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 __FBSDID("$FreeBSD$"); 36 37 #include "opt_inet.h" 38 #include "opt_inet6.h" 39 #include "opt_ipsec.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/eventhandler.h> 44 #include <sys/malloc.h> 45 #include <sys/libkern.h> 46 #include <sys/lock.h> 47 #include <sys/rwlock.h> 48 #include <sys/mbuf.h> 49 #include <sys/socket.h> 50 #include <sys/sockio.h> 51 #include <sys/time.h> 52 #include <sys/kernel.h> 53 #include <sys/errno.h> 54 #include <sys/sysctl.h> 55 #include <sys/syslog.h> 56 #include <sys/queue.h> 57 #include <sys/callout.h> 58 #include <sys/refcount.h> 59 60 #include <net/if.h> 61 #include <net/if_types.h> 62 #include <net/if_dl.h> 63 #include <net/if_var.h> 64 #include <net/route.h> 65 #include <net/vnet.h> 66 67 #include <netinet/in.h> 68 #include <netinet/in_var.h> 69 #include <net/if_llatbl.h> 70 #include <netinet6/in6_var.h> 71 #include <netinet6/in6_ifattach.h> 72 #include <netinet/ip6.h> 73 #include <netinet6/ip6_var.h> 74 #include <netinet6/scope6_var.h> 75 #include <netinet6/nd6.h> 76 #include <netinet/icmp6.h> 77 #include <netinet/ip_carp.h> 78 #include <netinet6/send.h> 79 80 #define SDL(s) ((struct sockaddr_dl *)s) 81 82 struct dadq; 83 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *); 84 static void nd6_dad_add(struct dadq *dp); 85 static void nd6_dad_del(struct dadq *dp); 86 static void nd6_dad_rele(struct dadq *); 87 static void nd6_dad_starttimer(struct dadq *, int, int); 88 static void nd6_dad_stoptimer(struct dadq *); 89 static void nd6_dad_timer(struct dadq *); 90 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *); 91 static void nd6_dad_ns_output(struct dadq *); 92 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *); 93 static void nd6_dad_na_input(struct ifaddr *); 94 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *, 95 const struct in6_addr *, u_long, int, struct sockaddr *, u_int); 96 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *, 97 const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int); 98 99 VNET_DEFINE_STATIC(int, dad_enhanced) = 1; 100 #define V_dad_enhanced VNET(dad_enhanced) 101 102 SYSCTL_DECL(_net_inet6_ip6); 103 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW, 104 &VNET_NAME(dad_enhanced), 0, 105 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD."); 106 107 VNET_DEFINE_STATIC(int, dad_maxtry) = 15; /* max # of *tries* to 108 transmit DAD packet */ 109 #define V_dad_maxtry VNET(dad_maxtry) 110 111 /* 112 * Input a Neighbor Solicitation Message. 113 * 114 * Based on RFC 2461 115 * Based on RFC 2462 (duplicate address detection) 116 */ 117 void 118 nd6_ns_input(struct mbuf *m, int off, int icmp6len) 119 { 120 struct ifnet *ifp; 121 struct ip6_hdr *ip6; 122 struct nd_neighbor_solicit *nd_ns; 123 struct in6_addr daddr6, myaddr6, saddr6, taddr6; 124 struct ifaddr *ifa; 125 struct sockaddr_dl proxydl; 126 union nd_opts ndopts; 127 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 128 char *lladdr; 129 int anycast, lladdrlen, proxy, rflag, tentative, tlladdr; 130 131 ifa = NULL; 132 133 /* RFC 6980: Nodes MUST silently ignore fragments */ 134 if(m->m_flags & M_FRAGMENTED) 135 goto freeit; 136 137 ifp = m->m_pkthdr.rcvif; 138 ip6 = mtod(m, struct ip6_hdr *); 139 if (ip6->ip6_hlim != 255) { 140 nd6log((LOG_ERR, 141 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n", 142 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 143 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 144 goto bads; 145 } 146 147 if (m->m_len < off + icmp6len) { 148 m = m_pullup(m, off + icmp6len); 149 if (m == NULL) { 150 IP6STAT_INC(ip6s_exthdrtoolong); 151 return; 152 } 153 } 154 ip6 = mtod(m, struct ip6_hdr *); 155 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off); 156 157 saddr6 = ip6->ip6_src; 158 daddr6 = ip6->ip6_dst; 159 taddr6 = nd_ns->nd_ns_target; 160 if (in6_setscope(&taddr6, ifp, NULL) != 0) 161 goto bad; 162 163 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0; 164 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif) 165 rflag = 0; 166 167 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 168 /* dst has to be a solicited node multicast address. */ 169 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL && 170 /* don't check ifindex portion */ 171 daddr6.s6_addr32[1] == 0 && 172 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE && 173 daddr6.s6_addr8[12] == 0xff) { 174 ; /* good */ 175 } else { 176 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 177 "(wrong ip6 dst)\n")); 178 goto bad; 179 } 180 } else if (!V_nd6_onlink_ns_rfc4861) { 181 struct sockaddr_in6 src_sa6; 182 183 /* 184 * According to recent IETF discussions, it is not a good idea 185 * to accept a NS from an address which would not be deemed 186 * to be a neighbor otherwise. This point is expected to be 187 * clarified in future revisions of the specification. 188 */ 189 bzero(&src_sa6, sizeof(src_sa6)); 190 src_sa6.sin6_family = AF_INET6; 191 src_sa6.sin6_len = sizeof(src_sa6); 192 src_sa6.sin6_addr = saddr6; 193 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) { 194 nd6log((LOG_INFO, "nd6_ns_input: " 195 "NS packet from non-neighbor\n")); 196 goto bad; 197 } 198 } 199 200 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 201 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n")); 202 goto bad; 203 } 204 205 icmp6len -= sizeof(*nd_ns); 206 nd6_option_init(nd_ns + 1, icmp6len, &ndopts); 207 if (nd6_options(&ndopts) < 0) { 208 nd6log((LOG_INFO, 209 "nd6_ns_input: invalid ND option, ignored\n")); 210 /* nd6_options have incremented stats */ 211 goto freeit; 212 } 213 214 lladdr = NULL; 215 lladdrlen = 0; 216 if (ndopts.nd_opts_src_lladdr) { 217 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 218 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 219 } 220 221 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) { 222 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 223 "(link-layer address option)\n")); 224 goto bad; 225 } 226 227 /* 228 * Attaching target link-layer address to the NA? 229 * (RFC 2461 7.2.4) 230 * 231 * NS IP dst is unicast/anycast MUST NOT add 232 * NS IP dst is solicited-node multicast MUST add 233 * 234 * In implementation, we add target link-layer address by default. 235 * We do not add one in MUST NOT cases. 236 */ 237 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) 238 tlladdr = 0; 239 else 240 tlladdr = 1; 241 242 /* 243 * Target address (taddr6) must be either: 244 * (1) Valid unicast/anycast address for my receiving interface, 245 * (2) Unicast address for which I'm offering proxy service, or 246 * (3) "tentative" address on which DAD is being performed. 247 */ 248 /* (1) and (3) check. */ 249 if (ifp->if_carp) 250 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 251 else 252 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 253 254 /* (2) check. */ 255 proxy = 0; 256 if (ifa == NULL) { 257 struct sockaddr_dl rt_gateway; 258 struct rt_addrinfo info; 259 struct sockaddr_in6 dst6; 260 261 bzero(&dst6, sizeof(dst6)); 262 dst6.sin6_len = sizeof(struct sockaddr_in6); 263 dst6.sin6_family = AF_INET6; 264 dst6.sin6_addr = taddr6; 265 266 bzero(&rt_gateway, sizeof(rt_gateway)); 267 rt_gateway.sdl_len = sizeof(rt_gateway); 268 bzero(&info, sizeof(info)); 269 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway; 270 271 if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6, 272 0, 0, &info) == 0) { 273 if ((info.rti_flags & RTF_ANNOUNCE) != 0 && 274 rt_gateway.sdl_family == AF_LINK) { 275 276 /* 277 * proxy NDP for single entry 278 */ 279 proxydl = *SDL(&rt_gateway); 280 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal( 281 ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST); 282 if (ifa) 283 proxy = 1; 284 } 285 } 286 } 287 if (ifa == NULL) { 288 /* 289 * We've got an NS packet, and we don't have that adddress 290 * assigned for us. We MUST silently ignore it. 291 * See RFC2461 7.2.3. 292 */ 293 goto freeit; 294 } 295 myaddr6 = *IFA_IN6(ifa); 296 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST; 297 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE; 298 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED) 299 goto freeit; 300 301 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 302 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s " 303 "(if %d, NS packet %d)\n", 304 ip6_sprintf(ip6bufs, &taddr6), 305 ifp->if_addrlen, lladdrlen - 2)); 306 goto bad; 307 } 308 309 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) { 310 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n", 311 ip6_sprintf(ip6bufs, &saddr6))); 312 goto freeit; 313 } 314 315 /* 316 * We have neighbor solicitation packet, with target address equals to 317 * one of my tentative address. 318 * 319 * src addr how to process? 320 * --- --- 321 * multicast of course, invalid (rejected in ip6_input) 322 * unicast somebody is doing address resolution -> ignore 323 * unspec dup address detection 324 * 325 * The processing is defined in RFC 2462. 326 */ 327 if (tentative) { 328 /* 329 * If source address is unspecified address, it is for 330 * duplicate address detection. 331 * 332 * If not, the packet is for addess resolution; 333 * silently ignore it. 334 */ 335 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 336 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce); 337 338 goto freeit; 339 } 340 341 /* 342 * If the source address is unspecified address, entries must not 343 * be created or updated. 344 * It looks that sender is performing DAD. Output NA toward 345 * all-node multicast address, to tell the sender that I'm using 346 * the address. 347 * S bit ("solicited") must be zero. 348 */ 349 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 350 struct in6_addr in6_all; 351 352 in6_all = in6addr_linklocal_allnodes; 353 if (in6_setscope(&in6_all, ifp, NULL) != 0) 354 goto bad; 355 nd6_na_output_fib(ifp, &in6_all, &taddr6, 356 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 357 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL, 358 M_GETFIB(m)); 359 goto freeit; 360 } 361 362 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, 363 ND_NEIGHBOR_SOLICIT, 0); 364 365 nd6_na_output_fib(ifp, &saddr6, &taddr6, 366 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 367 rflag | ND_NA_FLAG_SOLICITED, tlladdr, 368 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m)); 369 freeit: 370 if (ifa != NULL) 371 ifa_free(ifa); 372 m_freem(m); 373 return; 374 375 bad: 376 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n", 377 ip6_sprintf(ip6bufs, &saddr6))); 378 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n", 379 ip6_sprintf(ip6bufs, &daddr6))); 380 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n", 381 ip6_sprintf(ip6bufs, &taddr6))); 382 bads: 383 ICMP6STAT_INC(icp6s_badns); 384 if (ifa != NULL) 385 ifa_free(ifa); 386 m_freem(m); 387 } 388 389 /* 390 * Output a Neighbor Solicitation Message. Caller specifies: 391 * - ICMP6 header source IP6 address 392 * - ND6 header target IP6 address 393 * - ND6 header source datalink address 394 * 395 * Based on RFC 2461 396 * Based on RFC 2462 (duplicate address detection) 397 * 398 * ln - for source address determination 399 * nonce - If non-NULL, NS is used for duplicate address detection and 400 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce. 401 */ 402 static void 403 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6, 404 const struct in6_addr *daddr6, const struct in6_addr *taddr6, 405 uint8_t *nonce, u_int fibnum) 406 { 407 struct mbuf *m; 408 struct m_tag *mtag; 409 struct ip6_hdr *ip6; 410 struct nd_neighbor_solicit *nd_ns; 411 struct ip6_moptions im6o; 412 int icmp6len; 413 int maxlen; 414 caddr_t mac; 415 416 if (IN6_IS_ADDR_MULTICAST(taddr6)) 417 return; 418 419 /* estimate the size of message */ 420 maxlen = sizeof(*ip6) + sizeof(*nd_ns); 421 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; 422 KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( 423 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", 424 __func__, max_linkhdr, maxlen, MCLBYTES)); 425 426 if (max_linkhdr + maxlen > MHLEN) 427 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 428 else 429 m = m_gethdr(M_NOWAIT, MT_DATA); 430 if (m == NULL) 431 return; 432 M_SETFIB(m, fibnum); 433 434 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) { 435 m->m_flags |= M_MCAST; 436 im6o.im6o_multicast_ifp = ifp; 437 im6o.im6o_multicast_hlim = 255; 438 im6o.im6o_multicast_loop = 0; 439 } 440 441 icmp6len = sizeof(*nd_ns); 442 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len; 443 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ 444 445 /* fill neighbor solicitation packet */ 446 ip6 = mtod(m, struct ip6_hdr *); 447 ip6->ip6_flow = 0; 448 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 449 ip6->ip6_vfc |= IPV6_VERSION; 450 /* ip6->ip6_plen will be set later */ 451 ip6->ip6_nxt = IPPROTO_ICMPV6; 452 ip6->ip6_hlim = 255; 453 if (daddr6) 454 ip6->ip6_dst = *daddr6; 455 else { 456 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL; 457 ip6->ip6_dst.s6_addr16[1] = 0; 458 ip6->ip6_dst.s6_addr32[1] = 0; 459 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE; 460 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3]; 461 ip6->ip6_dst.s6_addr8[12] = 0xff; 462 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 463 goto bad; 464 } 465 if (nonce == NULL) { 466 struct ifaddr *ifa = NULL; 467 468 /* 469 * RFC2461 7.2.2: 470 * "If the source address of the packet prompting the 471 * solicitation is the same as one of the addresses assigned 472 * to the outgoing interface, that address SHOULD be placed 473 * in the IP Source Address of the outgoing solicitation. 474 * Otherwise, any one of the addresses assigned to the 475 * interface should be used." 476 * 477 * We use the source address for the prompting packet 478 * (saddr6), if saddr6 belongs to the outgoing interface. 479 * Otherwise, we perform the source address selection as usual. 480 */ 481 482 if (saddr6 != NULL) 483 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6); 484 if (ifa != NULL) { 485 /* ip6_src set already. */ 486 ip6->ip6_src = *saddr6; 487 ifa_free(ifa); 488 } else { 489 int error; 490 struct in6_addr dst6, src6; 491 uint32_t scopeid; 492 493 in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid); 494 error = in6_selectsrc_addr(fibnum, &dst6, 495 scopeid, ifp, &src6, NULL); 496 if (error) { 497 char ip6buf[INET6_ADDRSTRLEN]; 498 nd6log((LOG_DEBUG, "%s: source can't be " 499 "determined: dst=%s, error=%d\n", __func__, 500 ip6_sprintf(ip6buf, &dst6), 501 error)); 502 goto bad; 503 } 504 ip6->ip6_src = src6; 505 } 506 } else { 507 /* 508 * Source address for DAD packet must always be IPv6 509 * unspecified address. (0::0) 510 * We actually don't have to 0-clear the address (we did it 511 * above), but we do so here explicitly to make the intention 512 * clearer. 513 */ 514 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src)); 515 } 516 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1); 517 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT; 518 nd_ns->nd_ns_code = 0; 519 nd_ns->nd_ns_reserved = 0; 520 nd_ns->nd_ns_target = *taddr6; 521 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */ 522 523 /* 524 * Add source link-layer address option. 525 * 526 * spec implementation 527 * --- --- 528 * DAD packet MUST NOT do not add the option 529 * there's no link layer address: 530 * impossible do not add the option 531 * there's link layer address: 532 * Multicast NS MUST add one add the option 533 * Unicast NS SHOULD add one add the option 534 */ 535 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) { 536 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 537 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); 538 /* 8 byte alignments... */ 539 optlen = (optlen + 7) & ~7; 540 541 m->m_pkthdr.len += optlen; 542 m->m_len += optlen; 543 icmp6len += optlen; 544 bzero((caddr_t)nd_opt, optlen); 545 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; 546 nd_opt->nd_opt_len = optlen >> 3; 547 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); 548 } 549 /* 550 * Add a Nonce option (RFC 3971) to detect looped back NS messages. 551 * This behavior is documented as Enhanced Duplicate Address 552 * Detection in RFC 7527. 553 * net.inet6.ip6.dad_enhanced=0 disables this. 554 */ 555 if (V_dad_enhanced != 0 && nonce != NULL) { 556 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN; 557 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); 558 /* 8-byte alignment is required. */ 559 optlen = (optlen + 7) & ~7; 560 561 m->m_pkthdr.len += optlen; 562 m->m_len += optlen; 563 icmp6len += optlen; 564 bzero((caddr_t)nd_opt, optlen); 565 nd_opt->nd_opt_type = ND_OPT_NONCE; 566 nd_opt->nd_opt_len = optlen >> 3; 567 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN); 568 } 569 ip6->ip6_plen = htons((u_short)icmp6len); 570 nd_ns->nd_ns_cksum = 0; 571 nd_ns->nd_ns_cksum = 572 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len); 573 574 if (send_sendso_input_hook != NULL) { 575 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 576 sizeof(unsigned short), M_NOWAIT); 577 if (mtag == NULL) 578 goto bad; 579 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type; 580 m_tag_prepend(m, mtag); 581 } 582 583 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0, 584 &im6o, NULL, NULL); 585 icmp6_ifstat_inc(ifp, ifs6_out_msg); 586 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit); 587 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]); 588 589 return; 590 591 bad: 592 m_freem(m); 593 } 594 595 #ifndef BURN_BRIDGES 596 void 597 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6, 598 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce) 599 { 600 601 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB); 602 } 603 #endif 604 /* 605 * Neighbor advertisement input handling. 606 * 607 * Based on RFC 2461 608 * Based on RFC 2462 (duplicate address detection) 609 * 610 * the following items are not implemented yet: 611 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 612 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 613 */ 614 void 615 nd6_na_input(struct mbuf *m, int off, int icmp6len) 616 { 617 struct ifnet *ifp; 618 struct ip6_hdr *ip6; 619 struct ifaddr *ifa; 620 struct llentry *ln; 621 struct mbuf *chain; 622 struct nd_neighbor_advert *nd_na; 623 struct in6_addr daddr6, taddr6; 624 struct sockaddr_in6 sin6; 625 union nd_opts ndopts; 626 u_char linkhdr[LLE_MAX_LINKHDR]; 627 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 628 char *lladdr; 629 size_t linkhdrsize; 630 int flags, is_override, is_router, is_solicited; 631 int lladdr_off, lladdrlen, checklink; 632 633 NET_EPOCH_ASSERT(); 634 635 chain = NULL; 636 ln = NULL; 637 checklink = 0; 638 639 /* RFC 6980: Nodes MUST silently ignore fragments */ 640 if(m->m_flags & M_FRAGMENTED) 641 goto freeit; 642 643 ifp = m->m_pkthdr.rcvif; 644 ip6 = mtod(m, struct ip6_hdr *); 645 if (ip6->ip6_hlim != 255) { 646 nd6log((LOG_ERR, 647 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n", 648 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 649 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 650 goto bad; 651 } 652 653 if (m->m_len < off + icmp6len) { 654 m = m_pullup(m, off + icmp6len); 655 if (m == NULL) { 656 IP6STAT_INC(ip6s_exthdrtoolong); 657 return; 658 } 659 } 660 ip6 = mtod(m, struct ip6_hdr *); 661 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off); 662 663 flags = nd_na->nd_na_flags_reserved; 664 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0); 665 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0); 666 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0); 667 668 taddr6 = nd_na->nd_na_target; 669 if (in6_setscope(&taddr6, ifp, NULL)) 670 goto bad; /* XXX: impossible */ 671 672 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 673 nd6log((LOG_ERR, 674 "nd6_na_input: invalid target address %s\n", 675 ip6_sprintf(ip6bufs, &taddr6))); 676 goto bad; 677 } 678 679 daddr6 = ip6->ip6_dst; 680 if (IN6_IS_ADDR_MULTICAST(&daddr6)) 681 if (is_solicited) { 682 nd6log((LOG_ERR, 683 "nd6_na_input: a solicited adv is multicasted\n")); 684 goto bad; 685 } 686 687 icmp6len -= sizeof(*nd_na); 688 nd6_option_init(nd_na + 1, icmp6len, &ndopts); 689 if (nd6_options(&ndopts) < 0) { 690 nd6log((LOG_INFO, 691 "nd6_na_input: invalid ND option, ignored\n")); 692 /* nd6_options have incremented stats */ 693 goto freeit; 694 } 695 696 lladdr = NULL; 697 lladdrlen = 0; 698 if (ndopts.nd_opts_tgt_lladdr) { 699 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 700 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 701 } 702 703 /* 704 * This effectively disables the DAD check on a non-master CARP 705 * address. 706 */ 707 if (ifp->if_carp) 708 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 709 else 710 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 711 712 /* 713 * Target address matches one of my interface address. 714 * 715 * If my address is tentative, this means that there's somebody 716 * already using the same address as mine. This indicates DAD failure. 717 * This is defined in RFC 2462. 718 * 719 * Otherwise, process as defined in RFC 2461. 720 */ 721 if (ifa 722 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) { 723 nd6_dad_na_input(ifa); 724 ifa_free(ifa); 725 goto freeit; 726 } 727 728 /* Just for safety, maybe unnecessary. */ 729 if (ifa) { 730 ifa_free(ifa); 731 log(LOG_ERR, 732 "nd6_na_input: duplicate IP6 address %s\n", 733 ip6_sprintf(ip6bufs, &taddr6)); 734 goto freeit; 735 } 736 737 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 738 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s " 739 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6), 740 ifp->if_addrlen, lladdrlen - 2)); 741 goto bad; 742 } 743 744 /* 745 * If no neighbor cache entry is found, NA SHOULD silently be 746 * discarded. 747 */ 748 ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp); 749 if (ln == NULL) { 750 goto freeit; 751 } 752 753 /* 754 * Do not try to override static entry. 755 */ 756 if (ln->la_flags & LLE_STATIC) 757 goto freeit; 758 759 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 760 /* 761 * If the link-layer has address, and no lladdr option came, 762 * discard the packet. 763 */ 764 if (ifp->if_addrlen && lladdr == NULL) { 765 goto freeit; 766 } 767 768 /* 769 * Record link-layer address, and update the state. 770 */ 771 linkhdrsize = sizeof(linkhdr); 772 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 773 linkhdr, &linkhdrsize, &lladdr_off) != 0) 774 return; 775 776 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 777 lladdr_off) == 0) { 778 ln = NULL; 779 goto freeit; 780 } 781 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 782 if (is_solicited) 783 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 784 else 785 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 786 if ((ln->ln_router = is_router) != 0) { 787 /* 788 * This means a router's state has changed from 789 * non-reachable to probably reachable, and might 790 * affect the status of associated prefixes.. 791 */ 792 checklink = 1; 793 } 794 } else { 795 int llchange; 796 797 /* 798 * Check if the link-layer address has changed or not. 799 */ 800 if (lladdr == NULL) 801 llchange = 0; 802 else { 803 if (ln->la_flags & LLE_VALID) { 804 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen)) 805 llchange = 1; 806 else 807 llchange = 0; 808 } else 809 llchange = 1; 810 } 811 812 /* 813 * This is VERY complex. Look at it with care. 814 * 815 * override solicit lladdr llchange action 816 * (L: record lladdr) 817 * 818 * 0 0 n -- (2c) 819 * 0 0 y n (2b) L 820 * 0 0 y y (1) REACHABLE->STALE 821 * 0 1 n -- (2c) *->REACHABLE 822 * 0 1 y n (2b) L *->REACHABLE 823 * 0 1 y y (1) REACHABLE->STALE 824 * 1 0 n -- (2a) 825 * 1 0 y n (2a) L 826 * 1 0 y y (2a) L *->STALE 827 * 1 1 n -- (2a) *->REACHABLE 828 * 1 1 y n (2a) L *->REACHABLE 829 * 1 1 y y (2a) L *->REACHABLE 830 */ 831 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */ 832 /* 833 * If state is REACHABLE, make it STALE. 834 * no other updates should be done. 835 */ 836 if (ln->ln_state == ND6_LLINFO_REACHABLE) 837 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 838 goto freeit; 839 } else if (is_override /* (2a) */ 840 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */ 841 || lladdr == NULL) { /* (2c) */ 842 /* 843 * Update link-local address, if any. 844 */ 845 if (lladdr != NULL) { 846 linkhdrsize = sizeof(linkhdr); 847 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 848 linkhdr, &linkhdrsize, &lladdr_off) != 0) 849 goto freeit; 850 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, 851 linkhdrsize, lladdr_off) == 0) { 852 ln = NULL; 853 goto freeit; 854 } 855 EVENTHANDLER_INVOKE(lle_event, ln, 856 LLENTRY_RESOLVED); 857 } 858 859 /* 860 * If solicited, make the state REACHABLE. 861 * If not solicited and the link-layer address was 862 * changed, make it STALE. 863 */ 864 if (is_solicited) 865 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 866 else { 867 if (lladdr != NULL && llchange) 868 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 869 } 870 } 871 872 if (ln->ln_router && !is_router) { 873 /* 874 * The peer dropped the router flag. 875 * Remove the sender from the Default Router List and 876 * update the Destination Cache entries. 877 */ 878 struct ifnet *nd6_ifp; 879 880 nd6_ifp = lltable_get_ifp(ln->lle_tbl); 881 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) && 882 (ND_IFINFO(nd6_ifp)->flags & 883 ND6_IFF_ACCEPT_RTADV) != 0) 884 /* 885 * Even if the neighbor is not in the default 886 * router list, the neighbor may be used as a 887 * next hop for some destinations (e.g. redirect 888 * case). So we must call rt6_flush explicitly. 889 */ 890 rt6_flush(&ip6->ip6_src, ifp); 891 } 892 ln->ln_router = is_router; 893 } 894 /* XXX - QL 895 * Does this matter? 896 * rt->rt_flags &= ~RTF_REJECT; 897 */ 898 ln->la_asked = 0; 899 if (ln->la_hold != NULL) { 900 memset(&sin6, 0, sizeof(sin6)); 901 nd6_grab_holdchain(ln, &chain, &sin6); 902 } 903 freeit: 904 if (ln != NULL) 905 LLE_WUNLOCK(ln); 906 907 if (chain != NULL) 908 nd6_flush_holdchain(ifp, chain, &sin6); 909 910 if (checklink) 911 pfxlist_onlink_check(); 912 913 m_freem(m); 914 return; 915 916 bad: 917 if (ln != NULL) 918 LLE_WUNLOCK(ln); 919 920 ICMP6STAT_INC(icp6s_badna); 921 m_freem(m); 922 } 923 924 /* 925 * Neighbor advertisement output handling. 926 * 927 * Based on RFC 2461 928 * 929 * the following items are not implemented yet: 930 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 931 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 932 * 933 * tlladdr - 1 if include target link-layer address 934 * sdl0 - sockaddr_dl (= proxy NA) or NULL 935 */ 936 static void 937 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0, 938 const struct in6_addr *taddr6, u_long flags, int tlladdr, 939 struct sockaddr *sdl0, u_int fibnum) 940 { 941 struct mbuf *m; 942 struct m_tag *mtag; 943 struct ip6_hdr *ip6; 944 struct nd_neighbor_advert *nd_na; 945 struct ip6_moptions im6o; 946 struct in6_addr daddr6, dst6, src6; 947 uint32_t scopeid; 948 949 int icmp6len, maxlen, error; 950 caddr_t mac = NULL; 951 952 daddr6 = *daddr6_0; /* make a local copy for modification */ 953 954 /* estimate the size of message */ 955 maxlen = sizeof(*ip6) + sizeof(*nd_na); 956 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; 957 KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( 958 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", 959 __func__, max_linkhdr, maxlen, MCLBYTES)); 960 961 if (max_linkhdr + maxlen > MHLEN) 962 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 963 else 964 m = m_gethdr(M_NOWAIT, MT_DATA); 965 if (m == NULL) 966 return; 967 M_SETFIB(m, fibnum); 968 969 if (IN6_IS_ADDR_MULTICAST(&daddr6)) { 970 m->m_flags |= M_MCAST; 971 im6o.im6o_multicast_ifp = ifp; 972 im6o.im6o_multicast_hlim = 255; 973 im6o.im6o_multicast_loop = 0; 974 } 975 976 icmp6len = sizeof(*nd_na); 977 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len; 978 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ 979 980 /* fill neighbor advertisement packet */ 981 ip6 = mtod(m, struct ip6_hdr *); 982 ip6->ip6_flow = 0; 983 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 984 ip6->ip6_vfc |= IPV6_VERSION; 985 ip6->ip6_nxt = IPPROTO_ICMPV6; 986 ip6->ip6_hlim = 255; 987 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) { 988 /* reply to DAD */ 989 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL; 990 daddr6.s6_addr16[1] = 0; 991 daddr6.s6_addr32[1] = 0; 992 daddr6.s6_addr32[2] = 0; 993 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE; 994 if (in6_setscope(&daddr6, ifp, NULL)) 995 goto bad; 996 997 flags &= ~ND_NA_FLAG_SOLICITED; 998 } 999 ip6->ip6_dst = daddr6; 1000 1001 /* 1002 * Select a source whose scope is the same as that of the dest. 1003 */ 1004 in6_splitscope(&daddr6, &dst6, &scopeid); 1005 error = in6_selectsrc_addr(fibnum, &dst6, 1006 scopeid, ifp, &src6, NULL); 1007 if (error) { 1008 char ip6buf[INET6_ADDRSTRLEN]; 1009 nd6log((LOG_DEBUG, "nd6_na_output: source can't be " 1010 "determined: dst=%s, error=%d\n", 1011 ip6_sprintf(ip6buf, &daddr6), error)); 1012 goto bad; 1013 } 1014 ip6->ip6_src = src6; 1015 nd_na = (struct nd_neighbor_advert *)(ip6 + 1); 1016 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT; 1017 nd_na->nd_na_code = 0; 1018 nd_na->nd_na_target = *taddr6; 1019 in6_clearscope(&nd_na->nd_na_target); /* XXX */ 1020 1021 /* 1022 * "tlladdr" indicates NS's condition for adding tlladdr or not. 1023 * see nd6_ns_input() for details. 1024 * Basically, if NS packet is sent to unicast/anycast addr, 1025 * target lladdr option SHOULD NOT be included. 1026 */ 1027 if (tlladdr) { 1028 /* 1029 * sdl0 != NULL indicates proxy NA. If we do proxy, use 1030 * lladdr in sdl0. If we are not proxying (sending NA for 1031 * my address) use lladdr configured for the interface. 1032 */ 1033 if (sdl0 == NULL) { 1034 if (ifp->if_carp) 1035 mac = (*carp_macmatch6_p)(ifp, m, taddr6); 1036 if (mac == NULL) 1037 mac = nd6_ifptomac(ifp); 1038 } else if (sdl0->sa_family == AF_LINK) { 1039 struct sockaddr_dl *sdl; 1040 sdl = (struct sockaddr_dl *)sdl0; 1041 if (sdl->sdl_alen == ifp->if_addrlen) 1042 mac = LLADDR(sdl); 1043 } 1044 } 1045 if (tlladdr && mac) { 1046 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 1047 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1); 1048 1049 /* roundup to 8 bytes alignment! */ 1050 optlen = (optlen + 7) & ~7; 1051 1052 m->m_pkthdr.len += optlen; 1053 m->m_len += optlen; 1054 icmp6len += optlen; 1055 bzero((caddr_t)nd_opt, optlen); 1056 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 1057 nd_opt->nd_opt_len = optlen >> 3; 1058 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); 1059 } else 1060 flags &= ~ND_NA_FLAG_OVERRIDE; 1061 1062 ip6->ip6_plen = htons((u_short)icmp6len); 1063 nd_na->nd_na_flags_reserved = flags; 1064 nd_na->nd_na_cksum = 0; 1065 nd_na->nd_na_cksum = 1066 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len); 1067 1068 if (send_sendso_input_hook != NULL) { 1069 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 1070 sizeof(unsigned short), M_NOWAIT); 1071 if (mtag == NULL) 1072 goto bad; 1073 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type; 1074 m_tag_prepend(m, mtag); 1075 } 1076 1077 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL); 1078 icmp6_ifstat_inc(ifp, ifs6_out_msg); 1079 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert); 1080 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]); 1081 1082 return; 1083 1084 bad: 1085 m_freem(m); 1086 } 1087 1088 #ifndef BURN_BRIDGES 1089 void 1090 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0, 1091 const struct in6_addr *taddr6, u_long flags, int tlladdr, 1092 struct sockaddr *sdl0) 1093 { 1094 1095 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0, 1096 RT_DEFAULT_FIB); 1097 } 1098 #endif 1099 1100 caddr_t 1101 nd6_ifptomac(struct ifnet *ifp) 1102 { 1103 switch (ifp->if_type) { 1104 case IFT_ETHER: 1105 case IFT_IEEE1394: 1106 case IFT_L2VLAN: 1107 case IFT_INFINIBAND: 1108 case IFT_BRIDGE: 1109 return IF_LLADDR(ifp); 1110 default: 1111 return NULL; 1112 } 1113 } 1114 1115 struct dadq { 1116 TAILQ_ENTRY(dadq) dad_list; 1117 struct ifaddr *dad_ifa; 1118 int dad_count; /* max NS to send */ 1119 int dad_ns_tcount; /* # of trials to send NS */ 1120 int dad_ns_ocount; /* NS sent so far */ 1121 int dad_ns_icount; 1122 int dad_na_icount; 1123 int dad_ns_lcount; /* looped back NS */ 1124 int dad_loopbackprobe; /* probing state for loopback detection */ 1125 struct callout dad_timer_ch; 1126 struct vnet *dad_vnet; 1127 u_int dad_refcnt; 1128 #define ND_OPT_NONCE_LEN32 \ 1129 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t)) 1130 uint32_t dad_nonce[ND_OPT_NONCE_LEN32]; 1131 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */ 1132 }; 1133 1134 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq); 1135 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock); 1136 #define V_dadq VNET(dadq) 1137 #define V_dad_rwlock VNET(dad_rwlock) 1138 1139 #define DADQ_RLOCK() rw_rlock(&V_dad_rwlock) 1140 #define DADQ_RUNLOCK() rw_runlock(&V_dad_rwlock) 1141 #define DADQ_WLOCK() rw_wlock(&V_dad_rwlock) 1142 #define DADQ_WUNLOCK() rw_wunlock(&V_dad_rwlock) 1143 1144 static void 1145 nd6_dad_add(struct dadq *dp) 1146 { 1147 1148 DADQ_WLOCK(); 1149 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list); 1150 dp->dad_ondadq = true; 1151 DADQ_WUNLOCK(); 1152 } 1153 1154 static void 1155 nd6_dad_del(struct dadq *dp) 1156 { 1157 1158 DADQ_WLOCK(); 1159 if (dp->dad_ondadq) { 1160 /* 1161 * Remove dp from the dadq and release the dadq's 1162 * reference. 1163 */ 1164 TAILQ_REMOVE(&V_dadq, dp, dad_list); 1165 dp->dad_ondadq = false; 1166 DADQ_WUNLOCK(); 1167 nd6_dad_rele(dp); 1168 } else 1169 DADQ_WUNLOCK(); 1170 } 1171 1172 static struct dadq * 1173 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n) 1174 { 1175 struct dadq *dp; 1176 1177 DADQ_RLOCK(); 1178 TAILQ_FOREACH(dp, &V_dadq, dad_list) { 1179 if (dp->dad_ifa != ifa) 1180 continue; 1181 /* 1182 * Skip if the nonce matches the received one. 1183 * +2 in the length is required because of type and 1184 * length fields are included in a header. 1185 */ 1186 if (n != NULL && 1187 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 && 1188 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0], 1189 ND_OPT_NONCE_LEN) == 0) { 1190 dp->dad_ns_lcount++; 1191 continue; 1192 } 1193 refcount_acquire(&dp->dad_refcnt); 1194 break; 1195 } 1196 DADQ_RUNLOCK(); 1197 1198 return (dp); 1199 } 1200 1201 static void 1202 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns) 1203 { 1204 1205 NET_EPOCH_ASSERT(); 1206 1207 if (send_ns != 0) 1208 nd6_dad_ns_output(dp); 1209 callout_reset(&dp->dad_timer_ch, ticks, 1210 (void (*)(void *))nd6_dad_timer, (void *)dp); 1211 } 1212 1213 static void 1214 nd6_dad_stoptimer(struct dadq *dp) 1215 { 1216 1217 callout_drain(&dp->dad_timer_ch); 1218 } 1219 1220 static void 1221 nd6_dad_rele(struct dadq *dp) 1222 { 1223 1224 if (refcount_release(&dp->dad_refcnt)) { 1225 ifa_free(dp->dad_ifa); 1226 free(dp, M_IP6NDP); 1227 } 1228 } 1229 1230 void 1231 nd6_dad_init(void) 1232 { 1233 1234 rw_init(&V_dad_rwlock, "nd6 DAD queue"); 1235 TAILQ_INIT(&V_dadq); 1236 } 1237 1238 /* 1239 * Start Duplicate Address Detection (DAD) for specified interface address. 1240 */ 1241 void 1242 nd6_dad_start(struct ifaddr *ifa, int delay) 1243 { 1244 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1245 struct dadq *dp; 1246 char ip6buf[INET6_ADDRSTRLEN]; 1247 struct epoch_tracker et; 1248 1249 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0, 1250 ("starting DAD on non-tentative address %p", ifa)); 1251 1252 /* 1253 * If we don't need DAD, don't do it. 1254 * There are several cases: 1255 * - DAD is disabled globally or on the interface 1256 * - the interface address is anycast 1257 */ 1258 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 || 1259 V_ip6_dad_count == 0 || 1260 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) { 1261 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1262 return; 1263 } 1264 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 || 1265 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || 1266 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0) 1267 return; 1268 1269 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) { 1270 /* 1271 * DAD is already in progress. Let the existing entry 1272 * finish it. 1273 */ 1274 nd6_dad_rele(dp); 1275 return; 1276 } 1277 1278 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO); 1279 if (dp == NULL) { 1280 log(LOG_ERR, "nd6_dad_start: memory allocation failed for " 1281 "%s(%s)\n", 1282 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1283 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1284 return; 1285 } 1286 callout_init(&dp->dad_timer_ch, 0); 1287 #ifdef VIMAGE 1288 dp->dad_vnet = curvnet; 1289 #endif 1290 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp), 1291 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1292 1293 /* 1294 * Send NS packet for DAD, ip6_dad_count times. 1295 * Note that we must delay the first transmission, if this is the 1296 * first packet to be sent from the interface after interface 1297 * (re)initialization. 1298 */ 1299 dp->dad_ifa = ifa; 1300 ifa_ref(dp->dad_ifa); 1301 dp->dad_count = V_ip6_dad_count; 1302 dp->dad_ns_icount = dp->dad_na_icount = 0; 1303 dp->dad_ns_ocount = dp->dad_ns_tcount = 0; 1304 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0; 1305 1306 /* Add this to the dadq and add a reference for the dadq. */ 1307 refcount_init(&dp->dad_refcnt, 1); 1308 nd6_dad_add(dp); 1309 NET_EPOCH_ENTER(et); 1310 nd6_dad_starttimer(dp, delay, 0); 1311 NET_EPOCH_EXIT(et); 1312 } 1313 1314 /* 1315 * terminate DAD unconditionally. used for address removals. 1316 */ 1317 void 1318 nd6_dad_stop(struct ifaddr *ifa) 1319 { 1320 struct dadq *dp; 1321 1322 dp = nd6_dad_find(ifa, NULL); 1323 if (!dp) { 1324 /* DAD wasn't started yet */ 1325 return; 1326 } 1327 1328 nd6_dad_stoptimer(dp); 1329 nd6_dad_del(dp); 1330 1331 /* Release this function's reference, acquired by nd6_dad_find(). */ 1332 nd6_dad_rele(dp); 1333 } 1334 1335 static void 1336 nd6_dad_timer(struct dadq *dp) 1337 { 1338 CURVNET_SET(dp->dad_vnet); 1339 struct ifaddr *ifa = dp->dad_ifa; 1340 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1341 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1342 char ip6buf[INET6_ADDRSTRLEN]; 1343 struct epoch_tracker et; 1344 1345 KASSERT(ia != NULL, ("DAD entry %p with no address", dp)); 1346 1347 NET_EPOCH_ENTER(et); 1348 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) { 1349 /* Do not need DAD for ifdisabled interface. */ 1350 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of " 1351 "ND6_IFF_IFDISABLED.\n", ifp->if_xname); 1352 goto err; 1353 } 1354 if (ia->ia6_flags & IN6_IFF_DUPLICATED) { 1355 log(LOG_ERR, "nd6_dad_timer: called with duplicated address " 1356 "%s(%s)\n", 1357 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1358 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1359 goto err; 1360 } 1361 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) { 1362 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address " 1363 "%s(%s)\n", 1364 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1365 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1366 goto err; 1367 } 1368 1369 /* Stop DAD if the interface is down even after dad_maxtry attempts. */ 1370 if ((dp->dad_ns_tcount > V_dad_maxtry) && 1371 (((ifp->if_flags & IFF_UP) == 0) || 1372 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) { 1373 nd6log((LOG_INFO, "%s: could not run DAD " 1374 "because the interface was down or not running.\n", 1375 if_name(ifa->ifa_ifp))); 1376 goto err; 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_starttimer(dp, 1385 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1); 1386 goto done; 1387 } else { 1388 /* 1389 * We have transmitted sufficient number of DAD packets. 1390 * See what we've got. 1391 */ 1392 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) 1393 /* We've seen NS or NA, means DAD has failed. */ 1394 nd6_dad_duplicated(ifa, dp); 1395 else if (V_dad_enhanced != 0 && 1396 dp->dad_ns_lcount > 0 && 1397 dp->dad_ns_lcount > dp->dad_loopbackprobe) { 1398 /* 1399 * Sec. 4.1 in RFC 7527 requires transmission of 1400 * additional probes until the loopback condition 1401 * becomes clear when a looped back probe is detected. 1402 */ 1403 log(LOG_ERR, "%s: a looped back NS message is " 1404 "detected during DAD for %s. " 1405 "Another DAD probes are being sent.\n", 1406 if_name(ifa->ifa_ifp), 1407 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1408 dp->dad_loopbackprobe = dp->dad_ns_lcount; 1409 /* 1410 * Send an NS immediately and increase dad_count by 1411 * V_nd6_mmaxtries - 1. 1412 */ 1413 dp->dad_count = 1414 dp->dad_ns_ocount + V_nd6_mmaxtries - 1; 1415 nd6_dad_starttimer(dp, 1416 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1417 1); 1418 goto done; 1419 } else { 1420 /* 1421 * We are done with DAD. No NA came, no NS came. 1422 * No duplicate address found. Check IFDISABLED flag 1423 * again in case that it is changed between the 1424 * beginning of this function and here. 1425 */ 1426 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0) 1427 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1428 1429 nd6log((LOG_DEBUG, 1430 "%s: DAD complete for %s - no duplicates found\n", 1431 if_name(ifa->ifa_ifp), 1432 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1433 if (dp->dad_ns_lcount > 0) 1434 log(LOG_ERR, "%s: DAD completed while " 1435 "a looped back NS message is detected " 1436 "during DAD for %s.\n", 1437 if_name(ifa->ifa_ifp), 1438 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1439 } 1440 } 1441 err: 1442 nd6_dad_del(dp); 1443 done: 1444 NET_EPOCH_EXIT(et); 1445 CURVNET_RESTORE(); 1446 } 1447 1448 static void 1449 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp) 1450 { 1451 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1452 struct ifnet *ifp; 1453 char ip6buf[INET6_ADDRSTRLEN]; 1454 1455 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: " 1456 "NS in/out/loopback=%d/%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_ns_lcount, 1459 dp->dad_na_icount); 1460 1461 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1462 ia->ia6_flags |= IN6_IFF_DUPLICATED; 1463 1464 ifp = ifa->ifa_ifp; 1465 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n", 1466 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)); 1467 log(LOG_ERR, "%s: manual intervention required\n", 1468 if_name(ifp)); 1469 1470 /* 1471 * If the address is a link-local address formed from an interface 1472 * identifier based on the hardware address which is supposed to be 1473 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP 1474 * operation on the interface SHOULD be disabled. 1475 * [RFC 4862, Section 5.4.5] 1476 */ 1477 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1478 struct in6_addr in6; 1479 1480 /* 1481 * To avoid over-reaction, we only apply this logic when we are 1482 * very sure that hardware addresses are supposed to be unique. 1483 */ 1484 switch (ifp->if_type) { 1485 case IFT_ETHER: 1486 case IFT_ATM: 1487 case IFT_IEEE1394: 1488 case IFT_INFINIBAND: 1489 in6 = ia->ia_addr.sin6_addr; 1490 if (in6_get_hw_ifid(ifp, &in6) == 0 && 1491 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) { 1492 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1493 log(LOG_ERR, "%s: possible hardware address " 1494 "duplication detected, disable IPv6\n", 1495 if_name(ifp)); 1496 } 1497 break; 1498 } 1499 } 1500 } 1501 1502 static void 1503 nd6_dad_ns_output(struct dadq *dp) 1504 { 1505 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa; 1506 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1507 int i; 1508 1509 dp->dad_ns_tcount++; 1510 if ((ifp->if_flags & IFF_UP) == 0) { 1511 return; 1512 } 1513 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 1514 return; 1515 } 1516 1517 dp->dad_ns_ocount++; 1518 if (V_dad_enhanced != 0) { 1519 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) 1520 dp->dad_nonce[i] = arc4random(); 1521 /* 1522 * XXXHRS: Note that in the case that 1523 * DupAddrDetectTransmits > 1, multiple NS messages with 1524 * different nonces can be looped back in an unexpected 1525 * order. The current implementation recognizes only 1526 * the latest nonce on the sender side. Practically it 1527 * should work well in almost all cases. 1528 */ 1529 } 1530 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr, 1531 (uint8_t *)&dp->dad_nonce[0]); 1532 } 1533 1534 static void 1535 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce) 1536 { 1537 struct dadq *dp; 1538 1539 if (ifa == NULL) 1540 panic("ifa == NULL in nd6_dad_ns_input"); 1541 1542 /* Ignore Nonce option when Enhanced DAD is disabled. */ 1543 if (V_dad_enhanced == 0) 1544 ndopt_nonce = NULL; 1545 dp = nd6_dad_find(ifa, ndopt_nonce); 1546 if (dp == NULL) 1547 return; 1548 1549 dp->dad_ns_icount++; 1550 nd6_dad_rele(dp); 1551 } 1552 1553 static void 1554 nd6_dad_na_input(struct ifaddr *ifa) 1555 { 1556 struct dadq *dp; 1557 1558 if (ifa == NULL) 1559 panic("ifa == NULL in nd6_dad_na_input"); 1560 1561 dp = nd6_dad_find(ifa, NULL); 1562 if (dp != NULL) { 1563 dp->dad_na_icount++; 1564 nd6_dad_rele(dp); 1565 } 1566 } 1567