]>
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 | ||
25 | #define ASSERT_READ_LOCK(x) MUST_BE_READ_LOCKED(&ip_nat_lock) | |
26 | #define ASSERT_WRITE_LOCK(x) MUST_BE_WRITE_LOCKED(&ip_nat_lock) | |
27 | ||
28 | #include <linux/netfilter_ipv4/ip_conntrack.h> | |
29 | #include <linux/netfilter_ipv4/ip_conntrack_core.h> | |
30 | #include <linux/netfilter_ipv4/ip_conntrack_protocol.h> | |
31 | #include <linux/netfilter_ipv4/ip_nat.h> | |
32 | #include <linux/netfilter_ipv4/ip_nat_protocol.h> | |
33 | #include <linux/netfilter_ipv4/ip_nat_core.h> | |
34 | #include <linux/netfilter_ipv4/ip_nat_helper.h> | |
35 | #include <linux/netfilter_ipv4/ip_conntrack_helper.h> | |
36 | #include <linux/netfilter_ipv4/listhelp.h> | |
37 | ||
38 | #if 0 | |
39 | #define DEBUGP printk | |
40 | #else | |
41 | #define DEBUGP(format, args...) | |
42 | #endif | |
43 | ||
44 | DECLARE_RWLOCK(ip_nat_lock); | |
45 | ||
46 | /* Calculated at init based on memory size */ | |
47 | static unsigned int ip_nat_htable_size; | |
48 | ||
49 | static struct list_head *bysource; | |
50 | struct ip_nat_protocol *ip_nat_protos[MAX_IP_NAT_PROTO]; | |
51 | ||
52 | ||
53 | /* We keep an extra hash for each conntrack, for fast searching. */ | |
54 | static inline unsigned int | |
55 | hash_by_src(const struct ip_conntrack_tuple *tuple) | |
56 | { | |
57 | /* Original src, to ensure we map it consistently if poss. */ | |
58 | return jhash_3words(tuple->src.ip, tuple->src.u.all, | |
59 | tuple->dst.protonum, 0) % ip_nat_htable_size; | |
60 | } | |
61 | ||
62 | /* Noone using conntrack by the time this called. */ | |
63 | static void ip_nat_cleanup_conntrack(struct ip_conntrack *conn) | |
64 | { | |
65 | if (!(conn->status & IPS_NAT_DONE_MASK)) | |
66 | return; | |
67 | ||
68 | WRITE_LOCK(&ip_nat_lock); | |
69 | list_del(&conn->nat.info.bysource); | |
70 | WRITE_UNLOCK(&ip_nat_lock); | |
71 | } | |
72 | ||
73 | /* We do checksum mangling, so if they were wrong before they're still | |
74 | * wrong. Also works for incomplete packets (eg. ICMP dest | |
75 | * unreachables.) */ | |
76 | u_int16_t | |
77 | ip_nat_cheat_check(u_int32_t oldvalinv, u_int32_t newval, u_int16_t oldcheck) | |
78 | { | |
79 | u_int32_t diffs[] = { oldvalinv, newval }; | |
80 | return csum_fold(csum_partial((char *)diffs, sizeof(diffs), | |
81 | oldcheck^0xFFFF)); | |
82 | } | |
83 | ||
84 | /* Is this tuple already taken? (not by us) */ | |
85 | int | |
86 | ip_nat_used_tuple(const struct ip_conntrack_tuple *tuple, | |
87 | const struct ip_conntrack *ignored_conntrack) | |
88 | { | |
89 | /* Conntrack tracking doesn't keep track of outgoing tuples; only | |
90 | incoming ones. NAT means they don't have a fixed mapping, | |
91 | so we invert the tuple and look for the incoming reply. | |
92 | ||
93 | We could keep a separate hash if this proves too slow. */ | |
94 | struct ip_conntrack_tuple reply; | |
95 | ||
96 | invert_tuplepr(&reply, tuple); | |
97 | return ip_conntrack_tuple_taken(&reply, ignored_conntrack); | |
98 | } | |
99 | ||
100 | /* If we source map this tuple so reply looks like reply_tuple, will | |
101 | * that meet the constraints of range. */ | |
102 | static int | |
103 | in_range(const struct ip_conntrack_tuple *tuple, | |
104 | const struct ip_nat_range *range) | |
105 | { | |
106 | struct ip_nat_protocol *proto = ip_nat_find_proto(tuple->dst.protonum); | |
107 | ||
108 | /* If we are supposed to map IPs, then we must be in the | |
109 | range specified, otherwise let this drag us onto a new src IP. */ | |
110 | if (range->flags & IP_NAT_RANGE_MAP_IPS) { | |
111 | if (ntohl(tuple->src.ip) < ntohl(range->min_ip) | |
112 | || ntohl(tuple->src.ip) > ntohl(range->max_ip)) | |
113 | return 0; | |
114 | } | |
115 | ||
116 | if (!(range->flags & IP_NAT_RANGE_PROTO_SPECIFIED) | |
117 | || proto->in_range(tuple, IP_NAT_MANIP_SRC, | |
118 | &range->min, &range->max)) | |
119 | return 1; | |
120 | ||
121 | return 0; | |
122 | } | |
123 | ||
124 | static inline int | |
125 | same_src(const struct ip_conntrack *ct, | |
126 | const struct ip_conntrack_tuple *tuple) | |
127 | { | |
128 | return (ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum | |
129 | == tuple->dst.protonum | |
130 | && ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.ip | |
131 | == tuple->src.ip | |
132 | && ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u.all | |
133 | == tuple->src.u.all); | |
134 | } | |
135 | ||
136 | /* Only called for SRC manip */ | |
137 | static int | |
138 | find_appropriate_src(const struct ip_conntrack_tuple *tuple, | |
139 | struct ip_conntrack_tuple *result, | |
140 | const struct ip_nat_range *range) | |
141 | { | |
142 | unsigned int h = hash_by_src(tuple); | |
143 | struct ip_conntrack *ct; | |
144 | ||
145 | READ_LOCK(&ip_nat_lock); | |
146 | list_for_each_entry(ct, &bysource[h], nat.info.bysource) { | |
147 | if (same_src(ct, tuple)) { | |
148 | /* Copy source part from reply tuple. */ | |
149 | invert_tuplepr(result, | |
150 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
151 | result->dst = tuple->dst; | |
152 | ||
153 | if (in_range(result, range)) { | |
154 | READ_UNLOCK(&ip_nat_lock); | |
155 | return 1; | |
156 | } | |
157 | } | |
158 | } | |
159 | READ_UNLOCK(&ip_nat_lock); | |
160 | return 0; | |
161 | } | |
162 | ||
163 | /* For [FUTURE] fragmentation handling, we want the least-used | |
164 | src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus | |
165 | if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports | |
166 | 1-65535, we don't do pro-rata allocation based on ports; we choose | |
167 | the ip with the lowest src-ip/dst-ip/proto usage. | |
168 | */ | |
169 | static void | |
170 | find_best_ips_proto(struct ip_conntrack_tuple *tuple, | |
171 | const struct ip_nat_range *range, | |
172 | const struct ip_conntrack *conntrack, | |
173 | enum ip_nat_manip_type maniptype) | |
174 | { | |
175 | u_int32_t *var_ipp; | |
176 | /* Host order */ | |
177 | u_int32_t minip, maxip, j; | |
178 | ||
179 | /* No IP mapping? Do nothing. */ | |
180 | if (!(range->flags & IP_NAT_RANGE_MAP_IPS)) | |
181 | return; | |
182 | ||
183 | if (maniptype == IP_NAT_MANIP_SRC) | |
184 | var_ipp = &tuple->src.ip; | |
185 | else | |
186 | var_ipp = &tuple->dst.ip; | |
187 | ||
188 | /* Fast path: only one choice. */ | |
189 | if (range->min_ip == range->max_ip) { | |
190 | *var_ipp = range->min_ip; | |
191 | return; | |
192 | } | |
193 | ||
194 | /* Hashing source and destination IPs gives a fairly even | |
195 | * spread in practice (if there are a small number of IPs | |
196 | * involved, there usually aren't that many connections | |
197 | * anyway). The consistency means that servers see the same | |
198 | * client coming from the same IP (some Internet Banking sites | |
199 | * like this), even across reboots. */ | |
200 | minip = ntohl(range->min_ip); | |
201 | maxip = ntohl(range->max_ip); | |
202 | j = jhash_2words(tuple->src.ip, tuple->dst.ip, 0); | |
203 | *var_ipp = htonl(minip + j % (maxip - minip + 1)); | |
204 | } | |
205 | ||
206 | /* Manipulate the tuple into the range given. For NF_IP_POST_ROUTING, | |
207 | * we change the source to map into the range. For NF_IP_PRE_ROUTING | |
208 | * and NF_IP_LOCAL_OUT, we change the destination to map into the | |
209 | * range. It might not be possible to get a unique tuple, but we try. | |
210 | * At worst (or if we race), we will end up with a final duplicate in | |
211 | * __ip_conntrack_confirm and drop the packet. */ | |
212 | static void | |
213 | get_unique_tuple(struct ip_conntrack_tuple *tuple, | |
214 | const struct ip_conntrack_tuple *orig_tuple, | |
215 | const struct ip_nat_range *range, | |
216 | struct ip_conntrack *conntrack, | |
217 | enum ip_nat_manip_type maniptype) | |
218 | { | |
219 | struct ip_nat_protocol *proto | |
220 | = ip_nat_find_proto(orig_tuple->dst.protonum); | |
221 | ||
222 | /* 1) If this srcip/proto/src-proto-part is currently mapped, | |
223 | and that same mapping gives a unique tuple within the given | |
224 | range, use that. | |
225 | ||
226 | This is only required for source (ie. NAT/masq) mappings. | |
227 | So far, we don't do local source mappings, so multiple | |
228 | manips not an issue. */ | |
229 | if (maniptype == IP_NAT_MANIP_SRC) { | |
230 | if (find_appropriate_src(orig_tuple, tuple, range)) { | |
231 | DEBUGP("get_unique_tuple: Found current src map\n"); | |
232 | if (!ip_nat_used_tuple(tuple, conntrack)) | |
233 | return; | |
234 | } | |
235 | } | |
236 | ||
237 | /* 2) Select the least-used IP/proto combination in the given | |
238 | range. */ | |
239 | *tuple = *orig_tuple; | |
240 | find_best_ips_proto(tuple, range, conntrack, maniptype); | |
241 | ||
242 | /* 3) The per-protocol part of the manip is made to map into | |
243 | the range to make a unique tuple. */ | |
244 | ||
245 | /* Only bother mapping if it's not already in range and unique */ | |
246 | if ((!(range->flags & IP_NAT_RANGE_PROTO_SPECIFIED) | |
247 | || proto->in_range(tuple, maniptype, &range->min, &range->max)) | |
248 | && !ip_nat_used_tuple(tuple, conntrack)) | |
249 | return; | |
250 | ||
251 | /* Last change: get protocol to try to obtain unique tuple. */ | |
252 | proto->unique_tuple(tuple, range, maniptype, conntrack); | |
253 | } | |
254 | ||
255 | unsigned int | |
256 | ip_nat_setup_info(struct ip_conntrack *conntrack, | |
257 | const struct ip_nat_range *range, | |
258 | unsigned int hooknum) | |
259 | { | |
260 | struct ip_conntrack_tuple curr_tuple, new_tuple; | |
261 | struct ip_nat_info *info = &conntrack->nat.info; | |
262 | int have_to_hash = !(conntrack->status & IPS_NAT_DONE_MASK); | |
263 | enum ip_nat_manip_type maniptype = HOOK2MANIP(hooknum); | |
264 | ||
265 | IP_NF_ASSERT(hooknum == NF_IP_PRE_ROUTING | |
266 | || hooknum == NF_IP_POST_ROUTING | |
267 | || hooknum == NF_IP_LOCAL_IN | |
268 | || hooknum == NF_IP_LOCAL_OUT); | |
269 | BUG_ON(ip_nat_initialized(conntrack, maniptype)); | |
270 | ||
271 | /* What we've got will look like inverse of reply. Normally | |
272 | this is what is in the conntrack, except for prior | |
273 | manipulations (future optimization: if num_manips == 0, | |
274 | orig_tp = | |
275 | conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple) */ | |
276 | invert_tuplepr(&curr_tuple, | |
277 | &conntrack->tuplehash[IP_CT_DIR_REPLY].tuple); | |
278 | ||
279 | get_unique_tuple(&new_tuple, &curr_tuple, range, conntrack, maniptype); | |
280 | ||
281 | if (!ip_ct_tuple_equal(&new_tuple, &curr_tuple)) { | |
282 | struct ip_conntrack_tuple reply; | |
283 | ||
284 | /* Alter conntrack table so will recognize replies. */ | |
285 | invert_tuplepr(&reply, &new_tuple); | |
286 | ip_conntrack_alter_reply(conntrack, &reply); | |
287 | ||
288 | /* Non-atomic: we own this at the moment. */ | |
289 | if (maniptype == IP_NAT_MANIP_SRC) | |
290 | conntrack->status |= IPS_SRC_NAT; | |
291 | else | |
292 | conntrack->status |= IPS_DST_NAT; | |
293 | } | |
294 | ||
295 | /* Place in source hash if this is the first time. */ | |
296 | if (have_to_hash) { | |
297 | unsigned int srchash | |
298 | = hash_by_src(&conntrack->tuplehash[IP_CT_DIR_ORIGINAL] | |
299 | .tuple); | |
300 | WRITE_LOCK(&ip_nat_lock); | |
301 | list_add(&info->bysource, &bysource[srchash]); | |
302 | WRITE_UNLOCK(&ip_nat_lock); | |
303 | } | |
304 | ||
305 | /* It's done. */ | |
306 | if (maniptype == IP_NAT_MANIP_DST) | |
307 | set_bit(IPS_DST_NAT_DONE_BIT, &conntrack->status); | |
308 | else | |
309 | set_bit(IPS_SRC_NAT_DONE_BIT, &conntrack->status); | |
310 | ||
311 | return NF_ACCEPT; | |
312 | } | |
313 | ||
314 | /* Returns true if succeeded. */ | |
315 | static int | |
316 | manip_pkt(u_int16_t proto, | |
317 | struct sk_buff **pskb, | |
318 | unsigned int iphdroff, | |
319 | const struct ip_conntrack_tuple *target, | |
320 | enum ip_nat_manip_type maniptype) | |
321 | { | |
322 | struct iphdr *iph; | |
323 | ||
324 | (*pskb)->nfcache |= NFC_ALTERED; | |
325 | if (!skb_ip_make_writable(pskb, iphdroff + sizeof(*iph))) | |
326 | return 0; | |
327 | ||
328 | iph = (void *)(*pskb)->data + iphdroff; | |
329 | ||
330 | /* Manipulate protcol part. */ | |
331 | if (!ip_nat_find_proto(proto)->manip_pkt(pskb, iphdroff, | |
332 | target, maniptype)) | |
333 | return 0; | |
334 | ||
335 | iph = (void *)(*pskb)->data + iphdroff; | |
336 | ||
337 | if (maniptype == IP_NAT_MANIP_SRC) { | |
338 | iph->check = ip_nat_cheat_check(~iph->saddr, target->src.ip, | |
339 | iph->check); | |
340 | iph->saddr = target->src.ip; | |
341 | } else { | |
342 | iph->check = ip_nat_cheat_check(~iph->daddr, target->dst.ip, | |
343 | iph->check); | |
344 | iph->daddr = target->dst.ip; | |
345 | } | |
346 | return 1; | |
347 | } | |
348 | ||
349 | /* Do packet manipulations according to ip_nat_setup_info. */ | |
350 | unsigned int nat_packet(struct ip_conntrack *ct, | |
351 | enum ip_conntrack_info ctinfo, | |
352 | unsigned int hooknum, | |
353 | struct sk_buff **pskb) | |
354 | { | |
355 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); | |
356 | unsigned long statusbit; | |
357 | enum ip_nat_manip_type mtype = HOOK2MANIP(hooknum); | |
358 | ||
359 | if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) | |
360 | && (hooknum == NF_IP_POST_ROUTING || hooknum == NF_IP_LOCAL_IN)) { | |
361 | DEBUGP("ip_nat_core: adjusting sequence number\n"); | |
362 | /* future: put this in a l4-proto specific function, | |
363 | * and call this function here. */ | |
364 | if (!ip_nat_seq_adjust(pskb, ct, ctinfo)) | |
365 | return NF_DROP; | |
366 | } | |
367 | ||
368 | if (mtype == IP_NAT_MANIP_SRC) | |
369 | statusbit = IPS_SRC_NAT; | |
370 | else | |
371 | statusbit = IPS_DST_NAT; | |
372 | ||
373 | /* Invert if this is reply dir. */ | |
374 | if (dir == IP_CT_DIR_REPLY) | |
375 | statusbit ^= IPS_NAT_MASK; | |
376 | ||
377 | /* Non-atomic: these bits don't change. */ | |
378 | if (ct->status & statusbit) { | |
379 | struct ip_conntrack_tuple target; | |
380 | ||
381 | /* We are aiming to look like inverse of other direction. */ | |
382 | invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); | |
383 | ||
384 | if (!manip_pkt(target.dst.protonum, pskb, 0, &target, mtype)) | |
385 | return NF_DROP; | |
386 | } | |
387 | return NF_ACCEPT; | |
388 | } | |
389 | ||
390 | /* Dir is direction ICMP is coming from (opposite to packet it contains) */ | |
391 | int icmp_reply_translation(struct sk_buff **pskb, | |
392 | struct ip_conntrack *ct, | |
393 | enum ip_nat_manip_type manip, | |
394 | enum ip_conntrack_dir dir) | |
395 | { | |
396 | struct { | |
397 | struct icmphdr icmp; | |
398 | struct iphdr ip; | |
399 | } *inside; | |
400 | struct ip_conntrack_tuple inner, target; | |
401 | int hdrlen = (*pskb)->nh.iph->ihl * 4; | |
402 | ||
403 | if (!skb_ip_make_writable(pskb, hdrlen + sizeof(*inside))) | |
404 | return 0; | |
405 | ||
406 | inside = (void *)(*pskb)->data + (*pskb)->nh.iph->ihl*4; | |
407 | ||
408 | /* We're actually going to mangle it beyond trivial checksum | |
409 | adjustment, so make sure the current checksum is correct. */ | |
410 | if ((*pskb)->ip_summed != CHECKSUM_UNNECESSARY) { | |
411 | hdrlen = (*pskb)->nh.iph->ihl * 4; | |
412 | if ((u16)csum_fold(skb_checksum(*pskb, hdrlen, | |
413 | (*pskb)->len - hdrlen, 0))) | |
414 | return 0; | |
415 | } | |
416 | ||
417 | /* Must be RELATED */ | |
418 | IP_NF_ASSERT((*pskb)->nfctinfo == IP_CT_RELATED || | |
419 | (*pskb)->nfctinfo == IP_CT_RELATED+IP_CT_IS_REPLY); | |
420 | ||
421 | /* Redirects on non-null nats must be dropped, else they'll | |
422 | start talking to each other without our translation, and be | |
423 | confused... --RR */ | |
424 | if (inside->icmp.type == ICMP_REDIRECT) { | |
425 | /* If NAT isn't finished, assume it and drop. */ | |
426 | if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK) | |
427 | return 0; | |
428 | ||
429 | if (ct->status & IPS_NAT_MASK) | |
430 | return 0; | |
431 | } | |
432 | ||
433 | DEBUGP("icmp_reply_translation: translating error %p manp %u dir %s\n", | |
434 | *pskb, manip, dir == IP_CT_DIR_ORIGINAL ? "ORIG" : "REPLY"); | |
435 | ||
436 | if (!ip_ct_get_tuple(&inside->ip, *pskb, (*pskb)->nh.iph->ihl*4 + | |
437 | sizeof(struct icmphdr) + inside->ip.ihl*4, | |
438 | &inner, ip_ct_find_proto(inside->ip.protocol))) | |
439 | return 0; | |
440 | ||
441 | /* Change inner back to look like incoming packet. We do the | |
442 | opposite manip on this hook to normal, because it might not | |
443 | pass all hooks (locally-generated ICMP). Consider incoming | |
444 | packet: PREROUTING (DST manip), routing produces ICMP, goes | |
445 | through POSTROUTING (which must correct the DST manip). */ | |
446 | if (!manip_pkt(inside->ip.protocol, pskb, | |
447 | (*pskb)->nh.iph->ihl*4 | |
448 | + sizeof(inside->icmp), | |
449 | &ct->tuplehash[!dir].tuple, | |
450 | !manip)) | |
451 | return 0; | |
452 | ||
453 | /* Reloading "inside" here since manip_pkt inner. */ | |
454 | inside = (void *)(*pskb)->data + (*pskb)->nh.iph->ihl*4; | |
455 | inside->icmp.checksum = 0; | |
456 | inside->icmp.checksum = csum_fold(skb_checksum(*pskb, hdrlen, | |
457 | (*pskb)->len - hdrlen, | |
458 | 0)); | |
459 | ||
460 | /* Change outer to look the reply to an incoming packet | |
461 | * (proto 0 means don't invert per-proto part). */ | |
462 | ||
463 | /* Obviously, we need to NAT destination IP, but source IP | |
464 | should be NAT'ed only if it is from a NAT'd host. | |
465 | ||
466 | Explanation: some people use NAT for anonymizing. Also, | |
467 | CERT recommends dropping all packets from private IP | |
468 | addresses (although ICMP errors from internal links with | |
469 | such addresses are not too uncommon, as Alan Cox points | |
470 | out) */ | |
471 | if (manip != IP_NAT_MANIP_SRC | |
472 | || ((*pskb)->nh.iph->saddr == ct->tuplehash[dir].tuple.src.ip)) { | |
473 | invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); | |
474 | if (!manip_pkt(0, pskb, 0, &target, manip)) | |
475 | return 0; | |
476 | } | |
477 | ||
478 | return 1; | |
479 | } | |
480 | ||
481 | /* Protocol registration. */ | |
482 | int ip_nat_protocol_register(struct ip_nat_protocol *proto) | |
483 | { | |
484 | int ret = 0; | |
485 | ||
486 | WRITE_LOCK(&ip_nat_lock); | |
487 | if (ip_nat_protos[proto->protonum] != &ip_nat_unknown_protocol) { | |
488 | ret = -EBUSY; | |
489 | goto out; | |
490 | } | |
491 | ip_nat_protos[proto->protonum] = proto; | |
492 | out: | |
493 | WRITE_UNLOCK(&ip_nat_lock); | |
494 | return ret; | |
495 | } | |
496 | ||
497 | /* Noone stores the protocol anywhere; simply delete it. */ | |
498 | void ip_nat_protocol_unregister(struct ip_nat_protocol *proto) | |
499 | { | |
500 | WRITE_LOCK(&ip_nat_lock); | |
501 | ip_nat_protos[proto->protonum] = &ip_nat_unknown_protocol; | |
502 | WRITE_UNLOCK(&ip_nat_lock); | |
503 | ||
504 | /* Someone could be still looking at the proto in a bh. */ | |
505 | synchronize_net(); | |
506 | } | |
507 | ||
508 | int __init ip_nat_init(void) | |
509 | { | |
510 | size_t i; | |
511 | ||
512 | /* Leave them the same for the moment. */ | |
513 | ip_nat_htable_size = ip_conntrack_htable_size; | |
514 | ||
515 | /* One vmalloc for both hash tables */ | |
516 | bysource = vmalloc(sizeof(struct list_head) * ip_nat_htable_size); | |
517 | if (!bysource) | |
518 | return -ENOMEM; | |
519 | ||
520 | /* Sew in builtin protocols. */ | |
521 | WRITE_LOCK(&ip_nat_lock); | |
522 | for (i = 0; i < MAX_IP_NAT_PROTO; i++) | |
523 | ip_nat_protos[i] = &ip_nat_unknown_protocol; | |
524 | ip_nat_protos[IPPROTO_TCP] = &ip_nat_protocol_tcp; | |
525 | ip_nat_protos[IPPROTO_UDP] = &ip_nat_protocol_udp; | |
526 | ip_nat_protos[IPPROTO_ICMP] = &ip_nat_protocol_icmp; | |
527 | WRITE_UNLOCK(&ip_nat_lock); | |
528 | ||
529 | for (i = 0; i < ip_nat_htable_size; i++) { | |
530 | INIT_LIST_HEAD(&bysource[i]); | |
531 | } | |
532 | ||
533 | /* FIXME: Man, this is a hack. <SIGH> */ | |
534 | IP_NF_ASSERT(ip_conntrack_destroyed == NULL); | |
535 | ip_conntrack_destroyed = &ip_nat_cleanup_conntrack; | |
536 | ||
537 | /* Initialize fake conntrack so that NAT will skip it */ | |
538 | ip_conntrack_untracked.status |= IPS_NAT_DONE_MASK; | |
539 | return 0; | |
540 | } | |
541 | ||
542 | /* Clear NAT section of all conntracks, in case we're loaded again. */ | |
543 | static int clean_nat(struct ip_conntrack *i, void *data) | |
544 | { | |
545 | memset(&i->nat, 0, sizeof(i->nat)); | |
546 | i->status &= ~(IPS_NAT_MASK | IPS_NAT_DONE_MASK | IPS_SEQ_ADJUST); | |
547 | return 0; | |
548 | } | |
549 | ||
550 | /* Not __exit: called from ip_nat_standalone.c:init_or_cleanup() --RR */ | |
551 | void ip_nat_cleanup(void) | |
552 | { | |
553 | ip_ct_iterate_cleanup(&clean_nat, NULL); | |
554 | ip_conntrack_destroyed = NULL; | |
555 | vfree(bysource); | |
556 | } |