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 #define HOOKNAME(hooknum) ((hooknum) == NF_IP_POST_ROUTING ? "POST_ROUTING"  \
  48                            : ((hooknum) == NF_IP_PRE_ROUTING ? "PRE_ROUTING" \
  49                               : ((hooknum) == NF_IP_LOCAL_OUT ? "LOCAL_OUT"  \
  50                                  : ((hooknum) == NF_IP_LOCAL_IN ? "LOCAL_IN"  \
  51                                     : "*ERROR*")))
  52
  53 #ifdef CONFIG_XFRM
  54 static void nat_decode_session(struct sk_buff *skb, struct flowi *fl)
  55 {
  56         struct ip_conntrack *ct;
  57         struct ip_conntrack_tuple *t;
  58         enum ip_conntrack_info ctinfo;
  59         enum ip_conntrack_dir dir;
  60         unsigned long statusbit;
  61
  62         ct = ip_conntrack_get(skb, &ctinfo);
  63         if (ct == NULL)
  64                 return;
  65         dir = CTINFO2DIR(ctinfo);
  66         t = &ct->tuplehash[dir].tuple;
  67
  68         if (dir == IP_CT_DIR_ORIGINAL)
  69                 statusbit = IPS_DST_NAT;
  70         else
  71                 statusbit = IPS_SRC_NAT;
  72
  73         if (ct->status & statusbit) {
  74                 fl->fl4_dst = t->dst.ip;
  75                 if (t->dst.protonum == IPPROTO_TCP ||
  76                     t->dst.protonum == IPPROTO_UDP)
  77                         fl->fl_ip_dport = t->dst.u.tcp.port;
  78         }
  79
  80         statusbit ^= IPS_NAT_MASK;
  81
  82         if (ct->status & statusbit) {
  83                 fl->fl4_src = t->src.ip;
  84                 if (t->dst.protonum == IPPROTO_TCP ||
  85                     t->dst.protonum == IPPROTO_UDP)
  86                         fl->fl_ip_sport = t->src.u.tcp.port;
  87         }
  88 }
  89 #endif
  90
  91 static unsigned int
  92 ip_nat_fn(unsigned int hooknum,
  93           struct sk_buff **pskb,
  94           const struct net_device *in,
  95           const struct net_device *out,
  96           int (*okfn)(struct sk_buff *))
  97 {
  98         struct ip_conntrack *ct;
  99         enum ip_conntrack_info ctinfo;
 100         struct ip_nat_info *info;
 101         /* maniptype == SRC for postrouting. */
 102         enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
 103
 104         /* We never see fragments: conntrack defrags on pre-routing
 105            and local-out, and ip_nat_out protects post-routing. */
 106         IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
 107                        & htons(IP_MF|IP_OFFSET)));
 108
 109         ct = ip_conntrack_get(*pskb, &ctinfo);
 110         /* Can't track?  It's not due to stress, or conntrack would
 111            have dropped it.  Hence it's the user's responsibilty to
 112            packet filter it out, or implement conntrack/NAT for that
 113            protocol. 8) --RR */
 114         if (!ct) {
 115                 /* Exception: ICMP redirect to new connection (not in
 116                    hash table yet).  We must not let this through, in
 117                    case we're doing NAT to the same network. */
 118                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 119                         struct icmphdr _hdr, *hp;
 120
 121                         hp = skb_header_pointer(*pskb,
 122                                                 (*pskb)->nh.iph->ihl*4,
 123                                                 sizeof(_hdr), &_hdr);
 124                         if (hp != NULL &&
 125                             hp->type == ICMP_REDIRECT)
 126                                 return NF_DROP;
 127                 }
 128                 return NF_ACCEPT;
 129         }
 130
 131         /* Don't try to NAT if this packet is not conntracked */
 132         if (ct == &ip_conntrack_untracked)
 133                 return NF_ACCEPT;
 134
 135         switch (ctinfo) {
 136         case IP_CT_RELATED:
 137         case IP_CT_RELATED+IP_CT_IS_REPLY:
 138                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 139                         if (!ip_nat_icmp_reply_translation(ct, ctinfo,
 140                                                            hooknum, pskb))
 141                                 return NF_DROP;
 142                         else
 143                                 return NF_ACCEPT;
 144                 }
 145                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 146         case IP_CT_NEW:
 147                 info = &ct->nat.info;
 148
 149                 /* Seen it before?  This can happen for loopback, retrans,
 150                    or local packets.. */
 151                 if (!ip_nat_initialized(ct, maniptype)) {
 152                         unsigned int ret;
 153
 154                         if (unlikely(is_confirmed(ct)))
 155                                 /* NAT module was loaded late */
 156                                 ret = alloc_null_binding_confirmed(ct, info,
 157                                                                    hooknum);
 158                         else if (hooknum == NF_IP_LOCAL_IN)
 159                                 /* LOCAL_IN hook doesn't have a chain!  */
 160                                 ret = alloc_null_binding(ct, info, hooknum);
 161                         else
 162                                 ret = ip_nat_rule_find(pskb, hooknum,
 163                                                        in, out, ct,
 164                                                        info);
 165
 166                         if (ret != NF_ACCEPT) {
 167                                 return ret;
 168                         }
 169                 } else
 170                         DEBUGP("Already setup manip %s for ct %p\n",
 171                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 172                                ct);
 173                 break;
 174
 175         default:
 176                 /* ESTABLISHED */
 177                 IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
 178                              || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 179                 info = &ct->nat.info;
 180         }
 181
 182         IP_NF_ASSERT(info);
 183         return ip_nat_packet(ct, ctinfo, hooknum, pskb);
 184 }
 185
 186 static unsigned int
 187 ip_nat_in(unsigned int hooknum,
 188           struct sk_buff **pskb,
 189           const struct net_device *in,
 190           const struct net_device *out,
 191           int (*okfn)(struct sk_buff *))
 192 {
 193         unsigned int ret;
 194         u_int32_t daddr = (*pskb)->nh.iph->daddr;
 195
 196         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 197         if (ret != NF_DROP && ret != NF_STOLEN
 198             && daddr != (*pskb)->nh.iph->daddr) {
 199                 dst_release((*pskb)->dst);
 200                 (*pskb)->dst = NULL;
 201         }
 202         return ret;
 203 }
 204
 205 static unsigned int
 206 ip_nat_out(unsigned int hooknum,
 207            struct sk_buff **pskb,
 208            const struct net_device *in,
 209            const struct net_device *out,
 210            int (*okfn)(struct sk_buff *))
 211 {
 212 #ifdef CONFIG_XFRM
 213         struct ip_conntrack *ct;
 214         enum ip_conntrack_info ctinfo;
 215 #endif
 216         unsigned int ret;
 217
 218         /* root is playing with raw sockets. */
 219         if ((*pskb)->len < sizeof(struct iphdr)
 220             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 221                 return NF_ACCEPT;
 222
 223         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 224 #ifdef CONFIG_XFRM
 225         if (ret != NF_DROP && ret != NF_STOLEN
 226             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 227                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 228
 229                 if (ct->tuplehash[dir].tuple.src.ip !=
 230                     ct->tuplehash[!dir].tuple.dst.ip
 231                     || ct->tuplehash[dir].tuple.src.u.all !=
 232                        ct->tuplehash[!dir].tuple.dst.u.all
 233                     )
 234                         return ip_xfrm_me_harder(pskb) == 0 ? ret : NF_DROP;
 235         }
 236 #endif
 237         return ret;
 238 }
 239
 240 static unsigned int
 241 ip_nat_local_fn(unsigned int hooknum,
 242                 struct sk_buff **pskb,
 243                 const struct net_device *in,
 244                 const struct net_device *out,
 245                 int (*okfn)(struct sk_buff *))
 246 {
 247         struct ip_conntrack *ct;
 248         enum ip_conntrack_info ctinfo;
 249         unsigned int ret;
 250
 251         /* root is playing with raw sockets. */
 252         if ((*pskb)->len < sizeof(struct iphdr)
 253             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 254                 return NF_ACCEPT;
 255
 256         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 257         if (ret != NF_DROP && ret != NF_STOLEN
 258             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 259                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 260
 261                 if (ct->tuplehash[dir].tuple.dst.ip !=
 262                     ct->tuplehash[!dir].tuple.src.ip
 263 #ifdef CONFIG_XFRM
 264                     || ct->tuplehash[dir].tuple.dst.u.all !=
 265                        ct->tuplehash[!dir].tuple.src.u.all
 266 #endif
 267                     )
 268                         return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
 269         }
 270         return ret;
 271 }
 272
 273 static unsigned int
 274 ip_nat_adjust(unsigned int hooknum,
 275               struct sk_buff **pskb,
 276               const struct net_device *in,
 277               const struct net_device *out,
 278               int (*okfn)(struct sk_buff *))
 279 {
 280         struct ip_conntrack *ct;
 281         enum ip_conntrack_info ctinfo;
 282
 283         ct = ip_conntrack_get(*pskb, &ctinfo);
 284         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 285                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
 286                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
 287                         return NF_DROP;
 288         }
 289         return NF_ACCEPT;
 290 }
 291
 292 /* We must be after connection tracking and before packet filtering. */
 293
 294 static struct nf_hook_ops ip_nat_ops[] = {
 295         /* Before packet filtering, change destination */
 296         {
 297                 .hook           = ip_nat_in,
 298                 .owner          = THIS_MODULE,
 299                 .pf             = PF_INET,
 300                 .hooknum        = NF_IP_PRE_ROUTING,
 301                 .priority       = NF_IP_PRI_NAT_DST,
 302         },
 303         /* After packet filtering, change source */
 304         {
 305                 .hook           = ip_nat_out,
 306                 .owner          = THIS_MODULE,
 307                 .pf             = PF_INET,
 308                 .hooknum        = NF_IP_POST_ROUTING,
 309                 .priority       = NF_IP_PRI_NAT_SRC,
 310         },
 311         /* After conntrack, adjust sequence number */
 312         {
 313                 .hook           = ip_nat_adjust,
 314                 .owner          = THIS_MODULE,
 315                 .pf             = PF_INET,
 316                 .hooknum        = NF_IP_POST_ROUTING,
 317                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 318         },
 319         /* Before packet filtering, change destination */
 320         {
 321                 .hook           = ip_nat_local_fn,
 322                 .owner          = THIS_MODULE,
 323                 .pf             = PF_INET,
 324                 .hooknum        = NF_IP_LOCAL_OUT,
 325                 .priority       = NF_IP_PRI_NAT_DST,
 326         },
 327         /* After packet filtering, change source */
 328         {
 329                 .hook           = ip_nat_fn,
 330                 .owner          = THIS_MODULE,
 331                 .pf             = PF_INET,
 332                 .hooknum        = NF_IP_LOCAL_IN,
 333                 .priority       = NF_IP_PRI_NAT_SRC,
 334         },
 335         /* After conntrack, adjust sequence number */
 336         {
 337                 .hook           = ip_nat_adjust,
 338                 .owner          = THIS_MODULE,
 339                 .pf             = PF_INET,
 340                 .hooknum        = NF_IP_LOCAL_IN,
 341                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 342         },
 343 };
 344
 345 static int __init ip_nat_standalone_init(void)
 346 {
 347         int ret = 0;
 348
 349         need_conntrack();
 350
 351 #ifdef CONFIG_XFRM
 352         BUG_ON(ip_nat_decode_session != NULL);
 353         ip_nat_decode_session = nat_decode_session;
 354 #endif
 355         ret = ip_nat_rule_init();
 356         if (ret < 0) {
 357                 printk("ip_nat_init: can't setup rules.\n");
 358                 goto cleanup_decode_session;
 359         }
 360         ret = nf_register_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 361         if (ret < 0) {
 362                 printk("ip_nat_init: can't register hooks.\n");
 363                 goto cleanup_rule_init;
 364         }
 365         return ret;
 366
 367  cleanup_rule_init:
 368         ip_nat_rule_cleanup();
 369  cleanup_decode_session:
 370 #ifdef CONFIG_XFRM
 371         ip_nat_decode_session = NULL;
 372         synchronize_net();
 373 #endif
 374         return ret;
 375 }
 376
 377 static void __exit ip_nat_standalone_fini(void)
 378 {
 379         nf_unregister_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 380         ip_nat_rule_cleanup();
 381 #ifdef CONFIG_XFRM
 382         ip_nat_decode_session = NULL;
 383         synchronize_net();
 384 #endif
 385 }
 386
 387 module_init(ip_nat_standalone_init);
 388 module_exit(ip_nat_standalone_fini);
 389
 390 MODULE_LICENSE("GPL");