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