SAFE public projects git trees. - safe/jmp/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  * 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         u_int32_t _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 nat_module_is_loaded = 0;
 115 static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
 116 {
 117         if (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 #ifdef CONFIG_SYSCTL
 270 /* From nf_conntrack_proto_icmp.c */
 271 extern unsigned int nf_ct_icmp_timeout;
 272 static struct ctl_table_header *nf_ct_ipv4_sysctl_header;
 273
 274 static ctl_table nf_ct_sysctl_table[] = {
 275         {
 276                 .ctl_name       = NET_NF_CONNTRACK_ICMP_TIMEOUT,
 277                 .procname       = "nf_conntrack_icmp_timeout",
 278                 .data           = &nf_ct_icmp_timeout,
 279                 .maxlen         = sizeof(unsigned int),
 280                 .mode           = 0644,
 281                 .proc_handler   = &proc_dointvec_jiffies,
 282         },
 283         { .ctl_name = 0 }
 284 };
 285
 286 static ctl_table nf_ct_netfilter_table[] = {
 287         {
 288                 .ctl_name       = NET_NETFILTER,
 289                 .procname       = "netfilter",
 290                 .mode           = 0555,
 291                 .child          = nf_ct_sysctl_table,
 292         },
 293         { .ctl_name = 0 }
 294 };
 295
 296 static ctl_table nf_ct_net_table[] = {
 297         {
 298                 .ctl_name       = CTL_NET,
 299                 .procname       = "net",
 300                 .mode           = 0555,
 301                 .child          = nf_ct_netfilter_table,
 302         },
 303         { .ctl_name = 0 }
 304 };
 305 #endif
 306
 307 /* Fast function for those who don't want to parse /proc (and I don't
 308    blame them). */
 309 /* Reversing the socket's dst/src point of view gives us the reply
 310    mapping. */
 311 static int
 312 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 313 {
 314         struct inet_sock *inet = inet_sk(sk);
 315         struct nf_conntrack_tuple_hash *h;
 316         struct nf_conntrack_tuple tuple;
 317
 318         NF_CT_TUPLE_U_BLANK(&tuple);
 319         tuple.src.u3.ip = inet->rcv_saddr;
 320         tuple.src.u.tcp.port = inet->sport;
 321         tuple.dst.u3.ip = inet->daddr;
 322         tuple.dst.u.tcp.port = inet->dport;
 323         tuple.src.l3num = PF_INET;
 324         tuple.dst.protonum = IPPROTO_TCP;
 325
 326         /* We only do TCP at the moment: is there a better way? */
 327         if (strcmp(sk->sk_prot->name, "TCP")) {
 328                 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
 329                 return -ENOPROTOOPT;
 330         }
 331
 332         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 333                 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
 334                        *len, sizeof(struct sockaddr_in));
 335                 return -EINVAL;
 336         }
 337
 338         h = nf_conntrack_find_get(&tuple, NULL);
 339         if (h) {
 340                 struct sockaddr_in sin;
 341                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 342
 343                 sin.sin_family = AF_INET;
 344                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 345                         .tuple.dst.u.tcp.port;
 346                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 347                         .tuple.dst.u3.ip;
 348                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 349
 350                 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 351                        NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 352                 nf_ct_put(ct);
 353                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 354                         return -EFAULT;
 355                 else
 356                         return 0;
 357         }
 358         DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 359                NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 360                NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 361         return -ENOENT;
 362 }
 363
 364 #if defined(CONFIG_NF_CT_NETLINK) || \
 365     defined(CONFIG_NF_CT_NETLINK_MODULE)
 366
 367 #include <linux/netfilter/nfnetlink.h>
 368 #include <linux/netfilter/nfnetlink_conntrack.h>
 369
 370 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
 371                                 const struct nf_conntrack_tuple *tuple)
 372 {
 373         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
 374                 &tuple->src.u3.ip);
 375         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
 376                 &tuple->dst.u3.ip);
 377         return 0;
 378
 379 nfattr_failure:
 380         return -1;
 381 }
 382
 383 static const size_t cta_min_ip[CTA_IP_MAX] = {
 384         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 385         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 386 };
 387
 388 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
 389                                 struct nf_conntrack_tuple *t)
 390 {
 391         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
 392                 return -EINVAL;
 393
 394         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 395                 return -EINVAL;
 396
 397         t->src.u3.ip =
 398                 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 399         t->dst.u3.ip =
 400                 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 401
 402         return 0;
 403 }
 404 #endif
 405
 406 static struct nf_sockopt_ops so_getorigdst = {
 407         .pf             = PF_INET,
 408         .get_optmin     = SO_ORIGINAL_DST,
 409         .get_optmax     = SO_ORIGINAL_DST+1,
 410         .get            = &getorigdst,
 411 };
 412
 413 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
 414         .l3proto         = PF_INET,
 415         .name            = "ipv4",
 416         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 417         .invert_tuple    = ipv4_invert_tuple,
 418         .print_tuple     = ipv4_print_tuple,
 419         .print_conntrack = ipv4_print_conntrack,
 420         .prepare         = ipv4_prepare,
 421         .get_features    = ipv4_get_features,
 422 #if defined(CONFIG_NF_CT_NETLINK) || \
 423     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         .me              = THIS_MODULE,
 428 };
 429
 430 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 431 MODULE_LICENSE("GPL");
 432
 433 static int __init nf_conntrack_l3proto_ipv4_init(void)
 434 {
 435         int ret = 0;
 436
 437         need_conntrack();
 438
 439         ret = nf_register_sockopt(&so_getorigdst);
 440         if (ret < 0) {
 441                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 442                 return ret;
 443         }
 444
 445         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 446         if (ret < 0) {
 447                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 448                 goto cleanup_sockopt;
 449         }
 450
 451         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 452         if (ret < 0) {
 453                 printk("nf_conntrack_ipv4: can't register udp.\n");
 454                 goto cleanup_tcp;
 455         }
 456
 457         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 458         if (ret < 0) {
 459                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 460                 goto cleanup_udp;
 461         }
 462
 463         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 464         if (ret < 0) {
 465                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 466                 goto cleanup_icmp;
 467         }
 468
 469         ret = nf_register_hooks(ipv4_conntrack_ops,
 470                                 ARRAY_SIZE(ipv4_conntrack_ops));
 471         if (ret < 0) {
 472                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 473                 goto cleanup_ipv4;
 474         }
 475 #ifdef CONFIG_SYSCTL
 476         nf_ct_ipv4_sysctl_header = register_sysctl_table(nf_ct_net_table, 0);
 477         if (nf_ct_ipv4_sysctl_header == NULL) {
 478                 printk("nf_conntrack: can't register to sysctl.\n");
 479                 ret = -ENOMEM;
 480                 goto cleanup_hooks;
 481         }
 482 #endif
 483         return ret;
 484
 485 #ifdef CONFIG_SYSCTL
 486  cleanup_hooks:
 487         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 488 #endif
 489  cleanup_ipv4:
 490         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 491  cleanup_icmp:
 492         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 493  cleanup_udp:
 494         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 495  cleanup_tcp:
 496         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 497  cleanup_sockopt:
 498         nf_unregister_sockopt(&so_getorigdst);
 499         return ret;
 500 }
 501
 502 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 503 {
 504         synchronize_net();
 505 #ifdef CONFIG_SYSCTL
 506         unregister_sysctl_table(nf_ct_ipv4_sysctl_header);
 507 #endif
 508         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 509         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 510         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 511         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 512         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 513         nf_unregister_sockopt(&so_getorigdst);
 514 }
 515
 516 module_init(nf_conntrack_l3proto_ipv4_init);
 517 module_exit(nf_conntrack_l3proto_ipv4_fini);
 518
 519 EXPORT_SYMBOL(nf_ct_ipv4_gather_frags);