release/src-rt-6.x/linux/linux-2.6/net/ipv4/netfilter/nf_nat_standalone.c

   1 /* (C) 1999-2001 Paul `Rusty' Russell
   2  * (C) 2002-2006 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 #include <linux/types.h>
   9 #include <linux/icmp.h>
  10 #include <linux/ip.h>
  11 #include <linux/netfilter.h>
  12 #include <linux/netfilter_ipv4.h>
  13 #include <linux/module.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/proc_fs.h>
  16 #include <net/ip.h>
  17 #include <net/checksum.h>
  18 #include <linux/spinlock.h>
  19
  20 #include <net/netfilter/nf_conntrack.h>
  21 #include <net/netfilter/nf_conntrack_core.h>
  22 #include <net/netfilter/nf_nat.h>
  23 #include <net/netfilter/nf_nat_rule.h>
  24 #include <net/netfilter/nf_nat_protocol.h>
  25 #include <net/netfilter/nf_nat_core.h>
  26 #include <net/netfilter/nf_nat_helper.h>
  27 #include <linux/netfilter_ipv4/ip_tables.h>
  28 #include <linux/netfilter_ipv4/ipt_cone.h>
  29
  30 #if 0
  31 #define DEBUGP printk
  32 #else
  33 #define DEBUGP(format, args...)
  34 #endif
  35
  36 #ifdef CONFIG_XFRM
  37 static void nat_decode_session(struct sk_buff *skb, struct flowi *fl)
  38 {
  39         struct nf_conn *ct;
  40         struct nf_conntrack_tuple *t;
  41         enum ip_conntrack_info ctinfo;
  42         enum ip_conntrack_dir dir;
  43         unsigned long statusbit;
  44
  45         ct = nf_ct_get(skb, &ctinfo);
  46         if (ct == NULL)
  47                 return;
  48         dir = CTINFO2DIR(ctinfo);
  49         t = &ct->tuplehash[dir].tuple;
  50
  51         if (dir == IP_CT_DIR_ORIGINAL)
  52                 statusbit = IPS_DST_NAT;
  53         else
  54                 statusbit = IPS_SRC_NAT;
  55
  56         if (ct->status & statusbit) {
  57                 fl->fl4_dst = t->dst.u3.ip;
  58                 if (t->dst.protonum == IPPROTO_TCP ||
  59                     t->dst.protonum == IPPROTO_UDP)
  60                         fl->fl_ip_dport = t->dst.u.tcp.port;
  61         }
  62
  63         statusbit ^= IPS_NAT_MASK;
  64
  65         if (ct->status & statusbit) {
  66                 fl->fl4_src = t->src.u3.ip;
  67                 if (t->dst.protonum == IPPROTO_TCP ||
  68                     t->dst.protonum == IPPROTO_UDP)
  69                         fl->fl_ip_sport = t->src.u.tcp.port;
  70         }
  71 }
  72 #endif
  73
  74 static unsigned int
  75 nf_nat_fn(unsigned int hooknum,
  76           struct sk_buff *skb,
  77           const struct net_device *in,
  78           const struct net_device *out,
  79           int (*okfn)(struct sk_buff *))
  80 {
  81         struct nf_conn *ct;
  82         enum ip_conntrack_info ctinfo;
  83         struct nf_conn_nat *nat;
  84         /* maniptype == SRC for postrouting. */
  85         enum nf_nat_manip_type maniptype = HOOK2MANIP(hooknum);
  86
  87         /* We never see fragments: conntrack defrags on pre-routing
  88            and local-out, and nf_nat_out protects post-routing. */
  89         NF_CT_ASSERT(!(ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)));
  90
  91         ct = nf_ct_get(skb, &ctinfo);
  92         /* Can't track?  It's not due to stress, or conntrack would
  93            have dropped it.  Hence it's the user's responsibilty to
  94            packet filter it out, or implement conntrack/NAT for that
  95            protocol. 8) --RR */
  96         if (!ct) {
  97                 /* Exception: ICMP redirect to new connection (not in
  98                    hash table yet).  We must not let this through, in
  99                    case we're doing NAT to the same network. */
 100                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 101                         struct icmphdr _hdr, *hp;
 102
 103                         hp = skb_header_pointer(skb, ip_hdrlen(skb),
 104                                                 sizeof(_hdr), &_hdr);
 105                         if (hp != NULL &&
 106                             hp->type == ICMP_REDIRECT)
 107                                 return NF_DROP;
 108                 }
 109                 return NF_ACCEPT;
 110         }
 111
 112         /* Don't try to NAT if this packet is not conntracked */
 113         if (ct == &nf_conntrack_untracked)
 114                 return NF_ACCEPT;
 115
 116         nat = nfct_nat(ct);
 117         if (!nat)
 118                 return NF_ACCEPT;
 119
 120         switch (ctinfo) {
 121         case IP_CT_RELATED:
 122         case IP_CT_RELATED+IP_CT_IS_REPLY:
 123                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 124                         if (!nf_nat_icmp_reply_translation(ct, ctinfo,
 125                                                            hooknum, skb))
 126                                 return NF_DROP;
 127                         else
 128                                 return NF_ACCEPT;
 129                 }
 130                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 131         case IP_CT_NEW:
 132
 133                 /* Seen it before?  This can happen for loopback, retrans,
 134                    or local packets.. */
 135                 if (!nf_nat_initialized(ct, maniptype)) {
 136                         unsigned int ret;
 137
 138                         if (unlikely(nf_ct_is_confirmed(ct)))
 139                                 /* NAT module was loaded late */
 140                                 ret = alloc_null_binding_confirmed(ct, hooknum);
 141                         else if (hooknum == NF_IP_LOCAL_IN)
 142                                 /* LOCAL_IN hook doesn't have a chain!  */
 143                                 ret = alloc_null_binding(ct, hooknum);
 144                         else
 145                                 ret = nf_nat_rule_find(skb, hooknum, in, out,
 146                                                        ct);
 147
 148                         if (ret != NF_ACCEPT) {
 149                                 return ret;
 150                         }
 151
 152                         ipt_cone_place_in_hashes(ct);
 153
 154                 } else
 155                         DEBUGP("Already setup manip %s for ct %p\n",
 156                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 157                                ct);
 158                 break;
 159
 160         default:
 161                 /* ESTABLISHED */
 162                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 163                              ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 164         }
 165
 166         return nf_nat_packet(ct, ctinfo, hooknum, skb);
 167 }
 168
 169 static unsigned int
 170 nf_nat_in(unsigned int hooknum,
 171           struct sk_buff *skb,
 172           const struct net_device *in,
 173           const struct net_device *out,
 174           int (*okfn)(struct sk_buff *))
 175 {
 176         unsigned int ret;
 177         __be32 daddr = ip_hdr(skb)->daddr;
 178
 179         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 180         if (ret != NF_DROP && ret != NF_STOLEN &&
 181             daddr != ip_hdr(skb)->daddr) {
 182                 dst_release(skb->dst);
 183                 skb->dst = NULL;
 184         }
 185         return ret;
 186 }
 187
 188 static unsigned int
 189 nf_nat_out(unsigned int hooknum,
 190            struct sk_buff *skb,
 191            const struct net_device *in,
 192            const struct net_device *out,
 193            int (*okfn)(struct sk_buff *))
 194 {
 195 #ifdef CONFIG_XFRM
 196         struct nf_conn *ct;
 197         enum ip_conntrack_info ctinfo;
 198 #endif
 199         unsigned int ret;
 200
 201         /* root is playing with raw sockets. */
 202         if (skb->len < sizeof(struct iphdr) ||
 203             ip_hdrlen(skb) < sizeof(struct iphdr))
 204                 return NF_ACCEPT;
 205
 206         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 207 #ifdef CONFIG_XFRM
 208         if (ret != NF_DROP && ret != NF_STOLEN &&
 209             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 210                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 211
 212                 if (ct->tuplehash[dir].tuple.src.u3.ip !=
 213                     ct->tuplehash[!dir].tuple.dst.u3.ip
 214                     || ct->tuplehash[dir].tuple.src.u.all !=
 215                        ct->tuplehash[!dir].tuple.dst.u.all
 216                     )
 217                         return ip_xfrm_me_harder(skb) == 0 ? ret : NF_DROP;
 218         }
 219 #endif
 220         return ret;
 221 }
 222
 223 static unsigned int
 224 nf_nat_local_fn(unsigned int hooknum,
 225                 struct sk_buff *skb,
 226                 const struct net_device *in,
 227                 const struct net_device *out,
 228                 int (*okfn)(struct sk_buff *))
 229 {
 230         struct nf_conn *ct;
 231         enum ip_conntrack_info ctinfo;
 232         unsigned int ret;
 233
 234         /* root is playing with raw sockets. */
 235         if (skb->len < sizeof(struct iphdr) ||
 236             ip_hdrlen(skb) < sizeof(struct iphdr))
 237                 return NF_ACCEPT;
 238
 239         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 240         if (ret != NF_DROP && ret != NF_STOLEN &&
 241             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 242                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 243
 244                 if (ct->tuplehash[dir].tuple.dst.u3.ip !=
 245                     ct->tuplehash[!dir].tuple.src.u3.ip) {
 246                         if (ip_route_me_harder(skb, RTN_UNSPEC))
 247                                 ret = NF_DROP;
 248                 }
 249 #ifdef CONFIG_XFRM
 250                 else if (ct->tuplehash[dir].tuple.dst.u.all !=
 251                          ct->tuplehash[!dir].tuple.src.u.all)
 252                         if (ip_xfrm_me_harder(skb))
 253                                 ret = NF_DROP;
 254 #endif
 255         }
 256         return ret;
 257 }
 258
 259 static unsigned int
 260 nf_nat_adjust(unsigned int hooknum,
 261               struct sk_buff *skb,
 262               const struct net_device *in,
 263               const struct net_device *out,
 264               int (*okfn)(struct sk_buff *))
 265 {
 266         struct nf_conn *ct;
 267         enum ip_conntrack_info ctinfo;
 268
 269         ct = nf_ct_get(skb, &ctinfo);
 270         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 271                 DEBUGP("nf_nat_standalone: adjusting sequence number\n");
 272                 if (!nf_nat_seq_adjust(skb, ct, ctinfo))
 273                         return NF_DROP;
 274         }
 275         return NF_ACCEPT;
 276 }
 277
 278 /* We must be after connection tracking and before packet filtering. */
 279
 280 static struct nf_hook_ops nf_nat_ops[] = {
 281         /* Before packet filtering, change destination */
 282         {
 283                 .hook           = nf_nat_in,
 284                 .owner          = THIS_MODULE,
 285                 .pf             = PF_INET,
 286                 .hooknum        = NF_IP_PRE_ROUTING,
 287                 .priority       = NF_IP_PRI_NAT_DST,
 288         },
 289         /* After packet filtering, change source */
 290         {
 291                 .hook           = nf_nat_out,
 292                 .owner          = THIS_MODULE,
 293                 .pf             = PF_INET,
 294                 .hooknum        = NF_IP_POST_ROUTING,
 295                 .priority       = NF_IP_PRI_NAT_SRC,
 296         },
 297         /* After conntrack, adjust sequence number */
 298         {
 299                 .hook           = nf_nat_adjust,
 300                 .owner          = THIS_MODULE,
 301                 .pf             = PF_INET,
 302                 .hooknum        = NF_IP_POST_ROUTING,
 303                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 304         },
 305         /* Before packet filtering, change destination */
 306         {
 307                 .hook           = nf_nat_local_fn,
 308                 .owner          = THIS_MODULE,
 309                 .pf             = PF_INET,
 310                 .hooknum        = NF_IP_LOCAL_OUT,
 311                 .priority       = NF_IP_PRI_NAT_DST,
 312         },
 313         /* After packet filtering, change source */
 314         {
 315                 .hook           = nf_nat_fn,
 316                 .owner          = THIS_MODULE,
 317                 .pf             = PF_INET,
 318                 .hooknum        = NF_IP_LOCAL_IN,
 319                 .priority       = NF_IP_PRI_NAT_SRC,
 320         },
 321         /* After conntrack, adjust sequence number */
 322         {
 323                 .hook           = nf_nat_adjust,
 324                 .owner          = THIS_MODULE,
 325                 .pf             = PF_INET,
 326                 .hooknum        = NF_IP_LOCAL_IN,
 327                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 328         },
 329 };
 330
 331 static int __init nf_nat_standalone_init(void)
 332 {
 333         int size, ret = 0;
 334
 335         need_conntrack();
 336
 337         size = ALIGN(sizeof(struct nf_conn), __alignof__(struct nf_conn_nat)) +
 338                sizeof(struct nf_conn_nat);
 339         ret = nf_conntrack_register_cache(NF_CT_F_NAT, "nf_nat:base", size);
 340         if (ret < 0) {
 341                 printk(KERN_ERR "nf_nat_init: Unable to create slab cache\n");
 342                 return ret;
 343         }
 344
 345         size = ALIGN(size, __alignof__(struct nf_conn_help)) +
 346                sizeof(struct nf_conn_help);
 347         ret = nf_conntrack_register_cache(NF_CT_F_NAT|NF_CT_F_HELP,
 348                                           "nf_nat:help", size);
 349         if (ret < 0) {
 350                 printk(KERN_ERR "nf_nat_init: Unable to create slab cache\n");
 351                 goto cleanup_register_cache;
 352         }
 353 #ifdef CONFIG_XFRM
 354         BUG_ON(ip_nat_decode_session != NULL);
 355         ip_nat_decode_session = nat_decode_session;
 356 #endif
 357         ret = nf_nat_rule_init();
 358         if (ret < 0) {
 359                 printk("nf_nat_init: can't setup rules.\n");
 360                 goto cleanup_decode_session;
 361         }
 362         ret = nf_register_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 363         if (ret < 0) {
 364                 printk("nf_nat_init: can't register hooks.\n");
 365                 goto cleanup_rule_init;
 366         }
 367         nf_nat_module_is_loaded = 1;
 368         return ret;
 369
 370  cleanup_rule_init:
 371         nf_nat_rule_cleanup();
 372  cleanup_decode_session:
 373 #ifdef CONFIG_XFRM
 374         ip_nat_decode_session = NULL;
 375         synchronize_net();
 376 #endif
 377         nf_conntrack_unregister_cache(NF_CT_F_NAT|NF_CT_F_HELP);
 378  cleanup_register_cache:
 379         nf_conntrack_unregister_cache(NF_CT_F_NAT);
 380         return ret;
 381 }
 382
 383 static void __exit nf_nat_standalone_fini(void)
 384 {
 385         nf_unregister_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 386         nf_nat_rule_cleanup();
 387         nf_nat_module_is_loaded = 0;
 388 #ifdef CONFIG_XFRM
 389         ip_nat_decode_session = NULL;
 390         synchronize_net();
 391 #endif
 392         /* Conntrack caches are unregistered in nf_conntrack_cleanup */
 393 }
 394
 395 module_init(nf_nat_standalone_init);
 396 module_exit(nf_nat_standalone_fini);
 397
 398 MODULE_LICENSE("GPL");
 399 MODULE_ALIAS("ip_nat");