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