1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * Neighbour Discovery for IPv6 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * Mike Shaver <shaver@ingenia.com> 9 */ 10 11 /* 12 * Changes: 13 * 14 * Alexey I. Froloff : RFC6106 (DNSSL) support 15 * Pierre Ynard : export userland ND options 16 * through netlink (RDNSS support) 17 * Lars Fenneberg : fixed MTU setting on receipt 18 * of an RA. 19 * Janos Farkas : kmalloc failure checks 20 * Alexey Kuznetsov : state machine reworked 21 * and moved to net/core. 22 * Pekka Savola : RFC2461 validation 23 * YOSHIFUJI Hideaki @USAGI : Verify ND options properly 24 */ 25 26 #define pr_fmt(fmt) "ICMPv6: " fmt 27 28 #include <linux/module.h> 29 #include <linux/errno.h> 30 #include <linux/types.h> 31 #include <linux/socket.h> 32 #include <linux/sockios.h> 33 #include <linux/sched.h> 34 #include <linux/net.h> 35 #include <linux/in6.h> 36 #include <linux/route.h> 37 #include <linux/init.h> 38 #include <linux/rcupdate.h> 39 #include <linux/slab.h> 40 #ifdef CONFIG_SYSCTL 41 #include <linux/sysctl.h> 42 #endif 43 44 #include <linux/if_addr.h> 45 #include <linux/if_ether.h> 46 #include <linux/if_arp.h> 47 #include <linux/ipv6.h> 48 #include <linux/icmpv6.h> 49 #include <linux/jhash.h> 50 51 #include <net/sock.h> 52 #include <net/snmp.h> 53 54 #include <net/ipv6.h> 55 #include <net/protocol.h> 56 #include <net/ndisc.h> 57 #include <net/ip6_route.h> 58 #include <net/addrconf.h> 59 #include <net/icmp.h> 60 61 #include <net/netlink.h> 62 #include <linux/rtnetlink.h> 63 64 #include <net/flow.h> 65 #include <net/ip6_checksum.h> 66 #include <net/inet_common.h> 67 #include <linux/proc_fs.h> 68 69 #include <linux/netfilter.h> 70 #include <linux/netfilter_ipv6.h> 71 72 static u32 ndisc_hash(const void *pkey, 73 const struct net_device *dev, 74 __u32 *hash_rnd); 75 static bool ndisc_key_eq(const struct neighbour *neigh, const void *pkey); 76 static bool ndisc_allow_add(const struct net_device *dev, 77 struct netlink_ext_ack *extack); 78 static int ndisc_constructor(struct neighbour *neigh); 79 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb); 80 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb); 81 static int pndisc_constructor(struct pneigh_entry *n); 82 static void pndisc_destructor(struct pneigh_entry *n); 83 static void pndisc_redo(struct sk_buff *skb); 84 static int ndisc_is_multicast(const void *pkey); 85 86 static const struct neigh_ops ndisc_generic_ops = { 87 .family = AF_INET6, 88 .solicit = ndisc_solicit, 89 .error_report = ndisc_error_report, 90 .output = neigh_resolve_output, 91 .connected_output = neigh_connected_output, 92 }; 93 94 static const struct neigh_ops ndisc_hh_ops = { 95 .family = AF_INET6, 96 .solicit = ndisc_solicit, 97 .error_report = ndisc_error_report, 98 .output = neigh_resolve_output, 99 .connected_output = neigh_resolve_output, 100 }; 101 102 103 static const struct neigh_ops ndisc_direct_ops = { 104 .family = AF_INET6, 105 .output = neigh_direct_output, 106 .connected_output = neigh_direct_output, 107 }; 108 109 struct neigh_table nd_tbl = { 110 .family = AF_INET6, 111 .key_len = sizeof(struct in6_addr), 112 .protocol = cpu_to_be16(ETH_P_IPV6), 113 .hash = ndisc_hash, 114 .key_eq = ndisc_key_eq, 115 .constructor = ndisc_constructor, 116 .pconstructor = pndisc_constructor, 117 .pdestructor = pndisc_destructor, 118 .proxy_redo = pndisc_redo, 119 .is_multicast = ndisc_is_multicast, 120 .allow_add = ndisc_allow_add, 121 .id = "ndisc_cache", 122 .parms = { 123 .tbl = &nd_tbl, 124 .reachable_time = ND_REACHABLE_TIME, 125 .data = { 126 [NEIGH_VAR_MCAST_PROBES] = 3, 127 [NEIGH_VAR_UCAST_PROBES] = 3, 128 [NEIGH_VAR_RETRANS_TIME] = ND_RETRANS_TIMER, 129 [NEIGH_VAR_BASE_REACHABLE_TIME] = ND_REACHABLE_TIME, 130 [NEIGH_VAR_DELAY_PROBE_TIME] = 5 * HZ, 131 [NEIGH_VAR_INTERVAL_PROBE_TIME_MS] = 5 * HZ, 132 [NEIGH_VAR_GC_STALETIME] = 60 * HZ, 133 [NEIGH_VAR_QUEUE_LEN_BYTES] = SK_WMEM_DEFAULT, 134 [NEIGH_VAR_PROXY_QLEN] = 64, 135 [NEIGH_VAR_ANYCAST_DELAY] = 1 * HZ, 136 [NEIGH_VAR_PROXY_DELAY] = (8 * HZ) / 10, 137 }, 138 }, 139 .gc_interval = 30 * HZ, 140 .gc_thresh1 = 128, 141 .gc_thresh2 = 512, 142 .gc_thresh3 = 1024, 143 }; 144 EXPORT_SYMBOL_GPL(nd_tbl); 145 146 void __ndisc_fill_addr_option(struct sk_buff *skb, int type, const void *data, 147 int data_len, int pad) 148 { 149 int space = __ndisc_opt_addr_space(data_len, pad); 150 u8 *opt = skb_put(skb, space); 151 152 opt[0] = type; 153 opt[1] = space>>3; 154 155 memset(opt + 2, 0, pad); 156 opt += pad; 157 space -= pad; 158 159 memcpy(opt+2, data, data_len); 160 data_len += 2; 161 opt += data_len; 162 space -= data_len; 163 if (space > 0) 164 memset(opt, 0, space); 165 } 166 EXPORT_SYMBOL_GPL(__ndisc_fill_addr_option); 167 168 static inline void ndisc_fill_addr_option(struct sk_buff *skb, int type, 169 const void *data, u8 icmp6_type) 170 { 171 __ndisc_fill_addr_option(skb, type, data, skb->dev->addr_len, 172 ndisc_addr_option_pad(skb->dev->type)); 173 ndisc_ops_fill_addr_option(skb->dev, skb, icmp6_type); 174 } 175 176 static inline void ndisc_fill_redirect_addr_option(struct sk_buff *skb, 177 void *ha, 178 const u8 *ops_data) 179 { 180 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, ha, NDISC_REDIRECT); 181 ndisc_ops_fill_redirect_addr_option(skb->dev, skb, ops_data); 182 } 183 184 static struct nd_opt_hdr *ndisc_next_option(struct nd_opt_hdr *cur, 185 struct nd_opt_hdr *end) 186 { 187 int type; 188 if (!cur || !end || cur >= end) 189 return NULL; 190 type = cur->nd_opt_type; 191 do { 192 cur = ((void *)cur) + (cur->nd_opt_len << 3); 193 } while (cur < end && cur->nd_opt_type != type); 194 return cur <= end && cur->nd_opt_type == type ? cur : NULL; 195 } 196 197 static inline int ndisc_is_useropt(const struct net_device *dev, 198 struct nd_opt_hdr *opt) 199 { 200 return opt->nd_opt_type == ND_OPT_PREFIX_INFO || 201 opt->nd_opt_type == ND_OPT_RDNSS || 202 opt->nd_opt_type == ND_OPT_DNSSL || 203 opt->nd_opt_type == ND_OPT_6CO || 204 opt->nd_opt_type == ND_OPT_CAPTIVE_PORTAL || 205 opt->nd_opt_type == ND_OPT_PREF64; 206 } 207 208 static struct nd_opt_hdr *ndisc_next_useropt(const struct net_device *dev, 209 struct nd_opt_hdr *cur, 210 struct nd_opt_hdr *end) 211 { 212 if (!cur || !end || cur >= end) 213 return NULL; 214 do { 215 cur = ((void *)cur) + (cur->nd_opt_len << 3); 216 } while (cur < end && !ndisc_is_useropt(dev, cur)); 217 return cur <= end && ndisc_is_useropt(dev, cur) ? cur : NULL; 218 } 219 220 struct ndisc_options *ndisc_parse_options(const struct net_device *dev, 221 u8 *opt, int opt_len, 222 struct ndisc_options *ndopts) 223 { 224 struct nd_opt_hdr *nd_opt = (struct nd_opt_hdr *)opt; 225 226 if (!nd_opt || opt_len < 0 || !ndopts) 227 return NULL; 228 memset(ndopts, 0, sizeof(*ndopts)); 229 while (opt_len) { 230 bool unknown = false; 231 int l; 232 if (opt_len < sizeof(struct nd_opt_hdr)) 233 return NULL; 234 l = nd_opt->nd_opt_len << 3; 235 if (opt_len < l || l == 0) 236 return NULL; 237 if (ndisc_ops_parse_options(dev, nd_opt, ndopts)) 238 goto next_opt; 239 switch (nd_opt->nd_opt_type) { 240 case ND_OPT_SOURCE_LL_ADDR: 241 case ND_OPT_TARGET_LL_ADDR: 242 case ND_OPT_MTU: 243 case ND_OPT_NONCE: 244 case ND_OPT_REDIRECT_HDR: 245 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 246 net_dbg_ratelimited("%s: duplicated ND6 option found: type=%d\n", 247 __func__, nd_opt->nd_opt_type); 248 } else { 249 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt; 250 } 251 break; 252 case ND_OPT_PREFIX_INFO: 253 ndopts->nd_opts_pi_end = nd_opt; 254 if (!ndopts->nd_opt_array[nd_opt->nd_opt_type]) 255 ndopts->nd_opt_array[nd_opt->nd_opt_type] = nd_opt; 256 break; 257 #ifdef CONFIG_IPV6_ROUTE_INFO 258 case ND_OPT_ROUTE_INFO: 259 ndopts->nd_opts_ri_end = nd_opt; 260 if (!ndopts->nd_opts_ri) 261 ndopts->nd_opts_ri = nd_opt; 262 break; 263 #endif 264 default: 265 unknown = true; 266 } 267 if (ndisc_is_useropt(dev, nd_opt)) { 268 ndopts->nd_useropts_end = nd_opt; 269 if (!ndopts->nd_useropts) 270 ndopts->nd_useropts = nd_opt; 271 } else if (unknown) { 272 /* 273 * Unknown options must be silently ignored, 274 * to accommodate future extension to the 275 * protocol. 276 */ 277 net_dbg_ratelimited("%s: ignored unsupported option; type=%d, len=%d\n", 278 __func__, nd_opt->nd_opt_type, nd_opt->nd_opt_len); 279 } 280 next_opt: 281 opt_len -= l; 282 nd_opt = ((void *)nd_opt) + l; 283 } 284 return ndopts; 285 } 286 EXPORT_SYMBOL_GPL(ndisc_parse_options); 287 288 int ndisc_mc_map(const struct in6_addr *addr, char *buf, struct net_device *dev, int dir) 289 { 290 switch (dev->type) { 291 case ARPHRD_ETHER: 292 case ARPHRD_IEEE802: /* Not sure. Check it later. --ANK */ 293 case ARPHRD_FDDI: 294 ipv6_eth_mc_map(addr, buf); 295 return 0; 296 case ARPHRD_ARCNET: 297 ipv6_arcnet_mc_map(addr, buf); 298 return 0; 299 case ARPHRD_INFINIBAND: 300 ipv6_ib_mc_map(addr, dev->broadcast, buf); 301 return 0; 302 case ARPHRD_IPGRE: 303 return ipv6_ipgre_mc_map(addr, dev->broadcast, buf); 304 default: 305 if (dir) { 306 memcpy(buf, dev->broadcast, dev->addr_len); 307 return 0; 308 } 309 } 310 return -EINVAL; 311 } 312 EXPORT_SYMBOL(ndisc_mc_map); 313 314 static u32 ndisc_hash(const void *pkey, 315 const struct net_device *dev, 316 __u32 *hash_rnd) 317 { 318 return ndisc_hashfn(pkey, dev, hash_rnd); 319 } 320 321 static bool ndisc_key_eq(const struct neighbour *n, const void *pkey) 322 { 323 return neigh_key_eq128(n, pkey); 324 } 325 326 static int ndisc_constructor(struct neighbour *neigh) 327 { 328 struct in6_addr *addr = (struct in6_addr *)&neigh->primary_key; 329 struct net_device *dev = neigh->dev; 330 struct inet6_dev *in6_dev; 331 struct neigh_parms *parms; 332 bool is_multicast = ipv6_addr_is_multicast(addr); 333 334 in6_dev = in6_dev_get(dev); 335 if (!in6_dev) { 336 return -EINVAL; 337 } 338 339 parms = in6_dev->nd_parms; 340 __neigh_parms_put(neigh->parms); 341 neigh->parms = neigh_parms_clone(parms); 342 343 neigh->type = is_multicast ? RTN_MULTICAST : RTN_UNICAST; 344 if (!dev->header_ops) { 345 neigh->nud_state = NUD_NOARP; 346 neigh->ops = &ndisc_direct_ops; 347 neigh->output = neigh_direct_output; 348 } else { 349 if (is_multicast) { 350 neigh->nud_state = NUD_NOARP; 351 ndisc_mc_map(addr, neigh->ha, dev, 1); 352 } else if (dev->flags&(IFF_NOARP|IFF_LOOPBACK)) { 353 neigh->nud_state = NUD_NOARP; 354 memcpy(neigh->ha, dev->dev_addr, dev->addr_len); 355 if (dev->flags&IFF_LOOPBACK) 356 neigh->type = RTN_LOCAL; 357 } else if (dev->flags&IFF_POINTOPOINT) { 358 neigh->nud_state = NUD_NOARP; 359 memcpy(neigh->ha, dev->broadcast, dev->addr_len); 360 } 361 if (dev->header_ops->cache) 362 neigh->ops = &ndisc_hh_ops; 363 else 364 neigh->ops = &ndisc_generic_ops; 365 if (neigh->nud_state&NUD_VALID) 366 neigh->output = neigh->ops->connected_output; 367 else 368 neigh->output = neigh->ops->output; 369 } 370 in6_dev_put(in6_dev); 371 return 0; 372 } 373 374 static int pndisc_constructor(struct pneigh_entry *n) 375 { 376 struct in6_addr *addr = (struct in6_addr *)&n->key; 377 struct net_device *dev = n->dev; 378 struct in6_addr maddr; 379 380 if (!dev) 381 return -EINVAL; 382 383 addrconf_addr_solict_mult(addr, &maddr); 384 return ipv6_dev_mc_inc(dev, &maddr); 385 } 386 387 static void pndisc_destructor(struct pneigh_entry *n) 388 { 389 struct in6_addr *addr = (struct in6_addr *)&n->key; 390 struct net_device *dev = n->dev; 391 struct in6_addr maddr; 392 393 if (!dev) 394 return; 395 396 addrconf_addr_solict_mult(addr, &maddr); 397 ipv6_dev_mc_dec(dev, &maddr); 398 } 399 400 /* called with rtnl held */ 401 static bool ndisc_allow_add(const struct net_device *dev, 402 struct netlink_ext_ack *extack) 403 { 404 struct inet6_dev *idev = __in6_dev_get(dev); 405 406 if (!idev || idev->cnf.disable_ipv6) { 407 NL_SET_ERR_MSG(extack, "IPv6 is disabled on this device"); 408 return false; 409 } 410 411 return true; 412 } 413 414 static struct sk_buff *ndisc_alloc_skb(struct net_device *dev, 415 int len) 416 { 417 int hlen = LL_RESERVED_SPACE(dev); 418 int tlen = dev->needed_tailroom; 419 struct sk_buff *skb; 420 421 skb = alloc_skb(hlen + sizeof(struct ipv6hdr) + len + tlen, GFP_ATOMIC); 422 if (!skb) 423 return NULL; 424 425 skb->protocol = htons(ETH_P_IPV6); 426 skb->dev = dev; 427 428 skb_reserve(skb, hlen + sizeof(struct ipv6hdr)); 429 skb_reset_transport_header(skb); 430 431 /* Manually assign socket ownership as we avoid calling 432 * sock_alloc_send_pskb() to bypass wmem buffer limits 433 */ 434 rcu_read_lock(); 435 skb_set_owner_w(skb, dev_net_rcu(dev)->ipv6.ndisc_sk); 436 rcu_read_unlock(); 437 438 return skb; 439 } 440 441 static void ip6_nd_hdr(struct sk_buff *skb, 442 const struct in6_addr *saddr, 443 const struct in6_addr *daddr, 444 int hop_limit, int len) 445 { 446 struct ipv6hdr *hdr; 447 struct inet6_dev *idev; 448 unsigned tclass; 449 450 rcu_read_lock(); 451 idev = __in6_dev_get(skb->dev); 452 tclass = idev ? READ_ONCE(idev->cnf.ndisc_tclass) : 0; 453 rcu_read_unlock(); 454 455 skb_push(skb, sizeof(*hdr)); 456 skb_reset_network_header(skb); 457 hdr = ipv6_hdr(skb); 458 459 ip6_flow_hdr(hdr, tclass, 0); 460 461 hdr->payload_len = htons(len); 462 hdr->nexthdr = IPPROTO_ICMPV6; 463 hdr->hop_limit = hop_limit; 464 465 hdr->saddr = *saddr; 466 hdr->daddr = *daddr; 467 } 468 469 void ndisc_send_skb(struct sk_buff *skb, const struct in6_addr *daddr, 470 const struct in6_addr *saddr) 471 { 472 struct icmp6hdr *icmp6h = icmp6_hdr(skb); 473 struct dst_entry *dst = skb_dst(skb); 474 struct net_device *dev; 475 struct inet6_dev *idev; 476 struct net *net; 477 struct sock *sk; 478 int err; 479 u8 type; 480 481 type = icmp6h->icmp6_type; 482 483 rcu_read_lock(); 484 485 net = dev_net_rcu(skb->dev); 486 sk = net->ipv6.ndisc_sk; 487 if (!dst) { 488 struct flowi6 fl6; 489 int oif = skb->dev->ifindex; 490 491 icmpv6_flow_init(sk, &fl6, type, saddr, daddr, oif); 492 dst = icmp6_dst_alloc(skb->dev, &fl6); 493 if (IS_ERR(dst)) { 494 rcu_read_unlock(); 495 kfree_skb(skb); 496 return; 497 } 498 499 skb_dst_set(skb, dst); 500 } 501 502 icmp6h->icmp6_cksum = csum_ipv6_magic(saddr, daddr, skb->len, 503 IPPROTO_ICMPV6, 504 csum_partial(icmp6h, 505 skb->len, 0)); 506 507 ip6_nd_hdr(skb, saddr, daddr, READ_ONCE(inet6_sk(sk)->hop_limit), skb->len); 508 509 dev = dst_dev_rcu(dst); 510 idev = __in6_dev_get(dev); 511 IP6_INC_STATS(net, idev, IPSTATS_MIB_OUTREQUESTS); 512 513 err = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_OUT, 514 net, sk, skb, NULL, dev, 515 dst_output); 516 if (!err) { 517 ICMP6MSGOUT_INC_STATS(net, idev, type); 518 ICMP6_INC_STATS(net, idev, ICMP6_MIB_OUTMSGS); 519 } 520 521 rcu_read_unlock(); 522 } 523 EXPORT_SYMBOL(ndisc_send_skb); 524 525 void ndisc_send_na(struct net_device *dev, const struct in6_addr *daddr, 526 const struct in6_addr *solicited_addr, 527 bool router, bool solicited, bool override, bool inc_opt) 528 { 529 struct sk_buff *skb; 530 struct in6_addr tmpaddr; 531 struct inet6_ifaddr *ifp; 532 const struct in6_addr *src_addr; 533 struct nd_msg *msg; 534 int optlen = 0; 535 536 /* for anycast or proxy, solicited_addr != src_addr */ 537 ifp = ipv6_get_ifaddr(dev_net(dev), solicited_addr, dev, 1); 538 if (ifp) { 539 src_addr = solicited_addr; 540 if (ifp->flags & IFA_F_OPTIMISTIC) 541 override = false; 542 inc_opt |= READ_ONCE(ifp->idev->cnf.force_tllao); 543 in6_ifa_put(ifp); 544 } else { 545 if (ipv6_dev_get_saddr(dev_net(dev), dev, daddr, 546 inet6_sk(dev_net(dev)->ipv6.ndisc_sk)->srcprefs, 547 &tmpaddr)) 548 return; 549 src_addr = &tmpaddr; 550 } 551 552 if (!dev->addr_len) 553 inc_opt = false; 554 if (inc_opt) 555 optlen += ndisc_opt_addr_space(dev, 556 NDISC_NEIGHBOUR_ADVERTISEMENT); 557 558 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen); 559 if (!skb) 560 return; 561 562 msg = skb_put(skb, sizeof(*msg)); 563 *msg = (struct nd_msg) { 564 .icmph = { 565 .icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT, 566 .icmp6_router = router, 567 .icmp6_solicited = solicited, 568 .icmp6_override = override, 569 }, 570 .target = *solicited_addr, 571 }; 572 573 if (inc_opt) 574 ndisc_fill_addr_option(skb, ND_OPT_TARGET_LL_ADDR, 575 dev->dev_addr, 576 NDISC_NEIGHBOUR_ADVERTISEMENT); 577 578 ndisc_send_skb(skb, daddr, src_addr); 579 } 580 EXPORT_SYMBOL_GPL(ndisc_send_na); 581 582 static void ndisc_send_unsol_na(struct net_device *dev) 583 { 584 struct inet6_dev *idev; 585 struct inet6_ifaddr *ifa; 586 587 idev = in6_dev_get(dev); 588 if (!idev) 589 return; 590 591 read_lock_bh(&idev->lock); 592 list_for_each_entry(ifa, &idev->addr_list, if_list) { 593 /* skip tentative addresses until dad completes */ 594 if (ifa->flags & IFA_F_TENTATIVE && 595 !(ifa->flags & IFA_F_OPTIMISTIC)) 596 continue; 597 598 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifa->addr, 599 /*router=*/ !!idev->cnf.forwarding, 600 /*solicited=*/ false, /*override=*/ true, 601 /*inc_opt=*/ true); 602 } 603 read_unlock_bh(&idev->lock); 604 605 in6_dev_put(idev); 606 } 607 608 struct sk_buff *ndisc_ns_create(struct net_device *dev, const struct in6_addr *solicit, 609 const struct in6_addr *saddr, u64 nonce) 610 { 611 int inc_opt = dev->addr_len; 612 struct sk_buff *skb; 613 struct nd_msg *msg; 614 int optlen = 0; 615 616 if (!saddr) 617 return NULL; 618 619 if (ipv6_addr_any(saddr)) 620 inc_opt = false; 621 if (inc_opt) 622 optlen += ndisc_opt_addr_space(dev, 623 NDISC_NEIGHBOUR_SOLICITATION); 624 if (nonce != 0) 625 optlen += 8; 626 627 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen); 628 if (!skb) 629 return NULL; 630 631 msg = skb_put(skb, sizeof(*msg)); 632 *msg = (struct nd_msg) { 633 .icmph = { 634 .icmp6_type = NDISC_NEIGHBOUR_SOLICITATION, 635 }, 636 .target = *solicit, 637 }; 638 639 if (inc_opt) 640 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR, 641 dev->dev_addr, 642 NDISC_NEIGHBOUR_SOLICITATION); 643 if (nonce != 0) { 644 u8 *opt = skb_put(skb, 8); 645 646 opt[0] = ND_OPT_NONCE; 647 opt[1] = 8 >> 3; 648 memcpy(opt + 2, &nonce, 6); 649 } 650 651 return skb; 652 } 653 EXPORT_SYMBOL(ndisc_ns_create); 654 655 void ndisc_send_ns(struct net_device *dev, const struct in6_addr *solicit, 656 const struct in6_addr *daddr, const struct in6_addr *saddr, 657 u64 nonce) 658 { 659 struct in6_addr addr_buf; 660 struct sk_buff *skb; 661 662 if (!saddr) { 663 if (ipv6_get_lladdr(dev, &addr_buf, 664 (IFA_F_TENTATIVE | IFA_F_OPTIMISTIC))) 665 return; 666 saddr = &addr_buf; 667 } 668 669 skb = ndisc_ns_create(dev, solicit, saddr, nonce); 670 671 if (skb) 672 ndisc_send_skb(skb, daddr, saddr); 673 } 674 675 void ndisc_send_rs(struct net_device *dev, const struct in6_addr *saddr, 676 const struct in6_addr *daddr) 677 { 678 struct sk_buff *skb; 679 struct rs_msg *msg; 680 int send_sllao = dev->addr_len; 681 int optlen = 0; 682 683 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 684 /* 685 * According to section 2.2 of RFC 4429, we must not 686 * send router solicitations with a sllao from 687 * optimistic addresses, but we may send the solicitation 688 * if we don't include the sllao. So here we check 689 * if our address is optimistic, and if so, we 690 * suppress the inclusion of the sllao. 691 */ 692 if (send_sllao) { 693 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(dev_net(dev), saddr, 694 dev, 1); 695 if (ifp) { 696 if (ifp->flags & IFA_F_OPTIMISTIC) { 697 send_sllao = 0; 698 } 699 in6_ifa_put(ifp); 700 } else { 701 send_sllao = 0; 702 } 703 } 704 #endif 705 if (send_sllao) 706 optlen += ndisc_opt_addr_space(dev, NDISC_ROUTER_SOLICITATION); 707 708 skb = ndisc_alloc_skb(dev, sizeof(*msg) + optlen); 709 if (!skb) 710 return; 711 712 msg = skb_put(skb, sizeof(*msg)); 713 *msg = (struct rs_msg) { 714 .icmph = { 715 .icmp6_type = NDISC_ROUTER_SOLICITATION, 716 }, 717 }; 718 719 if (send_sllao) 720 ndisc_fill_addr_option(skb, ND_OPT_SOURCE_LL_ADDR, 721 dev->dev_addr, 722 NDISC_ROUTER_SOLICITATION); 723 724 ndisc_send_skb(skb, daddr, saddr); 725 } 726 727 728 static void ndisc_error_report(struct neighbour *neigh, struct sk_buff *skb) 729 { 730 /* 731 * "The sender MUST return an ICMP 732 * destination unreachable" 733 */ 734 dst_link_failure(skb); 735 kfree_skb(skb); 736 } 737 738 /* Called with locked neigh: either read or both */ 739 740 static void ndisc_solicit(struct neighbour *neigh, struct sk_buff *skb) 741 { 742 struct in6_addr *saddr = NULL; 743 struct in6_addr mcaddr; 744 struct net_device *dev = neigh->dev; 745 struct in6_addr *target = (struct in6_addr *)&neigh->primary_key; 746 int probes = atomic_read(&neigh->probes); 747 748 if (skb && ipv6_chk_addr_and_flags(dev_net(dev), &ipv6_hdr(skb)->saddr, 749 dev, false, 1, 750 IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) 751 saddr = &ipv6_hdr(skb)->saddr; 752 probes -= NEIGH_VAR(neigh->parms, UCAST_PROBES); 753 if (probes < 0) { 754 if (!(READ_ONCE(neigh->nud_state) & NUD_VALID)) { 755 net_dbg_ratelimited("%s: trying to ucast probe in NUD_INVALID: %pI6\n", 756 __func__, target); 757 } 758 ndisc_send_ns(dev, target, target, saddr, 0); 759 } else if ((probes -= NEIGH_VAR(neigh->parms, APP_PROBES)) < 0) { 760 neigh_app_ns(neigh); 761 } else { 762 addrconf_addr_solict_mult(target, &mcaddr); 763 ndisc_send_ns(dev, target, &mcaddr, saddr, 0); 764 } 765 } 766 767 static int pndisc_is_router(const void *pkey, 768 struct net_device *dev) 769 { 770 struct net *net = dev_net(dev); 771 struct pneigh_entry *n; 772 int ret = -1; 773 774 n = pneigh_lookup(nd_table(net), pkey, dev); 775 if (n) 776 ret = !!(READ_ONCE(n->flags) & NTF_ROUTER); 777 778 return ret; 779 } 780 781 void ndisc_update(const struct net_device *dev, struct neighbour *neigh, 782 const u8 *lladdr, u8 new, u32 flags, u8 icmp6_type, 783 struct ndisc_options *ndopts) 784 { 785 neigh_update(neigh, lladdr, new, flags, 0); 786 /* report ndisc ops about neighbour update */ 787 ndisc_ops_update(dev, neigh, flags, icmp6_type, ndopts); 788 } 789 790 static enum skb_drop_reason ndisc_recv_ns(struct sk_buff *skb) 791 { 792 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) + 793 offsetof(struct nd_msg, opt)); 794 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb); 795 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr; 796 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr; 797 struct net_device *dev = skb->dev; 798 struct net *net = dev_net(dev); 799 int dad = ipv6_addr_any(saddr); 800 struct inet6_dev *idev = NULL; 801 struct ndisc_options ndopts; 802 struct inet6_ifaddr *ifp; 803 struct neigh_table *tbl; 804 struct neighbour *neigh; 805 int is_router = -1; 806 u8 *lladdr = NULL; 807 SKB_DR(reason); 808 u64 nonce = 0; 809 bool inc; 810 811 if (skb->len < sizeof(struct nd_msg)) 812 return SKB_DROP_REASON_PKT_TOO_SMALL; 813 814 if (ipv6_addr_is_multicast(&msg->target)) { 815 net_dbg_ratelimited("NS: multicast target address\n"); 816 return reason; 817 } 818 819 /* 820 * RFC2461 7.1.1: 821 * DAD has to be destined for solicited node multicast address. 822 */ 823 if (dad && !ipv6_addr_is_solict_mult(daddr)) { 824 net_dbg_ratelimited("NS: bad DAD packet (wrong destination)\n"); 825 return reason; 826 } 827 828 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) 829 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS; 830 831 if (ndopts.nd_opts_src_lladdr) { 832 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev); 833 if (!lladdr) { 834 net_dbg_ratelimited("NS: invalid link-layer address length\n"); 835 return reason; 836 } 837 838 /* RFC2461 7.1.1: 839 * If the IP source address is the unspecified address, 840 * there MUST NOT be source link-layer address option 841 * in the message. 842 */ 843 if (dad) { 844 net_dbg_ratelimited("NS: bad DAD packet (link-layer address option)\n"); 845 return reason; 846 } 847 } 848 if (ndopts.nd_opts_nonce && ndopts.nd_opts_nonce->nd_opt_len == 1) 849 memcpy(&nonce, (u8 *)(ndopts.nd_opts_nonce + 1), 6); 850 851 tbl = nd_table(net); 852 853 inc = ipv6_addr_is_multicast(daddr); 854 ifp = ipv6_get_ifaddr(net, &msg->target, dev, 1); 855 if (ifp) { 856 have_ifp: 857 if (ifp->flags & (IFA_F_TENTATIVE|IFA_F_OPTIMISTIC)) { 858 if (dad) { 859 if (nonce != 0 && ifp->dad_nonce == nonce) { 860 u8 *np = (u8 *)&nonce; 861 /* Matching nonce if looped back */ 862 net_dbg_ratelimited("%s: IPv6 DAD loopback for address %pI6c nonce %pM ignored\n", 863 ifp->idev->dev->name, &ifp->addr, np); 864 goto out; 865 } 866 /* 867 * We are colliding with another node 868 * who is doing DAD 869 * so fail our DAD process 870 */ 871 addrconf_dad_failure(skb, ifp); 872 return reason; 873 } else { 874 /* 875 * This is not a dad solicitation. 876 * If we are an optimistic node, 877 * we should respond. 878 * Otherwise, we should ignore it. 879 */ 880 if (!(ifp->flags & IFA_F_OPTIMISTIC)) 881 goto out; 882 } 883 } 884 885 idev = ifp->idev; 886 } else { 887 /* perhaps an address on the master device */ 888 if (netif_is_l3_slave(dev)) { 889 struct net_device *mdev; 890 891 mdev = netdev_master_upper_dev_get_rcu(dev); 892 if (mdev) { 893 ifp = ipv6_get_ifaddr(net, &msg->target, mdev, 1); 894 if (ifp) 895 goto have_ifp; 896 } 897 } 898 899 idev = in6_dev_get(dev); 900 if (!idev) { 901 /* XXX: count this drop? */ 902 return reason; 903 } 904 905 if (ipv6_chk_acast_addr(net, dev, &msg->target) || 906 (READ_ONCE(idev->cnf.forwarding) && 907 (READ_ONCE(net->ipv6.devconf_all->proxy_ndp) || 908 READ_ONCE(idev->cnf.proxy_ndp)) && 909 (is_router = pndisc_is_router(&msg->target, dev)) >= 0)) { 910 if (!(NEIGH_CB(skb)->flags & LOCALLY_ENQUEUED) && 911 skb->pkt_type != PACKET_HOST && 912 inc && 913 NEIGH_VAR(idev->nd_parms, PROXY_DELAY) != 0) { 914 /* 915 * for anycast or proxy, 916 * sender should delay its response 917 * by a random time between 0 and 918 * MAX_ANYCAST_DELAY_TIME seconds. 919 * (RFC2461) -- yoshfuji 920 */ 921 struct sk_buff *n = skb_clone(skb, GFP_ATOMIC); 922 if (n) 923 pneigh_enqueue(tbl, idev->nd_parms, n); 924 goto out; 925 } 926 } else { 927 SKB_DR_SET(reason, IPV6_NDISC_NS_OTHERHOST); 928 goto out; 929 } 930 } 931 932 if (is_router < 0) 933 is_router = READ_ONCE(idev->cnf.forwarding); 934 935 if (dad) { 936 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &msg->target, 937 !!is_router, false, (ifp != NULL), true); 938 goto out; 939 } 940 941 if (inc) 942 NEIGH_CACHE_STAT_INC(tbl, rcv_probes_mcast); 943 else 944 NEIGH_CACHE_STAT_INC(tbl, rcv_probes_ucast); 945 946 /* 947 * update / create cache entry 948 * for the source address 949 */ 950 neigh = ipv6_neigh_lookup(dev, saddr); 951 if (!neigh && (!inc || lladdr || !dev->addr_len)) { 952 neigh = ipv6_neigh_create(dev, saddr); 953 if (IS_ERR(neigh)) 954 neigh = NULL; 955 } 956 if (neigh) 957 ndisc_update(dev, neigh, lladdr, NUD_STALE, 958 NEIGH_UPDATE_F_WEAK_OVERRIDE| 959 NEIGH_UPDATE_F_OVERRIDE, 960 NDISC_NEIGHBOUR_SOLICITATION, &ndopts); 961 if (neigh || !dev->header_ops) { 962 ndisc_send_na(dev, saddr, &msg->target, !!is_router, 963 true, (ifp != NULL && inc), inc); 964 if (neigh) 965 neigh_release(neigh); 966 reason = SKB_CONSUMED; 967 } 968 969 out: 970 if (ifp) 971 in6_ifa_put(ifp); 972 else 973 in6_dev_put(idev); 974 return reason; 975 } 976 977 static int accept_untracked_na(struct inet6_dev *idev, struct in6_addr *saddr) 978 { 979 switch (READ_ONCE(idev->cnf.accept_untracked_na)) { 980 case 0: /* Don't accept untracked NA (absent or FAILED) */ 981 return 0; 982 case 1: /* Create new or update FAILED entries from NA */ 983 return 1; 984 case 2: /* Create new or update FAILED entries only if saddr is in the 985 * same subnet as an address configured on the interface that 986 * received the NA 987 */ 988 return !!ipv6_chk_prefix(saddr, idev->dev); 989 default: 990 return 0; 991 } 992 } 993 994 static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb) 995 { 996 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) + 997 offsetof(struct nd_msg, opt)); 998 struct nd_msg *msg = (struct nd_msg *)skb_transport_header(skb); 999 const struct in6_addr *daddr = &ipv6_hdr(skb)->daddr; 1000 struct in6_addr *saddr = &ipv6_hdr(skb)->saddr; 1001 struct net_device *dev = skb->dev; 1002 struct net *net = dev_net(dev); 1003 struct ndisc_options ndopts; 1004 struct inet6_ifaddr *ifp; 1005 struct neighbour *neigh; 1006 struct inet6_dev *idev; 1007 u8 *lladdr = NULL; 1008 SKB_DR(reason); 1009 u8 new_state; 1010 1011 if (skb->len < sizeof(struct nd_msg)) 1012 return SKB_DROP_REASON_PKT_TOO_SMALL; 1013 1014 if (ipv6_addr_is_multicast(&msg->target)) { 1015 net_dbg_ratelimited("NA: target address is multicast\n"); 1016 return reason; 1017 } 1018 1019 if (ipv6_addr_is_multicast(daddr) && 1020 msg->icmph.icmp6_solicited) { 1021 net_dbg_ratelimited("NA: solicited NA is multicasted\n"); 1022 return reason; 1023 } 1024 1025 /* For some 802.11 wireless deployments (and possibly other networks), 1026 * there will be a NA proxy and unsolicitd packets are attacks 1027 * and thus should not be accepted. 1028 * drop_unsolicited_na takes precedence over accept_untracked_na 1029 */ 1030 idev = __in6_dev_get(dev); 1031 if (!msg->icmph.icmp6_solicited && idev && 1032 READ_ONCE(idev->cnf.drop_unsolicited_na)) 1033 return reason; 1034 1035 if (!ndisc_parse_options(dev, msg->opt, ndoptlen, &ndopts)) 1036 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS; 1037 1038 if (ndopts.nd_opts_tgt_lladdr) { 1039 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr, dev); 1040 if (!lladdr) { 1041 net_dbg_ratelimited("NA: invalid link-layer address length\n"); 1042 return reason; 1043 } 1044 } 1045 ifp = ipv6_get_ifaddr(net, &msg->target, dev, 1); 1046 if (ifp) { 1047 if (skb->pkt_type != PACKET_LOOPBACK 1048 && (ifp->flags & IFA_F_TENTATIVE)) { 1049 addrconf_dad_failure(skb, ifp); 1050 return reason; 1051 } 1052 /* What should we make now? The advertisement 1053 is invalid, but ndisc specs say nothing 1054 about it. It could be misconfiguration, or 1055 an smart proxy agent tries to help us :-) 1056 1057 We should not print the error if NA has been 1058 received from loopback - it is just our own 1059 unsolicited advertisement. 1060 */ 1061 if (skb->pkt_type != PACKET_LOOPBACK) 1062 net_warn_ratelimited("NA: %pM advertised our address %pI6c on %s!\n", 1063 eth_hdr(skb)->h_source, &ifp->addr, 1064 ifp->idev->dev->name); 1065 in6_ifa_put(ifp); 1066 return reason; 1067 } 1068 1069 neigh = ipv6_neigh_lookup(dev, &msg->target); 1070 1071 /* RFC 9131 updates original Neighbour Discovery RFC 4861. 1072 * NAs with Target LL Address option can now create a STALE neighbor 1073 * cache entry on routers if the NA does not have a corresponding entry 1074 * in the neighbour cache or has a corresponding FAILED entry. 1075 * 1076 * entry accept fwding solicited behaviour 1077 * ----------- ------ ------ --------- ---------------------- 1078 * non-FAILED X X 0 Set state to STALE 1079 * non-FAILED X X 1 Set state to REACHABLE 1080 * FAILED 0 X X Do nothing 1081 * FAILED 1 0 X Do nothing 1082 * FAILED 1 1 X Set state to STALE 1083 * absent 0 X X Do nothing 1084 * absent 1 0 X Do nothing 1085 * absent 1 1 X Add a new STALE entry 1086 * 1087 * Note that we don't do a (daddr == all-routers-mcast) check. 1088 */ 1089 new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE; 1090 if (!neigh || (READ_ONCE(neigh->nud_state) & NUD_FAILED)) { 1091 if (!lladdr || !idev || !READ_ONCE(idev->cnf.forwarding) || 1092 !accept_untracked_na(idev, saddr)) { 1093 if (neigh) 1094 neigh_release(neigh); 1095 return reason; 1096 } 1097 if (!neigh) 1098 neigh = ipv6_neigh_create(dev, &msg->target); 1099 new_state = NUD_STALE; 1100 } 1101 1102 if (neigh && !IS_ERR(neigh)) { 1103 u8 old_flags = neigh->flags; 1104 1105 /* 1106 * Don't update the neighbor cache entry on a proxy NA from 1107 * ourselves because either the proxied node is off link or it 1108 * has already sent a NA to us. 1109 */ 1110 if (lladdr && !memcmp(lladdr, dev->dev_addr, dev->addr_len) && 1111 READ_ONCE(net->ipv6.devconf_all->forwarding) && 1112 READ_ONCE(net->ipv6.devconf_all->proxy_ndp) && 1113 pneigh_lookup(nd_table(dev_net(dev)), &msg->target, dev)) { 1114 /* XXX: idev->cnf.proxy_ndp */ 1115 goto out; 1116 } 1117 1118 ndisc_update(dev, neigh, lladdr, 1119 new_state, 1120 NEIGH_UPDATE_F_WEAK_OVERRIDE| 1121 (msg->icmph.icmp6_override ? NEIGH_UPDATE_F_OVERRIDE : 0)| 1122 NEIGH_UPDATE_F_OVERRIDE_ISROUTER| 1123 (msg->icmph.icmp6_router ? NEIGH_UPDATE_F_ISROUTER : 0), 1124 NDISC_NEIGHBOUR_ADVERTISEMENT, &ndopts); 1125 1126 if ((old_flags & ~neigh->flags) & NTF_ROUTER) { 1127 /* 1128 * Change: router to host 1129 */ 1130 rt6_clean_tohost(dev_net(dev), saddr); 1131 } 1132 reason = SKB_CONSUMED; 1133 out: 1134 neigh_release(neigh); 1135 } 1136 return reason; 1137 } 1138 1139 static enum skb_drop_reason ndisc_recv_rs(struct sk_buff *skb) 1140 { 1141 struct rs_msg *rs_msg = (struct rs_msg *)skb_transport_header(skb); 1142 const struct in6_addr *saddr = &ipv6_hdr(skb)->saddr; 1143 unsigned long ndoptlen = skb->len - sizeof(*rs_msg); 1144 struct net_device *dev = skb->dev; 1145 struct ndisc_options ndopts; 1146 struct neighbour *neigh; 1147 struct inet6_dev *idev; 1148 u8 *lladdr = NULL; 1149 SKB_DR(reason); 1150 1151 if (skb->len < sizeof(*rs_msg)) 1152 return SKB_DROP_REASON_PKT_TOO_SMALL; 1153 1154 idev = __in6_dev_get(dev); 1155 if (!idev) { 1156 net_err_ratelimited("RS: can't find in6 device\n"); 1157 return reason; 1158 } 1159 1160 /* Don't accept RS if we're not in router mode */ 1161 if (!READ_ONCE(idev->cnf.forwarding)) 1162 goto out; 1163 1164 /* 1165 * Don't update NCE if src = ::; 1166 * this implies that the source node has no ip address assigned yet. 1167 */ 1168 if (ipv6_addr_any(saddr)) 1169 goto out; 1170 1171 /* Parse ND options */ 1172 if (!ndisc_parse_options(dev, rs_msg->opt, ndoptlen, &ndopts)) 1173 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS; 1174 1175 if (ndopts.nd_opts_src_lladdr) { 1176 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev); 1177 if (!lladdr) 1178 goto out; 1179 } 1180 1181 neigh = ipv6_neigh_lookup(dev, saddr); 1182 if (!neigh) { 1183 neigh = ipv6_neigh_create(dev, saddr); 1184 if (IS_ERR(neigh)) 1185 goto out; 1186 } 1187 if (neigh) { 1188 ndisc_update(dev, neigh, lladdr, NUD_STALE, 1189 NEIGH_UPDATE_F_WEAK_OVERRIDE| 1190 NEIGH_UPDATE_F_OVERRIDE| 1191 NEIGH_UPDATE_F_OVERRIDE_ISROUTER, 1192 NDISC_ROUTER_SOLICITATION, &ndopts); 1193 neigh_release(neigh); 1194 reason = SKB_CONSUMED; 1195 } 1196 out: 1197 return reason; 1198 } 1199 1200 static void ndisc_ra_useropt(struct sk_buff *ra, struct nd_opt_hdr *opt) 1201 { 1202 struct icmp6hdr *icmp6h = (struct icmp6hdr *)skb_transport_header(ra); 1203 struct sk_buff *skb; 1204 struct nlmsghdr *nlh; 1205 struct nduseroptmsg *ndmsg; 1206 struct net *net = dev_net(ra->dev); 1207 int err; 1208 int base_size = NLMSG_ALIGN(sizeof(struct nduseroptmsg) 1209 + (opt->nd_opt_len << 3)); 1210 size_t msg_size = base_size + nla_total_size(sizeof(struct in6_addr)); 1211 1212 skb = nlmsg_new(msg_size, GFP_ATOMIC); 1213 if (!skb) { 1214 err = -ENOBUFS; 1215 goto errout; 1216 } 1217 1218 nlh = nlmsg_put(skb, 0, 0, RTM_NEWNDUSEROPT, base_size, 0); 1219 if (!nlh) { 1220 goto nla_put_failure; 1221 } 1222 1223 ndmsg = nlmsg_data(nlh); 1224 ndmsg->nduseropt_family = AF_INET6; 1225 ndmsg->nduseropt_ifindex = ra->dev->ifindex; 1226 ndmsg->nduseropt_icmp_type = icmp6h->icmp6_type; 1227 ndmsg->nduseropt_icmp_code = icmp6h->icmp6_code; 1228 ndmsg->nduseropt_opts_len = opt->nd_opt_len << 3; 1229 ndmsg->nduseropt_pad1 = 0; 1230 ndmsg->nduseropt_pad2 = 0; 1231 ndmsg->nduseropt_pad3 = 0; 1232 1233 memcpy(ndmsg + 1, opt, opt->nd_opt_len << 3); 1234 1235 if (nla_put_in6_addr(skb, NDUSEROPT_SRCADDR, &ipv6_hdr(ra)->saddr)) 1236 goto nla_put_failure; 1237 nlmsg_end(skb, nlh); 1238 1239 rtnl_notify(skb, net, 0, RTNLGRP_ND_USEROPT, NULL, GFP_ATOMIC); 1240 return; 1241 1242 nla_put_failure: 1243 nlmsg_free(skb); 1244 err = -EMSGSIZE; 1245 errout: 1246 rtnl_set_sk_err(net, RTNLGRP_ND_USEROPT, err); 1247 } 1248 1249 static enum skb_drop_reason ndisc_router_discovery(struct sk_buff *skb) 1250 { 1251 struct ra_msg *ra_msg = (struct ra_msg *)skb_transport_header(skb); 1252 struct net *net = dev_net(skb->dev); 1253 bool send_ifinfo_notify = false; 1254 struct neighbour *neigh = NULL; 1255 struct ndisc_options ndopts; 1256 struct fib6_info *rt = NULL; 1257 struct inet6_dev *in6_dev; 1258 struct fib6_table *table; 1259 u32 defrtr_usr_metric; 1260 unsigned int pref = 0; 1261 __u32 old_if_flags; 1262 SKB_DR(reason); 1263 int lifetime; 1264 int optlen; 1265 1266 __u8 *opt = (__u8 *)(ra_msg + 1); 1267 1268 optlen = (skb_tail_pointer(skb) - skb_transport_header(skb)) - 1269 sizeof(struct ra_msg); 1270 1271 net_dbg_ratelimited("RA: %s, dev: %s\n", __func__, skb->dev->name); 1272 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) { 1273 net_dbg_ratelimited("RA: source address is not link-local\n"); 1274 return reason; 1275 } 1276 if (optlen < 0) 1277 return SKB_DROP_REASON_PKT_TOO_SMALL; 1278 1279 #ifdef CONFIG_IPV6_NDISC_NODETYPE 1280 if (skb->ndisc_nodetype == NDISC_NODETYPE_HOST) { 1281 net_dbg_ratelimited("RA: from host or unauthorized router\n"); 1282 return reason; 1283 } 1284 #endif 1285 1286 in6_dev = __in6_dev_get(skb->dev); 1287 if (!in6_dev) { 1288 net_err_ratelimited("RA: can't find inet6 device for %s\n", skb->dev->name); 1289 return reason; 1290 } 1291 1292 if (!ndisc_parse_options(skb->dev, opt, optlen, &ndopts)) 1293 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS; 1294 1295 if (!ipv6_accept_ra(in6_dev)) { 1296 net_dbg_ratelimited("RA: %s, did not accept ra for dev: %s\n", __func__, 1297 skb->dev->name); 1298 goto skip_linkparms; 1299 } 1300 1301 #ifdef CONFIG_IPV6_NDISC_NODETYPE 1302 /* skip link-specific parameters from interior routers */ 1303 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) { 1304 net_dbg_ratelimited("RA: %s, nodetype is NODEFAULT, dev: %s\n", __func__, 1305 skb->dev->name); 1306 goto skip_linkparms; 1307 } 1308 #endif 1309 1310 if (in6_dev->if_flags & IF_RS_SENT) { 1311 /* 1312 * flag that an RA was received after an RS was sent 1313 * out on this interface. 1314 */ 1315 in6_dev->if_flags |= IF_RA_RCVD; 1316 } 1317 1318 /* 1319 * Remember the managed/otherconf flags from most recently 1320 * received RA message (RFC 2462) -- yoshfuji 1321 */ 1322 old_if_flags = in6_dev->if_flags; 1323 in6_dev->if_flags = (in6_dev->if_flags & ~(IF_RA_MANAGED | 1324 IF_RA_OTHERCONF)) | 1325 (ra_msg->icmph.icmp6_addrconf_managed ? 1326 IF_RA_MANAGED : 0) | 1327 (ra_msg->icmph.icmp6_addrconf_other ? 1328 IF_RA_OTHERCONF : 0); 1329 1330 if (old_if_flags != in6_dev->if_flags) 1331 send_ifinfo_notify = true; 1332 1333 if (!READ_ONCE(in6_dev->cnf.accept_ra_defrtr)) { 1334 net_dbg_ratelimited("RA: %s, defrtr is false for dev: %s\n", __func__, 1335 skb->dev->name); 1336 goto skip_defrtr; 1337 } 1338 1339 lifetime = ntohs(ra_msg->icmph.icmp6_rt_lifetime); 1340 if (lifetime != 0 && 1341 lifetime < READ_ONCE(in6_dev->cnf.accept_ra_min_lft)) { 1342 net_dbg_ratelimited("RA: router lifetime (%ds) is too short: %s\n", lifetime, 1343 skb->dev->name); 1344 goto skip_defrtr; 1345 } 1346 1347 /* Do not accept RA with source-addr found on local machine unless 1348 * accept_ra_from_local is set to true. 1349 */ 1350 if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) && 1351 ipv6_chk_addr(net, &ipv6_hdr(skb)->saddr, in6_dev->dev, 0)) { 1352 net_dbg_ratelimited("RA from local address detected on dev: %s: default router ignored\n", 1353 skb->dev->name); 1354 goto skip_defrtr; 1355 } 1356 1357 #ifdef CONFIG_IPV6_ROUTER_PREF 1358 pref = ra_msg->icmph.icmp6_router_pref; 1359 /* 10b is handled as if it were 00b (medium) */ 1360 if (pref == ICMPV6_ROUTER_PREF_INVALID || 1361 !READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref)) 1362 pref = ICMPV6_ROUTER_PREF_MEDIUM; 1363 #endif 1364 /* routes added from RAs do not use nexthop objects */ 1365 rt = rt6_get_dflt_router(net, &ipv6_hdr(skb)->saddr, skb->dev); 1366 if (rt) { 1367 neigh = __ip6_dst_neigh_lookup(&rt->fib6_nh->fib_nh_gw6, 1368 rt->fib6_nh->fib_nh_dev, NULL, 1369 &ipv6_hdr(skb)->saddr); 1370 if (!neigh) { 1371 net_err_ratelimited("RA: %s got default router without neighbour\n", 1372 __func__); 1373 fib6_info_release(rt); 1374 return reason; 1375 } 1376 } 1377 /* Set default route metric as specified by user */ 1378 defrtr_usr_metric = in6_dev->cnf.ra_defrtr_metric; 1379 /* delete the route if lifetime is 0 or if metric needs change */ 1380 if (rt && (lifetime == 0 || rt->fib6_metric != defrtr_usr_metric)) { 1381 ip6_del_rt_reason(net, rt, 1382 lifetime == 0 ? RT_DEL_REASON_RA_WITHDRAWN : 1383 RT_DEL_REASON_UNSPEC); 1384 rt = NULL; 1385 } 1386 1387 net_dbg_ratelimited("RA: rt: %p lifetime: %d, metric: %d, for dev: %s\n", rt, lifetime, 1388 defrtr_usr_metric, skb->dev->name); 1389 if (!rt && lifetime) { 1390 net_dbg_ratelimited("RA: adding default router\n"); 1391 1392 if (neigh) 1393 neigh_release(neigh); 1394 1395 rt = rt6_add_dflt_router(net, &ipv6_hdr(skb)->saddr, 1396 skb->dev, pref, defrtr_usr_metric, 1397 lifetime); 1398 if (!rt) { 1399 net_err_ratelimited("RA: %s failed to add default route\n", __func__); 1400 return reason; 1401 } 1402 1403 neigh = __ip6_dst_neigh_lookup(&rt->fib6_nh->fib_nh_gw6, 1404 rt->fib6_nh->fib_nh_dev, NULL, 1405 &ipv6_hdr(skb)->saddr); 1406 if (!neigh) { 1407 net_err_ratelimited("RA: %s got default router without neighbour\n", 1408 __func__); 1409 fib6_info_release(rt); 1410 return reason; 1411 } 1412 neigh->flags |= NTF_ROUTER; 1413 } else if (rt && IPV6_EXTRACT_PREF(rt->fib6_flags) != pref) { 1414 struct nl_info nlinfo = { 1415 .nl_net = net, 1416 }; 1417 rt->fib6_flags = (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref); 1418 inet6_rt_notify(RTM_NEWROUTE, rt, &nlinfo, NLM_F_REPLACE); 1419 } 1420 1421 if (rt) { 1422 table = rt->fib6_table; 1423 spin_lock_bh(&table->tb6_lock); 1424 1425 fib6_set_expires(rt, jiffies + (HZ * lifetime)); 1426 fib6_add_gc_list(rt); 1427 1428 spin_unlock_bh(&table->tb6_lock); 1429 } 1430 if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) < 256 && 1431 ra_msg->icmph.icmp6_hop_limit) { 1432 if (READ_ONCE(in6_dev->cnf.accept_ra_min_hop_limit) <= 1433 ra_msg->icmph.icmp6_hop_limit) { 1434 WRITE_ONCE(in6_dev->cnf.hop_limit, 1435 ra_msg->icmph.icmp6_hop_limit); 1436 fib6_metric_set(rt, RTAX_HOPLIMIT, 1437 ra_msg->icmph.icmp6_hop_limit); 1438 } else { 1439 net_dbg_ratelimited("RA: Got route advertisement with lower hop_limit than minimum\n"); 1440 } 1441 } 1442 1443 skip_defrtr: 1444 1445 /* 1446 * Update Reachable Time and Retrans Timer 1447 */ 1448 1449 if (in6_dev->nd_parms) { 1450 unsigned long rtime = ntohl(ra_msg->retrans_timer); 1451 1452 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/HZ) { 1453 rtime = (rtime*HZ)/1000; 1454 if (rtime < HZ/100) 1455 rtime = HZ/100; 1456 NEIGH_VAR_SET(in6_dev->nd_parms, RETRANS_TIME, rtime); 1457 in6_dev->tstamp = jiffies; 1458 send_ifinfo_notify = true; 1459 } 1460 1461 rtime = ntohl(ra_msg->reachable_time); 1462 if (rtime && rtime/1000 < MAX_SCHEDULE_TIMEOUT/(3*HZ)) { 1463 rtime = (rtime*HZ)/1000; 1464 1465 if (rtime < HZ/10) 1466 rtime = HZ/10; 1467 1468 if (rtime != NEIGH_VAR(in6_dev->nd_parms, BASE_REACHABLE_TIME)) { 1469 NEIGH_VAR_SET(in6_dev->nd_parms, 1470 BASE_REACHABLE_TIME, rtime); 1471 NEIGH_VAR_SET(in6_dev->nd_parms, 1472 GC_STALETIME, 3 * rtime); 1473 neigh_set_reach_time(in6_dev->nd_parms); 1474 in6_dev->tstamp = jiffies; 1475 send_ifinfo_notify = true; 1476 } 1477 } 1478 } 1479 1480 skip_linkparms: 1481 1482 /* 1483 * Process options. 1484 */ 1485 1486 if (!neigh) { 1487 neigh = ipv6_neigh_lookup(skb->dev, &ipv6_hdr(skb)->saddr); 1488 if (!neigh) { 1489 neigh = ipv6_neigh_create(skb->dev, &ipv6_hdr(skb)->saddr); 1490 if (IS_ERR(neigh)) 1491 neigh = NULL; 1492 } 1493 } 1494 if (neigh) { 1495 u8 *lladdr = NULL; 1496 if (ndopts.nd_opts_src_lladdr) { 1497 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, 1498 skb->dev); 1499 if (!lladdr) { 1500 net_dbg_ratelimited("RA: invalid link-layer address length\n"); 1501 goto out; 1502 } 1503 } 1504 ndisc_update(skb->dev, neigh, lladdr, NUD_STALE, 1505 NEIGH_UPDATE_F_WEAK_OVERRIDE| 1506 NEIGH_UPDATE_F_OVERRIDE| 1507 NEIGH_UPDATE_F_OVERRIDE_ISROUTER| 1508 NEIGH_UPDATE_F_ISROUTER, 1509 NDISC_ROUTER_ADVERTISEMENT, &ndopts); 1510 reason = SKB_CONSUMED; 1511 } 1512 1513 if (!ipv6_accept_ra(in6_dev)) { 1514 net_dbg_ratelimited("RA: %s, accept_ra is false for dev: %s\n", __func__, 1515 skb->dev->name); 1516 goto out; 1517 } 1518 1519 #ifdef CONFIG_IPV6_ROUTE_INFO 1520 if (!READ_ONCE(in6_dev->cnf.accept_ra_from_local) && 1521 ipv6_chk_addr(dev_net(in6_dev->dev), &ipv6_hdr(skb)->saddr, 1522 in6_dev->dev, 0)) { 1523 net_dbg_ratelimited("RA from local address detected on dev: %s: router info ignored.\n", 1524 skb->dev->name); 1525 goto skip_routeinfo; 1526 } 1527 1528 if (READ_ONCE(in6_dev->cnf.accept_ra_rtr_pref) && ndopts.nd_opts_ri) { 1529 struct nd_opt_hdr *p; 1530 for (p = ndopts.nd_opts_ri; 1531 p; 1532 p = ndisc_next_option(p, ndopts.nd_opts_ri_end)) { 1533 struct route_info *ri = (struct route_info *)p; 1534 #ifdef CONFIG_IPV6_NDISC_NODETYPE 1535 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT && 1536 ri->prefix_len == 0) 1537 continue; 1538 #endif 1539 if (ri->prefix_len == 0 && 1540 !READ_ONCE(in6_dev->cnf.accept_ra_defrtr)) 1541 continue; 1542 if (ri->lifetime != 0 && 1543 ntohl(ri->lifetime) < READ_ONCE(in6_dev->cnf.accept_ra_min_lft)) 1544 continue; 1545 if (ri->prefix_len < READ_ONCE(in6_dev->cnf.accept_ra_rt_info_min_plen)) 1546 continue; 1547 if (ri->prefix_len > READ_ONCE(in6_dev->cnf.accept_ra_rt_info_max_plen)) 1548 continue; 1549 rt6_route_rcv(skb->dev, (u8 *)p, (p->nd_opt_len) << 3, 1550 &ipv6_hdr(skb)->saddr); 1551 } 1552 } 1553 1554 skip_routeinfo: 1555 #endif 1556 1557 #ifdef CONFIG_IPV6_NDISC_NODETYPE 1558 /* skip link-specific ndopts from interior routers */ 1559 if (skb->ndisc_nodetype == NDISC_NODETYPE_NODEFAULT) { 1560 net_dbg_ratelimited("RA: %s, nodetype is NODEFAULT (interior routes), dev: %s\n", 1561 __func__, skb->dev->name); 1562 goto out; 1563 } 1564 #endif 1565 1566 if (READ_ONCE(in6_dev->cnf.accept_ra_pinfo) && ndopts.nd_opts_pi) { 1567 struct nd_opt_hdr *p; 1568 for (p = ndopts.nd_opts_pi; 1569 p; 1570 p = ndisc_next_option(p, ndopts.nd_opts_pi_end)) { 1571 addrconf_prefix_rcv(skb->dev, (u8 *)p, 1572 (p->nd_opt_len) << 3, 1573 ndopts.nd_opts_src_lladdr != NULL); 1574 } 1575 } 1576 1577 if (ndopts.nd_opts_mtu && READ_ONCE(in6_dev->cnf.accept_ra_mtu)) { 1578 __be32 n; 1579 u32 mtu; 1580 1581 memcpy(&n, ((u8 *)(ndopts.nd_opts_mtu+1))+2, sizeof(mtu)); 1582 mtu = ntohl(n); 1583 1584 if (READ_ONCE(in6_dev->ra_mtu) != mtu) { 1585 WRITE_ONCE(in6_dev->ra_mtu, mtu); 1586 send_ifinfo_notify = true; 1587 } 1588 1589 if (mtu < IPV6_MIN_MTU || mtu > skb->dev->mtu) { 1590 net_dbg_ratelimited("RA: invalid mtu: %d\n", mtu); 1591 } else if (READ_ONCE(in6_dev->cnf.mtu6) != mtu) { 1592 WRITE_ONCE(in6_dev->cnf.mtu6, mtu); 1593 fib6_metric_set(rt, RTAX_MTU, mtu); 1594 rt6_mtu_change(skb->dev, mtu); 1595 } 1596 } 1597 1598 if (ndopts.nd_useropts) { 1599 struct nd_opt_hdr *p; 1600 for (p = ndopts.nd_useropts; 1601 p; 1602 p = ndisc_next_useropt(skb->dev, p, 1603 ndopts.nd_useropts_end)) { 1604 ndisc_ra_useropt(skb, p); 1605 } 1606 } 1607 1608 if (ndopts.nd_opts_tgt_lladdr || ndopts.nd_opts_rh) { 1609 net_dbg_ratelimited("RA: invalid RA options\n"); 1610 } 1611 out: 1612 /* Send a notify if RA changed managed/otherconf flags or 1613 * timer settings or ra_mtu value 1614 */ 1615 if (send_ifinfo_notify) 1616 inet6_ifinfo_notify(RTM_NEWLINK, in6_dev); 1617 1618 fib6_info_release(rt); 1619 if (neigh) 1620 neigh_release(neigh); 1621 return reason; 1622 } 1623 1624 static enum skb_drop_reason ndisc_redirect_rcv(struct sk_buff *skb) 1625 { 1626 struct rd_msg *msg = (struct rd_msg *)skb_transport_header(skb); 1627 u32 ndoptlen = skb_tail_pointer(skb) - (skb_transport_header(skb) + 1628 offsetof(struct rd_msg, opt)); 1629 struct ndisc_options ndopts; 1630 SKB_DR(reason); 1631 u8 *hdr; 1632 1633 #ifdef CONFIG_IPV6_NDISC_NODETYPE 1634 switch (skb->ndisc_nodetype) { 1635 case NDISC_NODETYPE_HOST: 1636 case NDISC_NODETYPE_NODEFAULT: 1637 net_dbg_ratelimited("Redirect: from host or unauthorized router\n"); 1638 return reason; 1639 } 1640 #endif 1641 1642 if (!(ipv6_addr_type(&ipv6_hdr(skb)->saddr) & IPV6_ADDR_LINKLOCAL)) { 1643 net_dbg_ratelimited("Redirect: source address is not link-local\n"); 1644 return reason; 1645 } 1646 1647 if (!ndisc_parse_options(skb->dev, msg->opt, ndoptlen, &ndopts)) 1648 return SKB_DROP_REASON_IPV6_NDISC_BAD_OPTIONS; 1649 1650 if (!ndopts.nd_opts_rh) { 1651 ip6_redirect_no_header(skb, dev_net(skb->dev), 1652 skb->dev->ifindex); 1653 return reason; 1654 } 1655 1656 hdr = (u8 *)ndopts.nd_opts_rh; 1657 hdr += 8; 1658 if (!pskb_pull(skb, hdr - skb_transport_header(skb))) 1659 return SKB_DROP_REASON_PKT_TOO_SMALL; 1660 1661 return icmpv6_notify(skb, NDISC_REDIRECT, 0, 0); 1662 } 1663 1664 static void ndisc_fill_redirect_hdr_option(struct sk_buff *skb, 1665 struct sk_buff *orig_skb, 1666 int rd_len) 1667 { 1668 u8 *opt = skb_put(skb, rd_len); 1669 1670 memset(opt, 0, 8); 1671 *(opt++) = ND_OPT_REDIRECT_HDR; 1672 *(opt++) = (rd_len >> 3); 1673 opt += 6; 1674 1675 skb_copy_bits(orig_skb, skb_network_offset(orig_skb), opt, 1676 rd_len - 8); 1677 } 1678 1679 void ndisc_send_redirect(struct sk_buff *skb, const struct in6_addr *target) 1680 { 1681 struct net_device *dev = skb->dev; 1682 struct net *net = dev_net_rcu(dev); 1683 struct sock *sk = net->ipv6.ndisc_sk; 1684 int optlen = 0; 1685 struct inet_peer *peer; 1686 struct sk_buff *buff; 1687 struct rd_msg *msg; 1688 struct in6_addr saddr_buf; 1689 struct rt6_info *rt; 1690 struct dst_entry *dst; 1691 struct flowi6 fl6; 1692 int rd_len; 1693 u8 ha_buf[MAX_ADDR_LEN], *ha = NULL, 1694 ops_data_buf[NDISC_OPS_REDIRECT_DATA_SPACE], *ops_data = NULL; 1695 bool ret; 1696 1697 if (netif_is_l3_master(dev)) { 1698 dev = dev_get_by_index_rcu(net, IPCB(skb)->iif); 1699 if (!dev) 1700 return; 1701 } 1702 1703 if (ipv6_get_lladdr(dev, &saddr_buf, IFA_F_TENTATIVE)) { 1704 net_dbg_ratelimited("Redirect: no link-local address on %s\n", dev->name); 1705 return; 1706 } 1707 1708 if (!ipv6_addr_equal(&ipv6_hdr(skb)->daddr, target) && 1709 ipv6_addr_type(target) != (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) { 1710 net_dbg_ratelimited("Redirect: target address is not link-local unicast\n"); 1711 return; 1712 } 1713 1714 icmpv6_flow_init(sk, &fl6, NDISC_REDIRECT, 1715 &saddr_buf, &ipv6_hdr(skb)->saddr, dev->ifindex); 1716 1717 dst = ip6_route_output(net, NULL, &fl6); 1718 if (dst->error) { 1719 dst_release(dst); 1720 return; 1721 } 1722 dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), NULL, 0); 1723 if (IS_ERR(dst)) 1724 return; 1725 1726 rt = dst_rt6_info(dst); 1727 1728 if (rt->rt6i_flags & RTF_GATEWAY) { 1729 net_dbg_ratelimited("Redirect: destination is not a neighbour\n"); 1730 goto release; 1731 } 1732 1733 peer = inet_getpeer_v6(net->ipv6.peers, &ipv6_hdr(skb)->saddr); 1734 if (!peer) 1735 goto release; 1736 ret = inet_peer_xrlim_allow(peer, 1*HZ); 1737 1738 if (!ret) 1739 goto release; 1740 1741 if (dev->addr_len) { 1742 struct neighbour *neigh = dst_neigh_lookup(skb_dst(skb), target); 1743 if (!neigh) { 1744 net_dbg_ratelimited("Redirect: no neigh for target address\n"); 1745 goto release; 1746 } 1747 1748 read_lock_bh(&neigh->lock); 1749 if (neigh->nud_state & NUD_VALID) { 1750 memcpy(ha_buf, neigh->ha, dev->addr_len); 1751 read_unlock_bh(&neigh->lock); 1752 ha = ha_buf; 1753 optlen += ndisc_redirect_opt_addr_space(dev, neigh, 1754 ops_data_buf, 1755 &ops_data); 1756 } else 1757 read_unlock_bh(&neigh->lock); 1758 1759 neigh_release(neigh); 1760 } 1761 1762 rd_len = min_t(unsigned int, 1763 IPV6_MIN_MTU - sizeof(struct ipv6hdr) - sizeof(*msg) - optlen, 1764 skb->len + 8); 1765 rd_len &= ~0x7; 1766 optlen += rd_len; 1767 1768 buff = ndisc_alloc_skb(dev, sizeof(*msg) + optlen); 1769 if (!buff) 1770 goto release; 1771 1772 msg = skb_put(buff, sizeof(*msg)); 1773 *msg = (struct rd_msg) { 1774 .icmph = { 1775 .icmp6_type = NDISC_REDIRECT, 1776 }, 1777 .target = *target, 1778 .dest = ipv6_hdr(skb)->daddr, 1779 }; 1780 1781 /* 1782 * include target_address option 1783 */ 1784 1785 if (ha) 1786 ndisc_fill_redirect_addr_option(buff, ha, ops_data); 1787 1788 /* 1789 * build redirect option and copy skb over to the new packet. 1790 */ 1791 1792 if (rd_len) 1793 ndisc_fill_redirect_hdr_option(buff, skb, rd_len); 1794 1795 skb_dst_set(buff, dst); 1796 ndisc_send_skb(buff, &ipv6_hdr(skb)->saddr, &saddr_buf); 1797 return; 1798 1799 release: 1800 dst_release(dst); 1801 } 1802 1803 static void pndisc_redo(struct sk_buff *skb) 1804 { 1805 enum skb_drop_reason reason = ndisc_recv_ns(skb); 1806 1807 kfree_skb_reason(skb, reason); 1808 } 1809 1810 static int ndisc_is_multicast(const void *pkey) 1811 { 1812 return ipv6_addr_is_multicast((struct in6_addr *)pkey); 1813 } 1814 1815 static bool ndisc_suppress_frag_ndisc(struct sk_buff *skb) 1816 { 1817 struct inet6_dev *idev = __in6_dev_get(skb->dev); 1818 1819 if (!idev) 1820 return true; 1821 if (IP6CB(skb)->flags & IP6SKB_FRAGMENTED && 1822 READ_ONCE(idev->cnf.suppress_frag_ndisc)) { 1823 net_warn_ratelimited("Received fragmented ndisc packet. Carefully consider disabling suppress_frag_ndisc.\n"); 1824 return true; 1825 } 1826 return false; 1827 } 1828 1829 enum skb_drop_reason ndisc_rcv(struct sk_buff *skb) 1830 { 1831 struct nd_msg *msg; 1832 SKB_DR(reason); 1833 1834 if (ndisc_suppress_frag_ndisc(skb)) 1835 return SKB_DROP_REASON_IPV6_NDISC_FRAG; 1836 1837 if (skb_linearize(skb)) 1838 return SKB_DROP_REASON_NOMEM; 1839 1840 msg = (struct nd_msg *)skb_transport_header(skb); 1841 1842 __skb_push(skb, skb->data - skb_transport_header(skb)); 1843 1844 if (ipv6_hdr(skb)->hop_limit != 255) { 1845 net_dbg_ratelimited("NDISC: invalid hop-limit: %d\n", ipv6_hdr(skb)->hop_limit); 1846 return SKB_DROP_REASON_IPV6_NDISC_HOP_LIMIT; 1847 } 1848 1849 if (msg->icmph.icmp6_code != 0) { 1850 net_dbg_ratelimited("NDISC: invalid ICMPv6 code: %d\n", msg->icmph.icmp6_code); 1851 return SKB_DROP_REASON_IPV6_NDISC_BAD_CODE; 1852 } 1853 1854 switch (msg->icmph.icmp6_type) { 1855 case NDISC_NEIGHBOUR_SOLICITATION: 1856 memset(NEIGH_CB(skb), 0, sizeof(struct neighbour_cb)); 1857 reason = ndisc_recv_ns(skb); 1858 break; 1859 1860 case NDISC_NEIGHBOUR_ADVERTISEMENT: 1861 reason = ndisc_recv_na(skb); 1862 break; 1863 1864 case NDISC_ROUTER_SOLICITATION: 1865 reason = ndisc_recv_rs(skb); 1866 break; 1867 1868 case NDISC_ROUTER_ADVERTISEMENT: 1869 reason = ndisc_router_discovery(skb); 1870 break; 1871 1872 case NDISC_REDIRECT: 1873 reason = ndisc_redirect_rcv(skb); 1874 break; 1875 } 1876 1877 return reason; 1878 } 1879 1880 static int ndisc_netdev_event(struct notifier_block *this, unsigned long event, void *ptr) 1881 { 1882 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1883 struct netdev_notifier_change_info *change_info; 1884 struct net *net = dev_net(dev); 1885 struct neigh_table *tbl; 1886 struct inet6_dev *idev; 1887 bool evict_nocarrier; 1888 1889 tbl = nd_table(net); 1890 1891 switch (event) { 1892 case NETDEV_CHANGEADDR: 1893 neigh_changeaddr(tbl, dev); 1894 fib6_run_gc(0, net, false); 1895 fallthrough; 1896 case NETDEV_UP: 1897 idev = in6_dev_get(dev); 1898 if (!idev) 1899 break; 1900 if (READ_ONCE(idev->cnf.ndisc_notify) || 1901 READ_ONCE(net->ipv6.devconf_all->ndisc_notify)) 1902 ndisc_send_unsol_na(dev); 1903 in6_dev_put(idev); 1904 break; 1905 case NETDEV_CHANGE: 1906 idev = in6_dev_get(dev); 1907 if (!idev) 1908 evict_nocarrier = true; 1909 else { 1910 evict_nocarrier = READ_ONCE(idev->cnf.ndisc_evict_nocarrier) && 1911 READ_ONCE(net->ipv6.devconf_all->ndisc_evict_nocarrier); 1912 in6_dev_put(idev); 1913 } 1914 1915 change_info = ptr; 1916 if (change_info->flags_changed & IFF_NOARP) 1917 neigh_changeaddr(tbl, dev); 1918 if (evict_nocarrier && !netif_carrier_ok(dev)) 1919 neigh_carrier_down(tbl, dev); 1920 break; 1921 case NETDEV_DOWN: 1922 neigh_ifdown(tbl, dev); 1923 fib6_run_gc(0, net, false); 1924 break; 1925 case NETDEV_NOTIFY_PEERS: 1926 ndisc_send_unsol_na(dev); 1927 break; 1928 default: 1929 break; 1930 } 1931 1932 return NOTIFY_DONE; 1933 } 1934 1935 static struct notifier_block ndisc_netdev_notifier = { 1936 .notifier_call = ndisc_netdev_event, 1937 .priority = ADDRCONF_NOTIFY_PRIORITY - 5, 1938 }; 1939 1940 #ifdef CONFIG_SYSCTL 1941 static void ndisc_warn_deprecated_sysctl(const struct ctl_table *ctl, 1942 const char *func, const char *dev_name) 1943 { 1944 static char warncomm[TASK_COMM_LEN]; 1945 static int warned; 1946 if (strcmp(warncomm, current->comm) && warned < 5) { 1947 strscpy(warncomm, current->comm); 1948 pr_warn("process `%s' is using deprecated sysctl (%s) net.ipv6.neigh.%s.%s - use net.ipv6.neigh.%s.%s_ms instead\n", 1949 warncomm, func, 1950 dev_name, ctl->procname, 1951 dev_name, ctl->procname); 1952 warned++; 1953 } 1954 } 1955 1956 int ndisc_ifinfo_sysctl_change(const struct ctl_table *ctl, int write, void *buffer, 1957 size_t *lenp, loff_t *ppos) 1958 { 1959 struct net_device *dev = ctl->extra1; 1960 struct inet6_dev *idev; 1961 int ret; 1962 1963 if ((strcmp(ctl->procname, "retrans_time") == 0) || 1964 (strcmp(ctl->procname, "base_reachable_time") == 0)) 1965 ndisc_warn_deprecated_sysctl(ctl, "syscall", dev ? dev->name : "default"); 1966 1967 if (strcmp(ctl->procname, "retrans_time") == 0) 1968 ret = neigh_proc_dointvec(ctl, write, buffer, lenp, ppos); 1969 1970 else if (strcmp(ctl->procname, "base_reachable_time") == 0) 1971 ret = neigh_proc_dointvec_jiffies(ctl, write, 1972 buffer, lenp, ppos); 1973 1974 else if ((strcmp(ctl->procname, "retrans_time_ms") == 0) || 1975 (strcmp(ctl->procname, "base_reachable_time_ms") == 0)) 1976 ret = neigh_proc_dointvec_ms_jiffies(ctl, write, 1977 buffer, lenp, ppos); 1978 else 1979 ret = -1; 1980 1981 if (write && ret == 0 && dev && (idev = in6_dev_get(dev)) != NULL) { 1982 if (ctl->data == NEIGH_VAR_PTR(idev->nd_parms, BASE_REACHABLE_TIME)) 1983 neigh_set_reach_time(idev->nd_parms); 1984 1985 WRITE_ONCE(idev->tstamp, jiffies); 1986 inet6_ifinfo_notify(RTM_NEWLINK, idev); 1987 in6_dev_put(idev); 1988 } 1989 return ret; 1990 } 1991 1992 1993 #endif 1994 1995 static int __net_init ndisc_net_init(struct net *net) 1996 { 1997 struct ipv6_pinfo *np; 1998 struct sock *sk; 1999 int err; 2000 2001 err = neigh_table_register(net, &nd_tbl, NEIGH_ND_TABLE); 2002 if (err) 2003 goto err; 2004 2005 #ifdef CONFIG_SYSCTL 2006 err = neigh_sysctl_register(NULL, &nd_table(net)->parms, 2007 ndisc_ifinfo_sysctl_change); 2008 if (err) 2009 goto err_sysctl; 2010 #endif 2011 2012 err = inet_ctl_sock_create(&sk, PF_INET6, 2013 SOCK_RAW, IPPROTO_ICMPV6, net); 2014 if (err < 0) { 2015 net_err_ratelimited("NDISC: Failed to initialize the control socket (err %d)\n", 2016 err); 2017 goto err_sock_create; 2018 } 2019 2020 net->ipv6.ndisc_sk = sk; 2021 2022 np = inet6_sk(sk); 2023 np->hop_limit = 255; 2024 /* Do not loopback ndisc messages */ 2025 inet6_clear_bit(MC6_LOOP, sk); 2026 2027 return 0; 2028 2029 err_sock_create: 2030 #ifdef CONFIG_SYSCTL 2031 neigh_sysctl_unregister(&nd_table(net)->parms); 2032 err_sysctl: 2033 #endif 2034 neigh_table_unregister(net, NEIGH_ND_TABLE); 2035 err: 2036 return err; 2037 } 2038 2039 static void __net_exit ndisc_net_exit(struct net *net) 2040 { 2041 inet_ctl_sock_destroy(net->ipv6.ndisc_sk); 2042 #ifdef CONFIG_SYSCTL 2043 neigh_sysctl_unregister(&nd_table(net)->parms); 2044 #endif 2045 neigh_table_unregister(net, NEIGH_ND_TABLE); 2046 } 2047 2048 static struct pernet_operations ndisc_net_ops = { 2049 .init = ndisc_net_init, 2050 .exit = ndisc_net_exit, 2051 }; 2052 2053 int __init ndisc_init(void) 2054 { 2055 return register_pernet_subsys(&ndisc_net_ops); 2056 } 2057 2058 int __init ndisc_late_init(void) 2059 { 2060 return register_netdevice_notifier(&ndisc_netdev_notifier); 2061 } 2062 2063 void ndisc_late_cleanup(void) 2064 { 2065 unregister_netdevice_notifier(&ndisc_netdev_notifier); 2066 } 2067 2068 void ndisc_cleanup(void) 2069 { 2070 unregister_pernet_subsys(&ndisc_net_ops); 2071 } 2072