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 static unsigned int
  59 ip_nat_fn(unsigned int hooknum,
  60           struct sk_buff **pskb,
  61           const struct net_device *in,
  62           const struct net_device *out,
  63           int (*okfn)(struct sk_buff *))
  64 {
  65         struct ip_conntrack *ct;
  66         enum ip_conntrack_info ctinfo;
  67         struct ip_nat_info *info;
  68         /* maniptype == SRC for postrouting. */
  69         enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
  70
  71         /* We never see fragments: conntrack defrags on pre-routing
  72            and local-out, and ip_nat_out protects post-routing. */
  73         IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
  74                        & htons(IP_MF|IP_OFFSET)));
  75
  76         /* If we had a hardware checksum before, it's now invalid */
  77         if ((*pskb)->ip_summed == CHECKSUM_HW)
  78                 if (skb_checksum_help(*pskb, (out == NULL)))
  79                         return NF_DROP;
  80
  81         ct = ip_conntrack_get(*pskb, &ctinfo);
  82         /* Can't track?  It's not due to stress, or conntrack would
  83            have dropped it.  Hence it's the user's responsibilty to
  84            packet filter it out, or implement conntrack/NAT for that
  85            protocol. 8) --RR */
  86         if (!ct) {
  87                 /* Exception: ICMP redirect to new connection (not in
  88                    hash table yet).  We must not let this through, in
  89                    case we're doing NAT to the same network. */
  90                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
  91                         struct icmphdr _hdr, *hp;
  92
  93                         hp = skb_header_pointer(*pskb,
  94                                                 (*pskb)->nh.iph->ihl*4,
  95                                                 sizeof(_hdr), &_hdr);
  96                         if (hp != NULL &&
  97                             hp->type == ICMP_REDIRECT)
  98                                 return NF_DROP;
  99                 }
 100                 return NF_ACCEPT;
 101         }
 102
 103         /* Don't try to NAT if this packet is not conntracked */
 104         if (ct == &ip_conntrack_untracked)
 105                 return NF_ACCEPT;
 106
 107         switch (ctinfo) {
 108         case IP_CT_RELATED:
 109         case IP_CT_RELATED+IP_CT_IS_REPLY:
 110                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
 111                         if (!ip_nat_icmp_reply_translation(pskb, ct, maniptype,
 112                                                            CTINFO2DIR(ctinfo)))
 113                                 return NF_DROP;
 114                         else
 115                                 return NF_ACCEPT;
 116                 }
 117                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
 118         case IP_CT_NEW:
 119                 info = &ct->nat.info;
 120
 121                 /* Seen it before?  This can happen for loopback, retrans,
 122                    or local packets.. */
 123                 if (!ip_nat_initialized(ct, maniptype)) {
 124                         unsigned int ret;
 125
 126                         if (unlikely(is_confirmed(ct)))
 127                                 /* NAT module was loaded late */
 128                                 ret = alloc_null_binding_confirmed(ct, info,
 129                                                                    hooknum);
 130                         else if (hooknum == NF_IP_LOCAL_IN)
 131                                 /* LOCAL_IN hook doesn't have a chain!  */
 132                                 ret = alloc_null_binding(ct, info, hooknum);
 133                         else
 134                                 ret = ip_nat_rule_find(pskb, hooknum,
 135                                                        in, out, ct,
 136                                                        info);
 137
 138                         if (ret != NF_ACCEPT) {
 139                                 return ret;
 140                         }
 141                 } else
 142                         DEBUGP("Already setup manip %s for ct %p\n",
 143                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
 144                                ct);
 145                 break;
 146
 147         default:
 148                 /* ESTABLISHED */
 149                 IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
 150                              || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
 151                 info = &ct->nat.info;
 152         }
 153
 154         IP_NF_ASSERT(info);
 155         return ip_nat_packet(ct, ctinfo, hooknum, pskb);
 156 }
 157
 158 static unsigned int
 159 ip_nat_in(unsigned int hooknum,
 160           struct sk_buff **pskb,
 161           const struct net_device *in,
 162           const struct net_device *out,
 163           int (*okfn)(struct sk_buff *))
 164 {
 165         u_int32_t saddr, daddr;
 166         unsigned int ret;
 167
 168         saddr = (*pskb)->nh.iph->saddr;
 169         daddr = (*pskb)->nh.iph->daddr;
 170
 171         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 172         if (ret != NF_DROP && ret != NF_STOLEN
 173             && ((*pskb)->nh.iph->saddr != saddr
 174                 || (*pskb)->nh.iph->daddr != daddr)) {
 175                 dst_release((*pskb)->dst);
 176                 (*pskb)->dst = NULL;
 177         }
 178         return ret;
 179 }
 180
 181 static unsigned int
 182 ip_nat_out(unsigned int hooknum,
 183            struct sk_buff **pskb,
 184            const struct net_device *in,
 185            const struct net_device *out,
 186            int (*okfn)(struct sk_buff *))
 187 {
 188         /* root is playing with raw sockets. */
 189         if ((*pskb)->len < sizeof(struct iphdr)
 190             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 191                 return NF_ACCEPT;
 192
 193         /* We can hit fragment here; forwarded packets get
 194            defragmented by connection tracking coming in, then
 195            fragmented (grr) by the forward code.
 196
 197            In future: If we have nfct != NULL, AND we have NAT
 198            initialized, AND there is no helper, then we can do full
 199            NAPT on the head, and IP-address-only NAT on the rest.
 200
 201            I'm starting to have nightmares about fragments.  */
 202
 203         if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
 204                 *pskb = ip_ct_gather_frags(*pskb, IP_DEFRAG_NAT_OUT);
 205
 206                 if (!*pskb)
 207                         return NF_STOLEN;
 208         }
 209
 210         return ip_nat_fn(hooknum, pskb, in, out, okfn);
 211 }
 212
 213 static unsigned int
 214 ip_nat_local_fn(unsigned int hooknum,
 215                 struct sk_buff **pskb,
 216                 const struct net_device *in,
 217                 const struct net_device *out,
 218                 int (*okfn)(struct sk_buff *))
 219 {
 220         u_int32_t saddr, daddr;
 221         unsigned int ret;
 222
 223         /* root is playing with raw sockets. */
 224         if ((*pskb)->len < sizeof(struct iphdr)
 225             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
 226                 return NF_ACCEPT;
 227
 228         saddr = (*pskb)->nh.iph->saddr;
 229         daddr = (*pskb)->nh.iph->daddr;
 230
 231         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
 232         if (ret != NF_DROP && ret != NF_STOLEN
 233             && ((*pskb)->nh.iph->saddr != saddr
 234                 || (*pskb)->nh.iph->daddr != daddr))
 235                 return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
 236         return ret;
 237 }
 238
 239 static unsigned int
 240 ip_nat_adjust(unsigned int hooknum,
 241               struct sk_buff **pskb,
 242               const struct net_device *in,
 243               const struct net_device *out,
 244               int (*okfn)(struct sk_buff *))
 245 {
 246         struct ip_conntrack *ct;
 247         enum ip_conntrack_info ctinfo;
 248
 249         ct = ip_conntrack_get(*pskb, &ctinfo);
 250         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
 251                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
 252                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
 253                         return NF_DROP;
 254         }
 255         return NF_ACCEPT;
 256 }
 257
 258 /* We must be after connection tracking and before packet filtering. */
 259
 260 /* Before packet filtering, change destination */
 261 static struct nf_hook_ops ip_nat_in_ops = {
 262         .hook           = ip_nat_in,
 263         .owner          = THIS_MODULE,
 264         .pf             = PF_INET,
 265         .hooknum        = NF_IP_PRE_ROUTING,
 266         .priority       = NF_IP_PRI_NAT_DST,
 267 };
 268
 269 /* After packet filtering, change source */
 270 static struct nf_hook_ops ip_nat_out_ops = {
 271         .hook           = ip_nat_out,
 272         .owner          = THIS_MODULE,
 273         .pf             = PF_INET,
 274         .hooknum        = NF_IP_POST_ROUTING,
 275         .priority       = NF_IP_PRI_NAT_SRC,
 276 };
 277
 278 /* After conntrack, adjust sequence number */
 279 static struct nf_hook_ops ip_nat_adjust_out_ops = {
 280         .hook           = ip_nat_adjust,
 281         .owner          = THIS_MODULE,
 282         .pf             = PF_INET,
 283         .hooknum        = NF_IP_POST_ROUTING,
 284         .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 285 };
 286
 287 /* Before packet filtering, change destination */
 288 static struct nf_hook_ops ip_nat_local_out_ops = {
 289         .hook           = ip_nat_local_fn,
 290         .owner          = THIS_MODULE,
 291         .pf             = PF_INET,
 292         .hooknum        = NF_IP_LOCAL_OUT,
 293         .priority       = NF_IP_PRI_NAT_DST,
 294 };
 295
 296 /* After packet filtering, change source for reply packets of LOCAL_OUT DNAT */
 297 static struct nf_hook_ops ip_nat_local_in_ops = {
 298         .hook           = ip_nat_fn,
 299         .owner          = THIS_MODULE,
 300         .pf             = PF_INET,
 301         .hooknum        = NF_IP_LOCAL_IN,
 302         .priority       = NF_IP_PRI_NAT_SRC,
 303 };
 304
 305 /* After conntrack, adjust sequence number */
 306 static struct nf_hook_ops ip_nat_adjust_in_ops = {
 307         .hook           = ip_nat_adjust,
 308         .owner          = THIS_MODULE,
 309         .pf             = PF_INET,
 310         .hooknum        = NF_IP_LOCAL_IN,
 311         .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
 312 };
 313
 314
 315 static int init_or_cleanup(int init)
 316 {
 317         int ret = 0;
 318
 319         need_ip_conntrack();
 320
 321         if (!init) goto cleanup;
 322
 323         ret = ip_nat_rule_init();
 324         if (ret < 0) {
 325                 printk("ip_nat_init: can't setup rules.\n");
 326                 goto cleanup_nothing;
 327         }
 328         ret = nf_register_hook(&ip_nat_in_ops);
 329         if (ret < 0) {
 330                 printk("ip_nat_init: can't register in hook.\n");
 331                 goto cleanup_rule_init;
 332         }
 333         ret = nf_register_hook(&ip_nat_out_ops);
 334         if (ret < 0) {
 335                 printk("ip_nat_init: can't register out hook.\n");
 336                 goto cleanup_inops;
 337         }
 338         ret = nf_register_hook(&ip_nat_adjust_in_ops);
 339         if (ret < 0) {
 340                 printk("ip_nat_init: can't register adjust in hook.\n");
 341                 goto cleanup_outops;
 342         }
 343         ret = nf_register_hook(&ip_nat_adjust_out_ops);
 344         if (ret < 0) {
 345                 printk("ip_nat_init: can't register adjust out hook.\n");
 346                 goto cleanup_adjustin_ops;
 347         }
 348         ret = nf_register_hook(&ip_nat_local_out_ops);
 349         if (ret < 0) {
 350                 printk("ip_nat_init: can't register local out hook.\n");
 351                 goto cleanup_adjustout_ops;;
 352         }
 353         ret = nf_register_hook(&ip_nat_local_in_ops);
 354         if (ret < 0) {
 355                 printk("ip_nat_init: can't register local in hook.\n");
 356                 goto cleanup_localoutops;
 357         }
 358         return ret;
 359
 360  cleanup:
 361         nf_unregister_hook(&ip_nat_local_in_ops);
 362  cleanup_localoutops:
 363         nf_unregister_hook(&ip_nat_local_out_ops);
 364  cleanup_adjustout_ops:
 365         nf_unregister_hook(&ip_nat_adjust_out_ops);
 366  cleanup_adjustin_ops:
 367         nf_unregister_hook(&ip_nat_adjust_in_ops);
 368  cleanup_outops:
 369         nf_unregister_hook(&ip_nat_out_ops);
 370  cleanup_inops:
 371         nf_unregister_hook(&ip_nat_in_ops);
 372  cleanup_rule_init:
 373         ip_nat_rule_cleanup();
 374  cleanup_nothing:
 375         return ret;
 376 }
 377
 378 static int __init init(void)
 379 {
 380         return init_or_cleanup(1);
 381 }
 382
 383 static void __exit fini(void)
 384 {
 385         init_or_cleanup(0);
 386 }
 387
 388 module_init(init);
 389 module_exit(fini);
 390
 391 MODULE_LICENSE("GPL");