]>
Commit | Line | Data |
---|---|---|
9fb9cbb1 YK |
1 | /* Connection state tracking for netfilter. This is separated from, |
2 | but required by, the NAT layer; it can also be used by an iptables | |
3 | extension. */ | |
4 | ||
5 | /* (C) 1999-2001 Paul `Rusty' Russell | |
dc808fe2 | 6 | * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org> |
9fb9cbb1 YK |
7 | * (C) 2003,2004 USAGI/WIDE Project <http://www.linux-ipv6.org> |
8 | * | |
9 | * This program is free software; you can redistribute it and/or modify | |
10 | * it under the terms of the GNU General Public License version 2 as | |
11 | * published by the Free Software Foundation. | |
9fb9cbb1 YK |
12 | */ |
13 | ||
9fb9cbb1 YK |
14 | #include <linux/types.h> |
15 | #include <linux/netfilter.h> | |
16 | #include <linux/module.h> | |
17 | #include <linux/skbuff.h> | |
18 | #include <linux/proc_fs.h> | |
19 | #include <linux/vmalloc.h> | |
20 | #include <linux/stddef.h> | |
21 | #include <linux/slab.h> | |
22 | #include <linux/random.h> | |
23 | #include <linux/jhash.h> | |
24 | #include <linux/err.h> | |
25 | #include <linux/percpu.h> | |
26 | #include <linux/moduleparam.h> | |
27 | #include <linux/notifier.h> | |
28 | #include <linux/kernel.h> | |
29 | #include <linux/netdevice.h> | |
30 | #include <linux/socket.h> | |
d7fe0f24 | 31 | #include <linux/mm.h> |
ea781f19 | 32 | #include <linux/rculist_nulls.h> |
9fb9cbb1 | 33 | |
9fb9cbb1 YK |
34 | #include <net/netfilter/nf_conntrack.h> |
35 | #include <net/netfilter/nf_conntrack_l3proto.h> | |
605dcad6 | 36 | #include <net/netfilter/nf_conntrack_l4proto.h> |
77ab9cff | 37 | #include <net/netfilter/nf_conntrack_expect.h> |
9fb9cbb1 YK |
38 | #include <net/netfilter/nf_conntrack_helper.h> |
39 | #include <net/netfilter/nf_conntrack_core.h> | |
ecfab2c9 | 40 | #include <net/netfilter/nf_conntrack_extend.h> |
58401572 | 41 | #include <net/netfilter/nf_conntrack_acct.h> |
e6a7d3c0 | 42 | #include <net/netfilter/nf_nat.h> |
e17b666a | 43 | #include <net/netfilter/nf_nat_core.h> |
9fb9cbb1 | 44 | |
dc808fe2 | 45 | #define NF_CONNTRACK_VERSION "0.5.0" |
9fb9cbb1 | 46 | |
e17b666a PM |
47 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
48 | enum nf_nat_manip_type manip, | |
49 | struct nlattr *attr) __read_mostly; | |
e6a7d3c0 PNA |
50 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
51 | ||
f8ba1aff | 52 | DEFINE_SPINLOCK(nf_conntrack_lock); |
13b18339 | 53 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); |
9fb9cbb1 | 54 | |
e2b7606c | 55 | unsigned int nf_conntrack_htable_size __read_mostly; |
13b18339 PM |
56 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
57 | ||
e478075c | 58 | unsigned int nf_conntrack_max __read_mostly; |
a999e683 | 59 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
13b18339 | 60 | |
e2b7606c | 61 | struct nf_conn nf_conntrack_untracked __read_mostly; |
13b18339 PM |
62 | EXPORT_SYMBOL_GPL(nf_conntrack_untracked); |
63 | ||
dacd2a1a | 64 | static struct kmem_cache *nf_conntrack_cachep __read_mostly; |
77ab9cff | 65 | |
9fb9cbb1 YK |
66 | static int nf_conntrack_hash_rnd_initted; |
67 | static unsigned int nf_conntrack_hash_rnd; | |
68 | ||
69 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, | |
70 | unsigned int size, unsigned int rnd) | |
71 | { | |
0794935e PM |
72 | unsigned int n; |
73 | u_int32_t h; | |
74 | ||
75 | /* The direction must be ignored, so we hash everything up to the | |
76 | * destination ports (which is a multiple of 4) and treat the last | |
77 | * three bytes manually. | |
78 | */ | |
79 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); | |
80 | h = jhash2((u32 *)tuple, n, | |
81 | rnd ^ (((__force __u16)tuple->dst.u.all << 16) | | |
82 | tuple->dst.protonum)); | |
83 | ||
84 | return ((u64)h * size) >> 32; | |
9fb9cbb1 YK |
85 | } |
86 | ||
87 | static inline u_int32_t hash_conntrack(const struct nf_conntrack_tuple *tuple) | |
88 | { | |
89 | return __hash_conntrack(tuple, nf_conntrack_htable_size, | |
90 | nf_conntrack_hash_rnd); | |
91 | } | |
92 | ||
5f2b4c90 | 93 | bool |
9fb9cbb1 YK |
94 | nf_ct_get_tuple(const struct sk_buff *skb, |
95 | unsigned int nhoff, | |
96 | unsigned int dataoff, | |
97 | u_int16_t l3num, | |
98 | u_int8_t protonum, | |
99 | struct nf_conntrack_tuple *tuple, | |
100 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 101 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 102 | { |
443a70d5 | 103 | memset(tuple, 0, sizeof(*tuple)); |
9fb9cbb1 YK |
104 | |
105 | tuple->src.l3num = l3num; | |
106 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | |
5f2b4c90 | 107 | return false; |
9fb9cbb1 YK |
108 | |
109 | tuple->dst.protonum = protonum; | |
110 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | |
111 | ||
605dcad6 | 112 | return l4proto->pkt_to_tuple(skb, dataoff, tuple); |
9fb9cbb1 | 113 | } |
13b18339 | 114 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
9fb9cbb1 | 115 | |
5f2b4c90 JE |
116 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
117 | u_int16_t l3num, struct nf_conntrack_tuple *tuple) | |
e2a3123f YK |
118 | { |
119 | struct nf_conntrack_l3proto *l3proto; | |
120 | struct nf_conntrack_l4proto *l4proto; | |
121 | unsigned int protoff; | |
122 | u_int8_t protonum; | |
123 | int ret; | |
124 | ||
125 | rcu_read_lock(); | |
126 | ||
127 | l3proto = __nf_ct_l3proto_find(l3num); | |
128 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | |
129 | if (ret != NF_ACCEPT) { | |
130 | rcu_read_unlock(); | |
5f2b4c90 | 131 | return false; |
e2a3123f YK |
132 | } |
133 | ||
134 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
135 | ||
136 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, tuple, | |
137 | l3proto, l4proto); | |
138 | ||
139 | rcu_read_unlock(); | |
140 | return ret; | |
141 | } | |
142 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | |
143 | ||
5f2b4c90 | 144 | bool |
9fb9cbb1 YK |
145 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
146 | const struct nf_conntrack_tuple *orig, | |
147 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 148 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 149 | { |
443a70d5 | 150 | memset(inverse, 0, sizeof(*inverse)); |
9fb9cbb1 YK |
151 | |
152 | inverse->src.l3num = orig->src.l3num; | |
153 | if (l3proto->invert_tuple(inverse, orig) == 0) | |
5f2b4c90 | 154 | return false; |
9fb9cbb1 YK |
155 | |
156 | inverse->dst.dir = !orig->dst.dir; | |
157 | ||
158 | inverse->dst.protonum = orig->dst.protonum; | |
605dcad6 | 159 | return l4proto->invert_tuple(inverse, orig); |
9fb9cbb1 | 160 | } |
13b18339 | 161 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
9fb9cbb1 | 162 | |
9fb9cbb1 YK |
163 | static void |
164 | clean_from_lists(struct nf_conn *ct) | |
165 | { | |
0d53778e | 166 | pr_debug("clean_from_lists(%p)\n", ct); |
ea781f19 ED |
167 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
168 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | |
9fb9cbb1 YK |
169 | |
170 | /* Destroy all pending expectations */ | |
c1d10adb | 171 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
172 | } |
173 | ||
174 | static void | |
175 | destroy_conntrack(struct nf_conntrack *nfct) | |
176 | { | |
177 | struct nf_conn *ct = (struct nf_conn *)nfct; | |
0d55af87 | 178 | struct net *net = nf_ct_net(ct); |
605dcad6 | 179 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 | 180 | |
0d53778e | 181 | pr_debug("destroy_conntrack(%p)\n", ct); |
9fb9cbb1 YK |
182 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
183 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); | |
184 | ||
19abb7b0 PNA |
185 | if (!test_bit(IPS_DYING_BIT, &ct->status)) |
186 | nf_conntrack_event(IPCT_DESTROY, ct); | |
9fb9cbb1 YK |
187 | set_bit(IPS_DYING_BIT, &ct->status); |
188 | ||
189 | /* To make sure we don't get any weird locking issues here: | |
190 | * destroy_conntrack() MUST NOT be called with a write lock | |
191 | * to nf_conntrack_lock!!! -HW */ | |
923f4902 | 192 | rcu_read_lock(); |
5e8fbe2a | 193 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
605dcad6 MJ |
194 | if (l4proto && l4proto->destroy) |
195 | l4proto->destroy(ct); | |
9fb9cbb1 | 196 | |
982d9a9c | 197 | rcu_read_unlock(); |
9fb9cbb1 | 198 | |
f8ba1aff | 199 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
200 | /* Expectations will have been removed in clean_from_lists, |
201 | * except TFTP can create an expectation on the first packet, | |
202 | * before connection is in the list, so we need to clean here, | |
203 | * too. */ | |
c1d10adb | 204 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
205 | |
206 | /* We overload first tuple to link into unconfirmed list. */ | |
207 | if (!nf_ct_is_confirmed(ct)) { | |
ea781f19 ED |
208 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); |
209 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | |
9fb9cbb1 YK |
210 | } |
211 | ||
0d55af87 | 212 | NF_CT_STAT_INC(net, delete); |
f8ba1aff | 213 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
214 | |
215 | if (ct->master) | |
216 | nf_ct_put(ct->master); | |
217 | ||
0d53778e | 218 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
9fb9cbb1 YK |
219 | nf_conntrack_free(ct); |
220 | } | |
221 | ||
222 | static void death_by_timeout(unsigned long ul_conntrack) | |
223 | { | |
224 | struct nf_conn *ct = (void *)ul_conntrack; | |
0d55af87 | 225 | struct net *net = nf_ct_net(ct); |
5397e97d | 226 | struct nf_conn_help *help = nfct_help(ct); |
3c158f7f | 227 | struct nf_conntrack_helper *helper; |
5397e97d | 228 | |
3c158f7f PM |
229 | if (help) { |
230 | rcu_read_lock(); | |
231 | helper = rcu_dereference(help->helper); | |
232 | if (helper && helper->destroy) | |
233 | helper->destroy(ct); | |
234 | rcu_read_unlock(); | |
235 | } | |
9fb9cbb1 | 236 | |
f8ba1aff | 237 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
238 | /* Inside lock so preempt is disabled on module removal path. |
239 | * Otherwise we can get spurious warnings. */ | |
0d55af87 | 240 | NF_CT_STAT_INC(net, delete_list); |
9fb9cbb1 | 241 | clean_from_lists(ct); |
f8ba1aff | 242 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
243 | nf_ct_put(ct); |
244 | } | |
245 | ||
ea781f19 ED |
246 | /* |
247 | * Warning : | |
248 | * - Caller must take a reference on returned object | |
249 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | |
250 | * OR | |
251 | * - Caller must lock nf_conntrack_lock before calling this function | |
252 | */ | |
c1d10adb | 253 | struct nf_conntrack_tuple_hash * |
400dad39 | 254 | __nf_conntrack_find(struct net *net, const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 YK |
255 | { |
256 | struct nf_conntrack_tuple_hash *h; | |
ea781f19 | 257 | struct hlist_nulls_node *n; |
9fb9cbb1 YK |
258 | unsigned int hash = hash_conntrack(tuple); |
259 | ||
4e29e9ec PM |
260 | /* Disable BHs the entire time since we normally need to disable them |
261 | * at least once for the stats anyway. | |
262 | */ | |
263 | local_bh_disable(); | |
ea781f19 ED |
264 | begin: |
265 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { | |
ba419aff | 266 | if (nf_ct_tuple_equal(tuple, &h->tuple)) { |
0d55af87 | 267 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 268 | local_bh_enable(); |
9fb9cbb1 YK |
269 | return h; |
270 | } | |
0d55af87 | 271 | NF_CT_STAT_INC(net, searched); |
9fb9cbb1 | 272 | } |
ea781f19 ED |
273 | /* |
274 | * if the nulls value we got at the end of this lookup is | |
275 | * not the expected one, we must restart lookup. | |
276 | * We probably met an item that was moved to another chain. | |
277 | */ | |
278 | if (get_nulls_value(n) != hash) | |
279 | goto begin; | |
4e29e9ec | 280 | local_bh_enable(); |
9fb9cbb1 YK |
281 | |
282 | return NULL; | |
283 | } | |
13b18339 | 284 | EXPORT_SYMBOL_GPL(__nf_conntrack_find); |
9fb9cbb1 YK |
285 | |
286 | /* Find a connection corresponding to a tuple. */ | |
287 | struct nf_conntrack_tuple_hash * | |
400dad39 | 288 | nf_conntrack_find_get(struct net *net, const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 YK |
289 | { |
290 | struct nf_conntrack_tuple_hash *h; | |
76507f69 | 291 | struct nf_conn *ct; |
9fb9cbb1 | 292 | |
76507f69 | 293 | rcu_read_lock(); |
ea781f19 | 294 | begin: |
400dad39 | 295 | h = __nf_conntrack_find(net, tuple); |
76507f69 PM |
296 | if (h) { |
297 | ct = nf_ct_tuplehash_to_ctrack(h); | |
298 | if (unlikely(!atomic_inc_not_zero(&ct->ct_general.use))) | |
299 | h = NULL; | |
ea781f19 ED |
300 | else { |
301 | if (unlikely(!nf_ct_tuple_equal(tuple, &h->tuple))) { | |
302 | nf_ct_put(ct); | |
303 | goto begin; | |
304 | } | |
305 | } | |
76507f69 PM |
306 | } |
307 | rcu_read_unlock(); | |
9fb9cbb1 YK |
308 | |
309 | return h; | |
310 | } | |
13b18339 | 311 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
9fb9cbb1 | 312 | |
c1d10adb PNA |
313 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
314 | unsigned int hash, | |
601e68e1 | 315 | unsigned int repl_hash) |
c1d10adb | 316 | { |
400dad39 AD |
317 | struct net *net = nf_ct_net(ct); |
318 | ||
ea781f19 | 319 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
400dad39 | 320 | &net->ct.hash[hash]); |
ea781f19 | 321 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
400dad39 | 322 | &net->ct.hash[repl_hash]); |
c1d10adb PNA |
323 | } |
324 | ||
325 | void nf_conntrack_hash_insert(struct nf_conn *ct) | |
326 | { | |
327 | unsigned int hash, repl_hash; | |
328 | ||
329 | hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
330 | repl_hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
331 | ||
c1d10adb | 332 | __nf_conntrack_hash_insert(ct, hash, repl_hash); |
c1d10adb | 333 | } |
13b18339 | 334 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_insert); |
c1d10adb | 335 | |
9fb9cbb1 YK |
336 | /* Confirm a connection given skb; places it in hash table */ |
337 | int | |
3db05fea | 338 | __nf_conntrack_confirm(struct sk_buff *skb) |
9fb9cbb1 YK |
339 | { |
340 | unsigned int hash, repl_hash; | |
df0933dc | 341 | struct nf_conntrack_tuple_hash *h; |
9fb9cbb1 | 342 | struct nf_conn *ct; |
df0933dc | 343 | struct nf_conn_help *help; |
ea781f19 | 344 | struct hlist_nulls_node *n; |
9fb9cbb1 | 345 | enum ip_conntrack_info ctinfo; |
400dad39 | 346 | struct net *net; |
9fb9cbb1 | 347 | |
3db05fea | 348 | ct = nf_ct_get(skb, &ctinfo); |
400dad39 | 349 | net = nf_ct_net(ct); |
9fb9cbb1 YK |
350 | |
351 | /* ipt_REJECT uses nf_conntrack_attach to attach related | |
352 | ICMP/TCP RST packets in other direction. Actual packet | |
353 | which created connection will be IP_CT_NEW or for an | |
354 | expected connection, IP_CT_RELATED. */ | |
355 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | |
356 | return NF_ACCEPT; | |
357 | ||
358 | hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
359 | repl_hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
360 | ||
361 | /* We're not in hash table, and we refuse to set up related | |
362 | connections for unconfirmed conns. But packet copies and | |
363 | REJECT will give spurious warnings here. */ | |
364 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ | |
365 | ||
366 | /* No external references means noone else could have | |
367 | confirmed us. */ | |
368 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
0d53778e | 369 | pr_debug("Confirming conntrack %p\n", ct); |
9fb9cbb1 | 370 | |
f8ba1aff | 371 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
372 | |
373 | /* See if there's one in the list already, including reverse: | |
374 | NAT could have grabbed it without realizing, since we're | |
375 | not in the hash. If there is, we lost race. */ | |
ea781f19 | 376 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
df0933dc PM |
377 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
378 | &h->tuple)) | |
379 | goto out; | |
ea781f19 | 380 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) |
df0933dc PM |
381 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
382 | &h->tuple)) | |
383 | goto out; | |
9fb9cbb1 | 384 | |
df0933dc | 385 | /* Remove from unconfirmed list */ |
ea781f19 | 386 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
df0933dc PM |
387 | |
388 | __nf_conntrack_hash_insert(ct, hash, repl_hash); | |
389 | /* Timer relative to confirmation time, not original | |
390 | setting time, otherwise we'd get timer wrap in | |
391 | weird delay cases. */ | |
392 | ct->timeout.expires += jiffies; | |
393 | add_timer(&ct->timeout); | |
394 | atomic_inc(&ct->ct_general.use); | |
395 | set_bit(IPS_CONFIRMED_BIT, &ct->status); | |
0d55af87 | 396 | NF_CT_STAT_INC(net, insert); |
f8ba1aff | 397 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc PM |
398 | help = nfct_help(ct); |
399 | if (help && help->helper) | |
a71996fc | 400 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
17e6e4ea | 401 | |
df0933dc | 402 | nf_conntrack_event_cache(master_ct(ct) ? |
a71996fc | 403 | IPCT_RELATED : IPCT_NEW, ct); |
df0933dc | 404 | return NF_ACCEPT; |
9fb9cbb1 | 405 | |
df0933dc | 406 | out: |
0d55af87 | 407 | NF_CT_STAT_INC(net, insert_failed); |
f8ba1aff | 408 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
409 | return NF_DROP; |
410 | } | |
13b18339 | 411 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
9fb9cbb1 YK |
412 | |
413 | /* Returns true if a connection correspondings to the tuple (required | |
414 | for NAT). */ | |
415 | int | |
416 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | |
417 | const struct nf_conn *ignored_conntrack) | |
418 | { | |
400dad39 | 419 | struct net *net = nf_ct_net(ignored_conntrack); |
9fb9cbb1 | 420 | struct nf_conntrack_tuple_hash *h; |
ea781f19 | 421 | struct hlist_nulls_node *n; |
ba419aff | 422 | unsigned int hash = hash_conntrack(tuple); |
9fb9cbb1 | 423 | |
4e29e9ec PM |
424 | /* Disable BHs the entire time since we need to disable them at |
425 | * least once for the stats anyway. | |
426 | */ | |
427 | rcu_read_lock_bh(); | |
ea781f19 | 428 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
ba419aff PM |
429 | if (nf_ct_tuplehash_to_ctrack(h) != ignored_conntrack && |
430 | nf_ct_tuple_equal(tuple, &h->tuple)) { | |
0d55af87 | 431 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 432 | rcu_read_unlock_bh(); |
ba419aff PM |
433 | return 1; |
434 | } | |
0d55af87 | 435 | NF_CT_STAT_INC(net, searched); |
ba419aff | 436 | } |
4e29e9ec | 437 | rcu_read_unlock_bh(); |
9fb9cbb1 | 438 | |
ba419aff | 439 | return 0; |
9fb9cbb1 | 440 | } |
13b18339 | 441 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
9fb9cbb1 | 442 | |
7ae7730f PM |
443 | #define NF_CT_EVICTION_RANGE 8 |
444 | ||
9fb9cbb1 YK |
445 | /* There's a small race here where we may free a just-assured |
446 | connection. Too bad: we're in trouble anyway. */ | |
400dad39 | 447 | static noinline int early_drop(struct net *net, unsigned int hash) |
9fb9cbb1 | 448 | { |
f205c5e0 | 449 | /* Use oldest entry, which is roughly LRU */ |
9fb9cbb1 | 450 | struct nf_conntrack_tuple_hash *h; |
df0933dc | 451 | struct nf_conn *ct = NULL, *tmp; |
ea781f19 | 452 | struct hlist_nulls_node *n; |
7ae7730f | 453 | unsigned int i, cnt = 0; |
9fb9cbb1 YK |
454 | int dropped = 0; |
455 | ||
76507f69 | 456 | rcu_read_lock(); |
7ae7730f | 457 | for (i = 0; i < nf_conntrack_htable_size; i++) { |
ea781f19 ED |
458 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
459 | hnnode) { | |
7ae7730f PM |
460 | tmp = nf_ct_tuplehash_to_ctrack(h); |
461 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) | |
462 | ct = tmp; | |
463 | cnt++; | |
464 | } | |
76507f69 PM |
465 | |
466 | if (ct && unlikely(!atomic_inc_not_zero(&ct->ct_general.use))) | |
467 | ct = NULL; | |
7ae7730f PM |
468 | if (ct || cnt >= NF_CT_EVICTION_RANGE) |
469 | break; | |
470 | hash = (hash + 1) % nf_conntrack_htable_size; | |
9fb9cbb1 | 471 | } |
76507f69 | 472 | rcu_read_unlock(); |
9fb9cbb1 YK |
473 | |
474 | if (!ct) | |
475 | return dropped; | |
476 | ||
477 | if (del_timer(&ct->timeout)) { | |
478 | death_by_timeout((unsigned long)ct); | |
479 | dropped = 1; | |
0d55af87 | 480 | NF_CT_STAT_INC_ATOMIC(net, early_drop); |
9fb9cbb1 YK |
481 | } |
482 | nf_ct_put(ct); | |
483 | return dropped; | |
484 | } | |
485 | ||
5a1fb391 AD |
486 | struct nf_conn *nf_conntrack_alloc(struct net *net, |
487 | const struct nf_conntrack_tuple *orig, | |
b891c5a8 PNA |
488 | const struct nf_conntrack_tuple *repl, |
489 | gfp_t gfp) | |
9fb9cbb1 | 490 | { |
cd7fcbf1 | 491 | struct nf_conn *ct; |
9fb9cbb1 | 492 | |
dc808fe2 | 493 | if (unlikely(!nf_conntrack_hash_rnd_initted)) { |
af07d241 HPP |
494 | get_random_bytes(&nf_conntrack_hash_rnd, |
495 | sizeof(nf_conntrack_hash_rnd)); | |
9fb9cbb1 YK |
496 | nf_conntrack_hash_rnd_initted = 1; |
497 | } | |
498 | ||
5251e2d2 | 499 | /* We don't want any race condition at early drop stage */ |
49ac8713 | 500 | atomic_inc(&net->ct.count); |
5251e2d2 | 501 | |
76eb9460 | 502 | if (nf_conntrack_max && |
49ac8713 | 503 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
9fb9cbb1 | 504 | unsigned int hash = hash_conntrack(orig); |
400dad39 | 505 | if (!early_drop(net, hash)) { |
49ac8713 | 506 | atomic_dec(&net->ct.count); |
9fb9cbb1 YK |
507 | if (net_ratelimit()) |
508 | printk(KERN_WARNING | |
509 | "nf_conntrack: table full, dropping" | |
510 | " packet.\n"); | |
511 | return ERR_PTR(-ENOMEM); | |
512 | } | |
513 | } | |
514 | ||
b891c5a8 | 515 | ct = kmem_cache_zalloc(nf_conntrack_cachep, gfp); |
c88130bc | 516 | if (ct == NULL) { |
0d53778e | 517 | pr_debug("nf_conntrack_alloc: Can't alloc conntrack.\n"); |
49ac8713 | 518 | atomic_dec(&net->ct.count); |
dacd2a1a | 519 | return ERR_PTR(-ENOMEM); |
9fb9cbb1 YK |
520 | } |
521 | ||
440f0d58 | 522 | spin_lock_init(&ct->lock); |
c88130bc PM |
523 | atomic_set(&ct->ct_general.use, 1); |
524 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; | |
525 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; | |
9fb9cbb1 | 526 | /* Don't set timer yet: wait for confirmation */ |
c88130bc | 527 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
5a1fb391 AD |
528 | #ifdef CONFIG_NET_NS |
529 | ct->ct_net = net; | |
530 | #endif | |
9fb9cbb1 | 531 | |
c88130bc | 532 | return ct; |
9fb9cbb1 | 533 | } |
13b18339 | 534 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
9fb9cbb1 | 535 | |
c88130bc | 536 | void nf_conntrack_free(struct nf_conn *ct) |
76507f69 | 537 | { |
1d45209d ED |
538 | struct net *net = nf_ct_net(ct); |
539 | ||
ceeff754 | 540 | nf_ct_ext_destroy(ct); |
1d45209d | 541 | atomic_dec(&net->ct.count); |
ea781f19 ED |
542 | nf_ct_ext_free(ct); |
543 | kmem_cache_free(nf_conntrack_cachep, ct); | |
76507f69 | 544 | } |
13b18339 | 545 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
9fb9cbb1 YK |
546 | |
547 | /* Allocate a new conntrack: we return -ENOMEM if classification | |
548 | failed due to stress. Otherwise it really is unclassifiable. */ | |
549 | static struct nf_conntrack_tuple_hash * | |
5a1fb391 AD |
550 | init_conntrack(struct net *net, |
551 | const struct nf_conntrack_tuple *tuple, | |
9fb9cbb1 | 552 | struct nf_conntrack_l3proto *l3proto, |
605dcad6 | 553 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 YK |
554 | struct sk_buff *skb, |
555 | unsigned int dataoff) | |
556 | { | |
c88130bc | 557 | struct nf_conn *ct; |
3c158f7f | 558 | struct nf_conn_help *help; |
9fb9cbb1 YK |
559 | struct nf_conntrack_tuple repl_tuple; |
560 | struct nf_conntrack_expect *exp; | |
561 | ||
605dcad6 | 562 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
0d53778e | 563 | pr_debug("Can't invert tuple.\n"); |
9fb9cbb1 YK |
564 | return NULL; |
565 | } | |
566 | ||
5a1fb391 | 567 | ct = nf_conntrack_alloc(net, tuple, &repl_tuple, GFP_ATOMIC); |
cd7fcbf1 | 568 | if (IS_ERR(ct)) { |
0d53778e | 569 | pr_debug("Can't allocate conntrack.\n"); |
c88130bc | 570 | return (struct nf_conntrack_tuple_hash *)ct; |
9fb9cbb1 YK |
571 | } |
572 | ||
c88130bc PM |
573 | if (!l4proto->new(ct, skb, dataoff)) { |
574 | nf_conntrack_free(ct); | |
0d53778e | 575 | pr_debug("init conntrack: can't track with proto module\n"); |
9fb9cbb1 YK |
576 | return NULL; |
577 | } | |
578 | ||
58401572 KPO |
579 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
580 | ||
f8ba1aff | 581 | spin_lock_bh(&nf_conntrack_lock); |
9b03f38d | 582 | exp = nf_ct_find_expectation(net, tuple); |
9fb9cbb1 | 583 | if (exp) { |
0d53778e | 584 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", |
c88130bc | 585 | ct, exp); |
9fb9cbb1 | 586 | /* Welcome, Mr. Bond. We've been expecting you... */ |
c88130bc PM |
587 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
588 | ct->master = exp->master; | |
ceceae1b | 589 | if (exp->helper) { |
c88130bc | 590 | help = nf_ct_helper_ext_add(ct, GFP_ATOMIC); |
ceceae1b YK |
591 | if (help) |
592 | rcu_assign_pointer(help->helper, exp->helper); | |
ceceae1b YK |
593 | } |
594 | ||
9fb9cbb1 | 595 | #ifdef CONFIG_NF_CONNTRACK_MARK |
c88130bc | 596 | ct->mark = exp->master->mark; |
7c9728c3 JM |
597 | #endif |
598 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | |
c88130bc | 599 | ct->secmark = exp->master->secmark; |
9fb9cbb1 | 600 | #endif |
c88130bc | 601 | nf_conntrack_get(&ct->master->ct_general); |
0d55af87 | 602 | NF_CT_STAT_INC(net, expect_new); |
22e7410b | 603 | } else { |
226c0c0e | 604 | __nf_ct_try_assign_helper(ct, GFP_ATOMIC); |
0d55af87 | 605 | NF_CT_STAT_INC(net, new); |
22e7410b | 606 | } |
9fb9cbb1 YK |
607 | |
608 | /* Overload tuple linked list to put us in unconfirmed list. */ | |
ea781f19 | 609 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
63c9a262 | 610 | &net->ct.unconfirmed); |
9fb9cbb1 | 611 | |
f8ba1aff | 612 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
613 | |
614 | if (exp) { | |
615 | if (exp->expectfn) | |
c88130bc | 616 | exp->expectfn(ct, exp); |
6823645d | 617 | nf_ct_expect_put(exp); |
9fb9cbb1 YK |
618 | } |
619 | ||
c88130bc | 620 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
9fb9cbb1 YK |
621 | } |
622 | ||
623 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | |
624 | static inline struct nf_conn * | |
a702a65f AD |
625 | resolve_normal_ct(struct net *net, |
626 | struct sk_buff *skb, | |
9fb9cbb1 YK |
627 | unsigned int dataoff, |
628 | u_int16_t l3num, | |
629 | u_int8_t protonum, | |
630 | struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 631 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 YK |
632 | int *set_reply, |
633 | enum ip_conntrack_info *ctinfo) | |
634 | { | |
635 | struct nf_conntrack_tuple tuple; | |
636 | struct nf_conntrack_tuple_hash *h; | |
637 | struct nf_conn *ct; | |
638 | ||
bbe735e4 | 639 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
9fb9cbb1 | 640 | dataoff, l3num, protonum, &tuple, l3proto, |
605dcad6 | 641 | l4proto)) { |
0d53778e | 642 | pr_debug("resolve_normal_ct: Can't get tuple\n"); |
9fb9cbb1 YK |
643 | return NULL; |
644 | } | |
645 | ||
646 | /* look for tuple match */ | |
a702a65f | 647 | h = nf_conntrack_find_get(net, &tuple); |
9fb9cbb1 | 648 | if (!h) { |
a702a65f | 649 | h = init_conntrack(net, &tuple, l3proto, l4proto, skb, dataoff); |
9fb9cbb1 YK |
650 | if (!h) |
651 | return NULL; | |
652 | if (IS_ERR(h)) | |
653 | return (void *)h; | |
654 | } | |
655 | ct = nf_ct_tuplehash_to_ctrack(h); | |
656 | ||
657 | /* It exists; we have (non-exclusive) reference. */ | |
658 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | |
659 | *ctinfo = IP_CT_ESTABLISHED + IP_CT_IS_REPLY; | |
660 | /* Please set reply bit if this packet OK */ | |
661 | *set_reply = 1; | |
662 | } else { | |
663 | /* Once we've had two way comms, always ESTABLISHED. */ | |
664 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | |
0d53778e | 665 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); |
9fb9cbb1 YK |
666 | *ctinfo = IP_CT_ESTABLISHED; |
667 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | |
0d53778e PM |
668 | pr_debug("nf_conntrack_in: related packet for %p\n", |
669 | ct); | |
9fb9cbb1 YK |
670 | *ctinfo = IP_CT_RELATED; |
671 | } else { | |
0d53778e | 672 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); |
9fb9cbb1 YK |
673 | *ctinfo = IP_CT_NEW; |
674 | } | |
675 | *set_reply = 0; | |
676 | } | |
677 | skb->nfct = &ct->ct_general; | |
678 | skb->nfctinfo = *ctinfo; | |
679 | return ct; | |
680 | } | |
681 | ||
682 | unsigned int | |
a702a65f AD |
683 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
684 | struct sk_buff *skb) | |
9fb9cbb1 YK |
685 | { |
686 | struct nf_conn *ct; | |
687 | enum ip_conntrack_info ctinfo; | |
688 | struct nf_conntrack_l3proto *l3proto; | |
605dcad6 | 689 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 YK |
690 | unsigned int dataoff; |
691 | u_int8_t protonum; | |
692 | int set_reply = 0; | |
693 | int ret; | |
694 | ||
695 | /* Previously seen (loopback or untracked)? Ignore. */ | |
3db05fea | 696 | if (skb->nfct) { |
0d55af87 | 697 | NF_CT_STAT_INC_ATOMIC(net, ignore); |
9fb9cbb1 YK |
698 | return NF_ACCEPT; |
699 | } | |
700 | ||
923f4902 | 701 | /* rcu_read_lock()ed by nf_hook_slow */ |
76108cea | 702 | l3proto = __nf_ct_l3proto_find(pf); |
3db05fea | 703 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
ffc30690 YK |
704 | &dataoff, &protonum); |
705 | if (ret <= 0) { | |
0d53778e | 706 | pr_debug("not prepared to track yet or error occured\n"); |
0d55af87 AD |
707 | NF_CT_STAT_INC_ATOMIC(net, error); |
708 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
9fb9cbb1 YK |
709 | return -ret; |
710 | } | |
711 | ||
76108cea | 712 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
9fb9cbb1 YK |
713 | |
714 | /* It may be an special packet, error, unclean... | |
715 | * inverse of the return code tells to the netfilter | |
716 | * core what to do with the packet. */ | |
74c51a14 AD |
717 | if (l4proto->error != NULL) { |
718 | ret = l4proto->error(net, skb, dataoff, &ctinfo, pf, hooknum); | |
719 | if (ret <= 0) { | |
0d55af87 AD |
720 | NF_CT_STAT_INC_ATOMIC(net, error); |
721 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
74c51a14 AD |
722 | return -ret; |
723 | } | |
9fb9cbb1 YK |
724 | } |
725 | ||
a702a65f AD |
726 | ct = resolve_normal_ct(net, skb, dataoff, pf, protonum, |
727 | l3proto, l4proto, &set_reply, &ctinfo); | |
9fb9cbb1 YK |
728 | if (!ct) { |
729 | /* Not valid part of a connection */ | |
0d55af87 | 730 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
9fb9cbb1 YK |
731 | return NF_ACCEPT; |
732 | } | |
733 | ||
734 | if (IS_ERR(ct)) { | |
735 | /* Too stressed to deal. */ | |
0d55af87 | 736 | NF_CT_STAT_INC_ATOMIC(net, drop); |
9fb9cbb1 YK |
737 | return NF_DROP; |
738 | } | |
739 | ||
3db05fea | 740 | NF_CT_ASSERT(skb->nfct); |
9fb9cbb1 | 741 | |
3db05fea | 742 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum); |
ec8d5409 | 743 | if (ret <= 0) { |
9fb9cbb1 YK |
744 | /* Invalid: inverse of the return code tells |
745 | * the netfilter core what to do */ | |
0d53778e | 746 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
3db05fea HX |
747 | nf_conntrack_put(skb->nfct); |
748 | skb->nfct = NULL; | |
0d55af87 | 749 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
7d1e0459 PNA |
750 | if (ret == -NF_DROP) |
751 | NF_CT_STAT_INC_ATOMIC(net, drop); | |
9fb9cbb1 YK |
752 | return -ret; |
753 | } | |
754 | ||
755 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
a71996fc | 756 | nf_conntrack_event_cache(IPCT_STATUS, ct); |
9fb9cbb1 YK |
757 | |
758 | return ret; | |
759 | } | |
13b18339 | 760 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
9fb9cbb1 | 761 | |
5f2b4c90 JE |
762 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
763 | const struct nf_conntrack_tuple *orig) | |
9fb9cbb1 | 764 | { |
5f2b4c90 | 765 | bool ret; |
923f4902 PM |
766 | |
767 | rcu_read_lock(); | |
768 | ret = nf_ct_invert_tuple(inverse, orig, | |
769 | __nf_ct_l3proto_find(orig->src.l3num), | |
770 | __nf_ct_l4proto_find(orig->src.l3num, | |
771 | orig->dst.protonum)); | |
772 | rcu_read_unlock(); | |
773 | return ret; | |
9fb9cbb1 | 774 | } |
13b18339 | 775 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
9fb9cbb1 | 776 | |
5b1158e9 JK |
777 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
778 | implicitly racy: see __nf_conntrack_confirm */ | |
779 | void nf_conntrack_alter_reply(struct nf_conn *ct, | |
780 | const struct nf_conntrack_tuple *newreply) | |
781 | { | |
782 | struct nf_conn_help *help = nfct_help(ct); | |
783 | ||
5b1158e9 JK |
784 | /* Should be unconfirmed, so not in hash table yet */ |
785 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
786 | ||
0d53778e | 787 | pr_debug("Altering reply tuple of %p to ", ct); |
3c9fba65 | 788 | nf_ct_dump_tuple(newreply); |
5b1158e9 JK |
789 | |
790 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | |
ef1a5a50 | 791 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
c52fbb41 | 792 | return; |
ceceae1b | 793 | |
c52fbb41 | 794 | rcu_read_lock(); |
226c0c0e | 795 | __nf_ct_try_assign_helper(ct, GFP_ATOMIC); |
c52fbb41 | 796 | rcu_read_unlock(); |
5b1158e9 | 797 | } |
13b18339 | 798 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
5b1158e9 | 799 | |
9fb9cbb1 YK |
800 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
801 | void __nf_ct_refresh_acct(struct nf_conn *ct, | |
802 | enum ip_conntrack_info ctinfo, | |
803 | const struct sk_buff *skb, | |
804 | unsigned long extra_jiffies, | |
805 | int do_acct) | |
806 | { | |
9fb9cbb1 YK |
807 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
808 | NF_CT_ASSERT(skb); | |
809 | ||
997ae831 | 810 | /* Only update if this is not a fixed timeout */ |
47d95045 PM |
811 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
812 | goto acct; | |
997ae831 | 813 | |
9fb9cbb1 YK |
814 | /* If not in hash table, timer will not be active yet */ |
815 | if (!nf_ct_is_confirmed(ct)) { | |
816 | ct->timeout.expires = extra_jiffies; | |
9fb9cbb1 | 817 | } else { |
be00c8e4 MJ |
818 | unsigned long newtime = jiffies + extra_jiffies; |
819 | ||
820 | /* Only update the timeout if the new timeout is at least | |
821 | HZ jiffies from the old timeout. Need del_timer for race | |
822 | avoidance (may already be dying). */ | |
65cb9fda PM |
823 | if (newtime - ct->timeout.expires >= HZ) |
824 | mod_timer_pending(&ct->timeout, newtime); | |
9fb9cbb1 YK |
825 | } |
826 | ||
47d95045 | 827 | acct: |
9fb9cbb1 | 828 | if (do_acct) { |
58401572 | 829 | struct nf_conn_counter *acct; |
3ffd5eeb | 830 | |
58401572 KPO |
831 | acct = nf_conn_acct_find(ct); |
832 | if (acct) { | |
65cb9fda | 833 | spin_lock_bh(&ct->lock); |
58401572 KPO |
834 | acct[CTINFO2DIR(ctinfo)].packets++; |
835 | acct[CTINFO2DIR(ctinfo)].bytes += | |
836 | skb->len - skb_network_offset(skb); | |
65cb9fda | 837 | spin_unlock_bh(&ct->lock); |
58401572 | 838 | } |
9fb9cbb1 | 839 | } |
9fb9cbb1 | 840 | } |
13b18339 | 841 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
9fb9cbb1 | 842 | |
4c889498 DM |
843 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
844 | enum ip_conntrack_info ctinfo, | |
845 | const struct sk_buff *skb, | |
846 | int do_acct) | |
51091764 | 847 | { |
718d4ad9 | 848 | if (do_acct) { |
58401572 KPO |
849 | struct nf_conn_counter *acct; |
850 | ||
58401572 KPO |
851 | acct = nf_conn_acct_find(ct); |
852 | if (acct) { | |
65cb9fda | 853 | spin_lock_bh(&ct->lock); |
58401572 KPO |
854 | acct[CTINFO2DIR(ctinfo)].packets++; |
855 | acct[CTINFO2DIR(ctinfo)].bytes += | |
856 | skb->len - skb_network_offset(skb); | |
65cb9fda | 857 | spin_unlock_bh(&ct->lock); |
58401572 | 858 | } |
718d4ad9 | 859 | } |
58401572 | 860 | |
4c889498 | 861 | if (del_timer(&ct->timeout)) { |
51091764 | 862 | ct->timeout.function((unsigned long)ct); |
4c889498 DM |
863 | return true; |
864 | } | |
865 | return false; | |
51091764 | 866 | } |
718d4ad9 | 867 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
51091764 | 868 | |
e281db5c | 869 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
c1d10adb PNA |
870 | |
871 | #include <linux/netfilter/nfnetlink.h> | |
872 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
57b47a53 IM |
873 | #include <linux/mutex.h> |
874 | ||
c1d10adb PNA |
875 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
876 | * in ip_conntrack_core, since we don't want the protocols to autoload | |
877 | * or depend on ctnetlink */ | |
fdf70832 | 878 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
c1d10adb PNA |
879 | const struct nf_conntrack_tuple *tuple) |
880 | { | |
77236b6e PM |
881 | NLA_PUT_BE16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port); |
882 | NLA_PUT_BE16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port); | |
c1d10adb PNA |
883 | return 0; |
884 | ||
df6fb868 | 885 | nla_put_failure: |
c1d10adb PNA |
886 | return -1; |
887 | } | |
fdf70832 | 888 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
c1d10adb | 889 | |
f73e924c PM |
890 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
891 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | |
892 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | |
c1d10adb | 893 | }; |
f73e924c | 894 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
c1d10adb | 895 | |
fdf70832 | 896 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
c1d10adb PNA |
897 | struct nf_conntrack_tuple *t) |
898 | { | |
df6fb868 | 899 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
c1d10adb PNA |
900 | return -EINVAL; |
901 | ||
77236b6e PM |
902 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
903 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | |
c1d10adb PNA |
904 | |
905 | return 0; | |
906 | } | |
fdf70832 | 907 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
5c0de29d HE |
908 | |
909 | int nf_ct_port_nlattr_tuple_size(void) | |
910 | { | |
911 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
912 | } | |
913 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | |
c1d10adb PNA |
914 | #endif |
915 | ||
9fb9cbb1 | 916 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
b334aadc | 917 | static void nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb) |
9fb9cbb1 YK |
918 | { |
919 | struct nf_conn *ct; | |
920 | enum ip_conntrack_info ctinfo; | |
921 | ||
922 | /* This ICMP is in reverse direction to the packet which caused it */ | |
923 | ct = nf_ct_get(skb, &ctinfo); | |
924 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | |
925 | ctinfo = IP_CT_RELATED + IP_CT_IS_REPLY; | |
926 | else | |
927 | ctinfo = IP_CT_RELATED; | |
928 | ||
929 | /* Attach to new skbuff, and increment count */ | |
930 | nskb->nfct = &ct->ct_general; | |
931 | nskb->nfctinfo = ctinfo; | |
932 | nf_conntrack_get(nskb->nfct); | |
933 | } | |
934 | ||
9fb9cbb1 | 935 | /* Bring out ya dead! */ |
df0933dc | 936 | static struct nf_conn * |
400dad39 | 937 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
9fb9cbb1 YK |
938 | void *data, unsigned int *bucket) |
939 | { | |
df0933dc PM |
940 | struct nf_conntrack_tuple_hash *h; |
941 | struct nf_conn *ct; | |
ea781f19 | 942 | struct hlist_nulls_node *n; |
9fb9cbb1 | 943 | |
f8ba1aff | 944 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 945 | for (; *bucket < nf_conntrack_htable_size; (*bucket)++) { |
ea781f19 | 946 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { |
df0933dc PM |
947 | ct = nf_ct_tuplehash_to_ctrack(h); |
948 | if (iter(ct, data)) | |
949 | goto found; | |
950 | } | |
601e68e1 | 951 | } |
ea781f19 | 952 | hlist_nulls_for_each_entry(h, n, &net->ct.unconfirmed, hnnode) { |
df0933dc PM |
953 | ct = nf_ct_tuplehash_to_ctrack(h); |
954 | if (iter(ct, data)) | |
ec68e97d | 955 | set_bit(IPS_DYING_BIT, &ct->status); |
df0933dc | 956 | } |
f8ba1aff | 957 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc PM |
958 | return NULL; |
959 | found: | |
c073e3fa | 960 | atomic_inc(&ct->ct_general.use); |
f8ba1aff | 961 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc | 962 | return ct; |
9fb9cbb1 YK |
963 | } |
964 | ||
400dad39 AD |
965 | void nf_ct_iterate_cleanup(struct net *net, |
966 | int (*iter)(struct nf_conn *i, void *data), | |
967 | void *data) | |
9fb9cbb1 | 968 | { |
df0933dc | 969 | struct nf_conn *ct; |
9fb9cbb1 YK |
970 | unsigned int bucket = 0; |
971 | ||
400dad39 | 972 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
9fb9cbb1 YK |
973 | /* Time to push up daises... */ |
974 | if (del_timer(&ct->timeout)) | |
975 | death_by_timeout((unsigned long)ct); | |
976 | /* ... else the timer will get him soon. */ | |
977 | ||
978 | nf_ct_put(ct); | |
979 | } | |
980 | } | |
13b18339 | 981 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
9fb9cbb1 | 982 | |
19abb7b0 PNA |
983 | struct __nf_ct_flush_report { |
984 | u32 pid; | |
985 | int report; | |
986 | }; | |
987 | ||
274d383b | 988 | static int kill_report(struct nf_conn *i, void *data) |
9fb9cbb1 | 989 | { |
19abb7b0 PNA |
990 | struct __nf_ct_flush_report *fr = (struct __nf_ct_flush_report *)data; |
991 | ||
992 | /* get_next_corpse sets the dying bit for us */ | |
993 | nf_conntrack_event_report(IPCT_DESTROY, | |
994 | i, | |
995 | fr->pid, | |
996 | fr->report); | |
9fb9cbb1 YK |
997 | return 1; |
998 | } | |
999 | ||
274d383b PNA |
1000 | static int kill_all(struct nf_conn *i, void *data) |
1001 | { | |
1002 | return 1; | |
1003 | } | |
1004 | ||
ea781f19 | 1005 | void nf_ct_free_hashtable(void *hash, int vmalloced, unsigned int size) |
9fb9cbb1 YK |
1006 | { |
1007 | if (vmalloced) | |
1008 | vfree(hash); | |
1009 | else | |
601e68e1 | 1010 | free_pages((unsigned long)hash, |
f205c5e0 | 1011 | get_order(sizeof(struct hlist_head) * size)); |
9fb9cbb1 | 1012 | } |
ac565e5f | 1013 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
9fb9cbb1 | 1014 | |
274d383b | 1015 | void nf_conntrack_flush_report(struct net *net, u32 pid, int report) |
c1d10adb | 1016 | { |
19abb7b0 PNA |
1017 | struct __nf_ct_flush_report fr = { |
1018 | .pid = pid, | |
1019 | .report = report, | |
1020 | }; | |
274d383b | 1021 | nf_ct_iterate_cleanup(net, kill_report, &fr); |
c1d10adb | 1022 | } |
274d383b | 1023 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); |
c1d10adb | 1024 | |
08f6547d | 1025 | static void nf_conntrack_cleanup_init_net(void) |
9fb9cbb1 | 1026 | { |
08f6547d AD |
1027 | nf_conntrack_helper_fini(); |
1028 | nf_conntrack_proto_fini(); | |
1029 | kmem_cache_destroy(nf_conntrack_cachep); | |
1030 | } | |
9fb9cbb1 | 1031 | |
08f6547d AD |
1032 | static void nf_conntrack_cleanup_net(struct net *net) |
1033 | { | |
6058fa6b AD |
1034 | nf_ct_event_cache_flush(net); |
1035 | nf_conntrack_ecache_fini(net); | |
9fb9cbb1 | 1036 | i_see_dead_people: |
274d383b | 1037 | nf_ct_iterate_cleanup(net, kill_all, NULL); |
49ac8713 | 1038 | if (atomic_read(&net->ct.count) != 0) { |
9fb9cbb1 YK |
1039 | schedule(); |
1040 | goto i_see_dead_people; | |
1041 | } | |
6636568c PM |
1042 | /* wait until all references to nf_conntrack_untracked are dropped */ |
1043 | while (atomic_read(&nf_conntrack_untracked.ct_general.use) > 1) | |
1044 | schedule(); | |
9fb9cbb1 | 1045 | |
400dad39 | 1046 | nf_ct_free_hashtable(net->ct.hash, net->ct.hash_vmalloc, |
ac565e5f | 1047 | nf_conntrack_htable_size); |
d716a4df | 1048 | nf_conntrack_acct_fini(net); |
9b03f38d | 1049 | nf_conntrack_expect_fini(net); |
0d55af87 | 1050 | free_percpu(net->ct.stat); |
08f6547d AD |
1051 | } |
1052 | ||
1053 | /* Mishearing the voices in his head, our hero wonders how he's | |
1054 | supposed to kill the mall. */ | |
1055 | void nf_conntrack_cleanup(struct net *net) | |
1056 | { | |
1057 | if (net_eq(net, &init_net)) | |
1058 | rcu_assign_pointer(ip_ct_attach, NULL); | |
1059 | ||
1060 | /* This makes sure all current packets have passed through | |
1061 | netfilter framework. Roll on, two-stage module | |
1062 | delete... */ | |
1063 | synchronize_net(); | |
1064 | ||
1065 | nf_conntrack_cleanup_net(net); | |
1066 | ||
1067 | if (net_eq(net, &init_net)) { | |
1068 | rcu_assign_pointer(nf_ct_destroy, NULL); | |
1069 | nf_conntrack_cleanup_init_net(); | |
1070 | } | |
9fb9cbb1 YK |
1071 | } |
1072 | ||
ea781f19 | 1073 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int *vmalloced, int nulls) |
9fb9cbb1 | 1074 | { |
ea781f19 ED |
1075 | struct hlist_nulls_head *hash; |
1076 | unsigned int nr_slots, i; | |
1077 | size_t sz; | |
9fb9cbb1 | 1078 | |
601e68e1 | 1079 | *vmalloced = 0; |
8e5105a0 | 1080 | |
ea781f19 ED |
1081 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1082 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | |
1083 | sz = nr_slots * sizeof(struct hlist_nulls_head); | |
1084 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | |
1085 | get_order(sz)); | |
601e68e1 | 1086 | if (!hash) { |
9fb9cbb1 YK |
1087 | *vmalloced = 1; |
1088 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); | |
ea781f19 | 1089 | hash = __vmalloc(sz, GFP_KERNEL | __GFP_ZERO, PAGE_KERNEL); |
9fb9cbb1 YK |
1090 | } |
1091 | ||
ea781f19 ED |
1092 | if (hash && nulls) |
1093 | for (i = 0; i < nr_slots; i++) | |
1094 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | |
9fb9cbb1 YK |
1095 | |
1096 | return hash; | |
1097 | } | |
ac565e5f | 1098 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
9fb9cbb1 | 1099 | |
fae718dd | 1100 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
9fb9cbb1 | 1101 | { |
96eb24d7 SH |
1102 | int i, bucket, vmalloced, old_vmalloced; |
1103 | unsigned int hashsize, old_size; | |
9fb9cbb1 | 1104 | int rnd; |
ea781f19 | 1105 | struct hlist_nulls_head *hash, *old_hash; |
9fb9cbb1 YK |
1106 | struct nf_conntrack_tuple_hash *h; |
1107 | ||
1108 | /* On boot, we can set this without any fancy locking. */ | |
1109 | if (!nf_conntrack_htable_size) | |
1110 | return param_set_uint(val, kp); | |
1111 | ||
96eb24d7 | 1112 | hashsize = simple_strtoul(val, NULL, 0); |
9fb9cbb1 YK |
1113 | if (!hashsize) |
1114 | return -EINVAL; | |
1115 | ||
ea781f19 | 1116 | hash = nf_ct_alloc_hashtable(&hashsize, &vmalloced, 1); |
9fb9cbb1 YK |
1117 | if (!hash) |
1118 | return -ENOMEM; | |
1119 | ||
1120 | /* We have to rehahs for the new table anyway, so we also can | |
1121 | * use a newrandom seed */ | |
af07d241 | 1122 | get_random_bytes(&rnd, sizeof(rnd)); |
9fb9cbb1 | 1123 | |
76507f69 PM |
1124 | /* Lookups in the old hash might happen in parallel, which means we |
1125 | * might get false negatives during connection lookup. New connections | |
1126 | * created because of a false negative won't make it into the hash | |
1127 | * though since that required taking the lock. | |
1128 | */ | |
f8ba1aff | 1129 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1130 | for (i = 0; i < nf_conntrack_htable_size; i++) { |
ea781f19 ED |
1131 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
1132 | h = hlist_nulls_entry(init_net.ct.hash[i].first, | |
1133 | struct nf_conntrack_tuple_hash, hnnode); | |
1134 | hlist_nulls_del_rcu(&h->hnnode); | |
9fb9cbb1 | 1135 | bucket = __hash_conntrack(&h->tuple, hashsize, rnd); |
ea781f19 | 1136 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
9fb9cbb1 YK |
1137 | } |
1138 | } | |
1139 | old_size = nf_conntrack_htable_size; | |
400dad39 AD |
1140 | old_vmalloced = init_net.ct.hash_vmalloc; |
1141 | old_hash = init_net.ct.hash; | |
9fb9cbb1 YK |
1142 | |
1143 | nf_conntrack_htable_size = hashsize; | |
400dad39 AD |
1144 | init_net.ct.hash_vmalloc = vmalloced; |
1145 | init_net.ct.hash = hash; | |
9fb9cbb1 | 1146 | nf_conntrack_hash_rnd = rnd; |
f8ba1aff | 1147 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1148 | |
ac565e5f | 1149 | nf_ct_free_hashtable(old_hash, old_vmalloced, old_size); |
9fb9cbb1 YK |
1150 | return 0; |
1151 | } | |
fae718dd | 1152 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
9fb9cbb1 | 1153 | |
fae718dd | 1154 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
9fb9cbb1 YK |
1155 | &nf_conntrack_htable_size, 0600); |
1156 | ||
08f6547d | 1157 | static int nf_conntrack_init_init_net(void) |
9fb9cbb1 | 1158 | { |
f205c5e0 | 1159 | int max_factor = 8; |
9fb9cbb1 YK |
1160 | int ret; |
1161 | ||
1162 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB | |
f205c5e0 | 1163 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ |
9fb9cbb1 YK |
1164 | if (!nf_conntrack_htable_size) { |
1165 | nf_conntrack_htable_size | |
1166 | = (((num_physpages << PAGE_SHIFT) / 16384) | |
f205c5e0 | 1167 | / sizeof(struct hlist_head)); |
9fb9cbb1 | 1168 | if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
f205c5e0 PM |
1169 | nf_conntrack_htable_size = 16384; |
1170 | if (nf_conntrack_htable_size < 32) | |
1171 | nf_conntrack_htable_size = 32; | |
1172 | ||
1173 | /* Use a max. factor of four by default to get the same max as | |
1174 | * with the old struct list_heads. When a table size is given | |
1175 | * we use the old value of 8 to avoid reducing the max. | |
1176 | * entries. */ | |
1177 | max_factor = 4; | |
9fb9cbb1 | 1178 | } |
f205c5e0 | 1179 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
8e5105a0 PM |
1180 | |
1181 | printk("nf_conntrack version %s (%u buckets, %d max)\n", | |
1182 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, | |
1183 | nf_conntrack_max); | |
1184 | ||
dacd2a1a YK |
1185 | nf_conntrack_cachep = kmem_cache_create("nf_conntrack", |
1186 | sizeof(struct nf_conn), | |
ea781f19 | 1187 | 0, SLAB_DESTROY_BY_RCU, NULL); |
dacd2a1a | 1188 | if (!nf_conntrack_cachep) { |
9fb9cbb1 | 1189 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); |
08f6547d AD |
1190 | ret = -ENOMEM; |
1191 | goto err_cache; | |
9fb9cbb1 YK |
1192 | } |
1193 | ||
e9c1b084 PM |
1194 | ret = nf_conntrack_proto_init(); |
1195 | if (ret < 0) | |
08f6547d | 1196 | goto err_proto; |
933a41e7 | 1197 | |
ceceae1b YK |
1198 | ret = nf_conntrack_helper_init(); |
1199 | if (ret < 0) | |
08f6547d AD |
1200 | goto err_helper; |
1201 | ||
1202 | return 0; | |
1203 | ||
1204 | err_helper: | |
1205 | nf_conntrack_proto_fini(); | |
1206 | err_proto: | |
1207 | kmem_cache_destroy(nf_conntrack_cachep); | |
1208 | err_cache: | |
1209 | return ret; | |
1210 | } | |
1211 | ||
1212 | static int nf_conntrack_init_net(struct net *net) | |
1213 | { | |
1214 | int ret; | |
ceceae1b | 1215 | |
08f6547d | 1216 | atomic_set(&net->ct.count, 0); |
ea781f19 | 1217 | INIT_HLIST_NULLS_HEAD(&net->ct.unconfirmed, 0); |
08f6547d AD |
1218 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1219 | if (!net->ct.stat) { | |
1220 | ret = -ENOMEM; | |
1221 | goto err_stat; | |
1222 | } | |
1223 | ret = nf_conntrack_ecache_init(net); | |
1224 | if (ret < 0) | |
1225 | goto err_ecache; | |
1226 | net->ct.hash = nf_ct_alloc_hashtable(&nf_conntrack_htable_size, | |
ea781f19 | 1227 | &net->ct.hash_vmalloc, 1); |
08f6547d AD |
1228 | if (!net->ct.hash) { |
1229 | ret = -ENOMEM; | |
1230 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); | |
1231 | goto err_hash; | |
1232 | } | |
1233 | ret = nf_conntrack_expect_init(net); | |
1234 | if (ret < 0) | |
1235 | goto err_expect; | |
d716a4df | 1236 | ret = nf_conntrack_acct_init(net); |
58401572 | 1237 | if (ret < 0) |
08f6547d | 1238 | goto err_acct; |
7d3cdc6b | 1239 | |
9fb9cbb1 YK |
1240 | /* Set up fake conntrack: |
1241 | - to never be deleted, not in any hashes */ | |
5a1fb391 AD |
1242 | #ifdef CONFIG_NET_NS |
1243 | nf_conntrack_untracked.ct_net = &init_net; | |
1244 | #endif | |
9fb9cbb1 YK |
1245 | atomic_set(&nf_conntrack_untracked.ct_general.use, 1); |
1246 | /* - and look it like as a confirmed connection */ | |
1247 | set_bit(IPS_CONFIRMED_BIT, &nf_conntrack_untracked.status); | |
1248 | ||
08f6547d | 1249 | return 0; |
9fb9cbb1 | 1250 | |
08f6547d | 1251 | err_acct: |
9b03f38d | 1252 | nf_conntrack_expect_fini(net); |
08f6547d | 1253 | err_expect: |
400dad39 | 1254 | nf_ct_free_hashtable(net->ct.hash, net->ct.hash_vmalloc, |
ac565e5f | 1255 | nf_conntrack_htable_size); |
6058fa6b AD |
1256 | err_hash: |
1257 | nf_conntrack_ecache_fini(net); | |
1258 | err_ecache: | |
0d55af87 AD |
1259 | free_percpu(net->ct.stat); |
1260 | err_stat: | |
08f6547d AD |
1261 | return ret; |
1262 | } | |
1263 | ||
1264 | int nf_conntrack_init(struct net *net) | |
1265 | { | |
1266 | int ret; | |
1267 | ||
1268 | if (net_eq(net, &init_net)) { | |
1269 | ret = nf_conntrack_init_init_net(); | |
1270 | if (ret < 0) | |
1271 | goto out_init_net; | |
1272 | } | |
1273 | ret = nf_conntrack_init_net(net); | |
1274 | if (ret < 0) | |
1275 | goto out_net; | |
1276 | ||
1277 | if (net_eq(net, &init_net)) { | |
1278 | /* For use by REJECT target */ | |
1279 | rcu_assign_pointer(ip_ct_attach, nf_conntrack_attach); | |
1280 | rcu_assign_pointer(nf_ct_destroy, destroy_conntrack); | |
1281 | } | |
1282 | return 0; | |
1283 | ||
1284 | out_net: | |
1285 | if (net_eq(net, &init_net)) | |
1286 | nf_conntrack_cleanup_init_net(); | |
1287 | out_init_net: | |
1288 | return ret; | |
9fb9cbb1 | 1289 | } |