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/config.h>
  18 #include <linux/types.h>
  19 #include <linux/ip.h>
  20 #include <linux/netfilter.h>
  21 #include <linux/module.h>
  22 #include <linux/skbuff.h>
  23 #include <linux/icmp.h>
  24 #include <linux/sysctl.h>
  25 #include <net/route.h>
  26 #include <net/ip.h>
  27
  28 #include <linux/netfilter_ipv4.h>
  29 #include <net/netfilter/nf_conntrack.h>
  30 #include <net/netfilter/nf_conntrack_helper.h>
  31 #include <net/netfilter/nf_conntrack_protocol.h>
  32 #include <net/netfilter/nf_conntrack_l3proto.h>
  33 #include <net/netfilter/nf_conntrack_core.h>
  34 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
  35
  36 #if 0
  37 #define DEBUGP printk
  38 #else
  39 #define DEBUGP(format, args...)
  40 #endif
  41
  42 DECLARE_PER_CPU(struct nf_conntrack_stat, nf_conntrack_stat);
  43
  44 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  45                              struct nf_conntrack_tuple *tuple)
  46 {
  47         u_int32_t _addrs[2], *ap;
  48         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  49                                 sizeof(u_int32_t) * 2, _addrs);
  50         if (ap == NULL)
  51                 return 0;
  52
  53         tuple->src.u3.ip = ap[0];
  54         tuple->dst.u3.ip = ap[1];
  55
  56         return 1;
  57 }
  58
  59 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
  60                            const struct nf_conntrack_tuple *orig)
  61 {
  62         tuple->src.u3.ip = orig->dst.u3.ip;
  63         tuple->dst.u3.ip = orig->src.u3.ip;
  64
  65         return 1;
  66 }
  67
  68 static int ipv4_print_tuple(struct seq_file *s,
  69                             const struct nf_conntrack_tuple *tuple)
  70 {
  71         return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
  72                           NIPQUAD(tuple->src.u3.ip),
  73                           NIPQUAD(tuple->dst.u3.ip));
  74 }
  75
  76 static int ipv4_print_conntrack(struct seq_file *s,
  77                                 const struct nf_conn *conntrack)
  78 {
  79         return 0;
  80 }
  81
  82 /* Returns new sk_buff, or NULL */
  83 static struct sk_buff *
  84 nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
  85 {
  86         skb_orphan(skb);
  87
  88         local_bh_disable();
  89         skb = ip_defrag(skb, user);
  90         local_bh_enable();
  91
  92         if (skb)
  93                 ip_send_check(skb->nh.iph);
  94
  95         return skb;
  96 }
  97
  98 static int
  99 ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
 100              u_int8_t *protonum)
 101 {
 102         /* Never happen */
 103         if ((*pskb)->nh.iph->frag_off & htons(IP_OFFSET)) {
 104                 if (net_ratelimit()) {
 105                         printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
 106                         (*pskb)->nh.iph->protocol, hooknum);
 107                 }
 108                 return -NF_DROP;
 109         }
 110
 111         *dataoff = (*pskb)->nh.raw - (*pskb)->data + (*pskb)->nh.iph->ihl*4;
 112         *protonum = (*pskb)->nh.iph->protocol;
 113
 114         return NF_ACCEPT;
 115 }
 116
 117 int nat_module_is_loaded = 0;
 118 static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
 119 {
 120         if (nat_module_is_loaded)
 121                 return NF_CT_F_NAT;
 122
 123         return NF_CT_F_BASIC;
 124 }
 125
 126 static unsigned int ipv4_confirm(unsigned int hooknum,
 127                                  struct sk_buff **pskb,
 128                                  const struct net_device *in,
 129                                  const struct net_device *out,
 130                                  int (*okfn)(struct sk_buff *))
 131 {
 132         /* We've seen it coming out the other side: confirm it */
 133         return nf_conntrack_confirm(pskb);
 134 }
 135
 136 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
 137                                       struct sk_buff **pskb,
 138                                       const struct net_device *in,
 139                                       const struct net_device *out,
 140                                       int (*okfn)(struct sk_buff *))
 141 {
 142         struct nf_conn *ct;
 143         enum ip_conntrack_info ctinfo;
 144         struct nf_conn_help *help;
 145
 146         /* This is where we call the helper: as the packet goes out. */
 147         ct = nf_ct_get(*pskb, &ctinfo);
 148         if (!ct)
 149                 return NF_ACCEPT;
 150
 151         help = nfct_help(ct);
 152         if (!help || !help->helper)
 153                 return NF_ACCEPT;
 154
 155         return help->helper->help(pskb,
 156                                (*pskb)->nh.raw - (*pskb)->data
 157                                                + (*pskb)->nh.iph->ihl*4,
 158                                ct, ctinfo);
 159 }
 160
 161 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
 162                                           struct sk_buff **pskb,
 163                                           const struct net_device *in,
 164                                           const struct net_device *out,
 165                                           int (*okfn)(struct sk_buff *))
 166 {
 167 #if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
 168         /* Previously seen (loopback)?  Ignore.  Do this before
 169            fragment check. */
 170         if ((*pskb)->nfct)
 171                 return NF_ACCEPT;
 172 #endif
 173
 174         /* Gather fragments. */
 175         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 176                 *pskb = nf_ct_ipv4_gather_frags(*pskb,
 177                                                 hooknum == NF_IP_PRE_ROUTING ?
 178                                                 IP_DEFRAG_CONNTRACK_IN :
 179                                                 IP_DEFRAG_CONNTRACK_OUT);
 180                 if (!*pskb)
 181                         return NF_STOLEN;
 182         }
 183         return NF_ACCEPT;
 184 }
 185
 186 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 187                                       struct sk_buff **pskb,
 188                                       const struct net_device *in,
 189                                       const struct net_device *out,
 190                                       int (*okfn)(struct sk_buff *))
 191 {
 192         return nf_conntrack_in(PF_INET, hooknum, pskb);
 193 }
 194
 195 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 196                                          struct sk_buff **pskb,
 197                                          const struct net_device *in,
 198                                          const struct net_device *out,
 199                                          int (*okfn)(struct sk_buff *))
 200 {
 201         /* root is playing with raw sockets. */
 202         if ((*pskb)->len < sizeof(struct iphdr)
 203             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
 204                 if (net_ratelimit())
 205                         printk("ipt_hook: happy cracking.\n");
 206                 return NF_ACCEPT;
 207         }
 208         return nf_conntrack_in(PF_INET, hooknum, pskb);
 209 }
 210
 211 /* Connection tracking may drop packets, but never alters them, so
 212    make it the first hook. */
 213 static struct nf_hook_ops ipv4_conntrack_ops[] = {
 214         {
 215                 .hook           = ipv4_conntrack_defrag,
 216                 .owner          = THIS_MODULE,
 217                 .pf             = PF_INET,
 218                 .hooknum        = NF_IP_PRE_ROUTING,
 219                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 220         },
 221         {
 222                 .hook           = ipv4_conntrack_in,
 223                 .owner          = THIS_MODULE,
 224                 .pf             = PF_INET,
 225                 .hooknum        = NF_IP_PRE_ROUTING,
 226                 .priority       = NF_IP_PRI_CONNTRACK,
 227         },
 228         {
 229                 .hook           = ipv4_conntrack_defrag,
 230                 .owner          = THIS_MODULE,
 231                 .pf             = PF_INET,
 232                 .hooknum        = NF_IP_LOCAL_OUT,
 233                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 234         },
 235         {
 236                 .hook           = ipv4_conntrack_local,
 237                 .owner          = THIS_MODULE,
 238                 .pf             = PF_INET,
 239                 .hooknum        = NF_IP_LOCAL_OUT,
 240                 .priority       = NF_IP_PRI_CONNTRACK,
 241         },
 242         {
 243                 .hook           = ipv4_conntrack_help,
 244                 .owner          = THIS_MODULE,
 245                 .pf             = PF_INET,
 246                 .hooknum        = NF_IP_POST_ROUTING,
 247                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 248         },
 249         {
 250                 .hook           = ipv4_conntrack_help,
 251                 .owner          = THIS_MODULE,
 252                 .pf             = PF_INET,
 253                 .hooknum        = NF_IP_LOCAL_IN,
 254                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 255         },
 256         {
 257                 .hook           = ipv4_confirm,
 258                 .owner          = THIS_MODULE,
 259                 .pf             = PF_INET,
 260                 .hooknum        = NF_IP_POST_ROUTING,
 261                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 262         },
 263         {
 264                 .hook           = ipv4_confirm,
 265                 .owner          = THIS_MODULE,
 266                 .pf             = PF_INET,
 267                 .hooknum        = NF_IP_LOCAL_IN,
 268                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 269         },
 270 };
 271
 272 #ifdef CONFIG_SYSCTL
 273 /* From nf_conntrack_proto_icmp.c */
 274 extern unsigned int nf_ct_icmp_timeout;
 275 static struct ctl_table_header *nf_ct_ipv4_sysctl_header;
 276
 277 static ctl_table nf_ct_sysctl_table[] = {
 278         {
 279                 .ctl_name       = NET_NF_CONNTRACK_ICMP_TIMEOUT,
 280                 .procname       = "nf_conntrack_icmp_timeout",
 281                 .data           = &nf_ct_icmp_timeout,
 282                 .maxlen         = sizeof(unsigned int),
 283                 .mode           = 0644,
 284                 .proc_handler   = &proc_dointvec_jiffies,
 285         },
 286         { .ctl_name = 0 }
 287 };
 288
 289 static ctl_table nf_ct_netfilter_table[] = {
 290         {
 291                 .ctl_name       = NET_NETFILTER,
 292                 .procname       = "netfilter",
 293                 .mode           = 0555,
 294                 .child          = nf_ct_sysctl_table,
 295         },
 296         { .ctl_name = 0 }
 297 };
 298
 299 static ctl_table nf_ct_net_table[] = {
 300         {
 301                 .ctl_name       = CTL_NET,
 302                 .procname       = "net",
 303                 .mode           = 0555,
 304                 .child          = nf_ct_netfilter_table,
 305         },
 306         { .ctl_name = 0 }
 307 };
 308 #endif
 309
 310 /* Fast function for those who don't want to parse /proc (and I don't
 311    blame them). */
 312 /* Reversing the socket's dst/src point of view gives us the reply
 313    mapping. */
 314 static int
 315 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 316 {
 317         struct inet_sock *inet = inet_sk(sk);
 318         struct nf_conntrack_tuple_hash *h;
 319         struct nf_conntrack_tuple tuple;
 320
 321         NF_CT_TUPLE_U_BLANK(&tuple);
 322         tuple.src.u3.ip = inet->rcv_saddr;
 323         tuple.src.u.tcp.port = inet->sport;
 324         tuple.dst.u3.ip = inet->daddr;
 325         tuple.dst.u.tcp.port = inet->dport;
 326         tuple.src.l3num = PF_INET;
 327         tuple.dst.protonum = IPPROTO_TCP;
 328
 329         /* We only do TCP at the moment: is there a better way? */
 330         if (strcmp(sk->sk_prot->name, "TCP")) {
 331                 DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
 332                 return -ENOPROTOOPT;
 333         }
 334
 335         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 336                 DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
 337                        *len, sizeof(struct sockaddr_in));
 338                 return -EINVAL;
 339         }
 340
 341         h = nf_conntrack_find_get(&tuple, NULL);
 342         if (h) {
 343                 struct sockaddr_in sin;
 344                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 345
 346                 sin.sin_family = AF_INET;
 347                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 348                         .tuple.dst.u.tcp.port;
 349                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 350                         .tuple.dst.u3.ip;
 351
 352                 DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 353                        NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 354                 nf_ct_put(ct);
 355                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 356                         return -EFAULT;
 357                 else
 358                         return 0;
 359         }
 360         DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 361                NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 362                NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 363         return -ENOENT;
 364 }
 365
 366 #if defined(CONFIG_NF_CT_NETLINK) || \
 367     defined(CONFIG_NF_CT_NETLINK_MODULE)
 368
 369 #include <linux/netfilter/nfnetlink.h>
 370 #include <linux/netfilter/nfnetlink_conntrack.h>
 371
 372 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
 373                                 const struct nf_conntrack_tuple *tuple)
 374 {
 375         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
 376                 &tuple->src.u3.ip);
 377         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
 378                 &tuple->dst.u3.ip);
 379         return 0;
 380
 381 nfattr_failure:
 382         return -1;
 383 }
 384
 385 static const size_t cta_min_ip[CTA_IP_MAX] = {
 386         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 387         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 388 };
 389
 390 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
 391                                 struct nf_conntrack_tuple *t)
 392 {
 393         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
 394                 return -EINVAL;
 395
 396         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 397                 return -EINVAL;
 398
 399         t->src.u3.ip =
 400                 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 401         t->dst.u3.ip =
 402                 *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 403
 404         return 0;
 405 }
 406 #endif
 407
 408 static struct nf_sockopt_ops so_getorigdst = {
 409         .pf             = PF_INET,
 410         .get_optmin     = SO_ORIGINAL_DST,
 411         .get_optmax     = SO_ORIGINAL_DST+1,
 412         .get            = &getorigdst,
 413 };
 414
 415 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
 416         .l3proto         = PF_INET,
 417         .name            = "ipv4",
 418         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 419         .invert_tuple    = ipv4_invert_tuple,
 420         .print_tuple     = ipv4_print_tuple,
 421         .print_conntrack = ipv4_print_conntrack,
 422         .prepare         = ipv4_prepare,
 423         .get_features    = ipv4_get_features,
 424 #if defined(CONFIG_NF_CT_NETLINK) || \
 425     defined(CONFIG_NF_CT_NETLINK_MODULE)
 426         .tuple_to_nfattr = ipv4_tuple_to_nfattr,
 427         .nfattr_to_tuple = ipv4_nfattr_to_tuple,
 428 #endif
 429         .me              = THIS_MODULE,
 430 };
 431
 432 extern struct nf_conntrack_protocol nf_conntrack_protocol_tcp4;
 433 extern struct nf_conntrack_protocol nf_conntrack_protocol_udp4;
 434 extern struct nf_conntrack_protocol nf_conntrack_protocol_icmp;
 435
 436 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 437 MODULE_LICENSE("GPL");
 438
 439 static int __init nf_conntrack_l3proto_ipv4_init(void)
 440 {
 441         int ret = 0;
 442
 443         need_conntrack();
 444
 445         ret = nf_register_sockopt(&so_getorigdst);
 446         if (ret < 0) {
 447                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 448                 return ret;
 449         }
 450
 451         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_tcp4);
 452         if (ret < 0) {
 453                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 454                 goto cleanup_sockopt;
 455         }
 456
 457         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_udp4);
 458         if (ret < 0) {
 459                 printk("nf_conntrack_ipv4: can't register udp.\n");
 460                 goto cleanup_tcp;
 461         }
 462
 463         ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_icmp);
 464         if (ret < 0) {
 465                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 466                 goto cleanup_udp;
 467         }
 468
 469         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 470         if (ret < 0) {
 471                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 472                 goto cleanup_icmp;
 473         }
 474
 475         ret = nf_register_hooks(ipv4_conntrack_ops,
 476                                 ARRAY_SIZE(ipv4_conntrack_ops));
 477         if (ret < 0) {
 478                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 479                 goto cleanup_ipv4;
 480         }
 481 #ifdef CONFIG_SYSCTL
 482         nf_ct_ipv4_sysctl_header = register_sysctl_table(nf_ct_net_table, 0);
 483         if (nf_ct_ipv4_sysctl_header == NULL) {
 484                 printk("nf_conntrack: can't register to sysctl.\n");
 485                 ret = -ENOMEM;
 486                 goto cleanup_hooks;
 487         }
 488 #endif
 489         return ret;
 490
 491 #ifdef CONFIG_SYSCTL
 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_protocol_unregister(&nf_conntrack_protocol_icmp);
 499  cleanup_udp:
 500         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp4);
 501  cleanup_tcp:
 502         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_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 #ifdef CONFIG_SYSCTL
 512         unregister_sysctl_table(nf_ct_ipv4_sysctl_header);
 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_protocol_unregister(&nf_conntrack_protocol_icmp);
 517         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp4);
 518         nf_conntrack_protocol_unregister(&nf_conntrack_protocol_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);
 524
 525 EXPORT_SYMBOL(nf_ct_ipv4_gather_frags);