]>
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 | ||
ccd63c20 WJ |
15 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
16 | ||
9fb9cbb1 YK |
17 | #include <linux/types.h> |
18 | #include <linux/netfilter.h> | |
19 | #include <linux/module.h> | |
d43c36dc | 20 | #include <linux/sched.h> |
9fb9cbb1 YK |
21 | #include <linux/skbuff.h> |
22 | #include <linux/proc_fs.h> | |
23 | #include <linux/vmalloc.h> | |
24 | #include <linux/stddef.h> | |
25 | #include <linux/slab.h> | |
26 | #include <linux/random.h> | |
27 | #include <linux/jhash.h> | |
28 | #include <linux/err.h> | |
29 | #include <linux/percpu.h> | |
30 | #include <linux/moduleparam.h> | |
31 | #include <linux/notifier.h> | |
32 | #include <linux/kernel.h> | |
33 | #include <linux/netdevice.h> | |
34 | #include <linux/socket.h> | |
d7fe0f24 | 35 | #include <linux/mm.h> |
d696c7bd | 36 | #include <linux/nsproxy.h> |
ea781f19 | 37 | #include <linux/rculist_nulls.h> |
9fb9cbb1 | 38 | |
9fb9cbb1 YK |
39 | #include <net/netfilter/nf_conntrack.h> |
40 | #include <net/netfilter/nf_conntrack_l3proto.h> | |
605dcad6 | 41 | #include <net/netfilter/nf_conntrack_l4proto.h> |
77ab9cff | 42 | #include <net/netfilter/nf_conntrack_expect.h> |
9fb9cbb1 | 43 | #include <net/netfilter/nf_conntrack_helper.h> |
41d73ec0 | 44 | #include <net/netfilter/nf_conntrack_seqadj.h> |
9fb9cbb1 | 45 | #include <net/netfilter/nf_conntrack_core.h> |
ecfab2c9 | 46 | #include <net/netfilter/nf_conntrack_extend.h> |
58401572 | 47 | #include <net/netfilter/nf_conntrack_acct.h> |
a0891aa6 | 48 | #include <net/netfilter/nf_conntrack_ecache.h> |
5d0aa2cc | 49 | #include <net/netfilter/nf_conntrack_zones.h> |
a992ca2a | 50 | #include <net/netfilter/nf_conntrack_timestamp.h> |
dd705072 | 51 | #include <net/netfilter/nf_conntrack_timeout.h> |
c539f017 | 52 | #include <net/netfilter/nf_conntrack_labels.h> |
48b1de4c | 53 | #include <net/netfilter/nf_conntrack_synproxy.h> |
e6a7d3c0 | 54 | #include <net/netfilter/nf_nat.h> |
e17b666a | 55 | #include <net/netfilter/nf_nat_core.h> |
49376368 | 56 | #include <net/netfilter/nf_nat_helper.h> |
9fb9cbb1 | 57 | |
dc808fe2 | 58 | #define NF_CONNTRACK_VERSION "0.5.0" |
9fb9cbb1 | 59 | |
e17b666a PM |
60 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
61 | enum nf_nat_manip_type manip, | |
39938324 | 62 | const struct nlattr *attr) __read_mostly; |
e6a7d3c0 PNA |
63 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
64 | ||
93bb0ceb JDB |
65 | __cacheline_aligned_in_smp spinlock_t nf_conntrack_locks[CONNTRACK_LOCKS]; |
66 | EXPORT_SYMBOL_GPL(nf_conntrack_locks); | |
9fb9cbb1 | 67 | |
ca7433df JDB |
68 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(nf_conntrack_expect_lock); |
69 | EXPORT_SYMBOL_GPL(nf_conntrack_expect_lock); | |
70 | ||
b16c2919 | 71 | static __read_mostly spinlock_t nf_conntrack_locks_all_lock; |
a3efd812 | 72 | static __read_mostly seqcount_t nf_conntrack_generation; |
b16c2919 SL |
73 | static __read_mostly bool nf_conntrack_locks_all; |
74 | ||
75 | void nf_conntrack_lock(spinlock_t *lock) __acquires(lock) | |
76 | { | |
77 | spin_lock(lock); | |
78 | while (unlikely(nf_conntrack_locks_all)) { | |
79 | spin_unlock(lock); | |
e39365be | 80 | spin_unlock_wait(&nf_conntrack_locks_all_lock); |
b16c2919 SL |
81 | spin_lock(lock); |
82 | } | |
83 | } | |
84 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); | |
85 | ||
93bb0ceb JDB |
86 | static void nf_conntrack_double_unlock(unsigned int h1, unsigned int h2) |
87 | { | |
88 | h1 %= CONNTRACK_LOCKS; | |
89 | h2 %= CONNTRACK_LOCKS; | |
90 | spin_unlock(&nf_conntrack_locks[h1]); | |
91 | if (h1 != h2) | |
92 | spin_unlock(&nf_conntrack_locks[h2]); | |
93 | } | |
94 | ||
95 | /* return true if we need to recompute hashes (in case hash table was resized) */ | |
96 | static bool nf_conntrack_double_lock(struct net *net, unsigned int h1, | |
97 | unsigned int h2, unsigned int sequence) | |
98 | { | |
99 | h1 %= CONNTRACK_LOCKS; | |
100 | h2 %= CONNTRACK_LOCKS; | |
101 | if (h1 <= h2) { | |
b16c2919 | 102 | nf_conntrack_lock(&nf_conntrack_locks[h1]); |
93bb0ceb JDB |
103 | if (h1 != h2) |
104 | spin_lock_nested(&nf_conntrack_locks[h2], | |
105 | SINGLE_DEPTH_NESTING); | |
106 | } else { | |
b16c2919 | 107 | nf_conntrack_lock(&nf_conntrack_locks[h2]); |
93bb0ceb JDB |
108 | spin_lock_nested(&nf_conntrack_locks[h1], |
109 | SINGLE_DEPTH_NESTING); | |
110 | } | |
a3efd812 | 111 | if (read_seqcount_retry(&nf_conntrack_generation, sequence)) { |
93bb0ceb JDB |
112 | nf_conntrack_double_unlock(h1, h2); |
113 | return true; | |
114 | } | |
115 | return false; | |
116 | } | |
117 | ||
118 | static void nf_conntrack_all_lock(void) | |
119 | { | |
120 | int i; | |
121 | ||
b16c2919 SL |
122 | spin_lock(&nf_conntrack_locks_all_lock); |
123 | nf_conntrack_locks_all = true; | |
124 | ||
125 | for (i = 0; i < CONNTRACK_LOCKS; i++) { | |
e39365be | 126 | spin_unlock_wait(&nf_conntrack_locks[i]); |
b16c2919 | 127 | } |
93bb0ceb JDB |
128 | } |
129 | ||
130 | static void nf_conntrack_all_unlock(void) | |
131 | { | |
b16c2919 SL |
132 | nf_conntrack_locks_all = false; |
133 | spin_unlock(&nf_conntrack_locks_all_lock); | |
93bb0ceb JDB |
134 | } |
135 | ||
e2b7606c | 136 | unsigned int nf_conntrack_htable_size __read_mostly; |
13b18339 PM |
137 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
138 | ||
e478075c | 139 | unsigned int nf_conntrack_max __read_mostly; |
a999e683 | 140 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
13b18339 | 141 | |
b3c5163f ED |
142 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); |
143 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); | |
13b18339 | 144 | |
141658fb | 145 | static unsigned int nf_conntrack_hash_rnd __read_mostly; |
9fb9cbb1 | 146 | |
deedb590 | 147 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 | 148 | { |
0794935e | 149 | unsigned int n; |
0794935e | 150 | |
141658fb FW |
151 | get_random_once(&nf_conntrack_hash_rnd, sizeof(nf_conntrack_hash_rnd)); |
152 | ||
0794935e PM |
153 | /* The direction must be ignored, so we hash everything up to the |
154 | * destination ports (which is a multiple of 4) and treat the last | |
155 | * three bytes manually. | |
156 | */ | |
157 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); | |
deedb590 | 158 | return jhash2((u32 *)tuple, n, nf_conntrack_hash_rnd ^ |
99f07e91 CG |
159 | (((__force __u16)tuple->dst.u.all << 16) | |
160 | tuple->dst.protonum)); | |
161 | } | |
162 | ||
163 | static u32 __hash_bucket(u32 hash, unsigned int size) | |
164 | { | |
8fc54f68 | 165 | return reciprocal_scale(hash, size); |
99f07e91 CG |
166 | } |
167 | ||
168 | static u32 hash_bucket(u32 hash, const struct net *net) | |
169 | { | |
170 | return __hash_bucket(hash, net->ct.htable_size); | |
171 | } | |
0794935e | 172 | |
99f07e91 | 173 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, |
308ac914 | 174 | unsigned int size) |
99f07e91 | 175 | { |
deedb590 | 176 | return __hash_bucket(hash_conntrack_raw(tuple), size); |
9fb9cbb1 YK |
177 | } |
178 | ||
308ac914 | 179 | static inline u_int32_t hash_conntrack(const struct net *net, |
d696c7bd | 180 | const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 | 181 | { |
deedb590 | 182 | return __hash_conntrack(tuple, net->ct.htable_size); |
9fb9cbb1 YK |
183 | } |
184 | ||
5f2b4c90 | 185 | bool |
9fb9cbb1 YK |
186 | nf_ct_get_tuple(const struct sk_buff *skb, |
187 | unsigned int nhoff, | |
188 | unsigned int dataoff, | |
189 | u_int16_t l3num, | |
190 | u_int8_t protonum, | |
a31f1adc | 191 | struct net *net, |
9fb9cbb1 YK |
192 | struct nf_conntrack_tuple *tuple, |
193 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 194 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 195 | { |
443a70d5 | 196 | memset(tuple, 0, sizeof(*tuple)); |
9fb9cbb1 YK |
197 | |
198 | tuple->src.l3num = l3num; | |
199 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | |
5f2b4c90 | 200 | return false; |
9fb9cbb1 YK |
201 | |
202 | tuple->dst.protonum = protonum; | |
203 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | |
204 | ||
a31f1adc | 205 | return l4proto->pkt_to_tuple(skb, dataoff, net, tuple); |
9fb9cbb1 | 206 | } |
13b18339 | 207 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
9fb9cbb1 | 208 | |
5f2b4c90 | 209 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
a31f1adc EB |
210 | u_int16_t l3num, |
211 | struct net *net, struct nf_conntrack_tuple *tuple) | |
e2a3123f YK |
212 | { |
213 | struct nf_conntrack_l3proto *l3proto; | |
214 | struct nf_conntrack_l4proto *l4proto; | |
215 | unsigned int protoff; | |
216 | u_int8_t protonum; | |
217 | int ret; | |
218 | ||
219 | rcu_read_lock(); | |
220 | ||
221 | l3proto = __nf_ct_l3proto_find(l3num); | |
222 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | |
223 | if (ret != NF_ACCEPT) { | |
224 | rcu_read_unlock(); | |
5f2b4c90 | 225 | return false; |
e2a3123f YK |
226 | } |
227 | ||
228 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
229 | ||
a31f1adc | 230 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, net, tuple, |
e2a3123f YK |
231 | l3proto, l4proto); |
232 | ||
233 | rcu_read_unlock(); | |
234 | return ret; | |
235 | } | |
236 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | |
237 | ||
5f2b4c90 | 238 | bool |
9fb9cbb1 YK |
239 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
240 | const struct nf_conntrack_tuple *orig, | |
241 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 242 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 243 | { |
443a70d5 | 244 | memset(inverse, 0, sizeof(*inverse)); |
9fb9cbb1 YK |
245 | |
246 | inverse->src.l3num = orig->src.l3num; | |
247 | if (l3proto->invert_tuple(inverse, orig) == 0) | |
5f2b4c90 | 248 | return false; |
9fb9cbb1 YK |
249 | |
250 | inverse->dst.dir = !orig->dst.dir; | |
251 | ||
252 | inverse->dst.protonum = orig->dst.protonum; | |
605dcad6 | 253 | return l4proto->invert_tuple(inverse, orig); |
9fb9cbb1 | 254 | } |
13b18339 | 255 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
9fb9cbb1 | 256 | |
9fb9cbb1 YK |
257 | static void |
258 | clean_from_lists(struct nf_conn *ct) | |
259 | { | |
0d53778e | 260 | pr_debug("clean_from_lists(%p)\n", ct); |
ea781f19 ED |
261 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
262 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | |
9fb9cbb1 YK |
263 | |
264 | /* Destroy all pending expectations */ | |
c1d10adb | 265 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
266 | } |
267 | ||
b7779d06 JDB |
268 | /* must be called with local_bh_disable */ |
269 | static void nf_ct_add_to_dying_list(struct nf_conn *ct) | |
270 | { | |
271 | struct ct_pcpu *pcpu; | |
272 | ||
273 | /* add this conntrack to the (per cpu) dying list */ | |
274 | ct->cpu = smp_processor_id(); | |
275 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
276 | ||
277 | spin_lock(&pcpu->lock); | |
278 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
279 | &pcpu->dying); | |
280 | spin_unlock(&pcpu->lock); | |
281 | } | |
282 | ||
283 | /* must be called with local_bh_disable */ | |
284 | static void nf_ct_add_to_unconfirmed_list(struct nf_conn *ct) | |
285 | { | |
286 | struct ct_pcpu *pcpu; | |
287 | ||
288 | /* add this conntrack to the (per cpu) unconfirmed list */ | |
289 | ct->cpu = smp_processor_id(); | |
290 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
291 | ||
292 | spin_lock(&pcpu->lock); | |
293 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
294 | &pcpu->unconfirmed); | |
295 | spin_unlock(&pcpu->lock); | |
296 | } | |
297 | ||
298 | /* must be called with local_bh_disable */ | |
299 | static void nf_ct_del_from_dying_or_unconfirmed_list(struct nf_conn *ct) | |
300 | { | |
301 | struct ct_pcpu *pcpu; | |
302 | ||
303 | /* We overload first tuple to link into unconfirmed or dying list.*/ | |
304 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
305 | ||
306 | spin_lock(&pcpu->lock); | |
307 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); | |
308 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | |
309 | spin_unlock(&pcpu->lock); | |
310 | } | |
311 | ||
0838aa7f | 312 | /* Released via destroy_conntrack() */ |
308ac914 DB |
313 | struct nf_conn *nf_ct_tmpl_alloc(struct net *net, |
314 | const struct nf_conntrack_zone *zone, | |
315 | gfp_t flags) | |
0838aa7f PNA |
316 | { |
317 | struct nf_conn *tmpl; | |
318 | ||
f58e5aa7 | 319 | tmpl = kzalloc(sizeof(*tmpl), flags); |
0838aa7f PNA |
320 | if (tmpl == NULL) |
321 | return NULL; | |
322 | ||
323 | tmpl->status = IPS_TEMPLATE; | |
324 | write_pnet(&tmpl->ct_net, net); | |
325 | ||
5e8018fc DB |
326 | if (nf_ct_zone_add(tmpl, flags, zone) < 0) |
327 | goto out_free; | |
328 | ||
0838aa7f PNA |
329 | atomic_set(&tmpl->ct_general.use, 0); |
330 | ||
331 | return tmpl; | |
0838aa7f PNA |
332 | out_free: |
333 | kfree(tmpl); | |
334 | return NULL; | |
0838aa7f PNA |
335 | } |
336 | EXPORT_SYMBOL_GPL(nf_ct_tmpl_alloc); | |
337 | ||
9cf94eab | 338 | void nf_ct_tmpl_free(struct nf_conn *tmpl) |
0838aa7f PNA |
339 | { |
340 | nf_ct_ext_destroy(tmpl); | |
341 | nf_ct_ext_free(tmpl); | |
342 | kfree(tmpl); | |
343 | } | |
9cf94eab | 344 | EXPORT_SYMBOL_GPL(nf_ct_tmpl_free); |
0838aa7f | 345 | |
9fb9cbb1 YK |
346 | static void |
347 | destroy_conntrack(struct nf_conntrack *nfct) | |
348 | { | |
349 | struct nf_conn *ct = (struct nf_conn *)nfct; | |
0d55af87 | 350 | struct net *net = nf_ct_net(ct); |
605dcad6 | 351 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 | 352 | |
0d53778e | 353 | pr_debug("destroy_conntrack(%p)\n", ct); |
9fb9cbb1 YK |
354 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
355 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); | |
356 | ||
0838aa7f PNA |
357 | if (unlikely(nf_ct_is_template(ct))) { |
358 | nf_ct_tmpl_free(ct); | |
359 | return; | |
360 | } | |
923f4902 | 361 | rcu_read_lock(); |
5e8fbe2a | 362 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
605dcad6 MJ |
363 | if (l4proto && l4proto->destroy) |
364 | l4proto->destroy(ct); | |
9fb9cbb1 | 365 | |
982d9a9c | 366 | rcu_read_unlock(); |
9fb9cbb1 | 367 | |
ca7433df | 368 | local_bh_disable(); |
9fb9cbb1 YK |
369 | /* Expectations will have been removed in clean_from_lists, |
370 | * except TFTP can create an expectation on the first packet, | |
371 | * before connection is in the list, so we need to clean here, | |
ca7433df JDB |
372 | * too. |
373 | */ | |
c1d10adb | 374 | nf_ct_remove_expectations(ct); |
9fb9cbb1 | 375 | |
b7779d06 | 376 | nf_ct_del_from_dying_or_unconfirmed_list(ct); |
9fb9cbb1 | 377 | |
0d55af87 | 378 | NF_CT_STAT_INC(net, delete); |
ca7433df | 379 | local_bh_enable(); |
9fb9cbb1 YK |
380 | |
381 | if (ct->master) | |
382 | nf_ct_put(ct->master); | |
383 | ||
0d53778e | 384 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
9fb9cbb1 YK |
385 | nf_conntrack_free(ct); |
386 | } | |
387 | ||
02982c27 | 388 | static void nf_ct_delete_from_lists(struct nf_conn *ct) |
9fb9cbb1 | 389 | { |
0d55af87 | 390 | struct net *net = nf_ct_net(ct); |
93bb0ceb | 391 | unsigned int hash, reply_hash; |
93bb0ceb | 392 | unsigned int sequence; |
9fb9cbb1 | 393 | |
9858a3ae | 394 | nf_ct_helper_destroy(ct); |
93bb0ceb JDB |
395 | |
396 | local_bh_disable(); | |
397 | do { | |
a3efd812 | 398 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
deedb590 | 399 | hash = hash_conntrack(net, |
93bb0ceb | 400 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
deedb590 | 401 | reply_hash = hash_conntrack(net, |
93bb0ceb JDB |
402 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
403 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | |
404 | ||
9fb9cbb1 | 405 | clean_from_lists(ct); |
93bb0ceb JDB |
406 | nf_conntrack_double_unlock(hash, reply_hash); |
407 | ||
b7779d06 | 408 | nf_ct_add_to_dying_list(ct); |
93bb0ceb JDB |
409 | |
410 | NF_CT_STAT_INC(net, delete_list); | |
411 | local_bh_enable(); | |
dd7669a9 | 412 | } |
dd7669a9 | 413 | |
02982c27 | 414 | bool nf_ct_delete(struct nf_conn *ct, u32 portid, int report) |
dd7669a9 | 415 | { |
a992ca2a PNA |
416 | struct nf_conn_tstamp *tstamp; |
417 | ||
418 | tstamp = nf_conn_tstamp_find(ct); | |
419 | if (tstamp && tstamp->stop == 0) | |
d2de875c | 420 | tstamp->stop = ktime_get_real_ns(); |
dd7669a9 | 421 | |
9500507c FW |
422 | if (nf_ct_is_dying(ct)) |
423 | goto delete; | |
424 | ||
425 | if (nf_conntrack_event_report(IPCT_DESTROY, ct, | |
426 | portid, report) < 0) { | |
dd7669a9 PNA |
427 | /* destroy event was not delivered */ |
428 | nf_ct_delete_from_lists(ct); | |
9500507c | 429 | nf_conntrack_ecache_delayed_work(nf_ct_net(ct)); |
02982c27 | 430 | return false; |
dd7669a9 | 431 | } |
9500507c FW |
432 | |
433 | nf_conntrack_ecache_work(nf_ct_net(ct)); | |
dd7669a9 | 434 | set_bit(IPS_DYING_BIT, &ct->status); |
9500507c | 435 | delete: |
dd7669a9 | 436 | nf_ct_delete_from_lists(ct); |
9fb9cbb1 | 437 | nf_ct_put(ct); |
02982c27 FW |
438 | return true; |
439 | } | |
440 | EXPORT_SYMBOL_GPL(nf_ct_delete); | |
441 | ||
442 | static void death_by_timeout(unsigned long ul_conntrack) | |
443 | { | |
444 | nf_ct_delete((struct nf_conn *)ul_conntrack, 0, 0); | |
9fb9cbb1 YK |
445 | } |
446 | ||
c6825c09 AV |
447 | static inline bool |
448 | nf_ct_key_equal(struct nf_conntrack_tuple_hash *h, | |
308ac914 DB |
449 | const struct nf_conntrack_tuple *tuple, |
450 | const struct nf_conntrack_zone *zone) | |
c6825c09 AV |
451 | { |
452 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | |
453 | ||
454 | /* A conntrack can be recreated with the equal tuple, | |
455 | * so we need to check that the conntrack is confirmed | |
456 | */ | |
457 | return nf_ct_tuple_equal(tuple, &h->tuple) && | |
deedb590 | 458 | nf_ct_zone_equal(ct, zone, NF_CT_DIRECTION(h)) && |
308ac914 | 459 | nf_ct_is_confirmed(ct); |
c6825c09 AV |
460 | } |
461 | ||
ea781f19 ED |
462 | /* |
463 | * Warning : | |
464 | * - Caller must take a reference on returned object | |
465 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | |
ea781f19 | 466 | */ |
99f07e91 | 467 | static struct nf_conntrack_tuple_hash * |
308ac914 | 468 | ____nf_conntrack_find(struct net *net, const struct nf_conntrack_zone *zone, |
99f07e91 | 469 | const struct nf_conntrack_tuple *tuple, u32 hash) |
9fb9cbb1 YK |
470 | { |
471 | struct nf_conntrack_tuple_hash *h; | |
ea781f19 | 472 | struct hlist_nulls_node *n; |
99f07e91 | 473 | unsigned int bucket = hash_bucket(hash, net); |
9fb9cbb1 | 474 | |
4e29e9ec PM |
475 | /* Disable BHs the entire time since we normally need to disable them |
476 | * at least once for the stats anyway. | |
477 | */ | |
478 | local_bh_disable(); | |
ea781f19 | 479 | begin: |
99f07e91 | 480 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[bucket], hnnode) { |
c6825c09 | 481 | if (nf_ct_key_equal(h, tuple, zone)) { |
0d55af87 | 482 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 483 | local_bh_enable(); |
9fb9cbb1 YK |
484 | return h; |
485 | } | |
0d55af87 | 486 | NF_CT_STAT_INC(net, searched); |
9fb9cbb1 | 487 | } |
ea781f19 ED |
488 | /* |
489 | * if the nulls value we got at the end of this lookup is | |
490 | * not the expected one, we must restart lookup. | |
491 | * We probably met an item that was moved to another chain. | |
492 | */ | |
99f07e91 | 493 | if (get_nulls_value(n) != bucket) { |
af740b2c | 494 | NF_CT_STAT_INC(net, search_restart); |
ea781f19 | 495 | goto begin; |
af740b2c | 496 | } |
4e29e9ec | 497 | local_bh_enable(); |
9fb9cbb1 YK |
498 | |
499 | return NULL; | |
500 | } | |
99f07e91 | 501 | |
9fb9cbb1 | 502 | /* Find a connection corresponding to a tuple. */ |
99f07e91 | 503 | static struct nf_conntrack_tuple_hash * |
308ac914 | 504 | __nf_conntrack_find_get(struct net *net, const struct nf_conntrack_zone *zone, |
99f07e91 | 505 | const struct nf_conntrack_tuple *tuple, u32 hash) |
9fb9cbb1 YK |
506 | { |
507 | struct nf_conntrack_tuple_hash *h; | |
76507f69 | 508 | struct nf_conn *ct; |
9fb9cbb1 | 509 | |
76507f69 | 510 | rcu_read_lock(); |
ea781f19 | 511 | begin: |
99f07e91 | 512 | h = ____nf_conntrack_find(net, zone, tuple, hash); |
76507f69 PM |
513 | if (h) { |
514 | ct = nf_ct_tuplehash_to_ctrack(h); | |
8d8890b7 PM |
515 | if (unlikely(nf_ct_is_dying(ct) || |
516 | !atomic_inc_not_zero(&ct->ct_general.use))) | |
76507f69 | 517 | h = NULL; |
ea781f19 | 518 | else { |
c6825c09 | 519 | if (unlikely(!nf_ct_key_equal(h, tuple, zone))) { |
ea781f19 ED |
520 | nf_ct_put(ct); |
521 | goto begin; | |
522 | } | |
523 | } | |
76507f69 PM |
524 | } |
525 | rcu_read_unlock(); | |
9fb9cbb1 YK |
526 | |
527 | return h; | |
528 | } | |
99f07e91 CG |
529 | |
530 | struct nf_conntrack_tuple_hash * | |
308ac914 | 531 | nf_conntrack_find_get(struct net *net, const struct nf_conntrack_zone *zone, |
99f07e91 CG |
532 | const struct nf_conntrack_tuple *tuple) |
533 | { | |
534 | return __nf_conntrack_find_get(net, zone, tuple, | |
deedb590 | 535 | hash_conntrack_raw(tuple)); |
99f07e91 | 536 | } |
13b18339 | 537 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
9fb9cbb1 | 538 | |
c1d10adb PNA |
539 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
540 | unsigned int hash, | |
b476b72a | 541 | unsigned int reply_hash) |
c1d10adb | 542 | { |
400dad39 AD |
543 | struct net *net = nf_ct_net(ct); |
544 | ||
ea781f19 | 545 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
400dad39 | 546 | &net->ct.hash[hash]); |
ea781f19 | 547 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
b476b72a | 548 | &net->ct.hash[reply_hash]); |
c1d10adb PNA |
549 | } |
550 | ||
7d367e06 JK |
551 | int |
552 | nf_conntrack_hash_check_insert(struct nf_conn *ct) | |
c1d10adb | 553 | { |
308ac914 | 554 | const struct nf_conntrack_zone *zone; |
d696c7bd | 555 | struct net *net = nf_ct_net(ct); |
b476b72a | 556 | unsigned int hash, reply_hash; |
7d367e06 JK |
557 | struct nf_conntrack_tuple_hash *h; |
558 | struct hlist_nulls_node *n; | |
93bb0ceb | 559 | unsigned int sequence; |
c1d10adb | 560 | |
5d0aa2cc | 561 | zone = nf_ct_zone(ct); |
7d367e06 | 562 | |
93bb0ceb JDB |
563 | local_bh_disable(); |
564 | do { | |
a3efd812 | 565 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
deedb590 | 566 | hash = hash_conntrack(net, |
93bb0ceb | 567 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
deedb590 | 568 | reply_hash = hash_conntrack(net, |
93bb0ceb JDB |
569 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
570 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | |
7d367e06 JK |
571 | |
572 | /* See if there's one in the list already, including reverse */ | |
573 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) | |
574 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | |
575 | &h->tuple) && | |
deedb590 DB |
576 | nf_ct_zone_equal(nf_ct_tuplehash_to_ctrack(h), zone, |
577 | NF_CT_DIRECTION(h))) | |
7d367e06 | 578 | goto out; |
b476b72a | 579 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[reply_hash], hnnode) |
7d367e06 JK |
580 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
581 | &h->tuple) && | |
deedb590 DB |
582 | nf_ct_zone_equal(nf_ct_tuplehash_to_ctrack(h), zone, |
583 | NF_CT_DIRECTION(h))) | |
7d367e06 | 584 | goto out; |
c1d10adb | 585 | |
7d367e06 | 586 | add_timer(&ct->timeout); |
e53376be PNA |
587 | smp_wmb(); |
588 | /* The caller holds a reference to this object */ | |
589 | atomic_set(&ct->ct_general.use, 2); | |
b476b72a | 590 | __nf_conntrack_hash_insert(ct, hash, reply_hash); |
93bb0ceb | 591 | nf_conntrack_double_unlock(hash, reply_hash); |
7d367e06 | 592 | NF_CT_STAT_INC(net, insert); |
93bb0ceb | 593 | local_bh_enable(); |
7d367e06 JK |
594 | return 0; |
595 | ||
596 | out: | |
93bb0ceb | 597 | nf_conntrack_double_unlock(hash, reply_hash); |
7d367e06 | 598 | NF_CT_STAT_INC(net, insert_failed); |
93bb0ceb | 599 | local_bh_enable(); |
7d367e06 | 600 | return -EEXIST; |
c1d10adb | 601 | } |
7d367e06 | 602 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); |
c1d10adb | 603 | |
9fb9cbb1 YK |
604 | /* Confirm a connection given skb; places it in hash table */ |
605 | int | |
3db05fea | 606 | __nf_conntrack_confirm(struct sk_buff *skb) |
9fb9cbb1 | 607 | { |
308ac914 | 608 | const struct nf_conntrack_zone *zone; |
b476b72a | 609 | unsigned int hash, reply_hash; |
df0933dc | 610 | struct nf_conntrack_tuple_hash *h; |
9fb9cbb1 | 611 | struct nf_conn *ct; |
df0933dc | 612 | struct nf_conn_help *help; |
a992ca2a | 613 | struct nf_conn_tstamp *tstamp; |
ea781f19 | 614 | struct hlist_nulls_node *n; |
9fb9cbb1 | 615 | enum ip_conntrack_info ctinfo; |
400dad39 | 616 | struct net *net; |
93bb0ceb | 617 | unsigned int sequence; |
9fb9cbb1 | 618 | |
3db05fea | 619 | ct = nf_ct_get(skb, &ctinfo); |
400dad39 | 620 | net = nf_ct_net(ct); |
9fb9cbb1 YK |
621 | |
622 | /* ipt_REJECT uses nf_conntrack_attach to attach related | |
623 | ICMP/TCP RST packets in other direction. Actual packet | |
624 | which created connection will be IP_CT_NEW or for an | |
625 | expected connection, IP_CT_RELATED. */ | |
626 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | |
627 | return NF_ACCEPT; | |
628 | ||
5d0aa2cc | 629 | zone = nf_ct_zone(ct); |
93bb0ceb JDB |
630 | local_bh_disable(); |
631 | ||
632 | do { | |
a3efd812 | 633 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
93bb0ceb JDB |
634 | /* reuse the hash saved before */ |
635 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; | |
636 | hash = hash_bucket(hash, net); | |
deedb590 | 637 | reply_hash = hash_conntrack(net, |
93bb0ceb JDB |
638 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
639 | ||
640 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | |
9fb9cbb1 YK |
641 | |
642 | /* We're not in hash table, and we refuse to set up related | |
93bb0ceb JDB |
643 | * connections for unconfirmed conns. But packet copies and |
644 | * REJECT will give spurious warnings here. | |
645 | */ | |
9fb9cbb1 YK |
646 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ |
647 | ||
25985edc | 648 | /* No external references means no one else could have |
93bb0ceb JDB |
649 | * confirmed us. |
650 | */ | |
9fb9cbb1 | 651 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); |
0d53778e | 652 | pr_debug("Confirming conntrack %p\n", ct); |
8ca3f5e9 PNA |
653 | /* We have to check the DYING flag after unlink to prevent |
654 | * a race against nf_ct_get_next_corpse() possibly called from | |
655 | * user context, else we insert an already 'dead' hash, blocking | |
656 | * further use of that particular connection -JM. | |
657 | */ | |
658 | nf_ct_del_from_dying_or_unconfirmed_list(ct); | |
659 | ||
660 | if (unlikely(nf_ct_is_dying(ct))) | |
661 | goto out; | |
fc350777 | 662 | |
9fb9cbb1 YK |
663 | /* See if there's one in the list already, including reverse: |
664 | NAT could have grabbed it without realizing, since we're | |
665 | not in the hash. If there is, we lost race. */ | |
ea781f19 | 666 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
df0933dc | 667 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
5d0aa2cc | 668 | &h->tuple) && |
deedb590 DB |
669 | nf_ct_zone_equal(nf_ct_tuplehash_to_ctrack(h), zone, |
670 | NF_CT_DIRECTION(h))) | |
df0933dc | 671 | goto out; |
b476b72a | 672 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[reply_hash], hnnode) |
df0933dc | 673 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
5d0aa2cc | 674 | &h->tuple) && |
deedb590 DB |
675 | nf_ct_zone_equal(nf_ct_tuplehash_to_ctrack(h), zone, |
676 | NF_CT_DIRECTION(h))) | |
df0933dc | 677 | goto out; |
9fb9cbb1 | 678 | |
df0933dc PM |
679 | /* Timer relative to confirmation time, not original |
680 | setting time, otherwise we'd get timer wrap in | |
681 | weird delay cases. */ | |
682 | ct->timeout.expires += jiffies; | |
683 | add_timer(&ct->timeout); | |
684 | atomic_inc(&ct->ct_general.use); | |
45eec341 | 685 | ct->status |= IPS_CONFIRMED; |
5c8ec910 | 686 | |
a992ca2a PNA |
687 | /* set conntrack timestamp, if enabled. */ |
688 | tstamp = nf_conn_tstamp_find(ct); | |
689 | if (tstamp) { | |
690 | if (skb->tstamp.tv64 == 0) | |
e3192690 | 691 | __net_timestamp(skb); |
a992ca2a PNA |
692 | |
693 | tstamp->start = ktime_to_ns(skb->tstamp); | |
694 | } | |
5c8ec910 PM |
695 | /* Since the lookup is lockless, hash insertion must be done after |
696 | * starting the timer and setting the CONFIRMED bit. The RCU barriers | |
697 | * guarantee that no other CPU can find the conntrack before the above | |
698 | * stores are visible. | |
699 | */ | |
b476b72a | 700 | __nf_conntrack_hash_insert(ct, hash, reply_hash); |
93bb0ceb | 701 | nf_conntrack_double_unlock(hash, reply_hash); |
0d55af87 | 702 | NF_CT_STAT_INC(net, insert); |
93bb0ceb | 703 | local_bh_enable(); |
5c8ec910 | 704 | |
df0933dc PM |
705 | help = nfct_help(ct); |
706 | if (help && help->helper) | |
a71996fc | 707 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
17e6e4ea | 708 | |
df0933dc | 709 | nf_conntrack_event_cache(master_ct(ct) ? |
a71996fc | 710 | IPCT_RELATED : IPCT_NEW, ct); |
df0933dc | 711 | return NF_ACCEPT; |
9fb9cbb1 | 712 | |
df0933dc | 713 | out: |
8ca3f5e9 | 714 | nf_ct_add_to_dying_list(ct); |
93bb0ceb | 715 | nf_conntrack_double_unlock(hash, reply_hash); |
0d55af87 | 716 | NF_CT_STAT_INC(net, insert_failed); |
93bb0ceb | 717 | local_bh_enable(); |
9fb9cbb1 YK |
718 | return NF_DROP; |
719 | } | |
13b18339 | 720 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
9fb9cbb1 YK |
721 | |
722 | /* Returns true if a connection correspondings to the tuple (required | |
723 | for NAT). */ | |
724 | int | |
725 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | |
726 | const struct nf_conn *ignored_conntrack) | |
727 | { | |
400dad39 | 728 | struct net *net = nf_ct_net(ignored_conntrack); |
308ac914 | 729 | const struct nf_conntrack_zone *zone; |
9fb9cbb1 | 730 | struct nf_conntrack_tuple_hash *h; |
ea781f19 | 731 | struct hlist_nulls_node *n; |
5d0aa2cc | 732 | struct nf_conn *ct; |
308ac914 DB |
733 | unsigned int hash; |
734 | ||
735 | zone = nf_ct_zone(ignored_conntrack); | |
deedb590 | 736 | hash = hash_conntrack(net, tuple); |
9fb9cbb1 | 737 | |
4e29e9ec PM |
738 | /* Disable BHs the entire time since we need to disable them at |
739 | * least once for the stats anyway. | |
740 | */ | |
741 | rcu_read_lock_bh(); | |
ea781f19 | 742 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
5d0aa2cc PM |
743 | ct = nf_ct_tuplehash_to_ctrack(h); |
744 | if (ct != ignored_conntrack && | |
745 | nf_ct_tuple_equal(tuple, &h->tuple) && | |
deedb590 | 746 | nf_ct_zone_equal(ct, zone, NF_CT_DIRECTION(h))) { |
0d55af87 | 747 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 748 | rcu_read_unlock_bh(); |
ba419aff PM |
749 | return 1; |
750 | } | |
0d55af87 | 751 | NF_CT_STAT_INC(net, searched); |
ba419aff | 752 | } |
4e29e9ec | 753 | rcu_read_unlock_bh(); |
9fb9cbb1 | 754 | |
ba419aff | 755 | return 0; |
9fb9cbb1 | 756 | } |
13b18339 | 757 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
9fb9cbb1 | 758 | |
7ae7730f PM |
759 | #define NF_CT_EVICTION_RANGE 8 |
760 | ||
9fb9cbb1 YK |
761 | /* There's a small race here where we may free a just-assured |
762 | connection. Too bad: we're in trouble anyway. */ | |
93bb0ceb | 763 | static noinline int early_drop(struct net *net, unsigned int _hash) |
9fb9cbb1 | 764 | { |
f205c5e0 | 765 | /* Use oldest entry, which is roughly LRU */ |
9fb9cbb1 | 766 | struct nf_conntrack_tuple_hash *h; |
df0933dc | 767 | struct nf_conn *ct = NULL, *tmp; |
ea781f19 | 768 | struct hlist_nulls_node *n; |
93bb0ceb | 769 | unsigned int i = 0, cnt = 0; |
9fb9cbb1 | 770 | int dropped = 0; |
93bb0ceb JDB |
771 | unsigned int hash, sequence; |
772 | spinlock_t *lockp; | |
9fb9cbb1 | 773 | |
93bb0ceb JDB |
774 | local_bh_disable(); |
775 | restart: | |
a3efd812 | 776 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
93bb0ceb JDB |
777 | hash = hash_bucket(_hash, net); |
778 | for (; i < net->ct.htable_size; i++) { | |
779 | lockp = &nf_conntrack_locks[hash % CONNTRACK_LOCKS]; | |
b16c2919 | 780 | nf_conntrack_lock(lockp); |
a3efd812 | 781 | if (read_seqcount_retry(&nf_conntrack_generation, sequence)) { |
93bb0ceb JDB |
782 | spin_unlock(lockp); |
783 | goto restart; | |
784 | } | |
ea781f19 ED |
785 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
786 | hnnode) { | |
7ae7730f | 787 | tmp = nf_ct_tuplehash_to_ctrack(h); |
93bb0ceb JDB |
788 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status) && |
789 | !nf_ct_is_dying(tmp) && | |
790 | atomic_inc_not_zero(&tmp->ct_general.use)) { | |
7ae7730f | 791 | ct = tmp; |
93bb0ceb JDB |
792 | break; |
793 | } | |
7ae7730f PM |
794 | cnt++; |
795 | } | |
76507f69 | 796 | |
93bb0ceb JDB |
797 | hash = (hash + 1) % net->ct.htable_size; |
798 | spin_unlock(lockp); | |
5ae27aa2 | 799 | |
93bb0ceb | 800 | if (ct || cnt >= NF_CT_EVICTION_RANGE) |
7ae7730f | 801 | break; |
5ae27aa2 | 802 | |
9fb9cbb1 | 803 | } |
93bb0ceb | 804 | local_bh_enable(); |
9fb9cbb1 YK |
805 | |
806 | if (!ct) | |
807 | return dropped; | |
808 | ||
809 | if (del_timer(&ct->timeout)) { | |
02982c27 | 810 | if (nf_ct_delete(ct, 0, 0)) { |
74138511 PNA |
811 | dropped = 1; |
812 | NF_CT_STAT_INC_ATOMIC(net, early_drop); | |
813 | } | |
9fb9cbb1 YK |
814 | } |
815 | nf_ct_put(ct); | |
816 | return dropped; | |
817 | } | |
818 | ||
99f07e91 | 819 | static struct nf_conn * |
308ac914 DB |
820 | __nf_conntrack_alloc(struct net *net, |
821 | const struct nf_conntrack_zone *zone, | |
99f07e91 CG |
822 | const struct nf_conntrack_tuple *orig, |
823 | const struct nf_conntrack_tuple *repl, | |
824 | gfp_t gfp, u32 hash) | |
9fb9cbb1 | 825 | { |
cd7fcbf1 | 826 | struct nf_conn *ct; |
9fb9cbb1 | 827 | |
5251e2d2 | 828 | /* We don't want any race condition at early drop stage */ |
49ac8713 | 829 | atomic_inc(&net->ct.count); |
5251e2d2 | 830 | |
76eb9460 | 831 | if (nf_conntrack_max && |
49ac8713 | 832 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
93bb0ceb | 833 | if (!early_drop(net, hash)) { |
49ac8713 | 834 | atomic_dec(&net->ct.count); |
e87cc472 | 835 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); |
9fb9cbb1 YK |
836 | return ERR_PTR(-ENOMEM); |
837 | } | |
838 | } | |
839 | ||
941297f4 ED |
840 | /* |
841 | * Do not use kmem_cache_zalloc(), as this cache uses | |
842 | * SLAB_DESTROY_BY_RCU. | |
843 | */ | |
5b3501fa | 844 | ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp); |
5e8018fc DB |
845 | if (ct == NULL) |
846 | goto out; | |
847 | ||
440f0d58 | 848 | spin_lock_init(&ct->lock); |
c88130bc | 849 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; |
941297f4 | 850 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; |
c88130bc | 851 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; |
99f07e91 CG |
852 | /* save hash for reusing when confirming */ |
853 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; | |
c41884ce | 854 | ct->status = 0; |
9fb9cbb1 | 855 | /* Don't set timer yet: wait for confirmation */ |
c88130bc | 856 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
c2d9ba9b | 857 | write_pnet(&ct->ct_net, net); |
c41884ce FW |
858 | memset(&ct->__nfct_init_offset[0], 0, |
859 | offsetof(struct nf_conn, proto) - | |
860 | offsetof(struct nf_conn, __nfct_init_offset[0])); | |
5e8018fc DB |
861 | |
862 | if (zone && nf_ct_zone_add(ct, GFP_ATOMIC, zone) < 0) | |
863 | goto out_free; | |
864 | ||
e53376be PNA |
865 | /* Because we use RCU lookups, we set ct_general.use to zero before |
866 | * this is inserted in any list. | |
941297f4 | 867 | */ |
e53376be | 868 | atomic_set(&ct->ct_general.use, 0); |
c88130bc | 869 | return ct; |
5d0aa2cc PM |
870 | out_free: |
871 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); | |
5e8018fc DB |
872 | out: |
873 | atomic_dec(&net->ct.count); | |
5d0aa2cc | 874 | return ERR_PTR(-ENOMEM); |
9fb9cbb1 | 875 | } |
99f07e91 | 876 | |
308ac914 DB |
877 | struct nf_conn *nf_conntrack_alloc(struct net *net, |
878 | const struct nf_conntrack_zone *zone, | |
99f07e91 CG |
879 | const struct nf_conntrack_tuple *orig, |
880 | const struct nf_conntrack_tuple *repl, | |
881 | gfp_t gfp) | |
882 | { | |
883 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); | |
884 | } | |
13b18339 | 885 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
9fb9cbb1 | 886 | |
c88130bc | 887 | void nf_conntrack_free(struct nf_conn *ct) |
76507f69 | 888 | { |
1d45209d ED |
889 | struct net *net = nf_ct_net(ct); |
890 | ||
e53376be PNA |
891 | /* A freed object has refcnt == 0, that's |
892 | * the golden rule for SLAB_DESTROY_BY_RCU | |
893 | */ | |
894 | NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 0); | |
895 | ||
ceeff754 | 896 | nf_ct_ext_destroy(ct); |
ea781f19 | 897 | nf_ct_ext_free(ct); |
5b3501fa | 898 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
4e857c58 | 899 | smp_mb__before_atomic(); |
0c3c6c00 | 900 | atomic_dec(&net->ct.count); |
76507f69 | 901 | } |
13b18339 | 902 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
9fb9cbb1 | 903 | |
c539f017 | 904 | |
9fb9cbb1 YK |
905 | /* Allocate a new conntrack: we return -ENOMEM if classification |
906 | failed due to stress. Otherwise it really is unclassifiable. */ | |
907 | static struct nf_conntrack_tuple_hash * | |
b2a15a60 | 908 | init_conntrack(struct net *net, struct nf_conn *tmpl, |
5a1fb391 | 909 | const struct nf_conntrack_tuple *tuple, |
9fb9cbb1 | 910 | struct nf_conntrack_l3proto *l3proto, |
605dcad6 | 911 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 912 | struct sk_buff *skb, |
60b5f8f7 | 913 | unsigned int dataoff, u32 hash) |
9fb9cbb1 | 914 | { |
c88130bc | 915 | struct nf_conn *ct; |
3c158f7f | 916 | struct nf_conn_help *help; |
9fb9cbb1 | 917 | struct nf_conntrack_tuple repl_tuple; |
b2a15a60 | 918 | struct nf_conntrack_ecache *ecache; |
ca7433df | 919 | struct nf_conntrack_expect *exp = NULL; |
308ac914 | 920 | const struct nf_conntrack_zone *zone; |
60b5f8f7 | 921 | struct nf_conn_timeout *timeout_ext; |
5e8018fc | 922 | struct nf_conntrack_zone tmp; |
60b5f8f7 | 923 | unsigned int *timeouts; |
9fb9cbb1 | 924 | |
605dcad6 | 925 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
0d53778e | 926 | pr_debug("Can't invert tuple.\n"); |
9fb9cbb1 YK |
927 | return NULL; |
928 | } | |
929 | ||
5e8018fc | 930 | zone = nf_ct_zone_tmpl(tmpl, skb, &tmp); |
99f07e91 CG |
931 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, |
932 | hash); | |
0a9ee813 | 933 | if (IS_ERR(ct)) |
c88130bc | 934 | return (struct nf_conntrack_tuple_hash *)ct; |
9fb9cbb1 | 935 | |
48b1de4c PM |
936 | if (tmpl && nfct_synproxy(tmpl)) { |
937 | nfct_seqadj_ext_add(ct); | |
938 | nfct_synproxy_ext_add(ct); | |
939 | } | |
940 | ||
60b5f8f7 | 941 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; |
ae2d708e PNA |
942 | if (timeout_ext) { |
943 | timeouts = nf_ct_timeout_data(timeout_ext); | |
944 | if (unlikely(!timeouts)) | |
945 | timeouts = l4proto->get_timeouts(net); | |
946 | } else { | |
60b5f8f7 | 947 | timeouts = l4proto->get_timeouts(net); |
ae2d708e | 948 | } |
60b5f8f7 | 949 | |
2c8503f5 | 950 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { |
c88130bc | 951 | nf_conntrack_free(ct); |
ccd63c20 | 952 | pr_debug("can't track with proto module\n"); |
9fb9cbb1 YK |
953 | return NULL; |
954 | } | |
955 | ||
60b5f8f7 | 956 | if (timeout_ext) |
ae2d708e PNA |
957 | nf_ct_timeout_ext_add(ct, rcu_dereference(timeout_ext->timeout), |
958 | GFP_ATOMIC); | |
60b5f8f7 | 959 | |
58401572 | 960 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
a992ca2a | 961 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); |
c539f017 | 962 | nf_ct_labels_ext_add(ct); |
b2a15a60 PM |
963 | |
964 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; | |
965 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, | |
966 | ecache ? ecache->expmask : 0, | |
967 | GFP_ATOMIC); | |
58401572 | 968 | |
ca7433df JDB |
969 | local_bh_disable(); |
970 | if (net->ct.expect_count) { | |
971 | spin_lock(&nf_conntrack_expect_lock); | |
972 | exp = nf_ct_find_expectation(net, zone, tuple); | |
973 | if (exp) { | |
ccd63c20 | 974 | pr_debug("expectation arrives ct=%p exp=%p\n", |
ca7433df JDB |
975 | ct, exp); |
976 | /* Welcome, Mr. Bond. We've been expecting you... */ | |
977 | __set_bit(IPS_EXPECTED_BIT, &ct->status); | |
978 | /* exp->master safe, refcnt bumped in nf_ct_find_expectation */ | |
979 | ct->master = exp->master; | |
980 | if (exp->helper) { | |
981 | help = nf_ct_helper_ext_add(ct, exp->helper, | |
982 | GFP_ATOMIC); | |
983 | if (help) | |
984 | rcu_assign_pointer(help->helper, exp->helper); | |
985 | } | |
ceceae1b | 986 | |
9fb9cbb1 | 987 | #ifdef CONFIG_NF_CONNTRACK_MARK |
ca7433df | 988 | ct->mark = exp->master->mark; |
7c9728c3 JM |
989 | #endif |
990 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | |
ca7433df | 991 | ct->secmark = exp->master->secmark; |
9fb9cbb1 | 992 | #endif |
ca7433df JDB |
993 | NF_CT_STAT_INC(net, expect_new); |
994 | } | |
995 | spin_unlock(&nf_conntrack_expect_lock); | |
996 | } | |
997 | if (!exp) { | |
b2a15a60 | 998 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); |
0d55af87 | 999 | NF_CT_STAT_INC(net, new); |
22e7410b | 1000 | } |
9fb9cbb1 | 1001 | |
e53376be PNA |
1002 | /* Now it is inserted into the unconfirmed list, bump refcount */ |
1003 | nf_conntrack_get(&ct->ct_general); | |
b7779d06 | 1004 | nf_ct_add_to_unconfirmed_list(ct); |
9fb9cbb1 | 1005 | |
ca7433df | 1006 | local_bh_enable(); |
9fb9cbb1 YK |
1007 | |
1008 | if (exp) { | |
1009 | if (exp->expectfn) | |
c88130bc | 1010 | exp->expectfn(ct, exp); |
6823645d | 1011 | nf_ct_expect_put(exp); |
9fb9cbb1 YK |
1012 | } |
1013 | ||
c88130bc | 1014 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
9fb9cbb1 YK |
1015 | } |
1016 | ||
1017 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | |
1018 | static inline struct nf_conn * | |
b2a15a60 | 1019 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, |
a702a65f | 1020 | struct sk_buff *skb, |
9fb9cbb1 YK |
1021 | unsigned int dataoff, |
1022 | u_int16_t l3num, | |
1023 | u_int8_t protonum, | |
1024 | struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 1025 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 1026 | int *set_reply, |
60b5f8f7 | 1027 | enum ip_conntrack_info *ctinfo) |
9fb9cbb1 | 1028 | { |
308ac914 | 1029 | const struct nf_conntrack_zone *zone; |
9fb9cbb1 YK |
1030 | struct nf_conntrack_tuple tuple; |
1031 | struct nf_conntrack_tuple_hash *h; | |
5e8018fc | 1032 | struct nf_conntrack_zone tmp; |
9fb9cbb1 | 1033 | struct nf_conn *ct; |
99f07e91 | 1034 | u32 hash; |
9fb9cbb1 | 1035 | |
bbe735e4 | 1036 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
a31f1adc | 1037 | dataoff, l3num, protonum, net, &tuple, l3proto, |
605dcad6 | 1038 | l4proto)) { |
ccd63c20 | 1039 | pr_debug("Can't get tuple\n"); |
9fb9cbb1 YK |
1040 | return NULL; |
1041 | } | |
1042 | ||
1043 | /* look for tuple match */ | |
5e8018fc | 1044 | zone = nf_ct_zone_tmpl(tmpl, skb, &tmp); |
deedb590 | 1045 | hash = hash_conntrack_raw(&tuple); |
99f07e91 | 1046 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); |
9fb9cbb1 | 1047 | if (!h) { |
b2a15a60 | 1048 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, |
60b5f8f7 | 1049 | skb, dataoff, hash); |
9fb9cbb1 YK |
1050 | if (!h) |
1051 | return NULL; | |
1052 | if (IS_ERR(h)) | |
1053 | return (void *)h; | |
1054 | } | |
1055 | ct = nf_ct_tuplehash_to_ctrack(h); | |
1056 | ||
1057 | /* It exists; we have (non-exclusive) reference. */ | |
1058 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | |
fb048833 | 1059 | *ctinfo = IP_CT_ESTABLISHED_REPLY; |
9fb9cbb1 YK |
1060 | /* Please set reply bit if this packet OK */ |
1061 | *set_reply = 1; | |
1062 | } else { | |
1063 | /* Once we've had two way comms, always ESTABLISHED. */ | |
1064 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | |
ccd63c20 | 1065 | pr_debug("normal packet for %p\n", ct); |
9fb9cbb1 YK |
1066 | *ctinfo = IP_CT_ESTABLISHED; |
1067 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | |
ccd63c20 | 1068 | pr_debug("related packet for %p\n", ct); |
9fb9cbb1 YK |
1069 | *ctinfo = IP_CT_RELATED; |
1070 | } else { | |
ccd63c20 | 1071 | pr_debug("new packet for %p\n", ct); |
9fb9cbb1 YK |
1072 | *ctinfo = IP_CT_NEW; |
1073 | } | |
1074 | *set_reply = 0; | |
1075 | } | |
1076 | skb->nfct = &ct->ct_general; | |
1077 | skb->nfctinfo = *ctinfo; | |
1078 | return ct; | |
1079 | } | |
1080 | ||
1081 | unsigned int | |
a702a65f AD |
1082 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
1083 | struct sk_buff *skb) | |
9fb9cbb1 | 1084 | { |
b2a15a60 | 1085 | struct nf_conn *ct, *tmpl = NULL; |
9fb9cbb1 YK |
1086 | enum ip_conntrack_info ctinfo; |
1087 | struct nf_conntrack_l3proto *l3proto; | |
605dcad6 | 1088 | struct nf_conntrack_l4proto *l4proto; |
2c8503f5 | 1089 | unsigned int *timeouts; |
9fb9cbb1 YK |
1090 | unsigned int dataoff; |
1091 | u_int8_t protonum; | |
1092 | int set_reply = 0; | |
1093 | int ret; | |
1094 | ||
3db05fea | 1095 | if (skb->nfct) { |
b2a15a60 PM |
1096 | /* Previously seen (loopback or untracked)? Ignore. */ |
1097 | tmpl = (struct nf_conn *)skb->nfct; | |
1098 | if (!nf_ct_is_template(tmpl)) { | |
1099 | NF_CT_STAT_INC_ATOMIC(net, ignore); | |
1100 | return NF_ACCEPT; | |
1101 | } | |
1102 | skb->nfct = NULL; | |
9fb9cbb1 YK |
1103 | } |
1104 | ||
923f4902 | 1105 | /* rcu_read_lock()ed by nf_hook_slow */ |
76108cea | 1106 | l3proto = __nf_ct_l3proto_find(pf); |
3db05fea | 1107 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
ffc30690 YK |
1108 | &dataoff, &protonum); |
1109 | if (ret <= 0) { | |
25985edc | 1110 | pr_debug("not prepared to track yet or error occurred\n"); |
0d55af87 AD |
1111 | NF_CT_STAT_INC_ATOMIC(net, error); |
1112 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
1113 | ret = -ret; |
1114 | goto out; | |
9fb9cbb1 YK |
1115 | } |
1116 | ||
76108cea | 1117 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
9fb9cbb1 YK |
1118 | |
1119 | /* It may be an special packet, error, unclean... | |
1120 | * inverse of the return code tells to the netfilter | |
1121 | * core what to do with the packet. */ | |
74c51a14 | 1122 | if (l4proto->error != NULL) { |
8fea97ec PM |
1123 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, |
1124 | pf, hooknum); | |
74c51a14 | 1125 | if (ret <= 0) { |
0d55af87 AD |
1126 | NF_CT_STAT_INC_ATOMIC(net, error); |
1127 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
1128 | ret = -ret; |
1129 | goto out; | |
74c51a14 | 1130 | } |
88ed01d1 PNA |
1131 | /* ICMP[v6] protocol trackers may assign one conntrack. */ |
1132 | if (skb->nfct) | |
1133 | goto out; | |
9fb9cbb1 YK |
1134 | } |
1135 | ||
b2a15a60 | 1136 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, |
60b5f8f7 | 1137 | l3proto, l4proto, &set_reply, &ctinfo); |
9fb9cbb1 YK |
1138 | if (!ct) { |
1139 | /* Not valid part of a connection */ | |
0d55af87 | 1140 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
b2a15a60 PM |
1141 | ret = NF_ACCEPT; |
1142 | goto out; | |
9fb9cbb1 YK |
1143 | } |
1144 | ||
1145 | if (IS_ERR(ct)) { | |
1146 | /* Too stressed to deal. */ | |
0d55af87 | 1147 | NF_CT_STAT_INC_ATOMIC(net, drop); |
b2a15a60 PM |
1148 | ret = NF_DROP; |
1149 | goto out; | |
9fb9cbb1 YK |
1150 | } |
1151 | ||
3db05fea | 1152 | NF_CT_ASSERT(skb->nfct); |
9fb9cbb1 | 1153 | |
60b5f8f7 | 1154 | /* Decide what timeout policy we want to apply to this flow. */ |
84b5ee93 | 1155 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); |
60b5f8f7 | 1156 | |
2c8503f5 | 1157 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); |
ec8d5409 | 1158 | if (ret <= 0) { |
9fb9cbb1 YK |
1159 | /* Invalid: inverse of the return code tells |
1160 | * the netfilter core what to do */ | |
0d53778e | 1161 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
3db05fea HX |
1162 | nf_conntrack_put(skb->nfct); |
1163 | skb->nfct = NULL; | |
0d55af87 | 1164 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
7d1e0459 PNA |
1165 | if (ret == -NF_DROP) |
1166 | NF_CT_STAT_INC_ATOMIC(net, drop); | |
b2a15a60 PM |
1167 | ret = -ret; |
1168 | goto out; | |
9fb9cbb1 YK |
1169 | } |
1170 | ||
1171 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
858b3133 | 1172 | nf_conntrack_event_cache(IPCT_REPLY, ct); |
b2a15a60 | 1173 | out: |
c3174286 PNA |
1174 | if (tmpl) { |
1175 | /* Special case: we have to repeat this hook, assign the | |
1176 | * template again to this packet. We assume that this packet | |
1177 | * has no conntrack assigned. This is used by nf_ct_tcp. */ | |
1178 | if (ret == NF_REPEAT) | |
1179 | skb->nfct = (struct nf_conntrack *)tmpl; | |
1180 | else | |
1181 | nf_ct_put(tmpl); | |
1182 | } | |
9fb9cbb1 YK |
1183 | |
1184 | return ret; | |
1185 | } | |
13b18339 | 1186 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
9fb9cbb1 | 1187 | |
5f2b4c90 JE |
1188 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
1189 | const struct nf_conntrack_tuple *orig) | |
9fb9cbb1 | 1190 | { |
5f2b4c90 | 1191 | bool ret; |
923f4902 PM |
1192 | |
1193 | rcu_read_lock(); | |
1194 | ret = nf_ct_invert_tuple(inverse, orig, | |
1195 | __nf_ct_l3proto_find(orig->src.l3num), | |
1196 | __nf_ct_l4proto_find(orig->src.l3num, | |
1197 | orig->dst.protonum)); | |
1198 | rcu_read_unlock(); | |
1199 | return ret; | |
9fb9cbb1 | 1200 | } |
13b18339 | 1201 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
9fb9cbb1 | 1202 | |
5b1158e9 JK |
1203 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
1204 | implicitly racy: see __nf_conntrack_confirm */ | |
1205 | void nf_conntrack_alter_reply(struct nf_conn *ct, | |
1206 | const struct nf_conntrack_tuple *newreply) | |
1207 | { | |
1208 | struct nf_conn_help *help = nfct_help(ct); | |
1209 | ||
5b1158e9 JK |
1210 | /* Should be unconfirmed, so not in hash table yet */ |
1211 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
1212 | ||
0d53778e | 1213 | pr_debug("Altering reply tuple of %p to ", ct); |
3c9fba65 | 1214 | nf_ct_dump_tuple(newreply); |
5b1158e9 JK |
1215 | |
1216 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | |
ef1a5a50 | 1217 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
c52fbb41 | 1218 | return; |
ceceae1b | 1219 | |
c52fbb41 | 1220 | rcu_read_lock(); |
b2a15a60 | 1221 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); |
c52fbb41 | 1222 | rcu_read_unlock(); |
5b1158e9 | 1223 | } |
13b18339 | 1224 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
5b1158e9 | 1225 | |
9fb9cbb1 YK |
1226 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
1227 | void __nf_ct_refresh_acct(struct nf_conn *ct, | |
1228 | enum ip_conntrack_info ctinfo, | |
1229 | const struct sk_buff *skb, | |
1230 | unsigned long extra_jiffies, | |
1231 | int do_acct) | |
1232 | { | |
9fb9cbb1 YK |
1233 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
1234 | NF_CT_ASSERT(skb); | |
1235 | ||
997ae831 | 1236 | /* Only update if this is not a fixed timeout */ |
47d95045 PM |
1237 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
1238 | goto acct; | |
997ae831 | 1239 | |
9fb9cbb1 YK |
1240 | /* If not in hash table, timer will not be active yet */ |
1241 | if (!nf_ct_is_confirmed(ct)) { | |
1242 | ct->timeout.expires = extra_jiffies; | |
9fb9cbb1 | 1243 | } else { |
be00c8e4 MJ |
1244 | unsigned long newtime = jiffies + extra_jiffies; |
1245 | ||
1246 | /* Only update the timeout if the new timeout is at least | |
1247 | HZ jiffies from the old timeout. Need del_timer for race | |
1248 | avoidance (may already be dying). */ | |
65cb9fda PM |
1249 | if (newtime - ct->timeout.expires >= HZ) |
1250 | mod_timer_pending(&ct->timeout, newtime); | |
9fb9cbb1 YK |
1251 | } |
1252 | ||
47d95045 | 1253 | acct: |
9fb9cbb1 | 1254 | if (do_acct) { |
f7b13e43 | 1255 | struct nf_conn_acct *acct; |
3ffd5eeb | 1256 | |
58401572 KPO |
1257 | acct = nf_conn_acct_find(ct); |
1258 | if (acct) { | |
f7b13e43 HE |
1259 | struct nf_conn_counter *counter = acct->counter; |
1260 | ||
1261 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); | |
1262 | atomic64_add(skb->len, &counter[CTINFO2DIR(ctinfo)].bytes); | |
58401572 | 1263 | } |
9fb9cbb1 | 1264 | } |
9fb9cbb1 | 1265 | } |
13b18339 | 1266 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
9fb9cbb1 | 1267 | |
4c889498 DM |
1268 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
1269 | enum ip_conntrack_info ctinfo, | |
1270 | const struct sk_buff *skb, | |
1271 | int do_acct) | |
51091764 | 1272 | { |
718d4ad9 | 1273 | if (do_acct) { |
f7b13e43 | 1274 | struct nf_conn_acct *acct; |
58401572 | 1275 | |
58401572 KPO |
1276 | acct = nf_conn_acct_find(ct); |
1277 | if (acct) { | |
f7b13e43 HE |
1278 | struct nf_conn_counter *counter = acct->counter; |
1279 | ||
1280 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); | |
b3e0bfa7 | 1281 | atomic64_add(skb->len - skb_network_offset(skb), |
f7b13e43 | 1282 | &counter[CTINFO2DIR(ctinfo)].bytes); |
58401572 | 1283 | } |
718d4ad9 | 1284 | } |
58401572 | 1285 | |
4c889498 | 1286 | if (del_timer(&ct->timeout)) { |
51091764 | 1287 | ct->timeout.function((unsigned long)ct); |
4c889498 DM |
1288 | return true; |
1289 | } | |
1290 | return false; | |
51091764 | 1291 | } |
718d4ad9 | 1292 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
51091764 | 1293 | |
5d0aa2cc PM |
1294 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1295 | static struct nf_ct_ext_type nf_ct_zone_extend __read_mostly = { | |
1296 | .len = sizeof(struct nf_conntrack_zone), | |
1297 | .align = __alignof__(struct nf_conntrack_zone), | |
1298 | .id = NF_CT_EXT_ZONE, | |
1299 | }; | |
1300 | #endif | |
1301 | ||
c0cd1156 | 1302 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
c1d10adb PNA |
1303 | |
1304 | #include <linux/netfilter/nfnetlink.h> | |
1305 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
57b47a53 IM |
1306 | #include <linux/mutex.h> |
1307 | ||
c1d10adb PNA |
1308 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
1309 | * in ip_conntrack_core, since we don't want the protocols to autoload | |
1310 | * or depend on ctnetlink */ | |
fdf70832 | 1311 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
c1d10adb PNA |
1312 | const struct nf_conntrack_tuple *tuple) |
1313 | { | |
bae65be8 DM |
1314 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || |
1315 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) | |
1316 | goto nla_put_failure; | |
c1d10adb PNA |
1317 | return 0; |
1318 | ||
df6fb868 | 1319 | nla_put_failure: |
c1d10adb PNA |
1320 | return -1; |
1321 | } | |
fdf70832 | 1322 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
c1d10adb | 1323 | |
f73e924c PM |
1324 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
1325 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | |
1326 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | |
c1d10adb | 1327 | }; |
f73e924c | 1328 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
c1d10adb | 1329 | |
fdf70832 | 1330 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
c1d10adb PNA |
1331 | struct nf_conntrack_tuple *t) |
1332 | { | |
df6fb868 | 1333 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
c1d10adb PNA |
1334 | return -EINVAL; |
1335 | ||
77236b6e PM |
1336 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
1337 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | |
c1d10adb PNA |
1338 | |
1339 | return 0; | |
1340 | } | |
fdf70832 | 1341 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
5c0de29d HE |
1342 | |
1343 | int nf_ct_port_nlattr_tuple_size(void) | |
1344 | { | |
1345 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
1346 | } | |
1347 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | |
c1d10adb PNA |
1348 | #endif |
1349 | ||
9fb9cbb1 | 1350 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
312a0c16 | 1351 | static void nf_conntrack_attach(struct sk_buff *nskb, const struct sk_buff *skb) |
9fb9cbb1 YK |
1352 | { |
1353 | struct nf_conn *ct; | |
1354 | enum ip_conntrack_info ctinfo; | |
1355 | ||
1356 | /* This ICMP is in reverse direction to the packet which caused it */ | |
1357 | ct = nf_ct_get(skb, &ctinfo); | |
1358 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | |
fb048833 | 1359 | ctinfo = IP_CT_RELATED_REPLY; |
9fb9cbb1 YK |
1360 | else |
1361 | ctinfo = IP_CT_RELATED; | |
1362 | ||
1363 | /* Attach to new skbuff, and increment count */ | |
1364 | nskb->nfct = &ct->ct_general; | |
1365 | nskb->nfctinfo = ctinfo; | |
1366 | nf_conntrack_get(nskb->nfct); | |
1367 | } | |
1368 | ||
9fb9cbb1 | 1369 | /* Bring out ya dead! */ |
df0933dc | 1370 | static struct nf_conn * |
400dad39 | 1371 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
9fb9cbb1 YK |
1372 | void *data, unsigned int *bucket) |
1373 | { | |
df0933dc PM |
1374 | struct nf_conntrack_tuple_hash *h; |
1375 | struct nf_conn *ct; | |
ea781f19 | 1376 | struct hlist_nulls_node *n; |
b7779d06 | 1377 | int cpu; |
93bb0ceb | 1378 | spinlock_t *lockp; |
9fb9cbb1 | 1379 | |
d696c7bd | 1380 | for (; *bucket < net->ct.htable_size; (*bucket)++) { |
93bb0ceb JDB |
1381 | lockp = &nf_conntrack_locks[*bucket % CONNTRACK_LOCKS]; |
1382 | local_bh_disable(); | |
b16c2919 | 1383 | nf_conntrack_lock(lockp); |
93bb0ceb JDB |
1384 | if (*bucket < net->ct.htable_size) { |
1385 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { | |
1386 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) | |
1387 | continue; | |
1388 | ct = nf_ct_tuplehash_to_ctrack(h); | |
1389 | if (iter(ct, data)) | |
1390 | goto found; | |
1391 | } | |
df0933dc | 1392 | } |
93bb0ceb JDB |
1393 | spin_unlock(lockp); |
1394 | local_bh_enable(); | |
d93c6258 | 1395 | cond_resched(); |
601e68e1 | 1396 | } |
b7779d06 JDB |
1397 | |
1398 | for_each_possible_cpu(cpu) { | |
1399 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | |
1400 | ||
1401 | spin_lock_bh(&pcpu->lock); | |
1402 | hlist_nulls_for_each_entry(h, n, &pcpu->unconfirmed, hnnode) { | |
1403 | ct = nf_ct_tuplehash_to_ctrack(h); | |
1404 | if (iter(ct, data)) | |
1405 | set_bit(IPS_DYING_BIT, &ct->status); | |
1406 | } | |
1407 | spin_unlock_bh(&pcpu->lock); | |
d93c6258 | 1408 | cond_resched(); |
b7779d06 | 1409 | } |
df0933dc PM |
1410 | return NULL; |
1411 | found: | |
c073e3fa | 1412 | atomic_inc(&ct->ct_general.use); |
93bb0ceb JDB |
1413 | spin_unlock(lockp); |
1414 | local_bh_enable(); | |
df0933dc | 1415 | return ct; |
9fb9cbb1 YK |
1416 | } |
1417 | ||
400dad39 AD |
1418 | void nf_ct_iterate_cleanup(struct net *net, |
1419 | int (*iter)(struct nf_conn *i, void *data), | |
c655bc68 | 1420 | void *data, u32 portid, int report) |
9fb9cbb1 | 1421 | { |
df0933dc | 1422 | struct nf_conn *ct; |
9fb9cbb1 YK |
1423 | unsigned int bucket = 0; |
1424 | ||
d93c6258 FW |
1425 | might_sleep(); |
1426 | ||
400dad39 | 1427 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
9fb9cbb1 YK |
1428 | /* Time to push up daises... */ |
1429 | if (del_timer(&ct->timeout)) | |
c655bc68 | 1430 | nf_ct_delete(ct, portid, report); |
02982c27 | 1431 | |
9fb9cbb1 YK |
1432 | /* ... else the timer will get him soon. */ |
1433 | ||
1434 | nf_ct_put(ct); | |
d93c6258 | 1435 | cond_resched(); |
9fb9cbb1 YK |
1436 | } |
1437 | } | |
13b18339 | 1438 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
9fb9cbb1 | 1439 | |
274d383b PNA |
1440 | static int kill_all(struct nf_conn *i, void *data) |
1441 | { | |
1442 | return 1; | |
1443 | } | |
1444 | ||
d862a662 | 1445 | void nf_ct_free_hashtable(void *hash, unsigned int size) |
9fb9cbb1 | 1446 | { |
d862a662 | 1447 | if (is_vmalloc_addr(hash)) |
9fb9cbb1 YK |
1448 | vfree(hash); |
1449 | else | |
601e68e1 | 1450 | free_pages((unsigned long)hash, |
f205c5e0 | 1451 | get_order(sizeof(struct hlist_head) * size)); |
9fb9cbb1 | 1452 | } |
ac565e5f | 1453 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
9fb9cbb1 | 1454 | |
b3c5163f ED |
1455 | static int untrack_refs(void) |
1456 | { | |
1457 | int cnt = 0, cpu; | |
1458 | ||
1459 | for_each_possible_cpu(cpu) { | |
1460 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
1461 | ||
1462 | cnt += atomic_read(&ct->ct_general.use) - 1; | |
1463 | } | |
1464 | return cnt; | |
1465 | } | |
1466 | ||
f94161c1 | 1467 | void nf_conntrack_cleanup_start(void) |
9fb9cbb1 | 1468 | { |
f94161c1 G |
1469 | RCU_INIT_POINTER(ip_ct_attach, NULL); |
1470 | } | |
1471 | ||
1472 | void nf_conntrack_cleanup_end(void) | |
1473 | { | |
1474 | RCU_INIT_POINTER(nf_ct_destroy, NULL); | |
b3c5163f | 1475 | while (untrack_refs() > 0) |
9edd7ca0 PM |
1476 | schedule(); |
1477 | ||
5d0aa2cc PM |
1478 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1479 | nf_ct_extend_unregister(&nf_ct_zone_extend); | |
1480 | #endif | |
04d87001 | 1481 | nf_conntrack_proto_fini(); |
41d73ec0 | 1482 | nf_conntrack_seqadj_fini(); |
5f69b8f5 | 1483 | nf_conntrack_labels_fini(); |
5e615b22 | 1484 | nf_conntrack_helper_fini(); |
8684094c | 1485 | nf_conntrack_timeout_fini(); |
3fe0f943 | 1486 | nf_conntrack_ecache_fini(); |
73f4001a | 1487 | nf_conntrack_tstamp_fini(); |
b7ff3a1f | 1488 | nf_conntrack_acct_fini(); |
83b4dbe1 | 1489 | nf_conntrack_expect_fini(); |
08f6547d | 1490 | } |
9fb9cbb1 | 1491 | |
f94161c1 G |
1492 | /* |
1493 | * Mishearing the voices in his head, our hero wonders how he's | |
1494 | * supposed to kill the mall. | |
1495 | */ | |
1496 | void nf_conntrack_cleanup_net(struct net *net) | |
08f6547d | 1497 | { |
dece40e8 VD |
1498 | LIST_HEAD(single); |
1499 | ||
1500 | list_add(&net->exit_list, &single); | |
1501 | nf_conntrack_cleanup_net_list(&single); | |
1502 | } | |
1503 | ||
1504 | void nf_conntrack_cleanup_net_list(struct list_head *net_exit_list) | |
1505 | { | |
1506 | int busy; | |
1507 | struct net *net; | |
1508 | ||
f94161c1 G |
1509 | /* |
1510 | * This makes sure all current packets have passed through | |
1511 | * netfilter framework. Roll on, two-stage module | |
1512 | * delete... | |
1513 | */ | |
1514 | synchronize_net(); | |
dece40e8 VD |
1515 | i_see_dead_people: |
1516 | busy = 0; | |
1517 | list_for_each_entry(net, net_exit_list, exit_list) { | |
c655bc68 | 1518 | nf_ct_iterate_cleanup(net, kill_all, NULL, 0, 0); |
dece40e8 VD |
1519 | if (atomic_read(&net->ct.count) != 0) |
1520 | busy = 1; | |
1521 | } | |
1522 | if (busy) { | |
9fb9cbb1 YK |
1523 | schedule(); |
1524 | goto i_see_dead_people; | |
1525 | } | |
1526 | ||
dece40e8 VD |
1527 | list_for_each_entry(net, net_exit_list, exit_list) { |
1528 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); | |
1529 | nf_conntrack_proto_pernet_fini(net); | |
1530 | nf_conntrack_helper_pernet_fini(net); | |
1531 | nf_conntrack_ecache_pernet_fini(net); | |
1532 | nf_conntrack_tstamp_pernet_fini(net); | |
1533 | nf_conntrack_acct_pernet_fini(net); | |
1534 | nf_conntrack_expect_pernet_fini(net); | |
1535 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); | |
1536 | kfree(net->ct.slabname); | |
1537 | free_percpu(net->ct.stat); | |
b7779d06 | 1538 | free_percpu(net->ct.pcpu_lists); |
dece40e8 | 1539 | } |
08f6547d AD |
1540 | } |
1541 | ||
d862a662 | 1542 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) |
9fb9cbb1 | 1543 | { |
ea781f19 ED |
1544 | struct hlist_nulls_head *hash; |
1545 | unsigned int nr_slots, i; | |
1546 | size_t sz; | |
9fb9cbb1 | 1547 | |
ea781f19 ED |
1548 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1549 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | |
1550 | sz = nr_slots * sizeof(struct hlist_nulls_head); | |
1551 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | |
1552 | get_order(sz)); | |
f0ad4621 | 1553 | if (!hash) |
966567b7 | 1554 | hash = vzalloc(sz); |
9fb9cbb1 | 1555 | |
ea781f19 ED |
1556 | if (hash && nulls) |
1557 | for (i = 0; i < nr_slots; i++) | |
1558 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | |
9fb9cbb1 YK |
1559 | |
1560 | return hash; | |
1561 | } | |
ac565e5f | 1562 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
9fb9cbb1 | 1563 | |
fae718dd | 1564 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
9fb9cbb1 | 1565 | { |
4b5511eb | 1566 | int i, bucket, rc; |
96eb24d7 | 1567 | unsigned int hashsize, old_size; |
ea781f19 | 1568 | struct hlist_nulls_head *hash, *old_hash; |
9fb9cbb1 | 1569 | struct nf_conntrack_tuple_hash *h; |
5d0aa2cc | 1570 | struct nf_conn *ct; |
9fb9cbb1 | 1571 | |
d696c7bd PM |
1572 | if (current->nsproxy->net_ns != &init_net) |
1573 | return -EOPNOTSUPP; | |
1574 | ||
9fb9cbb1 YK |
1575 | /* On boot, we can set this without any fancy locking. */ |
1576 | if (!nf_conntrack_htable_size) | |
1577 | return param_set_uint(val, kp); | |
1578 | ||
4b5511eb AP |
1579 | rc = kstrtouint(val, 0, &hashsize); |
1580 | if (rc) | |
1581 | return rc; | |
9fb9cbb1 YK |
1582 | if (!hashsize) |
1583 | return -EINVAL; | |
1584 | ||
d862a662 | 1585 | hash = nf_ct_alloc_hashtable(&hashsize, 1); |
9fb9cbb1 YK |
1586 | if (!hash) |
1587 | return -ENOMEM; | |
1588 | ||
93bb0ceb JDB |
1589 | local_bh_disable(); |
1590 | nf_conntrack_all_lock(); | |
a3efd812 | 1591 | write_seqcount_begin(&nf_conntrack_generation); |
93bb0ceb | 1592 | |
76507f69 PM |
1593 | /* Lookups in the old hash might happen in parallel, which means we |
1594 | * might get false negatives during connection lookup. New connections | |
1595 | * created because of a false negative won't make it into the hash | |
93bb0ceb | 1596 | * though since that required taking the locks. |
76507f69 | 1597 | */ |
93bb0ceb | 1598 | |
d696c7bd | 1599 | for (i = 0; i < init_net.ct.htable_size; i++) { |
ea781f19 ED |
1600 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
1601 | h = hlist_nulls_entry(init_net.ct.hash[i].first, | |
1602 | struct nf_conntrack_tuple_hash, hnnode); | |
5d0aa2cc | 1603 | ct = nf_ct_tuplehash_to_ctrack(h); |
ea781f19 | 1604 | hlist_nulls_del_rcu(&h->hnnode); |
deedb590 | 1605 | bucket = __hash_conntrack(&h->tuple, hashsize); |
ea781f19 | 1606 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
9fb9cbb1 YK |
1607 | } |
1608 | } | |
d696c7bd | 1609 | old_size = init_net.ct.htable_size; |
400dad39 | 1610 | old_hash = init_net.ct.hash; |
9fb9cbb1 | 1611 | |
d696c7bd | 1612 | init_net.ct.htable_size = nf_conntrack_htable_size = hashsize; |
400dad39 | 1613 | init_net.ct.hash = hash; |
93bb0ceb | 1614 | |
a3efd812 | 1615 | write_seqcount_end(&nf_conntrack_generation); |
93bb0ceb JDB |
1616 | nf_conntrack_all_unlock(); |
1617 | local_bh_enable(); | |
9fb9cbb1 | 1618 | |
d862a662 | 1619 | nf_ct_free_hashtable(old_hash, old_size); |
9fb9cbb1 YK |
1620 | return 0; |
1621 | } | |
fae718dd | 1622 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
9fb9cbb1 | 1623 | |
fae718dd | 1624 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
9fb9cbb1 YK |
1625 | &nf_conntrack_htable_size, 0600); |
1626 | ||
5bfddbd4 ED |
1627 | void nf_ct_untracked_status_or(unsigned long bits) |
1628 | { | |
b3c5163f ED |
1629 | int cpu; |
1630 | ||
1631 | for_each_possible_cpu(cpu) | |
1632 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; | |
5bfddbd4 ED |
1633 | } |
1634 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); | |
1635 | ||
f94161c1 | 1636 | int nf_conntrack_init_start(void) |
9fb9cbb1 | 1637 | { |
f205c5e0 | 1638 | int max_factor = 8; |
93bb0ceb JDB |
1639 | int i, ret, cpu; |
1640 | ||
a3efd812 FW |
1641 | seqcount_init(&nf_conntrack_generation); |
1642 | ||
d5d20912 | 1643 | for (i = 0; i < CONNTRACK_LOCKS; i++) |
93bb0ceb | 1644 | spin_lock_init(&nf_conntrack_locks[i]); |
9fb9cbb1 | 1645 | |
9fb9cbb1 | 1646 | if (!nf_conntrack_htable_size) { |
88eab472 ML |
1647 | /* Idea from tcp.c: use 1/16384 of memory. |
1648 | * On i386: 32MB machine has 512 buckets. | |
1649 | * >= 1GB machines have 16384 buckets. | |
1650 | * >= 4GB machines have 65536 buckets. | |
1651 | */ | |
9fb9cbb1 | 1652 | nf_conntrack_htable_size |
4481374c | 1653 | = (((totalram_pages << PAGE_SHIFT) / 16384) |
f205c5e0 | 1654 | / sizeof(struct hlist_head)); |
88eab472 ML |
1655 | if (totalram_pages > (4 * (1024 * 1024 * 1024 / PAGE_SIZE))) |
1656 | nf_conntrack_htable_size = 65536; | |
1657 | else if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) | |
f205c5e0 PM |
1658 | nf_conntrack_htable_size = 16384; |
1659 | if (nf_conntrack_htable_size < 32) | |
1660 | nf_conntrack_htable_size = 32; | |
1661 | ||
1662 | /* Use a max. factor of four by default to get the same max as | |
1663 | * with the old struct list_heads. When a table size is given | |
1664 | * we use the old value of 8 to avoid reducing the max. | |
1665 | * entries. */ | |
1666 | max_factor = 4; | |
9fb9cbb1 | 1667 | } |
f205c5e0 | 1668 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
8e5105a0 | 1669 | |
654d0fbd | 1670 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", |
8e5105a0 PM |
1671 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, |
1672 | nf_conntrack_max); | |
83b4dbe1 G |
1673 | |
1674 | ret = nf_conntrack_expect_init(); | |
1675 | if (ret < 0) | |
1676 | goto err_expect; | |
1677 | ||
b7ff3a1f G |
1678 | ret = nf_conntrack_acct_init(); |
1679 | if (ret < 0) | |
1680 | goto err_acct; | |
1681 | ||
73f4001a G |
1682 | ret = nf_conntrack_tstamp_init(); |
1683 | if (ret < 0) | |
1684 | goto err_tstamp; | |
1685 | ||
3fe0f943 G |
1686 | ret = nf_conntrack_ecache_init(); |
1687 | if (ret < 0) | |
1688 | goto err_ecache; | |
1689 | ||
8684094c G |
1690 | ret = nf_conntrack_timeout_init(); |
1691 | if (ret < 0) | |
1692 | goto err_timeout; | |
1693 | ||
5e615b22 G |
1694 | ret = nf_conntrack_helper_init(); |
1695 | if (ret < 0) | |
1696 | goto err_helper; | |
1697 | ||
5f69b8f5 G |
1698 | ret = nf_conntrack_labels_init(); |
1699 | if (ret < 0) | |
1700 | goto err_labels; | |
1701 | ||
41d73ec0 PM |
1702 | ret = nf_conntrack_seqadj_init(); |
1703 | if (ret < 0) | |
1704 | goto err_seqadj; | |
1705 | ||
5d0aa2cc PM |
1706 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1707 | ret = nf_ct_extend_register(&nf_ct_zone_extend); | |
1708 | if (ret < 0) | |
1709 | goto err_extend; | |
1710 | #endif | |
04d87001 G |
1711 | ret = nf_conntrack_proto_init(); |
1712 | if (ret < 0) | |
1713 | goto err_proto; | |
1714 | ||
9edd7ca0 | 1715 | /* Set up fake conntrack: to never be deleted, not in any hashes */ |
b3c5163f ED |
1716 | for_each_possible_cpu(cpu) { |
1717 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
b3c5163f ED |
1718 | write_pnet(&ct->ct_net, &init_net); |
1719 | atomic_set(&ct->ct_general.use, 1); | |
1720 | } | |
9edd7ca0 | 1721 | /* - and look it like as a confirmed connection */ |
5bfddbd4 | 1722 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); |
08f6547d AD |
1723 | return 0; |
1724 | ||
04d87001 | 1725 | err_proto: |
5d0aa2cc | 1726 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
04d87001 | 1727 | nf_ct_extend_unregister(&nf_ct_zone_extend); |
5d0aa2cc | 1728 | err_extend: |
a9006892 | 1729 | #endif |
41d73ec0 PM |
1730 | nf_conntrack_seqadj_fini(); |
1731 | err_seqadj: | |
04d87001 | 1732 | nf_conntrack_labels_fini(); |
5f69b8f5 G |
1733 | err_labels: |
1734 | nf_conntrack_helper_fini(); | |
5e615b22 G |
1735 | err_helper: |
1736 | nf_conntrack_timeout_fini(); | |
8684094c G |
1737 | err_timeout: |
1738 | nf_conntrack_ecache_fini(); | |
3fe0f943 G |
1739 | err_ecache: |
1740 | nf_conntrack_tstamp_fini(); | |
73f4001a G |
1741 | err_tstamp: |
1742 | nf_conntrack_acct_fini(); | |
b7ff3a1f G |
1743 | err_acct: |
1744 | nf_conntrack_expect_fini(); | |
83b4dbe1 | 1745 | err_expect: |
08f6547d AD |
1746 | return ret; |
1747 | } | |
1748 | ||
f94161c1 G |
1749 | void nf_conntrack_init_end(void) |
1750 | { | |
1751 | /* For use by REJECT target */ | |
1752 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); | |
1753 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); | |
f94161c1 G |
1754 | } |
1755 | ||
8cc20198 ED |
1756 | /* |
1757 | * We need to use special "null" values, not used in hash table | |
1758 | */ | |
1759 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) | |
1760 | #define DYING_NULLS_VAL ((1<<30)+1) | |
252b3e8c | 1761 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) |
8cc20198 | 1762 | |
f94161c1 | 1763 | int nf_conntrack_init_net(struct net *net) |
08f6547d | 1764 | { |
b7779d06 JDB |
1765 | int ret = -ENOMEM; |
1766 | int cpu; | |
ceceae1b | 1767 | |
08f6547d | 1768 | atomic_set(&net->ct.count, 0); |
b7779d06 JDB |
1769 | |
1770 | net->ct.pcpu_lists = alloc_percpu(struct ct_pcpu); | |
1771 | if (!net->ct.pcpu_lists) | |
08f6547d | 1772 | goto err_stat; |
b7779d06 JDB |
1773 | |
1774 | for_each_possible_cpu(cpu) { | |
1775 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | |
1776 | ||
1777 | spin_lock_init(&pcpu->lock); | |
1778 | INIT_HLIST_NULLS_HEAD(&pcpu->unconfirmed, UNCONFIRMED_NULLS_VAL); | |
1779 | INIT_HLIST_NULLS_HEAD(&pcpu->dying, DYING_NULLS_VAL); | |
08f6547d | 1780 | } |
5b3501fa | 1781 | |
b7779d06 JDB |
1782 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1783 | if (!net->ct.stat) | |
1784 | goto err_pcpu_lists; | |
1785 | ||
5b3501fa | 1786 | net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net); |
b7779d06 | 1787 | if (!net->ct.slabname) |
5b3501fa | 1788 | goto err_slabname; |
5b3501fa ED |
1789 | |
1790 | net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname, | |
1791 | sizeof(struct nf_conn), 0, | |
1792 | SLAB_DESTROY_BY_RCU, NULL); | |
1793 | if (!net->ct.nf_conntrack_cachep) { | |
1794 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); | |
5b3501fa ED |
1795 | goto err_cache; |
1796 | } | |
d696c7bd PM |
1797 | |
1798 | net->ct.htable_size = nf_conntrack_htable_size; | |
d862a662 | 1799 | net->ct.hash = nf_ct_alloc_hashtable(&net->ct.htable_size, 1); |
08f6547d | 1800 | if (!net->ct.hash) { |
08f6547d AD |
1801 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); |
1802 | goto err_hash; | |
1803 | } | |
83b4dbe1 | 1804 | ret = nf_conntrack_expect_pernet_init(net); |
08f6547d AD |
1805 | if (ret < 0) |
1806 | goto err_expect; | |
b7ff3a1f | 1807 | ret = nf_conntrack_acct_pernet_init(net); |
58401572 | 1808 | if (ret < 0) |
08f6547d | 1809 | goto err_acct; |
73f4001a | 1810 | ret = nf_conntrack_tstamp_pernet_init(net); |
a992ca2a PNA |
1811 | if (ret < 0) |
1812 | goto err_tstamp; | |
3fe0f943 | 1813 | ret = nf_conntrack_ecache_pernet_init(net); |
a0891aa6 PNA |
1814 | if (ret < 0) |
1815 | goto err_ecache; | |
5e615b22 | 1816 | ret = nf_conntrack_helper_pernet_init(net); |
a9006892 EL |
1817 | if (ret < 0) |
1818 | goto err_helper; | |
04d87001 | 1819 | ret = nf_conntrack_proto_pernet_init(net); |
f94161c1 G |
1820 | if (ret < 0) |
1821 | goto err_proto; | |
08f6547d | 1822 | return 0; |
c539f017 | 1823 | |
f94161c1 | 1824 | err_proto: |
5e615b22 | 1825 | nf_conntrack_helper_pernet_fini(net); |
a9006892 | 1826 | err_helper: |
3fe0f943 | 1827 | nf_conntrack_ecache_pernet_fini(net); |
a0891aa6 | 1828 | err_ecache: |
73f4001a | 1829 | nf_conntrack_tstamp_pernet_fini(net); |
a992ca2a | 1830 | err_tstamp: |
b7ff3a1f | 1831 | nf_conntrack_acct_pernet_fini(net); |
08f6547d | 1832 | err_acct: |
83b4dbe1 | 1833 | nf_conntrack_expect_pernet_fini(net); |
08f6547d | 1834 | err_expect: |
d862a662 | 1835 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
6058fa6b | 1836 | err_hash: |
5b3501fa ED |
1837 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
1838 | err_cache: | |
1839 | kfree(net->ct.slabname); | |
1840 | err_slabname: | |
0d55af87 | 1841 | free_percpu(net->ct.stat); |
b7779d06 JDB |
1842 | err_pcpu_lists: |
1843 | free_percpu(net->ct.pcpu_lists); | |
0d55af87 | 1844 | err_stat: |
08f6547d AD |
1845 | return ret; |
1846 | } |