SAFE public projects git trees. - safe/jmp/linux-2.6/blob - 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_conntrack_extend.h>
  23 #include <net/netfilter/nf_nat.h>
  24 #include <net/netfilter/nf_nat_rule.h>
  25 #include <net/netfilter/nf_nat_protocol.h>
  26 #include <net/netfilter/nf_nat_core.h>
  27 #include <net/netfilter/nf_nat_helper.h>
  28 #include <linux/netfilter_ipv4/ip_tables.h>
  29
  30 #ifdef CONFIG_XFRM
  31 static void nat_decode_session(struct sk_buff *skb, struct flowi *fl)
  32 {
  33         const struct nf_conn *ct;
  34         const struct nf_conntrack_tuple *t;
  35         enum ip_conntrack_info ctinfo;
  36         enum ip_conntrack_dir dir;
  37         unsigned long statusbit;
  38
  39         ct = nf_ct_get(skb, &ctinfo);
  40         if (ct == NULL)
  41                 return;
  42         dir = CTINFO2DIR(ctinfo);
  43         t = &ct->tuplehash[dir].tuple;
  44
  45         if (dir == IP_CT_DIR_ORIGINAL)
  46                 statusbit = IPS_DST_NAT;
  47         else
  48                 statusbit = IPS_SRC_NAT;
  49
  50         if (ct->status & statusbit) {
  51                 fl->fl4_dst = t->dst.u3.ip;
  52                 if (t->dst.protonum == IPPROTO_TCP ||
  53                     t->dst.protonum == IPPROTO_UDP ||
  54                     t->dst.protonum == IPPROTO_UDPLITE ||
  55                     t->dst.protonum == IPPROTO_DCCP)
  56                         fl->fl_ip_dport = t->dst.u.tcp.port;
  57         }
  58
  59         statusbit ^= IPS_NAT_MASK;
  60
  61         if (ct->status & statusbit) {
  62                 fl->fl4_src = t->src.u3.ip;
  63                 if (t->dst.protonum == IPPROTO_TCP ||
  64                     t->dst.protonum == IPPROTO_UDP ||
  65                     t->dst.protonum == IPPROTO_UDPLITE ||
  66                     t->dst.protonum == IPPROTO_DCCP)
  67                         fl->fl_ip_sport = t->src.u.tcp.port;
  68         }
  69 }
  70 #endif
  71
  72 static unsigned int
  73 nf_nat_fn(unsigned int hooknum,
  74           struct sk_buff *skb,
  75           const struct net_device *in,
  76           const struct net_device *out,
  77           int (*okfn)(struct sk_buff *))
  78 {
  79         struct nf_conn *ct;
  80         enum ip_conntrack_info ctinfo;
  81         struct nf_conn_nat *nat;
  82         /* maniptype == SRC for postrouting. */
  83         enum nf_nat_manip_type maniptype = HOOK2MANIP(hooknum);
  84
  85         /* We never see fragments: conntrack defrags on pre-routing
  86            and local-out, and nf_nat_out protects post-routing. */
  87         NF_CT_ASSERT(!(ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)));
  88
  89         ct = nf_ct_get(skb, &ctinfo);
  90         /* Can't track?  It's not due to stress, or conntrack would
  91            have dropped it.  Hence it's the user's responsibilty to
  92            packet filter it out, or implement conntrack/NAT for that
  93            protocol. 8) --RR */
  94         if (!ct) {
  95                 /* Exception: ICMP redirect to new connection (not in
  96                    hash table yet).  We must not let this through, in
  97                    case we're doing NAT to the same network. */
  98                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
  99                         struct icmphdr _hdr, *hp;
 100
 101                         hp = skb_header_pointer(skb, ip_hdrlen(skb),
 102                                                 sizeof(_hdr), &_hdr);
 103                         if (hp != NULL &&
 104                             hp->type == ICMP_REDIRECT)
 105                                 return NF_DROP;
 106                 }
 107                 return NF_ACCEPT;
 108         }
 109
 110         /* Don't try to NAT if this packet is not conntracked */
 111         if (ct == &nf_conntrack_untracked)
 112                 return NF_ACCEPT;
 113
 114         nat = nfct_nat(ct);
 115         if (!nat) {
 116                 nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC);
 117                 if (nat == NULL) {
 118                         pr_debug("failed to add NAT extension\n");
 119                         return NF_ACCEPT;
 120                 }
 121         }
 122
 123         switch (ctinfo) {
 124         case IP_CT_RELATED:
 125         case IP_CT_RELATED+IP_CT_IS_REPLY:
 126                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 127                         if (!nf_nat_icmp_reply_translation(ct, ctinfo,
 128                                                            hooknum, skb))
 129                                 return NF_DROP;
 130                         else
 131                                 return NF_ACCEPT;
 132                 }
 133                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 134         case IP_CT_NEW:
 135
 136                 /* Seen it before?  This can happen for loopback, retrans,
 137                    or local packets.. */
 138                 if (!nf_nat_initialized(ct, maniptype)) {
 139                         unsigned int ret;
 140
 141                         if (unlikely(nf_ct_is_confirmed(ct)))
 142                                 /* NAT module was loaded late */
 143                                 ret = alloc_null_binding_confirmed(ct, hooknum);
 144                         else if (hooknum == NF_INET_LOCAL_IN)
 145                                 /* LOCAL_IN hook doesn't have a chain!  */
 146                                 ret = alloc_null_binding(ct, hooknum);
 147                         else
 148                                 ret = nf_nat_rule_find(skb, hooknum, in, out,
 149                                                        ct);
 150
 151                         if (ret != NF_ACCEPT) {
 152                                 return ret;
 153                         }
 154                 } else
 155                         pr_debug("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         const 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         const 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                 pr_debug("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[] __read_mostly = {
 281         /* Before packet filtering, change destination */
 282         {
 283                 .hook           = nf_nat_in,
 284                 .owner          = THIS_MODULE,
 285                 .pf             = PF_INET,
 286                 .hooknum        = NF_INET_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_INET_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_INET_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_INET_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_INET_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_INET_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 ret = 0;
 334
 335         need_ipv4_conntrack();
 336
 337 #ifdef CONFIG_XFRM
 338         BUG_ON(ip_nat_decode_session != NULL);
 339         rcu_assign_pointer(ip_nat_decode_session, nat_decode_session);
 340 #endif
 341         ret = nf_nat_rule_init();
 342         if (ret < 0) {
 343                 printk("nf_nat_init: can't setup rules.\n");
 344                 goto cleanup_decode_session;
 345         }
 346         ret = nf_register_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 347         if (ret < 0) {
 348                 printk("nf_nat_init: can't register hooks.\n");
 349                 goto cleanup_rule_init;
 350         }
 351         return ret;
 352
 353  cleanup_rule_init:
 354         nf_nat_rule_cleanup();
 355  cleanup_decode_session:
 356 #ifdef CONFIG_XFRM
 357         rcu_assign_pointer(ip_nat_decode_session, NULL);
 358         synchronize_net();
 359 #endif
 360         return ret;
 361 }
 362
 363 static void __exit nf_nat_standalone_fini(void)
 364 {
 365         nf_unregister_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 366         nf_nat_rule_cleanup();
 367 #ifdef CONFIG_XFRM
 368         rcu_assign_pointer(ip_nat_decode_session, NULL);
 369         synchronize_net();
 370 #endif
 371         /* Conntrack caches are unregistered in nf_conntrack_cleanup */
 372 }
 373
 374 module_init(nf_nat_standalone_init);
 375 module_exit(nf_nat_standalone_fini);
 376
 377 MODULE_LICENSE("GPL");
 378 MODULE_ALIAS("ip_nat");