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 #include <net/udp.h> 36 37 #include <net/ipv6.h> 38 #include <net/protocol.h> 39 #include <net/transp_v6.h> 40 #include <net/rawv6.h> 41 #include <net/ndisc.h> 42 #include <net/ip6_route.h> 43 #include <net/addrconf.h> 44 #include <net/xfrm.h> 45 #include <net/inet_ecn.h> 46 #include <net/dst_metadata.h> 47 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 dev_core_stats_rx_otherhost_dropped_inc(skb->dev); 154 kfree_skb(skb); 155 return NULL; 156 } 157 158 rcu_read_lock(); 159 160 idev = __in6_dev_get(skb->dev); 161 162 __IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len); 163 164 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL || 165 !idev || unlikely(idev->cnf.disable_ipv6)) { 166 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); 167 goto drop; 168 } 169 170 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); 171 172 /* 173 * Store incoming device index. When the packet will 174 * be queued, we cannot refer to skb->dev anymore. 175 * 176 * BTW, when we send a packet for our own local address on a 177 * non-loopback interface (e.g. ethX), it is being delivered 178 * via the loopback interface (lo) here; skb->dev = loopback_dev. 179 * It, however, should be considered as if it is being 180 * arrived via the sending interface (ethX), because of the 181 * nature of scoping architecture. --yoshfuji 182 */ 183 IP6CB(skb)->iif = skb_valid_dst(skb) ? ip6_dst_idev(skb_dst(skb))->dev->ifindex : dev->ifindex; 184 185 if (unlikely(!pskb_may_pull(skb, sizeof(*hdr)))) 186 goto err; 187 188 hdr = ipv6_hdr(skb); 189 190 if (hdr->version != 6) 191 goto err; 192 193 __IP6_ADD_STATS(net, idev, 194 IPSTATS_MIB_NOECTPKTS + 195 (ipv6_get_dsfield(hdr) & INET_ECN_MASK), 196 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs)); 197 /* 198 * RFC4291 2.5.3 199 * The loopback address must not be used as the source address in IPv6 200 * packets that are sent outside of a single node. [..] 201 * A packet received on an interface with a destination address 202 * of loopback must be dropped. 203 */ 204 if ((ipv6_addr_loopback(&hdr->saddr) || 205 ipv6_addr_loopback(&hdr->daddr)) && 206 !(dev->flags & IFF_LOOPBACK) && 207 !netif_is_l3_master(dev)) 208 goto err; 209 210 /* RFC4291 Errata ID: 3480 211 * Interface-Local scope spans only a single interface on a 212 * node and is useful only for loopback transmission of 213 * multicast. Packets with interface-local scope received 214 * from another node must be discarded. 215 */ 216 if (!(skb->pkt_type == PACKET_LOOPBACK || 217 dev->flags & IFF_LOOPBACK) && 218 ipv6_addr_is_multicast(&hdr->daddr) && 219 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1) 220 goto err; 221 222 /* If enabled, drop unicast packets that were encapsulated in link-layer 223 * multicast or broadcast to protected against the so-called "hole-196" 224 * attack in 802.11 wireless. 225 */ 226 if (!ipv6_addr_is_multicast(&hdr->daddr) && 227 (skb->pkt_type == PACKET_BROADCAST || 228 skb->pkt_type == PACKET_MULTICAST) && 229 idev->cnf.drop_unicast_in_l2_multicast) 230 goto err; 231 232 /* RFC4291 2.7 233 * Nodes must not originate a packet to a multicast address whose scope 234 * field contains the reserved value 0; if such a packet is received, it 235 * must be silently dropped. 236 */ 237 if (ipv6_addr_is_multicast(&hdr->daddr) && 238 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0) 239 goto err; 240 241 /* 242 * RFC4291 2.7 243 * Multicast addresses must not be used as source addresses in IPv6 244 * packets or appear in any Routing header. 245 */ 246 if (ipv6_addr_is_multicast(&hdr->saddr)) 247 goto err; 248 249 skb->transport_header = skb->network_header + sizeof(*hdr); 250 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr); 251 252 pkt_len = ntohs(hdr->payload_len); 253 254 /* pkt_len may be zero if Jumbo payload option is present */ 255 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) { 256 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) { 257 __IP6_INC_STATS(net, 258 idev, IPSTATS_MIB_INTRUNCATEDPKTS); 259 goto drop; 260 } 261 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) { 262 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); 263 goto drop; 264 } 265 hdr = ipv6_hdr(skb); 266 } 267 268 if (hdr->nexthdr == NEXTHDR_HOP) { 269 if (ipv6_parse_hopopts(skb) < 0) { 270 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); 271 rcu_read_unlock(); 272 return NULL; 273 } 274 } 275 276 rcu_read_unlock(); 277 278 /* Must drop socket now because of tproxy. */ 279 if (!skb_sk_is_prefetched(skb)) 280 skb_orphan(skb); 281 282 return skb; 283 err: 284 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); 285 drop: 286 rcu_read_unlock(); 287 kfree_skb(skb); 288 return NULL; 289 } 290 291 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) 292 { 293 struct net *net = dev_net(skb->dev); 294 295 skb = ip6_rcv_core(skb, dev, net); 296 if (skb == NULL) 297 return NET_RX_DROP; 298 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, 299 net, NULL, skb, dev, NULL, 300 ip6_rcv_finish); 301 } 302 303 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev, 304 struct net *net) 305 { 306 NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL, 307 head, dev, NULL, ip6_rcv_finish); 308 ip6_list_rcv_finish(net, NULL, head); 309 } 310 311 /* Receive a list of IPv6 packets */ 312 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt, 313 struct net_device *orig_dev) 314 { 315 struct net_device *curr_dev = NULL; 316 struct net *curr_net = NULL; 317 struct sk_buff *skb, *next; 318 struct list_head sublist; 319 320 INIT_LIST_HEAD(&sublist); 321 list_for_each_entry_safe(skb, next, head, list) { 322 struct net_device *dev = skb->dev; 323 struct net *net = dev_net(dev); 324 325 skb_list_del_init(skb); 326 skb = ip6_rcv_core(skb, dev, net); 327 if (skb == NULL) 328 continue; 329 330 if (curr_dev != dev || curr_net != net) { 331 /* dispatch old sublist */ 332 if (!list_empty(&sublist)) 333 ip6_sublist_rcv(&sublist, curr_dev, curr_net); 334 /* start new sublist */ 335 INIT_LIST_HEAD(&sublist); 336 curr_dev = dev; 337 curr_net = net; 338 } 339 list_add_tail(&skb->list, &sublist); 340 } 341 /* dispatch final sublist */ 342 if (!list_empty(&sublist)) 343 ip6_sublist_rcv(&sublist, curr_dev, curr_net); 344 } 345 346 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *)); 347 348 /* 349 * Deliver the packet to the host 350 */ 351 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr, 352 bool have_final) 353 { 354 const struct inet6_protocol *ipprot; 355 struct inet6_dev *idev; 356 unsigned int nhoff; 357 bool raw; 358 359 /* 360 * Parse extension headers 361 */ 362 363 resubmit: 364 idev = ip6_dst_idev(skb_dst(skb)); 365 nhoff = IP6CB(skb)->nhoff; 366 if (!have_final) { 367 if (!pskb_pull(skb, skb_transport_offset(skb))) 368 goto discard; 369 nexthdr = skb_network_header(skb)[nhoff]; 370 } 371 372 resubmit_final: 373 raw = raw6_local_deliver(skb, nexthdr); 374 ipprot = rcu_dereference(inet6_protos[nexthdr]); 375 if (ipprot) { 376 int ret; 377 378 if (have_final) { 379 if (!(ipprot->flags & INET6_PROTO_FINAL)) { 380 /* Once we've seen a final protocol don't 381 * allow encapsulation on any non-final 382 * ones. This allows foo in UDP encapsulation 383 * to work. 384 */ 385 goto discard; 386 } 387 } else if (ipprot->flags & INET6_PROTO_FINAL) { 388 const struct ipv6hdr *hdr; 389 int sdif = inet6_sdif(skb); 390 struct net_device *dev; 391 392 /* Only do this once for first final protocol */ 393 have_final = true; 394 395 /* Free reference early: we don't need it any more, 396 and it may hold ip_conntrack module loaded 397 indefinitely. */ 398 nf_reset_ct(skb); 399 400 skb_postpull_rcsum(skb, skb_network_header(skb), 401 skb_network_header_len(skb)); 402 hdr = ipv6_hdr(skb); 403 404 /* skb->dev passed may be master dev for vrfs. */ 405 if (sdif) { 406 dev = dev_get_by_index_rcu(net, sdif); 407 if (!dev) 408 goto discard; 409 } else { 410 dev = skb->dev; 411 } 412 413 if (ipv6_addr_is_multicast(&hdr->daddr) && 414 !ipv6_chk_mcast_addr(dev, &hdr->daddr, 415 &hdr->saddr) && 416 !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb))) 417 goto discard; 418 } 419 if (!(ipprot->flags & INET6_PROTO_NOPOLICY) && 420 !xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 421 goto discard; 422 423 ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv, 424 skb); 425 if (ret > 0) { 426 if (ipprot->flags & INET6_PROTO_FINAL) { 427 /* Not an extension header, most likely UDP 428 * encapsulation. Use return value as nexthdr 429 * protocol not nhoff (which presumably is 430 * not set by handler). 431 */ 432 nexthdr = ret; 433 goto resubmit_final; 434 } else { 435 goto resubmit; 436 } 437 } else if (ret == 0) { 438 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS); 439 } 440 } else { 441 if (!raw) { 442 if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 443 __IP6_INC_STATS(net, idev, 444 IPSTATS_MIB_INUNKNOWNPROTOS); 445 icmpv6_send(skb, ICMPV6_PARAMPROB, 446 ICMPV6_UNK_NEXTHDR, nhoff); 447 } 448 kfree_skb(skb); 449 } else { 450 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS); 451 consume_skb(skb); 452 } 453 } 454 return; 455 456 discard: 457 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); 458 kfree_skb(skb); 459 } 460 461 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 462 { 463 skb_clear_delivery_time(skb); 464 rcu_read_lock(); 465 ip6_protocol_deliver_rcu(net, skb, 0, false); 466 rcu_read_unlock(); 467 468 return 0; 469 } 470 471 472 int ip6_input(struct sk_buff *skb) 473 { 474 return NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN, 475 dev_net(skb->dev), NULL, skb, skb->dev, NULL, 476 ip6_input_finish); 477 } 478 EXPORT_SYMBOL_GPL(ip6_input); 479 480 int ip6_mc_input(struct sk_buff *skb) 481 { 482 int sdif = inet6_sdif(skb); 483 const struct ipv6hdr *hdr; 484 struct net_device *dev; 485 bool deliver; 486 487 __IP6_UPD_PO_STATS(dev_net(skb_dst(skb)->dev), 488 __in6_dev_get_safely(skb->dev), IPSTATS_MIB_INMCAST, 489 skb->len); 490 491 /* skb->dev passed may be master dev for vrfs. */ 492 if (sdif) { 493 rcu_read_lock(); 494 dev = dev_get_by_index_rcu(dev_net(skb->dev), sdif); 495 if (!dev) { 496 rcu_read_unlock(); 497 kfree_skb(skb); 498 return -ENODEV; 499 } 500 } else { 501 dev = skb->dev; 502 } 503 504 hdr = ipv6_hdr(skb); 505 deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL); 506 if (sdif) 507 rcu_read_unlock(); 508 509 #ifdef CONFIG_IPV6_MROUTE 510 /* 511 * IPv6 multicast router mode is now supported ;) 512 */ 513 if (atomic_read(&dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding) && 514 !(ipv6_addr_type(&hdr->daddr) & 515 (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) && 516 likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) { 517 /* 518 * Okay, we try to forward - split and duplicate 519 * packets. 520 */ 521 struct sk_buff *skb2; 522 struct inet6_skb_parm *opt = IP6CB(skb); 523 524 /* Check for MLD */ 525 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) { 526 /* Check if this is a mld message */ 527 u8 nexthdr = hdr->nexthdr; 528 __be16 frag_off; 529 int offset; 530 531 /* Check if the value of Router Alert 532 * is for MLD (0x0000). 533 */ 534 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) { 535 deliver = false; 536 537 if (!ipv6_ext_hdr(nexthdr)) { 538 /* BUG */ 539 goto out; 540 } 541 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), 542 &nexthdr, &frag_off); 543 if (offset < 0) 544 goto out; 545 546 if (ipv6_is_mld(skb, nexthdr, offset)) 547 deliver = true; 548 549 goto out; 550 } 551 /* unknown RA - process it normally */ 552 } 553 554 if (deliver) 555 skb2 = skb_clone(skb, GFP_ATOMIC); 556 else { 557 skb2 = skb; 558 skb = NULL; 559 } 560 561 if (skb2) { 562 ip6_mr_input(skb2); 563 } 564 } 565 out: 566 #endif 567 if (likely(deliver)) 568 ip6_input(skb); 569 else { 570 /* discard */ 571 kfree_skb(skb); 572 } 573 574 return 0; 575 } 576