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/gfp.h>
  11 #include <linux/ip.h>
  12 #include <linux/netfilter.h>
  13 #include <linux/netfilter_ipv4.h>
  14 #include <linux/module.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/proc_fs.h>
  17 #include <net/ip.h>
  18 #include <net/checksum.h>
  19 #include <linux/spinlock.h>
  20
  21 #include <net/netfilter/nf_conntrack.h>
  22 #include <net/netfilter/nf_conntrack_core.h>
  23 #include <net/netfilter/nf_conntrack_extend.h>
  24 #include <net/netfilter/nf_nat.h>
  25 #include <net/netfilter/nf_nat_rule.h>
  26 #include <net/netfilter/nf_nat_protocol.h>
  27 #include <net/netfilter/nf_nat_core.h>
  28 #include <net/netfilter/nf_nat_helper.h>
  29 #include <linux/netfilter_ipv4/ip_tables.h>
  30
  31 #ifdef CONFIG_XFRM
  32 static void nat_decode_session(struct sk_buff *skb, struct flowi *fl)
  33 {
  34         const struct nf_conn *ct;
  35         const struct nf_conntrack_tuple *t;
  36         enum ip_conntrack_info ctinfo;
  37         enum ip_conntrack_dir dir;
  38         unsigned long statusbit;
  39
  40         ct = nf_ct_get(skb, &ctinfo);
  41         if (ct == NULL)
  42                 return;
  43         dir = CTINFO2DIR(ctinfo);
  44         t = &ct->tuplehash[dir].tuple;
  45
  46         if (dir == IP_CT_DIR_ORIGINAL)
  47                 statusbit = IPS_DST_NAT;
  48         else
  49                 statusbit = IPS_SRC_NAT;
  50
  51         if (ct->status & statusbit) {
  52                 fl->fl4_dst = t->dst.u3.ip;
  53                 if (t->dst.protonum == IPPROTO_TCP ||
  54                     t->dst.protonum == IPPROTO_UDP ||
  55                     t->dst.protonum == IPPROTO_UDPLITE ||
  56                     t->dst.protonum == IPPROTO_DCCP ||
  57                     t->dst.protonum == IPPROTO_SCTP)
  58                         fl->fl_ip_dport = t->dst.u.tcp.port;
  59         }
  60
  61         statusbit ^= IPS_NAT_MASK;
  62
  63         if (ct->status & statusbit) {
  64                 fl->fl4_src = t->src.u3.ip;
  65                 if (t->dst.protonum == IPPROTO_TCP ||
  66                     t->dst.protonum == IPPROTO_UDP ||
  67                     t->dst.protonum == IPPROTO_UDPLITE ||
  68                     t->dst.protonum == IPPROTO_DCCP ||
  69                     t->dst.protonum == IPPROTO_SCTP)
  70                         fl->fl_ip_sport = t->src.u.tcp.port;
  71         }
  72 }
  73 #endif
  74
  75 static unsigned int
  76 nf_nat_fn(unsigned int hooknum,
  77           struct sk_buff *skb,
  78           const struct net_device *in,
  79           const struct net_device *out,
  80           int (*okfn)(struct sk_buff *))
  81 {
  82         struct nf_conn *ct;
  83         enum ip_conntrack_info ctinfo;
  84         struct nf_conn_nat *nat;
  85         /* maniptype == SRC for postrouting. */
  86         enum nf_nat_manip_type maniptype = HOOK2MANIP(hooknum);
  87
  88         /* We never see fragments: conntrack defrags on pre-routing
  89            and local-out, and nf_nat_out protects post-routing. */
  90         NF_CT_ASSERT(!(ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)));
  91
  92         ct = nf_ct_get(skb, &ctinfo);
  93         /* Can't track?  It's not due to stress, or conntrack would
  94            have dropped it.  Hence it's the user's responsibilty to
  95            packet filter it out, or implement conntrack/NAT for that
  96            protocol. 8) --RR */
  97         if (!ct)
  98                 return NF_ACCEPT;
  99
 100         /* Don't try to NAT if this packet is not conntracked */
 101         if (nf_ct_is_untracked(ct))
 102                 return NF_ACCEPT;
 103
 104         nat = nfct_nat(ct);
 105         if (!nat) {
 106                 /* NAT module was loaded late. */
 107                 if (nf_ct_is_confirmed(ct))
 108                         return NF_ACCEPT;
 109                 nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC);
 110                 if (nat == NULL) {
 111                         pr_debug("failed to add NAT extension\n");
 112                         return NF_ACCEPT;
 113                 }
 114         }
 115
 116         switch (ctinfo) {
 117         case IP_CT_RELATED:
 118         case IP_CT_RELATED+IP_CT_IS_REPLY:
 119                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 120                         if (!nf_nat_icmp_reply_translation(ct, ctinfo,
 121                                                            hooknum, skb))
 122                                 return NF_DROP;
 123                         else
 124                                 return NF_ACCEPT;
 125                 }
 126                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 127         case IP_CT_NEW:
 128
 129                 /* Seen it before?  This can happen for loopback, retrans,
 130                    or local packets.. */
 131                 if (!nf_nat_initialized(ct, maniptype)) {
 132                         unsigned int ret;
 133
 134                         ret = nf_nat_rule_find(skb, hooknum, in, out, ct);
 135                         if (ret != NF_ACCEPT)
 136                                 return ret;
 137                 } else
 138                         pr_debug("Already setup manip %s for ct %p\n",
 139                                  maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 140                                  ct);
 141                 break;
 142
 143         default:
 144                 /* ESTABLISHED */
 145                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 146                              ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 147         }
 148
 149         return nf_nat_packet(ct, ctinfo, hooknum, skb);
 150 }
 151
 152 static unsigned int
 153 nf_nat_in(unsigned int hooknum,
 154           struct sk_buff *skb,
 155           const struct net_device *in,
 156           const struct net_device *out,
 157           int (*okfn)(struct sk_buff *))
 158 {
 159         unsigned int ret;
 160         __be32 daddr = ip_hdr(skb)->daddr;
 161
 162         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 163         if (ret != NF_DROP && ret != NF_STOLEN &&
 164             daddr != ip_hdr(skb)->daddr)
 165                 skb_dst_drop(skb);
 166
 167         return ret;
 168 }
 169
 170 static unsigned int
 171 nf_nat_out(unsigned int hooknum,
 172            struct sk_buff *skb,
 173            const struct net_device *in,
 174            const struct net_device *out,
 175            int (*okfn)(struct sk_buff *))
 176 {
 177 #ifdef CONFIG_XFRM
 178         const struct nf_conn *ct;
 179         enum ip_conntrack_info ctinfo;
 180 #endif
 181         unsigned int ret;
 182
 183         /* root is playing with raw sockets. */
 184         if (skb->len < sizeof(struct iphdr) ||
 185             ip_hdrlen(skb) < sizeof(struct iphdr))
 186                 return NF_ACCEPT;
 187
 188         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 189 #ifdef CONFIG_XFRM
 190         if (ret != NF_DROP && ret != NF_STOLEN &&
 191             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 192                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 193
 194                 if ((ct->tuplehash[dir].tuple.src.u3.ip !=
 195                      ct->tuplehash[!dir].tuple.dst.u3.ip) ||
 196                     (ct->tuplehash[dir].tuple.src.u.all !=
 197                      ct->tuplehash[!dir].tuple.dst.u.all)
 198                    )
 199                         return ip_xfrm_me_harder(skb) == 0 ? ret : NF_DROP;
 200         }
 201 #endif
 202         return ret;
 203 }
 204
 205 static unsigned int
 206 nf_nat_local_fn(unsigned int hooknum,
 207                 struct sk_buff *skb,
 208                 const struct net_device *in,
 209                 const struct net_device *out,
 210                 int (*okfn)(struct sk_buff *))
 211 {
 212         const struct nf_conn *ct;
 213         enum ip_conntrack_info ctinfo;
 214         unsigned int ret;
 215
 216         /* root is playing with raw sockets. */
 217         if (skb->len < sizeof(struct iphdr) ||
 218             ip_hdrlen(skb) < sizeof(struct iphdr))
 219                 return NF_ACCEPT;
 220
 221         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 222         if (ret != NF_DROP && ret != NF_STOLEN &&
 223             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 224                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 225
 226                 if (ct->tuplehash[dir].tuple.dst.u3.ip !=
 227                     ct->tuplehash[!dir].tuple.src.u3.ip) {
 228                         if (ip_route_me_harder(skb, RTN_UNSPEC))
 229                                 ret = NF_DROP;
 230                 }
 231 #ifdef CONFIG_XFRM
 232                 else if (ct->tuplehash[dir].tuple.dst.u.all !=
 233                          ct->tuplehash[!dir].tuple.src.u.all)
 234                         if (ip_xfrm_me_harder(skb))
 235                                 ret = NF_DROP;
 236 #endif
 237         }
 238         return ret;
 239 }
 240
 241 /* We must be after connection tracking and before packet filtering. */
 242
 243 static struct nf_hook_ops nf_nat_ops[] __read_mostly = {
 244         /* Before packet filtering, change destination */
 245         {
 246                 .hook           = nf_nat_in,
 247                 .owner          = THIS_MODULE,
 248                 .pf             = NFPROTO_IPV4,
 249                 .hooknum        = NF_INET_PRE_ROUTING,
 250                 .priority       = NF_IP_PRI_NAT_DST,
 251         },
 252         /* After packet filtering, change source */
 253         {
 254                 .hook           = nf_nat_out,
 255                 .owner          = THIS_MODULE,
 256                 .pf             = NFPROTO_IPV4,
 257                 .hooknum        = NF_INET_POST_ROUTING,
 258                 .priority       = NF_IP_PRI_NAT_SRC,
 259         },
 260         /* Before packet filtering, change destination */
 261         {
 262                 .hook           = nf_nat_local_fn,
 263                 .owner          = THIS_MODULE,
 264                 .pf             = NFPROTO_IPV4,
 265                 .hooknum        = NF_INET_LOCAL_OUT,
 266                 .priority       = NF_IP_PRI_NAT_DST,
 267         },
 268         /* After packet filtering, change source */
 269         {
 270                 .hook           = nf_nat_fn,
 271                 .owner          = THIS_MODULE,
 272                 .pf             = NFPROTO_IPV4,
 273                 .hooknum        = NF_INET_LOCAL_IN,
 274                 .priority       = NF_IP_PRI_NAT_SRC,
 275         },
 276 };
 277
 278 static int __init nf_nat_standalone_init(void)
 279 {
 280         int ret = 0;
 281
 282         need_ipv4_conntrack();
 283
 284 #ifdef CONFIG_XFRM
 285         BUG_ON(ip_nat_decode_session != NULL);
 286         rcu_assign_pointer(ip_nat_decode_session, nat_decode_session);
 287 #endif
 288         ret = nf_nat_rule_init();
 289         if (ret < 0) {
 290                 pr_err("nf_nat_init: can't setup rules.\n");
 291                 goto cleanup_decode_session;
 292         }
 293         ret = nf_register_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 294         if (ret < 0) {
 295                 pr_err("nf_nat_init: can't register hooks.\n");
 296                 goto cleanup_rule_init;
 297         }
 298         return ret;
 299
 300  cleanup_rule_init:
 301         nf_nat_rule_cleanup();
 302  cleanup_decode_session:
 303 #ifdef CONFIG_XFRM
 304         rcu_assign_pointer(ip_nat_decode_session, NULL);
 305         synchronize_net();
 306 #endif
 307         return ret;
 308 }
 309
 310 static void __exit nf_nat_standalone_fini(void)
 311 {
 312         nf_unregister_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 313         nf_nat_rule_cleanup();
 314 #ifdef CONFIG_XFRM
 315         rcu_assign_pointer(ip_nat_decode_session, NULL);
 316         synchronize_net();
 317 #endif
 318         /* Conntrack caches are unregistered in nf_conntrack_cleanup */
 319 }
 320
 321 module_init(nf_nat_standalone_init);
 322 module_exit(nf_nat_standalone_fini);
 323
 324 MODULE_LICENSE("GPL");
 325 MODULE_ALIAS("ip_nat");