[NETFILTER]: Call POST_ROUTING hook before fragmentation
[safe/jmp/linux-2.6] / 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         return ip_nat_fn(hooknum, pskb, in, out, okfn);
194 }
195
196 static unsigned int
197 ip_nat_local_fn(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         u_int32_t saddr, daddr;
204         unsigned int ret;
205
206         /* root is playing with raw sockets. */
207         if ((*pskb)->len < sizeof(struct iphdr)
208             || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
209                 return NF_ACCEPT;
210
211         saddr = (*pskb)->nh.iph->saddr;
212         daddr = (*pskb)->nh.iph->daddr;
213
214         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
215         if (ret != NF_DROP && ret != NF_STOLEN
216             && ((*pskb)->nh.iph->saddr != saddr
217                 || (*pskb)->nh.iph->daddr != daddr))
218                 return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
219         return ret;
220 }
221
222 static unsigned int
223 ip_nat_adjust(unsigned int hooknum,
224               struct sk_buff **pskb,
225               const struct net_device *in,
226               const struct net_device *out,
227               int (*okfn)(struct sk_buff *))
228 {
229         struct ip_conntrack *ct;
230         enum ip_conntrack_info ctinfo;
231
232         ct = ip_conntrack_get(*pskb, &ctinfo);
233         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
234                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
235                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
236                         return NF_DROP;
237         }
238         return NF_ACCEPT;
239 }
240
241 /* We must be after connection tracking and before packet filtering. */
242
243 /* Before packet filtering, change destination */
244 static struct nf_hook_ops ip_nat_in_ops = {
245         .hook           = ip_nat_in,
246         .owner          = THIS_MODULE,
247         .pf             = PF_INET,
248         .hooknum        = NF_IP_PRE_ROUTING,
249         .priority       = NF_IP_PRI_NAT_DST,
250 };
251
252 /* After packet filtering, change source */
253 static struct nf_hook_ops ip_nat_out_ops = {
254         .hook           = ip_nat_out,
255         .owner          = THIS_MODULE,
256         .pf             = PF_INET,
257         .hooknum        = NF_IP_POST_ROUTING,
258         .priority       = NF_IP_PRI_NAT_SRC,
259 };
260
261 /* After conntrack, adjust sequence number */
262 static struct nf_hook_ops ip_nat_adjust_out_ops = {
263         .hook           = ip_nat_adjust,
264         .owner          = THIS_MODULE,
265         .pf             = PF_INET,
266         .hooknum        = NF_IP_POST_ROUTING,
267         .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
268 };
269
270 /* Before packet filtering, change destination */
271 static struct nf_hook_ops ip_nat_local_out_ops = {
272         .hook           = ip_nat_local_fn,
273         .owner          = THIS_MODULE,
274         .pf             = PF_INET,
275         .hooknum        = NF_IP_LOCAL_OUT,
276         .priority       = NF_IP_PRI_NAT_DST,
277 };
278
279 /* After packet filtering, change source for reply packets of LOCAL_OUT DNAT */
280 static struct nf_hook_ops ip_nat_local_in_ops = {
281         .hook           = ip_nat_fn,
282         .owner          = THIS_MODULE,
283         .pf             = PF_INET,
284         .hooknum        = NF_IP_LOCAL_IN,
285         .priority       = NF_IP_PRI_NAT_SRC,
286 };
287
288 /* After conntrack, adjust sequence number */
289 static struct nf_hook_ops ip_nat_adjust_in_ops = {
290         .hook           = ip_nat_adjust,
291         .owner          = THIS_MODULE,
292         .pf             = PF_INET,
293         .hooknum        = NF_IP_LOCAL_IN,
294         .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
295 };
296
297
298 static int init_or_cleanup(int init)
299 {
300         int ret = 0;
301
302         need_ip_conntrack();
303
304         if (!init) goto cleanup;
305
306         ret = ip_nat_rule_init();
307         if (ret < 0) {
308                 printk("ip_nat_init: can't setup rules.\n");
309                 goto cleanup_nothing;
310         }
311         ret = nf_register_hook(&ip_nat_in_ops);
312         if (ret < 0) {
313                 printk("ip_nat_init: can't register in hook.\n");
314                 goto cleanup_rule_init;
315         }
316         ret = nf_register_hook(&ip_nat_out_ops);
317         if (ret < 0) {
318                 printk("ip_nat_init: can't register out hook.\n");
319                 goto cleanup_inops;
320         }
321         ret = nf_register_hook(&ip_nat_adjust_in_ops);
322         if (ret < 0) {
323                 printk("ip_nat_init: can't register adjust in hook.\n");
324                 goto cleanup_outops;
325         }
326         ret = nf_register_hook(&ip_nat_adjust_out_ops);
327         if (ret < 0) {
328                 printk("ip_nat_init: can't register adjust out hook.\n");
329                 goto cleanup_adjustin_ops;
330         }
331         ret = nf_register_hook(&ip_nat_local_out_ops);
332         if (ret < 0) {
333                 printk("ip_nat_init: can't register local out hook.\n");
334                 goto cleanup_adjustout_ops;;
335         }
336         ret = nf_register_hook(&ip_nat_local_in_ops);
337         if (ret < 0) {
338                 printk("ip_nat_init: can't register local in hook.\n");
339                 goto cleanup_localoutops;
340         }
341         return ret;
342
343  cleanup:
344         nf_unregister_hook(&ip_nat_local_in_ops);
345  cleanup_localoutops:
346         nf_unregister_hook(&ip_nat_local_out_ops);
347  cleanup_adjustout_ops:
348         nf_unregister_hook(&ip_nat_adjust_out_ops);
349  cleanup_adjustin_ops:
350         nf_unregister_hook(&ip_nat_adjust_in_ops);
351  cleanup_outops:
352         nf_unregister_hook(&ip_nat_out_ops);
353  cleanup_inops:
354         nf_unregister_hook(&ip_nat_in_ops);
355  cleanup_rule_init:
356         ip_nat_rule_cleanup();
357  cleanup_nothing:
358         return ret;
359 }
360
361 static int __init init(void)
362 {
363         return init_or_cleanup(1);
364 }
365
366 static void __exit fini(void)
367 {
368         init_or_cleanup(0);
369 }
370
371 module_init(init);
372 module_exit(fini);
373
374 MODULE_LICENSE("GPL");