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/types.h>
  22 #include <linux/icmp.h>
  23 #include <linux/ip.h>
  24 #include <linux/netfilter.h>
  25 #include <linux/netfilter_ipv4.h>
  26 #include <linux/module.h>
  27 #include <linux/skbuff.h>
  28 #include <linux/proc_fs.h>
  29 #include <net/ip.h>
  30 #include <net/checksum.h>
  31 #include <linux/spinlock.h>
  32
  33 #include <linux/netfilter_ipv4/ip_nat.h>
  34 #include <linux/netfilter_ipv4/ip_nat_rule.h>
  35 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  36 #include <linux/netfilter_ipv4/ip_nat_core.h>
  37 #include <linux/netfilter_ipv4/ip_nat_helper.h>
  38 #include <linux/netfilter_ipv4/ip_tables.h>
  39 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  40
  41 #if 0
  42 #define DEBUGP printk
  43 #else
  44 #define DEBUGP(format, args...)
  45 #endif
  46
  47 #ifdef CONFIG_XFRM
  48 static void nat_decode_session(struct sk_buff *skb, struct flowi *fl)
  49 {
  50         struct ip_conntrack *ct;
  51         struct ip_conntrack_tuple *t;
  52         enum ip_conntrack_info ctinfo;
  53         enum ip_conntrack_dir dir;
  54         unsigned long statusbit;
  55
  56         ct = ip_conntrack_get(skb, &ctinfo);
  57         if (ct == NULL)
  58                 return;
  59         dir = CTINFO2DIR(ctinfo);
  60         t = &ct->tuplehash[dir].tuple;
  61
  62         if (dir == IP_CT_DIR_ORIGINAL)
  63                 statusbit = IPS_DST_NAT;
  64         else
  65                 statusbit = IPS_SRC_NAT;
  66
  67         if (ct->status & statusbit) {
  68                 fl->fl4_dst = t->dst.ip;
  69                 if (t->dst.protonum == IPPROTO_TCP ||
  70                     t->dst.protonum == IPPROTO_UDP)
  71                         fl->fl_ip_dport = t->dst.u.tcp.port;
  72         }
  73
  74         statusbit ^= IPS_NAT_MASK;
  75
  76         if (ct->status & statusbit) {
  77                 fl->fl4_src = t->src.ip;
  78                 if (t->dst.protonum == IPPROTO_TCP ||
  79                     t->dst.protonum == IPPROTO_UDP)
  80                         fl->fl_ip_sport = t->src.u.tcp.port;
  81         }
  82 }
  83 #endif
  84
  85 static unsigned int
  86 ip_nat_fn(unsigned int hooknum,
  87           struct sk_buff **pskb,
  88           const struct net_device *in,
  89           const struct net_device *out,
  90           int (*okfn)(struct sk_buff *))
  91 {
  92         struct ip_conntrack *ct;
  93         enum ip_conntrack_info ctinfo;
  94         struct ip_nat_info *info;
  95         /* maniptype == SRC for postrouting. */
  96         enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
  97
  98         /* We never see fragments: conntrack defrags on pre-routing
  99            and local-out, and ip_nat_out protects post-routing. */
 100         IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
 101                        & htons(IP_MF|IP_OFFSET)));
 102
 103         ct = ip_conntrack_get(*pskb, &ctinfo);
 104         /* Can't track?  It's not due to stress, or conntrack would
 105            have dropped it.  Hence it's the user's responsibilty to
 106            packet filter it out, or implement conntrack/NAT for that
 107            protocol. 8) --RR */
 108         if (!ct) {
 109                 /* Exception: ICMP redirect to new connection (not in
 110                    hash table yet).  We must not let this through, in
 111                    case we're doing NAT to the same network. */
 112                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 113                         struct icmphdr _hdr, *hp;
 114
 115                         hp = skb_header_pointer(*pskb,
 116                                                 (*pskb)->nh.iph->ihl*4,
 117                                                 sizeof(_hdr), &_hdr);
 118                         if (hp != NULL &&
 119                             hp->type == ICMP_REDIRECT)
 120                                 return NF_DROP;
 121                 }
 122                 return NF_ACCEPT;
 123         }
 124
 125         /* Don't try to NAT if this packet is not conntracked */
 126         if (ct == &ip_conntrack_untracked)
 127                 return NF_ACCEPT;
 128
 129         switch (ctinfo) {
 130         case IP_CT_RELATED:
 131         case IP_CT_RELATED+IP_CT_IS_REPLY:
 132                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 133                         if (!ip_nat_icmp_reply_translation(ct, ctinfo,
 134                                                            hooknum, pskb))
 135                                 return NF_DROP;
 136                         else
 137                                 return NF_ACCEPT;
 138                 }
 139                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 140         case IP_CT_NEW:
 141                 info = &ct->nat.info;
 142
 143                 /* Seen it before?  This can happen for loopback, retrans,
 144                    or local packets.. */
 145                 if (!ip_nat_initialized(ct, maniptype)) {
 146                         unsigned int ret;
 147
 148                         if (unlikely(is_confirmed(ct)))
 149                                 /* NAT module was loaded late */
 150                                 ret = alloc_null_binding_confirmed(ct, info,
 151                                                                    hooknum);
 152                         else if (hooknum == NF_IP_LOCAL_IN)
 153                                 /* LOCAL_IN hook doesn't have a chain!  */
 154                                 ret = alloc_null_binding(ct, info, hooknum);
 155                         else
 156                                 ret = ip_nat_rule_find(pskb, hooknum,
 157                                                        in, out, ct,
 158                                                        info);
 159
 160                         if (ret != NF_ACCEPT) {
 161                                 return ret;
 162                         }
 163                 } else
 164                         DEBUGP("Already setup manip %s for ct %p\n",
 165                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 166                                ct);
 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 ip_nat_packet(ct, ctinfo, hooknum, pskb);
 178 }
 179
 180 static unsigned int
 181 ip_nat_in(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         unsigned int ret;
 188         __be32 daddr = (*pskb)->nh.iph->daddr;
 189
 190         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 191         if (ret != NF_DROP && ret != NF_STOLEN
 192             && daddr != (*pskb)->nh.iph->daddr) {
 193                 dst_release((*pskb)->dst);
 194                 (*pskb)->dst = NULL;
 195         }
 196         return ret;
 197 }
 198
 199 static unsigned int
 200 ip_nat_out(unsigned int hooknum,
 201            struct sk_buff **pskb,
 202            const struct net_device *in,
 203            const struct net_device *out,
 204            int (*okfn)(struct sk_buff *))
 205 {
 206 #ifdef CONFIG_XFRM
 207         struct ip_conntrack *ct;
 208         enum ip_conntrack_info ctinfo;
 209 #endif
 210         unsigned int ret;
 211
 212         /* root is playing with raw sockets. */
 213         if ((*pskb)->len < sizeof(struct iphdr)
 214             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 215                 return NF_ACCEPT;
 216
 217         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 218 #ifdef CONFIG_XFRM
 219         if (ret != NF_DROP && ret != NF_STOLEN
 220             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 221                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 222
 223                 if (ct->tuplehash[dir].tuple.src.ip !=
 224                     ct->tuplehash[!dir].tuple.dst.ip
 225                     || ct->tuplehash[dir].tuple.src.u.all !=
 226                        ct->tuplehash[!dir].tuple.dst.u.all
 227                     )
 228                         return ip_xfrm_me_harder(pskb) == 0 ? ret : NF_DROP;
 229         }
 230 #endif
 231         return ret;
 232 }
 233
 234 static unsigned int
 235 ip_nat_local_fn(unsigned int hooknum,
 236                 struct sk_buff **pskb,
 237                 const struct net_device *in,
 238                 const struct net_device *out,
 239                 int (*okfn)(struct sk_buff *))
 240 {
 241         struct ip_conntrack *ct;
 242         enum ip_conntrack_info ctinfo;
 243         unsigned int ret;
 244
 245         /* root is playing with raw sockets. */
 246         if ((*pskb)->len < sizeof(struct iphdr)
 247             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 248                 return NF_ACCEPT;
 249
 250         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 251         if (ret != NF_DROP && ret != NF_STOLEN
 252             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 253                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 254
 255                 if (ct->tuplehash[dir].tuple.dst.ip !=
 256                     ct->tuplehash[!dir].tuple.src.ip) {
 257                         if (ip_route_me_harder(pskb, RTN_UNSPEC))
 258                                 ret = NF_DROP;
 259                 }
 260 #ifdef CONFIG_XFRM
 261                 else if (ct->tuplehash[dir].tuple.dst.u.all !=
 262                          ct->tuplehash[!dir].tuple.src.u.all)
 263                         if (ip_xfrm_me_harder(pskb))
 264                                 ret = NF_DROP;
 265 #endif
 266
 267         }
 268         return ret;
 269 }
 270
 271 static unsigned int
 272 ip_nat_adjust(unsigned int hooknum,
 273               struct sk_buff **pskb,
 274               const struct net_device *in,
 275               const struct net_device *out,
 276               int (*okfn)(struct sk_buff *))
 277 {
 278         struct ip_conntrack *ct;
 279         enum ip_conntrack_info ctinfo;
 280
 281         ct = ip_conntrack_get(*pskb, &ctinfo);
 282         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 283                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
 284                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
 285                         return NF_DROP;
 286         }
 287         return NF_ACCEPT;
 288 }
 289
 290 /* We must be after connection tracking and before packet filtering. */
 291
 292 static struct nf_hook_ops ip_nat_ops[] = {
 293         /* Before packet filtering, change destination */
 294         {
 295                 .hook           = ip_nat_in,
 296                 .owner          = THIS_MODULE,
 297                 .pf             = PF_INET,
 298                 .hooknum        = NF_IP_PRE_ROUTING,
 299                 .priority       = NF_IP_PRI_NAT_DST,
 300         },
 301         /* After packet filtering, change source */
 302         {
 303                 .hook           = ip_nat_out,
 304                 .owner          = THIS_MODULE,
 305                 .pf             = PF_INET,
 306                 .hooknum        = NF_IP_POST_ROUTING,
 307                 .priority       = NF_IP_PRI_NAT_SRC,
 308         },
 309         /* After conntrack, adjust sequence number */
 310         {
 311                 .hook           = ip_nat_adjust,
 312                 .owner          = THIS_MODULE,
 313                 .pf             = PF_INET,
 314                 .hooknum        = NF_IP_POST_ROUTING,
 315                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 316         },
 317         /* Before packet filtering, change destination */
 318         {
 319                 .hook           = ip_nat_local_fn,
 320                 .owner          = THIS_MODULE,
 321                 .pf             = PF_INET,
 322                 .hooknum        = NF_IP_LOCAL_OUT,
 323                 .priority       = NF_IP_PRI_NAT_DST,
 324         },
 325         /* After packet filtering, change source */
 326         {
 327                 .hook           = ip_nat_fn,
 328                 .owner          = THIS_MODULE,
 329                 .pf             = PF_INET,
 330                 .hooknum        = NF_IP_LOCAL_IN,
 331                 .priority       = NF_IP_PRI_NAT_SRC,
 332         },
 333         /* After conntrack, adjust sequence number */
 334         {
 335                 .hook           = ip_nat_adjust,
 336                 .owner          = THIS_MODULE,
 337                 .pf             = PF_INET,
 338                 .hooknum        = NF_IP_LOCAL_IN,
 339                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 340         },
 341 };
 342
 343 static int __init ip_nat_standalone_init(void)
 344 {
 345         int ret = 0;
 346
 347         need_conntrack();
 348
 349 #ifdef CONFIG_XFRM
 350         BUG_ON(ip_nat_decode_session != NULL);
 351         ip_nat_decode_session = nat_decode_session;
 352 #endif
 353         ret = ip_nat_rule_init();
 354         if (ret < 0) {
 355                 printk("ip_nat_init: can't setup rules.\n");
 356                 goto cleanup_decode_session;
 357         }
 358         ret = nf_register_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 359         if (ret < 0) {
 360                 printk("ip_nat_init: can't register hooks.\n");
 361                 goto cleanup_rule_init;
 362         }
 363         return ret;
 364
 365  cleanup_rule_init:
 366         ip_nat_rule_cleanup();
 367  cleanup_decode_session:
 368 #ifdef CONFIG_XFRM
 369         ip_nat_decode_session = NULL;
 370         synchronize_net();
 371 #endif
 372         return ret;
 373 }
 374
 375 static void __exit ip_nat_standalone_fini(void)
 376 {
 377         nf_unregister_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 378         ip_nat_rule_cleanup();
 379 #ifdef CONFIG_XFRM
 380         ip_nat_decode_session = NULL;
 381         synchronize_net();
 382 #endif
 383 }
 384
 385 module_init(ip_nat_standalone_init);
 386 module_exit(ip_nat_standalone_fini);
 387
 388 MODULE_LICENSE("GPL");