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         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         struct ip_conntrack *ct;
 223         enum ip_conntrack_info ctinfo;
 224         unsigned int ret;
 225
 226         /* root is playing with raw sockets. */
 227         if ((*pskb)->len < sizeof(struct iphdr)
 228             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 229                 return NF_ACCEPT;
 230
 231         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 232 #ifdef CONFIG_XFRM
 233         if (ret != NF_DROP && ret != NF_STOLEN
 234             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 235                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 236
 237                 if (ct->tuplehash[dir].tuple.src.ip !=
 238                     ct->tuplehash[!dir].tuple.dst.ip
 239                     || ct->tuplehash[dir].tuple.src.u.all !=
 240                        ct->tuplehash[!dir].tuple.dst.u.all
 241                     )
 242                         return ip_xfrm_me_harder(pskb) == 0 ? ret : NF_DROP;
 243         }
 244 #endif
 245         return ret;
 246 }
 247
 248 static unsigned int
 249 ip_nat_local_fn(unsigned int hooknum,
 250                 struct sk_buff **pskb,
 251                 const struct net_device *in,
 252                 const struct net_device *out,
 253                 int (*okfn)(struct sk_buff *))
 254 {
 255         struct ip_conntrack *ct;
 256         enum ip_conntrack_info ctinfo;
 257         unsigned int ret;
 258
 259         /* root is playing with raw sockets. */
 260         if ((*pskb)->len < sizeof(struct iphdr)
 261             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 262                 return NF_ACCEPT;
 263
 264         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 265         if (ret != NF_DROP && ret != NF_STOLEN
 266             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
 267                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
 268
 269                 if (ct->tuplehash[dir].tuple.dst.ip !=
 270                     ct->tuplehash[!dir].tuple.src.ip
 271 #ifdef CONFIG_XFRM
 272                     || ct->tuplehash[dir].tuple.dst.u.all !=
 273                        ct->tuplehash[!dir].tuple.src.u.all
 274 #endif
 275                     )
 276                         return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
 277         }
 278         return ret;
 279 }
 280
 281 static unsigned int
 282 ip_nat_adjust(unsigned int hooknum,
 283               struct sk_buff **pskb,
 284               const struct net_device *in,
 285               const struct net_device *out,
 286               int (*okfn)(struct sk_buff *))
 287 {
 288         struct ip_conntrack *ct;
 289         enum ip_conntrack_info ctinfo;
 290
 291         ct = ip_conntrack_get(*pskb, &ctinfo);
 292         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 293                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
 294                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
 295                         return NF_DROP;
 296         }
 297         return NF_ACCEPT;
 298 }
 299
 300 /* We must be after connection tracking and before packet filtering. */
 301
 302 static struct nf_hook_ops ip_nat_ops[] = {
 303         /* Before packet filtering, change destination */
 304         {
 305                 .hook           = ip_nat_in,
 306                 .owner          = THIS_MODULE,
 307                 .pf             = PF_INET,
 308                 .hooknum        = NF_IP_PRE_ROUTING,
 309                 .priority       = NF_IP_PRI_NAT_DST,
 310         },
 311         /* After packet filtering, change source */
 312         {
 313                 .hook           = ip_nat_out,
 314                 .owner          = THIS_MODULE,
 315                 .pf             = PF_INET,
 316                 .hooknum        = NF_IP_POST_ROUTING,
 317                 .priority       = NF_IP_PRI_NAT_SRC,
 318         },
 319         /* After conntrack, adjust sequence number */
 320         {
 321                 .hook           = ip_nat_adjust,
 322                 .owner          = THIS_MODULE,
 323                 .pf             = PF_INET,
 324                 .hooknum        = NF_IP_POST_ROUTING,
 325                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 326         },
 327         /* Before packet filtering, change destination */
 328         {
 329                 .hook           = ip_nat_local_fn,
 330                 .owner          = THIS_MODULE,
 331                 .pf             = PF_INET,
 332                 .hooknum        = NF_IP_LOCAL_OUT,
 333                 .priority       = NF_IP_PRI_NAT_DST,
 334         },
 335         /* After packet filtering, change source */
 336         {
 337                 .hook           = ip_nat_fn,
 338                 .owner          = THIS_MODULE,
 339                 .pf             = PF_INET,
 340                 .hooknum        = NF_IP_LOCAL_IN,
 341                 .priority       = NF_IP_PRI_NAT_SRC,
 342         },
 343         /* After conntrack, adjust sequence number */
 344         {
 345                 .hook           = ip_nat_adjust,
 346                 .owner          = THIS_MODULE,
 347                 .pf             = PF_INET,
 348                 .hooknum        = NF_IP_LOCAL_IN,
 349                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 350         },
 351 };
 352
 353 static int __init ip_nat_standalone_init(void)
 354 {
 355         int ret = 0;
 356
 357         need_conntrack();
 358
 359 #ifdef CONFIG_XFRM
 360         BUG_ON(ip_nat_decode_session != NULL);
 361         ip_nat_decode_session = nat_decode_session;
 362 #endif
 363         ret = ip_nat_rule_init();
 364         if (ret < 0) {
 365                 printk("ip_nat_init: can't setup rules.\n");
 366                 goto cleanup_decode_session;
 367         }
 368         ret = nf_register_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 369         if (ret < 0) {
 370                 printk("ip_nat_init: can't register hooks.\n");
 371                 goto cleanup_rule_init;
 372         }
 373         return ret;
 374
 375  cleanup_rule_init:
 376         ip_nat_rule_cleanup();
 377  cleanup_decode_session:
 378 #ifdef CONFIG_XFRM
 379         ip_nat_decode_session = NULL;
 380         synchronize_net();
 381 #endif
 382         return ret;
 383 }
 384
 385 static void __exit ip_nat_standalone_fini(void)
 386 {
 387         nf_unregister_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
 388         ip_nat_rule_cleanup();
 389 #ifdef CONFIG_XFRM
 390         ip_nat_decode_session = NULL;
 391         synchronize_net();
 392 #endif
 393 }
 394
 395 module_init(ip_nat_standalone_init);
 396 module_exit(ip_nat_standalone_fini);
 397
 398 MODULE_LICENSE("GPL");