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)
  35 #define ASSERT_WRITE_LOCK(x)
  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 #ifdef CONFIG_XFRM
  59 static void nat_decode_session(struct sk_buff *skb, struct flowi *fl)
  60 {
  61         struct ip_conntrack *ct;
  62         struct ip_conntrack_tuple *t;
  63         enum ip_conntrack_info ctinfo;
  64         enum ip_conntrack_dir dir;
  65         unsigned long statusbit;
  66
  67         ct = ip_conntrack_get(skb, &ctinfo);
  68         if (ct == NULL)
  69                 return;
  70         dir = CTINFO2DIR(ctinfo);
  71         t = &ct->tuplehash[dir].tuple;
  72
  73         if (dir == IP_CT_DIR_ORIGINAL)
  74                 statusbit = IPS_DST_NAT;
  75         else
  76                 statusbit = IPS_SRC_NAT;
  77
  78         if (ct->status & statusbit) {
  79                 fl->fl4_dst = t->dst.ip;
  80                 if (t->dst.protonum == IPPROTO_TCP ||
  81                     t->dst.protonum == IPPROTO_UDP)
  82                         fl->fl_ip_dport = t->dst.u.tcp.port;
  83         }
  84
  85         statusbit ^= IPS_NAT_MASK;
  86
  87         if (ct->status & statusbit) {
  88                 fl->fl4_src = t->src.ip;
  89                 if (t->dst.protonum == IPPROTO_TCP ||
  90                     t->dst.protonum == IPPROTO_UDP)
  91                         fl->fl_ip_sport = t->src.u.tcp.port;
  92         }
  93 }
  94 #endif
  95
  96 static unsigned int
  97 ip_nat_fn(unsigned int hooknum,
  98           struct sk_buff **pskb,
  99           const struct net_device *in,
 100           const struct net_device *out,
 101           int (*okfn)(struct sk_buff *))
 102 {
 103         struct ip_conntrack *ct;
 104         enum ip_conntrack_info ctinfo;
 105         struct ip_nat_info *info;
 106         /* maniptype == SRC for postrouting. */
 107         enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
 108
 109         /* We never see fragments: conntrack defrags on pre-routing
 110            and local-out, and ip_nat_out protects post-routing. */
 111         IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
 112                        & htons(IP_MF|IP_OFFSET)));
 113
 114         /* If we had a hardware checksum before, it's now invalid */
 115         if ((*pskb)->ip_summed == CHECKSUM_HW)
 116                 if (skb_checksum_help(*pskb, (out == NULL)))
 117                         return NF_DROP;
 118
 119         ct = ip_conntrack_get(*pskb, &ctinfo);
 120         /* Can't track?  It's not due to stress, or conntrack would
 121            have dropped it.  Hence it's the user's responsibilty to
 122            packet filter it out, or implement conntrack/NAT for that
 123            protocol. 8) --RR */
 124         if (!ct) {
 125                 /* Exception: ICMP redirect to new connection (not in
 126                    hash table yet).  We must not let this through, in
 127                    case we're doing NAT to the same network. */
 128                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 129                         struct icmphdr _hdr, *hp;
 130
 131                         hp = skb_header_pointer(*pskb,
 132                                                 (*pskb)->nh.iph->ihl*4,
 133                                                 sizeof(_hdr), &_hdr);
 134                         if (hp != NULL &&
 135                             hp->type == ICMP_REDIRECT)
 136                                 return NF_DROP;
 137                 }
 138                 return NF_ACCEPT;
 139         }
 140
 141         /* Don't try to NAT if this packet is not conntracked */
 142         if (ct == &ip_conntrack_untracked)
 143                 return NF_ACCEPT;
 144
 145         switch (ctinfo) {
 146         case IP_CT_RELATED:
 147         case IP_CT_RELATED+IP_CT_IS_REPLY:
 148                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 149                         if (!ip_nat_icmp_reply_translation(pskb, ct, maniptype,
 150                                                            CTINFO2DIR(ctinfo)))
 151                                 return NF_DROP;
 152                         else
 153                                 return NF_ACCEPT;
 154                 }
 155                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 156         case IP_CT_NEW:
 157                 info = &ct->nat.info;
 158
 159                 /* Seen it before?  This can happen for loopback, retrans,
 160                    or local packets.. */
 161                 if (!ip_nat_initialized(ct, maniptype)) {
 162                         unsigned int ret;
 163
 164                         if (unlikely(is_confirmed(ct)))
 165                                 /* NAT module was loaded late */
 166                                 ret = alloc_null_binding_confirmed(ct, info,
 167                                                                    hooknum);
 168                         else if (hooknum == NF_IP_LOCAL_IN)
 169                                 /* LOCAL_IN hook doesn't have a chain!  */
 170                                 ret = alloc_null_binding(ct, info, hooknum);
 171                         else
 172                                 ret = ip_nat_rule_find(pskb, hooknum,
 173                                                        in, out, ct,
 174                                                        info);
 175
 176                         if (ret != NF_ACCEPT) {
 177                                 return ret;
 178                         }
 179                 } else
 180                         DEBUGP("Already setup manip %s for ct %p\n",
 181                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 182                                ct);
 183                 break;
 184
 185         default:
 186                 /* ESTABLISHED */
 187                 IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
 188                              || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 189                 info = &ct->nat.info;
 190         }
 191
 192         IP_NF_ASSERT(info);
 193         return ip_nat_packet(ct, ctinfo, hooknum, pskb);
 194 }
 195
 196 static unsigned int
 197 ip_nat_in(unsigned int hooknum,
 198           struct sk_buff **pskb,
 199           const struct net_device *in,
 200           const struct net_device *out,
 201           int (*okfn)(struct sk_buff *))
 202 {
 203         struct ip_conntrack *ct;
 204         enum ip_conntrack_info ctinfo;
 205         unsigned int ret;
 206
 207         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 208         if (ret != NF_DROP && ret != NF_STOLEN
 209             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 210                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 211
 212                 if (ct->tuplehash[dir].tuple.src.ip !=
 213                     ct->tuplehash[!dir].tuple.dst.ip) {
 214                         dst_release((*pskb)->dst);
 215                         (*pskb)->dst = NULL;
 216                 }
 217         }
 218         return ret;
 219 }
 220
 221 static unsigned int
 222 ip_nat_out(unsigned int hooknum,
 223            struct sk_buff **pskb,
 224            const struct net_device *in,
 225            const struct net_device *out,
 226            int (*okfn)(struct sk_buff *))
 227 {
 228         struct ip_conntrack *ct;
 229         enum ip_conntrack_info ctinfo;
 230         unsigned int ret;
 231
 232         /* root is playing with raw sockets. */
 233         if ((*pskb)->len < sizeof(struct iphdr)
 234             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 235                 return NF_ACCEPT;
 236
 237         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 238         if (ret != NF_DROP && ret != NF_STOLEN
 239             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 240                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 241
 242                 if (ct->tuplehash[dir].tuple.src.ip !=
 243                     ct->tuplehash[!dir].tuple.dst.ip
 244 #ifdef CONFIG_XFRM
 245                     || ct->tuplehash[dir].tuple.src.u.all !=
 246                        ct->tuplehash[!dir].tuple.dst.u.all
 247 #endif
 248                     )
 249                         return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
 250         }
 251         return ret;
 252 }
 253
 254 static unsigned int
 255 ip_nat_local_fn(unsigned int hooknum,
 256                 struct sk_buff **pskb,
 257                 const struct net_device *in,
 258                 const struct net_device *out,
 259                 int (*okfn)(struct sk_buff *))
 260 {
 261         struct ip_conntrack *ct;
 262         enum ip_conntrack_info ctinfo;
 263         unsigned int ret;
 264
 265         /* root is playing with raw sockets. */
 266         if ((*pskb)->len < sizeof(struct iphdr)
 267             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 268                 return NF_ACCEPT;
 269
 270         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 271         if (ret != NF_DROP && ret != NF_STOLEN
 272             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 273                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 274
 275                 if (ct->tuplehash[dir].tuple.dst.ip !=
 276                     ct->tuplehash[!dir].tuple.src.ip
 277 #ifdef CONFIG_XFRM
 278                     || ct->tuplehash[dir].tuple.dst.u.all !=
 279                        ct->tuplehash[dir].tuple.src.u.all
 280 #endif
 281                     )
 282                         return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
 283         }
 284         return ret;
 285 }
 286
 287 static unsigned int
 288 ip_nat_adjust(unsigned int hooknum,
 289               struct sk_buff **pskb,
 290               const struct net_device *in,
 291               const struct net_device *out,
 292               int (*okfn)(struct sk_buff *))
 293 {
 294         struct ip_conntrack *ct;
 295         enum ip_conntrack_info ctinfo;
 296
 297         ct = ip_conntrack_get(*pskb, &ctinfo);
 298         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 299                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
 300                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
 301                         return NF_DROP;
 302         }
 303         return NF_ACCEPT;
 304 }
 305
 306 /* We must be after connection tracking and before packet filtering. */
 307
 308 /* Before packet filtering, change destination */
 309 static struct nf_hook_ops ip_nat_in_ops = {
 310         .hook           = ip_nat_in,
 311         .owner          = THIS_MODULE,
 312         .pf             = PF_INET,
 313         .hooknum        = NF_IP_PRE_ROUTING,
 314         .priority       = NF_IP_PRI_NAT_DST,
 315 };
 316
 317 /* After packet filtering, change source */
 318 static struct nf_hook_ops ip_nat_out_ops = {
 319         .hook           = ip_nat_out,
 320         .owner          = THIS_MODULE,
 321         .pf             = PF_INET,
 322         .hooknum        = NF_IP_POST_ROUTING,
 323         .priority       = NF_IP_PRI_NAT_SRC,
 324 };
 325
 326 /* After conntrack, adjust sequence number */
 327 static struct nf_hook_ops ip_nat_adjust_out_ops = {
 328         .hook           = ip_nat_adjust,
 329         .owner          = THIS_MODULE,
 330         .pf             = PF_INET,
 331         .hooknum        = NF_IP_POST_ROUTING,
 332         .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 333 };
 334
 335 /* Before packet filtering, change destination */
 336 static struct nf_hook_ops ip_nat_local_out_ops = {
 337         .hook           = ip_nat_local_fn,
 338         .owner          = THIS_MODULE,
 339         .pf             = PF_INET,
 340         .hooknum        = NF_IP_LOCAL_OUT,
 341         .priority       = NF_IP_PRI_NAT_DST,
 342 };
 343
 344 /* After packet filtering, change source for reply packets of LOCAL_OUT DNAT */
 345 static struct nf_hook_ops ip_nat_local_in_ops = {
 346         .hook           = ip_nat_fn,
 347         .owner          = THIS_MODULE,
 348         .pf             = PF_INET,
 349         .hooknum        = NF_IP_LOCAL_IN,
 350         .priority       = NF_IP_PRI_NAT_SRC,
 351 };
 352
 353 /* After conntrack, adjust sequence number */
 354 static struct nf_hook_ops ip_nat_adjust_in_ops = {
 355         .hook           = ip_nat_adjust,
 356         .owner          = THIS_MODULE,
 357         .pf             = PF_INET,
 358         .hooknum        = NF_IP_LOCAL_IN,
 359         .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 360 };
 361
 362
 363 static int init_or_cleanup(int init)
 364 {
 365         int ret = 0;
 366
 367         need_conntrack();
 368
 369         if (!init) goto cleanup;
 370
 371 #ifdef CONFIG_XFRM
 372         BUG_ON(ip_nat_decode_session != NULL);
 373         ip_nat_decode_session = nat_decode_session;
 374 #endif
 375         ret = ip_nat_rule_init();
 376         if (ret < 0) {
 377                 printk("ip_nat_init: can't setup rules.\n");
 378                 goto cleanup_decode_session;
 379         }
 380         ret = nf_register_hook(&ip_nat_in_ops);
 381         if (ret < 0) {
 382                 printk("ip_nat_init: can't register in hook.\n");
 383                 goto cleanup_rule_init;
 384         }
 385         ret = nf_register_hook(&ip_nat_out_ops);
 386         if (ret < 0) {
 387                 printk("ip_nat_init: can't register out hook.\n");
 388                 goto cleanup_inops;
 389         }
 390         ret = nf_register_hook(&ip_nat_adjust_in_ops);
 391         if (ret < 0) {
 392                 printk("ip_nat_init: can't register adjust in hook.\n");
 393                 goto cleanup_outops;
 394         }
 395         ret = nf_register_hook(&ip_nat_adjust_out_ops);
 396         if (ret < 0) {
 397                 printk("ip_nat_init: can't register adjust out hook.\n");
 398                 goto cleanup_adjustin_ops;
 399         }
 400         ret = nf_register_hook(&ip_nat_local_out_ops);
 401         if (ret < 0) {
 402                 printk("ip_nat_init: can't register local out hook.\n");
 403                 goto cleanup_adjustout_ops;;
 404         }
 405         ret = nf_register_hook(&ip_nat_local_in_ops);
 406         if (ret < 0) {
 407                 printk("ip_nat_init: can't register local in hook.\n");
 408                 goto cleanup_localoutops;
 409         }
 410         return ret;
 411
 412  cleanup:
 413         nf_unregister_hook(&ip_nat_local_in_ops);
 414  cleanup_localoutops:
 415         nf_unregister_hook(&ip_nat_local_out_ops);
 416  cleanup_adjustout_ops:
 417         nf_unregister_hook(&ip_nat_adjust_out_ops);
 418  cleanup_adjustin_ops:
 419         nf_unregister_hook(&ip_nat_adjust_in_ops);
 420  cleanup_outops:
 421         nf_unregister_hook(&ip_nat_out_ops);
 422  cleanup_inops:
 423         nf_unregister_hook(&ip_nat_in_ops);
 424  cleanup_rule_init:
 425         ip_nat_rule_cleanup();
 426  cleanup_decode_session:
 427 #ifdef CONFIG_XFRM
 428         ip_nat_decode_session = NULL;
 429         synchronize_net();
 430 #endif
 431         return ret;
 432 }
 433
 434 static int __init init(void)
 435 {
 436         return init_or_cleanup(1);
 437 }
 438
 439 static void __exit fini(void)
 440 {
 441         init_or_cleanup(0);
 442 }
 443
 444 module_init(init);
 445 module_exit(fini);
 446
 447 MODULE_LICENSE("GPL");