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 int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb,
  33                               struct nf_conn *ct,
  34                               enum ip_conntrack_info ctinfo);
  35 EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook);
  36
  37 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  38                               struct nf_conntrack_tuple *tuple)
  39 {
  40         const __be32 *ap;
  41         __be32 _addrs[2];
  42         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  43                                 sizeof(u_int32_t) * 2, _addrs);
  44         if (ap == NULL)
  45                 return false;
  46
  47         tuple->src.u3.ip = ap[0];
  48         tuple->dst.u3.ip = ap[1];
  49
  50         return true;
  51 }
  52
  53 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
  54                               const struct nf_conntrack_tuple *orig)
  55 {
  56         tuple->src.u3.ip = orig->dst.u3.ip;
  57         tuple->dst.u3.ip = orig->src.u3.ip;
  58
  59         return true;
  60 }
  61
  62 static int ipv4_print_tuple(struct seq_file *s,
  63                             const struct nf_conntrack_tuple *tuple)
  64 {
  65         return seq_printf(s, "src=%pI4 dst=%pI4 ",
  66                           &tuple->src.u3.ip, &tuple->dst.u3.ip);
  67 }
  68
  69 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
  70                             unsigned int *dataoff, u_int8_t *protonum)
  71 {
  72         const struct iphdr *iph;
  73         struct iphdr _iph;
  74
  75         iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
  76         if (iph == NULL)
  77                 return -NF_DROP;
  78
  79         /* Conntrack defragments packets, we might still see fragments
  80          * inside ICMP packets though. */
  81         if (iph->frag_off & htons(IP_OFFSET))
  82                 return -NF_DROP;
  83
  84         *dataoff = nhoff + (iph->ihl << 2);
  85         *protonum = iph->protocol;
  86
  87         /* Check bogus IP headers */
  88         if (*dataoff > skb->len) {
  89                 pr_debug("nf_conntrack_ipv4: bogus IPv4 packet: "
  90                          "nhoff %u, ihl %u, skblen %u\n",
  91                          nhoff, iph->ihl << 2, skb->len);
  92                 return -NF_ACCEPT;
  93         }
  94
  95         return NF_ACCEPT;
  96 }
  97
  98 static unsigned int ipv4_confirm(unsigned int hooknum,
  99                                  struct sk_buff *skb,
 100                                  const struct net_device *in,
 101                                  const struct net_device *out,
 102                                  int (*okfn)(struct sk_buff *))
 103 {
 104         struct nf_conn *ct;
 105         enum ip_conntrack_info ctinfo;
 106         const struct nf_conn_help *help;
 107         const struct nf_conntrack_helper *helper;
 108         unsigned int ret;
 109
 110         /* This is where we call the helper: as the packet goes out. */
 111         ct = nf_ct_get(skb, &ctinfo);
 112         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 113                 goto out;
 114
 115         help = nfct_help(ct);
 116         if (!help)
 117                 goto out;
 118
 119         /* rcu_read_lock()ed by nf_hook_slow */
 120         helper = rcu_dereference(help->helper);
 121         if (!helper)
 122                 goto out;
 123
 124         ret = helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
 125                            ct, ctinfo);
 126         if (ret != NF_ACCEPT) {
 127                 nf_log_packet(NFPROTO_IPV4, hooknum, skb, in, out, NULL,
 128                               "nf_ct_%s: dropping packet", helper->name);
 129                 return ret;
 130         }
 131
 132         /* adjust seqs for loopback traffic only in outgoing direction */
 133         if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
 134             !nf_is_loopback_packet(skb)) {
 135                 typeof(nf_nat_seq_adjust_hook) seq_adjust;
 136
 137                 seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
 138                 if (!seq_adjust || !seq_adjust(skb, ct, ctinfo)) {
 139                         NF_CT_STAT_INC_ATOMIC(nf_ct_net(ct), drop);
 140                         return NF_DROP;
 141                 }
 142         }
 143 out:
 144         /* We've seen it coming out the other side: confirm it */
 145         return nf_conntrack_confirm(skb);
 146 }
 147
 148 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 149                                       struct sk_buff *skb,
 150                                       const struct net_device *in,
 151                                       const struct net_device *out,
 152                                       int (*okfn)(struct sk_buff *))
 153 {
 154         return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
 155 }
 156
 157 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 158                                          struct sk_buff *skb,
 159                                          const struct net_device *in,
 160                                          const struct net_device *out,
 161                                          int (*okfn)(struct sk_buff *))
 162 {
 163         /* root is playing with raw sockets. */
 164         if (skb->len < sizeof(struct iphdr) ||
 165             ip_hdrlen(skb) < sizeof(struct iphdr))
 166                 return NF_ACCEPT;
 167         return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
 168 }
 169
 170 /* Connection tracking may drop packets, but never alters them, so
 171    make it the first hook. */
 172 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
 173         {
 174                 .hook           = ipv4_conntrack_in,
 175                 .owner          = THIS_MODULE,
 176                 .pf             = NFPROTO_IPV4,
 177                 .hooknum        = NF_INET_PRE_ROUTING,
 178                 .priority       = NF_IP_PRI_CONNTRACK,
 179         },
 180         {
 181                 .hook           = ipv4_conntrack_local,
 182                 .owner          = THIS_MODULE,
 183                 .pf             = NFPROTO_IPV4,
 184                 .hooknum        = NF_INET_LOCAL_OUT,
 185                 .priority       = NF_IP_PRI_CONNTRACK,
 186         },
 187         {
 188                 .hook           = ipv4_confirm,
 189                 .owner          = THIS_MODULE,
 190                 .pf             = NFPROTO_IPV4,
 191                 .hooknum        = NF_INET_POST_ROUTING,
 192                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 193         },
 194         {
 195                 .hook           = ipv4_confirm,
 196                 .owner          = THIS_MODULE,
 197                 .pf             = NFPROTO_IPV4,
 198                 .hooknum        = NF_INET_LOCAL_IN,
 199                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 200         },
 201 };
 202
 203 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 204 static int log_invalid_proto_min = 0;
 205 static int log_invalid_proto_max = 255;
 206
 207 static ctl_table ip_ct_sysctl_table[] = {
 208         {
 209                 .procname       = "ip_conntrack_max",
 210                 .data           = &nf_conntrack_max,
 211                 .maxlen         = sizeof(int),
 212                 .mode           = 0644,
 213                 .proc_handler   = proc_dointvec,
 214         },
 215         {
 216                 .procname       = "ip_conntrack_count",
 217                 .data           = &init_net.ct.count,
 218                 .maxlen         = sizeof(int),
 219                 .mode           = 0444,
 220                 .proc_handler   = proc_dointvec,
 221         },
 222         {
 223                 .procname       = "ip_conntrack_buckets",
 224                 .data           = &init_net.ct.htable_size,
 225                 .maxlen         = sizeof(unsigned int),
 226                 .mode           = 0444,
 227                 .proc_handler   = proc_dointvec,
 228         },
 229         {
 230                 .procname       = "ip_conntrack_checksum",
 231                 .data           = &init_net.ct.sysctl_checksum,
 232                 .maxlen         = sizeof(int),
 233                 .mode           = 0644,
 234                 .proc_handler   = proc_dointvec,
 235         },
 236         {
 237                 .procname       = "ip_conntrack_log_invalid",
 238                 .data           = &init_net.ct.sysctl_log_invalid,
 239                 .maxlen         = sizeof(unsigned int),
 240                 .mode           = 0644,
 241                 .proc_handler   = proc_dointvec_minmax,
 242                 .extra1         = &log_invalid_proto_min,
 243                 .extra2         = &log_invalid_proto_max,
 244         },
 245         { }
 246 };
 247 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 248
 249 /* Fast function for those who don't want to parse /proc (and I don't
 250    blame them). */
 251 /* Reversing the socket's dst/src point of view gives us the reply
 252    mapping. */
 253 static int
 254 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 255 {
 256         const struct inet_sock *inet = inet_sk(sk);
 257         const struct nf_conntrack_tuple_hash *h;
 258         struct nf_conntrack_tuple tuple;
 259
 260         memset(&tuple, 0, sizeof(tuple));
 261
 262         lock_sock(sk);
 263         tuple.src.u3.ip = inet->inet_rcv_saddr;
 264         tuple.src.u.tcp.port = inet->inet_sport;
 265         tuple.dst.u3.ip = inet->inet_daddr;
 266         tuple.dst.u.tcp.port = inet->inet_dport;
 267         tuple.src.l3num = PF_INET;
 268         tuple.dst.protonum = sk->sk_protocol;
 269         release_sock(sk);
 270
 271         /* We only do TCP and SCTP at the moment: is there a better way? */
 272         if (tuple.dst.protonum != IPPROTO_TCP &&
 273             tuple.dst.protonum != IPPROTO_SCTP) {
 274                 pr_debug("SO_ORIGINAL_DST: Not a TCP/SCTP socket\n");
 275                 return -ENOPROTOOPT;
 276         }
 277
 278         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 279                 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
 280                          *len, sizeof(struct sockaddr_in));
 281                 return -EINVAL;
 282         }
 283
 284         h = nf_conntrack_find_get(sock_net(sk), NF_CT_DEFAULT_ZONE, &tuple);
 285         if (h) {
 286                 struct sockaddr_in sin;
 287                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 288
 289                 sin.sin_family = AF_INET;
 290                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 291                         .tuple.dst.u.tcp.port;
 292                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 293                         .tuple.dst.u3.ip;
 294                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 295
 296                 pr_debug("SO_ORIGINAL_DST: %pI4 %u\n",
 297                          &sin.sin_addr.s_addr, ntohs(sin.sin_port));
 298                 nf_ct_put(ct);
 299                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 300                         return -EFAULT;
 301                 else
 302                         return 0;
 303         }
 304         pr_debug("SO_ORIGINAL_DST: Can't find %pI4/%u-%pI4/%u.\n",
 305                  &tuple.src.u3.ip, ntohs(tuple.src.u.tcp.port),
 306                  &tuple.dst.u3.ip, ntohs(tuple.dst.u.tcp.port));
 307         return -ENOENT;
 308 }
 309
 310 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 311
 312 #include <linux/netfilter/nfnetlink.h>
 313 #include <linux/netfilter/nfnetlink_conntrack.h>
 314
 315 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
 316                                 const struct nf_conntrack_tuple *tuple)
 317 {
 318         NLA_PUT_BE32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip);
 319         NLA_PUT_BE32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip);
 320         return 0;
 321
 322 nla_put_failure:
 323         return -1;
 324 }
 325
 326 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
 327         [CTA_IP_V4_SRC] = { .type = NLA_U32 },
 328         [CTA_IP_V4_DST] = { .type = NLA_U32 },
 329 };
 330
 331 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
 332                                 struct nf_conntrack_tuple *t)
 333 {
 334         if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
 335                 return -EINVAL;
 336
 337         t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
 338         t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
 339
 340         return 0;
 341 }
 342
 343 static int ipv4_nlattr_tuple_size(void)
 344 {
 345         return nla_policy_len(ipv4_nla_policy, CTA_IP_MAX + 1);
 346 }
 347 #endif
 348
 349 static struct nf_sockopt_ops so_getorigdst = {
 350         .pf             = PF_INET,
 351         .get_optmin     = SO_ORIGINAL_DST,
 352         .get_optmax     = SO_ORIGINAL_DST+1,
 353         .get            = &getorigdst,
 354         .owner          = THIS_MODULE,
 355 };
 356
 357 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
 358         .l3proto         = PF_INET,
 359         .name            = "ipv4",
 360         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 361         .invert_tuple    = ipv4_invert_tuple,
 362         .print_tuple     = ipv4_print_tuple,
 363         .get_l4proto     = ipv4_get_l4proto,
 364 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 365         .tuple_to_nlattr = ipv4_tuple_to_nlattr,
 366         .nlattr_tuple_size = ipv4_nlattr_tuple_size,
 367         .nlattr_to_tuple = ipv4_nlattr_to_tuple,
 368         .nla_policy      = ipv4_nla_policy,
 369 #endif
 370 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 371         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
 372         .ctl_table       = ip_ct_sysctl_table,
 373 #endif
 374         .me              = THIS_MODULE,
 375 };
 376
 377 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
 378                   &nf_conntrack_htable_size, 0600);
 379
 380 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 381 MODULE_ALIAS("ip_conntrack");
 382 MODULE_LICENSE("GPL");
 383
 384 static int __init nf_conntrack_l3proto_ipv4_init(void)
 385 {
 386         int ret = 0;
 387
 388         need_conntrack();
 389         nf_defrag_ipv4_enable();
 390
 391         ret = nf_register_sockopt(&so_getorigdst);
 392         if (ret < 0) {
 393                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 394                 return ret;
 395         }
 396
 397         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 398         if (ret < 0) {
 399                 pr_err("nf_conntrack_ipv4: can't register tcp.\n");
 400                 goto cleanup_sockopt;
 401         }
 402
 403         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 404         if (ret < 0) {
 405                 pr_err("nf_conntrack_ipv4: can't register udp.\n");
 406                 goto cleanup_tcp;
 407         }
 408
 409         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 410         if (ret < 0) {
 411                 pr_err("nf_conntrack_ipv4: can't register icmp.\n");
 412                 goto cleanup_udp;
 413         }
 414
 415         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 416         if (ret < 0) {
 417                 pr_err("nf_conntrack_ipv4: can't register ipv4\n");
 418                 goto cleanup_icmp;
 419         }
 420
 421         ret = nf_register_hooks(ipv4_conntrack_ops,
 422                                 ARRAY_SIZE(ipv4_conntrack_ops));
 423         if (ret < 0) {
 424                 pr_err("nf_conntrack_ipv4: can't register hooks.\n");
 425                 goto cleanup_ipv4;
 426         }
 427 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 428         ret = nf_conntrack_ipv4_compat_init();
 429         if (ret < 0)
 430                 goto cleanup_hooks;
 431 #endif
 432         return ret;
 433 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 434  cleanup_hooks:
 435         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 436 #endif
 437  cleanup_ipv4:
 438         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 439  cleanup_icmp:
 440         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 441  cleanup_udp:
 442         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 443  cleanup_tcp:
 444         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 445  cleanup_sockopt:
 446         nf_unregister_sockopt(&so_getorigdst);
 447         return ret;
 448 }
 449
 450 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 451 {
 452         synchronize_net();
 453 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 454         nf_conntrack_ipv4_compat_fini();
 455 #endif
 456         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 457         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 458         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 459         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 460         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 461         nf_unregister_sockopt(&so_getorigdst);
 462 }
 463
 464 module_init(nf_conntrack_l3proto_ipv4_init);
 465 module_exit(nf_conntrack_l3proto_ipv4_fini);
 466
 467 void need_ipv4_conntrack(void)
 468 {
 469         return;
 470 }
 471 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);