1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * $KAME: nd6_nbr.c,v 1.86 2002/01/21 02:33:04 jinmei Exp $ 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_inet.h" 38 #include "opt_inet6.h" 39 #include "opt_ipsec.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/eventhandler.h> 44 #include <sys/malloc.h> 45 #include <sys/libkern.h> 46 #include <sys/lock.h> 47 #include <sys/rwlock.h> 48 #include <sys/mbuf.h> 49 #include <sys/socket.h> 50 #include <sys/sockio.h> 51 #include <sys/time.h> 52 #include <sys/kernel.h> 53 #include <sys/errno.h> 54 #include <sys/sysctl.h> 55 #include <sys/syslog.h> 56 #include <sys/queue.h> 57 #include <sys/callout.h> 58 #include <sys/refcount.h> 59 60 #include <net/if.h> 61 #include <net/if_types.h> 62 #include <net/if_dl.h> 63 #include <net/if_var.h> 64 #include <net/route.h> 65 #include <net/vnet.h> 66 67 #include <netinet/in.h> 68 #include <netinet/in_var.h> 69 #include <net/if_llatbl.h> 70 #include <netinet6/in6_var.h> 71 #include <netinet6/in6_ifattach.h> 72 #include <netinet/ip6.h> 73 #include <netinet6/ip6_var.h> 74 #include <netinet6/scope6_var.h> 75 #include <netinet6/nd6.h> 76 #include <netinet/icmp6.h> 77 #include <netinet/ip_carp.h> 78 #include <netinet6/send.h> 79 80 #define SDL(s) ((struct sockaddr_dl *)s) 81 82 struct dadq; 83 static struct dadq *nd6_dad_find(struct ifaddr *, struct nd_opt_nonce *); 84 static void nd6_dad_add(struct dadq *dp); 85 static void nd6_dad_del(struct dadq *dp); 86 static void nd6_dad_rele(struct dadq *); 87 static void nd6_dad_starttimer(struct dadq *, int, int); 88 static void nd6_dad_stoptimer(struct dadq *); 89 static void nd6_dad_timer(struct dadq *); 90 static void nd6_dad_duplicated(struct ifaddr *, struct dadq *); 91 static void nd6_dad_ns_output(struct dadq *); 92 static void nd6_dad_ns_input(struct ifaddr *, struct nd_opt_nonce *); 93 static void nd6_dad_na_input(struct ifaddr *); 94 static void nd6_na_output_fib(struct ifnet *, const struct in6_addr *, 95 const struct in6_addr *, u_long, int, struct sockaddr *, u_int); 96 static void nd6_ns_output_fib(struct ifnet *, const struct in6_addr *, 97 const struct in6_addr *, const struct in6_addr *, uint8_t *, u_int); 98 99 VNET_DEFINE_STATIC(int, dad_enhanced) = 1; 100 #define V_dad_enhanced VNET(dad_enhanced) 101 102 SYSCTL_DECL(_net_inet6_ip6); 103 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW, 104 &VNET_NAME(dad_enhanced), 0, 105 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD."); 106 107 VNET_DEFINE_STATIC(int, dad_maxtry) = 15; /* max # of *tries* to 108 transmit DAD packet */ 109 #define V_dad_maxtry VNET(dad_maxtry) 110 111 /* 112 * Input a Neighbor Solicitation Message. 113 * 114 * Based on RFC 2461 115 * Based on RFC 2462 (duplicate address detection) 116 */ 117 void 118 nd6_ns_input(struct mbuf *m, int off, int icmp6len) 119 { 120 struct ifnet *ifp; 121 struct ip6_hdr *ip6; 122 struct nd_neighbor_solicit *nd_ns; 123 struct in6_addr daddr6, myaddr6, saddr6, taddr6; 124 struct ifaddr *ifa; 125 struct sockaddr_dl proxydl; 126 union nd_opts ndopts; 127 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 128 char *lladdr; 129 int anycast, lladdrlen, proxy, rflag, tentative, tlladdr; 130 131 ifa = NULL; 132 133 /* RFC 6980: Nodes MUST silently ignore fragments */ 134 if(m->m_flags & M_FRAGMENTED) 135 goto freeit; 136 137 ifp = m->m_pkthdr.rcvif; 138 ip6 = mtod(m, struct ip6_hdr *); 139 if (__predict_false(ip6->ip6_hlim != 255)) { 140 ICMP6STAT_INC(icp6s_invlhlim); 141 nd6log((LOG_ERR, 142 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n", 143 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 144 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 145 goto bads; 146 } 147 148 if (m->m_len < off + icmp6len) { 149 m = m_pullup(m, off + icmp6len); 150 if (m == NULL) { 151 IP6STAT_INC(ip6s_exthdrtoolong); 152 return; 153 } 154 } 155 ip6 = mtod(m, struct ip6_hdr *); 156 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off); 157 158 saddr6 = ip6->ip6_src; 159 daddr6 = ip6->ip6_dst; 160 taddr6 = nd_ns->nd_ns_target; 161 if (in6_setscope(&taddr6, ifp, NULL) != 0) 162 goto bad; 163 164 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0; 165 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif) 166 rflag = 0; 167 168 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 169 /* dst has to be a solicited node multicast address. */ 170 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL && 171 /* don't check ifindex portion */ 172 daddr6.s6_addr32[1] == 0 && 173 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE && 174 daddr6.s6_addr8[12] == 0xff) { 175 ; /* good */ 176 } else { 177 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 178 "(wrong ip6 dst)\n")); 179 goto bad; 180 } 181 } else if (!V_nd6_onlink_ns_rfc4861) { 182 struct sockaddr_in6 src_sa6; 183 184 /* 185 * According to recent IETF discussions, it is not a good idea 186 * to accept a NS from an address which would not be deemed 187 * to be a neighbor otherwise. This point is expected to be 188 * clarified in future revisions of the specification. 189 */ 190 bzero(&src_sa6, sizeof(src_sa6)); 191 src_sa6.sin6_family = AF_INET6; 192 src_sa6.sin6_len = sizeof(src_sa6); 193 src_sa6.sin6_addr = saddr6; 194 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) { 195 nd6log((LOG_INFO, "nd6_ns_input: " 196 "NS packet from non-neighbor\n")); 197 goto bad; 198 } 199 } 200 201 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 202 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n")); 203 goto bad; 204 } 205 206 icmp6len -= sizeof(*nd_ns); 207 nd6_option_init(nd_ns + 1, icmp6len, &ndopts); 208 if (nd6_options(&ndopts) < 0) { 209 nd6log((LOG_INFO, 210 "nd6_ns_input: invalid ND option, ignored\n")); 211 /* nd6_options have incremented stats */ 212 goto freeit; 213 } 214 215 lladdr = NULL; 216 lladdrlen = 0; 217 if (ndopts.nd_opts_src_lladdr) { 218 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 219 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 220 } 221 222 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) { 223 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 224 "(link-layer address option)\n")); 225 goto bad; 226 } 227 228 /* 229 * Attaching target link-layer address to the NA? 230 * (RFC 2461 7.2.4) 231 * 232 * NS IP dst is unicast/anycast MUST NOT add 233 * NS IP dst is solicited-node multicast MUST add 234 * 235 * In implementation, we add target link-layer address by default. 236 * We do not add one in MUST NOT cases. 237 */ 238 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) 239 tlladdr = 0; 240 else 241 tlladdr = 1; 242 243 /* 244 * Target address (taddr6) must be either: 245 * (1) Valid unicast/anycast address for my receiving interface, 246 * (2) Unicast address for which I'm offering proxy service, or 247 * (3) "tentative" address on which DAD is being performed. 248 */ 249 /* (1) and (3) check. */ 250 if (ifp->if_carp) 251 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 252 else 253 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 254 255 /* (2) check. */ 256 proxy = 0; 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 * proxy NDP for single entry 278 */ 279 proxydl = *SDL(&rt_gateway); 280 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal( 281 ifp, IN6_IFF_NOTREADY|IN6_IFF_ANYCAST); 282 if (ifa) 283 proxy = 1; 284 } 285 } 286 } 287 if (ifa == NULL) { 288 /* 289 * We've got an NS packet, and we don't have that adddress 290 * assigned for us. We MUST silently ignore it. 291 * See RFC2461 7.2.3. 292 */ 293 goto freeit; 294 } 295 myaddr6 = *IFA_IN6(ifa); 296 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST; 297 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE; 298 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED) 299 goto freeit; 300 301 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 302 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s " 303 "(if %d, NS packet %d)\n", 304 ip6_sprintf(ip6bufs, &taddr6), 305 ifp->if_addrlen, lladdrlen - 2)); 306 goto bad; 307 } 308 309 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) { 310 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n", 311 ip6_sprintf(ip6bufs, &saddr6))); 312 goto freeit; 313 } 314 315 /* 316 * We have neighbor solicitation packet, with target address equals to 317 * one of my tentative address. 318 * 319 * src addr how to process? 320 * --- --- 321 * multicast of course, invalid (rejected in ip6_input) 322 * unicast somebody is doing address resolution -> ignore 323 * unspec dup address detection 324 * 325 * The processing is defined in RFC 2462. 326 */ 327 if (tentative) { 328 /* 329 * If source address is unspecified address, it is for 330 * duplicate address detection. 331 * 332 * If not, the packet is for addess resolution; 333 * silently ignore it. 334 */ 335 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 336 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce); 337 338 goto freeit; 339 } 340 341 /* 342 * If the source address is unspecified address, entries must not 343 * be created or updated. 344 * It looks that sender is performing DAD. Output NA toward 345 * all-node multicast address, to tell the sender that I'm using 346 * the address. 347 * S bit ("solicited") must be zero. 348 */ 349 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 350 struct in6_addr in6_all; 351 352 in6_all = in6addr_linklocal_allnodes; 353 if (in6_setscope(&in6_all, ifp, NULL) != 0) 354 goto bad; 355 nd6_na_output_fib(ifp, &in6_all, &taddr6, 356 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 357 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL, 358 M_GETFIB(m)); 359 goto freeit; 360 } 361 362 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, 363 ND_NEIGHBOR_SOLICIT, 0); 364 365 nd6_na_output_fib(ifp, &saddr6, &taddr6, 366 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 367 rflag | ND_NA_FLAG_SOLICITED, tlladdr, 368 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m)); 369 freeit: 370 if (ifa != NULL) 371 ifa_free(ifa); 372 m_freem(m); 373 return; 374 375 bad: 376 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n", 377 ip6_sprintf(ip6bufs, &saddr6))); 378 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n", 379 ip6_sprintf(ip6bufs, &daddr6))); 380 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n", 381 ip6_sprintf(ip6bufs, &taddr6))); 382 bads: 383 ICMP6STAT_INC(icp6s_badns); 384 if (ifa != NULL) 385 ifa_free(ifa); 386 m_freem(m); 387 } 388 389 /* 390 * Output a Neighbor Solicitation Message. Caller specifies: 391 * - ICMP6 header source IP6 address 392 * - ND6 header target IP6 address 393 * - ND6 header source datalink address 394 * 395 * Based on RFC 2461 396 * Based on RFC 2462 (duplicate address detection) 397 * 398 * ln - for source address determination 399 * nonce - If non-NULL, NS is used for duplicate address detection and 400 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce. 401 */ 402 static void 403 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6, 404 const struct in6_addr *daddr6, const struct in6_addr *taddr6, 405 uint8_t *nonce, u_int fibnum) 406 { 407 struct mbuf *m; 408 struct m_tag *mtag; 409 struct ip6_hdr *ip6; 410 struct nd_neighbor_solicit *nd_ns; 411 struct ip6_moptions im6o; 412 int icmp6len; 413 int maxlen; 414 caddr_t mac; 415 416 NET_EPOCH_ASSERT(); 417 418 if (IN6_IS_ADDR_MULTICAST(taddr6)) 419 return; 420 421 /* estimate the size of message */ 422 maxlen = sizeof(*ip6) + sizeof(*nd_ns); 423 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; 424 KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( 425 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", 426 __func__, max_linkhdr, maxlen, MCLBYTES)); 427 428 if (max_linkhdr + maxlen > MHLEN) 429 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 430 else 431 m = m_gethdr(M_NOWAIT, MT_DATA); 432 if (m == NULL) 433 return; 434 M_SETFIB(m, fibnum); 435 436 if (daddr6 == NULL || IN6_IS_ADDR_MULTICAST(daddr6)) { 437 m->m_flags |= M_MCAST; 438 im6o.im6o_multicast_ifp = ifp; 439 im6o.im6o_multicast_hlim = 255; 440 im6o.im6o_multicast_loop = 0; 441 } 442 443 icmp6len = sizeof(*nd_ns); 444 m->m_pkthdr.len = m->m_len = sizeof(*ip6) + icmp6len; 445 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ 446 447 /* fill neighbor solicitation packet */ 448 ip6 = mtod(m, struct ip6_hdr *); 449 ip6->ip6_flow = 0; 450 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 451 ip6->ip6_vfc |= IPV6_VERSION; 452 /* ip6->ip6_plen will be set later */ 453 ip6->ip6_nxt = IPPROTO_ICMPV6; 454 ip6->ip6_hlim = 255; 455 if (daddr6) 456 ip6->ip6_dst = *daddr6; 457 else { 458 ip6->ip6_dst.s6_addr16[0] = IPV6_ADDR_INT16_MLL; 459 ip6->ip6_dst.s6_addr16[1] = 0; 460 ip6->ip6_dst.s6_addr32[1] = 0; 461 ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_ONE; 462 ip6->ip6_dst.s6_addr32[3] = taddr6->s6_addr32[3]; 463 ip6->ip6_dst.s6_addr8[12] = 0xff; 464 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 465 goto bad; 466 } 467 if (nonce == NULL) { 468 struct ifaddr *ifa = NULL; 469 470 /* 471 * RFC2461 7.2.2: 472 * "If the source address of the packet prompting the 473 * solicitation is the same as one of the addresses assigned 474 * to the outgoing interface, that address SHOULD be placed 475 * in the IP Source Address of the outgoing solicitation. 476 * Otherwise, any one of the addresses assigned to the 477 * interface should be used." 478 * 479 * We use the source address for the prompting packet 480 * (saddr6), if saddr6 belongs to the outgoing interface. 481 * Otherwise, we perform the source address selection as usual. 482 */ 483 484 if (saddr6 != NULL) 485 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, saddr6); 486 if (ifa != NULL) { 487 /* ip6_src set already. */ 488 ip6->ip6_src = *saddr6; 489 ifa_free(ifa); 490 } else { 491 int error; 492 struct in6_addr dst6, src6; 493 uint32_t scopeid; 494 495 in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid); 496 error = in6_selectsrc_addr(fibnum, &dst6, 497 scopeid, ifp, &src6, NULL); 498 if (error) { 499 char ip6buf[INET6_ADDRSTRLEN]; 500 nd6log((LOG_DEBUG, "%s: source can't be " 501 "determined: dst=%s, error=%d\n", __func__, 502 ip6_sprintf(ip6buf, &dst6), 503 error)); 504 goto bad; 505 } 506 ip6->ip6_src = src6; 507 } 508 } else { 509 /* 510 * Source address for DAD packet must always be IPv6 511 * unspecified address. (0::0) 512 * We actually don't have to 0-clear the address (we did it 513 * above), but we do so here explicitly to make the intention 514 * clearer. 515 */ 516 bzero(&ip6->ip6_src, sizeof(ip6->ip6_src)); 517 } 518 nd_ns = (struct nd_neighbor_solicit *)(ip6 + 1); 519 nd_ns->nd_ns_type = ND_NEIGHBOR_SOLICIT; 520 nd_ns->nd_ns_code = 0; 521 nd_ns->nd_ns_reserved = 0; 522 nd_ns->nd_ns_target = *taddr6; 523 in6_clearscope(&nd_ns->nd_ns_target); /* XXX */ 524 525 /* 526 * Add source link-layer address option. 527 * 528 * spec implementation 529 * --- --- 530 * DAD packet MUST NOT do not add the option 531 * there's no link layer address: 532 * impossible do not add the option 533 * there's link layer address: 534 * Multicast NS MUST add one add the option 535 * Unicast NS SHOULD add one add the option 536 */ 537 if (nonce == NULL && (mac = nd6_ifptomac(ifp))) { 538 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 539 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); 540 /* 8 byte alignments... */ 541 optlen = (optlen + 7) & ~7; 542 543 m->m_pkthdr.len += optlen; 544 m->m_len += optlen; 545 icmp6len += optlen; 546 bzero((caddr_t)nd_opt, optlen); 547 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; 548 nd_opt->nd_opt_len = optlen >> 3; 549 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); 550 } 551 /* 552 * Add a Nonce option (RFC 3971) to detect looped back NS messages. 553 * This behavior is documented as Enhanced Duplicate Address 554 * Detection in RFC 7527. 555 * net.inet6.ip6.dad_enhanced=0 disables this. 556 */ 557 if (V_dad_enhanced != 0 && nonce != NULL) { 558 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN; 559 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); 560 /* 8-byte alignment is required. */ 561 optlen = (optlen + 7) & ~7; 562 563 m->m_pkthdr.len += optlen; 564 m->m_len += optlen; 565 icmp6len += optlen; 566 bzero((caddr_t)nd_opt, optlen); 567 nd_opt->nd_opt_type = ND_OPT_NONCE; 568 nd_opt->nd_opt_len = optlen >> 3; 569 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN); 570 } 571 ip6->ip6_plen = htons((u_short)icmp6len); 572 nd_ns->nd_ns_cksum = 0; 573 nd_ns->nd_ns_cksum = 574 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len); 575 576 if (send_sendso_input_hook != NULL) { 577 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 578 sizeof(unsigned short), M_NOWAIT); 579 if (mtag == NULL) 580 goto bad; 581 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type; 582 m_tag_prepend(m, mtag); 583 } 584 585 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0, 586 &im6o, NULL, NULL); 587 icmp6_ifstat_inc(ifp, ifs6_out_msg); 588 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit); 589 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]); 590 591 return; 592 593 bad: 594 m_freem(m); 595 } 596 597 #ifndef BURN_BRIDGES 598 void 599 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6, 600 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce) 601 { 602 603 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB); 604 } 605 #endif 606 /* 607 * Neighbor advertisement input handling. 608 * 609 * Based on RFC 2461 610 * Based on RFC 2462 (duplicate address detection) 611 * 612 * the following items are not implemented yet: 613 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 614 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 615 */ 616 void 617 nd6_na_input(struct mbuf *m, int off, int icmp6len) 618 { 619 struct ifnet *ifp; 620 struct ip6_hdr *ip6; 621 struct ifaddr *ifa; 622 struct llentry *ln; 623 struct mbuf *chain; 624 struct nd_neighbor_advert *nd_na; 625 struct in6_addr daddr6, taddr6; 626 union nd_opts ndopts; 627 u_char linkhdr[LLE_MAX_LINKHDR]; 628 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 629 char *lladdr; 630 size_t linkhdrsize; 631 int flags, is_override, is_router, is_solicited; 632 int lladdr_off, lladdrlen, checklink; 633 634 NET_EPOCH_ASSERT(); 635 636 chain = NULL; 637 ln = NULL; 638 checklink = 0; 639 640 /* RFC 6980: Nodes MUST silently ignore fragments */ 641 if(m->m_flags & M_FRAGMENTED) 642 goto freeit; 643 644 ifp = m->m_pkthdr.rcvif; 645 ip6 = mtod(m, struct ip6_hdr *); 646 if (__predict_false(ip6->ip6_hlim != 255)) { 647 ICMP6STAT_INC(icp6s_invlhlim); 648 nd6log((LOG_ERR, 649 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n", 650 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 651 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 652 goto bad; 653 } 654 655 if (m->m_len < off + icmp6len) { 656 m = m_pullup(m, off + icmp6len); 657 if (m == NULL) { 658 IP6STAT_INC(ip6s_exthdrtoolong); 659 return; 660 } 661 } 662 ip6 = mtod(m, struct ip6_hdr *); 663 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off); 664 665 flags = nd_na->nd_na_flags_reserved; 666 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0); 667 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0); 668 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0); 669 670 taddr6 = nd_na->nd_na_target; 671 if (in6_setscope(&taddr6, ifp, NULL)) 672 goto bad; /* XXX: impossible */ 673 674 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 675 nd6log((LOG_ERR, 676 "nd6_na_input: invalid target address %s\n", 677 ip6_sprintf(ip6bufs, &taddr6))); 678 goto bad; 679 } 680 681 daddr6 = ip6->ip6_dst; 682 if (IN6_IS_ADDR_MULTICAST(&daddr6)) 683 if (is_solicited) { 684 nd6log((LOG_ERR, 685 "nd6_na_input: a solicited adv is multicasted\n")); 686 goto bad; 687 } 688 689 icmp6len -= sizeof(*nd_na); 690 nd6_option_init(nd_na + 1, icmp6len, &ndopts); 691 if (nd6_options(&ndopts) < 0) { 692 nd6log((LOG_INFO, 693 "nd6_na_input: invalid ND option, ignored\n")); 694 /* nd6_options have incremented stats */ 695 goto freeit; 696 } 697 698 lladdr = NULL; 699 lladdrlen = 0; 700 if (ndopts.nd_opts_tgt_lladdr) { 701 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 702 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 703 } 704 705 /* 706 * This effectively disables the DAD check on a non-master CARP 707 * address. 708 */ 709 if (ifp->if_carp) 710 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 711 else 712 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 713 714 /* 715 * Target address matches one of my interface address. 716 * 717 * If my address is tentative, this means that there's somebody 718 * already using the same address as mine. This indicates DAD failure. 719 * This is defined in RFC 2462. 720 * 721 * Otherwise, process as defined in RFC 2461. 722 */ 723 if (ifa 724 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) { 725 nd6_dad_na_input(ifa); 726 ifa_free(ifa); 727 goto freeit; 728 } 729 730 /* Just for safety, maybe unnecessary. */ 731 if (ifa) { 732 ifa_free(ifa); 733 log(LOG_ERR, 734 "nd6_na_input: duplicate IP6 address %s\n", 735 ip6_sprintf(ip6bufs, &taddr6)); 736 goto freeit; 737 } 738 739 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 740 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s " 741 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6), 742 ifp->if_addrlen, lladdrlen - 2)); 743 goto bad; 744 } 745 746 /* 747 * If no neighbor cache entry is found, NA SHOULD silently be 748 * discarded. 749 */ 750 ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp); 751 if (ln == NULL) { 752 goto freeit; 753 } 754 755 /* 756 * Do not try to override static entry. 757 */ 758 if (ln->la_flags & LLE_STATIC) 759 goto freeit; 760 761 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 762 /* 763 * If the link-layer has address, and no lladdr option came, 764 * discard the packet. 765 */ 766 if (ifp->if_addrlen && lladdr == NULL) { 767 goto freeit; 768 } 769 770 /* 771 * Record link-layer address, and update the state. 772 */ 773 if (!nd6_try_set_entry_addr(ifp, ln, lladdr)) 774 goto freeit; 775 776 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 777 if (is_solicited) 778 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 779 else 780 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 781 if ((ln->ln_router = is_router) != 0) { 782 /* 783 * This means a router's state has changed from 784 * non-reachable to probably reachable, and might 785 * affect the status of associated prefixes.. 786 */ 787 checklink = 1; 788 } 789 } else { 790 int llchange; 791 792 /* 793 * Check if the link-layer address has changed or not. 794 */ 795 if (lladdr == NULL) 796 llchange = 0; 797 else { 798 if (ln->la_flags & LLE_VALID) { 799 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen)) 800 llchange = 1; 801 else 802 llchange = 0; 803 } else 804 llchange = 1; 805 } 806 807 /* 808 * This is VERY complex. Look at it with care. 809 * 810 * override solicit lladdr llchange action 811 * (L: record lladdr) 812 * 813 * 0 0 n -- (2c) 814 * 0 0 y n (2b) L 815 * 0 0 y y (1) REACHABLE->STALE 816 * 0 1 n -- (2c) *->REACHABLE 817 * 0 1 y n (2b) L *->REACHABLE 818 * 0 1 y y (1) REACHABLE->STALE 819 * 1 0 n -- (2a) 820 * 1 0 y n (2a) L 821 * 1 0 y y (2a) L *->STALE 822 * 1 1 n -- (2a) *->REACHABLE 823 * 1 1 y n (2a) L *->REACHABLE 824 * 1 1 y y (2a) L *->REACHABLE 825 */ 826 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */ 827 /* 828 * If state is REACHABLE, make it STALE. 829 * no other updates should be done. 830 */ 831 if (ln->ln_state == ND6_LLINFO_REACHABLE) 832 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 833 goto freeit; 834 } else if (is_override /* (2a) */ 835 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */ 836 || lladdr == NULL) { /* (2c) */ 837 /* 838 * Update link-local address, if any. 839 */ 840 if (lladdr != NULL) { 841 linkhdrsize = sizeof(linkhdr); 842 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 843 linkhdr, &linkhdrsize, &lladdr_off) != 0) 844 goto freeit; 845 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, 846 linkhdrsize, lladdr_off) == 0) { 847 ln = NULL; 848 goto freeit; 849 } 850 EVENTHANDLER_INVOKE(lle_event, ln, 851 LLENTRY_RESOLVED); 852 } 853 854 /* 855 * If solicited, make the state REACHABLE. 856 * If not solicited and the link-layer address was 857 * changed, make it STALE. 858 */ 859 if (is_solicited) 860 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 861 else { 862 if (lladdr != NULL && llchange) 863 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 864 } 865 } 866 867 if (ln->ln_router && !is_router) { 868 /* 869 * The peer dropped the router flag. 870 * Remove the sender from the Default Router List and 871 * update the Destination Cache entries. 872 */ 873 struct ifnet *nd6_ifp; 874 875 nd6_ifp = lltable_get_ifp(ln->lle_tbl); 876 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) && 877 (ND_IFINFO(nd6_ifp)->flags & 878 ND6_IFF_ACCEPT_RTADV) != 0) 879 /* 880 * Even if the neighbor is not in the default 881 * router list, the neighbor may be used as a 882 * next hop for some destinations (e.g. redirect 883 * case). So we must call rt6_flush explicitly. 884 */ 885 rt6_flush(&ip6->ip6_src, ifp); 886 } 887 ln->ln_router = is_router; 888 } 889 /* XXX - QL 890 * Does this matter? 891 * rt->rt_flags &= ~RTF_REJECT; 892 */ 893 ln->la_asked = 0; 894 if (ln->la_hold != NULL) 895 chain = nd6_grab_holdchain(ln); 896 freeit: 897 if (ln != NULL) 898 LLE_WUNLOCK(ln); 899 900 if (chain != NULL) 901 nd6_flush_holdchain(ifp, ln, chain); 902 903 if (checklink) 904 pfxlist_onlink_check(); 905 906 m_freem(m); 907 return; 908 909 bad: 910 if (ln != NULL) 911 LLE_WUNLOCK(ln); 912 913 ICMP6STAT_INC(icp6s_badna); 914 m_freem(m); 915 } 916 917 /* 918 * Neighbor advertisement output handling. 919 * 920 * Based on RFC 2461 921 * 922 * the following items are not implemented yet: 923 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 924 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 925 * 926 * tlladdr - 1 if include target link-layer address 927 * sdl0 - sockaddr_dl (= proxy NA) or NULL 928 */ 929 static void 930 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0, 931 const struct in6_addr *taddr6, u_long flags, int tlladdr, 932 struct sockaddr *sdl0, u_int fibnum) 933 { 934 struct mbuf *m; 935 struct m_tag *mtag; 936 struct ip6_hdr *ip6; 937 struct nd_neighbor_advert *nd_na; 938 struct ip6_moptions im6o; 939 struct in6_addr daddr6, dst6, src6; 940 uint32_t scopeid; 941 942 NET_EPOCH_ASSERT(); 943 944 int icmp6len, maxlen, error; 945 caddr_t mac = NULL; 946 947 daddr6 = *daddr6_0; /* make a local copy for modification */ 948 949 /* estimate the size of message */ 950 maxlen = sizeof(*ip6) + sizeof(*nd_na); 951 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; 952 KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( 953 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", 954 __func__, max_linkhdr, maxlen, MCLBYTES)); 955 956 if (max_linkhdr + maxlen > MHLEN) 957 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 958 else 959 m = m_gethdr(M_NOWAIT, MT_DATA); 960 if (m == NULL) 961 return; 962 M_SETFIB(m, fibnum); 963 964 if (IN6_IS_ADDR_MULTICAST(&daddr6)) { 965 m->m_flags |= M_MCAST; 966 im6o.im6o_multicast_ifp = ifp; 967 im6o.im6o_multicast_hlim = 255; 968 im6o.im6o_multicast_loop = 0; 969 } 970 971 icmp6len = sizeof(*nd_na); 972 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len; 973 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ 974 975 /* fill neighbor advertisement packet */ 976 ip6 = mtod(m, struct ip6_hdr *); 977 ip6->ip6_flow = 0; 978 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 979 ip6->ip6_vfc |= IPV6_VERSION; 980 ip6->ip6_nxt = IPPROTO_ICMPV6; 981 ip6->ip6_hlim = 255; 982 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) { 983 /* reply to DAD */ 984 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL; 985 daddr6.s6_addr16[1] = 0; 986 daddr6.s6_addr32[1] = 0; 987 daddr6.s6_addr32[2] = 0; 988 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE; 989 if (in6_setscope(&daddr6, ifp, NULL)) 990 goto bad; 991 992 flags &= ~ND_NA_FLAG_SOLICITED; 993 } 994 ip6->ip6_dst = daddr6; 995 996 /* 997 * Select a source whose scope is the same as that of the dest. 998 */ 999 in6_splitscope(&daddr6, &dst6, &scopeid); 1000 error = in6_selectsrc_addr(fibnum, &dst6, 1001 scopeid, ifp, &src6, NULL); 1002 if (error) { 1003 char ip6buf[INET6_ADDRSTRLEN]; 1004 nd6log((LOG_DEBUG, "nd6_na_output: source can't be " 1005 "determined: dst=%s, error=%d\n", 1006 ip6_sprintf(ip6buf, &daddr6), error)); 1007 goto bad; 1008 } 1009 ip6->ip6_src = src6; 1010 nd_na = (struct nd_neighbor_advert *)(ip6 + 1); 1011 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT; 1012 nd_na->nd_na_code = 0; 1013 nd_na->nd_na_target = *taddr6; 1014 in6_clearscope(&nd_na->nd_na_target); /* XXX */ 1015 1016 /* 1017 * "tlladdr" indicates NS's condition for adding tlladdr or not. 1018 * see nd6_ns_input() for details. 1019 * Basically, if NS packet is sent to unicast/anycast addr, 1020 * target lladdr option SHOULD NOT be included. 1021 */ 1022 if (tlladdr) { 1023 /* 1024 * sdl0 != NULL indicates proxy NA. If we do proxy, use 1025 * lladdr in sdl0. If we are not proxying (sending NA for 1026 * my address) use lladdr configured for the interface. 1027 */ 1028 if (sdl0 == NULL) { 1029 if (ifp->if_carp) 1030 mac = (*carp_macmatch6_p)(ifp, m, taddr6); 1031 if (mac == NULL) 1032 mac = nd6_ifptomac(ifp); 1033 } else if (sdl0->sa_family == AF_LINK) { 1034 struct sockaddr_dl *sdl; 1035 sdl = (struct sockaddr_dl *)sdl0; 1036 if (sdl->sdl_alen == ifp->if_addrlen) 1037 mac = LLADDR(sdl); 1038 } 1039 } 1040 if (tlladdr && mac) { 1041 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 1042 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1); 1043 1044 /* roundup to 8 bytes alignment! */ 1045 optlen = (optlen + 7) & ~7; 1046 1047 m->m_pkthdr.len += optlen; 1048 m->m_len += optlen; 1049 icmp6len += optlen; 1050 bzero((caddr_t)nd_opt, optlen); 1051 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 1052 nd_opt->nd_opt_len = optlen >> 3; 1053 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); 1054 } else 1055 flags &= ~ND_NA_FLAG_OVERRIDE; 1056 1057 ip6->ip6_plen = htons((u_short)icmp6len); 1058 nd_na->nd_na_flags_reserved = flags; 1059 nd_na->nd_na_cksum = 0; 1060 nd_na->nd_na_cksum = 1061 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len); 1062 1063 if (send_sendso_input_hook != NULL) { 1064 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 1065 sizeof(unsigned short), M_NOWAIT); 1066 if (mtag == NULL) 1067 goto bad; 1068 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type; 1069 m_tag_prepend(m, mtag); 1070 } 1071 1072 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL); 1073 icmp6_ifstat_inc(ifp, ifs6_out_msg); 1074 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert); 1075 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]); 1076 1077 return; 1078 1079 bad: 1080 m_freem(m); 1081 } 1082 1083 #ifndef BURN_BRIDGES 1084 void 1085 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0, 1086 const struct in6_addr *taddr6, u_long flags, int tlladdr, 1087 struct sockaddr *sdl0) 1088 { 1089 1090 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0, 1091 RT_DEFAULT_FIB); 1092 } 1093 #endif 1094 1095 caddr_t 1096 nd6_ifptomac(struct ifnet *ifp) 1097 { 1098 switch (ifp->if_type) { 1099 case IFT_ETHER: 1100 case IFT_IEEE1394: 1101 case IFT_L2VLAN: 1102 case IFT_INFINIBAND: 1103 case IFT_BRIDGE: 1104 return IF_LLADDR(ifp); 1105 default: 1106 return NULL; 1107 } 1108 } 1109 1110 struct dadq { 1111 TAILQ_ENTRY(dadq) dad_list; 1112 struct ifaddr *dad_ifa; 1113 int dad_count; /* max NS to send */ 1114 int dad_ns_tcount; /* # of trials to send NS */ 1115 int dad_ns_ocount; /* NS sent so far */ 1116 int dad_ns_icount; 1117 int dad_na_icount; 1118 int dad_ns_lcount; /* looped back NS */ 1119 int dad_loopbackprobe; /* probing state for loopback detection */ 1120 struct callout dad_timer_ch; 1121 struct vnet *dad_vnet; 1122 u_int dad_refcnt; 1123 #define ND_OPT_NONCE_LEN32 \ 1124 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t)) 1125 uint32_t dad_nonce[ND_OPT_NONCE_LEN32]; 1126 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */ 1127 }; 1128 1129 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq); 1130 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock); 1131 #define V_dadq VNET(dadq) 1132 #define V_dad_rwlock VNET(dad_rwlock) 1133 1134 #define DADQ_RLOCK() rw_rlock(&V_dad_rwlock) 1135 #define DADQ_RUNLOCK() rw_runlock(&V_dad_rwlock) 1136 #define DADQ_WLOCK() rw_wlock(&V_dad_rwlock) 1137 #define DADQ_WUNLOCK() rw_wunlock(&V_dad_rwlock) 1138 1139 static void 1140 nd6_dad_add(struct dadq *dp) 1141 { 1142 1143 DADQ_WLOCK(); 1144 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list); 1145 dp->dad_ondadq = true; 1146 DADQ_WUNLOCK(); 1147 } 1148 1149 static void 1150 nd6_dad_del(struct dadq *dp) 1151 { 1152 1153 DADQ_WLOCK(); 1154 if (dp->dad_ondadq) { 1155 /* 1156 * Remove dp from the dadq and release the dadq's 1157 * reference. 1158 */ 1159 TAILQ_REMOVE(&V_dadq, dp, dad_list); 1160 dp->dad_ondadq = false; 1161 DADQ_WUNLOCK(); 1162 nd6_dad_rele(dp); 1163 } else 1164 DADQ_WUNLOCK(); 1165 } 1166 1167 static struct dadq * 1168 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n) 1169 { 1170 struct dadq *dp; 1171 1172 DADQ_RLOCK(); 1173 TAILQ_FOREACH(dp, &V_dadq, dad_list) { 1174 if (dp->dad_ifa != ifa) 1175 continue; 1176 /* 1177 * Skip if the nonce matches the received one. 1178 * +2 in the length is required because of type and 1179 * length fields are included in a header. 1180 */ 1181 if (n != NULL && 1182 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 && 1183 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0], 1184 ND_OPT_NONCE_LEN) == 0) { 1185 dp->dad_ns_lcount++; 1186 continue; 1187 } 1188 refcount_acquire(&dp->dad_refcnt); 1189 break; 1190 } 1191 DADQ_RUNLOCK(); 1192 1193 return (dp); 1194 } 1195 1196 static void 1197 nd6_dad_starttimer(struct dadq *dp, int ticks, int send_ns) 1198 { 1199 1200 NET_EPOCH_ASSERT(); 1201 1202 if (send_ns != 0) 1203 nd6_dad_ns_output(dp); 1204 callout_reset(&dp->dad_timer_ch, ticks, 1205 (void (*)(void *))nd6_dad_timer, (void *)dp); 1206 } 1207 1208 static void 1209 nd6_dad_stoptimer(struct dadq *dp) 1210 { 1211 1212 callout_drain(&dp->dad_timer_ch); 1213 } 1214 1215 static void 1216 nd6_dad_rele(struct dadq *dp) 1217 { 1218 1219 if (refcount_release(&dp->dad_refcnt)) { 1220 ifa_free(dp->dad_ifa); 1221 free(dp, M_IP6NDP); 1222 } 1223 } 1224 1225 void 1226 nd6_dad_init(void) 1227 { 1228 1229 rw_init(&V_dad_rwlock, "nd6 DAD queue"); 1230 TAILQ_INIT(&V_dadq); 1231 } 1232 1233 /* 1234 * Start Duplicate Address Detection (DAD) for specified interface address. 1235 */ 1236 void 1237 nd6_dad_start(struct ifaddr *ifa, int delay) 1238 { 1239 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1240 struct dadq *dp; 1241 char ip6buf[INET6_ADDRSTRLEN]; 1242 struct epoch_tracker et; 1243 1244 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0, 1245 ("starting DAD on non-tentative address %p", ifa)); 1246 1247 /* 1248 * If we don't need DAD, don't do it. 1249 * There are several cases: 1250 * - DAD is disabled globally or on the interface 1251 * - the interface address is anycast 1252 */ 1253 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 || 1254 V_ip6_dad_count == 0 || 1255 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) { 1256 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1257 return; 1258 } 1259 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 || 1260 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || 1261 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0) 1262 return; 1263 1264 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) { 1265 /* 1266 * DAD is already in progress. Let the existing entry 1267 * finish it. 1268 */ 1269 nd6_dad_rele(dp); 1270 return; 1271 } 1272 1273 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO); 1274 if (dp == NULL) { 1275 log(LOG_ERR, "nd6_dad_start: memory allocation failed for " 1276 "%s(%s)\n", 1277 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1278 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1279 return; 1280 } 1281 callout_init(&dp->dad_timer_ch, 0); 1282 #ifdef VIMAGE 1283 dp->dad_vnet = curvnet; 1284 #endif 1285 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp), 1286 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1287 1288 /* 1289 * Send NS packet for DAD, ip6_dad_count times. 1290 * Note that we must delay the first transmission, if this is the 1291 * first packet to be sent from the interface after interface 1292 * (re)initialization. 1293 */ 1294 dp->dad_ifa = ifa; 1295 ifa_ref(dp->dad_ifa); 1296 dp->dad_count = V_ip6_dad_count; 1297 dp->dad_ns_icount = dp->dad_na_icount = 0; 1298 dp->dad_ns_ocount = dp->dad_ns_tcount = 0; 1299 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0; 1300 1301 /* Add this to the dadq and add a reference for the dadq. */ 1302 refcount_init(&dp->dad_refcnt, 1); 1303 nd6_dad_add(dp); 1304 NET_EPOCH_ENTER(et); 1305 nd6_dad_starttimer(dp, delay, 0); 1306 NET_EPOCH_EXIT(et); 1307 } 1308 1309 /* 1310 * terminate DAD unconditionally. used for address removals. 1311 */ 1312 void 1313 nd6_dad_stop(struct ifaddr *ifa) 1314 { 1315 struct dadq *dp; 1316 1317 dp = nd6_dad_find(ifa, NULL); 1318 if (!dp) { 1319 /* DAD wasn't started yet */ 1320 return; 1321 } 1322 1323 nd6_dad_stoptimer(dp); 1324 nd6_dad_del(dp); 1325 1326 /* Release this function's reference, acquired by nd6_dad_find(). */ 1327 nd6_dad_rele(dp); 1328 } 1329 1330 static void 1331 nd6_dad_timer(struct dadq *dp) 1332 { 1333 CURVNET_SET(dp->dad_vnet); 1334 struct ifaddr *ifa = dp->dad_ifa; 1335 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1336 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1337 char ip6buf[INET6_ADDRSTRLEN]; 1338 struct epoch_tracker et; 1339 1340 KASSERT(ia != NULL, ("DAD entry %p with no address", dp)); 1341 1342 NET_EPOCH_ENTER(et); 1343 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) { 1344 /* Do not need DAD for ifdisabled interface. */ 1345 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of " 1346 "ND6_IFF_IFDISABLED.\n", ifp->if_xname); 1347 goto err; 1348 } 1349 if (ia->ia6_flags & IN6_IFF_DUPLICATED) { 1350 log(LOG_ERR, "nd6_dad_timer: called with duplicated address " 1351 "%s(%s)\n", 1352 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1353 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1354 goto err; 1355 } 1356 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) { 1357 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address " 1358 "%s(%s)\n", 1359 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1360 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1361 goto err; 1362 } 1363 1364 /* Stop DAD if the interface is down even after dad_maxtry attempts. */ 1365 if ((dp->dad_ns_tcount > V_dad_maxtry) && 1366 (((ifp->if_flags & IFF_UP) == 0) || 1367 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) { 1368 nd6log((LOG_INFO, "%s: could not run DAD " 1369 "because the interface was down or not running.\n", 1370 if_name(ifa->ifa_ifp))); 1371 goto err; 1372 } 1373 1374 /* Need more checks? */ 1375 if (dp->dad_ns_ocount < dp->dad_count) { 1376 /* 1377 * We have more NS to go. Send NS packet for DAD. 1378 */ 1379 nd6_dad_starttimer(dp, 1380 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1); 1381 goto done; 1382 } else { 1383 /* 1384 * We have transmitted sufficient number of DAD packets. 1385 * See what we've got. 1386 */ 1387 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) 1388 /* We've seen NS or NA, means DAD has failed. */ 1389 nd6_dad_duplicated(ifa, dp); 1390 else if (V_dad_enhanced != 0 && 1391 dp->dad_ns_lcount > 0 && 1392 dp->dad_ns_lcount > dp->dad_loopbackprobe) { 1393 /* 1394 * Sec. 4.1 in RFC 7527 requires transmission of 1395 * additional probes until the loopback condition 1396 * becomes clear when a looped back probe is detected. 1397 */ 1398 log(LOG_ERR, "%s: a looped back NS message is " 1399 "detected during DAD for %s. " 1400 "Another DAD probes are being sent.\n", 1401 if_name(ifa->ifa_ifp), 1402 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1403 dp->dad_loopbackprobe = dp->dad_ns_lcount; 1404 /* 1405 * Send an NS immediately and increase dad_count by 1406 * V_nd6_mmaxtries - 1. 1407 */ 1408 dp->dad_count = 1409 dp->dad_ns_ocount + V_nd6_mmaxtries - 1; 1410 nd6_dad_starttimer(dp, 1411 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000, 1412 1); 1413 goto done; 1414 } else { 1415 /* 1416 * We are done with DAD. No NA came, no NS came. 1417 * No duplicate address found. Check IFDISABLED flag 1418 * again in case that it is changed between the 1419 * beginning of this function and here. 1420 */ 1421 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0) 1422 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1423 1424 nd6log((LOG_DEBUG, 1425 "%s: DAD complete for %s - no duplicates found\n", 1426 if_name(ifa->ifa_ifp), 1427 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1428 if (dp->dad_ns_lcount > 0) 1429 log(LOG_ERR, "%s: DAD completed while " 1430 "a looped back NS message is detected " 1431 "during DAD for %s.\n", 1432 if_name(ifa->ifa_ifp), 1433 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1434 } 1435 } 1436 err: 1437 nd6_dad_del(dp); 1438 done: 1439 NET_EPOCH_EXIT(et); 1440 CURVNET_RESTORE(); 1441 } 1442 1443 static void 1444 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp) 1445 { 1446 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1447 struct ifnet *ifp; 1448 char ip6buf[INET6_ADDRSTRLEN]; 1449 1450 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: " 1451 "NS in/out/loopback=%d/%d/%d, NA in=%d\n", 1452 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1453 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount, 1454 dp->dad_na_icount); 1455 1456 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1457 ia->ia6_flags |= IN6_IFF_DUPLICATED; 1458 1459 ifp = ifa->ifa_ifp; 1460 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n", 1461 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)); 1462 log(LOG_ERR, "%s: manual intervention required\n", 1463 if_name(ifp)); 1464 1465 /* 1466 * If the address is a link-local address formed from an interface 1467 * identifier based on the hardware address which is supposed to be 1468 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP 1469 * operation on the interface SHOULD be disabled. 1470 * [RFC 4862, Section 5.4.5] 1471 */ 1472 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1473 struct in6_addr in6; 1474 1475 /* 1476 * To avoid over-reaction, we only apply this logic when we are 1477 * very sure that hardware addresses are supposed to be unique. 1478 */ 1479 switch (ifp->if_type) { 1480 case IFT_ETHER: 1481 case IFT_ATM: 1482 case IFT_IEEE1394: 1483 case IFT_INFINIBAND: 1484 in6 = ia->ia_addr.sin6_addr; 1485 if (in6_get_hw_ifid(ifp, &in6) == 0 && 1486 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) { 1487 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1488 log(LOG_ERR, "%s: possible hardware address " 1489 "duplication detected, disable IPv6\n", 1490 if_name(ifp)); 1491 } 1492 break; 1493 } 1494 } 1495 } 1496 1497 static void 1498 nd6_dad_ns_output(struct dadq *dp) 1499 { 1500 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa; 1501 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1502 int i; 1503 1504 dp->dad_ns_tcount++; 1505 if ((ifp->if_flags & IFF_UP) == 0) { 1506 return; 1507 } 1508 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 1509 return; 1510 } 1511 1512 dp->dad_ns_ocount++; 1513 if (V_dad_enhanced != 0) { 1514 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) 1515 dp->dad_nonce[i] = arc4random(); 1516 /* 1517 * XXXHRS: Note that in the case that 1518 * DupAddrDetectTransmits > 1, multiple NS messages with 1519 * different nonces can be looped back in an unexpected 1520 * order. The current implementation recognizes only 1521 * the latest nonce on the sender side. Practically it 1522 * should work well in almost all cases. 1523 */ 1524 } 1525 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr, 1526 (uint8_t *)&dp->dad_nonce[0]); 1527 } 1528 1529 static void 1530 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce) 1531 { 1532 struct dadq *dp; 1533 1534 if (ifa == NULL) 1535 panic("ifa == NULL in nd6_dad_ns_input"); 1536 1537 /* Ignore Nonce option when Enhanced DAD is disabled. */ 1538 if (V_dad_enhanced == 0) 1539 ndopt_nonce = NULL; 1540 dp = nd6_dad_find(ifa, ndopt_nonce); 1541 if (dp == NULL) 1542 return; 1543 1544 dp->dad_ns_icount++; 1545 nd6_dad_rele(dp); 1546 } 1547 1548 static void 1549 nd6_dad_na_input(struct ifaddr *ifa) 1550 { 1551 struct dadq *dp; 1552 1553 if (ifa == NULL) 1554 panic("ifa == NULL in nd6_dad_na_input"); 1555 1556 dp = nd6_dad_find(ifa, NULL); 1557 if (dp != NULL) { 1558 dp->dad_na_icount++; 1559 nd6_dad_rele(dp); 1560 } 1561 } 1562