]>
Commit | Line | Data |
---|---|---|
c7232c99 PM |
1 | /* |
2 | * (C) 1999-2001 Paul `Rusty' Russell | |
5b1158e9 | 3 | * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org> |
c7232c99 | 4 | * (C) 2011 Patrick McHardy <kaber@trash.net> |
5b1158e9 JK |
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> |
c7232c99 | 16 | #include <net/xfrm.h> |
5b1158e9 | 17 | #include <linux/jhash.h> |
c7232c99 | 18 | #include <linux/rtnetlink.h> |
5b1158e9 | 19 | |
5b1158e9 JK |
20 | #include <net/netfilter/nf_conntrack.h> |
21 | #include <net/netfilter/nf_conntrack_core.h> | |
22 | #include <net/netfilter/nf_nat.h> | |
c7232c99 PM |
23 | #include <net/netfilter/nf_nat_l3proto.h> |
24 | #include <net/netfilter/nf_nat_l4proto.h> | |
5b1158e9 JK |
25 | #include <net/netfilter/nf_nat_core.h> |
26 | #include <net/netfilter/nf_nat_helper.h> | |
27 | #include <net/netfilter/nf_conntrack_helper.h> | |
28 | #include <net/netfilter/nf_conntrack_l3proto.h> | |
5d0aa2cc | 29 | #include <net/netfilter/nf_conntrack_zones.h> |
c7232c99 | 30 | #include <linux/netfilter/nf_nat.h> |
5b1158e9 | 31 | |
02502f62 | 32 | static DEFINE_SPINLOCK(nf_nat_lock); |
5b1158e9 | 33 | |
c7232c99 PM |
34 | static DEFINE_MUTEX(nf_nat_proto_mutex); |
35 | static const struct nf_nat_l3proto __rcu *nf_nat_l3protos[NFPROTO_NUMPROTO] | |
36 | __read_mostly; | |
37 | static const struct nf_nat_l4proto __rcu **nf_nat_l4protos[NFPROTO_NUMPROTO] | |
ce4b1ceb | 38 | __read_mostly; |
5b1158e9 | 39 | |
c7232c99 PM |
40 | |
41 | inline const struct nf_nat_l3proto * | |
42 | __nf_nat_l3proto_find(u8 family) | |
5b1158e9 | 43 | { |
c7232c99 | 44 | return rcu_dereference(nf_nat_l3protos[family]); |
5b1158e9 JK |
45 | } |
46 | ||
c7232c99 PM |
47 | inline const struct nf_nat_l4proto * |
48 | __nf_nat_l4proto_find(u8 family, u8 protonum) | |
49 | { | |
50 | return rcu_dereference(nf_nat_l4protos[family][protonum]); | |
51 | } | |
52 | EXPORT_SYMBOL_GPL(__nf_nat_l4proto_find); | |
53 | ||
54 | #ifdef CONFIG_XFRM | |
55 | static void __nf_nat_decode_session(struct sk_buff *skb, struct flowi *fl) | |
56 | { | |
57 | const struct nf_nat_l3proto *l3proto; | |
58 | const struct nf_conn *ct; | |
59 | enum ip_conntrack_info ctinfo; | |
60 | enum ip_conntrack_dir dir; | |
61 | unsigned long statusbit; | |
62 | u8 family; | |
63 | ||
64 | ct = nf_ct_get(skb, &ctinfo); | |
65 | if (ct == NULL) | |
66 | return; | |
67 | ||
68 | family = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num; | |
69 | rcu_read_lock(); | |
70 | l3proto = __nf_nat_l3proto_find(family); | |
71 | if (l3proto == NULL) | |
72 | goto out; | |
73 | ||
74 | dir = CTINFO2DIR(ctinfo); | |
75 | if (dir == IP_CT_DIR_ORIGINAL) | |
76 | statusbit = IPS_DST_NAT; | |
77 | else | |
78 | statusbit = IPS_SRC_NAT; | |
79 | ||
80 | l3proto->decode_session(skb, ct, dir, statusbit, fl); | |
81 | out: | |
82 | rcu_read_unlock(); | |
83 | } | |
84 | ||
85 | int nf_xfrm_me_harder(struct sk_buff *skb, unsigned int family) | |
86 | { | |
87 | struct flowi fl; | |
88 | unsigned int hh_len; | |
89 | struct dst_entry *dst; | |
aaa795ad | 90 | int err; |
c7232c99 | 91 | |
aaa795ad | 92 | err = xfrm_decode_session(skb, &fl, family); |
e7e6f630 | 93 | if (err < 0) |
aaa795ad | 94 | return err; |
c7232c99 PM |
95 | |
96 | dst = skb_dst(skb); | |
97 | if (dst->xfrm) | |
98 | dst = ((struct xfrm_dst *)dst)->route; | |
99 | dst_hold(dst); | |
100 | ||
101 | dst = xfrm_lookup(dev_net(dst->dev), dst, &fl, skb->sk, 0); | |
102 | if (IS_ERR(dst)) | |
aaa795ad | 103 | return PTR_ERR(dst); |
c7232c99 PM |
104 | |
105 | skb_dst_drop(skb); | |
106 | skb_dst_set(skb, dst); | |
107 | ||
108 | /* Change in oif may mean change in hh_len. */ | |
109 | hh_len = skb_dst(skb)->dev->hard_header_len; | |
110 | if (skb_headroom(skb) < hh_len && | |
111 | pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC)) | |
aaa795ad | 112 | return -ENOMEM; |
c7232c99 PM |
113 | return 0; |
114 | } | |
115 | EXPORT_SYMBOL(nf_xfrm_me_harder); | |
116 | #endif /* CONFIG_XFRM */ | |
117 | ||
5b1158e9 JK |
118 | /* We keep an extra hash for each conntrack, for fast searching. */ |
119 | static inline unsigned int | |
5d0aa2cc PM |
120 | hash_by_src(const struct net *net, u16 zone, |
121 | const struct nf_conntrack_tuple *tuple) | |
5b1158e9 | 122 | { |
34498825 PM |
123 | unsigned int hash; |
124 | ||
5b1158e9 | 125 | /* Original src, to ensure we map it consistently if poss. */ |
c7232c99 PM |
126 | hash = jhash2((u32 *)&tuple->src, sizeof(tuple->src) / sizeof(u32), |
127 | tuple->dst.protonum ^ zone ^ nf_conntrack_hash_rnd); | |
128 | return ((u64)hash * net->ct.nat_htable_size) >> 32; | |
5b1158e9 JK |
129 | } |
130 | ||
5b1158e9 JK |
131 | /* Is this tuple already taken? (not by us) */ |
132 | int | |
133 | nf_nat_used_tuple(const struct nf_conntrack_tuple *tuple, | |
134 | const struct nf_conn *ignored_conntrack) | |
135 | { | |
136 | /* Conntrack tracking doesn't keep track of outgoing tuples; only | |
c7232c99 PM |
137 | * incoming ones. NAT means they don't have a fixed mapping, |
138 | * so we invert the tuple and look for the incoming reply. | |
139 | * | |
140 | * We could keep a separate hash if this proves too slow. | |
141 | */ | |
5b1158e9 JK |
142 | struct nf_conntrack_tuple reply; |
143 | ||
144 | nf_ct_invert_tuplepr(&reply, tuple); | |
145 | return nf_conntrack_tuple_taken(&reply, ignored_conntrack); | |
146 | } | |
147 | EXPORT_SYMBOL(nf_nat_used_tuple); | |
148 | ||
149 | /* If we source map this tuple so reply looks like reply_tuple, will | |
c7232c99 PM |
150 | * that meet the constraints of range. |
151 | */ | |
152 | static int in_range(const struct nf_nat_l3proto *l3proto, | |
153 | const struct nf_nat_l4proto *l4proto, | |
154 | const struct nf_conntrack_tuple *tuple, | |
155 | const struct nf_nat_range *range) | |
5b1158e9 | 156 | { |
5b1158e9 | 157 | /* If we are supposed to map IPs, then we must be in the |
c7232c99 PM |
158 | * range specified, otherwise let this drag us onto a new src IP. |
159 | */ | |
160 | if (range->flags & NF_NAT_RANGE_MAP_IPS && | |
161 | !l3proto->in_range(tuple, range)) | |
162 | return 0; | |
5b1158e9 | 163 | |
cbc9f2f4 | 164 | if (!(range->flags & NF_NAT_RANGE_PROTO_SPECIFIED) || |
c7232c99 PM |
165 | l4proto->in_range(tuple, NF_NAT_MANIP_SRC, |
166 | &range->min_proto, &range->max_proto)) | |
167 | return 1; | |
5b1158e9 | 168 | |
c7232c99 | 169 | return 0; |
5b1158e9 JK |
170 | } |
171 | ||
172 | static inline int | |
173 | same_src(const struct nf_conn *ct, | |
174 | const struct nf_conntrack_tuple *tuple) | |
175 | { | |
176 | const struct nf_conntrack_tuple *t; | |
177 | ||
178 | t = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; | |
179 | return (t->dst.protonum == tuple->dst.protonum && | |
c7232c99 | 180 | nf_inet_addr_cmp(&t->src.u3, &tuple->src.u3) && |
5b1158e9 JK |
181 | t->src.u.all == tuple->src.u.all); |
182 | } | |
183 | ||
184 | /* Only called for SRC manip */ | |
185 | static int | |
5d0aa2cc | 186 | find_appropriate_src(struct net *net, u16 zone, |
c7232c99 PM |
187 | const struct nf_nat_l3proto *l3proto, |
188 | const struct nf_nat_l4proto *l4proto, | |
0c4c9288 | 189 | const struct nf_conntrack_tuple *tuple, |
5b1158e9 | 190 | struct nf_conntrack_tuple *result, |
c7232c99 | 191 | const struct nf_nat_range *range) |
5b1158e9 | 192 | { |
5d0aa2cc | 193 | unsigned int h = hash_by_src(net, zone, tuple); |
72b72949 JE |
194 | const struct nf_conn_nat *nat; |
195 | const struct nf_conn *ct; | |
5b1158e9 | 196 | |
b67bfe0d | 197 | hlist_for_each_entry_rcu(nat, &net->ct.nat_bysource[h], bysource) { |
b6b84d4a | 198 | ct = nat->ct; |
5d0aa2cc | 199 | if (same_src(ct, tuple) && nf_ct_zone(ct) == zone) { |
5b1158e9 JK |
200 | /* Copy source part from reply tuple. */ |
201 | nf_ct_invert_tuplepr(result, | |
202 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
203 | result->dst = tuple->dst; | |
204 | ||
136251d0 | 205 | if (in_range(l3proto, l4proto, result, range)) |
5b1158e9 | 206 | return 1; |
5b1158e9 JK |
207 | } |
208 | } | |
5b1158e9 JK |
209 | return 0; |
210 | } | |
211 | ||
212 | /* For [FUTURE] fragmentation handling, we want the least-used | |
c7232c99 PM |
213 | * src-ip/dst-ip/proto triple. Fairness doesn't come into it. Thus |
214 | * if the range specifies 1.2.3.4 ports 10000-10005 and 1.2.3.5 ports | |
215 | * 1-65535, we don't do pro-rata allocation based on ports; we choose | |
216 | * the ip with the lowest src-ip/dst-ip/proto usage. | |
217 | */ | |
5b1158e9 | 218 | static void |
5d0aa2cc | 219 | find_best_ips_proto(u16 zone, struct nf_conntrack_tuple *tuple, |
c7232c99 | 220 | const struct nf_nat_range *range, |
5b1158e9 JK |
221 | const struct nf_conn *ct, |
222 | enum nf_nat_manip_type maniptype) | |
223 | { | |
c7232c99 PM |
224 | union nf_inet_addr *var_ipp; |
225 | unsigned int i, max; | |
5b1158e9 | 226 | /* Host order */ |
c7232c99 PM |
227 | u32 minip, maxip, j, dist; |
228 | bool full_range; | |
5b1158e9 JK |
229 | |
230 | /* No IP mapping? Do nothing. */ | |
cbc9f2f4 | 231 | if (!(range->flags & NF_NAT_RANGE_MAP_IPS)) |
5b1158e9 JK |
232 | return; |
233 | ||
cbc9f2f4 | 234 | if (maniptype == NF_NAT_MANIP_SRC) |
c7232c99 | 235 | var_ipp = &tuple->src.u3; |
5b1158e9 | 236 | else |
c7232c99 | 237 | var_ipp = &tuple->dst.u3; |
5b1158e9 JK |
238 | |
239 | /* Fast path: only one choice. */ | |
c7232c99 PM |
240 | if (nf_inet_addr_cmp(&range->min_addr, &range->max_addr)) { |
241 | *var_ipp = range->min_addr; | |
5b1158e9 JK |
242 | return; |
243 | } | |
244 | ||
c7232c99 PM |
245 | if (nf_ct_l3num(ct) == NFPROTO_IPV4) |
246 | max = sizeof(var_ipp->ip) / sizeof(u32) - 1; | |
247 | else | |
248 | max = sizeof(var_ipp->ip6) / sizeof(u32) - 1; | |
249 | ||
5b1158e9 JK |
250 | /* Hashing source and destination IPs gives a fairly even |
251 | * spread in practice (if there are a small number of IPs | |
252 | * involved, there usually aren't that many connections | |
253 | * anyway). The consistency means that servers see the same | |
254 | * client coming from the same IP (some Internet Banking sites | |
c7232c99 PM |
255 | * like this), even across reboots. |
256 | */ | |
5693d68d | 257 | j = jhash2((u32 *)&tuple->src.u3, sizeof(tuple->src.u3) / sizeof(u32), |
c7232c99 PM |
258 | range->flags & NF_NAT_RANGE_PERSISTENT ? |
259 | 0 : (__force u32)tuple->dst.u3.all[max] ^ zone); | |
260 | ||
261 | full_range = false; | |
262 | for (i = 0; i <= max; i++) { | |
263 | /* If first bytes of the address are at the maximum, use the | |
264 | * distance. Otherwise use the full range. | |
265 | */ | |
266 | if (!full_range) { | |
267 | minip = ntohl((__force __be32)range->min_addr.all[i]); | |
268 | maxip = ntohl((__force __be32)range->max_addr.all[i]); | |
269 | dist = maxip - minip + 1; | |
270 | } else { | |
271 | minip = 0; | |
272 | dist = ~0; | |
273 | } | |
274 | ||
275 | var_ipp->all[i] = (__force __u32) | |
276 | htonl(minip + (((u64)j * dist) >> 32)); | |
277 | if (var_ipp->all[i] != range->max_addr.all[i]) | |
278 | full_range = true; | |
279 | ||
280 | if (!(range->flags & NF_NAT_RANGE_PERSISTENT)) | |
281 | j ^= (__force u32)tuple->dst.u3.all[i]; | |
282 | } | |
5b1158e9 JK |
283 | } |
284 | ||
c7232c99 PM |
285 | /* Manipulate the tuple into the range given. For NF_INET_POST_ROUTING, |
286 | * we change the source to map into the range. For NF_INET_PRE_ROUTING | |
6e23ae2a | 287 | * and NF_INET_LOCAL_OUT, we change the destination to map into the |
c7232c99 | 288 | * range. It might not be possible to get a unique tuple, but we try. |
5b1158e9 JK |
289 | * At worst (or if we race), we will end up with a final duplicate in |
290 | * __ip_conntrack_confirm and drop the packet. */ | |
291 | static void | |
292 | get_unique_tuple(struct nf_conntrack_tuple *tuple, | |
293 | const struct nf_conntrack_tuple *orig_tuple, | |
c7232c99 | 294 | const struct nf_nat_range *range, |
5b1158e9 JK |
295 | struct nf_conn *ct, |
296 | enum nf_nat_manip_type maniptype) | |
297 | { | |
c7232c99 PM |
298 | const struct nf_nat_l3proto *l3proto; |
299 | const struct nf_nat_l4proto *l4proto; | |
0c4c9288 | 300 | struct net *net = nf_ct_net(ct); |
5d0aa2cc | 301 | u16 zone = nf_ct_zone(ct); |
5b1158e9 | 302 | |
c7232c99 PM |
303 | rcu_read_lock(); |
304 | l3proto = __nf_nat_l3proto_find(orig_tuple->src.l3num); | |
305 | l4proto = __nf_nat_l4proto_find(orig_tuple->src.l3num, | |
306 | orig_tuple->dst.protonum); | |
5b1158e9 | 307 | |
c7232c99 PM |
308 | /* 1) If this srcip/proto/src-proto-part is currently mapped, |
309 | * and that same mapping gives a unique tuple within the given | |
310 | * range, use that. | |
311 | * | |
312 | * This is only required for source (ie. NAT/masq) mappings. | |
313 | * So far, we don't do local source mappings, so multiple | |
314 | * manips not an issue. | |
315 | */ | |
cbc9f2f4 PM |
316 | if (maniptype == NF_NAT_MANIP_SRC && |
317 | !(range->flags & NF_NAT_RANGE_PROTO_RANDOM)) { | |
41a7cab6 | 318 | /* try the original tuple first */ |
c7232c99 | 319 | if (in_range(l3proto, l4proto, orig_tuple, range)) { |
41a7cab6 CG |
320 | if (!nf_nat_used_tuple(orig_tuple, ct)) { |
321 | *tuple = *orig_tuple; | |
c7232c99 | 322 | goto out; |
41a7cab6 | 323 | } |
c7232c99 PM |
324 | } else if (find_appropriate_src(net, zone, l3proto, l4proto, |
325 | orig_tuple, tuple, range)) { | |
0d53778e | 326 | pr_debug("get_unique_tuple: Found current src map\n"); |
0dbff689 | 327 | if (!nf_nat_used_tuple(tuple, ct)) |
c7232c99 | 328 | goto out; |
5b1158e9 JK |
329 | } |
330 | } | |
331 | ||
c7232c99 | 332 | /* 2) Select the least-used IP/proto combination in the given range */ |
5b1158e9 | 333 | *tuple = *orig_tuple; |
5d0aa2cc | 334 | find_best_ips_proto(zone, tuple, range, ct, maniptype); |
5b1158e9 JK |
335 | |
336 | /* 3) The per-protocol part of the manip is made to map into | |
c7232c99 PM |
337 | * the range to make a unique tuple. |
338 | */ | |
5b1158e9 JK |
339 | |
340 | /* Only bother mapping if it's not already in range and unique */ | |
cbc9f2f4 PM |
341 | if (!(range->flags & NF_NAT_RANGE_PROTO_RANDOM)) { |
342 | if (range->flags & NF_NAT_RANGE_PROTO_SPECIFIED) { | |
c7232c99 PM |
343 | if (l4proto->in_range(tuple, maniptype, |
344 | &range->min_proto, | |
345 | &range->max_proto) && | |
346 | (range->min_proto.all == range->max_proto.all || | |
99ad3c53 CG |
347 | !nf_nat_used_tuple(tuple, ct))) |
348 | goto out; | |
349 | } else if (!nf_nat_used_tuple(tuple, ct)) { | |
350 | goto out; | |
351 | } | |
352 | } | |
5b1158e9 JK |
353 | |
354 | /* Last change: get protocol to try to obtain unique tuple. */ | |
c7232c99 | 355 | l4proto->unique_tuple(l3proto, tuple, range, maniptype, ct); |
e22a0548 PM |
356 | out: |
357 | rcu_read_unlock(); | |
5b1158e9 JK |
358 | } |
359 | ||
360 | unsigned int | |
361 | nf_nat_setup_info(struct nf_conn *ct, | |
c7232c99 | 362 | const struct nf_nat_range *range, |
cc01dcbd | 363 | enum nf_nat_manip_type maniptype) |
5b1158e9 | 364 | { |
0c4c9288 | 365 | struct net *net = nf_ct_net(ct); |
5b1158e9 | 366 | struct nf_conntrack_tuple curr_tuple, new_tuple; |
2d59e5ca | 367 | struct nf_conn_nat *nat; |
5b1158e9 | 368 | |
2d59e5ca YK |
369 | /* nat helper or nfctnetlink also setup binding */ |
370 | nat = nfct_nat(ct); | |
371 | if (!nat) { | |
372 | nat = nf_ct_ext_add(ct, NF_CT_EXT_NAT, GFP_ATOMIC); | |
373 | if (nat == NULL) { | |
0d53778e | 374 | pr_debug("failed to add NAT extension\n"); |
2d59e5ca YK |
375 | return NF_ACCEPT; |
376 | } | |
377 | } | |
378 | ||
cbc9f2f4 PM |
379 | NF_CT_ASSERT(maniptype == NF_NAT_MANIP_SRC || |
380 | maniptype == NF_NAT_MANIP_DST); | |
5b1158e9 JK |
381 | BUG_ON(nf_nat_initialized(ct, maniptype)); |
382 | ||
383 | /* What we've got will look like inverse of reply. Normally | |
c7232c99 PM |
384 | * this is what is in the conntrack, except for prior |
385 | * manipulations (future optimization: if num_manips == 0, | |
386 | * orig_tp = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple) | |
387 | */ | |
5b1158e9 JK |
388 | nf_ct_invert_tuplepr(&curr_tuple, |
389 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
390 | ||
391 | get_unique_tuple(&new_tuple, &curr_tuple, range, ct, maniptype); | |
392 | ||
393 | if (!nf_ct_tuple_equal(&new_tuple, &curr_tuple)) { | |
394 | struct nf_conntrack_tuple reply; | |
395 | ||
396 | /* Alter conntrack table so will recognize replies. */ | |
397 | nf_ct_invert_tuplepr(&reply, &new_tuple); | |
398 | nf_conntrack_alter_reply(ct, &reply); | |
399 | ||
400 | /* Non-atomic: we own this at the moment. */ | |
cbc9f2f4 | 401 | if (maniptype == NF_NAT_MANIP_SRC) |
5b1158e9 JK |
402 | ct->status |= IPS_SRC_NAT; |
403 | else | |
404 | ct->status |= IPS_DST_NAT; | |
405 | } | |
406 | ||
cbc9f2f4 | 407 | if (maniptype == NF_NAT_MANIP_SRC) { |
5b1158e9 JK |
408 | unsigned int srchash; |
409 | ||
5d0aa2cc PM |
410 | srchash = hash_by_src(net, nf_ct_zone(ct), |
411 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
02502f62 | 412 | spin_lock_bh(&nf_nat_lock); |
c7232c99 | 413 | /* nf_conntrack_alter_reply might re-allocate extension aera */ |
b6b84d4a YK |
414 | nat = nfct_nat(ct); |
415 | nat->ct = ct; | |
0c4c9288 | 416 | hlist_add_head_rcu(&nat->bysource, |
c7232c99 | 417 | &net->ct.nat_bysource[srchash]); |
02502f62 | 418 | spin_unlock_bh(&nf_nat_lock); |
5b1158e9 JK |
419 | } |
420 | ||
421 | /* It's done. */ | |
cbc9f2f4 | 422 | if (maniptype == NF_NAT_MANIP_DST) |
a7c2f4d7 | 423 | ct->status |= IPS_DST_NAT_DONE; |
5b1158e9 | 424 | else |
a7c2f4d7 | 425 | ct->status |= IPS_SRC_NAT_DONE; |
5b1158e9 JK |
426 | |
427 | return NF_ACCEPT; | |
428 | } | |
429 | EXPORT_SYMBOL(nf_nat_setup_info); | |
430 | ||
5b1158e9 JK |
431 | /* Do packet manipulations according to nf_nat_setup_info. */ |
432 | unsigned int nf_nat_packet(struct nf_conn *ct, | |
433 | enum ip_conntrack_info ctinfo, | |
434 | unsigned int hooknum, | |
3db05fea | 435 | struct sk_buff *skb) |
5b1158e9 | 436 | { |
c7232c99 PM |
437 | const struct nf_nat_l3proto *l3proto; |
438 | const struct nf_nat_l4proto *l4proto; | |
5b1158e9 JK |
439 | enum ip_conntrack_dir dir = CTINFO2DIR(ctinfo); |
440 | unsigned long statusbit; | |
441 | enum nf_nat_manip_type mtype = HOOK2MANIP(hooknum); | |
442 | ||
cbc9f2f4 | 443 | if (mtype == NF_NAT_MANIP_SRC) |
5b1158e9 JK |
444 | statusbit = IPS_SRC_NAT; |
445 | else | |
446 | statusbit = IPS_DST_NAT; | |
447 | ||
448 | /* Invert if this is reply dir. */ | |
449 | if (dir == IP_CT_DIR_REPLY) | |
450 | statusbit ^= IPS_NAT_MASK; | |
451 | ||
452 | /* Non-atomic: these bits don't change. */ | |
453 | if (ct->status & statusbit) { | |
454 | struct nf_conntrack_tuple target; | |
455 | ||
456 | /* We are aiming to look like inverse of other direction. */ | |
457 | nf_ct_invert_tuplepr(&target, &ct->tuplehash[!dir].tuple); | |
458 | ||
c7232c99 PM |
459 | l3proto = __nf_nat_l3proto_find(target.src.l3num); |
460 | l4proto = __nf_nat_l4proto_find(target.src.l3num, | |
461 | target.dst.protonum); | |
462 | if (!l3proto->manip_pkt(skb, 0, l4proto, &target, mtype)) | |
5b1158e9 JK |
463 | return NF_DROP; |
464 | } | |
465 | return NF_ACCEPT; | |
466 | } | |
467 | EXPORT_SYMBOL_GPL(nf_nat_packet); | |
468 | ||
c7232c99 PM |
469 | struct nf_nat_proto_clean { |
470 | u8 l3proto; | |
471 | u8 l4proto; | |
c7232c99 PM |
472 | }; |
473 | ||
c2d421e1 FW |
474 | /* kill conntracks with affected NAT section */ |
475 | static int nf_nat_proto_remove(struct nf_conn *i, void *data) | |
5b1158e9 | 476 | { |
c7232c99 PM |
477 | const struct nf_nat_proto_clean *clean = data; |
478 | struct nf_conn_nat *nat = nfct_nat(i); | |
5b1158e9 | 479 | |
c7232c99 | 480 | if (!nat) |
5b1158e9 | 481 | return 0; |
c2d421e1 | 482 | |
c7232c99 PM |
483 | if ((clean->l3proto && nf_ct_l3num(i) != clean->l3proto) || |
484 | (clean->l4proto && nf_ct_protonum(i) != clean->l4proto)) | |
5b1158e9 JK |
485 | return 0; |
486 | ||
c2d421e1 | 487 | return i->status & IPS_NAT_MASK ? 1 : 0; |
c7232c99 | 488 | } |
5b1158e9 | 489 | |
c7232c99 PM |
490 | static void nf_nat_l4proto_clean(u8 l3proto, u8 l4proto) |
491 | { | |
492 | struct nf_nat_proto_clean clean = { | |
493 | .l3proto = l3proto, | |
494 | .l4proto = l4proto, | |
495 | }; | |
496 | struct net *net; | |
497 | ||
498 | rtnl_lock(); | |
c7232c99 | 499 | for_each_net(net) |
c655bc68 | 500 | nf_ct_iterate_cleanup(net, nf_nat_proto_remove, &clean, 0, 0); |
c7232c99 PM |
501 | rtnl_unlock(); |
502 | } | |
5b1158e9 | 503 | |
c7232c99 PM |
504 | static void nf_nat_l3proto_clean(u8 l3proto) |
505 | { | |
506 | struct nf_nat_proto_clean clean = { | |
507 | .l3proto = l3proto, | |
508 | }; | |
509 | struct net *net; | |
510 | ||
511 | rtnl_lock(); | |
5b1158e9 | 512 | |
c7232c99 | 513 | for_each_net(net) |
c655bc68 | 514 | nf_ct_iterate_cleanup(net, nf_nat_proto_remove, &clean, 0, 0); |
c7232c99 | 515 | rtnl_unlock(); |
5b1158e9 | 516 | } |
5b1158e9 JK |
517 | |
518 | /* Protocol registration. */ | |
c7232c99 | 519 | int nf_nat_l4proto_register(u8 l3proto, const struct nf_nat_l4proto *l4proto) |
5b1158e9 | 520 | { |
c7232c99 PM |
521 | const struct nf_nat_l4proto **l4protos; |
522 | unsigned int i; | |
5b1158e9 JK |
523 | int ret = 0; |
524 | ||
c7232c99 PM |
525 | mutex_lock(&nf_nat_proto_mutex); |
526 | if (nf_nat_l4protos[l3proto] == NULL) { | |
527 | l4protos = kmalloc(IPPROTO_MAX * sizeof(struct nf_nat_l4proto *), | |
528 | GFP_KERNEL); | |
529 | if (l4protos == NULL) { | |
530 | ret = -ENOMEM; | |
531 | goto out; | |
532 | } | |
533 | ||
534 | for (i = 0; i < IPPROTO_MAX; i++) | |
535 | RCU_INIT_POINTER(l4protos[i], &nf_nat_l4proto_unknown); | |
536 | ||
537 | /* Before making proto_array visible to lockless readers, | |
538 | * we must make sure its content is committed to memory. | |
539 | */ | |
540 | smp_wmb(); | |
541 | ||
542 | nf_nat_l4protos[l3proto] = l4protos; | |
543 | } | |
544 | ||
eb733162 | 545 | if (rcu_dereference_protected( |
c7232c99 PM |
546 | nf_nat_l4protos[l3proto][l4proto->l4proto], |
547 | lockdep_is_held(&nf_nat_proto_mutex) | |
548 | ) != &nf_nat_l4proto_unknown) { | |
5b1158e9 JK |
549 | ret = -EBUSY; |
550 | goto out; | |
551 | } | |
c7232c99 | 552 | RCU_INIT_POINTER(nf_nat_l4protos[l3proto][l4proto->l4proto], l4proto); |
5b1158e9 | 553 | out: |
c7232c99 | 554 | mutex_unlock(&nf_nat_proto_mutex); |
5b1158e9 JK |
555 | return ret; |
556 | } | |
c7232c99 | 557 | EXPORT_SYMBOL_GPL(nf_nat_l4proto_register); |
5b1158e9 | 558 | |
25985edc | 559 | /* No one stores the protocol anywhere; simply delete it. */ |
c7232c99 | 560 | void nf_nat_l4proto_unregister(u8 l3proto, const struct nf_nat_l4proto *l4proto) |
5b1158e9 | 561 | { |
c7232c99 PM |
562 | mutex_lock(&nf_nat_proto_mutex); |
563 | RCU_INIT_POINTER(nf_nat_l4protos[l3proto][l4proto->l4proto], | |
564 | &nf_nat_l4proto_unknown); | |
565 | mutex_unlock(&nf_nat_proto_mutex); | |
e22a0548 | 566 | synchronize_rcu(); |
c7232c99 PM |
567 | |
568 | nf_nat_l4proto_clean(l3proto, l4proto->l4proto); | |
5b1158e9 | 569 | } |
c7232c99 PM |
570 | EXPORT_SYMBOL_GPL(nf_nat_l4proto_unregister); |
571 | ||
572 | int nf_nat_l3proto_register(const struct nf_nat_l3proto *l3proto) | |
573 | { | |
574 | int err; | |
575 | ||
576 | err = nf_ct_l3proto_try_module_get(l3proto->l3proto); | |
577 | if (err < 0) | |
578 | return err; | |
579 | ||
580 | mutex_lock(&nf_nat_proto_mutex); | |
581 | RCU_INIT_POINTER(nf_nat_l4protos[l3proto->l3proto][IPPROTO_TCP], | |
582 | &nf_nat_l4proto_tcp); | |
583 | RCU_INIT_POINTER(nf_nat_l4protos[l3proto->l3proto][IPPROTO_UDP], | |
584 | &nf_nat_l4proto_udp); | |
585 | mutex_unlock(&nf_nat_proto_mutex); | |
586 | ||
587 | RCU_INIT_POINTER(nf_nat_l3protos[l3proto->l3proto], l3proto); | |
588 | return 0; | |
589 | } | |
590 | EXPORT_SYMBOL_GPL(nf_nat_l3proto_register); | |
591 | ||
592 | void nf_nat_l3proto_unregister(const struct nf_nat_l3proto *l3proto) | |
593 | { | |
594 | mutex_lock(&nf_nat_proto_mutex); | |
595 | RCU_INIT_POINTER(nf_nat_l3protos[l3proto->l3proto], NULL); | |
596 | mutex_unlock(&nf_nat_proto_mutex); | |
597 | synchronize_rcu(); | |
598 | ||
599 | nf_nat_l3proto_clean(l3proto->l3proto); | |
600 | nf_ct_l3proto_module_put(l3proto->l3proto); | |
601 | } | |
602 | EXPORT_SYMBOL_GPL(nf_nat_l3proto_unregister); | |
5b1158e9 | 603 | |
25985edc | 604 | /* No one using conntrack by the time this called. */ |
d8a0509a YK |
605 | static void nf_nat_cleanup_conntrack(struct nf_conn *ct) |
606 | { | |
607 | struct nf_conn_nat *nat = nf_ct_ext_find(ct, NF_CT_EXT_NAT); | |
608 | ||
b6b84d4a | 609 | if (nat == NULL || nat->ct == NULL) |
d8a0509a YK |
610 | return; |
611 | ||
41a7cab6 | 612 | NF_CT_ASSERT(nat->ct->status & IPS_SRC_NAT_DONE); |
d8a0509a | 613 | |
02502f62 | 614 | spin_lock_bh(&nf_nat_lock); |
4d354c57 | 615 | hlist_del_rcu(&nat->bysource); |
02502f62 | 616 | spin_unlock_bh(&nf_nat_lock); |
d8a0509a YK |
617 | } |
618 | ||
86577c66 | 619 | static void nf_nat_move_storage(void *new, void *old) |
2d59e5ca | 620 | { |
86577c66 PM |
621 | struct nf_conn_nat *new_nat = new; |
622 | struct nf_conn_nat *old_nat = old; | |
b6b84d4a | 623 | struct nf_conn *ct = old_nat->ct; |
2d59e5ca | 624 | |
41a7cab6 | 625 | if (!ct || !(ct->status & IPS_SRC_NAT_DONE)) |
2d59e5ca YK |
626 | return; |
627 | ||
02502f62 | 628 | spin_lock_bh(&nf_nat_lock); |
68b80f11 | 629 | hlist_replace_rcu(&old_nat->bysource, &new_nat->bysource); |
02502f62 | 630 | spin_unlock_bh(&nf_nat_lock); |
2d59e5ca YK |
631 | } |
632 | ||
61eb3107 | 633 | static struct nf_ct_ext_type nat_extend __read_mostly = { |
d8a0509a YK |
634 | .len = sizeof(struct nf_conn_nat), |
635 | .align = __alignof__(struct nf_conn_nat), | |
636 | .destroy = nf_nat_cleanup_conntrack, | |
637 | .move = nf_nat_move_storage, | |
638 | .id = NF_CT_EXT_NAT, | |
639 | .flags = NF_CT_EXT_F_PREALLOC, | |
2d59e5ca YK |
640 | }; |
641 | ||
e6a7d3c0 PNA |
642 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
643 | ||
644 | #include <linux/netfilter/nfnetlink.h> | |
645 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
646 | ||
647 | static const struct nla_policy protonat_nla_policy[CTA_PROTONAT_MAX+1] = { | |
648 | [CTA_PROTONAT_PORT_MIN] = { .type = NLA_U16 }, | |
649 | [CTA_PROTONAT_PORT_MAX] = { .type = NLA_U16 }, | |
650 | }; | |
651 | ||
652 | static int nfnetlink_parse_nat_proto(struct nlattr *attr, | |
653 | const struct nf_conn *ct, | |
c7232c99 | 654 | struct nf_nat_range *range) |
e6a7d3c0 PNA |
655 | { |
656 | struct nlattr *tb[CTA_PROTONAT_MAX+1]; | |
c7232c99 | 657 | const struct nf_nat_l4proto *l4proto; |
e6a7d3c0 PNA |
658 | int err; |
659 | ||
660 | err = nla_parse_nested(tb, CTA_PROTONAT_MAX, attr, protonat_nla_policy); | |
661 | if (err < 0) | |
662 | return err; | |
663 | ||
c7232c99 PM |
664 | l4proto = __nf_nat_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
665 | if (l4proto->nlattr_to_range) | |
666 | err = l4proto->nlattr_to_range(tb, range); | |
667 | ||
e6a7d3c0 PNA |
668 | return err; |
669 | } | |
670 | ||
671 | static const struct nla_policy nat_nla_policy[CTA_NAT_MAX+1] = { | |
c7232c99 PM |
672 | [CTA_NAT_V4_MINIP] = { .type = NLA_U32 }, |
673 | [CTA_NAT_V4_MAXIP] = { .type = NLA_U32 }, | |
58a317f1 PM |
674 | [CTA_NAT_V6_MINIP] = { .len = sizeof(struct in6_addr) }, |
675 | [CTA_NAT_V6_MAXIP] = { .len = sizeof(struct in6_addr) }, | |
329fb58a | 676 | [CTA_NAT_PROTO] = { .type = NLA_NESTED }, |
e6a7d3c0 PNA |
677 | }; |
678 | ||
679 | static int | |
39938324 | 680 | nfnetlink_parse_nat(const struct nlattr *nat, |
c7232c99 | 681 | const struct nf_conn *ct, struct nf_nat_range *range) |
e6a7d3c0 | 682 | { |
c7232c99 | 683 | const struct nf_nat_l3proto *l3proto; |
e6a7d3c0 PNA |
684 | struct nlattr *tb[CTA_NAT_MAX+1]; |
685 | int err; | |
686 | ||
687 | memset(range, 0, sizeof(*range)); | |
688 | ||
689 | err = nla_parse_nested(tb, CTA_NAT_MAX, nat, nat_nla_policy); | |
690 | if (err < 0) | |
691 | return err; | |
692 | ||
c7232c99 PM |
693 | rcu_read_lock(); |
694 | l3proto = __nf_nat_l3proto_find(nf_ct_l3num(ct)); | |
695 | if (l3proto == NULL) { | |
696 | err = -EAGAIN; | |
697 | goto out; | |
698 | } | |
699 | err = l3proto->nlattr_to_range(tb, range); | |
700 | if (err < 0) | |
701 | goto out; | |
e6a7d3c0 PNA |
702 | |
703 | if (!tb[CTA_NAT_PROTO]) | |
c7232c99 | 704 | goto out; |
e6a7d3c0 PNA |
705 | |
706 | err = nfnetlink_parse_nat_proto(tb[CTA_NAT_PROTO], ct, range); | |
c7232c99 PM |
707 | out: |
708 | rcu_read_unlock(); | |
709 | return err; | |
e6a7d3c0 PNA |
710 | } |
711 | ||
712 | static int | |
713 | nfnetlink_parse_nat_setup(struct nf_conn *ct, | |
714 | enum nf_nat_manip_type manip, | |
39938324 | 715 | const struct nlattr *attr) |
e6a7d3c0 | 716 | { |
c7232c99 PM |
717 | struct nf_nat_range range; |
718 | int err; | |
e6a7d3c0 | 719 | |
c7232c99 PM |
720 | err = nfnetlink_parse_nat(attr, ct, &range); |
721 | if (err < 0) | |
722 | return err; | |
e6a7d3c0 PNA |
723 | if (nf_nat_initialized(ct, manip)) |
724 | return -EEXIST; | |
725 | ||
726 | return nf_nat_setup_info(ct, &range, manip); | |
727 | } | |
728 | #else | |
729 | static int | |
730 | nfnetlink_parse_nat_setup(struct nf_conn *ct, | |
731 | enum nf_nat_manip_type manip, | |
39938324 | 732 | const struct nlattr *attr) |
e6a7d3c0 PNA |
733 | { |
734 | return -EOPNOTSUPP; | |
735 | } | |
736 | #endif | |
737 | ||
0c4c9288 AD |
738 | static int __net_init nf_nat_net_init(struct net *net) |
739 | { | |
d696c7bd | 740 | /* Leave them the same for the moment. */ |
c7232c99 PM |
741 | net->ct.nat_htable_size = net->ct.htable_size; |
742 | net->ct.nat_bysource = nf_ct_alloc_hashtable(&net->ct.nat_htable_size, 0); | |
743 | if (!net->ct.nat_bysource) | |
0c4c9288 AD |
744 | return -ENOMEM; |
745 | return 0; | |
746 | } | |
747 | ||
0c4c9288 AD |
748 | static void __net_exit nf_nat_net_exit(struct net *net) |
749 | { | |
c7232c99 PM |
750 | struct nf_nat_proto_clean clean = {}; |
751 | ||
c655bc68 | 752 | nf_ct_iterate_cleanup(net, &nf_nat_proto_remove, &clean, 0, 0); |
0c4c9288 | 753 | synchronize_rcu(); |
c7232c99 | 754 | nf_ct_free_hashtable(net->ct.nat_bysource, net->ct.nat_htable_size); |
0c4c9288 AD |
755 | } |
756 | ||
757 | static struct pernet_operations nf_nat_net_ops = { | |
758 | .init = nf_nat_net_init, | |
759 | .exit = nf_nat_net_exit, | |
760 | }; | |
761 | ||
544d5c7d PNA |
762 | static struct nf_ct_helper_expectfn follow_master_nat = { |
763 | .name = "nat-follow-master", | |
764 | .expectfn = nf_nat_follow_master, | |
765 | }; | |
766 | ||
d584a61a PNA |
767 | static struct nfq_ct_nat_hook nfq_ct_nat = { |
768 | .seq_adjust = nf_nat_tcp_seq_adjust, | |
769 | }; | |
770 | ||
5b1158e9 JK |
771 | static int __init nf_nat_init(void) |
772 | { | |
2d59e5ca YK |
773 | int ret; |
774 | ||
775 | ret = nf_ct_extend_register(&nat_extend); | |
776 | if (ret < 0) { | |
777 | printk(KERN_ERR "nf_nat_core: Unable to register extension\n"); | |
778 | return ret; | |
779 | } | |
5b1158e9 | 780 | |
0c4c9288 AD |
781 | ret = register_pernet_subsys(&nf_nat_net_ops); |
782 | if (ret < 0) | |
2d59e5ca | 783 | goto cleanup_extend; |
5b1158e9 | 784 | |
c7232c99 | 785 | nf_ct_helper_expectfn_register(&follow_master_nat); |
5b1158e9 | 786 | |
5b1158e9 | 787 | /* Initialize fake conntrack so that NAT will skip it */ |
5bfddbd4 | 788 | nf_ct_untracked_status_or(IPS_NAT_DONE_MASK); |
5b1158e9 | 789 | |
dd13b010 | 790 | BUG_ON(nf_nat_seq_adjust_hook != NULL); |
a9b3cd7f | 791 | RCU_INIT_POINTER(nf_nat_seq_adjust_hook, nf_nat_seq_adjust); |
e6a7d3c0 | 792 | BUG_ON(nfnetlink_parse_nat_setup_hook != NULL); |
a9b3cd7f | 793 | RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook, |
e6a7d3c0 | 794 | nfnetlink_parse_nat_setup); |
f9dd09c7 | 795 | BUG_ON(nf_ct_nat_offset != NULL); |
a9b3cd7f | 796 | RCU_INIT_POINTER(nf_ct_nat_offset, nf_nat_get_offset); |
d584a61a | 797 | RCU_INIT_POINTER(nfq_ct_nat_hook, &nfq_ct_nat); |
c7232c99 PM |
798 | #ifdef CONFIG_XFRM |
799 | BUG_ON(nf_nat_decode_session_hook != NULL); | |
800 | RCU_INIT_POINTER(nf_nat_decode_session_hook, __nf_nat_decode_session); | |
801 | #endif | |
5b1158e9 | 802 | return 0; |
2d59e5ca YK |
803 | |
804 | cleanup_extend: | |
805 | nf_ct_extend_unregister(&nat_extend); | |
806 | return ret; | |
5b1158e9 JK |
807 | } |
808 | ||
5b1158e9 JK |
809 | static void __exit nf_nat_cleanup(void) |
810 | { | |
c7232c99 PM |
811 | unsigned int i; |
812 | ||
0c4c9288 | 813 | unregister_pernet_subsys(&nf_nat_net_ops); |
2d59e5ca | 814 | nf_ct_extend_unregister(&nat_extend); |
544d5c7d | 815 | nf_ct_helper_expectfn_unregister(&follow_master_nat); |
a9b3cd7f SH |
816 | RCU_INIT_POINTER(nf_nat_seq_adjust_hook, NULL); |
817 | RCU_INIT_POINTER(nfnetlink_parse_nat_setup_hook, NULL); | |
818 | RCU_INIT_POINTER(nf_ct_nat_offset, NULL); | |
d584a61a | 819 | RCU_INIT_POINTER(nfq_ct_nat_hook, NULL); |
c7232c99 PM |
820 | #ifdef CONFIG_XFRM |
821 | RCU_INIT_POINTER(nf_nat_decode_session_hook, NULL); | |
822 | #endif | |
823 | for (i = 0; i < NFPROTO_NUMPROTO; i++) | |
824 | kfree(nf_nat_l4protos[i]); | |
dd13b010 | 825 | synchronize_net(); |
5b1158e9 JK |
826 | } |
827 | ||
828 | MODULE_LICENSE("GPL"); | |
829 | ||
830 | module_init(nf_nat_init); | |
831 | module_exit(nf_nat_cleanup); |