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 #include "opt_mpath.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/eventhandler.h> 45 #include <sys/malloc.h> 46 #include <sys/libkern.h> 47 #include <sys/lock.h> 48 #include <sys/rwlock.h> 49 #include <sys/mbuf.h> 50 #include <sys/socket.h> 51 #include <sys/sockio.h> 52 #include <sys/time.h> 53 #include <sys/kernel.h> 54 #include <sys/errno.h> 55 #include <sys/sysctl.h> 56 #include <sys/syslog.h> 57 #include <sys/queue.h> 58 #include <sys/callout.h> 59 #include <sys/refcount.h> 60 61 #include <net/if.h> 62 #include <net/if_types.h> 63 #include <net/if_dl.h> 64 #include <net/if_var.h> 65 #include <net/route.h> 66 #ifdef RADIX_MPATH 67 #include <net/radix_mpath.h> 68 #endif 69 #include <net/vnet.h> 70 71 #include <netinet/in.h> 72 #include <netinet/in_var.h> 73 #include <net/if_llatbl.h> 74 #include <netinet6/in6_var.h> 75 #include <netinet6/in6_ifattach.h> 76 #include <netinet/ip6.h> 77 #include <netinet6/ip6_var.h> 78 #include <netinet6/scope6_var.h> 79 #include <netinet6/nd6.h> 80 #include <netinet/icmp6.h> 81 #include <netinet/ip_carp.h> 82 #include <netinet6/send.h> 83 84 #define SDL(s) ((struct sockaddr_dl *)s) 85 86 struct dadq; 87 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *); 88 static void nd6_dad_add(struct dadq *dp); 89 static void nd6_dad_del(struct dadq *dp); 90 static void nd6_dad_rele(struct dadq *); 91 static void nd6_dad_starttimer(struct dadq *, int, int); 92 static void nd6_dad_stoptimer(struct dadq *); 93 static void nd6_dad_timer(struct dadq *); 94 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *); 95 static void nd6_dad_ns_output(struct dadq *); 96 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *); 97 static void nd6_dad_na_input(struct ifaddr *); 98 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *, 99 const struct in6_addr *, u_long, int, struct sockaddr *, u_int); 100 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *, 101 const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int); 102 103 VNET_DEFINE_STATIC(int, dad_enhanced) = 1; 104 #define V_dad_enhanced VNET(dad_enhanced) 105 106 SYSCTL_DECL(_net_inet6_ip6); 107 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW, 108 &VNET_NAME(dad_enhanced), 0, 109 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD."); 110 111 VNET_DEFINE_STATIC(int, dad_maxtry) = 15; /* max # of *tries* to 112 transmit DAD packet */ 113 #define V_dad_maxtry VNET(dad_maxtry) 114 115 /* 116 * Input a Neighbor Solicitation Message. 117 * 118 * Based on RFC 2461 119 * Based on RFC 2462 (duplicate address detection) 120 */ 121 void 122 nd6_ns_input(struct mbuf *m, int off, int icmp6len) 123 { 124 struct ifnet *ifp; 125 struct ip6_hdr *ip6; 126 struct nd_neighbor_solicit *nd_ns; 127 struct in6_addr daddr6, myaddr6, saddr6, taddr6; 128 struct ifaddr *ifa; 129 struct sockaddr_dl proxydl; 130 union nd_opts ndopts; 131 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 132 char *lladdr; 133 int anycast, lladdrlen, proxy, rflag, tentative, tlladdr; 134 135 ifa = NULL; 136 137 /* RFC 6980: Nodes MUST silently ignore fragments */ 138 if(m->m_flags & M_FRAGMENTED) 139 goto freeit; 140 141 ifp = m->m_pkthdr.rcvif; 142 ip6 = mtod(m, struct ip6_hdr *); 143 if (ip6->ip6_hlim != 255) { 144 nd6log((LOG_ERR, 145 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n", 146 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 147 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 148 goto bads; 149 } 150 151 if (m->m_len < off + icmp6len) { 152 m = m_pullup(m, off + icmp6len); 153 if (m == NULL) { 154 IP6STAT_INC(ip6s_exthdrtoolong); 155 return; 156 } 157 } 158 ip6 = mtod(m, struct ip6_hdr *); 159 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off); 160 161 saddr6 = ip6->ip6_src; 162 daddr6 = ip6->ip6_dst; 163 taddr6 = nd_ns->nd_ns_target; 164 if (in6_setscope(&taddr6, ifp, NULL) != 0) 165 goto bad; 166 167 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0; 168 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif) 169 rflag = 0; 170 171 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 172 /* dst has to be a solicited node multicast address. */ 173 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL && 174 /* don't check ifindex portion */ 175 daddr6.s6_addr32[1] == 0 && 176 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE && 177 daddr6.s6_addr8[12] == 0xff) { 178 ; /* good */ 179 } else { 180 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 181 "(wrong ip6 dst)\n")); 182 goto bad; 183 } 184 } else if (!V_nd6_onlink_ns_rfc4861) { 185 struct sockaddr_in6 src_sa6; 186 187 /* 188 * According to recent IETF discussions, it is not a good idea 189 * to accept a NS from an address which would not be deemed 190 * to be a neighbor otherwise. This point is expected to be 191 * clarified in future revisions of the specification. 192 */ 193 bzero(&src_sa6, sizeof(src_sa6)); 194 src_sa6.sin6_family = AF_INET6; 195 src_sa6.sin6_len = sizeof(src_sa6); 196 src_sa6.sin6_addr = saddr6; 197 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) { 198 nd6log((LOG_INFO, "nd6_ns_input: " 199 "NS packet from non-neighbor\n")); 200 goto bad; 201 } 202 } 203 204 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 205 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n")); 206 goto bad; 207 } 208 209 icmp6len -= sizeof(*nd_ns); 210 nd6_option_init(nd_ns + 1, icmp6len, &ndopts); 211 if (nd6_options(&ndopts) < 0) { 212 nd6log((LOG_INFO, 213 "nd6_ns_input: invalid ND option, ignored\n")); 214 /* nd6_options have incremented stats */ 215 goto freeit; 216 } 217 218 lladdr = NULL; 219 lladdrlen = 0; 220 if (ndopts.nd_opts_src_lladdr) { 221 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 222 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 223 } 224 225 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) { 226 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 227 "(link-layer address option)\n")); 228 goto bad; 229 } 230 231 /* 232 * Attaching target link-layer address to the NA? 233 * (RFC 2461 7.2.4) 234 * 235 * NS IP dst is unicast/anycast MUST NOT add 236 * NS IP dst is solicited-node multicast MUST add 237 * 238 * In implementation, we add target link-layer address by default. 239 * We do not add one in MUST NOT cases. 240 */ 241 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) 242 tlladdr = 0; 243 else 244 tlladdr = 1; 245 246 /* 247 * Target address (taddr6) must be either: 248 * (1) Valid unicast/anycast address for my receiving interface, 249 * (2) Unicast address for which I'm offering proxy service, or 250 * (3) "tentative" address on which DAD is being performed. 251 */ 252 /* (1) and (3) check. */ 253 if (ifp->if_carp) 254 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 255 else 256 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 257 258 /* (2) check. */ 259 proxy = 0; 260 if (ifa == NULL) { 261 struct sockaddr_dl rt_gateway; 262 struct rt_addrinfo info; 263 struct sockaddr_in6 dst6; 264 265 bzero(&dst6, sizeof(dst6)); 266 dst6.sin6_len = sizeof(struct sockaddr_in6); 267 dst6.sin6_family = AF_INET6; 268 dst6.sin6_addr = taddr6; 269 270 bzero(&rt_gateway, sizeof(rt_gateway)); 271 rt_gateway.sdl_len = sizeof(rt_gateway); 272 bzero(&info, sizeof(info)); 273 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway; 274 275 if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6, 276 0, 0, &info) == 0) { 277 if ((info.rti_flags & RTF_ANNOUNCE) != 0 && 278 rt_gateway.sdl_family == AF_LINK) { 279 280 /* 281 * proxy NDP for single entry 282 */ 283 proxydl = *SDL(&rt_gateway); 284 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal( 285 ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST); 286 if (ifa) 287 proxy = 1; 288 } 289 } 290 } 291 if (ifa == NULL) { 292 /* 293 * We've got an NS packet, and we don't have that adddress 294 * assigned for us. We MUST silently ignore it. 295 * See RFC2461 7.2.3. 296 */ 297 goto freeit; 298 } 299 myaddr6 = *IFA_IN6(ifa); 300 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST; 301 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE; 302 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED) 303 goto freeit; 304 305 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 306 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s " 307 "(if %d, NS packet %d)\n", 308 ip6_sprintf(ip6bufs, &taddr6), 309 ifp->if_addrlen, lladdrlen - 2)); 310 goto bad; 311 } 312 313 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) { 314 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n", 315 ip6_sprintf(ip6bufs, &saddr6))); 316 goto freeit; 317 } 318 319 /* 320 * We have neighbor solicitation packet, with target address equals to 321 * one of my tentative address. 322 * 323 * src addr how to process? 324 * --- --- 325 * multicast of course, invalid (rejected in ip6_input) 326 * unicast somebody is doing address resolution -> ignore 327 * unspec dup address detection 328 * 329 * The processing is defined in RFC 2462. 330 */ 331 if (tentative) { 332 /* 333 * If source address is unspecified address, it is for 334 * duplicate address detection. 335 * 336 * If not, the packet is for addess resolution; 337 * silently ignore it. 338 */ 339 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 340 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce); 341 342 goto freeit; 343 } 344 345 /* 346 * If the source address is unspecified address, entries must not 347 * be created or updated. 348 * It looks that sender is performing DAD. Output NA toward 349 * all-node multicast address, to tell the sender that I'm using 350 * the address. 351 * S bit ("solicited") must be zero. 352 */ 353 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 354 struct in6_addr in6_all; 355 356 in6_all = in6addr_linklocal_allnodes; 357 if (in6_setscope(&in6_all, ifp, NULL) != 0) 358 goto bad; 359 nd6_na_output_fib(ifp, &in6_all, &taddr6, 360 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 361 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL, 362 M_GETFIB(m)); 363 goto freeit; 364 } 365 366 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, 367 ND_NEIGHBOR_SOLICIT, 0); 368 369 nd6_na_output_fib(ifp, &saddr6, &taddr6, 370 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 371 rflag | ND_NA_FLAG_SOLICITED, tlladdr, 372 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m)); 373 freeit: 374 if (ifa != NULL) 375 ifa_free(ifa); 376 m_freem(m); 377 return; 378 379 bad: 380 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n", 381 ip6_sprintf(ip6bufs, &saddr6))); 382 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n", 383 ip6_sprintf(ip6bufs, &daddr6))); 384 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n", 385 ip6_sprintf(ip6bufs, &taddr6))); 386 bads: 387 ICMP6STAT_INC(icp6s_badns); 388 if (ifa != NULL) 389 ifa_free(ifa); 390 m_freem(m); 391 } 392 393 /* 394 * Output a Neighbor Solicitation Message. Caller specifies: 395 * - ICMP6 header source IP6 address 396 * - ND6 header target IP6 address 397 * - ND6 header source datalink address 398 * 399 * Based on RFC 2461 400 * Based on RFC 2462 (duplicate address detection) 401 * 402 * ln - for source address determination 403 * nonce - If non-NULL, NS is used for duplicate address detection and 404 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce. 405 */ 406 static void 407 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6, 408 const struct in6_addr *daddr6, const struct in6_addr *taddr6, 409 uint8_t *nonce, u_int fibnum) 410 { 411 struct mbuf *m; 412 struct m_tag *mtag; 413 struct ip6_hdr *ip6; 414 struct nd_neighbor_solicit *nd_ns; 415 struct ip6_moptions im6o; 416 int icmp6len; 417 int maxlen; 418 caddr_t mac; 419 420 if (IN6_IS_ADDR_MULTICAST(taddr6)) 421 return; 422 423 /* estimate the size of message */ 424 maxlen = sizeof(*ip6) + sizeof(*nd_ns); 425 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; 426 KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( 427 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", 428 __func__, max_linkhdr, maxlen, MCLBYTES)); 429 430 if (max_linkhdr + maxlen > MHLEN) 431 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 432 else 433 m = m_gethdr(M_NOWAIT, MT_DATA); 434 if (m == NULL) 435 return; 436 M_SETFIB(m, fibnum); 437 438 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) { 439 m->m_flags |= M_MCAST; 440 im6o.im6o_multicast_ifp = ifp; 441 im6o.im6o_multicast_hlim = 255; 442 im6o.im6o_multicast_loop = 0; 443 } 444 445 icmp6len = sizeof(*nd_ns); 446 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len; 447 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ 448 449 /* fill neighbor solicitation packet */ 450 ip6 = mtod(m, struct ip6_hdr *); 451 ip6->ip6_flow = 0; 452 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 453 ip6->ip6_vfc |= IPV6_VERSION; 454 /* ip6->ip6_plen will be set later */ 455 ip6->ip6_nxt = IPPROTO_ICMPV6; 456 ip6->ip6_hlim = 255; 457 if (daddr6) 458 ip6->ip6_dst = *daddr6; 459 else { 460 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL; 461 ip6->ip6_dst.s6_addr16[1] = 0; 462 ip6->ip6_dst.s6_addr32[1] = 0; 463 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE; 464 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3]; 465 ip6->ip6_dst.s6_addr8[12] = 0xff; 466 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 467 goto bad; 468 } 469 if (nonce == NULL) { 470 struct ifaddr *ifa = NULL; 471 472 /* 473 * RFC2461 7.2.2: 474 * "If the source address of the packet prompting the 475 * solicitation is the same as one of the addresses assigned 476 * to the outgoing interface, that address SHOULD be placed 477 * in the IP Source Address of the outgoing solicitation. 478 * Otherwise, any one of the addresses assigned to the 479 * interface should be used." 480 * 481 * We use the source address for the prompting packet 482 * (saddr6), if saddr6 belongs to the outgoing interface. 483 * Otherwise, we perform the source address selection as usual. 484 */ 485 486 if (saddr6 != NULL) 487 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6); 488 if (ifa != NULL) { 489 /* ip6_src set already. */ 490 ip6->ip6_src = *saddr6; 491 ifa_free(ifa); 492 } else { 493 int error; 494 struct in6_addr dst6, src6; 495 uint32_t scopeid; 496 497 in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid); 498 error = in6_selectsrc_addr(fibnum, &dst6, 499 scopeid, ifp, &src6, NULL); 500 if (error) { 501 char ip6buf[INET6_ADDRSTRLEN]; 502 nd6log((LOG_DEBUG, "%s: source can't be " 503 "determined: dst=%s, error=%d\n", __func__, 504 ip6_sprintf(ip6buf, &dst6), 505 error)); 506 goto bad; 507 } 508 ip6->ip6_src = src6; 509 } 510 } else { 511 /* 512 * Source address for DAD packet must always be IPv6 513 * unspecified address. (0::0) 514 * We actually don't have to 0-clear the address (we did it 515 * above), but we do so here explicitly to make the intention 516 * clearer. 517 */ 518 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src)); 519 } 520 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1); 521 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT; 522 nd_ns->nd_ns_code = 0; 523 nd_ns->nd_ns_reserved = 0; 524 nd_ns->nd_ns_target = *taddr6; 525 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */ 526 527 /* 528 * Add source link-layer address option. 529 * 530 * spec implementation 531 * --- --- 532 * DAD packet MUST NOT do not add the option 533 * there's no link layer address: 534 * impossible do not add the option 535 * there's link layer address: 536 * Multicast NS MUST add one add the option 537 * Unicast NS SHOULD add one add the option 538 */ 539 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) { 540 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 541 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); 542 /* 8 byte alignments... */ 543 optlen = (optlen + 7) & ~7; 544 545 m->m_pkthdr.len += optlen; 546 m->m_len += optlen; 547 icmp6len += optlen; 548 bzero((caddr_t)nd_opt, optlen); 549 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; 550 nd_opt->nd_opt_len = optlen >> 3; 551 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); 552 } 553 /* 554 * Add a Nonce option (RFC 3971) to detect looped back NS messages. 555 * This behavior is documented as Enhanced Duplicate Address 556 * Detection in RFC 7527. 557 * net.inet6.ip6.dad_enhanced=0 disables this. 558 */ 559 if (V_dad_enhanced != 0 && nonce != NULL) { 560 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN; 561 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); 562 /* 8-byte alignment is required. */ 563 optlen = (optlen + 7) & ~7; 564 565 m->m_pkthdr.len += optlen; 566 m->m_len += optlen; 567 icmp6len += optlen; 568 bzero((caddr_t)nd_opt, optlen); 569 nd_opt->nd_opt_type = ND_OPT_NONCE; 570 nd_opt->nd_opt_len = optlen >> 3; 571 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN); 572 } 573 ip6->ip6_plen = htons((u_short)icmp6len); 574 nd_ns->nd_ns_cksum = 0; 575 nd_ns->nd_ns_cksum = 576 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len); 577 578 if (send_sendso_input_hook != NULL) { 579 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 580 sizeof(unsigned short), M_NOWAIT); 581 if (mtag == NULL) 582 goto bad; 583 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type; 584 m_tag_prepend(m, mtag); 585 } 586 587 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0, 588 &im6o, NULL, NULL); 589 icmp6_ifstat_inc(ifp, ifs6_out_msg); 590 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit); 591 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]); 592 593 return; 594 595 bad: 596 m_freem(m); 597 } 598 599 #ifndef BURN_BRIDGES 600 void 601 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6, 602 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce) 603 { 604 605 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB); 606 } 607 #endif 608 /* 609 * Neighbor advertisement input handling. 610 * 611 * Based on RFC 2461 612 * Based on RFC 2462 (duplicate address detection) 613 * 614 * the following items are not implemented yet: 615 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 616 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 617 */ 618 void 619 nd6_na_input(struct mbuf *m, int off, int icmp6len) 620 { 621 struct ifnet *ifp; 622 struct ip6_hdr *ip6; 623 struct ifaddr *ifa; 624 struct llentry *ln; 625 struct mbuf *chain; 626 struct nd_neighbor_advert *nd_na; 627 struct in6_addr daddr6, taddr6; 628 struct sockaddr_in6 sin6; 629 union nd_opts ndopts; 630 u_char linkhdr[LLE_MAX_LINKHDR]; 631 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 632 char *lladdr; 633 size_t linkhdrsize; 634 int flags, is_override, is_router, is_solicited; 635 int lladdr_off, lladdrlen, checklink; 636 637 NET_EPOCH_ASSERT(); 638 639 chain = NULL; 640 ln = NULL; 641 checklink = 0; 642 643 /* RFC 6980: Nodes MUST silently ignore fragments */ 644 if(m->m_flags & M_FRAGMENTED) 645 goto freeit; 646 647 ifp = m->m_pkthdr.rcvif; 648 ip6 = mtod(m, struct ip6_hdr *); 649 if (ip6->ip6_hlim != 255) { 650 nd6log((LOG_ERR, 651 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n", 652 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 653 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 654 goto bad; 655 } 656 657 if (m->m_len < off + icmp6len) { 658 m = m_pullup(m, off + icmp6len); 659 if (m == NULL) { 660 IP6STAT_INC(ip6s_exthdrtoolong); 661 return; 662 } 663 } 664 ip6 = mtod(m, struct ip6_hdr *); 665 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off); 666 667 flags = nd_na->nd_na_flags_reserved; 668 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0); 669 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0); 670 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0); 671 672 taddr6 = nd_na->nd_na_target; 673 if (in6_setscope(&taddr6, ifp, NULL)) 674 goto bad; /* XXX: impossible */ 675 676 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 677 nd6log((LOG_ERR, 678 "nd6_na_input: invalid target address %s\n", 679 ip6_sprintf(ip6bufs, &taddr6))); 680 goto bad; 681 } 682 683 daddr6 = ip6->ip6_dst; 684 if (IN6_IS_ADDR_MULTICAST(&daddr6)) 685 if (is_solicited) { 686 nd6log((LOG_ERR, 687 "nd6_na_input: a solicited adv is multicasted\n")); 688 goto bad; 689 } 690 691 icmp6len -= sizeof(*nd_na); 692 nd6_option_init(nd_na + 1, icmp6len, &ndopts); 693 if (nd6_options(&ndopts) < 0) { 694 nd6log((LOG_INFO, 695 "nd6_na_input: invalid ND option, ignored\n")); 696 /* nd6_options have incremented stats */ 697 goto freeit; 698 } 699 700 lladdr = NULL; 701 lladdrlen = 0; 702 if (ndopts.nd_opts_tgt_lladdr) { 703 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 704 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 705 } 706 707 /* 708 * This effectively disables the DAD check on a non-master CARP 709 * address. 710 */ 711 if (ifp->if_carp) 712 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 713 else 714 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 715 716 /* 717 * Target address matches one of my interface address. 718 * 719 * If my address is tentative, this means that there's somebody 720 * already using the same address as mine. This indicates DAD failure. 721 * This is defined in RFC 2462. 722 * 723 * Otherwise, process as defined in RFC 2461. 724 */ 725 if (ifa 726 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) { 727 nd6_dad_na_input(ifa); 728 ifa_free(ifa); 729 goto freeit; 730 } 731 732 /* Just for safety, maybe unnecessary. */ 733 if (ifa) { 734 ifa_free(ifa); 735 log(LOG_ERR, 736 "nd6_na_input: duplicate IP6 address %s\n", 737 ip6_sprintf(ip6bufs, &taddr6)); 738 goto freeit; 739 } 740 741 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 742 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s " 743 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6), 744 ifp->if_addrlen, lladdrlen - 2)); 745 goto bad; 746 } 747 748 /* 749 * If no neighbor cache entry is found, NA SHOULD silently be 750 * discarded. 751 */ 752 ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp); 753 if (ln == NULL) { 754 goto freeit; 755 } 756 757 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 758 /* 759 * If the link-layer has address, and no lladdr option came, 760 * discard the packet. 761 */ 762 if (ifp->if_addrlen && lladdr == NULL) { 763 goto freeit; 764 } 765 766 /* 767 * Record link-layer address, and update the state. 768 */ 769 linkhdrsize = sizeof(linkhdr); 770 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 771 linkhdr, &linkhdrsize, &lladdr_off) != 0) 772 return; 773 774 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 775 lladdr_off) == 0) { 776 ln = NULL; 777 goto freeit; 778 } 779 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 780 if (is_solicited) 781 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 782 else 783 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 784 if ((ln->ln_router = is_router) != 0) { 785 /* 786 * This means a router's state has changed from 787 * non-reachable to probably reachable, and might 788 * affect the status of associated prefixes.. 789 */ 790 checklink = 1; 791 } 792 } else { 793 int llchange; 794 795 /* 796 * Check if the link-layer address has changed or not. 797 */ 798 if (lladdr == NULL) 799 llchange = 0; 800 else { 801 if (ln->la_flags & LLE_VALID) { 802 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen)) 803 llchange = 1; 804 else 805 llchange = 0; 806 } else 807 llchange = 1; 808 } 809 810 /* 811 * This is VERY complex. Look at it with care. 812 * 813 * override solicit lladdr llchange action 814 * (L: record lladdr) 815 * 816 * 0 0 n -- (2c) 817 * 0 0 y n (2b) L 818 * 0 0 y y (1) REACHABLE->STALE 819 * 0 1 n -- (2c) *->REACHABLE 820 * 0 1 y n (2b) L *->REACHABLE 821 * 0 1 y y (1) REACHABLE->STALE 822 * 1 0 n -- (2a) 823 * 1 0 y n (2a) L 824 * 1 0 y y (2a) L *->STALE 825 * 1 1 n -- (2a) *->REACHABLE 826 * 1 1 y n (2a) L *->REACHABLE 827 * 1 1 y y (2a) L *->REACHABLE 828 */ 829 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */ 830 /* 831 * If state is REACHABLE, make it STALE. 832 * no other updates should be done. 833 */ 834 if (ln->ln_state == ND6_LLINFO_REACHABLE) 835 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 836 goto freeit; 837 } else if (is_override /* (2a) */ 838 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */ 839 || lladdr == NULL) { /* (2c) */ 840 /* 841 * Update link-local address, if any. 842 */ 843 if (lladdr != NULL) { 844 linkhdrsize = sizeof(linkhdr); 845 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 846 linkhdr, &linkhdrsize, &lladdr_off) != 0) 847 goto freeit; 848 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, 849 linkhdrsize, lladdr_off) == 0) { 850 ln = NULL; 851 goto freeit; 852 } 853 EVENTHANDLER_INVOKE(lle_event, ln, 854 LLENTRY_RESOLVED); 855 } 856 857 /* 858 * If solicited, make the state REACHABLE. 859 * If not solicited and the link-layer address was 860 * changed, make it STALE. 861 */ 862 if (is_solicited) 863 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 864 else { 865 if (lladdr != NULL && llchange) 866 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 867 } 868 } 869 870 if (ln->ln_router && !is_router) { 871 /* 872 * The peer dropped the router flag. 873 * Remove the sender from the Default Router List and 874 * update the Destination Cache entries. 875 */ 876 struct ifnet *nd6_ifp; 877 878 nd6_ifp = lltable_get_ifp(ln->lle_tbl); 879 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) && 880 (ND_IFINFO(nd6_ifp)->flags & 881 ND6_IFF_ACCEPT_RTADV) != 0) 882 /* 883 * Even if the neighbor is not in the default 884 * router list, the neighbor may be used as a 885 * next hop for some destinations (e.g. redirect 886 * case). So we must call rt6_flush explicitly. 887 */ 888 rt6_flush(&ip6->ip6_src, ifp); 889 } 890 ln->ln_router = is_router; 891 } 892 /* XXX - QL 893 * Does this matter? 894 * rt->rt_flags &= ~RTF_REJECT; 895 */ 896 ln->la_asked = 0; 897 if (ln->la_hold != NULL) { 898 memset(&sin6, 0, sizeof(sin6)); 899 nd6_grab_holdchain(ln, &chain, &sin6); 900 } 901 freeit: 902 if (ln != NULL) 903 LLE_WUNLOCK(ln); 904 905 if (chain != NULL) 906 nd6_flush_holdchain(ifp, chain, &sin6); 907 908 if (checklink) 909 pfxlist_onlink_check(); 910 911 m_freem(m); 912 return; 913 914 bad: 915 if (ln != NULL) 916 LLE_WUNLOCK(ln); 917 918 ICMP6STAT_INC(icp6s_badna); 919 m_freem(m); 920 } 921 922 /* 923 * Neighbor advertisement output handling. 924 * 925 * Based on RFC 2461 926 * 927 * the following items are not implemented yet: 928 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 929 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 930 * 931 * tlladdr - 1 if include target link-layer address 932 * sdl0 - sockaddr_dl (= proxy NA) or NULL 933 */ 934 static void 935 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0, 936 const struct in6_addr *taddr6, u_long flags, int tlladdr, 937 struct sockaddr *sdl0, u_int fibnum) 938 { 939 struct mbuf *m; 940 struct m_tag *mtag; 941 struct ip6_hdr *ip6; 942 struct nd_neighbor_advert *nd_na; 943 struct ip6_moptions im6o; 944 struct in6_addr daddr6, dst6, src6; 945 uint32_t scopeid; 946 947 int icmp6len, maxlen, error; 948 caddr_t mac = NULL; 949 950 daddr6 = *daddr6_0; /* make a local copy for modification */ 951 952 /* estimate the size of message */ 953 maxlen = sizeof(*ip6) + sizeof(*nd_na); 954 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; 955 KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( 956 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", 957 __func__, max_linkhdr, maxlen, MCLBYTES)); 958 959 if (max_linkhdr + maxlen > MHLEN) 960 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 961 else 962 m = m_gethdr(M_NOWAIT, MT_DATA); 963 if (m == NULL) 964 return; 965 M_SETFIB(m, fibnum); 966 967 if (IN6_IS_ADDR_MULTICAST(&daddr6)) { 968 m->m_flags |= M_MCAST; 969 im6o.im6o_multicast_ifp = ifp; 970 im6o.im6o_multicast_hlim = 255; 971 im6o.im6o_multicast_loop = 0; 972 } 973 974 icmp6len = sizeof(*nd_na); 975 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len; 976 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ 977 978 /* fill neighbor advertisement packet */ 979 ip6 = mtod(m, struct ip6_hdr *); 980 ip6->ip6_flow = 0; 981 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 982 ip6->ip6_vfc |= IPV6_VERSION; 983 ip6->ip6_nxt = IPPROTO_ICMPV6; 984 ip6->ip6_hlim = 255; 985 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) { 986 /* reply to DAD */ 987 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL; 988 daddr6.s6_addr16[1] = 0; 989 daddr6.s6_addr32[1] = 0; 990 daddr6.s6_addr32[2] = 0; 991 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE; 992 if (in6_setscope(&daddr6, ifp, NULL)) 993 goto bad; 994 995 flags &= ~ND_NA_FLAG_SOLICITED; 996 } 997 ip6->ip6_dst = daddr6; 998 999 /* 1000 * Select a source whose scope is the same as that of the dest. 1001 */ 1002 in6_splitscope(&daddr6, &dst6, &scopeid); 1003 error = in6_selectsrc_addr(fibnum, &dst6, 1004 scopeid, ifp, &src6, NULL); 1005 if (error) { 1006 char ip6buf[INET6_ADDRSTRLEN]; 1007 nd6log((LOG_DEBUG, "nd6_na_output: source can't be " 1008 "determined: dst=%s, error=%d\n", 1009 ip6_sprintf(ip6buf, &daddr6), error)); 1010 goto bad; 1011 } 1012 ip6->ip6_src = src6; 1013 nd_na = (struct nd_neighbor_advert *)(ip6 + 1); 1014 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT; 1015 nd_na->nd_na_code = 0; 1016 nd_na->nd_na_target = *taddr6; 1017 in6_clearscope(&nd_na->nd_na_target); /* XXX */ 1018 1019 /* 1020 * "tlladdr" indicates NS's condition for adding tlladdr or not. 1021 * see nd6_ns_input() for details. 1022 * Basically, if NS packet is sent to unicast/anycast addr, 1023 * target lladdr option SHOULD NOT be included. 1024 */ 1025 if (tlladdr) { 1026 /* 1027 * sdl0 != NULL indicates proxy NA. If we do proxy, use 1028 * lladdr in sdl0. If we are not proxying (sending NA for 1029 * my address) use lladdr configured for the interface. 1030 */ 1031 if (sdl0 == NULL) { 1032 if (ifp->if_carp) 1033 mac = (*carp_macmatch6_p)(ifp, m, taddr6); 1034 if (mac == NULL) 1035 mac = nd6_ifptomac(ifp); 1036 } else if (sdl0->sa_family == AF_LINK) { 1037 struct sockaddr_dl *sdl; 1038 sdl = (struct sockaddr_dl *)sdl0; 1039 if (sdl->sdl_alen == ifp->if_addrlen) 1040 mac = LLADDR(sdl); 1041 } 1042 } 1043 if (tlladdr && mac) { 1044 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 1045 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1); 1046 1047 /* roundup to 8 bytes alignment! */ 1048 optlen = (optlen + 7) & ~7; 1049 1050 m->m_pkthdr.len += optlen; 1051 m->m_len += optlen; 1052 icmp6len += optlen; 1053 bzero((caddr_t)nd_opt, optlen); 1054 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 1055 nd_opt->nd_opt_len = optlen >> 3; 1056 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); 1057 } else 1058 flags &= ~ND_NA_FLAG_OVERRIDE; 1059 1060 ip6->ip6_plen = htons((u_short)icmp6len); 1061 nd_na->nd_na_flags_reserved = flags; 1062 nd_na->nd_na_cksum = 0; 1063 nd_na->nd_na_cksum = 1064 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len); 1065 1066 if (send_sendso_input_hook != NULL) { 1067 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 1068 sizeof(unsigned short), M_NOWAIT); 1069 if (mtag == NULL) 1070 goto bad; 1071 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type; 1072 m_tag_prepend(m, mtag); 1073 } 1074 1075 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL); 1076 icmp6_ifstat_inc(ifp, ifs6_out_msg); 1077 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert); 1078 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]); 1079 1080 return; 1081 1082 bad: 1083 m_freem(m); 1084 } 1085 1086 #ifndef BURN_BRIDGES 1087 void 1088 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0, 1089 const struct in6_addr *taddr6, u_long flags, int tlladdr, 1090 struct sockaddr *sdl0) 1091 { 1092 1093 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0, 1094 RT_DEFAULT_FIB); 1095 } 1096 #endif 1097 1098 caddr_t 1099 nd6_ifptomac(struct ifnet *ifp) 1100 { 1101 switch (ifp->if_type) { 1102 case IFT_ETHER: 1103 case IFT_IEEE1394: 1104 case IFT_L2VLAN: 1105 case IFT_INFINIBAND: 1106 case IFT_BRIDGE: 1107 return IF_LLADDR(ifp); 1108 default: 1109 return NULL; 1110 } 1111 } 1112 1113 struct dadq { 1114 TAILQ_ENTRY(dadq) dad_list; 1115 struct ifaddr *dad_ifa; 1116 int dad_count; /* max NS to send */ 1117 int dad_ns_tcount; /* # of trials to send NS */ 1118 int dad_ns_ocount; /* NS sent so far */ 1119 int dad_ns_icount; 1120 int dad_na_icount; 1121 int dad_ns_lcount; /* looped back NS */ 1122 int dad_loopbackprobe; /* probing state for loopback detection */ 1123 struct callout dad_timer_ch; 1124 struct vnet *dad_vnet; 1125 u_int dad_refcnt; 1126 #define ND_OPT_NONCE_LEN32 \ 1127 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t)) 1128 uint32_t dad_nonce[ND_OPT_NONCE_LEN32]; 1129 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */ 1130 }; 1131 1132 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq); 1133 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock); 1134 #define V_dadq VNET(dadq) 1135 #define V_dad_rwlock VNET(dad_rwlock) 1136 1137 #define DADQ_RLOCK() rw_rlock(&V_dad_rwlock) 1138 #define DADQ_RUNLOCK() rw_runlock(&V_dad_rwlock) 1139 #define DADQ_WLOCK() rw_wlock(&V_dad_rwlock) 1140 #define DADQ_WUNLOCK() rw_wunlock(&V_dad_rwlock) 1141 1142 static void 1143 nd6_dad_add(struct dadq *dp) 1144 { 1145 1146 DADQ_WLOCK(); 1147 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list); 1148 dp->dad_ondadq = true; 1149 DADQ_WUNLOCK(); 1150 } 1151 1152 static void 1153 nd6_dad_del(struct dadq *dp) 1154 { 1155 1156 DADQ_WLOCK(); 1157 if (dp->dad_ondadq) { 1158 /* 1159 * Remove dp from the dadq and release the dadq's 1160 * reference. 1161 */ 1162 TAILQ_REMOVE(&V_dadq, dp, dad_list); 1163 dp->dad_ondadq = false; 1164 DADQ_WUNLOCK(); 1165 nd6_dad_rele(dp); 1166 } else 1167 DADQ_WUNLOCK(); 1168 } 1169 1170 static struct dadq * 1171 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n) 1172 { 1173 struct dadq *dp; 1174 1175 DADQ_RLOCK(); 1176 TAILQ_FOREACH(dp, &V_dadq, dad_list) { 1177 if (dp->dad_ifa != ifa) 1178 continue; 1179 /* 1180 * Skip if the nonce matches the received one. 1181 * +2 in the length is required because of type and 1182 * length fields are included in a header. 1183 */ 1184 if (n != NULL && 1185 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 && 1186 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0], 1187 ND_OPT_NONCE_LEN) == 0) { 1188 dp->dad_ns_lcount++; 1189 continue; 1190 } 1191 refcount_acquire(&dp->dad_refcnt); 1192 break; 1193 } 1194 DADQ_RUNLOCK(); 1195 1196 return (dp); 1197 } 1198 1199 static void 1200 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns) 1201 { 1202 1203 NET_EPOCH_ASSERT(); 1204 1205 if (send_ns != 0) 1206 nd6_dad_ns_output(dp); 1207 callout_reset(&dp->dad_timer_ch, ticks, 1208 (void (*)(void *))nd6_dad_timer, (void *)dp); 1209 } 1210 1211 static void 1212 nd6_dad_stoptimer(struct dadq *dp) 1213 { 1214 1215 callout_drain(&dp->dad_timer_ch); 1216 } 1217 1218 static void 1219 nd6_dad_rele(struct dadq *dp) 1220 { 1221 1222 if (refcount_release(&dp->dad_refcnt)) { 1223 ifa_free(dp->dad_ifa); 1224 free(dp, M_IP6NDP); 1225 } 1226 } 1227 1228 void 1229 nd6_dad_init(void) 1230 { 1231 1232 rw_init(&V_dad_rwlock, "nd6 DAD queue"); 1233 TAILQ_INIT(&V_dadq); 1234 } 1235 1236 /* 1237 * Start Duplicate Address Detection (DAD) for specified interface address. 1238 */ 1239 void 1240 nd6_dad_start(struct ifaddr *ifa, int delay) 1241 { 1242 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1243 struct dadq *dp; 1244 char ip6buf[INET6_ADDRSTRLEN]; 1245 struct epoch_tracker et; 1246 1247 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0, 1248 ("starting DAD on non-tentative address %p", ifa)); 1249 1250 /* 1251 * If we don't need DAD, don't do it. 1252 * There are several cases: 1253 * - DAD is disabled globally or on the interface 1254 * - the interface address is anycast 1255 */ 1256 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 || 1257 V_ip6_dad_count == 0 || 1258 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) { 1259 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1260 return; 1261 } 1262 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 || 1263 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || 1264 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0) 1265 return; 1266 1267 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) { 1268 /* 1269 * DAD is already in progress. Let the existing entry 1270 * finish it. 1271 */ 1272 nd6_dad_rele(dp); 1273 return; 1274 } 1275 1276 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO); 1277 if (dp == NULL) { 1278 log(LOG_ERR, "nd6_dad_start: memory allocation failed for " 1279 "%s(%s)\n", 1280 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1281 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1282 return; 1283 } 1284 callout_init(&dp->dad_timer_ch, 0); 1285 #ifdef VIMAGE 1286 dp->dad_vnet = curvnet; 1287 #endif 1288 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp), 1289 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1290 1291 /* 1292 * Send NS packet for DAD, ip6_dad_count times. 1293 * Note that we must delay the first transmission, if this is the 1294 * first packet to be sent from the interface after interface 1295 * (re)initialization. 1296 */ 1297 dp->dad_ifa = ifa; 1298 ifa_ref(dp->dad_ifa); 1299 dp->dad_count = V_ip6_dad_count; 1300 dp->dad_ns_icount = dp->dad_na_icount = 0; 1301 dp->dad_ns_ocount = dp->dad_ns_tcount = 0; 1302 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0; 1303 1304 /* Add this to the dadq and add a reference for the dadq. */ 1305 refcount_init(&dp->dad_refcnt, 1); 1306 nd6_dad_add(dp); 1307 NET_EPOCH_ENTER(et); 1308 nd6_dad_starttimer(dp, delay, 0); 1309 NET_EPOCH_EXIT(et); 1310 } 1311 1312 /* 1313 * terminate DAD unconditionally. used for address removals. 1314 */ 1315 void 1316 nd6_dad_stop(struct ifaddr *ifa) 1317 { 1318 struct dadq *dp; 1319 1320 dp = nd6_dad_find(ifa, NULL); 1321 if (!dp) { 1322 /* DAD wasn't started yet */ 1323 return; 1324 } 1325 1326 nd6_dad_stoptimer(dp); 1327 nd6_dad_del(dp); 1328 1329 /* Release this function's reference, acquired by nd6_dad_find(). */ 1330 nd6_dad_rele(dp); 1331 } 1332 1333 static void 1334 nd6_dad_timer(struct dadq *dp) 1335 { 1336 CURVNET_SET(dp->dad_vnet); 1337 struct ifaddr *ifa = dp->dad_ifa; 1338 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1339 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1340 char ip6buf[INET6_ADDRSTRLEN]; 1341 struct epoch_tracker et; 1342 1343 KASSERT(ia != NULL, ("DAD entry %p with no address", dp)); 1344 1345 NET_EPOCH_ENTER(et); 1346 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) { 1347 /* Do not need DAD for ifdisabled interface. */ 1348 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of " 1349 "ND6_IFF_IFDISABLED.\n", ifp->if_xname); 1350 goto err; 1351 } 1352 if (ia->ia6_flags & IN6_IFF_DUPLICATED) { 1353 log(LOG_ERR, "nd6_dad_timer: called with duplicated address " 1354 "%s(%s)\n", 1355 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1356 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1357 goto err; 1358 } 1359 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) { 1360 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address " 1361 "%s(%s)\n", 1362 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1363 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1364 goto err; 1365 } 1366 1367 /* Stop DAD if the interface is down even after dad_maxtry attempts. */ 1368 if ((dp->dad_ns_tcount > V_dad_maxtry) && 1369 (((ifp->if_flags & IFF_UP) == 0) || 1370 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) { 1371 nd6log((LOG_INFO, "%s: could not run DAD " 1372 "because the interface was down or not running.\n", 1373 if_name(ifa->ifa_ifp))); 1374 goto err; 1375 } 1376 1377 /* Need more checks? */ 1378 if (dp->dad_ns_ocount < dp->dad_count) { 1379 /* 1380 * We have more NS to go. Send NS packet for DAD. 1381 */ 1382 nd6_dad_starttimer(dp, 1383 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1); 1384 goto done; 1385 } else { 1386 /* 1387 * We have transmitted sufficient number of DAD packets. 1388 * See what we've got. 1389 */ 1390 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) 1391 /* We've seen NS or NA, means DAD has failed. */ 1392 nd6_dad_duplicated(ifa, dp); 1393 else if (V_dad_enhanced != 0 && 1394 dp->dad_ns_lcount > 0 && 1395 dp->dad_ns_lcount > dp->dad_loopbackprobe) { 1396 /* 1397 * Sec. 4.1 in RFC 7527 requires transmission of 1398 * additional probes until the loopback condition 1399 * becomes clear when a looped back probe is detected. 1400 */ 1401 log(LOG_ERR, "%s: a looped back NS message is " 1402 "detected during DAD for %s. " 1403 "Another DAD probes are being sent.\n", 1404 if_name(ifa->ifa_ifp), 1405 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1406 dp->dad_loopbackprobe = dp->dad_ns_lcount; 1407 /* 1408 * Send an NS immediately and increase dad_count by 1409 * V_nd6_mmaxtries - 1. 1410 */ 1411 dp->dad_count = 1412 dp->dad_ns_ocount + V_nd6_mmaxtries - 1; 1413 nd6_dad_starttimer(dp, 1414 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1415 1); 1416 goto done; 1417 } else { 1418 /* 1419 * We are done with DAD. No NA came, no NS came. 1420 * No duplicate address found. Check IFDISABLED flag 1421 * again in case that it is changed between the 1422 * beginning of this function and here. 1423 */ 1424 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0) 1425 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1426 1427 nd6log((LOG_DEBUG, 1428 "%s: DAD complete for %s - no duplicates found\n", 1429 if_name(ifa->ifa_ifp), 1430 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1431 if (dp->dad_ns_lcount > 0) 1432 log(LOG_ERR, "%s: DAD completed while " 1433 "a looped back NS message is detected " 1434 "during DAD for %s.\n", 1435 if_name(ifa->ifa_ifp), 1436 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1437 } 1438 } 1439 err: 1440 nd6_dad_del(dp); 1441 done: 1442 NET_EPOCH_EXIT(et); 1443 CURVNET_RESTORE(); 1444 } 1445 1446 static void 1447 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp) 1448 { 1449 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1450 struct ifnet *ifp; 1451 char ip6buf[INET6_ADDRSTRLEN]; 1452 1453 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: " 1454 "NS in/out/loopback=%d/%d/%d, NA in=%d\n", 1455 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1456 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount, 1457 dp->dad_na_icount); 1458 1459 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1460 ia->ia6_flags |= IN6_IFF_DUPLICATED; 1461 1462 ifp = ifa->ifa_ifp; 1463 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n", 1464 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)); 1465 log(LOG_ERR, "%s: manual intervention required\n", 1466 if_name(ifp)); 1467 1468 /* 1469 * If the address is a link-local address formed from an interface 1470 * identifier based on the hardware address which is supposed to be 1471 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP 1472 * operation on the interface SHOULD be disabled. 1473 * [RFC 4862, Section 5.4.5] 1474 */ 1475 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1476 struct in6_addr in6; 1477 1478 /* 1479 * To avoid over-reaction, we only apply this logic when we are 1480 * very sure that hardware addresses are supposed to be unique. 1481 */ 1482 switch (ifp->if_type) { 1483 case IFT_ETHER: 1484 case IFT_ATM: 1485 case IFT_IEEE1394: 1486 case IFT_INFINIBAND: 1487 in6 = ia->ia_addr.sin6_addr; 1488 if (in6_get_hw_ifid(ifp, &in6) == 0 && 1489 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) { 1490 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1491 log(LOG_ERR, "%s: possible hardware address " 1492 "duplication detected, disable IPv6\n", 1493 if_name(ifp)); 1494 } 1495 break; 1496 } 1497 } 1498 } 1499 1500 static void 1501 nd6_dad_ns_output(struct dadq *dp) 1502 { 1503 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa; 1504 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1505 int i; 1506 1507 dp->dad_ns_tcount++; 1508 if ((ifp->if_flags & IFF_UP) == 0) { 1509 return; 1510 } 1511 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 1512 return; 1513 } 1514 1515 dp->dad_ns_ocount++; 1516 if (V_dad_enhanced != 0) { 1517 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) 1518 dp->dad_nonce[i] = arc4random(); 1519 /* 1520 * XXXHRS: Note that in the case that 1521 * DupAddrDetectTransmits > 1, multiple NS messages with 1522 * different nonces can be looped back in an unexpected 1523 * order. The current implementation recognizes only 1524 * the latest nonce on the sender side. Practically it 1525 * should work well in almost all cases. 1526 */ 1527 } 1528 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr, 1529 (uint8_t *)&dp->dad_nonce[0]); 1530 } 1531 1532 static void 1533 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce) 1534 { 1535 struct dadq *dp; 1536 1537 if (ifa == NULL) 1538 panic("ifa == NULL in nd6_dad_ns_input"); 1539 1540 /* Ignore Nonce option when Enhanced DAD is disabled. */ 1541 if (V_dad_enhanced == 0) 1542 ndopt_nonce = NULL; 1543 dp = nd6_dad_find(ifa, ndopt_nonce); 1544 if (dp == NULL) 1545 return; 1546 1547 dp->dad_ns_icount++; 1548 nd6_dad_rele(dp); 1549 } 1550 1551 static void 1552 nd6_dad_na_input(struct ifaddr *ifa) 1553 { 1554 struct dadq *dp; 1555 1556 if (ifa == NULL) 1557 panic("ifa == NULL in nd6_dad_na_input"); 1558 1559 dp = nd6_dad_find(ifa, NULL); 1560 if (dp != NULL) { 1561 dp->dad_na_icount++; 1562 nd6_dad_rele(dp); 1563 } 1564 } 1565