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         struct flowi4 *fl4 = &fl->u.ip4;
  35         const struct nf_conn *ct;
  36         const struct nf_conntrack_tuple *t;
  37         enum ip_conntrack_info ctinfo;
  38         enum ip_conntrack_dir dir;
  39         unsigned long statusbit;
  40
  41         ct = nf_ct_get(skb, &ctinfo);
  42         if (ct == NULL)
  43                 return;
  44         dir = CTINFO2DIR(ctinfo);
  45         t = &ct->tuplehash[dir].tuple;
  46
  47         if (dir == IP_CT_DIR_ORIGINAL)
  48                 statusbit = IPS_DST_NAT;
  49         else
  50                 statusbit = IPS_SRC_NAT;
  51
  52         if (ct->status & statusbit) {
  53                 fl4->daddr = t->dst.u3.ip;
  54                 if (t->dst.protonum == IPPROTO_TCP ||
  55                     t->dst.protonum == IPPROTO_UDP ||
  56                     t->dst.protonum == IPPROTO_UDPLITE ||
  57                     t->dst.protonum == IPPROTO_DCCP ||
  58                     t->dst.protonum == IPPROTO_SCTP)
  59                         fl4->fl4_dport = t->dst.u.tcp.port;
  60         }
  61
  62         statusbit ^= IPS_NAT_MASK;
  63
  64         if (ct->status & statusbit) {
  65                 fl4->saddr = t->src.u3.ip;
  66                 if (t->dst.protonum == IPPROTO_TCP ||
  67                     t->dst.protonum == IPPROTO_UDP ||
  68                     t->dst.protonum == IPPROTO_UDPLITE ||
  69                     t->dst.protonum == IPPROTO_DCCP ||
  70                     t->dst.protonum == IPPROTO_SCTP)
  71                         fl4->fl4_sport = t->src.u.tcp.port;
  72         }
  73 }
  74 #endif
  75
  76 static unsigned int
  77 nf_nat_fn(unsigned int hooknum,
  78           struct sk_buff *skb,
  79           const struct net_device *in,
  80           const struct net_device *out,
  81           int (*okfn)(struct sk_buff *))
  82 {
  83         struct nf_conn *ct;
  84         enum ip_conntrack_info ctinfo;
  85         struct nf_conn_nat *nat;
  86         /* maniptype == SRC for postrouting. */
  87         enum nf_nat_manip_type maniptype = HOOK2MANIP(hooknum);
  88
  89         /* We never see fragments: conntrack defrags on pre-routing
  90            and local-out, and nf_nat_out protects post-routing. */
  91         NF_CT_ASSERT(!ip_is_fragment(ip_hdr(skb)));
  92
  93         ct = nf_ct_get(skb, &ctinfo);
  94         /* Can't track?  It's not due to stress, or conntrack would
  95            have dropped it.  Hence it's the user's responsibilty to
  96            packet filter it out, or implement conntrack/NAT for that
  97            protocol. 8) --RR */
  98         if (!ct)
  99                 return NF_ACCEPT;
 100
 101         /* Don't try to NAT if this packet is not conntracked */
 102         if (nf_ct_is_untracked(ct))
 103                 return NF_ACCEPT;
 104
 105         nat = nfct_nat(ct);
 106         if (!nat) {
 107                 /* NAT module was loaded late. */
 108                 if (nf_ct_is_confirmed(ct))
 109                         return NF_ACCEPT;
 110                 nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC);
 111                 if (nat == NULL) {
 112                         pr_debug("failed to add NAT extension\n");
 113                         return NF_ACCEPT;
 114                 }
 115         }
 116
 117         switch (ctinfo) {
 118         case IP_CT_RELATED:
 119         case IP_CT_RELATED_REPLY:
 120                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 121                         if (!nf_nat_icmp_reply_translation(ct, ctinfo,
 122                                                            hooknum, skb))
 123                                 return NF_DROP;
 124                         else
 125                                 return NF_ACCEPT;
 126                 }
 127                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 128         case IP_CT_NEW:
 129
 130                 /* Seen it before?  This can happen for loopback, retrans,
 131                    or local packets.. */
 132                 if (!nf_nat_initialized(ct, maniptype)) {
 133                         unsigned int ret;
 134
 135                         ret = nf_nat_rule_find(skb, hooknum, in, out, ct);
 136                         if (ret != NF_ACCEPT)
 137                                 return ret;
 138                 } else
 139                         pr_debug("Already setup manip %s for ct %p\n",
 140                                  maniptype == NF_NAT_MANIP_SRC ? "SRC" : "DST",
 141                                  ct);
 142                 break;
 143
 144         default:
 145                 /* ESTABLISHED */
 146                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 147                              ctinfo == IP_CT_ESTABLISHED_REPLY);
 148         }
 149
 150         return nf_nat_packet(ct, ctinfo, hooknum, skb);
 151 }
 152
 153 static unsigned int
 154 nf_nat_in(unsigned int hooknum,
 155           struct sk_buff *skb,
 156           const struct net_device *in,
 157           const struct net_device *out,
 158           int (*okfn)(struct sk_buff *))
 159 {
 160         unsigned int ret;
 161         __be32 daddr = ip_hdr(skb)->daddr;
 162
 163         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 164         if (ret != NF_DROP && ret != NF_STOLEN &&
 165             daddr != ip_hdr(skb)->daddr)
 166                 skb_dst_drop(skb);
 167
 168         return ret;
 169 }
 170
 171 static unsigned int
 172 nf_nat_out(unsigned int hooknum,
 173            struct sk_buff *skb,
 174            const struct net_device *in,
 175            const struct net_device *out,
 176            int (*okfn)(struct sk_buff *))
 177 {
 178 #ifdef CONFIG_XFRM
 179         const struct nf_conn *ct;
 180         enum ip_conntrack_info ctinfo;
 181 #endif
 182         unsigned int ret;
 183
 184         /* root is playing with raw sockets. */
 185         if (skb->len < sizeof(struct iphdr) ||
 186             ip_hdrlen(skb) < sizeof(struct iphdr))
 187                 return NF_ACCEPT;
 188
 189         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 190 #ifdef CONFIG_XFRM
 191         if (ret != NF_DROP && ret != NF_STOLEN &&
 192             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 193                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 194
 195                 if ((ct->tuplehash[dir].tuple.src.u3.ip !=
 196                      ct->tuplehash[!dir].tuple.dst.u3.ip) ||
 197                     (ct->tuplehash[dir].tuple.src.u.all !=
 198                      ct->tuplehash[!dir].tuple.dst.u.all)
 199                    )
 200                         return ip_xfrm_me_harder(skb) == 0 ? ret : NF_DROP;
 201         }
 202 #endif
 203         return ret;
 204 }
 205
 206 static unsigned int
 207 nf_nat_local_fn(unsigned int hooknum,
 208                 struct sk_buff *skb,
 209                 const struct net_device *in,
 210                 const struct net_device *out,
 211                 int (*okfn)(struct sk_buff *))
 212 {
 213         const struct nf_conn *ct;
 214         enum ip_conntrack_info ctinfo;
 215         unsigned int ret;
 216
 217         /* root is playing with raw sockets. */
 218         if (skb->len < sizeof(struct iphdr) ||
 219             ip_hdrlen(skb) < sizeof(struct iphdr))
 220                 return NF_ACCEPT;
 221
 222         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 223         if (ret != NF_DROP && ret != NF_STOLEN &&
 224             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 225                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 226
 227                 if (ct->tuplehash[dir].tuple.dst.u3.ip !=
 228                     ct->tuplehash[!dir].tuple.src.u3.ip) {
 229                         if (ip_route_me_harder(skb, RTN_UNSPEC))
 230                                 ret = NF_DROP;
 231                 }
 232 #ifdef CONFIG_XFRM
 233                 else if (ct->tuplehash[dir].tuple.dst.u.all !=
 234                          ct->tuplehash[!dir].tuple.src.u.all)
 235                         if (ip_xfrm_me_harder(skb))
 236                                 ret = NF_DROP;
 237 #endif
 238         }
 239         return ret;
 240 }
 241
 242 /* We must be after connection tracking and before packet filtering. */
 243
 244 static struct nf_hook_ops nf_nat_ops[] __read_mostly = {
 245         /* Before packet filtering, change destination */
 246         {
 247                 .hook           = nf_nat_in,
 248                 .owner          = THIS_MODULE,
 249                 .pf             = NFPROTO_IPV4,
 250                 .hooknum        = NF_INET_PRE_ROUTING,
 251                 .priority       = NF_IP_PRI_NAT_DST,
 252         },
 253         /* After packet filtering, change source */
 254         {
 255                 .hook           = nf_nat_out,
 256                 .owner          = THIS_MODULE,
 257                 .pf             = NFPROTO_IPV4,
 258                 .hooknum        = NF_INET_POST_ROUTING,
 259                 .priority       = NF_IP_PRI_NAT_SRC,
 260         },
 261         /* Before packet filtering, change destination */
 262         {
 263                 .hook           = nf_nat_local_fn,
 264                 .owner          = THIS_MODULE,
 265                 .pf             = NFPROTO_IPV4,
 266                 .hooknum        = NF_INET_LOCAL_OUT,
 267                 .priority       = NF_IP_PRI_NAT_DST,
 268         },
 269         /* After packet filtering, change source */
 270         {
 271                 .hook           = nf_nat_fn,
 272                 .owner          = THIS_MODULE,
 273                 .pf             = NFPROTO_IPV4,
 274                 .hooknum        = NF_INET_LOCAL_IN,
 275                 .priority       = NF_IP_PRI_NAT_SRC,
 276         },
 277 };
 278
 279 static int __init nf_nat_standalone_init(void)
 280 {
 281         int ret = 0;
 282
 283         need_ipv4_conntrack();
 284
 285 #ifdef CONFIG_XFRM
 286         BUG_ON(ip_nat_decode_session != NULL);
 287         RCU_INIT_POINTER(ip_nat_decode_session, nat_decode_session);
 288 #endif
 289         ret = nf_nat_rule_init();
 290         if (ret < 0) {
 291                 pr_err("nf_nat_init: can't setup rules.\n");
 292                 goto cleanup_decode_session;
 293         }
 294         ret = nf_register_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 295         if (ret < 0) {
 296                 pr_err("nf_nat_init: can't register hooks.\n");
 297                 goto cleanup_rule_init;
 298         }
 299         return ret;
 300
 301  cleanup_rule_init:
 302         nf_nat_rule_cleanup();
 303  cleanup_decode_session:
 304 #ifdef CONFIG_XFRM
 305         RCU_INIT_POINTER(ip_nat_decode_session, NULL);
 306         synchronize_net();
 307 #endif
 308         return ret;
 309 }
 310
 311 static void __exit nf_nat_standalone_fini(void)
 312 {
 313         nf_unregister_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 314         nf_nat_rule_cleanup();
 315 #ifdef CONFIG_XFRM
 316         RCU_INIT_POINTER(ip_nat_decode_session, NULL);
 317         synchronize_net();
 318 #endif
 319         /* Conntrack caches are unregistered in nf_conntrack_cleanup */
 320 }
 321
 322 module_init(nf_nat_standalone_init);
 323 module_exit(nf_nat_standalone_fini);
 324
 325 MODULE_LICENSE("GPL");
 326 MODULE_ALIAS("ip_nat");