git.oblomov.eu Git - linux-2.6/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_core.h>
  26 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
  27 #include <net/netfilter/nf_nat_helper.h>
  28 #include <net/netfilter/ipv4/nf_defrag_ipv4.h>
  29
  30 int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb,
  31                               struct nf_conn *ct,
  32                               enum ip_conntrack_info ctinfo);
  33 EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook);
  34
  35 static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  36                               struct nf_conntrack_tuple *tuple)
  37 {
  38         const __be32 *ap;
  39         __be32 _addrs[2];
  40         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  41                                 sizeof(u_int32_t) * 2, _addrs);
  42         if (ap == NULL)
  43                 return false;
  44
  45         tuple->src.u3.ip = ap[0];
  46         tuple->dst.u3.ip = ap[1];
  47
  48         return true;
  49 }
  50
  51 static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
  52                               const struct nf_conntrack_tuple *orig)
  53 {
  54         tuple->src.u3.ip = orig->dst.u3.ip;
  55         tuple->dst.u3.ip = orig->src.u3.ip;
  56
  57         return true;
  58 }
  59
  60 static int ipv4_print_tuple(struct seq_file *s,
  61                             const struct nf_conntrack_tuple *tuple)
  62 {
  63         return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
  64                           NIPQUAD(tuple->src.u3.ip),
  65                           NIPQUAD(tuple->dst.u3.ip));
  66 }
  67
  68 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
  69                             unsigned int *dataoff, u_int8_t *protonum)
  70 {
  71         const struct iphdr *iph;
  72         struct iphdr _iph;
  73
  74         iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
  75         if (iph == NULL)
  76                 return -NF_DROP;
  77
  78         /* Conntrack defragments packets, we might still see fragments
  79          * inside ICMP packets though. */
  80         if (iph->frag_off & htons(IP_OFFSET))
  81                 return -NF_DROP;
  82
  83         *dataoff = nhoff + (iph->ihl << 2);
  84         *protonum = iph->protocol;
  85
  86         return NF_ACCEPT;
  87 }
  88
  89 static unsigned int ipv4_confirm(unsigned int hooknum,
  90                                  struct sk_buff *skb,
  91                                  const struct net_device *in,
  92                                  const struct net_device *out,
  93                                  int (*okfn)(struct sk_buff *))
  94 {
  95         struct nf_conn *ct;
  96         enum ip_conntrack_info ctinfo;
  97         const struct nf_conn_help *help;
  98         const struct nf_conntrack_helper *helper;
  99         unsigned int ret;
 100
 101         /* This is where we call the helper: as the packet goes out. */
 102         ct = nf_ct_get(skb, &ctinfo);
 103         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 104                 goto out;
 105
 106         help = nfct_help(ct);
 107         if (!help)
 108                 goto out;
 109
 110         /* rcu_read_lock()ed by nf_hook_slow */
 111         helper = rcu_dereference(help->helper);
 112         if (!helper)
 113                 goto out;
 114
 115         ret = helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
 116                            ct, ctinfo);
 117         if (ret != NF_ACCEPT)
 118                 return ret;
 119
 120         if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 121                 typeof(nf_nat_seq_adjust_hook) seq_adjust;
 122
 123                 seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
 124                 if (!seq_adjust || !seq_adjust(skb, ct, ctinfo))
 125                         return NF_DROP;
 126         }
 127 out:
 128         /* We've seen it coming out the other side: confirm it */
 129         return nf_conntrack_confirm(skb);
 130 }
 131
 132 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 133                                       struct sk_buff *skb,
 134                                       const struct net_device *in,
 135                                       const struct net_device *out,
 136                                       int (*okfn)(struct sk_buff *))
 137 {
 138         return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
 139 }
 140
 141 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 142                                          struct sk_buff *skb,
 143                                          const struct net_device *in,
 144                                          const struct net_device *out,
 145                                          int (*okfn)(struct sk_buff *))
 146 {
 147         /* root is playing with raw sockets. */
 148         if (skb->len < sizeof(struct iphdr) ||
 149             ip_hdrlen(skb) < sizeof(struct iphdr)) {
 150                 if (net_ratelimit())
 151                         printk("ipt_hook: happy cracking.\n");
 152                 return NF_ACCEPT;
 153         }
 154         return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
 155 }
 156
 157 /* Connection tracking may drop packets, but never alters them, so
 158    make it the first hook. */
 159 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
 160         {
 161                 .hook           = ipv4_conntrack_in,
 162                 .owner          = THIS_MODULE,
 163                 .pf             = PF_INET,
 164                 .hooknum        = NF_INET_PRE_ROUTING,
 165                 .priority       = NF_IP_PRI_CONNTRACK,
 166         },
 167         {
 168                 .hook           = ipv4_conntrack_local,
 169                 .owner          = THIS_MODULE,
 170                 .pf             = PF_INET,
 171                 .hooknum        = NF_INET_LOCAL_OUT,
 172                 .priority       = NF_IP_PRI_CONNTRACK,
 173         },
 174         {
 175                 .hook           = ipv4_confirm,
 176                 .owner          = THIS_MODULE,
 177                 .pf             = PF_INET,
 178                 .hooknum        = NF_INET_POST_ROUTING,
 179                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 180         },
 181         {
 182                 .hook           = ipv4_confirm,
 183                 .owner          = THIS_MODULE,
 184                 .pf             = PF_INET,
 185                 .hooknum        = NF_INET_LOCAL_IN,
 186                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 187         },
 188 };
 189
 190 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 191 static int log_invalid_proto_min = 0;
 192 static int log_invalid_proto_max = 255;
 193
 194 static ctl_table ip_ct_sysctl_table[] = {
 195         {
 196                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 197                 .procname       = "ip_conntrack_max",
 198                 .data           = &nf_conntrack_max,
 199                 .maxlen         = sizeof(int),
 200                 .mode           = 0644,
 201                 .proc_handler   = &proc_dointvec,
 202         },
 203         {
 204                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 205                 .procname       = "ip_conntrack_count",
 206                 .data           = &init_net.ct.count,
 207                 .maxlen         = sizeof(int),
 208                 .mode           = 0444,
 209                 .proc_handler   = &proc_dointvec,
 210         },
 211         {
 212                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 213                 .procname       = "ip_conntrack_buckets",
 214                 .data           = &nf_conntrack_htable_size,
 215                 .maxlen         = sizeof(unsigned int),
 216                 .mode           = 0444,
 217                 .proc_handler   = &proc_dointvec,
 218         },
 219         {
 220                 .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
 221                 .procname       = "ip_conntrack_checksum",
 222                 .data           = &init_net.ct.sysctl_checksum,
 223                 .maxlen         = sizeof(int),
 224                 .mode           = 0644,
 225                 .proc_handler   = &proc_dointvec,
 226         },
 227         {
 228                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 229                 .procname       = "ip_conntrack_log_invalid",
 230                 .data           = &init_net.ct.sysctl_log_invalid,
 231                 .maxlen         = sizeof(unsigned int),
 232                 .mode           = 0644,
 233                 .proc_handler   = &proc_dointvec_minmax,
 234                 .strategy       = &sysctl_intvec,
 235                 .extra1         = &log_invalid_proto_min,
 236                 .extra2         = &log_invalid_proto_max,
 237         },
 238         {
 239                 .ctl_name       = 0
 240         }
 241 };
 242 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 243
 244 /* Fast function for those who don't want to parse /proc (and I don't
 245    blame them). */
 246 /* Reversing the socket's dst/src point of view gives us the reply
 247    mapping. */
 248 static int
 249 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 250 {
 251         const struct inet_sock *inet = inet_sk(sk);
 252         const struct nf_conntrack_tuple_hash *h;
 253         struct nf_conntrack_tuple tuple;
 254
 255         memset(&tuple, 0, sizeof(tuple));
 256         tuple.src.u3.ip = inet->rcv_saddr;
 257         tuple.src.u.tcp.port = inet->sport;
 258         tuple.dst.u3.ip = inet->daddr;
 259         tuple.dst.u.tcp.port = inet->dport;
 260         tuple.src.l3num = PF_INET;
 261         tuple.dst.protonum = IPPROTO_TCP;
 262
 263         /* We only do TCP at the moment: is there a better way? */
 264         if (strcmp(sk->sk_prot->name, "TCP")) {
 265                 pr_debug("SO_ORIGINAL_DST: Not a TCP socket\n");
 266                 return -ENOPROTOOPT;
 267         }
 268
 269         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 270                 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
 271                          *len, sizeof(struct sockaddr_in));
 272                 return -EINVAL;
 273         }
 274
 275         h = nf_conntrack_find_get(sock_net(sk), &tuple);
 276         if (h) {
 277                 struct sockaddr_in sin;
 278                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 279
 280                 sin.sin_family = AF_INET;
 281                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 282                         .tuple.dst.u.tcp.port;
 283                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 284                         .tuple.dst.u3.ip;
 285                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 286
 287                 pr_debug("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 288                          NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 289                 nf_ct_put(ct);
 290                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 291                         return -EFAULT;
 292                 else
 293                         return 0;
 294         }
 295         pr_debug("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 296                  NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 297                  NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 298         return -ENOENT;
 299 }
 300
 301 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 302
 303 #include <linux/netfilter/nfnetlink.h>
 304 #include <linux/netfilter/nfnetlink_conntrack.h>
 305
 306 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
 307                                 const struct nf_conntrack_tuple *tuple)
 308 {
 309         NLA_PUT_BE32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip);
 310         NLA_PUT_BE32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip);
 311         return 0;
 312
 313 nla_put_failure:
 314         return -1;
 315 }
 316
 317 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
 318         [CTA_IP_V4_SRC] = { .type = NLA_U32 },
 319         [CTA_IP_V4_DST] = { .type = NLA_U32 },
 320 };
 321
 322 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
 323                                 struct nf_conntrack_tuple *t)
 324 {
 325         if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
 326                 return -EINVAL;
 327
 328         t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
 329         t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
 330
 331         return 0;
 332 }
 333 #endif
 334
 335 static struct nf_sockopt_ops so_getorigdst = {
 336         .pf             = PF_INET,
 337         .get_optmin     = SO_ORIGINAL_DST,
 338         .get_optmax     = SO_ORIGINAL_DST+1,
 339         .get            = &getorigdst,
 340         .owner          = THIS_MODULE,
 341 };
 342
 343 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
 344         .l3proto         = PF_INET,
 345         .name            = "ipv4",
 346         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 347         .invert_tuple    = ipv4_invert_tuple,
 348         .print_tuple     = ipv4_print_tuple,
 349         .get_l4proto     = ipv4_get_l4proto,
 350 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 351         .tuple_to_nlattr = ipv4_tuple_to_nlattr,
 352         .nlattr_to_tuple = ipv4_nlattr_to_tuple,
 353         .nla_policy      = ipv4_nla_policy,
 354 #endif
 355 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 356         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
 357         .ctl_table       = ip_ct_sysctl_table,
 358 #endif
 359         .me              = THIS_MODULE,
 360 };
 361
 362 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
 363                   &nf_conntrack_htable_size, 0600);
 364
 365 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 366 MODULE_ALIAS("ip_conntrack");
 367 MODULE_LICENSE("GPL");
 368
 369 static int __init nf_conntrack_l3proto_ipv4_init(void)
 370 {
 371         int ret = 0;
 372
 373         need_conntrack();
 374         nf_defrag_ipv4_enable();
 375
 376         ret = nf_register_sockopt(&so_getorigdst);
 377         if (ret < 0) {
 378                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 379                 return ret;
 380         }
 381
 382         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 383         if (ret < 0) {
 384                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 385                 goto cleanup_sockopt;
 386         }
 387
 388         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 389         if (ret < 0) {
 390                 printk("nf_conntrack_ipv4: can't register udp.\n");
 391                 goto cleanup_tcp;
 392         }
 393
 394         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 395         if (ret < 0) {
 396                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 397                 goto cleanup_udp;
 398         }
 399
 400         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 401         if (ret < 0) {
 402                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 403                 goto cleanup_icmp;
 404         }
 405
 406         ret = nf_register_hooks(ipv4_conntrack_ops,
 407                                 ARRAY_SIZE(ipv4_conntrack_ops));
 408         if (ret < 0) {
 409                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 410                 goto cleanup_ipv4;
 411         }
 412 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 413         ret = nf_conntrack_ipv4_compat_init();
 414         if (ret < 0)
 415                 goto cleanup_hooks;
 416 #endif
 417         return ret;
 418 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 419  cleanup_hooks:
 420         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 421 #endif
 422  cleanup_ipv4:
 423         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 424  cleanup_icmp:
 425         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 426  cleanup_udp:
 427         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 428  cleanup_tcp:
 429         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 430  cleanup_sockopt:
 431         nf_unregister_sockopt(&so_getorigdst);
 432         return ret;
 433 }
 434
 435 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 436 {
 437         synchronize_net();
 438 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 439         nf_conntrack_ipv4_compat_fini();
 440 #endif
 441         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 442         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 443         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 444         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 445         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 446         nf_unregister_sockopt(&so_getorigdst);
 447 }
 448
 449 module_init(nf_conntrack_l3proto_ipv4_init);
 450 module_exit(nf_conntrack_l3proto_ipv4_fini);
 451
 452 void need_ipv4_conntrack(void)
 453 {
 454         return;
 455 }
 456 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);