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                     t->dst.protonum == IPPROTO_SCTP)
  57                         fl->fl_ip_dport = t->dst.u.tcp.port;
  58         }
  59
  60         statusbit ^= IPS_NAT_MASK;
  61
  62         if (ct->status & statusbit) {
  63                 fl->fl4_src = t->src.u3.ip;
  64                 if (t->dst.protonum == IPPROTO_TCP ||
  65                     t->dst.protonum == IPPROTO_UDP ||
  66                     t->dst.protonum == IPPROTO_UDPLITE ||
  67                     t->dst.protonum == IPPROTO_DCCP ||
  68                     t->dst.protonum == IPPROTO_SCTP)
  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                 return NF_ACCEPT;
  98
  99         /* Don't try to NAT if this packet is not conntracked */
 100         if (ct == &nf_conntrack_untracked)
 101                 return NF_ACCEPT;
 102
 103         nat = nfct_nat(ct);
 104         if (!nat) {
 105                 /* NAT module was loaded late. */
 106                 if (nf_ct_is_confirmed(ct))
 107                         return NF_ACCEPT;
 108                 nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC);
 109                 if (nat == NULL) {
 110                         pr_debug("failed to add NAT extension\n");
 111                         return NF_ACCEPT;
 112                 }
 113         }
 114
 115         switch (ctinfo) {
 116         case IP_CT_RELATED:
 117         case IP_CT_RELATED+IP_CT_IS_REPLY:
 118                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 119                         if (!nf_nat_icmp_reply_translation(ct, ctinfo,
 120                                                            hooknum, skb))
 121                                 return NF_DROP;
 122                         else
 123                                 return NF_ACCEPT;
 124                 }
 125                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 126         case IP_CT_NEW:
 127
 128                 /* Seen it before?  This can happen for loopback, retrans,
 129                    or local packets.. */
 130                 if (!nf_nat_initialized(ct, maniptype)) {
 131                         unsigned int ret;
 132
 133                         if (hooknum == NF_INET_LOCAL_IN)
 134                                 /* LOCAL_IN hook doesn't have a chain!  */
 135                                 ret = alloc_null_binding(ct, hooknum);
 136                         else
 137                                 ret = nf_nat_rule_find(skb, hooknum, in, out,
 138                                                        ct);
 139
 140                         if (ret != NF_ACCEPT) {
 141                                 return ret;
 142                         }
 143                 } else
 144                         pr_debug("Already setup manip %s for ct %p\n",
 145                                  maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 146                                  ct);
 147                 break;
 148
 149         default:
 150                 /* ESTABLISHED */
 151                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 152                              ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 153         }
 154
 155         return nf_nat_packet(ct, ctinfo, hooknum, skb);
 156 }
 157
 158 static unsigned int
 159 nf_nat_in(unsigned int hooknum,
 160           struct sk_buff *skb,
 161           const struct net_device *in,
 162           const struct net_device *out,
 163           int (*okfn)(struct sk_buff *))
 164 {
 165         unsigned int ret;
 166         __be32 daddr = ip_hdr(skb)->daddr;
 167
 168         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 169         if (ret != NF_DROP && ret != NF_STOLEN &&
 170             daddr != ip_hdr(skb)->daddr) {
 171                 dst_release(skb->dst);
 172                 skb->dst = NULL;
 173         }
 174         return ret;
 175 }
 176
 177 static unsigned int
 178 nf_nat_out(unsigned int hooknum,
 179            struct sk_buff *skb,
 180            const struct net_device *in,
 181            const struct net_device *out,
 182            int (*okfn)(struct sk_buff *))
 183 {
 184 #ifdef CONFIG_XFRM
 185         const struct nf_conn *ct;
 186         enum ip_conntrack_info ctinfo;
 187 #endif
 188         unsigned int ret;
 189
 190         /* root is playing with raw sockets. */
 191         if (skb->len < sizeof(struct iphdr) ||
 192             ip_hdrlen(skb) < sizeof(struct iphdr))
 193                 return NF_ACCEPT;
 194
 195         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 196 #ifdef CONFIG_XFRM
 197         if (ret != NF_DROP && ret != NF_STOLEN &&
 198             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 199                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 200
 201                 if (ct->tuplehash[dir].tuple.src.u3.ip !=
 202                     ct->tuplehash[!dir].tuple.dst.u3.ip
 203                     || ct->tuplehash[dir].tuple.src.u.all !=
 204                        ct->tuplehash[!dir].tuple.dst.u.all
 205                     )
 206                         return ip_xfrm_me_harder(skb) == 0 ? ret : NF_DROP;
 207         }
 208 #endif
 209         return ret;
 210 }
 211
 212 static unsigned int
 213 nf_nat_local_fn(unsigned int hooknum,
 214                 struct sk_buff *skb,
 215                 const struct net_device *in,
 216                 const struct net_device *out,
 217                 int (*okfn)(struct sk_buff *))
 218 {
 219         const struct nf_conn *ct;
 220         enum ip_conntrack_info ctinfo;
 221         unsigned int ret;
 222
 223         /* root is playing with raw sockets. */
 224         if (skb->len < sizeof(struct iphdr) ||
 225             ip_hdrlen(skb) < sizeof(struct iphdr))
 226                 return NF_ACCEPT;
 227
 228         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 229         if (ret != NF_DROP && ret != NF_STOLEN &&
 230             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 231                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 232
 233                 if (ct->tuplehash[dir].tuple.dst.u3.ip !=
 234                     ct->tuplehash[!dir].tuple.src.u3.ip) {
 235                         if (ip_route_me_harder(skb, RTN_UNSPEC))
 236                                 ret = NF_DROP;
 237                 }
 238 #ifdef CONFIG_XFRM
 239                 else if (ct->tuplehash[dir].tuple.dst.u.all !=
 240                          ct->tuplehash[!dir].tuple.src.u.all)
 241                         if (ip_xfrm_me_harder(skb))
 242                                 ret = NF_DROP;
 243 #endif
 244         }
 245         return ret;
 246 }
 247
 248 static unsigned int
 249 nf_nat_adjust(unsigned int hooknum,
 250               struct sk_buff *skb,
 251               const struct net_device *in,
 252               const struct net_device *out,
 253               int (*okfn)(struct sk_buff *))
 254 {
 255         struct nf_conn *ct;
 256         enum ip_conntrack_info ctinfo;
 257
 258         ct = nf_ct_get(skb, &ctinfo);
 259         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 260                 pr_debug("nf_nat_standalone: adjusting sequence number\n");
 261                 if (!nf_nat_seq_adjust(skb, ct, ctinfo))
 262                         return NF_DROP;
 263         }
 264         return NF_ACCEPT;
 265 }
 266
 267 /* We must be after connection tracking and before packet filtering. */
 268
 269 static struct nf_hook_ops nf_nat_ops[] __read_mostly = {
 270         /* Before packet filtering, change destination */
 271         {
 272                 .hook           = nf_nat_in,
 273                 .owner          = THIS_MODULE,
 274                 .pf             = PF_INET,
 275                 .hooknum        = NF_INET_PRE_ROUTING,
 276                 .priority       = NF_IP_PRI_NAT_DST,
 277         },
 278         /* After packet filtering, change source */
 279         {
 280                 .hook           = nf_nat_out,
 281                 .owner          = THIS_MODULE,
 282                 .pf             = PF_INET,
 283                 .hooknum        = NF_INET_POST_ROUTING,
 284                 .priority       = NF_IP_PRI_NAT_SRC,
 285         },
 286         /* After conntrack, adjust sequence number */
 287         {
 288                 .hook           = nf_nat_adjust,
 289                 .owner          = THIS_MODULE,
 290                 .pf             = PF_INET,
 291                 .hooknum        = NF_INET_POST_ROUTING,
 292                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 293         },
 294         /* Before packet filtering, change destination */
 295         {
 296                 .hook           = nf_nat_local_fn,
 297                 .owner          = THIS_MODULE,
 298                 .pf             = PF_INET,
 299                 .hooknum        = NF_INET_LOCAL_OUT,
 300                 .priority       = NF_IP_PRI_NAT_DST,
 301         },
 302         /* After packet filtering, change source */
 303         {
 304                 .hook           = nf_nat_fn,
 305                 .owner          = THIS_MODULE,
 306                 .pf             = PF_INET,
 307                 .hooknum        = NF_INET_LOCAL_IN,
 308                 .priority       = NF_IP_PRI_NAT_SRC,
 309         },
 310         /* After conntrack, adjust sequence number */
 311         {
 312                 .hook           = nf_nat_adjust,
 313                 .owner          = THIS_MODULE,
 314                 .pf             = PF_INET,
 315                 .hooknum        = NF_INET_LOCAL_IN,
 316                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 317         },
 318 };
 319
 320 static int __init nf_nat_standalone_init(void)
 321 {
 322         int ret = 0;
 323
 324         need_ipv4_conntrack();
 325
 326 #ifdef CONFIG_XFRM
 327         BUG_ON(ip_nat_decode_session != NULL);
 328         rcu_assign_pointer(ip_nat_decode_session, nat_decode_session);
 329 #endif
 330         ret = nf_nat_rule_init();
 331         if (ret < 0) {
 332                 printk("nf_nat_init: can't setup rules.\n");
 333                 goto cleanup_decode_session;
 334         }
 335         ret = nf_register_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 336         if (ret < 0) {
 337                 printk("nf_nat_init: can't register hooks.\n");
 338                 goto cleanup_rule_init;
 339         }
 340         return ret;
 341
 342  cleanup_rule_init:
 343         nf_nat_rule_cleanup();
 344  cleanup_decode_session:
 345 #ifdef CONFIG_XFRM
 346         rcu_assign_pointer(ip_nat_decode_session, NULL);
 347         synchronize_net();
 348 #endif
 349         return ret;
 350 }
 351
 352 static void __exit nf_nat_standalone_fini(void)
 353 {
 354         nf_unregister_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 355         nf_nat_rule_cleanup();
 356 #ifdef CONFIG_XFRM
 357         rcu_assign_pointer(ip_nat_decode_session, NULL);
 358         synchronize_net();
 359 #endif
 360         /* Conntrack caches are unregistered in nf_conntrack_cleanup */
 361 }
 362
 363 module_init(nf_nat_standalone_init);
 364 module_exit(nf_nat_standalone_fini);
 365
 366 MODULE_LICENSE("GPL");
 367 MODULE_ALIAS("ip_nat");