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                     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                 /* 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                 nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC);
 119                 if (nat == NULL) {
 120                         pr_debug("failed to add NAT extension\n");
 121                         return NF_ACCEPT;
 122                 }
 123         }
 124
 125         switch (ctinfo) {
 126         case IP_CT_RELATED:
 127         case IP_CT_RELATED+IP_CT_IS_REPLY:
 128                 if (ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 129                         if (!nf_nat_icmp_reply_translation(ct, ctinfo,
 130                                                            hooknum, skb))
 131                                 return NF_DROP;
 132                         else
 133                                 return NF_ACCEPT;
 134                 }
 135                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 136         case IP_CT_NEW:
 137
 138                 /* Seen it before?  This can happen for loopback, retrans,
 139                    or local packets.. */
 140                 if (!nf_nat_initialized(ct, maniptype)) {
 141                         unsigned int ret;
 142
 143                         if (unlikely(nf_ct_is_confirmed(ct)))
 144                                 /* NAT module was loaded late */
 145                                 ret = alloc_null_binding_confirmed(ct, hooknum);
 146                         else if (hooknum == NF_INET_LOCAL_IN)
 147                                 /* LOCAL_IN hook doesn't have a chain!  */
 148                                 ret = alloc_null_binding(ct, hooknum);
 149                         else
 150                                 ret = nf_nat_rule_find(skb, hooknum, in, out,
 151                                                        ct);
 152
 153                         if (ret != NF_ACCEPT) {
 154                                 return ret;
 155                         }
 156                 } else
 157                         pr_debug("Already setup manip %s for ct %p\n",
 158                                  maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 159                                  ct);
 160                 break;
 161
 162         default:
 163                 /* ESTABLISHED */
 164                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 165                              ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 166         }
 167
 168         return nf_nat_packet(ct, ctinfo, hooknum, skb);
 169 }
 170
 171 static unsigned int
 172 nf_nat_in(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         unsigned int ret;
 179         __be32 daddr = ip_hdr(skb)->daddr;
 180
 181         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 182         if (ret != NF_DROP && ret != NF_STOLEN &&
 183             daddr != ip_hdr(skb)->daddr) {
 184                 dst_release(skb->dst);
 185                 skb->dst = NULL;
 186         }
 187         return ret;
 188 }
 189
 190 static unsigned int
 191 nf_nat_out(unsigned int hooknum,
 192            struct sk_buff *skb,
 193            const struct net_device *in,
 194            const struct net_device *out,
 195            int (*okfn)(struct sk_buff *))
 196 {
 197 #ifdef CONFIG_XFRM
 198         const struct nf_conn *ct;
 199         enum ip_conntrack_info ctinfo;
 200 #endif
 201         unsigned int ret;
 202
 203         /* root is playing with raw sockets. */
 204         if (skb->len < sizeof(struct iphdr) ||
 205             ip_hdrlen(skb) < sizeof(struct iphdr))
 206                 return NF_ACCEPT;
 207
 208         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 209 #ifdef CONFIG_XFRM
 210         if (ret != NF_DROP && ret != NF_STOLEN &&
 211             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 212                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 213
 214                 if (ct->tuplehash[dir].tuple.src.u3.ip !=
 215                     ct->tuplehash[!dir].tuple.dst.u3.ip
 216                     || ct->tuplehash[dir].tuple.src.u.all !=
 217                        ct->tuplehash[!dir].tuple.dst.u.all
 218                     )
 219                         return ip_xfrm_me_harder(skb) == 0 ? ret : NF_DROP;
 220         }
 221 #endif
 222         return ret;
 223 }
 224
 225 static unsigned int
 226 nf_nat_local_fn(unsigned int hooknum,
 227                 struct sk_buff *skb,
 228                 const struct net_device *in,
 229                 const struct net_device *out,
 230                 int (*okfn)(struct sk_buff *))
 231 {
 232         const struct nf_conn *ct;
 233         enum ip_conntrack_info ctinfo;
 234         unsigned int ret;
 235
 236         /* root is playing with raw sockets. */
 237         if (skb->len < sizeof(struct iphdr) ||
 238             ip_hdrlen(skb) < sizeof(struct iphdr))
 239                 return NF_ACCEPT;
 240
 241         ret = nf_nat_fn(hooknum, skb, in, out, okfn);
 242         if (ret != NF_DROP && ret != NF_STOLEN &&
 243             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 244                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 245
 246                 if (ct->tuplehash[dir].tuple.dst.u3.ip !=
 247                     ct->tuplehash[!dir].tuple.src.u3.ip) {
 248                         if (ip_route_me_harder(skb, RTN_UNSPEC))
 249                                 ret = NF_DROP;
 250                 }
 251 #ifdef CONFIG_XFRM
 252                 else if (ct->tuplehash[dir].tuple.dst.u.all !=
 253                          ct->tuplehash[!dir].tuple.src.u.all)
 254                         if (ip_xfrm_me_harder(skb))
 255                                 ret = NF_DROP;
 256 #endif
 257         }
 258         return ret;
 259 }
 260
 261 static unsigned int
 262 nf_nat_adjust(unsigned int hooknum,
 263               struct sk_buff *skb,
 264               const struct net_device *in,
 265               const struct net_device *out,
 266               int (*okfn)(struct sk_buff *))
 267 {
 268         struct nf_conn *ct;
 269         enum ip_conntrack_info ctinfo;
 270
 271         ct = nf_ct_get(skb, &ctinfo);
 272         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 273                 pr_debug("nf_nat_standalone: adjusting sequence number\n");
 274                 if (!nf_nat_seq_adjust(skb, ct, ctinfo))
 275                         return NF_DROP;
 276         }
 277         return NF_ACCEPT;
 278 }
 279
 280 /* We must be after connection tracking and before packet filtering. */
 281
 282 static struct nf_hook_ops nf_nat_ops[] __read_mostly = {
 283         /* Before packet filtering, change destination */
 284         {
 285                 .hook           = nf_nat_in,
 286                 .owner          = THIS_MODULE,
 287                 .pf             = PF_INET,
 288                 .hooknum        = NF_INET_PRE_ROUTING,
 289                 .priority       = NF_IP_PRI_NAT_DST,
 290         },
 291         /* After packet filtering, change source */
 292         {
 293                 .hook           = nf_nat_out,
 294                 .owner          = THIS_MODULE,
 295                 .pf             = PF_INET,
 296                 .hooknum        = NF_INET_POST_ROUTING,
 297                 .priority       = NF_IP_PRI_NAT_SRC,
 298         },
 299         /* After conntrack, adjust sequence number */
 300         {
 301                 .hook           = nf_nat_adjust,
 302                 .owner          = THIS_MODULE,
 303                 .pf             = PF_INET,
 304                 .hooknum        = NF_INET_POST_ROUTING,
 305                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 306         },
 307         /* Before packet filtering, change destination */
 308         {
 309                 .hook           = nf_nat_local_fn,
 310                 .owner          = THIS_MODULE,
 311                 .pf             = PF_INET,
 312                 .hooknum        = NF_INET_LOCAL_OUT,
 313                 .priority       = NF_IP_PRI_NAT_DST,
 314         },
 315         /* After packet filtering, change source */
 316         {
 317                 .hook           = nf_nat_fn,
 318                 .owner          = THIS_MODULE,
 319                 .pf             = PF_INET,
 320                 .hooknum        = NF_INET_LOCAL_IN,
 321                 .priority       = NF_IP_PRI_NAT_SRC,
 322         },
 323         /* After conntrack, adjust sequence number */
 324         {
 325                 .hook           = nf_nat_adjust,
 326                 .owner          = THIS_MODULE,
 327                 .pf             = PF_INET,
 328                 .hooknum        = NF_INET_LOCAL_IN,
 329                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 330         },
 331 };
 332
 333 static int __init nf_nat_standalone_init(void)
 334 {
 335         int ret = 0;
 336
 337         need_ipv4_conntrack();
 338
 339 #ifdef CONFIG_XFRM
 340         BUG_ON(ip_nat_decode_session != NULL);
 341         rcu_assign_pointer(ip_nat_decode_session, nat_decode_session);
 342 #endif
 343         ret = nf_nat_rule_init();
 344         if (ret < 0) {
 345                 printk("nf_nat_init: can't setup rules.\n");
 346                 goto cleanup_decode_session;
 347         }
 348         ret = nf_register_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 349         if (ret < 0) {
 350                 printk("nf_nat_init: can't register hooks.\n");
 351                 goto cleanup_rule_init;
 352         }
 353         return ret;
 354
 355  cleanup_rule_init:
 356         nf_nat_rule_cleanup();
 357  cleanup_decode_session:
 358 #ifdef CONFIG_XFRM
 359         rcu_assign_pointer(ip_nat_decode_session, NULL);
 360         synchronize_net();
 361 #endif
 362         return ret;
 363 }
 364
 365 static void __exit nf_nat_standalone_fini(void)
 366 {
 367         nf_unregister_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 368         nf_nat_rule_cleanup();
 369 #ifdef CONFIG_XFRM
 370         rcu_assign_pointer(ip_nat_decode_session, NULL);
 371         synchronize_net();
 372 #endif
 373         /* Conntrack caches are unregistered in nf_conntrack_cleanup */
 374 }
 375
 376 module_init(nf_nat_standalone_init);
 377 module_exit(nf_nat_standalone_fini);
 378
 379 MODULE_LICENSE("GPL");
 380 MODULE_ALIAS("ip_nat");