1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPv6 input 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * Ian P. Morris <I.P.Morris@soton.ac.uk> 9 * 10 * Based in linux/net/ipv4/ip_input.c 11 */ 12 /* Changes 13 * 14 * Mitsuru KANDA @USAGI and 15 * YOSHIFUJI Hideaki @USAGI: Remove ipv6_parse_exthdrs(). 16 */ 17 18 #include <linux/errno.h> 19 #include <linux/types.h> 20 #include <linux/socket.h> 21 #include <linux/sockios.h> 22 #include <linux/net.h> 23 #include <linux/netdevice.h> 24 #include <linux/in6.h> 25 #include <linux/icmpv6.h> 26 #include <linux/mroute6.h> 27 #include <linux/slab.h> 28 #include <linux/indirect_call_wrapper.h> 29 30 #include <linux/netfilter.h> 31 #include <linux/netfilter_ipv6.h> 32 33 #include <net/sock.h> 34 #include <net/snmp.h> 35 36 #include <net/ipv6.h> 37 #include <net/protocol.h> 38 #include <net/transp_v6.h> 39 #include <net/rawv6.h> 40 #include <net/ndisc.h> 41 #include <net/ip6_route.h> 42 #include <net/addrconf.h> 43 #include <net/xfrm.h> 44 #include <net/inet_ecn.h> 45 #include <net/dst_metadata.h> 46 47 INDIRECT_CALLABLE_DECLARE(void udp_v6_early_demux(struct sk_buff *)); 48 INDIRECT_CALLABLE_DECLARE(void tcp_v6_early_demux(struct sk_buff *)); 49 static void ip6_rcv_finish_core(struct net *net, struct sock *sk, 50 struct sk_buff *skb) 51 { 52 void (*edemux)(struct sk_buff *skb); 53 54 if (net->ipv4.sysctl_ip_early_demux && !skb_dst(skb) && skb->sk == NULL) { 55 const struct inet6_protocol *ipprot; 56 57 ipprot = rcu_dereference(inet6_protos[ipv6_hdr(skb)->nexthdr]); 58 if (ipprot && (edemux = READ_ONCE(ipprot->early_demux))) 59 INDIRECT_CALL_2(edemux, tcp_v6_early_demux, 60 udp_v6_early_demux, skb); 61 } 62 if (!skb_valid_dst(skb)) 63 ip6_route_input(skb); 64 } 65 66 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 67 { 68 /* if ingress device is enslaved to an L3 master device pass the 69 * skb to its handler for processing 70 */ 71 skb = l3mdev_ip6_rcv(skb); 72 if (!skb) 73 return NET_RX_SUCCESS; 74 ip6_rcv_finish_core(net, sk, skb); 75 76 return dst_input(skb); 77 } 78 79 static void ip6_sublist_rcv_finish(struct list_head *head) 80 { 81 struct sk_buff *skb, *next; 82 83 list_for_each_entry_safe(skb, next, head, list) { 84 skb_list_del_init(skb); 85 dst_input(skb); 86 } 87 } 88 89 static bool ip6_can_use_hint(const struct sk_buff *skb, 90 const struct sk_buff *hint) 91 { 92 return hint && !skb_dst(skb) && 93 ipv6_addr_equal(&ipv6_hdr(hint)->daddr, &ipv6_hdr(skb)->daddr); 94 } 95 96 static struct sk_buff *ip6_extract_route_hint(const struct net *net, 97 struct sk_buff *skb) 98 { 99 if (fib6_routes_require_src(net) || fib6_has_custom_rules(net)) 100 return NULL; 101 102 return skb; 103 } 104 105 static void ip6_list_rcv_finish(struct net *net, struct sock *sk, 106 struct list_head *head) 107 { 108 struct sk_buff *skb, *next, *hint = NULL; 109 struct dst_entry *curr_dst = NULL; 110 struct list_head sublist; 111 112 INIT_LIST_HEAD(&sublist); 113 list_for_each_entry_safe(skb, next, head, list) { 114 struct dst_entry *dst; 115 116 skb_list_del_init(skb); 117 /* if ingress device is enslaved to an L3 master device pass the 118 * skb to its handler for processing 119 */ 120 skb = l3mdev_ip6_rcv(skb); 121 if (!skb) 122 continue; 123 124 if (ip6_can_use_hint(skb, hint)) 125 skb_dst_copy(skb, hint); 126 else 127 ip6_rcv_finish_core(net, sk, skb); 128 dst = skb_dst(skb); 129 if (curr_dst != dst) { 130 hint = ip6_extract_route_hint(net, skb); 131 132 /* dispatch old sublist */ 133 if (!list_empty(&sublist)) 134 ip6_sublist_rcv_finish(&sublist); 135 /* start new sublist */ 136 INIT_LIST_HEAD(&sublist); 137 curr_dst = dst; 138 } 139 list_add_tail(&skb->list, &sublist); 140 } 141 /* dispatch final sublist */ 142 ip6_sublist_rcv_finish(&sublist); 143 } 144 145 static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev, 146 struct net *net) 147 { 148 const struct ipv6hdr *hdr; 149 u32 pkt_len; 150 struct inet6_dev *idev; 151 152 if (skb->pkt_type == PACKET_OTHERHOST) { 153 kfree_skb(skb); 154 return NULL; 155 } 156 157 rcu_read_lock(); 158 159 idev = __in6_dev_get(skb->dev); 160 161 __IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len); 162 163 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL || 164 !idev || unlikely(idev->cnf.disable_ipv6)) { 165 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); 166 goto drop; 167 } 168 169 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); 170 171 /* 172 * Store incoming device index. When the packet will 173 * be queued, we cannot refer to skb->dev anymore. 174 * 175 * BTW, when we send a packet for our own local address on a 176 * non-loopback interface (e.g. ethX), it is being delivered 177 * via the loopback interface (lo) here; skb->dev = loopback_dev. 178 * It, however, should be considered as if it is being 179 * arrived via the sending interface (ethX), because of the 180 * nature of scoping architecture. --yoshfuji 181 */ 182 IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex; 183 184 if (unlikely(!pskb_may_pull(skb, sizeof(*hdr)))) 185 goto err; 186 187 hdr = ipv6_hdr(skb); 188 189 if (hdr->version != 6) 190 goto err; 191 192 __IP6_ADD_STATS(net, idev, 193 IPSTATS_MIB_NOECTPKTS + 194 (ipv6_get_dsfield(hdr) & INET_ECN_MASK), 195 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs)); 196 /* 197 * RFC4291 2.5.3 198 * The loopback address must not be used as the source address in IPv6 199 * packets that are sent outside of a single node. [..] 200 * A packet received on an interface with a destination address 201 * of loopback must be dropped. 202 */ 203 if ((ipv6_addr_loopback(&hdr->saddr) || 204 ipv6_addr_loopback(&hdr->daddr)) && 205 !(dev->flags & IFF_LOOPBACK) && 206 !netif_is_l3_master(dev)) 207 goto err; 208 209 /* RFC4291 Errata ID: 3480 210 * Interface-Local scope spans only a single interface on a 211 * node and is useful only for loopback transmission of 212 * multicast. Packets with interface-local scope received 213 * from another node must be discarded. 214 */ 215 if (!(skb->pkt_type == PACKET_LOOPBACK || 216 dev->flags & IFF_LOOPBACK) && 217 ipv6_addr_is_multicast(&hdr->daddr) && 218 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1) 219 goto err; 220 221 /* If enabled, drop unicast packets that were encapsulated in link-layer 222 * multicast or broadcast to protected against the so-called "hole-196" 223 * attack in 802.11 wireless. 224 */ 225 if (!ipv6_addr_is_multicast(&hdr->daddr) && 226 (skb->pkt_type == PACKET_BROADCAST || 227 skb->pkt_type == PACKET_MULTICAST) && 228 idev->cnf.drop_unicast_in_l2_multicast) 229 goto err; 230 231 /* RFC4291 2.7 232 * Nodes must not originate a packet to a multicast address whose scope 233 * field contains the reserved value 0; if such a packet is received, it 234 * must be silently dropped. 235 */ 236 if (ipv6_addr_is_multicast(&hdr->daddr) && 237 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0) 238 goto err; 239 240 /* 241 * RFC4291 2.7 242 * Multicast addresses must not be used as source addresses in IPv6 243 * packets or appear in any Routing header. 244 */ 245 if (ipv6_addr_is_multicast(&hdr->saddr)) 246 goto err; 247 248 /* While RFC4291 is not explicit about v4mapped addresses 249 * in IPv6 headers, it seems clear linux dual-stack 250 * model can not deal properly with these. 251 * Security models could be fooled by ::ffff:127.0.0.1 for example. 252 * 253 * https://tools.ietf.org/html/draft-itojun-v6ops-v4mapped-harmful-02 254 */ 255 if (ipv6_addr_v4mapped(&hdr->saddr)) 256 goto err; 257 258 skb->transport_header = skb->network_header + sizeof(*hdr); 259 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr); 260 261 pkt_len = ntohs(hdr->payload_len); 262 263 /* pkt_len may be zero if Jumbo payload option is present */ 264 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) { 265 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) { 266 __IP6_INC_STATS(net, 267 idev, IPSTATS_MIB_INTRUNCATEDPKTS); 268 goto drop; 269 } 270 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) { 271 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); 272 goto drop; 273 } 274 hdr = ipv6_hdr(skb); 275 } 276 277 if (hdr->nexthdr == NEXTHDR_HOP) { 278 if (ipv6_parse_hopopts(skb) < 0) { 279 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); 280 rcu_read_unlock(); 281 return NULL; 282 } 283 } 284 285 rcu_read_unlock(); 286 287 /* Must drop socket now because of tproxy. */ 288 if (!skb_sk_is_prefetched(skb)) 289 skb_orphan(skb); 290 291 return skb; 292 err: 293 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); 294 drop: 295 rcu_read_unlock(); 296 kfree_skb(skb); 297 return NULL; 298 } 299 300 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) 301 { 302 struct net *net = dev_net(skb->dev); 303 304 skb = ip6_rcv_core(skb, dev, net); 305 if (skb == NULL) 306 return NET_RX_DROP; 307 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, 308 net, NULL, skb, dev, NULL, 309 ip6_rcv_finish); 310 } 311 312 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev, 313 struct net *net) 314 { 315 NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL, 316 head, dev, NULL, ip6_rcv_finish); 317 ip6_list_rcv_finish(net, NULL, head); 318 } 319 320 /* Receive a list of IPv6 packets */ 321 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt, 322 struct net_device *orig_dev) 323 { 324 struct net_device *curr_dev = NULL; 325 struct net *curr_net = NULL; 326 struct sk_buff *skb, *next; 327 struct list_head sublist; 328 329 INIT_LIST_HEAD(&sublist); 330 list_for_each_entry_safe(skb, next, head, list) { 331 struct net_device *dev = skb->dev; 332 struct net *net = dev_net(dev); 333 334 skb_list_del_init(skb); 335 skb = ip6_rcv_core(skb, dev, net); 336 if (skb == NULL) 337 continue; 338 339 if (curr_dev != dev || curr_net != net) { 340 /* dispatch old sublist */ 341 if (!list_empty(&sublist)) 342 ip6_sublist_rcv(&sublist, curr_dev, curr_net); 343 /* start new sublist */ 344 INIT_LIST_HEAD(&sublist); 345 curr_dev = dev; 346 curr_net = net; 347 } 348 list_add_tail(&skb->list, &sublist); 349 } 350 /* dispatch final sublist */ 351 if (!list_empty(&sublist)) 352 ip6_sublist_rcv(&sublist, curr_dev, curr_net); 353 } 354 355 INDIRECT_CALLABLE_DECLARE(int udpv6_rcv(struct sk_buff *)); 356 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *)); 357 358 /* 359 * Deliver the packet to the host 360 */ 361 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr, 362 bool have_final) 363 { 364 const struct inet6_protocol *ipprot; 365 struct inet6_dev *idev; 366 unsigned int nhoff; 367 bool raw; 368 369 /* 370 * Parse extension headers 371 */ 372 373 resubmit: 374 idev = ip6_dst_idev(skb_dst(skb)); 375 nhoff = IP6CB(skb)->nhoff; 376 if (!have_final) { 377 if (!pskb_pull(skb, skb_transport_offset(skb))) 378 goto discard; 379 nexthdr = skb_network_header(skb)[nhoff]; 380 } 381 382 resubmit_final: 383 raw = raw6_local_deliver(skb, nexthdr); 384 ipprot = rcu_dereference(inet6_protos[nexthdr]); 385 if (ipprot) { 386 int ret; 387 388 if (have_final) { 389 if (!(ipprot->flags & INET6_PROTO_FINAL)) { 390 /* Once we've seen a final protocol don't 391 * allow encapsulation on any non-final 392 * ones. This allows foo in UDP encapsulation 393 * to work. 394 */ 395 goto discard; 396 } 397 } else if (ipprot->flags & INET6_PROTO_FINAL) { 398 const struct ipv6hdr *hdr; 399 int sdif = inet6_sdif(skb); 400 struct net_device *dev; 401 402 /* Only do this once for first final protocol */ 403 have_final = true; 404 405 /* Free reference early: we don't need it any more, 406 and it may hold ip_conntrack module loaded 407 indefinitely. */ 408 nf_reset_ct(skb); 409 410 skb_postpull_rcsum(skb, skb_network_header(skb), 411 skb_network_header_len(skb)); 412 hdr = ipv6_hdr(skb); 413 414 /* skb->dev passed may be master dev for vrfs. */ 415 if (sdif) { 416 dev = dev_get_by_index_rcu(net, sdif); 417 if (!dev) 418 goto discard; 419 } else { 420 dev = skb->dev; 421 } 422 423 if (ipv6_addr_is_multicast(&hdr->daddr) && 424 !ipv6_chk_mcast_addr(dev, &hdr->daddr, 425 &hdr->saddr) && 426 !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb))) 427 goto discard; 428 } 429 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) && 430 !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 431 goto discard; 432 433 ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv, 434 skb); 435 if (ret > 0) { 436 if (ipprot->flags & INET6_PROTO_FINAL) { 437 /* Not an extension header, most likely UDP 438 * encapsulation. Use return value as nexthdr 439 * protocol not nhoff (which presumably is 440 * not set by handler). 441 */ 442 nexthdr = ret; 443 goto resubmit_final; 444 } else { 445 goto resubmit; 446 } 447 } else if (ret == 0) { 448 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS); 449 } 450 } else { 451 if (!raw) { 452 if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 453 __IP6_INC_STATS(net, idev, 454 IPSTATS_MIB_INUNKNOWNPROTOS); 455 icmpv6_send(skb, ICMPV6_PARAMPROB, 456 ICMPV6_UNK_NEXTHDR, nhoff); 457 } 458 kfree_skb(skb); 459 } else { 460 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS); 461 consume_skb(skb); 462 } 463 } 464 return; 465 466 discard: 467 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); 468 kfree_skb(skb); 469 } 470 471 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 472 { 473 rcu_read_lock(); 474 ip6_protocol_deliver_rcu(net, skb, 0, false); 475 rcu_read_unlock(); 476 477 return 0; 478 } 479 480 481 int ip6_input(struct sk_buff *skb) 482 { 483 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN, 484 dev_net(skb->dev), NULL, skb, skb->dev, NULL, 485 ip6_input_finish); 486 } 487 EXPORT_SYMBOL_GPL(ip6_input); 488 489 int ip6_mc_input(struct sk_buff *skb) 490 { 491 int sdif = inet6_sdif(skb); 492 const struct ipv6hdr *hdr; 493 struct net_device *dev; 494 bool deliver; 495 496 __IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev), 497 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST, 498 skb->len); 499 500 /* skb->dev passed may be master dev for vrfs. */ 501 if (sdif) { 502 rcu_read_lock(); 503 dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif); 504 if (!dev) { 505 rcu_read_unlock(); 506 kfree_skb(skb); 507 return -ENODEV; 508 } 509 } else { 510 dev = skb->dev; 511 } 512 513 hdr = ipv6_hdr(skb); 514 deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL); 515 if (sdif) 516 rcu_read_unlock(); 517 518 #ifdef CONFIG_IPV6_MROUTE 519 /* 520 * IPv6 multicast router mode is now supported ;) 521 */ 522 if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding && 523 !(ipv6_addr_type(&hdr->daddr) & 524 (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) && 525 likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) { 526 /* 527 * Okay, we try to forward - split and duplicate 528 * packets. 529 */ 530 struct sk_buff *skb2; 531 struct inet6_skb_parm *opt = IP6CB(skb); 532 533 /* Check for MLD */ 534 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) { 535 /* Check if this is a mld message */ 536 u8 nexthdr = hdr->nexthdr; 537 __be16 frag_off; 538 int offset; 539 540 /* Check if the value of Router Alert 541 * is for MLD (0x0000). 542 */ 543 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) { 544 deliver = false; 545 546 if (!ipv6_ext_hdr(nexthdr)) { 547 /* BUG */ 548 goto out; 549 } 550 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), 551 &nexthdr, &frag_off); 552 if (offset < 0) 553 goto out; 554 555 if (ipv6_is_mld(skb, nexthdr, offset)) 556 deliver = true; 557 558 goto out; 559 } 560 /* unknown RA - process it normally */ 561 } 562 563 if (deliver) 564 skb2 = skb_clone(skb, GFP_ATOMIC); 565 else { 566 skb2 = skb; 567 skb = NULL; 568 } 569 570 if (skb2) { 571 ip6_mr_input(skb2); 572 } 573 } 574 out: 575 #endif 576 if (likely(deliver)) 577 ip6_input(skb); 578 else { 579 /* discard */ 580 kfree_skb(skb); 581 } 582 583 return 0; 584 } 585