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 ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
  82                             unsigned int *dataoff, u_int8_t *protonum)
  83 {
  84         struct iphdr _iph, *iph;
  85
  86         iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
  87         if (iph == NULL)
  88                 return -NF_DROP;
  89
  90         /* Never happen */
  91         if (iph->frag_off & htons(IP_OFFSET)) {
  92                 if (net_ratelimit()) {
  93                         printk(KERN_ERR "ipv4_get_l4proto: Frag of proto %u\n",
  94                         iph->protocol);
  95                 }
  96                 return -NF_DROP;
  97         }
  98
  99         *dataoff = nhoff + (iph->ihl << 2);
 100         *protonum = iph->protocol;
 101
 102         return NF_ACCEPT;
 103 }
 104
 105 static unsigned int ipv4_confirm(unsigned int hooknum,
 106                                  struct sk_buff **pskb,
 107                                  const struct net_device *in,
 108                                  const struct net_device *out,
 109                                  int (*okfn)(struct sk_buff *))
 110 {
 111         /* We've seen it coming out the other side: confirm it */
 112         return nf_conntrack_confirm(pskb);
 113 }
 114
 115 static unsigned int ipv4_conntrack_help(unsigned int hooknum,
 116                                       struct sk_buff **pskb,
 117                                       const struct net_device *in,
 118                                       const struct net_device *out,
 119                                       int (*okfn)(struct sk_buff *))
 120 {
 121         struct nf_conn *ct;
 122         enum ip_conntrack_info ctinfo;
 123         struct nf_conn_help *help;
 124         struct nf_conntrack_helper *helper;
 125
 126         /* This is where we call the helper: as the packet goes out. */
 127         ct = nf_ct_get(*pskb, &ctinfo);
 128         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
 129                 return NF_ACCEPT;
 130
 131         help = nfct_help(ct);
 132         if (!help)
 133                 return NF_ACCEPT;
 134         /* rcu_read_lock()ed by nf_hook_slow */
 135         helper = rcu_dereference(help->helper);
 136         if (!helper)
 137                 return NF_ACCEPT;
 138         return helper->help(pskb, skb_network_offset(*pskb) + ip_hdrlen(*pskb),
 139                             ct, ctinfo);
 140 }
 141
 142 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
 143                                           struct sk_buff **pskb,
 144                                           const struct net_device *in,
 145                                           const struct net_device *out,
 146                                           int (*okfn)(struct sk_buff *))
 147 {
 148         /* Previously seen (loopback)?  Ignore.  Do this before
 149            fragment check. */
 150         if ((*pskb)->nfct)
 151                 return NF_ACCEPT;
 152
 153         /* Gather fragments. */
 154         if (ip_hdr(*pskb)->frag_off & htons(IP_MF | IP_OFFSET)) {
 155                 *pskb = nf_ct_ipv4_gather_frags(*pskb,
 156                                                 hooknum == NF_IP_PRE_ROUTING ?
 157                                                 IP_DEFRAG_CONNTRACK_IN :
 158                                                 IP_DEFRAG_CONNTRACK_OUT);
 159                 if (!*pskb)
 160                         return NF_STOLEN;
 161         }
 162         return NF_ACCEPT;
 163 }
 164
 165 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
 166                                       struct sk_buff **pskb,
 167                                       const struct net_device *in,
 168                                       const struct net_device *out,
 169                                       int (*okfn)(struct sk_buff *))
 170 {
 171         return nf_conntrack_in(PF_INET, hooknum, pskb);
 172 }
 173
 174 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
 175                                          struct sk_buff **pskb,
 176                                          const struct net_device *in,
 177                                          const struct net_device *out,
 178                                          int (*okfn)(struct sk_buff *))
 179 {
 180         /* root is playing with raw sockets. */
 181         if ((*pskb)->len < sizeof(struct iphdr)
 182             || ip_hdrlen(*pskb) < sizeof(struct iphdr)) {
 183                 if (net_ratelimit())
 184                         printk("ipt_hook: happy cracking.\n");
 185                 return NF_ACCEPT;
 186         }
 187         return nf_conntrack_in(PF_INET, hooknum, pskb);
 188 }
 189
 190 /* Connection tracking may drop packets, but never alters them, so
 191    make it the first hook. */
 192 static struct nf_hook_ops ipv4_conntrack_ops[] = {
 193         {
 194                 .hook           = ipv4_conntrack_defrag,
 195                 .owner          = THIS_MODULE,
 196                 .pf             = PF_INET,
 197                 .hooknum        = NF_IP_PRE_ROUTING,
 198                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 199         },
 200         {
 201                 .hook           = ipv4_conntrack_in,
 202                 .owner          = THIS_MODULE,
 203                 .pf             = PF_INET,
 204                 .hooknum        = NF_IP_PRE_ROUTING,
 205                 .priority       = NF_IP_PRI_CONNTRACK,
 206         },
 207         {
 208                 .hook           = ipv4_conntrack_defrag,
 209                 .owner          = THIS_MODULE,
 210                 .pf             = PF_INET,
 211                 .hooknum        = NF_IP_LOCAL_OUT,
 212                 .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
 213         },
 214         {
 215                 .hook           = ipv4_conntrack_local,
 216                 .owner          = THIS_MODULE,
 217                 .pf             = PF_INET,
 218                 .hooknum        = NF_IP_LOCAL_OUT,
 219                 .priority       = NF_IP_PRI_CONNTRACK,
 220         },
 221         {
 222                 .hook           = ipv4_conntrack_help,
 223                 .owner          = THIS_MODULE,
 224                 .pf             = PF_INET,
 225                 .hooknum        = NF_IP_POST_ROUTING,
 226                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 227         },
 228         {
 229                 .hook           = ipv4_conntrack_help,
 230                 .owner          = THIS_MODULE,
 231                 .pf             = PF_INET,
 232                 .hooknum        = NF_IP_LOCAL_IN,
 233                 .priority       = NF_IP_PRI_CONNTRACK_HELPER,
 234         },
 235         {
 236                 .hook           = ipv4_confirm,
 237                 .owner          = THIS_MODULE,
 238                 .pf             = PF_INET,
 239                 .hooknum        = NF_IP_POST_ROUTING,
 240                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 241         },
 242         {
 243                 .hook           = ipv4_confirm,
 244                 .owner          = THIS_MODULE,
 245                 .pf             = PF_INET,
 246                 .hooknum        = NF_IP_LOCAL_IN,
 247                 .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
 248         },
 249 };
 250
 251 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 252 static int log_invalid_proto_min = 0;
 253 static int log_invalid_proto_max = 255;
 254
 255 static ctl_table ip_ct_sysctl_table[] = {
 256         {
 257                 .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
 258                 .procname       = "ip_conntrack_max",
 259                 .data           = &nf_conntrack_max,
 260                 .maxlen         = sizeof(int),
 261                 .mode           = 0644,
 262                 .proc_handler   = &proc_dointvec,
 263         },
 264         {
 265                 .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
 266                 .procname       = "ip_conntrack_count",
 267                 .data           = &nf_conntrack_count,
 268                 .maxlen         = sizeof(int),
 269                 .mode           = 0444,
 270                 .proc_handler   = &proc_dointvec,
 271         },
 272         {
 273                 .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
 274                 .procname       = "ip_conntrack_buckets",
 275                 .data           = &nf_conntrack_htable_size,
 276                 .maxlen         = sizeof(unsigned int),
 277                 .mode           = 0444,
 278                 .proc_handler   = &proc_dointvec,
 279         },
 280         {
 281                 .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
 282                 .procname       = "ip_conntrack_checksum",
 283                 .data           = &nf_conntrack_checksum,
 284                 .maxlen         = sizeof(int),
 285                 .mode           = 0644,
 286                 .proc_handler   = &proc_dointvec,
 287         },
 288         {
 289                 .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
 290                 .procname       = "ip_conntrack_log_invalid",
 291                 .data           = &nf_ct_log_invalid,
 292                 .maxlen         = sizeof(unsigned int),
 293                 .mode           = 0644,
 294                 .proc_handler   = &proc_dointvec_minmax,
 295                 .strategy       = &sysctl_intvec,
 296                 .extra1         = &log_invalid_proto_min,
 297                 .extra2         = &log_invalid_proto_max,
 298         },
 299         {
 300                 .ctl_name       = 0
 301         }
 302 };
 303 #endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 304
 305 /* Fast function for those who don't want to parse /proc (and I don't
 306    blame them). */
 307 /* Reversing the socket's dst/src point of view gives us the reply
 308    mapping. */
 309 static int
 310 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 311 {
 312         struct inet_sock *inet = inet_sk(sk);
 313         struct nf_conntrack_tuple_hash *h;
 314         struct nf_conntrack_tuple tuple;
 315
 316         NF_CT_TUPLE_U_BLANK(&tuple);
 317         tuple.src.u3.ip = inet->rcv_saddr;
 318         tuple.src.u.tcp.port = inet->sport;
 319         tuple.dst.u3.ip = inet->daddr;
 320         tuple.dst.u.tcp.port = inet->dport;
 321         tuple.src.l3num = PF_INET;
 322         tuple.dst.protonum = IPPROTO_TCP;
 323
 324         /* We only do TCP at the moment: is there a better way? */
 325         if (strcmp(sk->sk_prot->name, "TCP")) {
 326                 pr_debug("SO_ORIGINAL_DST: Not a TCP socket\n");
 327                 return -ENOPROTOOPT;
 328         }
 329
 330         if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
 331                 pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
 332                          *len, sizeof(struct sockaddr_in));
 333                 return -EINVAL;
 334         }
 335
 336         h = nf_conntrack_find_get(&tuple);
 337         if (h) {
 338                 struct sockaddr_in sin;
 339                 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 340
 341                 sin.sin_family = AF_INET;
 342                 sin.sin_port = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 343                         .tuple.dst.u.tcp.port;
 344                 sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
 345                         .tuple.dst.u3.ip;
 346                 memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 347
 348                 pr_debug("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
 349                          NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
 350                 nf_ct_put(ct);
 351                 if (copy_to_user(user, &sin, sizeof(sin)) != 0)
 352                         return -EFAULT;
 353                 else
 354                         return 0;
 355         }
 356         pr_debug("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
 357                  NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
 358                  NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
 359         return -ENOENT;
 360 }
 361
 362 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 363
 364 #include <linux/netfilter/nfnetlink.h>
 365 #include <linux/netfilter/nfnetlink_conntrack.h>
 366
 367 static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
 368                                 const struct nf_conntrack_tuple *tuple)
 369 {
 370         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
 371                 &tuple->src.u3.ip);
 372         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
 373                 &tuple->dst.u3.ip);
 374         return 0;
 375
 376 nfattr_failure:
 377         return -1;
 378 }
 379
 380 static const size_t cta_min_ip[CTA_IP_MAX] = {
 381         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 382         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 383 };
 384
 385 static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
 386                                 struct nf_conntrack_tuple *t)
 387 {
 388         if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
 389                 return -EINVAL;
 390
 391         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 392                 return -EINVAL;
 393
 394         t->src.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 395         t->dst.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 396
 397         return 0;
 398 }
 399 #endif
 400
 401 static struct nf_sockopt_ops so_getorigdst = {
 402         .pf             = PF_INET,
 403         .get_optmin     = SO_ORIGINAL_DST,
 404         .get_optmax     = SO_ORIGINAL_DST+1,
 405         .get            = &getorigdst,
 406 };
 407
 408 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
 409         .l3proto         = PF_INET,
 410         .name            = "ipv4",
 411         .pkt_to_tuple    = ipv4_pkt_to_tuple,
 412         .invert_tuple    = ipv4_invert_tuple,
 413         .print_tuple     = ipv4_print_tuple,
 414         .print_conntrack = ipv4_print_conntrack,
 415         .get_l4proto     = ipv4_get_l4proto,
 416 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 417         .tuple_to_nfattr = ipv4_tuple_to_nfattr,
 418         .nfattr_to_tuple = ipv4_nfattr_to_tuple,
 419 #endif
 420 #if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 421         .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
 422         .ctl_table       = ip_ct_sysctl_table,
 423 #endif
 424         .me              = THIS_MODULE,
 425 };
 426
 427 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
 428 MODULE_ALIAS("ip_conntrack");
 429 MODULE_LICENSE("GPL");
 430
 431 static int __init nf_conntrack_l3proto_ipv4_init(void)
 432 {
 433         int ret = 0;
 434
 435         need_conntrack();
 436
 437         ret = nf_register_sockopt(&so_getorigdst);
 438         if (ret < 0) {
 439                 printk(KERN_ERR "Unable to register netfilter socket option\n");
 440                 return ret;
 441         }
 442
 443         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
 444         if (ret < 0) {
 445                 printk("nf_conntrack_ipv4: can't register tcp.\n");
 446                 goto cleanup_sockopt;
 447         }
 448
 449         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
 450         if (ret < 0) {
 451                 printk("nf_conntrack_ipv4: can't register udp.\n");
 452                 goto cleanup_tcp;
 453         }
 454
 455         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
 456         if (ret < 0) {
 457                 printk("nf_conntrack_ipv4: can't register icmp.\n");
 458                 goto cleanup_udp;
 459         }
 460
 461         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv4);
 462         if (ret < 0) {
 463                 printk("nf_conntrack_ipv4: can't register ipv4\n");
 464                 goto cleanup_icmp;
 465         }
 466
 467         ret = nf_register_hooks(ipv4_conntrack_ops,
 468                                 ARRAY_SIZE(ipv4_conntrack_ops));
 469         if (ret < 0) {
 470                 printk("nf_conntrack_ipv4: can't register hooks.\n");
 471                 goto cleanup_ipv4;
 472         }
 473 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 474         ret = nf_conntrack_ipv4_compat_init();
 475         if (ret < 0)
 476                 goto cleanup_hooks;
 477 #endif
 478         return ret;
 479 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 480  cleanup_hooks:
 481         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 482 #endif
 483  cleanup_ipv4:
 484         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 485  cleanup_icmp:
 486         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 487  cleanup_udp:
 488         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 489  cleanup_tcp:
 490         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 491  cleanup_sockopt:
 492         nf_unregister_sockopt(&so_getorigdst);
 493         return ret;
 494 }
 495
 496 static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 497 {
 498         synchronize_net();
 499 #if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
 500         nf_conntrack_ipv4_compat_fini();
 501 #endif
 502         nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 503         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
 504         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
 505         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
 506         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
 507         nf_unregister_sockopt(&so_getorigdst);
 508 }
 509
 510 module_init(nf_conntrack_l3proto_ipv4_init);
 511 module_exit(nf_conntrack_l3proto_ipv4_fini);