net/ipv6/ip6_input.c

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