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 static void ip6_rcv_finish_core(struct net *net, struct sock *sk, 49 struct sk_buff *skb) 50 { 51 if (READ_ONCE(net->ipv4.sysctl_ip_early_demux) && 52 !skb_dst(skb) && !skb->sk) { 53 switch (ipv6_hdr(skb)->nexthdr) { 54 case IPPROTO_TCP: 55 if (READ_ONCE(net->ipv4.sysctl_tcp_early_demux)) 56 tcp_v6_early_demux(skb); 57 break; 58 case IPPROTO_UDP: 59 if (READ_ONCE(net->ipv4.sysctl_udp_early_demux)) 60 udp_v6_early_demux(skb); 61 break; 62 } 63 } 64 65 if (!skb_valid_dst(skb)) 66 ip6_route_input(skb); 67 } 68 69 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 70 { 71 /* if ingress device is enslaved to an L3 master device pass the 72 * skb to its handler for processing 73 */ 74 skb = l3mdev_ip6_rcv(skb); 75 if (!skb) 76 return NET_RX_SUCCESS; 77 ip6_rcv_finish_core(net, sk, skb); 78 79 return dst_input(skb); 80 } 81 82 static void ip6_sublist_rcv_finish(struct list_head *head) 83 { 84 struct sk_buff *skb, *next; 85 86 list_for_each_entry_safe(skb, next, head, list) { 87 skb_list_del_init(skb); 88 dst_input(skb); 89 } 90 } 91 92 static bool ip6_can_use_hint(const struct sk_buff *skb, 93 const struct sk_buff *hint) 94 { 95 return hint && !skb_dst(skb) && 96 ipv6_addr_equal(&ipv6_hdr(hint)->daddr, &ipv6_hdr(skb)->daddr); 97 } 98 99 static struct sk_buff *ip6_extract_route_hint(const struct net *net, 100 struct sk_buff *skb) 101 { 102 if (fib6_routes_require_src(net) || fib6_has_custom_rules(net) || 103 IP6CB(skb)->flags & IP6SKB_MULTIPATH) 104 return NULL; 105 106 return skb; 107 } 108 109 static void ip6_list_rcv_finish(struct net *net, struct sock *sk, 110 struct list_head *head) 111 { 112 struct sk_buff *skb, *next, *hint = NULL; 113 struct dst_entry *curr_dst = NULL; 114 LIST_HEAD(sublist); 115 116 list_for_each_entry_safe(skb, next, head, list) { 117 struct dst_entry *dst; 118 119 skb_list_del_init(skb); 120 /* if ingress device is enslaved to an L3 master device pass the 121 * skb to its handler for processing 122 */ 123 skb = l3mdev_ip6_rcv(skb); 124 if (!skb) 125 continue; 126 127 if (ip6_can_use_hint(skb, hint)) 128 skb_dst_copy(skb, hint); 129 else 130 ip6_rcv_finish_core(net, sk, skb); 131 dst = skb_dst(skb); 132 if (curr_dst != dst) { 133 hint = ip6_extract_route_hint(net, skb); 134 135 /* dispatch old sublist */ 136 if (!list_empty(&sublist)) 137 ip6_sublist_rcv_finish(&sublist); 138 /* start new sublist */ 139 INIT_LIST_HEAD(&sublist); 140 curr_dst = dst; 141 } 142 list_add_tail(&skb->list, &sublist); 143 } 144 /* dispatch final sublist */ 145 ip6_sublist_rcv_finish(&sublist); 146 } 147 148 static struct sk_buff *ip6_rcv_core(struct sk_buff *skb, struct net_device *dev, 149 struct net *net) 150 { 151 enum skb_drop_reason reason; 152 const struct ipv6hdr *hdr; 153 u32 pkt_len; 154 struct inet6_dev *idev; 155 156 if (skb->pkt_type == PACKET_OTHERHOST) { 157 dev_core_stats_rx_otherhost_dropped_inc(skb->dev); 158 kfree_skb_reason(skb, SKB_DROP_REASON_OTHERHOST); 159 return NULL; 160 } 161 162 rcu_read_lock(); 163 164 idev = __in6_dev_get(skb->dev); 165 166 __IP6_UPD_PO_STATS(net, idev, IPSTATS_MIB_IN, skb->len); 167 168 SKB_DR_SET(reason, NOT_SPECIFIED); 169 if ((skb = skb_share_check(skb, GFP_ATOMIC)) == NULL || 170 !idev || unlikely(READ_ONCE(idev->cnf.disable_ipv6))) { 171 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); 172 if (idev && unlikely(READ_ONCE(idev->cnf.disable_ipv6))) 173 SKB_DR_SET(reason, IPV6DISABLED); 174 goto drop; 175 } 176 177 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm)); 178 179 /* 180 * Store incoming device index. When the packet will 181 * be queued, we cannot refer to skb->dev anymore. 182 * 183 * BTW, when we send a packet for our own local address on a 184 * non-loopback interface (e.g. ethX), it is being delivered 185 * via the loopback interface (lo) here; skb->dev = loopback_dev. 186 * It, however, should be considered as if it is being 187 * arrived via the sending interface (ethX), because of the 188 * nature of scoping architecture. --yoshfuji 189 */ 190 IP6CB(skb)->iif = skb_valid_dst(skb) ? 191 ip6_dst_idev(skb_dst(skb))->dev->ifindex : 192 dev->ifindex; 193 194 if (unlikely(!pskb_may_pull(skb, sizeof(*hdr)))) 195 goto err; 196 197 hdr = ipv6_hdr(skb); 198 199 if (hdr->version != 6) { 200 SKB_DR_SET(reason, UNHANDLED_PROTO); 201 goto err; 202 } 203 204 __IP6_ADD_STATS(net, idev, 205 IPSTATS_MIB_NOECTPKTS + 206 (ipv6_get_dsfield(hdr) & INET_ECN_MASK), 207 max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs)); 208 /* 209 * RFC4291 2.5.3 210 * The loopback address must not be used as the source address in IPv6 211 * packets that are sent outside of a single node. [..] 212 * A packet received on an interface with a destination address 213 * of loopback must be dropped. 214 */ 215 if ((ipv6_addr_loopback(&hdr->saddr) || 216 ipv6_addr_loopback(&hdr->daddr)) && 217 !(dev->flags & IFF_LOOPBACK) && 218 !netif_is_l3_master(dev)) 219 goto err; 220 221 /* RFC4291 Errata ID: 3480 222 * Interface-Local scope spans only a single interface on a 223 * node and is useful only for loopback transmission of 224 * multicast. Packets with interface-local scope received 225 * from another node must be discarded. 226 */ 227 if (!(skb->pkt_type == PACKET_LOOPBACK || 228 dev->flags & IFF_LOOPBACK) && 229 ipv6_addr_is_multicast(&hdr->daddr) && 230 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 1) 231 goto err; 232 233 /* If enabled, drop unicast packets that were encapsulated in link-layer 234 * multicast or broadcast to protected against the so-called "hole-196" 235 * attack in 802.11 wireless. 236 */ 237 if (!ipv6_addr_is_multicast(&hdr->daddr) && 238 (skb->pkt_type == PACKET_BROADCAST || 239 skb->pkt_type == PACKET_MULTICAST) && 240 READ_ONCE(idev->cnf.drop_unicast_in_l2_multicast)) { 241 SKB_DR_SET(reason, UNICAST_IN_L2_MULTICAST); 242 goto err; 243 } 244 245 /* RFC4291 2.7 246 * Nodes must not originate a packet to a multicast address whose scope 247 * field contains the reserved value 0; if such a packet is received, it 248 * must be silently dropped. 249 */ 250 if (ipv6_addr_is_multicast(&hdr->daddr) && 251 IPV6_ADDR_MC_SCOPE(&hdr->daddr) == 0) 252 goto err; 253 254 /* 255 * RFC4291 2.7 256 * Multicast addresses must not be used as source addresses in IPv6 257 * packets or appear in any Routing header. 258 */ 259 if (ipv6_addr_is_multicast(&hdr->saddr)) 260 goto err; 261 262 skb->transport_header = skb->network_header + sizeof(*hdr); 263 IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr); 264 265 pkt_len = ntohs(hdr->payload_len); 266 267 /* pkt_len may be zero if Jumbo payload option is present */ 268 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) { 269 if (pkt_len + sizeof(struct ipv6hdr) > skb->len) { 270 __IP6_INC_STATS(net, 271 idev, IPSTATS_MIB_INTRUNCATEDPKTS); 272 SKB_DR_SET(reason, PKT_TOO_SMALL); 273 goto drop; 274 } 275 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr))) 276 goto err; 277 hdr = ipv6_hdr(skb); 278 } 279 280 if (hdr->nexthdr == NEXTHDR_HOP) { 281 if (ipv6_parse_hopopts(skb) < 0) { 282 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); 283 rcu_read_unlock(); 284 return NULL; 285 } 286 } 287 288 rcu_read_unlock(); 289 290 /* Must drop socket now because of tproxy. */ 291 if (!skb_sk_is_prefetched(skb)) 292 skb_orphan(skb); 293 294 return skb; 295 err: 296 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INHDRERRORS); 297 SKB_DR_OR(reason, IP_INHDR); 298 drop: 299 rcu_read_unlock(); 300 kfree_skb_reason(skb, reason); 301 return NULL; 302 } 303 304 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev) 305 { 306 struct net *net = dev_net(skb->dev); 307 308 skb = ip6_rcv_core(skb, dev, net); 309 if (skb == NULL) 310 return NET_RX_DROP; 311 return NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, 312 net, NULL, skb, dev, NULL, 313 ip6_rcv_finish); 314 } 315 316 static void ip6_sublist_rcv(struct list_head *head, struct net_device *dev, 317 struct net *net) 318 { 319 NF_HOOK_LIST(NFPROTO_IPV6, NF_INET_PRE_ROUTING, net, NULL, 320 head, dev, NULL, ip6_rcv_finish); 321 ip6_list_rcv_finish(net, NULL, head); 322 } 323 324 /* Receive a list of IPv6 packets */ 325 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt, 326 struct net_device *orig_dev) 327 { 328 struct net_device *curr_dev = NULL; 329 struct net *curr_net = NULL; 330 struct sk_buff *skb, *next; 331 LIST_HEAD(sublist); 332 333 list_for_each_entry_safe(skb, next, head, list) { 334 struct net_device *dev = skb->dev; 335 struct net *net = dev_net(dev); 336 337 skb_list_del_init(skb); 338 skb = ip6_rcv_core(skb, dev, net); 339 if (skb == NULL) 340 continue; 341 342 if (curr_dev != dev || curr_net != net) { 343 /* dispatch old sublist */ 344 if (!list_empty(&sublist)) 345 ip6_sublist_rcv(&sublist, curr_dev, curr_net); 346 /* start new sublist */ 347 INIT_LIST_HEAD(&sublist); 348 curr_dev = dev; 349 curr_net = net; 350 } 351 list_add_tail(&skb->list, &sublist); 352 } 353 /* dispatch final sublist */ 354 if (!list_empty(&sublist)) 355 ip6_sublist_rcv(&sublist, curr_dev, curr_net); 356 } 357 358 INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *)); 359 360 /* 361 * Deliver the packet to the host 362 */ 363 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr, 364 bool have_final) 365 { 366 const struct inet6_protocol *ipprot; 367 struct inet6_dev *idev; 368 unsigned int nhoff; 369 SKB_DR(reason); 370 bool raw; 371 372 /* 373 * Parse extension headers 374 */ 375 376 resubmit: 377 idev = ip6_dst_idev(skb_dst(skb)); 378 nhoff = IP6CB(skb)->nhoff; 379 if (!have_final) { 380 if (!pskb_pull(skb, skb_transport_offset(skb))) 381 goto discard; 382 nexthdr = skb_network_header(skb)[nhoff]; 383 } 384 385 resubmit_final: 386 raw = raw6_local_deliver(skb, nexthdr); 387 ipprot = rcu_dereference(inet6_protos[nexthdr]); 388 if (ipprot) { 389 int ret; 390 391 if (have_final) { 392 if (!(ipprot->flags & INET6_PROTO_FINAL)) { 393 /* Once we've seen a final protocol don't 394 * allow encapsulation on any non-final 395 * ones. This allows foo in UDP encapsulation 396 * to work. 397 */ 398 goto discard; 399 } 400 } else if (ipprot->flags & INET6_PROTO_FINAL) { 401 const struct ipv6hdr *hdr; 402 int sdif = inet6_sdif(skb); 403 struct net_device *dev; 404 405 /* Only do this once for first final protocol */ 406 have_final = true; 407 408 409 skb_postpull_rcsum(skb, skb_network_header(skb), 410 skb_network_header_len(skb)); 411 hdr = ipv6_hdr(skb); 412 413 /* skb->dev passed may be master dev for vrfs. */ 414 if (sdif) { 415 dev = dev_get_by_index_rcu(net, sdif); 416 if (!dev) 417 goto discard; 418 } else { 419 dev = skb->dev; 420 } 421 422 if (ipv6_addr_is_multicast(&hdr->daddr) && 423 !ipv6_chk_mcast_addr(dev, &hdr->daddr, 424 &hdr->saddr) && 425 !ipv6_is_mld(skb, nexthdr, skb_network_header_len(skb))) { 426 SKB_DR_SET(reason, IP_INADDRERRORS); 427 goto discard; 428 } 429 } 430 if (!(ipprot->flags & INET6_PROTO_NOPOLICY)) { 431 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 432 SKB_DR_SET(reason, XFRM_POLICY); 433 goto discard; 434 } 435 nf_reset_ct(skb); 436 } 437 438 ret = INDIRECT_CALL_2(ipprot->handler, tcp_v6_rcv, udpv6_rcv, 439 skb); 440 if (ret > 0) { 441 if (ipprot->flags & INET6_PROTO_FINAL) { 442 /* Not an extension header, most likely UDP 443 * encapsulation. Use return value as nexthdr 444 * protocol not nhoff (which presumably is 445 * not set by handler). 446 */ 447 nexthdr = ret; 448 goto resubmit_final; 449 } else { 450 goto resubmit; 451 } 452 } else if (ret == 0) { 453 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS); 454 } 455 } else { 456 if (!raw) { 457 if (xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) { 458 __IP6_INC_STATS(net, idev, 459 IPSTATS_MIB_INUNKNOWNPROTOS); 460 icmpv6_send(skb, ICMPV6_PARAMPROB, 461 ICMPV6_UNK_NEXTHDR, nhoff); 462 SKB_DR_SET(reason, IP_NOPROTO); 463 } else { 464 SKB_DR_SET(reason, XFRM_POLICY); 465 } 466 kfree_skb_reason(skb, reason); 467 } else { 468 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDELIVERS); 469 consume_skb(skb); 470 } 471 } 472 return; 473 474 discard: 475 __IP6_INC_STATS(net, idev, IPSTATS_MIB_INDISCARDS); 476 kfree_skb_reason(skb, reason); 477 } 478 479 static int ip6_input_finish(struct net *net, struct sock *sk, struct sk_buff *skb) 480 { 481 if (unlikely(skb_orphan_frags_rx(skb, GFP_ATOMIC))) { 482 __IP6_INC_STATS(net, ip6_dst_idev(skb_dst(skb)), 483 IPSTATS_MIB_INDISCARDS); 484 kfree_skb_reason(skb, SKB_DROP_REASON_NOMEM); 485 return 0; 486 } 487 488 skb_clear_delivery_time(skb); 489 ip6_protocol_deliver_rcu(net, skb, 0, false); 490 491 return 0; 492 } 493 494 495 int ip6_input(struct sk_buff *skb) 496 { 497 int res; 498 499 rcu_read_lock(); 500 res = NF_HOOK(NFPROTO_IPV6, NF_INET_LOCAL_IN, 501 dev_net_rcu(skb->dev), NULL, skb, skb->dev, NULL, 502 ip6_input_finish); 503 rcu_read_unlock(); 504 505 return res; 506 } 507 EXPORT_SYMBOL_GPL(ip6_input); 508 509 int ip6_mc_input(struct sk_buff *skb) 510 { 511 struct net_device *dev = skb->dev; 512 int sdif = inet6_sdif(skb); 513 const struct ipv6hdr *hdr; 514 bool deliver; 515 516 __IP6_UPD_PO_STATS(skb_dst_dev_net_rcu(skb), 517 __in6_dev_get_safely(dev), IPSTATS_MIB_INMCAST, 518 skb->len); 519 520 /* skb->dev passed may be master dev for vrfs. */ 521 if (sdif) { 522 dev = dev_get_by_index_rcu(dev_net_rcu(dev), sdif); 523 if (!dev) { 524 kfree_skb(skb); 525 return -ENODEV; 526 } 527 } 528 529 hdr = ipv6_hdr(skb); 530 deliver = ipv6_chk_mcast_addr(dev, &hdr->daddr, NULL); 531 532 #ifdef CONFIG_IPV6_MROUTE 533 /* 534 * IPv6 multicast router mode is now supported ;) 535 */ 536 if (atomic_read(&dev_net_rcu(skb->dev)->ipv6.devconf_all->mc_forwarding) && 537 !(ipv6_addr_type(&hdr->daddr) & 538 (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)) && 539 likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) { 540 /* 541 * Okay, we try to forward - split and duplicate 542 * packets. 543 */ 544 struct sk_buff *skb2; 545 struct inet6_skb_parm *opt = IP6CB(skb); 546 547 /* Check for MLD */ 548 if (unlikely(opt->flags & IP6SKB_ROUTERALERT)) { 549 /* Check if this is a mld message */ 550 u8 nexthdr = hdr->nexthdr; 551 __be16 frag_off; 552 int offset; 553 554 /* Check if the value of Router Alert 555 * is for MLD (0x0000). 556 */ 557 if (opt->ra == htons(IPV6_OPT_ROUTERALERT_MLD)) { 558 deliver = false; 559 560 if (!ipv6_ext_hdr(nexthdr)) { 561 /* BUG */ 562 goto out; 563 } 564 offset = ipv6_skip_exthdr(skb, sizeof(*hdr), 565 &nexthdr, &frag_off); 566 if (offset < 0) 567 goto out; 568 569 if (ipv6_is_mld(skb, nexthdr, offset)) 570 deliver = true; 571 572 goto out; 573 } 574 /* unknown RA - process it normally */ 575 } 576 577 if (deliver) { 578 skb2 = skb_clone(skb, GFP_ATOMIC); 579 } else { 580 skb2 = skb; 581 skb = NULL; 582 } 583 584 if (skb2) 585 ip6_mr_input(skb2); 586 } 587 out: 588 #endif 589 if (likely(deliver)) { 590 ip6_input(skb); 591 } else { 592 /* discard */ 593 kfree_skb(skb); 594 } 595 596 return 0; 597 } 598