net/ipv6/netfilter/ip6table_nat.c

   1 /*
   2  * Copyright (c) 2011 Patrick McHardy <[email protected]>
   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 = nf_ct_nat_ext_add(ct);
  94         if (nat == NULL)
  95                 return NF_ACCEPT;
  96
  97         switch (ctinfo) {
  98         case IP_CT_RELATED:
  99         case IP_CT_RELATED_REPLY:
 100                 nexthdr = ipv6_hdr(skb)->nexthdr;
 101                 hdrlen = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr),
 102                                           &nexthdr, &frag_off);
 103
 104                 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
 105                         if (!nf_nat_icmpv6_reply_translation(skb, ct, ctinfo,
 106                                                              ops->hooknum,
 107                                                              hdrlen))
 108                                 return NF_DROP;
 109                         else
 110                                 return NF_ACCEPT;
 111                 }
 112                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 113         case IP_CT_NEW:
 114                 /* Seen it before?  This can happen for loopback, retrans,
 115                  * or local packets.
 116                  */
 117                 if (!nf_nat_initialized(ct, maniptype)) {
 118                         unsigned int ret;
 119
 120                         ret = nf_nat_rule_find(skb, ops->hooknum, in, out, ct);
 121                         if (ret != NF_ACCEPT)
 122                                 return ret;
 123                 } else {
 124                         pr_debug("Already setup manip %s for ct %p\n",
 125                                  maniptype == NF_NAT_MANIP_SRC ? "SRC" : "DST",
 126                                  ct);
 127                         if (nf_nat_oif_changed(ops->hooknum, ctinfo, nat, out))
 128                                 goto oif_changed;
 129                 }
 130                 break;
 131
 132         default:
 133                 /* ESTABLISHED */
 134                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 135                              ctinfo == IP_CT_ESTABLISHED_REPLY);
 136                 if (nf_nat_oif_changed(ops->hooknum, ctinfo, nat, out))
 137                         goto oif_changed;
 138         }
 139
 140         return nf_nat_packet(ct, ctinfo, ops->hooknum, skb);
 141
 142 oif_changed:
 143         nf_ct_kill_acct(ct, ctinfo, skb);
 144         return NF_DROP;
 145 }
 146
 147 static unsigned int
 148 nf_nat_ipv6_in(const struct nf_hook_ops *ops,
 149                struct sk_buff *skb,
 150                const struct net_device *in,
 151                const struct net_device *out,
 152                int (*okfn)(struct sk_buff *))
 153 {
 154         unsigned int ret;
 155         struct in6_addr daddr = ipv6_hdr(skb)->daddr;
 156
 157         ret = nf_nat_ipv6_fn(ops, skb, in, out, okfn);
 158         if (ret != NF_DROP && ret != NF_STOLEN &&
 159             ipv6_addr_cmp(&daddr, &ipv6_hdr(skb)->daddr))
 160                 skb_dst_drop(skb);
 161
 162         return ret;
 163 }
 164
 165 static unsigned int
 166 nf_nat_ipv6_out(const struct nf_hook_ops *ops,
 167                 struct sk_buff *skb,
 168                 const struct net_device *in,
 169                 const struct net_device *out,
 170                 int (*okfn)(struct sk_buff *))
 171 {
 172 #ifdef CONFIG_XFRM
 173         const struct nf_conn *ct;
 174         enum ip_conntrack_info ctinfo;
 175         int err;
 176 #endif
 177         unsigned int ret;
 178
 179         /* root is playing with raw sockets. */
 180         if (skb->len < sizeof(struct ipv6hdr))
 181                 return NF_ACCEPT;
 182
 183         ret = nf_nat_ipv6_fn(ops, skb, in, out, okfn);
 184 #ifdef CONFIG_XFRM
 185         if (ret != NF_DROP && ret != NF_STOLEN &&
 186             !(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 187             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 188                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 189
 190                 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.src.u3,
 191                                       &ct->tuplehash[!dir].tuple.dst.u3) ||
 192                     (ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
 193                      ct->tuplehash[dir].tuple.src.u.all !=
 194                      ct->tuplehash[!dir].tuple.dst.u.all)) {
 195                         err = nf_xfrm_me_harder(skb, AF_INET6);
 196                         if (err < 0)
 197                                 ret = NF_DROP_ERR(err);
 198                 }
 199         }
 200 #endif
 201         return ret;
 202 }
 203
 204 static unsigned int
 205 nf_nat_ipv6_local_fn(const struct nf_hook_ops *ops,
 206                      struct sk_buff *skb,
 207                      const struct net_device *in,
 208                      const struct net_device *out,
 209                      int (*okfn)(struct sk_buff *))
 210 {
 211         const struct nf_conn *ct;
 212         enum ip_conntrack_info ctinfo;
 213         unsigned int ret;
 214         int err;
 215
 216         /* root is playing with raw sockets. */
 217         if (skb->len < sizeof(struct ipv6hdr))
 218                 return NF_ACCEPT;
 219
 220         ret = nf_nat_ipv6_fn(ops, skb, in, out, okfn);
 221         if (ret != NF_DROP && ret != NF_STOLEN &&
 222             (ct = nf_ct_get(skb, &ctinfo)) != NULL) {
 223                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 224
 225                 if (!nf_inet_addr_cmp(&ct->tuplehash[dir].tuple.dst.u3,
 226                                       &ct->tuplehash[!dir].tuple.src.u3)) {
 227                         err = ip6_route_me_harder(skb);
 228                         if (err < 0)
 229                                 ret = NF_DROP_ERR(err);
 230                 }
 231 #ifdef CONFIG_XFRM
 232                 else if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
 233                          ct->tuplehash[dir].tuple.dst.protonum != IPPROTO_ICMPV6 &&
 234                          ct->tuplehash[dir].tuple.dst.u.all !=
 235                          ct->tuplehash[!dir].tuple.src.u.all) {
 236                         err = nf_xfrm_me_harder(skb, AF_INET6);
 237                         if (err < 0)
 238                                 ret = NF_DROP_ERR(err);
 239                 }
 240 #endif
 241         }
 242         return ret;
 243 }
 244
 245 static struct nf_hook_ops nf_nat_ipv6_ops[] __read_mostly = {
 246         /* Before packet filtering, change destination */
 247         {
 248                 .hook           = nf_nat_ipv6_in,
 249                 .owner          = THIS_MODULE,
 250                 .pf             = NFPROTO_IPV6,
 251                 .hooknum        = NF_INET_PRE_ROUTING,
 252                 .priority       = NF_IP6_PRI_NAT_DST,
 253         },
 254         /* After packet filtering, change source */
 255         {
 256                 .hook           = nf_nat_ipv6_out,
 257                 .owner          = THIS_MODULE,
 258                 .pf             = NFPROTO_IPV6,
 259                 .hooknum        = NF_INET_POST_ROUTING,
 260                 .priority       = NF_IP6_PRI_NAT_SRC,
 261         },
 262         /* Before packet filtering, change destination */
 263         {
 264                 .hook           = nf_nat_ipv6_local_fn,
 265                 .owner          = THIS_MODULE,
 266                 .pf             = NFPROTO_IPV6,
 267                 .hooknum        = NF_INET_LOCAL_OUT,
 268                 .priority       = NF_IP6_PRI_NAT_DST,
 269         },
 270         /* After packet filtering, change source */
 271         {
 272                 .hook           = nf_nat_ipv6_fn,
 273                 .owner          = THIS_MODULE,
 274                 .pf             = NFPROTO_IPV6,
 275                 .hooknum        = NF_INET_LOCAL_IN,
 276                 .priority       = NF_IP6_PRI_NAT_SRC,
 277         },
 278 };
 279
 280 static int __net_init ip6table_nat_net_init(struct net *net)
 281 {
 282         struct ip6t_replace *repl;
 283
 284         repl = ip6t_alloc_initial_table(&nf_nat_ipv6_table);
 285         if (repl == NULL)
 286                 return -ENOMEM;
 287         net->ipv6.ip6table_nat = ip6t_register_table(net, &nf_nat_ipv6_table, repl);
 288         kfree(repl);
 289         return PTR_ERR_OR_ZERO(net->ipv6.ip6table_nat);
 290 }
 291
 292 static void __net_exit ip6table_nat_net_exit(struct net *net)
 293 {
 294         ip6t_unregister_table(net, net->ipv6.ip6table_nat);
 295 }
 296
 297 static struct pernet_operations ip6table_nat_net_ops = {
 298         .init   = ip6table_nat_net_init,
 299         .exit   = ip6table_nat_net_exit,
 300 };
 301
 302 static int __init ip6table_nat_init(void)
 303 {
 304         int err;
 305
 306         err = register_pernet_subsys(&ip6table_nat_net_ops);
 307         if (err < 0)
 308                 goto err1;
 309
 310         err = nf_register_hooks(nf_nat_ipv6_ops, ARRAY_SIZE(nf_nat_ipv6_ops));
 311         if (err < 0)
 312                 goto err2;
 313         return 0;
 314
 315 err2:
 316         unregister_pernet_subsys(&ip6table_nat_net_ops);
 317 err1:
 318         return err;
 319 }
 320
 321 static void __exit ip6table_nat_exit(void)
 322 {
 323         nf_unregister_hooks(nf_nat_ipv6_ops, ARRAY_SIZE(nf_nat_ipv6_ops));
 324         unregister_pernet_subsys(&ip6table_nat_net_ops);
 325 }
 326
 327 module_init(ip6table_nat_init);
 328 module_exit(ip6table_nat_exit);
 329
 330 MODULE_LICENSE("GPL");