net/ipv6/netfilter/ip6table_nat.c

   1 /*
   2  * Copyright (c) 2011 Patrick McHardy <kaber@trash.net>
   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  * Based on Rusty Russell's IPv4 NAT code. Development of IPv6 NAT
   9  * funded by Astaro.
  10  */
  11
  12 #include <linux/module.h>
  13 #include <linux/netfilter.h>
  14 #include <linux/netfilter_ipv6.h>
  15 #include <linux/netfilter_ipv6/ip6_tables.h>
  16 #include <linux/ipv6.h>
  17 #include <net/ipv6.h>
  18
  19 #include <net/netfilter/nf_nat.h>
  20 #include <net/netfilter/nf_nat_core.h>
  21 #include <net/netfilter/nf_nat_l3proto.h>
  22
  23 static const struct xt_table nf_nat_ipv6_table = {
  24         .name           = "nat",
  25         .valid_hooks    = (1 << NF_INET_PRE_ROUTING) |
  26                           (1 << NF_INET_POST_ROUTING) |
  27                           (1 << NF_INET_LOCAL_OUT) |
  28                           (1 << NF_INET_LOCAL_IN),
  29         .me             = THIS_MODULE,
  30         .af             = NFPROTO_IPV6,
  31 };
  32
  33 static unsigned int alloc_null_binding(struct nf_conn *ct, unsigned int hooknum)
  34 {
  35         /* Force range to this IP; let proto decide mapping for
  36          * per-proto parts (hence not IP_NAT_RANGE_PROTO_SPECIFIED).
  37          */
  38         struct nf_nat_range range;
  39
  40         range.flags = 0;
  41         pr_debug("Allocating NULL binding for %p (%pI6)\n", ct,
  42                  HOOK2MANIP(hooknum) == NF_NAT_MANIP_SRC ?
  43                  &ct->tuplehash[IP_CT_DIR_REPLY].tuple.dst.u3.ip6 :
  44                  &ct->tuplehash[IP_CT_DIR_REPLY].tuple.src.u3.ip6);
  45
  46         return nf_nat_setup_info(ct, &range, HOOK2MANIP(hooknum));
  47 }
  48
  49 static unsigned int nf_nat_rule_find(struct sk_buff *skb, unsigned int hooknum,
  50                                      const struct net_device *in,
  51                                      const struct net_device *out,
  52                                      struct nf_conn *ct)
  53 {
  54         struct net *net = nf_ct_net(ct);
  55         unsigned int ret;
  56
  57         ret = ip6t_do_table(skb, hooknum, in, out, net->ipv6.ip6table_nat);
  58         if (ret == NF_ACCEPT) {
  59                 if (!nf_nat_initialized(ct, HOOK2MANIP(hooknum)))
  60                         ret = alloc_null_binding(ct, hooknum);
  61         }
  62         return ret;
  63 }
  64
  65 static unsigned int
  66 nf_nat_ipv6_fn(const struct nf_hook_ops *ops,
  67                struct sk_buff *skb,
  68                const struct net_device *in,
  69                const struct net_device *out,
  70                int (*okfn)(struct sk_buff *))
  71 {
  72         struct nf_conn *ct;
  73         enum ip_conntrack_info ctinfo;
  74         struct nf_conn_nat *nat;
  75         enum nf_nat_manip_type maniptype = HOOK2MANIP(ops->hooknum);
  76         __be16 frag_off;
  77         int hdrlen;
  78         u8 nexthdr;
  79
  80         ct = nf_ct_get(skb, &ctinfo);
  81         /* Can't track?  It's not due to stress, or conntrack would
  82          * have dropped it.  Hence it's the user's responsibilty to
  83          * packet filter it out, or implement conntrack/NAT for that
  84          * protocol. 8) --RR
  85          */
  86         if (!ct)
  87                 return NF_ACCEPT;
  88
  89         /* Don't try to NAT if this packet is not conntracked */
  90         if (nf_ct_is_untracked(ct))
  91                 return NF_ACCEPT;
  92
  93         nat = nfct_nat(ct);
  94         if (!nat) {
  95                 /* NAT module was loaded late. */
  96                 if (nf_ct_is_confirmed(ct))
  97                         return NF_ACCEPT;
  98                 nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC);
  99                 if (nat == NULL) {
 100                         pr_debug("failed to add NAT extension\n");
 101                         return NF_ACCEPT;
 102                 }
 103         }
 104
 105         switch (ctinfo) {
 106         case IP_CT_RELATED:
 107         case IP_CT_RELATED_REPLY:
 108                 nexthdr = ipv6_hdr(skb)->nexthdr;
 109                 hdrlen = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
 110                                           &nexthdr, &frag_off);
 111
 112                 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
 113                         if (!nf_nat_icmpv6_reply_translation(skb, ct, ctinfo,
 114                                                              ops->hooknum,
 115                                                              hdrlen))
 116                                 return NF_DROP;
 117                         else
 118                                 return NF_ACCEPT;
 119                 }
 120                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 121         case IP_CT_NEW:
 122                 /* Seen it before?  This can happen for loopback, retrans,
 123                  * or local packets.
 124                  */
 125                 if (!nf_nat_initialized(ct, maniptype)) {
 126                         unsigned int ret;
 127
 128                         ret = nf_nat_rule_find(skb, ops->hooknum, in, out, ct);
 129                         if (ret != NF_ACCEPT)
 130                                 return ret;
 131                 } else {
 132                         pr_debug("Already setup manip %s for ct %p\n",
 133                                  maniptype == NF_NAT_MANIP_SRC ? "SRC" : "DST",
 134                                  ct);
 135                         if (nf_nat_oif_changed(ops->hooknum, ctinfo, nat, out))
 136                                 goto oif_changed;
 137                 }
 138                 break;
 139
 140         default:
 141                 /* ESTABLISHED */
 142                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 143                              ctinfo == IP_CT_ESTABLISHED_REPLY);
 144                 if (nf_nat_oif_changed(ops->hooknum, ctinfo, nat, out))
 145                         goto oif_changed;
 146         }
 147
 148         return nf_nat_packet(ct, ctinfo, ops->hooknum, skb);
 149
 150 oif_changed:
 151         nf_ct_kill_acct(ct, ctinfo, skb);
 152         return NF_DROP;
 153 }
 154
 155 static unsigned int
 156 nf_nat_ipv6_in(const struct nf_hook_ops *ops,
 157                struct sk_buff *skb,
 158                const struct net_device *in,
 159                const struct net_device *out,
 160                int (*okfn)(struct sk_buff *))
 161 {
 162         unsigned int ret;
 163         struct in6_addr daddr = ipv6_hdr(skb)->daddr;
 164
 165         ret = nf_nat_ipv6_fn(ops, skb, in, out, okfn);
 166         if (ret != NF_DROP && ret != NF_STOLEN &&
 167             ipv6_addr_cmp(&daddr, &ipv6_hdr(skb)->daddr))
 168                 skb_dst_drop(skb);
 169
 170         return ret;
 171 }
 172
 173 static unsigned int
 174 nf_nat_ipv6_out(const struct nf_hook_ops *ops,
 175                 struct sk_buff *skb,
 176                 const struct net_device *in,
 177                 const struct net_device *out,
 178                 int (*okfn)(struct sk_buff *))
 179 {
 180 #ifdef CONFIG_XFRM
 181         const struct nf_conn *ct;
 182         enum ip_conntrack_info ctinfo;
 183         int err;
 184 #endif
 185         unsigned int ret;
 186
 187         /* root is playing with raw sockets. */
 188         if (skb->len < sizeof(struct ipv6hdr))
 189                 return NF_ACCEPT;
 190
 191         ret = nf_nat_ipv6_fn(ops, skb, in, out, okfn);
 192 #ifdef CONFIG_XFRM
 193         if (ret != NF_DROP && ret != NF_STOLEN &&
 194             !(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 195             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 196                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 197
 198                 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.src.u3,
 199                                       &ct->tuplehash[!dir].tuple.dst.u3) ||
 200                     (ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
 201                      ct->tuplehash[dir].tuple.src.u.all !=
 202                      ct->tuplehash[!dir].tuple.dst.u.all)) {
 203                         err = nf_xfrm_me_harder(skb, AF_INET6);
 204                         if (err < 0)
 205                                 ret = NF_DROP_ERR(err);
 206                 }
 207         }
 208 #endif
 209         return ret;
 210 }
 211
 212 static unsigned int
 213 nf_nat_ipv6_local_fn(const struct nf_hook_ops *ops,
 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         int err;
 223
 224         /* root is playing with raw sockets. */
 225         if (skb->len < sizeof(struct ipv6hdr))
 226                 return NF_ACCEPT;
 227
 228         ret = nf_nat_ipv6_fn(ops, 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 (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.dst.u3,
 234                                       &ct->tuplehash[!dir].tuple.src.u3)) {
 235                         err = ip6_route_me_harder(skb);
 236                         if (err < 0)
 237                                 ret = NF_DROP_ERR(err);
 238                 }
 239 #ifdef CONFIG_XFRM
 240                 else if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 241                          ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
 242                          ct->tuplehash[dir].tuple.dst.u.all !=
 243                          ct->tuplehash[!dir].tuple.src.u.all) {
 244                         err = nf_xfrm_me_harder(skb, AF_INET6);
 245                         if (err < 0)
 246                                 ret = NF_DROP_ERR(err);
 247                 }
 248 #endif
 249         }
 250         return ret;
 251 }
 252
 253 static struct nf_hook_ops nf_nat_ipv6_ops[] __read_mostly = {
 254         /* Before packet filtering, change destination */
 255         {
 256                 .hook           = nf_nat_ipv6_in,
 257                 .owner          = THIS_MODULE,
 258                 .pf             = NFPROTO_IPV6,
 259                 .hooknum        = NF_INET_PRE_ROUTING,
 260                 .priority       = NF_IP6_PRI_NAT_DST,
 261         },
 262         /* After packet filtering, change source */
 263         {
 264                 .hook           = nf_nat_ipv6_out,
 265                 .owner          = THIS_MODULE,
 266                 .pf             = NFPROTO_IPV6,
 267                 .hooknum        = NF_INET_POST_ROUTING,
 268                 .priority       = NF_IP6_PRI_NAT_SRC,
 269         },
 270         /* Before packet filtering, change destination */
 271         {
 272                 .hook           = nf_nat_ipv6_local_fn,
 273                 .owner          = THIS_MODULE,
 274                 .pf             = NFPROTO_IPV6,
 275                 .hooknum        = NF_INET_LOCAL_OUT,
 276                 .priority       = NF_IP6_PRI_NAT_DST,
 277         },
 278         /* After packet filtering, change source */
 279         {
 280                 .hook           = nf_nat_ipv6_fn,
 281                 .owner          = THIS_MODULE,
 282                 .pf             = NFPROTO_IPV6,
 283                 .hooknum        = NF_INET_LOCAL_IN,
 284                 .priority       = NF_IP6_PRI_NAT_SRC,
 285         },
 286 };
 287
 288 static int __net_init ip6table_nat_net_init(struct net *net)
 289 {
 290         struct ip6t_replace *repl;
 291
 292         repl = ip6t_alloc_initial_table(&nf_nat_ipv6_table);
 293         if (repl == NULL)
 294                 return -ENOMEM;
 295         net->ipv6.ip6table_nat = ip6t_register_table(net, &nf_nat_ipv6_table, repl);
 296         kfree(repl);
 297         return PTR_ERR_OR_ZERO(net->ipv6.ip6table_nat);
 298 }
 299
 300 static void __net_exit ip6table_nat_net_exit(struct net *net)
 301 {
 302         ip6t_unregister_table(net, net->ipv6.ip6table_nat);
 303 }
 304
 305 static struct pernet_operations ip6table_nat_net_ops = {
 306         .init   = ip6table_nat_net_init,
 307         .exit   = ip6table_nat_net_exit,
 308 };
 309
 310 static int __init ip6table_nat_init(void)
 311 {
 312         int err;
 313
 314         err = register_pernet_subsys(&ip6table_nat_net_ops);
 315         if (err < 0)
 316                 goto err1;
 317
 318         err = nf_register_hooks(nf_nat_ipv6_ops, ARRAY_SIZE(nf_nat_ipv6_ops));
 319         if (err < 0)
 320                 goto err2;
 321         return 0;
 322
 323 err2:
 324         unregister_pernet_subsys(&ip6table_nat_net_ops);
 325 err1:
 326         return err;
 327 }
 328
 329 static void __exit ip6table_nat_exit(void)
 330 {
 331         nf_unregister_hooks(nf_nat_ipv6_ops, ARRAY_SIZE(nf_nat_ipv6_ops));
 332         unregister_pernet_subsys(&ip6table_nat_net_ops);
 333 }
 334
 335 module_init(ip6table_nat_init);
 336 module_exit(ip6table_nat_exit);
 337
 338 MODULE_LICENSE("GPL");