1 |
1275 |
phoenix |
/* 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 Paul `Rusty' Russell. Licenced under the GNU General
|
8 |
|
|
* Public Licence.
|
9 |
|
|
*
|
10 |
|
|
* 23 Apr 2001: Harald Welte <laforge@gnumonks.org>
|
11 |
|
|
* - new API and handling of conntrack/nat helpers
|
12 |
|
|
* - now capable of multiple expectations for one master
|
13 |
|
|
* */
|
14 |
|
|
|
15 |
|
|
#include <linux/config.h>
|
16 |
|
|
#include <linux/types.h>
|
17 |
|
|
#include <linux/ip.h>
|
18 |
|
|
#include <linux/netfilter.h>
|
19 |
|
|
#include <linux/netfilter_ipv4.h>
|
20 |
|
|
#include <linux/module.h>
|
21 |
|
|
#include <linux/skbuff.h>
|
22 |
|
|
#include <linux/proc_fs.h>
|
23 |
|
|
#include <net/checksum.h>
|
24 |
|
|
#include <linux/spinlock.h>
|
25 |
|
|
#include <linux/version.h>
|
26 |
|
|
#include <linux/brlock.h>
|
27 |
|
|
|
28 |
|
|
#define ASSERT_READ_LOCK(x) MUST_BE_READ_LOCKED(&ip_nat_lock)
|
29 |
|
|
#define ASSERT_WRITE_LOCK(x) MUST_BE_WRITE_LOCKED(&ip_nat_lock)
|
30 |
|
|
|
31 |
|
|
#include <linux/netfilter_ipv4/ip_nat.h>
|
32 |
|
|
#include <linux/netfilter_ipv4/ip_nat_rule.h>
|
33 |
|
|
#include <linux/netfilter_ipv4/ip_nat_protocol.h>
|
34 |
|
|
#include <linux/netfilter_ipv4/ip_nat_core.h>
|
35 |
|
|
#include <linux/netfilter_ipv4/ip_nat_helper.h>
|
36 |
|
|
#include <linux/netfilter_ipv4/ip_tables.h>
|
37 |
|
|
#include <linux/netfilter_ipv4/ip_conntrack_core.h>
|
38 |
|
|
#include <linux/netfilter_ipv4/listhelp.h>
|
39 |
|
|
|
40 |
|
|
#if 0
|
41 |
|
|
#define DEBUGP printk
|
42 |
|
|
#else
|
43 |
|
|
#define DEBUGP(format, args...)
|
44 |
|
|
#endif
|
45 |
|
|
|
46 |
|
|
#define HOOKNAME(hooknum) ((hooknum) == NF_IP_POST_ROUTING ? "POST_ROUTING" \
|
47 |
|
|
: ((hooknum) == NF_IP_PRE_ROUTING ? "PRE_ROUTING" \
|
48 |
|
|
: ((hooknum) == NF_IP_LOCAL_OUT ? "LOCAL_OUT" \
|
49 |
|
|
: ((hooknum) == NF_IP_LOCAL_IN ? "LOCAL_IN" \
|
50 |
|
|
: "*ERROR*")))
|
51 |
|
|
|
52 |
|
|
static inline int call_expect(struct ip_conntrack *master,
|
53 |
|
|
struct sk_buff **pskb,
|
54 |
|
|
unsigned int hooknum,
|
55 |
|
|
struct ip_conntrack *ct,
|
56 |
|
|
struct ip_nat_info *info)
|
57 |
|
|
{
|
58 |
|
|
return master->nat.info.helper->expect(pskb, hooknum, ct, info);
|
59 |
|
|
}
|
60 |
|
|
|
61 |
|
|
static unsigned int
|
62 |
|
|
ip_nat_fn(unsigned int hooknum,
|
63 |
|
|
struct sk_buff **pskb,
|
64 |
|
|
const struct net_device *in,
|
65 |
|
|
const struct net_device *out,
|
66 |
|
|
int (*okfn)(struct sk_buff *))
|
67 |
|
|
{
|
68 |
|
|
struct ip_conntrack *ct;
|
69 |
|
|
enum ip_conntrack_info ctinfo;
|
70 |
|
|
struct ip_nat_info *info;
|
71 |
|
|
/* maniptype == SRC for postrouting. */
|
72 |
|
|
enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum);
|
73 |
|
|
|
74 |
|
|
/* We never see fragments: conntrack defrags on pre-routing
|
75 |
|
|
and local-out, and ip_nat_out protects post-routing. */
|
76 |
|
|
IP_NF_ASSERT(!((*pskb)->nh.iph->frag_off
|
77 |
|
|
& htons(IP_MF|IP_OFFSET)));
|
78 |
|
|
|
79 |
|
|
(*pskb)->nfcache |= NFC_UNKNOWN;
|
80 |
|
|
|
81 |
|
|
/* If we had a hardware checksum before, it's now invalid */
|
82 |
|
|
if ((*pskb)->ip_summed == CHECKSUM_HW)
|
83 |
|
|
(*pskb)->ip_summed = CHECKSUM_NONE;
|
84 |
|
|
|
85 |
|
|
ct = ip_conntrack_get(*pskb, &ctinfo);
|
86 |
|
|
/* Can't track? It's not due to stress, or conntrack would
|
87 |
|
|
have dropped it. Hence it's the user's responsibilty to
|
88 |
|
|
packet filter it out, or implement conntrack/NAT for that
|
89 |
|
|
protocol. 8) --RR */
|
90 |
|
|
if (!ct) {
|
91 |
|
|
/* Exception: ICMP redirect to new connection (not in
|
92 |
|
|
hash table yet). We must not let this through, in
|
93 |
|
|
case we're doing NAT to the same network. */
|
94 |
|
|
struct iphdr *iph = (*pskb)->nh.iph;
|
95 |
|
|
struct icmphdr *hdr = (struct icmphdr *)
|
96 |
|
|
((u_int32_t *)iph + iph->ihl);
|
97 |
|
|
if (iph->protocol == IPPROTO_ICMP
|
98 |
|
|
&& hdr->type == ICMP_REDIRECT)
|
99 |
|
|
return NF_DROP;
|
100 |
|
|
return NF_ACCEPT;
|
101 |
|
|
}
|
102 |
|
|
|
103 |
|
|
switch (ctinfo) {
|
104 |
|
|
case IP_CT_RELATED:
|
105 |
|
|
case IP_CT_RELATED+IP_CT_IS_REPLY:
|
106 |
|
|
if ((*pskb)->nh.iph->protocol == IPPROTO_ICMP) {
|
107 |
|
|
return icmp_reply_translation(*pskb, ct, hooknum,
|
108 |
|
|
CTINFO2DIR(ctinfo));
|
109 |
|
|
}
|
110 |
|
|
/* Fall thru... (Only ICMPs can be IP_CT_IS_REPLY) */
|
111 |
|
|
case IP_CT_NEW:
|
112 |
|
|
info = &ct->nat.info;
|
113 |
|
|
|
114 |
|
|
WRITE_LOCK(&ip_nat_lock);
|
115 |
|
|
/* Seen it before? This can happen for loopback, retrans,
|
116 |
|
|
or local packets.. */
|
117 |
|
|
if (!(info->initialized & (1 << maniptype))
|
118 |
|
|
#ifndef CONFIG_IP_NF_NAT_LOCAL
|
119 |
|
|
/* If this session has already been confirmed we must not
|
120 |
|
|
* touch it again even if there is no mapping set up.
|
121 |
|
|
* Can only happen on local->local traffic with
|
122 |
|
|
* CONFIG_IP_NF_NAT_LOCAL disabled.
|
123 |
|
|
*/
|
124 |
|
|
&& !(ct->status & IPS_CONFIRMED)
|
125 |
|
|
#endif
|
126 |
|
|
) {
|
127 |
|
|
unsigned int ret;
|
128 |
|
|
|
129 |
|
|
if (ct->master
|
130 |
|
|
&& master_ct(ct)->nat.info.helper
|
131 |
|
|
&& master_ct(ct)->nat.info.helper->expect) {
|
132 |
|
|
ret = call_expect(master_ct(ct), pskb,
|
133 |
|
|
hooknum, ct, info);
|
134 |
|
|
} else {
|
135 |
|
|
#ifdef CONFIG_IP_NF_NAT_LOCAL
|
136 |
|
|
/* LOCAL_IN hook doesn't have a chain! */
|
137 |
|
|
if (hooknum == NF_IP_LOCAL_IN)
|
138 |
|
|
ret = alloc_null_binding(ct, info,
|
139 |
|
|
hooknum);
|
140 |
|
|
else
|
141 |
|
|
#endif
|
142 |
|
|
ret = ip_nat_rule_find(pskb, hooknum, in, out,
|
143 |
|
|
ct, info);
|
144 |
|
|
}
|
145 |
|
|
|
146 |
|
|
if (ret != NF_ACCEPT) {
|
147 |
|
|
WRITE_UNLOCK(&ip_nat_lock);
|
148 |
|
|
return ret;
|
149 |
|
|
}
|
150 |
|
|
} else
|
151 |
|
|
DEBUGP("Already setup manip %s for ct %p\n",
|
152 |
|
|
maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
|
153 |
|
|
ct);
|
154 |
|
|
WRITE_UNLOCK(&ip_nat_lock);
|
155 |
|
|
break;
|
156 |
|
|
|
157 |
|
|
default:
|
158 |
|
|
/* ESTABLISHED */
|
159 |
|
|
IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
|
160 |
|
|
|| ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
|
161 |
|
|
info = &ct->nat.info;
|
162 |
|
|
}
|
163 |
|
|
|
164 |
|
|
IP_NF_ASSERT(info);
|
165 |
|
|
return do_bindings(ct, ctinfo, info, hooknum, pskb);
|
166 |
|
|
}
|
167 |
|
|
|
168 |
|
|
static unsigned int
|
169 |
|
|
ip_nat_out(unsigned int hooknum,
|
170 |
|
|
struct sk_buff **pskb,
|
171 |
|
|
const struct net_device *in,
|
172 |
|
|
const struct net_device *out,
|
173 |
|
|
int (*okfn)(struct sk_buff *))
|
174 |
|
|
{
|
175 |
|
|
/* root is playing with raw sockets. */
|
176 |
|
|
if ((*pskb)->len < sizeof(struct iphdr)
|
177 |
|
|
|| (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
|
178 |
|
|
return NF_ACCEPT;
|
179 |
|
|
|
180 |
|
|
/* We can hit fragment here; forwarded packets get
|
181 |
|
|
defragmented by connection tracking coming in, then
|
182 |
|
|
fragmented (grr) by the forward code.
|
183 |
|
|
|
184 |
|
|
In future: If we have nfct != NULL, AND we have NAT
|
185 |
|
|
initialized, AND there is no helper, then we can do full
|
186 |
|
|
NAPT on the head, and IP-address-only NAT on the rest.
|
187 |
|
|
|
188 |
|
|
I'm starting to have nightmares about fragments. */
|
189 |
|
|
|
190 |
|
|
if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
|
191 |
|
|
*pskb = ip_ct_gather_frags(*pskb);
|
192 |
|
|
|
193 |
|
|
if (!*pskb)
|
194 |
|
|
return NF_STOLEN;
|
195 |
|
|
}
|
196 |
|
|
|
197 |
|
|
return ip_nat_fn(hooknum, pskb, in, out, okfn);
|
198 |
|
|
}
|
199 |
|
|
|
200 |
|
|
#ifdef CONFIG_IP_NF_NAT_LOCAL
|
201 |
|
|
static unsigned int
|
202 |
|
|
ip_nat_local_fn(unsigned int hooknum,
|
203 |
|
|
struct sk_buff **pskb,
|
204 |
|
|
const struct net_device *in,
|
205 |
|
|
const struct net_device *out,
|
206 |
|
|
int (*okfn)(struct sk_buff *))
|
207 |
|
|
{
|
208 |
|
|
u_int32_t saddr, daddr;
|
209 |
|
|
unsigned int ret;
|
210 |
|
|
|
211 |
|
|
/* root is playing with raw sockets. */
|
212 |
|
|
if ((*pskb)->len < sizeof(struct iphdr)
|
213 |
|
|
|| (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
|
214 |
|
|
return NF_ACCEPT;
|
215 |
|
|
|
216 |
|
|
saddr = (*pskb)->nh.iph->saddr;
|
217 |
|
|
daddr = (*pskb)->nh.iph->daddr;
|
218 |
|
|
|
219 |
|
|
ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
|
220 |
|
|
if (ret != NF_DROP && ret != NF_STOLEN
|
221 |
|
|
&& ((*pskb)->nh.iph->saddr != saddr
|
222 |
|
|
|| (*pskb)->nh.iph->daddr != daddr))
|
223 |
|
|
return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
|
224 |
|
|
return ret;
|
225 |
|
|
}
|
226 |
|
|
#endif
|
227 |
|
|
|
228 |
|
|
/* We must be after connection tracking and before packet filtering. */
|
229 |
|
|
|
230 |
|
|
/* Before packet filtering, change destination */
|
231 |
|
|
static struct nf_hook_ops ip_nat_in_ops
|
232 |
|
|
= { { NULL, NULL }, ip_nat_fn, PF_INET, NF_IP_PRE_ROUTING, NF_IP_PRI_NAT_DST };
|
233 |
|
|
/* After packet filtering, change source */
|
234 |
|
|
static struct nf_hook_ops ip_nat_out_ops
|
235 |
|
|
= { { NULL, NULL }, ip_nat_out, PF_INET, NF_IP_POST_ROUTING, NF_IP_PRI_NAT_SRC};
|
236 |
|
|
|
237 |
|
|
#ifdef CONFIG_IP_NF_NAT_LOCAL
|
238 |
|
|
/* Before packet filtering, change destination */
|
239 |
|
|
static struct nf_hook_ops ip_nat_local_out_ops
|
240 |
|
|
= { { NULL, NULL }, ip_nat_local_fn, PF_INET, NF_IP_LOCAL_OUT, NF_IP_PRI_NAT_DST };
|
241 |
|
|
/* After packet filtering, change source for reply packets of LOCAL_OUT DNAT */
|
242 |
|
|
static struct nf_hook_ops ip_nat_local_in_ops
|
243 |
|
|
= { { NULL, NULL }, ip_nat_fn, PF_INET, NF_IP_LOCAL_IN, NF_IP_PRI_NAT_SRC };
|
244 |
|
|
#endif
|
245 |
|
|
|
246 |
|
|
/* Protocol registration. */
|
247 |
|
|
int ip_nat_protocol_register(struct ip_nat_protocol *proto)
|
248 |
|
|
{
|
249 |
|
|
int ret = 0;
|
250 |
|
|
struct list_head *i;
|
251 |
|
|
|
252 |
|
|
WRITE_LOCK(&ip_nat_lock);
|
253 |
|
|
for (i = protos.next; i != &protos; i = i->next) {
|
254 |
|
|
if (((struct ip_nat_protocol *)i)->protonum
|
255 |
|
|
== proto->protonum) {
|
256 |
|
|
ret = -EBUSY;
|
257 |
|
|
goto out;
|
258 |
|
|
}
|
259 |
|
|
}
|
260 |
|
|
|
261 |
|
|
list_prepend(&protos, proto);
|
262 |
|
|
MOD_INC_USE_COUNT;
|
263 |
|
|
|
264 |
|
|
out:
|
265 |
|
|
WRITE_UNLOCK(&ip_nat_lock);
|
266 |
|
|
return ret;
|
267 |
|
|
}
|
268 |
|
|
|
269 |
|
|
/* Noone stores the protocol anywhere; simply delete it. */
|
270 |
|
|
void ip_nat_protocol_unregister(struct ip_nat_protocol *proto)
|
271 |
|
|
{
|
272 |
|
|
WRITE_LOCK(&ip_nat_lock);
|
273 |
|
|
LIST_DELETE(&protos, proto);
|
274 |
|
|
WRITE_UNLOCK(&ip_nat_lock);
|
275 |
|
|
|
276 |
|
|
/* Someone could be still looking at the proto in a bh. */
|
277 |
|
|
br_write_lock_bh(BR_NETPROTO_LOCK);
|
278 |
|
|
br_write_unlock_bh(BR_NETPROTO_LOCK);
|
279 |
|
|
|
280 |
|
|
MOD_DEC_USE_COUNT;
|
281 |
|
|
}
|
282 |
|
|
|
283 |
|
|
static int init_or_cleanup(int init)
|
284 |
|
|
{
|
285 |
|
|
int ret = 0;
|
286 |
|
|
|
287 |
|
|
if (!init) goto cleanup;
|
288 |
|
|
|
289 |
|
|
ret = ip_nat_rule_init();
|
290 |
|
|
if (ret < 0) {
|
291 |
|
|
printk("ip_nat_init: can't setup rules.\n");
|
292 |
|
|
goto cleanup_nothing;
|
293 |
|
|
}
|
294 |
|
|
ret = ip_nat_init();
|
295 |
|
|
if (ret < 0) {
|
296 |
|
|
printk("ip_nat_init: can't setup rules.\n");
|
297 |
|
|
goto cleanup_rule_init;
|
298 |
|
|
}
|
299 |
|
|
ret = nf_register_hook(&ip_nat_in_ops);
|
300 |
|
|
if (ret < 0) {
|
301 |
|
|
printk("ip_nat_init: can't register in hook.\n");
|
302 |
|
|
goto cleanup_nat;
|
303 |
|
|
}
|
304 |
|
|
ret = nf_register_hook(&ip_nat_out_ops);
|
305 |
|
|
if (ret < 0) {
|
306 |
|
|
printk("ip_nat_init: can't register out hook.\n");
|
307 |
|
|
goto cleanup_inops;
|
308 |
|
|
}
|
309 |
|
|
#ifdef CONFIG_IP_NF_NAT_LOCAL
|
310 |
|
|
ret = nf_register_hook(&ip_nat_local_out_ops);
|
311 |
|
|
if (ret < 0) {
|
312 |
|
|
printk("ip_nat_init: can't register local out hook.\n");
|
313 |
|
|
goto cleanup_outops;
|
314 |
|
|
}
|
315 |
|
|
ret = nf_register_hook(&ip_nat_local_in_ops);
|
316 |
|
|
if (ret < 0) {
|
317 |
|
|
printk("ip_nat_init: can't register local in hook.\n");
|
318 |
|
|
goto cleanup_localoutops;
|
319 |
|
|
}
|
320 |
|
|
#endif
|
321 |
|
|
return ret;
|
322 |
|
|
|
323 |
|
|
cleanup:
|
324 |
|
|
#ifdef CONFIG_IP_NF_NAT_LOCAL
|
325 |
|
|
nf_unregister_hook(&ip_nat_local_in_ops);
|
326 |
|
|
cleanup_localoutops:
|
327 |
|
|
nf_unregister_hook(&ip_nat_local_out_ops);
|
328 |
|
|
cleanup_outops:
|
329 |
|
|
#endif
|
330 |
|
|
nf_unregister_hook(&ip_nat_out_ops);
|
331 |
|
|
cleanup_inops:
|
332 |
|
|
nf_unregister_hook(&ip_nat_in_ops);
|
333 |
|
|
cleanup_nat:
|
334 |
|
|
ip_nat_cleanup();
|
335 |
|
|
cleanup_rule_init:
|
336 |
|
|
ip_nat_rule_cleanup();
|
337 |
|
|
cleanup_nothing:
|
338 |
|
|
MUST_BE_READ_WRITE_UNLOCKED(&ip_nat_lock);
|
339 |
|
|
return ret;
|
340 |
|
|
}
|
341 |
|
|
|
342 |
|
|
static int __init init(void)
|
343 |
|
|
{
|
344 |
|
|
return init_or_cleanup(1);
|
345 |
|
|
}
|
346 |
|
|
|
347 |
|
|
static void __exit fini(void)
|
348 |
|
|
{
|
349 |
|
|
init_or_cleanup(0);
|
350 |
|
|
}
|
351 |
|
|
|
352 |
|
|
module_init(init);
|
353 |
|
|
module_exit(fini);
|
354 |
|
|
|
355 |
|
|
EXPORT_SYMBOL(ip_nat_setup_info);
|
356 |
|
|
EXPORT_SYMBOL(ip_nat_protocol_register);
|
357 |
|
|
EXPORT_SYMBOL(ip_nat_protocol_unregister);
|
358 |
|
|
EXPORT_SYMBOL(ip_nat_helper_register);
|
359 |
|
|
EXPORT_SYMBOL(ip_nat_helper_unregister);
|
360 |
|
|
EXPORT_SYMBOL(ip_nat_cheat_check);
|
361 |
|
|
EXPORT_SYMBOL(ip_nat_mangle_tcp_packet);
|
362 |
|
|
EXPORT_SYMBOL(ip_nat_mangle_udp_packet);
|
363 |
|
|
EXPORT_SYMBOL(ip_nat_used_tuple);
|
364 |
|
|
MODULE_LICENSE("GPL");
|