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