]>
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); |
9fb9cbb1 | 401 | #ifdef CONFIG_NF_NAT_NEEDED |
df0933dc PM |
402 | if (test_bit(IPS_SRC_NAT_DONE_BIT, &ct->status) || |
403 | test_bit(IPS_DST_NAT_DONE_BIT, &ct->status)) | |
a71996fc | 404 | nf_conntrack_event_cache(IPCT_NATINFO, ct); |
9fb9cbb1 | 405 | #endif |
df0933dc | 406 | nf_conntrack_event_cache(master_ct(ct) ? |
a71996fc | 407 | IPCT_RELATED : IPCT_NEW, ct); |
df0933dc | 408 | return NF_ACCEPT; |
9fb9cbb1 | 409 | |
df0933dc | 410 | out: |
0d55af87 | 411 | NF_CT_STAT_INC(net, insert_failed); |
f8ba1aff | 412 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
413 | return NF_DROP; |
414 | } | |
13b18339 | 415 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
9fb9cbb1 YK |
416 | |
417 | /* Returns true if a connection correspondings to the tuple (required | |
418 | for NAT). */ | |
419 | int | |
420 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | |
421 | const struct nf_conn *ignored_conntrack) | |
422 | { | |
400dad39 | 423 | struct net *net = nf_ct_net(ignored_conntrack); |
9fb9cbb1 | 424 | struct nf_conntrack_tuple_hash *h; |
ea781f19 | 425 | struct hlist_nulls_node *n; |
ba419aff | 426 | unsigned int hash = hash_conntrack(tuple); |
9fb9cbb1 | 427 | |
4e29e9ec PM |
428 | /* Disable BHs the entire time since we need to disable them at |
429 | * least once for the stats anyway. | |
430 | */ | |
431 | rcu_read_lock_bh(); | |
ea781f19 | 432 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
ba419aff PM |
433 | if (nf_ct_tuplehash_to_ctrack(h) != ignored_conntrack && |
434 | nf_ct_tuple_equal(tuple, &h->tuple)) { | |
0d55af87 | 435 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 436 | rcu_read_unlock_bh(); |
ba419aff PM |
437 | return 1; |
438 | } | |
0d55af87 | 439 | NF_CT_STAT_INC(net, searched); |
ba419aff | 440 | } |
4e29e9ec | 441 | rcu_read_unlock_bh(); |
9fb9cbb1 | 442 | |
ba419aff | 443 | return 0; |
9fb9cbb1 | 444 | } |
13b18339 | 445 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
9fb9cbb1 | 446 | |
7ae7730f PM |
447 | #define NF_CT_EVICTION_RANGE 8 |
448 | ||
9fb9cbb1 YK |
449 | /* There's a small race here where we may free a just-assured |
450 | connection. Too bad: we're in trouble anyway. */ | |
400dad39 | 451 | static noinline int early_drop(struct net *net, unsigned int hash) |
9fb9cbb1 | 452 | { |
f205c5e0 | 453 | /* Use oldest entry, which is roughly LRU */ |
9fb9cbb1 | 454 | struct nf_conntrack_tuple_hash *h; |
df0933dc | 455 | struct nf_conn *ct = NULL, *tmp; |
ea781f19 | 456 | struct hlist_nulls_node *n; |
7ae7730f | 457 | unsigned int i, cnt = 0; |
9fb9cbb1 YK |
458 | int dropped = 0; |
459 | ||
76507f69 | 460 | rcu_read_lock(); |
7ae7730f | 461 | for (i = 0; i < nf_conntrack_htable_size; i++) { |
ea781f19 ED |
462 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
463 | hnnode) { | |
7ae7730f PM |
464 | tmp = nf_ct_tuplehash_to_ctrack(h); |
465 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) | |
466 | ct = tmp; | |
467 | cnt++; | |
468 | } | |
76507f69 PM |
469 | |
470 | if (ct && unlikely(!atomic_inc_not_zero(&ct->ct_general.use))) | |
471 | ct = NULL; | |
7ae7730f PM |
472 | if (ct || cnt >= NF_CT_EVICTION_RANGE) |
473 | break; | |
474 | hash = (hash + 1) % nf_conntrack_htable_size; | |
9fb9cbb1 | 475 | } |
76507f69 | 476 | rcu_read_unlock(); |
9fb9cbb1 YK |
477 | |
478 | if (!ct) | |
479 | return dropped; | |
480 | ||
481 | if (del_timer(&ct->timeout)) { | |
482 | death_by_timeout((unsigned long)ct); | |
483 | dropped = 1; | |
0d55af87 | 484 | NF_CT_STAT_INC_ATOMIC(net, early_drop); |
9fb9cbb1 YK |
485 | } |
486 | nf_ct_put(ct); | |
487 | return dropped; | |
488 | } | |
489 | ||
5a1fb391 AD |
490 | struct nf_conn *nf_conntrack_alloc(struct net *net, |
491 | const struct nf_conntrack_tuple *orig, | |
b891c5a8 PNA |
492 | const struct nf_conntrack_tuple *repl, |
493 | gfp_t gfp) | |
9fb9cbb1 | 494 | { |
cd7fcbf1 | 495 | struct nf_conn *ct; |
9fb9cbb1 | 496 | |
dc808fe2 | 497 | if (unlikely(!nf_conntrack_hash_rnd_initted)) { |
af07d241 HPP |
498 | get_random_bytes(&nf_conntrack_hash_rnd, |
499 | sizeof(nf_conntrack_hash_rnd)); | |
9fb9cbb1 YK |
500 | nf_conntrack_hash_rnd_initted = 1; |
501 | } | |
502 | ||
5251e2d2 | 503 | /* We don't want any race condition at early drop stage */ |
49ac8713 | 504 | atomic_inc(&net->ct.count); |
5251e2d2 | 505 | |
76eb9460 | 506 | if (nf_conntrack_max && |
49ac8713 | 507 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
9fb9cbb1 | 508 | unsigned int hash = hash_conntrack(orig); |
400dad39 | 509 | if (!early_drop(net, hash)) { |
49ac8713 | 510 | atomic_dec(&net->ct.count); |
9fb9cbb1 YK |
511 | if (net_ratelimit()) |
512 | printk(KERN_WARNING | |
513 | "nf_conntrack: table full, dropping" | |
514 | " packet.\n"); | |
515 | return ERR_PTR(-ENOMEM); | |
516 | } | |
517 | } | |
518 | ||
b891c5a8 | 519 | ct = kmem_cache_zalloc(nf_conntrack_cachep, gfp); |
c88130bc | 520 | if (ct == NULL) { |
0d53778e | 521 | pr_debug("nf_conntrack_alloc: Can't alloc conntrack.\n"); |
49ac8713 | 522 | atomic_dec(&net->ct.count); |
dacd2a1a | 523 | return ERR_PTR(-ENOMEM); |
9fb9cbb1 YK |
524 | } |
525 | ||
c88130bc PM |
526 | atomic_set(&ct->ct_general.use, 1); |
527 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; | |
528 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; | |
9fb9cbb1 | 529 | /* Don't set timer yet: wait for confirmation */ |
c88130bc | 530 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
5a1fb391 AD |
531 | #ifdef CONFIG_NET_NS |
532 | ct->ct_net = net; | |
533 | #endif | |
9fb9cbb1 | 534 | |
c88130bc | 535 | return ct; |
9fb9cbb1 | 536 | } |
13b18339 | 537 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
9fb9cbb1 | 538 | |
c88130bc | 539 | void nf_conntrack_free(struct nf_conn *ct) |
76507f69 | 540 | { |
1d45209d ED |
541 | struct net *net = nf_ct_net(ct); |
542 | ||
ceeff754 | 543 | nf_ct_ext_destroy(ct); |
1d45209d | 544 | atomic_dec(&net->ct.count); |
ea781f19 ED |
545 | nf_ct_ext_free(ct); |
546 | kmem_cache_free(nf_conntrack_cachep, ct); | |
76507f69 | 547 | } |
13b18339 | 548 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
9fb9cbb1 YK |
549 | |
550 | /* Allocate a new conntrack: we return -ENOMEM if classification | |
551 | failed due to stress. Otherwise it really is unclassifiable. */ | |
552 | static struct nf_conntrack_tuple_hash * | |
5a1fb391 AD |
553 | init_conntrack(struct net *net, |
554 | const struct nf_conntrack_tuple *tuple, | |
9fb9cbb1 | 555 | struct nf_conntrack_l3proto *l3proto, |
605dcad6 | 556 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 YK |
557 | struct sk_buff *skb, |
558 | unsigned int dataoff) | |
559 | { | |
c88130bc | 560 | struct nf_conn *ct; |
3c158f7f | 561 | struct nf_conn_help *help; |
9fb9cbb1 YK |
562 | struct nf_conntrack_tuple repl_tuple; |
563 | struct nf_conntrack_expect *exp; | |
564 | ||
605dcad6 | 565 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
0d53778e | 566 | pr_debug("Can't invert tuple.\n"); |
9fb9cbb1 YK |
567 | return NULL; |
568 | } | |
569 | ||
5a1fb391 | 570 | ct = nf_conntrack_alloc(net, tuple, &repl_tuple, GFP_ATOMIC); |
cd7fcbf1 | 571 | if (IS_ERR(ct)) { |
0d53778e | 572 | pr_debug("Can't allocate conntrack.\n"); |
c88130bc | 573 | return (struct nf_conntrack_tuple_hash *)ct; |
9fb9cbb1 YK |
574 | } |
575 | ||
c88130bc PM |
576 | if (!l4proto->new(ct, skb, dataoff)) { |
577 | nf_conntrack_free(ct); | |
0d53778e | 578 | pr_debug("init conntrack: can't track with proto module\n"); |
9fb9cbb1 YK |
579 | return NULL; |
580 | } | |
581 | ||
58401572 KPO |
582 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
583 | ||
f8ba1aff | 584 | spin_lock_bh(&nf_conntrack_lock); |
9b03f38d | 585 | exp = nf_ct_find_expectation(net, tuple); |
9fb9cbb1 | 586 | if (exp) { |
0d53778e | 587 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", |
c88130bc | 588 | ct, exp); |
9fb9cbb1 | 589 | /* Welcome, Mr. Bond. We've been expecting you... */ |
c88130bc PM |
590 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
591 | ct->master = exp->master; | |
ceceae1b | 592 | if (exp->helper) { |
c88130bc | 593 | help = nf_ct_helper_ext_add(ct, GFP_ATOMIC); |
ceceae1b YK |
594 | if (help) |
595 | rcu_assign_pointer(help->helper, exp->helper); | |
ceceae1b YK |
596 | } |
597 | ||
9fb9cbb1 | 598 | #ifdef CONFIG_NF_CONNTRACK_MARK |
c88130bc | 599 | ct->mark = exp->master->mark; |
7c9728c3 JM |
600 | #endif |
601 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | |
c88130bc | 602 | ct->secmark = exp->master->secmark; |
9fb9cbb1 | 603 | #endif |
c88130bc | 604 | nf_conntrack_get(&ct->master->ct_general); |
0d55af87 | 605 | NF_CT_STAT_INC(net, expect_new); |
22e7410b | 606 | } else { |
226c0c0e | 607 | __nf_ct_try_assign_helper(ct, GFP_ATOMIC); |
0d55af87 | 608 | NF_CT_STAT_INC(net, new); |
22e7410b | 609 | } |
9fb9cbb1 YK |
610 | |
611 | /* Overload tuple linked list to put us in unconfirmed list. */ | |
ea781f19 | 612 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
63c9a262 | 613 | &net->ct.unconfirmed); |
9fb9cbb1 | 614 | |
f8ba1aff | 615 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
616 | |
617 | if (exp) { | |
618 | if (exp->expectfn) | |
c88130bc | 619 | exp->expectfn(ct, exp); |
6823645d | 620 | nf_ct_expect_put(exp); |
9fb9cbb1 YK |
621 | } |
622 | ||
c88130bc | 623 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
9fb9cbb1 YK |
624 | } |
625 | ||
626 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | |
627 | static inline struct nf_conn * | |
a702a65f AD |
628 | resolve_normal_ct(struct net *net, |
629 | struct sk_buff *skb, | |
9fb9cbb1 YK |
630 | unsigned int dataoff, |
631 | u_int16_t l3num, | |
632 | u_int8_t protonum, | |
633 | struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 634 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 YK |
635 | int *set_reply, |
636 | enum ip_conntrack_info *ctinfo) | |
637 | { | |
638 | struct nf_conntrack_tuple tuple; | |
639 | struct nf_conntrack_tuple_hash *h; | |
640 | struct nf_conn *ct; | |
641 | ||
bbe735e4 | 642 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
9fb9cbb1 | 643 | dataoff, l3num, protonum, &tuple, l3proto, |
605dcad6 | 644 | l4proto)) { |
0d53778e | 645 | pr_debug("resolve_normal_ct: Can't get tuple\n"); |
9fb9cbb1 YK |
646 | return NULL; |
647 | } | |
648 | ||
649 | /* look for tuple match */ | |
a702a65f | 650 | h = nf_conntrack_find_get(net, &tuple); |
9fb9cbb1 | 651 | if (!h) { |
a702a65f | 652 | h = init_conntrack(net, &tuple, l3proto, l4proto, skb, dataoff); |
9fb9cbb1 YK |
653 | if (!h) |
654 | return NULL; | |
655 | if (IS_ERR(h)) | |
656 | return (void *)h; | |
657 | } | |
658 | ct = nf_ct_tuplehash_to_ctrack(h); | |
659 | ||
660 | /* It exists; we have (non-exclusive) reference. */ | |
661 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | |
662 | *ctinfo = IP_CT_ESTABLISHED + IP_CT_IS_REPLY; | |
663 | /* Please set reply bit if this packet OK */ | |
664 | *set_reply = 1; | |
665 | } else { | |
666 | /* Once we've had two way comms, always ESTABLISHED. */ | |
667 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | |
0d53778e | 668 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); |
9fb9cbb1 YK |
669 | *ctinfo = IP_CT_ESTABLISHED; |
670 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | |
0d53778e PM |
671 | pr_debug("nf_conntrack_in: related packet for %p\n", |
672 | ct); | |
9fb9cbb1 YK |
673 | *ctinfo = IP_CT_RELATED; |
674 | } else { | |
0d53778e | 675 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); |
9fb9cbb1 YK |
676 | *ctinfo = IP_CT_NEW; |
677 | } | |
678 | *set_reply = 0; | |
679 | } | |
680 | skb->nfct = &ct->ct_general; | |
681 | skb->nfctinfo = *ctinfo; | |
682 | return ct; | |
683 | } | |
684 | ||
685 | unsigned int | |
a702a65f AD |
686 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
687 | struct sk_buff *skb) | |
9fb9cbb1 YK |
688 | { |
689 | struct nf_conn *ct; | |
690 | enum ip_conntrack_info ctinfo; | |
691 | struct nf_conntrack_l3proto *l3proto; | |
605dcad6 | 692 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 YK |
693 | unsigned int dataoff; |
694 | u_int8_t protonum; | |
695 | int set_reply = 0; | |
696 | int ret; | |
697 | ||
698 | /* Previously seen (loopback or untracked)? Ignore. */ | |
3db05fea | 699 | if (skb->nfct) { |
0d55af87 | 700 | NF_CT_STAT_INC_ATOMIC(net, ignore); |
9fb9cbb1 YK |
701 | return NF_ACCEPT; |
702 | } | |
703 | ||
923f4902 | 704 | /* rcu_read_lock()ed by nf_hook_slow */ |
76108cea | 705 | l3proto = __nf_ct_l3proto_find(pf); |
3db05fea | 706 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
ffc30690 YK |
707 | &dataoff, &protonum); |
708 | if (ret <= 0) { | |
0d53778e | 709 | pr_debug("not prepared to track yet or error occured\n"); |
0d55af87 AD |
710 | NF_CT_STAT_INC_ATOMIC(net, error); |
711 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
9fb9cbb1 YK |
712 | return -ret; |
713 | } | |
714 | ||
76108cea | 715 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
9fb9cbb1 YK |
716 | |
717 | /* It may be an special packet, error, unclean... | |
718 | * inverse of the return code tells to the netfilter | |
719 | * core what to do with the packet. */ | |
74c51a14 AD |
720 | if (l4proto->error != NULL) { |
721 | ret = l4proto->error(net, skb, dataoff, &ctinfo, pf, hooknum); | |
722 | if (ret <= 0) { | |
0d55af87 AD |
723 | NF_CT_STAT_INC_ATOMIC(net, error); |
724 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
74c51a14 AD |
725 | return -ret; |
726 | } | |
9fb9cbb1 YK |
727 | } |
728 | ||
a702a65f AD |
729 | ct = resolve_normal_ct(net, skb, dataoff, pf, protonum, |
730 | l3proto, l4proto, &set_reply, &ctinfo); | |
9fb9cbb1 YK |
731 | if (!ct) { |
732 | /* Not valid part of a connection */ | |
0d55af87 | 733 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
9fb9cbb1 YK |
734 | return NF_ACCEPT; |
735 | } | |
736 | ||
737 | if (IS_ERR(ct)) { | |
738 | /* Too stressed to deal. */ | |
0d55af87 | 739 | NF_CT_STAT_INC_ATOMIC(net, drop); |
9fb9cbb1 YK |
740 | return NF_DROP; |
741 | } | |
742 | ||
3db05fea | 743 | NF_CT_ASSERT(skb->nfct); |
9fb9cbb1 | 744 | |
3db05fea | 745 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum); |
ec8d5409 | 746 | if (ret <= 0) { |
9fb9cbb1 YK |
747 | /* Invalid: inverse of the return code tells |
748 | * the netfilter core what to do */ | |
0d53778e | 749 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
3db05fea HX |
750 | nf_conntrack_put(skb->nfct); |
751 | skb->nfct = NULL; | |
0d55af87 | 752 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
7d1e0459 PNA |
753 | if (ret == -NF_DROP) |
754 | NF_CT_STAT_INC_ATOMIC(net, drop); | |
9fb9cbb1 YK |
755 | return -ret; |
756 | } | |
757 | ||
758 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
a71996fc | 759 | nf_conntrack_event_cache(IPCT_STATUS, ct); |
9fb9cbb1 YK |
760 | |
761 | return ret; | |
762 | } | |
13b18339 | 763 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
9fb9cbb1 | 764 | |
5f2b4c90 JE |
765 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
766 | const struct nf_conntrack_tuple *orig) | |
9fb9cbb1 | 767 | { |
5f2b4c90 | 768 | bool ret; |
923f4902 PM |
769 | |
770 | rcu_read_lock(); | |
771 | ret = nf_ct_invert_tuple(inverse, orig, | |
772 | __nf_ct_l3proto_find(orig->src.l3num), | |
773 | __nf_ct_l4proto_find(orig->src.l3num, | |
774 | orig->dst.protonum)); | |
775 | rcu_read_unlock(); | |
776 | return ret; | |
9fb9cbb1 | 777 | } |
13b18339 | 778 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
9fb9cbb1 | 779 | |
5b1158e9 JK |
780 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
781 | implicitly racy: see __nf_conntrack_confirm */ | |
782 | void nf_conntrack_alter_reply(struct nf_conn *ct, | |
783 | const struct nf_conntrack_tuple *newreply) | |
784 | { | |
785 | struct nf_conn_help *help = nfct_help(ct); | |
786 | ||
5b1158e9 JK |
787 | /* Should be unconfirmed, so not in hash table yet */ |
788 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
789 | ||
0d53778e | 790 | pr_debug("Altering reply tuple of %p to ", ct); |
3c9fba65 | 791 | nf_ct_dump_tuple(newreply); |
5b1158e9 JK |
792 | |
793 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | |
ef1a5a50 | 794 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
c52fbb41 | 795 | return; |
ceceae1b | 796 | |
c52fbb41 | 797 | rcu_read_lock(); |
226c0c0e | 798 | __nf_ct_try_assign_helper(ct, GFP_ATOMIC); |
c52fbb41 | 799 | rcu_read_unlock(); |
5b1158e9 | 800 | } |
13b18339 | 801 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
5b1158e9 | 802 | |
9fb9cbb1 YK |
803 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
804 | void __nf_ct_refresh_acct(struct nf_conn *ct, | |
805 | enum ip_conntrack_info ctinfo, | |
806 | const struct sk_buff *skb, | |
807 | unsigned long extra_jiffies, | |
808 | int do_acct) | |
809 | { | |
810 | int event = 0; | |
811 | ||
812 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); | |
813 | NF_CT_ASSERT(skb); | |
814 | ||
f8ba1aff | 815 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 816 | |
997ae831 | 817 | /* Only update if this is not a fixed timeout */ |
47d95045 PM |
818 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
819 | goto acct; | |
997ae831 | 820 | |
9fb9cbb1 YK |
821 | /* If not in hash table, timer will not be active yet */ |
822 | if (!nf_ct_is_confirmed(ct)) { | |
823 | ct->timeout.expires = extra_jiffies; | |
824 | event = IPCT_REFRESH; | |
825 | } else { | |
be00c8e4 MJ |
826 | unsigned long newtime = jiffies + extra_jiffies; |
827 | ||
828 | /* Only update the timeout if the new timeout is at least | |
829 | HZ jiffies from the old timeout. Need del_timer for race | |
830 | avoidance (may already be dying). */ | |
831 | if (newtime - ct->timeout.expires >= HZ | |
832 | && del_timer(&ct->timeout)) { | |
833 | ct->timeout.expires = newtime; | |
9fb9cbb1 YK |
834 | add_timer(&ct->timeout); |
835 | event = IPCT_REFRESH; | |
836 | } | |
837 | } | |
838 | ||
47d95045 | 839 | acct: |
9fb9cbb1 | 840 | if (do_acct) { |
58401572 | 841 | struct nf_conn_counter *acct; |
3ffd5eeb | 842 | |
58401572 KPO |
843 | acct = nf_conn_acct_find(ct); |
844 | if (acct) { | |
845 | acct[CTINFO2DIR(ctinfo)].packets++; | |
846 | acct[CTINFO2DIR(ctinfo)].bytes += | |
847 | skb->len - skb_network_offset(skb); | |
848 | } | |
9fb9cbb1 | 849 | } |
9fb9cbb1 | 850 | |
f8ba1aff | 851 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
852 | |
853 | /* must be unlocked when calling event cache */ | |
854 | if (event) | |
a71996fc | 855 | nf_conntrack_event_cache(event, ct); |
9fb9cbb1 | 856 | } |
13b18339 | 857 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
9fb9cbb1 | 858 | |
4c889498 DM |
859 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
860 | enum ip_conntrack_info ctinfo, | |
861 | const struct sk_buff *skb, | |
862 | int do_acct) | |
51091764 | 863 | { |
718d4ad9 | 864 | if (do_acct) { |
58401572 KPO |
865 | struct nf_conn_counter *acct; |
866 | ||
718d4ad9 | 867 | spin_lock_bh(&nf_conntrack_lock); |
58401572 KPO |
868 | acct = nf_conn_acct_find(ct); |
869 | if (acct) { | |
870 | acct[CTINFO2DIR(ctinfo)].packets++; | |
871 | acct[CTINFO2DIR(ctinfo)].bytes += | |
872 | skb->len - skb_network_offset(skb); | |
873 | } | |
718d4ad9 FH |
874 | spin_unlock_bh(&nf_conntrack_lock); |
875 | } | |
58401572 | 876 | |
4c889498 | 877 | if (del_timer(&ct->timeout)) { |
51091764 | 878 | ct->timeout.function((unsigned long)ct); |
4c889498 DM |
879 | return true; |
880 | } | |
881 | return false; | |
51091764 | 882 | } |
718d4ad9 | 883 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
51091764 | 884 | |
e281db5c | 885 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
c1d10adb PNA |
886 | |
887 | #include <linux/netfilter/nfnetlink.h> | |
888 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
57b47a53 IM |
889 | #include <linux/mutex.h> |
890 | ||
c1d10adb PNA |
891 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
892 | * in ip_conntrack_core, since we don't want the protocols to autoload | |
893 | * or depend on ctnetlink */ | |
fdf70832 | 894 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
c1d10adb PNA |
895 | const struct nf_conntrack_tuple *tuple) |
896 | { | |
77236b6e PM |
897 | NLA_PUT_BE16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port); |
898 | NLA_PUT_BE16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port); | |
c1d10adb PNA |
899 | return 0; |
900 | ||
df6fb868 | 901 | nla_put_failure: |
c1d10adb PNA |
902 | return -1; |
903 | } | |
fdf70832 | 904 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
c1d10adb | 905 | |
f73e924c PM |
906 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
907 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | |
908 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | |
c1d10adb | 909 | }; |
f73e924c | 910 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
c1d10adb | 911 | |
fdf70832 | 912 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
c1d10adb PNA |
913 | struct nf_conntrack_tuple *t) |
914 | { | |
df6fb868 | 915 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
c1d10adb PNA |
916 | return -EINVAL; |
917 | ||
77236b6e PM |
918 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
919 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | |
c1d10adb PNA |
920 | |
921 | return 0; | |
922 | } | |
fdf70832 | 923 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
5c0de29d HE |
924 | |
925 | int nf_ct_port_nlattr_tuple_size(void) | |
926 | { | |
927 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
928 | } | |
929 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | |
c1d10adb PNA |
930 | #endif |
931 | ||
9fb9cbb1 | 932 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
b334aadc | 933 | static void nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb) |
9fb9cbb1 YK |
934 | { |
935 | struct nf_conn *ct; | |
936 | enum ip_conntrack_info ctinfo; | |
937 | ||
938 | /* This ICMP is in reverse direction to the packet which caused it */ | |
939 | ct = nf_ct_get(skb, &ctinfo); | |
940 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | |
941 | ctinfo = IP_CT_RELATED + IP_CT_IS_REPLY; | |
942 | else | |
943 | ctinfo = IP_CT_RELATED; | |
944 | ||
945 | /* Attach to new skbuff, and increment count */ | |
946 | nskb->nfct = &ct->ct_general; | |
947 | nskb->nfctinfo = ctinfo; | |
948 | nf_conntrack_get(nskb->nfct); | |
949 | } | |
950 | ||
9fb9cbb1 | 951 | /* Bring out ya dead! */ |
df0933dc | 952 | static struct nf_conn * |
400dad39 | 953 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
9fb9cbb1 YK |
954 | void *data, unsigned int *bucket) |
955 | { | |
df0933dc PM |
956 | struct nf_conntrack_tuple_hash *h; |
957 | struct nf_conn *ct; | |
ea781f19 | 958 | struct hlist_nulls_node *n; |
9fb9cbb1 | 959 | |
f8ba1aff | 960 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 961 | for (; *bucket < nf_conntrack_htable_size; (*bucket)++) { |
ea781f19 | 962 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { |
df0933dc PM |
963 | ct = nf_ct_tuplehash_to_ctrack(h); |
964 | if (iter(ct, data)) | |
965 | goto found; | |
966 | } | |
601e68e1 | 967 | } |
ea781f19 | 968 | hlist_nulls_for_each_entry(h, n, &net->ct.unconfirmed, hnnode) { |
df0933dc PM |
969 | ct = nf_ct_tuplehash_to_ctrack(h); |
970 | if (iter(ct, data)) | |
ec68e97d | 971 | set_bit(IPS_DYING_BIT, &ct->status); |
df0933dc | 972 | } |
f8ba1aff | 973 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc PM |
974 | return NULL; |
975 | found: | |
c073e3fa | 976 | atomic_inc(&ct->ct_general.use); |
f8ba1aff | 977 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc | 978 | return ct; |
9fb9cbb1 YK |
979 | } |
980 | ||
400dad39 AD |
981 | void nf_ct_iterate_cleanup(struct net *net, |
982 | int (*iter)(struct nf_conn *i, void *data), | |
983 | void *data) | |
9fb9cbb1 | 984 | { |
df0933dc | 985 | struct nf_conn *ct; |
9fb9cbb1 YK |
986 | unsigned int bucket = 0; |
987 | ||
400dad39 | 988 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
9fb9cbb1 YK |
989 | /* Time to push up daises... */ |
990 | if (del_timer(&ct->timeout)) | |
991 | death_by_timeout((unsigned long)ct); | |
992 | /* ... else the timer will get him soon. */ | |
993 | ||
994 | nf_ct_put(ct); | |
995 | } | |
996 | } | |
13b18339 | 997 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
9fb9cbb1 | 998 | |
19abb7b0 PNA |
999 | struct __nf_ct_flush_report { |
1000 | u32 pid; | |
1001 | int report; | |
1002 | }; | |
1003 | ||
274d383b | 1004 | static int kill_report(struct nf_conn *i, void *data) |
9fb9cbb1 | 1005 | { |
19abb7b0 PNA |
1006 | struct __nf_ct_flush_report *fr = (struct __nf_ct_flush_report *)data; |
1007 | ||
1008 | /* get_next_corpse sets the dying bit for us */ | |
1009 | nf_conntrack_event_report(IPCT_DESTROY, | |
1010 | i, | |
1011 | fr->pid, | |
1012 | fr->report); | |
9fb9cbb1 YK |
1013 | return 1; |
1014 | } | |
1015 | ||
274d383b PNA |
1016 | static int kill_all(struct nf_conn *i, void *data) |
1017 | { | |
1018 | return 1; | |
1019 | } | |
1020 | ||
ea781f19 | 1021 | void nf_ct_free_hashtable(void *hash, int vmalloced, unsigned int size) |
9fb9cbb1 YK |
1022 | { |
1023 | if (vmalloced) | |
1024 | vfree(hash); | |
1025 | else | |
601e68e1 | 1026 | free_pages((unsigned long)hash, |
f205c5e0 | 1027 | get_order(sizeof(struct hlist_head) * size)); |
9fb9cbb1 | 1028 | } |
ac565e5f | 1029 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
9fb9cbb1 | 1030 | |
274d383b | 1031 | void nf_conntrack_flush_report(struct net *net, u32 pid, int report) |
c1d10adb | 1032 | { |
19abb7b0 PNA |
1033 | struct __nf_ct_flush_report fr = { |
1034 | .pid = pid, | |
1035 | .report = report, | |
1036 | }; | |
274d383b | 1037 | nf_ct_iterate_cleanup(net, kill_report, &fr); |
c1d10adb | 1038 | } |
274d383b | 1039 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); |
c1d10adb | 1040 | |
08f6547d | 1041 | static void nf_conntrack_cleanup_init_net(void) |
9fb9cbb1 | 1042 | { |
08f6547d AD |
1043 | nf_conntrack_helper_fini(); |
1044 | nf_conntrack_proto_fini(); | |
1045 | kmem_cache_destroy(nf_conntrack_cachep); | |
1046 | } | |
9fb9cbb1 | 1047 | |
08f6547d AD |
1048 | static void nf_conntrack_cleanup_net(struct net *net) |
1049 | { | |
6058fa6b AD |
1050 | nf_ct_event_cache_flush(net); |
1051 | nf_conntrack_ecache_fini(net); | |
9fb9cbb1 | 1052 | i_see_dead_people: |
274d383b | 1053 | nf_ct_iterate_cleanup(net, kill_all, NULL); |
49ac8713 | 1054 | if (atomic_read(&net->ct.count) != 0) { |
9fb9cbb1 YK |
1055 | schedule(); |
1056 | goto i_see_dead_people; | |
1057 | } | |
6636568c PM |
1058 | /* wait until all references to nf_conntrack_untracked are dropped */ |
1059 | while (atomic_read(&nf_conntrack_untracked.ct_general.use) > 1) | |
1060 | schedule(); | |
9fb9cbb1 | 1061 | |
400dad39 | 1062 | nf_ct_free_hashtable(net->ct.hash, net->ct.hash_vmalloc, |
ac565e5f | 1063 | nf_conntrack_htable_size); |
d716a4df | 1064 | nf_conntrack_acct_fini(net); |
9b03f38d | 1065 | nf_conntrack_expect_fini(net); |
0d55af87 | 1066 | free_percpu(net->ct.stat); |
08f6547d AD |
1067 | } |
1068 | ||
1069 | /* Mishearing the voices in his head, our hero wonders how he's | |
1070 | supposed to kill the mall. */ | |
1071 | void nf_conntrack_cleanup(struct net *net) | |
1072 | { | |
1073 | if (net_eq(net, &init_net)) | |
1074 | rcu_assign_pointer(ip_ct_attach, NULL); | |
1075 | ||
1076 | /* This makes sure all current packets have passed through | |
1077 | netfilter framework. Roll on, two-stage module | |
1078 | delete... */ | |
1079 | synchronize_net(); | |
1080 | ||
1081 | nf_conntrack_cleanup_net(net); | |
1082 | ||
1083 | if (net_eq(net, &init_net)) { | |
1084 | rcu_assign_pointer(nf_ct_destroy, NULL); | |
1085 | nf_conntrack_cleanup_init_net(); | |
1086 | } | |
9fb9cbb1 YK |
1087 | } |
1088 | ||
ea781f19 | 1089 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int *vmalloced, int nulls) |
9fb9cbb1 | 1090 | { |
ea781f19 ED |
1091 | struct hlist_nulls_head *hash; |
1092 | unsigned int nr_slots, i; | |
1093 | size_t sz; | |
9fb9cbb1 | 1094 | |
601e68e1 | 1095 | *vmalloced = 0; |
8e5105a0 | 1096 | |
ea781f19 ED |
1097 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1098 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | |
1099 | sz = nr_slots * sizeof(struct hlist_nulls_head); | |
1100 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | |
1101 | get_order(sz)); | |
601e68e1 | 1102 | if (!hash) { |
9fb9cbb1 YK |
1103 | *vmalloced = 1; |
1104 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); | |
ea781f19 | 1105 | hash = __vmalloc(sz, GFP_KERNEL | __GFP_ZERO, PAGE_KERNEL); |
9fb9cbb1 YK |
1106 | } |
1107 | ||
ea781f19 ED |
1108 | if (hash && nulls) |
1109 | for (i = 0; i < nr_slots; i++) | |
1110 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | |
9fb9cbb1 YK |
1111 | |
1112 | return hash; | |
1113 | } | |
ac565e5f | 1114 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
9fb9cbb1 | 1115 | |
fae718dd | 1116 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
9fb9cbb1 | 1117 | { |
96eb24d7 SH |
1118 | int i, bucket, vmalloced, old_vmalloced; |
1119 | unsigned int hashsize, old_size; | |
9fb9cbb1 | 1120 | int rnd; |
ea781f19 | 1121 | struct hlist_nulls_head *hash, *old_hash; |
9fb9cbb1 YK |
1122 | struct nf_conntrack_tuple_hash *h; |
1123 | ||
1124 | /* On boot, we can set this without any fancy locking. */ | |
1125 | if (!nf_conntrack_htable_size) | |
1126 | return param_set_uint(val, kp); | |
1127 | ||
96eb24d7 | 1128 | hashsize = simple_strtoul(val, NULL, 0); |
9fb9cbb1 YK |
1129 | if (!hashsize) |
1130 | return -EINVAL; | |
1131 | ||
ea781f19 | 1132 | hash = nf_ct_alloc_hashtable(&hashsize, &vmalloced, 1); |
9fb9cbb1 YK |
1133 | if (!hash) |
1134 | return -ENOMEM; | |
1135 | ||
1136 | /* We have to rehahs for the new table anyway, so we also can | |
1137 | * use a newrandom seed */ | |
af07d241 | 1138 | get_random_bytes(&rnd, sizeof(rnd)); |
9fb9cbb1 | 1139 | |
76507f69 PM |
1140 | /* Lookups in the old hash might happen in parallel, which means we |
1141 | * might get false negatives during connection lookup. New connections | |
1142 | * created because of a false negative won't make it into the hash | |
1143 | * though since that required taking the lock. | |
1144 | */ | |
f8ba1aff | 1145 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1146 | for (i = 0; i < nf_conntrack_htable_size; i++) { |
ea781f19 ED |
1147 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
1148 | h = hlist_nulls_entry(init_net.ct.hash[i].first, | |
1149 | struct nf_conntrack_tuple_hash, hnnode); | |
1150 | hlist_nulls_del_rcu(&h->hnnode); | |
9fb9cbb1 | 1151 | bucket = __hash_conntrack(&h->tuple, hashsize, rnd); |
ea781f19 | 1152 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
9fb9cbb1 YK |
1153 | } |
1154 | } | |
1155 | old_size = nf_conntrack_htable_size; | |
400dad39 AD |
1156 | old_vmalloced = init_net.ct.hash_vmalloc; |
1157 | old_hash = init_net.ct.hash; | |
9fb9cbb1 YK |
1158 | |
1159 | nf_conntrack_htable_size = hashsize; | |
400dad39 AD |
1160 | init_net.ct.hash_vmalloc = vmalloced; |
1161 | init_net.ct.hash = hash; | |
9fb9cbb1 | 1162 | nf_conntrack_hash_rnd = rnd; |
f8ba1aff | 1163 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1164 | |
ac565e5f | 1165 | nf_ct_free_hashtable(old_hash, old_vmalloced, old_size); |
9fb9cbb1 YK |
1166 | return 0; |
1167 | } | |
fae718dd | 1168 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
9fb9cbb1 | 1169 | |
fae718dd | 1170 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
9fb9cbb1 YK |
1171 | &nf_conntrack_htable_size, 0600); |
1172 | ||
08f6547d | 1173 | static int nf_conntrack_init_init_net(void) |
9fb9cbb1 | 1174 | { |
f205c5e0 | 1175 | int max_factor = 8; |
9fb9cbb1 YK |
1176 | int ret; |
1177 | ||
1178 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB | |
f205c5e0 | 1179 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ |
9fb9cbb1 YK |
1180 | if (!nf_conntrack_htable_size) { |
1181 | nf_conntrack_htable_size | |
1182 | = (((num_physpages << PAGE_SHIFT) / 16384) | |
f205c5e0 | 1183 | / sizeof(struct hlist_head)); |
9fb9cbb1 | 1184 | if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
f205c5e0 PM |
1185 | nf_conntrack_htable_size = 16384; |
1186 | if (nf_conntrack_htable_size < 32) | |
1187 | nf_conntrack_htable_size = 32; | |
1188 | ||
1189 | /* Use a max. factor of four by default to get the same max as | |
1190 | * with the old struct list_heads. When a table size is given | |
1191 | * we use the old value of 8 to avoid reducing the max. | |
1192 | * entries. */ | |
1193 | max_factor = 4; | |
9fb9cbb1 | 1194 | } |
f205c5e0 | 1195 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
8e5105a0 PM |
1196 | |
1197 | printk("nf_conntrack version %s (%u buckets, %d max)\n", | |
1198 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, | |
1199 | nf_conntrack_max); | |
1200 | ||
dacd2a1a YK |
1201 | nf_conntrack_cachep = kmem_cache_create("nf_conntrack", |
1202 | sizeof(struct nf_conn), | |
ea781f19 | 1203 | 0, SLAB_DESTROY_BY_RCU, NULL); |
dacd2a1a | 1204 | if (!nf_conntrack_cachep) { |
9fb9cbb1 | 1205 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); |
08f6547d AD |
1206 | ret = -ENOMEM; |
1207 | goto err_cache; | |
9fb9cbb1 YK |
1208 | } |
1209 | ||
e9c1b084 PM |
1210 | ret = nf_conntrack_proto_init(); |
1211 | if (ret < 0) | |
08f6547d | 1212 | goto err_proto; |
933a41e7 | 1213 | |
ceceae1b YK |
1214 | ret = nf_conntrack_helper_init(); |
1215 | if (ret < 0) | |
08f6547d AD |
1216 | goto err_helper; |
1217 | ||
1218 | return 0; | |
1219 | ||
1220 | err_helper: | |
1221 | nf_conntrack_proto_fini(); | |
1222 | err_proto: | |
1223 | kmem_cache_destroy(nf_conntrack_cachep); | |
1224 | err_cache: | |
1225 | return ret; | |
1226 | } | |
1227 | ||
1228 | static int nf_conntrack_init_net(struct net *net) | |
1229 | { | |
1230 | int ret; | |
ceceae1b | 1231 | |
08f6547d | 1232 | atomic_set(&net->ct.count, 0); |
ea781f19 | 1233 | INIT_HLIST_NULLS_HEAD(&net->ct.unconfirmed, 0); |
08f6547d AD |
1234 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1235 | if (!net->ct.stat) { | |
1236 | ret = -ENOMEM; | |
1237 | goto err_stat; | |
1238 | } | |
1239 | ret = nf_conntrack_ecache_init(net); | |
1240 | if (ret < 0) | |
1241 | goto err_ecache; | |
1242 | net->ct.hash = nf_ct_alloc_hashtable(&nf_conntrack_htable_size, | |
ea781f19 | 1243 | &net->ct.hash_vmalloc, 1); |
08f6547d AD |
1244 | if (!net->ct.hash) { |
1245 | ret = -ENOMEM; | |
1246 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); | |
1247 | goto err_hash; | |
1248 | } | |
1249 | ret = nf_conntrack_expect_init(net); | |
1250 | if (ret < 0) | |
1251 | goto err_expect; | |
d716a4df | 1252 | ret = nf_conntrack_acct_init(net); |
58401572 | 1253 | if (ret < 0) |
08f6547d | 1254 | goto err_acct; |
7d3cdc6b | 1255 | |
9fb9cbb1 YK |
1256 | /* Set up fake conntrack: |
1257 | - to never be deleted, not in any hashes */ | |
5a1fb391 AD |
1258 | #ifdef CONFIG_NET_NS |
1259 | nf_conntrack_untracked.ct_net = &init_net; | |
1260 | #endif | |
9fb9cbb1 YK |
1261 | atomic_set(&nf_conntrack_untracked.ct_general.use, 1); |
1262 | /* - and look it like as a confirmed connection */ | |
1263 | set_bit(IPS_CONFIRMED_BIT, &nf_conntrack_untracked.status); | |
1264 | ||
08f6547d | 1265 | return 0; |
9fb9cbb1 | 1266 | |
08f6547d | 1267 | err_acct: |
9b03f38d | 1268 | nf_conntrack_expect_fini(net); |
08f6547d | 1269 | err_expect: |
400dad39 | 1270 | nf_ct_free_hashtable(net->ct.hash, net->ct.hash_vmalloc, |
ac565e5f | 1271 | nf_conntrack_htable_size); |
6058fa6b AD |
1272 | err_hash: |
1273 | nf_conntrack_ecache_fini(net); | |
1274 | err_ecache: | |
0d55af87 AD |
1275 | free_percpu(net->ct.stat); |
1276 | err_stat: | |
08f6547d AD |
1277 | return ret; |
1278 | } | |
1279 | ||
1280 | int nf_conntrack_init(struct net *net) | |
1281 | { | |
1282 | int ret; | |
1283 | ||
1284 | if (net_eq(net, &init_net)) { | |
1285 | ret = nf_conntrack_init_init_net(); | |
1286 | if (ret < 0) | |
1287 | goto out_init_net; | |
1288 | } | |
1289 | ret = nf_conntrack_init_net(net); | |
1290 | if (ret < 0) | |
1291 | goto out_net; | |
1292 | ||
1293 | if (net_eq(net, &init_net)) { | |
1294 | /* For use by REJECT target */ | |
1295 | rcu_assign_pointer(ip_ct_attach, nf_conntrack_attach); | |
1296 | rcu_assign_pointer(nf_ct_destroy, destroy_conntrack); | |
1297 | } | |
1298 | return 0; | |
1299 | ||
1300 | out_net: | |
1301 | if (net_eq(net, &init_net)) | |
1302 | nf_conntrack_cleanup_init_net(); | |
1303 | out_init_net: | |
1304 | return ret; | |
9fb9cbb1 | 1305 | } |