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 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
  28                              struct nf_conntrack_tuple *tuple)
  29 {
  30         const __be32 *ap;
  31         __be32 _addrs[2];
  32         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  33                                 sizeof(u_int32_t) * 2, _addrs);
  34         if (ap == NULL)
  35                 return 0;
  36
  37         tuple->src.u3.ip = ap[0];
  38         tuple->dst.u3.ip = ap[1];
  39
  40         return 1;
  41 }
  42
  43 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
  44                            const struct nf_conntrack_tuple *orig)
  45 {
  46         tuple->src.u3.ip = orig->dst.u3.ip;
  47         tuple->dst.u3.ip = orig->src.u3.ip;
  48
  49         return 1;
  50 }
  51
  52 static int ipv4_print_tuple(struct seq_file *s,
  53                             const struct nf_conntrack_tuple *tuple)
  54 {
  55         return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
  56                           NIPQUAD(tuple->src.u3.ip),
  57                           NIPQUAD(tuple->dst.u3.ip));
  58 }
  59
  60 /* Returns new sk_buff, or NULL */
  61 static int nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
  62 {
  63         int err;
  64
  65         skb_orphan(skb);
  66
  67         local_bh_disable();
  68         err = ip_defrag(skb, user);
  69         local_bh_enable();
  70
  71         if (!err)
  72                 ip_send_check(ip_hdr(skb));
  73
  74         return err;
  75 }
  76
  77 static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
  78                             unsigned int *dataoff, u_int8_t *protonum)
  79 {
  80         const struct iphdr *iph;
  81         struct iphdr _iph;
  82
  83         iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
  84         if (iph == NULL)
  85                 return -NF_DROP;
  86
  87         /* Conntrack defragments packets, we might still see fragments
  88          * inside ICMP packets though. */
  89         if (iph->frag_off & htons(IP_OFFSET))
  90                 return -NF_DROP;
  91
  92         *dataoff = nhoff + (iph->ihl << 2);
  93         *protonum = iph->protocol;
  94
  95         return NF_ACCEPT;
  96 }
  97
  98 static unsigned int ipv4_confirm(unsigned int hooknum,
  99                                  struct sk_buff *skb,
 100                                  const struct net_device *in,
 101                                  const struct net_device *out,
 102                                  int (*okfn)(struct sk_buff *))
 103 {
 104         /* We've seen it coming out the other side: confirm it */
 105         return nf_conntrack_confirm(skb);
 106 }
 107
 108 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
 109                                       struct sk_buff *skb,
 110                                       const struct net_device *in,
 111                                       const struct net_device *out,
 112                                       int (*okfn)(struct sk_buff *))
 113 {
 114         struct nf_conn *ct;
 115         enum ip_conntrack_info ctinfo;
 116         const struct nf_conn_help *help;
 117         const struct nf_conntrack_helper *helper;
 118
 119         /* This is where we call the helper: as the packet goes out. */
 120         ct = nf_ct_get(skb, &ctinfo);
 121         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 122                 return NF_ACCEPT;
 123
 124         help = nfct_help(ct);
 125         if (!help)
 126                 return NF_ACCEPT;
 127         /* rcu_read_lock()ed by nf_hook_slow */
 128         helper = rcu_dereference(help->helper);
 129         if (!helper)
 130                 return NF_ACCEPT;
 131         return helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
 132                             ct, ctinfo);
 133 }
 134
 135 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
 136                                           struct sk_buff *skb,
 137                                           const struct net_device *in,
 138                                           const struct net_device *out,
 139                                           int (*okfn)(struct sk_buff *))
 140 {
 141         /* Previously seen (loopback)?  Ignore.  Do this before
 142            fragment check. */
 143         if (skb->nfct)
 144                 return NF_ACCEPT;
 145
 146         /* Gather fragments. */
 147         if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
 148                 if (nf_ct_ipv4_gather_frags(skb,
 149                                             hooknum == NF_INET_PRE_ROUTING ?
 150                                             IP_DEFRAG_CONNTRACK_IN :
 151                                             IP_DEFRAG_CONNTRACK_OUT))
 152                         return NF_STOLEN;
 153         }
 154         return NF_ACCEPT;
 155 }
 156
 157 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 158                                       struct sk_buff *skb,
 159                                       const struct net_device *in,
 160                                       const struct net_device *out,
 161                                       int (*okfn)(struct sk_buff *))
 162 {
 163         return nf_conntrack_in(PF_INET, hooknum, skb);
 164 }
 165
 166 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 167                                          struct sk_buff *skb,
 168                                          const struct net_device *in,
 169                                          const struct net_device *out,
 170                                          int (*okfn)(struct sk_buff *))
 171 {
 172         /* root is playing with raw sockets. */
 173         if (skb->len < sizeof(struct iphdr) ||
 174             ip_hdrlen(skb) < sizeof(struct iphdr)) {
 175                 if (net_ratelimit())
 176                         printk("ipt_hook: happy cracking.\n");
 177                 return NF_ACCEPT;
 178         }
 179         return nf_conntrack_in(PF_INET, hooknum, skb);
 180 }
 181
 182 /* Connection tracking may drop packets, but never alters them, so
 183    make it the first hook. */
 184 static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
 185         {
 186                 .hook           = ipv4_conntrack_defrag,
 187                 .owner          = THIS_MODULE,
 188                 .pf             = PF_INET,
 189                 .hooknum        = NF_INET_PRE_ROUTING,
 190                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 191         },
 192         {
 193                 .hook           = ipv4_conntrack_in,
 194                 .owner          = THIS_MODULE,
 195                 .pf             = PF_INET,
 196                 .hooknum        = NF_INET_PRE_ROUTING,
 197                 .priority       = NF_IP_PRI_CONNTRACK,
 198         },
 199         {
 200                 .hook           = ipv4_conntrack_defrag,
 201                 .owner          = THIS_MODULE,
 202                 .pf             = PF_INET,
 203                 .hooknum        = NF_INET_LOCAL_OUT,
 204                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 205         },
 206         {
 207                 .hook           = ipv4_conntrack_local,
 208                 .owner          = THIS_MODULE,
 209                 .pf             = PF_INET,
 210                 .hooknum        = NF_INET_LOCAL_OUT,
 211                 .priority       = NF_IP_PRI_CONNTRACK,
 212         },
 213         {
 214                 .hook           = ipv4_conntrack_help,
 215                 .owner          = THIS_MODULE,
 216                 .pf             = PF_INET,
 217                 .hooknum        = NF_INET_POST_ROUTING,
 218                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 219         },
 220         {
 221                 .hook           = ipv4_conntrack_help,
 222                 .owner          = THIS_MODULE,
 223                 .pf             = PF_INET,
 224                 .hooknum        = NF_INET_LOCAL_IN,
 225                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 226         },
 227         {
 228                 .hook           = ipv4_confirm,
 229                 .owner          = THIS_MODULE,
 230                 .pf             = PF_INET,
 231                 .hooknum        = NF_INET_POST_ROUTING,
 232                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 233         },
 234         {
 235                 .hook           = ipv4_confirm,
 236                 .owner          = THIS_MODULE,
 237                 .pf             = PF_INET,
 238                 .hooknum        = NF_INET_LOCAL_IN,
 239                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 240         },
 241 };
 242
 243 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 244 static int log_invalid_proto_min = 0;
 245 static int log_invalid_proto_max = 255;
 246
 247 static ctl_table ip_ct_sysctl_table[] = {
 248         {
 249                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 250                 .procname       = "ip_conntrack_max",
 251                 .data           = &nf_conntrack_max,
 252                 .maxlen         = sizeof(int),
 253                 .mode           = 0644,
 254                 .proc_handler   = &proc_dointvec,
 255         },
 256         {
 257                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 258                 .procname       = "ip_conntrack_count",
 259                 .data           = &nf_conntrack_count,
 260                 .maxlen         = sizeof(int),
 261                 .mode           = 0444,
 262                 .proc_handler   = &proc_dointvec,
 263         },
 264         {
 265                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 266                 .procname       = "ip_conntrack_buckets",
 267                 .data           = &nf_conntrack_htable_size,
 268                 .maxlen         = sizeof(unsigned int),
 269                 .mode           = 0444,
 270                 .proc_handler   = &proc_dointvec,
 271         },
 272         {
 273                 .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
 274                 .procname       = "ip_conntrack_checksum",
 275                 .data           = &nf_conntrack_checksum,
 276                 .maxlen         = sizeof(int),
 277                 .mode           = 0644,
 278                 .proc_handler   = &proc_dointvec,
 279         },
 280         {
 281                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 282                 .procname       = "ip_conntrack_log_invalid",
 283                 .data           = &nf_ct_log_invalid,
 284                 .maxlen         = sizeof(unsigned int),
 285                 .mode           = 0644,
 286                 .proc_handler   = &proc_dointvec_minmax,
 287                 .strategy       = &sysctl_intvec,
 288                 .extra1         = &log_invalid_proto_min,
 289                 .extra2         = &log_invalid_proto_max,
 290         },
 291         {
 292                 .ctl_name       = 0
 293         }
 294 };
 295 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 296
 297 /* Fast function for those who don't want to parse /proc (and I don't
 298    blame them). */
 299 /* Reversing the socket's dst/src point of view gives us the reply
 300    mapping. */
 301 static int
 302 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 303 {
 304         const struct inet_sock *inet = inet_sk(sk);
 305         const struct nf_conntrack_tuple_hash *h;
 306         struct nf_conntrack_tuple tuple;
 307
 308         NF_CT_TUPLE_U_BLANK(&tuple);
 309         tuple.src.u3.ip = inet->rcv_saddr;
 310         tuple.src.u.tcp.port = inet->sport;
 311         tuple.dst.u3.ip = inet->daddr;
 312         tuple.dst.u.tcp.port = inet->dport;
 313         tuple.src.l3num = PF_INET;
 314         tuple.dst.protonum = IPPROTO_TCP;
 315
 316         /* We only do TCP at the moment: is there a better way? */
 317         if (strcmp(sk->sk_prot->name, "TCP")) {
 318                 pr_debug("SO_ORIGINAL_DST: Not a TCP socket\n");
 319                 return -ENOPROTOOPT;
 320         }
 321
 322         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 323                 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
 324                          *len, sizeof(struct sockaddr_in));
 325                 return -EINVAL;
 326         }
 327
 328         h = nf_conntrack_find_get(&tuple);
 329         if (h) {
 330                 struct sockaddr_in sin;
 331                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 332
 333                 sin.sin_family = AF_INET;
 334                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 335                         .tuple.dst.u.tcp.port;
 336                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 337                         .tuple.dst.u3.ip;
 338                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 339
 340                 pr_debug("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 341                          NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 342                 nf_ct_put(ct);
 343                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 344                         return -EFAULT;
 345                 else
 346                         return 0;
 347         }
 348         pr_debug("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 349                  NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 350                  NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 351         return -ENOENT;
 352 }
 353
 354 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 355
 356 #include <linux/netfilter/nfnetlink.h>
 357 #include <linux/netfilter/nfnetlink_conntrack.h>
 358
 359 static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
 360                                 const struct nf_conntrack_tuple *tuple)
 361 {
 362         NLA_PUT_BE32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip);
 363         NLA_PUT_BE32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip);
 364         return 0;
 365
 366 nla_put_failure:
 367         return -1;
 368 }
 369
 370 static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
 371         [CTA_IP_V4_SRC] = { .type = NLA_U32 },
 372         [CTA_IP_V4_DST] = { .type = NLA_U32 },
 373 };
 374
 375 static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
 376                                 struct nf_conntrack_tuple *t)
 377 {
 378         if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
 379                 return -EINVAL;
 380
 381         t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
 382         t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
 383
 384         return 0;
 385 }
 386 #endif
 387
 388 static struct nf_sockopt_ops so_getorigdst = {
 389         .pf             = PF_INET,
 390         .get_optmin     = SO_ORIGINAL_DST,
 391         .get_optmax     = SO_ORIGINAL_DST+1,
 392         .get            = &getorigdst,
 393         .owner          = THIS_MODULE,
 394 };
 395
 396 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
 397         .l3proto         = PF_INET,
 398         .name            = "ipv4",
 399         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 400         .invert_tuple    = ipv4_invert_tuple,
 401         .print_tuple     = ipv4_print_tuple,
 402         .get_l4proto     = ipv4_get_l4proto,
 403 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 404         .tuple_to_nlattr = ipv4_tuple_to_nlattr,
 405         .nlattr_to_tuple = ipv4_nlattr_to_tuple,
 406         .nla_policy      = ipv4_nla_policy,
 407 #endif
 408 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 409         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
 410         .ctl_table       = ip_ct_sysctl_table,
 411 #endif
 412         .me              = THIS_MODULE,
 413 };
 414
 415 module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
 416                   &nf_conntrack_htable_size, 0600);
 417
 418 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 419 MODULE_ALIAS("ip_conntrack");
 420 MODULE_LICENSE("GPL");
 421
 422 static int __init nf_conntrack_l3proto_ipv4_init(void)
 423 {
 424         int ret = 0;
 425
 426         need_conntrack();
 427
 428         ret = nf_register_sockopt(&so_getorigdst);
 429         if (ret < 0) {
 430                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 431                 return ret;
 432         }
 433
 434         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 435         if (ret < 0) {
 436                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 437                 goto cleanup_sockopt;
 438         }
 439
 440         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 441         if (ret < 0) {
 442                 printk("nf_conntrack_ipv4: can't register udp.\n");
 443                 goto cleanup_tcp;
 444         }
 445
 446         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 447         if (ret < 0) {
 448                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 449                 goto cleanup_udp;
 450         }
 451
 452         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 453         if (ret < 0) {
 454                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 455                 goto cleanup_icmp;
 456         }
 457
 458         ret = nf_register_hooks(ipv4_conntrack_ops,
 459                                 ARRAY_SIZE(ipv4_conntrack_ops));
 460         if (ret < 0) {
 461                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 462                 goto cleanup_ipv4;
 463         }
 464 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 465         ret = nf_conntrack_ipv4_compat_init();
 466         if (ret < 0)
 467                 goto cleanup_hooks;
 468 #endif
 469         return ret;
 470 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 471  cleanup_hooks:
 472         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 473 #endif
 474  cleanup_ipv4:
 475         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 476  cleanup_icmp:
 477         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 478  cleanup_udp:
 479         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 480  cleanup_tcp:
 481         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 482  cleanup_sockopt:
 483         nf_unregister_sockopt(&so_getorigdst);
 484         return ret;
 485 }
 486
 487 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 488 {
 489         synchronize_net();
 490 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 491         nf_conntrack_ipv4_compat_fini();
 492 #endif
 493         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 494         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 495         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 496         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 497         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 498         nf_unregister_sockopt(&so_getorigdst);
 499 }
 500
 501 module_init(nf_conntrack_l3proto_ipv4_init);
 502 module_exit(nf_conntrack_l3proto_ipv4_fini);
 503
 504 void need_ipv4_conntrack(void)
 505 {
 506         return;
 507 }
 508 EXPORT_SYMBOL_GPL(need_ipv4_conntrack);