Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* NAT for netfilter; shared with compatibility layer. */ |
2 | ||
3 | /* (C) 1999-2001 Paul `Rusty' Russell | |
4 | * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org> | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License version 2 as | |
8 | * published by the Free Software Foundation. | |
9 | */ | |
10 | ||
11 | #include <linux/module.h> | |
12 | #include <linux/types.h> | |
13 | #include <linux/timer.h> | |
14 | #include <linux/skbuff.h> | |
15 | #include <linux/netfilter_ipv4.h> | |
16 | #include <linux/vmalloc.h> | |
17 | #include <net/checksum.h> | |
18 | #include <net/icmp.h> | |
19 | #include <net/ip.h> | |
20 | #include <net/tcp.h> /* For tcp_prot in getorigdst */ | |
21 | #include <linux/icmp.h> | |
22 | #include <linux/udp.h> | |
23 | #include <linux/jhash.h> | |
24 | ||
1da177e4 LT |
25 | #include <linux/netfilter_ipv4/ip_conntrack.h> |
26 | #include <linux/netfilter_ipv4/ip_conntrack_core.h> | |
27 | #include <linux/netfilter_ipv4/ip_conntrack_protocol.h> | |
28 | #include <linux/netfilter_ipv4/ip_nat.h> | |
29 | #include <linux/netfilter_ipv4/ip_nat_protocol.h> | |
30 | #include <linux/netfilter_ipv4/ip_nat_core.h> | |
31 | #include <linux/netfilter_ipv4/ip_nat_helper.h> | |
32 | #include <linux/netfilter_ipv4/ip_conntrack_helper.h> | |
1da177e4 LT |
33 | |
34 | #if 0 | |
35 | #define DEBUGP printk | |
36 | #else | |
37 | #define DEBUGP(format, args...) | |
38 | #endif | |
39 | ||
e45b1be8 | 40 | DEFINE_RWLOCK(ip_nat_lock); |
1da177e4 LT |
41 | |
42 | /* Calculated at init based on memory size */ | |
43 | static unsigned int ip_nat_htable_size; | |
44 | ||
45 | static struct list_head *bysource; | |
080774a2 HW |
46 | |
47 | #define MAX_IP_NAT_PROTO 256 | |
d127e94a | 48 | static struct ip_nat_protocol *ip_nat_protos[MAX_IP_NAT_PROTO]; |
1da177e4 | 49 | |
080774a2 HW |
50 | static inline struct ip_nat_protocol * |
51 | __ip_nat_proto_find(u_int8_t protonum) | |
52 | { | |
53 | return ip_nat_protos[protonum]; | |
54 | } | |
55 | ||
56 | struct ip_nat_protocol * | |
57 | ip_nat_proto_find_get(u_int8_t protonum) | |
58 | { | |
59 | struct ip_nat_protocol *p; | |
60 | ||
61 | /* we need to disable preemption to make sure 'p' doesn't get | |
62 | * removed until we've grabbed the reference */ | |
63 | preempt_disable(); | |
64 | p = __ip_nat_proto_find(protonum); | |
d2a7bb71 HW |
65 | if (!try_module_get(p->me)) |
66 | p = &ip_nat_unknown_protocol; | |
080774a2 HW |
67 | preempt_enable(); |
68 | ||
69 | return p; | |
70 | } | |
188bab3a | 71 | EXPORT_SYMBOL_GPL(ip_nat_proto_find_get); |
080774a2 HW |
72 | |
73 | void | |
74 | ip_nat_proto_put(struct ip_nat_protocol *p) | |
75 | { | |
76 | module_put(p->me); | |
77 | } | |
188bab3a | 78 | EXPORT_SYMBOL_GPL(ip_nat_proto_put); |
1da177e4 LT |
79 | |
80 | /* We keep an extra hash for each conntrack, for fast searching. */ | |
81 | static inline unsigned int | |
82 | hash_by_src(const struct ip_conntrack_tuple *tuple) | |
83 | { | |
84 | /* Original src, to ensure we map it consistently if poss. */ | |
a76b11dd | 85 | return jhash_3words((__force u32)tuple->src.ip, tuple->src.u.all, |
1da177e4 LT |
86 | tuple->dst.protonum, 0) % ip_nat_htable_size; |
87 | } | |
88 | ||
89 | /* Noone using conntrack by the time this called. */ | |
90 | static void ip_nat_cleanup_conntrack(struct ip_conntrack *conn) | |
91 | { | |
92 | if (!(conn->status & IPS_NAT_DONE_MASK)) | |
93 | return; | |
94 | ||
e45b1be8 | 95 | write_lock_bh(&ip_nat_lock); |
1da177e4 | 96 | list_del(&conn->nat.info.bysource); |
e45b1be8 | 97 | write_unlock_bh(&ip_nat_lock); |
1da177e4 LT |
98 | } |
99 | ||
1da177e4 LT |
100 | /* Is this tuple already taken? (not by us) */ |
101 | int | |
102 | ip_nat_used_tuple(const struct ip_conntrack_tuple *tuple, | |
103 | const struct ip_conntrack *ignored_conntrack) | |
104 | { | |
105 | /* Conntrack tracking doesn't keep track of outgoing tuples; only | |
106 | incoming ones. NAT means they don't have a fixed mapping, | |
107 | so we invert the tuple and look for the incoming reply. | |
108 | ||
109 | We could keep a separate hash if this proves too slow. */ | |
110 | struct ip_conntrack_tuple reply; | |
111 | ||
112 | invert_tuplepr(&reply, tuple); | |
113 | return ip_conntrack_tuple_taken(&reply, ignored_conntrack); | |
114 | } | |
188bab3a | 115 | EXPORT_SYMBOL(ip_nat_used_tuple); |
1da177e4 LT |
116 | |
117 | /* If we source map this tuple so reply looks like reply_tuple, will | |
118 | * that meet the constraints of range. */ | |
119 | static int | |
120 | in_range(const struct ip_conntrack_tuple *tuple, | |
121 | const struct ip_nat_range *range) | |
122 | { | |
e905a9ed | 123 | struct ip_nat_protocol *proto = |
080774a2 | 124 | __ip_nat_proto_find(tuple->dst.protonum); |
1da177e4 LT |
125 | |
126 | /* If we are supposed to map IPs, then we must be in the | |
127 | range specified, otherwise let this drag us onto a new src IP. */ | |
128 | if (range->flags & IP_NAT_RANGE_MAP_IPS) { | |
129 | if (ntohl(tuple->src.ip) < ntohl(range->min_ip) | |
130 | || ntohl(tuple->src.ip) > ntohl(range->max_ip)) | |
131 | return 0; | |
132 | } | |
133 | ||
134 | if (!(range->flags & IP_NAT_RANGE_PROTO_SPECIFIED) | |
135 | || proto->in_range(tuple, IP_NAT_MANIP_SRC, | |
136 | &range->min, &range->max)) | |
137 | return 1; | |
138 | ||
139 | return 0; | |
140 | } | |
141 | ||
142 | static inline int | |
143 | same_src(const struct ip_conntrack *ct, | |
144 | const struct ip_conntrack_tuple *tuple) | |
145 | { | |
146 | return (ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum | |
147 | == tuple->dst.protonum | |
148 | && ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.ip | |
149 | == tuple->src.ip | |
150 | && ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u.all | |
151 | == tuple->src.u.all); | |
152 | } | |
153 | ||
154 | /* Only called for SRC manip */ | |
155 | static int | |
156 | find_appropriate_src(const struct ip_conntrack_tuple *tuple, | |
157 | struct ip_conntrack_tuple *result, | |
158 | const struct ip_nat_range *range) | |
159 | { | |
160 | unsigned int h = hash_by_src(tuple); | |
161 | struct ip_conntrack *ct; | |
162 | ||
e45b1be8 | 163 | read_lock_bh(&ip_nat_lock); |
1da177e4 LT |
164 | list_for_each_entry(ct, &bysource[h], nat.info.bysource) { |
165 | if (same_src(ct, tuple)) { | |
166 | /* Copy source part from reply tuple. */ | |
167 | invert_tuplepr(result, | |
168 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
169 | result->dst = tuple->dst; | |
170 | ||
171 | if (in_range(result, range)) { | |
e45b1be8 | 172 | read_unlock_bh(&ip_nat_lock); |
1da177e4 LT |
173 | return 1; |
174 | } | |
175 | } | |
176 | } | |
e45b1be8 | 177 | read_unlock_bh(&ip_nat_lock); |
1da177e4 LT |
178 | return 0; |
179 | } | |
180 | ||
181 | /* For [FUTURE] fragmentation handling, we want the least-used | |
182 | src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus | |
183 | if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports | |
184 | 1-65535, we don't do pro-rata allocation based on ports; we choose | |
185 | the ip with the lowest src-ip/dst-ip/proto usage. | |
186 | */ | |
187 | static void | |
188 | find_best_ips_proto(struct ip_conntrack_tuple *tuple, | |
189 | const struct ip_nat_range *range, | |
190 | const struct ip_conntrack *conntrack, | |
191 | enum ip_nat_manip_type maniptype) | |
192 | { | |
a76b11dd | 193 | __be32 *var_ipp; |
1da177e4 LT |
194 | /* Host order */ |
195 | u_int32_t minip, maxip, j; | |
196 | ||
197 | /* No IP mapping? Do nothing. */ | |
198 | if (!(range->flags & IP_NAT_RANGE_MAP_IPS)) | |
199 | return; | |
200 | ||
201 | if (maniptype == IP_NAT_MANIP_SRC) | |
202 | var_ipp = &tuple->src.ip; | |
203 | else | |
204 | var_ipp = &tuple->dst.ip; | |
205 | ||
206 | /* Fast path: only one choice. */ | |
207 | if (range->min_ip == range->max_ip) { | |
208 | *var_ipp = range->min_ip; | |
209 | return; | |
210 | } | |
211 | ||
212 | /* Hashing source and destination IPs gives a fairly even | |
213 | * spread in practice (if there are a small number of IPs | |
214 | * involved, there usually aren't that many connections | |
215 | * anyway). The consistency means that servers see the same | |
216 | * client coming from the same IP (some Internet Banking sites | |
217 | * like this), even across reboots. */ | |
218 | minip = ntohl(range->min_ip); | |
219 | maxip = ntohl(range->max_ip); | |
a76b11dd | 220 | j = jhash_2words((__force u32)tuple->src.ip, (__force u32)tuple->dst.ip, 0); |
1da177e4 LT |
221 | *var_ipp = htonl(minip + j % (maxip - minip + 1)); |
222 | } | |
223 | ||
224 | /* Manipulate the tuple into the range given. For NF_IP_POST_ROUTING, | |
225 | * we change the source to map into the range. For NF_IP_PRE_ROUTING | |
226 | * and NF_IP_LOCAL_OUT, we change the destination to map into the | |
227 | * range. It might not be possible to get a unique tuple, but we try. | |
228 | * At worst (or if we race), we will end up with a final duplicate in | |
229 | * __ip_conntrack_confirm and drop the packet. */ | |
230 | static void | |
231 | get_unique_tuple(struct ip_conntrack_tuple *tuple, | |
232 | const struct ip_conntrack_tuple *orig_tuple, | |
233 | const struct ip_nat_range *range, | |
234 | struct ip_conntrack *conntrack, | |
235 | enum ip_nat_manip_type maniptype) | |
236 | { | |
080774a2 | 237 | struct ip_nat_protocol *proto; |
1da177e4 LT |
238 | |
239 | /* 1) If this srcip/proto/src-proto-part is currently mapped, | |
240 | and that same mapping gives a unique tuple within the given | |
241 | range, use that. | |
242 | ||
243 | This is only required for source (ie. NAT/masq) mappings. | |
244 | So far, we don't do local source mappings, so multiple | |
245 | manips not an issue. */ | |
246 | if (maniptype == IP_NAT_MANIP_SRC) { | |
247 | if (find_appropriate_src(orig_tuple, tuple, range)) { | |
248 | DEBUGP("get_unique_tuple: Found current src map\n"); | |
41f4689a EL |
249 | if (!(range->flags & IP_NAT_RANGE_PROTO_RANDOM)) |
250 | if (!ip_nat_used_tuple(tuple, conntrack)) | |
251 | return; | |
1da177e4 LT |
252 | } |
253 | } | |
254 | ||
255 | /* 2) Select the least-used IP/proto combination in the given | |
256 | range. */ | |
257 | *tuple = *orig_tuple; | |
258 | find_best_ips_proto(tuple, range, conntrack, maniptype); | |
259 | ||
260 | /* 3) The per-protocol part of the manip is made to map into | |
261 | the range to make a unique tuple. */ | |
262 | ||
080774a2 HW |
263 | proto = ip_nat_proto_find_get(orig_tuple->dst.protonum); |
264 | ||
41f4689a EL |
265 | /* Change protocol info to have some randomization */ |
266 | if (range->flags & IP_NAT_RANGE_PROTO_RANDOM) { | |
267 | proto->unique_tuple(tuple, range, maniptype, conntrack); | |
268 | ip_nat_proto_put(proto); | |
269 | return; | |
270 | } | |
271 | ||
1da177e4 LT |
272 | /* Only bother mapping if it's not already in range and unique */ |
273 | if ((!(range->flags & IP_NAT_RANGE_PROTO_SPECIFIED) | |
274 | || proto->in_range(tuple, maniptype, &range->min, &range->max)) | |
080774a2 HW |
275 | && !ip_nat_used_tuple(tuple, conntrack)) { |
276 | ip_nat_proto_put(proto); | |
1da177e4 | 277 | return; |
080774a2 | 278 | } |
1da177e4 LT |
279 | |
280 | /* Last change: get protocol to try to obtain unique tuple. */ | |
281 | proto->unique_tuple(tuple, range, maniptype, conntrack); | |
080774a2 HW |
282 | |
283 | ip_nat_proto_put(proto); | |
1da177e4 LT |
284 | } |
285 | ||
286 | unsigned int | |
287 | ip_nat_setup_info(struct ip_conntrack *conntrack, | |
288 | const struct ip_nat_range *range, | |
289 | unsigned int hooknum) | |
290 | { | |
291 | struct ip_conntrack_tuple curr_tuple, new_tuple; | |
292 | struct ip_nat_info *info = &conntrack->nat.info; | |
293 | int have_to_hash = !(conntrack->status & IPS_NAT_DONE_MASK); | |
294 | enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum); | |
295 | ||
296 | IP_NF_ASSERT(hooknum == NF_IP_PRE_ROUTING | |
297 | || hooknum == NF_IP_POST_ROUTING | |
298 | || hooknum == NF_IP_LOCAL_IN | |
299 | || hooknum == NF_IP_LOCAL_OUT); | |
300 | BUG_ON(ip_nat_initialized(conntrack, maniptype)); | |
301 | ||
302 | /* What we've got will look like inverse of reply. Normally | |
303 | this is what is in the conntrack, except for prior | |
304 | manipulations (future optimization: if num_manips == 0, | |
305 | orig_tp = | |
306 | conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple) */ | |
307 | invert_tuplepr(&curr_tuple, | |
308 | &conntrack->tuplehash[IP_CT_DIR_REPLY].tuple); | |
309 | ||
310 | get_unique_tuple(&new_tuple, &curr_tuple, range, conntrack, maniptype); | |
311 | ||
312 | if (!ip_ct_tuple_equal(&new_tuple, &curr_tuple)) { | |
313 | struct ip_conntrack_tuple reply; | |
314 | ||
315 | /* Alter conntrack table so will recognize replies. */ | |
316 | invert_tuplepr(&reply, &new_tuple); | |
317 | ip_conntrack_alter_reply(conntrack, &reply); | |
318 | ||
319 | /* Non-atomic: we own this at the moment. */ | |
320 | if (maniptype == IP_NAT_MANIP_SRC) | |
321 | conntrack->status |= IPS_SRC_NAT; | |
322 | else | |
323 | conntrack->status |= IPS_DST_NAT; | |
324 | } | |
325 | ||
326 | /* Place in source hash if this is the first time. */ | |
327 | if (have_to_hash) { | |
328 | unsigned int srchash | |
329 | = hash_by_src(&conntrack->tuplehash[IP_CT_DIR_ORIGINAL] | |
330 | .tuple); | |
e45b1be8 | 331 | write_lock_bh(&ip_nat_lock); |
1da177e4 | 332 | list_add(&info->bysource, &bysource[srchash]); |
e45b1be8 | 333 | write_unlock_bh(&ip_nat_lock); |
1da177e4 LT |
334 | } |
335 | ||
336 | /* It's done. */ | |
337 | if (maniptype == IP_NAT_MANIP_DST) | |
338 | set_bit(IPS_DST_NAT_DONE_BIT, &conntrack->status); | |
339 | else | |
340 | set_bit(IPS_SRC_NAT_DONE_BIT, &conntrack->status); | |
341 | ||
342 | return NF_ACCEPT; | |
343 | } | |
188bab3a | 344 | EXPORT_SYMBOL(ip_nat_setup_info); |
1da177e4 LT |
345 | |
346 | /* Returns true if succeeded. */ | |
347 | static int | |
348 | manip_pkt(u_int16_t proto, | |
349 | struct sk_buff **pskb, | |
350 | unsigned int iphdroff, | |
351 | const struct ip_conntrack_tuple *target, | |
352 | enum ip_nat_manip_type maniptype) | |
353 | { | |
354 | struct iphdr *iph; | |
080774a2 | 355 | struct ip_nat_protocol *p; |
1da177e4 | 356 | |
089af26c | 357 | if (!skb_make_writable(pskb, iphdroff + sizeof(*iph))) |
1da177e4 LT |
358 | return 0; |
359 | ||
360 | iph = (void *)(*pskb)->data + iphdroff; | |
361 | ||
362 | /* Manipulate protcol part. */ | |
080774a2 HW |
363 | p = ip_nat_proto_find_get(proto); |
364 | if (!p->manip_pkt(pskb, iphdroff, target, maniptype)) { | |
365 | ip_nat_proto_put(p); | |
1da177e4 | 366 | return 0; |
080774a2 HW |
367 | } |
368 | ip_nat_proto_put(p); | |
1da177e4 LT |
369 | |
370 | iph = (void *)(*pskb)->data + iphdroff; | |
371 | ||
372 | if (maniptype == IP_NAT_MANIP_SRC) { | |
43bc0ca7 | 373 | nf_csum_replace4(&iph->check, iph->saddr, target->src.ip); |
1da177e4 LT |
374 | iph->saddr = target->src.ip; |
375 | } else { | |
43bc0ca7 | 376 | nf_csum_replace4(&iph->check, iph->daddr, target->dst.ip); |
1da177e4 LT |
377 | iph->daddr = target->dst.ip; |
378 | } | |
379 | return 1; | |
380 | } | |
381 | ||
382 | /* Do packet manipulations according to ip_nat_setup_info. */ | |
188bab3a HW |
383 | unsigned int ip_nat_packet(struct ip_conntrack *ct, |
384 | enum ip_conntrack_info ctinfo, | |
385 | unsigned int hooknum, | |
386 | struct sk_buff **pskb) | |
1da177e4 LT |
387 | { |
388 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); | |
389 | unsigned long statusbit; | |
390 | enum ip_nat_manip_type mtype = HOOK2MANIP(hooknum); | |
391 | ||
1da177e4 LT |
392 | if (mtype == IP_NAT_MANIP_SRC) |
393 | statusbit = IPS_SRC_NAT; | |
394 | else | |
395 | statusbit = IPS_DST_NAT; | |
396 | ||
397 | /* Invert if this is reply dir. */ | |
398 | if (dir == IP_CT_DIR_REPLY) | |
399 | statusbit ^= IPS_NAT_MASK; | |
400 | ||
401 | /* Non-atomic: these bits don't change. */ | |
402 | if (ct->status & statusbit) { | |
403 | struct ip_conntrack_tuple target; | |
404 | ||
405 | /* We are aiming to look like inverse of other direction. */ | |
406 | invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); | |
407 | ||
408 | if (!manip_pkt(target.dst.protonum, pskb, 0, &target, mtype)) | |
409 | return NF_DROP; | |
410 | } | |
411 | return NF_ACCEPT; | |
412 | } | |
188bab3a | 413 | EXPORT_SYMBOL_GPL(ip_nat_packet); |
1da177e4 LT |
414 | |
415 | /* Dir is direction ICMP is coming from (opposite to packet it contains) */ | |
4cf411de PM |
416 | int ip_nat_icmp_reply_translation(struct ip_conntrack *ct, |
417 | enum ip_conntrack_info ctinfo, | |
418 | unsigned int hooknum, | |
419 | struct sk_buff **pskb) | |
1da177e4 LT |
420 | { |
421 | struct { | |
422 | struct icmphdr icmp; | |
423 | struct iphdr ip; | |
424 | } *inside; | |
425 | struct ip_conntrack_tuple inner, target; | |
426 | int hdrlen = (*pskb)->nh.iph->ihl * 4; | |
4cf411de | 427 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); |
bc6e14b6 | 428 | unsigned long statusbit; |
4cf411de | 429 | enum ip_nat_manip_type manip = HOOK2MANIP(hooknum); |
1da177e4 | 430 | |
089af26c | 431 | if (!skb_make_writable(pskb, hdrlen + sizeof(*inside))) |
1da177e4 LT |
432 | return 0; |
433 | ||
434 | inside = (void *)(*pskb)->data + (*pskb)->nh.iph->ihl*4; | |
435 | ||
436 | /* We're actually going to mangle it beyond trivial checksum | |
437 | adjustment, so make sure the current checksum is correct. */ | |
4cf411de PM |
438 | if (nf_ip_checksum(*pskb, hooknum, hdrlen, 0)) |
439 | return 0; | |
1da177e4 LT |
440 | |
441 | /* Must be RELATED */ | |
442 | IP_NF_ASSERT((*pskb)->nfctinfo == IP_CT_RELATED || | |
443 | (*pskb)->nfctinfo == IP_CT_RELATED+IP_CT_IS_REPLY); | |
444 | ||
445 | /* Redirects on non-null nats must be dropped, else they'll | |
e905a9ed YH |
446 | start talking to each other without our translation, and be |
447 | confused... --RR */ | |
1da177e4 LT |
448 | if (inside->icmp.type == ICMP_REDIRECT) { |
449 | /* If NAT isn't finished, assume it and drop. */ | |
450 | if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK) | |
451 | return 0; | |
452 | ||
453 | if (ct->status & IPS_NAT_MASK) | |
454 | return 0; | |
455 | } | |
456 | ||
457 | DEBUGP("icmp_reply_translation: translating error %p manp %u dir %s\n", | |
458 | *pskb, manip, dir == IP_CT_DIR_ORIGINAL ? "ORIG" : "REPLY"); | |
459 | ||
460 | if (!ip_ct_get_tuple(&inside->ip, *pskb, (*pskb)->nh.iph->ihl*4 + | |
e905a9ed YH |
461 | sizeof(struct icmphdr) + inside->ip.ihl*4, |
462 | &inner, | |
080774a2 | 463 | __ip_conntrack_proto_find(inside->ip.protocol))) |
1da177e4 LT |
464 | return 0; |
465 | ||
466 | /* Change inner back to look like incoming packet. We do the | |
467 | opposite manip on this hook to normal, because it might not | |
468 | pass all hooks (locally-generated ICMP). Consider incoming | |
469 | packet: PREROUTING (DST manip), routing produces ICMP, goes | |
470 | through POSTROUTING (which must correct the DST manip). */ | |
471 | if (!manip_pkt(inside->ip.protocol, pskb, | |
472 | (*pskb)->nh.iph->ihl*4 | |
473 | + sizeof(inside->icmp), | |
474 | &ct->tuplehash[!dir].tuple, | |
475 | !manip)) | |
476 | return 0; | |
477 | ||
4cf411de PM |
478 | if ((*pskb)->ip_summed != CHECKSUM_PARTIAL) { |
479 | /* Reloading "inside" here since manip_pkt inner. */ | |
480 | inside = (void *)(*pskb)->data + (*pskb)->nh.iph->ihl*4; | |
481 | inside->icmp.checksum = 0; | |
482 | inside->icmp.checksum = csum_fold(skb_checksum(*pskb, hdrlen, | |
483 | (*pskb)->len - hdrlen, | |
484 | 0)); | |
485 | } | |
1da177e4 LT |
486 | |
487 | /* Change outer to look the reply to an incoming packet | |
488 | * (proto 0 means don't invert per-proto part). */ | |
bc6e14b6 PM |
489 | if (manip == IP_NAT_MANIP_SRC) |
490 | statusbit = IPS_SRC_NAT; | |
491 | else | |
492 | statusbit = IPS_DST_NAT; | |
1da177e4 | 493 | |
bc6e14b6 PM |
494 | /* Invert if this is reply dir. */ |
495 | if (dir == IP_CT_DIR_REPLY) | |
496 | statusbit ^= IPS_NAT_MASK; | |
1da177e4 | 497 | |
bc6e14b6 | 498 | if (ct->status & statusbit) { |
1da177e4 LT |
499 | invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); |
500 | if (!manip_pkt(0, pskb, 0, &target, manip)) | |
501 | return 0; | |
502 | } | |
503 | ||
504 | return 1; | |
505 | } | |
188bab3a | 506 | EXPORT_SYMBOL_GPL(ip_nat_icmp_reply_translation); |
1da177e4 LT |
507 | |
508 | /* Protocol registration. */ | |
509 | int ip_nat_protocol_register(struct ip_nat_protocol *proto) | |
510 | { | |
511 | int ret = 0; | |
512 | ||
e45b1be8 | 513 | write_lock_bh(&ip_nat_lock); |
1da177e4 LT |
514 | if (ip_nat_protos[proto->protonum] != &ip_nat_unknown_protocol) { |
515 | ret = -EBUSY; | |
516 | goto out; | |
517 | } | |
518 | ip_nat_protos[proto->protonum] = proto; | |
519 | out: | |
e45b1be8 | 520 | write_unlock_bh(&ip_nat_lock); |
1da177e4 LT |
521 | return ret; |
522 | } | |
188bab3a | 523 | EXPORT_SYMBOL(ip_nat_protocol_register); |
1da177e4 LT |
524 | |
525 | /* Noone stores the protocol anywhere; simply delete it. */ | |
526 | void ip_nat_protocol_unregister(struct ip_nat_protocol *proto) | |
527 | { | |
e45b1be8 | 528 | write_lock_bh(&ip_nat_lock); |
1da177e4 | 529 | ip_nat_protos[proto->protonum] = &ip_nat_unknown_protocol; |
e45b1be8 | 530 | write_unlock_bh(&ip_nat_lock); |
1da177e4 LT |
531 | |
532 | /* Someone could be still looking at the proto in a bh. */ | |
533 | synchronize_net(); | |
534 | } | |
188bab3a | 535 | EXPORT_SYMBOL(ip_nat_protocol_unregister); |
1da177e4 | 536 | |
080774a2 HW |
537 | #if defined(CONFIG_IP_NF_CONNTRACK_NETLINK) || \ |
538 | defined(CONFIG_IP_NF_CONNTRACK_NETLINK_MODULE) | |
539 | int | |
e905a9ed | 540 | ip_nat_port_range_to_nfattr(struct sk_buff *skb, |
080774a2 HW |
541 | const struct ip_nat_range *range) |
542 | { | |
a76b11dd | 543 | NFA_PUT(skb, CTA_PROTONAT_PORT_MIN, sizeof(__be16), |
080774a2 | 544 | &range->min.tcp.port); |
a76b11dd | 545 | NFA_PUT(skb, CTA_PROTONAT_PORT_MAX, sizeof(__be16), |
080774a2 HW |
546 | &range->max.tcp.port); |
547 | ||
548 | return 0; | |
549 | ||
550 | nfattr_failure: | |
551 | return -1; | |
552 | } | |
553 | ||
554 | int | |
555 | ip_nat_port_nfattr_to_range(struct nfattr *tb[], struct ip_nat_range *range) | |
556 | { | |
557 | int ret = 0; | |
e905a9ed | 558 | |
080774a2 HW |
559 | /* we have to return whether we actually parsed something or not */ |
560 | ||
561 | if (tb[CTA_PROTONAT_PORT_MIN-1]) { | |
562 | ret = 1; | |
e905a9ed | 563 | range->min.tcp.port = |
a76b11dd | 564 | *(__be16 *)NFA_DATA(tb[CTA_PROTONAT_PORT_MIN-1]); |
080774a2 | 565 | } |
e905a9ed | 566 | |
080774a2 | 567 | if (!tb[CTA_PROTONAT_PORT_MAX-1]) { |
e905a9ed | 568 | if (ret) |
080774a2 HW |
569 | range->max.tcp.port = range->min.tcp.port; |
570 | } else { | |
571 | ret = 1; | |
e905a9ed | 572 | range->max.tcp.port = |
a76b11dd | 573 | *(__be16 *)NFA_DATA(tb[CTA_PROTONAT_PORT_MAX-1]); |
080774a2 HW |
574 | } |
575 | ||
576 | return ret; | |
577 | } | |
8922bc93 HW |
578 | EXPORT_SYMBOL_GPL(ip_nat_port_nfattr_to_range); |
579 | EXPORT_SYMBOL_GPL(ip_nat_port_range_to_nfattr); | |
080774a2 HW |
580 | #endif |
581 | ||
188bab3a | 582 | static int __init ip_nat_init(void) |
1da177e4 LT |
583 | { |
584 | size_t i; | |
585 | ||
586 | /* Leave them the same for the moment. */ | |
587 | ip_nat_htable_size = ip_conntrack_htable_size; | |
588 | ||
589 | /* One vmalloc for both hash tables */ | |
590 | bysource = vmalloc(sizeof(struct list_head) * ip_nat_htable_size); | |
591 | if (!bysource) | |
592 | return -ENOMEM; | |
593 | ||
594 | /* Sew in builtin protocols. */ | |
e45b1be8 | 595 | write_lock_bh(&ip_nat_lock); |
1da177e4 LT |
596 | for (i = 0; i < MAX_IP_NAT_PROTO; i++) |
597 | ip_nat_protos[i] = &ip_nat_unknown_protocol; | |
598 | ip_nat_protos[IPPROTO_TCP] = &ip_nat_protocol_tcp; | |
599 | ip_nat_protos[IPPROTO_UDP] = &ip_nat_protocol_udp; | |
600 | ip_nat_protos[IPPROTO_ICMP] = &ip_nat_protocol_icmp; | |
e45b1be8 | 601 | write_unlock_bh(&ip_nat_lock); |
1da177e4 LT |
602 | |
603 | for (i = 0; i < ip_nat_htable_size; i++) { | |
604 | INIT_LIST_HEAD(&bysource[i]); | |
605 | } | |
606 | ||
607 | /* FIXME: Man, this is a hack. <SIGH> */ | |
608 | IP_NF_ASSERT(ip_conntrack_destroyed == NULL); | |
609 | ip_conntrack_destroyed = &ip_nat_cleanup_conntrack; | |
610 | ||
611 | /* Initialize fake conntrack so that NAT will skip it */ | |
612 | ip_conntrack_untracked.status |= IPS_NAT_DONE_MASK; | |
613 | return 0; | |
614 | } | |
615 | ||
616 | /* Clear NAT section of all conntracks, in case we're loaded again. */ | |
617 | static int clean_nat(struct ip_conntrack *i, void *data) | |
618 | { | |
619 | memset(&i->nat, 0, sizeof(i->nat)); | |
620 | i->status &= ~(IPS_NAT_MASK | IPS_NAT_DONE_MASK | IPS_SEQ_ADJUST); | |
621 | return 0; | |
622 | } | |
623 | ||
188bab3a | 624 | static void __exit ip_nat_cleanup(void) |
1da177e4 LT |
625 | { |
626 | ip_ct_iterate_cleanup(&clean_nat, NULL); | |
627 | ip_conntrack_destroyed = NULL; | |
628 | vfree(bysource); | |
629 | } | |
188bab3a HW |
630 | |
631 | MODULE_LICENSE("GPL"); | |
632 | ||
633 | module_init(ip_nat_init); | |
634 | module_exit(ip_nat_cleanup); |