Pileus Git - ~andy/linux/blob - net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c

   1
   2 /* (C) 1999-2001 Paul `Rusty' Russell
   3  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License version 2 as
   7  * published by the Free Software Foundation.
   8  */
   9
  10 #include <linux/types.h>
  11 #include <linux/ip.h>
  12 #include <linux/netfilter.h>
  13 #include <linux/module.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/icmp.h>
  16 #include <linux/sysctl.h>
  17 #include <net/route.h>
  18 #include <net/ip.h>
  19
  20 #include <linux/netfilter_ipv4.h>
  21 #include <net/netfilter/nf_conntrack.h>
  22 #include <net/netfilter/nf_conntrack_helper.h>
  23 #include <net/netfilter/nf_conntrack_l4proto.h>
  24 #include <net/netfilter/nf_conntrack_l3proto.h>
  25 #include <net/netfilter/nf_conntrack_zones.h>
  26 #include <net/netfilter/nf_conntrack_core.h>
  27 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
  28 #include <net/netfilter/nf_nat_helper.h>
  29 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
  30 #include <net/netfilter/nf_log.h>
  31
  32 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  33                               struct nf_conntrack_tuple *tuple)
  34 {
  35         const __be32 *ap;
  36         __be32 _addrs[2];
  37         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  38                                 sizeof(u_int32_t) * 2, _addrs);
  39         if (ap == NULL)
  40                 return false;
  41
  42         tuple->src.u3.ip = ap[0];
  43         tuple->dst.u3.ip = ap[1];
  44
  45         return true;
  46 }
  47
  48 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
  49                               const struct nf_conntrack_tuple *orig)
  50 {
  51         tuple->src.u3.ip = orig->dst.u3.ip;
  52         tuple->dst.u3.ip = orig->src.u3.ip;
  53
  54         return true;
  55 }
  56
  57 static int ipv4_print_tuple(struct seq_file *s,
  58                             const struct nf_conntrack_tuple *tuple)
  59 {
  60         return seq_printf(s, "src=%pI4 dst=%pI4 ",
  61                           &tuple->src.u3.ip, &tuple->dst.u3.ip);
  62 }
  63
  64 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
  65                             unsigned int *dataoff, u_int8_t *protonum)
  66 {
  67         const struct iphdr *iph;
  68         struct iphdr _iph;
  69
  70         iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
  71         if (iph == NULL)
  72                 return -NF_ACCEPT;
  73
  74         /* Conntrack defragments packets, we might still see fragments
  75          * inside ICMP packets though. */
  76         if (iph->frag_off & htons(IP_OFFSET))
  77                 return -NF_ACCEPT;
  78
  79         *dataoff = nhoff + (iph->ihl << 2);
  80         *protonum = iph->protocol;
  81
  82         /* Check bogus IP headers */
  83         if (*dataoff > skb->len) {
  84                 pr_debug("nf_conntrack_ipv4: bogus IPv4 packet: "
  85                          "nhoff %u, ihl %u, skblen %u\n",
  86                          nhoff, iph->ihl << 2, skb->len);
  87                 return -NF_ACCEPT;
  88         }
  89
  90         return NF_ACCEPT;
  91 }
  92
  93 static unsigned int ipv4_helper(unsigned int hooknum,
  94                                 struct sk_buff *skb,
  95                                 const struct net_device *in,
  96                                 const struct net_device *out,
  97                                 int (*okfn)(struct sk_buff *))
  98 {
  99         struct nf_conn *ct;
 100         enum ip_conntrack_info ctinfo;
 101         const struct nf_conn_help *help;
 102         const struct nf_conntrack_helper *helper;
 103         unsigned int ret;
 104
 105         /* This is where we call the helper: as the packet goes out. */
 106         ct = nf_ct_get(skb, &ctinfo);
 107         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 108                 return NF_ACCEPT;
 109
 110         help = nfct_help(ct);
 111         if (!help)
 112                 return NF_ACCEPT;
 113
 114         /* rcu_read_lock()ed by nf_hook_slow */
 115         helper = rcu_dereference(help->helper);
 116         if (!helper)
 117                 return NF_ACCEPT;
 118
 119         ret = helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
 120                            ct, ctinfo);
 121         if (ret != NF_ACCEPT && (ret & NF_VERDICT_MASK) != NF_QUEUE) {
 122                 nf_log_packet(NFPROTO_IPV4, hooknum, skb, in, out, NULL,
 123                               "nf_ct_%s: dropping packet", helper->name);
 124         }
 125         return ret;
 126 }
 127
 128 static unsigned int ipv4_confirm(unsigned int hooknum,
 129                                  struct sk_buff *skb,
 130                                  const struct net_device *in,
 131                                  const struct net_device *out,
 132                                  int (*okfn)(struct sk_buff *))
 133 {
 134         struct nf_conn *ct;
 135         enum ip_conntrack_info ctinfo;
 136
 137         ct = nf_ct_get(skb, &ctinfo);
 138         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 139                 goto out;
 140
 141         /* adjust seqs for loopback traffic only in outgoing direction */
 142         if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
 143             !nf_is_loopback_packet(skb)) {
 144                 typeof(nf_nat_seq_adjust_hook) seq_adjust;
 145
 146                 seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
 147                 if (!seq_adjust ||
 148                     !seq_adjust(skb, ct, ctinfo, ip_hdrlen(skb))) {
 149                         NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
 150                         return NF_DROP;
 151                 }
 152         }
 153 out:
 154         /* We've seen it coming out the other side: confirm it */
 155         return nf_conntrack_confirm(skb);
 156 }
 157
 158 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 159                                       struct sk_buff *skb,
 160                                       const struct net_device *in,
 161                                       const struct net_device *out,
 162                                       int (*okfn)(struct sk_buff *))
 163 {
 164         return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
 165 }
 166
 167 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 168                                          struct sk_buff *skb,
 169                                          const struct net_device *in,
 170                                          const struct net_device *out,
 171                                          int (*okfn)(struct sk_buff *))
 172 {
 173         /* root is playing with raw sockets. */
 174         if (skb->len < sizeof(struct iphdr) ||
 175             ip_hdrlen(skb) < sizeof(struct iphdr))
 176                 return NF_ACCEPT;
 177         return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
 178 }
 179
 180 /* Connection tracking may drop packets, but never alters them, so
 181    make it the first hook. */
 182 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
 183         {
 184                 .hook           = ipv4_conntrack_in,
 185                 .owner          = THIS_MODULE,
 186                 .pf             = NFPROTO_IPV4,
 187                 .hooknum        = NF_INET_PRE_ROUTING,
 188                 .priority       = NF_IP_PRI_CONNTRACK,
 189         },
 190         {
 191                 .hook           = ipv4_conntrack_local,
 192                 .owner          = THIS_MODULE,
 193                 .pf             = NFPROTO_IPV4,
 194                 .hooknum        = NF_INET_LOCAL_OUT,
 195                 .priority       = NF_IP_PRI_CONNTRACK,
 196         },
 197         {
 198                 .hook           = ipv4_helper,
 199                 .owner          = THIS_MODULE,
 200                 .pf             = NFPROTO_IPV4,
 201                 .hooknum        = NF_INET_POST_ROUTING,
 202                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 203         },
 204         {
 205                 .hook           = ipv4_confirm,
 206                 .owner          = THIS_MODULE,
 207                 .pf             = NFPROTO_IPV4,
 208                 .hooknum        = NF_INET_POST_ROUTING,
 209                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 210         },
 211         {
 212                 .hook           = ipv4_helper,
 213                 .owner          = THIS_MODULE,
 214                 .pf             = NFPROTO_IPV4,
 215                 .hooknum        = NF_INET_LOCAL_IN,
 216                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 217         },
 218         {
 219                 .hook           = ipv4_confirm,
 220                 .owner          = THIS_MODULE,
 221                 .pf             = NFPROTO_IPV4,
 222                 .hooknum        = NF_INET_LOCAL_IN,
 223                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 224         },
 225 };
 226
 227 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 228 static int log_invalid_proto_min = 0;
 229 static int log_invalid_proto_max = 255;
 230
 231 static ctl_table ip_ct_sysctl_table[] = {
 232         {
 233                 .procname       = "ip_conntrack_max",
 234                 .maxlen         = sizeof(int),
 235                 .mode           = 0644,
 236                 .proc_handler   = proc_dointvec,
 237         },
 238         {
 239                 .procname       = "ip_conntrack_count",
 240                 .maxlen         = sizeof(int),
 241                 .mode           = 0444,
 242                 .proc_handler   = proc_dointvec,
 243         },
 244         {
 245                 .procname       = "ip_conntrack_buckets",
 246                 .maxlen         = sizeof(unsigned int),
 247                 .mode           = 0444,
 248                 .proc_handler   = proc_dointvec,
 249         },
 250         {
 251                 .procname       = "ip_conntrack_checksum",
 252                 .maxlen         = sizeof(int),
 253                 .mode           = 0644,
 254                 .proc_handler   = proc_dointvec,
 255         },
 256         {
 257                 .procname       = "ip_conntrack_log_invalid",
 258                 .maxlen         = sizeof(unsigned int),
 259                 .mode           = 0644,
 260                 .proc_handler   = proc_dointvec_minmax,
 261                 .extra1         = &log_invalid_proto_min,
 262                 .extra2         = &log_invalid_proto_max,
 263         },
 264         { }
 265 };
 266 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 267
 268 /* Fast function for those who don't want to parse /proc (and I don't
 269    blame them). */
 270 /* Reversing the socket's dst/src point of view gives us the reply
 271    mapping. */
 272 static int
 273 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 274 {
 275         const struct inet_sock *inet = inet_sk(sk);
 276         const struct nf_conntrack_tuple_hash *h;
 277         struct nf_conntrack_tuple tuple;
 278
 279         memset(&tuple, 0, sizeof(tuple));
 280         tuple.src.u3.ip = inet->inet_rcv_saddr;
 281         tuple.src.u.tcp.port = inet->inet_sport;
 282         tuple.dst.u3.ip = inet->inet_daddr;
 283         tuple.dst.u.tcp.port = inet->inet_dport;
 284         tuple.src.l3num = PF_INET;
 285         tuple.dst.protonum = sk->sk_protocol;
 286
 287         /* We only do TCP and SCTP at the moment: is there a better way? */
 288         if (sk->sk_protocol != IPPROTO_TCP && sk->sk_protocol != IPPROTO_SCTP) {
 289                 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
 290                 return -ENOPROTOOPT;
 291         }
 292
 293         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 294                 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
 295                          *len, sizeof(struct sockaddr_in));
 296                 return -EINVAL;
 297         }
 298
 299         h = nf_conntrack_find_get(sock_net(sk), NF_CT_DEFAULT_ZONE, &tuple);
 300         if (h) {
 301                 struct sockaddr_in sin;
 302                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 303
 304                 sin.sin_family = AF_INET;
 305                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 306                         .tuple.dst.u.tcp.port;
 307                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 308                         .tuple.dst.u3.ip;
 309                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 310
 311                 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
 312                          &sin.sin_addr.s_addr, ntohs(sin.sin_port));
 313                 nf_ct_put(ct);
 314                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 315                         return -EFAULT;
 316                 else
 317                         return 0;
 318         }
 319         pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
 320                  &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
 321                  &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
 322         return -ENOENT;
 323 }
 324
 325 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 326
 327 #include <linux/netfilter/nfnetlink.h>
 328 #include <linux/netfilter/nfnetlink_conntrack.h>
 329
 330 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
 331                                 const struct nf_conntrack_tuple *tuple)
 332 {
 333         if (nla_put_be32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) ||
 334             nla_put_be32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip))
 335                 goto nla_put_failure;
 336         return 0;
 337
 338 nla_put_failure:
 339         return -1;
 340 }
 341
 342 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
 343         [CTA_IP_V4_SRC] = { .type = NLA_U32 },
 344         [CTA_IP_V4_DST] = { .type = NLA_U32 },
 345 };
 346
 347 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
 348                                 struct nf_conntrack_tuple *t)
 349 {
 350         if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
 351                 return -EINVAL;
 352
 353         t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
 354         t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
 355
 356         return 0;
 357 }
 358
 359 static int ipv4_nlattr_tuple_size(void)
 360 {
 361         return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
 362 }
 363 #endif
 364
 365 static struct nf_sockopt_ops so_getorigdst = {
 366         .pf             = PF_INET,
 367         .get_optmin     = SO_ORIGINAL_DST,
 368         .get_optmax     = SO_ORIGINAL_DST+1,
 369         .get            = &getorigdst,
 370         .owner          = THIS_MODULE,
 371 };
 372
 373 static int ipv4_init_net(struct net *net)
 374 {
 375 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 376         struct nf_ip_net *in = &net->ct.nf_ct_proto;
 377         in->ctl_table = kmemdup(ip_ct_sysctl_table,
 378                                 sizeof(ip_ct_sysctl_table),
 379                                 GFP_KERNEL);
 380         if (!in->ctl_table)
 381                 return -ENOMEM;
 382
 383         in->ctl_table[0].data = &nf_conntrack_max;
 384         in->ctl_table[1].data = &net->ct.count;
 385         in->ctl_table[2].data = &net->ct.htable_size;
 386         in->ctl_table[3].data = &net->ct.sysctl_checksum;
 387         in->ctl_table[4].data = &net->ct.sysctl_log_invalid;
 388 #endif
 389         return 0;
 390 }
 391
 392 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
 393         .l3proto         = PF_INET,
 394         .name            = "ipv4",
 395         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 396         .invert_tuple    = ipv4_invert_tuple,
 397         .print_tuple     = ipv4_print_tuple,
 398         .get_l4proto     = ipv4_get_l4proto,
 399 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 400         .tuple_to_nlattr = ipv4_tuple_to_nlattr,
 401         .nlattr_tuple_size = ipv4_nlattr_tuple_size,
 402         .nlattr_to_tuple = ipv4_nlattr_to_tuple,
 403         .nla_policy      = ipv4_nla_policy,
 404 #endif
 405 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 406         .ctl_table_path  = "net/ipv4/netfilter",
 407 #endif
 408         .init_net        = ipv4_init_net,
 409         .me              = THIS_MODULE,
 410 };
 411
 412 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
 413                   &nf_conntrack_htable_size, 0600);
 414
 415 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 416 MODULE_ALIAS("ip_conntrack");
 417 MODULE_LICENSE("GPL");
 418
 419 static int ipv4_net_init(struct net *net)
 420 {
 421         int ret = 0;
 422
 423         ret = nf_conntrack_l4proto_register(net,
 424                                             &nf_conntrack_l4proto_tcp4);
 425         if (ret < 0) {
 426                 pr_err("nf_conntrack_l4proto_tcp4 :protocol register failed\n");
 427                 goto out_tcp;
 428         }
 429         ret = nf_conntrack_l4proto_register(net,
 430                                             &nf_conntrack_l4proto_udp4);
 431         if (ret < 0) {
 432                 pr_err("nf_conntrack_l4proto_udp4 :protocol register failed\n");
 433                 goto out_udp;
 434         }
 435         ret = nf_conntrack_l4proto_register(net,
 436                                             &nf_conntrack_l4proto_icmp);
 437         if (ret < 0) {
 438                 pr_err("nf_conntrack_l4proto_icmp4 :protocol register failed\n");
 439                 goto out_icmp;
 440         }
 441         ret = nf_conntrack_l3proto_register(net,
 442                                             &nf_conntrack_l3proto_ipv4);
 443         if (ret < 0) {
 444                 pr_err("nf_conntrack_l3proto_ipv4 :protocol register failed\n");
 445                 goto out_ipv4;
 446         }
 447         return 0;
 448 out_ipv4:
 449         nf_conntrack_l4proto_unregister(net,
 450                                         &nf_conntrack_l4proto_icmp);
 451 out_icmp:
 452         nf_conntrack_l4proto_unregister(net,
 453                                         &nf_conntrack_l4proto_udp4);
 454 out_udp:
 455         nf_conntrack_l4proto_unregister(net,
 456                                         &nf_conntrack_l4proto_tcp4);
 457 out_tcp:
 458         return ret;
 459 }
 460
 461 static void ipv4_net_exit(struct net *net)
 462 {
 463         nf_conntrack_l3proto_unregister(net,
 464                                         &nf_conntrack_l3proto_ipv4);
 465         nf_conntrack_l4proto_unregister(net,
 466                                         &nf_conntrack_l4proto_icmp);
 467         nf_conntrack_l4proto_unregister(net,
 468                                         &nf_conntrack_l4proto_udp4);
 469         nf_conntrack_l4proto_unregister(net,
 470                                         &nf_conntrack_l4proto_tcp4);
 471 }
 472
 473 static struct pernet_operations ipv4_net_ops = {
 474         .init = ipv4_net_init,
 475         .exit = ipv4_net_exit,
 476 };
 477
 478 static int __init nf_conntrack_l3proto_ipv4_init(void)
 479 {
 480         int ret = 0;
 481
 482         need_conntrack();
 483         nf_defrag_ipv4_enable();
 484
 485         ret = nf_register_sockopt(&so_getorigdst);
 486         if (ret < 0) {
 487                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 488                 return ret;
 489         }
 490
 491         ret = register_pernet_subsys(&ipv4_net_ops);
 492         if (ret < 0) {
 493                 pr_err("nf_conntrack_ipv4: can't register pernet ops\n");
 494                 goto cleanup_sockopt;
 495         }
 496
 497         ret = nf_register_hooks(ipv4_conntrack_ops,
 498                                 ARRAY_SIZE(ipv4_conntrack_ops));
 499         if (ret < 0) {
 500                 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
 501                 goto cleanup_pernet;
 502         }
 503 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 504         ret = nf_conntrack_ipv4_compat_init();
 505         if (ret < 0)
 506                 goto cleanup_hooks;
 507 #endif
 508         return ret;
 509 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 510  cleanup_hooks:
 511         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 512 #endif
 513  cleanup_pernet:
 514         unregister_pernet_subsys(&ipv4_net_ops);
 515  cleanup_sockopt:
 516         nf_unregister_sockopt(&so_getorigdst);
 517         return ret;
 518 }
 519
 520 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 521 {
 522         synchronize_net();
 523 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 524         nf_conntrack_ipv4_compat_fini();
 525 #endif
 526         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 527         unregister_pernet_subsys(&ipv4_net_ops);
 528         nf_unregister_sockopt(&so_getorigdst);
 529 }
 530
 531 module_init(nf_conntrack_l3proto_ipv4_init);
 532 module_exit(nf_conntrack_l3proto_ipv4_fini);
 533
 534 void need_ipv4_conntrack(void)
 535 {
 536         return;
 537 }
 538 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);