Commit | Line | Data |
---|---|---|
5b1158e9 JK |
1 | /* NAT for netfilter; shared with compatibility layer. */ |
2 | ||
3 | /* (C) 1999-2001 Paul `Rusty' Russell | |
4 | * (C) 2002-2006 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> | |
5b1158e9 JK |
15 | #include <net/checksum.h> |
16 | #include <net/icmp.h> | |
17 | #include <net/ip.h> | |
18 | #include <net/tcp.h> /* For tcp_prot in getorigdst */ | |
19 | #include <linux/icmp.h> | |
20 | #include <linux/udp.h> | |
21 | #include <linux/jhash.h> | |
22 | ||
23 | #include <linux/netfilter_ipv4.h> | |
24 | #include <net/netfilter/nf_conntrack.h> | |
25 | #include <net/netfilter/nf_conntrack_core.h> | |
26 | #include <net/netfilter/nf_nat.h> | |
27 | #include <net/netfilter/nf_nat_protocol.h> | |
28 | #include <net/netfilter/nf_nat_core.h> | |
29 | #include <net/netfilter/nf_nat_helper.h> | |
30 | #include <net/netfilter/nf_conntrack_helper.h> | |
31 | #include <net/netfilter/nf_conntrack_l3proto.h> | |
32 | #include <net/netfilter/nf_conntrack_l4proto.h> | |
33 | ||
5b1158e9 JK |
34 | static DEFINE_RWLOCK(nf_nat_lock); |
35 | ||
36 | static struct nf_conntrack_l3proto *l3proto = NULL; | |
37 | ||
38 | /* Calculated at init based on memory size */ | |
39 | static unsigned int nf_nat_htable_size; | |
53aba597 | 40 | static int nf_nat_vmalloced; |
5b1158e9 | 41 | |
53aba597 | 42 | static struct hlist_head *bysource; |
5b1158e9 JK |
43 | |
44 | #define MAX_IP_NAT_PROTO 256 | |
45 | static struct nf_nat_protocol *nf_nat_protos[MAX_IP_NAT_PROTO]; | |
46 | ||
47 | static inline struct nf_nat_protocol * | |
48 | __nf_nat_proto_find(u_int8_t protonum) | |
49 | { | |
e22a0548 | 50 | return rcu_dereference(nf_nat_protos[protonum]); |
5b1158e9 JK |
51 | } |
52 | ||
53 | struct nf_nat_protocol * | |
54 | nf_nat_proto_find_get(u_int8_t protonum) | |
55 | { | |
56 | struct nf_nat_protocol *p; | |
57 | ||
e22a0548 | 58 | rcu_read_lock(); |
5b1158e9 JK |
59 | p = __nf_nat_proto_find(protonum); |
60 | if (!try_module_get(p->me)) | |
61 | p = &nf_nat_unknown_protocol; | |
e22a0548 | 62 | rcu_read_unlock(); |
5b1158e9 JK |
63 | |
64 | return p; | |
65 | } | |
66 | EXPORT_SYMBOL_GPL(nf_nat_proto_find_get); | |
67 | ||
68 | void | |
69 | nf_nat_proto_put(struct nf_nat_protocol *p) | |
70 | { | |
71 | module_put(p->me); | |
72 | } | |
73 | EXPORT_SYMBOL_GPL(nf_nat_proto_put); | |
74 | ||
75 | /* We keep an extra hash for each conntrack, for fast searching. */ | |
76 | static inline unsigned int | |
77 | hash_by_src(const struct nf_conntrack_tuple *tuple) | |
78 | { | |
79 | /* Original src, to ensure we map it consistently if poss. */ | |
a34c4589 AV |
80 | return jhash_3words((__force u32)tuple->src.u3.ip, |
81 | (__force u32)tuple->src.u.all, | |
5b1158e9 JK |
82 | tuple->dst.protonum, 0) % nf_nat_htable_size; |
83 | } | |
84 | ||
5b1158e9 JK |
85 | /* Is this tuple already taken? (not by us) */ |
86 | int | |
87 | nf_nat_used_tuple(const struct nf_conntrack_tuple *tuple, | |
88 | const struct nf_conn *ignored_conntrack) | |
89 | { | |
90 | /* Conntrack tracking doesn't keep track of outgoing tuples; only | |
91 | incoming ones. NAT means they don't have a fixed mapping, | |
92 | so we invert the tuple and look for the incoming reply. | |
93 | ||
94 | We could keep a separate hash if this proves too slow. */ | |
95 | struct nf_conntrack_tuple reply; | |
96 | ||
97 | nf_ct_invert_tuplepr(&reply, tuple); | |
98 | return nf_conntrack_tuple_taken(&reply, ignored_conntrack); | |
99 | } | |
100 | EXPORT_SYMBOL(nf_nat_used_tuple); | |
101 | ||
102 | /* If we source map this tuple so reply looks like reply_tuple, will | |
103 | * that meet the constraints of range. */ | |
104 | static int | |
105 | in_range(const struct nf_conntrack_tuple *tuple, | |
106 | const struct nf_nat_range *range) | |
107 | { | |
108 | struct nf_nat_protocol *proto; | |
e22a0548 | 109 | int ret = 0; |
5b1158e9 | 110 | |
5b1158e9 JK |
111 | /* If we are supposed to map IPs, then we must be in the |
112 | range specified, otherwise let this drag us onto a new src IP. */ | |
113 | if (range->flags & IP_NAT_RANGE_MAP_IPS) { | |
114 | if (ntohl(tuple->src.u3.ip) < ntohl(range->min_ip) || | |
115 | ntohl(tuple->src.u3.ip) > ntohl(range->max_ip)) | |
116 | return 0; | |
117 | } | |
118 | ||
e22a0548 PM |
119 | rcu_read_lock(); |
120 | proto = __nf_nat_proto_find(tuple->dst.protonum); | |
5b1158e9 JK |
121 | if (!(range->flags & IP_NAT_RANGE_PROTO_SPECIFIED) || |
122 | proto->in_range(tuple, IP_NAT_MANIP_SRC, | |
123 | &range->min, &range->max)) | |
e22a0548 PM |
124 | ret = 1; |
125 | rcu_read_unlock(); | |
5b1158e9 | 126 | |
e22a0548 | 127 | return ret; |
5b1158e9 JK |
128 | } |
129 | ||
130 | static inline int | |
131 | same_src(const struct nf_conn *ct, | |
132 | const struct nf_conntrack_tuple *tuple) | |
133 | { | |
134 | const struct nf_conntrack_tuple *t; | |
135 | ||
136 | t = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; | |
137 | return (t->dst.protonum == tuple->dst.protonum && | |
138 | t->src.u3.ip == tuple->src.u3.ip && | |
139 | t->src.u.all == tuple->src.u.all); | |
140 | } | |
141 | ||
142 | /* Only called for SRC manip */ | |
143 | static int | |
144 | find_appropriate_src(const struct nf_conntrack_tuple *tuple, | |
145 | struct nf_conntrack_tuple *result, | |
146 | const struct nf_nat_range *range) | |
147 | { | |
148 | unsigned int h = hash_by_src(tuple); | |
149 | struct nf_conn_nat *nat; | |
150 | struct nf_conn *ct; | |
53aba597 | 151 | struct hlist_node *n; |
5b1158e9 JK |
152 | |
153 | read_lock_bh(&nf_nat_lock); | |
53aba597 | 154 | hlist_for_each_entry(nat, n, &bysource[h], bysource) { |
b6b84d4a | 155 | ct = nat->ct; |
5b1158e9 JK |
156 | if (same_src(ct, tuple)) { |
157 | /* Copy source part from reply tuple. */ | |
158 | nf_ct_invert_tuplepr(result, | |
159 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
160 | result->dst = tuple->dst; | |
161 | ||
162 | if (in_range(result, range)) { | |
163 | read_unlock_bh(&nf_nat_lock); | |
164 | return 1; | |
165 | } | |
166 | } | |
167 | } | |
168 | read_unlock_bh(&nf_nat_lock); | |
169 | return 0; | |
170 | } | |
171 | ||
172 | /* For [FUTURE] fragmentation handling, we want the least-used | |
173 | src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus | |
174 | if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports | |
175 | 1-65535, we don't do pro-rata allocation based on ports; we choose | |
176 | the ip with the lowest src-ip/dst-ip/proto usage. | |
177 | */ | |
178 | static void | |
179 | find_best_ips_proto(struct nf_conntrack_tuple *tuple, | |
180 | const struct nf_nat_range *range, | |
181 | const struct nf_conn *ct, | |
182 | enum nf_nat_manip_type maniptype) | |
183 | { | |
184 | __be32 *var_ipp; | |
185 | /* Host order */ | |
186 | u_int32_t minip, maxip, j; | |
187 | ||
188 | /* No IP mapping? Do nothing. */ | |
189 | if (!(range->flags & IP_NAT_RANGE_MAP_IPS)) | |
190 | return; | |
191 | ||
192 | if (maniptype == IP_NAT_MANIP_SRC) | |
193 | var_ipp = &tuple->src.u3.ip; | |
194 | else | |
195 | var_ipp = &tuple->dst.u3.ip; | |
196 | ||
197 | /* Fast path: only one choice. */ | |
198 | if (range->min_ip == range->max_ip) { | |
199 | *var_ipp = range->min_ip; | |
200 | return; | |
201 | } | |
202 | ||
203 | /* Hashing source and destination IPs gives a fairly even | |
204 | * spread in practice (if there are a small number of IPs | |
205 | * involved, there usually aren't that many connections | |
206 | * anyway). The consistency means that servers see the same | |
207 | * client coming from the same IP (some Internet Banking sites | |
208 | * like this), even across reboots. */ | |
209 | minip = ntohl(range->min_ip); | |
210 | maxip = ntohl(range->max_ip); | |
211 | j = jhash_2words((__force u32)tuple->src.u3.ip, | |
212 | (__force u32)tuple->dst.u3.ip, 0); | |
213 | *var_ipp = htonl(minip + j % (maxip - minip + 1)); | |
214 | } | |
215 | ||
216 | /* Manipulate the tuple into the range given. For NF_IP_POST_ROUTING, | |
217 | * we change the source to map into the range. For NF_IP_PRE_ROUTING | |
218 | * and NF_IP_LOCAL_OUT, we change the destination to map into the | |
219 | * range. It might not be possible to get a unique tuple, but we try. | |
220 | * At worst (or if we race), we will end up with a final duplicate in | |
221 | * __ip_conntrack_confirm and drop the packet. */ | |
222 | static void | |
223 | get_unique_tuple(struct nf_conntrack_tuple *tuple, | |
224 | const struct nf_conntrack_tuple *orig_tuple, | |
225 | const struct nf_nat_range *range, | |
226 | struct nf_conn *ct, | |
227 | enum nf_nat_manip_type maniptype) | |
228 | { | |
229 | struct nf_nat_protocol *proto; | |
230 | ||
231 | /* 1) If this srcip/proto/src-proto-part is currently mapped, | |
232 | and that same mapping gives a unique tuple within the given | |
233 | range, use that. | |
234 | ||
235 | This is only required for source (ie. NAT/masq) mappings. | |
236 | So far, we don't do local source mappings, so multiple | |
237 | manips not an issue. */ | |
238 | if (maniptype == IP_NAT_MANIP_SRC) { | |
239 | if (find_appropriate_src(orig_tuple, tuple, range)) { | |
0d53778e | 240 | pr_debug("get_unique_tuple: Found current src map\n"); |
41f4689a EL |
241 | if (!(range->flags & IP_NAT_RANGE_PROTO_RANDOM)) |
242 | if (!nf_nat_used_tuple(tuple, ct)) | |
243 | return; | |
5b1158e9 JK |
244 | } |
245 | } | |
246 | ||
247 | /* 2) Select the least-used IP/proto combination in the given | |
248 | range. */ | |
249 | *tuple = *orig_tuple; | |
250 | find_best_ips_proto(tuple, range, ct, maniptype); | |
251 | ||
252 | /* 3) The per-protocol part of the manip is made to map into | |
253 | the range to make a unique tuple. */ | |
254 | ||
e22a0548 PM |
255 | rcu_read_lock(); |
256 | proto = __nf_nat_proto_find(orig_tuple->dst.protonum); | |
5b1158e9 | 257 | |
41f4689a EL |
258 | /* Change protocol info to have some randomization */ |
259 | if (range->flags & IP_NAT_RANGE_PROTO_RANDOM) { | |
260 | proto->unique_tuple(tuple, range, maniptype, ct); | |
e22a0548 | 261 | goto out; |
41f4689a EL |
262 | } |
263 | ||
5b1158e9 JK |
264 | /* Only bother mapping if it's not already in range and unique */ |
265 | if ((!(range->flags & IP_NAT_RANGE_PROTO_SPECIFIED) || | |
266 | proto->in_range(tuple, maniptype, &range->min, &range->max)) && | |
e22a0548 PM |
267 | !nf_nat_used_tuple(tuple, ct)) |
268 | goto out; | |
5b1158e9 JK |
269 | |
270 | /* Last change: get protocol to try to obtain unique tuple. */ | |
271 | proto->unique_tuple(tuple, range, maniptype, ct); | |
e22a0548 PM |
272 | out: |
273 | rcu_read_unlock(); | |
5b1158e9 JK |
274 | } |
275 | ||
276 | unsigned int | |
277 | nf_nat_setup_info(struct nf_conn *ct, | |
278 | const struct nf_nat_range *range, | |
279 | unsigned int hooknum) | |
280 | { | |
281 | struct nf_conntrack_tuple curr_tuple, new_tuple; | |
2d59e5ca | 282 | struct nf_conn_nat *nat; |
5b1158e9 JK |
283 | int have_to_hash = !(ct->status & IPS_NAT_DONE_MASK); |
284 | enum nf_nat_manip_type maniptype = HOOK2MANIP(hooknum); | |
285 | ||
2d59e5ca YK |
286 | /* nat helper or nfctnetlink also setup binding */ |
287 | nat = nfct_nat(ct); | |
288 | if (!nat) { | |
289 | nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC); | |
290 | if (nat == NULL) { | |
0d53778e | 291 | pr_debug("failed to add NAT extension\n"); |
2d59e5ca YK |
292 | return NF_ACCEPT; |
293 | } | |
294 | } | |
295 | ||
5b1158e9 JK |
296 | NF_CT_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(nf_nat_initialized(ct, 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 | nf_ct_invert_tuplepr(&curr_tuple, | |
308 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
309 | ||
310 | get_unique_tuple(&new_tuple, &curr_tuple, range, ct, maniptype); | |
311 | ||
312 | if (!nf_ct_tuple_equal(&new_tuple, &curr_tuple)) { | |
313 | struct nf_conntrack_tuple reply; | |
314 | ||
315 | /* Alter conntrack table so will recognize replies. */ | |
316 | nf_ct_invert_tuplepr(&reply, &new_tuple); | |
317 | nf_conntrack_alter_reply(ct, &reply); | |
318 | ||
319 | /* Non-atomic: we own this at the moment. */ | |
320 | if (maniptype == IP_NAT_MANIP_SRC) | |
321 | ct->status |= IPS_SRC_NAT; | |
322 | else | |
323 | ct->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 | ||
330 | srchash = hash_by_src(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
331 | write_lock_bh(&nf_nat_lock); | |
2d59e5ca | 332 | /* nf_conntrack_alter_reply might re-allocate exntension aera */ |
b6b84d4a YK |
333 | nat = nfct_nat(ct); |
334 | nat->ct = ct; | |
53aba597 | 335 | hlist_add_head(&nat->bysource, &bysource[srchash]); |
5b1158e9 JK |
336 | write_unlock_bh(&nf_nat_lock); |
337 | } | |
338 | ||
339 | /* It's done. */ | |
340 | if (maniptype == IP_NAT_MANIP_DST) | |
341 | set_bit(IPS_DST_NAT_DONE_BIT, &ct->status); | |
342 | else | |
343 | set_bit(IPS_SRC_NAT_DONE_BIT, &ct->status); | |
344 | ||
345 | return NF_ACCEPT; | |
346 | } | |
347 | EXPORT_SYMBOL(nf_nat_setup_info); | |
348 | ||
349 | /* Returns true if succeeded. */ | |
350 | static int | |
351 | manip_pkt(u_int16_t proto, | |
352 | struct sk_buff **pskb, | |
353 | unsigned int iphdroff, | |
354 | const struct nf_conntrack_tuple *target, | |
355 | enum nf_nat_manip_type maniptype) | |
356 | { | |
357 | struct iphdr *iph; | |
358 | struct nf_nat_protocol *p; | |
359 | ||
37d41879 | 360 | if (!skb_make_writable(*pskb, iphdroff + sizeof(*iph))) |
5b1158e9 JK |
361 | return 0; |
362 | ||
363 | iph = (void *)(*pskb)->data + iphdroff; | |
364 | ||
365 | /* Manipulate protcol part. */ | |
e22a0548 PM |
366 | |
367 | /* rcu_read_lock()ed by nf_hook_slow */ | |
368 | p = __nf_nat_proto_find(proto); | |
369 | if (!p->manip_pkt(pskb, iphdroff, target, maniptype)) | |
5b1158e9 | 370 | return 0; |
5b1158e9 JK |
371 | |
372 | iph = (void *)(*pskb)->data + iphdroff; | |
373 | ||
374 | if (maniptype == IP_NAT_MANIP_SRC) { | |
375 | nf_csum_replace4(&iph->check, iph->saddr, target->src.u3.ip); | |
376 | iph->saddr = target->src.u3.ip; | |
377 | } else { | |
378 | nf_csum_replace4(&iph->check, iph->daddr, target->dst.u3.ip); | |
379 | iph->daddr = target->dst.u3.ip; | |
380 | } | |
381 | return 1; | |
382 | } | |
383 | ||
384 | /* Do packet manipulations according to nf_nat_setup_info. */ | |
385 | unsigned int nf_nat_packet(struct nf_conn *ct, | |
386 | enum ip_conntrack_info ctinfo, | |
387 | unsigned int hooknum, | |
388 | struct sk_buff **pskb) | |
389 | { | |
390 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); | |
391 | unsigned long statusbit; | |
392 | enum nf_nat_manip_type mtype = HOOK2MANIP(hooknum); | |
393 | ||
394 | if (mtype == IP_NAT_MANIP_SRC) | |
395 | statusbit = IPS_SRC_NAT; | |
396 | else | |
397 | statusbit = IPS_DST_NAT; | |
398 | ||
399 | /* Invert if this is reply dir. */ | |
400 | if (dir == IP_CT_DIR_REPLY) | |
401 | statusbit ^= IPS_NAT_MASK; | |
402 | ||
403 | /* Non-atomic: these bits don't change. */ | |
404 | if (ct->status & statusbit) { | |
405 | struct nf_conntrack_tuple target; | |
406 | ||
407 | /* We are aiming to look like inverse of other direction. */ | |
408 | nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); | |
409 | ||
410 | if (!manip_pkt(target.dst.protonum, pskb, 0, &target, mtype)) | |
411 | return NF_DROP; | |
412 | } | |
413 | return NF_ACCEPT; | |
414 | } | |
415 | EXPORT_SYMBOL_GPL(nf_nat_packet); | |
416 | ||
417 | /* Dir is direction ICMP is coming from (opposite to packet it contains) */ | |
418 | int nf_nat_icmp_reply_translation(struct nf_conn *ct, | |
419 | enum ip_conntrack_info ctinfo, | |
420 | unsigned int hooknum, | |
421 | struct sk_buff **pskb) | |
422 | { | |
423 | struct { | |
424 | struct icmphdr icmp; | |
425 | struct iphdr ip; | |
426 | } *inside; | |
923f4902 | 427 | struct nf_conntrack_l4proto *l4proto; |
5b1158e9 | 428 | struct nf_conntrack_tuple inner, target; |
c9bdd4b5 | 429 | int hdrlen = ip_hdrlen(*pskb); |
5b1158e9 JK |
430 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); |
431 | unsigned long statusbit; | |
432 | enum nf_nat_manip_type manip = HOOK2MANIP(hooknum); | |
433 | ||
37d41879 | 434 | if (!skb_make_writable(*pskb, hdrlen + sizeof(*inside))) |
5b1158e9 JK |
435 | return 0; |
436 | ||
c9bdd4b5 | 437 | inside = (void *)(*pskb)->data + ip_hdrlen(*pskb); |
5b1158e9 JK |
438 | |
439 | /* We're actually going to mangle it beyond trivial checksum | |
440 | adjustment, so make sure the current checksum is correct. */ | |
441 | if (nf_ip_checksum(*pskb, hooknum, hdrlen, 0)) | |
442 | return 0; | |
443 | ||
444 | /* Must be RELATED */ | |
445 | NF_CT_ASSERT((*pskb)->nfctinfo == IP_CT_RELATED || | |
446 | (*pskb)->nfctinfo == IP_CT_RELATED+IP_CT_IS_REPLY); | |
447 | ||
448 | /* Redirects on non-null nats must be dropped, else they'll | |
e905a9ed YH |
449 | start talking to each other without our translation, and be |
450 | confused... --RR */ | |
5b1158e9 JK |
451 | if (inside->icmp.type == ICMP_REDIRECT) { |
452 | /* If NAT isn't finished, assume it and drop. */ | |
453 | if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK) | |
454 | return 0; | |
455 | ||
456 | if (ct->status & IPS_NAT_MASK) | |
457 | return 0; | |
458 | } | |
459 | ||
0d53778e PM |
460 | pr_debug("icmp_reply_translation: translating error %p manip %u " |
461 | "dir %s\n", *pskb, manip, | |
462 | dir == IP_CT_DIR_ORIGINAL ? "ORIG" : "REPLY"); | |
5b1158e9 | 463 | |
923f4902 PM |
464 | /* rcu_read_lock()ed by nf_hook_slow */ |
465 | l4proto = __nf_ct_l4proto_find(PF_INET, inside->ip.protocol); | |
466 | ||
5b1158e9 | 467 | if (!nf_ct_get_tuple(*pskb, |
c9bdd4b5 ACM |
468 | ip_hdrlen(*pskb) + sizeof(struct icmphdr), |
469 | (ip_hdrlen(*pskb) + | |
470 | sizeof(struct icmphdr) + inside->ip.ihl * 4), | |
e905a9ed YH |
471 | (u_int16_t)AF_INET, |
472 | inside->ip.protocol, | |
923f4902 | 473 | &inner, l3proto, l4proto)) |
5b1158e9 JK |
474 | return 0; |
475 | ||
476 | /* Change inner back to look like incoming packet. We do the | |
477 | opposite manip on this hook to normal, because it might not | |
478 | pass all hooks (locally-generated ICMP). Consider incoming | |
479 | packet: PREROUTING (DST manip), routing produces ICMP, goes | |
480 | through POSTROUTING (which must correct the DST manip). */ | |
481 | if (!manip_pkt(inside->ip.protocol, pskb, | |
c9bdd4b5 | 482 | ip_hdrlen(*pskb) + sizeof(inside->icmp), |
5b1158e9 JK |
483 | &ct->tuplehash[!dir].tuple, |
484 | !manip)) | |
485 | return 0; | |
486 | ||
487 | if ((*pskb)->ip_summed != CHECKSUM_PARTIAL) { | |
488 | /* Reloading "inside" here since manip_pkt inner. */ | |
c9bdd4b5 | 489 | inside = (void *)(*pskb)->data + ip_hdrlen(*pskb); |
5b1158e9 JK |
490 | inside->icmp.checksum = 0; |
491 | inside->icmp.checksum = | |
492 | csum_fold(skb_checksum(*pskb, hdrlen, | |
493 | (*pskb)->len - hdrlen, 0)); | |
494 | } | |
495 | ||
496 | /* Change outer to look the reply to an incoming packet | |
497 | * (proto 0 means don't invert per-proto part). */ | |
498 | if (manip == IP_NAT_MANIP_SRC) | |
499 | statusbit = IPS_SRC_NAT; | |
500 | else | |
501 | statusbit = IPS_DST_NAT; | |
502 | ||
503 | /* Invert if this is reply dir. */ | |
504 | if (dir == IP_CT_DIR_REPLY) | |
505 | statusbit ^= IPS_NAT_MASK; | |
506 | ||
507 | if (ct->status & statusbit) { | |
508 | nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); | |
509 | if (!manip_pkt(0, pskb, 0, &target, manip)) | |
510 | return 0; | |
511 | } | |
512 | ||
513 | return 1; | |
514 | } | |
515 | EXPORT_SYMBOL_GPL(nf_nat_icmp_reply_translation); | |
516 | ||
517 | /* Protocol registration. */ | |
518 | int nf_nat_protocol_register(struct nf_nat_protocol *proto) | |
519 | { | |
520 | int ret = 0; | |
521 | ||
522 | write_lock_bh(&nf_nat_lock); | |
523 | if (nf_nat_protos[proto->protonum] != &nf_nat_unknown_protocol) { | |
524 | ret = -EBUSY; | |
525 | goto out; | |
526 | } | |
e22a0548 | 527 | rcu_assign_pointer(nf_nat_protos[proto->protonum], proto); |
5b1158e9 JK |
528 | out: |
529 | write_unlock_bh(&nf_nat_lock); | |
530 | return ret; | |
531 | } | |
532 | EXPORT_SYMBOL(nf_nat_protocol_register); | |
533 | ||
534 | /* Noone stores the protocol anywhere; simply delete it. */ | |
535 | void nf_nat_protocol_unregister(struct nf_nat_protocol *proto) | |
536 | { | |
537 | write_lock_bh(&nf_nat_lock); | |
e22a0548 PM |
538 | rcu_assign_pointer(nf_nat_protos[proto->protonum], |
539 | &nf_nat_unknown_protocol); | |
5b1158e9 | 540 | write_unlock_bh(&nf_nat_lock); |
e22a0548 | 541 | synchronize_rcu(); |
5b1158e9 JK |
542 | } |
543 | EXPORT_SYMBOL(nf_nat_protocol_unregister); | |
544 | ||
e281db5c | 545 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
5b1158e9 | 546 | int |
fdf70832 | 547 | nf_nat_port_range_to_nlattr(struct sk_buff *skb, |
5b1158e9 JK |
548 | const struct nf_nat_range *range) |
549 | { | |
df6fb868 | 550 | NLA_PUT(skb, CTA_PROTONAT_PORT_MIN, sizeof(__be16), |
5b1158e9 | 551 | &range->min.tcp.port); |
df6fb868 | 552 | NLA_PUT(skb, CTA_PROTONAT_PORT_MAX, sizeof(__be16), |
5b1158e9 JK |
553 | &range->max.tcp.port); |
554 | ||
555 | return 0; | |
556 | ||
df6fb868 | 557 | nla_put_failure: |
5b1158e9 JK |
558 | return -1; |
559 | } | |
fdf70832 | 560 | EXPORT_SYMBOL_GPL(nf_nat_port_nlattr_to_range); |
5b1158e9 JK |
561 | |
562 | int | |
fdf70832 | 563 | nf_nat_port_nlattr_to_range(struct nlattr *tb[], struct nf_nat_range *range) |
5b1158e9 JK |
564 | { |
565 | int ret = 0; | |
566 | ||
567 | /* we have to return whether we actually parsed something or not */ | |
568 | ||
df6fb868 | 569 | if (tb[CTA_PROTONAT_PORT_MIN]) { |
5b1158e9 JK |
570 | ret = 1; |
571 | range->min.tcp.port = | |
df6fb868 | 572 | *(__be16 *)nla_data(tb[CTA_PROTONAT_PORT_MIN]); |
5b1158e9 JK |
573 | } |
574 | ||
df6fb868 | 575 | if (!tb[CTA_PROTONAT_PORT_MAX]) { |
5b1158e9 JK |
576 | if (ret) |
577 | range->max.tcp.port = range->min.tcp.port; | |
578 | } else { | |
579 | ret = 1; | |
580 | range->max.tcp.port = | |
df6fb868 | 581 | *(__be16 *)nla_data(tb[CTA_PROTONAT_PORT_MAX]); |
5b1158e9 JK |
582 | } |
583 | ||
584 | return ret; | |
585 | } | |
fdf70832 | 586 | EXPORT_SYMBOL_GPL(nf_nat_port_range_to_nlattr); |
5b1158e9 JK |
587 | #endif |
588 | ||
d8a0509a YK |
589 | /* Noone using conntrack by the time this called. */ |
590 | static void nf_nat_cleanup_conntrack(struct nf_conn *ct) | |
591 | { | |
592 | struct nf_conn_nat *nat = nf_ct_ext_find(ct, NF_CT_EXT_NAT); | |
593 | ||
b6b84d4a | 594 | if (nat == NULL || nat->ct == NULL) |
d8a0509a YK |
595 | return; |
596 | ||
b6b84d4a | 597 | NF_CT_ASSERT(nat->ct->status & IPS_NAT_DONE_MASK); |
d8a0509a YK |
598 | |
599 | write_lock_bh(&nf_nat_lock); | |
53aba597 | 600 | hlist_del(&nat->bysource); |
b6b84d4a | 601 | nat->ct = NULL; |
d8a0509a YK |
602 | write_unlock_bh(&nf_nat_lock); |
603 | } | |
604 | ||
2d59e5ca YK |
605 | static void nf_nat_move_storage(struct nf_conn *conntrack, void *old) |
606 | { | |
607 | struct nf_conn_nat *new_nat = nf_ct_ext_find(conntrack, NF_CT_EXT_NAT); | |
608 | struct nf_conn_nat *old_nat = (struct nf_conn_nat *)old; | |
b6b84d4a | 609 | struct nf_conn *ct = old_nat->ct; |
2d59e5ca YK |
610 | unsigned int srchash; |
611 | ||
612 | if (!(ct->status & IPS_NAT_DONE_MASK)) | |
613 | return; | |
614 | ||
615 | srchash = hash_by_src(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
616 | ||
617 | write_lock_bh(&nf_nat_lock); | |
53aba597 | 618 | hlist_replace_rcu(&old_nat->bysource, &new_nat->bysource); |
b6b84d4a | 619 | new_nat->ct = ct; |
2d59e5ca YK |
620 | write_unlock_bh(&nf_nat_lock); |
621 | } | |
622 | ||
61eb3107 | 623 | static struct nf_ct_ext_type nat_extend __read_mostly = { |
d8a0509a YK |
624 | .len = sizeof(struct nf_conn_nat), |
625 | .align = __alignof__(struct nf_conn_nat), | |
626 | .destroy = nf_nat_cleanup_conntrack, | |
627 | .move = nf_nat_move_storage, | |
628 | .id = NF_CT_EXT_NAT, | |
629 | .flags = NF_CT_EXT_F_PREALLOC, | |
2d59e5ca YK |
630 | }; |
631 | ||
5b1158e9 JK |
632 | static int __init nf_nat_init(void) |
633 | { | |
634 | size_t i; | |
2d59e5ca YK |
635 | int ret; |
636 | ||
637 | ret = nf_ct_extend_register(&nat_extend); | |
638 | if (ret < 0) { | |
639 | printk(KERN_ERR "nf_nat_core: Unable to register extension\n"); | |
640 | return ret; | |
641 | } | |
5b1158e9 JK |
642 | |
643 | /* Leave them the same for the moment. */ | |
644 | nf_nat_htable_size = nf_conntrack_htable_size; | |
645 | ||
53aba597 PM |
646 | bysource = nf_ct_alloc_hashtable(&nf_nat_htable_size, |
647 | &nf_nat_vmalloced); | |
2d59e5ca YK |
648 | if (!bysource) { |
649 | ret = -ENOMEM; | |
650 | goto cleanup_extend; | |
651 | } | |
5b1158e9 JK |
652 | |
653 | /* Sew in builtin protocols. */ | |
654 | write_lock_bh(&nf_nat_lock); | |
655 | for (i = 0; i < MAX_IP_NAT_PROTO; i++) | |
e22a0548 PM |
656 | rcu_assign_pointer(nf_nat_protos[i], &nf_nat_unknown_protocol); |
657 | rcu_assign_pointer(nf_nat_protos[IPPROTO_TCP], &nf_nat_protocol_tcp); | |
658 | rcu_assign_pointer(nf_nat_protos[IPPROTO_UDP], &nf_nat_protocol_udp); | |
659 | rcu_assign_pointer(nf_nat_protos[IPPROTO_ICMP], &nf_nat_protocol_icmp); | |
5b1158e9 JK |
660 | write_unlock_bh(&nf_nat_lock); |
661 | ||
662 | for (i = 0; i < nf_nat_htable_size; i++) { | |
53aba597 | 663 | INIT_HLIST_HEAD(&bysource[i]); |
5b1158e9 JK |
664 | } |
665 | ||
5b1158e9 JK |
666 | /* Initialize fake conntrack so that NAT will skip it */ |
667 | nf_conntrack_untracked.status |= IPS_NAT_DONE_MASK; | |
668 | ||
669 | l3proto = nf_ct_l3proto_find_get((u_int16_t)AF_INET); | |
670 | return 0; | |
2d59e5ca YK |
671 | |
672 | cleanup_extend: | |
673 | nf_ct_extend_unregister(&nat_extend); | |
674 | return ret; | |
5b1158e9 JK |
675 | } |
676 | ||
677 | /* Clear NAT section of all conntracks, in case we're loaded again. */ | |
678 | static int clean_nat(struct nf_conn *i, void *data) | |
679 | { | |
680 | struct nf_conn_nat *nat = nfct_nat(i); | |
681 | ||
682 | if (!nat) | |
683 | return 0; | |
684 | memset(nat, 0, sizeof(nat)); | |
685 | i->status &= ~(IPS_NAT_MASK | IPS_NAT_DONE_MASK | IPS_SEQ_ADJUST); | |
686 | return 0; | |
687 | } | |
688 | ||
689 | static void __exit nf_nat_cleanup(void) | |
690 | { | |
691 | nf_ct_iterate_cleanup(&clean_nat, NULL); | |
982d9a9c | 692 | synchronize_rcu(); |
53aba597 | 693 | nf_ct_free_hashtable(bysource, nf_nat_vmalloced, nf_nat_htable_size); |
5b1158e9 | 694 | nf_ct_l3proto_put(l3proto); |
2d59e5ca | 695 | nf_ct_extend_unregister(&nat_extend); |
5b1158e9 JK |
696 | } |
697 | ||
698 | MODULE_LICENSE("GPL"); | |
699 | ||
700 | module_init(nf_nat_init); | |
701 | module_exit(nf_nat_cleanup); |