git.oblomov.eu Git - linux-2.6/blob - 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
 137         /* This is where we call the helper: as the packet goes out. */
 138         ct = nf_ct_get(*pskb, &ctinfo);
 139         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 140                 return NF_ACCEPT;
 141
 142         help = nfct_help(ct);
 143         if (!help || !help->helper)
 144                 return NF_ACCEPT;
 145
 146         return help->helper->help(pskb,
 147                                   skb_network_offset(*pskb) + ip_hdrlen(*pskb),
 148                                   ct, ctinfo);
 149 }
 150
 151 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
 152                                           struct sk_buff **pskb,
 153                                           const struct net_device *in,
 154                                           const struct net_device *out,
 155                                           int (*okfn)(struct sk_buff *))
 156 {
 157         /* Previously seen (loopback)?  Ignore.  Do this before
 158            fragment check. */
 159         if ((*pskb)->nfct)
 160                 return NF_ACCEPT;
 161
 162         /* Gather fragments. */
 163         if (ip_hdr(*pskb)->frag_off & htons(IP_MF | IP_OFFSET)) {
 164                 *pskb = nf_ct_ipv4_gather_frags(*pskb,
 165                                                 hooknum == NF_IP_PRE_ROUTING ?
 166                                                 IP_DEFRAG_CONNTRACK_IN :
 167                                                 IP_DEFRAG_CONNTRACK_OUT);
 168                 if (!*pskb)
 169                         return NF_STOLEN;
 170         }
 171         return NF_ACCEPT;
 172 }
 173
 174 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 175                                       struct sk_buff **pskb,
 176                                       const struct net_device *in,
 177                                       const struct net_device *out,
 178                                       int (*okfn)(struct sk_buff *))
 179 {
 180         return nf_conntrack_in(PF_INET, hooknum, pskb);
 181 }
 182
 183 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 184                                          struct sk_buff **pskb,
 185                                          const struct net_device *in,
 186                                          const struct net_device *out,
 187                                          int (*okfn)(struct sk_buff *))
 188 {
 189         /* root is playing with raw sockets. */
 190         if ((*pskb)->len < sizeof(struct iphdr)
 191             || ip_hdrlen(*pskb) < sizeof(struct iphdr)) {
 192                 if (net_ratelimit())
 193                         printk("ipt_hook: happy cracking.\n");
 194                 return NF_ACCEPT;
 195         }
 196         return nf_conntrack_in(PF_INET, hooknum, pskb);
 197 }
 198
 199 /* Connection tracking may drop packets, but never alters them, so
 200    make it the first hook. */
 201 static struct nf_hook_ops ipv4_conntrack_ops[] = {
 202         {
 203                 .hook           = ipv4_conntrack_defrag,
 204                 .owner          = THIS_MODULE,
 205                 .pf             = PF_INET,
 206                 .hooknum        = NF_IP_PRE_ROUTING,
 207                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 208         },
 209         {
 210                 .hook           = ipv4_conntrack_in,
 211                 .owner          = THIS_MODULE,
 212                 .pf             = PF_INET,
 213                 .hooknum        = NF_IP_PRE_ROUTING,
 214                 .priority       = NF_IP_PRI_CONNTRACK,
 215         },
 216         {
 217                 .hook           = ipv4_conntrack_defrag,
 218                 .owner          = THIS_MODULE,
 219                 .pf             = PF_INET,
 220                 .hooknum        = NF_IP_LOCAL_OUT,
 221                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 222         },
 223         {
 224                 .hook           = ipv4_conntrack_local,
 225                 .owner          = THIS_MODULE,
 226                 .pf             = PF_INET,
 227                 .hooknum        = NF_IP_LOCAL_OUT,
 228                 .priority       = NF_IP_PRI_CONNTRACK,
 229         },
 230         {
 231                 .hook           = ipv4_conntrack_help,
 232                 .owner          = THIS_MODULE,
 233                 .pf             = PF_INET,
 234                 .hooknum        = NF_IP_POST_ROUTING,
 235                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 236         },
 237         {
 238                 .hook           = ipv4_conntrack_help,
 239                 .owner          = THIS_MODULE,
 240                 .pf             = PF_INET,
 241                 .hooknum        = NF_IP_LOCAL_IN,
 242                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 243         },
 244         {
 245                 .hook           = ipv4_confirm,
 246                 .owner          = THIS_MODULE,
 247                 .pf             = PF_INET,
 248                 .hooknum        = NF_IP_POST_ROUTING,
 249                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 250         },
 251         {
 252                 .hook           = ipv4_confirm,
 253                 .owner          = THIS_MODULE,
 254                 .pf             = PF_INET,
 255                 .hooknum        = NF_IP_LOCAL_IN,
 256                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 257         },
 258 };
 259
 260 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 261 static int log_invalid_proto_min = 0;
 262 static int log_invalid_proto_max = 255;
 263
 264 static ctl_table ip_ct_sysctl_table[] = {
 265         {
 266                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 267                 .procname       = "ip_conntrack_max",
 268                 .data           = &nf_conntrack_max,
 269                 .maxlen         = sizeof(int),
 270                 .mode           = 0644,
 271                 .proc_handler   = &proc_dointvec,
 272         },
 273         {
 274                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 275                 .procname       = "ip_conntrack_count",
 276                 .data           = &nf_conntrack_count,
 277                 .maxlen         = sizeof(int),
 278                 .mode           = 0444,
 279                 .proc_handler   = &proc_dointvec,
 280         },
 281         {
 282                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 283                 .procname       = "ip_conntrack_buckets",
 284                 .data           = &nf_conntrack_htable_size,
 285                 .maxlen         = sizeof(unsigned int),
 286                 .mode           = 0444,
 287                 .proc_handler   = &proc_dointvec,
 288         },
 289         {
 290                 .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
 291                 .procname       = "ip_conntrack_checksum",
 292                 .data           = &nf_conntrack_checksum,
 293                 .maxlen         = sizeof(int),
 294                 .mode           = 0644,
 295                 .proc_handler   = &proc_dointvec,
 296         },
 297         {
 298                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 299                 .procname       = "ip_conntrack_log_invalid",
 300                 .data           = &nf_ct_log_invalid,
 301                 .maxlen         = sizeof(unsigned int),
 302                 .mode           = 0644,
 303                 .proc_handler   = &proc_dointvec_minmax,
 304                 .strategy       = &sysctl_intvec,
 305                 .extra1         = &log_invalid_proto_min,
 306                 .extra2         = &log_invalid_proto_max,
 307         },
 308         {
 309                 .ctl_name       = 0
 310         }
 311 };
 312 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 313
 314 /* Fast function for those who don't want to parse /proc (and I don't
 315    blame them). */
 316 /* Reversing the socket's dst/src point of view gives us the reply
 317    mapping. */
 318 static int
 319 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 320 {
 321         struct inet_sock *inet = inet_sk(sk);
 322         struct nf_conntrack_tuple_hash *h;
 323         struct nf_conntrack_tuple tuple;
 324
 325         NF_CT_TUPLE_U_BLANK(&tuple);
 326         tuple.src.u3.ip = inet->rcv_saddr;
 327         tuple.src.u.tcp.port = inet->sport;
 328         tuple.dst.u3.ip = inet->daddr;
 329         tuple.dst.u.tcp.port = inet->dport;
 330         tuple.src.l3num = PF_INET;
 331         tuple.dst.protonum = IPPROTO_TCP;
 332
 333         /* We only do TCP at the moment: is there a better way? */
 334         if (strcmp(sk->sk_prot->name, "TCP")) {
 335                 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
 336                 return -ENOPROTOOPT;
 337         }
 338
 339         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 340                 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
 341                        *len, sizeof(struct sockaddr_in));
 342                 return -EINVAL;
 343         }
 344
 345         h = nf_conntrack_find_get(&tuple, NULL);
 346         if (h) {
 347                 struct sockaddr_in sin;
 348                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 349
 350                 sin.sin_family = AF_INET;
 351                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 352                         .tuple.dst.u.tcp.port;
 353                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 354                         .tuple.dst.u3.ip;
 355                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 356
 357                 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 358                        NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 359                 nf_ct_put(ct);
 360                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 361                         return -EFAULT;
 362                 else
 363                         return 0;
 364         }
 365         DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 366                NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 367                NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 368         return -ENOENT;
 369 }
 370
 371 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 372
 373 #include <linux/netfilter/nfnetlink.h>
 374 #include <linux/netfilter/nfnetlink_conntrack.h>
 375
 376 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
 377                                 const struct nf_conntrack_tuple *tuple)
 378 {
 379         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
 380                 &tuple->src.u3.ip);
 381         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
 382                 &tuple->dst.u3.ip);
 383         return 0;
 384
 385 nfattr_failure:
 386         return -1;
 387 }
 388
 389 static const size_t cta_min_ip[CTA_IP_MAX] = {
 390         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 391         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 392 };
 393
 394 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
 395                                 struct nf_conntrack_tuple *t)
 396 {
 397         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
 398                 return -EINVAL;
 399
 400         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 401                 return -EINVAL;
 402
 403         t->src.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 404         t->dst.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 405
 406         return 0;
 407 }
 408 #endif
 409
 410 static struct nf_sockopt_ops so_getorigdst = {
 411         .pf             = PF_INET,
 412         .get_optmin     = SO_ORIGINAL_DST,
 413         .get_optmax     = SO_ORIGINAL_DST+1,
 414         .get            = &getorigdst,
 415 };
 416
 417 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
 418         .l3proto         = PF_INET,
 419         .name            = "ipv4",
 420         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 421         .invert_tuple    = ipv4_invert_tuple,
 422         .print_tuple     = ipv4_print_tuple,
 423         .print_conntrack = ipv4_print_conntrack,
 424         .prepare         = ipv4_prepare,
 425         .get_features    = ipv4_get_features,
 426 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 427         .tuple_to_nfattr = ipv4_tuple_to_nfattr,
 428         .nfattr_to_tuple = ipv4_nfattr_to_tuple,
 429 #endif
 430 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 431         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
 432         .ctl_table       = ip_ct_sysctl_table,
 433 #endif
 434         .me              = THIS_MODULE,
 435 };
 436
 437 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 438 MODULE_ALIAS("ip_conntrack");
 439 MODULE_LICENSE("GPL");
 440
 441 static int __init nf_conntrack_l3proto_ipv4_init(void)
 442 {
 443         int ret = 0;
 444
 445         need_conntrack();
 446
 447         ret = nf_register_sockopt(&so_getorigdst);
 448         if (ret < 0) {
 449                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 450                 return ret;
 451         }
 452
 453         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 454         if (ret < 0) {
 455                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 456                 goto cleanup_sockopt;
 457         }
 458
 459         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 460         if (ret < 0) {
 461                 printk("nf_conntrack_ipv4: can't register udp.\n");
 462                 goto cleanup_tcp;
 463         }
 464
 465         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 466         if (ret < 0) {
 467                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 468                 goto cleanup_udp;
 469         }
 470
 471         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 472         if (ret < 0) {
 473                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 474                 goto cleanup_icmp;
 475         }
 476
 477         ret = nf_register_hooks(ipv4_conntrack_ops,
 478                                 ARRAY_SIZE(ipv4_conntrack_ops));
 479         if (ret < 0) {
 480                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 481                 goto cleanup_ipv4;
 482         }
 483 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 484         ret = nf_conntrack_ipv4_compat_init();
 485         if (ret < 0)
 486                 goto cleanup_hooks;
 487 #endif
 488         return ret;
 489 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 490  cleanup_hooks:
 491         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 492 #endif
 493  cleanup_ipv4:
 494         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 495  cleanup_icmp:
 496         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 497  cleanup_udp:
 498         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 499  cleanup_tcp:
 500         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 501  cleanup_sockopt:
 502         nf_unregister_sockopt(&so_getorigdst);
 503         return ret;
 504 }
 505
 506 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 507 {
 508         synchronize_net();
 509 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 510         nf_conntrack_ipv4_compat_fini();
 511 #endif
 512         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 513         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 514         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 515         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 516         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 517         nf_unregister_sockopt(&so_getorigdst);
 518 }
 519
 520 module_init(nf_conntrack_l3proto_ipv4_init);
 521 module_exit(nf_conntrack_l3proto_ipv4_fini);