]>
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> | |
d43c36dc | 17 | #include <linux/sched.h> |
9fb9cbb1 YK |
18 | #include <linux/skbuff.h> |
19 | #include <linux/proc_fs.h> | |
20 | #include <linux/vmalloc.h> | |
21 | #include <linux/stddef.h> | |
22 | #include <linux/slab.h> | |
23 | #include <linux/random.h> | |
24 | #include <linux/jhash.h> | |
25 | #include <linux/err.h> | |
26 | #include <linux/percpu.h> | |
27 | #include <linux/moduleparam.h> | |
28 | #include <linux/notifier.h> | |
29 | #include <linux/kernel.h> | |
30 | #include <linux/netdevice.h> | |
31 | #include <linux/socket.h> | |
d7fe0f24 | 32 | #include <linux/mm.h> |
d696c7bd | 33 | #include <linux/nsproxy.h> |
ea781f19 | 34 | #include <linux/rculist_nulls.h> |
9fb9cbb1 | 35 | |
9fb9cbb1 YK |
36 | #include <net/netfilter/nf_conntrack.h> |
37 | #include <net/netfilter/nf_conntrack_l3proto.h> | |
605dcad6 | 38 | #include <net/netfilter/nf_conntrack_l4proto.h> |
77ab9cff | 39 | #include <net/netfilter/nf_conntrack_expect.h> |
9fb9cbb1 YK |
40 | #include <net/netfilter/nf_conntrack_helper.h> |
41 | #include <net/netfilter/nf_conntrack_core.h> | |
ecfab2c9 | 42 | #include <net/netfilter/nf_conntrack_extend.h> |
58401572 | 43 | #include <net/netfilter/nf_conntrack_acct.h> |
a0891aa6 | 44 | #include <net/netfilter/nf_conntrack_ecache.h> |
5d0aa2cc | 45 | #include <net/netfilter/nf_conntrack_zones.h> |
a992ca2a | 46 | #include <net/netfilter/nf_conntrack_timestamp.h> |
dd705072 | 47 | #include <net/netfilter/nf_conntrack_timeout.h> |
c539f017 | 48 | #include <net/netfilter/nf_conntrack_labels.h> |
e6a7d3c0 | 49 | #include <net/netfilter/nf_nat.h> |
e17b666a | 50 | #include <net/netfilter/nf_nat_core.h> |
49376368 | 51 | #include <net/netfilter/nf_nat_helper.h> |
9fb9cbb1 | 52 | |
dc808fe2 | 53 | #define NF_CONNTRACK_VERSION "0.5.0" |
9fb9cbb1 | 54 | |
e17b666a PM |
55 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
56 | enum nf_nat_manip_type manip, | |
39938324 | 57 | const struct nlattr *attr) __read_mostly; |
e6a7d3c0 PNA |
58 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
59 | ||
c7232c99 PM |
60 | int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb, |
61 | struct nf_conn *ct, | |
62 | enum ip_conntrack_info ctinfo, | |
63 | unsigned int protoff); | |
64 | EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook); | |
65 | ||
f8ba1aff | 66 | DEFINE_SPINLOCK(nf_conntrack_lock); |
13b18339 | 67 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); |
9fb9cbb1 | 68 | |
e2b7606c | 69 | unsigned int nf_conntrack_htable_size __read_mostly; |
13b18339 PM |
70 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
71 | ||
e478075c | 72 | unsigned int nf_conntrack_max __read_mostly; |
a999e683 | 73 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
13b18339 | 74 | |
b3c5163f ED |
75 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); |
76 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); | |
13b18339 | 77 | |
f682cefa | 78 | unsigned int nf_conntrack_hash_rnd __read_mostly; |
4d4e61c6 | 79 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_rnd); |
9fb9cbb1 | 80 | |
99f07e91 | 81 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, u16 zone) |
9fb9cbb1 | 82 | { |
0794935e | 83 | unsigned int n; |
0794935e PM |
84 | |
85 | /* The direction must be ignored, so we hash everything up to the | |
86 | * destination ports (which is a multiple of 4) and treat the last | |
87 | * three bytes manually. | |
88 | */ | |
89 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); | |
99f07e91 CG |
90 | return jhash2((u32 *)tuple, n, zone ^ nf_conntrack_hash_rnd ^ |
91 | (((__force __u16)tuple->dst.u.all << 16) | | |
92 | tuple->dst.protonum)); | |
93 | } | |
94 | ||
95 | static u32 __hash_bucket(u32 hash, unsigned int size) | |
96 | { | |
97 | return ((u64)hash * size) >> 32; | |
98 | } | |
99 | ||
100 | static u32 hash_bucket(u32 hash, const struct net *net) | |
101 | { | |
102 | return __hash_bucket(hash, net->ct.htable_size); | |
103 | } | |
0794935e | 104 | |
99f07e91 CG |
105 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, |
106 | u16 zone, unsigned int size) | |
107 | { | |
108 | return __hash_bucket(hash_conntrack_raw(tuple, zone), size); | |
9fb9cbb1 YK |
109 | } |
110 | ||
5d0aa2cc | 111 | static inline u_int32_t hash_conntrack(const struct net *net, u16 zone, |
d696c7bd | 112 | const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 | 113 | { |
99f07e91 | 114 | return __hash_conntrack(tuple, zone, net->ct.htable_size); |
9fb9cbb1 YK |
115 | } |
116 | ||
5f2b4c90 | 117 | bool |
9fb9cbb1 YK |
118 | nf_ct_get_tuple(const struct sk_buff *skb, |
119 | unsigned int nhoff, | |
120 | unsigned int dataoff, | |
121 | u_int16_t l3num, | |
122 | u_int8_t protonum, | |
123 | struct nf_conntrack_tuple *tuple, | |
124 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 125 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 126 | { |
443a70d5 | 127 | memset(tuple, 0, sizeof(*tuple)); |
9fb9cbb1 YK |
128 | |
129 | tuple->src.l3num = l3num; | |
130 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | |
5f2b4c90 | 131 | return false; |
9fb9cbb1 YK |
132 | |
133 | tuple->dst.protonum = protonum; | |
134 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | |
135 | ||
605dcad6 | 136 | return l4proto->pkt_to_tuple(skb, dataoff, tuple); |
9fb9cbb1 | 137 | } |
13b18339 | 138 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
9fb9cbb1 | 139 | |
5f2b4c90 JE |
140 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
141 | u_int16_t l3num, struct nf_conntrack_tuple *tuple) | |
e2a3123f YK |
142 | { |
143 | struct nf_conntrack_l3proto *l3proto; | |
144 | struct nf_conntrack_l4proto *l4proto; | |
145 | unsigned int protoff; | |
146 | u_int8_t protonum; | |
147 | int ret; | |
148 | ||
149 | rcu_read_lock(); | |
150 | ||
151 | l3proto = __nf_ct_l3proto_find(l3num); | |
152 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | |
153 | if (ret != NF_ACCEPT) { | |
154 | rcu_read_unlock(); | |
5f2b4c90 | 155 | return false; |
e2a3123f YK |
156 | } |
157 | ||
158 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
159 | ||
160 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, tuple, | |
161 | l3proto, l4proto); | |
162 | ||
163 | rcu_read_unlock(); | |
164 | return ret; | |
165 | } | |
166 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | |
167 | ||
5f2b4c90 | 168 | bool |
9fb9cbb1 YK |
169 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
170 | const struct nf_conntrack_tuple *orig, | |
171 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 172 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 173 | { |
443a70d5 | 174 | memset(inverse, 0, sizeof(*inverse)); |
9fb9cbb1 YK |
175 | |
176 | inverse->src.l3num = orig->src.l3num; | |
177 | if (l3proto->invert_tuple(inverse, orig) == 0) | |
5f2b4c90 | 178 | return false; |
9fb9cbb1 YK |
179 | |
180 | inverse->dst.dir = !orig->dst.dir; | |
181 | ||
182 | inverse->dst.protonum = orig->dst.protonum; | |
605dcad6 | 183 | return l4proto->invert_tuple(inverse, orig); |
9fb9cbb1 | 184 | } |
13b18339 | 185 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
9fb9cbb1 | 186 | |
9fb9cbb1 YK |
187 | static void |
188 | clean_from_lists(struct nf_conn *ct) | |
189 | { | |
0d53778e | 190 | pr_debug("clean_from_lists(%p)\n", ct); |
ea781f19 ED |
191 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
192 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | |
9fb9cbb1 YK |
193 | |
194 | /* Destroy all pending expectations */ | |
c1d10adb | 195 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
196 | } |
197 | ||
198 | static void | |
199 | destroy_conntrack(struct nf_conntrack *nfct) | |
200 | { | |
201 | struct nf_conn *ct = (struct nf_conn *)nfct; | |
0d55af87 | 202 | struct net *net = nf_ct_net(ct); |
605dcad6 | 203 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 | 204 | |
0d53778e | 205 | pr_debug("destroy_conntrack(%p)\n", ct); |
9fb9cbb1 YK |
206 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
207 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); | |
208 | ||
9fb9cbb1 YK |
209 | /* To make sure we don't get any weird locking issues here: |
210 | * destroy_conntrack() MUST NOT be called with a write lock | |
211 | * to nf_conntrack_lock!!! -HW */ | |
923f4902 | 212 | rcu_read_lock(); |
5e8fbe2a | 213 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
605dcad6 MJ |
214 | if (l4proto && l4proto->destroy) |
215 | l4proto->destroy(ct); | |
9fb9cbb1 | 216 | |
982d9a9c | 217 | rcu_read_unlock(); |
9fb9cbb1 | 218 | |
f8ba1aff | 219 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
220 | /* Expectations will have been removed in clean_from_lists, |
221 | * except TFTP can create an expectation on the first packet, | |
222 | * before connection is in the list, so we need to clean here, | |
223 | * too. */ | |
c1d10adb | 224 | nf_ct_remove_expectations(ct); |
9fb9cbb1 | 225 | |
04dac011 PNA |
226 | /* We overload first tuple to link into unconfirmed or dying list.*/ |
227 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); | |
228 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | |
9fb9cbb1 | 229 | |
0d55af87 | 230 | NF_CT_STAT_INC(net, delete); |
f8ba1aff | 231 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
232 | |
233 | if (ct->master) | |
234 | nf_ct_put(ct->master); | |
235 | ||
0d53778e | 236 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
9fb9cbb1 YK |
237 | nf_conntrack_free(ct); |
238 | } | |
239 | ||
dd7669a9 | 240 | void nf_ct_delete_from_lists(struct nf_conn *ct) |
9fb9cbb1 | 241 | { |
0d55af87 | 242 | struct net *net = nf_ct_net(ct); |
9fb9cbb1 | 243 | |
9858a3ae | 244 | nf_ct_helper_destroy(ct); |
f8ba1aff | 245 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
246 | /* Inside lock so preempt is disabled on module removal path. |
247 | * Otherwise we can get spurious warnings. */ | |
0d55af87 | 248 | NF_CT_STAT_INC(net, delete_list); |
9fb9cbb1 | 249 | clean_from_lists(ct); |
04dac011 PNA |
250 | /* add this conntrack to the dying list */ |
251 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
252 | &net->ct.dying); | |
f8ba1aff | 253 | spin_unlock_bh(&nf_conntrack_lock); |
dd7669a9 PNA |
254 | } |
255 | EXPORT_SYMBOL_GPL(nf_ct_delete_from_lists); | |
256 | ||
257 | static void death_by_event(unsigned long ul_conntrack) | |
258 | { | |
259 | struct nf_conn *ct = (void *)ul_conntrack; | |
260 | struct net *net = nf_ct_net(ct); | |
5b423f6a PNA |
261 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); |
262 | ||
263 | BUG_ON(ecache == NULL); | |
dd7669a9 PNA |
264 | |
265 | if (nf_conntrack_event(IPCT_DESTROY, ct) < 0) { | |
266 | /* bad luck, let's retry again */ | |
5b423f6a | 267 | ecache->timeout.expires = jiffies + |
dd7669a9 | 268 | (random32() % net->ct.sysctl_events_retry_timeout); |
5b423f6a | 269 | add_timer(&ecache->timeout); |
dd7669a9 PNA |
270 | return; |
271 | } | |
272 | /* we've got the event delivered, now it's dying */ | |
273 | set_bit(IPS_DYING_BIT, &ct->status); | |
dd7669a9 PNA |
274 | nf_ct_put(ct); |
275 | } | |
276 | ||
04dac011 | 277 | void nf_ct_dying_timeout(struct nf_conn *ct) |
dd7669a9 PNA |
278 | { |
279 | struct net *net = nf_ct_net(ct); | |
5b423f6a PNA |
280 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); |
281 | ||
282 | BUG_ON(ecache == NULL); | |
dd7669a9 | 283 | |
dd7669a9 | 284 | /* set a new timer to retry event delivery */ |
5b423f6a PNA |
285 | setup_timer(&ecache->timeout, death_by_event, (unsigned long)ct); |
286 | ecache->timeout.expires = jiffies + | |
dd7669a9 | 287 | (random32() % net->ct.sysctl_events_retry_timeout); |
5b423f6a | 288 | add_timer(&ecache->timeout); |
dd7669a9 | 289 | } |
04dac011 | 290 | EXPORT_SYMBOL_GPL(nf_ct_dying_timeout); |
dd7669a9 PNA |
291 | |
292 | static void death_by_timeout(unsigned long ul_conntrack) | |
293 | { | |
294 | struct nf_conn *ct = (void *)ul_conntrack; | |
a992ca2a PNA |
295 | struct nf_conn_tstamp *tstamp; |
296 | ||
297 | tstamp = nf_conn_tstamp_find(ct); | |
298 | if (tstamp && tstamp->stop == 0) | |
299 | tstamp->stop = ktime_to_ns(ktime_get_real()); | |
dd7669a9 PNA |
300 | |
301 | if (!test_bit(IPS_DYING_BIT, &ct->status) && | |
302 | unlikely(nf_conntrack_event(IPCT_DESTROY, ct) < 0)) { | |
303 | /* destroy event was not delivered */ | |
304 | nf_ct_delete_from_lists(ct); | |
04dac011 | 305 | nf_ct_dying_timeout(ct); |
dd7669a9 PNA |
306 | return; |
307 | } | |
308 | set_bit(IPS_DYING_BIT, &ct->status); | |
309 | nf_ct_delete_from_lists(ct); | |
9fb9cbb1 YK |
310 | nf_ct_put(ct); |
311 | } | |
312 | ||
ea781f19 ED |
313 | /* |
314 | * Warning : | |
315 | * - Caller must take a reference on returned object | |
316 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | |
317 | * OR | |
318 | * - Caller must lock nf_conntrack_lock before calling this function | |
319 | */ | |
99f07e91 CG |
320 | static struct nf_conntrack_tuple_hash * |
321 | ____nf_conntrack_find(struct net *net, u16 zone, | |
322 | const struct nf_conntrack_tuple *tuple, u32 hash) | |
9fb9cbb1 YK |
323 | { |
324 | struct nf_conntrack_tuple_hash *h; | |
ea781f19 | 325 | struct hlist_nulls_node *n; |
99f07e91 | 326 | unsigned int bucket = hash_bucket(hash, net); |
9fb9cbb1 | 327 | |
4e29e9ec PM |
328 | /* Disable BHs the entire time since we normally need to disable them |
329 | * at least once for the stats anyway. | |
330 | */ | |
331 | local_bh_disable(); | |
ea781f19 | 332 | begin: |
99f07e91 | 333 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[bucket], hnnode) { |
5d0aa2cc PM |
334 | if (nf_ct_tuple_equal(tuple, &h->tuple) && |
335 | nf_ct_zone(nf_ct_tuplehash_to_ctrack(h)) == zone) { | |
0d55af87 | 336 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 337 | local_bh_enable(); |
9fb9cbb1 YK |
338 | return h; |
339 | } | |
0d55af87 | 340 | NF_CT_STAT_INC(net, searched); |
9fb9cbb1 | 341 | } |
ea781f19 ED |
342 | /* |
343 | * if the nulls value we got at the end of this lookup is | |
344 | * not the expected one, we must restart lookup. | |
345 | * We probably met an item that was moved to another chain. | |
346 | */ | |
99f07e91 | 347 | if (get_nulls_value(n) != bucket) { |
af740b2c | 348 | NF_CT_STAT_INC(net, search_restart); |
ea781f19 | 349 | goto begin; |
af740b2c | 350 | } |
4e29e9ec | 351 | local_bh_enable(); |
9fb9cbb1 YK |
352 | |
353 | return NULL; | |
354 | } | |
99f07e91 CG |
355 | |
356 | struct nf_conntrack_tuple_hash * | |
357 | __nf_conntrack_find(struct net *net, u16 zone, | |
358 | const struct nf_conntrack_tuple *tuple) | |
359 | { | |
360 | return ____nf_conntrack_find(net, zone, tuple, | |
361 | hash_conntrack_raw(tuple, zone)); | |
362 | } | |
13b18339 | 363 | EXPORT_SYMBOL_GPL(__nf_conntrack_find); |
9fb9cbb1 YK |
364 | |
365 | /* Find a connection corresponding to a tuple. */ | |
99f07e91 CG |
366 | static struct nf_conntrack_tuple_hash * |
367 | __nf_conntrack_find_get(struct net *net, u16 zone, | |
368 | const struct nf_conntrack_tuple *tuple, u32 hash) | |
9fb9cbb1 YK |
369 | { |
370 | struct nf_conntrack_tuple_hash *h; | |
76507f69 | 371 | struct nf_conn *ct; |
9fb9cbb1 | 372 | |
76507f69 | 373 | rcu_read_lock(); |
ea781f19 | 374 | begin: |
99f07e91 | 375 | h = ____nf_conntrack_find(net, zone, tuple, hash); |
76507f69 PM |
376 | if (h) { |
377 | ct = nf_ct_tuplehash_to_ctrack(h); | |
8d8890b7 PM |
378 | if (unlikely(nf_ct_is_dying(ct) || |
379 | !atomic_inc_not_zero(&ct->ct_general.use))) | |
76507f69 | 380 | h = NULL; |
ea781f19 | 381 | else { |
5d0aa2cc PM |
382 | if (unlikely(!nf_ct_tuple_equal(tuple, &h->tuple) || |
383 | nf_ct_zone(ct) != zone)) { | |
ea781f19 ED |
384 | nf_ct_put(ct); |
385 | goto begin; | |
386 | } | |
387 | } | |
76507f69 PM |
388 | } |
389 | rcu_read_unlock(); | |
9fb9cbb1 YK |
390 | |
391 | return h; | |
392 | } | |
99f07e91 CG |
393 | |
394 | struct nf_conntrack_tuple_hash * | |
395 | nf_conntrack_find_get(struct net *net, u16 zone, | |
396 | const struct nf_conntrack_tuple *tuple) | |
397 | { | |
398 | return __nf_conntrack_find_get(net, zone, tuple, | |
399 | hash_conntrack_raw(tuple, zone)); | |
400 | } | |
13b18339 | 401 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
9fb9cbb1 | 402 | |
c1d10adb PNA |
403 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
404 | unsigned int hash, | |
601e68e1 | 405 | unsigned int repl_hash) |
c1d10adb | 406 | { |
400dad39 AD |
407 | struct net *net = nf_ct_net(ct); |
408 | ||
ea781f19 | 409 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
400dad39 | 410 | &net->ct.hash[hash]); |
ea781f19 | 411 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
400dad39 | 412 | &net->ct.hash[repl_hash]); |
c1d10adb PNA |
413 | } |
414 | ||
7d367e06 JK |
415 | int |
416 | nf_conntrack_hash_check_insert(struct nf_conn *ct) | |
c1d10adb | 417 | { |
d696c7bd | 418 | struct net *net = nf_ct_net(ct); |
c1d10adb | 419 | unsigned int hash, repl_hash; |
7d367e06 JK |
420 | struct nf_conntrack_tuple_hash *h; |
421 | struct hlist_nulls_node *n; | |
5d0aa2cc | 422 | u16 zone; |
c1d10adb | 423 | |
5d0aa2cc | 424 | zone = nf_ct_zone(ct); |
7d367e06 JK |
425 | hash = hash_conntrack(net, zone, |
426 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
427 | repl_hash = hash_conntrack(net, zone, | |
428 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
429 | ||
430 | spin_lock_bh(&nf_conntrack_lock); | |
431 | ||
432 | /* See if there's one in the list already, including reverse */ | |
433 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) | |
434 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | |
435 | &h->tuple) && | |
436 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
437 | goto out; | |
438 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) | |
439 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, | |
440 | &h->tuple) && | |
441 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
442 | goto out; | |
c1d10adb | 443 | |
7d367e06 JK |
444 | add_timer(&ct->timeout); |
445 | nf_conntrack_get(&ct->ct_general); | |
c1d10adb | 446 | __nf_conntrack_hash_insert(ct, hash, repl_hash); |
7d367e06 JK |
447 | NF_CT_STAT_INC(net, insert); |
448 | spin_unlock_bh(&nf_conntrack_lock); | |
449 | ||
450 | return 0; | |
451 | ||
452 | out: | |
453 | NF_CT_STAT_INC(net, insert_failed); | |
454 | spin_unlock_bh(&nf_conntrack_lock); | |
455 | return -EEXIST; | |
c1d10adb | 456 | } |
7d367e06 | 457 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); |
c1d10adb | 458 | |
9fb9cbb1 YK |
459 | /* Confirm a connection given skb; places it in hash table */ |
460 | int | |
3db05fea | 461 | __nf_conntrack_confirm(struct sk_buff *skb) |
9fb9cbb1 YK |
462 | { |
463 | unsigned int hash, repl_hash; | |
df0933dc | 464 | struct nf_conntrack_tuple_hash *h; |
9fb9cbb1 | 465 | struct nf_conn *ct; |
df0933dc | 466 | struct nf_conn_help *help; |
a992ca2a | 467 | struct nf_conn_tstamp *tstamp; |
ea781f19 | 468 | struct hlist_nulls_node *n; |
9fb9cbb1 | 469 | enum ip_conntrack_info ctinfo; |
400dad39 | 470 | struct net *net; |
5d0aa2cc | 471 | u16 zone; |
9fb9cbb1 | 472 | |
3db05fea | 473 | ct = nf_ct_get(skb, &ctinfo); |
400dad39 | 474 | net = nf_ct_net(ct); |
9fb9cbb1 YK |
475 | |
476 | /* ipt_REJECT uses nf_conntrack_attach to attach related | |
477 | ICMP/TCP RST packets in other direction. Actual packet | |
478 | which created connection will be IP_CT_NEW or for an | |
479 | expected connection, IP_CT_RELATED. */ | |
480 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | |
481 | return NF_ACCEPT; | |
482 | ||
5d0aa2cc | 483 | zone = nf_ct_zone(ct); |
99f07e91 CG |
484 | /* reuse the hash saved before */ |
485 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; | |
486 | hash = hash_bucket(hash, net); | |
487 | repl_hash = hash_conntrack(net, zone, | |
488 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
9fb9cbb1 YK |
489 | |
490 | /* We're not in hash table, and we refuse to set up related | |
491 | connections for unconfirmed conns. But packet copies and | |
492 | REJECT will give spurious warnings here. */ | |
493 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ | |
494 | ||
25985edc | 495 | /* No external references means no one else could have |
9fb9cbb1 YK |
496 | confirmed us. */ |
497 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
0d53778e | 498 | pr_debug("Confirming conntrack %p\n", ct); |
9fb9cbb1 | 499 | |
f8ba1aff | 500 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 501 | |
fc350777 JM |
502 | /* We have to check the DYING flag inside the lock to prevent |
503 | a race against nf_ct_get_next_corpse() possibly called from | |
504 | user context, else we insert an already 'dead' hash, blocking | |
505 | further use of that particular connection -JM */ | |
506 | ||
507 | if (unlikely(nf_ct_is_dying(ct))) { | |
508 | spin_unlock_bh(&nf_conntrack_lock); | |
509 | return NF_ACCEPT; | |
510 | } | |
511 | ||
9fb9cbb1 YK |
512 | /* See if there's one in the list already, including reverse: |
513 | NAT could have grabbed it without realizing, since we're | |
514 | not in the hash. If there is, we lost race. */ | |
ea781f19 | 515 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
df0933dc | 516 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
5d0aa2cc PM |
517 | &h->tuple) && |
518 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
df0933dc | 519 | goto out; |
ea781f19 | 520 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) |
df0933dc | 521 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
5d0aa2cc PM |
522 | &h->tuple) && |
523 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
df0933dc | 524 | goto out; |
9fb9cbb1 | 525 | |
df0933dc | 526 | /* Remove from unconfirmed list */ |
ea781f19 | 527 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
df0933dc | 528 | |
df0933dc PM |
529 | /* Timer relative to confirmation time, not original |
530 | setting time, otherwise we'd get timer wrap in | |
531 | weird delay cases. */ | |
532 | ct->timeout.expires += jiffies; | |
533 | add_timer(&ct->timeout); | |
534 | atomic_inc(&ct->ct_general.use); | |
45eec341 | 535 | ct->status |= IPS_CONFIRMED; |
5c8ec910 | 536 | |
a992ca2a PNA |
537 | /* set conntrack timestamp, if enabled. */ |
538 | tstamp = nf_conn_tstamp_find(ct); | |
539 | if (tstamp) { | |
540 | if (skb->tstamp.tv64 == 0) | |
e3192690 | 541 | __net_timestamp(skb); |
a992ca2a PNA |
542 | |
543 | tstamp->start = ktime_to_ns(skb->tstamp); | |
544 | } | |
5c8ec910 PM |
545 | /* Since the lookup is lockless, hash insertion must be done after |
546 | * starting the timer and setting the CONFIRMED bit. The RCU barriers | |
547 | * guarantee that no other CPU can find the conntrack before the above | |
548 | * stores are visible. | |
549 | */ | |
550 | __nf_conntrack_hash_insert(ct, hash, repl_hash); | |
0d55af87 | 551 | NF_CT_STAT_INC(net, insert); |
f8ba1aff | 552 | spin_unlock_bh(&nf_conntrack_lock); |
5c8ec910 | 553 | |
df0933dc PM |
554 | help = nfct_help(ct); |
555 | if (help && help->helper) | |
a71996fc | 556 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
17e6e4ea | 557 | |
df0933dc | 558 | nf_conntrack_event_cache(master_ct(ct) ? |
a71996fc | 559 | IPCT_RELATED : IPCT_NEW, ct); |
df0933dc | 560 | return NF_ACCEPT; |
9fb9cbb1 | 561 | |
df0933dc | 562 | out: |
0d55af87 | 563 | NF_CT_STAT_INC(net, insert_failed); |
f8ba1aff | 564 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
565 | return NF_DROP; |
566 | } | |
13b18339 | 567 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
9fb9cbb1 YK |
568 | |
569 | /* Returns true if a connection correspondings to the tuple (required | |
570 | for NAT). */ | |
571 | int | |
572 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | |
573 | const struct nf_conn *ignored_conntrack) | |
574 | { | |
400dad39 | 575 | struct net *net = nf_ct_net(ignored_conntrack); |
9fb9cbb1 | 576 | struct nf_conntrack_tuple_hash *h; |
ea781f19 | 577 | struct hlist_nulls_node *n; |
5d0aa2cc PM |
578 | struct nf_conn *ct; |
579 | u16 zone = nf_ct_zone(ignored_conntrack); | |
580 | unsigned int hash = hash_conntrack(net, zone, tuple); | |
9fb9cbb1 | 581 | |
4e29e9ec PM |
582 | /* Disable BHs the entire time since we need to disable them at |
583 | * least once for the stats anyway. | |
584 | */ | |
585 | rcu_read_lock_bh(); | |
ea781f19 | 586 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
5d0aa2cc PM |
587 | ct = nf_ct_tuplehash_to_ctrack(h); |
588 | if (ct != ignored_conntrack && | |
589 | nf_ct_tuple_equal(tuple, &h->tuple) && | |
590 | nf_ct_zone(ct) == zone) { | |
0d55af87 | 591 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 592 | rcu_read_unlock_bh(); |
ba419aff PM |
593 | return 1; |
594 | } | |
0d55af87 | 595 | NF_CT_STAT_INC(net, searched); |
ba419aff | 596 | } |
4e29e9ec | 597 | rcu_read_unlock_bh(); |
9fb9cbb1 | 598 | |
ba419aff | 599 | return 0; |
9fb9cbb1 | 600 | } |
13b18339 | 601 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
9fb9cbb1 | 602 | |
7ae7730f PM |
603 | #define NF_CT_EVICTION_RANGE 8 |
604 | ||
9fb9cbb1 YK |
605 | /* There's a small race here where we may free a just-assured |
606 | connection. Too bad: we're in trouble anyway. */ | |
400dad39 | 607 | static noinline int early_drop(struct net *net, unsigned int hash) |
9fb9cbb1 | 608 | { |
f205c5e0 | 609 | /* Use oldest entry, which is roughly LRU */ |
9fb9cbb1 | 610 | struct nf_conntrack_tuple_hash *h; |
df0933dc | 611 | struct nf_conn *ct = NULL, *tmp; |
ea781f19 | 612 | struct hlist_nulls_node *n; |
7ae7730f | 613 | unsigned int i, cnt = 0; |
9fb9cbb1 YK |
614 | int dropped = 0; |
615 | ||
76507f69 | 616 | rcu_read_lock(); |
d696c7bd | 617 | for (i = 0; i < net->ct.htable_size; i++) { |
ea781f19 ED |
618 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
619 | hnnode) { | |
7ae7730f PM |
620 | tmp = nf_ct_tuplehash_to_ctrack(h); |
621 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) | |
622 | ct = tmp; | |
623 | cnt++; | |
624 | } | |
76507f69 | 625 | |
5ae27aa2 CG |
626 | if (ct != NULL) { |
627 | if (likely(!nf_ct_is_dying(ct) && | |
628 | atomic_inc_not_zero(&ct->ct_general.use))) | |
629 | break; | |
630 | else | |
631 | ct = NULL; | |
632 | } | |
633 | ||
634 | if (cnt >= NF_CT_EVICTION_RANGE) | |
7ae7730f | 635 | break; |
5ae27aa2 | 636 | |
d696c7bd | 637 | hash = (hash + 1) % net->ct.htable_size; |
9fb9cbb1 | 638 | } |
76507f69 | 639 | rcu_read_unlock(); |
9fb9cbb1 YK |
640 | |
641 | if (!ct) | |
642 | return dropped; | |
643 | ||
644 | if (del_timer(&ct->timeout)) { | |
645 | death_by_timeout((unsigned long)ct); | |
74138511 PNA |
646 | /* Check if we indeed killed this entry. Reliable event |
647 | delivery may have inserted it into the dying list. */ | |
648 | if (test_bit(IPS_DYING_BIT, &ct->status)) { | |
649 | dropped = 1; | |
650 | NF_CT_STAT_INC_ATOMIC(net, early_drop); | |
651 | } | |
9fb9cbb1 YK |
652 | } |
653 | nf_ct_put(ct); | |
654 | return dropped; | |
655 | } | |
656 | ||
f682cefa CG |
657 | void init_nf_conntrack_hash_rnd(void) |
658 | { | |
659 | unsigned int rand; | |
660 | ||
661 | /* | |
662 | * Why not initialize nf_conntrack_rnd in a "init()" function ? | |
663 | * Because there isn't enough entropy when system initializing, | |
664 | * and we initialize it as late as possible. | |
665 | */ | |
666 | do { | |
667 | get_random_bytes(&rand, sizeof(rand)); | |
668 | } while (!rand); | |
669 | cmpxchg(&nf_conntrack_hash_rnd, 0, rand); | |
670 | } | |
671 | ||
99f07e91 CG |
672 | static struct nf_conn * |
673 | __nf_conntrack_alloc(struct net *net, u16 zone, | |
674 | const struct nf_conntrack_tuple *orig, | |
675 | const struct nf_conntrack_tuple *repl, | |
676 | gfp_t gfp, u32 hash) | |
9fb9cbb1 | 677 | { |
cd7fcbf1 | 678 | struct nf_conn *ct; |
9fb9cbb1 | 679 | |
b2390969 | 680 | if (unlikely(!nf_conntrack_hash_rnd)) { |
f682cefa | 681 | init_nf_conntrack_hash_rnd(); |
99f07e91 CG |
682 | /* recompute the hash as nf_conntrack_hash_rnd is initialized */ |
683 | hash = hash_conntrack_raw(orig, zone); | |
9fb9cbb1 YK |
684 | } |
685 | ||
5251e2d2 | 686 | /* We don't want any race condition at early drop stage */ |
49ac8713 | 687 | atomic_inc(&net->ct.count); |
5251e2d2 | 688 | |
76eb9460 | 689 | if (nf_conntrack_max && |
49ac8713 | 690 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
99f07e91 | 691 | if (!early_drop(net, hash_bucket(hash, net))) { |
49ac8713 | 692 | atomic_dec(&net->ct.count); |
e87cc472 | 693 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); |
9fb9cbb1 YK |
694 | return ERR_PTR(-ENOMEM); |
695 | } | |
696 | } | |
697 | ||
941297f4 ED |
698 | /* |
699 | * Do not use kmem_cache_zalloc(), as this cache uses | |
700 | * SLAB_DESTROY_BY_RCU. | |
701 | */ | |
5b3501fa | 702 | ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp); |
c88130bc | 703 | if (ct == NULL) { |
49ac8713 | 704 | atomic_dec(&net->ct.count); |
dacd2a1a | 705 | return ERR_PTR(-ENOMEM); |
9fb9cbb1 | 706 | } |
941297f4 ED |
707 | /* |
708 | * Let ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.next | |
709 | * and ct->tuplehash[IP_CT_DIR_REPLY].hnnode.next unchanged. | |
710 | */ | |
711 | memset(&ct->tuplehash[IP_CT_DIR_MAX], 0, | |
e5fc9e7a CG |
712 | offsetof(struct nf_conn, proto) - |
713 | offsetof(struct nf_conn, tuplehash[IP_CT_DIR_MAX])); | |
440f0d58 | 714 | spin_lock_init(&ct->lock); |
c88130bc | 715 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; |
941297f4 | 716 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; |
c88130bc | 717 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; |
99f07e91 CG |
718 | /* save hash for reusing when confirming */ |
719 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; | |
9fb9cbb1 | 720 | /* Don't set timer yet: wait for confirmation */ |
c88130bc | 721 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
c2d9ba9b | 722 | write_pnet(&ct->ct_net, net); |
5d0aa2cc PM |
723 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
724 | if (zone) { | |
725 | struct nf_conntrack_zone *nf_ct_zone; | |
726 | ||
727 | nf_ct_zone = nf_ct_ext_add(ct, NF_CT_EXT_ZONE, GFP_ATOMIC); | |
728 | if (!nf_ct_zone) | |
729 | goto out_free; | |
730 | nf_ct_zone->id = zone; | |
731 | } | |
732 | #endif | |
941297f4 ED |
733 | /* |
734 | * changes to lookup keys must be done before setting refcnt to 1 | |
735 | */ | |
736 | smp_wmb(); | |
737 | atomic_set(&ct->ct_general.use, 1); | |
c88130bc | 738 | return ct; |
5d0aa2cc PM |
739 | |
740 | #ifdef CONFIG_NF_CONNTRACK_ZONES | |
741 | out_free: | |
d96fc659 | 742 | atomic_dec(&net->ct.count); |
5d0aa2cc PM |
743 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
744 | return ERR_PTR(-ENOMEM); | |
745 | #endif | |
9fb9cbb1 | 746 | } |
99f07e91 CG |
747 | |
748 | struct nf_conn *nf_conntrack_alloc(struct net *net, u16 zone, | |
749 | const struct nf_conntrack_tuple *orig, | |
750 | const struct nf_conntrack_tuple *repl, | |
751 | gfp_t gfp) | |
752 | { | |
753 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); | |
754 | } | |
13b18339 | 755 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
9fb9cbb1 | 756 | |
c88130bc | 757 | void nf_conntrack_free(struct nf_conn *ct) |
76507f69 | 758 | { |
1d45209d ED |
759 | struct net *net = nf_ct_net(ct); |
760 | ||
ceeff754 | 761 | nf_ct_ext_destroy(ct); |
1d45209d | 762 | atomic_dec(&net->ct.count); |
ea781f19 | 763 | nf_ct_ext_free(ct); |
5b3501fa | 764 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
76507f69 | 765 | } |
13b18339 | 766 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
9fb9cbb1 | 767 | |
c539f017 | 768 | |
9fb9cbb1 YK |
769 | /* Allocate a new conntrack: we return -ENOMEM if classification |
770 | failed due to stress. Otherwise it really is unclassifiable. */ | |
771 | static struct nf_conntrack_tuple_hash * | |
b2a15a60 | 772 | init_conntrack(struct net *net, struct nf_conn *tmpl, |
5a1fb391 | 773 | const struct nf_conntrack_tuple *tuple, |
9fb9cbb1 | 774 | struct nf_conntrack_l3proto *l3proto, |
605dcad6 | 775 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 776 | struct sk_buff *skb, |
60b5f8f7 | 777 | unsigned int dataoff, u32 hash) |
9fb9cbb1 | 778 | { |
c88130bc | 779 | struct nf_conn *ct; |
3c158f7f | 780 | struct nf_conn_help *help; |
9fb9cbb1 | 781 | struct nf_conntrack_tuple repl_tuple; |
b2a15a60 | 782 | struct nf_conntrack_ecache *ecache; |
9fb9cbb1 | 783 | struct nf_conntrack_expect *exp; |
5d0aa2cc | 784 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
60b5f8f7 PNA |
785 | struct nf_conn_timeout *timeout_ext; |
786 | unsigned int *timeouts; | |
9fb9cbb1 | 787 | |
605dcad6 | 788 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
0d53778e | 789 | pr_debug("Can't invert tuple.\n"); |
9fb9cbb1 YK |
790 | return NULL; |
791 | } | |
792 | ||
99f07e91 CG |
793 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, |
794 | hash); | |
0a9ee813 | 795 | if (IS_ERR(ct)) |
c88130bc | 796 | return (struct nf_conntrack_tuple_hash *)ct; |
9fb9cbb1 | 797 | |
60b5f8f7 PNA |
798 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; |
799 | if (timeout_ext) | |
800 | timeouts = NF_CT_TIMEOUT_EXT_DATA(timeout_ext); | |
801 | else | |
802 | timeouts = l4proto->get_timeouts(net); | |
803 | ||
2c8503f5 | 804 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { |
c88130bc | 805 | nf_conntrack_free(ct); |
0d53778e | 806 | pr_debug("init conntrack: can't track with proto module\n"); |
9fb9cbb1 YK |
807 | return NULL; |
808 | } | |
809 | ||
60b5f8f7 PNA |
810 | if (timeout_ext) |
811 | nf_ct_timeout_ext_add(ct, timeout_ext->timeout, GFP_ATOMIC); | |
812 | ||
58401572 | 813 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
a992ca2a | 814 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); |
c539f017 | 815 | nf_ct_labels_ext_add(ct); |
b2a15a60 PM |
816 | |
817 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; | |
818 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, | |
819 | ecache ? ecache->expmask : 0, | |
820 | GFP_ATOMIC); | |
58401572 | 821 | |
f8ba1aff | 822 | spin_lock_bh(&nf_conntrack_lock); |
5d0aa2cc | 823 | exp = nf_ct_find_expectation(net, zone, tuple); |
9fb9cbb1 | 824 | if (exp) { |
0d53778e | 825 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", |
c88130bc | 826 | ct, exp); |
9fb9cbb1 | 827 | /* Welcome, Mr. Bond. We've been expecting you... */ |
c88130bc PM |
828 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
829 | ct->master = exp->master; | |
ceceae1b | 830 | if (exp->helper) { |
1afc5679 PNA |
831 | help = nf_ct_helper_ext_add(ct, exp->helper, |
832 | GFP_ATOMIC); | |
ceceae1b | 833 | if (help) |
cf778b00 | 834 | rcu_assign_pointer(help->helper, exp->helper); |
ceceae1b YK |
835 | } |
836 | ||
9fb9cbb1 | 837 | #ifdef CONFIG_NF_CONNTRACK_MARK |
c88130bc | 838 | ct->mark = exp->master->mark; |
7c9728c3 JM |
839 | #endif |
840 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | |
c88130bc | 841 | ct->secmark = exp->master->secmark; |
9fb9cbb1 | 842 | #endif |
c88130bc | 843 | nf_conntrack_get(&ct->master->ct_general); |
0d55af87 | 844 | NF_CT_STAT_INC(net, expect_new); |
22e7410b | 845 | } else { |
b2a15a60 | 846 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); |
0d55af87 | 847 | NF_CT_STAT_INC(net, new); |
22e7410b | 848 | } |
9fb9cbb1 YK |
849 | |
850 | /* Overload tuple linked list to put us in unconfirmed list. */ | |
ea781f19 | 851 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
63c9a262 | 852 | &net->ct.unconfirmed); |
9fb9cbb1 | 853 | |
f8ba1aff | 854 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
855 | |
856 | if (exp) { | |
857 | if (exp->expectfn) | |
c88130bc | 858 | exp->expectfn(ct, exp); |
6823645d | 859 | nf_ct_expect_put(exp); |
9fb9cbb1 YK |
860 | } |
861 | ||
c88130bc | 862 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
9fb9cbb1 YK |
863 | } |
864 | ||
865 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | |
866 | static inline struct nf_conn * | |
b2a15a60 | 867 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, |
a702a65f | 868 | struct sk_buff *skb, |
9fb9cbb1 YK |
869 | unsigned int dataoff, |
870 | u_int16_t l3num, | |
871 | u_int8_t protonum, | |
872 | struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 873 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 874 | int *set_reply, |
60b5f8f7 | 875 | enum ip_conntrack_info *ctinfo) |
9fb9cbb1 YK |
876 | { |
877 | struct nf_conntrack_tuple tuple; | |
878 | struct nf_conntrack_tuple_hash *h; | |
879 | struct nf_conn *ct; | |
5d0aa2cc | 880 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
99f07e91 | 881 | u32 hash; |
9fb9cbb1 | 882 | |
bbe735e4 | 883 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
9fb9cbb1 | 884 | dataoff, l3num, protonum, &tuple, l3proto, |
605dcad6 | 885 | l4proto)) { |
0d53778e | 886 | pr_debug("resolve_normal_ct: Can't get tuple\n"); |
9fb9cbb1 YK |
887 | return NULL; |
888 | } | |
889 | ||
890 | /* look for tuple match */ | |
99f07e91 CG |
891 | hash = hash_conntrack_raw(&tuple, zone); |
892 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); | |
9fb9cbb1 | 893 | if (!h) { |
b2a15a60 | 894 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, |
60b5f8f7 | 895 | skb, dataoff, hash); |
9fb9cbb1 YK |
896 | if (!h) |
897 | return NULL; | |
898 | if (IS_ERR(h)) | |
899 | return (void *)h; | |
900 | } | |
901 | ct = nf_ct_tuplehash_to_ctrack(h); | |
902 | ||
903 | /* It exists; we have (non-exclusive) reference. */ | |
904 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | |
fb048833 | 905 | *ctinfo = IP_CT_ESTABLISHED_REPLY; |
9fb9cbb1 YK |
906 | /* Please set reply bit if this packet OK */ |
907 | *set_reply = 1; | |
908 | } else { | |
909 | /* Once we've had two way comms, always ESTABLISHED. */ | |
910 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | |
0d53778e | 911 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); |
9fb9cbb1 YK |
912 | *ctinfo = IP_CT_ESTABLISHED; |
913 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | |
0d53778e PM |
914 | pr_debug("nf_conntrack_in: related packet for %p\n", |
915 | ct); | |
9fb9cbb1 YK |
916 | *ctinfo = IP_CT_RELATED; |
917 | } else { | |
0d53778e | 918 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); |
9fb9cbb1 YK |
919 | *ctinfo = IP_CT_NEW; |
920 | } | |
921 | *set_reply = 0; | |
922 | } | |
923 | skb->nfct = &ct->ct_general; | |
924 | skb->nfctinfo = *ctinfo; | |
925 | return ct; | |
926 | } | |
927 | ||
928 | unsigned int | |
a702a65f AD |
929 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
930 | struct sk_buff *skb) | |
9fb9cbb1 | 931 | { |
b2a15a60 | 932 | struct nf_conn *ct, *tmpl = NULL; |
9fb9cbb1 YK |
933 | enum ip_conntrack_info ctinfo; |
934 | struct nf_conntrack_l3proto *l3proto; | |
605dcad6 | 935 | struct nf_conntrack_l4proto *l4proto; |
2c8503f5 | 936 | unsigned int *timeouts; |
9fb9cbb1 YK |
937 | unsigned int dataoff; |
938 | u_int8_t protonum; | |
939 | int set_reply = 0; | |
940 | int ret; | |
941 | ||
3db05fea | 942 | if (skb->nfct) { |
b2a15a60 PM |
943 | /* Previously seen (loopback or untracked)? Ignore. */ |
944 | tmpl = (struct nf_conn *)skb->nfct; | |
945 | if (!nf_ct_is_template(tmpl)) { | |
946 | NF_CT_STAT_INC_ATOMIC(net, ignore); | |
947 | return NF_ACCEPT; | |
948 | } | |
949 | skb->nfct = NULL; | |
9fb9cbb1 YK |
950 | } |
951 | ||
923f4902 | 952 | /* rcu_read_lock()ed by nf_hook_slow */ |
76108cea | 953 | l3proto = __nf_ct_l3proto_find(pf); |
3db05fea | 954 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
ffc30690 YK |
955 | &dataoff, &protonum); |
956 | if (ret <= 0) { | |
25985edc | 957 | pr_debug("not prepared to track yet or error occurred\n"); |
0d55af87 AD |
958 | NF_CT_STAT_INC_ATOMIC(net, error); |
959 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
960 | ret = -ret; |
961 | goto out; | |
9fb9cbb1 YK |
962 | } |
963 | ||
76108cea | 964 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
9fb9cbb1 YK |
965 | |
966 | /* It may be an special packet, error, unclean... | |
967 | * inverse of the return code tells to the netfilter | |
968 | * core what to do with the packet. */ | |
74c51a14 | 969 | if (l4proto->error != NULL) { |
8fea97ec PM |
970 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, |
971 | pf, hooknum); | |
74c51a14 | 972 | if (ret <= 0) { |
0d55af87 AD |
973 | NF_CT_STAT_INC_ATOMIC(net, error); |
974 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
975 | ret = -ret; |
976 | goto out; | |
74c51a14 | 977 | } |
88ed01d1 PNA |
978 | /* ICMP[v6] protocol trackers may assign one conntrack. */ |
979 | if (skb->nfct) | |
980 | goto out; | |
9fb9cbb1 YK |
981 | } |
982 | ||
b2a15a60 | 983 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, |
60b5f8f7 | 984 | l3proto, l4proto, &set_reply, &ctinfo); |
9fb9cbb1 YK |
985 | if (!ct) { |
986 | /* Not valid part of a connection */ | |
0d55af87 | 987 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
b2a15a60 PM |
988 | ret = NF_ACCEPT; |
989 | goto out; | |
9fb9cbb1 YK |
990 | } |
991 | ||
992 | if (IS_ERR(ct)) { | |
993 | /* Too stressed to deal. */ | |
0d55af87 | 994 | NF_CT_STAT_INC_ATOMIC(net, drop); |
b2a15a60 PM |
995 | ret = NF_DROP; |
996 | goto out; | |
9fb9cbb1 YK |
997 | } |
998 | ||
3db05fea | 999 | NF_CT_ASSERT(skb->nfct); |
9fb9cbb1 | 1000 | |
60b5f8f7 | 1001 | /* Decide what timeout policy we want to apply to this flow. */ |
84b5ee93 | 1002 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); |
60b5f8f7 | 1003 | |
2c8503f5 | 1004 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); |
ec8d5409 | 1005 | if (ret <= 0) { |
9fb9cbb1 YK |
1006 | /* Invalid: inverse of the return code tells |
1007 | * the netfilter core what to do */ | |
0d53778e | 1008 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
3db05fea HX |
1009 | nf_conntrack_put(skb->nfct); |
1010 | skb->nfct = NULL; | |
0d55af87 | 1011 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
7d1e0459 PNA |
1012 | if (ret == -NF_DROP) |
1013 | NF_CT_STAT_INC_ATOMIC(net, drop); | |
b2a15a60 PM |
1014 | ret = -ret; |
1015 | goto out; | |
9fb9cbb1 YK |
1016 | } |
1017 | ||
1018 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
858b3133 | 1019 | nf_conntrack_event_cache(IPCT_REPLY, ct); |
b2a15a60 | 1020 | out: |
c3174286 PNA |
1021 | if (tmpl) { |
1022 | /* Special case: we have to repeat this hook, assign the | |
1023 | * template again to this packet. We assume that this packet | |
1024 | * has no conntrack assigned. This is used by nf_ct_tcp. */ | |
1025 | if (ret == NF_REPEAT) | |
1026 | skb->nfct = (struct nf_conntrack *)tmpl; | |
1027 | else | |
1028 | nf_ct_put(tmpl); | |
1029 | } | |
9fb9cbb1 YK |
1030 | |
1031 | return ret; | |
1032 | } | |
13b18339 | 1033 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
9fb9cbb1 | 1034 | |
5f2b4c90 JE |
1035 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
1036 | const struct nf_conntrack_tuple *orig) | |
9fb9cbb1 | 1037 | { |
5f2b4c90 | 1038 | bool ret; |
923f4902 PM |
1039 | |
1040 | rcu_read_lock(); | |
1041 | ret = nf_ct_invert_tuple(inverse, orig, | |
1042 | __nf_ct_l3proto_find(orig->src.l3num), | |
1043 | __nf_ct_l4proto_find(orig->src.l3num, | |
1044 | orig->dst.protonum)); | |
1045 | rcu_read_unlock(); | |
1046 | return ret; | |
9fb9cbb1 | 1047 | } |
13b18339 | 1048 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
9fb9cbb1 | 1049 | |
5b1158e9 JK |
1050 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
1051 | implicitly racy: see __nf_conntrack_confirm */ | |
1052 | void nf_conntrack_alter_reply(struct nf_conn *ct, | |
1053 | const struct nf_conntrack_tuple *newreply) | |
1054 | { | |
1055 | struct nf_conn_help *help = nfct_help(ct); | |
1056 | ||
5b1158e9 JK |
1057 | /* Should be unconfirmed, so not in hash table yet */ |
1058 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
1059 | ||
0d53778e | 1060 | pr_debug("Altering reply tuple of %p to ", ct); |
3c9fba65 | 1061 | nf_ct_dump_tuple(newreply); |
5b1158e9 JK |
1062 | |
1063 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | |
ef1a5a50 | 1064 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
c52fbb41 | 1065 | return; |
ceceae1b | 1066 | |
c52fbb41 | 1067 | rcu_read_lock(); |
b2a15a60 | 1068 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); |
c52fbb41 | 1069 | rcu_read_unlock(); |
5b1158e9 | 1070 | } |
13b18339 | 1071 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
5b1158e9 | 1072 | |
9fb9cbb1 YK |
1073 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
1074 | void __nf_ct_refresh_acct(struct nf_conn *ct, | |
1075 | enum ip_conntrack_info ctinfo, | |
1076 | const struct sk_buff *skb, | |
1077 | unsigned long extra_jiffies, | |
1078 | int do_acct) | |
1079 | { | |
9fb9cbb1 YK |
1080 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
1081 | NF_CT_ASSERT(skb); | |
1082 | ||
997ae831 | 1083 | /* Only update if this is not a fixed timeout */ |
47d95045 PM |
1084 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
1085 | goto acct; | |
997ae831 | 1086 | |
9fb9cbb1 YK |
1087 | /* If not in hash table, timer will not be active yet */ |
1088 | if (!nf_ct_is_confirmed(ct)) { | |
1089 | ct->timeout.expires = extra_jiffies; | |
9fb9cbb1 | 1090 | } else { |
be00c8e4 MJ |
1091 | unsigned long newtime = jiffies + extra_jiffies; |
1092 | ||
1093 | /* Only update the timeout if the new timeout is at least | |
1094 | HZ jiffies from the old timeout. Need del_timer for race | |
1095 | avoidance (may already be dying). */ | |
65cb9fda PM |
1096 | if (newtime - ct->timeout.expires >= HZ) |
1097 | mod_timer_pending(&ct->timeout, newtime); | |
9fb9cbb1 YK |
1098 | } |
1099 | ||
47d95045 | 1100 | acct: |
9fb9cbb1 | 1101 | if (do_acct) { |
58401572 | 1102 | struct nf_conn_counter *acct; |
3ffd5eeb | 1103 | |
58401572 KPO |
1104 | acct = nf_conn_acct_find(ct); |
1105 | if (acct) { | |
b3e0bfa7 ED |
1106 | atomic64_inc(&acct[CTINFO2DIR(ctinfo)].packets); |
1107 | atomic64_add(skb->len, &acct[CTINFO2DIR(ctinfo)].bytes); | |
58401572 | 1108 | } |
9fb9cbb1 | 1109 | } |
9fb9cbb1 | 1110 | } |
13b18339 | 1111 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
9fb9cbb1 | 1112 | |
4c889498 DM |
1113 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
1114 | enum ip_conntrack_info ctinfo, | |
1115 | const struct sk_buff *skb, | |
1116 | int do_acct) | |
51091764 | 1117 | { |
718d4ad9 | 1118 | if (do_acct) { |
58401572 KPO |
1119 | struct nf_conn_counter *acct; |
1120 | ||
58401572 KPO |
1121 | acct = nf_conn_acct_find(ct); |
1122 | if (acct) { | |
b3e0bfa7 ED |
1123 | atomic64_inc(&acct[CTINFO2DIR(ctinfo)].packets); |
1124 | atomic64_add(skb->len - skb_network_offset(skb), | |
1125 | &acct[CTINFO2DIR(ctinfo)].bytes); | |
58401572 | 1126 | } |
718d4ad9 | 1127 | } |
58401572 | 1128 | |
4c889498 | 1129 | if (del_timer(&ct->timeout)) { |
51091764 | 1130 | ct->timeout.function((unsigned long)ct); |
4c889498 DM |
1131 | return true; |
1132 | } | |
1133 | return false; | |
51091764 | 1134 | } |
718d4ad9 | 1135 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
51091764 | 1136 | |
5d0aa2cc PM |
1137 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1138 | static struct nf_ct_ext_type nf_ct_zone_extend __read_mostly = { | |
1139 | .len = sizeof(struct nf_conntrack_zone), | |
1140 | .align = __alignof__(struct nf_conntrack_zone), | |
1141 | .id = NF_CT_EXT_ZONE, | |
1142 | }; | |
1143 | #endif | |
1144 | ||
c0cd1156 | 1145 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
c1d10adb PNA |
1146 | |
1147 | #include <linux/netfilter/nfnetlink.h> | |
1148 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
57b47a53 IM |
1149 | #include <linux/mutex.h> |
1150 | ||
c1d10adb PNA |
1151 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
1152 | * in ip_conntrack_core, since we don't want the protocols to autoload | |
1153 | * or depend on ctnetlink */ | |
fdf70832 | 1154 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
c1d10adb PNA |
1155 | const struct nf_conntrack_tuple *tuple) |
1156 | { | |
bae65be8 DM |
1157 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || |
1158 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) | |
1159 | goto nla_put_failure; | |
c1d10adb PNA |
1160 | return 0; |
1161 | ||
df6fb868 | 1162 | nla_put_failure: |
c1d10adb PNA |
1163 | return -1; |
1164 | } | |
fdf70832 | 1165 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
c1d10adb | 1166 | |
f73e924c PM |
1167 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
1168 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | |
1169 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | |
c1d10adb | 1170 | }; |
f73e924c | 1171 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
c1d10adb | 1172 | |
fdf70832 | 1173 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
c1d10adb PNA |
1174 | struct nf_conntrack_tuple *t) |
1175 | { | |
df6fb868 | 1176 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
c1d10adb PNA |
1177 | return -EINVAL; |
1178 | ||
77236b6e PM |
1179 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
1180 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | |
c1d10adb PNA |
1181 | |
1182 | return 0; | |
1183 | } | |
fdf70832 | 1184 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
5c0de29d HE |
1185 | |
1186 | int nf_ct_port_nlattr_tuple_size(void) | |
1187 | { | |
1188 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
1189 | } | |
1190 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | |
c1d10adb PNA |
1191 | #endif |
1192 | ||
9fb9cbb1 | 1193 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
b334aadc | 1194 | static void nf_conntrack_attach(struct sk_buff *nskb, struct sk_buff *skb) |
9fb9cbb1 YK |
1195 | { |
1196 | struct nf_conn *ct; | |
1197 | enum ip_conntrack_info ctinfo; | |
1198 | ||
1199 | /* This ICMP is in reverse direction to the packet which caused it */ | |
1200 | ct = nf_ct_get(skb, &ctinfo); | |
1201 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | |
fb048833 | 1202 | ctinfo = IP_CT_RELATED_REPLY; |
9fb9cbb1 YK |
1203 | else |
1204 | ctinfo = IP_CT_RELATED; | |
1205 | ||
1206 | /* Attach to new skbuff, and increment count */ | |
1207 | nskb->nfct = &ct->ct_general; | |
1208 | nskb->nfctinfo = ctinfo; | |
1209 | nf_conntrack_get(nskb->nfct); | |
1210 | } | |
1211 | ||
9fb9cbb1 | 1212 | /* Bring out ya dead! */ |
df0933dc | 1213 | static struct nf_conn * |
400dad39 | 1214 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
9fb9cbb1 YK |
1215 | void *data, unsigned int *bucket) |
1216 | { | |
df0933dc PM |
1217 | struct nf_conntrack_tuple_hash *h; |
1218 | struct nf_conn *ct; | |
ea781f19 | 1219 | struct hlist_nulls_node *n; |
9fb9cbb1 | 1220 | |
f8ba1aff | 1221 | spin_lock_bh(&nf_conntrack_lock); |
d696c7bd | 1222 | for (; *bucket < net->ct.htable_size; (*bucket)++) { |
ea781f19 | 1223 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { |
b0cdb1d9 PM |
1224 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) |
1225 | continue; | |
df0933dc PM |
1226 | ct = nf_ct_tuplehash_to_ctrack(h); |
1227 | if (iter(ct, data)) | |
1228 | goto found; | |
1229 | } | |
601e68e1 | 1230 | } |
ea781f19 | 1231 | hlist_nulls_for_each_entry(h, n, &net->ct.unconfirmed, hnnode) { |
df0933dc PM |
1232 | ct = nf_ct_tuplehash_to_ctrack(h); |
1233 | if (iter(ct, data)) | |
ec68e97d | 1234 | set_bit(IPS_DYING_BIT, &ct->status); |
df0933dc | 1235 | } |
f8ba1aff | 1236 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc PM |
1237 | return NULL; |
1238 | found: | |
c073e3fa | 1239 | atomic_inc(&ct->ct_general.use); |
f8ba1aff | 1240 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc | 1241 | return ct; |
9fb9cbb1 YK |
1242 | } |
1243 | ||
400dad39 AD |
1244 | void nf_ct_iterate_cleanup(struct net *net, |
1245 | int (*iter)(struct nf_conn *i, void *data), | |
1246 | void *data) | |
9fb9cbb1 | 1247 | { |
df0933dc | 1248 | struct nf_conn *ct; |
9fb9cbb1 YK |
1249 | unsigned int bucket = 0; |
1250 | ||
400dad39 | 1251 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
9fb9cbb1 YK |
1252 | /* Time to push up daises... */ |
1253 | if (del_timer(&ct->timeout)) | |
1254 | death_by_timeout((unsigned long)ct); | |
1255 | /* ... else the timer will get him soon. */ | |
1256 | ||
1257 | nf_ct_put(ct); | |
1258 | } | |
1259 | } | |
13b18339 | 1260 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
9fb9cbb1 | 1261 | |
19abb7b0 PNA |
1262 | struct __nf_ct_flush_report { |
1263 | u32 pid; | |
1264 | int report; | |
1265 | }; | |
1266 | ||
274d383b | 1267 | static int kill_report(struct nf_conn *i, void *data) |
9fb9cbb1 | 1268 | { |
19abb7b0 | 1269 | struct __nf_ct_flush_report *fr = (struct __nf_ct_flush_report *)data; |
a992ca2a PNA |
1270 | struct nf_conn_tstamp *tstamp; |
1271 | ||
1272 | tstamp = nf_conn_tstamp_find(i); | |
1273 | if (tstamp && tstamp->stop == 0) | |
1274 | tstamp->stop = ktime_to_ns(ktime_get_real()); | |
19abb7b0 | 1275 | |
dd7669a9 PNA |
1276 | /* If we fail to deliver the event, death_by_timeout() will retry */ |
1277 | if (nf_conntrack_event_report(IPCT_DESTROY, i, | |
1278 | fr->pid, fr->report) < 0) | |
1279 | return 1; | |
1280 | ||
1281 | /* Avoid the delivery of the destroy event in death_by_timeout(). */ | |
1282 | set_bit(IPS_DYING_BIT, &i->status); | |
9fb9cbb1 YK |
1283 | return 1; |
1284 | } | |
1285 | ||
274d383b PNA |
1286 | static int kill_all(struct nf_conn *i, void *data) |
1287 | { | |
1288 | return 1; | |
1289 | } | |
1290 | ||
d862a662 | 1291 | void nf_ct_free_hashtable(void *hash, unsigned int size) |
9fb9cbb1 | 1292 | { |
d862a662 | 1293 | if (is_vmalloc_addr(hash)) |
9fb9cbb1 YK |
1294 | vfree(hash); |
1295 | else | |
601e68e1 | 1296 | free_pages((unsigned long)hash, |
f205c5e0 | 1297 | get_order(sizeof(struct hlist_head) * size)); |
9fb9cbb1 | 1298 | } |
ac565e5f | 1299 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
9fb9cbb1 | 1300 | |
274d383b | 1301 | void nf_conntrack_flush_report(struct net *net, u32 pid, int report) |
c1d10adb | 1302 | { |
19abb7b0 PNA |
1303 | struct __nf_ct_flush_report fr = { |
1304 | .pid = pid, | |
1305 | .report = report, | |
1306 | }; | |
274d383b | 1307 | nf_ct_iterate_cleanup(net, kill_report, &fr); |
c1d10adb | 1308 | } |
274d383b | 1309 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); |
c1d10adb | 1310 | |
ee254fa4 | 1311 | static void nf_ct_release_dying_list(struct net *net) |
dd7669a9 PNA |
1312 | { |
1313 | struct nf_conntrack_tuple_hash *h; | |
1314 | struct nf_conn *ct; | |
1315 | struct hlist_nulls_node *n; | |
1316 | ||
1317 | spin_lock_bh(&nf_conntrack_lock); | |
ee254fa4 | 1318 | hlist_nulls_for_each_entry(h, n, &net->ct.dying, hnnode) { |
dd7669a9 PNA |
1319 | ct = nf_ct_tuplehash_to_ctrack(h); |
1320 | /* never fails to remove them, no listeners at this point */ | |
1321 | nf_ct_kill(ct); | |
1322 | } | |
1323 | spin_unlock_bh(&nf_conntrack_lock); | |
1324 | } | |
1325 | ||
b3c5163f ED |
1326 | static int untrack_refs(void) |
1327 | { | |
1328 | int cnt = 0, cpu; | |
1329 | ||
1330 | for_each_possible_cpu(cpu) { | |
1331 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
1332 | ||
1333 | cnt += atomic_read(&ct->ct_general.use) - 1; | |
1334 | } | |
1335 | return cnt; | |
1336 | } | |
1337 | ||
f94161c1 | 1338 | void nf_conntrack_cleanup_start(void) |
9fb9cbb1 | 1339 | { |
f94161c1 G |
1340 | RCU_INIT_POINTER(ip_ct_attach, NULL); |
1341 | } | |
1342 | ||
1343 | void nf_conntrack_cleanup_end(void) | |
1344 | { | |
1345 | RCU_INIT_POINTER(nf_ct_destroy, NULL); | |
b3c5163f | 1346 | while (untrack_refs() > 0) |
9edd7ca0 PM |
1347 | schedule(); |
1348 | ||
5d0aa2cc PM |
1349 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1350 | nf_ct_extend_unregister(&nf_ct_zone_extend); | |
1351 | #endif | |
04d87001 | 1352 | nf_conntrack_proto_fini(); |
5f69b8f5 | 1353 | nf_conntrack_labels_fini(); |
5e615b22 | 1354 | nf_conntrack_helper_fini(); |
8684094c | 1355 | nf_conntrack_timeout_fini(); |
3fe0f943 | 1356 | nf_conntrack_ecache_fini(); |
73f4001a | 1357 | nf_conntrack_tstamp_fini(); |
b7ff3a1f | 1358 | nf_conntrack_acct_fini(); |
83b4dbe1 | 1359 | nf_conntrack_expect_fini(); |
08f6547d | 1360 | } |
9fb9cbb1 | 1361 | |
f94161c1 G |
1362 | /* |
1363 | * Mishearing the voices in his head, our hero wonders how he's | |
1364 | * supposed to kill the mall. | |
1365 | */ | |
1366 | void nf_conntrack_cleanup_net(struct net *net) | |
08f6547d | 1367 | { |
f94161c1 G |
1368 | /* |
1369 | * This makes sure all current packets have passed through | |
1370 | * netfilter framework. Roll on, two-stage module | |
1371 | * delete... | |
1372 | */ | |
1373 | synchronize_net(); | |
9fb9cbb1 | 1374 | i_see_dead_people: |
274d383b | 1375 | nf_ct_iterate_cleanup(net, kill_all, NULL); |
ee254fa4 | 1376 | nf_ct_release_dying_list(net); |
49ac8713 | 1377 | if (atomic_read(&net->ct.count) != 0) { |
9fb9cbb1 YK |
1378 | schedule(); |
1379 | goto i_see_dead_people; | |
1380 | } | |
1381 | ||
d862a662 | 1382 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
04d87001 | 1383 | nf_conntrack_proto_pernet_fini(net); |
5e615b22 | 1384 | nf_conntrack_helper_pernet_fini(net); |
3fe0f943 | 1385 | nf_conntrack_ecache_pernet_fini(net); |
73f4001a | 1386 | nf_conntrack_tstamp_pernet_fini(net); |
b7ff3a1f | 1387 | nf_conntrack_acct_pernet_fini(net); |
83b4dbe1 | 1388 | nf_conntrack_expect_pernet_fini(net); |
5b3501fa ED |
1389 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
1390 | kfree(net->ct.slabname); | |
0d55af87 | 1391 | free_percpu(net->ct.stat); |
08f6547d AD |
1392 | } |
1393 | ||
d862a662 | 1394 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) |
9fb9cbb1 | 1395 | { |
ea781f19 ED |
1396 | struct hlist_nulls_head *hash; |
1397 | unsigned int nr_slots, i; | |
1398 | size_t sz; | |
9fb9cbb1 | 1399 | |
ea781f19 ED |
1400 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1401 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | |
1402 | sz = nr_slots * sizeof(struct hlist_nulls_head); | |
1403 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | |
1404 | get_order(sz)); | |
601e68e1 | 1405 | if (!hash) { |
9fb9cbb1 | 1406 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); |
966567b7 | 1407 | hash = vzalloc(sz); |
9fb9cbb1 YK |
1408 | } |
1409 | ||
ea781f19 ED |
1410 | if (hash && nulls) |
1411 | for (i = 0; i < nr_slots; i++) | |
1412 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | |
9fb9cbb1 YK |
1413 | |
1414 | return hash; | |
1415 | } | |
ac565e5f | 1416 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
9fb9cbb1 | 1417 | |
fae718dd | 1418 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
9fb9cbb1 | 1419 | { |
4b5511eb | 1420 | int i, bucket, rc; |
96eb24d7 | 1421 | unsigned int hashsize, old_size; |
ea781f19 | 1422 | struct hlist_nulls_head *hash, *old_hash; |
9fb9cbb1 | 1423 | struct nf_conntrack_tuple_hash *h; |
5d0aa2cc | 1424 | struct nf_conn *ct; |
9fb9cbb1 | 1425 | |
d696c7bd PM |
1426 | if (current->nsproxy->net_ns != &init_net) |
1427 | return -EOPNOTSUPP; | |
1428 | ||
9fb9cbb1 YK |
1429 | /* On boot, we can set this without any fancy locking. */ |
1430 | if (!nf_conntrack_htable_size) | |
1431 | return param_set_uint(val, kp); | |
1432 | ||
4b5511eb AP |
1433 | rc = kstrtouint(val, 0, &hashsize); |
1434 | if (rc) | |
1435 | return rc; | |
9fb9cbb1 YK |
1436 | if (!hashsize) |
1437 | return -EINVAL; | |
1438 | ||
d862a662 | 1439 | hash = nf_ct_alloc_hashtable(&hashsize, 1); |
9fb9cbb1 YK |
1440 | if (!hash) |
1441 | return -ENOMEM; | |
1442 | ||
76507f69 PM |
1443 | /* Lookups in the old hash might happen in parallel, which means we |
1444 | * might get false negatives during connection lookup. New connections | |
1445 | * created because of a false negative won't make it into the hash | |
1446 | * though since that required taking the lock. | |
1447 | */ | |
f8ba1aff | 1448 | spin_lock_bh(&nf_conntrack_lock); |
d696c7bd | 1449 | for (i = 0; i < init_net.ct.htable_size; i++) { |
ea781f19 ED |
1450 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
1451 | h = hlist_nulls_entry(init_net.ct.hash[i].first, | |
1452 | struct nf_conntrack_tuple_hash, hnnode); | |
5d0aa2cc | 1453 | ct = nf_ct_tuplehash_to_ctrack(h); |
ea781f19 | 1454 | hlist_nulls_del_rcu(&h->hnnode); |
5d0aa2cc | 1455 | bucket = __hash_conntrack(&h->tuple, nf_ct_zone(ct), |
99f07e91 | 1456 | hashsize); |
ea781f19 | 1457 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
9fb9cbb1 YK |
1458 | } |
1459 | } | |
d696c7bd | 1460 | old_size = init_net.ct.htable_size; |
400dad39 | 1461 | old_hash = init_net.ct.hash; |
9fb9cbb1 | 1462 | |
d696c7bd | 1463 | init_net.ct.htable_size = nf_conntrack_htable_size = hashsize; |
400dad39 | 1464 | init_net.ct.hash = hash; |
f8ba1aff | 1465 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1466 | |
d862a662 | 1467 | nf_ct_free_hashtable(old_hash, old_size); |
9fb9cbb1 YK |
1468 | return 0; |
1469 | } | |
fae718dd | 1470 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
9fb9cbb1 | 1471 | |
fae718dd | 1472 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
9fb9cbb1 YK |
1473 | &nf_conntrack_htable_size, 0600); |
1474 | ||
5bfddbd4 ED |
1475 | void nf_ct_untracked_status_or(unsigned long bits) |
1476 | { | |
b3c5163f ED |
1477 | int cpu; |
1478 | ||
1479 | for_each_possible_cpu(cpu) | |
1480 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; | |
5bfddbd4 ED |
1481 | } |
1482 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); | |
1483 | ||
f94161c1 | 1484 | int nf_conntrack_init_start(void) |
9fb9cbb1 | 1485 | { |
f205c5e0 | 1486 | int max_factor = 8; |
b3c5163f | 1487 | int ret, cpu; |
9fb9cbb1 YK |
1488 | |
1489 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB | |
f205c5e0 | 1490 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ |
9fb9cbb1 YK |
1491 | if (!nf_conntrack_htable_size) { |
1492 | nf_conntrack_htable_size | |
4481374c | 1493 | = (((totalram_pages << PAGE_SHIFT) / 16384) |
f205c5e0 | 1494 | / sizeof(struct hlist_head)); |
4481374c | 1495 | if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
f205c5e0 PM |
1496 | nf_conntrack_htable_size = 16384; |
1497 | if (nf_conntrack_htable_size < 32) | |
1498 | nf_conntrack_htable_size = 32; | |
1499 | ||
1500 | /* Use a max. factor of four by default to get the same max as | |
1501 | * with the old struct list_heads. When a table size is given | |
1502 | * we use the old value of 8 to avoid reducing the max. | |
1503 | * entries. */ | |
1504 | max_factor = 4; | |
9fb9cbb1 | 1505 | } |
f205c5e0 | 1506 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
8e5105a0 | 1507 | |
654d0fbd | 1508 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", |
8e5105a0 PM |
1509 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, |
1510 | nf_conntrack_max); | |
83b4dbe1 G |
1511 | |
1512 | ret = nf_conntrack_expect_init(); | |
1513 | if (ret < 0) | |
1514 | goto err_expect; | |
1515 | ||
b7ff3a1f G |
1516 | ret = nf_conntrack_acct_init(); |
1517 | if (ret < 0) | |
1518 | goto err_acct; | |
1519 | ||
73f4001a G |
1520 | ret = nf_conntrack_tstamp_init(); |
1521 | if (ret < 0) | |
1522 | goto err_tstamp; | |
1523 | ||
3fe0f943 G |
1524 | ret = nf_conntrack_ecache_init(); |
1525 | if (ret < 0) | |
1526 | goto err_ecache; | |
1527 | ||
8684094c G |
1528 | ret = nf_conntrack_timeout_init(); |
1529 | if (ret < 0) | |
1530 | goto err_timeout; | |
1531 | ||
5e615b22 G |
1532 | ret = nf_conntrack_helper_init(); |
1533 | if (ret < 0) | |
1534 | goto err_helper; | |
1535 | ||
5f69b8f5 G |
1536 | ret = nf_conntrack_labels_init(); |
1537 | if (ret < 0) | |
1538 | goto err_labels; | |
1539 | ||
5d0aa2cc PM |
1540 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1541 | ret = nf_ct_extend_register(&nf_ct_zone_extend); | |
1542 | if (ret < 0) | |
1543 | goto err_extend; | |
1544 | #endif | |
04d87001 G |
1545 | ret = nf_conntrack_proto_init(); |
1546 | if (ret < 0) | |
1547 | goto err_proto; | |
1548 | ||
9edd7ca0 | 1549 | /* Set up fake conntrack: to never be deleted, not in any hashes */ |
b3c5163f ED |
1550 | for_each_possible_cpu(cpu) { |
1551 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
b3c5163f ED |
1552 | write_pnet(&ct->ct_net, &init_net); |
1553 | atomic_set(&ct->ct_general.use, 1); | |
1554 | } | |
9edd7ca0 | 1555 | /* - and look it like as a confirmed connection */ |
5bfddbd4 | 1556 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); |
08f6547d AD |
1557 | return 0; |
1558 | ||
04d87001 | 1559 | err_proto: |
5d0aa2cc | 1560 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
04d87001 | 1561 | nf_ct_extend_unregister(&nf_ct_zone_extend); |
5d0aa2cc | 1562 | err_extend: |
a9006892 | 1563 | #endif |
04d87001 | 1564 | nf_conntrack_labels_fini(); |
5f69b8f5 G |
1565 | err_labels: |
1566 | nf_conntrack_helper_fini(); | |
5e615b22 G |
1567 | err_helper: |
1568 | nf_conntrack_timeout_fini(); | |
8684094c G |
1569 | err_timeout: |
1570 | nf_conntrack_ecache_fini(); | |
3fe0f943 G |
1571 | err_ecache: |
1572 | nf_conntrack_tstamp_fini(); | |
73f4001a G |
1573 | err_tstamp: |
1574 | nf_conntrack_acct_fini(); | |
b7ff3a1f G |
1575 | err_acct: |
1576 | nf_conntrack_expect_fini(); | |
83b4dbe1 | 1577 | err_expect: |
08f6547d AD |
1578 | return ret; |
1579 | } | |
1580 | ||
f94161c1 G |
1581 | void nf_conntrack_init_end(void) |
1582 | { | |
1583 | /* For use by REJECT target */ | |
1584 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); | |
1585 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); | |
1586 | ||
1587 | /* Howto get NAT offsets */ | |
1588 | RCU_INIT_POINTER(nf_ct_nat_offset, NULL); | |
1589 | } | |
1590 | ||
8cc20198 ED |
1591 | /* |
1592 | * We need to use special "null" values, not used in hash table | |
1593 | */ | |
1594 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) | |
1595 | #define DYING_NULLS_VAL ((1<<30)+1) | |
252b3e8c | 1596 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) |
8cc20198 | 1597 | |
f94161c1 | 1598 | int nf_conntrack_init_net(struct net *net) |
08f6547d AD |
1599 | { |
1600 | int ret; | |
ceceae1b | 1601 | |
08f6547d | 1602 | atomic_set(&net->ct.count, 0); |
8cc20198 ED |
1603 | INIT_HLIST_NULLS_HEAD(&net->ct.unconfirmed, UNCONFIRMED_NULLS_VAL); |
1604 | INIT_HLIST_NULLS_HEAD(&net->ct.dying, DYING_NULLS_VAL); | |
252b3e8c | 1605 | INIT_HLIST_NULLS_HEAD(&net->ct.tmpl, TEMPLATE_NULLS_VAL); |
08f6547d AD |
1606 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1607 | if (!net->ct.stat) { | |
1608 | ret = -ENOMEM; | |
1609 | goto err_stat; | |
1610 | } | |
5b3501fa ED |
1611 | |
1612 | net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net); | |
1613 | if (!net->ct.slabname) { | |
1614 | ret = -ENOMEM; | |
1615 | goto err_slabname; | |
1616 | } | |
1617 | ||
1618 | net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname, | |
1619 | sizeof(struct nf_conn), 0, | |
1620 | SLAB_DESTROY_BY_RCU, NULL); | |
1621 | if (!net->ct.nf_conntrack_cachep) { | |
1622 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); | |
1623 | ret = -ENOMEM; | |
1624 | goto err_cache; | |
1625 | } | |
d696c7bd PM |
1626 | |
1627 | net->ct.htable_size = nf_conntrack_htable_size; | |
d862a662 | 1628 | net->ct.hash = nf_ct_alloc_hashtable(&net->ct.htable_size, 1); |
08f6547d AD |
1629 | if (!net->ct.hash) { |
1630 | ret = -ENOMEM; | |
1631 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); | |
1632 | goto err_hash; | |
1633 | } | |
83b4dbe1 | 1634 | ret = nf_conntrack_expect_pernet_init(net); |
08f6547d AD |
1635 | if (ret < 0) |
1636 | goto err_expect; | |
b7ff3a1f | 1637 | ret = nf_conntrack_acct_pernet_init(net); |
58401572 | 1638 | if (ret < 0) |
08f6547d | 1639 | goto err_acct; |
73f4001a | 1640 | ret = nf_conntrack_tstamp_pernet_init(net); |
a992ca2a PNA |
1641 | if (ret < 0) |
1642 | goto err_tstamp; | |
3fe0f943 | 1643 | ret = nf_conntrack_ecache_pernet_init(net); |
a0891aa6 PNA |
1644 | if (ret < 0) |
1645 | goto err_ecache; | |
5e615b22 | 1646 | ret = nf_conntrack_helper_pernet_init(net); |
a9006892 EL |
1647 | if (ret < 0) |
1648 | goto err_helper; | |
04d87001 | 1649 | ret = nf_conntrack_proto_pernet_init(net); |
f94161c1 G |
1650 | if (ret < 0) |
1651 | goto err_proto; | |
08f6547d | 1652 | return 0; |
c539f017 | 1653 | |
f94161c1 | 1654 | err_proto: |
5e615b22 | 1655 | nf_conntrack_helper_pernet_fini(net); |
a9006892 | 1656 | err_helper: |
3fe0f943 | 1657 | nf_conntrack_ecache_pernet_fini(net); |
a0891aa6 | 1658 | err_ecache: |
73f4001a | 1659 | nf_conntrack_tstamp_pernet_fini(net); |
a992ca2a | 1660 | err_tstamp: |
b7ff3a1f | 1661 | nf_conntrack_acct_pernet_fini(net); |
08f6547d | 1662 | err_acct: |
83b4dbe1 | 1663 | nf_conntrack_expect_pernet_fini(net); |
08f6547d | 1664 | err_expect: |
d862a662 | 1665 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
6058fa6b | 1666 | err_hash: |
5b3501fa ED |
1667 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
1668 | err_cache: | |
1669 | kfree(net->ct.slabname); | |
1670 | err_slabname: | |
0d55af87 AD |
1671 | free_percpu(net->ct.stat); |
1672 | err_stat: | |
08f6547d AD |
1673 | return ret; |
1674 | } | |
1675 | ||
f9dd09c7 JK |
1676 | s16 (*nf_ct_nat_offset)(const struct nf_conn *ct, |
1677 | enum ip_conntrack_dir dir, | |
1678 | u32 seq); | |
1679 | EXPORT_SYMBOL_GPL(nf_ct_nat_offset); |