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