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 = m->m_pkthdr.rcvif; 125 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 126 struct nd_neighbor_solicit *nd_ns; 127 struct in6_addr saddr6 = ip6->ip6_src; 128 struct in6_addr daddr6 = ip6->ip6_dst; 129 struct in6_addr taddr6; 130 struct in6_addr myaddr6; 131 char *lladdr = NULL; 132 struct ifaddr *ifa = NULL; 133 int lladdrlen = 0; 134 int anycast = 0, proxy = 0, tentative = 0; 135 int tlladdr; 136 int rflag; 137 union nd_opts ndopts; 138 struct sockaddr_dl proxydl; 139 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 140 141 /* RFC 6980: Nodes MUST silently ignore fragments */ 142 if(m->m_flags & M_FRAGMENTED) 143 goto freeit; 144 145 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0; 146 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif) 147 rflag = 0; 148 #ifndef PULLDOWN_TEST 149 IP6_EXTHDR_CHECK(m, off, icmp6len,); 150 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off); 151 #else 152 IP6_EXTHDR_GET(nd_ns, struct nd_neighbor_solicit *, m, off, icmp6len); 153 if (nd_ns == NULL) { 154 ICMP6STAT_INC(icp6s_tooshort); 155 return; 156 } 157 #endif 158 ip6 = mtod(m, struct ip6_hdr *); /* adjust pointer for safety */ 159 taddr6 = nd_ns->nd_ns_target; 160 if (in6_setscope(&taddr6, ifp, NULL) != 0) 161 goto bad; 162 163 if (ip6->ip6_hlim != 255) { 164 nd6log((LOG_ERR, 165 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n", 166 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 167 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 168 goto bad; 169 } 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 if (ndopts.nd_opts_src_lladdr) { 219 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 220 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 221 } 222 223 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) { 224 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 225 "(link-layer address option)\n")); 226 goto bad; 227 } 228 229 /* 230 * Attaching target link-layer address to the NA? 231 * (RFC 2461 7.2.4) 232 * 233 * NS IP dst is unicast/anycast MUST NOT add 234 * NS IP dst is solicited-node multicast MUST add 235 * 236 * In implementation, we add target link-layer address by default. 237 * We do not add one in MUST NOT cases. 238 */ 239 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) 240 tlladdr = 0; 241 else 242 tlladdr = 1; 243 244 /* 245 * Target address (taddr6) must be either: 246 * (1) Valid unicast/anycast address for my receiving interface, 247 * (2) Unicast address for which I'm offering proxy service, or 248 * (3) "tentative" address on which DAD is being performed. 249 */ 250 /* (1) and (3) check. */ 251 if (ifp->if_carp) 252 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 253 else 254 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 255 256 /* (2) check. */ 257 if (ifa == NULL) { 258 struct sockaddr_dl rt_gateway; 259 struct rt_addrinfo info; 260 struct sockaddr_in6 dst6; 261 262 bzero(&dst6, sizeof(dst6)); 263 dst6.sin6_len = sizeof(struct sockaddr_in6); 264 dst6.sin6_family = AF_INET6; 265 dst6.sin6_addr = taddr6; 266 267 bzero(&rt_gateway, sizeof(rt_gateway)); 268 rt_gateway.sdl_len = sizeof(rt_gateway); 269 bzero(&info, sizeof(info)); 270 info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&rt_gateway; 271 272 if (rib_lookup_info(ifp->if_fib, (struct sockaddr *)&dst6, 273 0, 0, &info) == 0) { 274 if ((info.rti_flags & RTF_ANNOUNCE) != 0 && 275 rt_gateway.sdl_family == AF_LINK) { 276 277 /* 278 * proxy NDP for single entry 279 */ 280 proxydl = *SDL(&rt_gateway); 281 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal( 282 ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST); 283 if (ifa) 284 proxy = 1; 285 } 286 } 287 } 288 if (ifa == NULL) { 289 /* 290 * We've got an NS packet, and we don't have that adddress 291 * assigned for us. We MUST silently ignore it. 292 * See RFC2461 7.2.3. 293 */ 294 goto freeit; 295 } 296 myaddr6 = *IFA_IN6(ifa); 297 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST; 298 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE; 299 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED) 300 goto freeit; 301 302 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 303 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s " 304 "(if %d, NS packet %d)\n", 305 ip6_sprintf(ip6bufs, &taddr6), 306 ifp->if_addrlen, lladdrlen - 2)); 307 goto bad; 308 } 309 310 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) { 311 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n", 312 ip6_sprintf(ip6bufs, &saddr6))); 313 goto freeit; 314 } 315 316 /* 317 * We have neighbor solicitation packet, with target address equals to 318 * one of my tentative address. 319 * 320 * src addr how to process? 321 * --- --- 322 * multicast of course, invalid (rejected in ip6_input) 323 * unicast somebody is doing address resolution -> ignore 324 * unspec dup address detection 325 * 326 * The processing is defined in RFC 2462. 327 */ 328 if (tentative) { 329 /* 330 * If source address is unspecified address, it is for 331 * duplicate address detection. 332 * 333 * If not, the packet is for addess resolution; 334 * silently ignore it. 335 */ 336 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 337 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce); 338 339 goto freeit; 340 } 341 342 /* 343 * If the source address is unspecified address, entries must not 344 * be created or updated. 345 * It looks that sender is performing DAD. Output NA toward 346 * all-node multicast address, to tell the sender that I'm using 347 * the address. 348 * S bit ("solicited") must be zero. 349 */ 350 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 351 struct in6_addr in6_all; 352 353 in6_all = in6addr_linklocal_allnodes; 354 if (in6_setscope(&in6_all, ifp, NULL) != 0) 355 goto bad; 356 nd6_na_output_fib(ifp, &in6_all, &taddr6, 357 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 358 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL, 359 M_GETFIB(m)); 360 goto freeit; 361 } 362 363 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, 364 ND_NEIGHBOR_SOLICIT, 0); 365 366 nd6_na_output_fib(ifp, &saddr6, &taddr6, 367 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 368 rflag | ND_NA_FLAG_SOLICITED, tlladdr, 369 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m)); 370 freeit: 371 if (ifa != NULL) 372 ifa_free(ifa); 373 m_freem(m); 374 return; 375 376 bad: 377 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n", 378 ip6_sprintf(ip6bufs, &saddr6))); 379 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n", 380 ip6_sprintf(ip6bufs, &daddr6))); 381 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n", 382 ip6_sprintf(ip6bufs, &taddr6))); 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 epoch_tracker et; 618 struct ifnet *ifp = m->m_pkthdr.rcvif; 619 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 620 struct nd_neighbor_advert *nd_na; 621 struct in6_addr daddr6 = ip6->ip6_dst; 622 struct in6_addr taddr6; 623 int flags; 624 int is_router; 625 int is_solicited; 626 int is_override; 627 char *lladdr = NULL; 628 int lladdrlen = 0; 629 int checklink = 0; 630 struct ifaddr *ifa; 631 struct llentry *ln = NULL; 632 union nd_opts ndopts; 633 struct mbuf *chain = NULL; 634 struct sockaddr_in6 sin6; 635 u_char linkhdr[LLE_MAX_LINKHDR]; 636 size_t linkhdrsize; 637 int lladdr_off; 638 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 639 640 /* RFC 6980: Nodes MUST silently ignore fragments */ 641 if(m->m_flags & M_FRAGMENTED) 642 goto freeit; 643 644 if (ip6->ip6_hlim != 255) { 645 nd6log((LOG_ERR, 646 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n", 647 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 648 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 649 goto bad; 650 } 651 652 #ifndef PULLDOWN_TEST 653 IP6_EXTHDR_CHECK(m, off, icmp6len,); 654 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off); 655 #else 656 IP6_EXTHDR_GET(nd_na, struct nd_neighbor_advert *, m, off, icmp6len); 657 if (nd_na == NULL) { 658 ICMP6STAT_INC(icp6s_tooshort); 659 return; 660 } 661 #endif 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 memset(&sin6, 0, sizeof(sin6)); 668 669 taddr6 = nd_na->nd_na_target; 670 if (in6_setscope(&taddr6, ifp, NULL)) 671 goto bad; /* XXX: impossible */ 672 673 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 674 nd6log((LOG_ERR, 675 "nd6_na_input: invalid target address %s\n", 676 ip6_sprintf(ip6bufs, &taddr6))); 677 goto bad; 678 } 679 if (IN6_IS_ADDR_MULTICAST(&daddr6)) 680 if (is_solicited) { 681 nd6log((LOG_ERR, 682 "nd6_na_input: a solicited adv is multicasted\n")); 683 goto bad; 684 } 685 686 icmp6len -= sizeof(*nd_na); 687 nd6_option_init(nd_na + 1, icmp6len, &ndopts); 688 if (nd6_options(&ndopts) < 0) { 689 nd6log((LOG_INFO, 690 "nd6_na_input: invalid ND option, ignored\n")); 691 /* nd6_options have incremented stats */ 692 goto freeit; 693 } 694 695 if (ndopts.nd_opts_tgt_lladdr) { 696 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 697 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 698 } 699 700 /* 701 * This effectively disables the DAD check on a non-master CARP 702 * address. 703 */ 704 if (ifp->if_carp) 705 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 706 else 707 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 708 709 /* 710 * Target address matches one of my interface address. 711 * 712 * If my address is tentative, this means that there's somebody 713 * already using the same address as mine. This indicates DAD failure. 714 * This is defined in RFC 2462. 715 * 716 * Otherwise, process as defined in RFC 2461. 717 */ 718 if (ifa 719 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) { 720 nd6_dad_na_input(ifa); 721 ifa_free(ifa); 722 goto freeit; 723 } 724 725 /* Just for safety, maybe unnecessary. */ 726 if (ifa) { 727 ifa_free(ifa); 728 log(LOG_ERR, 729 "nd6_na_input: duplicate IP6 address %s\n", 730 ip6_sprintf(ip6bufs, &taddr6)); 731 goto freeit; 732 } 733 734 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 735 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s " 736 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6), 737 ifp->if_addrlen, lladdrlen - 2)); 738 goto bad; 739 } 740 741 /* 742 * If no neighbor cache entry is found, NA SHOULD silently be 743 * discarded. 744 */ 745 NET_EPOCH_ENTER(et); 746 ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp); 747 NET_EPOCH_EXIT(et); 748 if (ln == NULL) { 749 goto freeit; 750 } 751 752 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 753 /* 754 * If the link-layer has address, and no lladdr option came, 755 * discard the packet. 756 */ 757 if (ifp->if_addrlen && lladdr == NULL) { 758 goto freeit; 759 } 760 761 /* 762 * Record link-layer address, and update the state. 763 */ 764 linkhdrsize = sizeof(linkhdr); 765 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 766 linkhdr, &linkhdrsize, &lladdr_off) != 0) 767 return; 768 769 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 770 lladdr_off) == 0) { 771 ln = NULL; 772 goto freeit; 773 } 774 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 775 if (is_solicited) 776 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 777 else 778 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 779 if ((ln->ln_router = is_router) != 0) { 780 /* 781 * This means a router's state has changed from 782 * non-reachable to probably reachable, and might 783 * affect the status of associated prefixes.. 784 */ 785 checklink = 1; 786 } 787 } else { 788 int llchange; 789 790 /* 791 * Check if the link-layer address has changed or not. 792 */ 793 if (lladdr == NULL) 794 llchange = 0; 795 else { 796 if (ln->la_flags & LLE_VALID) { 797 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen)) 798 llchange = 1; 799 else 800 llchange = 0; 801 } else 802 llchange = 1; 803 } 804 805 /* 806 * This is VERY complex. Look at it with care. 807 * 808 * override solicit lladdr llchange action 809 * (L: record lladdr) 810 * 811 * 0 0 n -- (2c) 812 * 0 0 y n (2b) L 813 * 0 0 y y (1) REACHABLE->STALE 814 * 0 1 n -- (2c) *->REACHABLE 815 * 0 1 y n (2b) L *->REACHABLE 816 * 0 1 y y (1) REACHABLE->STALE 817 * 1 0 n -- (2a) 818 * 1 0 y n (2a) L 819 * 1 0 y y (2a) L *->STALE 820 * 1 1 n -- (2a) *->REACHABLE 821 * 1 1 y n (2a) L *->REACHABLE 822 * 1 1 y y (2a) L *->REACHABLE 823 */ 824 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */ 825 /* 826 * If state is REACHABLE, make it STALE. 827 * no other updates should be done. 828 */ 829 if (ln->ln_state == ND6_LLINFO_REACHABLE) 830 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 831 goto freeit; 832 } else if (is_override /* (2a) */ 833 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */ 834 || lladdr == NULL) { /* (2c) */ 835 /* 836 * Update link-local address, if any. 837 */ 838 if (lladdr != NULL) { 839 linkhdrsize = sizeof(linkhdr); 840 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 841 linkhdr, &linkhdrsize, &lladdr_off) != 0) 842 goto freeit; 843 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, 844 linkhdrsize, lladdr_off) == 0) { 845 ln = NULL; 846 goto freeit; 847 } 848 EVENTHANDLER_INVOKE(lle_event, ln, 849 LLENTRY_RESOLVED); 850 } 851 852 /* 853 * If solicited, make the state REACHABLE. 854 * If not solicited and the link-layer address was 855 * changed, make it STALE. 856 */ 857 if (is_solicited) 858 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 859 else { 860 if (lladdr != NULL && llchange) 861 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 862 } 863 } 864 865 if (ln->ln_router && !is_router) { 866 /* 867 * The peer dropped the router flag. 868 * Remove the sender from the Default Router List and 869 * update the Destination Cache entries. 870 */ 871 struct ifnet *nd6_ifp; 872 873 nd6_ifp = lltable_get_ifp(ln->lle_tbl); 874 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) && 875 (ND_IFINFO(nd6_ifp)->flags & 876 ND6_IFF_ACCEPT_RTADV) != 0) 877 /* 878 * Even if the neighbor is not in the default 879 * router list, the neighbor may be used as a 880 * next hop for some destinations (e.g. redirect 881 * case). So we must call rt6_flush explicitly. 882 */ 883 rt6_flush(&ip6->ip6_src, ifp); 884 } 885 ln->ln_router = is_router; 886 } 887 /* XXX - QL 888 * Does this matter? 889 * rt->rt_flags &= ~RTF_REJECT; 890 */ 891 ln->la_asked = 0; 892 if (ln->la_hold != NULL) 893 nd6_grab_holdchain(ln, &chain, &sin6); 894 freeit: 895 if (ln != NULL) 896 LLE_WUNLOCK(ln); 897 898 if (chain != NULL) 899 nd6_flush_holdchain(ifp, chain, &sin6); 900 901 if (checklink) 902 pfxlist_onlink_check(); 903 904 m_freem(m); 905 return; 906 907 bad: 908 if (ln != NULL) 909 LLE_WUNLOCK(ln); 910 911 ICMP6STAT_INC(icp6s_badna); 912 m_freem(m); 913 } 914 915 /* 916 * Neighbor advertisement output handling. 917 * 918 * Based on RFC 2461 919 * 920 * the following items are not implemented yet: 921 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 922 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 923 * 924 * tlladdr - 1 if include target link-layer address 925 * sdl0 - sockaddr_dl (= proxy NA) or NULL 926 */ 927 static void 928 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0, 929 const struct in6_addr *taddr6, u_long flags, int tlladdr, 930 struct sockaddr *sdl0, u_int fibnum) 931 { 932 struct mbuf *m; 933 struct m_tag *mtag; 934 struct ip6_hdr *ip6; 935 struct nd_neighbor_advert *nd_na; 936 struct ip6_moptions im6o; 937 struct in6_addr daddr6, dst6, src6; 938 uint32_t scopeid; 939 940 int icmp6len, maxlen, error; 941 caddr_t mac = NULL; 942 943 daddr6 = *daddr6_0; /* make a local copy for modification */ 944 945 /* estimate the size of message */ 946 maxlen = sizeof(*ip6) + sizeof(*nd_na); 947 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; 948 KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( 949 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", 950 __func__, max_linkhdr, maxlen, MCLBYTES)); 951 952 if (max_linkhdr + maxlen > MHLEN) 953 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 954 else 955 m = m_gethdr(M_NOWAIT, MT_DATA); 956 if (m == NULL) 957 return; 958 M_SETFIB(m, fibnum); 959 960 if (IN6_IS_ADDR_MULTICAST(&daddr6)) { 961 m->m_flags |= M_MCAST; 962 im6o.im6o_multicast_ifp = ifp; 963 im6o.im6o_multicast_hlim = 255; 964 im6o.im6o_multicast_loop = 0; 965 } 966 967 icmp6len = sizeof(*nd_na); 968 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len; 969 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ 970 971 /* fill neighbor advertisement packet */ 972 ip6 = mtod(m, struct ip6_hdr *); 973 ip6->ip6_flow = 0; 974 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 975 ip6->ip6_vfc |= IPV6_VERSION; 976 ip6->ip6_nxt = IPPROTO_ICMPV6; 977 ip6->ip6_hlim = 255; 978 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) { 979 /* reply to DAD */ 980 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL; 981 daddr6.s6_addr16[1] = 0; 982 daddr6.s6_addr32[1] = 0; 983 daddr6.s6_addr32[2] = 0; 984 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE; 985 if (in6_setscope(&daddr6, ifp, NULL)) 986 goto bad; 987 988 flags &= ~ND_NA_FLAG_SOLICITED; 989 } 990 ip6->ip6_dst = daddr6; 991 992 /* 993 * Select a source whose scope is the same as that of the dest. 994 */ 995 in6_splitscope(&daddr6, &dst6, &scopeid); 996 error = in6_selectsrc_addr(fibnum, &dst6, 997 scopeid, ifp, &src6, NULL); 998 if (error) { 999 char ip6buf[INET6_ADDRSTRLEN]; 1000 nd6log((LOG_DEBUG, "nd6_na_output: source can't be " 1001 "determined: dst=%s, error=%d\n", 1002 ip6_sprintf(ip6buf, &daddr6), error)); 1003 goto bad; 1004 } 1005 ip6->ip6_src = src6; 1006 nd_na = (struct nd_neighbor_advert *)(ip6 + 1); 1007 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT; 1008 nd_na->nd_na_code = 0; 1009 nd_na->nd_na_target = *taddr6; 1010 in6_clearscope(&nd_na->nd_na_target); /* XXX */ 1011 1012 /* 1013 * "tlladdr" indicates NS's condition for adding tlladdr or not. 1014 * see nd6_ns_input() for details. 1015 * Basically, if NS packet is sent to unicast/anycast addr, 1016 * target lladdr option SHOULD NOT be included. 1017 */ 1018 if (tlladdr) { 1019 /* 1020 * sdl0 != NULL indicates proxy NA. If we do proxy, use 1021 * lladdr in sdl0. If we are not proxying (sending NA for 1022 * my address) use lladdr configured for the interface. 1023 */ 1024 if (sdl0 == NULL) { 1025 if (ifp->if_carp) 1026 mac = (*carp_macmatch6_p)(ifp, m, taddr6); 1027 if (mac == NULL) 1028 mac = nd6_ifptomac(ifp); 1029 } else if (sdl0->sa_family == AF_LINK) { 1030 struct sockaddr_dl *sdl; 1031 sdl = (struct sockaddr_dl *)sdl0; 1032 if (sdl->sdl_alen == ifp->if_addrlen) 1033 mac = LLADDR(sdl); 1034 } 1035 } 1036 if (tlladdr && mac) { 1037 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 1038 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1); 1039 1040 /* roundup to 8 bytes alignment! */ 1041 optlen = (optlen + 7) & ~7; 1042 1043 m->m_pkthdr.len += optlen; 1044 m->m_len += optlen; 1045 icmp6len += optlen; 1046 bzero((caddr_t)nd_opt, optlen); 1047 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 1048 nd_opt->nd_opt_len = optlen >> 3; 1049 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); 1050 } else 1051 flags &= ~ND_NA_FLAG_OVERRIDE; 1052 1053 ip6->ip6_plen = htons((u_short)icmp6len); 1054 nd_na->nd_na_flags_reserved = flags; 1055 nd_na->nd_na_cksum = 0; 1056 nd_na->nd_na_cksum = 1057 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len); 1058 1059 if (send_sendso_input_hook != NULL) { 1060 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 1061 sizeof(unsigned short), M_NOWAIT); 1062 if (mtag == NULL) 1063 goto bad; 1064 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type; 1065 m_tag_prepend(m, mtag); 1066 } 1067 1068 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL); 1069 icmp6_ifstat_inc(ifp, ifs6_out_msg); 1070 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert); 1071 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]); 1072 1073 return; 1074 1075 bad: 1076 m_freem(m); 1077 } 1078 1079 #ifndef BURN_BRIDGES 1080 void 1081 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0, 1082 const struct in6_addr *taddr6, u_long flags, int tlladdr, 1083 struct sockaddr *sdl0) 1084 { 1085 1086 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0, 1087 RT_DEFAULT_FIB); 1088 } 1089 #endif 1090 1091 caddr_t 1092 nd6_ifptomac(struct ifnet *ifp) 1093 { 1094 switch (ifp->if_type) { 1095 case IFT_ETHER: 1096 case IFT_IEEE1394: 1097 case IFT_L2VLAN: 1098 case IFT_INFINIBAND: 1099 case IFT_BRIDGE: 1100 return IF_LLADDR(ifp); 1101 default: 1102 return NULL; 1103 } 1104 } 1105 1106 struct dadq { 1107 TAILQ_ENTRY(dadq) dad_list; 1108 struct ifaddr *dad_ifa; 1109 int dad_count; /* max NS to send */ 1110 int dad_ns_tcount; /* # of trials to send NS */ 1111 int dad_ns_ocount; /* NS sent so far */ 1112 int dad_ns_icount; 1113 int dad_na_icount; 1114 int dad_ns_lcount; /* looped back NS */ 1115 int dad_loopbackprobe; /* probing state for loopback detection */ 1116 struct callout dad_timer_ch; 1117 struct vnet *dad_vnet; 1118 u_int dad_refcnt; 1119 #define ND_OPT_NONCE_LEN32 \ 1120 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t)) 1121 uint32_t dad_nonce[ND_OPT_NONCE_LEN32]; 1122 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */ 1123 }; 1124 1125 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq); 1126 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock); 1127 #define V_dadq VNET(dadq) 1128 #define V_dad_rwlock VNET(dad_rwlock) 1129 1130 #define DADQ_RLOCK() rw_rlock(&V_dad_rwlock) 1131 #define DADQ_RUNLOCK() rw_runlock(&V_dad_rwlock) 1132 #define DADQ_WLOCK() rw_wlock(&V_dad_rwlock) 1133 #define DADQ_WUNLOCK() rw_wunlock(&V_dad_rwlock) 1134 1135 static void 1136 nd6_dad_add(struct dadq *dp) 1137 { 1138 1139 DADQ_WLOCK(); 1140 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list); 1141 dp->dad_ondadq = true; 1142 DADQ_WUNLOCK(); 1143 } 1144 1145 static void 1146 nd6_dad_del(struct dadq *dp) 1147 { 1148 1149 DADQ_WLOCK(); 1150 if (dp->dad_ondadq) { 1151 /* 1152 * Remove dp from the dadq and release the dadq's 1153 * reference. 1154 */ 1155 TAILQ_REMOVE(&V_dadq, dp, dad_list); 1156 dp->dad_ondadq = false; 1157 DADQ_WUNLOCK(); 1158 nd6_dad_rele(dp); 1159 } else 1160 DADQ_WUNLOCK(); 1161 } 1162 1163 static struct dadq * 1164 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n) 1165 { 1166 struct dadq *dp; 1167 1168 DADQ_RLOCK(); 1169 TAILQ_FOREACH(dp, &V_dadq, dad_list) { 1170 if (dp->dad_ifa != ifa) 1171 continue; 1172 /* 1173 * Skip if the nonce matches the received one. 1174 * +2 in the length is required because of type and 1175 * length fields are included in a header. 1176 */ 1177 if (n != NULL && 1178 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 && 1179 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0], 1180 ND_OPT_NONCE_LEN) == 0) { 1181 dp->dad_ns_lcount++; 1182 continue; 1183 } 1184 refcount_acquire(&dp->dad_refcnt); 1185 break; 1186 } 1187 DADQ_RUNLOCK(); 1188 1189 return (dp); 1190 } 1191 1192 static void 1193 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns) 1194 { 1195 1196 if (send_ns != 0) 1197 nd6_dad_ns_output(dp); 1198 callout_reset(&dp->dad_timer_ch, ticks, 1199 (void (*)(void *))nd6_dad_timer, (void *)dp); 1200 } 1201 1202 static void 1203 nd6_dad_stoptimer(struct dadq *dp) 1204 { 1205 1206 callout_drain(&dp->dad_timer_ch); 1207 } 1208 1209 static void 1210 nd6_dad_rele(struct dadq *dp) 1211 { 1212 1213 if (refcount_release(&dp->dad_refcnt)) { 1214 ifa_free(dp->dad_ifa); 1215 free(dp, M_IP6NDP); 1216 } 1217 } 1218 1219 void 1220 nd6_dad_init(void) 1221 { 1222 1223 rw_init(&V_dad_rwlock, "nd6 DAD queue"); 1224 TAILQ_INIT(&V_dadq); 1225 } 1226 1227 /* 1228 * Start Duplicate Address Detection (DAD) for specified interface address. 1229 */ 1230 void 1231 nd6_dad_start(struct ifaddr *ifa, int delay) 1232 { 1233 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1234 struct dadq *dp; 1235 char ip6buf[INET6_ADDRSTRLEN]; 1236 1237 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0, 1238 ("starting DAD on non-tentative address %p", ifa)); 1239 1240 /* 1241 * If we don't need DAD, don't do it. 1242 * There are several cases: 1243 * - DAD is disabled globally or on the interface 1244 * - the interface address is anycast 1245 */ 1246 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 || 1247 V_ip6_dad_count == 0 || 1248 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) { 1249 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1250 return; 1251 } 1252 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 || 1253 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || 1254 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0) 1255 return; 1256 1257 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) { 1258 /* 1259 * DAD is already in progress. Let the existing entry 1260 * finish it. 1261 */ 1262 nd6_dad_rele(dp); 1263 return; 1264 } 1265 1266 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO); 1267 if (dp == NULL) { 1268 log(LOG_ERR, "nd6_dad_start: memory allocation failed for " 1269 "%s(%s)\n", 1270 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1271 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1272 return; 1273 } 1274 callout_init(&dp->dad_timer_ch, 0); 1275 #ifdef VIMAGE 1276 dp->dad_vnet = curvnet; 1277 #endif 1278 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp), 1279 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1280 1281 /* 1282 * Send NS packet for DAD, ip6_dad_count times. 1283 * Note that we must delay the first transmission, if this is the 1284 * first packet to be sent from the interface after interface 1285 * (re)initialization. 1286 */ 1287 dp->dad_ifa = ifa; 1288 ifa_ref(dp->dad_ifa); 1289 dp->dad_count = V_ip6_dad_count; 1290 dp->dad_ns_icount = dp->dad_na_icount = 0; 1291 dp->dad_ns_ocount = dp->dad_ns_tcount = 0; 1292 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0; 1293 1294 /* Add this to the dadq and add a reference for the dadq. */ 1295 refcount_init(&dp->dad_refcnt, 1); 1296 nd6_dad_add(dp); 1297 nd6_dad_starttimer(dp, delay, 0); 1298 } 1299 1300 /* 1301 * terminate DAD unconditionally. used for address removals. 1302 */ 1303 void 1304 nd6_dad_stop(struct ifaddr *ifa) 1305 { 1306 struct dadq *dp; 1307 1308 dp = nd6_dad_find(ifa, NULL); 1309 if (!dp) { 1310 /* DAD wasn't started yet */ 1311 return; 1312 } 1313 1314 nd6_dad_stoptimer(dp); 1315 nd6_dad_del(dp); 1316 1317 /* Release this function's reference, acquired by nd6_dad_find(). */ 1318 nd6_dad_rele(dp); 1319 } 1320 1321 static void 1322 nd6_dad_timer(struct dadq *dp) 1323 { 1324 CURVNET_SET(dp->dad_vnet); 1325 struct ifaddr *ifa = dp->dad_ifa; 1326 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1327 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1328 char ip6buf[INET6_ADDRSTRLEN]; 1329 1330 KASSERT(ia != NULL, ("DAD entry %p with no address", dp)); 1331 1332 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) { 1333 /* Do not need DAD for ifdisabled interface. */ 1334 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of " 1335 "ND6_IFF_IFDISABLED.\n", ifp->if_xname); 1336 goto err; 1337 } 1338 if (ia->ia6_flags & IN6_IFF_DUPLICATED) { 1339 log(LOG_ERR, "nd6_dad_timer: called with duplicated address " 1340 "%s(%s)\n", 1341 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1342 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1343 goto err; 1344 } 1345 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) { 1346 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address " 1347 "%s(%s)\n", 1348 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1349 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1350 goto err; 1351 } 1352 1353 /* Stop DAD if the interface is down even after dad_maxtry attempts. */ 1354 if ((dp->dad_ns_tcount > V_dad_maxtry) && 1355 (((ifp->if_flags & IFF_UP) == 0) || 1356 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) { 1357 nd6log((LOG_INFO, "%s: could not run DAD " 1358 "because the interface was down or not running.\n", 1359 if_name(ifa->ifa_ifp))); 1360 goto err; 1361 } 1362 1363 /* Need more checks? */ 1364 if (dp->dad_ns_ocount < dp->dad_count) { 1365 /* 1366 * We have more NS to go. Send NS packet for DAD. 1367 */ 1368 nd6_dad_starttimer(dp, 1369 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1); 1370 goto done; 1371 } else { 1372 /* 1373 * We have transmitted sufficient number of DAD packets. 1374 * See what we've got. 1375 */ 1376 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) 1377 /* We've seen NS or NA, means DAD has failed. */ 1378 nd6_dad_duplicated(ifa, dp); 1379 else if (V_dad_enhanced != 0 && 1380 dp->dad_ns_lcount > 0 && 1381 dp->dad_ns_lcount > dp->dad_loopbackprobe) { 1382 /* 1383 * Sec. 4.1 in RFC 7527 requires transmission of 1384 * additional probes until the loopback condition 1385 * becomes clear when a looped back probe is detected. 1386 */ 1387 log(LOG_ERR, "%s: a looped back NS message is " 1388 "detected during DAD for %s. " 1389 "Another DAD probes are being sent.\n", 1390 if_name(ifa->ifa_ifp), 1391 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1392 dp->dad_loopbackprobe = dp->dad_ns_lcount; 1393 /* 1394 * Send an NS immediately and increase dad_count by 1395 * V_nd6_mmaxtries - 1. 1396 */ 1397 dp->dad_count = 1398 dp->dad_ns_ocount + V_nd6_mmaxtries - 1; 1399 nd6_dad_starttimer(dp, 1400 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1401 1); 1402 goto done; 1403 } else { 1404 /* 1405 * We are done with DAD. No NA came, no NS came. 1406 * No duplicate address found. Check IFDISABLED flag 1407 * again in case that it is changed between the 1408 * beginning of this function and here. 1409 */ 1410 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0) 1411 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1412 1413 nd6log((LOG_DEBUG, 1414 "%s: DAD complete for %s - no duplicates found\n", 1415 if_name(ifa->ifa_ifp), 1416 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1417 if (dp->dad_ns_lcount > 0) 1418 log(LOG_ERR, "%s: DAD completed while " 1419 "a looped back NS message is detected " 1420 "during DAD for %s.\n", 1421 if_name(ifa->ifa_ifp), 1422 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1423 } 1424 } 1425 err: 1426 nd6_dad_del(dp); 1427 done: 1428 CURVNET_RESTORE(); 1429 } 1430 1431 static void 1432 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp) 1433 { 1434 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1435 struct ifnet *ifp; 1436 char ip6buf[INET6_ADDRSTRLEN]; 1437 1438 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: " 1439 "NS in/out/loopback=%d/%d/%d, NA in=%d\n", 1440 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1441 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount, 1442 dp->dad_na_icount); 1443 1444 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1445 ia->ia6_flags |= IN6_IFF_DUPLICATED; 1446 1447 ifp = ifa->ifa_ifp; 1448 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n", 1449 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)); 1450 log(LOG_ERR, "%s: manual intervention required\n", 1451 if_name(ifp)); 1452 1453 /* 1454 * If the address is a link-local address formed from an interface 1455 * identifier based on the hardware address which is supposed to be 1456 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP 1457 * operation on the interface SHOULD be disabled. 1458 * [RFC 4862, Section 5.4.5] 1459 */ 1460 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1461 struct in6_addr in6; 1462 1463 /* 1464 * To avoid over-reaction, we only apply this logic when we are 1465 * very sure that hardware addresses are supposed to be unique. 1466 */ 1467 switch (ifp->if_type) { 1468 case IFT_ETHER: 1469 case IFT_ATM: 1470 case IFT_IEEE1394: 1471 case IFT_INFINIBAND: 1472 in6 = ia->ia_addr.sin6_addr; 1473 if (in6_get_hw_ifid(ifp, &in6) == 0 && 1474 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) { 1475 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1476 log(LOG_ERR, "%s: possible hardware address " 1477 "duplication detected, disable IPv6\n", 1478 if_name(ifp)); 1479 } 1480 break; 1481 } 1482 } 1483 } 1484 1485 static void 1486 nd6_dad_ns_output(struct dadq *dp) 1487 { 1488 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa; 1489 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1490 int i; 1491 1492 dp->dad_ns_tcount++; 1493 if ((ifp->if_flags & IFF_UP) == 0) { 1494 return; 1495 } 1496 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 1497 return; 1498 } 1499 1500 dp->dad_ns_ocount++; 1501 if (V_dad_enhanced != 0) { 1502 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) 1503 dp->dad_nonce[i] = arc4random(); 1504 /* 1505 * XXXHRS: Note that in the case that 1506 * DupAddrDetectTransmits > 1, multiple NS messages with 1507 * different nonces can be looped back in an unexpected 1508 * order. The current implementation recognizes only 1509 * the latest nonce on the sender side. Practically it 1510 * should work well in almost all cases. 1511 */ 1512 } 1513 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr, 1514 (uint8_t *)&dp->dad_nonce[0]); 1515 } 1516 1517 static void 1518 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce) 1519 { 1520 struct dadq *dp; 1521 1522 if (ifa == NULL) 1523 panic("ifa == NULL in nd6_dad_ns_input"); 1524 1525 /* Ignore Nonce option when Enhanced DAD is disabled. */ 1526 if (V_dad_enhanced == 0) 1527 ndopt_nonce = NULL; 1528 dp = nd6_dad_find(ifa, ndopt_nonce); 1529 if (dp == NULL) 1530 return; 1531 1532 dp->dad_ns_icount++; 1533 nd6_dad_rele(dp); 1534 } 1535 1536 static void 1537 nd6_dad_na_input(struct ifaddr *ifa) 1538 { 1539 struct dadq *dp; 1540 1541 if (ifa == NULL) 1542 panic("ifa == NULL in nd6_dad_na_input"); 1543 1544 dp = nd6_dad_find(ifa, NULL); 1545 if (dp != NULL) { 1546 dp->dad_na_icount++; 1547 nd6_dad_rele(dp); 1548 } 1549 } 1550