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 | 7 | * (C) 2003,2004 USAGI/WIDE Project <http://www.linux-ipv6.org> |
f229f6ce | 8 | * (C) 2005-2012 Patrick McHardy <kaber@trash.net> |
9fb9cbb1 YK |
9 | * |
10 | * This program is free software; you can redistribute it and/or modify | |
11 | * it under the terms of the GNU General Public License version 2 as | |
12 | * published by the Free Software Foundation. | |
9fb9cbb1 YK |
13 | */ |
14 | ||
9fb9cbb1 YK |
15 | #include <linux/types.h> |
16 | #include <linux/netfilter.h> | |
17 | #include <linux/module.h> | |
d43c36dc | 18 | #include <linux/sched.h> |
9fb9cbb1 YK |
19 | #include <linux/skbuff.h> |
20 | #include <linux/proc_fs.h> | |
21 | #include <linux/vmalloc.h> | |
22 | #include <linux/stddef.h> | |
23 | #include <linux/slab.h> | |
24 | #include <linux/random.h> | |
25 | #include <linux/jhash.h> | |
26 | #include <linux/err.h> | |
27 | #include <linux/percpu.h> | |
28 | #include <linux/moduleparam.h> | |
29 | #include <linux/notifier.h> | |
30 | #include <linux/kernel.h> | |
31 | #include <linux/netdevice.h> | |
32 | #include <linux/socket.h> | |
d7fe0f24 | 33 | #include <linux/mm.h> |
d696c7bd | 34 | #include <linux/nsproxy.h> |
ea781f19 | 35 | #include <linux/rculist_nulls.h> |
9fb9cbb1 | 36 | |
9fb9cbb1 YK |
37 | #include <net/netfilter/nf_conntrack.h> |
38 | #include <net/netfilter/nf_conntrack_l3proto.h> | |
605dcad6 | 39 | #include <net/netfilter/nf_conntrack_l4proto.h> |
77ab9cff | 40 | #include <net/netfilter/nf_conntrack_expect.h> |
9fb9cbb1 | 41 | #include <net/netfilter/nf_conntrack_helper.h> |
41d73ec0 | 42 | #include <net/netfilter/nf_conntrack_seqadj.h> |
9fb9cbb1 | 43 | #include <net/netfilter/nf_conntrack_core.h> |
ecfab2c9 | 44 | #include <net/netfilter/nf_conntrack_extend.h> |
58401572 | 45 | #include <net/netfilter/nf_conntrack_acct.h> |
a0891aa6 | 46 | #include <net/netfilter/nf_conntrack_ecache.h> |
5d0aa2cc | 47 | #include <net/netfilter/nf_conntrack_zones.h> |
a992ca2a | 48 | #include <net/netfilter/nf_conntrack_timestamp.h> |
dd705072 | 49 | #include <net/netfilter/nf_conntrack_timeout.h> |
c539f017 | 50 | #include <net/netfilter/nf_conntrack_labels.h> |
48b1de4c | 51 | #include <net/netfilter/nf_conntrack_synproxy.h> |
e6a7d3c0 | 52 | #include <net/netfilter/nf_nat.h> |
e17b666a | 53 | #include <net/netfilter/nf_nat_core.h> |
49376368 | 54 | #include <net/netfilter/nf_nat_helper.h> |
9fb9cbb1 | 55 | |
dc808fe2 | 56 | #define NF_CONNTRACK_VERSION "0.5.0" |
9fb9cbb1 | 57 | |
e17b666a PM |
58 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
59 | enum nf_nat_manip_type manip, | |
39938324 | 60 | const struct nlattr *attr) __read_mostly; |
e6a7d3c0 PNA |
61 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
62 | ||
f8ba1aff | 63 | DEFINE_SPINLOCK(nf_conntrack_lock); |
13b18339 | 64 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); |
9fb9cbb1 | 65 | |
e2b7606c | 66 | unsigned int nf_conntrack_htable_size __read_mostly; |
13b18339 PM |
67 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
68 | ||
e478075c | 69 | unsigned int nf_conntrack_max __read_mostly; |
a999e683 | 70 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
13b18339 | 71 | |
b3c5163f ED |
72 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); |
73 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); | |
13b18339 | 74 | |
f682cefa | 75 | unsigned int nf_conntrack_hash_rnd __read_mostly; |
4d4e61c6 | 76 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_rnd); |
9fb9cbb1 | 77 | |
99f07e91 | 78 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, u16 zone) |
9fb9cbb1 | 79 | { |
0794935e | 80 | unsigned int n; |
0794935e PM |
81 | |
82 | /* The direction must be ignored, so we hash everything up to the | |
83 | * destination ports (which is a multiple of 4) and treat the last | |
84 | * three bytes manually. | |
85 | */ | |
86 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); | |
99f07e91 CG |
87 | return jhash2((u32 *)tuple, n, zone ^ nf_conntrack_hash_rnd ^ |
88 | (((__force __u16)tuple->dst.u.all << 16) | | |
89 | tuple->dst.protonum)); | |
90 | } | |
91 | ||
92 | static u32 __hash_bucket(u32 hash, unsigned int size) | |
93 | { | |
94 | return ((u64)hash * size) >> 32; | |
95 | } | |
96 | ||
97 | static u32 hash_bucket(u32 hash, const struct net *net) | |
98 | { | |
99 | return __hash_bucket(hash, net->ct.htable_size); | |
100 | } | |
0794935e | 101 | |
99f07e91 CG |
102 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, |
103 | u16 zone, unsigned int size) | |
104 | { | |
105 | return __hash_bucket(hash_conntrack_raw(tuple, zone), size); | |
9fb9cbb1 YK |
106 | } |
107 | ||
5d0aa2cc | 108 | static inline u_int32_t hash_conntrack(const struct net *net, u16 zone, |
d696c7bd | 109 | const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 | 110 | { |
99f07e91 | 111 | return __hash_conntrack(tuple, zone, net->ct.htable_size); |
9fb9cbb1 YK |
112 | } |
113 | ||
5f2b4c90 | 114 | bool |
9fb9cbb1 YK |
115 | nf_ct_get_tuple(const struct sk_buff *skb, |
116 | unsigned int nhoff, | |
117 | unsigned int dataoff, | |
118 | u_int16_t l3num, | |
119 | u_int8_t protonum, | |
120 | struct nf_conntrack_tuple *tuple, | |
121 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 122 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 123 | { |
443a70d5 | 124 | memset(tuple, 0, sizeof(*tuple)); |
9fb9cbb1 YK |
125 | |
126 | tuple->src.l3num = l3num; | |
127 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | |
5f2b4c90 | 128 | return false; |
9fb9cbb1 YK |
129 | |
130 | tuple->dst.protonum = protonum; | |
131 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | |
132 | ||
605dcad6 | 133 | return l4proto->pkt_to_tuple(skb, dataoff, tuple); |
9fb9cbb1 | 134 | } |
13b18339 | 135 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
9fb9cbb1 | 136 | |
5f2b4c90 JE |
137 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
138 | u_int16_t l3num, struct nf_conntrack_tuple *tuple) | |
e2a3123f YK |
139 | { |
140 | struct nf_conntrack_l3proto *l3proto; | |
141 | struct nf_conntrack_l4proto *l4proto; | |
142 | unsigned int protoff; | |
143 | u_int8_t protonum; | |
144 | int ret; | |
145 | ||
146 | rcu_read_lock(); | |
147 | ||
148 | l3proto = __nf_ct_l3proto_find(l3num); | |
149 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | |
150 | if (ret != NF_ACCEPT) { | |
151 | rcu_read_unlock(); | |
5f2b4c90 | 152 | return false; |
e2a3123f YK |
153 | } |
154 | ||
155 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
156 | ||
157 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, tuple, | |
158 | l3proto, l4proto); | |
159 | ||
160 | rcu_read_unlock(); | |
161 | return ret; | |
162 | } | |
163 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | |
164 | ||
5f2b4c90 | 165 | bool |
9fb9cbb1 YK |
166 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
167 | const struct nf_conntrack_tuple *orig, | |
168 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 169 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 170 | { |
443a70d5 | 171 | memset(inverse, 0, sizeof(*inverse)); |
9fb9cbb1 YK |
172 | |
173 | inverse->src.l3num = orig->src.l3num; | |
174 | if (l3proto->invert_tuple(inverse, orig) == 0) | |
5f2b4c90 | 175 | return false; |
9fb9cbb1 YK |
176 | |
177 | inverse->dst.dir = !orig->dst.dir; | |
178 | ||
179 | inverse->dst.protonum = orig->dst.protonum; | |
605dcad6 | 180 | return l4proto->invert_tuple(inverse, orig); |
9fb9cbb1 | 181 | } |
13b18339 | 182 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
9fb9cbb1 | 183 | |
9fb9cbb1 YK |
184 | static void |
185 | clean_from_lists(struct nf_conn *ct) | |
186 | { | |
0d53778e | 187 | pr_debug("clean_from_lists(%p)\n", ct); |
ea781f19 ED |
188 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
189 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | |
9fb9cbb1 YK |
190 | |
191 | /* Destroy all pending expectations */ | |
c1d10adb | 192 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
193 | } |
194 | ||
195 | static void | |
196 | destroy_conntrack(struct nf_conntrack *nfct) | |
197 | { | |
198 | struct nf_conn *ct = (struct nf_conn *)nfct; | |
0d55af87 | 199 | struct net *net = nf_ct_net(ct); |
605dcad6 | 200 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 | 201 | |
0d53778e | 202 | pr_debug("destroy_conntrack(%p)\n", ct); |
9fb9cbb1 YK |
203 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
204 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); | |
205 | ||
9fb9cbb1 YK |
206 | /* To make sure we don't get any weird locking issues here: |
207 | * destroy_conntrack() MUST NOT be called with a write lock | |
208 | * to nf_conntrack_lock!!! -HW */ | |
923f4902 | 209 | rcu_read_lock(); |
5e8fbe2a | 210 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
605dcad6 MJ |
211 | if (l4proto && l4proto->destroy) |
212 | l4proto->destroy(ct); | |
9fb9cbb1 | 213 | |
982d9a9c | 214 | rcu_read_unlock(); |
9fb9cbb1 | 215 | |
f8ba1aff | 216 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
217 | /* Expectations will have been removed in clean_from_lists, |
218 | * except TFTP can create an expectation on the first packet, | |
219 | * before connection is in the list, so we need to clean here, | |
220 | * too. */ | |
c1d10adb | 221 | nf_ct_remove_expectations(ct); |
9fb9cbb1 | 222 | |
04dac011 PNA |
223 | /* We overload first tuple to link into unconfirmed or dying list.*/ |
224 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); | |
225 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | |
9fb9cbb1 | 226 | |
0d55af87 | 227 | NF_CT_STAT_INC(net, delete); |
f8ba1aff | 228 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
229 | |
230 | if (ct->master) | |
231 | nf_ct_put(ct->master); | |
232 | ||
0d53778e | 233 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
9fb9cbb1 YK |
234 | nf_conntrack_free(ct); |
235 | } | |
236 | ||
02982c27 | 237 | static void nf_ct_delete_from_lists(struct nf_conn *ct) |
9fb9cbb1 | 238 | { |
0d55af87 | 239 | struct net *net = nf_ct_net(ct); |
9fb9cbb1 | 240 | |
9858a3ae | 241 | nf_ct_helper_destroy(ct); |
f8ba1aff | 242 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
243 | /* Inside lock so preempt is disabled on module removal path. |
244 | * Otherwise we can get spurious warnings. */ | |
0d55af87 | 245 | NF_CT_STAT_INC(net, delete_list); |
9fb9cbb1 | 246 | clean_from_lists(ct); |
04dac011 PNA |
247 | /* add this conntrack to the dying list */ |
248 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
249 | &net->ct.dying); | |
f8ba1aff | 250 | spin_unlock_bh(&nf_conntrack_lock); |
dd7669a9 | 251 | } |
dd7669a9 PNA |
252 | |
253 | static void death_by_event(unsigned long ul_conntrack) | |
254 | { | |
255 | struct nf_conn *ct = (void *)ul_conntrack; | |
256 | struct net *net = nf_ct_net(ct); | |
5b423f6a PNA |
257 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); |
258 | ||
259 | BUG_ON(ecache == NULL); | |
dd7669a9 PNA |
260 | |
261 | if (nf_conntrack_event(IPCT_DESTROY, ct) < 0) { | |
262 | /* bad luck, let's retry again */ | |
5b423f6a | 263 | ecache->timeout.expires = jiffies + |
ca3d41a5 | 264 | (prandom_u32() % net->ct.sysctl_events_retry_timeout); |
5b423f6a | 265 | add_timer(&ecache->timeout); |
dd7669a9 PNA |
266 | return; |
267 | } | |
268 | /* we've got the event delivered, now it's dying */ | |
269 | set_bit(IPS_DYING_BIT, &ct->status); | |
dd7669a9 PNA |
270 | nf_ct_put(ct); |
271 | } | |
272 | ||
02982c27 | 273 | static void nf_ct_dying_timeout(struct nf_conn *ct) |
dd7669a9 PNA |
274 | { |
275 | struct net *net = nf_ct_net(ct); | |
5b423f6a PNA |
276 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); |
277 | ||
278 | BUG_ON(ecache == NULL); | |
dd7669a9 | 279 | |
dd7669a9 | 280 | /* set a new timer to retry event delivery */ |
5b423f6a PNA |
281 | setup_timer(&ecache->timeout, death_by_event, (unsigned long)ct); |
282 | ecache->timeout.expires = jiffies + | |
ca3d41a5 | 283 | (prandom_u32() % net->ct.sysctl_events_retry_timeout); |
5b423f6a | 284 | add_timer(&ecache->timeout); |
dd7669a9 | 285 | } |
dd7669a9 | 286 | |
02982c27 | 287 | bool nf_ct_delete(struct nf_conn *ct, u32 portid, int report) |
dd7669a9 | 288 | { |
a992ca2a PNA |
289 | struct nf_conn_tstamp *tstamp; |
290 | ||
291 | tstamp = nf_conn_tstamp_find(ct); | |
292 | if (tstamp && tstamp->stop == 0) | |
293 | tstamp->stop = ktime_to_ns(ktime_get_real()); | |
dd7669a9 | 294 | |
02982c27 FW |
295 | if (!nf_ct_is_dying(ct) && |
296 | unlikely(nf_conntrack_event_report(IPCT_DESTROY, ct, | |
297 | portid, report) < 0)) { | |
dd7669a9 PNA |
298 | /* destroy event was not delivered */ |
299 | nf_ct_delete_from_lists(ct); | |
04dac011 | 300 | nf_ct_dying_timeout(ct); |
02982c27 | 301 | return false; |
dd7669a9 PNA |
302 | } |
303 | set_bit(IPS_DYING_BIT, &ct->status); | |
304 | nf_ct_delete_from_lists(ct); | |
9fb9cbb1 | 305 | nf_ct_put(ct); |
02982c27 FW |
306 | return true; |
307 | } | |
308 | EXPORT_SYMBOL_GPL(nf_ct_delete); | |
309 | ||
310 | static void death_by_timeout(unsigned long ul_conntrack) | |
311 | { | |
312 | nf_ct_delete((struct nf_conn *)ul_conntrack, 0, 0); | |
9fb9cbb1 YK |
313 | } |
314 | ||
c6825c09 AV |
315 | static inline bool |
316 | nf_ct_key_equal(struct nf_conntrack_tuple_hash *h, | |
317 | const struct nf_conntrack_tuple *tuple, | |
318 | u16 zone) | |
319 | { | |
320 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | |
321 | ||
322 | /* A conntrack can be recreated with the equal tuple, | |
323 | * so we need to check that the conntrack is confirmed | |
324 | */ | |
325 | return nf_ct_tuple_equal(tuple, &h->tuple) && | |
326 | nf_ct_zone(ct) == zone && | |
327 | nf_ct_is_confirmed(ct); | |
328 | } | |
329 | ||
ea781f19 ED |
330 | /* |
331 | * Warning : | |
332 | * - Caller must take a reference on returned object | |
333 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | |
334 | * OR | |
335 | * - Caller must lock nf_conntrack_lock before calling this function | |
336 | */ | |
99f07e91 CG |
337 | static struct nf_conntrack_tuple_hash * |
338 | ____nf_conntrack_find(struct net *net, u16 zone, | |
339 | const struct nf_conntrack_tuple *tuple, u32 hash) | |
9fb9cbb1 YK |
340 | { |
341 | struct nf_conntrack_tuple_hash *h; | |
ea781f19 | 342 | struct hlist_nulls_node *n; |
99f07e91 | 343 | unsigned int bucket = hash_bucket(hash, net); |
9fb9cbb1 | 344 | |
4e29e9ec PM |
345 | /* Disable BHs the entire time since we normally need to disable them |
346 | * at least once for the stats anyway. | |
347 | */ | |
348 | local_bh_disable(); | |
ea781f19 | 349 | begin: |
99f07e91 | 350 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[bucket], hnnode) { |
c6825c09 | 351 | if (nf_ct_key_equal(h, tuple, zone)) { |
0d55af87 | 352 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 353 | local_bh_enable(); |
9fb9cbb1 YK |
354 | return h; |
355 | } | |
0d55af87 | 356 | NF_CT_STAT_INC(net, searched); |
9fb9cbb1 | 357 | } |
ea781f19 ED |
358 | /* |
359 | * if the nulls value we got at the end of this lookup is | |
360 | * not the expected one, we must restart lookup. | |
361 | * We probably met an item that was moved to another chain. | |
362 | */ | |
99f07e91 | 363 | if (get_nulls_value(n) != bucket) { |
af740b2c | 364 | NF_CT_STAT_INC(net, search_restart); |
ea781f19 | 365 | goto begin; |
af740b2c | 366 | } |
4e29e9ec | 367 | local_bh_enable(); |
9fb9cbb1 YK |
368 | |
369 | return NULL; | |
370 | } | |
99f07e91 | 371 | |
9fb9cbb1 | 372 | /* Find a connection corresponding to a tuple. */ |
99f07e91 CG |
373 | static struct nf_conntrack_tuple_hash * |
374 | __nf_conntrack_find_get(struct net *net, u16 zone, | |
375 | const struct nf_conntrack_tuple *tuple, u32 hash) | |
9fb9cbb1 YK |
376 | { |
377 | struct nf_conntrack_tuple_hash *h; | |
76507f69 | 378 | struct nf_conn *ct; |
9fb9cbb1 | 379 | |
76507f69 | 380 | rcu_read_lock(); |
ea781f19 | 381 | begin: |
99f07e91 | 382 | h = ____nf_conntrack_find(net, zone, tuple, hash); |
76507f69 PM |
383 | if (h) { |
384 | ct = nf_ct_tuplehash_to_ctrack(h); | |
8d8890b7 PM |
385 | if (unlikely(nf_ct_is_dying(ct) || |
386 | !atomic_inc_not_zero(&ct->ct_general.use))) | |
76507f69 | 387 | h = NULL; |
ea781f19 | 388 | else { |
c6825c09 | 389 | if (unlikely(!nf_ct_key_equal(h, tuple, zone))) { |
ea781f19 ED |
390 | nf_ct_put(ct); |
391 | goto begin; | |
392 | } | |
393 | } | |
76507f69 PM |
394 | } |
395 | rcu_read_unlock(); | |
9fb9cbb1 YK |
396 | |
397 | return h; | |
398 | } | |
99f07e91 CG |
399 | |
400 | struct nf_conntrack_tuple_hash * | |
401 | nf_conntrack_find_get(struct net *net, u16 zone, | |
402 | const struct nf_conntrack_tuple *tuple) | |
403 | { | |
404 | return __nf_conntrack_find_get(net, zone, tuple, | |
405 | hash_conntrack_raw(tuple, zone)); | |
406 | } | |
13b18339 | 407 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
9fb9cbb1 | 408 | |
c1d10adb PNA |
409 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
410 | unsigned int hash, | |
601e68e1 | 411 | unsigned int repl_hash) |
c1d10adb | 412 | { |
400dad39 AD |
413 | struct net *net = nf_ct_net(ct); |
414 | ||
ea781f19 | 415 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
400dad39 | 416 | &net->ct.hash[hash]); |
ea781f19 | 417 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
400dad39 | 418 | &net->ct.hash[repl_hash]); |
c1d10adb PNA |
419 | } |
420 | ||
7d367e06 JK |
421 | int |
422 | nf_conntrack_hash_check_insert(struct nf_conn *ct) | |
c1d10adb | 423 | { |
d696c7bd | 424 | struct net *net = nf_ct_net(ct); |
c1d10adb | 425 | unsigned int hash, repl_hash; |
7d367e06 JK |
426 | struct nf_conntrack_tuple_hash *h; |
427 | struct hlist_nulls_node *n; | |
5d0aa2cc | 428 | u16 zone; |
c1d10adb | 429 | |
5d0aa2cc | 430 | zone = nf_ct_zone(ct); |
7d367e06 JK |
431 | hash = hash_conntrack(net, zone, |
432 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
433 | repl_hash = hash_conntrack(net, zone, | |
434 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
435 | ||
436 | spin_lock_bh(&nf_conntrack_lock); | |
437 | ||
438 | /* See if there's one in the list already, including reverse */ | |
439 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) | |
440 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | |
441 | &h->tuple) && | |
442 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
443 | goto out; | |
444 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) | |
445 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, | |
446 | &h->tuple) && | |
447 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
448 | goto out; | |
c1d10adb | 449 | |
7d367e06 | 450 | add_timer(&ct->timeout); |
e53376be PNA |
451 | smp_wmb(); |
452 | /* The caller holds a reference to this object */ | |
453 | atomic_set(&ct->ct_general.use, 2); | |
c1d10adb | 454 | __nf_conntrack_hash_insert(ct, hash, repl_hash); |
7d367e06 JK |
455 | NF_CT_STAT_INC(net, insert); |
456 | spin_unlock_bh(&nf_conntrack_lock); | |
457 | ||
458 | return 0; | |
459 | ||
460 | out: | |
461 | NF_CT_STAT_INC(net, insert_failed); | |
462 | spin_unlock_bh(&nf_conntrack_lock); | |
463 | return -EEXIST; | |
c1d10adb | 464 | } |
7d367e06 | 465 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); |
c1d10adb | 466 | |
e53376be PNA |
467 | /* deletion from this larval template list happens via nf_ct_put() */ |
468 | void nf_conntrack_tmpl_insert(struct net *net, struct nf_conn *tmpl) | |
469 | { | |
470 | __set_bit(IPS_TEMPLATE_BIT, &tmpl->status); | |
471 | __set_bit(IPS_CONFIRMED_BIT, &tmpl->status); | |
472 | nf_conntrack_get(&tmpl->ct_general); | |
473 | ||
474 | spin_lock_bh(&nf_conntrack_lock); | |
475 | /* Overload tuple linked list to put us in template list. */ | |
476 | hlist_nulls_add_head_rcu(&tmpl->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
477 | &net->ct.tmpl); | |
478 | spin_unlock_bh(&nf_conntrack_lock); | |
479 | } | |
480 | EXPORT_SYMBOL_GPL(nf_conntrack_tmpl_insert); | |
481 | ||
9fb9cbb1 YK |
482 | /* Confirm a connection given skb; places it in hash table */ |
483 | int | |
3db05fea | 484 | __nf_conntrack_confirm(struct sk_buff *skb) |
9fb9cbb1 YK |
485 | { |
486 | unsigned int hash, repl_hash; | |
df0933dc | 487 | struct nf_conntrack_tuple_hash *h; |
9fb9cbb1 | 488 | struct nf_conn *ct; |
df0933dc | 489 | struct nf_conn_help *help; |
a992ca2a | 490 | struct nf_conn_tstamp *tstamp; |
ea781f19 | 491 | struct hlist_nulls_node *n; |
9fb9cbb1 | 492 | enum ip_conntrack_info ctinfo; |
400dad39 | 493 | struct net *net; |
5d0aa2cc | 494 | u16 zone; |
9fb9cbb1 | 495 | |
3db05fea | 496 | ct = nf_ct_get(skb, &ctinfo); |
400dad39 | 497 | net = nf_ct_net(ct); |
9fb9cbb1 YK |
498 | |
499 | /* ipt_REJECT uses nf_conntrack_attach to attach related | |
500 | ICMP/TCP RST packets in other direction. Actual packet | |
501 | which created connection will be IP_CT_NEW or for an | |
502 | expected connection, IP_CT_RELATED. */ | |
503 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | |
504 | return NF_ACCEPT; | |
505 | ||
5d0aa2cc | 506 | zone = nf_ct_zone(ct); |
99f07e91 CG |
507 | /* reuse the hash saved before */ |
508 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; | |
509 | hash = hash_bucket(hash, net); | |
510 | repl_hash = hash_conntrack(net, zone, | |
511 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); | |
9fb9cbb1 YK |
512 | |
513 | /* We're not in hash table, and we refuse to set up related | |
514 | connections for unconfirmed conns. But packet copies and | |
515 | REJECT will give spurious warnings here. */ | |
516 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ | |
517 | ||
25985edc | 518 | /* No external references means no one else could have |
9fb9cbb1 YK |
519 | confirmed us. */ |
520 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
0d53778e | 521 | pr_debug("Confirming conntrack %p\n", ct); |
9fb9cbb1 | 522 | |
f8ba1aff | 523 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 524 | |
fc350777 JM |
525 | /* We have to check the DYING flag inside the lock to prevent |
526 | a race against nf_ct_get_next_corpse() possibly called from | |
527 | user context, else we insert an already 'dead' hash, blocking | |
528 | further use of that particular connection -JM */ | |
529 | ||
530 | if (unlikely(nf_ct_is_dying(ct))) { | |
531 | spin_unlock_bh(&nf_conntrack_lock); | |
532 | return NF_ACCEPT; | |
533 | } | |
534 | ||
9fb9cbb1 YK |
535 | /* See if there's one in the list already, including reverse: |
536 | NAT could have grabbed it without realizing, since we're | |
537 | not in the hash. If there is, we lost race. */ | |
ea781f19 | 538 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
df0933dc | 539 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
5d0aa2cc PM |
540 | &h->tuple) && |
541 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
df0933dc | 542 | goto out; |
ea781f19 | 543 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[repl_hash], hnnode) |
df0933dc | 544 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
5d0aa2cc PM |
545 | &h->tuple) && |
546 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
df0933dc | 547 | goto out; |
9fb9cbb1 | 548 | |
df0933dc | 549 | /* Remove from unconfirmed list */ |
ea781f19 | 550 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
df0933dc | 551 | |
df0933dc PM |
552 | /* Timer relative to confirmation time, not original |
553 | setting time, otherwise we'd get timer wrap in | |
554 | weird delay cases. */ | |
555 | ct->timeout.expires += jiffies; | |
556 | add_timer(&ct->timeout); | |
557 | atomic_inc(&ct->ct_general.use); | |
45eec341 | 558 | ct->status |= IPS_CONFIRMED; |
5c8ec910 | 559 | |
a992ca2a PNA |
560 | /* set conntrack timestamp, if enabled. */ |
561 | tstamp = nf_conn_tstamp_find(ct); | |
562 | if (tstamp) { | |
563 | if (skb->tstamp.tv64 == 0) | |
e3192690 | 564 | __net_timestamp(skb); |
a992ca2a PNA |
565 | |
566 | tstamp->start = ktime_to_ns(skb->tstamp); | |
567 | } | |
5c8ec910 PM |
568 | /* Since the lookup is lockless, hash insertion must be done after |
569 | * starting the timer and setting the CONFIRMED bit. The RCU barriers | |
570 | * guarantee that no other CPU can find the conntrack before the above | |
571 | * stores are visible. | |
572 | */ | |
573 | __nf_conntrack_hash_insert(ct, hash, repl_hash); | |
0d55af87 | 574 | NF_CT_STAT_INC(net, insert); |
f8ba1aff | 575 | spin_unlock_bh(&nf_conntrack_lock); |
5c8ec910 | 576 | |
df0933dc PM |
577 | help = nfct_help(ct); |
578 | if (help && help->helper) | |
a71996fc | 579 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
17e6e4ea | 580 | |
df0933dc | 581 | nf_conntrack_event_cache(master_ct(ct) ? |
a71996fc | 582 | IPCT_RELATED : IPCT_NEW, ct); |
df0933dc | 583 | return NF_ACCEPT; |
9fb9cbb1 | 584 | |
df0933dc | 585 | out: |
0d55af87 | 586 | NF_CT_STAT_INC(net, insert_failed); |
f8ba1aff | 587 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
588 | return NF_DROP; |
589 | } | |
13b18339 | 590 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
9fb9cbb1 YK |
591 | |
592 | /* Returns true if a connection correspondings to the tuple (required | |
593 | for NAT). */ | |
594 | int | |
595 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | |
596 | const struct nf_conn *ignored_conntrack) | |
597 | { | |
400dad39 | 598 | struct net *net = nf_ct_net(ignored_conntrack); |
9fb9cbb1 | 599 | struct nf_conntrack_tuple_hash *h; |
ea781f19 | 600 | struct hlist_nulls_node *n; |
5d0aa2cc PM |
601 | struct nf_conn *ct; |
602 | u16 zone = nf_ct_zone(ignored_conntrack); | |
603 | unsigned int hash = hash_conntrack(net, zone, tuple); | |
9fb9cbb1 | 604 | |
4e29e9ec PM |
605 | /* Disable BHs the entire time since we need to disable them at |
606 | * least once for the stats anyway. | |
607 | */ | |
608 | rcu_read_lock_bh(); | |
ea781f19 | 609 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
5d0aa2cc PM |
610 | ct = nf_ct_tuplehash_to_ctrack(h); |
611 | if (ct != ignored_conntrack && | |
612 | nf_ct_tuple_equal(tuple, &h->tuple) && | |
613 | nf_ct_zone(ct) == zone) { | |
0d55af87 | 614 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 615 | rcu_read_unlock_bh(); |
ba419aff PM |
616 | return 1; |
617 | } | |
0d55af87 | 618 | NF_CT_STAT_INC(net, searched); |
ba419aff | 619 | } |
4e29e9ec | 620 | rcu_read_unlock_bh(); |
9fb9cbb1 | 621 | |
ba419aff | 622 | return 0; |
9fb9cbb1 | 623 | } |
13b18339 | 624 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
9fb9cbb1 | 625 | |
7ae7730f PM |
626 | #define NF_CT_EVICTION_RANGE 8 |
627 | ||
9fb9cbb1 YK |
628 | /* There's a small race here where we may free a just-assured |
629 | connection. Too bad: we're in trouble anyway. */ | |
400dad39 | 630 | static noinline int early_drop(struct net *net, unsigned int hash) |
9fb9cbb1 | 631 | { |
f205c5e0 | 632 | /* Use oldest entry, which is roughly LRU */ |
9fb9cbb1 | 633 | struct nf_conntrack_tuple_hash *h; |
df0933dc | 634 | struct nf_conn *ct = NULL, *tmp; |
ea781f19 | 635 | struct hlist_nulls_node *n; |
7ae7730f | 636 | unsigned int i, cnt = 0; |
9fb9cbb1 YK |
637 | int dropped = 0; |
638 | ||
76507f69 | 639 | rcu_read_lock(); |
d696c7bd | 640 | for (i = 0; i < net->ct.htable_size; i++) { |
ea781f19 ED |
641 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
642 | hnnode) { | |
7ae7730f PM |
643 | tmp = nf_ct_tuplehash_to_ctrack(h); |
644 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) | |
645 | ct = tmp; | |
646 | cnt++; | |
647 | } | |
76507f69 | 648 | |
5ae27aa2 CG |
649 | if (ct != NULL) { |
650 | if (likely(!nf_ct_is_dying(ct) && | |
651 | atomic_inc_not_zero(&ct->ct_general.use))) | |
652 | break; | |
653 | else | |
654 | ct = NULL; | |
655 | } | |
656 | ||
657 | if (cnt >= NF_CT_EVICTION_RANGE) | |
7ae7730f | 658 | break; |
5ae27aa2 | 659 | |
d696c7bd | 660 | hash = (hash + 1) % net->ct.htable_size; |
9fb9cbb1 | 661 | } |
76507f69 | 662 | rcu_read_unlock(); |
9fb9cbb1 YK |
663 | |
664 | if (!ct) | |
665 | return dropped; | |
666 | ||
667 | if (del_timer(&ct->timeout)) { | |
02982c27 | 668 | if (nf_ct_delete(ct, 0, 0)) { |
74138511 PNA |
669 | dropped = 1; |
670 | NF_CT_STAT_INC_ATOMIC(net, early_drop); | |
671 | } | |
9fb9cbb1 YK |
672 | } |
673 | nf_ct_put(ct); | |
674 | return dropped; | |
675 | } | |
676 | ||
f682cefa CG |
677 | void init_nf_conntrack_hash_rnd(void) |
678 | { | |
679 | unsigned int rand; | |
680 | ||
681 | /* | |
682 | * Why not initialize nf_conntrack_rnd in a "init()" function ? | |
683 | * Because there isn't enough entropy when system initializing, | |
684 | * and we initialize it as late as possible. | |
685 | */ | |
686 | do { | |
687 | get_random_bytes(&rand, sizeof(rand)); | |
688 | } while (!rand); | |
689 | cmpxchg(&nf_conntrack_hash_rnd, 0, rand); | |
690 | } | |
691 | ||
99f07e91 CG |
692 | static struct nf_conn * |
693 | __nf_conntrack_alloc(struct net *net, u16 zone, | |
694 | const struct nf_conntrack_tuple *orig, | |
695 | const struct nf_conntrack_tuple *repl, | |
696 | gfp_t gfp, u32 hash) | |
9fb9cbb1 | 697 | { |
cd7fcbf1 | 698 | struct nf_conn *ct; |
9fb9cbb1 | 699 | |
b2390969 | 700 | if (unlikely(!nf_conntrack_hash_rnd)) { |
f682cefa | 701 | init_nf_conntrack_hash_rnd(); |
99f07e91 CG |
702 | /* recompute the hash as nf_conntrack_hash_rnd is initialized */ |
703 | hash = hash_conntrack_raw(orig, zone); | |
9fb9cbb1 YK |
704 | } |
705 | ||
5251e2d2 | 706 | /* We don't want any race condition at early drop stage */ |
49ac8713 | 707 | atomic_inc(&net->ct.count); |
5251e2d2 | 708 | |
76eb9460 | 709 | if (nf_conntrack_max && |
49ac8713 | 710 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
99f07e91 | 711 | if (!early_drop(net, hash_bucket(hash, net))) { |
49ac8713 | 712 | atomic_dec(&net->ct.count); |
e87cc472 | 713 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); |
9fb9cbb1 YK |
714 | return ERR_PTR(-ENOMEM); |
715 | } | |
716 | } | |
717 | ||
941297f4 ED |
718 | /* |
719 | * Do not use kmem_cache_zalloc(), as this cache uses | |
720 | * SLAB_DESTROY_BY_RCU. | |
721 | */ | |
5b3501fa | 722 | ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp); |
c88130bc | 723 | if (ct == NULL) { |
49ac8713 | 724 | atomic_dec(&net->ct.count); |
dacd2a1a | 725 | return ERR_PTR(-ENOMEM); |
9fb9cbb1 | 726 | } |
941297f4 ED |
727 | /* |
728 | * Let ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.next | |
729 | * and ct->tuplehash[IP_CT_DIR_REPLY].hnnode.next unchanged. | |
730 | */ | |
731 | memset(&ct->tuplehash[IP_CT_DIR_MAX], 0, | |
e5fc9e7a CG |
732 | offsetof(struct nf_conn, proto) - |
733 | offsetof(struct nf_conn, tuplehash[IP_CT_DIR_MAX])); | |
440f0d58 | 734 | spin_lock_init(&ct->lock); |
c88130bc | 735 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; |
941297f4 | 736 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; |
c88130bc | 737 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; |
99f07e91 CG |
738 | /* save hash for reusing when confirming */ |
739 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; | |
9fb9cbb1 | 740 | /* Don't set timer yet: wait for confirmation */ |
c88130bc | 741 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
c2d9ba9b | 742 | write_pnet(&ct->ct_net, net); |
5d0aa2cc PM |
743 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
744 | if (zone) { | |
745 | struct nf_conntrack_zone *nf_ct_zone; | |
746 | ||
747 | nf_ct_zone = nf_ct_ext_add(ct, NF_CT_EXT_ZONE, GFP_ATOMIC); | |
748 | if (!nf_ct_zone) | |
749 | goto out_free; | |
750 | nf_ct_zone->id = zone; | |
751 | } | |
752 | #endif | |
e53376be PNA |
753 | /* Because we use RCU lookups, we set ct_general.use to zero before |
754 | * this is inserted in any list. | |
941297f4 | 755 | */ |
e53376be | 756 | atomic_set(&ct->ct_general.use, 0); |
c88130bc | 757 | return ct; |
5d0aa2cc PM |
758 | |
759 | #ifdef CONFIG_NF_CONNTRACK_ZONES | |
760 | out_free: | |
d96fc659 | 761 | atomic_dec(&net->ct.count); |
5d0aa2cc PM |
762 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
763 | return ERR_PTR(-ENOMEM); | |
764 | #endif | |
9fb9cbb1 | 765 | } |
99f07e91 CG |
766 | |
767 | struct nf_conn *nf_conntrack_alloc(struct net *net, u16 zone, | |
768 | const struct nf_conntrack_tuple *orig, | |
769 | const struct nf_conntrack_tuple *repl, | |
770 | gfp_t gfp) | |
771 | { | |
772 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); | |
773 | } | |
13b18339 | 774 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
9fb9cbb1 | 775 | |
c88130bc | 776 | void nf_conntrack_free(struct nf_conn *ct) |
76507f69 | 777 | { |
1d45209d ED |
778 | struct net *net = nf_ct_net(ct); |
779 | ||
e53376be PNA |
780 | /* A freed object has refcnt == 0, that's |
781 | * the golden rule for SLAB_DESTROY_BY_RCU | |
782 | */ | |
783 | NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 0); | |
784 | ||
ceeff754 | 785 | nf_ct_ext_destroy(ct); |
ea781f19 | 786 | nf_ct_ext_free(ct); |
5b3501fa | 787 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
0c3c6c00 PNA |
788 | smp_mb__before_atomic_dec(); |
789 | atomic_dec(&net->ct.count); | |
76507f69 | 790 | } |
13b18339 | 791 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
9fb9cbb1 | 792 | |
c539f017 | 793 | |
9fb9cbb1 YK |
794 | /* Allocate a new conntrack: we return -ENOMEM if classification |
795 | failed due to stress. Otherwise it really is unclassifiable. */ | |
796 | static struct nf_conntrack_tuple_hash * | |
b2a15a60 | 797 | init_conntrack(struct net *net, struct nf_conn *tmpl, |
5a1fb391 | 798 | const struct nf_conntrack_tuple *tuple, |
9fb9cbb1 | 799 | struct nf_conntrack_l3proto *l3proto, |
605dcad6 | 800 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 801 | struct sk_buff *skb, |
60b5f8f7 | 802 | unsigned int dataoff, u32 hash) |
9fb9cbb1 | 803 | { |
c88130bc | 804 | struct nf_conn *ct; |
3c158f7f | 805 | struct nf_conn_help *help; |
9fb9cbb1 | 806 | struct nf_conntrack_tuple repl_tuple; |
b2a15a60 | 807 | struct nf_conntrack_ecache *ecache; |
9fb9cbb1 | 808 | struct nf_conntrack_expect *exp; |
5d0aa2cc | 809 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
60b5f8f7 PNA |
810 | struct nf_conn_timeout *timeout_ext; |
811 | unsigned int *timeouts; | |
9fb9cbb1 | 812 | |
605dcad6 | 813 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
0d53778e | 814 | pr_debug("Can't invert tuple.\n"); |
9fb9cbb1 YK |
815 | return NULL; |
816 | } | |
817 | ||
99f07e91 CG |
818 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, |
819 | hash); | |
0a9ee813 | 820 | if (IS_ERR(ct)) |
c88130bc | 821 | return (struct nf_conntrack_tuple_hash *)ct; |
9fb9cbb1 | 822 | |
48b1de4c PM |
823 | if (tmpl && nfct_synproxy(tmpl)) { |
824 | nfct_seqadj_ext_add(ct); | |
825 | nfct_synproxy_ext_add(ct); | |
826 | } | |
827 | ||
60b5f8f7 PNA |
828 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; |
829 | if (timeout_ext) | |
830 | timeouts = NF_CT_TIMEOUT_EXT_DATA(timeout_ext); | |
831 | else | |
832 | timeouts = l4proto->get_timeouts(net); | |
833 | ||
2c8503f5 | 834 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { |
c88130bc | 835 | nf_conntrack_free(ct); |
0d53778e | 836 | pr_debug("init conntrack: can't track with proto module\n"); |
9fb9cbb1 YK |
837 | return NULL; |
838 | } | |
839 | ||
60b5f8f7 PNA |
840 | if (timeout_ext) |
841 | nf_ct_timeout_ext_add(ct, timeout_ext->timeout, GFP_ATOMIC); | |
842 | ||
58401572 | 843 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
a992ca2a | 844 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); |
c539f017 | 845 | nf_ct_labels_ext_add(ct); |
b2a15a60 PM |
846 | |
847 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; | |
848 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, | |
849 | ecache ? ecache->expmask : 0, | |
850 | GFP_ATOMIC); | |
58401572 | 851 | |
f8ba1aff | 852 | spin_lock_bh(&nf_conntrack_lock); |
5d0aa2cc | 853 | exp = nf_ct_find_expectation(net, zone, tuple); |
9fb9cbb1 | 854 | if (exp) { |
0d53778e | 855 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", |
c88130bc | 856 | ct, exp); |
9fb9cbb1 | 857 | /* Welcome, Mr. Bond. We've been expecting you... */ |
c88130bc PM |
858 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
859 | ct->master = exp->master; | |
ceceae1b | 860 | if (exp->helper) { |
1afc5679 PNA |
861 | help = nf_ct_helper_ext_add(ct, exp->helper, |
862 | GFP_ATOMIC); | |
ceceae1b | 863 | if (help) |
cf778b00 | 864 | rcu_assign_pointer(help->helper, exp->helper); |
ceceae1b YK |
865 | } |
866 | ||
9fb9cbb1 | 867 | #ifdef CONFIG_NF_CONNTRACK_MARK |
c88130bc | 868 | ct->mark = exp->master->mark; |
7c9728c3 JM |
869 | #endif |
870 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | |
c88130bc | 871 | ct->secmark = exp->master->secmark; |
9fb9cbb1 | 872 | #endif |
c88130bc | 873 | nf_conntrack_get(&ct->master->ct_general); |
0d55af87 | 874 | NF_CT_STAT_INC(net, expect_new); |
22e7410b | 875 | } else { |
b2a15a60 | 876 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); |
0d55af87 | 877 | NF_CT_STAT_INC(net, new); |
22e7410b | 878 | } |
9fb9cbb1 | 879 | |
e53376be PNA |
880 | /* Now it is inserted into the unconfirmed list, bump refcount */ |
881 | nf_conntrack_get(&ct->ct_general); | |
882 | ||
9fb9cbb1 | 883 | /* Overload tuple linked list to put us in unconfirmed list. */ |
ea781f19 | 884 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
63c9a262 | 885 | &net->ct.unconfirmed); |
9fb9cbb1 | 886 | |
f8ba1aff | 887 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
888 | |
889 | if (exp) { | |
890 | if (exp->expectfn) | |
c88130bc | 891 | exp->expectfn(ct, exp); |
6823645d | 892 | nf_ct_expect_put(exp); |
9fb9cbb1 YK |
893 | } |
894 | ||
c88130bc | 895 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
9fb9cbb1 YK |
896 | } |
897 | ||
898 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | |
899 | static inline struct nf_conn * | |
b2a15a60 | 900 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, |
a702a65f | 901 | struct sk_buff *skb, |
9fb9cbb1 YK |
902 | unsigned int dataoff, |
903 | u_int16_t l3num, | |
904 | u_int8_t protonum, | |
905 | struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 906 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 907 | int *set_reply, |
60b5f8f7 | 908 | enum ip_conntrack_info *ctinfo) |
9fb9cbb1 YK |
909 | { |
910 | struct nf_conntrack_tuple tuple; | |
911 | struct nf_conntrack_tuple_hash *h; | |
912 | struct nf_conn *ct; | |
5d0aa2cc | 913 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
99f07e91 | 914 | u32 hash; |
9fb9cbb1 | 915 | |
bbe735e4 | 916 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
9fb9cbb1 | 917 | dataoff, l3num, protonum, &tuple, l3proto, |
605dcad6 | 918 | l4proto)) { |
0d53778e | 919 | pr_debug("resolve_normal_ct: Can't get tuple\n"); |
9fb9cbb1 YK |
920 | return NULL; |
921 | } | |
922 | ||
923 | /* look for tuple match */ | |
99f07e91 CG |
924 | hash = hash_conntrack_raw(&tuple, zone); |
925 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); | |
9fb9cbb1 | 926 | if (!h) { |
b2a15a60 | 927 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, |
60b5f8f7 | 928 | skb, dataoff, hash); |
9fb9cbb1 YK |
929 | if (!h) |
930 | return NULL; | |
931 | if (IS_ERR(h)) | |
932 | return (void *)h; | |
933 | } | |
934 | ct = nf_ct_tuplehash_to_ctrack(h); | |
935 | ||
936 | /* It exists; we have (non-exclusive) reference. */ | |
937 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | |
fb048833 | 938 | *ctinfo = IP_CT_ESTABLISHED_REPLY; |
9fb9cbb1 YK |
939 | /* Please set reply bit if this packet OK */ |
940 | *set_reply = 1; | |
941 | } else { | |
942 | /* Once we've had two way comms, always ESTABLISHED. */ | |
943 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | |
0d53778e | 944 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); |
9fb9cbb1 YK |
945 | *ctinfo = IP_CT_ESTABLISHED; |
946 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | |
0d53778e PM |
947 | pr_debug("nf_conntrack_in: related packet for %p\n", |
948 | ct); | |
9fb9cbb1 YK |
949 | *ctinfo = IP_CT_RELATED; |
950 | } else { | |
0d53778e | 951 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); |
9fb9cbb1 YK |
952 | *ctinfo = IP_CT_NEW; |
953 | } | |
954 | *set_reply = 0; | |
955 | } | |
956 | skb->nfct = &ct->ct_general; | |
957 | skb->nfctinfo = *ctinfo; | |
958 | return ct; | |
959 | } | |
960 | ||
961 | unsigned int | |
a702a65f AD |
962 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
963 | struct sk_buff *skb) | |
9fb9cbb1 | 964 | { |
b2a15a60 | 965 | struct nf_conn *ct, *tmpl = NULL; |
9fb9cbb1 YK |
966 | enum ip_conntrack_info ctinfo; |
967 | struct nf_conntrack_l3proto *l3proto; | |
605dcad6 | 968 | struct nf_conntrack_l4proto *l4proto; |
2c8503f5 | 969 | unsigned int *timeouts; |
9fb9cbb1 YK |
970 | unsigned int dataoff; |
971 | u_int8_t protonum; | |
972 | int set_reply = 0; | |
973 | int ret; | |
974 | ||
3db05fea | 975 | if (skb->nfct) { |
b2a15a60 PM |
976 | /* Previously seen (loopback or untracked)? Ignore. */ |
977 | tmpl = (struct nf_conn *)skb->nfct; | |
978 | if (!nf_ct_is_template(tmpl)) { | |
979 | NF_CT_STAT_INC_ATOMIC(net, ignore); | |
980 | return NF_ACCEPT; | |
981 | } | |
982 | skb->nfct = NULL; | |
9fb9cbb1 YK |
983 | } |
984 | ||
923f4902 | 985 | /* rcu_read_lock()ed by nf_hook_slow */ |
76108cea | 986 | l3proto = __nf_ct_l3proto_find(pf); |
3db05fea | 987 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
ffc30690 YK |
988 | &dataoff, &protonum); |
989 | if (ret <= 0) { | |
25985edc | 990 | pr_debug("not prepared to track yet or error occurred\n"); |
0d55af87 AD |
991 | NF_CT_STAT_INC_ATOMIC(net, error); |
992 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
993 | ret = -ret; |
994 | goto out; | |
9fb9cbb1 YK |
995 | } |
996 | ||
76108cea | 997 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
9fb9cbb1 YK |
998 | |
999 | /* It may be an special packet, error, unclean... | |
1000 | * inverse of the return code tells to the netfilter | |
1001 | * core what to do with the packet. */ | |
74c51a14 | 1002 | if (l4proto->error != NULL) { |
8fea97ec PM |
1003 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, |
1004 | pf, hooknum); | |
74c51a14 | 1005 | if (ret <= 0) { |
0d55af87 AD |
1006 | NF_CT_STAT_INC_ATOMIC(net, error); |
1007 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
1008 | ret = -ret; |
1009 | goto out; | |
74c51a14 | 1010 | } |
88ed01d1 PNA |
1011 | /* ICMP[v6] protocol trackers may assign one conntrack. */ |
1012 | if (skb->nfct) | |
1013 | goto out; | |
9fb9cbb1 YK |
1014 | } |
1015 | ||
b2a15a60 | 1016 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, |
60b5f8f7 | 1017 | l3proto, l4proto, &set_reply, &ctinfo); |
9fb9cbb1 YK |
1018 | if (!ct) { |
1019 | /* Not valid part of a connection */ | |
0d55af87 | 1020 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
b2a15a60 PM |
1021 | ret = NF_ACCEPT; |
1022 | goto out; | |
9fb9cbb1 YK |
1023 | } |
1024 | ||
1025 | if (IS_ERR(ct)) { | |
1026 | /* Too stressed to deal. */ | |
0d55af87 | 1027 | NF_CT_STAT_INC_ATOMIC(net, drop); |
b2a15a60 PM |
1028 | ret = NF_DROP; |
1029 | goto out; | |
9fb9cbb1 YK |
1030 | } |
1031 | ||
3db05fea | 1032 | NF_CT_ASSERT(skb->nfct); |
9fb9cbb1 | 1033 | |
60b5f8f7 | 1034 | /* Decide what timeout policy we want to apply to this flow. */ |
84b5ee93 | 1035 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); |
60b5f8f7 | 1036 | |
2c8503f5 | 1037 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); |
ec8d5409 | 1038 | if (ret <= 0) { |
9fb9cbb1 YK |
1039 | /* Invalid: inverse of the return code tells |
1040 | * the netfilter core what to do */ | |
0d53778e | 1041 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
3db05fea HX |
1042 | nf_conntrack_put(skb->nfct); |
1043 | skb->nfct = NULL; | |
0d55af87 | 1044 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
7d1e0459 PNA |
1045 | if (ret == -NF_DROP) |
1046 | NF_CT_STAT_INC_ATOMIC(net, drop); | |
b2a15a60 PM |
1047 | ret = -ret; |
1048 | goto out; | |
9fb9cbb1 YK |
1049 | } |
1050 | ||
1051 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
858b3133 | 1052 | nf_conntrack_event_cache(IPCT_REPLY, ct); |
b2a15a60 | 1053 | out: |
c3174286 PNA |
1054 | if (tmpl) { |
1055 | /* Special case: we have to repeat this hook, assign the | |
1056 | * template again to this packet. We assume that this packet | |
1057 | * has no conntrack assigned. This is used by nf_ct_tcp. */ | |
1058 | if (ret == NF_REPEAT) | |
1059 | skb->nfct = (struct nf_conntrack *)tmpl; | |
1060 | else | |
1061 | nf_ct_put(tmpl); | |
1062 | } | |
9fb9cbb1 YK |
1063 | |
1064 | return ret; | |
1065 | } | |
13b18339 | 1066 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
9fb9cbb1 | 1067 | |
5f2b4c90 JE |
1068 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
1069 | const struct nf_conntrack_tuple *orig) | |
9fb9cbb1 | 1070 | { |
5f2b4c90 | 1071 | bool ret; |
923f4902 PM |
1072 | |
1073 | rcu_read_lock(); | |
1074 | ret = nf_ct_invert_tuple(inverse, orig, | |
1075 | __nf_ct_l3proto_find(orig->src.l3num), | |
1076 | __nf_ct_l4proto_find(orig->src.l3num, | |
1077 | orig->dst.protonum)); | |
1078 | rcu_read_unlock(); | |
1079 | return ret; | |
9fb9cbb1 | 1080 | } |
13b18339 | 1081 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
9fb9cbb1 | 1082 | |
5b1158e9 JK |
1083 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
1084 | implicitly racy: see __nf_conntrack_confirm */ | |
1085 | void nf_conntrack_alter_reply(struct nf_conn *ct, | |
1086 | const struct nf_conntrack_tuple *newreply) | |
1087 | { | |
1088 | struct nf_conn_help *help = nfct_help(ct); | |
1089 | ||
5b1158e9 JK |
1090 | /* Should be unconfirmed, so not in hash table yet */ |
1091 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
1092 | ||
0d53778e | 1093 | pr_debug("Altering reply tuple of %p to ", ct); |
3c9fba65 | 1094 | nf_ct_dump_tuple(newreply); |
5b1158e9 JK |
1095 | |
1096 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | |
ef1a5a50 | 1097 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
c52fbb41 | 1098 | return; |
ceceae1b | 1099 | |
c52fbb41 | 1100 | rcu_read_lock(); |
b2a15a60 | 1101 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); |
c52fbb41 | 1102 | rcu_read_unlock(); |
5b1158e9 | 1103 | } |
13b18339 | 1104 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
5b1158e9 | 1105 | |
9fb9cbb1 YK |
1106 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
1107 | void __nf_ct_refresh_acct(struct nf_conn *ct, | |
1108 | enum ip_conntrack_info ctinfo, | |
1109 | const struct sk_buff *skb, | |
1110 | unsigned long extra_jiffies, | |
1111 | int do_acct) | |
1112 | { | |
9fb9cbb1 YK |
1113 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
1114 | NF_CT_ASSERT(skb); | |
1115 | ||
997ae831 | 1116 | /* Only update if this is not a fixed timeout */ |
47d95045 PM |
1117 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
1118 | goto acct; | |
997ae831 | 1119 | |
9fb9cbb1 YK |
1120 | /* If not in hash table, timer will not be active yet */ |
1121 | if (!nf_ct_is_confirmed(ct)) { | |
1122 | ct->timeout.expires = extra_jiffies; | |
9fb9cbb1 | 1123 | } else { |
be00c8e4 MJ |
1124 | unsigned long newtime = jiffies + extra_jiffies; |
1125 | ||
1126 | /* Only update the timeout if the new timeout is at least | |
1127 | HZ jiffies from the old timeout. Need del_timer for race | |
1128 | avoidance (may already be dying). */ | |
65cb9fda PM |
1129 | if (newtime - ct->timeout.expires >= HZ) |
1130 | mod_timer_pending(&ct->timeout, newtime); | |
9fb9cbb1 YK |
1131 | } |
1132 | ||
47d95045 | 1133 | acct: |
9fb9cbb1 | 1134 | if (do_acct) { |
f7b13e43 | 1135 | struct nf_conn_acct *acct; |
3ffd5eeb | 1136 | |
58401572 KPO |
1137 | acct = nf_conn_acct_find(ct); |
1138 | if (acct) { | |
f7b13e43 HE |
1139 | struct nf_conn_counter *counter = acct->counter; |
1140 | ||
1141 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); | |
1142 | atomic64_add(skb->len, &counter[CTINFO2DIR(ctinfo)].bytes); | |
58401572 | 1143 | } |
9fb9cbb1 | 1144 | } |
9fb9cbb1 | 1145 | } |
13b18339 | 1146 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
9fb9cbb1 | 1147 | |
4c889498 DM |
1148 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
1149 | enum ip_conntrack_info ctinfo, | |
1150 | const struct sk_buff *skb, | |
1151 | int do_acct) | |
51091764 | 1152 | { |
718d4ad9 | 1153 | if (do_acct) { |
f7b13e43 | 1154 | struct nf_conn_acct *acct; |
58401572 | 1155 | |
58401572 KPO |
1156 | acct = nf_conn_acct_find(ct); |
1157 | if (acct) { | |
f7b13e43 HE |
1158 | struct nf_conn_counter *counter = acct->counter; |
1159 | ||
1160 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); | |
b3e0bfa7 | 1161 | atomic64_add(skb->len - skb_network_offset(skb), |
f7b13e43 | 1162 | &counter[CTINFO2DIR(ctinfo)].bytes); |
58401572 | 1163 | } |
718d4ad9 | 1164 | } |
58401572 | 1165 | |
4c889498 | 1166 | if (del_timer(&ct->timeout)) { |
51091764 | 1167 | ct->timeout.function((unsigned long)ct); |
4c889498 DM |
1168 | return true; |
1169 | } | |
1170 | return false; | |
51091764 | 1171 | } |
718d4ad9 | 1172 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
51091764 | 1173 | |
5d0aa2cc PM |
1174 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1175 | static struct nf_ct_ext_type nf_ct_zone_extend __read_mostly = { | |
1176 | .len = sizeof(struct nf_conntrack_zone), | |
1177 | .align = __alignof__(struct nf_conntrack_zone), | |
1178 | .id = NF_CT_EXT_ZONE, | |
1179 | }; | |
1180 | #endif | |
1181 | ||
c0cd1156 | 1182 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
c1d10adb PNA |
1183 | |
1184 | #include <linux/netfilter/nfnetlink.h> | |
1185 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
57b47a53 IM |
1186 | #include <linux/mutex.h> |
1187 | ||
c1d10adb PNA |
1188 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
1189 | * in ip_conntrack_core, since we don't want the protocols to autoload | |
1190 | * or depend on ctnetlink */ | |
fdf70832 | 1191 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
c1d10adb PNA |
1192 | const struct nf_conntrack_tuple *tuple) |
1193 | { | |
bae65be8 DM |
1194 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || |
1195 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) | |
1196 | goto nla_put_failure; | |
c1d10adb PNA |
1197 | return 0; |
1198 | ||
df6fb868 | 1199 | nla_put_failure: |
c1d10adb PNA |
1200 | return -1; |
1201 | } | |
fdf70832 | 1202 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
c1d10adb | 1203 | |
f73e924c PM |
1204 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
1205 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | |
1206 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | |
c1d10adb | 1207 | }; |
f73e924c | 1208 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
c1d10adb | 1209 | |
fdf70832 | 1210 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
c1d10adb PNA |
1211 | struct nf_conntrack_tuple *t) |
1212 | { | |
df6fb868 | 1213 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
c1d10adb PNA |
1214 | return -EINVAL; |
1215 | ||
77236b6e PM |
1216 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
1217 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | |
c1d10adb PNA |
1218 | |
1219 | return 0; | |
1220 | } | |
fdf70832 | 1221 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
5c0de29d HE |
1222 | |
1223 | int nf_ct_port_nlattr_tuple_size(void) | |
1224 | { | |
1225 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
1226 | } | |
1227 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | |
c1d10adb PNA |
1228 | #endif |
1229 | ||
9fb9cbb1 | 1230 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
312a0c16 | 1231 | static void nf_conntrack_attach(struct sk_buff *nskb, const struct sk_buff *skb) |
9fb9cbb1 YK |
1232 | { |
1233 | struct nf_conn *ct; | |
1234 | enum ip_conntrack_info ctinfo; | |
1235 | ||
1236 | /* This ICMP is in reverse direction to the packet which caused it */ | |
1237 | ct = nf_ct_get(skb, &ctinfo); | |
1238 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | |
fb048833 | 1239 | ctinfo = IP_CT_RELATED_REPLY; |
9fb9cbb1 YK |
1240 | else |
1241 | ctinfo = IP_CT_RELATED; | |
1242 | ||
1243 | /* Attach to new skbuff, and increment count */ | |
1244 | nskb->nfct = &ct->ct_general; | |
1245 | nskb->nfctinfo = ctinfo; | |
1246 | nf_conntrack_get(nskb->nfct); | |
1247 | } | |
1248 | ||
9fb9cbb1 | 1249 | /* Bring out ya dead! */ |
df0933dc | 1250 | static struct nf_conn * |
400dad39 | 1251 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
9fb9cbb1 YK |
1252 | void *data, unsigned int *bucket) |
1253 | { | |
df0933dc PM |
1254 | struct nf_conntrack_tuple_hash *h; |
1255 | struct nf_conn *ct; | |
ea781f19 | 1256 | struct hlist_nulls_node *n; |
9fb9cbb1 | 1257 | |
f8ba1aff | 1258 | spin_lock_bh(&nf_conntrack_lock); |
d696c7bd | 1259 | for (; *bucket < net->ct.htable_size; (*bucket)++) { |
ea781f19 | 1260 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { |
b0cdb1d9 PM |
1261 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) |
1262 | continue; | |
df0933dc PM |
1263 | ct = nf_ct_tuplehash_to_ctrack(h); |
1264 | if (iter(ct, data)) | |
1265 | goto found; | |
1266 | } | |
601e68e1 | 1267 | } |
ea781f19 | 1268 | hlist_nulls_for_each_entry(h, n, &net->ct.unconfirmed, hnnode) { |
df0933dc PM |
1269 | ct = nf_ct_tuplehash_to_ctrack(h); |
1270 | if (iter(ct, data)) | |
ec68e97d | 1271 | set_bit(IPS_DYING_BIT, &ct->status); |
df0933dc | 1272 | } |
f8ba1aff | 1273 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc PM |
1274 | return NULL; |
1275 | found: | |
c073e3fa | 1276 | atomic_inc(&ct->ct_general.use); |
f8ba1aff | 1277 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc | 1278 | return ct; |
9fb9cbb1 YK |
1279 | } |
1280 | ||
400dad39 AD |
1281 | void nf_ct_iterate_cleanup(struct net *net, |
1282 | int (*iter)(struct nf_conn *i, void *data), | |
c655bc68 | 1283 | void *data, u32 portid, int report) |
9fb9cbb1 | 1284 | { |
df0933dc | 1285 | struct nf_conn *ct; |
9fb9cbb1 YK |
1286 | unsigned int bucket = 0; |
1287 | ||
400dad39 | 1288 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
9fb9cbb1 YK |
1289 | /* Time to push up daises... */ |
1290 | if (del_timer(&ct->timeout)) | |
c655bc68 | 1291 | nf_ct_delete(ct, portid, report); |
02982c27 | 1292 | |
9fb9cbb1 YK |
1293 | /* ... else the timer will get him soon. */ |
1294 | ||
1295 | nf_ct_put(ct); | |
1296 | } | |
1297 | } | |
13b18339 | 1298 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
9fb9cbb1 | 1299 | |
274d383b PNA |
1300 | static int kill_all(struct nf_conn *i, void *data) |
1301 | { | |
1302 | return 1; | |
1303 | } | |
1304 | ||
d862a662 | 1305 | void nf_ct_free_hashtable(void *hash, unsigned int size) |
9fb9cbb1 | 1306 | { |
d862a662 | 1307 | if (is_vmalloc_addr(hash)) |
9fb9cbb1 YK |
1308 | vfree(hash); |
1309 | else | |
601e68e1 | 1310 | free_pages((unsigned long)hash, |
f205c5e0 | 1311 | get_order(sizeof(struct hlist_head) * size)); |
9fb9cbb1 | 1312 | } |
ac565e5f | 1313 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
9fb9cbb1 | 1314 | |
ec464e5d | 1315 | void nf_conntrack_flush_report(struct net *net, u32 portid, int report) |
c1d10adb | 1316 | { |
c655bc68 | 1317 | nf_ct_iterate_cleanup(net, kill_all, NULL, portid, report); |
c1d10adb | 1318 | } |
274d383b | 1319 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); |
c1d10adb | 1320 | |
ee254fa4 | 1321 | static void nf_ct_release_dying_list(struct net *net) |
dd7669a9 PNA |
1322 | { |
1323 | struct nf_conntrack_tuple_hash *h; | |
1324 | struct nf_conn *ct; | |
1325 | struct hlist_nulls_node *n; | |
1326 | ||
1327 | spin_lock_bh(&nf_conntrack_lock); | |
ee254fa4 | 1328 | hlist_nulls_for_each_entry(h, n, &net->ct.dying, hnnode) { |
dd7669a9 PNA |
1329 | ct = nf_ct_tuplehash_to_ctrack(h); |
1330 | /* never fails to remove them, no listeners at this point */ | |
1331 | nf_ct_kill(ct); | |
1332 | } | |
1333 | spin_unlock_bh(&nf_conntrack_lock); | |
1334 | } | |
1335 | ||
b3c5163f ED |
1336 | static int untrack_refs(void) |
1337 | { | |
1338 | int cnt = 0, cpu; | |
1339 | ||
1340 | for_each_possible_cpu(cpu) { | |
1341 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
1342 | ||
1343 | cnt += atomic_read(&ct->ct_general.use) - 1; | |
1344 | } | |
1345 | return cnt; | |
1346 | } | |
1347 | ||
f94161c1 | 1348 | void nf_conntrack_cleanup_start(void) |
9fb9cbb1 | 1349 | { |
f94161c1 G |
1350 | RCU_INIT_POINTER(ip_ct_attach, NULL); |
1351 | } | |
1352 | ||
1353 | void nf_conntrack_cleanup_end(void) | |
1354 | { | |
1355 | RCU_INIT_POINTER(nf_ct_destroy, NULL); | |
b3c5163f | 1356 | while (untrack_refs() > 0) |
9edd7ca0 PM |
1357 | schedule(); |
1358 | ||
5d0aa2cc PM |
1359 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1360 | nf_ct_extend_unregister(&nf_ct_zone_extend); | |
1361 | #endif | |
04d87001 | 1362 | nf_conntrack_proto_fini(); |
41d73ec0 | 1363 | nf_conntrack_seqadj_fini(); |
5f69b8f5 | 1364 | nf_conntrack_labels_fini(); |
5e615b22 | 1365 | nf_conntrack_helper_fini(); |
8684094c | 1366 | nf_conntrack_timeout_fini(); |
3fe0f943 | 1367 | nf_conntrack_ecache_fini(); |
73f4001a | 1368 | nf_conntrack_tstamp_fini(); |
b7ff3a1f | 1369 | nf_conntrack_acct_fini(); |
83b4dbe1 | 1370 | nf_conntrack_expect_fini(); |
08f6547d | 1371 | } |
9fb9cbb1 | 1372 | |
f94161c1 G |
1373 | /* |
1374 | * Mishearing the voices in his head, our hero wonders how he's | |
1375 | * supposed to kill the mall. | |
1376 | */ | |
1377 | void nf_conntrack_cleanup_net(struct net *net) | |
08f6547d | 1378 | { |
dece40e8 VD |
1379 | LIST_HEAD(single); |
1380 | ||
1381 | list_add(&net->exit_list, &single); | |
1382 | nf_conntrack_cleanup_net_list(&single); | |
1383 | } | |
1384 | ||
1385 | void nf_conntrack_cleanup_net_list(struct list_head *net_exit_list) | |
1386 | { | |
1387 | int busy; | |
1388 | struct net *net; | |
1389 | ||
f94161c1 G |
1390 | /* |
1391 | * This makes sure all current packets have passed through | |
1392 | * netfilter framework. Roll on, two-stage module | |
1393 | * delete... | |
1394 | */ | |
1395 | synchronize_net(); | |
dece40e8 VD |
1396 | i_see_dead_people: |
1397 | busy = 0; | |
1398 | list_for_each_entry(net, net_exit_list, exit_list) { | |
c655bc68 | 1399 | nf_ct_iterate_cleanup(net, kill_all, NULL, 0, 0); |
dece40e8 VD |
1400 | nf_ct_release_dying_list(net); |
1401 | if (atomic_read(&net->ct.count) != 0) | |
1402 | busy = 1; | |
1403 | } | |
1404 | if (busy) { | |
9fb9cbb1 YK |
1405 | schedule(); |
1406 | goto i_see_dead_people; | |
1407 | } | |
1408 | ||
dece40e8 VD |
1409 | list_for_each_entry(net, net_exit_list, exit_list) { |
1410 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); | |
1411 | nf_conntrack_proto_pernet_fini(net); | |
1412 | nf_conntrack_helper_pernet_fini(net); | |
1413 | nf_conntrack_ecache_pernet_fini(net); | |
1414 | nf_conntrack_tstamp_pernet_fini(net); | |
1415 | nf_conntrack_acct_pernet_fini(net); | |
1416 | nf_conntrack_expect_pernet_fini(net); | |
1417 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); | |
1418 | kfree(net->ct.slabname); | |
1419 | free_percpu(net->ct.stat); | |
1420 | } | |
08f6547d AD |
1421 | } |
1422 | ||
d862a662 | 1423 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) |
9fb9cbb1 | 1424 | { |
ea781f19 ED |
1425 | struct hlist_nulls_head *hash; |
1426 | unsigned int nr_slots, i; | |
1427 | size_t sz; | |
9fb9cbb1 | 1428 | |
ea781f19 ED |
1429 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1430 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | |
1431 | sz = nr_slots * sizeof(struct hlist_nulls_head); | |
1432 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | |
1433 | get_order(sz)); | |
601e68e1 | 1434 | if (!hash) { |
9fb9cbb1 | 1435 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); |
966567b7 | 1436 | hash = vzalloc(sz); |
9fb9cbb1 YK |
1437 | } |
1438 | ||
ea781f19 ED |
1439 | if (hash && nulls) |
1440 | for (i = 0; i < nr_slots; i++) | |
1441 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | |
9fb9cbb1 YK |
1442 | |
1443 | return hash; | |
1444 | } | |
ac565e5f | 1445 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
9fb9cbb1 | 1446 | |
fae718dd | 1447 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
9fb9cbb1 | 1448 | { |
4b5511eb | 1449 | int i, bucket, rc; |
96eb24d7 | 1450 | unsigned int hashsize, old_size; |
ea781f19 | 1451 | struct hlist_nulls_head *hash, *old_hash; |
9fb9cbb1 | 1452 | struct nf_conntrack_tuple_hash *h; |
5d0aa2cc | 1453 | struct nf_conn *ct; |
9fb9cbb1 | 1454 | |
d696c7bd PM |
1455 | if (current->nsproxy->net_ns != &init_net) |
1456 | return -EOPNOTSUPP; | |
1457 | ||
9fb9cbb1 YK |
1458 | /* On boot, we can set this without any fancy locking. */ |
1459 | if (!nf_conntrack_htable_size) | |
1460 | return param_set_uint(val, kp); | |
1461 | ||
4b5511eb AP |
1462 | rc = kstrtouint(val, 0, &hashsize); |
1463 | if (rc) | |
1464 | return rc; | |
9fb9cbb1 YK |
1465 | if (!hashsize) |
1466 | return -EINVAL; | |
1467 | ||
d862a662 | 1468 | hash = nf_ct_alloc_hashtable(&hashsize, 1); |
9fb9cbb1 YK |
1469 | if (!hash) |
1470 | return -ENOMEM; | |
1471 | ||
76507f69 PM |
1472 | /* Lookups in the old hash might happen in parallel, which means we |
1473 | * might get false negatives during connection lookup. New connections | |
1474 | * created because of a false negative won't make it into the hash | |
1475 | * though since that required taking the lock. | |
1476 | */ | |
f8ba1aff | 1477 | spin_lock_bh(&nf_conntrack_lock); |
d696c7bd | 1478 | for (i = 0; i < init_net.ct.htable_size; i++) { |
ea781f19 ED |
1479 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
1480 | h = hlist_nulls_entry(init_net.ct.hash[i].first, | |
1481 | struct nf_conntrack_tuple_hash, hnnode); | |
5d0aa2cc | 1482 | ct = nf_ct_tuplehash_to_ctrack(h); |
ea781f19 | 1483 | hlist_nulls_del_rcu(&h->hnnode); |
5d0aa2cc | 1484 | bucket = __hash_conntrack(&h->tuple, nf_ct_zone(ct), |
99f07e91 | 1485 | hashsize); |
ea781f19 | 1486 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
9fb9cbb1 YK |
1487 | } |
1488 | } | |
d696c7bd | 1489 | old_size = init_net.ct.htable_size; |
400dad39 | 1490 | old_hash = init_net.ct.hash; |
9fb9cbb1 | 1491 | |
d696c7bd | 1492 | init_net.ct.htable_size = nf_conntrack_htable_size = hashsize; |
400dad39 | 1493 | init_net.ct.hash = hash; |
f8ba1aff | 1494 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1495 | |
d862a662 | 1496 | nf_ct_free_hashtable(old_hash, old_size); |
9fb9cbb1 YK |
1497 | return 0; |
1498 | } | |
fae718dd | 1499 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
9fb9cbb1 | 1500 | |
fae718dd | 1501 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
9fb9cbb1 YK |
1502 | &nf_conntrack_htable_size, 0600); |
1503 | ||
5bfddbd4 ED |
1504 | void nf_ct_untracked_status_or(unsigned long bits) |
1505 | { | |
b3c5163f ED |
1506 | int cpu; |
1507 | ||
1508 | for_each_possible_cpu(cpu) | |
1509 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; | |
5bfddbd4 ED |
1510 | } |
1511 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); | |
1512 | ||
f94161c1 | 1513 | int nf_conntrack_init_start(void) |
9fb9cbb1 | 1514 | { |
f205c5e0 | 1515 | int max_factor = 8; |
b3c5163f | 1516 | int ret, cpu; |
9fb9cbb1 YK |
1517 | |
1518 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB | |
f205c5e0 | 1519 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ |
9fb9cbb1 YK |
1520 | if (!nf_conntrack_htable_size) { |
1521 | nf_conntrack_htable_size | |
4481374c | 1522 | = (((totalram_pages << PAGE_SHIFT) / 16384) |
f205c5e0 | 1523 | / sizeof(struct hlist_head)); |
4481374c | 1524 | if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
f205c5e0 PM |
1525 | nf_conntrack_htable_size = 16384; |
1526 | if (nf_conntrack_htable_size < 32) | |
1527 | nf_conntrack_htable_size = 32; | |
1528 | ||
1529 | /* Use a max. factor of four by default to get the same max as | |
1530 | * with the old struct list_heads. When a table size is given | |
1531 | * we use the old value of 8 to avoid reducing the max. | |
1532 | * entries. */ | |
1533 | max_factor = 4; | |
9fb9cbb1 | 1534 | } |
f205c5e0 | 1535 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
8e5105a0 | 1536 | |
654d0fbd | 1537 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", |
8e5105a0 PM |
1538 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, |
1539 | nf_conntrack_max); | |
83b4dbe1 G |
1540 | |
1541 | ret = nf_conntrack_expect_init(); | |
1542 | if (ret < 0) | |
1543 | goto err_expect; | |
1544 | ||
b7ff3a1f G |
1545 | ret = nf_conntrack_acct_init(); |
1546 | if (ret < 0) | |
1547 | goto err_acct; | |
1548 | ||
73f4001a G |
1549 | ret = nf_conntrack_tstamp_init(); |
1550 | if (ret < 0) | |
1551 | goto err_tstamp; | |
1552 | ||
3fe0f943 G |
1553 | ret = nf_conntrack_ecache_init(); |
1554 | if (ret < 0) | |
1555 | goto err_ecache; | |
1556 | ||
8684094c G |
1557 | ret = nf_conntrack_timeout_init(); |
1558 | if (ret < 0) | |
1559 | goto err_timeout; | |
1560 | ||
5e615b22 G |
1561 | ret = nf_conntrack_helper_init(); |
1562 | if (ret < 0) | |
1563 | goto err_helper; | |
1564 | ||
5f69b8f5 G |
1565 | ret = nf_conntrack_labels_init(); |
1566 | if (ret < 0) | |
1567 | goto err_labels; | |
1568 | ||
41d73ec0 PM |
1569 | ret = nf_conntrack_seqadj_init(); |
1570 | if (ret < 0) | |
1571 | goto err_seqadj; | |
1572 | ||
5d0aa2cc PM |
1573 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1574 | ret = nf_ct_extend_register(&nf_ct_zone_extend); | |
1575 | if (ret < 0) | |
1576 | goto err_extend; | |
1577 | #endif | |
04d87001 G |
1578 | ret = nf_conntrack_proto_init(); |
1579 | if (ret < 0) | |
1580 | goto err_proto; | |
1581 | ||
9edd7ca0 | 1582 | /* Set up fake conntrack: to never be deleted, not in any hashes */ |
b3c5163f ED |
1583 | for_each_possible_cpu(cpu) { |
1584 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
b3c5163f ED |
1585 | write_pnet(&ct->ct_net, &init_net); |
1586 | atomic_set(&ct->ct_general.use, 1); | |
1587 | } | |
9edd7ca0 | 1588 | /* - and look it like as a confirmed connection */ |
5bfddbd4 | 1589 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); |
08f6547d AD |
1590 | return 0; |
1591 | ||
04d87001 | 1592 | err_proto: |
5d0aa2cc | 1593 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
04d87001 | 1594 | nf_ct_extend_unregister(&nf_ct_zone_extend); |
5d0aa2cc | 1595 | err_extend: |
a9006892 | 1596 | #endif |
41d73ec0 PM |
1597 | nf_conntrack_seqadj_fini(); |
1598 | err_seqadj: | |
04d87001 | 1599 | nf_conntrack_labels_fini(); |
5f69b8f5 G |
1600 | err_labels: |
1601 | nf_conntrack_helper_fini(); | |
5e615b22 G |
1602 | err_helper: |
1603 | nf_conntrack_timeout_fini(); | |
8684094c G |
1604 | err_timeout: |
1605 | nf_conntrack_ecache_fini(); | |
3fe0f943 G |
1606 | err_ecache: |
1607 | nf_conntrack_tstamp_fini(); | |
73f4001a G |
1608 | err_tstamp: |
1609 | nf_conntrack_acct_fini(); | |
b7ff3a1f G |
1610 | err_acct: |
1611 | nf_conntrack_expect_fini(); | |
83b4dbe1 | 1612 | err_expect: |
08f6547d AD |
1613 | return ret; |
1614 | } | |
1615 | ||
f94161c1 G |
1616 | void nf_conntrack_init_end(void) |
1617 | { | |
1618 | /* For use by REJECT target */ | |
1619 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); | |
1620 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); | |
f94161c1 G |
1621 | } |
1622 | ||
8cc20198 ED |
1623 | /* |
1624 | * We need to use special "null" values, not used in hash table | |
1625 | */ | |
1626 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) | |
1627 | #define DYING_NULLS_VAL ((1<<30)+1) | |
252b3e8c | 1628 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) |
8cc20198 | 1629 | |
f94161c1 | 1630 | int nf_conntrack_init_net(struct net *net) |
08f6547d AD |
1631 | { |
1632 | int ret; | |
ceceae1b | 1633 | |
08f6547d | 1634 | atomic_set(&net->ct.count, 0); |
8cc20198 ED |
1635 | INIT_HLIST_NULLS_HEAD(&net->ct.unconfirmed, UNCONFIRMED_NULLS_VAL); |
1636 | INIT_HLIST_NULLS_HEAD(&net->ct.dying, DYING_NULLS_VAL); | |
252b3e8c | 1637 | INIT_HLIST_NULLS_HEAD(&net->ct.tmpl, TEMPLATE_NULLS_VAL); |
08f6547d AD |
1638 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1639 | if (!net->ct.stat) { | |
1640 | ret = -ENOMEM; | |
1641 | goto err_stat; | |
1642 | } | |
5b3501fa ED |
1643 | |
1644 | net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net); | |
1645 | if (!net->ct.slabname) { | |
1646 | ret = -ENOMEM; | |
1647 | goto err_slabname; | |
1648 | } | |
1649 | ||
1650 | net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname, | |
1651 | sizeof(struct nf_conn), 0, | |
1652 | SLAB_DESTROY_BY_RCU, NULL); | |
1653 | if (!net->ct.nf_conntrack_cachep) { | |
1654 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); | |
1655 | ret = -ENOMEM; | |
1656 | goto err_cache; | |
1657 | } | |
d696c7bd PM |
1658 | |
1659 | net->ct.htable_size = nf_conntrack_htable_size; | |
d862a662 | 1660 | net->ct.hash = nf_ct_alloc_hashtable(&net->ct.htable_size, 1); |
08f6547d AD |
1661 | if (!net->ct.hash) { |
1662 | ret = -ENOMEM; | |
1663 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); | |
1664 | goto err_hash; | |
1665 | } | |
83b4dbe1 | 1666 | ret = nf_conntrack_expect_pernet_init(net); |
08f6547d AD |
1667 | if (ret < 0) |
1668 | goto err_expect; | |
b7ff3a1f | 1669 | ret = nf_conntrack_acct_pernet_init(net); |
58401572 | 1670 | if (ret < 0) |
08f6547d | 1671 | goto err_acct; |
73f4001a | 1672 | ret = nf_conntrack_tstamp_pernet_init(net); |
a992ca2a PNA |
1673 | if (ret < 0) |
1674 | goto err_tstamp; | |
3fe0f943 | 1675 | ret = nf_conntrack_ecache_pernet_init(net); |
a0891aa6 PNA |
1676 | if (ret < 0) |
1677 | goto err_ecache; | |
5e615b22 | 1678 | ret = nf_conntrack_helper_pernet_init(net); |
a9006892 EL |
1679 | if (ret < 0) |
1680 | goto err_helper; | |
04d87001 | 1681 | ret = nf_conntrack_proto_pernet_init(net); |
f94161c1 G |
1682 | if (ret < 0) |
1683 | goto err_proto; | |
08f6547d | 1684 | return 0; |
c539f017 | 1685 | |
f94161c1 | 1686 | err_proto: |
5e615b22 | 1687 | nf_conntrack_helper_pernet_fini(net); |
a9006892 | 1688 | err_helper: |
3fe0f943 | 1689 | nf_conntrack_ecache_pernet_fini(net); |
a0891aa6 | 1690 | err_ecache: |
73f4001a | 1691 | nf_conntrack_tstamp_pernet_fini(net); |
a992ca2a | 1692 | err_tstamp: |
b7ff3a1f | 1693 | nf_conntrack_acct_pernet_fini(net); |
08f6547d | 1694 | err_acct: |
83b4dbe1 | 1695 | nf_conntrack_expect_pernet_fini(net); |
08f6547d | 1696 | err_expect: |
d862a662 | 1697 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
6058fa6b | 1698 | err_hash: |
5b3501fa ED |
1699 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
1700 | err_cache: | |
1701 | kfree(net->ct.slabname); | |
1702 | err_slabname: | |
0d55af87 AD |
1703 | free_percpu(net->ct.stat); |
1704 | err_stat: | |
08f6547d AD |
1705 | return ret; |
1706 | } |