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); 88 static void nd6_dad_stoptimer(struct dadq *); 89 static void nd6_dad_timer(void *); 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 static struct ifaddr *nd6_proxy_fill_sdl(struct ifnet *, 100 const struct in6_addr *, struct sockaddr_dl *); 101 102 VNET_DEFINE_STATIC(int, dad_enhanced) = 1; 103 #define V_dad_enhanced VNET(dad_enhanced) 104 105 SYSCTL_DECL(_net_inet6_ip6); 106 SYSCTL_INT(_net_inet6_ip6, OID_AUTO, dad_enhanced, CTLFLAG_VNET | CTLFLAG_RW, 107 &VNET_NAME(dad_enhanced), 0, 108 "Enable Enhanced DAD, which adds a random nonce to NS messages for DAD."); 109 110 VNET_DEFINE_STATIC(int, dad_maxtry) = 15; /* max # of *tries* to 111 transmit DAD packet */ 112 #define V_dad_maxtry VNET(dad_maxtry) 113 114 /* 115 * Input a Neighbor Solicitation Message. 116 * 117 * Based on RFC 2461 118 * Based on RFC 2462 (duplicate address detection) 119 */ 120 void 121 nd6_ns_input(struct mbuf *m, int off, int icmp6len) 122 { 123 struct ifnet *ifp; 124 struct ip6_hdr *ip6; 125 struct nd_neighbor_solicit *nd_ns; 126 struct in6_addr daddr6, myaddr6, saddr6, taddr6; 127 struct ifaddr *ifa; 128 struct sockaddr_dl proxydl; 129 union nd_opts ndopts; 130 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 131 char *lladdr; 132 int anycast, lladdrlen, proxy, rflag, tentative, tlladdr; 133 134 ifa = NULL; 135 136 /* RFC 6980: Nodes MUST silently ignore fragments */ 137 if(m->m_flags & M_FRAGMENTED) 138 goto freeit; 139 140 ifp = m->m_pkthdr.rcvif; 141 ip6 = mtod(m, struct ip6_hdr *); 142 if (__predict_false(ip6->ip6_hlim != 255)) { 143 ICMP6STAT_INC(icp6s_invlhlim); 144 nd6log((LOG_ERR, 145 "nd6_ns_input: invalid hlim (%d) from %s to %s on %s\n", 146 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 147 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 148 goto bads; 149 } 150 151 if (m->m_len < off + icmp6len) { 152 m = m_pullup(m, off + icmp6len); 153 if (m == NULL) { 154 IP6STAT_INC(ip6s_exthdrtoolong); 155 return; 156 } 157 } 158 ip6 = mtod(m, struct ip6_hdr *); 159 nd_ns = (struct nd_neighbor_solicit *)((caddr_t)ip6 + off); 160 161 saddr6 = ip6->ip6_src; 162 daddr6 = ip6->ip6_dst; 163 taddr6 = nd_ns->nd_ns_target; 164 if (in6_setscope(&taddr6, ifp, NULL) != 0) 165 goto bad; 166 167 rflag = (V_ip6_forwarding) ? ND_NA_FLAG_ROUTER : 0; 168 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && V_ip6_norbit_raif) 169 rflag = 0; 170 171 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 172 /* dst has to be a solicited node multicast address. */ 173 if (daddr6.s6_addr16[0] == IPV6_ADDR_INT16_MLL && 174 /* don't check ifindex portion */ 175 daddr6.s6_addr32[1] == 0 && 176 daddr6.s6_addr32[2] == IPV6_ADDR_INT32_ONE && 177 daddr6.s6_addr8[12] == 0xff) { 178 ; /* good */ 179 } else { 180 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 181 "(wrong ip6 dst)\n")); 182 goto bad; 183 } 184 } else if (!V_nd6_onlink_ns_rfc4861) { 185 struct sockaddr_in6 src_sa6; 186 187 /* 188 * According to recent IETF discussions, it is not a good idea 189 * to accept a NS from an address which would not be deemed 190 * to be a neighbor otherwise. This point is expected to be 191 * clarified in future revisions of the specification. 192 */ 193 bzero(&src_sa6, sizeof(src_sa6)); 194 src_sa6.sin6_family = AF_INET6; 195 src_sa6.sin6_len = sizeof(src_sa6); 196 src_sa6.sin6_addr = saddr6; 197 if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) { 198 nd6log((LOG_INFO, "nd6_ns_input: " 199 "NS packet from non-neighbor\n")); 200 goto bad; 201 } 202 } 203 204 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 205 nd6log((LOG_INFO, "nd6_ns_input: bad NS target (multicast)\n")); 206 goto bad; 207 } 208 209 icmp6len -= sizeof(*nd_ns); 210 nd6_option_init(nd_ns + 1, icmp6len, &ndopts); 211 if (nd6_options(&ndopts) < 0) { 212 nd6log((LOG_INFO, 213 "nd6_ns_input: invalid ND option, ignored\n")); 214 /* nd6_options have incremented stats */ 215 goto freeit; 216 } 217 218 lladdr = NULL; 219 lladdrlen = 0; 220 if (ndopts.nd_opts_src_lladdr) { 221 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1); 222 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3; 223 } 224 225 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) && lladdr) { 226 nd6log((LOG_INFO, "nd6_ns_input: bad DAD packet " 227 "(link-layer address option)\n")); 228 goto bad; 229 } 230 231 /* 232 * Attaching target link-layer address to the NA? 233 * (RFC 2461 7.2.4) 234 * 235 * NS IP dst is unicast/anycast MUST NOT add 236 * NS IP dst is solicited-node multicast MUST add 237 * 238 * In implementation, we add target link-layer address by default. 239 * We do not add one in MUST NOT cases. 240 */ 241 if (!IN6_IS_ADDR_MULTICAST(&daddr6)) 242 tlladdr = 0; 243 else 244 tlladdr = 1; 245 246 /* 247 * Target address (taddr6) must be either: 248 * (1) Valid unicast/anycast address for my receiving interface, 249 * (2) Unicast address for which I'm offering proxy service, or 250 * (3) "tentative" address on which DAD is being performed. 251 */ 252 /* (1) and (3) check. */ 253 if (ifp->if_carp) 254 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 255 else 256 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 257 258 /* (2) check. */ 259 proxy = 0; 260 if (ifa == NULL) { 261 if ((ifa = nd6_proxy_fill_sdl(ifp, &taddr6, &proxydl)) != NULL) 262 proxy = 1; 263 } 264 if (ifa == NULL) { 265 /* 266 * We've got an NS packet, and we don't have that adddress 267 * assigned for us. We MUST silently ignore it. 268 * See RFC2461 7.2.3. 269 */ 270 goto freeit; 271 } 272 myaddr6 = *IFA_IN6(ifa); 273 anycast = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST; 274 tentative = ((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE; 275 if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DUPLICATED) 276 goto freeit; 277 278 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 279 nd6log((LOG_INFO, "nd6_ns_input: lladdrlen mismatch for %s " 280 "(if %d, NS packet %d)\n", 281 ip6_sprintf(ip6bufs, &taddr6), 282 ifp->if_addrlen, lladdrlen - 2)); 283 goto bad; 284 } 285 286 if (IN6_ARE_ADDR_EQUAL(&myaddr6, &saddr6)) { 287 nd6log((LOG_INFO, "nd6_ns_input: duplicate IP6 address %s\n", 288 ip6_sprintf(ip6bufs, &saddr6))); 289 goto freeit; 290 } 291 292 /* 293 * We have neighbor solicitation packet, with target address equals to 294 * one of my tentative address. 295 * 296 * src addr how to process? 297 * --- --- 298 * multicast of course, invalid (rejected in ip6_input) 299 * unicast somebody is doing address resolution -> ignore 300 * unspec dup address detection 301 * 302 * The processing is defined in RFC 2462. 303 */ 304 if (tentative) { 305 /* 306 * If source address is unspecified address, it is for 307 * duplicate address detection. 308 * 309 * If not, the packet is for addess resolution; 310 * silently ignore it. 311 */ 312 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) 313 nd6_dad_ns_input(ifa, ndopts.nd_opts_nonce); 314 315 goto freeit; 316 } 317 318 /* 319 * If the source address is unspecified address, entries must not 320 * be created or updated. 321 * It looks that sender is performing DAD. Output NA toward 322 * all-node multicast address, to tell the sender that I'm using 323 * the address. 324 * S bit ("solicited") must be zero. 325 */ 326 if (IN6_IS_ADDR_UNSPECIFIED(&saddr6)) { 327 struct in6_addr in6_all; 328 329 in6_all = in6addr_linklocal_allnodes; 330 if (in6_setscope(&in6_all, ifp, NULL) != 0) 331 goto bad; 332 nd6_na_output_fib(ifp, &in6_all, &taddr6, 333 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 334 rflag, tlladdr, proxy ? (struct sockaddr *)&proxydl : NULL, 335 M_GETFIB(m)); 336 goto freeit; 337 } 338 339 nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, 340 ND_NEIGHBOR_SOLICIT, 0); 341 342 nd6_na_output_fib(ifp, &saddr6, &taddr6, 343 ((anycast || proxy || !tlladdr) ? 0 : ND_NA_FLAG_OVERRIDE) | 344 rflag | ND_NA_FLAG_SOLICITED, tlladdr, 345 proxy ? (struct sockaddr *)&proxydl : NULL, M_GETFIB(m)); 346 freeit: 347 if (ifa != NULL) 348 ifa_free(ifa); 349 m_freem(m); 350 return; 351 352 bad: 353 nd6log((LOG_ERR, "nd6_ns_input: src=%s\n", 354 ip6_sprintf(ip6bufs, &saddr6))); 355 nd6log((LOG_ERR, "nd6_ns_input: dst=%s\n", 356 ip6_sprintf(ip6bufs, &daddr6))); 357 nd6log((LOG_ERR, "nd6_ns_input: tgt=%s\n", 358 ip6_sprintf(ip6bufs, &taddr6))); 359 bads: 360 ICMP6STAT_INC(icp6s_badns); 361 if (ifa != NULL) 362 ifa_free(ifa); 363 m_freem(m); 364 } 365 366 static struct ifaddr * 367 nd6_proxy_fill_sdl(struct ifnet *ifp, const struct in6_addr *taddr6, 368 struct sockaddr_dl *sdl) 369 { 370 struct ifaddr *ifa; 371 struct llentry *ln; 372 373 ifa = NULL; 374 ln = nd6_lookup(taddr6, LLE_SF(AF_INET6, 0), ifp); 375 if (ln == NULL) 376 return (ifa); 377 if ((ln->la_flags & (LLE_PUB | LLE_VALID)) == (LLE_PUB | LLE_VALID)) { 378 link_init_sdl(ifp, (struct sockaddr *)sdl, ifp->if_type); 379 sdl->sdl_alen = ifp->if_addrlen; 380 bcopy(ln->ll_addr, &sdl->sdl_data, ifp->if_addrlen); 381 LLE_RUNLOCK(ln); 382 ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 383 IN6_IFF_NOTREADY|IN6_IFF_ANYCAST); 384 } else 385 LLE_RUNLOCK(ln); 386 387 return (ifa); 388 } 389 390 /* 391 * Output a Neighbor Solicitation Message. Caller specifies: 392 * - ICMP6 header source IP6 address 393 * - ND6 header target IP6 address 394 * - ND6 header source datalink address 395 * 396 * Based on RFC 2461 397 * Based on RFC 2462 (duplicate address detection) 398 * 399 * ln - for source address determination 400 * nonce - If non-NULL, NS is used for duplicate address detection and 401 * the value (length is ND_OPT_NONCE_LEN) is used as a random nonce. 402 */ 403 static void 404 nd6_ns_output_fib(struct ifnet *ifp, const struct in6_addr *saddr6, 405 const struct in6_addr *daddr6, const struct in6_addr *taddr6, 406 uint8_t *nonce, u_int fibnum) 407 { 408 struct mbuf *m; 409 struct m_tag *mtag; 410 struct ip6_hdr *ip6; 411 struct nd_neighbor_solicit *nd_ns; 412 struct ip6_moptions im6o; 413 int icmp6len; 414 int maxlen; 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) { 538 struct nd_opt_hdr *nd_opt; 539 char *mac; 540 int optlen; 541 542 mac = NULL; 543 if (ifp->if_carp) 544 mac = (*carp_macmatch6_p)(ifp, m, &ip6->ip6_src); 545 if (mac == NULL) 546 mac = nd6_ifptomac(ifp); 547 548 if (mac != NULL) { 549 nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); 550 optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 551 /* 8 byte alignments... */ 552 optlen = (optlen + 7) & ~7; 553 m->m_pkthdr.len += optlen; 554 m->m_len += optlen; 555 icmp6len += optlen; 556 bzero(nd_opt, optlen); 557 nd_opt->nd_opt_type = ND_OPT_SOURCE_LINKADDR; 558 nd_opt->nd_opt_len = optlen >> 3; 559 bcopy(mac, nd_opt + 1, ifp->if_addrlen); 560 } 561 } 562 /* 563 * Add a Nonce option (RFC 3971) to detect looped back NS messages. 564 * This behavior is documented as Enhanced Duplicate Address 565 * Detection in RFC 7527. 566 * net.inet6.ip6.dad_enhanced=0 disables this. 567 */ 568 if (V_dad_enhanced != 0 && nonce != NULL) { 569 int optlen = sizeof(struct nd_opt_hdr) + ND_OPT_NONCE_LEN; 570 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_ns + 1); 571 /* 8-byte alignment is required. */ 572 optlen = (optlen + 7) & ~7; 573 574 m->m_pkthdr.len += optlen; 575 m->m_len += optlen; 576 icmp6len += optlen; 577 bzero((caddr_t)nd_opt, optlen); 578 nd_opt->nd_opt_type = ND_OPT_NONCE; 579 nd_opt->nd_opt_len = optlen >> 3; 580 bcopy(nonce, (caddr_t)(nd_opt + 1), ND_OPT_NONCE_LEN); 581 } 582 ip6->ip6_plen = htons((u_short)icmp6len); 583 nd_ns->nd_ns_cksum = 0; 584 nd_ns->nd_ns_cksum = 585 in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), icmp6len); 586 587 if (send_sendso_input_hook != NULL) { 588 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 589 sizeof(unsigned short), M_NOWAIT); 590 if (mtag == NULL) 591 goto bad; 592 *(unsigned short *)(mtag + 1) = nd_ns->nd_ns_type; 593 m_tag_prepend(m, mtag); 594 } 595 596 ip6_output(m, NULL, NULL, (nonce != NULL) ? IPV6_UNSPECSRC : 0, 597 &im6o, NULL, NULL); 598 icmp6_ifstat_inc(ifp, ifs6_out_msg); 599 icmp6_ifstat_inc(ifp, ifs6_out_neighborsolicit); 600 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_SOLICIT]); 601 602 return; 603 604 bad: 605 m_freem(m); 606 } 607 608 #ifndef BURN_BRIDGES 609 void 610 nd6_ns_output(struct ifnet *ifp, const struct in6_addr *saddr6, 611 const struct in6_addr *daddr6, const struct in6_addr *taddr6,uint8_t *nonce) 612 { 613 614 nd6_ns_output_fib(ifp, saddr6, daddr6, taddr6, nonce, RT_DEFAULT_FIB); 615 } 616 #endif 617 /* 618 * Neighbor advertisement input handling. 619 * 620 * Based on RFC 2461 621 * Based on RFC 2462 (duplicate address detection) 622 * 623 * the following items are not implemented yet: 624 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 625 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 626 */ 627 void 628 nd6_na_input(struct mbuf *m, int off, int icmp6len) 629 { 630 struct ifnet *ifp; 631 struct ip6_hdr *ip6; 632 struct ifaddr *ifa; 633 struct llentry *ln; 634 struct mbuf *chain; 635 struct nd_neighbor_advert *nd_na; 636 struct in6_addr daddr6, taddr6; 637 union nd_opts ndopts; 638 u_char linkhdr[LLE_MAX_LINKHDR]; 639 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 640 char *lladdr; 641 size_t linkhdrsize; 642 int flags, is_override, is_router, is_solicited; 643 int lladdr_off, lladdrlen, checklink; 644 bool flush_holdchain = false; 645 646 NET_EPOCH_ASSERT(); 647 648 chain = NULL; 649 ln = NULL; 650 checklink = 0; 651 652 /* RFC 6980: Nodes MUST silently ignore fragments */ 653 if(m->m_flags & M_FRAGMENTED) 654 goto freeit; 655 656 ifp = m->m_pkthdr.rcvif; 657 ip6 = mtod(m, struct ip6_hdr *); 658 if (__predict_false(ip6->ip6_hlim != 255)) { 659 ICMP6STAT_INC(icp6s_invlhlim); 660 nd6log((LOG_ERR, 661 "nd6_na_input: invalid hlim (%d) from %s to %s on %s\n", 662 ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src), 663 ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp))); 664 goto bad; 665 } 666 667 if (m->m_len < off + icmp6len) { 668 m = m_pullup(m, off + icmp6len); 669 if (m == NULL) { 670 IP6STAT_INC(ip6s_exthdrtoolong); 671 return; 672 } 673 } 674 ip6 = mtod(m, struct ip6_hdr *); 675 nd_na = (struct nd_neighbor_advert *)((caddr_t)ip6 + off); 676 677 flags = nd_na->nd_na_flags_reserved; 678 is_router = ((flags & ND_NA_FLAG_ROUTER) != 0); 679 is_solicited = ((flags & ND_NA_FLAG_SOLICITED) != 0); 680 is_override = ((flags & ND_NA_FLAG_OVERRIDE) != 0); 681 682 taddr6 = nd_na->nd_na_target; 683 if (in6_setscope(&taddr6, ifp, NULL)) 684 goto bad; /* XXX: impossible */ 685 686 if (IN6_IS_ADDR_MULTICAST(&taddr6)) { 687 nd6log((LOG_ERR, 688 "nd6_na_input: invalid target address %s\n", 689 ip6_sprintf(ip6bufs, &taddr6))); 690 goto bad; 691 } 692 693 daddr6 = ip6->ip6_dst; 694 if (IN6_IS_ADDR_MULTICAST(&daddr6)) 695 if (is_solicited) { 696 nd6log((LOG_ERR, 697 "nd6_na_input: a solicited adv is multicasted\n")); 698 goto bad; 699 } 700 701 icmp6len -= sizeof(*nd_na); 702 nd6_option_init(nd_na + 1, icmp6len, &ndopts); 703 if (nd6_options(&ndopts) < 0) { 704 nd6log((LOG_INFO, 705 "nd6_na_input: invalid ND option, ignored\n")); 706 /* nd6_options have incremented stats */ 707 goto freeit; 708 } 709 710 lladdr = NULL; 711 lladdrlen = 0; 712 if (ndopts.nd_opts_tgt_lladdr) { 713 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 714 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 715 } 716 717 /* 718 * This effectively disables the DAD check on a non-master CARP 719 * address. 720 */ 721 if (ifp->if_carp) 722 ifa = (*carp_iamatch6_p)(ifp, &taddr6); 723 else 724 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp, &taddr6); 725 726 /* 727 * Target address matches one of my interface address. 728 * 729 * If my address is tentative, this means that there's somebody 730 * already using the same address as mine. This indicates DAD failure. 731 * This is defined in RFC 2462. 732 * 733 * Otherwise, process as defined in RFC 2461. 734 */ 735 if (ifa 736 && (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_TENTATIVE)) { 737 nd6_dad_na_input(ifa); 738 ifa_free(ifa); 739 goto freeit; 740 } 741 742 /* Just for safety, maybe unnecessary. */ 743 if (ifa) { 744 ifa_free(ifa); 745 log(LOG_ERR, 746 "nd6_na_input: duplicate IP6 address %s\n", 747 ip6_sprintf(ip6bufs, &taddr6)); 748 goto freeit; 749 } 750 751 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 752 nd6log((LOG_INFO, "nd6_na_input: lladdrlen mismatch for %s " 753 "(if %d, NA packet %d)\n", ip6_sprintf(ip6bufs, &taddr6), 754 ifp->if_addrlen, lladdrlen - 2)); 755 goto bad; 756 } 757 758 /* 759 * If no neighbor cache entry is found, NA SHOULD silently be 760 * discarded. 761 */ 762 ln = nd6_lookup(&taddr6, LLE_SF(AF_INET6, LLE_EXCLUSIVE), ifp); 763 if (ln == NULL) { 764 goto freeit; 765 } 766 767 /* 768 * Do not try to override static entry. 769 */ 770 if (ln->la_flags & LLE_STATIC) 771 goto freeit; 772 773 if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 774 /* 775 * If the link-layer has address, and no lladdr option came, 776 * discard the packet. 777 */ 778 if (ifp->if_addrlen && lladdr == NULL) { 779 goto freeit; 780 } 781 782 /* 783 * Record link-layer address, and update the state. 784 */ 785 if (!nd6_try_set_entry_addr(ifp, ln, lladdr)) 786 goto freeit; 787 788 flush_holdchain = true; 789 if (is_solicited) 790 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 791 else 792 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 793 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 794 if ((ln->ln_router = is_router) != 0) { 795 /* 796 * This means a router's state has changed from 797 * non-reachable to probably reachable, and might 798 * affect the status of associated prefixes.. 799 */ 800 checklink = 1; 801 } 802 } else { 803 int llchange; 804 805 /* 806 * Check if the link-layer address has changed or not. 807 */ 808 if (lladdr == NULL) 809 llchange = 0; 810 else { 811 if (ln->la_flags & LLE_VALID) { 812 if (bcmp(lladdr, ln->ll_addr, ifp->if_addrlen)) 813 llchange = 1; 814 else 815 llchange = 0; 816 } else 817 llchange = 1; 818 } 819 820 /* 821 * This is VERY complex. Look at it with care. 822 * 823 * override solicit lladdr llchange action 824 * (L: record lladdr) 825 * 826 * 0 0 n -- (2c) 827 * 0 0 y n (2b) L 828 * 0 0 y y (1) REACHABLE->STALE 829 * 0 1 n -- (2c) *->REACHABLE 830 * 0 1 y n (2b) L *->REACHABLE 831 * 0 1 y y (1) REACHABLE->STALE 832 * 1 0 n -- (2a) 833 * 1 0 y n (2a) L 834 * 1 0 y y (2a) L *->STALE 835 * 1 1 n -- (2a) *->REACHABLE 836 * 1 1 y n (2a) L *->REACHABLE 837 * 1 1 y y (2a) L *->REACHABLE 838 */ 839 if (!is_override && (lladdr != NULL && llchange)) { /* (1) */ 840 /* 841 * If state is REACHABLE, make it STALE. 842 * no other updates should be done. 843 */ 844 if (ln->ln_state == ND6_LLINFO_REACHABLE) 845 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 846 goto freeit; 847 } else if (is_override /* (2a) */ 848 || (!is_override && (lladdr != NULL && !llchange)) /* (2b) */ 849 || lladdr == NULL) { /* (2c) */ 850 /* 851 * Update link-local address, if any. 852 */ 853 if (lladdr != NULL) { 854 linkhdrsize = sizeof(linkhdr); 855 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 856 linkhdr, &linkhdrsize, &lladdr_off) != 0) 857 goto freeit; 858 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, 859 linkhdrsize, lladdr_off) == 0) 860 goto freeit; 861 EVENTHANDLER_INVOKE(lle_event, ln, 862 LLENTRY_RESOLVED); 863 } 864 865 /* 866 * If solicited, make the state REACHABLE. 867 * If not solicited and the link-layer address was 868 * changed, make it STALE. 869 */ 870 if (is_solicited) 871 nd6_llinfo_setstate(ln, ND6_LLINFO_REACHABLE); 872 else { 873 if (lladdr != NULL && llchange) 874 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 875 } 876 } 877 878 if (ln->ln_router && !is_router) { 879 /* 880 * The peer dropped the router flag. 881 * Remove the sender from the Default Router List and 882 * update the Destination Cache entries. 883 */ 884 struct ifnet *nd6_ifp; 885 886 nd6_ifp = lltable_get_ifp(ln->lle_tbl); 887 if (!defrouter_remove(&ln->r_l3addr.addr6, nd6_ifp) && 888 (ND_IFINFO(nd6_ifp)->flags & 889 ND6_IFF_ACCEPT_RTADV) != 0) 890 /* 891 * Even if the neighbor is not in the default 892 * router list, the neighbor may be used as a 893 * next hop for some destinations (e.g. redirect 894 * case). So we must call rt6_flush explicitly. 895 */ 896 rt6_flush(&ip6->ip6_src, ifp); 897 } 898 ln->ln_router = is_router; 899 } 900 /* XXX - QL 901 * Does this matter? 902 * rt->rt_flags &= ~RTF_REJECT; 903 */ 904 ln->la_asked = 0; 905 if (ln->la_hold != NULL) 906 chain = nd6_grab_holdchain(ln); 907 freeit: 908 if (ln != NULL) 909 LLE_WUNLOCK(ln); 910 911 if (chain != NULL) 912 nd6_flush_holdchain(ifp, ln, chain); 913 if (flush_holdchain) 914 nd6_flush_children_holdchain(ifp, ln); 915 916 if (checklink) 917 pfxlist_onlink_check(); 918 919 m_freem(m); 920 return; 921 922 bad: 923 if (ln != NULL) 924 LLE_WUNLOCK(ln); 925 926 ICMP6STAT_INC(icp6s_badna); 927 m_freem(m); 928 } 929 930 /* 931 * Neighbor advertisement output handling. 932 * 933 * Based on RFC 2461 934 * 935 * the following items are not implemented yet: 936 * - proxy advertisement delay rule (RFC2461 7.2.8, last paragraph, SHOULD) 937 * - anycast advertisement delay rule (RFC2461 7.2.7, SHOULD) 938 * 939 * tlladdr - 1 if include target link-layer address 940 * sdl0 - sockaddr_dl (= proxy NA) or NULL 941 */ 942 static void 943 nd6_na_output_fib(struct ifnet *ifp, const struct in6_addr *daddr6_0, 944 const struct in6_addr *taddr6, u_long flags, int tlladdr, 945 struct sockaddr *sdl0, u_int fibnum) 946 { 947 struct mbuf *m; 948 struct m_tag *mtag; 949 struct ip6_hdr *ip6; 950 struct nd_neighbor_advert *nd_na; 951 struct ip6_moptions im6o; 952 struct in6_addr daddr6, dst6, src6; 953 uint32_t scopeid; 954 955 NET_EPOCH_ASSERT(); 956 957 int icmp6len, maxlen, error; 958 caddr_t mac = NULL; 959 960 daddr6 = *daddr6_0; /* make a local copy for modification */ 961 962 /* estimate the size of message */ 963 maxlen = sizeof(*ip6) + sizeof(*nd_na); 964 maxlen += (sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7; 965 KASSERT(max_linkhdr + maxlen <= MCLBYTES, ( 966 "%s: max_linkhdr + maxlen > MCLBYTES (%d + %d > %d)", 967 __func__, max_linkhdr, maxlen, MCLBYTES)); 968 969 if (max_linkhdr + maxlen > MHLEN) 970 m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR); 971 else 972 m = m_gethdr(M_NOWAIT, MT_DATA); 973 if (m == NULL) 974 return; 975 M_SETFIB(m, fibnum); 976 977 if (IN6_IS_ADDR_MULTICAST(&daddr6)) { 978 m->m_flags |= M_MCAST; 979 im6o.im6o_multicast_ifp = ifp; 980 im6o.im6o_multicast_hlim = 255; 981 im6o.im6o_multicast_loop = 0; 982 } 983 984 icmp6len = sizeof(*nd_na); 985 m->m_pkthdr.len = m->m_len = sizeof(struct ip6_hdr) + icmp6len; 986 m->m_data += max_linkhdr; /* or M_ALIGN() equivalent? */ 987 988 /* fill neighbor advertisement packet */ 989 ip6 = mtod(m, struct ip6_hdr *); 990 ip6->ip6_flow = 0; 991 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 992 ip6->ip6_vfc |= IPV6_VERSION; 993 ip6->ip6_nxt = IPPROTO_ICMPV6; 994 ip6->ip6_hlim = 255; 995 if (IN6_IS_ADDR_UNSPECIFIED(&daddr6)) { 996 /* reply to DAD */ 997 daddr6.s6_addr16[0] = IPV6_ADDR_INT16_MLL; 998 daddr6.s6_addr16[1] = 0; 999 daddr6.s6_addr32[1] = 0; 1000 daddr6.s6_addr32[2] = 0; 1001 daddr6.s6_addr32[3] = IPV6_ADDR_INT32_ONE; 1002 if (in6_setscope(&daddr6, ifp, NULL)) 1003 goto bad; 1004 1005 flags &= ~ND_NA_FLAG_SOLICITED; 1006 } 1007 ip6->ip6_dst = daddr6; 1008 1009 /* 1010 * Select a source whose scope is the same as that of the dest. 1011 */ 1012 in6_splitscope(&daddr6, &dst6, &scopeid); 1013 error = in6_selectsrc_addr(fibnum, &dst6, 1014 scopeid, ifp, &src6, NULL); 1015 if (error) { 1016 char ip6buf[INET6_ADDRSTRLEN]; 1017 nd6log((LOG_DEBUG, "nd6_na_output: source can't be " 1018 "determined: dst=%s, error=%d\n", 1019 ip6_sprintf(ip6buf, &daddr6), error)); 1020 goto bad; 1021 } 1022 ip6->ip6_src = src6; 1023 nd_na = (struct nd_neighbor_advert *)(ip6 + 1); 1024 nd_na->nd_na_type = ND_NEIGHBOR_ADVERT; 1025 nd_na->nd_na_code = 0; 1026 nd_na->nd_na_target = *taddr6; 1027 in6_clearscope(&nd_na->nd_na_target); /* XXX */ 1028 1029 /* 1030 * "tlladdr" indicates NS's condition for adding tlladdr or not. 1031 * see nd6_ns_input() for details. 1032 * Basically, if NS packet is sent to unicast/anycast addr, 1033 * target lladdr option SHOULD NOT be included. 1034 */ 1035 if (tlladdr) { 1036 /* 1037 * sdl0 != NULL indicates proxy NA. If we do proxy, use 1038 * lladdr in sdl0. If we are not proxying (sending NA for 1039 * my address) use lladdr configured for the interface. 1040 */ 1041 if (sdl0 == NULL) { 1042 if (ifp->if_carp) 1043 mac = (*carp_macmatch6_p)(ifp, m, taddr6); 1044 if (mac == NULL) 1045 mac = nd6_ifptomac(ifp); 1046 } else if (sdl0->sa_family == AF_LINK) { 1047 struct sockaddr_dl *sdl; 1048 sdl = (struct sockaddr_dl *)sdl0; 1049 if (sdl->sdl_alen == ifp->if_addrlen) 1050 mac = LLADDR(sdl); 1051 } 1052 } 1053 if (tlladdr && mac) { 1054 int optlen = sizeof(struct nd_opt_hdr) + ifp->if_addrlen; 1055 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)(nd_na + 1); 1056 1057 /* roundup to 8 bytes alignment! */ 1058 optlen = (optlen + 7) & ~7; 1059 1060 m->m_pkthdr.len += optlen; 1061 m->m_len += optlen; 1062 icmp6len += optlen; 1063 bzero((caddr_t)nd_opt, optlen); 1064 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 1065 nd_opt->nd_opt_len = optlen >> 3; 1066 bcopy(mac, (caddr_t)(nd_opt + 1), ifp->if_addrlen); 1067 } else 1068 flags &= ~ND_NA_FLAG_OVERRIDE; 1069 1070 ip6->ip6_plen = htons((u_short)icmp6len); 1071 nd_na->nd_na_flags_reserved = flags; 1072 nd_na->nd_na_cksum = 0; 1073 nd_na->nd_na_cksum = 1074 in6_cksum(m, IPPROTO_ICMPV6, sizeof(struct ip6_hdr), icmp6len); 1075 1076 if (send_sendso_input_hook != NULL) { 1077 mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, 1078 sizeof(unsigned short), M_NOWAIT); 1079 if (mtag == NULL) 1080 goto bad; 1081 *(unsigned short *)(mtag + 1) = nd_na->nd_na_type; 1082 m_tag_prepend(m, mtag); 1083 } 1084 1085 ip6_output(m, NULL, NULL, 0, &im6o, NULL, NULL); 1086 icmp6_ifstat_inc(ifp, ifs6_out_msg); 1087 icmp6_ifstat_inc(ifp, ifs6_out_neighboradvert); 1088 ICMP6STAT_INC(icp6s_outhist[ND_NEIGHBOR_ADVERT]); 1089 1090 return; 1091 1092 bad: 1093 m_freem(m); 1094 } 1095 1096 #ifndef BURN_BRIDGES 1097 void 1098 nd6_na_output(struct ifnet *ifp, const struct in6_addr *daddr6_0, 1099 const struct in6_addr *taddr6, u_long flags, int tlladdr, 1100 struct sockaddr *sdl0) 1101 { 1102 1103 nd6_na_output_fib(ifp, daddr6_0, taddr6, flags, tlladdr, sdl0, 1104 RT_DEFAULT_FIB); 1105 } 1106 #endif 1107 1108 caddr_t 1109 nd6_ifptomac(struct ifnet *ifp) 1110 { 1111 switch (ifp->if_type) { 1112 case IFT_ETHER: 1113 case IFT_IEEE1394: 1114 case IFT_L2VLAN: 1115 case IFT_INFINIBAND: 1116 case IFT_BRIDGE: 1117 return IF_LLADDR(ifp); 1118 default: 1119 return NULL; 1120 } 1121 } 1122 1123 struct dadq { 1124 TAILQ_ENTRY(dadq) dad_list; 1125 struct ifaddr *dad_ifa; 1126 int dad_count; /* max NS to send */ 1127 int dad_ns_tcount; /* # of trials to send NS */ 1128 int dad_ns_ocount; /* NS sent so far */ 1129 int dad_ns_icount; 1130 int dad_na_icount; 1131 int dad_ns_lcount; /* looped back NS */ 1132 int dad_loopbackprobe; /* probing state for loopback detection */ 1133 struct callout dad_timer_ch; 1134 struct vnet *dad_vnet; 1135 u_int dad_refcnt; 1136 #define ND_OPT_NONCE_LEN32 \ 1137 ((ND_OPT_NONCE_LEN + sizeof(uint32_t) - 1)/sizeof(uint32_t)) 1138 uint32_t dad_nonce[ND_OPT_NONCE_LEN32]; 1139 bool dad_ondadq; /* on dadq? Protected by DADQ_WLOCK. */ 1140 }; 1141 1142 VNET_DEFINE_STATIC(TAILQ_HEAD(, dadq), dadq); 1143 VNET_DEFINE_STATIC(struct rwlock, dad_rwlock); 1144 #define V_dadq VNET(dadq) 1145 #define V_dad_rwlock VNET(dad_rwlock) 1146 1147 #define DADQ_LOCKPTR() (&V_dad_rwlock) 1148 #define DADQ_LOCK_INIT() rw_init(DADQ_LOCKPTR(), "nd6 DAD queue") 1149 #define DADQ_RLOCK() rw_rlock(DADQ_LOCKPTR()) 1150 #define DADQ_RUNLOCK() rw_runlock(DADQ_LOCKPTR()) 1151 #define DADQ_WLOCK() rw_wlock(DADQ_LOCKPTR()) 1152 #define DADQ_WUNLOCK() rw_wunlock(DADQ_LOCKPTR()) 1153 1154 #define DADQ_LOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_LOCKED); 1155 #define DADQ_RLOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_RLOCKED); 1156 #define DADQ_WLOCK_ASSERT() rw_assert(DADQ_LOCKPTR(), RA_WLOCKED); 1157 1158 static void 1159 nd6_dad_add(struct dadq *dp) 1160 { 1161 DADQ_WLOCK_ASSERT(); 1162 1163 TAILQ_INSERT_TAIL(&V_dadq, dp, dad_list); 1164 dp->dad_ondadq = true; 1165 } 1166 1167 static void 1168 nd6_dad_del(struct dadq *dp) 1169 { 1170 DADQ_WLOCK_ASSERT(); 1171 1172 if (dp->dad_ondadq) { 1173 /* 1174 * Remove dp from the dadq and release the dadq's 1175 * reference. 1176 */ 1177 TAILQ_REMOVE(&V_dadq, dp, dad_list); 1178 dp->dad_ondadq = false; 1179 nd6_dad_rele(dp); 1180 } 1181 } 1182 1183 static struct dadq * 1184 nd6_dad_find(struct ifaddr *ifa, struct nd_opt_nonce *n) 1185 { 1186 struct dadq *dp; 1187 1188 DADQ_LOCK_ASSERT(); 1189 1190 TAILQ_FOREACH(dp, &V_dadq, dad_list) { 1191 if (dp->dad_ifa != ifa) 1192 continue; 1193 1194 /* 1195 * Skip if the nonce matches the received one. 1196 * +2 in the length is required because of type and 1197 * length fields are included in a header. 1198 */ 1199 if (n != NULL && 1200 n->nd_opt_nonce_len == (ND_OPT_NONCE_LEN + 2) / 8 && 1201 memcmp(&n->nd_opt_nonce[0], &dp->dad_nonce[0], 1202 ND_OPT_NONCE_LEN) == 0) { 1203 dp->dad_ns_lcount++; 1204 continue; 1205 } 1206 break; 1207 } 1208 1209 return (dp); 1210 } 1211 1212 static void 1213 nd6_dad_starttimer(struct dadq *dp, int ticks) 1214 { 1215 DADQ_WLOCK_ASSERT(); 1216 1217 callout_reset(&dp->dad_timer_ch, ticks, nd6_dad_timer, dp); 1218 } 1219 1220 static void 1221 nd6_dad_stoptimer(struct dadq *dp) 1222 { 1223 callout_drain(&dp->dad_timer_ch); 1224 } 1225 1226 static void 1227 nd6_dad_rele(struct dadq *dp) 1228 { 1229 if (refcount_release(&dp->dad_refcnt)) { 1230 KASSERT(!dp->dad_ondadq, ("dp %p still on DAD queue", dp)); 1231 ifa_free(dp->dad_ifa); 1232 free(dp, M_IP6NDP); 1233 } 1234 } 1235 1236 void 1237 nd6_dad_init(void) 1238 { 1239 DADQ_LOCK_INIT(); 1240 TAILQ_INIT(&V_dadq); 1241 } 1242 1243 /* 1244 * Start Duplicate Address Detection (DAD) for specified interface address. 1245 */ 1246 void 1247 nd6_dad_start(struct ifaddr *ifa, int delay) 1248 { 1249 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1250 struct dadq *dp; 1251 char ip6buf[INET6_ADDRSTRLEN]; 1252 1253 KASSERT((ia->ia6_flags & IN6_IFF_TENTATIVE) != 0, 1254 ("starting DAD on non-tentative address %p", ifa)); 1255 1256 /* 1257 * If we don't need DAD, don't do it. 1258 * There are several cases: 1259 * - DAD is disabled globally or on the interface 1260 * - the interface address is anycast 1261 */ 1262 if ((ia->ia6_flags & IN6_IFF_ANYCAST) != 0 || 1263 V_ip6_dad_count == 0 || 1264 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_NO_DAD) != 0) { 1265 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1266 return; 1267 } 1268 if ((ifa->ifa_ifp->if_flags & IFF_UP) == 0 || 1269 (ifa->ifa_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || 1270 (ND_IFINFO(ifa->ifa_ifp)->flags & ND6_IFF_IFDISABLED) != 0) 1271 return; 1272 1273 DADQ_WLOCK(); 1274 if ((dp = nd6_dad_find(ifa, NULL)) != NULL) { 1275 /* 1276 * DAD is already in progress. Let the existing entry 1277 * finish it. 1278 */ 1279 DADQ_WUNLOCK(); 1280 return; 1281 } 1282 1283 dp = malloc(sizeof(*dp), M_IP6NDP, M_NOWAIT | M_ZERO); 1284 if (dp == NULL) { 1285 log(LOG_ERR, "nd6_dad_start: memory allocation failed for " 1286 "%s(%s)\n", 1287 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1288 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1289 return; 1290 } 1291 callout_init_rw(&dp->dad_timer_ch, DADQ_LOCKPTR(), 1292 CALLOUT_RETURNUNLOCKED); 1293 #ifdef VIMAGE 1294 dp->dad_vnet = curvnet; 1295 #endif 1296 nd6log((LOG_DEBUG, "%s: starting DAD for %s\n", if_name(ifa->ifa_ifp), 1297 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1298 1299 /* 1300 * Send NS packet for DAD, ip6_dad_count times. 1301 * Note that we must delay the first transmission, if this is the 1302 * first packet to be sent from the interface after interface 1303 * (re)initialization. 1304 */ 1305 dp->dad_ifa = ifa; 1306 ifa_ref(dp->dad_ifa); 1307 dp->dad_count = V_ip6_dad_count; 1308 dp->dad_ns_icount = dp->dad_na_icount = 0; 1309 dp->dad_ns_ocount = dp->dad_ns_tcount = 0; 1310 dp->dad_ns_lcount = dp->dad_loopbackprobe = 0; 1311 1312 /* Add this to the dadq and add a reference for the dadq. */ 1313 refcount_init(&dp->dad_refcnt, 1); 1314 nd6_dad_add(dp); 1315 nd6_dad_starttimer(dp, delay); 1316 DADQ_WUNLOCK(); 1317 } 1318 1319 /* 1320 * terminate DAD unconditionally. used for address removals. 1321 */ 1322 void 1323 nd6_dad_stop(struct ifaddr *ifa) 1324 { 1325 struct dadq *dp; 1326 1327 DADQ_WLOCK(); 1328 dp = nd6_dad_find(ifa, NULL); 1329 if (dp == NULL) { 1330 DADQ_WUNLOCK(); 1331 /* DAD wasn't started yet */ 1332 return; 1333 } 1334 1335 /* 1336 * Acquire a temporary reference so that we can safely stop the callout. 1337 */ 1338 (void)refcount_acquire(&dp->dad_refcnt); 1339 nd6_dad_del(dp); 1340 DADQ_WUNLOCK(); 1341 1342 nd6_dad_stoptimer(dp); 1343 nd6_dad_rele(dp); 1344 } 1345 1346 static void 1347 nd6_dad_timer(void *arg) 1348 { 1349 struct dadq *dp = arg; 1350 struct ifaddr *ifa = dp->dad_ifa; 1351 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1352 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1353 char ip6buf[INET6_ADDRSTRLEN]; 1354 struct epoch_tracker et; 1355 1356 CURVNET_SET(dp->dad_vnet); 1357 KASSERT(ia != NULL, ("DAD entry %p with no address", dp)); 1358 1359 NET_EPOCH_ENTER(et); 1360 if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) { 1361 /* Do not need DAD for ifdisabled interface. */ 1362 log(LOG_ERR, "nd6_dad_timer: cancel DAD on %s because of " 1363 "ND6_IFF_IFDISABLED.\n", ifp->if_xname); 1364 goto err; 1365 } 1366 if (ia->ia6_flags & IN6_IFF_DUPLICATED) { 1367 log(LOG_ERR, "nd6_dad_timer: called with duplicated address " 1368 "%s(%s)\n", 1369 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1370 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1371 goto err; 1372 } 1373 if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0) { 1374 log(LOG_ERR, "nd6_dad_timer: called with non-tentative address " 1375 "%s(%s)\n", 1376 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1377 ifa->ifa_ifp ? if_name(ifa->ifa_ifp) : "???"); 1378 goto err; 1379 } 1380 1381 /* Stop DAD if the interface is down even after dad_maxtry attempts. */ 1382 if ((dp->dad_ns_tcount > V_dad_maxtry) && 1383 (((ifp->if_flags & IFF_UP) == 0) || 1384 ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0))) { 1385 nd6log((LOG_INFO, "%s: could not run DAD " 1386 "because the interface was down or not running.\n", 1387 if_name(ifa->ifa_ifp))); 1388 goto err; 1389 } 1390 1391 /* Need more checks? */ 1392 if (dp->dad_ns_ocount < dp->dad_count) { 1393 /* 1394 * We have more NS to go. Send NS packet for DAD. 1395 */ 1396 nd6_dad_starttimer(dp, 1397 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000); 1398 nd6_dad_ns_output(dp); 1399 goto done; 1400 } else { 1401 /* 1402 * We have transmitted sufficient number of DAD packets. 1403 * See what we've got. 1404 */ 1405 if (dp->dad_ns_icount > 0 || dp->dad_na_icount > 0) { 1406 /* We've seen NS or NA, means DAD has failed. */ 1407 nd6_dad_duplicated(ifa, dp); 1408 } else if (V_dad_enhanced != 0 && 1409 dp->dad_ns_lcount > 0 && 1410 dp->dad_ns_lcount > dp->dad_loopbackprobe) { 1411 /* 1412 * Sec. 4.1 in RFC 7527 requires transmission of 1413 * additional probes until the loopback condition 1414 * becomes clear when a looped back probe is detected. 1415 */ 1416 log(LOG_ERR, "%s: a looped back NS message is " 1417 "detected during DAD for %s. " 1418 "Another DAD probes are being sent.\n", 1419 if_name(ifa->ifa_ifp), 1420 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1421 dp->dad_loopbackprobe = dp->dad_ns_lcount; 1422 /* 1423 * Send an NS immediately and increase dad_count by 1424 * V_nd6_mmaxtries - 1. 1425 */ 1426 dp->dad_count = 1427 dp->dad_ns_ocount + V_nd6_mmaxtries - 1; 1428 nd6_dad_starttimer(dp, 1429 (long)ND_IFINFO(ifa->ifa_ifp)->retrans * hz / 1000); 1430 nd6_dad_ns_output(dp); 1431 goto done; 1432 } else { 1433 /* 1434 * We are done with DAD. No NA came, no NS came. 1435 * No duplicate address found. Check IFDISABLED flag 1436 * again in case that it is changed between the 1437 * beginning of this function and here. 1438 */ 1439 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) == 0) 1440 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1441 1442 nd6log((LOG_DEBUG, 1443 "%s: DAD complete for %s - no duplicates found\n", 1444 if_name(ifa->ifa_ifp), 1445 ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr))); 1446 if (dp->dad_ns_lcount > 0) 1447 log(LOG_ERR, "%s: DAD completed while " 1448 "a looped back NS message is detected " 1449 "during DAD for %s.\n", 1450 if_name(ifa->ifa_ifp), 1451 ip6_sprintf(ip6buf, IFA_IN6(ifa))); 1452 } 1453 } 1454 err: 1455 nd6_dad_del(dp); 1456 DADQ_WUNLOCK(); 1457 done: 1458 NET_EPOCH_EXIT(et); 1459 CURVNET_RESTORE(); 1460 } 1461 1462 static void 1463 nd6_dad_duplicated(struct ifaddr *ifa, struct dadq *dp) 1464 { 1465 struct in6_ifaddr *ia = (struct in6_ifaddr *)ifa; 1466 struct ifnet *ifp; 1467 char ip6buf[INET6_ADDRSTRLEN]; 1468 1469 log(LOG_ERR, "%s: DAD detected duplicate IPv6 address %s: " 1470 "NS in/out/loopback=%d/%d/%d, NA in=%d\n", 1471 if_name(ifa->ifa_ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), 1472 dp->dad_ns_icount, dp->dad_ns_ocount, dp->dad_ns_lcount, 1473 dp->dad_na_icount); 1474 1475 ia->ia6_flags &= ~IN6_IFF_TENTATIVE; 1476 ia->ia6_flags |= IN6_IFF_DUPLICATED; 1477 1478 ifp = ifa->ifa_ifp; 1479 log(LOG_ERR, "%s: DAD complete for %s - duplicate found\n", 1480 if_name(ifp), ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)); 1481 log(LOG_ERR, "%s: manual intervention required\n", 1482 if_name(ifp)); 1483 1484 /* 1485 * If the address is a link-local address formed from an interface 1486 * identifier based on the hardware address which is supposed to be 1487 * uniquely assigned (e.g., EUI-64 for an Ethernet interface), IP 1488 * operation on the interface SHOULD be disabled. 1489 * [RFC 4862, Section 5.4.5] 1490 */ 1491 if (IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) { 1492 struct in6_addr in6; 1493 1494 /* 1495 * To avoid over-reaction, we only apply this logic when we are 1496 * very sure that hardware addresses are supposed to be unique. 1497 */ 1498 switch (ifp->if_type) { 1499 case IFT_ETHER: 1500 case IFT_ATM: 1501 case IFT_IEEE1394: 1502 case IFT_INFINIBAND: 1503 in6 = ia->ia_addr.sin6_addr; 1504 if (in6_get_hw_ifid(ifp, &in6) == 0 && 1505 IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr, &in6)) { 1506 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1507 log(LOG_ERR, "%s: possible hardware address " 1508 "duplication detected, disable IPv6\n", 1509 if_name(ifp)); 1510 } 1511 break; 1512 } 1513 } 1514 } 1515 1516 /* 1517 * Transmit a neighbour solicitation for the purpose of DAD. Returns with the 1518 * DAD queue unlocked. 1519 */ 1520 static void 1521 nd6_dad_ns_output(struct dadq *dp) 1522 { 1523 struct in6_ifaddr *ia = (struct in6_ifaddr *)dp->dad_ifa; 1524 struct ifnet *ifp = dp->dad_ifa->ifa_ifp; 1525 int i; 1526 1527 DADQ_WLOCK_ASSERT(); 1528 1529 dp->dad_ns_tcount++; 1530 if ((ifp->if_flags & IFF_UP) == 0) { 1531 DADQ_WUNLOCK(); 1532 return; 1533 } 1534 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) { 1535 DADQ_WUNLOCK(); 1536 return; 1537 } 1538 1539 dp->dad_ns_ocount++; 1540 if (V_dad_enhanced != 0) { 1541 for (i = 0; i < ND_OPT_NONCE_LEN32; i++) 1542 dp->dad_nonce[i] = arc4random(); 1543 /* 1544 * XXXHRS: Note that in the case that 1545 * DupAddrDetectTransmits > 1, multiple NS messages with 1546 * different nonces can be looped back in an unexpected 1547 * order. The current implementation recognizes only 1548 * the latest nonce on the sender side. Practically it 1549 * should work well in almost all cases. 1550 */ 1551 } 1552 DADQ_WUNLOCK(); 1553 nd6_ns_output(ifp, NULL, NULL, &ia->ia_addr.sin6_addr, 1554 (uint8_t *)&dp->dad_nonce[0]); 1555 } 1556 1557 static void 1558 nd6_dad_ns_input(struct ifaddr *ifa, struct nd_opt_nonce *ndopt_nonce) 1559 { 1560 struct dadq *dp; 1561 1562 if (ifa == NULL) 1563 panic("ifa == NULL in nd6_dad_ns_input"); 1564 1565 /* Ignore Nonce option when Enhanced DAD is disabled. */ 1566 if (V_dad_enhanced == 0) 1567 ndopt_nonce = NULL; 1568 DADQ_RLOCK(); 1569 dp = nd6_dad_find(ifa, ndopt_nonce); 1570 if (dp != NULL) 1571 dp->dad_ns_icount++; 1572 DADQ_RUNLOCK(); 1573 } 1574 1575 static void 1576 nd6_dad_na_input(struct ifaddr *ifa) 1577 { 1578 struct dadq *dp; 1579 1580 if (ifa == NULL) 1581 panic("ifa == NULL in nd6_dad_na_input"); 1582 1583 DADQ_RLOCK(); 1584 dp = nd6_dad_find(ifa, NULL); 1585 if (dp != NULL) 1586 dp->dad_na_icount++; 1587 DADQ_RUNLOCK(); 1588 } 1589