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