[NETFILTER]: Use conntrack information to determine if packet was NATed
[linux-2.6.git] / 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         struct ip_conntrack *ct;
166         enum ip_conntrack_info ctinfo;
167         unsigned int ret;
168
169         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
170         if (ret != NF_DROP && ret != NF_STOLEN
171             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
172                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
173
174                 if (ct->tuplehash[dir].tuple.src.ip !=
175                     ct->tuplehash[!dir].tuple.dst.ip) {
176                         dst_release((*pskb)->dst);
177                         (*pskb)->dst = NULL;
178                 }
179         }
180         return ret;
181 }
182
183 static unsigned int
184 ip_nat_out(unsigned int hooknum,
185            struct sk_buff **pskb,
186            const struct net_device *in,
187            const struct net_device *out,
188            int (*okfn)(struct sk_buff *))
189 {
190         /* root is playing with raw sockets. */
191         if ((*pskb)->len < sizeof(struct iphdr)
192             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
193                 return NF_ACCEPT;
194
195         return ip_nat_fn(hooknum, pskb, in, out, okfn);
196 }
197
198 static unsigned int
199 ip_nat_local_fn(unsigned int hooknum,
200                 struct sk_buff **pskb,
201                 const struct net_device *in,
202                 const struct net_device *out,
203                 int (*okfn)(struct sk_buff *))
204 {
205         struct ip_conntrack *ct;
206         enum ip_conntrack_info ctinfo;
207         unsigned int ret;
208
209         /* root is playing with raw sockets. */
210         if ((*pskb)->len < sizeof(struct iphdr)
211             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
212                 return NF_ACCEPT;
213
214         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
215         if (ret != NF_DROP && ret != NF_STOLEN
216             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
217                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
218
219                 if (ct->tuplehash[dir].tuple.dst.ip !=
220                     ct->tuplehash[!dir].tuple.src.ip)
221                         return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
222         }
223         return ret;
224 }
225
226 static unsigned int
227 ip_nat_adjust(unsigned int hooknum,
228               struct sk_buff **pskb,
229               const struct net_device *in,
230               const struct net_device *out,
231               int (*okfn)(struct sk_buff *))
232 {
233         struct ip_conntrack *ct;
234         enum ip_conntrack_info ctinfo;
235
236         ct = ip_conntrack_get(*pskb, &ctinfo);
237         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
238                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
239                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
240                         return NF_DROP;
241         }
242         return NF_ACCEPT;
243 }
244
245 /* We must be after connection tracking and before packet filtering. */
246
247 /* Before packet filtering, change destination */
248 static struct nf_hook_ops ip_nat_in_ops = {
249         .hook           = ip_nat_in,
250         .owner          = THIS_MODULE,
251         .pf             = PF_INET,
252         .hooknum        = NF_IP_PRE_ROUTING,
253         .priority       = NF_IP_PRI_NAT_DST,
254 };
255
256 /* After packet filtering, change source */
257 static struct nf_hook_ops ip_nat_out_ops = {
258         .hook           = ip_nat_out,
259         .owner          = THIS_MODULE,
260         .pf             = PF_INET,
261         .hooknum        = NF_IP_POST_ROUTING,
262         .priority       = NF_IP_PRI_NAT_SRC,
263 };
264
265 /* After conntrack, adjust sequence number */
266 static struct nf_hook_ops ip_nat_adjust_out_ops = {
267         .hook           = ip_nat_adjust,
268         .owner          = THIS_MODULE,
269         .pf             = PF_INET,
270         .hooknum        = NF_IP_POST_ROUTING,
271         .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
272 };
273
274 /* Before packet filtering, change destination */
275 static struct nf_hook_ops ip_nat_local_out_ops = {
276         .hook           = ip_nat_local_fn,
277         .owner          = THIS_MODULE,
278         .pf             = PF_INET,
279         .hooknum        = NF_IP_LOCAL_OUT,
280         .priority       = NF_IP_PRI_NAT_DST,
281 };
282
283 /* After packet filtering, change source for reply packets of LOCAL_OUT DNAT */
284 static struct nf_hook_ops ip_nat_local_in_ops = {
285         .hook           = ip_nat_fn,
286         .owner          = THIS_MODULE,
287         .pf             = PF_INET,
288         .hooknum        = NF_IP_LOCAL_IN,
289         .priority       = NF_IP_PRI_NAT_SRC,
290 };
291
292 /* After conntrack, adjust sequence number */
293 static struct nf_hook_ops ip_nat_adjust_in_ops = {
294         .hook           = ip_nat_adjust,
295         .owner          = THIS_MODULE,
296         .pf             = PF_INET,
297         .hooknum        = NF_IP_LOCAL_IN,
298         .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
299 };
300
301
302 static int init_or_cleanup(int init)
303 {
304         int ret = 0;
305
306         need_ip_conntrack();
307
308         if (!init) goto cleanup;
309
310         ret = ip_nat_rule_init();
311         if (ret < 0) {
312                 printk("ip_nat_init: can't setup rules.\n");
313                 goto cleanup_nothing;
314         }
315         ret = nf_register_hook(&ip_nat_in_ops);
316         if (ret < 0) {
317                 printk("ip_nat_init: can't register in hook.\n");
318                 goto cleanup_rule_init;
319         }
320         ret = nf_register_hook(&ip_nat_out_ops);
321         if (ret < 0) {
322                 printk("ip_nat_init: can't register out hook.\n");
323                 goto cleanup_inops;
324         }
325         ret = nf_register_hook(&ip_nat_adjust_in_ops);
326         if (ret < 0) {
327                 printk("ip_nat_init: can't register adjust in hook.\n");
328                 goto cleanup_outops;
329         }
330         ret = nf_register_hook(&ip_nat_adjust_out_ops);
331         if (ret < 0) {
332                 printk("ip_nat_init: can't register adjust out hook.\n");
333                 goto cleanup_adjustin_ops;
334         }
335         ret = nf_register_hook(&ip_nat_local_out_ops);
336         if (ret < 0) {
337                 printk("ip_nat_init: can't register local out hook.\n");
338                 goto cleanup_adjustout_ops;;
339         }
340         ret = nf_register_hook(&ip_nat_local_in_ops);
341         if (ret < 0) {
342                 printk("ip_nat_init: can't register local in hook.\n");
343                 goto cleanup_localoutops;
344         }
345         return ret;
346
347  cleanup:
348         nf_unregister_hook(&ip_nat_local_in_ops);
349  cleanup_localoutops:
350         nf_unregister_hook(&ip_nat_local_out_ops);
351  cleanup_adjustout_ops:
352         nf_unregister_hook(&ip_nat_adjust_out_ops);
353  cleanup_adjustin_ops:
354         nf_unregister_hook(&ip_nat_adjust_in_ops);
355  cleanup_outops:
356         nf_unregister_hook(&ip_nat_out_ops);
357  cleanup_inops:
358         nf_unregister_hook(&ip_nat_in_ops);
359  cleanup_rule_init:
360         ip_nat_rule_cleanup();
361  cleanup_nothing:
362         return ret;
363 }
364
365 static int __init init(void)
366 {
367         return init_or_cleanup(1);
368 }
369
370 static void __exit fini(void)
371 {
372         init_or_cleanup(0);
373 }
374
375 module_init(init);
376 module_exit(fini);
377
378 MODULE_LICENSE("GPL");