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  * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
   9  *      - move L3 protocol dependent part to this file.
  10  * 23 Mar 2004: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
  11  *      - add get_features() to support various size of conntrack
  12  *        structures.
  13  *
  14  * Derived from net/ipv4/netfilter/ip_conntrack_standalone.c
  15  */
  16
  17 #include <linux/types.h>
  18 #include <linux/ip.h>
  19 #include <linux/netfilter.h>
  20 #include <linux/module.h>
  21 #include <linux/skbuff.h>
  22 #include <linux/icmp.h>
  23 #include <linux/sysctl.h>
  24 #include <net/route.h>
  25 #include <net/ip.h>
  26
  27 #include <linux/netfilter_ipv4.h>
  28 #include <net/netfilter/nf_conntrack.h>
  29 #include <net/netfilter/nf_conntrack_helper.h>
  30 #include <net/netfilter/nf_conntrack_l4proto.h>
  31 #include <net/netfilter/nf_conntrack_l3proto.h>
  32 #include <net/netfilter/nf_conntrack_core.h>
  33 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
  34
  35 #if 0
  36 #define DEBUGP printk
  37 #else
  38 #define DEBUGP(format, args...)
  39 #endif
  40
  41 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  42                              struct nf_conntrack_tuple *tuple)
  43 {
  44         __be32 _addrs[2], *ap;
  45         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  46                                 sizeof(u_int32_t) * 2, _addrs);
  47         if (ap == NULL)
  48                 return 0;
  49
  50         tuple->src.u3.ip = ap[0];
  51         tuple->dst.u3.ip = ap[1];
  52
  53         return 1;
  54 }
  55
  56 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
  57                            const struct nf_conntrack_tuple *orig)
  58 {
  59         tuple->src.u3.ip = orig->dst.u3.ip;
  60         tuple->dst.u3.ip = orig->src.u3.ip;
  61
  62         return 1;
  63 }
  64
  65 static int ipv4_print_tuple(struct seq_file *s,
  66                             const struct nf_conntrack_tuple *tuple)
  67 {
  68         return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
  69                           NIPQUAD(tuple->src.u3.ip),
  70                           NIPQUAD(tuple->dst.u3.ip));
  71 }
  72
  73 static int ipv4_print_conntrack(struct seq_file *s,
  74                                 const struct nf_conn *conntrack)
  75 {
  76         return 0;
  77 }
  78
  79 /* Returns new sk_buff, or NULL */
  80 static struct sk_buff *
  81 nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
  82 {
  83         skb_orphan(skb);
  84
  85         local_bh_disable();
  86         skb = ip_defrag(skb, user);
  87         local_bh_enable();
  88
  89         if (skb)
  90                 ip_send_check(skb->nh.iph);
  91
  92         return skb;
  93 }
  94
  95 static int
  96 ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
  97              u_int8_t *protonum)
  98 {
  99         /* Never happen */
 100         if ((*pskb)->nh.iph->frag_off & htons(IP_OFFSET)) {
 101                 if (net_ratelimit()) {
 102                         printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
 103                         (*pskb)->nh.iph->protocol, hooknum);
 104                 }
 105                 return -NF_DROP;
 106         }
 107
 108         *dataoff = (*pskb)->nh.raw - (*pskb)->data + (*pskb)->nh.iph->ihl*4;
 109         *protonum = (*pskb)->nh.iph->protocol;
 110
 111         return NF_ACCEPT;
 112 }
 113
 114 int nf_nat_module_is_loaded = 0;
 115 static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
 116 {
 117         if (nf_nat_module_is_loaded)
 118                 return NF_CT_F_NAT;
 119
 120         return NF_CT_F_BASIC;
 121 }
 122
 123 static unsigned int ipv4_confirm(unsigned int hooknum,
 124                                  struct sk_buff **pskb,
 125                                  const struct net_device *in,
 126                                  const struct net_device *out,
 127                                  int (*okfn)(struct sk_buff *))
 128 {
 129         /* We've seen it coming out the other side: confirm it */
 130         return nf_conntrack_confirm(pskb);
 131 }
 132
 133 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
 134                                       struct sk_buff **pskb,
 135                                       const struct net_device *in,
 136                                       const struct net_device *out,
 137                                       int (*okfn)(struct sk_buff *))
 138 {
 139         struct nf_conn *ct;
 140         enum ip_conntrack_info ctinfo;
 141         struct nf_conn_help *help;
 142
 143         /* This is where we call the helper: as the packet goes out. */
 144         ct = nf_ct_get(*pskb, &ctinfo);
 145         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 146                 return NF_ACCEPT;
 147
 148         help = nfct_help(ct);
 149         if (!help || !help->helper)
 150                 return NF_ACCEPT;
 151
 152         return help->helper->help(pskb,
 153                                (*pskb)->nh.raw - (*pskb)->data
 154                                                + (*pskb)->nh.iph->ihl*4,
 155                                ct, ctinfo);
 156 }
 157
 158 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
 159                                           struct sk_buff **pskb,
 160                                           const struct net_device *in,
 161                                           const struct net_device *out,
 162                                           int (*okfn)(struct sk_buff *))
 163 {
 164 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
 165         /* Previously seen (loopback)?  Ignore.  Do this before
 166            fragment check. */
 167         if ((*pskb)->nfct)
 168                 return NF_ACCEPT;
 169 #endif
 170
 171         /* Gather fragments. */
 172         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 173                 *pskb = nf_ct_ipv4_gather_frags(*pskb,
 174                                                 hooknum == NF_IP_PRE_ROUTING ?
 175                                                 IP_DEFRAG_CONNTRACK_IN :
 176                                                 IP_DEFRAG_CONNTRACK_OUT);
 177                 if (!*pskb)
 178                         return NF_STOLEN;
 179         }
 180         return NF_ACCEPT;
 181 }
 182
 183 static unsigned int ipv4_conntrack_in(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         return nf_conntrack_in(PF_INET, hooknum, pskb);
 190 }
 191
 192 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 193                                          struct sk_buff **pskb,
 194                                          const struct net_device *in,
 195                                          const struct net_device *out,
 196                                          int (*okfn)(struct sk_buff *))
 197 {
 198         /* root is playing with raw sockets. */
 199         if ((*pskb)->len < sizeof(struct iphdr)
 200             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 201                 if (net_ratelimit())
 202                         printk("ipt_hook: happy cracking.\n");
 203                 return NF_ACCEPT;
 204         }
 205         return nf_conntrack_in(PF_INET, hooknum, pskb);
 206 }
 207
 208 /* Connection tracking may drop packets, but never alters them, so
 209    make it the first hook. */
 210 static struct nf_hook_ops ipv4_conntrack_ops[] = {
 211         {
 212                 .hook           = ipv4_conntrack_defrag,
 213                 .owner          = THIS_MODULE,
 214                 .pf             = PF_INET,
 215                 .hooknum        = NF_IP_PRE_ROUTING,
 216                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 217         },
 218         {
 219                 .hook           = ipv4_conntrack_in,
 220                 .owner          = THIS_MODULE,
 221                 .pf             = PF_INET,
 222                 .hooknum        = NF_IP_PRE_ROUTING,
 223                 .priority       = NF_IP_PRI_CONNTRACK,
 224         },
 225         {
 226                 .hook           = ipv4_conntrack_defrag,
 227                 .owner          = THIS_MODULE,
 228                 .pf             = PF_INET,
 229                 .hooknum        = NF_IP_LOCAL_OUT,
 230                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 231         },
 232         {
 233                 .hook           = ipv4_conntrack_local,
 234                 .owner          = THIS_MODULE,
 235                 .pf             = PF_INET,
 236                 .hooknum        = NF_IP_LOCAL_OUT,
 237                 .priority       = NF_IP_PRI_CONNTRACK,
 238         },
 239         {
 240                 .hook           = ipv4_conntrack_help,
 241                 .owner          = THIS_MODULE,
 242                 .pf             = PF_INET,
 243                 .hooknum        = NF_IP_POST_ROUTING,
 244                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 245         },
 246         {
 247                 .hook           = ipv4_conntrack_help,
 248                 .owner          = THIS_MODULE,
 249                 .pf             = PF_INET,
 250                 .hooknum        = NF_IP_LOCAL_IN,
 251                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 252         },
 253         {
 254                 .hook           = ipv4_confirm,
 255                 .owner          = THIS_MODULE,
 256                 .pf             = PF_INET,
 257                 .hooknum        = NF_IP_POST_ROUTING,
 258                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 259         },
 260         {
 261                 .hook           = ipv4_confirm,
 262                 .owner          = THIS_MODULE,
 263                 .pf             = PF_INET,
 264                 .hooknum        = NF_IP_LOCAL_IN,
 265                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 266         },
 267 };
 268
 269 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 270 static int log_invalid_proto_min = 0;
 271 static int log_invalid_proto_max = 255;
 272
 273 static ctl_table ip_ct_sysctl_table[] = {
 274         {
 275                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 276                 .procname       = "ip_conntrack_max",
 277                 .data           = &nf_conntrack_max,
 278                 .maxlen         = sizeof(int),
 279                 .mode           = 0644,
 280                 .proc_handler   = &proc_dointvec,
 281         },
 282         {
 283                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 284                 .procname       = "ip_conntrack_count",
 285                 .data           = &nf_conntrack_count,
 286                 .maxlen         = sizeof(int),
 287                 .mode           = 0444,
 288                 .proc_handler   = &proc_dointvec,
 289         },
 290         {
 291                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 292                 .procname       = "ip_conntrack_buckets",
 293                 .data           = &nf_conntrack_htable_size,
 294                 .maxlen         = sizeof(unsigned int),
 295                 .mode           = 0444,
 296                 .proc_handler   = &proc_dointvec,
 297         },
 298         {
 299                 .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
 300                 .procname       = "ip_conntrack_checksum",
 301                 .data           = &nf_conntrack_checksum,
 302                 .maxlen         = sizeof(int),
 303                 .mode           = 0644,
 304                 .proc_handler   = &proc_dointvec,
 305         },
 306         {
 307                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 308                 .procname       = "ip_conntrack_log_invalid",
 309                 .data           = &nf_ct_log_invalid,
 310                 .maxlen         = sizeof(unsigned int),
 311                 .mode           = 0644,
 312                 .proc_handler   = &proc_dointvec_minmax,
 313                 .strategy       = &sysctl_intvec,
 314                 .extra1         = &log_invalid_proto_min,
 315                 .extra2         = &log_invalid_proto_max,
 316         },
 317         {
 318                 .ctl_name       = 0
 319         }
 320 };
 321 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 322
 323 /* Fast function for those who don't want to parse /proc (and I don't
 324    blame them). */
 325 /* Reversing the socket's dst/src point of view gives us the reply
 326    mapping. */
 327 static int
 328 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 329 {
 330         struct inet_sock *inet = inet_sk(sk);
 331         struct nf_conntrack_tuple_hash *h;
 332         struct nf_conntrack_tuple tuple;
 333
 334         NF_CT_TUPLE_U_BLANK(&tuple);
 335         tuple.src.u3.ip = inet->rcv_saddr;
 336         tuple.src.u.tcp.port = inet->sport;
 337         tuple.dst.u3.ip = inet->daddr;
 338         tuple.dst.u.tcp.port = inet->dport;
 339         tuple.src.l3num = PF_INET;
 340         tuple.dst.protonum = IPPROTO_TCP;
 341
 342         /* We only do TCP at the moment: is there a better way? */
 343         if (strcmp(sk->sk_prot->name, "TCP")) {
 344                 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
 345                 return -ENOPROTOOPT;
 346         }
 347
 348         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 349                 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
 350                        *len, sizeof(struct sockaddr_in));
 351                 return -EINVAL;
 352         }
 353
 354         h = nf_conntrack_find_get(&tuple, NULL);
 355         if (h) {
 356                 struct sockaddr_in sin;
 357                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 358
 359                 sin.sin_family = AF_INET;
 360                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 361                         .tuple.dst.u.tcp.port;
 362                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 363                         .tuple.dst.u3.ip;
 364                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 365
 366                 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 367                        NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 368                 nf_ct_put(ct);
 369                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 370                         return -EFAULT;
 371                 else
 372                         return 0;
 373         }
 374         DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 375                NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 376                NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 377         return -ENOENT;
 378 }
 379
 380 #if defined(CONFIG_NF_CT_NETLINK) || \
 381     defined(CONFIG_NF_CT_NETLINK_MODULE)
 382
 383 #include <linux/netfilter/nfnetlink.h>
 384 #include <linux/netfilter/nfnetlink_conntrack.h>
 385
 386 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
 387                                 const struct nf_conntrack_tuple *tuple)
 388 {
 389         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
 390                 &tuple->src.u3.ip);
 391         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
 392                 &tuple->dst.u3.ip);
 393         return 0;
 394
 395 nfattr_failure:
 396         return -1;
 397 }
 398
 399 static const size_t cta_min_ip[CTA_IP_MAX] = {
 400         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 401         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 402 };
 403
 404 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
 405                                 struct nf_conntrack_tuple *t)
 406 {
 407         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
 408                 return -EINVAL;
 409
 410         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 411                 return -EINVAL;
 412
 413         t->src.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 414         t->dst.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 415
 416         return 0;
 417 }
 418 #endif
 419
 420 static struct nf_sockopt_ops so_getorigdst = {
 421         .pf             = PF_INET,
 422         .get_optmin     = SO_ORIGINAL_DST,
 423         .get_optmax     = SO_ORIGINAL_DST+1,
 424         .get            = &getorigdst,
 425 };
 426
 427 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
 428         .l3proto         = PF_INET,
 429         .name            = "ipv4",
 430         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 431         .invert_tuple    = ipv4_invert_tuple,
 432         .print_tuple     = ipv4_print_tuple,
 433         .print_conntrack = ipv4_print_conntrack,
 434         .prepare         = ipv4_prepare,
 435         .get_features    = ipv4_get_features,
 436 #if defined(CONFIG_NF_CT_NETLINK) || \
 437     defined(CONFIG_NF_CT_NETLINK_MODULE)
 438         .tuple_to_nfattr = ipv4_tuple_to_nfattr,
 439         .nfattr_to_tuple = ipv4_nfattr_to_tuple,
 440 #endif
 441 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 442         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
 443         .ctl_table       = ip_ct_sysctl_table,
 444 #endif
 445         .me              = THIS_MODULE,
 446 };
 447
 448 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 449 MODULE_ALIAS("ip_conntrack");
 450 MODULE_LICENSE("GPL");
 451
 452 static int __init nf_conntrack_l3proto_ipv4_init(void)
 453 {
 454         int ret = 0;
 455
 456         need_conntrack();
 457
 458         ret = nf_register_sockopt(&so_getorigdst);
 459         if (ret < 0) {
 460                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 461                 return ret;
 462         }
 463
 464         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 465         if (ret < 0) {
 466                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 467                 goto cleanup_sockopt;
 468         }
 469
 470         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 471         if (ret < 0) {
 472                 printk("nf_conntrack_ipv4: can't register udp.\n");
 473                 goto cleanup_tcp;
 474         }
 475
 476         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 477         if (ret < 0) {
 478                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 479                 goto cleanup_udp;
 480         }
 481
 482         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 483         if (ret < 0) {
 484                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 485                 goto cleanup_icmp;
 486         }
 487
 488         ret = nf_register_hooks(ipv4_conntrack_ops,
 489                                 ARRAY_SIZE(ipv4_conntrack_ops));
 490         if (ret < 0) {
 491                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 492                 goto cleanup_ipv4;
 493         }
 494 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 495         ret = nf_conntrack_ipv4_compat_init();
 496         if (ret < 0)
 497                 goto cleanup_hooks;
 498 #endif
 499         return ret;
 500 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 501  cleanup_hooks:
 502         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 503 #endif
 504  cleanup_ipv4:
 505         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 506  cleanup_icmp:
 507         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 508  cleanup_udp:
 509         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 510  cleanup_tcp:
 511         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 512  cleanup_sockopt:
 513         nf_unregister_sockopt(&so_getorigdst);
 514         return ret;
 515 }
 516
 517 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 518 {
 519         synchronize_net();
 520 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 521         nf_conntrack_ipv4_compat_fini();
 522 #endif
 523         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 524         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 525         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 526         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 527         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 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 EXPORT_SYMBOL(nf_ct_ipv4_gather_frags);
 535 #ifdef CONFIG_NF_NAT_NEEDED
 536 EXPORT_SYMBOL(nf_nat_module_is_loaded);
 537 #endif