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> | |
5a0e3ad6 | 15 | #include <linux/gfp.h> |
5b1158e9 JK |
16 | #include <net/checksum.h> |
17 | #include <net/icmp.h> | |
18 | #include <net/ip.h> | |
19 | #include <net/tcp.h> /* For tcp_prot in getorigdst */ | |
20 | #include <linux/icmp.h> | |
21 | #include <linux/udp.h> | |
22 | #include <linux/jhash.h> | |
23 | ||
24 | #include <linux/netfilter_ipv4.h> | |
25 | #include <net/netfilter/nf_conntrack.h> | |
26 | #include <net/netfilter/nf_conntrack_core.h> | |
27 | #include <net/netfilter/nf_nat.h> | |
28 | #include <net/netfilter/nf_nat_protocol.h> | |
29 | #include <net/netfilter/nf_nat_core.h> | |
30 | #include <net/netfilter/nf_nat_helper.h> | |
31 | #include <net/netfilter/nf_conntrack_helper.h> | |
32 | #include <net/netfilter/nf_conntrack_l3proto.h> | |
33 | #include <net/netfilter/nf_conntrack_l4proto.h> | |
5d0aa2cc | 34 | #include <net/netfilter/nf_conntrack_zones.h> |
5b1158e9 | 35 | |
02502f62 | 36 | static DEFINE_SPINLOCK(nf_nat_lock); |
5b1158e9 | 37 | |
ce4b1ceb | 38 | static struct nf_conntrack_l3proto *l3proto __read_mostly; |
5b1158e9 | 39 | |
5b1158e9 | 40 | #define MAX_IP_NAT_PROTO 256 |
0906a372 | 41 | static const struct nf_nat_protocol __rcu *nf_nat_protos[MAX_IP_NAT_PROTO] |
ce4b1ceb | 42 | __read_mostly; |
5b1158e9 | 43 | |
2b628a08 | 44 | static inline const struct nf_nat_protocol * |
5b1158e9 JK |
45 | __nf_nat_proto_find(u_int8_t protonum) |
46 | { | |
e22a0548 | 47 | return rcu_dereference(nf_nat_protos[protonum]); |
5b1158e9 JK |
48 | } |
49 | ||
5b1158e9 JK |
50 | /* We keep an extra hash for each conntrack, for fast searching. */ |
51 | static inline unsigned int | |
5d0aa2cc PM |
52 | hash_by_src(const struct net *net, u16 zone, |
53 | const struct nf_conntrack_tuple *tuple) | |
5b1158e9 | 54 | { |
34498825 PM |
55 | unsigned int hash; |
56 | ||
5b1158e9 | 57 | /* Original src, to ensure we map it consistently if poss. */ |
34498825 | 58 | hash = jhash_3words((__force u32)tuple->src.u3.ip, |
5d0aa2cc | 59 | (__force u32)tuple->src.u.all ^ zone, |
34498825 | 60 | tuple->dst.protonum, 0); |
d696c7bd | 61 | return ((u64)hash * net->ipv4.nat_htable_size) >> 32; |
5b1158e9 JK |
62 | } |
63 | ||
5b1158e9 JK |
64 | /* Is this tuple already taken? (not by us) */ |
65 | int | |
66 | nf_nat_used_tuple(const struct nf_conntrack_tuple *tuple, | |
67 | const struct nf_conn *ignored_conntrack) | |
68 | { | |
69 | /* Conntrack tracking doesn't keep track of outgoing tuples; only | |
70 | incoming ones. NAT means they don't have a fixed mapping, | |
71 | so we invert the tuple and look for the incoming reply. | |
72 | ||
73 | We could keep a separate hash if this proves too slow. */ | |
74 | struct nf_conntrack_tuple reply; | |
75 | ||
76 | nf_ct_invert_tuplepr(&reply, tuple); | |
77 | return nf_conntrack_tuple_taken(&reply, ignored_conntrack); | |
78 | } | |
79 | EXPORT_SYMBOL(nf_nat_used_tuple); | |
80 | ||
81 | /* If we source map this tuple so reply looks like reply_tuple, will | |
82 | * that meet the constraints of range. */ | |
83 | static int | |
84 | in_range(const struct nf_conntrack_tuple *tuple, | |
85 | const struct nf_nat_range *range) | |
86 | { | |
2b628a08 | 87 | const struct nf_nat_protocol *proto; |
e22a0548 | 88 | int ret = 0; |
5b1158e9 | 89 | |
5b1158e9 JK |
90 | /* If we are supposed to map IPs, then we must be in the |
91 | range specified, otherwise let this drag us onto a new src IP. */ | |
92 | if (range->flags & IP_NAT_RANGE_MAP_IPS) { | |
93 | if (ntohl(tuple->src.u3.ip) < ntohl(range->min_ip) || | |
94 | ntohl(tuple->src.u3.ip) > ntohl(range->max_ip)) | |
95 | return 0; | |
96 | } | |
97 | ||
e22a0548 PM |
98 | rcu_read_lock(); |
99 | proto = __nf_nat_proto_find(tuple->dst.protonum); | |
5b1158e9 JK |
100 | if (!(range->flags & IP_NAT_RANGE_PROTO_SPECIFIED) || |
101 | proto->in_range(tuple, IP_NAT_MANIP_SRC, | |
102 | &range->min, &range->max)) | |
e22a0548 PM |
103 | ret = 1; |
104 | rcu_read_unlock(); | |
5b1158e9 | 105 | |
e22a0548 | 106 | return ret; |
5b1158e9 JK |
107 | } |
108 | ||
109 | static inline int | |
110 | same_src(const struct nf_conn *ct, | |
111 | const struct nf_conntrack_tuple *tuple) | |
112 | { | |
113 | const struct nf_conntrack_tuple *t; | |
114 | ||
115 | t = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; | |
116 | return (t->dst.protonum == tuple->dst.protonum && | |
117 | t->src.u3.ip == tuple->src.u3.ip && | |
118 | t->src.u.all == tuple->src.u.all); | |
119 | } | |
120 | ||
121 | /* Only called for SRC manip */ | |
122 | static int | |
5d0aa2cc | 123 | find_appropriate_src(struct net *net, u16 zone, |
0c4c9288 | 124 | const struct nf_conntrack_tuple *tuple, |
5b1158e9 JK |
125 | struct nf_conntrack_tuple *result, |
126 | const struct nf_nat_range *range) | |
127 | { | |
5d0aa2cc | 128 | unsigned int h = hash_by_src(net, zone, tuple); |
72b72949 JE |
129 | const struct nf_conn_nat *nat; |
130 | const struct nf_conn *ct; | |
131 | const struct hlist_node *n; | |
5b1158e9 | 132 | |
4d354c57 | 133 | rcu_read_lock(); |
0c4c9288 | 134 | hlist_for_each_entry_rcu(nat, n, &net->ipv4.nat_bysource[h], bysource) { |
b6b84d4a | 135 | ct = nat->ct; |
5d0aa2cc | 136 | if (same_src(ct, tuple) && nf_ct_zone(ct) == zone) { |
5b1158e9 JK |
137 | /* Copy source part from reply tuple. */ |
138 | nf_ct_invert_tuplepr(result, | |
139 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
140 | result->dst = tuple->dst; | |
141 | ||
142 | if (in_range(result, range)) { | |
4d354c57 | 143 | rcu_read_unlock(); |
5b1158e9 JK |
144 | return 1; |
145 | } | |
146 | } | |
147 | } | |
4d354c57 | 148 | rcu_read_unlock(); |
5b1158e9 JK |
149 | return 0; |
150 | } | |
151 | ||
152 | /* For [FUTURE] fragmentation handling, we want the least-used | |
153 | src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus | |
154 | if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports | |
155 | 1-65535, we don't do pro-rata allocation based on ports; we choose | |
156 | the ip with the lowest src-ip/dst-ip/proto usage. | |
157 | */ | |
158 | static void | |
5d0aa2cc | 159 | find_best_ips_proto(u16 zone, struct nf_conntrack_tuple *tuple, |
5b1158e9 JK |
160 | const struct nf_nat_range *range, |
161 | const struct nf_conn *ct, | |
162 | enum nf_nat_manip_type maniptype) | |
163 | { | |
164 | __be32 *var_ipp; | |
165 | /* Host order */ | |
166 | u_int32_t minip, maxip, j; | |
167 | ||
168 | /* No IP mapping? Do nothing. */ | |
169 | if (!(range->flags & IP_NAT_RANGE_MAP_IPS)) | |
170 | return; | |
171 | ||
172 | if (maniptype == IP_NAT_MANIP_SRC) | |
173 | var_ipp = &tuple->src.u3.ip; | |
174 | else | |
175 | var_ipp = &tuple->dst.u3.ip; | |
176 | ||
177 | /* Fast path: only one choice. */ | |
178 | if (range->min_ip == range->max_ip) { | |
179 | *var_ipp = range->min_ip; | |
180 | return; | |
181 | } | |
182 | ||
183 | /* Hashing source and destination IPs gives a fairly even | |
184 | * spread in practice (if there are a small number of IPs | |
185 | * involved, there usually aren't that many connections | |
186 | * anyway). The consistency means that servers see the same | |
187 | * client coming from the same IP (some Internet Banking sites | |
188 | * like this), even across reboots. */ | |
189 | minip = ntohl(range->min_ip); | |
190 | maxip = ntohl(range->max_ip); | |
191 | j = jhash_2words((__force u32)tuple->src.u3.ip, | |
98d500d6 | 192 | range->flags & IP_NAT_RANGE_PERSISTENT ? |
5d0aa2cc | 193 | 0 : (__force u32)tuple->dst.u3.ip ^ zone, 0); |
34498825 PM |
194 | j = ((u64)j * (maxip - minip + 1)) >> 32; |
195 | *var_ipp = htonl(minip + j); | |
5b1158e9 JK |
196 | } |
197 | ||
6e23ae2a PM |
198 | /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING, |
199 | * we change the source to map into the range. For NF_INET_PRE_ROUTING | |
200 | * and NF_INET_LOCAL_OUT, we change the destination to map into the | |
5b1158e9 JK |
201 | * range. It might not be possible to get a unique tuple, but we try. |
202 | * At worst (or if we race), we will end up with a final duplicate in | |
203 | * __ip_conntrack_confirm and drop the packet. */ | |
204 | static void | |
205 | get_unique_tuple(struct nf_conntrack_tuple *tuple, | |
206 | const struct nf_conntrack_tuple *orig_tuple, | |
207 | const struct nf_nat_range *range, | |
208 | struct nf_conn *ct, | |
209 | enum nf_nat_manip_type maniptype) | |
210 | { | |
0c4c9288 | 211 | struct net *net = nf_ct_net(ct); |
2b628a08 | 212 | const struct nf_nat_protocol *proto; |
5d0aa2cc | 213 | u16 zone = nf_ct_zone(ct); |
5b1158e9 JK |
214 | |
215 | /* 1) If this srcip/proto/src-proto-part is currently mapped, | |
216 | and that same mapping gives a unique tuple within the given | |
217 | range, use that. | |
218 | ||
219 | This is only required for source (ie. NAT/masq) mappings. | |
220 | So far, we don't do local source mappings, so multiple | |
221 | manips not an issue. */ | |
0dbff689 CG |
222 | if (maniptype == IP_NAT_MANIP_SRC && |
223 | !(range->flags & IP_NAT_RANGE_PROTO_RANDOM)) { | |
5d0aa2cc | 224 | if (find_appropriate_src(net, zone, orig_tuple, tuple, range)) { |
0d53778e | 225 | pr_debug("get_unique_tuple: Found current src map\n"); |
0dbff689 CG |
226 | if (!nf_nat_used_tuple(tuple, ct)) |
227 | return; | |
5b1158e9 JK |
228 | } |
229 | } | |
230 | ||
231 | /* 2) Select the least-used IP/proto combination in the given | |
232 | range. */ | |
233 | *tuple = *orig_tuple; | |
5d0aa2cc | 234 | find_best_ips_proto(zone, tuple, range, ct, maniptype); |
5b1158e9 JK |
235 | |
236 | /* 3) The per-protocol part of the manip is made to map into | |
237 | the range to make a unique tuple. */ | |
238 | ||
e22a0548 PM |
239 | rcu_read_lock(); |
240 | proto = __nf_nat_proto_find(orig_tuple->dst.protonum); | |
5b1158e9 JK |
241 | |
242 | /* Only bother mapping if it's not already in range and unique */ | |
99ad3c53 CG |
243 | if (!(range->flags & IP_NAT_RANGE_PROTO_RANDOM)) { |
244 | if (range->flags & IP_NAT_RANGE_PROTO_SPECIFIED) { | |
245 | if (proto->in_range(tuple, maniptype, &range->min, | |
246 | &range->max) && | |
247 | (range->min.all == range->max.all || | |
248 | !nf_nat_used_tuple(tuple, ct))) | |
249 | goto out; | |
250 | } else if (!nf_nat_used_tuple(tuple, ct)) { | |
251 | goto out; | |
252 | } | |
253 | } | |
5b1158e9 JK |
254 | |
255 | /* Last change: get protocol to try to obtain unique tuple. */ | |
256 | proto->unique_tuple(tuple, range, maniptype, ct); | |
e22a0548 PM |
257 | out: |
258 | rcu_read_unlock(); | |
5b1158e9 JK |
259 | } |
260 | ||
261 | unsigned int | |
262 | nf_nat_setup_info(struct nf_conn *ct, | |
263 | const struct nf_nat_range *range, | |
cc01dcbd | 264 | enum nf_nat_manip_type maniptype) |
5b1158e9 | 265 | { |
0c4c9288 | 266 | struct net *net = nf_ct_net(ct); |
5b1158e9 | 267 | struct nf_conntrack_tuple curr_tuple, new_tuple; |
2d59e5ca | 268 | struct nf_conn_nat *nat; |
5b1158e9 | 269 | int have_to_hash = !(ct->status & IPS_NAT_DONE_MASK); |
5b1158e9 | 270 | |
2d59e5ca YK |
271 | /* nat helper or nfctnetlink also setup binding */ |
272 | nat = nfct_nat(ct); | |
273 | if (!nat) { | |
274 | nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC); | |
275 | if (nat == NULL) { | |
0d53778e | 276 | pr_debug("failed to add NAT extension\n"); |
2d59e5ca YK |
277 | return NF_ACCEPT; |
278 | } | |
279 | } | |
280 | ||
cc01dcbd PM |
281 | NF_CT_ASSERT(maniptype == IP_NAT_MANIP_SRC || |
282 | maniptype == IP_NAT_MANIP_DST); | |
5b1158e9 JK |
283 | BUG_ON(nf_nat_initialized(ct, maniptype)); |
284 | ||
285 | /* What we've got will look like inverse of reply. Normally | |
286 | this is what is in the conntrack, except for prior | |
287 | manipulations (future optimization: if num_manips == 0, | |
288 | orig_tp = | |
289 | conntrack->tuplehash[IP_CT_DIR_ORIGINAL].tuple) */ | |
290 | nf_ct_invert_tuplepr(&curr_tuple, | |
291 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
292 | ||
293 | get_unique_tuple(&new_tuple, &curr_tuple, range, ct, maniptype); | |
294 | ||
295 | if (!nf_ct_tuple_equal(&new_tuple, &curr_tuple)) { | |
296 | struct nf_conntrack_tuple reply; | |
297 | ||
298 | /* Alter conntrack table so will recognize replies. */ | |
299 | nf_ct_invert_tuplepr(&reply, &new_tuple); | |
300 | nf_conntrack_alter_reply(ct, &reply); | |
301 | ||
302 | /* Non-atomic: we own this at the moment. */ | |
303 | if (maniptype == IP_NAT_MANIP_SRC) | |
304 | ct->status |= IPS_SRC_NAT; | |
305 | else | |
306 | ct->status |= IPS_DST_NAT; | |
307 | } | |
308 | ||
309 | /* Place in source hash if this is the first time. */ | |
310 | if (have_to_hash) { | |
311 | unsigned int srchash; | |
312 | ||
5d0aa2cc PM |
313 | srchash = hash_by_src(net, nf_ct_zone(ct), |
314 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
02502f62 | 315 | spin_lock_bh(&nf_nat_lock); |
2d59e5ca | 316 | /* nf_conntrack_alter_reply might re-allocate exntension aera */ |
b6b84d4a YK |
317 | nat = nfct_nat(ct); |
318 | nat->ct = ct; | |
0c4c9288 AD |
319 | hlist_add_head_rcu(&nat->bysource, |
320 | &net->ipv4.nat_bysource[srchash]); | |
02502f62 | 321 | spin_unlock_bh(&nf_nat_lock); |
5b1158e9 JK |
322 | } |
323 | ||
324 | /* It's done. */ | |
325 | if (maniptype == IP_NAT_MANIP_DST) | |
76a2d3bc | 326 | ct->status |= IPS_DST_NAT_DONE_BIT; |
5b1158e9 | 327 | else |
76a2d3bc | 328 | ct->status |= IPS_SRC_NAT_DONE_BIT; |
5b1158e9 JK |
329 | |
330 | return NF_ACCEPT; | |
331 | } | |
332 | EXPORT_SYMBOL(nf_nat_setup_info); | |
333 | ||
334 | /* Returns true if succeeded. */ | |
f2ea825f | 335 | static bool |
5b1158e9 | 336 | manip_pkt(u_int16_t proto, |
3db05fea | 337 | struct sk_buff *skb, |
5b1158e9 JK |
338 | unsigned int iphdroff, |
339 | const struct nf_conntrack_tuple *target, | |
340 | enum nf_nat_manip_type maniptype) | |
341 | { | |
342 | struct iphdr *iph; | |
2b628a08 | 343 | const struct nf_nat_protocol *p; |
5b1158e9 | 344 | |
3db05fea | 345 | if (!skb_make_writable(skb, iphdroff + sizeof(*iph))) |
f2ea825f | 346 | return false; |
5b1158e9 | 347 | |
3db05fea | 348 | iph = (void *)skb->data + iphdroff; |
5b1158e9 JK |
349 | |
350 | /* Manipulate protcol part. */ | |
e22a0548 PM |
351 | |
352 | /* rcu_read_lock()ed by nf_hook_slow */ | |
353 | p = __nf_nat_proto_find(proto); | |
3db05fea | 354 | if (!p->manip_pkt(skb, iphdroff, target, maniptype)) |
f2ea825f | 355 | return false; |
5b1158e9 | 356 | |
3db05fea | 357 | iph = (void *)skb->data + iphdroff; |
5b1158e9 JK |
358 | |
359 | if (maniptype == IP_NAT_MANIP_SRC) { | |
be0ea7d5 | 360 | csum_replace4(&iph->check, iph->saddr, target->src.u3.ip); |
5b1158e9 JK |
361 | iph->saddr = target->src.u3.ip; |
362 | } else { | |
be0ea7d5 | 363 | csum_replace4(&iph->check, iph->daddr, target->dst.u3.ip); |
5b1158e9 JK |
364 | iph->daddr = target->dst.u3.ip; |
365 | } | |
f2ea825f | 366 | return true; |
5b1158e9 JK |
367 | } |
368 | ||
369 | /* Do packet manipulations according to nf_nat_setup_info. */ | |
370 | unsigned int nf_nat_packet(struct nf_conn *ct, | |
371 | enum ip_conntrack_info ctinfo, | |
372 | unsigned int hooknum, | |
3db05fea | 373 | struct sk_buff *skb) |
5b1158e9 JK |
374 | { |
375 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); | |
376 | unsigned long statusbit; | |
377 | enum nf_nat_manip_type mtype = HOOK2MANIP(hooknum); | |
378 | ||
379 | if (mtype == IP_NAT_MANIP_SRC) | |
380 | statusbit = IPS_SRC_NAT; | |
381 | else | |
382 | statusbit = IPS_DST_NAT; | |
383 | ||
384 | /* Invert if this is reply dir. */ | |
385 | if (dir == IP_CT_DIR_REPLY) | |
386 | statusbit ^= IPS_NAT_MASK; | |
387 | ||
388 | /* Non-atomic: these bits don't change. */ | |
389 | if (ct->status & statusbit) { | |
390 | struct nf_conntrack_tuple target; | |
391 | ||
392 | /* We are aiming to look like inverse of other direction. */ | |
393 | nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); | |
394 | ||
3db05fea | 395 | if (!manip_pkt(target.dst.protonum, skb, 0, &target, mtype)) |
5b1158e9 JK |
396 | return NF_DROP; |
397 | } | |
398 | return NF_ACCEPT; | |
399 | } | |
400 | EXPORT_SYMBOL_GPL(nf_nat_packet); | |
401 | ||
402 | /* Dir is direction ICMP is coming from (opposite to packet it contains) */ | |
403 | int nf_nat_icmp_reply_translation(struct nf_conn *ct, | |
404 | enum ip_conntrack_info ctinfo, | |
405 | unsigned int hooknum, | |
3db05fea | 406 | struct sk_buff *skb) |
5b1158e9 JK |
407 | { |
408 | struct { | |
409 | struct icmphdr icmp; | |
410 | struct iphdr ip; | |
411 | } *inside; | |
72b72949 | 412 | const struct nf_conntrack_l4proto *l4proto; |
5b1158e9 | 413 | struct nf_conntrack_tuple inner, target; |
3db05fea | 414 | int hdrlen = ip_hdrlen(skb); |
5b1158e9 JK |
415 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); |
416 | unsigned long statusbit; | |
417 | enum nf_nat_manip_type manip = HOOK2MANIP(hooknum); | |
418 | ||
3db05fea | 419 | if (!skb_make_writable(skb, hdrlen + sizeof(*inside))) |
5b1158e9 JK |
420 | return 0; |
421 | ||
794dbc1d | 422 | inside = (void *)skb->data + hdrlen; |
5b1158e9 JK |
423 | |
424 | /* We're actually going to mangle it beyond trivial checksum | |
425 | adjustment, so make sure the current checksum is correct. */ | |
3db05fea | 426 | if (nf_ip_checksum(skb, hooknum, hdrlen, 0)) |
5b1158e9 JK |
427 | return 0; |
428 | ||
429 | /* Must be RELATED */ | |
3db05fea HX |
430 | NF_CT_ASSERT(skb->nfctinfo == IP_CT_RELATED || |
431 | skb->nfctinfo == IP_CT_RELATED+IP_CT_IS_REPLY); | |
5b1158e9 JK |
432 | |
433 | /* Redirects on non-null nats must be dropped, else they'll | |
e905a9ed YH |
434 | start talking to each other without our translation, and be |
435 | confused... --RR */ | |
5b1158e9 JK |
436 | if (inside->icmp.type == ICMP_REDIRECT) { |
437 | /* If NAT isn't finished, assume it and drop. */ | |
438 | if ((ct->status & IPS_NAT_DONE_MASK) != IPS_NAT_DONE_MASK) | |
439 | return 0; | |
440 | ||
441 | if (ct->status & IPS_NAT_MASK) | |
442 | return 0; | |
443 | } | |
444 | ||
b0aeef30 JA |
445 | if (manip == IP_NAT_MANIP_SRC) |
446 | statusbit = IPS_SRC_NAT; | |
447 | else | |
448 | statusbit = IPS_DST_NAT; | |
449 | ||
450 | /* Invert if this is reply dir. */ | |
451 | if (dir == IP_CT_DIR_REPLY) | |
452 | statusbit ^= IPS_NAT_MASK; | |
453 | ||
454 | if (!(ct->status & statusbit)) | |
455 | return 1; | |
456 | ||
0d53778e | 457 | pr_debug("icmp_reply_translation: translating error %p manip %u " |
3db05fea | 458 | "dir %s\n", skb, manip, |
0d53778e | 459 | dir == IP_CT_DIR_ORIGINAL ? "ORIG" : "REPLY"); |
5b1158e9 | 460 | |
923f4902 PM |
461 | /* rcu_read_lock()ed by nf_hook_slow */ |
462 | l4proto = __nf_ct_l4proto_find(PF_INET, inside->ip.protocol); | |
463 | ||
794dbc1d CG |
464 | if (!nf_ct_get_tuple(skb, hdrlen + sizeof(struct icmphdr), |
465 | (hdrlen + | |
c9bdd4b5 | 466 | sizeof(struct icmphdr) + inside->ip.ihl * 4), |
794dbc1d | 467 | (u_int16_t)AF_INET, inside->ip.protocol, |
923f4902 | 468 | &inner, l3proto, l4proto)) |
5b1158e9 JK |
469 | return 0; |
470 | ||
471 | /* Change inner back to look like incoming packet. We do the | |
472 | opposite manip on this hook to normal, because it might not | |
473 | pass all hooks (locally-generated ICMP). Consider incoming | |
474 | packet: PREROUTING (DST manip), routing produces ICMP, goes | |
475 | through POSTROUTING (which must correct the DST manip). */ | |
794dbc1d CG |
476 | if (!manip_pkt(inside->ip.protocol, skb, hdrlen + sizeof(inside->icmp), |
477 | &ct->tuplehash[!dir].tuple, !manip)) | |
5b1158e9 JK |
478 | return 0; |
479 | ||
3db05fea | 480 | if (skb->ip_summed != CHECKSUM_PARTIAL) { |
5b1158e9 | 481 | /* Reloading "inside" here since manip_pkt inner. */ |
794dbc1d | 482 | inside = (void *)skb->data + hdrlen; |
5b1158e9 JK |
483 | inside->icmp.checksum = 0; |
484 | inside->icmp.checksum = | |
3db05fea HX |
485 | csum_fold(skb_checksum(skb, hdrlen, |
486 | skb->len - hdrlen, 0)); | |
5b1158e9 JK |
487 | } |
488 | ||
489 | /* Change outer to look the reply to an incoming packet | |
490 | * (proto 0 means don't invert per-proto part). */ | |
b0aeef30 JA |
491 | nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); |
492 | if (!manip_pkt(0, skb, 0, &target, manip)) | |
493 | return 0; | |
5b1158e9 JK |
494 | |
495 | return 1; | |
496 | } | |
497 | EXPORT_SYMBOL_GPL(nf_nat_icmp_reply_translation); | |
498 | ||
499 | /* Protocol registration. */ | |
2b628a08 | 500 | int nf_nat_protocol_register(const struct nf_nat_protocol *proto) |
5b1158e9 JK |
501 | { |
502 | int ret = 0; | |
503 | ||
02502f62 | 504 | spin_lock_bh(&nf_nat_lock); |
eb733162 ED |
505 | if (rcu_dereference_protected( |
506 | nf_nat_protos[proto->protonum], | |
507 | lockdep_is_held(&nf_nat_lock) | |
508 | ) != &nf_nat_unknown_protocol) { | |
5b1158e9 JK |
509 | ret = -EBUSY; |
510 | goto out; | |
511 | } | |
e22a0548 | 512 | rcu_assign_pointer(nf_nat_protos[proto->protonum], proto); |
5b1158e9 | 513 | out: |
02502f62 | 514 | spin_unlock_bh(&nf_nat_lock); |
5b1158e9 JK |
515 | return ret; |
516 | } | |
517 | EXPORT_SYMBOL(nf_nat_protocol_register); | |
518 | ||
519 | /* Noone stores the protocol anywhere; simply delete it. */ | |
2b628a08 | 520 | void nf_nat_protocol_unregister(const struct nf_nat_protocol *proto) |
5b1158e9 | 521 | { |
02502f62 | 522 | spin_lock_bh(&nf_nat_lock); |
e22a0548 PM |
523 | rcu_assign_pointer(nf_nat_protos[proto->protonum], |
524 | &nf_nat_unknown_protocol); | |
02502f62 | 525 | spin_unlock_bh(&nf_nat_lock); |
e22a0548 | 526 | synchronize_rcu(); |
5b1158e9 JK |
527 | } |
528 | EXPORT_SYMBOL(nf_nat_protocol_unregister); | |
529 | ||
d8a0509a YK |
530 | /* Noone using conntrack by the time this called. */ |
531 | static void nf_nat_cleanup_conntrack(struct nf_conn *ct) | |
532 | { | |
533 | struct nf_conn_nat *nat = nf_ct_ext_find(ct, NF_CT_EXT_NAT); | |
534 | ||
b6b84d4a | 535 | if (nat == NULL || nat->ct == NULL) |
d8a0509a YK |
536 | return; |
537 | ||
b6b84d4a | 538 | NF_CT_ASSERT(nat->ct->status & IPS_NAT_DONE_MASK); |
d8a0509a | 539 | |
02502f62 | 540 | spin_lock_bh(&nf_nat_lock); |
4d354c57 | 541 | hlist_del_rcu(&nat->bysource); |
02502f62 | 542 | spin_unlock_bh(&nf_nat_lock); |
d8a0509a YK |
543 | } |
544 | ||
86577c66 | 545 | static void nf_nat_move_storage(void *new, void *old) |
2d59e5ca | 546 | { |
86577c66 PM |
547 | struct nf_conn_nat *new_nat = new; |
548 | struct nf_conn_nat *old_nat = old; | |
b6b84d4a | 549 | struct nf_conn *ct = old_nat->ct; |
2d59e5ca | 550 | |
1f305323 | 551 | if (!ct || !(ct->status & IPS_NAT_DONE_MASK)) |
2d59e5ca YK |
552 | return; |
553 | ||
02502f62 | 554 | spin_lock_bh(&nf_nat_lock); |
b6b84d4a | 555 | new_nat->ct = ct; |
68b80f11 | 556 | hlist_replace_rcu(&old_nat->bysource, &new_nat->bysource); |
02502f62 | 557 | spin_unlock_bh(&nf_nat_lock); |
2d59e5ca YK |
558 | } |
559 | ||
61eb3107 | 560 | static struct nf_ct_ext_type nat_extend __read_mostly = { |
d8a0509a YK |
561 | .len = sizeof(struct nf_conn_nat), |
562 | .align = __alignof__(struct nf_conn_nat), | |
563 | .destroy = nf_nat_cleanup_conntrack, | |
564 | .move = nf_nat_move_storage, | |
565 | .id = NF_CT_EXT_NAT, | |
566 | .flags = NF_CT_EXT_F_PREALLOC, | |
2d59e5ca YK |
567 | }; |
568 | ||
e6a7d3c0 PNA |
569 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
570 | ||
571 | #include <linux/netfilter/nfnetlink.h> | |
572 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
573 | ||
64e46749 PM |
574 | static const struct nf_nat_protocol * |
575 | nf_nat_proto_find_get(u_int8_t protonum) | |
576 | { | |
577 | const struct nf_nat_protocol *p; | |
578 | ||
579 | rcu_read_lock(); | |
580 | p = __nf_nat_proto_find(protonum); | |
581 | if (!try_module_get(p->me)) | |
582 | p = &nf_nat_unknown_protocol; | |
583 | rcu_read_unlock(); | |
584 | ||
585 | return p; | |
586 | } | |
587 | ||
588 | static void | |
589 | nf_nat_proto_put(const struct nf_nat_protocol *p) | |
590 | { | |
591 | module_put(p->me); | |
592 | } | |
593 | ||
e6a7d3c0 PNA |
594 | static const struct nla_policy protonat_nla_policy[CTA_PROTONAT_MAX+1] = { |
595 | [CTA_PROTONAT_PORT_MIN] = { .type = NLA_U16 }, | |
596 | [CTA_PROTONAT_PORT_MAX] = { .type = NLA_U16 }, | |
597 | }; | |
598 | ||
599 | static int nfnetlink_parse_nat_proto(struct nlattr *attr, | |
600 | const struct nf_conn *ct, | |
601 | struct nf_nat_range *range) | |
602 | { | |
603 | struct nlattr *tb[CTA_PROTONAT_MAX+1]; | |
604 | const struct nf_nat_protocol *npt; | |
605 | int err; | |
606 | ||
607 | err = nla_parse_nested(tb, CTA_PROTONAT_MAX, attr, protonat_nla_policy); | |
608 | if (err < 0) | |
609 | return err; | |
610 | ||
611 | npt = nf_nat_proto_find_get(nf_ct_protonum(ct)); | |
612 | if (npt->nlattr_to_range) | |
613 | err = npt->nlattr_to_range(tb, range); | |
614 | nf_nat_proto_put(npt); | |
615 | return err; | |
616 | } | |
617 | ||
618 | static const struct nla_policy nat_nla_policy[CTA_NAT_MAX+1] = { | |
619 | [CTA_NAT_MINIP] = { .type = NLA_U32 }, | |
620 | [CTA_NAT_MAXIP] = { .type = NLA_U32 }, | |
621 | }; | |
622 | ||
623 | static int | |
39938324 | 624 | nfnetlink_parse_nat(const struct nlattr *nat, |
e6a7d3c0 PNA |
625 | const struct nf_conn *ct, struct nf_nat_range *range) |
626 | { | |
627 | struct nlattr *tb[CTA_NAT_MAX+1]; | |
628 | int err; | |
629 | ||
630 | memset(range, 0, sizeof(*range)); | |
631 | ||
632 | err = nla_parse_nested(tb, CTA_NAT_MAX, nat, nat_nla_policy); | |
633 | if (err < 0) | |
634 | return err; | |
635 | ||
636 | if (tb[CTA_NAT_MINIP]) | |
637 | range->min_ip = nla_get_be32(tb[CTA_NAT_MINIP]); | |
638 | ||
639 | if (!tb[CTA_NAT_MAXIP]) | |
640 | range->max_ip = range->min_ip; | |
641 | else | |
642 | range->max_ip = nla_get_be32(tb[CTA_NAT_MAXIP]); | |
643 | ||
644 | if (range->min_ip) | |
645 | range->flags |= IP_NAT_RANGE_MAP_IPS; | |
646 | ||
647 | if (!tb[CTA_NAT_PROTO]) | |
648 | return 0; | |
649 | ||
650 | err = nfnetlink_parse_nat_proto(tb[CTA_NAT_PROTO], ct, range); | |
651 | if (err < 0) | |
652 | return err; | |
653 | ||
654 | return 0; | |
655 | } | |
656 | ||
657 | static int | |
658 | nfnetlink_parse_nat_setup(struct nf_conn *ct, | |
659 | enum nf_nat_manip_type manip, | |
39938324 | 660 | const struct nlattr *attr) |
e6a7d3c0 PNA |
661 | { |
662 | struct nf_nat_range range; | |
663 | ||
664 | if (nfnetlink_parse_nat(attr, ct, &range) < 0) | |
665 | return -EINVAL; | |
666 | if (nf_nat_initialized(ct, manip)) | |
667 | return -EEXIST; | |
668 | ||
669 | return nf_nat_setup_info(ct, &range, manip); | |
670 | } | |
671 | #else | |
672 | static int | |
673 | nfnetlink_parse_nat_setup(struct nf_conn *ct, | |
674 | enum nf_nat_manip_type manip, | |
39938324 | 675 | const struct nlattr *attr) |
e6a7d3c0 PNA |
676 | { |
677 | return -EOPNOTSUPP; | |
678 | } | |
679 | #endif | |
680 | ||
0c4c9288 AD |
681 | static int __net_init nf_nat_net_init(struct net *net) |
682 | { | |
d696c7bd PM |
683 | /* Leave them the same for the moment. */ |
684 | net->ipv4.nat_htable_size = net->ct.htable_size; | |
d862a662 | 685 | net->ipv4.nat_bysource = nf_ct_alloc_hashtable(&net->ipv4.nat_htable_size, 0); |
0c4c9288 AD |
686 | if (!net->ipv4.nat_bysource) |
687 | return -ENOMEM; | |
688 | return 0; | |
689 | } | |
690 | ||
691 | /* Clear NAT section of all conntracks, in case we're loaded again. */ | |
692 | static int clean_nat(struct nf_conn *i, void *data) | |
693 | { | |
694 | struct nf_conn_nat *nat = nfct_nat(i); | |
695 | ||
696 | if (!nat) | |
697 | return 0; | |
698 | memset(nat, 0, sizeof(*nat)); | |
699 | i->status &= ~(IPS_NAT_MASK | IPS_NAT_DONE_MASK | IPS_SEQ_ADJUST); | |
700 | return 0; | |
701 | } | |
702 | ||
703 | static void __net_exit nf_nat_net_exit(struct net *net) | |
704 | { | |
705 | nf_ct_iterate_cleanup(net, &clean_nat, NULL); | |
706 | synchronize_rcu(); | |
d862a662 | 707 | nf_ct_free_hashtable(net->ipv4.nat_bysource, net->ipv4.nat_htable_size); |
0c4c9288 AD |
708 | } |
709 | ||
710 | static struct pernet_operations nf_nat_net_ops = { | |
711 | .init = nf_nat_net_init, | |
712 | .exit = nf_nat_net_exit, | |
713 | }; | |
714 | ||
5b1158e9 JK |
715 | static int __init nf_nat_init(void) |
716 | { | |
717 | size_t i; | |
2d59e5ca YK |
718 | int ret; |
719 | ||
475959d4 JE |
720 | need_ipv4_conntrack(); |
721 | ||
2d59e5ca YK |
722 | ret = nf_ct_extend_register(&nat_extend); |
723 | if (ret < 0) { | |
724 | printk(KERN_ERR "nf_nat_core: Unable to register extension\n"); | |
725 | return ret; | |
726 | } | |
5b1158e9 | 727 | |
0c4c9288 AD |
728 | ret = register_pernet_subsys(&nf_nat_net_ops); |
729 | if (ret < 0) | |
2d59e5ca | 730 | goto cleanup_extend; |
5b1158e9 JK |
731 | |
732 | /* Sew in builtin protocols. */ | |
02502f62 | 733 | spin_lock_bh(&nf_nat_lock); |
5b1158e9 | 734 | for (i = 0; i < MAX_IP_NAT_PROTO; i++) |
e22a0548 PM |
735 | rcu_assign_pointer(nf_nat_protos[i], &nf_nat_unknown_protocol); |
736 | rcu_assign_pointer(nf_nat_protos[IPPROTO_TCP], &nf_nat_protocol_tcp); | |
737 | rcu_assign_pointer(nf_nat_protos[IPPROTO_UDP], &nf_nat_protocol_udp); | |
738 | rcu_assign_pointer(nf_nat_protos[IPPROTO_ICMP], &nf_nat_protocol_icmp); | |
02502f62 | 739 | spin_unlock_bh(&nf_nat_lock); |
5b1158e9 | 740 | |
5b1158e9 | 741 | /* Initialize fake conntrack so that NAT will skip it */ |
5bfddbd4 | 742 | nf_ct_untracked_status_or(IPS_NAT_DONE_MASK); |
5b1158e9 JK |
743 | |
744 | l3proto = nf_ct_l3proto_find_get((u_int16_t)AF_INET); | |
dd13b010 PM |
745 | |
746 | BUG_ON(nf_nat_seq_adjust_hook != NULL); | |
747 | rcu_assign_pointer(nf_nat_seq_adjust_hook, nf_nat_seq_adjust); | |
e6a7d3c0 PNA |
748 | BUG_ON(nfnetlink_parse_nat_setup_hook != NULL); |
749 | rcu_assign_pointer(nfnetlink_parse_nat_setup_hook, | |
750 | nfnetlink_parse_nat_setup); | |
f9dd09c7 JK |
751 | BUG_ON(nf_ct_nat_offset != NULL); |
752 | rcu_assign_pointer(nf_ct_nat_offset, nf_nat_get_offset); | |
5b1158e9 | 753 | return 0; |
2d59e5ca YK |
754 | |
755 | cleanup_extend: | |
756 | nf_ct_extend_unregister(&nat_extend); | |
757 | return ret; | |
5b1158e9 JK |
758 | } |
759 | ||
5b1158e9 JK |
760 | static void __exit nf_nat_cleanup(void) |
761 | { | |
0c4c9288 | 762 | unregister_pernet_subsys(&nf_nat_net_ops); |
5b1158e9 | 763 | nf_ct_l3proto_put(l3proto); |
2d59e5ca | 764 | nf_ct_extend_unregister(&nat_extend); |
dd13b010 | 765 | rcu_assign_pointer(nf_nat_seq_adjust_hook, NULL); |
e6a7d3c0 | 766 | rcu_assign_pointer(nfnetlink_parse_nat_setup_hook, NULL); |
f9dd09c7 | 767 | rcu_assign_pointer(nf_ct_nat_offset, NULL); |
dd13b010 | 768 | synchronize_net(); |
5b1158e9 JK |
769 | } |
770 | ||
771 | MODULE_LICENSE("GPL"); | |
e6a7d3c0 | 772 | MODULE_ALIAS("nf-nat-ipv4"); |
5b1158e9 JK |
773 | |
774 | module_init(nf_nat_init); | |
775 | module_exit(nf_nat_cleanup); |