]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - net/ipv4/netfilter/ip_nat_standalone.c
[NETFILTER]: reduce netfilter sk_buff enlargement
[mirror_ubuntu-zesty-kernel.git] / net / ipv4 / netfilter / ip_nat_standalone.c
CommitLineData
1da177e4
LT
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
e45b1be8
PM
34#define ASSERT_READ_LOCK(x)
35#define ASSERT_WRITE_LOCK(x)
1da177e4
LT
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
58static unsigned int
59ip_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
1da177e4
LT
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
8b83bc77
HW
103 /* Don't try to NAT if this packet is not conntracked */
104 if (ct == &ip_conntrack_untracked)
105 return NF_ACCEPT;
106
1da177e4
LT
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 (!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 /* LOCAL_IN hook doesn't have a chain! */
127 if (hooknum == NF_IP_LOCAL_IN)
128 ret = alloc_null_binding(ct, info, hooknum);
129 else
130 ret = ip_nat_rule_find(pskb, hooknum,
131 in, out, ct,
132 info);
133
134 if (ret != NF_ACCEPT) {
135 return ret;
136 }
137 } else
138 DEBUGP("Already setup manip %s for ct %p\n",
139 maniptype == IP_NAT_MANIP_SRC ? "SRC" : "DST",
140 ct);
141 break;
142
143 default:
144 /* ESTABLISHED */
145 IP_NF_ASSERT(ctinfo == IP_CT_ESTABLISHED
146 || ctinfo == (IP_CT_ESTABLISHED+IP_CT_IS_REPLY));
147 info = &ct->nat.info;
148 }
149
150 IP_NF_ASSERT(info);
151 return nat_packet(ct, ctinfo, hooknum, pskb);
152}
153
154static unsigned int
155ip_nat_in(unsigned int hooknum,
156 struct sk_buff **pskb,
157 const struct net_device *in,
158 const struct net_device *out,
159 int (*okfn)(struct sk_buff *))
160{
161 u_int32_t saddr, daddr;
162 unsigned int ret;
163
164 saddr = (*pskb)->nh.iph->saddr;
165 daddr = (*pskb)->nh.iph->daddr;
166
167 ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
168 if (ret != NF_DROP && ret != NF_STOLEN
169 && ((*pskb)->nh.iph->saddr != saddr
170 || (*pskb)->nh.iph->daddr != daddr)) {
171 dst_release((*pskb)->dst);
172 (*pskb)->dst = NULL;
173 }
174 return ret;
175}
176
177static unsigned int
178ip_nat_out(unsigned int hooknum,
179 struct sk_buff **pskb,
180 const struct net_device *in,
181 const struct net_device *out,
182 int (*okfn)(struct sk_buff *))
183{
184 /* root is playing with raw sockets. */
185 if ((*pskb)->len < sizeof(struct iphdr)
186 || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
187 return NF_ACCEPT;
188
189 /* We can hit fragment here; forwarded packets get
190 defragmented by connection tracking coming in, then
191 fragmented (grr) by the forward code.
192
193 In future: If we have nfct != NULL, AND we have NAT
194 initialized, AND there is no helper, then we can do full
195 NAPT on the head, and IP-address-only NAT on the rest.
196
197 I'm starting to have nightmares about fragments. */
198
199 if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
200 *pskb = ip_ct_gather_frags(*pskb, IP_DEFRAG_NAT_OUT);
201
202 if (!*pskb)
203 return NF_STOLEN;
204 }
205
206 return ip_nat_fn(hooknum, pskb, in, out, okfn);
207}
208
209static unsigned int
210ip_nat_local_fn(unsigned int hooknum,
211 struct sk_buff **pskb,
212 const struct net_device *in,
213 const struct net_device *out,
214 int (*okfn)(struct sk_buff *))
215{
216 u_int32_t saddr, daddr;
217 unsigned int ret;
218
219 /* root is playing with raw sockets. */
220 if ((*pskb)->len < sizeof(struct iphdr)
221 || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr))
222 return NF_ACCEPT;
223
224 saddr = (*pskb)->nh.iph->saddr;
225 daddr = (*pskb)->nh.iph->daddr;
226
227 ret = ip_nat_fn(hooknum, pskb, in, out, okfn);
228 if (ret != NF_DROP && ret != NF_STOLEN
229 && ((*pskb)->nh.iph->saddr != saddr
230 || (*pskb)->nh.iph->daddr != daddr))
231 return ip_route_me_harder(pskb) == 0 ? ret : NF_DROP;
232 return ret;
233}
234
e281e3ac
PM
235static unsigned int
236ip_nat_adjust(unsigned int hooknum,
237 struct sk_buff **pskb,
238 const struct net_device *in,
239 const struct net_device *out,
240 int (*okfn)(struct sk_buff *))
241{
242 struct ip_conntrack *ct;
243 enum ip_conntrack_info ctinfo;
244
245 ct = ip_conntrack_get(*pskb, &ctinfo);
246 if (ct && test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
247 DEBUGP("ip_nat_standalone: adjusting sequence number\n");
248 if (!ip_nat_seq_adjust(pskb, ct, ctinfo))
249 return NF_DROP;
250 }
251 return NF_ACCEPT;
252}
253
1da177e4
LT
254/* We must be after connection tracking and before packet filtering. */
255
256/* Before packet filtering, change destination */
257static struct nf_hook_ops ip_nat_in_ops = {
258 .hook = ip_nat_in,
259 .owner = THIS_MODULE,
260 .pf = PF_INET,
261 .hooknum = NF_IP_PRE_ROUTING,
262 .priority = NF_IP_PRI_NAT_DST,
263};
264
265/* After packet filtering, change source */
266static struct nf_hook_ops ip_nat_out_ops = {
267 .hook = ip_nat_out,
268 .owner = THIS_MODULE,
269 .pf = PF_INET,
270 .hooknum = NF_IP_POST_ROUTING,
271 .priority = NF_IP_PRI_NAT_SRC,
272};
273
e281e3ac
PM
274/* After conntrack, adjust sequence number */
275static struct nf_hook_ops ip_nat_adjust_out_ops = {
276 .hook = ip_nat_adjust,
277 .owner = THIS_MODULE,
278 .pf = PF_INET,
279 .hooknum = NF_IP_POST_ROUTING,
280 .priority = NF_IP_PRI_NAT_SEQ_ADJUST,
281};
282
1da177e4
LT
283/* Before packet filtering, change destination */
284static struct nf_hook_ops ip_nat_local_out_ops = {
285 .hook = ip_nat_local_fn,
286 .owner = THIS_MODULE,
287 .pf = PF_INET,
288 .hooknum = NF_IP_LOCAL_OUT,
289 .priority = NF_IP_PRI_NAT_DST,
290};
291
292/* After packet filtering, change source for reply packets of LOCAL_OUT DNAT */
293static struct nf_hook_ops ip_nat_local_in_ops = {
294 .hook = ip_nat_fn,
295 .owner = THIS_MODULE,
296 .pf = PF_INET,
297 .hooknum = NF_IP_LOCAL_IN,
298 .priority = NF_IP_PRI_NAT_SRC,
299};
300
e281e3ac
PM
301/* After conntrack, adjust sequence number */
302static struct nf_hook_ops ip_nat_adjust_in_ops = {
303 .hook = ip_nat_adjust,
304 .owner = THIS_MODULE,
305 .pf = PF_INET,
306 .hooknum = NF_IP_LOCAL_IN,
307 .priority = NF_IP_PRI_NAT_SEQ_ADJUST,
308};
309
310
1da177e4
LT
311static int init_or_cleanup(int init)
312{
313 int ret = 0;
314
315 need_ip_conntrack();
316
317 if (!init) goto cleanup;
318
319 ret = ip_nat_rule_init();
320 if (ret < 0) {
321 printk("ip_nat_init: can't setup rules.\n");
322 goto cleanup_nothing;
323 }
324 ret = ip_nat_init();
325 if (ret < 0) {
326 printk("ip_nat_init: can't setup rules.\n");
327 goto cleanup_rule_init;
328 }
329 ret = nf_register_hook(&ip_nat_in_ops);
330 if (ret < 0) {
331 printk("ip_nat_init: can't register in hook.\n");
332 goto cleanup_nat;
333 }
334 ret = nf_register_hook(&ip_nat_out_ops);
335 if (ret < 0) {
336 printk("ip_nat_init: can't register out hook.\n");
337 goto cleanup_inops;
338 }
e281e3ac
PM
339 ret = nf_register_hook(&ip_nat_adjust_in_ops);
340 if (ret < 0) {
341 printk("ip_nat_init: can't register adjust in hook.\n");
342 goto cleanup_outops;
343 }
344 ret = nf_register_hook(&ip_nat_adjust_out_ops);
345 if (ret < 0) {
346 printk("ip_nat_init: can't register adjust out hook.\n");
347 goto cleanup_adjustin_ops;
348 }
1da177e4
LT
349 ret = nf_register_hook(&ip_nat_local_out_ops);
350 if (ret < 0) {
351 printk("ip_nat_init: can't register local out hook.\n");
e281e3ac 352 goto cleanup_adjustout_ops;;
1da177e4
LT
353 }
354 ret = nf_register_hook(&ip_nat_local_in_ops);
355 if (ret < 0) {
356 printk("ip_nat_init: can't register local in hook.\n");
357 goto cleanup_localoutops;
358 }
359 return ret;
360
361 cleanup:
362 nf_unregister_hook(&ip_nat_local_in_ops);
363 cleanup_localoutops:
364 nf_unregister_hook(&ip_nat_local_out_ops);
e281e3ac
PM
365 cleanup_adjustout_ops:
366 nf_unregister_hook(&ip_nat_adjust_out_ops);
367 cleanup_adjustin_ops:
368 nf_unregister_hook(&ip_nat_adjust_in_ops);
1da177e4
LT
369 cleanup_outops:
370 nf_unregister_hook(&ip_nat_out_ops);
371 cleanup_inops:
372 nf_unregister_hook(&ip_nat_in_ops);
373 cleanup_nat:
374 ip_nat_cleanup();
375 cleanup_rule_init:
376 ip_nat_rule_cleanup();
377 cleanup_nothing:
1da177e4
LT
378 return ret;
379}
380
381static int __init init(void)
382{
383 return init_or_cleanup(1);
384}
385
386static void __exit fini(void)
387{
388 init_or_cleanup(0);
389}
390
391module_init(init);
392module_exit(fini);
393
394EXPORT_SYMBOL(ip_nat_setup_info);
395EXPORT_SYMBOL(ip_nat_protocol_register);
396EXPORT_SYMBOL(ip_nat_protocol_unregister);
397EXPORT_SYMBOL(ip_nat_cheat_check);
398EXPORT_SYMBOL(ip_nat_mangle_tcp_packet);
399EXPORT_SYMBOL(ip_nat_mangle_udp_packet);
400EXPORT_SYMBOL(ip_nat_used_tuple);
401EXPORT_SYMBOL(ip_nat_follow_master);
402MODULE_LICENSE("GPL");