net/ipv4/ip_input.c

   1 /*
   2  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   3  *              operating system.  INET is implemented using the  BSD Socket
   4  *              interface as the means of communication with the user level.
   5  *
   6  *              The Internet Protocol (IP) module.
   7  *
   8  * Authors:     Ross Biro
   9  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  10  *              Donald Becker, <becker@super.org>
  11  *              Alan Cox, <alan@lxorguk.ukuu.org.uk>
  12  *              Richard Underwood
  13  *              Stefan Becker, <stefanb@yello.ping.de>
  14  *              Jorge Cwik, <jorge@laser.satlink.net>
  15  *              Arnt Gulbrandsen, <agulbra@nvg.unit.no>
  16  *
  17  *
  18  * Fixes:
  19  *              Alan Cox        :       Commented a couple of minor bits of surplus code
  20  *              Alan Cox        :       Undefining IP_FORWARD doesn't include the code
  21  *                                      (just stops a compiler warning).
  22  *              Alan Cox        :       Frames with >=MAX_ROUTE record routes, strict routes or loose routes
  23  *                                      are junked rather than corrupting things.
  24  *              Alan Cox        :       Frames to bad broadcast subnets are dumped
  25  *                                      We used to process them non broadcast and
  26  *                                      boy could that cause havoc.
  27  *              Alan Cox        :       ip_forward sets the free flag on the
  28  *                                      new frame it queues. Still crap because
  29  *                                      it copies the frame but at least it
  30  *                                      doesn't eat memory too.
  31  *              Alan Cox        :       Generic queue code and memory fixes.
  32  *              Fred Van Kempen :       IP fragment support (borrowed from NET2E)
  33  *              Gerhard Koerting:       Forward fragmented frames correctly.
  34  *              Gerhard Koerting:       Fixes to my fix of the above 8-).
  35  *              Gerhard Koerting:       IP interface addressing fix.
  36  *              Linus Torvalds  :       More robustness checks
  37  *              Alan Cox        :       Even more checks: Still not as robust as it ought to be
  38  *              Alan Cox        :       Save IP header pointer for later
  39  *              Alan Cox        :       ip option setting
  40  *              Alan Cox        :       Use ip_tos/ip_ttl settings
  41  *              Alan Cox        :       Fragmentation bogosity removed
  42  *                                      (Thanks to Mark.Bush@prg.ox.ac.uk)
  43  *              Dmitry Gorodchanin :    Send of a raw packet crash fix.
  44  *              Alan Cox        :       Silly ip bug when an overlength
  45  *                                      fragment turns up. Now frees the
  46  *                                      queue.
  47  *              Linus Torvalds/ :       Memory leakage on fragmentation
  48  *              Alan Cox        :       handling.
  49  *              Gerhard Koerting:       Forwarding uses IP priority hints
  50  *              Teemu Rantanen  :       Fragment problems.
  51  *              Alan Cox        :       General cleanup, comments and reformat
  52  *              Alan Cox        :       SNMP statistics
  53  *              Alan Cox        :       BSD address rule semantics. Also see
  54  *                                      UDP as there is a nasty checksum issue
  55  *                                      if you do things the wrong way.
  56  *              Alan Cox        :       Always defrag, moved IP_FORWARD to the config.in file
  57  *              Alan Cox        :       IP options adjust sk->priority.
  58  *              Pedro Roque     :       Fix mtu/length error in ip_forward.
  59  *              Alan Cox        :       Avoid ip_chk_addr when possible.
  60  *      Richard Underwood       :       IP multicasting.
  61  *              Alan Cox        :       Cleaned up multicast handlers.
  62  *              Alan Cox        :       RAW sockets demultiplex in the BSD style.
  63  *              Gunther Mayer   :       Fix the SNMP reporting typo
  64  *              Alan Cox        :       Always in group 224.0.0.1
  65  *      Pauline Middelink       :       Fast ip_checksum update when forwarding
  66  *                                      Masquerading support.
  67  *              Alan Cox        :       Multicast loopback error for 224.0.0.1
  68  *              Alan Cox        :       IP_MULTICAST_LOOP option.
  69  *              Alan Cox        :       Use notifiers.
  70  *              Bjorn Ekwall    :       Removed ip_csum (from slhc.c too)
  71  *              Bjorn Ekwall    :       Moved ip_fast_csum to ip.h (inline!)
  72  *              Stefan Becker   :       Send out ICMP HOST REDIRECT
  73  *      Arnt Gulbrandsen        :       ip_build_xmit
  74  *              Alan Cox        :       Per socket routing cache
  75  *              Alan Cox        :       Fixed routing cache, added header cache.
  76  *              Alan Cox        :       Loopback didn't work right in original ip_build_xmit - fixed it.
  77  *              Alan Cox        :       Only send ICMP_REDIRECT if src/dest are the same net.
  78  *              Alan Cox        :       Incoming IP option handling.
  79  *              Alan Cox        :       Set saddr on raw output frames as per BSD.
  80  *              Alan Cox        :       Stopped broadcast source route explosions.
  81  *              Alan Cox        :       Can disable source routing
  82  *              Takeshi Sone    :       Masquerading didn't work.
  83  *      Dave Bonn,Alan Cox      :       Faster IP forwarding whenever possible.
  84  *              Alan Cox        :       Memory leaks, tramples, misc debugging.
  85  *              Alan Cox        :       Fixed multicast (by popular demand 8))
  86  *              Alan Cox        :       Fixed forwarding (by even more popular demand 8))
  87  *              Alan Cox        :       Fixed SNMP statistics [I think]
  88  *      Gerhard Koerting        :       IP fragmentation forwarding fix
  89  *              Alan Cox        :       Device lock against page fault.
  90  *              Alan Cox        :       IP_HDRINCL facility.
  91  *      Werner Almesberger      :       Zero fragment bug
  92  *              Alan Cox        :       RAW IP frame length bug
  93  *              Alan Cox        :       Outgoing firewall on build_xmit
  94  *              A.N.Kuznetsov   :       IP_OPTIONS support throughout the kernel
  95  *              Alan Cox        :       Multicast routing hooks
  96  *              Jos Vos         :       Do accounting *before* call_in_firewall
  97  *      Willy Konynenberg       :       Transparent proxying support
  98  *
  99  *
 100  *
 101  * To Fix:
 102  *              IP fragmentation wants rewriting cleanly. The RFC815 algorithm is much more efficient
 103  *              and could be made very efficient with the addition of some virtual memory hacks to permit
 104  *              the allocation of a buffer that can then be 'grown' by twiddling page tables.
 105  *              Output fragmentation wants updating along with the buffer management to use a single
 106  *              interleaved copy algorithm so that fragmenting has a one copy overhead. Actual packet
 107  *              output should probably do its own fragmentation at the UDP/RAW layer. TCP shouldn't cause
 108  *              fragmentation anyway.
 109  *
 110  *              This program is free software; you can redistribute it and/or
 111  *              modify it under the terms of the GNU General Public License
 112  *              as published by the Free Software Foundation; either version
 113  *              2 of the License, or (at your option) any later version.
 114  */
 115
 116 #define pr_fmt(fmt) "IPv4: " fmt
 117
 118 #include <linux/module.h>
 119 #include <linux/types.h>
 120 #include <linux/kernel.h>
 121 #include <linux/string.h>
 122 #include <linux/errno.h>
 123 #include <linux/slab.h>
 124
 125 #include <linux/net.h>
 126 #include <linux/socket.h>
 127 #include <linux/sockios.h>
 128 #include <linux/in.h>
 129 #include <linux/inet.h>
 130 #include <linux/inetdevice.h>
 131 #include <linux/netdevice.h>
 132 #include <linux/etherdevice.h>
 133
 134 #include <net/snmp.h>
 135 #include <net/ip.h>
 136 #include <net/protocol.h>
 137 #include <net/route.h>
 138 #include <linux/skbuff.h>
 139 #include <net/sock.h>
 140 #include <net/arp.h>
 141 #include <net/icmp.h>
 142 #include <net/raw.h>
 143 #include <net/checksum.h>
 144 #include <net/inet_ecn.h>
 145 #include <linux/netfilter_ipv4.h>
 146 #include <net/xfrm.h>
 147 #include <linux/mroute.h>
 148 #include <linux/netlink.h>
 149 #include <net/dst_metadata.h>
 150
 151 /*
 152  *      Process Router Attention IP option (RFC 2113)
 153  */
 154 bool ip_call_ra_chain(struct sk_buff *skb)
 155 {
 156         struct ip_ra_chain *ra;
 157         u8 protocol = ip_hdr(skb)->protocol;
 158         struct sock *last = NULL;
 159         struct net_device *dev = skb->dev;
 160
 161         for (ra = rcu_dereference(ip_ra_chain); ra; ra = rcu_dereference(ra->next)) {
 162                 struct sock *sk = ra->sk;
 163
 164                 /* If socket is bound to an interface, only report
 165                  * the packet if it came  from that interface.
 166                  */
 167                 if (sk && inet_sk(sk)->inet_num == protocol &&
 168                     (!sk->sk_bound_dev_if ||
 169                      sk->sk_bound_dev_if == dev->ifindex) &&
 170                     net_eq(sock_net(sk), dev_net(dev))) {
 171                         if (ip_is_fragment(ip_hdr(skb))) {
 172                                 if (ip_defrag(skb, IP_DEFRAG_CALL_RA_CHAIN))
 173                                         return true;
 174                         }
 175                         if (last) {
 176                                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
 177                                 if (skb2)
 178                                         raw_rcv(last, skb2);
 179                         }
 180                         last = sk;
 181                 }
 182         }
 183
 184         if (last) {
 185                 raw_rcv(last, skb);
 186                 return true;
 187         }
 188         return false;
 189 }
 190
 191 static int ip_local_deliver_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 192 {
 193         __skb_pull(skb, skb_network_header_len(skb));
 194
 195         rcu_read_lock();
 196         {
 197                 int protocol = ip_hdr(skb)->protocol;
 198                 const struct net_protocol *ipprot;
 199                 int raw;
 200
 201         resubmit:
 202                 raw = raw_local_deliver(skb, protocol);
 203
 204                 ipprot = rcu_dereference(inet_protos[protocol]);
 205                 if (ipprot) {
 206                         int ret;
 207
 208                         if (!ipprot->no_policy) {
 209                                 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 210                                         kfree_skb(skb);
 211                                         goto out;
 212                                 }
 213                                 nf_reset(skb);
 214                         }
 215                         ret = ipprot->handler(skb);
 216                         if (ret < 0) {
 217                                 protocol = -ret;
 218                                 goto resubmit;
 219                         }
 220                         IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
 221                 } else {
 222                         if (!raw) {
 223                                 if (xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb)) {
 224                                         IP_INC_STATS_BH(net, IPSTATS_MIB_INUNKNOWNPROTOS);
 225                                         icmp_send(skb, ICMP_DEST_UNREACH,
 226                                                   ICMP_PROT_UNREACH, 0);
 227                                 }
 228                                 kfree_skb(skb);
 229                         } else {
 230                                 IP_INC_STATS_BH(net, IPSTATS_MIB_INDELIVERS);
 231                                 consume_skb(skb);
 232                         }
 233                 }
 234         }
 235  out:
 236         rcu_read_unlock();
 237
 238         return 0;
 239 }
 240
 241 /*
 242  *      Deliver IP Packets to the higher protocol layers.
 243  */
 244 int ip_local_deliver(struct sk_buff *skb)
 245 {
 246         /*
 247          *      Reassemble IP fragments.
 248          */
 249
 250         if (ip_is_fragment(ip_hdr(skb))) {
 251                 if (ip_defrag(skb, IP_DEFRAG_LOCAL_DELIVER))
 252                         return 0;
 253         }
 254
 255         return NF_HOOK(NFPROTO_IPV4, NF_INET_LOCAL_IN,
 256                        dev_net(skb->dev), NULL, skb, skb->dev, NULL,
 257                        ip_local_deliver_finish);
 258 }
 259
 260 static inline bool ip_rcv_options(struct sk_buff *skb)
 261 {
 262         struct ip_options *opt;
 263         const struct iphdr *iph;
 264         struct net_device *dev = skb->dev;
 265
 266         /* It looks as overkill, because not all
 267            IP options require packet mangling.
 268            But it is the easiest for now, especially taking
 269            into account that combination of IP options
 270            and running sniffer is extremely rare condition.
 271                                               --ANK (980813)
 272         */
 273         if (skb_cow(skb, skb_headroom(skb))) {
 274                 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
 275                 goto drop;
 276         }
 277
 278         iph = ip_hdr(skb);
 279         opt = &(IPCB(skb)->opt);
 280         opt->optlen = iph->ihl*4 - sizeof(struct iphdr);
 281
 282         if (ip_options_compile(dev_net(dev), opt, skb)) {
 283                 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
 284                 goto drop;
 285         }
 286
 287         if (unlikely(opt->srr)) {
 288                 struct in_device *in_dev = __in_dev_get_rcu(dev);
 289
 290                 if (in_dev) {
 291                         if (!IN_DEV_SOURCE_ROUTE(in_dev)) {
 292                                 if (IN_DEV_LOG_MARTIANS(in_dev))
 293                                         net_info_ratelimited("source route option %pI4 -> %pI4\n",
 294                                                              &iph->saddr,
 295                                                              &iph->daddr);
 296                                 goto drop;
 297                         }
 298                 }
 299
 300                 if (ip_options_rcv_srr(skb))
 301                         goto drop;
 302         }
 303
 304         return false;
 305 drop:
 306         return true;
 307 }
 308
 309 int sysctl_ip_early_demux __read_mostly = 1;
 310 EXPORT_SYMBOL(sysctl_ip_early_demux);
 311
 312 static int ip_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
 313 {
 314         const struct iphdr *iph = ip_hdr(skb);
 315         struct rtable *rt;
 316
 317         if (sysctl_ip_early_demux && !skb_dst(skb) && !skb->sk) {
 318                 const struct net_protocol *ipprot;
 319                 int protocol = iph->protocol;
 320
 321                 ipprot = rcu_dereference(inet_protos[protocol]);
 322                 if (ipprot && ipprot->early_demux) {
 323                         ipprot->early_demux(skb);
 324                         /* must reload iph, skb->head might have changed */
 325                         iph = ip_hdr(skb);
 326                 }
 327         }
 328
 329         /*
 330          *      Initialise the virtual path cache for the packet. It describes
 331          *      how the packet travels inside Linux networking.
 332          */
 333         if (!skb_valid_dst(skb)) {
 334                 int err = ip_route_input_noref(skb, iph->daddr, iph->saddr,
 335                                                iph->tos, skb->dev);
 336                 if (unlikely(err)) {
 337                         if (err == -EXDEV)
 338                                 NET_INC_STATS_BH(net, LINUX_MIB_IPRPFILTER);
 339                         goto drop;
 340                 }
 341         }
 342
 343 #ifdef CONFIG_IP_ROUTE_CLASSID
 344         if (unlikely(skb_dst(skb)->tclassid)) {
 345                 struct ip_rt_acct *st = this_cpu_ptr(ip_rt_acct);
 346                 u32 idx = skb_dst(skb)->tclassid;
 347                 st[idx&0xFF].o_packets++;
 348                 st[idx&0xFF].o_bytes += skb->len;
 349                 st[(idx>>16)&0xFF].i_packets++;
 350                 st[(idx>>16)&0xFF].i_bytes += skb->len;
 351         }
 352 #endif
 353
 354         if (iph->ihl > 5 && ip_rcv_options(skb))
 355                 goto drop;
 356
 357         rt = skb_rtable(skb);
 358         if (rt->rt_type == RTN_MULTICAST) {
 359                 IP_UPD_PO_STATS_BH(net, IPSTATS_MIB_INMCAST, skb->len);
 360         } else if (rt->rt_type == RTN_BROADCAST)
 361                 IP_UPD_PO_STATS_BH(net, IPSTATS_MIB_INBCAST, skb->len);
 362
 363         return dst_input(skb);
 364
 365 drop:
 366         kfree_skb(skb);
 367         return NET_RX_DROP;
 368 }
 369
 370 /*
 371  *      Main IP Receive routine.
 372  */
 373 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt, struct net_device *orig_dev)
 374 {
 375         const struct iphdr *iph;
 376         struct net *net;
 377         u32 len;
 378
 379         /* When the interface is in promisc. mode, drop all the crap
 380          * that it receives, do not try to analyse it.
 381          */
 382         if (skb->pkt_type == PACKET_OTHERHOST)
 383                 goto drop;
 384
 385
 386         net = dev_net(dev);
 387         IP_UPD_PO_STATS_BH(net, IPSTATS_MIB_IN, skb->len);
 388
 389         skb = skb_share_check(skb, GFP_ATOMIC);
 390         if (!skb) {
 391                 IP_INC_STATS_BH(net, IPSTATS_MIB_INDISCARDS);
 392                 goto out;
 393         }
 394
 395         if (!pskb_may_pull(skb, sizeof(struct iphdr)))
 396                 goto inhdr_error;
 397
 398         iph = ip_hdr(skb);
 399
 400         /*
 401          *      RFC1122: 3.2.1.2 MUST silently discard any IP frame that fails the checksum.
 402          *
 403          *      Is the datagram acceptable?
 404          *
 405          *      1.      Length at least the size of an ip header
 406          *      2.      Version of 4
 407          *      3.      Checksums correctly. [Speed optimisation for later, skip loopback checksums]
 408          *      4.      Doesn't have a bogus length
 409          */
 410
 411         if (iph->ihl < 5 || iph->version != 4)
 412                 goto inhdr_error;
 413
 414         BUILD_BUG_ON(IPSTATS_MIB_ECT1PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_1);
 415         BUILD_BUG_ON(IPSTATS_MIB_ECT0PKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_ECT_0);
 416         BUILD_BUG_ON(IPSTATS_MIB_CEPKTS != IPSTATS_MIB_NOECTPKTS + INET_ECN_CE);
 417         IP_ADD_STATS_BH(net,
 418                         IPSTATS_MIB_NOECTPKTS + (iph->tos & INET_ECN_MASK),
 419                         max_t(unsigned short, 1, skb_shinfo(skb)->gso_segs));
 420
 421         if (!pskb_may_pull(skb, iph->ihl*4))
 422                 goto inhdr_error;
 423
 424         iph = ip_hdr(skb);
 425
 426         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
 427                 goto csum_error;
 428
 429         len = ntohs(iph->tot_len);
 430         if (skb->len < len) {
 431                 IP_INC_STATS_BH(net, IPSTATS_MIB_INTRUNCATEDPKTS);
 432                 goto drop;
 433         } else if (len < (iph->ihl*4))
 434                 goto inhdr_error;
 435
 436         /* Our transport medium may have padded the buffer out. Now we know it
 437          * is IP we can trim to the true length of the frame.
 438          * Note this now means skb->len holds ntohs(iph->tot_len).
 439          */
 440         if (pskb_trim_rcsum(skb, len)) {
 441                 IP_INC_STATS_BH(net, IPSTATS_MIB_INDISCARDS);
 442                 goto drop;
 443         }
 444
 445         skb->transport_header = skb->network_header + iph->ihl*4;
 446
 447         /* Remove any debris in the socket control block */
 448         memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
 449
 450         /* Must drop socket now because of tproxy. */
 451         skb_orphan(skb);
 452
 453         return NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING,
 454                        net, NULL, skb, dev, NULL,
 455                        ip_rcv_finish);
 456
 457 csum_error:
 458         IP_INC_STATS_BH(net, IPSTATS_MIB_CSUMERRORS);
 459 inhdr_error:
 460         IP_INC_STATS_BH(net, IPSTATS_MIB_INHDRERRORS);
 461 drop:
 462         kfree_skb(skb);
 463 out:
 464         return NET_RX_DROP;
 465 }