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