xref: /freebsd-13-stable/sys/netinet6/nd6_nbr.c (revision e347b72a3914be25ff77f987511165fc1af4562c)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $
32  */
33 
34 #include <sys/cdefs.h>
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/eventhandler.h>
42 #include <sys/malloc.h>
43 #include <sys/libkern.h>
44 #include <sys/lock.h>
45 #include <sys/rwlock.h>
46 #include <sys/mbuf.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/time.h>
50 #include <sys/kernel.h>
51 #include <sys/errno.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/queue.h>
55 #include <sys/callout.h>
56 #include <sys/refcount.h>
57 
58 #include <net/if.h>
59 #include <net/if_types.h>
60 #include <net/if_dl.h>
61 #include <net/if_var.h>
62 #include <net/route.h>
63 #include <net/vnet.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <net/if_llatbl.h>
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 
80 struct dadq;
81 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *);
82 static void nd6_dad_add(struct dadq *dp);
83 static void nd6_dad_del(struct dadq *dp);
84 static void nd6_dad_rele(struct dadq *);
85 static void nd6_dad_starttimer(struct dadq *, int);
86 static void nd6_dad_stoptimer(struct dadq *);
87 static void nd6_dad_timer(void *);
88 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *);
89 static void nd6_dad_ns_output(struct dadq *);
90 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *);
91 static void nd6_dad_na_input(struct ifaddr *);
92 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *,
93     const struct in6_addr *, u_long, int, struct sockaddr *, u_int);
94 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *,
95     const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int);
96 
97 VNET_DEFINE_STATIC(int, dad_enhanced) = 1;
98 #define	V_dad_enhanced			VNET(dad_enhanced)
99 
100 SYSCTL_DECL(_net_inet6_ip6);
101 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW,
102     &VNET_NAME(dad_enhanced), 0,
103     "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD.");
104 
105 VNET_DEFINE_STATIC(int, dad_maxtry) = 15;	/* max # of *tries* to
106 						   transmit DAD packet */
107 #define	V_dad_maxtry			VNET(dad_maxtry)
108 
109 /*
110  * Input a Neighbor Solicitation Message.
111  *
112  * Based on RFC 2461
113  * Based on RFC 2462 (duplicate address detection)
114  */
115 void
nd6_ns_input(struct mbuf * m,int off,int icmp6len)116 nd6_ns_input(struct mbuf *m, int off, int icmp6len)
117 {
118 	struct ifnet *ifp;
119 	struct ip6_hdr *ip6;
120 	struct nd_neighbor_solicit *nd_ns;
121 	struct in6_addr daddr6, myaddr6, saddr6, taddr6;
122 	struct ifaddr *ifa;
123 	struct sockaddr_dl proxydl;
124 	union nd_opts ndopts;
125 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
126 	char *lladdr;
127 	int anycast, lladdrlen, proxy, rflag, tentative, tlladdr;
128 
129 	ifa = NULL;
130 
131 	/* RFC 6980: Nodes MUST silently ignore fragments */
132 	if(m->m_flags & M_FRAGMENTED)
133 		goto freeit;
134 
135 	ifp = m->m_pkthdr.rcvif;
136 	ip6 = mtod(m, struct ip6_hdr *);
137 	if (__predict_false(ip6->ip6_hlim != 255)) {
138 		ICMP6STAT_INC(icp6s_invlhlim);
139 		nd6log((LOG_ERR,
140 		    "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n",
141 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
142 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
143 		goto bads;
144 	}
145 
146 	if (m->m_len < off + icmp6len) {
147 		m = m_pullup(m, off + icmp6len);
148 		if (m == NULL) {
149 			IP6STAT_INC(ip6s_exthdrtoolong);
150 			return;
151 		}
152 	}
153 	ip6 = mtod(m, struct ip6_hdr *);
154 	nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off);
155 
156 	saddr6 = ip6->ip6_src;
157 	daddr6 = ip6->ip6_dst;
158 	taddr6 = nd_ns->nd_ns_target;
159 	if (in6_setscope(&taddr6, ifp, NULL) != 0)
160 		goto bad;
161 
162 	rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0;
163 	if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif)
164 		rflag = 0;
165 
166 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
167 		/* dst has to be a solicited node multicast address. */
168 		if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL &&
169 		    /* don't check ifindex portion */
170 		    daddr6.s6_addr32[1] == 0 &&
171 		    daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE &&
172 		    daddr6.s6_addr8[12] == 0xff) {
173 			; /* good */
174 		} else {
175 			nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
176 			    "(wrong ip6 dst)\n"));
177 			goto bad;
178 		}
179 	} else if (!V_nd6_onlink_ns_rfc4861) {
180 		struct sockaddr_in6 src_sa6;
181 
182 		/*
183 		 * According to recent IETF discussions, it is not a good idea
184 		 * to accept a NS from an address which would not be deemed
185 		 * to be a neighbor otherwise.  This point is expected to be
186 		 * clarified in future revisions of the specification.
187 		 */
188 		bzero(&src_sa6, sizeof(src_sa6));
189 		src_sa6.sin6_family = AF_INET6;
190 		src_sa6.sin6_len = sizeof(src_sa6);
191 		src_sa6.sin6_addr = saddr6;
192 		if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) {
193 			nd6log((LOG_INFO, "nd6_ns_input: "
194 				"NS packet from non-neighbor\n"));
195 			goto bad;
196 		}
197 	}
198 
199 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
200 		nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n"));
201 		goto bad;
202 	}
203 
204 	icmp6len -= sizeof(*nd_ns);
205 	nd6_option_init(nd_ns + 1, icmp6len, &ndopts);
206 	if (nd6_options(&ndopts) < 0) {
207 		nd6log((LOG_INFO,
208 		    "nd6_ns_input: invalid ND option, ignored\n"));
209 		/* nd6_options have incremented stats */
210 		goto freeit;
211 	}
212 
213 	lladdr = NULL;
214 	lladdrlen = 0;
215 	if (ndopts.nd_opts_src_lladdr) {
216 		lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
217 		lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
218 	}
219 
220 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) {
221 		nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet "
222 		    "(link-layer address option)\n"));
223 		goto bad;
224 	}
225 
226 	/*
227 	 * Attaching target link-layer address to the NA?
228 	 * (RFC 2461 7.2.4)
229 	 *
230 	 * NS IP dst is unicast/anycast			MUST NOT add
231 	 * NS IP dst is solicited-node multicast	MUST add
232 	 *
233 	 * In implementation, we add target link-layer address by default.
234 	 * We do not add one in MUST NOT cases.
235 	 */
236 	if (!IN6_IS_ADDR_MULTICAST(&daddr6))
237 		tlladdr = 0;
238 	else
239 		tlladdr = 1;
240 
241 	/*
242 	 * Target address (taddr6) must be either:
243 	 * (1) Valid unicast/anycast address for my receiving interface,
244 	 * (2) Unicast address for which I'm offering proxy service, or
245 	 * (3) "tentative" address on which DAD is being performed.
246 	 */
247 	/* (1) and (3) check. */
248 	if (ifp->if_carp)
249 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
250 	else
251 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
252 
253 	/* (2) check. */
254 	proxy = 0;
255 	if (ifa == NULL) {
256 		struct sockaddr_dl rt_gateway;
257 		struct rt_addrinfo info;
258 		struct sockaddr_in6 dst6;
259 
260 		bzero(&dst6, sizeof(dst6));
261 		dst6.sin6_len = sizeof(struct sockaddr_in6);
262 		dst6.sin6_family = AF_INET6;
263 		dst6.sin6_addr = taddr6;
264 
265 		bzero(&rt_gateway, sizeof(rt_gateway));
266 		rt_gateway.sdl_len = sizeof(rt_gateway);
267 		bzero(&info, sizeof(info));
268 		info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway;
269 
270 		if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6,
271 		    0, 0, &info) == 0) {
272 			if ((info.rti_flags & RTF_ANNOUNCE) != 0 &&
273 			    rt_gateway.sdl_family == AF_LINK) {
274 				/*
275 				 * proxy NDP for single entry
276 				 */
277 				proxydl = *SDL(&rt_gateway);
278 				ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(
279 				    ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST);
280 				if (ifa)
281 					proxy = 1;
282 			}
283 		}
284 	}
285 	if (ifa == NULL) {
286 		/*
287 		 * We've got an NS packet, and we don't have that address
288 		 * assigned for us.  We MUST silently ignore it.
289 		 * See RFC2461 7.2.3.
290 		 */
291 		goto freeit;
292 	}
293 	myaddr6 = *IFA_IN6(ifa);
294 	anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST;
295 	tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE;
296 	if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED)
297 		goto freeit;
298 
299 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
300 		nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s "
301 		    "(if %d, NS packet %d)\n",
302 		    ip6_sprintf(ip6bufs, &taddr6),
303 		    ifp->if_addrlen, lladdrlen - 2));
304 		goto bad;
305 	}
306 
307 	if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) {
308 		nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n",
309 		    ip6_sprintf(ip6bufs, &saddr6)));
310 		goto freeit;
311 	}
312 
313 	/*
314 	 * We have neighbor solicitation packet, with target address equals to
315 	 * one of my tentative address.
316 	 *
317 	 * src addr	how to process?
318 	 * ---		---
319 	 * multicast	of course, invalid (rejected in ip6_input)
320 	 * unicast	somebody is doing address resolution -> ignore
321 	 * unspec	dup address detection
322 	 *
323 	 * The processing is defined in RFC 2462.
324 	 */
325 	if (tentative) {
326 		/*
327 		 * If source address is unspecified address, it is for
328 		 * duplicate address detection.
329 		 *
330 		 * If not, the packet is for addess resolution;
331 		 * silently ignore it.
332 		 */
333 		if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
334 			nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce);
335 
336 		goto freeit;
337 	}
338 
339 	/*
340 	 * If the source address is unspecified address, entries must not
341 	 * be created or updated.
342 	 * It looks that sender is performing DAD.  Output NA toward
343 	 * all-node multicast address, to tell the sender that I'm using
344 	 * the address.
345 	 * S bit ("solicited") must be zero.
346 	 */
347 	if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) {
348 		struct in6_addr in6_all;
349 
350 		in6_all = in6addr_linklocal_allnodes;
351 		if (in6_setscope(&in6_all, ifp, NULL) != 0)
352 			goto bad;
353 		nd6_na_output_fib(ifp, &in6_all, &taddr6,
354 		    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
355 		    rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL,
356 		    M_GETFIB(m));
357 		goto freeit;
358 	}
359 
360 	nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen,
361 	    ND_NEIGHBOR_SOLICIT, 0);
362 
363 	nd6_na_output_fib(ifp, &saddr6, &taddr6,
364 	    ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) |
365 	    rflag | ND_NA_FLAG_SOLICITED, tlladdr,
366 	    proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m));
367  freeit:
368 	if (ifa != NULL)
369 		ifa_free(ifa);
370 	m_freem(m);
371 	return;
372 
373  bad:
374 	nd6log((LOG_ERR, "nd6_ns_input: src=%s\n",
375 		ip6_sprintf(ip6bufs, &saddr6)));
376 	nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n",
377 		ip6_sprintf(ip6bufs, &daddr6)));
378 	nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n",
379 		ip6_sprintf(ip6bufs, &taddr6)));
380  bads:
381 	ICMP6STAT_INC(icp6s_badns);
382 	if (ifa != NULL)
383 		ifa_free(ifa);
384 	m_freem(m);
385 }
386 
387 /*
388  * Output a Neighbor Solicitation Message. Caller specifies:
389  *	- ICMP6 header source IP6 address
390  *	- ND6 header target IP6 address
391  *	- ND6 header source datalink address
392  *
393  * Based on RFC 2461
394  * Based on RFC 2462 (duplicate address detection)
395  *
396  *    ln - for source address determination
397  * nonce - If non-NULL, NS is used for duplicate address detection and
398  *         the value (length is ND_OPT_NONCE_LEN) is used as a random nonce.
399  */
400 static void
nd6_ns_output_fib(struct ifnet * ifp,const struct in6_addr * saddr6,const struct in6_addr * daddr6,const struct in6_addr * taddr6,uint8_t * nonce,u_int fibnum)401 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6,
402     const struct in6_addr *daddr6, const struct in6_addr *taddr6,
403     uint8_t *nonce, u_int fibnum)
404 {
405 	struct mbuf *m;
406 	struct m_tag *mtag;
407 	struct ip6_hdr *ip6;
408 	struct nd_neighbor_solicit *nd_ns;
409 	struct ip6_moptions im6o;
410 	int icmp6len;
411 	int maxlen;
412 
413 	NET_EPOCH_ASSERT();
414 
415 	if (IN6_IS_ADDR_MULTICAST(taddr6))
416 		return;
417 
418 	/* estimate the size of message */
419 	maxlen = sizeof(*ip6) + sizeof(*nd_ns);
420 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
421 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
422 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
423 	    __func__, max_linkhdr, maxlen, MCLBYTES));
424 
425 	if (max_linkhdr + maxlen > MHLEN)
426 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
427 	else
428 		m = m_gethdr(M_NOWAIT, MT_DATA);
429 	if (m == NULL)
430 		return;
431 	M_SETFIB(m, fibnum);
432 
433 	if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) {
434 		m->m_flags |= M_MCAST;
435 		im6o.im6o_multicast_ifp = ifp;
436 		im6o.im6o_multicast_hlim = 255;
437 		im6o.im6o_multicast_loop = 0;
438 	}
439 
440 	icmp6len = sizeof(*nd_ns);
441 	m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len;
442 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
443 
444 	/* fill neighbor solicitation packet */
445 	ip6 = mtod(m, struct ip6_hdr *);
446 	ip6->ip6_flow = 0;
447 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
448 	ip6->ip6_vfc |= IPV6_VERSION;
449 	/* ip6->ip6_plen will be set later */
450 	ip6->ip6_nxt = IPPROTO_ICMPV6;
451 	ip6->ip6_hlim = 255;
452 	if (daddr6)
453 		ip6->ip6_dst = *daddr6;
454 	else {
455 		ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
456 		ip6->ip6_dst.s6_addr16[1] = 0;
457 		ip6->ip6_dst.s6_addr32[1] = 0;
458 		ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE;
459 		ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3];
460 		ip6->ip6_dst.s6_addr8[12] = 0xff;
461 		if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
462 			goto bad;
463 	}
464 	if (nonce == NULL) {
465 		struct ifaddr *ifa = NULL;
466 
467 		/*
468 		 * RFC2461 7.2.2:
469 		 * "If the source address of the packet prompting the
470 		 * solicitation is the same as one of the addresses assigned
471 		 * to the outgoing interface, that address SHOULD be placed
472 		 * in the IP Source Address of the outgoing solicitation.
473 		 * Otherwise, any one of the addresses assigned to the
474 		 * interface should be used."
475 		 *
476 		 * We use the source address for the prompting packet
477 		 * (saddr6), if saddr6 belongs to the outgoing interface.
478 		 * Otherwise, we perform the source address selection as usual.
479 		 */
480 
481 		if (saddr6 != NULL)
482 			ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6);
483 		if (ifa != NULL) {
484 			/* ip6_src set already. */
485 			ip6->ip6_src = *saddr6;
486 			ifa_free(ifa);
487 		} else {
488 			int error;
489 			struct in6_addr dst6, src6;
490 			uint32_t scopeid;
491 
492 			in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
493 			error = in6_selectsrc_addr(fibnum, &dst6,
494 			    scopeid, ifp, &src6, NULL);
495 			if (error) {
496 				char ip6buf[INET6_ADDRSTRLEN];
497 				nd6log((LOG_DEBUG, "%s: source can't be "
498 				    "determined: dst=%s, error=%d\n", __func__,
499 				    ip6_sprintf(ip6buf, &dst6),
500 				    error));
501 				goto bad;
502 			}
503 			ip6->ip6_src = src6;
504 		}
505 	} else {
506 		/*
507 		 * Source address for DAD packet must always be IPv6
508 		 * unspecified address. (0::0)
509 		 * We actually don't have to 0-clear the address (we did it
510 		 * above), but we do so here explicitly to make the intention
511 		 * clearer.
512 		 */
513 		bzero(&ip6->ip6_src, sizeof(ip6->ip6_src));
514 	}
515 	nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1);
516 	nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT;
517 	nd_ns->nd_ns_code = 0;
518 	nd_ns->nd_ns_reserved = 0;
519 	nd_ns->nd_ns_target = *taddr6;
520 	in6_clearscope(&nd_ns->nd_ns_target); /* XXX */
521 
522 	/*
523 	 * Add source link-layer address option.
524 	 *
525 	 *				spec		implementation
526 	 *				---		---
527 	 * DAD packet			MUST NOT	do not add the option
528 	 * there's no link layer address:
529 	 *				impossible	do not add the option
530 	 * there's link layer address:
531 	 *	Multicast NS		MUST add one	add the option
532 	 *	Unicast NS		SHOULD add one	add the option
533 	 */
534 	if (nonce == NULL) {
535 		struct nd_opt_hdr *nd_opt;
536 		char *mac;
537 		int optlen;
538 
539 		mac = NULL;
540 		if (ifp->if_carp)
541 			mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src);
542 		if (mac == NULL)
543 			mac = nd6_ifptomac(ifp);
544 
545 		if (mac != NULL) {
546 			nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
547 			optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
548 			/* 8 byte alignments... */
549 			optlen = (optlen + 7) & ~7;
550 			m->m_pkthdr.len += optlen;
551 			m->m_len += optlen;
552 			icmp6len += optlen;
553 			bzero(nd_opt, optlen);
554 			nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR;
555 			nd_opt->nd_opt_len = optlen >> 3;
556 			bcopy(mac, nd_opt + 1, ifp->if_addrlen);
557 		}
558 	}
559 	/*
560 	 * Add a Nonce option (RFC 3971) to detect looped back NS messages.
561 	 * This behavior is documented as Enhanced Duplicate Address
562 	 * Detection in RFC 7527.
563 	 * net.inet6.ip6.dad_enhanced=0 disables this.
564 	 */
565 	if (V_dad_enhanced != 0 && nonce != NULL) {
566 		int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN;
567 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1);
568 		/* 8-byte alignment is required. */
569 		optlen = (optlen + 7) & ~7;
570 
571 		m->m_pkthdr.len += optlen;
572 		m->m_len += optlen;
573 		icmp6len += optlen;
574 		bzero((caddr_t)nd_opt, optlen);
575 		nd_opt->nd_opt_type = ND_OPT_NONCE;
576 		nd_opt->nd_opt_len = optlen >> 3;
577 		bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN);
578 	}
579 	ip6->ip6_plen = htons((u_short)icmp6len);
580 	nd_ns->nd_ns_cksum = 0;
581 	nd_ns->nd_ns_cksum =
582 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len);
583 
584 	if (send_sendso_input_hook != NULL) {
585 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
586 			sizeof(unsigned short), M_NOWAIT);
587 		if (mtag == NULL)
588 			goto bad;
589 		*(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type;
590 		m_tag_prepend(m, mtag);
591 	}
592 
593 	ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0,
594 	    &im6o, NULL, NULL);
595 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
596 	icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit);
597 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]);
598 
599 	return;
600 
601   bad:
602 	m_freem(m);
603 }
604 
605 #ifndef BURN_BRIDGES
606 void
nd6_ns_output(struct ifnet * ifp,const struct in6_addr * saddr6,const struct in6_addr * daddr6,const struct in6_addr * taddr6,uint8_t * nonce)607 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6,
608     const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce)
609 {
610 
611 	nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB);
612 }
613 #endif
614 /*
615  * Neighbor advertisement input handling.
616  *
617  * Based on RFC 2461
618  * Based on RFC 2462 (duplicate address detection)
619  *
620  * the following items are not implemented yet:
621  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
622  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
623  */
624 void
nd6_na_input(struct mbuf * m,int off,int icmp6len)625 nd6_na_input(struct mbuf *m, int off, int icmp6len)
626 {
627 	struct ifnet *ifp;
628 	struct ip6_hdr *ip6;
629 	struct ifaddr *ifa;
630 	struct llentry *ln;
631 	struct mbuf *chain;
632 	struct nd_neighbor_advert *nd_na;
633 	struct in6_addr daddr6, taddr6;
634 	union nd_opts ndopts;
635 	u_char linkhdr[LLE_MAX_LINKHDR];
636 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
637 	char *lladdr;
638 	size_t linkhdrsize;
639 	int flags, is_override, is_router, is_solicited;
640 	int lladdr_off, lladdrlen, checklink;
641 	bool flush_holdchain = false;
642 
643 	NET_EPOCH_ASSERT();
644 
645 	chain = NULL;
646 	ln = NULL;
647 	checklink = 0;
648 
649 	/* RFC 6980: Nodes MUST silently ignore fragments */
650 	if(m->m_flags & M_FRAGMENTED)
651 		goto freeit;
652 
653 	ifp = m->m_pkthdr.rcvif;
654 	ip6 = mtod(m, struct ip6_hdr *);
655 	if (__predict_false(ip6->ip6_hlim != 255)) {
656 		ICMP6STAT_INC(icp6s_invlhlim);
657 		nd6log((LOG_ERR,
658 		    "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n",
659 		    ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
660 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
661 		goto bad;
662 	}
663 
664 	if (m->m_len < off + icmp6len) {
665 		m = m_pullup(m, off + icmp6len);
666 		if (m == NULL) {
667 			IP6STAT_INC(ip6s_exthdrtoolong);
668 			return;
669 		}
670 	}
671 	ip6 = mtod(m, struct ip6_hdr *);
672 	nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off);
673 
674 	flags = nd_na->nd_na_flags_reserved;
675 	is_router = ((flags & ND_NA_FLAG_ROUTER) != 0);
676 	is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0);
677 	is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0);
678 
679 	taddr6 = nd_na->nd_na_target;
680 	if (in6_setscope(&taddr6, ifp, NULL))
681 		goto bad;	/* XXX: impossible */
682 
683 	if (IN6_IS_ADDR_MULTICAST(&taddr6)) {
684 		nd6log((LOG_ERR,
685 		    "nd6_na_input: invalid target address %s\n",
686 		    ip6_sprintf(ip6bufs, &taddr6)));
687 		goto bad;
688 	}
689 
690 	daddr6 = ip6->ip6_dst;
691 	if (IN6_IS_ADDR_MULTICAST(&daddr6))
692 		if (is_solicited) {
693 			nd6log((LOG_ERR,
694 			    "nd6_na_input: a solicited adv is multicasted\n"));
695 			goto bad;
696 		}
697 
698 	icmp6len -= sizeof(*nd_na);
699 	nd6_option_init(nd_na + 1, icmp6len, &ndopts);
700 	if (nd6_options(&ndopts) < 0) {
701 		nd6log((LOG_INFO,
702 		    "nd6_na_input: invalid ND option, ignored\n"));
703 		/* nd6_options have incremented stats */
704 		goto freeit;
705 	}
706 
707 	lladdr = NULL;
708 	lladdrlen = 0;
709 	if (ndopts.nd_opts_tgt_lladdr) {
710 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
711 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
712 	}
713 
714 	/*
715 	 * This effectively disables the DAD check on a non-master CARP
716 	 * address.
717 	 */
718 	if (ifp->if_carp)
719 		ifa = (*carp_iamatch6_p)(ifp, &taddr6);
720 	else
721 		ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6);
722 
723 	/*
724 	 * Target address matches one of my interface address.
725 	 *
726 	 * If my address is tentative, this means that there's somebody
727 	 * already using the same address as mine.  This indicates DAD failure.
728 	 * This is defined in RFC 2462.
729 	 *
730 	 * Otherwise, process as defined in RFC 2461.
731 	 */
732 	if (ifa
733 	 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) {
734 		nd6_dad_na_input(ifa);
735 		ifa_free(ifa);
736 		goto freeit;
737 	}
738 
739 	/* Just for safety, maybe unnecessary. */
740 	if (ifa) {
741 		ifa_free(ifa);
742 		log(LOG_ERR,
743 		    "nd6_na_input: duplicate IP6 address %s\n",
744 		    ip6_sprintf(ip6bufs, &taddr6));
745 		goto freeit;
746 	}
747 
748 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
749 		nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s "
750 		    "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6),
751 		    ifp->if_addrlen, lladdrlen - 2));
752 		goto bad;
753 	}
754 
755 	/*
756 	 * If no neighbor cache entry is found, NA SHOULD silently be
757 	 * discarded.
758 	 */
759 	ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp);
760 	if (ln == NULL) {
761 		goto freeit;
762 	}
763 
764 	/*
765 	 * Do not try to override static entry.
766 	 */
767 	if (ln->la_flags & LLE_STATIC)
768 		goto freeit;
769 
770 	if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
771 		/*
772 		 * If the link-layer has address, and no lladdr option came,
773 		 * discard the packet.
774 		 */
775 		if (ifp->if_addrlen && lladdr == NULL) {
776 			goto freeit;
777 		}
778 
779 		/*
780 		 * Record link-layer address, and update the state.
781 		 */
782 		if (!nd6_try_set_entry_addr(ifp, ln, lladdr))
783 			goto freeit;
784 
785 		flush_holdchain = true;
786 		if (is_solicited)
787 			nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
788 		else
789 			nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
790 		EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
791 		if ((ln->ln_router = is_router) != 0) {
792 			/*
793 			 * This means a router's state has changed from
794 			 * non-reachable to probably reachable, and might
795 			 * affect the status of associated prefixes..
796 			 */
797 			checklink = 1;
798 		}
799 	} else {
800 		int llchange;
801 
802 		/*
803 		 * Check if the link-layer address has changed or not.
804 		 */
805 		if (lladdr == NULL)
806 			llchange = 0;
807 		else {
808 			if (ln->la_flags & LLE_VALID) {
809 				if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen))
810 					llchange = 1;
811 				else
812 					llchange = 0;
813 			} else
814 				llchange = 1;
815 		}
816 
817 		/*
818 		 * This is VERY complex.  Look at it with care.
819 		 *
820 		 * override solicit lladdr llchange	action
821 		 *					(L: record lladdr)
822 		 *
823 		 *	0	0	n	--	(2c)
824 		 *	0	0	y	n	(2b) L
825 		 *	0	0	y	y	(1)    REACHABLE->STALE
826 		 *	0	1	n	--	(2c)   *->REACHABLE
827 		 *	0	1	y	n	(2b) L *->REACHABLE
828 		 *	0	1	y	y	(1)    REACHABLE->STALE
829 		 *	1	0	n	--	(2a)
830 		 *	1	0	y	n	(2a) L
831 		 *	1	0	y	y	(2a) L *->STALE
832 		 *	1	1	n	--	(2a)   *->REACHABLE
833 		 *	1	1	y	n	(2a) L *->REACHABLE
834 		 *	1	1	y	y	(2a) L *->REACHABLE
835 		 */
836 		if (!is_override && (lladdr != NULL && llchange)) {  /* (1) */
837 			/*
838 			 * If state is REACHABLE, make it STALE.
839 			 * no other updates should be done.
840 			 */
841 			if (ln->ln_state == ND6_LLINFO_REACHABLE)
842 				nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
843 			goto freeit;
844 		} else if (is_override				   /* (2a) */
845 			|| (!is_override && (lladdr != NULL && !llchange)) /* (2b) */
846 			|| lladdr == NULL) {			   /* (2c) */
847 			/*
848 			 * Update link-local address, if any.
849 			 */
850 			if (lladdr != NULL) {
851 				linkhdrsize = sizeof(linkhdr);
852 				if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
853 				    linkhdr, &linkhdrsize, &lladdr_off) != 0)
854 					goto freeit;
855 				if (lltable_try_set_entry_addr(ifp, ln, linkhdr,
856 				    linkhdrsize, lladdr_off) == 0)
857 					goto freeit;
858 				EVENTHANDLER_INVOKE(lle_event, ln,
859 				    LLENTRY_RESOLVED);
860 			}
861 
862 			/*
863 			 * If solicited, make the state REACHABLE.
864 			 * If not solicited and the link-layer address was
865 			 * changed, make it STALE.
866 			 */
867 			if (is_solicited)
868 				nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE);
869 			else {
870 				if (lladdr != NULL && llchange)
871 					nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
872 			}
873 		}
874 
875 		if (ln->ln_router && !is_router) {
876 			/*
877 			 * The peer dropped the router flag.
878 			 * Remove the sender from the Default Router List and
879 			 * update the Destination Cache entries.
880 			 */
881 			struct ifnet *nd6_ifp;
882 
883 			nd6_ifp = lltable_get_ifp(ln->lle_tbl);
884 			if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) &&
885 			    (ND_IFINFO(nd6_ifp)->flags &
886 			     ND6_IFF_ACCEPT_RTADV) != 0)
887 				/*
888 				 * Even if the neighbor is not in the default
889 				 * router list, the neighbor may be used as a
890 				 * next hop for some destinations (e.g. redirect
891 				 * case). So we must call rt6_flush explicitly.
892 				 */
893 				rt6_flush(&ip6->ip6_src, ifp);
894 		}
895 		ln->ln_router = is_router;
896 	}
897         /* XXX - QL
898 	 *  Does this matter?
899 	 *  rt->rt_flags &= ~RTF_REJECT;
900 	 */
901 	ln->la_asked = 0;
902 	if (ln->la_hold != NULL)
903 		chain = nd6_grab_holdchain(ln);
904  freeit:
905 	if (ln != NULL)
906 		LLE_WUNLOCK(ln);
907 
908 	if (chain != NULL)
909 		nd6_flush_holdchain(ifp, ln, chain);
910 	if (flush_holdchain)
911 		nd6_flush_children_holdchain(ifp, ln);
912 
913 	if (checklink)
914 		pfxlist_onlink_check();
915 
916 	m_freem(m);
917 	return;
918 
919  bad:
920 	if (ln != NULL)
921 		LLE_WUNLOCK(ln);
922 
923 	ICMP6STAT_INC(icp6s_badna);
924 	m_freem(m);
925 }
926 
927 /*
928  * Neighbor advertisement output handling.
929  *
930  * Based on RFC 2461
931  *
932  * the following items are not implemented yet:
933  * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD)
934  * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD)
935  *
936  * tlladdr - 1 if include target link-layer address
937  * sdl0 - sockaddr_dl (= proxy NA) or NULL
938  */
939 static void
nd6_na_output_fib(struct ifnet * ifp,const struct in6_addr * daddr6_0,const struct in6_addr * taddr6,u_long flags,int tlladdr,struct sockaddr * sdl0,u_int fibnum)940 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0,
941     const struct in6_addr *taddr6, u_long flags, int tlladdr,
942     struct sockaddr *sdl0, u_int fibnum)
943 {
944 	struct mbuf *m;
945 	struct m_tag *mtag;
946 	struct ip6_hdr *ip6;
947 	struct nd_neighbor_advert *nd_na;
948 	struct ip6_moptions im6o;
949 	struct in6_addr daddr6, dst6, src6;
950 	uint32_t scopeid;
951 
952 	NET_EPOCH_ASSERT();
953 
954 	int icmp6len, maxlen, error;
955 	caddr_t mac = NULL;
956 
957 	daddr6 = *daddr6_0;	/* make a local copy for modification */
958 
959 	/* estimate the size of message */
960 	maxlen = sizeof(*ip6) + sizeof(*nd_na);
961 	maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7;
962 	KASSERT(max_linkhdr + maxlen <= MCLBYTES, (
963 	    "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)",
964 	    __func__, max_linkhdr, maxlen, MCLBYTES));
965 
966 	if (max_linkhdr + maxlen > MHLEN)
967 		m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
968 	else
969 		m = m_gethdr(M_NOWAIT, MT_DATA);
970 	if (m == NULL)
971 		return;
972 	M_SETFIB(m, fibnum);
973 
974 	if (IN6_IS_ADDR_MULTICAST(&daddr6)) {
975 		m->m_flags |= M_MCAST;
976 		im6o.im6o_multicast_ifp = ifp;
977 		im6o.im6o_multicast_hlim = 255;
978 		im6o.im6o_multicast_loop = 0;
979 	}
980 
981 	icmp6len = sizeof(*nd_na);
982 	m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len;
983 	m->m_data += max_linkhdr;	/* or M_ALIGN() equivalent? */
984 
985 	/* fill neighbor advertisement packet */
986 	ip6 = mtod(m, struct ip6_hdr *);
987 	ip6->ip6_flow = 0;
988 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
989 	ip6->ip6_vfc |= IPV6_VERSION;
990 	ip6->ip6_nxt = IPPROTO_ICMPV6;
991 	ip6->ip6_hlim = 255;
992 	if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) {
993 		/* reply to DAD */
994 		daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL;
995 		daddr6.s6_addr16[1] = 0;
996 		daddr6.s6_addr32[1] = 0;
997 		daddr6.s6_addr32[2] = 0;
998 		daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE;
999 		if (in6_setscope(&daddr6, ifp, NULL))
1000 			goto bad;
1001 
1002 		flags &= ~ND_NA_FLAG_SOLICITED;
1003 	}
1004 	ip6->ip6_dst = daddr6;
1005 
1006 	/*
1007 	 * Select a source whose scope is the same as that of the dest.
1008 	 */
1009 	in6_splitscope(&daddr6, &dst6, &scopeid);
1010 	error = in6_selectsrc_addr(fibnum, &dst6,
1011 	    scopeid, ifp, &src6, NULL);
1012 	if (error) {
1013 		char ip6buf[INET6_ADDRSTRLEN];
1014 		nd6log((LOG_DEBUG, "nd6_na_output: source can't be "
1015 		    "determined: dst=%s, error=%d\n",
1016 		    ip6_sprintf(ip6buf, &daddr6), error));
1017 		goto bad;
1018 	}
1019 	ip6->ip6_src = src6;
1020 	nd_na = (struct nd_neighbor_advert *)(ip6 + 1);
1021 	nd_na->nd_na_type = ND_NEIGHBOR_ADVERT;
1022 	nd_na->nd_na_code = 0;
1023 	nd_na->nd_na_target = *taddr6;
1024 	in6_clearscope(&nd_na->nd_na_target); /* XXX */
1025 
1026 	/*
1027 	 * "tlladdr" indicates NS's condition for adding tlladdr or not.
1028 	 * see nd6_ns_input() for details.
1029 	 * Basically, if NS packet is sent to unicast/anycast addr,
1030 	 * target lladdr option SHOULD NOT be included.
1031 	 */
1032 	if (tlladdr) {
1033 		/*
1034 		 * sdl0 != NULL indicates proxy NA.  If we do proxy, use
1035 		 * lladdr in sdl0.  If we are not proxying (sending NA for
1036 		 * my address) use lladdr configured for the interface.
1037 		 */
1038 		if (sdl0 == NULL) {
1039 			if (ifp->if_carp)
1040 				mac = (*carp_macmatch6_p)(ifp, m, taddr6);
1041 			if (mac == NULL)
1042 				mac = nd6_ifptomac(ifp);
1043 		} else if (sdl0->sa_family == AF_LINK) {
1044 			struct sockaddr_dl *sdl;
1045 			sdl = (struct sockaddr_dl *)sdl0;
1046 			if (sdl->sdl_alen == ifp->if_addrlen)
1047 				mac = LLADDR(sdl);
1048 		}
1049 	}
1050 	if (tlladdr && mac) {
1051 		int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen;
1052 		struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1);
1053 
1054 		/* roundup to 8 bytes alignment! */
1055 		optlen = (optlen + 7) & ~7;
1056 
1057 		m->m_pkthdr.len += optlen;
1058 		m->m_len += optlen;
1059 		icmp6len += optlen;
1060 		bzero((caddr_t)nd_opt, optlen);
1061 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
1062 		nd_opt->nd_opt_len = optlen >> 3;
1063 		bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen);
1064 	} else
1065 		flags &= ~ND_NA_FLAG_OVERRIDE;
1066 
1067 	ip6->ip6_plen = htons((u_short)icmp6len);
1068 	nd_na->nd_na_flags_reserved = flags;
1069 	nd_na->nd_na_cksum = 0;
1070 	nd_na->nd_na_cksum =
1071 	    in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len);
1072 
1073 	if (send_sendso_input_hook != NULL) {
1074 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING,
1075 		    sizeof(unsigned short), M_NOWAIT);
1076 		if (mtag == NULL)
1077 			goto bad;
1078 		*(unsigned short *)(mtag + 1) = nd_na->nd_na_type;
1079 		m_tag_prepend(m, mtag);
1080 	}
1081 
1082 	ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL);
1083 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
1084 	icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert);
1085 	ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]);
1086 
1087 	return;
1088 
1089   bad:
1090 	m_freem(m);
1091 }
1092 
1093 #ifndef BURN_BRIDGES
1094 void
nd6_na_output(struct ifnet * ifp,const struct in6_addr * daddr6_0,const struct in6_addr * taddr6,u_long flags,int tlladdr,struct sockaddr * sdl0)1095 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0,
1096     const struct in6_addr *taddr6, u_long flags, int tlladdr,
1097     struct sockaddr *sdl0)
1098 {
1099 
1100 	nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0,
1101 	    RT_DEFAULT_FIB);
1102 }
1103 #endif
1104 
1105 caddr_t
nd6_ifptomac(struct ifnet * ifp)1106 nd6_ifptomac(struct ifnet *ifp)
1107 {
1108 	switch (ifp->if_type) {
1109 	case IFT_ETHER:
1110 	case IFT_IEEE1394:
1111 	case IFT_L2VLAN:
1112 	case IFT_INFINIBAND:
1113 	case IFT_BRIDGE:
1114 		return IF_LLADDR(ifp);
1115 	default:
1116 		return NULL;
1117 	}
1118 }
1119 
1120 struct dadq {
1121 	TAILQ_ENTRY(dadq) dad_list;
1122 	struct ifaddr *dad_ifa;
1123 	int dad_count;		/* max NS to send */
1124 	int dad_ns_tcount;	/* # of trials to send NS */
1125 	int dad_ns_ocount;	/* NS sent so far */
1126 	int dad_ns_icount;
1127 	int dad_na_icount;
1128 	int dad_ns_lcount;	/* looped back NS */
1129 	int dad_loopbackprobe;	/* probing state for loopback detection */
1130 	struct callout dad_timer_ch;
1131 	struct vnet *dad_vnet;
1132 	u_int dad_refcnt;
1133 #define	ND_OPT_NONCE_LEN32 \
1134 		((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t))
1135 	uint32_t dad_nonce[ND_OPT_NONCE_LEN32];
1136 	bool dad_ondadq;	/* on dadq? Protected by DADQ_WLOCK. */
1137 };
1138 
1139 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq);
1140 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock);
1141 #define	V_dadq			VNET(dadq)
1142 #define	V_dad_rwlock		VNET(dad_rwlock)
1143 
1144 #define	DADQ_LOCKPTR()		(&V_dad_rwlock)
1145 #define	DADQ_LOCK_INIT()	rw_init(DADQ_LOCKPTR(), "nd6 DAD queue")
1146 #define	DADQ_RLOCK()		rw_rlock(DADQ_LOCKPTR())
1147 #define	DADQ_RUNLOCK()		rw_runlock(DADQ_LOCKPTR())
1148 #define	DADQ_WLOCK()		rw_wlock(DADQ_LOCKPTR())
1149 #define	DADQ_WUNLOCK()		rw_wunlock(DADQ_LOCKPTR())
1150 
1151 #define	DADQ_LOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_LOCKED);
1152 #define	DADQ_RLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_RLOCKED);
1153 #define	DADQ_WLOCK_ASSERT()	rw_assert(DADQ_LOCKPTR(), RA_WLOCKED);
1154 
1155 static void
nd6_dad_add(struct dadq * dp)1156 nd6_dad_add(struct dadq *dp)
1157 {
1158 	DADQ_WLOCK_ASSERT();
1159 
1160 	TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list);
1161 	dp->dad_ondadq = true;
1162 }
1163 
1164 static void
nd6_dad_del(struct dadq * dp)1165 nd6_dad_del(struct dadq *dp)
1166 {
1167 	DADQ_WLOCK_ASSERT();
1168 
1169 	if (dp->dad_ondadq) {
1170 		/*
1171 		 * Remove dp from the dadq and release the dadq's
1172 		 * reference.
1173 		 */
1174 		TAILQ_REMOVE(&V_dadq, dp, dad_list);
1175 		dp->dad_ondadq = false;
1176 		nd6_dad_rele(dp);
1177 	}
1178 }
1179 
1180 static struct dadq *
nd6_dad_find(struct ifaddr * ifa,struct nd_opt_nonce * n)1181 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n)
1182 {
1183 	struct dadq *dp;
1184 
1185 	DADQ_LOCK_ASSERT();
1186 
1187 	TAILQ_FOREACH(dp, &V_dadq, dad_list) {
1188 		if (dp->dad_ifa != ifa)
1189 			continue;
1190 
1191 		/*
1192 		 * Skip if the nonce matches the received one.
1193 		 * +2 in the length is required because of type and
1194 		 * length fields are included in a header.
1195 		 */
1196 		if (n != NULL &&
1197 		    n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 &&
1198 		    memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0],
1199 		    ND_OPT_NONCE_LEN) == 0) {
1200 			dp->dad_ns_lcount++;
1201 			continue;
1202 		}
1203 		break;
1204 	}
1205 
1206 	return (dp);
1207 }
1208 
1209 static void
nd6_dad_starttimer(struct dadq * dp,int ticks)1210 nd6_dad_starttimer(struct dadq *dp, int ticks)
1211 {
1212 	DADQ_WLOCK_ASSERT();
1213 
1214 	callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp);
1215 }
1216 
1217 static void
nd6_dad_stoptimer(struct dadq * dp)1218 nd6_dad_stoptimer(struct dadq *dp)
1219 {
1220 	callout_drain(&dp->dad_timer_ch);
1221 }
1222 
1223 static void
nd6_dad_rele(struct dadq * dp)1224 nd6_dad_rele(struct dadq *dp)
1225 {
1226 	if (refcount_release(&dp->dad_refcnt)) {
1227 		KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp));
1228 		ifa_free(dp->dad_ifa);
1229 		free(dp, M_IP6NDP);
1230 	}
1231 }
1232 
1233 void
nd6_dad_init(void)1234 nd6_dad_init(void)
1235 {
1236 	DADQ_LOCK_INIT();
1237 	TAILQ_INIT(&V_dadq);
1238 }
1239 
1240 /*
1241  * Start Duplicate Address Detection (DAD) for specified interface address.
1242  */
1243 void
nd6_dad_start(struct ifaddr * ifa,int delay)1244 nd6_dad_start(struct ifaddr *ifa, int delay)
1245 {
1246 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1247 	struct dadq *dp;
1248 	char ip6buf[INET6_ADDRSTRLEN];
1249 
1250 	KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0,
1251 	    ("starting DAD on non-tentative address %p", ifa));
1252 
1253 	/*
1254 	 * If we don't need DAD, don't do it.
1255 	 * There are several cases:
1256 	 * - DAD is disabled globally or on the interface
1257 	 * - the interface address is anycast
1258 	 */
1259 	if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 ||
1260 	    V_ip6_dad_count == 0 ||
1261 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) {
1262 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1263 		return;
1264 	}
1265 	if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 ||
1266 	    (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
1267 	    (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0)
1268 		return;
1269 
1270 	DADQ_WLOCK();
1271 	if ((dp = nd6_dad_find(ifa, NULL)) != NULL) {
1272 		/*
1273 		 * DAD is already in progress.  Let the existing entry
1274 		 * finish it.
1275 		 */
1276 		DADQ_WUNLOCK();
1277 		return;
1278 	}
1279 
1280 	dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO);
1281 	if (dp == NULL) {
1282 		log(LOG_ERR, "nd6_dad_start: memory allocation failed for "
1283 			"%s(%s)\n",
1284 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1285 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1286 		return;
1287 	}
1288 	callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(),
1289 	    CALLOUT_RETURNUNLOCKED);
1290 #ifdef VIMAGE
1291 	dp->dad_vnet = curvnet;
1292 #endif
1293 	nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp),
1294 	    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1295 
1296 	/*
1297 	 * Send NS packet for DAD, ip6_dad_count times.
1298 	 * Note that we must delay the first transmission, if this is the
1299 	 * first packet to be sent from the interface after interface
1300 	 * (re)initialization.
1301 	 */
1302 	dp->dad_ifa = ifa;
1303 	ifa_ref(dp->dad_ifa);
1304 	dp->dad_count = V_ip6_dad_count;
1305 	dp->dad_ns_icount = dp->dad_na_icount = 0;
1306 	dp->dad_ns_ocount = dp->dad_ns_tcount = 0;
1307 	dp->dad_ns_lcount = dp->dad_loopbackprobe = 0;
1308 
1309 	/* Add this to the dadq and add a reference for the dadq. */
1310 	refcount_init(&dp->dad_refcnt, 1);
1311 	nd6_dad_add(dp);
1312 	nd6_dad_starttimer(dp, delay);
1313 	DADQ_WUNLOCK();
1314 }
1315 
1316 /*
1317  * terminate DAD unconditionally.  used for address removals.
1318  */
1319 void
nd6_dad_stop(struct ifaddr * ifa)1320 nd6_dad_stop(struct ifaddr *ifa)
1321 {
1322 	struct dadq *dp;
1323 
1324 	DADQ_WLOCK();
1325 	dp = nd6_dad_find(ifa, NULL);
1326 	if (dp == NULL) {
1327 		DADQ_WUNLOCK();
1328 		/* DAD wasn't started yet */
1329 		return;
1330 	}
1331 
1332 	/*
1333 	 * Acquire a temporary reference so that we can safely stop the callout.
1334 	 */
1335 	(void)refcount_acquire(&dp->dad_refcnt);
1336 	nd6_dad_del(dp);
1337 	DADQ_WUNLOCK();
1338 
1339 	nd6_dad_stoptimer(dp);
1340 	nd6_dad_rele(dp);
1341 }
1342 
1343 static void
nd6_dad_timer(void * arg)1344 nd6_dad_timer(void *arg)
1345 {
1346 	struct dadq *dp = arg;
1347 	struct ifaddr *ifa = dp->dad_ifa;
1348 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1349 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1350 	char ip6buf[INET6_ADDRSTRLEN];
1351 	struct epoch_tracker et;
1352 
1353 	CURVNET_SET(dp->dad_vnet);
1354 	KASSERT(ia != NULL, ("DAD entry %p with no address", dp));
1355 
1356 	NET_EPOCH_ENTER(et);
1357 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
1358 		/* Do not need DAD for ifdisabled interface. */
1359 		log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of "
1360 		    "ND6_IFF_IFDISABLED.\n", ifp->if_xname);
1361 		goto err;
1362 	}
1363 	if (ia->ia6_flags & IN6_IFF_DUPLICATED) {
1364 		log(LOG_ERR, "nd6_dad_timer: called with duplicated address "
1365 			"%s(%s)\n",
1366 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1367 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1368 		goto err;
1369 	}
1370 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) {
1371 		log(LOG_ERR, "nd6_dad_timer: called with non-tentative address "
1372 			"%s(%s)\n",
1373 			ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1374 			ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???");
1375 		goto err;
1376 	}
1377 
1378 	/* Stop DAD if the interface is down even after dad_maxtry attempts. */
1379 	if ((dp->dad_ns_tcount > V_dad_maxtry) &&
1380 	    (((ifp->if_flags & IFF_UP) == 0) ||
1381 	     ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) {
1382 		nd6log((LOG_INFO, "%s: could not run DAD "
1383 		    "because the interface was down or not running.\n",
1384 		    if_name(ifa->ifa_ifp)));
1385 		goto err;
1386 	}
1387 
1388 	/* Need more checks? */
1389 	if (dp->dad_ns_ocount < dp->dad_count) {
1390 		/*
1391 		 * We have more NS to go.  Send NS packet for DAD.
1392 		 */
1393 		nd6_dad_starttimer(dp,
1394 		    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1395 		nd6_dad_ns_output(dp);
1396 		goto done;
1397 	} else {
1398 		/*
1399 		 * We have transmitted sufficient number of DAD packets.
1400 		 * See what we've got.
1401 		 */
1402 		if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) {
1403 			/* We've seen NS or NA, means DAD has failed. */
1404 			nd6_dad_duplicated(ifa, dp);
1405 		} else if (V_dad_enhanced != 0 &&
1406 		    dp->dad_ns_lcount > 0 &&
1407 		    dp->dad_ns_lcount > dp->dad_loopbackprobe) {
1408 			/*
1409 			 * Sec. 4.1 in RFC 7527 requires transmission of
1410 			 * additional probes until the loopback condition
1411 			 * becomes clear when a looped back probe is detected.
1412 			 */
1413 			log(LOG_ERR, "%s: a looped back NS message is "
1414 			    "detected during DAD for %s.  "
1415 			    "Another DAD probes are being sent.\n",
1416 			    if_name(ifa->ifa_ifp),
1417 			    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1418 			dp->dad_loopbackprobe = dp->dad_ns_lcount;
1419 			/*
1420 			 * Send an NS immediately and increase dad_count by
1421 			 * V_nd6_mmaxtries - 1.
1422 			 */
1423 			dp->dad_count =
1424 			    dp->dad_ns_ocount + V_nd6_mmaxtries - 1;
1425 			nd6_dad_starttimer(dp,
1426 			    (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000);
1427 			nd6_dad_ns_output(dp);
1428 			goto done;
1429 		} else {
1430 			/*
1431 			 * We are done with DAD.  No NA came, no NS came.
1432 			 * No duplicate address found.  Check IFDISABLED flag
1433 			 * again in case that it is changed between the
1434 			 * beginning of this function and here.
1435 			 */
1436 			if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0)
1437 				ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1438 
1439 			nd6log((LOG_DEBUG,
1440 			    "%s: DAD complete for %s - no duplicates found\n",
1441 			    if_name(ifa->ifa_ifp),
1442 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1443 			if (dp->dad_ns_lcount > 0)
1444 				log(LOG_ERR, "%s: DAD completed while "
1445 				    "a looped back NS message is detected "
1446 				    "during DAD for %s.\n",
1447 				    if_name(ifa->ifa_ifp),
1448 				    ip6_sprintf(ip6buf, IFA_IN6(ifa)));
1449 		}
1450 	}
1451 err:
1452 	nd6_dad_del(dp);
1453 	DADQ_WUNLOCK();
1454 done:
1455 	NET_EPOCH_EXIT(et);
1456 	CURVNET_RESTORE();
1457 }
1458 
1459 static void
nd6_dad_duplicated(struct ifaddr * ifa,struct dadq * dp)1460 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp)
1461 {
1462 	struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa;
1463 	struct ifnet *ifp;
1464 	char ip6buf[INET6_ADDRSTRLEN];
1465 
1466 	log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: "
1467 	    "NS in/out/loopback=%d/%d/%d, NA in=%d\n",
1468 	    if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr),
1469 	    dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount,
1470 	    dp->dad_na_icount);
1471 
1472 	ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1473 	ia->ia6_flags |= IN6_IFF_DUPLICATED;
1474 
1475 	ifp = ifa->ifa_ifp;
1476 	log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n",
1477 	    if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr));
1478 	log(LOG_ERR, "%s: manual intervention required\n",
1479 	    if_name(ifp));
1480 
1481 	/*
1482 	 * If the address is a link-local address formed from an interface
1483 	 * identifier based on the hardware address which is supposed to be
1484 	 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP
1485 	 * operation on the interface SHOULD be disabled.
1486 	 * [RFC 4862, Section 5.4.5]
1487 	 */
1488 	if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) {
1489 		struct in6_addr in6;
1490 
1491 		/*
1492 		 * To avoid over-reaction, we only apply this logic when we are
1493 		 * very sure that hardware addresses are supposed to be unique.
1494 		 */
1495 		switch (ifp->if_type) {
1496 		case IFT_ETHER:
1497 		case IFT_ATM:
1498 		case IFT_IEEE1394:
1499 		case IFT_INFINIBAND:
1500 			in6 = ia->ia_addr.sin6_addr;
1501 			if (in6_get_hw_ifid(ifp, &in6) == 0 &&
1502 			    IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) {
1503 				ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
1504 				log(LOG_ERR, "%s: possible hardware address "
1505 				    "duplication detected, disable IPv6\n",
1506 				    if_name(ifp));
1507 			}
1508 			break;
1509 		}
1510 	}
1511 }
1512 
1513 /*
1514  * Transmit a neighbour solicitation for the purpose of DAD.  Returns with the
1515  * DAD queue unlocked.
1516  */
1517 static void
nd6_dad_ns_output(struct dadq * dp)1518 nd6_dad_ns_output(struct dadq *dp)
1519 {
1520 	struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa;
1521 	struct ifnet *ifp = dp->dad_ifa->ifa_ifp;
1522 	int i;
1523 
1524 	DADQ_WLOCK_ASSERT();
1525 
1526 	dp->dad_ns_tcount++;
1527 	if ((ifp->if_flags & IFF_UP) == 0) {
1528 		DADQ_WUNLOCK();
1529 		return;
1530 	}
1531 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1532 		DADQ_WUNLOCK();
1533 		return;
1534 	}
1535 
1536 	dp->dad_ns_ocount++;
1537 	if (V_dad_enhanced != 0) {
1538 		for (i = 0; i < ND_OPT_NONCE_LEN32; i++)
1539 			dp->dad_nonce[i] = arc4random();
1540 		/*
1541 		 * XXXHRS: Note that in the case that
1542 		 * DupAddrDetectTransmits > 1, multiple NS messages with
1543 		 * different nonces can be looped back in an unexpected
1544 		 * order.  The current implementation recognizes only
1545 		 * the latest nonce on the sender side.  Practically it
1546 		 * should work well in almost all cases.
1547 		 */
1548 	}
1549 	DADQ_WUNLOCK();
1550 	nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr,
1551 	    (uint8_t *)&dp->dad_nonce[0]);
1552 }
1553 
1554 static void
nd6_dad_ns_input(struct ifaddr * ifa,struct nd_opt_nonce * ndopt_nonce)1555 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce)
1556 {
1557 	struct dadq *dp;
1558 
1559 	if (ifa == NULL)
1560 		panic("ifa == NULL in nd6_dad_ns_input");
1561 
1562 	/* Ignore Nonce option when Enhanced DAD is disabled. */
1563 	if (V_dad_enhanced == 0)
1564 		ndopt_nonce = NULL;
1565 	DADQ_RLOCK();
1566 	dp = nd6_dad_find(ifa, ndopt_nonce);
1567 	if (dp != NULL)
1568 		dp->dad_ns_icount++;
1569 	DADQ_RUNLOCK();
1570 }
1571 
1572 static void
nd6_dad_na_input(struct ifaddr * ifa)1573 nd6_dad_na_input(struct ifaddr *ifa)
1574 {
1575 	struct dadq *dp;
1576 
1577 	if (ifa == NULL)
1578 		panic("ifa == NULL in nd6_dad_na_input");
1579 
1580 	DADQ_RLOCK();
1581 	dp = nd6_dad_find(ifa, NULL);
1582 	if (dp != NULL)
1583 		dp->dad_na_icount++;
1584 	DADQ_RUNLOCK();
1585 }
1586