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         struct nf_conn *ct;
  34         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                         fl->fl_ip_dport = t->dst.u.tcp.port;
  55         }
  56
  57         statusbit ^= IPS_NAT_MASK;
  58
  59         if (ct->status & statusbit) {
  60                 fl->fl4_src = t->src.u3.ip;
  61                 if (t->dst.protonum == IPPROTO_TCP ||
  62                     t->dst.protonum == IPPROTO_UDP)
  63                         fl->fl_ip_sport = t->src.u.tcp.port;
  64         }
  65 }
  66 #endif
  67
  68 static unsigned int
  69 nf_nat_fn(unsigned int hooknum,
  70           struct sk_buff **pskb,
  71           const struct net_device *in,
  72           const struct net_device *out,
  73           int (*okfn)(struct sk_buff *))
  74 {
  75         struct nf_conn *ct;
  76         enum ip_conntrack_info ctinfo;
  77         struct nf_conn_nat *nat;
  78         /* maniptype == SRC for postrouting. */
  79         enum nf_nat_manip_type maniptype = HOOK2MANIP(hooknum);
  80
  81         /* We never see fragments: conntrack defrags on pre-routing
  82            and local-out, and nf_nat_out protects post-routing. */
  83         NF_CT_ASSERT(!(ip_hdr(*pskb)->frag_off & htons(IP_MF | IP_OFFSET)));
  84
  85         ct = nf_ct_get(*pskb, &ctinfo);
  86         /* Can't track?  It's not due to stress, or conntrack would
  87            have dropped it.  Hence it's the user's responsibilty to
  88            packet filter it out, or implement conntrack/NAT for that
  89            protocol. 8) --RR */
  90         if (!ct) {
  91                 /* Exception: ICMP redirect to new connection (not in
  92                    hash table yet).  We must not let this through, in
  93                    case we're doing NAT to the same network. */
  94                 if (ip_hdr(*pskb)->protocol == IPPROTO_ICMP) {
  95                         struct icmphdr _hdr, *hp;
  96
  97                         hp = skb_header_pointer(*pskb, ip_hdrlen(*pskb),
  98                                                 sizeof(_hdr), &_hdr);
  99                         if (hp != NULL &&
 100                             hp->type == ICMP_REDIRECT)
 101                                 return NF_DROP;
 102                 }
 103                 return NF_ACCEPT;
 104         }
 105
 106         /* Don't try to NAT if this packet is not conntracked */
 107         if (ct == &nf_conntrack_untracked)
 108                 return NF_ACCEPT;
 109
 110         nat = nfct_nat(ct);
 111         if (!nat) {
 112                 nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC);
 113                 if (nat == NULL) {
 114                         pr_debug("failed to add NAT extension\n");
 115                         return NF_ACCEPT;
 116                 }
 117         }
 118
 119         switch (ctinfo) {
 120         case IP_CT_RELATED:
 121         case IP_CT_RELATED+IP_CT_IS_REPLY:
 122                 if (ip_hdr(*pskb)->protocol == IPPROTO_ICMP) {
 123                         if (!nf_nat_icmp_reply_translation(ct, ctinfo,
 124                                                            hooknum, pskb))
 125                                 return NF_DROP;
 126                         else
 127                                 return NF_ACCEPT;
 128                 }
 129                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 130         case IP_CT_NEW:
 131
 132                 /* Seen it before?  This can happen for loopback, retrans,
 133                    or local packets.. */
 134                 if (!nf_nat_initialized(ct, maniptype)) {
 135                         unsigned int ret;
 136
 137                         if (unlikely(nf_ct_is_confirmed(ct)))
 138                                 /* NAT module was loaded late */
 139                                 ret = alloc_null_binding_confirmed(ct, hooknum);
 140                         else if (hooknum == NF_IP_LOCAL_IN)
 141                                 /* LOCAL_IN hook doesn't have a chain!  */
 142                                 ret = alloc_null_binding(ct, hooknum);
 143                         else
 144                                 ret = nf_nat_rule_find(pskb, hooknum, in, out,
 145                                                        ct);
 146
 147                         if (ret != NF_ACCEPT) {
 148                                 return ret;
 149                         }
 150                 } else
 151                         pr_debug("Already setup manip %s for ct %p\n",
 152                                  maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 153                                  ct);
 154                 break;
 155
 156         default:
 157                 /* ESTABLISHED */
 158                 NF_CT_ASSERT(ctinfo == IP_CT_ESTABLISHED ||
 159                              ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 160         }
 161
 162         return nf_nat_packet(ct, ctinfo, hooknum, pskb);
 163 }
 164
 165 static unsigned int
 166 nf_nat_in(unsigned int hooknum,
 167           struct sk_buff **pskb,
 168           const struct net_device *in,
 169           const struct net_device *out,
 170           int (*okfn)(struct sk_buff *))
 171 {
 172         unsigned int ret;
 173         __be32 daddr = ip_hdr(*pskb)->daddr;
 174
 175         ret = nf_nat_fn(hooknum, pskb, in, out, okfn);
 176         if (ret != NF_DROP && ret != NF_STOLEN &&
 177             daddr != ip_hdr(*pskb)->daddr) {
 178                 dst_release((*pskb)->dst);
 179                 (*pskb)->dst = NULL;
 180         }
 181         return ret;
 182 }
 183
 184 static unsigned int
 185 nf_nat_out(unsigned int hooknum,
 186            struct sk_buff **pskb,
 187            const struct net_device *in,
 188            const struct net_device *out,
 189            int (*okfn)(struct sk_buff *))
 190 {
 191 #ifdef CONFIG_XFRM
 192         struct nf_conn *ct;
 193         enum ip_conntrack_info ctinfo;
 194 #endif
 195         unsigned int ret;
 196
 197         /* root is playing with raw sockets. */
 198         if ((*pskb)->len < sizeof(struct iphdr) ||
 199             ip_hdrlen(*pskb) < sizeof(struct iphdr))
 200                 return NF_ACCEPT;
 201
 202         ret = nf_nat_fn(hooknum, pskb, in, out, okfn);
 203 #ifdef CONFIG_XFRM
 204         if (ret != NF_DROP && ret != NF_STOLEN &&
 205             (ct = nf_ct_get(*pskb, &ctinfo)) != NULL) {
 206                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 207
 208                 if (ct->tuplehash[dir].tuple.src.u3.ip !=
 209                     ct->tuplehash[!dir].tuple.dst.u3.ip
 210                     || ct->tuplehash[dir].tuple.src.u.all !=
 211                        ct->tuplehash[!dir].tuple.dst.u.all
 212                     )
 213                         return ip_xfrm_me_harder(pskb) == 0 ? ret : NF_DROP;
 214         }
 215 #endif
 216         return ret;
 217 }
 218
 219 static unsigned int
 220 nf_nat_local_fn(unsigned int hooknum,
 221                 struct sk_buff **pskb,
 222                 const struct net_device *in,
 223                 const struct net_device *out,
 224                 int (*okfn)(struct sk_buff *))
 225 {
 226         struct nf_conn *ct;
 227         enum ip_conntrack_info ctinfo;
 228         unsigned int ret;
 229
 230         /* root is playing with raw sockets. */
 231         if ((*pskb)->len < sizeof(struct iphdr) ||
 232             ip_hdrlen(*pskb) < sizeof(struct iphdr))
 233                 return NF_ACCEPT;
 234
 235         ret = nf_nat_fn(hooknum, pskb, in, out, okfn);
 236         if (ret != NF_DROP && ret != NF_STOLEN &&
 237             (ct = nf_ct_get(*pskb, &ctinfo)) != NULL) {
 238                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 239
 240                 if (ct->tuplehash[dir].tuple.dst.u3.ip !=
 241                     ct->tuplehash[!dir].tuple.src.u3.ip) {
 242                         if (ip_route_me_harder(pskb, RTN_UNSPEC))
 243                                 ret = NF_DROP;
 244                 }
 245 #ifdef CONFIG_XFRM
 246                 else if (ct->tuplehash[dir].tuple.dst.u.all !=
 247                          ct->tuplehash[!dir].tuple.src.u.all)
 248                         if (ip_xfrm_me_harder(pskb))
 249                                 ret = NF_DROP;
 250 #endif
 251         }
 252         return ret;
 253 }
 254
 255 static unsigned int
 256 nf_nat_adjust(unsigned int hooknum,
 257               struct sk_buff **pskb,
 258               const struct net_device *in,
 259               const struct net_device *out,
 260               int (*okfn)(struct sk_buff *))
 261 {
 262         struct nf_conn *ct;
 263         enum ip_conntrack_info ctinfo;
 264
 265         ct = nf_ct_get(*pskb, &ctinfo);
 266         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 267                 pr_debug("nf_nat_standalone: adjusting sequence number\n");
 268                 if (!nf_nat_seq_adjust(pskb, ct, ctinfo))
 269                         return NF_DROP;
 270         }
 271         return NF_ACCEPT;
 272 }
 273
 274 /* We must be after connection tracking and before packet filtering. */
 275
 276 static struct nf_hook_ops nf_nat_ops[] = {
 277         /* Before packet filtering, change destination */
 278         {
 279                 .hook           = nf_nat_in,
 280                 .owner          = THIS_MODULE,
 281                 .pf             = PF_INET,
 282                 .hooknum        = NF_IP_PRE_ROUTING,
 283                 .priority       = NF_IP_PRI_NAT_DST,
 284         },
 285         /* After packet filtering, change source */
 286         {
 287                 .hook           = nf_nat_out,
 288                 .owner          = THIS_MODULE,
 289                 .pf             = PF_INET,
 290                 .hooknum        = NF_IP_POST_ROUTING,
 291                 .priority       = NF_IP_PRI_NAT_SRC,
 292         },
 293         /* After conntrack, adjust sequence number */
 294         {
 295                 .hook           = nf_nat_adjust,
 296                 .owner          = THIS_MODULE,
 297                 .pf             = PF_INET,
 298                 .hooknum        = NF_IP_POST_ROUTING,
 299                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 300         },
 301         /* Before packet filtering, change destination */
 302         {
 303                 .hook           = nf_nat_local_fn,
 304                 .owner          = THIS_MODULE,
 305                 .pf             = PF_INET,
 306                 .hooknum        = NF_IP_LOCAL_OUT,
 307                 .priority       = NF_IP_PRI_NAT_DST,
 308         },
 309         /* After packet filtering, change source */
 310         {
 311                 .hook           = nf_nat_fn,
 312                 .owner          = THIS_MODULE,
 313                 .pf             = PF_INET,
 314                 .hooknum        = NF_IP_LOCAL_IN,
 315                 .priority       = NF_IP_PRI_NAT_SRC,
 316         },
 317         /* After conntrack, adjust sequence number */
 318         {
 319                 .hook           = nf_nat_adjust,
 320                 .owner          = THIS_MODULE,
 321                 .pf             = PF_INET,
 322                 .hooknum        = NF_IP_LOCAL_IN,
 323                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 324         },
 325 };
 326
 327 static int __init nf_nat_standalone_init(void)
 328 {
 329         int ret = 0;
 330
 331         need_ipv4_conntrack();
 332
 333 #ifdef CONFIG_XFRM
 334         BUG_ON(ip_nat_decode_session != NULL);
 335         ip_nat_decode_session = nat_decode_session;
 336 #endif
 337         ret = nf_nat_rule_init();
 338         if (ret < 0) {
 339                 printk("nf_nat_init: can't setup rules.\n");
 340                 goto cleanup_decode_session;
 341         }
 342         ret = nf_register_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 343         if (ret < 0) {
 344                 printk("nf_nat_init: can't register hooks.\n");
 345                 goto cleanup_rule_init;
 346         }
 347         return ret;
 348
 349  cleanup_rule_init:
 350         nf_nat_rule_cleanup();
 351  cleanup_decode_session:
 352 #ifdef CONFIG_XFRM
 353         ip_nat_decode_session = NULL;
 354         synchronize_net();
 355 #endif
 356         return ret;
 357 }
 358
 359 static void __exit nf_nat_standalone_fini(void)
 360 {
 361         nf_unregister_hooks(nf_nat_ops, ARRAY_SIZE(nf_nat_ops));
 362         nf_nat_rule_cleanup();
 363 #ifdef CONFIG_XFRM
 364         ip_nat_decode_session = NULL;
 365         synchronize_net();
 366 #endif
 367         /* Conntrack caches are unregistered in nf_conntrack_cleanup */
 368 }
 369
 370 module_init(nf_nat_standalone_init);
 371 module_exit(nf_nat_standalone_fini);
 372
 373 MODULE_LICENSE("GPL");
 374 MODULE_ALIAS("ip_nat");