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 #define ASSERT_READ_LOCK(x)
  34 #define ASSERT_WRITE_LOCK(x)
  35
  36 #include <linux/netfilter_ipv4/ip_nat.h>
  37 #include <linux/netfilter_ipv4/ip_nat_rule.h>
  38 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  39 #include <linux/netfilter_ipv4/ip_nat_core.h>
  40 #include <linux/netfilter_ipv4/ip_nat_helper.h>
  41 #include <linux/netfilter_ipv4/ip_tables.h>
  42 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  43 #include <linux/netfilter_ipv4/listhelp.h>
  44
  45 #if 0
  46 #define DEBUGP printk
  47 #else
  48 #define DEBUGP(format, args...)
  49 #endif
  50
  51 #define HOOKNAME(hooknum) ((hooknum) == NF_IP_POST_ROUTING ? "POST_ROUTING"  \
  52                            : ((hooknum) == NF_IP_PRE_ROUTING ? "PRE_ROUTING" \
  53                               : ((hooknum) == NF_IP_LOCAL_OUT ? "LOCAL_OUT"  \
  54                                  : ((hooknum) == NF_IP_LOCAL_IN ? "LOCAL_IN"  \
  55                                     : "*ERROR*")))
  56
  57 #ifdef CONFIG_XFRM
  58 static void nat_decode_session(struct sk_buff *skb, struct flowi *fl)
  59 {
  60         struct ip_conntrack *ct;
  61         struct ip_conntrack_tuple *t;
  62         enum ip_conntrack_info ctinfo;
  63         enum ip_conntrack_dir dir;
  64         unsigned long statusbit;
  65
  66         ct = ip_conntrack_get(skb, &ctinfo);
  67         if (ct == NULL)
  68                 return;
  69         dir = CTINFO2DIR(ctinfo);
  70         t = &ct->tuplehash[dir].tuple;
  71
  72         if (dir == IP_CT_DIR_ORIGINAL)
  73                 statusbit = IPS_DST_NAT;
  74         else
  75                 statusbit = IPS_SRC_NAT;
  76
  77         if (ct->status & statusbit) {
  78                 fl->fl4_dst = t->dst.ip;
  79                 if (t->dst.protonum == IPPROTO_TCP ||
  80                     t->dst.protonum == IPPROTO_UDP)
  81                         fl->fl_ip_dport = t->dst.u.tcp.port;
  82         }
  83
  84         statusbit ^= IPS_NAT_MASK;
  85
  86         if (ct->status & statusbit) {
  87                 fl->fl4_src = t->src.ip;
  88                 if (t->dst.protonum == IPPROTO_TCP ||
  89                     t->dst.protonum == IPPROTO_UDP)
  90                         fl->fl_ip_sport = t->src.u.tcp.port;
  91         }
  92 }
  93 #endif
  94
  95 static unsigned int
  96 ip_nat_fn(unsigned int hooknum,
  97           struct sk_buff **pskb,
  98           const struct net_device *in,
  99           const struct net_device *out,
 100           int (*okfn)(struct sk_buff *))
 101 {
 102         struct ip_conntrack *ct;
 103         enum ip_conntrack_info ctinfo;
 104         struct ip_nat_info *info;
 105         /* maniptype == SRC for postrouting. */
 106         enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
 107
 108         /* We never see fragments: conntrack defrags on pre-routing
 109            and local-out, and ip_nat_out protects post-routing. */
 110         IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
 111                        & htons(IP_MF|IP_OFFSET)));
 112
 113         /* If we had a hardware checksum before, it's now invalid */
 114         if ((*pskb)->ip_summed == CHECKSUM_HW)
 115                 if (skb_checksum_help(*pskb, (out == NULL)))
 116                         return NF_DROP;
 117
 118         ct = ip_conntrack_get(*pskb, &ctinfo);
 119         /* Can't track?  It's not due to stress, or conntrack would
 120            have dropped it.  Hence it's the user's responsibilty to
 121            packet filter it out, or implement conntrack/NAT for that
 122            protocol. 8) --RR */
 123         if (!ct) {
 124                 /* Exception: ICMP redirect to new connection (not in
 125                    hash table yet).  We must not let this through, in
 126                    case we're doing NAT to the same network. */
 127                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 128                         struct icmphdr _hdr, *hp;
 129
 130                         hp = skb_header_pointer(*pskb,
 131                                                 (*pskb)->nh.iph->ihl*4,
 132                                                 sizeof(_hdr), &_hdr);
 133                         if (hp != NULL &&
 134                             hp->type == ICMP_REDIRECT)
 135                                 return NF_DROP;
 136                 }
 137                 return NF_ACCEPT;
 138         }
 139
 140         /* Don't try to NAT if this packet is not conntracked */
 141         if (ct == &ip_conntrack_untracked)
 142                 return NF_ACCEPT;
 143
 144         switch (ctinfo) {
 145         case IP_CT_RELATED:
 146         case IP_CT_RELATED+IP_CT_IS_REPLY:
 147                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 148                         if (!ip_nat_icmp_reply_translation(pskb, ct, maniptype,
 149                                                            CTINFO2DIR(ctinfo)))
 150                                 return NF_DROP;
 151                         else
 152                                 return NF_ACCEPT;
 153                 }
 154                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 155         case IP_CT_NEW:
 156                 info = &ct->nat.info;
 157
 158                 /* Seen it before?  This can happen for loopback, retrans,
 159                    or local packets.. */
 160                 if (!ip_nat_initialized(ct, maniptype)) {
 161                         unsigned int ret;
 162
 163                         if (unlikely(is_confirmed(ct)))
 164                                 /* NAT module was loaded late */
 165                                 ret = alloc_null_binding_confirmed(ct, info,
 166                                                                    hooknum);
 167                         else if (hooknum == NF_IP_LOCAL_IN)
 168                                 /* LOCAL_IN hook doesn't have a chain!  */
 169                                 ret = alloc_null_binding(ct, info, hooknum);
 170                         else
 171                                 ret = ip_nat_rule_find(pskb, hooknum,
 172                                                        in, out, ct,
 173                                                        info);
 174
 175                         if (ret != NF_ACCEPT) {
 176                                 return ret;
 177                         }
 178                 } else
 179                         DEBUGP("Already setup manip %s for ct %p\n",
 180                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 181                                ct);
 182                 break;
 183
 184         default:
 185                 /* ESTABLISHED */
 186                 IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
 187                              || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 188                 info = &ct->nat.info;
 189         }
 190
 191         IP_NF_ASSERT(info);
 192         return ip_nat_packet(ct, ctinfo, hooknum, pskb);
 193 }
 194
 195 static unsigned int
 196 ip_nat_in(unsigned int hooknum,
 197           struct sk_buff **pskb,
 198           const struct net_device *in,
 199           const struct net_device *out,
 200           int (*okfn)(struct sk_buff *))
 201 {
 202         unsigned int ret;
 203         u_int32_t daddr = (*pskb)->nh.iph->daddr;
 204
 205         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 206         if (ret != NF_DROP && ret != NF_STOLEN
 207             && daddr != (*pskb)->nh.iph->daddr) {
 208                 dst_release((*pskb)->dst);
 209                 (*pskb)->dst = NULL;
 210         }
 211         return ret;
 212 }
 213
 214 static unsigned int
 215 ip_nat_out(unsigned int hooknum,
 216            struct sk_buff **pskb,
 217            const struct net_device *in,
 218            const struct net_device *out,
 219            int (*okfn)(struct sk_buff *))
 220 {
 221 #ifdef CONFIG_XFRM
 222         struct ip_conntrack *ct;
 223         enum ip_conntrack_info ctinfo;
 224 #endif
 225         unsigned int ret;
 226
 227         /* root is playing with raw sockets. */
 228         if ((*pskb)->len < sizeof(struct iphdr)
 229             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 230                 return NF_ACCEPT;
 231
 232         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 233 #ifdef CONFIG_XFRM
 234         if (ret != NF_DROP && ret != NF_STOLEN
 235             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 236                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 237
 238                 if (ct->tuplehash[dir].tuple.src.ip !=
 239                     ct->tuplehash[!dir].tuple.dst.ip
 240                     || ct->tuplehash[dir].tuple.src.u.all !=
 241                        ct->tuplehash[!dir].tuple.dst.u.all
 242                     )
 243                         return ip_xfrm_me_harder(pskb) == 0 ? ret : NF_DROP;
 244         }
 245 #endif
 246         return ret;
 247 }
 248
 249 static unsigned int
 250 ip_nat_local_fn(unsigned int hooknum,
 251                 struct sk_buff **pskb,
 252                 const struct net_device *in,
 253                 const struct net_device *out,
 254                 int (*okfn)(struct sk_buff *))
 255 {
 256         struct ip_conntrack *ct;
 257         enum ip_conntrack_info ctinfo;
 258         unsigned int ret;
 259
 260         /* root is playing with raw sockets. */
 261         if ((*pskb)->len < sizeof(struct iphdr)
 262             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 263                 return NF_ACCEPT;
 264
 265         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 266         if (ret != NF_DROP && ret != NF_STOLEN
 267             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 268                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 269
 270                 if (ct->tuplehash[dir].tuple.dst.ip !=
 271                     ct->tuplehash[!dir].tuple.src.ip
 272 #ifdef CONFIG_XFRM
 273                     || ct->tuplehash[dir].tuple.dst.u.all !=
 274                        ct->tuplehash[!dir].tuple.src.u.all
 275 #endif
 276                     )
 277                         return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
 278         }
 279         return ret;
 280 }
 281
 282 static unsigned int
 283 ip_nat_adjust(unsigned int hooknum,
 284               struct sk_buff **pskb,
 285               const struct net_device *in,
 286               const struct net_device *out,
 287               int (*okfn)(struct sk_buff *))
 288 {
 289         struct ip_conntrack *ct;
 290         enum ip_conntrack_info ctinfo;
 291
 292         ct = ip_conntrack_get(*pskb, &ctinfo);
 293         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 294                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
 295                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
 296                         return NF_DROP;
 297         }
 298         return NF_ACCEPT;
 299 }
 300
 301 /* We must be after connection tracking and before packet filtering. */
 302
 303 static struct nf_hook_ops ip_nat_ops[] = {
 304         /* Before packet filtering, change destination */
 305         {
 306                 .hook           = ip_nat_in,
 307                 .owner          = THIS_MODULE,
 308                 .pf             = PF_INET,
 309                 .hooknum        = NF_IP_PRE_ROUTING,
 310                 .priority       = NF_IP_PRI_NAT_DST,
 311         },
 312         /* After packet filtering, change source */
 313         {
 314                 .hook           = ip_nat_out,
 315                 .owner          = THIS_MODULE,
 316                 .pf             = PF_INET,
 317                 .hooknum        = NF_IP_POST_ROUTING,
 318                 .priority       = NF_IP_PRI_NAT_SRC,
 319         },
 320         /* After conntrack, adjust sequence number */
 321         {
 322                 .hook           = ip_nat_adjust,
 323                 .owner          = THIS_MODULE,
 324                 .pf             = PF_INET,
 325                 .hooknum        = NF_IP_POST_ROUTING,
 326                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 327         },
 328         /* Before packet filtering, change destination */
 329         {
 330                 .hook           = ip_nat_local_fn,
 331                 .owner          = THIS_MODULE,
 332                 .pf             = PF_INET,
 333                 .hooknum        = NF_IP_LOCAL_OUT,
 334                 .priority       = NF_IP_PRI_NAT_DST,
 335         },
 336         /* After packet filtering, change source */
 337         {
 338                 .hook           = ip_nat_fn,
 339                 .owner          = THIS_MODULE,
 340                 .pf             = PF_INET,
 341                 .hooknum        = NF_IP_LOCAL_IN,
 342                 .priority       = NF_IP_PRI_NAT_SRC,
 343         },
 344         /* After conntrack, adjust sequence number */
 345         {
 346                 .hook           = ip_nat_adjust,
 347                 .owner          = THIS_MODULE,
 348                 .pf             = PF_INET,
 349                 .hooknum        = NF_IP_LOCAL_IN,
 350                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 351         },
 352 };
 353
 354 static int __init ip_nat_standalone_init(void)
 355 {
 356         int ret = 0;
 357
 358         need_conntrack();
 359
 360 #ifdef CONFIG_XFRM
 361         BUG_ON(ip_nat_decode_session != NULL);
 362         ip_nat_decode_session = nat_decode_session;
 363 #endif
 364         ret = ip_nat_rule_init();
 365         if (ret < 0) {
 366                 printk("ip_nat_init: can't setup rules.\n");
 367                 goto cleanup_decode_session;
 368         }
 369         ret = nf_register_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 370         if (ret < 0) {
 371                 printk("ip_nat_init: can't register hooks.\n");
 372                 goto cleanup_rule_init;
 373         }
 374         return ret;
 375
 376  cleanup_rule_init:
 377         ip_nat_rule_cleanup();
 378  cleanup_decode_session:
 379 #ifdef CONFIG_XFRM
 380         ip_nat_decode_session = NULL;
 381         synchronize_net();
 382 #endif
 383         return ret;
 384 }
 385
 386 static void __exit ip_nat_standalone_fini(void)
 387 {
 388         nf_unregister_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 389         ip_nat_rule_cleanup();
 390 #ifdef CONFIG_XFRM
 391         ip_nat_decode_session = NULL;
 392         synchronize_net();
 393 #endif
 394 }
 395
 396 module_init(ip_nat_standalone_init);
 397 module_exit(ip_nat_standalone_fini);
 398
 399 MODULE_LICENSE("GPL");