[IP]: Introduce ip_hdrlen()
[linux-3.10.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/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 #include <linux/netfilter_ipv4/ip_nat.h>
34 #include <linux/netfilter_ipv4/ip_nat_rule.h>
35 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
36 #include <linux/netfilter_ipv4/ip_nat_core.h>
37 #include <linux/netfilter_ipv4/ip_nat_helper.h>
38 #include <linux/netfilter_ipv4/ip_tables.h>
39 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
40
41 #if 0
42 #define DEBUGP printk
43 #else
44 #define DEBUGP(format, args...)
45 #endif
46
47 #ifdef CONFIG_XFRM
48 static void nat_decode_session(struct sk_buff *skb, struct flowi *fl)
49 {
50         struct ip_conntrack *ct;
51         struct ip_conntrack_tuple *t;
52         enum ip_conntrack_info ctinfo;
53         enum ip_conntrack_dir dir;
54         unsigned long statusbit;
55
56         ct = ip_conntrack_get(skb, &ctinfo);
57         if (ct == NULL)
58                 return;
59         dir = CTINFO2DIR(ctinfo);
60         t = &ct->tuplehash[dir].tuple;
61
62         if (dir == IP_CT_DIR_ORIGINAL)
63                 statusbit = IPS_DST_NAT;
64         else
65                 statusbit = IPS_SRC_NAT;
66
67         if (ct->status & statusbit) {
68                 fl->fl4_dst = t->dst.ip;
69                 if (t->dst.protonum == IPPROTO_TCP ||
70                     t->dst.protonum == IPPROTO_UDP)
71                         fl->fl_ip_dport = t->dst.u.tcp.port;
72         }
73
74         statusbit ^= IPS_NAT_MASK;
75
76         if (ct->status & statusbit) {
77                 fl->fl4_src = t->src.ip;
78                 if (t->dst.protonum == IPPROTO_TCP ||
79                     t->dst.protonum == IPPROTO_UDP)
80                         fl->fl_ip_sport = t->src.u.tcp.port;
81         }
82 }
83 #endif
84
85 static unsigned int
86 ip_nat_fn(unsigned int hooknum,
87           struct sk_buff **pskb,
88           const struct net_device *in,
89           const struct net_device *out,
90           int (*okfn)(struct sk_buff *))
91 {
92         struct ip_conntrack *ct;
93         enum ip_conntrack_info ctinfo;
94         struct ip_nat_info *info;
95         /* maniptype == SRC for postrouting. */
96         enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
97
98         /* We never see fragments: conntrack defrags on pre-routing
99            and local-out, and ip_nat_out protects post-routing. */
100         IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
101                        & htons(IP_MF|IP_OFFSET)));
102
103         ct = ip_conntrack_get(*pskb, &ctinfo);
104         /* Can't track?  It's not due to stress, or conntrack would
105            have dropped it.  Hence it's the user's responsibilty to
106            packet filter it out, or implement conntrack/NAT for that
107            protocol. 8) --RR */
108         if (!ct) {
109                 /* Exception: ICMP redirect to new connection (not in
110                    hash table yet).  We must not let this through, in
111                    case we're doing NAT to the same network. */
112                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
113                         struct icmphdr _hdr, *hp;
114
115                         hp = skb_header_pointer(*pskb, ip_hdrlen(*pskb),
116                                                 sizeof(_hdr), &_hdr);
117                         if (hp != NULL &&
118                             hp->type == ICMP_REDIRECT)
119                                 return NF_DROP;
120                 }
121                 return NF_ACCEPT;
122         }
123
124         /* Don't try to NAT if this packet is not conntracked */
125         if (ct == &ip_conntrack_untracked)
126                 return NF_ACCEPT;
127
128         switch (ctinfo) {
129         case IP_CT_RELATED:
130         case IP_CT_RELATED+IP_CT_IS_REPLY:
131                 if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
132                         if (!ip_nat_icmp_reply_translation(ct, ctinfo,
133                                                            hooknum, pskb))
134                                 return NF_DROP;
135                         else
136                                 return NF_ACCEPT;
137                 }
138                 /* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
139         case IP_CT_NEW:
140                 info = &ct->nat.info;
141
142                 /* Seen it before?  This can happen for loopback, retrans,
143                    or local packets.. */
144                 if (!ip_nat_initialized(ct, maniptype)) {
145                         unsigned int ret;
146
147                         if (unlikely(is_confirmed(ct)))
148                                 /* NAT module was loaded late */
149                                 ret = alloc_null_binding_confirmed(ct, info,
150                                                                    hooknum);
151                         else if (hooknum == NF_IP_LOCAL_IN)
152                                 /* LOCAL_IN hook doesn't have a chain!  */
153                                 ret = alloc_null_binding(ct, info, hooknum);
154                         else
155                                 ret = ip_nat_rule_find(pskb, hooknum,
156                                                        in, out, ct,
157                                                        info);
158
159                         if (ret != NF_ACCEPT) {
160                                 return ret;
161                         }
162                 } else
163                         DEBUGP("Already setup manip %s for ct %p\n",
164                                maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
165                                ct);
166                 break;
167
168         default:
169                 /* ESTABLISHED */
170                 IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
171                              || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
172                 info = &ct->nat.info;
173         }
174
175         IP_NF_ASSERT(info);
176         return ip_nat_packet(ct, ctinfo, hooknum, pskb);
177 }
178
179 static unsigned int
180 ip_nat_in(unsigned int hooknum,
181           struct sk_buff **pskb,
182           const struct net_device *in,
183           const struct net_device *out,
184           int (*okfn)(struct sk_buff *))
185 {
186         unsigned int ret;
187         __be32 daddr = (*pskb)->nh.iph->daddr;
188
189         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
190         if (ret != NF_DROP && ret != NF_STOLEN
191             && daddr != (*pskb)->nh.iph->daddr) {
192                 dst_release((*pskb)->dst);
193                 (*pskb)->dst = NULL;
194         }
195         return ret;
196 }
197
198 static unsigned int
199 ip_nat_out(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 #ifdef CONFIG_XFRM
206         struct ip_conntrack *ct;
207         enum ip_conntrack_info ctinfo;
208 #endif
209         unsigned int ret;
210
211         /* root is playing with raw sockets. */
212         if ((*pskb)->len < sizeof(struct iphdr)
213             || ip_hdrlen(*pskb) < sizeof(struct iphdr))
214                 return NF_ACCEPT;
215
216         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
217 #ifdef CONFIG_XFRM
218         if (ret != NF_DROP && ret != NF_STOLEN
219             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
220                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
221
222                 if (ct->tuplehash[dir].tuple.src.ip !=
223                     ct->tuplehash[!dir].tuple.dst.ip
224                     || ct->tuplehash[dir].tuple.src.u.all !=
225                        ct->tuplehash[!dir].tuple.dst.u.all
226                     )
227                         return ip_xfrm_me_harder(pskb) == 0 ? ret : NF_DROP;
228         }
229 #endif
230         return ret;
231 }
232
233 static unsigned int
234 ip_nat_local_fn(unsigned int hooknum,
235                 struct sk_buff **pskb,
236                 const struct net_device *in,
237                 const struct net_device *out,
238                 int (*okfn)(struct sk_buff *))
239 {
240         struct ip_conntrack *ct;
241         enum ip_conntrack_info ctinfo;
242         unsigned int ret;
243
244         /* root is playing with raw sockets. */
245         if ((*pskb)->len < sizeof(struct iphdr)
246             || ip_hdrlen(*pskb) < sizeof(struct iphdr))
247                 return NF_ACCEPT;
248
249         ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
250         if (ret != NF_DROP && ret != NF_STOLEN
251             && (ct = ip_conntrack_get(*pskb, &ctinfo)) != NULL) {
252                 enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo);
253
254                 if (ct->tuplehash[dir].tuple.dst.ip !=
255                     ct->tuplehash[!dir].tuple.src.ip) {
256                         if (ip_route_me_harder(pskb, RTN_UNSPEC))
257                                 ret = NF_DROP;
258                 }
259 #ifdef CONFIG_XFRM
260                 else if (ct->tuplehash[dir].tuple.dst.u.all !=
261                          ct->tuplehash[!dir].tuple.src.u.all)
262                         if (ip_xfrm_me_harder(pskb))
263                                 ret = NF_DROP;
264 #endif
265
266         }
267         return ret;
268 }
269
270 static unsigned int
271 ip_nat_adjust(unsigned int hooknum,
272               struct sk_buff **pskb,
273               const struct net_device *in,
274               const struct net_device *out,
275               int (*okfn)(struct sk_buff *))
276 {
277         struct ip_conntrack *ct;
278         enum ip_conntrack_info ctinfo;
279
280         ct = ip_conntrack_get(*pskb, &ctinfo);
281         if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
282                 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
283                 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
284                         return NF_DROP;
285         }
286         return NF_ACCEPT;
287 }
288
289 /* We must be after connection tracking and before packet filtering. */
290
291 static struct nf_hook_ops ip_nat_ops[] = {
292         /* Before packet filtering, change destination */
293         {
294                 .hook           = ip_nat_in,
295                 .owner          = THIS_MODULE,
296                 .pf             = PF_INET,
297                 .hooknum        = NF_IP_PRE_ROUTING,
298                 .priority       = NF_IP_PRI_NAT_DST,
299         },
300         /* After packet filtering, change source */
301         {
302                 .hook           = ip_nat_out,
303                 .owner          = THIS_MODULE,
304                 .pf             = PF_INET,
305                 .hooknum        = NF_IP_POST_ROUTING,
306                 .priority       = NF_IP_PRI_NAT_SRC,
307         },
308         /* After conntrack, adjust sequence number */
309         {
310                 .hook           = ip_nat_adjust,
311                 .owner          = THIS_MODULE,
312                 .pf             = PF_INET,
313                 .hooknum        = NF_IP_POST_ROUTING,
314                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
315         },
316         /* Before packet filtering, change destination */
317         {
318                 .hook           = ip_nat_local_fn,
319                 .owner          = THIS_MODULE,
320                 .pf             = PF_INET,
321                 .hooknum        = NF_IP_LOCAL_OUT,
322                 .priority       = NF_IP_PRI_NAT_DST,
323         },
324         /* After packet filtering, change source */
325         {
326                 .hook           = ip_nat_fn,
327                 .owner          = THIS_MODULE,
328                 .pf             = PF_INET,
329                 .hooknum        = NF_IP_LOCAL_IN,
330                 .priority       = NF_IP_PRI_NAT_SRC,
331         },
332         /* After conntrack, adjust sequence number */
333         {
334                 .hook           = ip_nat_adjust,
335                 .owner          = THIS_MODULE,
336                 .pf             = PF_INET,
337                 .hooknum        = NF_IP_LOCAL_IN,
338                 .priority       = NF_IP_PRI_NAT_SEQ_ADJUST,
339         },
340 };
341
342 static int __init ip_nat_standalone_init(void)
343 {
344         int ret = 0;
345
346         need_conntrack();
347
348 #ifdef CONFIG_XFRM
349         BUG_ON(ip_nat_decode_session != NULL);
350         ip_nat_decode_session = nat_decode_session;
351 #endif
352         ret = ip_nat_rule_init();
353         if (ret < 0) {
354                 printk("ip_nat_init: can't setup rules.\n");
355                 goto cleanup_decode_session;
356         }
357         ret = nf_register_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
358         if (ret < 0) {
359                 printk("ip_nat_init: can't register hooks.\n");
360                 goto cleanup_rule_init;
361         }
362         return ret;
363
364  cleanup_rule_init:
365         ip_nat_rule_cleanup();
366  cleanup_decode_session:
367 #ifdef CONFIG_XFRM
368         ip_nat_decode_session = NULL;
369         synchronize_net();
370 #endif
371         return ret;
372 }
373
374 static void __exit ip_nat_standalone_fini(void)
375 {
376         nf_unregister_hooks(ip_nat_ops, ARRAY_SIZE(ip_nat_ops));
377         ip_nat_rule_cleanup();
378 #ifdef CONFIG_XFRM
379         ip_nat_decode_session = NULL;
380         synchronize_net();
381 #endif
382 }
383
384 module_init(ip_nat_standalone_init);
385 module_exit(ip_nat_standalone_fini);
386
387 MODULE_LICENSE("GPL");