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