kernel/linux/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;
 103
 104                         if (skb_copy_bits(*pskb, (*pskb)->nh.iph->ihl*4,
 105                                           &hdr, sizeof(hdr)) == 0
 106                             && hdr.type == ICMP_REDIRECT)
 107                                 return NF_DROP;
 108                 }
 109                 return NF_ACCEPT;
 110         }
 111
 112         switch (ctinfo) {
 113         case IP_CT_RELATED:
 114         case IP_CT_RELATED+IP_CT_IS_REPLY:
 115                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 116                         if (!icmp_reply_translation(pskb, ct, hooknum,
 117                                                     CTINFO2DIR(ctinfo)))
 118                                 return NF_DROP;
 119                         else
 120                                 return NF_ACCEPT;
 121                 }
 122                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 123         case IP_CT_NEW:
 124                 info = &ct->nat.info;
 125
 126                 WRITE_LOCK(&ip_nat_lock);
 127                 /* Seen it before?  This can happen for loopback, retrans,
 128                    or local packets.. */
 129                 if (!(info->initialized & (1 << maniptype))
 130 #ifndef CONFIG_IP_NF_NAT_LOCAL
 131                     /* If this session has already been confirmed we must not
 132                      * touch it again even if there is no mapping set up.
 133                      * Can only happen on local->local traffic with
 134                      * CONFIG_IP_NF_NAT_LOCAL disabled.
 135                      */
 136                     && !(ct->status & IPS_CONFIRMED)
 137 #endif
 138                     ) {
 139                         unsigned int ret;
 140
 141                         if (ct->master
 142                             && master_ct(ct)->nat.info.helper
 143                             && master_ct(ct)->nat.info.helper->expect) {
 144                                 ret = call_expect(master_ct(ct), pskb,
 145                                                   hooknum, ct, info);
 146                         } else {
 147 #ifdef CONFIG_IP_NF_NAT_LOCAL
 148                                 /* LOCAL_IN hook doesn't have a chain!  */
 149                                 if (hooknum == NF_IP_LOCAL_IN)
 150                                         ret = alloc_null_binding(ct, info,
 151                                                                  hooknum);
 152                                 else
 153 #endif
 154                                 ret = ip_nat_rule_find(pskb, hooknum, in, out,
 155                                                        ct, info);
 156                         }
 157
 158                         if (ret != NF_ACCEPT) {
 159                                 WRITE_UNLOCK(&ip_nat_lock);
 160                                 return ret;
 161                         }
 162                 } else
 163                         DEBUGP("Already setup manip %s for ct %p\n",
 164                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 165                                ct);
 166                 WRITE_UNLOCK(&ip_nat_lock);
 167                 break;
 168
 169         default:
 170                 /* ESTABLISHED */
 171                 IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
 172                              || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 173                 info = &ct->nat.info;
 174         }
 175
 176         IP_NF_ASSERT(info);
 177         return do_bindings(ct, ctinfo, info, hooknum, pskb);
 178 }
 179
 180 static unsigned int
 181 ip_nat_out(unsigned int hooknum,
 182            struct sk_buff **pskb,
 183            const struct net_device *in,
 184            const struct net_device *out,
 185            int (*okfn)(struct sk_buff *))
 186 {
 187         /* root is playing with raw sockets. */
 188         if ((*pskb)->len < sizeof(struct iphdr)
 189             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 190                 return NF_ACCEPT;
 191
 192         /* We can hit fragment here; forwarded packets get
 193            defragmented by connection tracking coming in, then
 194            fragmented (grr) by the forward code.
 195
 196            In future: If we have nfct != NULL, AND we have NAT
 197            initialized, AND there is no helper, then we can do full
 198            NAPT on the head, and IP-address-only NAT on the rest.
 199
 200            I'm starting to have nightmares about fragments.  */
 201
 202         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 203                 *pskb = ip_ct_gather_frags(*pskb);
 204
 205                 if (!*pskb)
 206                         return NF_STOLEN;
 207         }
 208
 209         return ip_nat_fn(hooknum, pskb, in, out, okfn);
 210 }
 211
 212 #ifdef CONFIG_IP_NF_NAT_LOCAL
 213 static unsigned int
 214 ip_nat_local_fn(unsigned int hooknum,
 215                 struct sk_buff **pskb,
 216                 const struct net_device *in,
 217                 const struct net_device *out,
 218                 int (*okfn)(struct sk_buff *))
 219 {
 220         u_int32_t saddr, daddr;
 221         unsigned int ret;
 222
 223         /* root is playing with raw sockets. */
 224         if ((*pskb)->len < sizeof(struct iphdr)
 225             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 226                 return NF_ACCEPT;
 227
 228         saddr = (*pskb)->nh.iph->saddr;
 229         daddr = (*pskb)->nh.iph->daddr;
 230
 231         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 232         if (ret != NF_DROP && ret != NF_STOLEN
 233             && ((*pskb)->nh.iph->saddr != saddr
 234                 || (*pskb)->nh.iph->daddr != daddr))
 235                 return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
 236         return ret;
 237 }
 238 #endif
 239
 240 /* We must be after connection tracking and before packet filtering. */
 241
 242 /* Before packet filtering, change destination */
 243 static struct nf_hook_ops ip_nat_in_ops = {
 244         .hook           = ip_nat_fn,
 245         .owner          = THIS_MODULE,
 246         .pf             = PF_INET,
 247         .hooknum        = NF_IP_PRE_ROUTING,
 248         .priority       = NF_IP_PRI_NAT_DST,
 249 };
 250
 251 /* After packet filtering, change source */
 252 static struct nf_hook_ops ip_nat_out_ops = {
 253         .hook           = ip_nat_out,
 254         .owner          = THIS_MODULE,
 255         .pf             = PF_INET,
 256         .hooknum        = NF_IP_POST_ROUTING,
 257         .priority       = NF_IP_PRI_NAT_SRC,
 258 };
 259
 260 #ifdef CONFIG_IP_NF_NAT_LOCAL
 261 /* Before packet filtering, change destination */
 262 static struct nf_hook_ops ip_nat_local_out_ops = {
 263         .hook           = ip_nat_local_fn,
 264         .owner          = THIS_MODULE,
 265         .pf             = PF_INET,
 266         .hooknum        = NF_IP_LOCAL_OUT,
 267         .priority       = NF_IP_PRI_NAT_DST,
 268 };
 269
 270 /* After packet filtering, change source for reply packets of LOCAL_OUT DNAT */
 271 static struct nf_hook_ops ip_nat_local_in_ops = {
 272         .hook           = ip_nat_fn,
 273         .owner          = THIS_MODULE,
 274         .pf             = PF_INET,
 275         .hooknum        = NF_IP_LOCAL_IN,
 276         .priority       = NF_IP_PRI_NAT_SRC,
 277 };
 278 #endif
 279
 280 /* Protocol registration. */
 281 int ip_nat_protocol_register(struct ip_nat_protocol *proto)
 282 {
 283         int ret = 0;
 284         struct list_head *i;
 285
 286         WRITE_LOCK(&ip_nat_lock);
 287         list_for_each(i, &protos) {
 288                 if (((struct ip_nat_protocol *)i)->protonum
 289                     == proto->protonum) {
 290                         ret = -EBUSY;
 291                         goto out;
 292                 }
 293         }
 294
 295         list_prepend(&protos, proto);
 296  out:
 297         WRITE_UNLOCK(&ip_nat_lock);
 298         return ret;
 299 }
 300
 301 /* Noone stores the protocol anywhere; simply delete it. */
 302 void ip_nat_protocol_unregister(struct ip_nat_protocol *proto)
 303 {
 304         WRITE_LOCK(&ip_nat_lock);
 305         LIST_DELETE(&protos, proto);
 306         WRITE_UNLOCK(&ip_nat_lock);
 307
 308         /* Someone could be still looking at the proto in a bh. */
 309         synchronize_net();
 310 }
 311
 312 static int init_or_cleanup(int init)
 313 {
 314         int ret = 0;
 315
 316         need_ip_conntrack();
 317
 318         if (!init) goto cleanup;
 319
 320         ret = ip_nat_rule_init();
 321         if (ret < 0) {
 322                 printk("ip_nat_init: can't setup rules.\n");
 323                 goto cleanup_nothing;
 324         }
 325         ret = ip_nat_init();
 326         if (ret < 0) {
 327                 printk("ip_nat_init: can't setup rules.\n");
 328                 goto cleanup_rule_init;
 329         }
 330         ret = nf_register_hook(&ip_nat_in_ops);
 331         if (ret < 0) {
 332                 printk("ip_nat_init: can't register in hook.\n");
 333                 goto cleanup_nat;
 334         }
 335         ret = nf_register_hook(&ip_nat_out_ops);
 336         if (ret < 0) {
 337                 printk("ip_nat_init: can't register out hook.\n");
 338                 goto cleanup_inops;
 339         }
 340 #ifdef CONFIG_IP_NF_NAT_LOCAL
 341         ret = nf_register_hook(&ip_nat_local_out_ops);
 342         if (ret < 0) {
 343                 printk("ip_nat_init: can't register local out hook.\n");
 344                 goto cleanup_outops;
 345         }
 346         ret = nf_register_hook(&ip_nat_local_in_ops);
 347         if (ret < 0) {
 348                 printk("ip_nat_init: can't register local in hook.\n");
 349                 goto cleanup_localoutops;
 350         }
 351 #endif
 352         return ret;
 353
 354  cleanup:
 355 #ifdef CONFIG_IP_NF_NAT_LOCAL
 356         nf_unregister_hook(&ip_nat_local_in_ops);
 357  cleanup_localoutops:
 358         nf_unregister_hook(&ip_nat_local_out_ops);
 359  cleanup_outops:
 360 #endif
 361         nf_unregister_hook(&ip_nat_out_ops);
 362  cleanup_inops:
 363         nf_unregister_hook(&ip_nat_in_ops);
 364  cleanup_nat:
 365         ip_nat_cleanup();
 366  cleanup_rule_init:
 367         ip_nat_rule_cleanup();
 368  cleanup_nothing:
 369         MUST_BE_READ_WRITE_UNLOCKED(&ip_nat_lock);
 370         return ret;
 371 }
 372
 373 static int __init init(void)
 374 {
 375         return init_or_cleanup(1);
 376 }
 377
 378 static void __exit fini(void)
 379 {
 380         init_or_cleanup(0);
 381 }
 382
 383 module_init(init);
 384 module_exit(fini);
 385
 386 EXPORT_SYMBOL(ip_nat_setup_info);
 387 EXPORT_SYMBOL(ip_nat_protocol_register);
 388 EXPORT_SYMBOL(ip_nat_protocol_unregister);
 389 EXPORT_SYMBOL(ip_nat_helper_register);
 390 EXPORT_SYMBOL(ip_nat_helper_unregister);
 391 EXPORT_SYMBOL(ip_nat_cheat_check);
 392 EXPORT_SYMBOL(ip_nat_mangle_tcp_packet);
 393 EXPORT_SYMBOL(ip_nat_mangle_udp_packet);
 394 EXPORT_SYMBOL(ip_nat_used_tuple);
 395 MODULE_LICENSE("GPL");