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         __be32 _addrs[2], *ap;
  31         ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
  32                                 sizeof(u_int32_t) * 2, _addrs);
  33         if (ap == NULL)
  34                 return 0;
  35
  36         tuple->src.u3.ip = ap[0];
  37         tuple->dst.u3.ip = ap[1];
  38
  39         return 1;
  40 }
  41
  42 static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
  43                            const struct nf_conntrack_tuple *orig)
  44 {
  45         tuple->src.u3.ip = orig->dst.u3.ip;
  46         tuple->dst.u3.ip = orig->src.u3.ip;
  47
  48         return 1;
  49 }
  50
  51 static int ipv4_print_tuple(struct seq_file *s,
  52                             const struct nf_conntrack_tuple *tuple)
  53 {
  54         return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
  55                           NIPQUAD(tuple->src.u3.ip),
  56                           NIPQUAD(tuple->dst.u3.ip));
  57 }
  58
  59 static int ipv4_print_conntrack(struct seq_file *s,
  60                                 const struct nf_conn *conntrack)
  61 {
  62         return 0;
  63 }
  64
  65 /* Returns new sk_buff, or NULL */
  66 static struct sk_buff *
  67 nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
  68 {
  69         skb_orphan(skb);
  70
  71         local_bh_disable();
  72         skb = ip_defrag(skb, user);
  73         local_bh_enable();
  74
  75         if (skb)
  76                 ip_send_check(ip_hdr(skb));
  77
  78         return skb;
  79 }
  80
  81 static int
  82 ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
  83              u_int8_t *protonum)
  84 {
  85         /* Never happen */
  86         if (ip_hdr(*pskb)->frag_off & htons(IP_OFFSET)) {
  87                 if (net_ratelimit()) {
  88                         printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
  89                         ip_hdr(*pskb)->protocol, hooknum);
  90                 }
  91                 return -NF_DROP;
  92         }
  93
  94         *dataoff = skb_network_offset(*pskb) + ip_hdrlen(*pskb);
  95         *protonum = ip_hdr(*pskb)->protocol;
  96
  97         return NF_ACCEPT;
  98 }
  99
 100 static unsigned int ipv4_confirm(unsigned int hooknum,
 101                                  struct sk_buff **pskb,
 102                                  const struct net_device *in,
 103                                  const struct net_device *out,
 104                                  int (*okfn)(struct sk_buff *))
 105 {
 106         /* We've seen it coming out the other side: confirm it */
 107         return nf_conntrack_confirm(pskb);
 108 }
 109
 110 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
 111                                       struct sk_buff **pskb,
 112                                       const struct net_device *in,
 113                                       const struct net_device *out,
 114                                       int (*okfn)(struct sk_buff *))
 115 {
 116         struct nf_conn *ct;
 117         enum ip_conntrack_info ctinfo;
 118         struct nf_conn_help *help;
 119         struct nf_conntrack_helper *helper;
 120
 121         /* This is where we call the helper: as the packet goes out. */
 122         ct = nf_ct_get(*pskb, &ctinfo);
 123         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 124                 return NF_ACCEPT;
 125
 126         help = nfct_help(ct);
 127         if (!help)
 128                 return NF_ACCEPT;
 129         /* rcu_read_lock()ed by nf_hook_slow */
 130         helper = rcu_dereference(help->helper);
 131         if (!helper)
 132                 return NF_ACCEPT;
 133         return helper->help(pskb, skb_network_offset(*pskb) + ip_hdrlen(*pskb),
 134                             ct, ctinfo);
 135 }
 136
 137 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
 138                                           struct sk_buff **pskb,
 139                                           const struct net_device *in,
 140                                           const struct net_device *out,
 141                                           int (*okfn)(struct sk_buff *))
 142 {
 143         /* Previously seen (loopback)?  Ignore.  Do this before
 144            fragment check. */
 145         if ((*pskb)->nfct)
 146                 return NF_ACCEPT;
 147
 148         /* Gather fragments. */
 149         if (ip_hdr(*pskb)->frag_off & htons(IP_MF | IP_OFFSET)) {
 150                 *pskb = nf_ct_ipv4_gather_frags(*pskb,
 151                                                 hooknum == NF_IP_PRE_ROUTING ?
 152                                                 IP_DEFRAG_CONNTRACK_IN :
 153                                                 IP_DEFRAG_CONNTRACK_OUT);
 154                 if (!*pskb)
 155                         return NF_STOLEN;
 156         }
 157         return NF_ACCEPT;
 158 }
 159
 160 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 161                                       struct sk_buff **pskb,
 162                                       const struct net_device *in,
 163                                       const struct net_device *out,
 164                                       int (*okfn)(struct sk_buff *))
 165 {
 166         return nf_conntrack_in(PF_INET, hooknum, pskb);
 167 }
 168
 169 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 170                                          struct sk_buff **pskb,
 171                                          const struct net_device *in,
 172                                          const struct net_device *out,
 173                                          int (*okfn)(struct sk_buff *))
 174 {
 175         /* root is playing with raw sockets. */
 176         if ((*pskb)->len < sizeof(struct iphdr)
 177             || ip_hdrlen(*pskb) < sizeof(struct iphdr)) {
 178                 if (net_ratelimit())
 179                         printk("ipt_hook: happy cracking.\n");
 180                 return NF_ACCEPT;
 181         }
 182         return nf_conntrack_in(PF_INET, hooknum, pskb);
 183 }
 184
 185 /* Connection tracking may drop packets, but never alters them, so
 186    make it the first hook. */
 187 static struct nf_hook_ops ipv4_conntrack_ops[] = {
 188         {
 189                 .hook           = ipv4_conntrack_defrag,
 190                 .owner          = THIS_MODULE,
 191                 .pf             = PF_INET,
 192                 .hooknum        = NF_IP_PRE_ROUTING,
 193                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 194         },
 195         {
 196                 .hook           = ipv4_conntrack_in,
 197                 .owner          = THIS_MODULE,
 198                 .pf             = PF_INET,
 199                 .hooknum        = NF_IP_PRE_ROUTING,
 200                 .priority       = NF_IP_PRI_CONNTRACK,
 201         },
 202         {
 203                 .hook           = ipv4_conntrack_defrag,
 204                 .owner          = THIS_MODULE,
 205                 .pf             = PF_INET,
 206                 .hooknum        = NF_IP_LOCAL_OUT,
 207                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 208         },
 209         {
 210                 .hook           = ipv4_conntrack_local,
 211                 .owner          = THIS_MODULE,
 212                 .pf             = PF_INET,
 213                 .hooknum        = NF_IP_LOCAL_OUT,
 214                 .priority       = NF_IP_PRI_CONNTRACK,
 215         },
 216         {
 217                 .hook           = ipv4_conntrack_help,
 218                 .owner          = THIS_MODULE,
 219                 .pf             = PF_INET,
 220                 .hooknum        = NF_IP_POST_ROUTING,
 221                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 222         },
 223         {
 224                 .hook           = ipv4_conntrack_help,
 225                 .owner          = THIS_MODULE,
 226                 .pf             = PF_INET,
 227                 .hooknum        = NF_IP_LOCAL_IN,
 228                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 229         },
 230         {
 231                 .hook           = ipv4_confirm,
 232                 .owner          = THIS_MODULE,
 233                 .pf             = PF_INET,
 234                 .hooknum        = NF_IP_POST_ROUTING,
 235                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 236         },
 237         {
 238                 .hook           = ipv4_confirm,
 239                 .owner          = THIS_MODULE,
 240                 .pf             = PF_INET,
 241                 .hooknum        = NF_IP_LOCAL_IN,
 242                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 243         },
 244 };
 245
 246 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 247 static int log_invalid_proto_min = 0;
 248 static int log_invalid_proto_max = 255;
 249
 250 static ctl_table ip_ct_sysctl_table[] = {
 251         {
 252                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 253                 .procname       = "ip_conntrack_max",
 254                 .data           = &nf_conntrack_max,
 255                 .maxlen         = sizeof(int),
 256                 .mode           = 0644,
 257                 .proc_handler   = &proc_dointvec,
 258         },
 259         {
 260                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 261                 .procname       = "ip_conntrack_count",
 262                 .data           = &nf_conntrack_count,
 263                 .maxlen         = sizeof(int),
 264                 .mode           = 0444,
 265                 .proc_handler   = &proc_dointvec,
 266         },
 267         {
 268                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 269                 .procname       = "ip_conntrack_buckets",
 270                 .data           = &nf_conntrack_htable_size,
 271                 .maxlen         = sizeof(unsigned int),
 272                 .mode           = 0444,
 273                 .proc_handler   = &proc_dointvec,
 274         },
 275         {
 276                 .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
 277                 .procname       = "ip_conntrack_checksum",
 278                 .data           = &nf_conntrack_checksum,
 279                 .maxlen         = sizeof(int),
 280                 .mode           = 0644,
 281                 .proc_handler   = &proc_dointvec,
 282         },
 283         {
 284                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 285                 .procname       = "ip_conntrack_log_invalid",
 286                 .data           = &nf_ct_log_invalid,
 287                 .maxlen         = sizeof(unsigned int),
 288                 .mode           = 0644,
 289                 .proc_handler   = &proc_dointvec_minmax,
 290                 .strategy       = &sysctl_intvec,
 291                 .extra1         = &log_invalid_proto_min,
 292                 .extra2         = &log_invalid_proto_max,
 293         },
 294         {
 295                 .ctl_name       = 0
 296         }
 297 };
 298 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 299
 300 /* Fast function for those who don't want to parse /proc (and I don't
 301    blame them). */
 302 /* Reversing the socket's dst/src point of view gives us the reply
 303    mapping. */
 304 static int
 305 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 306 {
 307         struct inet_sock *inet = inet_sk(sk);
 308         struct nf_conntrack_tuple_hash *h;
 309         struct nf_conntrack_tuple tuple;
 310
 311         NF_CT_TUPLE_U_BLANK(&tuple);
 312         tuple.src.u3.ip = inet->rcv_saddr;
 313         tuple.src.u.tcp.port = inet->sport;
 314         tuple.dst.u3.ip = inet->daddr;
 315         tuple.dst.u.tcp.port = inet->dport;
 316         tuple.src.l3num = PF_INET;
 317         tuple.dst.protonum = IPPROTO_TCP;
 318
 319         /* We only do TCP at the moment: is there a better way? */
 320         if (strcmp(sk->sk_prot->name, "TCP")) {
 321                 pr_debug("SO_ORIGINAL_DST: Not a TCP socket\n");
 322                 return -ENOPROTOOPT;
 323         }
 324
 325         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 326                 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
 327                          *len, sizeof(struct sockaddr_in));
 328                 return -EINVAL;
 329         }
 330
 331         h = nf_conntrack_find_get(&tuple);
 332         if (h) {
 333                 struct sockaddr_in sin;
 334                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 335
 336                 sin.sin_family = AF_INET;
 337                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 338                         .tuple.dst.u.tcp.port;
 339                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 340                         .tuple.dst.u3.ip;
 341                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 342
 343                 pr_debug("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 344                          NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 345                 nf_ct_put(ct);
 346                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 347                         return -EFAULT;
 348                 else
 349                         return 0;
 350         }
 351         pr_debug("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 352                  NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 353                  NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 354         return -ENOENT;
 355 }
 356
 357 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 358
 359 #include <linux/netfilter/nfnetlink.h>
 360 #include <linux/netfilter/nfnetlink_conntrack.h>
 361
 362 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
 363                                 const struct nf_conntrack_tuple *tuple)
 364 {
 365         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
 366                 &tuple->src.u3.ip);
 367         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
 368                 &tuple->dst.u3.ip);
 369         return 0;
 370
 371 nfattr_failure:
 372         return -1;
 373 }
 374
 375 static const size_t cta_min_ip[CTA_IP_MAX] = {
 376         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 377         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 378 };
 379
 380 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
 381                                 struct nf_conntrack_tuple *t)
 382 {
 383         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
 384                 return -EINVAL;
 385
 386         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 387                 return -EINVAL;
 388
 389         t->src.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 390         t->dst.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 391
 392         return 0;
 393 }
 394 #endif
 395
 396 static struct nf_sockopt_ops so_getorigdst = {
 397         .pf             = PF_INET,
 398         .get_optmin     = SO_ORIGINAL_DST,
 399         .get_optmax     = SO_ORIGINAL_DST+1,
 400         .get            = &getorigdst,
 401 };
 402
 403 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
 404         .l3proto         = PF_INET,
 405         .name            = "ipv4",
 406         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 407         .invert_tuple    = ipv4_invert_tuple,
 408         .print_tuple     = ipv4_print_tuple,
 409         .print_conntrack = ipv4_print_conntrack,
 410         .prepare         = ipv4_prepare,
 411 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 412         .tuple_to_nfattr = ipv4_tuple_to_nfattr,
 413         .nfattr_to_tuple = ipv4_nfattr_to_tuple,
 414 #endif
 415 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 416         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
 417         .ctl_table       = ip_ct_sysctl_table,
 418 #endif
 419         .me              = THIS_MODULE,
 420 };
 421
 422 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 423 MODULE_ALIAS("ip_conntrack");
 424 MODULE_LICENSE("GPL");
 425
 426 static int __init nf_conntrack_l3proto_ipv4_init(void)
 427 {
 428         int ret = 0;
 429
 430         need_conntrack();
 431
 432         ret = nf_register_sockopt(&so_getorigdst);
 433         if (ret < 0) {
 434                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 435                 return ret;
 436         }
 437
 438         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 439         if (ret < 0) {
 440                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 441                 goto cleanup_sockopt;
 442         }
 443
 444         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 445         if (ret < 0) {
 446                 printk("nf_conntrack_ipv4: can't register udp.\n");
 447                 goto cleanup_tcp;
 448         }
 449
 450         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 451         if (ret < 0) {
 452                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 453                 goto cleanup_udp;
 454         }
 455
 456         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 457         if (ret < 0) {
 458                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 459                 goto cleanup_icmp;
 460         }
 461
 462         ret = nf_register_hooks(ipv4_conntrack_ops,
 463                                 ARRAY_SIZE(ipv4_conntrack_ops));
 464         if (ret < 0) {
 465                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 466                 goto cleanup_ipv4;
 467         }
 468 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 469         ret = nf_conntrack_ipv4_compat_init();
 470         if (ret < 0)
 471                 goto cleanup_hooks;
 472 #endif
 473         return ret;
 474 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 475  cleanup_hooks:
 476         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 477 #endif
 478  cleanup_ipv4:
 479         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 480  cleanup_icmp:
 481         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 482  cleanup_udp:
 483         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 484  cleanup_tcp:
 485         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 486  cleanup_sockopt:
 487         nf_unregister_sockopt(&so_getorigdst);
 488         return ret;
 489 }
 490
 491 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 492 {
 493         synchronize_net();
 494 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 495         nf_conntrack_ipv4_compat_fini();
 496 #endif
 497         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 498         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 499         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 500         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 501         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 502         nf_unregister_sockopt(&so_getorigdst);
 503 }
 504
 505 module_init(nf_conntrack_l3proto_ipv4_init);
 506 module_exit(nf_conntrack_l3proto_ipv4_fini);