net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c

   1 /* (C) 1999-2001 Paul `Rusty' Russell
   2  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
   3  *
   4  * This program is free software; you can redistribute it and/or modify
   5  * it under the terms of the GNU General Public License version 2 as
   6  * published by the Free Software Foundation.
   7  */
   8
   9 #include <linux/types.h>
  10 #include <linux/ip.h>
  11 #include <linux/netfilter.h>
  12 #include <linux/module.h>
  13 #include <linux/skbuff.h>
  14 #include <linux/icmp.h>
  15 #include <linux/sysctl.h>
  16 #include <net/route.h>
  17 #include <net/ip.h>
  18
  19 #include <linux/netfilter_ipv4.h>
  20 #include <net/netfilter/nf_conntrack.h>
  21 #include <net/netfilter/nf_conntrack_helper.h>
  22 #include <net/netfilter/nf_conntrack_l4proto.h>
  23 #include <net/netfilter/nf_conntrack_l3proto.h>
  24 #include <net/netfilter/nf_conntrack_core.h>
  25 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
  26
  27 #if 0
  28 #define DEBUGP printk
  29 #else
  30 #define DEBUGP(format, args...)
  31 #endif
  32
  33 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  34                              struct nf_conntrack_tuple *tuple)
  35 {
  36         __be32 _addrs[2], *ap;
  37         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  38                                 sizeof(u_int32_t) * 2, _addrs);
  39         if (ap == NULL)
  40                 return 0;
  41
  42         tuple->src.u3.ip = ap[0];
  43         tuple->dst.u3.ip = ap[1];
  44
  45         return 1;
  46 }
  47
  48 static int 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 1;
  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=%u.%u.%u.%u dst=%u.%u.%u.%u ",
  61                           NIPQUAD(tuple->src.u3.ip),
  62                           NIPQUAD(tuple->dst.u3.ip));
  63 }
  64
  65 static int ipv4_print_conntrack(struct seq_file *s,
  66                                 const struct nf_conn *conntrack)
  67 {
  68         return 0;
  69 }
  70
  71 /* Returns new sk_buff, or NULL */
  72 static struct sk_buff *
  73 nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
  74 {
  75         skb_orphan(skb);
  76
  77         local_bh_disable();
  78         skb = ip_defrag(skb, user);
  79         local_bh_enable();
  80
  81         if (skb)
  82                 ip_send_check(ip_hdr(skb));
  83
  84         return skb;
  85 }
  86
  87 static int
  88 ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
  89              u_int8_t *protonum)
  90 {
  91         /* Never happen */
  92         if (ip_hdr(*pskb)->frag_off & htons(IP_OFFSET)) {
  93                 if (net_ratelimit()) {
  94                         printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
  95                         ip_hdr(*pskb)->protocol, hooknum);
  96                 }
  97                 return -NF_DROP;
  98         }
  99
 100         *dataoff = skb_network_offset(*pskb) + ip_hdrlen(*pskb);
 101         *protonum = ip_hdr(*pskb)->protocol;
 102
 103         return NF_ACCEPT;
 104 }
 105
 106 int nf_nat_module_is_loaded = 0;
 107 EXPORT_SYMBOL_GPL(nf_nat_module_is_loaded);
 108
 109 static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
 110 {
 111         if (nf_nat_module_is_loaded)
 112                 return NF_CT_F_NAT;
 113
 114         return NF_CT_F_BASIC;
 115 }
 116
 117 static unsigned int ipv4_confirm(unsigned int hooknum,
 118                                  struct sk_buff **pskb,
 119                                  const struct net_device *in,
 120                                  const struct net_device *out,
 121                                  int (*okfn)(struct sk_buff *))
 122 {
 123         /* We've seen it coming out the other side: confirm it */
 124         return nf_conntrack_confirm(pskb);
 125 }
 126
 127 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
 128                                       struct sk_buff **pskb,
 129                                       const struct net_device *in,
 130                                       const struct net_device *out,
 131                                       int (*okfn)(struct sk_buff *))
 132 {
 133         struct nf_conn *ct;
 134         enum ip_conntrack_info ctinfo;
 135         struct nf_conn_help *help;
 136         struct nf_conntrack_helper *helper;
 137
 138         /* This is where we call the helper: as the packet goes out. */
 139         ct = nf_ct_get(*pskb, &ctinfo);
 140         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 141                 return NF_ACCEPT;
 142
 143         help = nfct_help(ct);
 144         if (!help)
 145                 return NF_ACCEPT;
 146         /* rcu_read_lock()ed by nf_hook_slow */
 147         helper = rcu_dereference(help->helper);
 148         if (!helper)
 149                 return NF_ACCEPT;
 150         return helper->help(pskb, skb_network_offset(*pskb) + ip_hdrlen(*pskb),
 151                             ct, ctinfo);
 152 }
 153
 154 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
 155                                           struct sk_buff **pskb,
 156                                           const struct net_device *in,
 157                                           const struct net_device *out,
 158                                           int (*okfn)(struct sk_buff *))
 159 {
 160         /* Previously seen (loopback)?  Ignore.  Do this before
 161            fragment check. */
 162         if ((*pskb)->nfct)
 163                 return NF_ACCEPT;
 164
 165         /* Gather fragments. */
 166         if (ip_hdr(*pskb)->frag_off & htons(IP_MF | IP_OFFSET)) {
 167                 *pskb = nf_ct_ipv4_gather_frags(*pskb,
 168                                                 hooknum == NF_IP_PRE_ROUTING ?
 169                                                 IP_DEFRAG_CONNTRACK_IN :
 170                                                 IP_DEFRAG_CONNTRACK_OUT);
 171                 if (!*pskb)
 172                         return NF_STOLEN;
 173         }
 174         return NF_ACCEPT;
 175 }
 176
 177 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 178                                       struct sk_buff **pskb,
 179                                       const struct net_device *in,
 180                                       const struct net_device *out,
 181                                       int (*okfn)(struct sk_buff *))
 182 {
 183         return nf_conntrack_in(PF_INET, hooknum, pskb);
 184 }
 185
 186 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 187                                          struct sk_buff **pskb,
 188                                          const struct net_device *in,
 189                                          const struct net_device *out,
 190                                          int (*okfn)(struct sk_buff *))
 191 {
 192         /* root is playing with raw sockets. */
 193         if ((*pskb)->len < sizeof(struct iphdr)
 194             || ip_hdrlen(*pskb) < sizeof(struct iphdr)) {
 195                 if (net_ratelimit())
 196                         printk("ipt_hook: happy cracking.\n");
 197                 return NF_ACCEPT;
 198         }
 199         return nf_conntrack_in(PF_INET, hooknum, pskb);
 200 }
 201
 202 /* Connection tracking may drop packets, but never alters them, so
 203    make it the first hook. */
 204 static struct nf_hook_ops ipv4_conntrack_ops[] = {
 205         {
 206                 .hook           = ipv4_conntrack_defrag,
 207                 .owner          = THIS_MODULE,
 208                 .pf             = PF_INET,
 209                 .hooknum        = NF_IP_PRE_ROUTING,
 210                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 211         },
 212         {
 213                 .hook           = ipv4_conntrack_in,
 214                 .owner          = THIS_MODULE,
 215                 .pf             = PF_INET,
 216                 .hooknum        = NF_IP_PRE_ROUTING,
 217                 .priority       = NF_IP_PRI_CONNTRACK,
 218         },
 219         {
 220                 .hook           = ipv4_conntrack_defrag,
 221                 .owner          = THIS_MODULE,
 222                 .pf             = PF_INET,
 223                 .hooknum        = NF_IP_LOCAL_OUT,
 224                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 225         },
 226         {
 227                 .hook           = ipv4_conntrack_local,
 228                 .owner          = THIS_MODULE,
 229                 .pf             = PF_INET,
 230                 .hooknum        = NF_IP_LOCAL_OUT,
 231                 .priority       = NF_IP_PRI_CONNTRACK,
 232         },
 233         {
 234                 .hook           = ipv4_conntrack_help,
 235                 .owner          = THIS_MODULE,
 236                 .pf             = PF_INET,
 237                 .hooknum        = NF_IP_POST_ROUTING,
 238                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 239         },
 240         {
 241                 .hook           = ipv4_conntrack_help,
 242                 .owner          = THIS_MODULE,
 243                 .pf             = PF_INET,
 244                 .hooknum        = NF_IP_LOCAL_IN,
 245                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 246         },
 247         {
 248                 .hook           = ipv4_confirm,
 249                 .owner          = THIS_MODULE,
 250                 .pf             = PF_INET,
 251                 .hooknum        = NF_IP_POST_ROUTING,
 252                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 253         },
 254         {
 255                 .hook           = ipv4_confirm,
 256                 .owner          = THIS_MODULE,
 257                 .pf             = PF_INET,
 258                 .hooknum        = NF_IP_LOCAL_IN,
 259                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 260         },
 261 };
 262
 263 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 264 static int log_invalid_proto_min = 0;
 265 static int log_invalid_proto_max = 255;
 266
 267 static ctl_table ip_ct_sysctl_table[] = {
 268         {
 269                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 270                 .procname       = "ip_conntrack_max",
 271                 .data           = &nf_conntrack_max,
 272                 .maxlen         = sizeof(int),
 273                 .mode           = 0644,
 274                 .proc_handler   = &proc_dointvec,
 275         },
 276         {
 277                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 278                 .procname       = "ip_conntrack_count",
 279                 .data           = &nf_conntrack_count,
 280                 .maxlen         = sizeof(int),
 281                 .mode           = 0444,
 282                 .proc_handler   = &proc_dointvec,
 283         },
 284         {
 285                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 286                 .procname       = "ip_conntrack_buckets",
 287                 .data           = &nf_conntrack_htable_size,
 288                 .maxlen         = sizeof(unsigned int),
 289                 .mode           = 0444,
 290                 .proc_handler   = &proc_dointvec,
 291         },
 292         {
 293                 .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
 294                 .procname       = "ip_conntrack_checksum",
 295                 .data           = &nf_conntrack_checksum,
 296                 .maxlen         = sizeof(int),
 297                 .mode           = 0644,
 298                 .proc_handler   = &proc_dointvec,
 299         },
 300         {
 301                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 302                 .procname       = "ip_conntrack_log_invalid",
 303                 .data           = &nf_ct_log_invalid,
 304                 .maxlen         = sizeof(unsigned int),
 305                 .mode           = 0644,
 306                 .proc_handler   = &proc_dointvec_minmax,
 307                 .strategy       = &sysctl_intvec,
 308                 .extra1         = &log_invalid_proto_min,
 309                 .extra2         = &log_invalid_proto_max,
 310         },
 311         {
 312                 .ctl_name       = 0
 313         }
 314 };
 315 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 316
 317 /* Fast function for those who don't want to parse /proc (and I don't
 318    blame them). */
 319 /* Reversing the socket's dst/src point of view gives us the reply
 320    mapping. */
 321 static int
 322 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 323 {
 324         struct inet_sock *inet = inet_sk(sk);
 325         struct nf_conntrack_tuple_hash *h;
 326         struct nf_conntrack_tuple tuple;
 327
 328         NF_CT_TUPLE_U_BLANK(&tuple);
 329         tuple.src.u3.ip = inet->rcv_saddr;
 330         tuple.src.u.tcp.port = inet->sport;
 331         tuple.dst.u3.ip = inet->daddr;
 332         tuple.dst.u.tcp.port = inet->dport;
 333         tuple.src.l3num = PF_INET;
 334         tuple.dst.protonum = IPPROTO_TCP;
 335
 336         /* We only do TCP at the moment: is there a better way? */
 337         if (strcmp(sk->sk_prot->name, "TCP")) {
 338                 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
 339                 return -ENOPROTOOPT;
 340         }
 341
 342         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 343                 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
 344                        *len, sizeof(struct sockaddr_in));
 345                 return -EINVAL;
 346         }
 347
 348         h = nf_conntrack_find_get(&tuple, NULL);
 349         if (h) {
 350                 struct sockaddr_in sin;
 351                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 352
 353                 sin.sin_family = AF_INET;
 354                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 355                         .tuple.dst.u.tcp.port;
 356                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 357                         .tuple.dst.u3.ip;
 358                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 359
 360                 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 361                        NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 362                 nf_ct_put(ct);
 363                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 364                         return -EFAULT;
 365                 else
 366                         return 0;
 367         }
 368         DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 369                NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 370                NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 371         return -ENOENT;
 372 }
 373
 374 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 375
 376 #include <linux/netfilter/nfnetlink.h>
 377 #include <linux/netfilter/nfnetlink_conntrack.h>
 378
 379 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
 380                                 const struct nf_conntrack_tuple *tuple)
 381 {
 382         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
 383                 &tuple->src.u3.ip);
 384         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
 385                 &tuple->dst.u3.ip);
 386         return 0;
 387
 388 nfattr_failure:
 389         return -1;
 390 }
 391
 392 static const size_t cta_min_ip[CTA_IP_MAX] = {
 393         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 394         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 395 };
 396
 397 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
 398                                 struct nf_conntrack_tuple *t)
 399 {
 400         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
 401                 return -EINVAL;
 402
 403         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 404                 return -EINVAL;
 405
 406         t->src.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 407         t->dst.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 408
 409         return 0;
 410 }
 411 #endif
 412
 413 static struct nf_sockopt_ops so_getorigdst = {
 414         .pf             = PF_INET,
 415         .get_optmin     = SO_ORIGINAL_DST,
 416         .get_optmax     = SO_ORIGINAL_DST+1,
 417         .get            = &getorigdst,
 418 };
 419
 420 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
 421         .l3proto         = PF_INET,
 422         .name            = "ipv4",
 423         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 424         .invert_tuple    = ipv4_invert_tuple,
 425         .print_tuple     = ipv4_print_tuple,
 426         .print_conntrack = ipv4_print_conntrack,
 427         .prepare         = ipv4_prepare,
 428         .get_features    = ipv4_get_features,
 429 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 430         .tuple_to_nfattr = ipv4_tuple_to_nfattr,
 431         .nfattr_to_tuple = ipv4_nfattr_to_tuple,
 432 #endif
 433 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 434         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
 435         .ctl_table       = ip_ct_sysctl_table,
 436 #endif
 437         .me              = THIS_MODULE,
 438 };
 439
 440 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 441 MODULE_ALIAS("ip_conntrack");
 442 MODULE_LICENSE("GPL");
 443
 444 static int __init nf_conntrack_l3proto_ipv4_init(void)
 445 {
 446         int ret = 0;
 447
 448         need_conntrack();
 449
 450         ret = nf_register_sockopt(&so_getorigdst);
 451         if (ret < 0) {
 452                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 453                 return ret;
 454         }
 455
 456         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 457         if (ret < 0) {
 458                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 459                 goto cleanup_sockopt;
 460         }
 461
 462         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 463         if (ret < 0) {
 464                 printk("nf_conntrack_ipv4: can't register udp.\n");
 465                 goto cleanup_tcp;
 466         }
 467
 468         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 469         if (ret < 0) {
 470                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 471                 goto cleanup_udp;
 472         }
 473
 474         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 475         if (ret < 0) {
 476                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 477                 goto cleanup_icmp;
 478         }
 479
 480         ret = nf_register_hooks(ipv4_conntrack_ops,
 481                                 ARRAY_SIZE(ipv4_conntrack_ops));
 482         if (ret < 0) {
 483                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 484                 goto cleanup_ipv4;
 485         }
 486 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 487         ret = nf_conntrack_ipv4_compat_init();
 488         if (ret < 0)
 489                 goto cleanup_hooks;
 490 #endif
 491         return ret;
 492 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 493  cleanup_hooks:
 494         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 495 #endif
 496  cleanup_ipv4:
 497         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 498  cleanup_icmp:
 499         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 500  cleanup_udp:
 501         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 502  cleanup_tcp:
 503         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 504  cleanup_sockopt:
 505         nf_unregister_sockopt(&so_getorigdst);
 506         return ret;
 507 }
 508
 509 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 510 {
 511         synchronize_net();
 512 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 513         nf_conntrack_ipv4_compat_fini();
 514 #endif
 515         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 516         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 517         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 518         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 519         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 520         nf_unregister_sockopt(&so_getorigdst);
 521 }
 522
 523 module_init(nf_conntrack_l3proto_ipv4_init);
 524 module_exit(nf_conntrack_l3proto_ipv4_fini);