
333		struct in6_addr in6_all;
334
335		in6_all = in6addr_linklocal_allnodes;
336		if (in6_setscope(&in6_all, ifp, NULL) != 0)
337			goto bad;
338		nd6_na_output_fib(ifp, &in6_all, &taddr6,
339		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
340		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
341		    M_GETFIB(m));
342		goto freeit;
343	}
344
345	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
346	    ND_NEIGHBOR_SOLICIT, 0);
347
348	nd6_na_output_fib(ifp, &saddr6, &taddr6,
349	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
350	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
351	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
352 freeit:
353	if (ifa != NULL)
354		ifa_free(ifa);
355	m_freem(m);
356	return;
357
358 bad:
359	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
360		ip6_sprintf(ip6bufs, &saddr6)));
361	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
362		ip6_sprintf(ip6bufs, &daddr6)));
363	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
364		ip6_sprintf(ip6bufs, &taddr6)));
365	ICMP6STAT_INC(icp6s_badns);
366	if (ifa != NULL)
367		ifa_free(ifa);
368	m_freem(m);
369}
370
371/*
372 * Output a Neighbor Solicitation Message. Caller specifies:
373 *	- ICMP6 header source IP6 address
374 *	- ND6 header target IP6 address
375 *	- ND6 header source datalink address
376 *
377 * Based on RFC 2461
378 * Based on RFC 2462 (duplicate address detection)
379 *
380 *   ln - for source address determination
381 *  dad - duplicate address detection
382 */
383void
384nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6,
385    const struct in6_addr *taddr6, struct llentry *ln, int dad)
386{
387	struct mbuf *m;
388	struct m_tag *mtag;
389	struct ip6_hdr *ip6;
390	struct nd_neighbor_solicit *nd_ns;
391	struct ip6_moptions im6o;
392	int icmp6len;
393	int maxlen;
394	caddr_t mac;
395	struct route_in6 ro;
396
397	if (IN6_IS_ADDR_MULTICAST(taddr6))
398		return;
399
400	/* estimate the size of message */
401	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
402	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
403	if (max_linkhdr + maxlen >= MCLBYTES) {
404#ifdef DIAGNOSTIC
405		printf("nd6_ns_output: max_linkhdr + maxlen >= MCLBYTES "
406		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
407#endif
408		return;
409	}
410
411	if (max_linkhdr + maxlen > MHLEN)
412		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
413	else
414		m = m_gethdr(M_NOWAIT, MT_DATA);
415	if (m == NULL)
416		return;
417
418	bzero(&ro, sizeof(ro));
419
420	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
421		m->m_flags |= M_MCAST;
422		im6o.im6o_multicast_ifp = ifp;
423		im6o.im6o_multicast_hlim = 255;
424		im6o.im6o_multicast_loop = 0;
425	}
426
427	icmp6len = sizeof(*nd_ns);
428	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
429	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
430
431	/* fill neighbor solicitation packet */
432	ip6 = mtod(m, struct ip6_hdr *);
433	ip6->ip6_flow = 0;
434	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
435	ip6->ip6_vfc |= IPV6_VERSION;
436	/* ip6->ip6_plen will be set later */
437	ip6->ip6_nxt = IPPROTO_ICMPV6;
438	ip6->ip6_hlim = 255;
439	if (daddr6)
440		ip6->ip6_dst = *daddr6;
441	else {
442		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
443		ip6->ip6_dst.s6_addr16[1] = 0;
444		ip6->ip6_dst.s6_addr32[1] = 0;
445		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
446		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
447		ip6->ip6_dst.s6_addr8[12] = 0xff;
448		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
449			goto bad;
450	}
451	if (!dad) {
452		struct ifaddr *ifa;
453
454		/*
455		 * RFC2461 7.2.2:
456		 * "If the source address of the packet prompting the
457		 * solicitation is the same as one of the addresses assigned
458		 * to the outgoing interface, that address SHOULD be placed
459		 * in the IP Source Address of the outgoing solicitation.
460		 * Otherwise, any one of the addresses assigned to the
461		 * interface should be used."
462		 *
463		 * We use the source address for the prompting packet
464		 * (saddr6), if:
465		 * - saddr6 is given from the caller (by giving "ln"), and
466		 * - saddr6 belongs to the outgoing interface.
467		 * Otherwise, we perform the source address selection as usual.
468		 */
469		struct in6_addr *hsrc;
470
471		hsrc = NULL;
472		if (ln != NULL) {
473			LLE_RLOCK(ln);
474			if (ln->la_hold != NULL) {
475				struct ip6_hdr *hip6;		/* hold ip6 */
476
477				/*
478				 * assuming every packet in la_hold has the same IP
479				 * header
480				 */
481				hip6 = mtod(ln->la_hold, struct ip6_hdr *);
482				/* XXX pullup? */
483				if (sizeof(*hip6) < ln->la_hold->m_len) {
484					ip6->ip6_src = hip6->ip6_src;
485					hsrc = &hip6->ip6_src;
486				}
487			}
488			LLE_RUNLOCK(ln);
489		}
490		if (hsrc && (ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
491		    hsrc)) != NULL) {
492			/* ip6_src set already. */
493			ifa_free(ifa);
494		} else {
495			int error;
496			struct sockaddr_in6 dst_sa;
497			struct in6_addr src_in;
498			struct ifnet *oifp;
499
500			bzero(&dst_sa, sizeof(dst_sa));
501			dst_sa.sin6_family = AF_INET6;
502			dst_sa.sin6_len = sizeof(dst_sa);
503			dst_sa.sin6_addr = ip6->ip6_dst;
504
505			oifp = ifp;
506			error = in6_selectsrc(&dst_sa, NULL,
507			    NULL, &ro, NULL, &oifp, &src_in);
508			if (error) {
509				char ip6buf[INET6_ADDRSTRLEN];
510				nd6log((LOG_DEBUG,
511				    "nd6_ns_output: source can't be "
512				    "determined: dst=%s, error=%d\n",
513				    ip6_sprintf(ip6buf, &dst_sa.sin6_addr),
514				    error));
515				goto bad;
516			}
517			ip6->ip6_src = src_in;
518		}
519	} else {
520		/*
521		 * Source address for DAD packet must always be IPv6
522		 * unspecified address. (0::0)
523		 * We actually don't have to 0-clear the address (we did it
524		 * above), but we do so here explicitly to make the intention
525		 * clearer.
526		 */
527		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
528	}
529	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
530	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
531	nd_ns->nd_ns_code = 0;
532	nd_ns->nd_ns_reserved = 0;
533	nd_ns->nd_ns_target = *taddr6;
534	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
535
536	/*
537	 * Add source link-layer address option.
538	 *
539	 *				spec		implementation
540	 *				---		---
541	 * DAD packet			MUST NOT	do not add the option
542	 * there's no link layer address:
543	 *				impossible	do not add the option
544	 * there's link layer address:
545	 *	Multicast NS		MUST add one	add the option
546	 *	Unicast NS		SHOULD add one	add the option
547	 */
548	if (!dad && (mac = nd6_ifptomac(ifp))) {
549		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
550		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
551		/* 8 byte alignments... */
552		optlen = (optlen + 7) & ~7;
553
554		m->m_pkthdr.len += optlen;
555		m->m_len += optlen;
556		icmp6len += optlen;
557		bzero((caddr_t)nd_opt, optlen);
558		nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
559		nd_opt->nd_opt_len = optlen >> 3;
560		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
561	}
562
563	ip6->ip6_plen = htons((u_short)icmp6len);
564	nd_ns->nd_ns_cksum = 0;
565	nd_ns->nd_ns_cksum =
566	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
567
568	if (send_sendso_input_hook != NULL) {
569		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
570			sizeof(unsigned short), M_NOWAIT);
571		if (mtag == NULL)
572			goto bad;
573		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
574		m_tag_prepend(m, mtag);
575	}
576
577	ip6_output(m, NULL, &ro, dad ? IPV6_UNSPECSRC : 0, &im6o, NULL, NULL);
578	icmp6_ifstat_inc(ifp, ifs6_out_msg);
579	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
580	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
581
582	/* We don't cache this route. */
583	RO_RTFREE(&ro);
584
585	return;
586
587  bad:
588	if (ro.ro_rt) {
589		RTFREE(ro.ro_rt);
590	}
591	m_freem(m);
592	return;
593}
594
595/*
596 * Neighbor advertisement input handling.
597 *
598 * Based on RFC 2461
599 * Based on RFC 2462 (duplicate address detection)
600 *
601 * the following items are not implemented yet:
602 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
603 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
604 */
605void
606nd6_na_input(struct mbuf *m, int off, int icmp6len)
607{
608	struct ifnet *ifp = m->m_pkthdr.rcvif;
609	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
610	struct nd_neighbor_advert *nd_na;
611	struct in6_addr daddr6 = ip6->ip6_dst;
612	struct in6_addr taddr6;
613	int flags;
614	int is_router;
615	int is_solicited;
616	int is_override;
617	char *lladdr = NULL;
618	int lladdrlen = 0;
619	int checklink = 0;
620	struct ifaddr *ifa;
621	struct llentry *ln = NULL;
622	union nd_opts ndopts;
623	struct mbuf *chain = NULL;
624	struct m_tag *mtag;
625	struct sockaddr_in6 sin6;
626	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
627
628	if (ip6->ip6_hlim != 255) {
629		nd6log((LOG_ERR,
630		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
631		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
632		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
633		goto bad;
634	}
635
636#ifndef PULLDOWN_TEST
637	IP6_EXTHDR_CHECK(m, off, icmp6len,);
638	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
639#else
640	IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len);
641	if (nd_na == NULL) {
642		ICMP6STAT_INC(icp6s_tooshort);
643		return;
644	}
645#endif
646
647	flags = nd_na->nd_na_flags_reserved;
648	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
649	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
650	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
651
652	taddr6 = nd_na->nd_na_target;
653	if (in6_setscope(&taddr6, ifp, NULL))
654		goto bad;	/* XXX: impossible */
655
656	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
657		nd6log((LOG_ERR,
658		    "nd6_na_input: invalid target address %s\n",
659		    ip6_sprintf(ip6bufs, &taddr6)));
660		goto bad;
661	}
662	if (IN6_IS_ADDR_MULTICAST(&daddr6))
663		if (is_solicited) {
664			nd6log((LOG_ERR,
665			    "nd6_na_input: a solicited adv is multicasted\n"));
666			goto bad;
667		}
668
669	icmp6len -= sizeof(*nd_na);
670	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
671	if (nd6_options(&ndopts) < 0) {
672		nd6log((LOG_INFO,
673		    "nd6_na_input: invalid ND option, ignored\n"));
674		/* nd6_options have incremented stats */
675		goto freeit;
676	}
677
678	if (ndopts.nd_opts_tgt_lladdr) {
679		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
680		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
681	}
682
683	/*
684	 * This effectively disables the DAD check on a non-master CARP
685	 * address.
686	 */
687	if (ifp->if_carp)
688		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
689	else
690		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
691
692	/*
693	 * Target address matches one of my interface address.
694	 *
695	 * If my address is tentative, this means that there's somebody
696	 * already using the same address as mine.  This indicates DAD failure.
697	 * This is defined in RFC 2462.
698	 *
699	 * Otherwise, process as defined in RFC 2461.
700	 */
701	if (ifa
702	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
703		ifa_free(ifa);
704		nd6_dad_na_input(ifa);
705		goto freeit;
706	}
707
708	/* Just for safety, maybe unnecessary. */
709	if (ifa) {
710		ifa_free(ifa);
711		log(LOG_ERR,
712		    "nd6_na_input: duplicate IP6 address %s\n",
713		    ip6_sprintf(ip6bufs, &taddr6));
714		goto freeit;
715	}
716
717	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
718		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
719		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
720		    ifp->if_addrlen, lladdrlen - 2));
721		goto bad;
722	}
723
724	/*
725	 * If no neighbor cache entry is found, NA SHOULD silently be
726	 * discarded.
727	 */
728	IF_AFDATA_RLOCK(ifp);
729	ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp);
730	IF_AFDATA_RUNLOCK(ifp);
731	if (ln == NULL) {
732		goto freeit;
733	}
734
735	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
736		/*
737		 * If the link-layer has address, and no lladdr option came,
738		 * discard the packet.
739		 */
740		if (ifp->if_addrlen && lladdr == NULL) {
741			goto freeit;
742		}
743
744		/*
745		 * Record link-layer address, and update the state.
746		 */
747		bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
748		ln->la_flags |= LLE_VALID;
749		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
750		if (is_solicited) {
751			ln->ln_state = ND6_LLINFO_REACHABLE;
752			ln->ln_byhint = 0;
753			if (!ND6_LLINFO_PERMANENT(ln)) {
754				nd6_llinfo_settimer_locked(ln,
755				    (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz);
756			}
757		} else {
758			ln->ln_state = ND6_LLINFO_STALE;
759			nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
760		}
761		if ((ln->ln_router = is_router) != 0) {
762			/*
763			 * This means a router's state has changed from
764			 * non-reachable to probably reachable, and might
765			 * affect the status of associated prefixes..
766			 */
767			checklink = 1;
768		}
769	} else {
770		int llchange;
771
772		/*
773		 * Check if the link-layer address has changed or not.
774		 */
775		if (lladdr == NULL)
776			llchange = 0;
777		else {
778			if (ln->la_flags & LLE_VALID) {
779				if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen))
780					llchange = 1;
781				else
782					llchange = 0;
783			} else
784				llchange = 1;
785		}
786
787		/*
788		 * This is VERY complex.  Look at it with care.
789		 *
790		 * override solicit lladdr llchange	action
791		 *					(L: record lladdr)
792		 *
793		 *	0	0	n	--	(2c)
794		 *	0	0	y	n	(2b) L
795		 *	0	0	y	y	(1)    REACHABLE->STALE
796		 *	0	1	n	--	(2c)   *->REACHABLE
797		 *	0	1	y	n	(2b) L *->REACHABLE
798		 *	0	1	y	y	(1)    REACHABLE->STALE
799		 *	1	0	n	--	(2a)
800		 *	1	0	y	n	(2a) L
801		 *	1	0	y	y	(2a) L *->STALE
802		 *	1	1	n	--	(2a)   *->REACHABLE
803		 *	1	1	y	n	(2a) L *->REACHABLE
804		 *	1	1	y	y	(2a) L *->REACHABLE
805		 */
806		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
807			/*
808			 * If state is REACHABLE, make it STALE.
809			 * no other updates should be done.
810			 */
811			if (ln->ln_state == ND6_LLINFO_REACHABLE) {
812				ln->ln_state = ND6_LLINFO_STALE;
813				nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
814			}
815			goto freeit;
816		} else if (is_override				   /* (2a) */
817			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
818			|| lladdr == NULL) {			   /* (2c) */
819			/*
820			 * Update link-local address, if any.
821			 */
822			if (lladdr != NULL) {
823				bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
824				ln->la_flags |= LLE_VALID;
825				EVENTHANDLER_INVOKE(lle_event, ln,
826				    LLENTRY_RESOLVED);
827			}
828
829			/*
830			 * If solicited, make the state REACHABLE.
831			 * If not solicited and the link-layer address was
832			 * changed, make it STALE.
833			 */
834			if (is_solicited) {
835				ln->ln_state = ND6_LLINFO_REACHABLE;
836				ln->ln_byhint = 0;
837				if (!ND6_LLINFO_PERMANENT(ln)) {
838					nd6_llinfo_settimer_locked(ln,
839					    (long)ND_IFINFO(ifp)->reachable * hz);
840				}
841			} else {
842				if (lladdr != NULL && llchange) {
843					ln->ln_state = ND6_LLINFO_STALE;
844					nd6_llinfo_settimer_locked(ln,
845					    (long)V_nd6_gctimer * hz);
846				}
847			}
848		}
849
850		if (ln->ln_router && !is_router) {
851			/*
852			 * The peer dropped the router flag.
853			 * Remove the sender from the Default Router List and
854			 * update the Destination Cache entries.
855			 */
856			struct nd_defrouter *dr;
857			struct in6_addr *in6;
858
859			in6 = &L3_ADDR_SIN6(ln)->sin6_addr;
860
861			/*
862			 * Lock to protect the default router list.
863			 * XXX: this might be unnecessary, since this function
864			 * is only called under the network software interrupt
865			 * context.  However, we keep it just for safety.
866			 */
867			dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp);
868			if (dr)
869				defrtrlist_del(dr);
870			else if (ND_IFINFO(ln->lle_tbl->llt_ifp)->flags &
871			    ND6_IFF_ACCEPT_RTADV) {
872				/*
873				 * Even if the neighbor is not in the default
874				 * router list, the neighbor may be used
875				 * as a next hop for some destinations
876				 * (e.g. redirect case). So we must
877				 * call rt6_flush explicitly.
878				 */
879				rt6_flush(&ip6->ip6_src, ifp);
880			}
881		}
882		ln->ln_router = is_router;
883	}
884        /* XXX - QL
885	 *  Does this matter?
886	 *  rt->rt_flags &= ~RTF_REJECT;
887	 */
888	ln->la_asked = 0;
889	if (ln->la_hold) {
890		struct mbuf *m_hold, *m_hold_next;
891
892		/*
893		 * reset the la_hold in advance, to explicitly
894		 * prevent a la_hold lookup in nd6_output()
895		 * (wouldn't happen, though...)
896		 */
897		for (m_hold = ln->la_hold, ln->la_hold = NULL;
898		    m_hold; m_hold = m_hold_next) {
899			m_hold_next = m_hold->m_nextpkt;
900			m_hold->m_nextpkt = NULL;
901			/*
902			 * we assume ifp is not a loopback here, so just set
903			 * the 2nd argument as the 1st one.
904			 */
905
906			if (send_sendso_input_hook != NULL) {
907				mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
908				    sizeof(unsigned short), M_NOWAIT);
909				if (mtag == NULL)
910					goto bad;
911				m_tag_prepend(m, mtag);
912			}
913
914			nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
915		}
916	}
917 freeit:
918	if (ln != NULL) {
919		if (chain)
920			memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
921		LLE_WUNLOCK(ln);
922
923		if (chain)
924			nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
925	}
926	if (checklink)
927		pfxlist_onlink_check();
928
929	m_freem(m);
930	return;
931
932 bad:
933	if (ln != NULL)
934		LLE_WUNLOCK(ln);
935
936	ICMP6STAT_INC(icp6s_badna);
937	m_freem(m);
938}
939
940/*
941 * Neighbor advertisement output handling.
942 *
943 * Based on RFC 2461
944 *
945 * the following items are not implemented yet:
946 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
947 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
948 *
949 * tlladdr - 1 if include target link-layer address
950 * sdl0 - sockaddr_dl (= proxy NA) or NULL
951 */
952static void
953nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
954    const struct in6_addr *taddr6, u_long flags, int tlladdr,
955    struct sockaddr *sdl0, u_int fibnum)
956{
957	struct mbuf *m;
958	struct m_tag *mtag;
959	struct ifnet *oifp;
960	struct ip6_hdr *ip6;
961	struct nd_neighbor_advert *nd_na;
962	struct ip6_moptions im6o;
963	struct in6_addr src, daddr6;
964	struct sockaddr_in6 dst_sa;
965	int icmp6len, maxlen, error;
966	caddr_t mac = NULL;
967	struct route_in6 ro;
968
969	bzero(&ro, sizeof(ro));
970
971	daddr6 = *daddr6_0;	/* make a local copy for modification */
972
973	/* estimate the size of message */
974	maxlen = sizeof(*ip6) + sizeof(*nd_na);
975	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
976	if (max_linkhdr + maxlen >= MCLBYTES) {
977#ifdef DIAGNOSTIC
978		printf("nd6_na_output: max_linkhdr + maxlen >= MCLBYTES "
979		    "(%d + %d > %d)\n", max_linkhdr, maxlen, MCLBYTES);
980#endif
981		return;
982	}
983
984	if (max_linkhdr + maxlen > MHLEN)
985		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
986	else
987		m = m_gethdr(M_NOWAIT, MT_DATA);
988	if (m == NULL)
989		return;
990	M_SETFIB(m, fibnum);
991
992	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
993		m->m_flags |= M_MCAST;
994		im6o.im6o_multicast_ifp = ifp;
995		im6o.im6o_multicast_hlim = 255;
996		im6o.im6o_multicast_loop = 0;
997	}
998
999	icmp6len = sizeof(*nd_na);
1000	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
1001	m->m_data += max_linkhdr;	/* or MH_ALIGN() equivalent? */
1002
1003	/* fill neighbor advertisement packet */
1004	ip6 = mtod(m, struct ip6_hdr *);
1005	ip6->ip6_flow = 0;
1006	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1007	ip6->ip6_vfc |= IPV6_VERSION;
1008	ip6->ip6_nxt = IPPROTO_ICMPV6;
1009	ip6->ip6_hlim = 255;
1010	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
1011		/* reply to DAD */
1012		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
1013		daddr6.s6_addr16[1] = 0;
1014		daddr6.s6_addr32[1] = 0;
1015		daddr6.s6_addr32[2] = 0;
1016		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
1017		if (in6_setscope(&daddr6, ifp, NULL))
1018			goto bad;
1019
1020		flags &= ~ND_NA_FLAG_SOLICITED;
1021	}
1022	ip6->ip6_dst = daddr6;
1023	bzero(&dst_sa, sizeof(struct sockaddr_in6));
1024	dst_sa.sin6_family = AF_INET6;
1025	dst_sa.sin6_len = sizeof(struct sockaddr_in6);
1026	dst_sa.sin6_addr = daddr6;
1027
1028	/*
1029	 * Select a source whose scope is the same as that of the dest.
1030	 */
1031	bcopy(&dst_sa, &ro.ro_dst, sizeof(dst_sa));
1032	oifp = ifp;
1033	error = in6_selectsrc(&dst_sa, NULL, NULL, &ro, NULL, &oifp, &src);
1034	if (error) {
1035		char ip6buf[INET6_ADDRSTRLEN];
1036		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1037		    "determined: dst=%s, error=%d\n",
1038		    ip6_sprintf(ip6buf, &dst_sa.sin6_addr), error));
1039		goto bad;
1040	}
1041	ip6->ip6_src = src;
1042	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1043	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1044	nd_na->nd_na_code = 0;
1045	nd_na->nd_na_target = *taddr6;
1046	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1047
1048	/*
1049	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1050	 * see nd6_ns_input() for details.
1051	 * Basically, if NS packet is sent to unicast/anycast addr,
1052	 * target lladdr option SHOULD NOT be included.
1053	 */
1054	if (tlladdr) {
1055		/*
1056		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1057		 * lladdr in sdl0.  If we are not proxying (sending NA for
1058		 * my address) use lladdr configured for the interface.
1059		 */
1060		if (sdl0 == NULL) {
1061			if (ifp->if_carp)
1062				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1063			if (mac == NULL)
1064				mac = nd6_ifptomac(ifp);
1065		} else if (sdl0->sa_family == AF_LINK) {
1066			struct sockaddr_dl *sdl;
1067			sdl = (struct sockaddr_dl *)sdl0;
1068			if (sdl->sdl_alen == ifp->if_addrlen)
1069				mac = LLADDR(sdl);
1070		}
1071	}
1072	if (tlladdr && mac) {
1073		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1074		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1075
1076		/* roundup to 8 bytes alignment! */
1077		optlen = (optlen + 7) & ~7;
1078
1079		m->m_pkthdr.len += optlen;
1080		m->m_len += optlen;
1081		icmp6len += optlen;
1082		bzero((caddr_t)nd_opt, optlen);
1083		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1084		nd_opt->nd_opt_len = optlen >> 3;
1085		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1086	} else
1087		flags &= ~ND_NA_FLAG_OVERRIDE;
1088
1089	ip6->ip6_plen = htons((u_short)icmp6len);
1090	nd_na->nd_na_flags_reserved = flags;
1091	nd_na->nd_na_cksum = 0;
1092	nd_na->nd_na_cksum =
1093	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1094
1095	if (send_sendso_input_hook != NULL) {
1096		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1097		    sizeof(unsigned short), M_NOWAIT);
1098		if (mtag == NULL)
1099			goto bad;
1100		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1101		m_tag_prepend(m, mtag);
1102	}
1103
1104	ip6_output(m, NULL, &ro, 0, &im6o, NULL, NULL);
1105	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1106	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1107	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1108
1109	/* We don't cache this route. */
1110	RO_RTFREE(&ro);
1111
1112	return;
1113
1114  bad:
1115	if (ro.ro_rt) {
1116		RTFREE(ro.ro_rt);
1117	}
1118	m_freem(m);
1119	return;
1120}
1121
1122#ifndef BURN_BRIDGES
1123void
1124nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1125    const struct in6_addr *taddr6, u_long flags, int tlladdr,
1126    struct sockaddr *sdl0)
1127{
1128
1129	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1130	    RT_DEFAULT_FIB);
1131}
1132#endif
1133
1134caddr_t
1135nd6_ifptomac(struct ifnet *ifp)
1136{
1137	switch (ifp->if_type) {
1138	case IFT_ARCNET:
1139	case IFT_ETHER:
1140	case IFT_FDDI:
1141	case IFT_IEEE1394:
1142#ifdef IFT_L2VLAN
1143	case IFT_L2VLAN:
1144#endif
1145#ifdef IFT_IEEE80211
1146	case IFT_IEEE80211:
1147#endif
1148	case IFT_INFINIBAND:
1149	case IFT_BRIDGE:
1150	case IFT_ISO88025:
1151		return IF_LLADDR(ifp);
1152	default:
1153		return NULL;
1154	}
1155}
1156
1157struct dadq {
1158	TAILQ_ENTRY(dadq) dad_list;
1159	struct ifaddr *dad_ifa;
1160	int dad_count;		/* max NS to send */
1161	int dad_ns_tcount;	/* # of trials to send NS */
1162	int dad_ns_ocount;	/* NS sent so far */
1163	int dad_ns_icount;
1164	int dad_na_icount;
1165	struct callout dad_timer_ch;
1166	struct vnet *dad_vnet;
1167};
1168
1169static VNET_DEFINE(TAILQ_HEAD(, dadq), dadq);
1170static VNET_DEFINE(struct rwlock, dad_rwlock);
1171#define	V_dadq			VNET(dadq)
1172#define	V_dad_rwlock		VNET(dad_rwlock)
1173
1174#define	DADQ_LOCK_INIT()	rw_init(&V_dad_rwlock, "nd6 DAD queue")
1175#define	DADQ_LOCK_DESTROY()	rw_destroy(&V_dad_rwlock)
1176#define	DADQ_LOCK_INITIALIZED()	rw_initialized(&V_dad_rwlock)
1177#define	DADQ_RLOCK()		rw_rlock(&V_dad_rwlock)
1178#define	DADQ_RUNLOCK()		rw_runlock(&V_dad_rwlock)
1179#define	DADQ_WLOCK()		rw_wlock(&V_dad_rwlock)
1180#define	DADQ_WUNLOCK()		rw_wunlock(&V_dad_rwlock)
1181
1182static struct dadq *
1183nd6_dad_find(struct ifaddr *ifa)
1184{
1185	struct dadq *dp;
1186
1187	DADQ_RLOCK();
1188	TAILQ_FOREACH(dp, &V_dadq, dad_list)
1189		if (dp->dad_ifa == ifa)
1190			break;
1191	DADQ_RUNLOCK();
1192
1193	return (dp);
1194}
1195
1196static void
1197nd6_dad_starttimer(struct dadq *dp, int ticks)
1198{
1199
1200	callout_reset(&dp->dad_timer_ch, ticks,
1201	    (void (*)(void *))nd6_dad_timer, (void *)dp);
1202}
1203
1204static void
1205nd6_dad_stoptimer(struct dadq *dp)
1206{
1207
1208	callout_stop(&dp->dad_timer_ch);
1209}
1210
1211/*
1212 * Start Duplicate Address Detection (DAD) for specified interface address.
1213 */
1214void
1215nd6_dad_start(struct ifaddr *ifa, int delay)
1216{
1217	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1218	struct dadq *dp;
1219	char ip6buf[INET6_ADDRSTRLEN];
1220
1221	if (DADQ_LOCK_INITIALIZED() == 0) {
1222		DADQ_LOCK_INIT();
1223		TAILQ_INIT(&V_dadq);
1224	}
1225
1226	/*
1227	 * If we don't need DAD, don't do it.
1228	 * There are several cases:
1229	 * - DAD is disabled (ip6_dad_count == 0)
1230	 * - the interface address is anycast
1231	 */
1232	if (!(ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1233		log(LOG_DEBUG,
1234			"nd6_dad_start: called with non-tentative address "
1235			"%s(%s)\n",
1236			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1237			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1238		return;
1239	}
1240	if (ia->ia6_flags & IN6_IFF_ANYCAST) {
1241		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1242		return;
1243	}
1244	if (!V_ip6_dad_count) {
1245		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1246		return;
1247	}
1248	if (ifa->ifa_ifp == NULL)
1249		panic("nd6_dad_start: ifa->ifa_ifp == NULL");
1250	if (!(ifa->ifa_ifp->if_flags & IFF_UP)) {
1251		return;
1252	}
1253	if (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED)
1254		return;
1255	if (nd6_dad_find(ifa) != NULL) {
1256		/* DAD already in progress */
1257		return;
1258	}
1259
1260	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT);
1261	if (dp == NULL) {
1262		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1263			"%s(%s)\n",
1264			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1265			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1266		return;
1267	}
1268	bzero(dp, sizeof(*dp));
1269	callout_init(&dp->dad_timer_ch, 0);
1270#ifdef VIMAGE
1271	dp->dad_vnet = curvnet;
1272#endif
1273	DADQ_WLOCK();
1274	TAILQ_INSERT_TAIL(&V_dadq, (struct dadq *)dp, dad_list);
1275	DADQ_WUNLOCK();
1276
1277	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1278	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1279
1280	/*
1281	 * Send NS packet for DAD, ip6_dad_count times.
1282	 * Note that we must delay the first transmission, if this is the
1283	 * first packet to be sent from the interface after interface
1284	 * (re)initialization.
1285	 */
1286	dp->dad_ifa = ifa;
1287	ifa_ref(ifa);	/* just for safety */
1288	dp->dad_count = V_ip6_dad_count;
1289	dp->dad_ns_icount = dp->dad_na_icount = 0;
1290	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1291	if (delay == 0) {
1292		nd6_dad_ns_output(dp, ifa);
1293		nd6_dad_starttimer(dp,
1294		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1295	} else {
1296		nd6_dad_starttimer(dp, delay);
1297	}
1298}
1299
1300/*
1301 * terminate DAD unconditionally.  used for address removals.
1302 */
1303void
1304nd6_dad_stop(struct ifaddr *ifa)
1305{
1306	struct dadq *dp;
1307
1308	if (DADQ_LOCK_INITIALIZED() == 0)
1309		return;
1310	dp = nd6_dad_find(ifa);
1311	if (!dp) {
1312		/* DAD wasn't started yet */
1313		return;
1314	}
1315
1316	nd6_dad_stoptimer(dp);
1317
1318	DADQ_WLOCK();
1319	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1320	DADQ_WUNLOCK();
1321	free(dp, M_IP6NDP);
1322	dp = NULL;
1323	ifa_free(ifa);
1324}
1325
1326static void
1327nd6_dad_timer(struct dadq *dp)
1328{
1329	CURVNET_SET(dp->dad_vnet);
1330	struct ifaddr *ifa = dp->dad_ifa;
1331	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1332	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1333	char ip6buf[INET6_ADDRSTRLEN];
1334
1335	/* Sanity check */
1336	if (ia == NULL) {
1337		log(LOG_ERR, "nd6_dad_timer: called with null parameter\n");
1338		goto done;
1339	}
1340	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1341		/* Do not need DAD for ifdisabled interface. */
1342		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1343		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1344		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1345		free(dp, M_IP6NDP);
1346		dp = NULL;
1347		ifa_free(ifa);
1348		goto done;
1349	}
1350	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1351		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1352			"%s(%s)\n",
1353			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1354			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1355		goto done;
1356	}
1357	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1358		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1359			"%s(%s)\n",
1360			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1361			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1362		goto done;
1363	}
1364
1365	/* timeouted with IFF_{RUNNING,UP} check */
1366	if (dp->dad_ns_tcount > V_dad_maxtry) {
1367		nd6log((LOG_INFO, "%s: could not run DAD, driver problem?\n",
1368		    if_name(ifa->ifa_ifp)));
1369
1370		DADQ_WLOCK();
1371		TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1372		DADQ_WUNLOCK();
1373		free(dp, M_IP6NDP);
1374		dp = NULL;
1375		ifa_free(ifa);
1376		goto done;
1377	}
1378
1379	/* Need more checks? */
1380	if (dp->dad_ns_ocount < dp->dad_count) {
1381		/*
1382		 * We have more NS to go.  Send NS packet for DAD.
1383		 */
1384		nd6_dad_ns_output(dp, ifa);
1385		nd6_dad_starttimer(dp,
1386		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1387	} else {
1388		/*
1389		 * We have transmitted sufficient number of DAD packets.
1390		 * See what we've got.
1391		 */
1392		int duplicate;
1393
1394		duplicate = 0;
1395
1396		if (dp->dad_na_icount) {
1397			/*
1398			 * the check is in nd6_dad_na_input(),
1399			 * but just in case
1400			 */
1401			duplicate++;
1402		}
1403
1404		if (dp->dad_ns_icount) {
1405			/* We've seen NS, means DAD has failed. */
1406			duplicate++;
1407		}
1408
1409		if (duplicate) {
1410			/* (*dp) will be freed in nd6_dad_duplicated() */
1411			dp = NULL;
1412			nd6_dad_duplicated(ifa);
1413		} else {
1414			/*
1415			 * We are done with DAD.  No NA came, no NS came.
1416			 * No duplicate address found.  Check IFDISABLED flag
1417			 * again in case that it is changed between the
1418			 * beginning of this function and here.
1419			 */
1420			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1421				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1422
1423			nd6log((LOG_DEBUG,
1424			    "%s: DAD complete for %s - no duplicates found\n",
1425			    if_name(ifa->ifa_ifp),
1426			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1427
1428			DADQ_WLOCK();
1429			TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1430			DADQ_WUNLOCK();
1431			free(dp, M_IP6NDP);
1432			dp = NULL;
1433			ifa_free(ifa);
1434		}
1435	}
1436
1437done:
1438	CURVNET_RESTORE();
1439}
1440
1441void
1442nd6_dad_duplicated(struct ifaddr *ifa)
1443{
1444	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1445	struct ifnet *ifp;
1446	struct dadq *dp;
1447	char ip6buf[INET6_ADDRSTRLEN];
1448
1449	dp = nd6_dad_find(ifa);
1450	if (dp == NULL) {
1451		log(LOG_ERR, "nd6_dad_duplicated: DAD structure not found\n");
1452		return;
1453	}
1454
1455	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1456	    "NS in/out=%d/%d, NA in=%d\n",
1457	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1458	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_na_icount);
1459
1460	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1461	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1462
1463	/* We are done with DAD, with duplicate address found. (failure) */
1464	nd6_dad_stoptimer(dp);
1465
1466	ifp = ifa->ifa_ifp;
1467	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1468	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1469	log(LOG_ERR, "%s: manual intervention required\n",
1470	    if_name(ifp));
1471
1472	/*
1473	 * If the address is a link-local address formed from an interface
1474	 * identifier based on the hardware address which is supposed to be
1475	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1476	 * operation on the interface SHOULD be disabled.
1477	 * [RFC 4862, Section 5.4.5]
1478	 */
1479	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1480		struct in6_addr in6;
1481
1482		/*
1483		 * To avoid over-reaction, we only apply this logic when we are
1484		 * very sure that hardware addresses are supposed to be unique.
1485		 */
1486		switch (ifp->if_type) {
1487		case IFT_ETHER:
1488		case IFT_FDDI:
1489		case IFT_ATM:
1490		case IFT_IEEE1394:
1491#ifdef IFT_IEEE80211
1492		case IFT_IEEE80211:
1493#endif
1494		case IFT_INFINIBAND:
1495			in6 = ia->ia_addr.sin6_addr;
1496			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1497			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1498				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1499				log(LOG_ERR, "%s: possible hardware address "
1500				    "duplication detected, disable IPv6\n",
1501				    if_name(ifp));
1502			}
1503			break;
1504		}
1505	}
1506
1507	DADQ_WLOCK();
1508	TAILQ_REMOVE(&V_dadq, (struct dadq *)dp, dad_list);
1509	DADQ_WUNLOCK();
1510	free(dp, M_IP6NDP);
1511	dp = NULL;
1512	ifa_free(ifa);
1513}
1514
1515static void
1516nd6_dad_ns_output(struct dadq *dp, struct ifaddr *ifa)
1517{
1518	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1519	struct ifnet *ifp = ifa->ifa_ifp;
1520
1521	dp->dad_ns_tcount++;
1522	if ((ifp->if_flags & IFF_UP) == 0) {
1523		return;
1524	}
1525	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1526		return;
1527	}
1528
1529	dp->dad_ns_ocount++;
1530	nd6_ns_output(ifp, NULL, &ia->ia_addr.sin6_addr, NULL, 1);
1531}
1532
1533static void
1534nd6_dad_ns_input(struct ifaddr *ifa)
1535{
1536	struct in6_ifaddr *ia;
1537	struct ifnet *ifp;
1538	const struct in6_addr *taddr6;
1539	struct dadq *dp;
1540	int duplicate;
1541
1542	if (ifa == NULL)
1543		panic("ifa == NULL in nd6_dad_ns_input");
1544
1545	ia = (struct in6_ifaddr *)ifa;
1546	ifp = ifa->ifa_ifp;
1547	taddr6 = &ia->ia_addr.sin6_addr;
1548	duplicate = 0;
1549	dp = nd6_dad_find(ifa);
1550
1551	/* Quickhack - completely ignore DAD NS packets */
1552	if (V_dad_ignore_ns) {
1553		char ip6buf[INET6_ADDRSTRLEN];
1554		nd6log((LOG_INFO,
1555		    "nd6_dad_ns_input: ignoring DAD NS packet for "
1556		    "address %s(%s)\n", ip6_sprintf(ip6buf, taddr6),
1557		    if_name(ifa->ifa_ifp)));
1558		return;
1559	}
1560
1561	/*
1562	 * if I'm yet to start DAD, someone else started using this address
1563	 * first.  I have a duplicate and you win.
1564	 */
1565	if (dp == NULL || dp->dad_ns_ocount == 0)
1566		duplicate++;
1567
1568	/* XXX more checks for loopback situation - see nd6_dad_timer too */
1569
1570	if (duplicate) {
1571		dp = NULL;	/* will be freed in nd6_dad_duplicated() */
1572		nd6_dad_duplicated(ifa);
1573	} else {
1574		/*
1575		 * not sure if I got a duplicate.
1576		 * increment ns count and see what happens.
1577		 */
1578		if (dp)
1579			dp->dad_ns_icount++;
1580	}
1581}
1582
1583static void
1584nd6_dad_na_input(struct ifaddr *ifa)
1585{
1586	struct dadq *dp;
1587
1588	if (ifa == NULL)
1589		panic("ifa == NULL in nd6_dad_na_input");
1590
1591	dp = nd6_dad_find(ifa);
1592	if (dp)
1593		dp->dad_na_icount++;
1594
1595	/* remove the address. */
1596	nd6_dad_duplicated(ifa);
1597}
1598