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> |
a0891aa6 | 42 | #include <net/netfilter/nf_conntrack_ecache.h> |
e6a7d3c0 | 43 | #include <net/netfilter/nf_nat.h> |
e17b666a | 44 | #include <net/netfilter/nf_nat_core.h> |
9fb9cbb1 | 45 | |
dc808fe2 | 46 | #define NF_CONNTRACK_VERSION "0.5.0" |
9fb9cbb1 | 47 | |
e17b666a PM |
48 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
49 | enum nf_nat_manip_type manip, | |
50 | struct nlattr *attr) __read_mostly; | |
e6a7d3c0 PNA |
51 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
52 | ||
f8ba1aff | 53 | DEFINE_SPINLOCK(nf_conntrack_lock); |
13b18339 | 54 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); |
9fb9cbb1 | 55 | |
e2b7606c | 56 | unsigned int nf_conntrack_htable_size __read_mostly; |
13b18339 PM |
57 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
58 | ||
e478075c | 59 | unsigned int nf_conntrack_max __read_mostly; |
a999e683 | 60 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
13b18339 | 61 | |
e2b7606c | 62 | struct nf_conn nf_conntrack_untracked __read_mostly; |
13b18339 PM |
63 | EXPORT_SYMBOL_GPL(nf_conntrack_untracked); |
64 | ||
dacd2a1a | 65 | static struct kmem_cache *nf_conntrack_cachep __read_mostly; |
77ab9cff | 66 | |
9fb9cbb1 YK |
67 | static int nf_conntrack_hash_rnd_initted; |
68 | static unsigned int nf_conntrack_hash_rnd; | |
69 | ||
70 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, | |
71 | unsigned int size, unsigned int rnd) | |
72 | { | |
0794935e PM |
73 | unsigned int n; |
74 | u_int32_t h; | |
75 | ||
76 | /* The direction must be ignored, so we hash everything up to the | |
77 | * destination ports (which is a multiple of 4) and treat the last | |
78 | * three bytes manually. | |
79 | */ | |
80 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); | |
81 | h = jhash2((u32 *)tuple, n, | |
82 | rnd ^ (((__force __u16)tuple->dst.u.all << 16) | | |
83 | tuple->dst.protonum)); | |
84 | ||
85 | return ((u64)h * size) >> 32; | |
9fb9cbb1 YK |
86 | } |
87 | ||
88 | static inline u_int32_t hash_conntrack(const struct nf_conntrack_tuple *tuple) | |
89 | { | |
90 | return __hash_conntrack(tuple, nf_conntrack_htable_size, | |
91 | nf_conntrack_hash_rnd); | |
92 | } | |
93 | ||
5f2b4c90 | 94 | bool |
9fb9cbb1 YK |
95 | nf_ct_get_tuple(const struct sk_buff *skb, |
96 | unsigned int nhoff, | |
97 | unsigned int dataoff, | |
98 | u_int16_t l3num, | |
99 | u_int8_t protonum, | |
100 | struct nf_conntrack_tuple *tuple, | |
101 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 102 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 103 | { |
443a70d5 | 104 | memset(tuple, 0, sizeof(*tuple)); |
9fb9cbb1 YK |
105 | |
106 | tuple->src.l3num = l3num; | |
107 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | |
5f2b4c90 | 108 | return false; |
9fb9cbb1 YK |
109 | |
110 | tuple->dst.protonum = protonum; | |
111 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | |
112 | ||
605dcad6 | 113 | return l4proto->pkt_to_tuple(skb, dataoff, tuple); |
9fb9cbb1 | 114 | } |
13b18339 | 115 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
9fb9cbb1 | 116 | |
5f2b4c90 JE |
117 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
118 | u_int16_t l3num, struct nf_conntrack_tuple *tuple) | |
e2a3123f YK |
119 | { |
120 | struct nf_conntrack_l3proto *l3proto; | |
121 | struct nf_conntrack_l4proto *l4proto; | |
122 | unsigned int protoff; | |
123 | u_int8_t protonum; | |
124 | int ret; | |
125 | ||
126 | rcu_read_lock(); | |
127 | ||
128 | l3proto = __nf_ct_l3proto_find(l3num); | |
129 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | |
130 | if (ret != NF_ACCEPT) { | |
131 | rcu_read_unlock(); | |
5f2b4c90 | 132 | return false; |
e2a3123f YK |
133 | } |
134 | ||
135 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
136 | ||
137 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, tuple, | |
138 | l3proto, l4proto); | |
139 | ||
140 | rcu_read_unlock(); | |
141 | return ret; | |
142 | } | |
143 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | |
144 | ||
5f2b4c90 | 145 | bool |
9fb9cbb1 YK |
146 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
147 | const struct nf_conntrack_tuple *orig, | |
148 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 149 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 150 | { |
443a70d5 | 151 | memset(inverse, 0, sizeof(*inverse)); |
9fb9cbb1 YK |
152 | |
153 | inverse->src.l3num = orig->src.l3num; | |
154 | if (l3proto->invert_tuple(inverse, orig) == 0) | |
5f2b4c90 | 155 | return false; |
9fb9cbb1 YK |
156 | |
157 | inverse->dst.dir = !orig->dst.dir; | |
158 | ||
159 | inverse->dst.protonum = orig->dst.protonum; | |
605dcad6 | 160 | return l4proto->invert_tuple(inverse, orig); |
9fb9cbb1 | 161 | } |
13b18339 | 162 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
9fb9cbb1 | 163 | |
9fb9cbb1 YK |
164 | static void |
165 | clean_from_lists(struct nf_conn *ct) | |
166 | { | |
0d53778e | 167 | pr_debug("clean_from_lists(%p)\n", ct); |
ea781f19 ED |
168 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
169 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | |
9fb9cbb1 YK |
170 | |
171 | /* Destroy all pending expectations */ | |
c1d10adb | 172 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
173 | } |
174 | ||
175 | static void | |
176 | destroy_conntrack(struct nf_conntrack *nfct) | |
177 | { | |
178 | struct nf_conn *ct = (struct nf_conn *)nfct; | |
0d55af87 | 179 | struct net *net = nf_ct_net(ct); |
605dcad6 | 180 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 | 181 | |
0d53778e | 182 | pr_debug("destroy_conntrack(%p)\n", ct); |
9fb9cbb1 YK |
183 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
184 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); | |
185 | ||
9fb9cbb1 YK |
186 | /* To make sure we don't get any weird locking issues here: |
187 | * destroy_conntrack() MUST NOT be called with a write lock | |
188 | * to nf_conntrack_lock!!! -HW */ | |
923f4902 | 189 | rcu_read_lock(); |
5e8fbe2a | 190 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
605dcad6 MJ |
191 | if (l4proto && l4proto->destroy) |
192 | l4proto->destroy(ct); | |
9fb9cbb1 | 193 | |
982d9a9c | 194 | rcu_read_unlock(); |
9fb9cbb1 | 195 | |
f8ba1aff | 196 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
197 | /* Expectations will have been removed in clean_from_lists, |
198 | * except TFTP can create an expectation on the first packet, | |
199 | * before connection is in the list, so we need to clean here, | |
200 | * too. */ | |
c1d10adb | 201 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
202 | |
203 | /* We overload first tuple to link into unconfirmed list. */ | |
204 | if (!nf_ct_is_confirmed(ct)) { | |
ea781f19 ED |
205 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); |
206 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | |
9fb9cbb1 YK |
207 | } |
208 | ||
0d55af87 | 209 | NF_CT_STAT_INC(net, delete); |
f8ba1aff | 210 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
211 | |
212 | if (ct->master) | |
213 | nf_ct_put(ct->master); | |
214 | ||
0d53778e | 215 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
9fb9cbb1 YK |
216 | nf_conntrack_free(ct); |
217 | } | |
218 | ||
dd7669a9 | 219 | void nf_ct_delete_from_lists(struct nf_conn *ct) |
9fb9cbb1 | 220 | { |
0d55af87 | 221 | struct net *net = nf_ct_net(ct); |
9fb9cbb1 | 222 | |
9858a3ae | 223 | nf_ct_helper_destroy(ct); |
f8ba1aff | 224 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
225 | /* Inside lock so preempt is disabled on module removal path. |
226 | * Otherwise we can get spurious warnings. */ | |
0d55af87 | 227 | NF_CT_STAT_INC(net, delete_list); |
9fb9cbb1 | 228 | clean_from_lists(ct); |
f8ba1aff | 229 | spin_unlock_bh(&nf_conntrack_lock); |
dd7669a9 PNA |
230 | } |
231 | EXPORT_SYMBOL_GPL(nf_ct_delete_from_lists); | |
232 | ||
233 | static void death_by_event(unsigned long ul_conntrack) | |
234 | { | |
235 | struct nf_conn *ct = (void *)ul_conntrack; | |
236 | struct net *net = nf_ct_net(ct); | |
237 | ||
238 | if (nf_conntrack_event(IPCT_DESTROY, ct) < 0) { | |
239 | /* bad luck, let's retry again */ | |
240 | ct->timeout.expires = jiffies + | |
241 | (random32() % net->ct.sysctl_events_retry_timeout); | |
242 | add_timer(&ct->timeout); | |
243 | return; | |
244 | } | |
245 | /* we've got the event delivered, now it's dying */ | |
246 | set_bit(IPS_DYING_BIT, &ct->status); | |
247 | spin_lock(&nf_conntrack_lock); | |
248 | hlist_nulls_del(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | |
249 | spin_unlock(&nf_conntrack_lock); | |
250 | nf_ct_put(ct); | |
251 | } | |
252 | ||
253 | void nf_ct_insert_dying_list(struct nf_conn *ct) | |
254 | { | |
255 | struct net *net = nf_ct_net(ct); | |
256 | ||
257 | /* add this conntrack to the dying list */ | |
258 | spin_lock_bh(&nf_conntrack_lock); | |
259 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
260 | &net->ct.dying); | |
261 | spin_unlock_bh(&nf_conntrack_lock); | |
262 | /* set a new timer to retry event delivery */ | |
263 | setup_timer(&ct->timeout, death_by_event, (unsigned long)ct); | |
264 | ct->timeout.expires = jiffies + | |
265 | (random32() % net->ct.sysctl_events_retry_timeout); | |
266 | add_timer(&ct->timeout); | |
267 | } | |
268 | EXPORT_SYMBOL_GPL(nf_ct_insert_dying_list); | |
269 | ||
270 | static void death_by_timeout(unsigned long ul_conntrack) | |
271 | { | |
272 | struct nf_conn *ct = (void *)ul_conntrack; | |
273 | ||
274 | if (!test_bit(IPS_DYING_BIT, &ct->status) && | |
275 | unlikely(nf_conntrack_event(IPCT_DESTROY, ct) < 0)) { | |
276 | /* destroy event was not delivered */ | |
277 | nf_ct_delete_from_lists(ct); | |
278 | nf_ct_insert_dying_list(ct); | |
279 | return; | |
280 | } | |
281 | set_bit(IPS_DYING_BIT, &ct->status); | |
282 | nf_ct_delete_from_lists(ct); | |
9fb9cbb1 YK |
283 | nf_ct_put(ct); |
284 | } | |
285 | ||
ea781f19 ED |
286 | /* |
287 | * Warning : | |
288 | * - Caller must take a reference on returned object | |
289 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | |
290 | * OR | |
291 | * - Caller must lock nf_conntrack_lock before calling this function | |
292 | */ | |
c1d10adb | 293 | struct nf_conntrack_tuple_hash * |
400dad39 | 294 | __nf_conntrack_find(struct net *net, const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 YK |
295 | { |
296 | struct nf_conntrack_tuple_hash *h; | |
ea781f19 | 297 | struct hlist_nulls_node *n; |
9fb9cbb1 YK |
298 | unsigned int hash = hash_conntrack(tuple); |
299 | ||
4e29e9ec PM |
300 | /* Disable BHs the entire time since we normally need to disable them |
301 | * at least once for the stats anyway. | |
302 | */ | |
303 | local_bh_disable(); | |
ea781f19 ED |
304 | begin: |
305 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { | |
ba419aff | 306 | if (nf_ct_tuple_equal(tuple, &h->tuple)) { |
0d55af87 | 307 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 308 | local_bh_enable(); |
9fb9cbb1 YK |
309 | return h; |
310 | } | |
0d55af87 | 311 | NF_CT_STAT_INC(net, searched); |
9fb9cbb1 | 312 | } |
ea781f19 ED |
313 | /* |
314 | * if the nulls value we got at the end of this lookup is | |
315 | * not the expected one, we must restart lookup. | |
316 | * We probably met an item that was moved to another chain. | |
317 | */ | |
318 | if (get_nulls_value(n) != hash) | |
319 | goto begin; | |
4e29e9ec | 320 | local_bh_enable(); |
9fb9cbb1 YK |
321 | |
322 | return NULL; | |
323 | } | |
13b18339 | 324 | EXPORT_SYMBOL_GPL(__nf_conntrack_find); |
9fb9cbb1 YK |
325 | |
326 | /* Find a connection corresponding to a tuple. */ | |
327 | struct nf_conntrack_tuple_hash * | |
400dad39 | 328 | nf_conntrack_find_get(struct net *net, const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 YK |
329 | { |
330 | struct nf_conntrack_tuple_hash *h; | |
76507f69 | 331 | struct nf_conn *ct; |
9fb9cbb1 | 332 | |
76507f69 | 333 | rcu_read_lock(); |
ea781f19 | 334 | begin: |
400dad39 | 335 | h = __nf_conntrack_find(net, tuple); |
76507f69 PM |
336 | if (h) { |
337 | ct = nf_ct_tuplehash_to_ctrack(h); | |
8d8890b7 PM |
338 | if (unlikely(nf_ct_is_dying(ct) || |
339 | !atomic_inc_not_zero(&ct->ct_general.use))) | |
76507f69 | 340 | h = NULL; |
ea781f19 ED |
341 | else { |
342 | if (unlikely(!nf_ct_tuple_equal(tuple, &h->tuple))) { | |
343 | nf_ct_put(ct); | |
344 | goto begin; | |
345 | } | |
346 | } | |
76507f69 PM |
347 | } |
348 | rcu_read_unlock(); | |
9fb9cbb1 YK |
349 | |
350 | return h; | |
351 | } | |
13b18339 | 352 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
9fb9cbb1 | 353 | |
c1d10adb PNA |
354 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
355 | unsigned int hash, | |
601e68e1 | 356 | unsigned int repl_hash) |
c1d10adb | 357 | { |
400dad39 AD |
358 | struct net *net = nf_ct_net(ct); |
359 | ||
ea781f19 | 360 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
400dad39 | 361 | &net->ct.hash[hash]); |
ea781f19 | 362 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
400dad39 | 363 | &net->ct.hash[repl_hash]); |
c1d10adb PNA |
364 | } |
365 | ||
366 | void nf_conntrack_hash_insert(struct nf_conn *ct) | |
367 | { | |
368 | unsigned int hash, repl_hash; | |
369 | ||
370 | hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
371 | repl_hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
372 | ||
c1d10adb | 373 | __nf_conntrack_hash_insert(ct, hash, repl_hash); |
c1d10adb | 374 | } |
13b18339 | 375 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_insert); |
c1d10adb | 376 | |
9fb9cbb1 YK |
377 | /* Confirm a connection given skb; places it in hash table */ |
378 | int | |
3db05fea | 379 | __nf_conntrack_confirm(struct sk_buff *skb) |
9fb9cbb1 YK |
380 | { |
381 | unsigned int hash, repl_hash; | |
df0933dc | 382 | struct nf_conntrack_tuple_hash *h; |
9fb9cbb1 | 383 | struct nf_conn *ct; |
df0933dc | 384 | struct nf_conn_help *help; |
ea781f19 | 385 | struct hlist_nulls_node *n; |
9fb9cbb1 | 386 | enum ip_conntrack_info ctinfo; |
400dad39 | 387 | struct net *net; |
9fb9cbb1 | 388 | |
3db05fea | 389 | ct = nf_ct_get(skb, &ctinfo); |
400dad39 | 390 | net = nf_ct_net(ct); |
9fb9cbb1 YK |
391 | |
392 | /* ipt_REJECT uses nf_conntrack_attach to attach related | |
393 | ICMP/TCP RST packets in other direction. Actual packet | |
394 | which created connection will be IP_CT_NEW or for an | |
395 | expected connection, IP_CT_RELATED. */ | |
396 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | |
397 | return NF_ACCEPT; | |
398 | ||
399 | hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
400 | repl_hash = hash_conntrack(&ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
401 | ||
402 | /* We're not in hash table, and we refuse to set up related | |
403 | connections for unconfirmed conns. But packet copies and | |
404 | REJECT will give spurious warnings here. */ | |
405 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ | |
406 | ||
407 | /* No external references means noone else could have | |
408 | confirmed us. */ | |
409 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
0d53778e | 410 | pr_debug("Confirming conntrack %p\n", ct); |
9fb9cbb1 | 411 | |
f8ba1aff | 412 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
413 | |
414 | /* See if there's one in the list already, including reverse: | |
415 | NAT could have grabbed it without realizing, since we're | |
416 | not in the hash. If there is, we lost race. */ | |
ea781f19 | 417 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
df0933dc PM |
418 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
419 | &h->tuple)) | |
420 | goto out; | |
ea781f19 | 421 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) |
df0933dc PM |
422 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
423 | &h->tuple)) | |
424 | goto out; | |
9fb9cbb1 | 425 | |
df0933dc | 426 | /* Remove from unconfirmed list */ |
ea781f19 | 427 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
df0933dc | 428 | |
df0933dc PM |
429 | /* Timer relative to confirmation time, not original |
430 | setting time, otherwise we'd get timer wrap in | |
431 | weird delay cases. */ | |
432 | ct->timeout.expires += jiffies; | |
433 | add_timer(&ct->timeout); | |
434 | atomic_inc(&ct->ct_general.use); | |
435 | set_bit(IPS_CONFIRMED_BIT, &ct->status); | |
5c8ec910 PM |
436 | |
437 | /* Since the lookup is lockless, hash insertion must be done after | |
438 | * starting the timer and setting the CONFIRMED bit. The RCU barriers | |
439 | * guarantee that no other CPU can find the conntrack before the above | |
440 | * stores are visible. | |
441 | */ | |
442 | __nf_conntrack_hash_insert(ct, hash, repl_hash); | |
0d55af87 | 443 | NF_CT_STAT_INC(net, insert); |
f8ba1aff | 444 | spin_unlock_bh(&nf_conntrack_lock); |
5c8ec910 | 445 | |
df0933dc PM |
446 | help = nfct_help(ct); |
447 | if (help && help->helper) | |
a71996fc | 448 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
17e6e4ea | 449 | |
df0933dc | 450 | nf_conntrack_event_cache(master_ct(ct) ? |
a71996fc | 451 | IPCT_RELATED : IPCT_NEW, ct); |
df0933dc | 452 | return NF_ACCEPT; |
9fb9cbb1 | 453 | |
df0933dc | 454 | out: |
0d55af87 | 455 | NF_CT_STAT_INC(net, insert_failed); |
f8ba1aff | 456 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
457 | return NF_DROP; |
458 | } | |
13b18339 | 459 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
9fb9cbb1 YK |
460 | |
461 | /* Returns true if a connection correspondings to the tuple (required | |
462 | for NAT). */ | |
463 | int | |
464 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | |
465 | const struct nf_conn *ignored_conntrack) | |
466 | { | |
400dad39 | 467 | struct net *net = nf_ct_net(ignored_conntrack); |
9fb9cbb1 | 468 | struct nf_conntrack_tuple_hash *h; |
ea781f19 | 469 | struct hlist_nulls_node *n; |
ba419aff | 470 | unsigned int hash = hash_conntrack(tuple); |
9fb9cbb1 | 471 | |
4e29e9ec PM |
472 | /* Disable BHs the entire time since we need to disable them at |
473 | * least once for the stats anyway. | |
474 | */ | |
475 | rcu_read_lock_bh(); | |
ea781f19 | 476 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
ba419aff PM |
477 | if (nf_ct_tuplehash_to_ctrack(h) != ignored_conntrack && |
478 | nf_ct_tuple_equal(tuple, &h->tuple)) { | |
0d55af87 | 479 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 480 | rcu_read_unlock_bh(); |
ba419aff PM |
481 | return 1; |
482 | } | |
0d55af87 | 483 | NF_CT_STAT_INC(net, searched); |
ba419aff | 484 | } |
4e29e9ec | 485 | rcu_read_unlock_bh(); |
9fb9cbb1 | 486 | |
ba419aff | 487 | return 0; |
9fb9cbb1 | 488 | } |
13b18339 | 489 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
9fb9cbb1 | 490 | |
7ae7730f PM |
491 | #define NF_CT_EVICTION_RANGE 8 |
492 | ||
9fb9cbb1 YK |
493 | /* There's a small race here where we may free a just-assured |
494 | connection. Too bad: we're in trouble anyway. */ | |
400dad39 | 495 | static noinline int early_drop(struct net *net, unsigned int hash) |
9fb9cbb1 | 496 | { |
f205c5e0 | 497 | /* Use oldest entry, which is roughly LRU */ |
9fb9cbb1 | 498 | struct nf_conntrack_tuple_hash *h; |
df0933dc | 499 | struct nf_conn *ct = NULL, *tmp; |
ea781f19 | 500 | struct hlist_nulls_node *n; |
7ae7730f | 501 | unsigned int i, cnt = 0; |
9fb9cbb1 YK |
502 | int dropped = 0; |
503 | ||
76507f69 | 504 | rcu_read_lock(); |
7ae7730f | 505 | for (i = 0; i < nf_conntrack_htable_size; i++) { |
ea781f19 ED |
506 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
507 | hnnode) { | |
7ae7730f PM |
508 | tmp = nf_ct_tuplehash_to_ctrack(h); |
509 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) | |
510 | ct = tmp; | |
511 | cnt++; | |
512 | } | |
76507f69 | 513 | |
8d8890b7 PM |
514 | if (ct && unlikely(nf_ct_is_dying(ct) || |
515 | !atomic_inc_not_zero(&ct->ct_general.use))) | |
76507f69 | 516 | ct = NULL; |
7ae7730f PM |
517 | if (ct || cnt >= NF_CT_EVICTION_RANGE) |
518 | break; | |
519 | hash = (hash + 1) % nf_conntrack_htable_size; | |
9fb9cbb1 | 520 | } |
76507f69 | 521 | rcu_read_unlock(); |
9fb9cbb1 YK |
522 | |
523 | if (!ct) | |
524 | return dropped; | |
525 | ||
526 | if (del_timer(&ct->timeout)) { | |
527 | death_by_timeout((unsigned long)ct); | |
528 | dropped = 1; | |
0d55af87 | 529 | NF_CT_STAT_INC_ATOMIC(net, early_drop); |
9fb9cbb1 YK |
530 | } |
531 | nf_ct_put(ct); | |
532 | return dropped; | |
533 | } | |
534 | ||
5a1fb391 AD |
535 | struct nf_conn *nf_conntrack_alloc(struct net *net, |
536 | const struct nf_conntrack_tuple *orig, | |
b891c5a8 PNA |
537 | const struct nf_conntrack_tuple *repl, |
538 | gfp_t gfp) | |
9fb9cbb1 | 539 | { |
cd7fcbf1 | 540 | struct nf_conn *ct; |
9fb9cbb1 | 541 | |
dc808fe2 | 542 | if (unlikely(!nf_conntrack_hash_rnd_initted)) { |
af07d241 HPP |
543 | get_random_bytes(&nf_conntrack_hash_rnd, |
544 | sizeof(nf_conntrack_hash_rnd)); | |
9fb9cbb1 YK |
545 | nf_conntrack_hash_rnd_initted = 1; |
546 | } | |
547 | ||
5251e2d2 | 548 | /* We don't want any race condition at early drop stage */ |
49ac8713 | 549 | atomic_inc(&net->ct.count); |
5251e2d2 | 550 | |
76eb9460 | 551 | if (nf_conntrack_max && |
49ac8713 | 552 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
9fb9cbb1 | 553 | unsigned int hash = hash_conntrack(orig); |
400dad39 | 554 | if (!early_drop(net, hash)) { |
49ac8713 | 555 | atomic_dec(&net->ct.count); |
9fb9cbb1 YK |
556 | if (net_ratelimit()) |
557 | printk(KERN_WARNING | |
558 | "nf_conntrack: table full, dropping" | |
559 | " packet.\n"); | |
560 | return ERR_PTR(-ENOMEM); | |
561 | } | |
562 | } | |
563 | ||
b891c5a8 | 564 | ct = kmem_cache_zalloc(nf_conntrack_cachep, gfp); |
c88130bc | 565 | if (ct == NULL) { |
0d53778e | 566 | pr_debug("nf_conntrack_alloc: Can't alloc conntrack.\n"); |
49ac8713 | 567 | atomic_dec(&net->ct.count); |
dacd2a1a | 568 | return ERR_PTR(-ENOMEM); |
9fb9cbb1 YK |
569 | } |
570 | ||
440f0d58 | 571 | spin_lock_init(&ct->lock); |
c88130bc PM |
572 | atomic_set(&ct->ct_general.use, 1); |
573 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; | |
574 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; | |
9fb9cbb1 | 575 | /* Don't set timer yet: wait for confirmation */ |
c88130bc | 576 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
5a1fb391 AD |
577 | #ifdef CONFIG_NET_NS |
578 | ct->ct_net = net; | |
579 | #endif | |
9fb9cbb1 | 580 | |
c88130bc | 581 | return ct; |
9fb9cbb1 | 582 | } |
13b18339 | 583 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
9fb9cbb1 | 584 | |
c88130bc | 585 | void nf_conntrack_free(struct nf_conn *ct) |
76507f69 | 586 | { |
1d45209d ED |
587 | struct net *net = nf_ct_net(ct); |
588 | ||
ceeff754 | 589 | nf_ct_ext_destroy(ct); |
1d45209d | 590 | atomic_dec(&net->ct.count); |
ea781f19 ED |
591 | nf_ct_ext_free(ct); |
592 | kmem_cache_free(nf_conntrack_cachep, ct); | |
76507f69 | 593 | } |
13b18339 | 594 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
9fb9cbb1 YK |
595 | |
596 | /* Allocate a new conntrack: we return -ENOMEM if classification | |
597 | failed due to stress. Otherwise it really is unclassifiable. */ | |
598 | static struct nf_conntrack_tuple_hash * | |
5a1fb391 AD |
599 | init_conntrack(struct net *net, |
600 | const struct nf_conntrack_tuple *tuple, | |
9fb9cbb1 | 601 | struct nf_conntrack_l3proto *l3proto, |
605dcad6 | 602 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 YK |
603 | struct sk_buff *skb, |
604 | unsigned int dataoff) | |
605 | { | |
c88130bc | 606 | struct nf_conn *ct; |
3c158f7f | 607 | struct nf_conn_help *help; |
9fb9cbb1 YK |
608 | struct nf_conntrack_tuple repl_tuple; |
609 | struct nf_conntrack_expect *exp; | |
610 | ||
605dcad6 | 611 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
0d53778e | 612 | pr_debug("Can't invert tuple.\n"); |
9fb9cbb1 YK |
613 | return NULL; |
614 | } | |
615 | ||
5a1fb391 | 616 | ct = nf_conntrack_alloc(net, tuple, &repl_tuple, GFP_ATOMIC); |
cd7fcbf1 | 617 | if (IS_ERR(ct)) { |
0d53778e | 618 | pr_debug("Can't allocate conntrack.\n"); |
c88130bc | 619 | return (struct nf_conntrack_tuple_hash *)ct; |
9fb9cbb1 YK |
620 | } |
621 | ||
c88130bc PM |
622 | if (!l4proto->new(ct, skb, dataoff)) { |
623 | nf_conntrack_free(ct); | |
0d53778e | 624 | pr_debug("init conntrack: can't track with proto module\n"); |
9fb9cbb1 YK |
625 | return NULL; |
626 | } | |
627 | ||
58401572 | 628 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
a0891aa6 | 629 | nf_ct_ecache_ext_add(ct, GFP_ATOMIC); |
58401572 | 630 | |
f8ba1aff | 631 | spin_lock_bh(&nf_conntrack_lock); |
9b03f38d | 632 | exp = nf_ct_find_expectation(net, tuple); |
9fb9cbb1 | 633 | if (exp) { |
0d53778e | 634 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", |
c88130bc | 635 | ct, exp); |
9fb9cbb1 | 636 | /* Welcome, Mr. Bond. We've been expecting you... */ |
c88130bc PM |
637 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
638 | ct->master = exp->master; | |
ceceae1b | 639 | if (exp->helper) { |
c88130bc | 640 | help = nf_ct_helper_ext_add(ct, GFP_ATOMIC); |
ceceae1b YK |
641 | if (help) |
642 | rcu_assign_pointer(help->helper, exp->helper); | |
ceceae1b YK |
643 | } |
644 | ||
9fb9cbb1 | 645 | #ifdef CONFIG_NF_CONNTRACK_MARK |
c88130bc | 646 | ct->mark = exp->master->mark; |
7c9728c3 JM |
647 | #endif |
648 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | |
c88130bc | 649 | ct->secmark = exp->master->secmark; |
9fb9cbb1 | 650 | #endif |
c88130bc | 651 | nf_conntrack_get(&ct->master->ct_general); |
0d55af87 | 652 | NF_CT_STAT_INC(net, expect_new); |
22e7410b | 653 | } else { |
226c0c0e | 654 | __nf_ct_try_assign_helper(ct, GFP_ATOMIC); |
0d55af87 | 655 | NF_CT_STAT_INC(net, new); |
22e7410b | 656 | } |
9fb9cbb1 YK |
657 | |
658 | /* Overload tuple linked list to put us in unconfirmed list. */ | |
ea781f19 | 659 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
63c9a262 | 660 | &net->ct.unconfirmed); |
9fb9cbb1 | 661 | |
f8ba1aff | 662 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
663 | |
664 | if (exp) { | |
665 | if (exp->expectfn) | |
c88130bc | 666 | exp->expectfn(ct, exp); |
6823645d | 667 | nf_ct_expect_put(exp); |
9fb9cbb1 YK |
668 | } |
669 | ||
c88130bc | 670 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
9fb9cbb1 YK |
671 | } |
672 | ||
673 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | |
674 | static inline struct nf_conn * | |
a702a65f AD |
675 | resolve_normal_ct(struct net *net, |
676 | struct sk_buff *skb, | |
9fb9cbb1 YK |
677 | unsigned int dataoff, |
678 | u_int16_t l3num, | |
679 | u_int8_t protonum, | |
680 | struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 681 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 YK |
682 | int *set_reply, |
683 | enum ip_conntrack_info *ctinfo) | |
684 | { | |
685 | struct nf_conntrack_tuple tuple; | |
686 | struct nf_conntrack_tuple_hash *h; | |
687 | struct nf_conn *ct; | |
688 | ||
bbe735e4 | 689 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
9fb9cbb1 | 690 | dataoff, l3num, protonum, &tuple, l3proto, |
605dcad6 | 691 | l4proto)) { |
0d53778e | 692 | pr_debug("resolve_normal_ct: Can't get tuple\n"); |
9fb9cbb1 YK |
693 | return NULL; |
694 | } | |
695 | ||
696 | /* look for tuple match */ | |
a702a65f | 697 | h = nf_conntrack_find_get(net, &tuple); |
9fb9cbb1 | 698 | if (!h) { |
a702a65f | 699 | h = init_conntrack(net, &tuple, l3proto, l4proto, skb, dataoff); |
9fb9cbb1 YK |
700 | if (!h) |
701 | return NULL; | |
702 | if (IS_ERR(h)) | |
703 | return (void *)h; | |
704 | } | |
705 | ct = nf_ct_tuplehash_to_ctrack(h); | |
706 | ||
707 | /* It exists; we have (non-exclusive) reference. */ | |
708 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | |
709 | *ctinfo = IP_CT_ESTABLISHED + IP_CT_IS_REPLY; | |
710 | /* Please set reply bit if this packet OK */ | |
711 | *set_reply = 1; | |
712 | } else { | |
713 | /* Once we've had two way comms, always ESTABLISHED. */ | |
714 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | |
0d53778e | 715 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); |
9fb9cbb1 YK |
716 | *ctinfo = IP_CT_ESTABLISHED; |
717 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | |
0d53778e PM |
718 | pr_debug("nf_conntrack_in: related packet for %p\n", |
719 | ct); | |
9fb9cbb1 YK |
720 | *ctinfo = IP_CT_RELATED; |
721 | } else { | |
0d53778e | 722 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); |
9fb9cbb1 YK |
723 | *ctinfo = IP_CT_NEW; |
724 | } | |
725 | *set_reply = 0; | |
726 | } | |
727 | skb->nfct = &ct->ct_general; | |
728 | skb->nfctinfo = *ctinfo; | |
729 | return ct; | |
730 | } | |
731 | ||
732 | unsigned int | |
a702a65f AD |
733 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
734 | struct sk_buff *skb) | |
9fb9cbb1 YK |
735 | { |
736 | struct nf_conn *ct; | |
737 | enum ip_conntrack_info ctinfo; | |
738 | struct nf_conntrack_l3proto *l3proto; | |
605dcad6 | 739 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 YK |
740 | unsigned int dataoff; |
741 | u_int8_t protonum; | |
742 | int set_reply = 0; | |
743 | int ret; | |
744 | ||
745 | /* Previously seen (loopback or untracked)? Ignore. */ | |
3db05fea | 746 | if (skb->nfct) { |
0d55af87 | 747 | NF_CT_STAT_INC_ATOMIC(net, ignore); |
9fb9cbb1 YK |
748 | return NF_ACCEPT; |
749 | } | |
750 | ||
923f4902 | 751 | /* rcu_read_lock()ed by nf_hook_slow */ |
76108cea | 752 | l3proto = __nf_ct_l3proto_find(pf); |
3db05fea | 753 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
ffc30690 YK |
754 | &dataoff, &protonum); |
755 | if (ret <= 0) { | |
0d53778e | 756 | pr_debug("not prepared to track yet or error occured\n"); |
0d55af87 AD |
757 | NF_CT_STAT_INC_ATOMIC(net, error); |
758 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
9fb9cbb1 YK |
759 | return -ret; |
760 | } | |
761 | ||
76108cea | 762 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
9fb9cbb1 YK |
763 | |
764 | /* It may be an special packet, error, unclean... | |
765 | * inverse of the return code tells to the netfilter | |
766 | * core what to do with the packet. */ | |
74c51a14 AD |
767 | if (l4proto->error != NULL) { |
768 | ret = l4proto->error(net, skb, dataoff, &ctinfo, pf, hooknum); | |
769 | if (ret <= 0) { | |
0d55af87 AD |
770 | NF_CT_STAT_INC_ATOMIC(net, error); |
771 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
74c51a14 AD |
772 | return -ret; |
773 | } | |
9fb9cbb1 YK |
774 | } |
775 | ||
a702a65f AD |
776 | ct = resolve_normal_ct(net, skb, dataoff, pf, protonum, |
777 | l3proto, l4proto, &set_reply, &ctinfo); | |
9fb9cbb1 YK |
778 | if (!ct) { |
779 | /* Not valid part of a connection */ | |
0d55af87 | 780 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
9fb9cbb1 YK |
781 | return NF_ACCEPT; |
782 | } | |
783 | ||
784 | if (IS_ERR(ct)) { | |
785 | /* Too stressed to deal. */ | |
0d55af87 | 786 | NF_CT_STAT_INC_ATOMIC(net, drop); |
9fb9cbb1 YK |
787 | return NF_DROP; |
788 | } | |
789 | ||
3db05fea | 790 | NF_CT_ASSERT(skb->nfct); |
9fb9cbb1 | 791 | |
3db05fea | 792 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum); |
ec8d5409 | 793 | if (ret <= 0) { |
9fb9cbb1 YK |
794 | /* Invalid: inverse of the return code tells |
795 | * the netfilter core what to do */ | |
0d53778e | 796 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
3db05fea HX |
797 | nf_conntrack_put(skb->nfct); |
798 | skb->nfct = NULL; | |
0d55af87 | 799 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
7d1e0459 PNA |
800 | if (ret == -NF_DROP) |
801 | NF_CT_STAT_INC_ATOMIC(net, drop); | |
9fb9cbb1 YK |
802 | return -ret; |
803 | } | |
804 | ||
805 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
a71996fc | 806 | nf_conntrack_event_cache(IPCT_STATUS, ct); |
9fb9cbb1 YK |
807 | |
808 | return ret; | |
809 | } | |
13b18339 | 810 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
9fb9cbb1 | 811 | |
5f2b4c90 JE |
812 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
813 | const struct nf_conntrack_tuple *orig) | |
9fb9cbb1 | 814 | { |
5f2b4c90 | 815 | bool ret; |
923f4902 PM |
816 | |
817 | rcu_read_lock(); | |
818 | ret = nf_ct_invert_tuple(inverse, orig, | |
819 | __nf_ct_l3proto_find(orig->src.l3num), | |
820 | __nf_ct_l4proto_find(orig->src.l3num, | |
821 | orig->dst.protonum)); | |
822 | rcu_read_unlock(); | |
823 | return ret; | |
9fb9cbb1 | 824 | } |
13b18339 | 825 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
9fb9cbb1 | 826 | |
5b1158e9 JK |
827 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
828 | implicitly racy: see __nf_conntrack_confirm */ | |
829 | void nf_conntrack_alter_reply(struct nf_conn *ct, | |
830 | const struct nf_conntrack_tuple *newreply) | |
831 | { | |
832 | struct nf_conn_help *help = nfct_help(ct); | |
833 | ||
5b1158e9 JK |
834 | /* Should be unconfirmed, so not in hash table yet */ |
835 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
836 | ||
0d53778e | 837 | pr_debug("Altering reply tuple of %p to ", ct); |
3c9fba65 | 838 | nf_ct_dump_tuple(newreply); |
5b1158e9 JK |
839 | |
840 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | |
ef1a5a50 | 841 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
c52fbb41 | 842 | return; |
ceceae1b | 843 | |
c52fbb41 | 844 | rcu_read_lock(); |
226c0c0e | 845 | __nf_ct_try_assign_helper(ct, GFP_ATOMIC); |
c52fbb41 | 846 | rcu_read_unlock(); |
5b1158e9 | 847 | } |
13b18339 | 848 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
5b1158e9 | 849 | |
9fb9cbb1 YK |
850 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
851 | void __nf_ct_refresh_acct(struct nf_conn *ct, | |
852 | enum ip_conntrack_info ctinfo, | |
853 | const struct sk_buff *skb, | |
854 | unsigned long extra_jiffies, | |
855 | int do_acct) | |
856 | { | |
9fb9cbb1 YK |
857 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
858 | NF_CT_ASSERT(skb); | |
859 | ||
997ae831 | 860 | /* Only update if this is not a fixed timeout */ |
47d95045 PM |
861 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
862 | goto acct; | |
997ae831 | 863 | |
9fb9cbb1 YK |
864 | /* If not in hash table, timer will not be active yet */ |
865 | if (!nf_ct_is_confirmed(ct)) { | |
866 | ct->timeout.expires = extra_jiffies; | |
9fb9cbb1 | 867 | } else { |
be00c8e4 MJ |
868 | unsigned long newtime = jiffies + extra_jiffies; |
869 | ||
870 | /* Only update the timeout if the new timeout is at least | |
871 | HZ jiffies from the old timeout. Need del_timer for race | |
872 | avoidance (may already be dying). */ | |
65cb9fda PM |
873 | if (newtime - ct->timeout.expires >= HZ) |
874 | mod_timer_pending(&ct->timeout, newtime); | |
9fb9cbb1 YK |
875 | } |
876 | ||
47d95045 | 877 | acct: |
9fb9cbb1 | 878 | if (do_acct) { |
58401572 | 879 | struct nf_conn_counter *acct; |
3ffd5eeb | 880 | |
58401572 KPO |
881 | acct = nf_conn_acct_find(ct); |
882 | if (acct) { | |
65cb9fda | 883 | spin_lock_bh(&ct->lock); |
58401572 KPO |
884 | acct[CTINFO2DIR(ctinfo)].packets++; |
885 | acct[CTINFO2DIR(ctinfo)].bytes += | |
886 | skb->len - skb_network_offset(skb); | |
65cb9fda | 887 | spin_unlock_bh(&ct->lock); |
58401572 | 888 | } |
9fb9cbb1 | 889 | } |
9fb9cbb1 | 890 | } |
13b18339 | 891 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
9fb9cbb1 | 892 | |
4c889498 DM |
893 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
894 | enum ip_conntrack_info ctinfo, | |
895 | const struct sk_buff *skb, | |
896 | int do_acct) | |
51091764 | 897 | { |
718d4ad9 | 898 | if (do_acct) { |
58401572 KPO |
899 | struct nf_conn_counter *acct; |
900 | ||
58401572 KPO |
901 | acct = nf_conn_acct_find(ct); |
902 | if (acct) { | |
65cb9fda | 903 | spin_lock_bh(&ct->lock); |
58401572 KPO |
904 | acct[CTINFO2DIR(ctinfo)].packets++; |
905 | acct[CTINFO2DIR(ctinfo)].bytes += | |
906 | skb->len - skb_network_offset(skb); | |
65cb9fda | 907 | spin_unlock_bh(&ct->lock); |
58401572 | 908 | } |
718d4ad9 | 909 | } |
58401572 | 910 | |
4c889498 | 911 | if (del_timer(&ct->timeout)) { |
51091764 | 912 | ct->timeout.function((unsigned long)ct); |
4c889498 DM |
913 | return true; |
914 | } | |
915 | return false; | |
51091764 | 916 | } |
718d4ad9 | 917 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
51091764 | 918 | |
e281db5c | 919 | #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE) |
c1d10adb PNA |
920 | |
921 | #include <linux/netfilter/nfnetlink.h> | |
922 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
57b47a53 IM |
923 | #include <linux/mutex.h> |
924 | ||
c1d10adb PNA |
925 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
926 | * in ip_conntrack_core, since we don't want the protocols to autoload | |
927 | * or depend on ctnetlink */ | |
fdf70832 | 928 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
c1d10adb PNA |
929 | const struct nf_conntrack_tuple *tuple) |
930 | { | |
77236b6e PM |
931 | NLA_PUT_BE16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port); |
932 | NLA_PUT_BE16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port); | |
c1d10adb PNA |
933 | return 0; |
934 | ||
df6fb868 | 935 | nla_put_failure: |
c1d10adb PNA |
936 | return -1; |
937 | } | |
fdf70832 | 938 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
c1d10adb | 939 | |
f73e924c PM |
940 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
941 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | |
942 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | |
c1d10adb | 943 | }; |
f73e924c | 944 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
c1d10adb | 945 | |
fdf70832 | 946 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
c1d10adb PNA |
947 | struct nf_conntrack_tuple *t) |
948 | { | |
df6fb868 | 949 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
c1d10adb PNA |
950 | return -EINVAL; |
951 | ||
77236b6e PM |
952 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
953 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | |
c1d10adb PNA |
954 | |
955 | return 0; | |
956 | } | |
fdf70832 | 957 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
5c0de29d HE |
958 | |
959 | int nf_ct_port_nlattr_tuple_size(void) | |
960 | { | |
961 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
962 | } | |
963 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | |
c1d10adb PNA |
964 | #endif |
965 | ||
9fb9cbb1 | 966 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
b334aadc | 967 | static void nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb) |
9fb9cbb1 YK |
968 | { |
969 | struct nf_conn *ct; | |
970 | enum ip_conntrack_info ctinfo; | |
971 | ||
972 | /* This ICMP is in reverse direction to the packet which caused it */ | |
973 | ct = nf_ct_get(skb, &ctinfo); | |
974 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | |
975 | ctinfo = IP_CT_RELATED + IP_CT_IS_REPLY; | |
976 | else | |
977 | ctinfo = IP_CT_RELATED; | |
978 | ||
979 | /* Attach to new skbuff, and increment count */ | |
980 | nskb->nfct = &ct->ct_general; | |
981 | nskb->nfctinfo = ctinfo; | |
982 | nf_conntrack_get(nskb->nfct); | |
983 | } | |
984 | ||
9fb9cbb1 | 985 | /* Bring out ya dead! */ |
df0933dc | 986 | static struct nf_conn * |
400dad39 | 987 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
9fb9cbb1 YK |
988 | void *data, unsigned int *bucket) |
989 | { | |
df0933dc PM |
990 | struct nf_conntrack_tuple_hash *h; |
991 | struct nf_conn *ct; | |
ea781f19 | 992 | struct hlist_nulls_node *n; |
9fb9cbb1 | 993 | |
f8ba1aff | 994 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 995 | for (; *bucket < nf_conntrack_htable_size; (*bucket)++) { |
ea781f19 | 996 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { |
df0933dc PM |
997 | ct = nf_ct_tuplehash_to_ctrack(h); |
998 | if (iter(ct, data)) | |
999 | goto found; | |
1000 | } | |
601e68e1 | 1001 | } |
ea781f19 | 1002 | hlist_nulls_for_each_entry(h, n, &net->ct.unconfirmed, hnnode) { |
df0933dc PM |
1003 | ct = nf_ct_tuplehash_to_ctrack(h); |
1004 | if (iter(ct, data)) | |
ec68e97d | 1005 | set_bit(IPS_DYING_BIT, &ct->status); |
df0933dc | 1006 | } |
f8ba1aff | 1007 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc PM |
1008 | return NULL; |
1009 | found: | |
c073e3fa | 1010 | atomic_inc(&ct->ct_general.use); |
f8ba1aff | 1011 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc | 1012 | return ct; |
9fb9cbb1 YK |
1013 | } |
1014 | ||
400dad39 AD |
1015 | void nf_ct_iterate_cleanup(struct net *net, |
1016 | int (*iter)(struct nf_conn *i, void *data), | |
1017 | void *data) | |
9fb9cbb1 | 1018 | { |
df0933dc | 1019 | struct nf_conn *ct; |
9fb9cbb1 YK |
1020 | unsigned int bucket = 0; |
1021 | ||
400dad39 | 1022 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
9fb9cbb1 YK |
1023 | /* Time to push up daises... */ |
1024 | if (del_timer(&ct->timeout)) | |
1025 | death_by_timeout((unsigned long)ct); | |
1026 | /* ... else the timer will get him soon. */ | |
1027 | ||
1028 | nf_ct_put(ct); | |
1029 | } | |
1030 | } | |
13b18339 | 1031 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
9fb9cbb1 | 1032 | |
19abb7b0 PNA |
1033 | struct __nf_ct_flush_report { |
1034 | u32 pid; | |
1035 | int report; | |
1036 | }; | |
1037 | ||
274d383b | 1038 | static int kill_report(struct nf_conn *i, void *data) |
9fb9cbb1 | 1039 | { |
19abb7b0 PNA |
1040 | struct __nf_ct_flush_report *fr = (struct __nf_ct_flush_report *)data; |
1041 | ||
dd7669a9 PNA |
1042 | /* If we fail to deliver the event, death_by_timeout() will retry */ |
1043 | if (nf_conntrack_event_report(IPCT_DESTROY, i, | |
1044 | fr->pid, fr->report) < 0) | |
1045 | return 1; | |
1046 | ||
1047 | /* Avoid the delivery of the destroy event in death_by_timeout(). */ | |
1048 | set_bit(IPS_DYING_BIT, &i->status); | |
9fb9cbb1 YK |
1049 | return 1; |
1050 | } | |
1051 | ||
274d383b PNA |
1052 | static int kill_all(struct nf_conn *i, void *data) |
1053 | { | |
1054 | return 1; | |
1055 | } | |
1056 | ||
ea781f19 | 1057 | void nf_ct_free_hashtable(void *hash, int vmalloced, unsigned int size) |
9fb9cbb1 YK |
1058 | { |
1059 | if (vmalloced) | |
1060 | vfree(hash); | |
1061 | else | |
601e68e1 | 1062 | free_pages((unsigned long)hash, |
f205c5e0 | 1063 | get_order(sizeof(struct hlist_head) * size)); |
9fb9cbb1 | 1064 | } |
ac565e5f | 1065 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
9fb9cbb1 | 1066 | |
274d383b | 1067 | void nf_conntrack_flush_report(struct net *net, u32 pid, int report) |
c1d10adb | 1068 | { |
19abb7b0 PNA |
1069 | struct __nf_ct_flush_report fr = { |
1070 | .pid = pid, | |
1071 | .report = report, | |
1072 | }; | |
274d383b | 1073 | nf_ct_iterate_cleanup(net, kill_report, &fr); |
c1d10adb | 1074 | } |
274d383b | 1075 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); |
c1d10adb | 1076 | |
dd7669a9 PNA |
1077 | static void nf_ct_release_dying_list(void) |
1078 | { | |
1079 | struct nf_conntrack_tuple_hash *h; | |
1080 | struct nf_conn *ct; | |
1081 | struct hlist_nulls_node *n; | |
1082 | ||
1083 | spin_lock_bh(&nf_conntrack_lock); | |
1084 | hlist_nulls_for_each_entry(h, n, &init_net.ct.dying, hnnode) { | |
1085 | ct = nf_ct_tuplehash_to_ctrack(h); | |
1086 | /* never fails to remove them, no listeners at this point */ | |
1087 | nf_ct_kill(ct); | |
1088 | } | |
1089 | spin_unlock_bh(&nf_conntrack_lock); | |
1090 | } | |
1091 | ||
08f6547d | 1092 | static void nf_conntrack_cleanup_init_net(void) |
9fb9cbb1 | 1093 | { |
08f6547d AD |
1094 | nf_conntrack_helper_fini(); |
1095 | nf_conntrack_proto_fini(); | |
1096 | kmem_cache_destroy(nf_conntrack_cachep); | |
1097 | } | |
9fb9cbb1 | 1098 | |
08f6547d AD |
1099 | static void nf_conntrack_cleanup_net(struct net *net) |
1100 | { | |
9fb9cbb1 | 1101 | i_see_dead_people: |
274d383b | 1102 | nf_ct_iterate_cleanup(net, kill_all, NULL); |
dd7669a9 | 1103 | nf_ct_release_dying_list(); |
49ac8713 | 1104 | if (atomic_read(&net->ct.count) != 0) { |
9fb9cbb1 YK |
1105 | schedule(); |
1106 | goto i_see_dead_people; | |
1107 | } | |
6636568c PM |
1108 | /* wait until all references to nf_conntrack_untracked are dropped */ |
1109 | while (atomic_read(&nf_conntrack_untracked.ct_general.use) > 1) | |
1110 | schedule(); | |
9fb9cbb1 | 1111 | |
400dad39 | 1112 | nf_ct_free_hashtable(net->ct.hash, net->ct.hash_vmalloc, |
ac565e5f | 1113 | nf_conntrack_htable_size); |
a0891aa6 | 1114 | nf_conntrack_ecache_fini(net); |
d716a4df | 1115 | nf_conntrack_acct_fini(net); |
9b03f38d | 1116 | nf_conntrack_expect_fini(net); |
0d55af87 | 1117 | free_percpu(net->ct.stat); |
08f6547d AD |
1118 | } |
1119 | ||
1120 | /* Mishearing the voices in his head, our hero wonders how he's | |
1121 | supposed to kill the mall. */ | |
1122 | void nf_conntrack_cleanup(struct net *net) | |
1123 | { | |
1124 | if (net_eq(net, &init_net)) | |
1125 | rcu_assign_pointer(ip_ct_attach, NULL); | |
1126 | ||
1127 | /* This makes sure all current packets have passed through | |
1128 | netfilter framework. Roll on, two-stage module | |
1129 | delete... */ | |
1130 | synchronize_net(); | |
1131 | ||
1132 | nf_conntrack_cleanup_net(net); | |
1133 | ||
1134 | if (net_eq(net, &init_net)) { | |
1135 | rcu_assign_pointer(nf_ct_destroy, NULL); | |
1136 | nf_conntrack_cleanup_init_net(); | |
1137 | } | |
9fb9cbb1 YK |
1138 | } |
1139 | ||
ea781f19 | 1140 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int *vmalloced, int nulls) |
9fb9cbb1 | 1141 | { |
ea781f19 ED |
1142 | struct hlist_nulls_head *hash; |
1143 | unsigned int nr_slots, i; | |
1144 | size_t sz; | |
9fb9cbb1 | 1145 | |
601e68e1 | 1146 | *vmalloced = 0; |
8e5105a0 | 1147 | |
ea781f19 ED |
1148 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1149 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | |
1150 | sz = nr_slots * sizeof(struct hlist_nulls_head); | |
1151 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | |
1152 | get_order(sz)); | |
601e68e1 | 1153 | if (!hash) { |
9fb9cbb1 YK |
1154 | *vmalloced = 1; |
1155 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); | |
ea781f19 | 1156 | hash = __vmalloc(sz, GFP_KERNEL | __GFP_ZERO, PAGE_KERNEL); |
9fb9cbb1 YK |
1157 | } |
1158 | ||
ea781f19 ED |
1159 | if (hash && nulls) |
1160 | for (i = 0; i < nr_slots; i++) | |
1161 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | |
9fb9cbb1 YK |
1162 | |
1163 | return hash; | |
1164 | } | |
ac565e5f | 1165 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
9fb9cbb1 | 1166 | |
fae718dd | 1167 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
9fb9cbb1 | 1168 | { |
96eb24d7 SH |
1169 | int i, bucket, vmalloced, old_vmalloced; |
1170 | unsigned int hashsize, old_size; | |
9fb9cbb1 | 1171 | int rnd; |
ea781f19 | 1172 | struct hlist_nulls_head *hash, *old_hash; |
9fb9cbb1 YK |
1173 | struct nf_conntrack_tuple_hash *h; |
1174 | ||
1175 | /* On boot, we can set this without any fancy locking. */ | |
1176 | if (!nf_conntrack_htable_size) | |
1177 | return param_set_uint(val, kp); | |
1178 | ||
96eb24d7 | 1179 | hashsize = simple_strtoul(val, NULL, 0); |
9fb9cbb1 YK |
1180 | if (!hashsize) |
1181 | return -EINVAL; | |
1182 | ||
ea781f19 | 1183 | hash = nf_ct_alloc_hashtable(&hashsize, &vmalloced, 1); |
9fb9cbb1 YK |
1184 | if (!hash) |
1185 | return -ENOMEM; | |
1186 | ||
1187 | /* We have to rehahs for the new table anyway, so we also can | |
1188 | * use a newrandom seed */ | |
af07d241 | 1189 | get_random_bytes(&rnd, sizeof(rnd)); |
9fb9cbb1 | 1190 | |
76507f69 PM |
1191 | /* Lookups in the old hash might happen in parallel, which means we |
1192 | * might get false negatives during connection lookup. New connections | |
1193 | * created because of a false negative won't make it into the hash | |
1194 | * though since that required taking the lock. | |
1195 | */ | |
f8ba1aff | 1196 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1197 | for (i = 0; i < nf_conntrack_htable_size; i++) { |
ea781f19 ED |
1198 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
1199 | h = hlist_nulls_entry(init_net.ct.hash[i].first, | |
1200 | struct nf_conntrack_tuple_hash, hnnode); | |
1201 | hlist_nulls_del_rcu(&h->hnnode); | |
9fb9cbb1 | 1202 | bucket = __hash_conntrack(&h->tuple, hashsize, rnd); |
ea781f19 | 1203 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
9fb9cbb1 YK |
1204 | } |
1205 | } | |
1206 | old_size = nf_conntrack_htable_size; | |
400dad39 AD |
1207 | old_vmalloced = init_net.ct.hash_vmalloc; |
1208 | old_hash = init_net.ct.hash; | |
9fb9cbb1 YK |
1209 | |
1210 | nf_conntrack_htable_size = hashsize; | |
400dad39 AD |
1211 | init_net.ct.hash_vmalloc = vmalloced; |
1212 | init_net.ct.hash = hash; | |
9fb9cbb1 | 1213 | nf_conntrack_hash_rnd = rnd; |
f8ba1aff | 1214 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1215 | |
ac565e5f | 1216 | nf_ct_free_hashtable(old_hash, old_vmalloced, old_size); |
9fb9cbb1 YK |
1217 | return 0; |
1218 | } | |
fae718dd | 1219 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
9fb9cbb1 | 1220 | |
fae718dd | 1221 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
9fb9cbb1 YK |
1222 | &nf_conntrack_htable_size, 0600); |
1223 | ||
08f6547d | 1224 | static int nf_conntrack_init_init_net(void) |
9fb9cbb1 | 1225 | { |
f205c5e0 | 1226 | int max_factor = 8; |
9fb9cbb1 YK |
1227 | int ret; |
1228 | ||
1229 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB | |
f205c5e0 | 1230 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ |
9fb9cbb1 YK |
1231 | if (!nf_conntrack_htable_size) { |
1232 | nf_conntrack_htable_size | |
1233 | = (((num_physpages << PAGE_SHIFT) / 16384) | |
f205c5e0 | 1234 | / sizeof(struct hlist_head)); |
9fb9cbb1 | 1235 | if (num_physpages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
f205c5e0 PM |
1236 | nf_conntrack_htable_size = 16384; |
1237 | if (nf_conntrack_htable_size < 32) | |
1238 | nf_conntrack_htable_size = 32; | |
1239 | ||
1240 | /* Use a max. factor of four by default to get the same max as | |
1241 | * with the old struct list_heads. When a table size is given | |
1242 | * we use the old value of 8 to avoid reducing the max. | |
1243 | * entries. */ | |
1244 | max_factor = 4; | |
9fb9cbb1 | 1245 | } |
f205c5e0 | 1246 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
8e5105a0 PM |
1247 | |
1248 | printk("nf_conntrack version %s (%u buckets, %d max)\n", | |
1249 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, | |
1250 | nf_conntrack_max); | |
1251 | ||
dacd2a1a YK |
1252 | nf_conntrack_cachep = kmem_cache_create("nf_conntrack", |
1253 | sizeof(struct nf_conn), | |
ea781f19 | 1254 | 0, SLAB_DESTROY_BY_RCU, NULL); |
dacd2a1a | 1255 | if (!nf_conntrack_cachep) { |
9fb9cbb1 | 1256 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); |
08f6547d AD |
1257 | ret = -ENOMEM; |
1258 | goto err_cache; | |
9fb9cbb1 YK |
1259 | } |
1260 | ||
e9c1b084 PM |
1261 | ret = nf_conntrack_proto_init(); |
1262 | if (ret < 0) | |
08f6547d | 1263 | goto err_proto; |
933a41e7 | 1264 | |
ceceae1b YK |
1265 | ret = nf_conntrack_helper_init(); |
1266 | if (ret < 0) | |
08f6547d AD |
1267 | goto err_helper; |
1268 | ||
1269 | return 0; | |
1270 | ||
1271 | err_helper: | |
1272 | nf_conntrack_proto_fini(); | |
1273 | err_proto: | |
1274 | kmem_cache_destroy(nf_conntrack_cachep); | |
1275 | err_cache: | |
1276 | return ret; | |
1277 | } | |
1278 | ||
8cc20198 ED |
1279 | /* |
1280 | * We need to use special "null" values, not used in hash table | |
1281 | */ | |
1282 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) | |
1283 | #define DYING_NULLS_VAL ((1<<30)+1) | |
1284 | ||
08f6547d AD |
1285 | static int nf_conntrack_init_net(struct net *net) |
1286 | { | |
1287 | int ret; | |
ceceae1b | 1288 | |
08f6547d | 1289 | atomic_set(&net->ct.count, 0); |
8cc20198 ED |
1290 | INIT_HLIST_NULLS_HEAD(&net->ct.unconfirmed, UNCONFIRMED_NULLS_VAL); |
1291 | INIT_HLIST_NULLS_HEAD(&net->ct.dying, DYING_NULLS_VAL); | |
08f6547d AD |
1292 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1293 | if (!net->ct.stat) { | |
1294 | ret = -ENOMEM; | |
1295 | goto err_stat; | |
1296 | } | |
08f6547d | 1297 | net->ct.hash = nf_ct_alloc_hashtable(&nf_conntrack_htable_size, |
ea781f19 | 1298 | &net->ct.hash_vmalloc, 1); |
08f6547d AD |
1299 | if (!net->ct.hash) { |
1300 | ret = -ENOMEM; | |
1301 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); | |
1302 | goto err_hash; | |
1303 | } | |
1304 | ret = nf_conntrack_expect_init(net); | |
1305 | if (ret < 0) | |
1306 | goto err_expect; | |
d716a4df | 1307 | ret = nf_conntrack_acct_init(net); |
58401572 | 1308 | if (ret < 0) |
08f6547d | 1309 | goto err_acct; |
a0891aa6 PNA |
1310 | ret = nf_conntrack_ecache_init(net); |
1311 | if (ret < 0) | |
1312 | goto err_ecache; | |
7d3cdc6b | 1313 | |
9fb9cbb1 YK |
1314 | /* Set up fake conntrack: |
1315 | - to never be deleted, not in any hashes */ | |
5a1fb391 AD |
1316 | #ifdef CONFIG_NET_NS |
1317 | nf_conntrack_untracked.ct_net = &init_net; | |
1318 | #endif | |
9fb9cbb1 YK |
1319 | atomic_set(&nf_conntrack_untracked.ct_general.use, 1); |
1320 | /* - and look it like as a confirmed connection */ | |
1321 | set_bit(IPS_CONFIRMED_BIT, &nf_conntrack_untracked.status); | |
1322 | ||
08f6547d | 1323 | return 0; |
9fb9cbb1 | 1324 | |
a0891aa6 PNA |
1325 | err_ecache: |
1326 | nf_conntrack_acct_fini(net); | |
08f6547d | 1327 | err_acct: |
9b03f38d | 1328 | nf_conntrack_expect_fini(net); |
08f6547d | 1329 | err_expect: |
400dad39 | 1330 | nf_ct_free_hashtable(net->ct.hash, net->ct.hash_vmalloc, |
ac565e5f | 1331 | nf_conntrack_htable_size); |
6058fa6b | 1332 | err_hash: |
0d55af87 AD |
1333 | free_percpu(net->ct.stat); |
1334 | err_stat: | |
08f6547d AD |
1335 | return ret; |
1336 | } | |
1337 | ||
1338 | int nf_conntrack_init(struct net *net) | |
1339 | { | |
1340 | int ret; | |
1341 | ||
1342 | if (net_eq(net, &init_net)) { | |
1343 | ret = nf_conntrack_init_init_net(); | |
1344 | if (ret < 0) | |
1345 | goto out_init_net; | |
1346 | } | |
1347 | ret = nf_conntrack_init_net(net); | |
1348 | if (ret < 0) | |
1349 | goto out_net; | |
1350 | ||
1351 | if (net_eq(net, &init_net)) { | |
1352 | /* For use by REJECT target */ | |
1353 | rcu_assign_pointer(ip_ct_attach, nf_conntrack_attach); | |
1354 | rcu_assign_pointer(nf_ct_destroy, destroy_conntrack); | |
1355 | } | |
1356 | return 0; | |
1357 | ||
1358 | out_net: | |
1359 | if (net_eq(net, &init_net)) | |
1360 | nf_conntrack_cleanup_init_net(); | |
1361 | out_init_net: | |
1362 | return ret; | |
9fb9cbb1 | 1363 | } |