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         ct = ip_conntrack_get(*pskb, &ctinfo);
 114
 115         /* Don't try to NAT if this packet is not conntracked */
 116         if (ct == &ip_conntrack_untracked)
 117                 return NF_ACCEPT;
 118
 119         /* If we had a hardware checksum before, it's now invalid */
 120         if ((*pskb)->ip_summed == CHECKSUM_HW)
 121                 if (skb_checksum_help(*pskb, (out == NULL)))
 122                         return NF_DROP;
 123
 124         /* Can't track?  It's not due to stress, or conntrack would
 125            have dropped it.  Hence it's the user's responsibilty to
 126            packet filter it out, or implement conntrack/NAT for that
 127            protocol. 8) --RR */
 128         if (!ct) {
 129                 /* Exception: ICMP redirect to new connection (not in
 130                    hash table yet).  We must not let this through, in
 131                    case we're doing NAT to the same network. */
 132                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 133                         struct icmphdr _hdr, *hp;
 134
 135                         hp = skb_header_pointer(*pskb,
 136                                                 (*pskb)->nh.iph->ihl*4,
 137                                                 sizeof(_hdr), &_hdr);
 138                         if (hp != NULL &&
 139                             hp->type == ICMP_REDIRECT)
 140                                 return NF_DROP;
 141                 }
 142                 return NF_ACCEPT;
 143         }
 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         unsigned int ret;
 204         u_int32_t daddr = (*pskb)->nh.iph->daddr;
 205
 206         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 207         if (ret != NF_DROP && ret != NF_STOLEN
 208             && daddr != (*pskb)->nh.iph->daddr) {
 209                 dst_release((*pskb)->dst);
 210                 (*pskb)->dst = NULL;
 211         }
 212         return ret;
 213 }
 214
 215 static unsigned int
 216 ip_nat_out(unsigned int hooknum,
 217            struct sk_buff **pskb,
 218            const struct net_device *in,
 219            const struct net_device *out,
 220            int (*okfn)(struct sk_buff *))
 221 {
 222 #ifdef CONFIG_XFRM
 223         struct ip_conntrack *ct;
 224         enum ip_conntrack_info ctinfo;
 225 #endif
 226         unsigned int ret;
 227
 228         /* root is playing with raw sockets. */
 229         if ((*pskb)->len < sizeof(struct iphdr)
 230             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 231                 return NF_ACCEPT;
 232
 233         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 234 #ifdef CONFIG_XFRM
 235         if (ret != NF_DROP && ret != NF_STOLEN
 236             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 237                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 238
 239                 if (ct->tuplehash[dir].tuple.src.ip !=
 240                     ct->tuplehash[!dir].tuple.dst.ip
 241                     || ct->tuplehash[dir].tuple.src.u.all !=
 242                        ct->tuplehash[!dir].tuple.dst.u.all
 243                     )
 244                         return ip_xfrm_me_harder(pskb) == 0 ? ret : NF_DROP;
 245         }
 246 #endif
 247         return ret;
 248 }
 249
 250 static unsigned int
 251 ip_nat_local_fn(unsigned int hooknum,
 252                 struct sk_buff **pskb,
 253                 const struct net_device *in,
 254                 const struct net_device *out,
 255                 int (*okfn)(struct sk_buff *))
 256 {
 257         struct ip_conntrack *ct;
 258         enum ip_conntrack_info ctinfo;
 259         unsigned int ret;
 260
 261         /* root is playing with raw sockets. */
 262         if ((*pskb)->len < sizeof(struct iphdr)
 263             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 264                 return NF_ACCEPT;
 265
 266         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 267         if (ret != NF_DROP && ret != NF_STOLEN
 268             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 269                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 270
 271                 if (ct->tuplehash[dir].tuple.dst.ip !=
 272                     ct->tuplehash[!dir].tuple.src.ip
 273 #ifdef CONFIG_XFRM
 274                     || ct->tuplehash[dir].tuple.dst.u.all !=
 275                        ct->tuplehash[!dir].tuple.src.u.all
 276 #endif
 277                     )
 278                         if (ip_route_me_harder(pskb, RTN_UNSPEC))
 279                                 ret = NF_DROP;
 280         }
 281         return ret;
 282 }
 283
 284 static unsigned int
 285 ip_nat_adjust(unsigned int hooknum,
 286               struct sk_buff **pskb,
 287               const struct net_device *in,
 288               const struct net_device *out,
 289               int (*okfn)(struct sk_buff *))
 290 {
 291         struct ip_conntrack *ct;
 292         enum ip_conntrack_info ctinfo;
 293
 294         ct = ip_conntrack_get(*pskb, &ctinfo);
 295         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 296                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
 297                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
 298                         return NF_DROP;
 299         }
 300         return NF_ACCEPT;
 301 }
 302
 303 /* We must be after connection tracking and before packet filtering. */
 304
 305 static struct nf_hook_ops ip_nat_ops[] = {
 306         /* Before packet filtering, change destination */
 307         {
 308                 .hook           = ip_nat_in,
 309                 .owner          = THIS_MODULE,
 310                 .pf             = PF_INET,
 311                 .hooknum        = NF_IP_PRE_ROUTING,
 312                 .priority       = NF_IP_PRI_NAT_DST,
 313         },
 314         /* After packet filtering, change source */
 315         {
 316                 .hook           = ip_nat_out,
 317                 .owner          = THIS_MODULE,
 318                 .pf             = PF_INET,
 319                 .hooknum        = NF_IP_POST_ROUTING,
 320                 .priority       = NF_IP_PRI_NAT_SRC,
 321         },
 322         /* After conntrack, adjust sequence number */
 323         {
 324                 .hook           = ip_nat_adjust,
 325                 .owner          = THIS_MODULE,
 326                 .pf             = PF_INET,
 327                 .hooknum        = NF_IP_POST_ROUTING,
 328                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 329         },
 330         /* Before packet filtering, change destination */
 331         {
 332                 .hook           = ip_nat_local_fn,
 333                 .owner          = THIS_MODULE,
 334                 .pf             = PF_INET,
 335                 .hooknum        = NF_IP_LOCAL_OUT,
 336                 .priority       = NF_IP_PRI_NAT_DST,
 337         },
 338         /* After packet filtering, change source */
 339         {
 340                 .hook           = ip_nat_fn,
 341                 .owner          = THIS_MODULE,
 342                 .pf             = PF_INET,
 343                 .hooknum        = NF_IP_LOCAL_IN,
 344                 .priority       = NF_IP_PRI_NAT_SRC,
 345         },
 346         /* After conntrack, adjust sequence number */
 347         {
 348                 .hook           = ip_nat_adjust,
 349                 .owner          = THIS_MODULE,
 350                 .pf             = PF_INET,
 351                 .hooknum        = NF_IP_LOCAL_IN,
 352                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 353         },
 354 };
 355
 356 static int __init ip_nat_standalone_init(void)
 357 {
 358         int ret = 0;
 359
 360         need_conntrack();
 361
 362 #ifdef CONFIG_XFRM
 363         BUG_ON(ip_nat_decode_session != NULL);
 364         ip_nat_decode_session = nat_decode_session;
 365 #endif
 366         ret = ip_nat_rule_init();
 367         if (ret < 0) {
 368                 printk("ip_nat_init: can't setup rules.\n");
 369                 goto cleanup_decode_session;
 370         }
 371         ret = nf_register_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 372         if (ret < 0) {
 373                 printk("ip_nat_init: can't register hooks.\n");
 374                 goto cleanup_rule_init;
 375         }
 376         return ret;
 377
 378  cleanup_rule_init:
 379         ip_nat_rule_cleanup();
 380  cleanup_decode_session:
 381 #ifdef CONFIG_XFRM
 382         ip_nat_decode_session = NULL;
 383         synchronize_net();
 384 #endif
 385         return ret;
 386 }
 387
 388 static void __exit ip_nat_standalone_fini(void)
 389 {
 390         nf_unregister_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 391         ip_nat_rule_cleanup();
 392 #ifdef CONFIG_XFRM
 393         ip_nat_decode_session = NULL;
 394         synchronize_net();
 395 #endif
 396 }
 397
 398 module_init(ip_nat_standalone_init);
 399 module_exit(ip_nat_standalone_fini);
 400
 401 MODULE_LICENSE("GPL");