net/ipv4/netfilter/ip_nat_standalone.c

   1 /* This file contains all the functions required for the standalone
   2    ip_nat module.
   3
   4    These are not required by the compatibility layer.
   5 */
   6
   7 /* (C) 1999-2001 Paul `Rusty' Russell
   8  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
   9  *
  10  * This program is free software; you can redistribute it and/or modify
  11  * it under the terms of the GNU General Public License version 2 as
  12  * published by the Free Software Foundation.
  13  */
  14
  15 /*
  16  * 23 Apr 2001: Harald Welte <laforge@gnumonks.org>
  17  *      - new API and handling of conntrack/nat helpers
  18  *      - now capable of multiple expectations for one master
  19  * */
  20
  21 #include <linux/config.h>
  22 #include <linux/types.h>
  23 #include <linux/icmp.h>
  24 #include <linux/ip.h>
  25 #include <linux/netfilter.h>
  26 #include <linux/netfilter_ipv4.h>
  27 #include <linux/module.h>
  28 #include <linux/skbuff.h>
  29 #include <linux/proc_fs.h>
  30 #include <net/ip.h>
  31 #include <net/checksum.h>
  32 #include <linux/spinlock.h>
  33
  34 #define ASSERT_READ_LOCK(x) MUST_BE_READ_LOCKED(&ip_nat_lock)
  35 #define ASSERT_WRITE_LOCK(x) MUST_BE_WRITE_LOCKED(&ip_nat_lock)
  36
  37 #include <linux/netfilter_ipv4/ip_nat.h>
  38 #include <linux/netfilter_ipv4/ip_nat_rule.h>
  39 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  40 #include <linux/netfilter_ipv4/ip_nat_core.h>
  41 #include <linux/netfilter_ipv4/ip_nat_helper.h>
  42 #include <linux/netfilter_ipv4/ip_tables.h>
  43 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  44 #include <linux/netfilter_ipv4/listhelp.h>
  45
  46 #if 0
  47 #define DEBUGP printk
  48 #else
  49 #define DEBUGP(format, args...)
  50 #endif
  51
  52 #define HOOKNAME(hooknum) ((hooknum) == NF_IP_POST_ROUTING ? "POST_ROUTING"  \
  53                            : ((hooknum) == NF_IP_PRE_ROUTING ? "PRE_ROUTING" \
  54                               : ((hooknum) == NF_IP_LOCAL_OUT ? "LOCAL_OUT"  \
  55                                  : ((hooknum) == NF_IP_LOCAL_IN ? "LOCAL_IN"  \
  56                                     : "*ERROR*")))
  57
  58 static inline int call_expect(struct ip_conntrack *master,
  59                               struct sk_buff **pskb,
  60                               unsigned int hooknum,
  61                               struct ip_conntrack *ct,
  62                               struct ip_nat_info *info)
  63 {
  64         return master->nat.info.helper->expect(pskb, hooknum, ct, info);
  65 }
  66
  67 static unsigned int
  68 ip_nat_fn(unsigned int hooknum,
  69           struct sk_buff **pskb,
  70           const struct net_device *in,
  71           const struct net_device *out,
  72           int (*okfn)(struct sk_buff *))
  73 {
  74         struct ip_conntrack *ct;
  75         enum ip_conntrack_info ctinfo;
  76         struct ip_nat_info *info;
  77         /* maniptype == SRC for postrouting. */
  78         enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
  79
  80         /* We never see fragments: conntrack defrags on pre-routing
  81            and local-out, and ip_nat_out protects post-routing. */
  82         IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
  83                        & htons(IP_MF|IP_OFFSET)));
  84
  85         (*pskb)->nfcache |= NFC_UNKNOWN;
  86
  87         /* If we had a hardware checksum before, it's now invalid */
  88         if ((*pskb)->ip_summed == CHECKSUM_HW)
  89                 if (skb_checksum_help(pskb, (out == NULL)))
  90                         return NF_DROP;
  91
  92         ct = ip_conntrack_get(*pskb, &ctinfo);
  93         /* Can't track?  It's not due to stress, or conntrack would
  94            have dropped it.  Hence it's the user's responsibilty to
  95            packet filter it out, or implement conntrack/NAT for that
  96            protocol. 8) --RR */
  97         if (!ct) {
  98                 /* Exception: ICMP redirect to new connection (not in
  99                    hash table yet).  We must not let this through, in
 100                    case we're doing NAT to the same network. */
 101                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 102                         struct icmphdr _hdr, *hp;
 103
 104                         hp = skb_header_pointer(*pskb,
 105                                                 (*pskb)->nh.iph->ihl*4,
 106                                                 sizeof(_hdr), &_hdr);
 107                         if (hp != NULL &&
 108                             hp->type == ICMP_REDIRECT)
 109                                 return NF_DROP;
 110                 }
 111                 return NF_ACCEPT;
 112         }
 113
 114         switch (ctinfo) {
 115         case IP_CT_RELATED:
 116         case IP_CT_RELATED+IP_CT_IS_REPLY:
 117                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 118                         if (!icmp_reply_translation(pskb, ct, hooknum,
 119                                                     CTINFO2DIR(ctinfo)))
 120                                 return NF_DROP;
 121                         else
 122                                 return NF_ACCEPT;
 123                 }
 124                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 125         case IP_CT_NEW:
 126                 info = &ct->nat.info;
 127
 128                 WRITE_LOCK(&ip_nat_lock);
 129                 /* Seen it before?  This can happen for loopback, retrans,
 130                    or local packets.. */
 131                 if (!(info->initialized & (1 << maniptype))
 132 #ifndef CONFIG_IP_NF_NAT_LOCAL
 133                     /* If this session has already been confirmed we must not
 134                      * touch it again even if there is no mapping set up.
 135                      * Can only happen on local->local traffic with
 136                      * CONFIG_IP_NF_NAT_LOCAL disabled.
 137                      */
 138                     && !(ct->status & IPS_CONFIRMED)
 139 #endif
 140                     ) {
 141                         unsigned int ret;
 142
 143                         if (ct->master
 144                             && master_ct(ct)->nat.info.helper
 145                             && master_ct(ct)->nat.info.helper->expect) {
 146                                 ret = call_expect(master_ct(ct), pskb,
 147                                                   hooknum, ct, info);
 148                         } else {
 149 #ifdef CONFIG_IP_NF_NAT_LOCAL
 150                                 /* LOCAL_IN hook doesn't have a chain!  */
 151                                 if (hooknum == NF_IP_LOCAL_IN)
 152                                         ret = alloc_null_binding(ct, info,
 153                                                                  hooknum);
 154                                 else
 155 #endif
 156                                 ret = ip_nat_rule_find(pskb, hooknum, in, out,
 157                                                        ct, info);
 158                         }
 159
 160                         if (ret != NF_ACCEPT) {
 161                                 WRITE_UNLOCK(&ip_nat_lock);
 162                                 return ret;
 163                         }
 164                 } else
 165                         DEBUGP("Already setup manip %s for ct %p\n",
 166                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 167                                ct);
 168                 WRITE_UNLOCK(&ip_nat_lock);
 169                 break;
 170
 171         default:
 172                 /* ESTABLISHED */
 173                 IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
 174                              || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 175                 info = &ct->nat.info;
 176         }
 177
 178         IP_NF_ASSERT(info);
 179         return do_bindings(ct, ctinfo, info, hooknum, pskb);
 180 }
 181
 182 static unsigned int
 183 ip_nat_out(unsigned int hooknum,
 184            struct sk_buff **pskb,
 185            const struct net_device *in,
 186            const struct net_device *out,
 187            int (*okfn)(struct sk_buff *))
 188 {
 189         /* root is playing with raw sockets. */
 190         if ((*pskb)->len < sizeof(struct iphdr)
 191             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 192                 return NF_ACCEPT;
 193
 194         /* We can hit fragment here; forwarded packets get
 195            defragmented by connection tracking coming in, then
 196            fragmented (grr) by the forward code.
 197
 198            In future: If we have nfct != NULL, AND we have NAT
 199            initialized, AND there is no helper, then we can do full
 200            NAPT on the head, and IP-address-only NAT on the rest.
 201
 202            I'm starting to have nightmares about fragments.  */
 203
 204         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 205                 *pskb = ip_ct_gather_frags(*pskb);
 206
 207                 if (!*pskb)
 208                         return NF_STOLEN;
 209         }
 210
 211         return ip_nat_fn(hooknum, pskb, in, out, okfn);
 212 }
 213
 214 #ifdef CONFIG_IP_NF_NAT_LOCAL
 215 static unsigned int
 216 ip_nat_local_fn(unsigned int hooknum,
 217                 struct sk_buff **pskb,
 218                 const struct net_device *in,
 219                 const struct net_device *out,
 220                 int (*okfn)(struct sk_buff *))
 221 {
 222         u_int32_t saddr, daddr;
 223         unsigned int ret;
 224
 225         /* root is playing with raw sockets. */
 226         if ((*pskb)->len < sizeof(struct iphdr)
 227             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 228                 return NF_ACCEPT;
 229
 230         saddr = (*pskb)->nh.iph->saddr;
 231         daddr = (*pskb)->nh.iph->daddr;
 232
 233         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 234         if (ret != NF_DROP && ret != NF_STOLEN
 235             && ((*pskb)->nh.iph->saddr != saddr
 236                 || (*pskb)->nh.iph->daddr != daddr))
 237                 return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
 238         return ret;
 239 }
 240 #endif
 241
 242 /* We must be after connection tracking and before packet filtering. */
 243
 244 /* Before packet filtering, change destination */
 245 static struct nf_hook_ops ip_nat_in_ops = {
 246         .hook           = ip_nat_fn,
 247         .owner          = THIS_MODULE,
 248         .pf             = PF_INET,
 249         .hooknum        = NF_IP_PRE_ROUTING,
 250         .priority       = NF_IP_PRI_NAT_DST,
 251 };
 252
 253 /* After packet filtering, change source */
 254 static struct nf_hook_ops ip_nat_out_ops = {
 255         .hook           = ip_nat_out,
 256         .owner          = THIS_MODULE,
 257         .pf             = PF_INET,
 258         .hooknum        = NF_IP_POST_ROUTING,
 259         .priority       = NF_IP_PRI_NAT_SRC,
 260 };
 261
 262 #ifdef CONFIG_IP_NF_NAT_LOCAL
 263 /* Before packet filtering, change destination */
 264 static struct nf_hook_ops ip_nat_local_out_ops = {
 265         .hook           = ip_nat_local_fn,
 266         .owner          = THIS_MODULE,
 267         .pf             = PF_INET,
 268         .hooknum        = NF_IP_LOCAL_OUT,
 269         .priority       = NF_IP_PRI_NAT_DST,
 270 };
 271
 272 /* After packet filtering, change source for reply packets of LOCAL_OUT DNAT */
 273 static struct nf_hook_ops ip_nat_local_in_ops = {
 274         .hook           = ip_nat_fn,
 275         .owner          = THIS_MODULE,
 276         .pf             = PF_INET,
 277         .hooknum        = NF_IP_LOCAL_IN,
 278         .priority       = NF_IP_PRI_NAT_SRC,
 279 };
 280 #endif
 281
 282 /* Protocol registration. */
 283 int ip_nat_protocol_register(struct ip_nat_protocol *proto)
 284 {
 285         int ret = 0;
 286
 287         WRITE_LOCK(&ip_nat_lock);
 288         if (ip_nat_protos[proto->protonum] != &ip_nat_unknown_protocol) {
 289                 ret = -EBUSY;
 290                 goto out;
 291         }
 292         ip_nat_protos[proto->protonum] = proto;
 293  out:
 294         WRITE_UNLOCK(&ip_nat_lock);
 295         return ret;
 296 }
 297
 298 /* Noone stores the protocol anywhere; simply delete it. */
 299 void ip_nat_protocol_unregister(struct ip_nat_protocol *proto)
 300 {
 301         WRITE_LOCK(&ip_nat_lock);
 302         ip_nat_protos[proto->protonum] = &ip_nat_unknown_protocol;
 303         WRITE_UNLOCK(&ip_nat_lock);
 304
 305         /* Someone could be still looking at the proto in a bh. */
 306         synchronize_net();
 307 }
 308
 309 static int init_or_cleanup(int init)
 310 {
 311         int ret = 0;
 312
 313         need_ip_conntrack();
 314
 315         if (!init) goto cleanup;
 316
 317         ret = ip_nat_rule_init();
 318         if (ret < 0) {
 319                 printk("ip_nat_init: can't setup rules.\n");
 320                 goto cleanup_nothing;
 321         }
 322         ret = ip_nat_init();
 323         if (ret < 0) {
 324                 printk("ip_nat_init: can't setup rules.\n");
 325                 goto cleanup_rule_init;
 326         }
 327         ret = nf_register_hook(&ip_nat_in_ops);
 328         if (ret < 0) {
 329                 printk("ip_nat_init: can't register in hook.\n");
 330                 goto cleanup_nat;
 331         }
 332         ret = nf_register_hook(&ip_nat_out_ops);
 333         if (ret < 0) {
 334                 printk("ip_nat_init: can't register out hook.\n");
 335                 goto cleanup_inops;
 336         }
 337 #ifdef CONFIG_IP_NF_NAT_LOCAL
 338         ret = nf_register_hook(&ip_nat_local_out_ops);
 339         if (ret < 0) {
 340                 printk("ip_nat_init: can't register local out hook.\n");
 341                 goto cleanup_outops;
 342         }
 343         ret = nf_register_hook(&ip_nat_local_in_ops);
 344         if (ret < 0) {
 345                 printk("ip_nat_init: can't register local in hook.\n");
 346                 goto cleanup_localoutops;
 347         }
 348 #endif
 349         return ret;
 350
 351  cleanup:
 352 #ifdef CONFIG_IP_NF_NAT_LOCAL
 353         nf_unregister_hook(&ip_nat_local_in_ops);
 354  cleanup_localoutops:
 355         nf_unregister_hook(&ip_nat_local_out_ops);
 356  cleanup_outops:
 357 #endif
 358         nf_unregister_hook(&ip_nat_out_ops);
 359  cleanup_inops:
 360         nf_unregister_hook(&ip_nat_in_ops);
 361  cleanup_nat:
 362         ip_nat_cleanup();
 363  cleanup_rule_init:
 364         ip_nat_rule_cleanup();
 365  cleanup_nothing:
 366         MUST_BE_READ_WRITE_UNLOCKED(&ip_nat_lock);
 367         return ret;
 368 }
 369
 370 static int __init init(void)
 371 {
 372         return init_or_cleanup(1);
 373 }
 374
 375 static void __exit fini(void)
 376 {
 377         init_or_cleanup(0);
 378 }
 379
 380 module_init(init);
 381 module_exit(fini);
 382
 383 EXPORT_SYMBOL(ip_nat_setup_info);
 384 EXPORT_SYMBOL(ip_nat_protocol_register);
 385 EXPORT_SYMBOL(ip_nat_protocol_unregister);
 386 EXPORT_SYMBOL(ip_nat_helper_register);
 387 EXPORT_SYMBOL(ip_nat_helper_unregister);
 388 EXPORT_SYMBOL(ip_nat_cheat_check);
 389 EXPORT_SYMBOL(ip_nat_mangle_tcp_packet);
 390 EXPORT_SYMBOL(ip_nat_mangle_udp_packet);
 391 EXPORT_SYMBOL(ip_nat_used_tuple);
 392 EXPORT_SYMBOL(ip_nat_find_helper);
 393 EXPORT_SYMBOL(__ip_nat_find_helper);
 394 MODULE_LICENSE("GPL");