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> |
1b8c8a9f | 57 | #include <net/netns/hash.h> |
9fb9cbb1 | 58 | |
dc808fe2 | 59 | #define NF_CONNTRACK_VERSION "0.5.0" |
9fb9cbb1 | 60 | |
e17b666a PM |
61 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
62 | enum nf_nat_manip_type manip, | |
39938324 | 63 | const struct nlattr *attr) __read_mostly; |
e6a7d3c0 PNA |
64 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
65 | ||
93bb0ceb JDB |
66 | __cacheline_aligned_in_smp spinlock_t nf_conntrack_locks[CONNTRACK_LOCKS]; |
67 | EXPORT_SYMBOL_GPL(nf_conntrack_locks); | |
9fb9cbb1 | 68 | |
ca7433df JDB |
69 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(nf_conntrack_expect_lock); |
70 | EXPORT_SYMBOL_GPL(nf_conntrack_expect_lock); | |
71 | ||
56d52d48 FW |
72 | struct hlist_nulls_head *nf_conntrack_hash __read_mostly; |
73 | EXPORT_SYMBOL_GPL(nf_conntrack_hash); | |
74 | ||
b16c2919 | 75 | static __read_mostly spinlock_t nf_conntrack_locks_all_lock; |
a3efd812 | 76 | static __read_mostly seqcount_t nf_conntrack_generation; |
b16c2919 SL |
77 | static __read_mostly bool nf_conntrack_locks_all; |
78 | ||
79 | void nf_conntrack_lock(spinlock_t *lock) __acquires(lock) | |
80 | { | |
81 | spin_lock(lock); | |
82 | while (unlikely(nf_conntrack_locks_all)) { | |
83 | spin_unlock(lock); | |
e39365be | 84 | spin_unlock_wait(&nf_conntrack_locks_all_lock); |
b16c2919 SL |
85 | spin_lock(lock); |
86 | } | |
87 | } | |
88 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); | |
89 | ||
93bb0ceb JDB |
90 | static void nf_conntrack_double_unlock(unsigned int h1, unsigned int h2) |
91 | { | |
92 | h1 %= CONNTRACK_LOCKS; | |
93 | h2 %= CONNTRACK_LOCKS; | |
94 | spin_unlock(&nf_conntrack_locks[h1]); | |
95 | if (h1 != h2) | |
96 | spin_unlock(&nf_conntrack_locks[h2]); | |
97 | } | |
98 | ||
99 | /* return true if we need to recompute hashes (in case hash table was resized) */ | |
100 | static bool nf_conntrack_double_lock(struct net *net, unsigned int h1, | |
101 | unsigned int h2, unsigned int sequence) | |
102 | { | |
103 | h1 %= CONNTRACK_LOCKS; | |
104 | h2 %= CONNTRACK_LOCKS; | |
105 | if (h1 <= h2) { | |
b16c2919 | 106 | nf_conntrack_lock(&nf_conntrack_locks[h1]); |
93bb0ceb JDB |
107 | if (h1 != h2) |
108 | spin_lock_nested(&nf_conntrack_locks[h2], | |
109 | SINGLE_DEPTH_NESTING); | |
110 | } else { | |
b16c2919 | 111 | nf_conntrack_lock(&nf_conntrack_locks[h2]); |
93bb0ceb JDB |
112 | spin_lock_nested(&nf_conntrack_locks[h1], |
113 | SINGLE_DEPTH_NESTING); | |
114 | } | |
a3efd812 | 115 | if (read_seqcount_retry(&nf_conntrack_generation, sequence)) { |
93bb0ceb JDB |
116 | nf_conntrack_double_unlock(h1, h2); |
117 | return true; | |
118 | } | |
119 | return false; | |
120 | } | |
121 | ||
122 | static void nf_conntrack_all_lock(void) | |
123 | { | |
124 | int i; | |
125 | ||
b16c2919 SL |
126 | spin_lock(&nf_conntrack_locks_all_lock); |
127 | nf_conntrack_locks_all = true; | |
128 | ||
129 | for (i = 0; i < CONNTRACK_LOCKS; i++) { | |
e39365be | 130 | spin_unlock_wait(&nf_conntrack_locks[i]); |
b16c2919 | 131 | } |
93bb0ceb JDB |
132 | } |
133 | ||
134 | static void nf_conntrack_all_unlock(void) | |
135 | { | |
b16c2919 SL |
136 | nf_conntrack_locks_all = false; |
137 | spin_unlock(&nf_conntrack_locks_all_lock); | |
93bb0ceb JDB |
138 | } |
139 | ||
e2b7606c | 140 | unsigned int nf_conntrack_htable_size __read_mostly; |
13b18339 PM |
141 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
142 | ||
e478075c | 143 | unsigned int nf_conntrack_max __read_mostly; |
a999e683 | 144 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
13b18339 | 145 | |
b3c5163f ED |
146 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); |
147 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); | |
13b18339 | 148 | |
141658fb | 149 | static unsigned int nf_conntrack_hash_rnd __read_mostly; |
9fb9cbb1 | 150 | |
1b8c8a9f FW |
151 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, |
152 | const struct net *net) | |
9fb9cbb1 | 153 | { |
0794935e | 154 | unsigned int n; |
1b8c8a9f | 155 | u32 seed; |
0794935e | 156 | |
141658fb FW |
157 | get_random_once(&nf_conntrack_hash_rnd, sizeof(nf_conntrack_hash_rnd)); |
158 | ||
0794935e PM |
159 | /* The direction must be ignored, so we hash everything up to the |
160 | * destination ports (which is a multiple of 4) and treat the last | |
161 | * three bytes manually. | |
162 | */ | |
1b8c8a9f | 163 | seed = nf_conntrack_hash_rnd ^ net_hash_mix(net); |
0794935e | 164 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); |
1b8c8a9f | 165 | return jhash2((u32 *)tuple, n, seed ^ |
99f07e91 CG |
166 | (((__force __u16)tuple->dst.u.all << 16) | |
167 | tuple->dst.protonum)); | |
168 | } | |
169 | ||
56d52d48 | 170 | static u32 scale_hash(u32 hash) |
99f07e91 | 171 | { |
56d52d48 | 172 | return reciprocal_scale(hash, nf_conntrack_htable_size); |
99f07e91 | 173 | } |
0794935e | 174 | |
1b8c8a9f FW |
175 | static u32 __hash_conntrack(const struct net *net, |
176 | const struct nf_conntrack_tuple *tuple, | |
177 | unsigned int size) | |
99f07e91 | 178 | { |
1b8c8a9f | 179 | return reciprocal_scale(hash_conntrack_raw(tuple, net), size); |
9fb9cbb1 YK |
180 | } |
181 | ||
1b8c8a9f FW |
182 | static u32 hash_conntrack(const struct net *net, |
183 | const struct nf_conntrack_tuple *tuple) | |
9fb9cbb1 | 184 | { |
56d52d48 | 185 | return scale_hash(hash_conntrack_raw(tuple, net)); |
9fb9cbb1 YK |
186 | } |
187 | ||
5f2b4c90 | 188 | bool |
9fb9cbb1 YK |
189 | nf_ct_get_tuple(const struct sk_buff *skb, |
190 | unsigned int nhoff, | |
191 | unsigned int dataoff, | |
192 | u_int16_t l3num, | |
193 | u_int8_t protonum, | |
a31f1adc | 194 | struct net *net, |
9fb9cbb1 YK |
195 | struct nf_conntrack_tuple *tuple, |
196 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 197 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 198 | { |
443a70d5 | 199 | memset(tuple, 0, sizeof(*tuple)); |
9fb9cbb1 YK |
200 | |
201 | tuple->src.l3num = l3num; | |
202 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | |
5f2b4c90 | 203 | return false; |
9fb9cbb1 YK |
204 | |
205 | tuple->dst.protonum = protonum; | |
206 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | |
207 | ||
a31f1adc | 208 | return l4proto->pkt_to_tuple(skb, dataoff, net, tuple); |
9fb9cbb1 | 209 | } |
13b18339 | 210 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
9fb9cbb1 | 211 | |
5f2b4c90 | 212 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
a31f1adc EB |
213 | u_int16_t l3num, |
214 | struct net *net, struct nf_conntrack_tuple *tuple) | |
e2a3123f YK |
215 | { |
216 | struct nf_conntrack_l3proto *l3proto; | |
217 | struct nf_conntrack_l4proto *l4proto; | |
218 | unsigned int protoff; | |
219 | u_int8_t protonum; | |
220 | int ret; | |
221 | ||
222 | rcu_read_lock(); | |
223 | ||
224 | l3proto = __nf_ct_l3proto_find(l3num); | |
225 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | |
226 | if (ret != NF_ACCEPT) { | |
227 | rcu_read_unlock(); | |
5f2b4c90 | 228 | return false; |
e2a3123f YK |
229 | } |
230 | ||
231 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
232 | ||
a31f1adc | 233 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, net, tuple, |
e2a3123f YK |
234 | l3proto, l4proto); |
235 | ||
236 | rcu_read_unlock(); | |
237 | return ret; | |
238 | } | |
239 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | |
240 | ||
5f2b4c90 | 241 | bool |
9fb9cbb1 YK |
242 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
243 | const struct nf_conntrack_tuple *orig, | |
244 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 245 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 246 | { |
443a70d5 | 247 | memset(inverse, 0, sizeof(*inverse)); |
9fb9cbb1 YK |
248 | |
249 | inverse->src.l3num = orig->src.l3num; | |
250 | if (l3proto->invert_tuple(inverse, orig) == 0) | |
5f2b4c90 | 251 | return false; |
9fb9cbb1 YK |
252 | |
253 | inverse->dst.dir = !orig->dst.dir; | |
254 | ||
255 | inverse->dst.protonum = orig->dst.protonum; | |
605dcad6 | 256 | return l4proto->invert_tuple(inverse, orig); |
9fb9cbb1 | 257 | } |
13b18339 | 258 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
9fb9cbb1 | 259 | |
9fb9cbb1 YK |
260 | static void |
261 | clean_from_lists(struct nf_conn *ct) | |
262 | { | |
0d53778e | 263 | pr_debug("clean_from_lists(%p)\n", ct); |
ea781f19 ED |
264 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
265 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | |
9fb9cbb1 YK |
266 | |
267 | /* Destroy all pending expectations */ | |
c1d10adb | 268 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
269 | } |
270 | ||
b7779d06 JDB |
271 | /* must be called with local_bh_disable */ |
272 | static void nf_ct_add_to_dying_list(struct nf_conn *ct) | |
273 | { | |
274 | struct ct_pcpu *pcpu; | |
275 | ||
276 | /* add this conntrack to the (per cpu) dying list */ | |
277 | ct->cpu = smp_processor_id(); | |
278 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
279 | ||
280 | spin_lock(&pcpu->lock); | |
281 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
282 | &pcpu->dying); | |
283 | spin_unlock(&pcpu->lock); | |
284 | } | |
285 | ||
286 | /* must be called with local_bh_disable */ | |
287 | static void nf_ct_add_to_unconfirmed_list(struct nf_conn *ct) | |
288 | { | |
289 | struct ct_pcpu *pcpu; | |
290 | ||
291 | /* add this conntrack to the (per cpu) unconfirmed list */ | |
292 | ct->cpu = smp_processor_id(); | |
293 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
294 | ||
295 | spin_lock(&pcpu->lock); | |
296 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
297 | &pcpu->unconfirmed); | |
298 | spin_unlock(&pcpu->lock); | |
299 | } | |
300 | ||
301 | /* must be called with local_bh_disable */ | |
302 | static void nf_ct_del_from_dying_or_unconfirmed_list(struct nf_conn *ct) | |
303 | { | |
304 | struct ct_pcpu *pcpu; | |
305 | ||
306 | /* We overload first tuple to link into unconfirmed or dying list.*/ | |
307 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
308 | ||
309 | spin_lock(&pcpu->lock); | |
310 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); | |
311 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | |
312 | spin_unlock(&pcpu->lock); | |
313 | } | |
314 | ||
0838aa7f | 315 | /* Released via destroy_conntrack() */ |
308ac914 DB |
316 | struct nf_conn *nf_ct_tmpl_alloc(struct net *net, |
317 | const struct nf_conntrack_zone *zone, | |
318 | gfp_t flags) | |
0838aa7f PNA |
319 | { |
320 | struct nf_conn *tmpl; | |
321 | ||
f58e5aa7 | 322 | tmpl = kzalloc(sizeof(*tmpl), flags); |
0838aa7f PNA |
323 | if (tmpl == NULL) |
324 | return NULL; | |
325 | ||
326 | tmpl->status = IPS_TEMPLATE; | |
327 | write_pnet(&tmpl->ct_net, net); | |
328 | ||
5e8018fc DB |
329 | if (nf_ct_zone_add(tmpl, flags, zone) < 0) |
330 | goto out_free; | |
331 | ||
0838aa7f PNA |
332 | atomic_set(&tmpl->ct_general.use, 0); |
333 | ||
334 | return tmpl; | |
0838aa7f PNA |
335 | out_free: |
336 | kfree(tmpl); | |
337 | return NULL; | |
0838aa7f PNA |
338 | } |
339 | EXPORT_SYMBOL_GPL(nf_ct_tmpl_alloc); | |
340 | ||
9cf94eab | 341 | void nf_ct_tmpl_free(struct nf_conn *tmpl) |
0838aa7f PNA |
342 | { |
343 | nf_ct_ext_destroy(tmpl); | |
344 | nf_ct_ext_free(tmpl); | |
345 | kfree(tmpl); | |
346 | } | |
9cf94eab | 347 | EXPORT_SYMBOL_GPL(nf_ct_tmpl_free); |
0838aa7f | 348 | |
9fb9cbb1 YK |
349 | static void |
350 | destroy_conntrack(struct nf_conntrack *nfct) | |
351 | { | |
352 | struct nf_conn *ct = (struct nf_conn *)nfct; | |
0d55af87 | 353 | struct net *net = nf_ct_net(ct); |
605dcad6 | 354 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 | 355 | |
0d53778e | 356 | pr_debug("destroy_conntrack(%p)\n", ct); |
9fb9cbb1 YK |
357 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
358 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); | |
359 | ||
0838aa7f PNA |
360 | if (unlikely(nf_ct_is_template(ct))) { |
361 | nf_ct_tmpl_free(ct); | |
362 | return; | |
363 | } | |
923f4902 | 364 | rcu_read_lock(); |
5e8fbe2a | 365 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
4b4ceb9d | 366 | if (l4proto->destroy) |
605dcad6 | 367 | l4proto->destroy(ct); |
9fb9cbb1 | 368 | |
982d9a9c | 369 | rcu_read_unlock(); |
9fb9cbb1 | 370 | |
ca7433df | 371 | local_bh_disable(); |
9fb9cbb1 YK |
372 | /* Expectations will have been removed in clean_from_lists, |
373 | * except TFTP can create an expectation on the first packet, | |
374 | * before connection is in the list, so we need to clean here, | |
ca7433df JDB |
375 | * too. |
376 | */ | |
c1d10adb | 377 | nf_ct_remove_expectations(ct); |
9fb9cbb1 | 378 | |
b7779d06 | 379 | nf_ct_del_from_dying_or_unconfirmed_list(ct); |
9fb9cbb1 | 380 | |
0d55af87 | 381 | NF_CT_STAT_INC(net, delete); |
ca7433df | 382 | local_bh_enable(); |
9fb9cbb1 YK |
383 | |
384 | if (ct->master) | |
385 | nf_ct_put(ct->master); | |
386 | ||
0d53778e | 387 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
9fb9cbb1 YK |
388 | nf_conntrack_free(ct); |
389 | } | |
390 | ||
02982c27 | 391 | static void nf_ct_delete_from_lists(struct nf_conn *ct) |
9fb9cbb1 | 392 | { |
0d55af87 | 393 | struct net *net = nf_ct_net(ct); |
93bb0ceb | 394 | unsigned int hash, reply_hash; |
93bb0ceb | 395 | unsigned int sequence; |
9fb9cbb1 | 396 | |
9858a3ae | 397 | nf_ct_helper_destroy(ct); |
93bb0ceb JDB |
398 | |
399 | local_bh_disable(); | |
400 | do { | |
a3efd812 | 401 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
deedb590 | 402 | hash = hash_conntrack(net, |
93bb0ceb | 403 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
deedb590 | 404 | reply_hash = hash_conntrack(net, |
93bb0ceb JDB |
405 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
406 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | |
407 | ||
9fb9cbb1 | 408 | clean_from_lists(ct); |
93bb0ceb JDB |
409 | nf_conntrack_double_unlock(hash, reply_hash); |
410 | ||
b7779d06 | 411 | nf_ct_add_to_dying_list(ct); |
93bb0ceb JDB |
412 | |
413 | NF_CT_STAT_INC(net, delete_list); | |
414 | local_bh_enable(); | |
dd7669a9 | 415 | } |
dd7669a9 | 416 | |
02982c27 | 417 | bool nf_ct_delete(struct nf_conn *ct, u32 portid, int report) |
dd7669a9 | 418 | { |
a992ca2a PNA |
419 | struct nf_conn_tstamp *tstamp; |
420 | ||
421 | tstamp = nf_conn_tstamp_find(ct); | |
422 | if (tstamp && tstamp->stop == 0) | |
d2de875c | 423 | tstamp->stop = ktime_get_real_ns(); |
dd7669a9 | 424 | |
9500507c FW |
425 | if (nf_ct_is_dying(ct)) |
426 | goto delete; | |
427 | ||
428 | if (nf_conntrack_event_report(IPCT_DESTROY, ct, | |
429 | portid, report) < 0) { | |
dd7669a9 PNA |
430 | /* destroy event was not delivered */ |
431 | nf_ct_delete_from_lists(ct); | |
9500507c | 432 | nf_conntrack_ecache_delayed_work(nf_ct_net(ct)); |
02982c27 | 433 | return false; |
dd7669a9 | 434 | } |
9500507c FW |
435 | |
436 | nf_conntrack_ecache_work(nf_ct_net(ct)); | |
dd7669a9 | 437 | set_bit(IPS_DYING_BIT, &ct->status); |
9500507c | 438 | delete: |
dd7669a9 | 439 | nf_ct_delete_from_lists(ct); |
9fb9cbb1 | 440 | nf_ct_put(ct); |
02982c27 FW |
441 | return true; |
442 | } | |
443 | EXPORT_SYMBOL_GPL(nf_ct_delete); | |
444 | ||
445 | static void death_by_timeout(unsigned long ul_conntrack) | |
446 | { | |
447 | nf_ct_delete((struct nf_conn *)ul_conntrack, 0, 0); | |
9fb9cbb1 YK |
448 | } |
449 | ||
c6825c09 AV |
450 | static inline bool |
451 | nf_ct_key_equal(struct nf_conntrack_tuple_hash *h, | |
308ac914 | 452 | const struct nf_conntrack_tuple *tuple, |
e0c7d472 FW |
453 | const struct nf_conntrack_zone *zone, |
454 | const struct net *net) | |
c6825c09 AV |
455 | { |
456 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | |
457 | ||
458 | /* A conntrack can be recreated with the equal tuple, | |
459 | * so we need to check that the conntrack is confirmed | |
460 | */ | |
461 | return nf_ct_tuple_equal(tuple, &h->tuple) && | |
deedb590 | 462 | nf_ct_zone_equal(ct, zone, NF_CT_DIRECTION(h)) && |
e0c7d472 FW |
463 | nf_ct_is_confirmed(ct) && |
464 | net_eq(net, nf_ct_net(ct)); | |
c6825c09 AV |
465 | } |
466 | ||
ea781f19 ED |
467 | /* |
468 | * Warning : | |
469 | * - Caller must take a reference on returned object | |
470 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | |
ea781f19 | 471 | */ |
99f07e91 | 472 | static struct nf_conntrack_tuple_hash * |
308ac914 | 473 | ____nf_conntrack_find(struct net *net, const struct nf_conntrack_zone *zone, |
99f07e91 | 474 | const struct nf_conntrack_tuple *tuple, u32 hash) |
9fb9cbb1 YK |
475 | { |
476 | struct nf_conntrack_tuple_hash *h; | |
5e3c61f9 | 477 | struct hlist_nulls_head *ct_hash; |
ea781f19 | 478 | struct hlist_nulls_node *n; |
5e3c61f9 | 479 | unsigned int bucket, sequence; |
9fb9cbb1 | 480 | |
ea781f19 | 481 | begin: |
5e3c61f9 FW |
482 | do { |
483 | sequence = read_seqcount_begin(&nf_conntrack_generation); | |
56d52d48 FW |
484 | bucket = scale_hash(hash); |
485 | ct_hash = nf_conntrack_hash; | |
5e3c61f9 FW |
486 | } while (read_seqcount_retry(&nf_conntrack_generation, sequence)); |
487 | ||
488 | hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[bucket], hnnode) { | |
e0c7d472 | 489 | if (nf_ct_key_equal(h, tuple, zone, net)) { |
2cf12348 | 490 | NF_CT_STAT_INC_ATOMIC(net, found); |
9fb9cbb1 YK |
491 | return h; |
492 | } | |
2cf12348 | 493 | NF_CT_STAT_INC_ATOMIC(net, searched); |
9fb9cbb1 | 494 | } |
ea781f19 ED |
495 | /* |
496 | * if the nulls value we got at the end of this lookup is | |
497 | * not the expected one, we must restart lookup. | |
498 | * We probably met an item that was moved to another chain. | |
499 | */ | |
99f07e91 | 500 | if (get_nulls_value(n) != bucket) { |
2cf12348 | 501 | NF_CT_STAT_INC_ATOMIC(net, search_restart); |
ea781f19 | 502 | goto begin; |
af740b2c | 503 | } |
9fb9cbb1 YK |
504 | |
505 | return NULL; | |
506 | } | |
99f07e91 | 507 | |
9fb9cbb1 | 508 | /* Find a connection corresponding to a tuple. */ |
99f07e91 | 509 | static struct nf_conntrack_tuple_hash * |
308ac914 | 510 | __nf_conntrack_find_get(struct net *net, const struct nf_conntrack_zone *zone, |
99f07e91 | 511 | const struct nf_conntrack_tuple *tuple, u32 hash) |
9fb9cbb1 YK |
512 | { |
513 | struct nf_conntrack_tuple_hash *h; | |
76507f69 | 514 | struct nf_conn *ct; |
9fb9cbb1 | 515 | |
76507f69 | 516 | rcu_read_lock(); |
ea781f19 | 517 | begin: |
99f07e91 | 518 | h = ____nf_conntrack_find(net, zone, tuple, hash); |
76507f69 PM |
519 | if (h) { |
520 | ct = nf_ct_tuplehash_to_ctrack(h); | |
8d8890b7 PM |
521 | if (unlikely(nf_ct_is_dying(ct) || |
522 | !atomic_inc_not_zero(&ct->ct_general.use))) | |
76507f69 | 523 | h = NULL; |
ea781f19 | 524 | else { |
e0c7d472 | 525 | if (unlikely(!nf_ct_key_equal(h, tuple, zone, net))) { |
ea781f19 ED |
526 | nf_ct_put(ct); |
527 | goto begin; | |
528 | } | |
529 | } | |
76507f69 PM |
530 | } |
531 | rcu_read_unlock(); | |
9fb9cbb1 YK |
532 | |
533 | return h; | |
534 | } | |
99f07e91 CG |
535 | |
536 | struct nf_conntrack_tuple_hash * | |
308ac914 | 537 | nf_conntrack_find_get(struct net *net, const struct nf_conntrack_zone *zone, |
99f07e91 CG |
538 | const struct nf_conntrack_tuple *tuple) |
539 | { | |
540 | return __nf_conntrack_find_get(net, zone, tuple, | |
1b8c8a9f | 541 | hash_conntrack_raw(tuple, net)); |
99f07e91 | 542 | } |
13b18339 | 543 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
9fb9cbb1 | 544 | |
c1d10adb PNA |
545 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
546 | unsigned int hash, | |
b476b72a | 547 | unsigned int reply_hash) |
c1d10adb | 548 | { |
ea781f19 | 549 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
56d52d48 | 550 | &nf_conntrack_hash[hash]); |
ea781f19 | 551 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
56d52d48 | 552 | &nf_conntrack_hash[reply_hash]); |
c1d10adb PNA |
553 | } |
554 | ||
7d367e06 JK |
555 | int |
556 | nf_conntrack_hash_check_insert(struct nf_conn *ct) | |
c1d10adb | 557 | { |
308ac914 | 558 | const struct nf_conntrack_zone *zone; |
d696c7bd | 559 | struct net *net = nf_ct_net(ct); |
b476b72a | 560 | unsigned int hash, reply_hash; |
7d367e06 JK |
561 | struct nf_conntrack_tuple_hash *h; |
562 | struct hlist_nulls_node *n; | |
93bb0ceb | 563 | unsigned int sequence; |
c1d10adb | 564 | |
5d0aa2cc | 565 | zone = nf_ct_zone(ct); |
7d367e06 | 566 | |
93bb0ceb JDB |
567 | local_bh_disable(); |
568 | do { | |
a3efd812 | 569 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
deedb590 | 570 | hash = hash_conntrack(net, |
93bb0ceb | 571 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); |
deedb590 | 572 | reply_hash = hash_conntrack(net, |
93bb0ceb JDB |
573 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
574 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | |
7d367e06 JK |
575 | |
576 | /* See if there's one in the list already, including reverse */ | |
56d52d48 | 577 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode) |
86804348 | 578 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
e0c7d472 | 579 | zone, net)) |
7d367e06 | 580 | goto out; |
86804348 | 581 | |
56d52d48 | 582 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode) |
86804348 | 583 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
e0c7d472 | 584 | zone, net)) |
7d367e06 | 585 | goto out; |
c1d10adb | 586 | |
7d367e06 | 587 | add_timer(&ct->timeout); |
e53376be PNA |
588 | smp_wmb(); |
589 | /* The caller holds a reference to this object */ | |
590 | atomic_set(&ct->ct_general.use, 2); | |
b476b72a | 591 | __nf_conntrack_hash_insert(ct, hash, reply_hash); |
93bb0ceb | 592 | nf_conntrack_double_unlock(hash, reply_hash); |
7d367e06 | 593 | NF_CT_STAT_INC(net, insert); |
93bb0ceb | 594 | local_bh_enable(); |
7d367e06 JK |
595 | return 0; |
596 | ||
597 | out: | |
93bb0ceb | 598 | nf_conntrack_double_unlock(hash, reply_hash); |
7d367e06 | 599 | NF_CT_STAT_INC(net, insert_failed); |
93bb0ceb | 600 | local_bh_enable(); |
7d367e06 | 601 | return -EEXIST; |
c1d10adb | 602 | } |
7d367e06 | 603 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); |
c1d10adb | 604 | |
ba76738c PNA |
605 | static inline void nf_ct_acct_update(struct nf_conn *ct, |
606 | enum ip_conntrack_info ctinfo, | |
607 | unsigned int len) | |
608 | { | |
609 | struct nf_conn_acct *acct; | |
610 | ||
611 | acct = nf_conn_acct_find(ct); | |
612 | if (acct) { | |
613 | struct nf_conn_counter *counter = acct->counter; | |
614 | ||
615 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); | |
616 | atomic64_add(len, &counter[CTINFO2DIR(ctinfo)].bytes); | |
617 | } | |
618 | } | |
619 | ||
71d8c47f PNA |
620 | static void nf_ct_acct_merge(struct nf_conn *ct, enum ip_conntrack_info ctinfo, |
621 | const struct nf_conn *loser_ct) | |
622 | { | |
623 | struct nf_conn_acct *acct; | |
624 | ||
625 | acct = nf_conn_acct_find(loser_ct); | |
626 | if (acct) { | |
627 | struct nf_conn_counter *counter = acct->counter; | |
628 | enum ip_conntrack_info ctinfo; | |
629 | unsigned int bytes; | |
630 | ||
631 | /* u32 should be fine since we must have seen one packet. */ | |
632 | bytes = atomic64_read(&counter[CTINFO2DIR(ctinfo)].bytes); | |
633 | nf_ct_acct_update(ct, ctinfo, bytes); | |
634 | } | |
635 | } | |
636 | ||
637 | /* Resolve race on insertion if this protocol allows this. */ | |
638 | static int nf_ct_resolve_clash(struct net *net, struct sk_buff *skb, | |
639 | enum ip_conntrack_info ctinfo, | |
640 | struct nf_conntrack_tuple_hash *h) | |
641 | { | |
642 | /* This is the conntrack entry already in hashes that won race. */ | |
643 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | |
644 | struct nf_conntrack_l4proto *l4proto; | |
645 | ||
646 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); | |
647 | if (l4proto->allow_clash && | |
648 | !nf_ct_is_dying(ct) && | |
649 | atomic_inc_not_zero(&ct->ct_general.use)) { | |
650 | nf_ct_acct_merge(ct, ctinfo, (struct nf_conn *)skb->nfct); | |
651 | nf_conntrack_put(skb->nfct); | |
652 | /* Assign conntrack already in hashes to this skbuff. Don't | |
653 | * modify skb->nfctinfo to ensure consistent stateful filtering. | |
654 | */ | |
655 | skb->nfct = &ct->ct_general; | |
656 | return NF_ACCEPT; | |
657 | } | |
658 | NF_CT_STAT_INC(net, drop); | |
659 | return NF_DROP; | |
660 | } | |
661 | ||
9fb9cbb1 YK |
662 | /* Confirm a connection given skb; places it in hash table */ |
663 | int | |
3db05fea | 664 | __nf_conntrack_confirm(struct sk_buff *skb) |
9fb9cbb1 | 665 | { |
308ac914 | 666 | const struct nf_conntrack_zone *zone; |
b476b72a | 667 | unsigned int hash, reply_hash; |
df0933dc | 668 | struct nf_conntrack_tuple_hash *h; |
9fb9cbb1 | 669 | struct nf_conn *ct; |
df0933dc | 670 | struct nf_conn_help *help; |
a992ca2a | 671 | struct nf_conn_tstamp *tstamp; |
ea781f19 | 672 | struct hlist_nulls_node *n; |
9fb9cbb1 | 673 | enum ip_conntrack_info ctinfo; |
400dad39 | 674 | struct net *net; |
93bb0ceb | 675 | unsigned int sequence; |
71d8c47f | 676 | int ret = NF_DROP; |
9fb9cbb1 | 677 | |
3db05fea | 678 | ct = nf_ct_get(skb, &ctinfo); |
400dad39 | 679 | net = nf_ct_net(ct); |
9fb9cbb1 YK |
680 | |
681 | /* ipt_REJECT uses nf_conntrack_attach to attach related | |
682 | ICMP/TCP RST packets in other direction. Actual packet | |
683 | which created connection will be IP_CT_NEW or for an | |
684 | expected connection, IP_CT_RELATED. */ | |
685 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | |
686 | return NF_ACCEPT; | |
687 | ||
5d0aa2cc | 688 | zone = nf_ct_zone(ct); |
93bb0ceb JDB |
689 | local_bh_disable(); |
690 | ||
691 | do { | |
a3efd812 | 692 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
93bb0ceb JDB |
693 | /* reuse the hash saved before */ |
694 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; | |
56d52d48 | 695 | hash = scale_hash(hash); |
deedb590 | 696 | reply_hash = hash_conntrack(net, |
93bb0ceb JDB |
697 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
698 | ||
699 | } while (nf_conntrack_double_lock(net, hash, reply_hash, sequence)); | |
9fb9cbb1 YK |
700 | |
701 | /* We're not in hash table, and we refuse to set up related | |
93bb0ceb JDB |
702 | * connections for unconfirmed conns. But packet copies and |
703 | * REJECT will give spurious warnings here. | |
704 | */ | |
9fb9cbb1 YK |
705 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ |
706 | ||
25985edc | 707 | /* No external references means no one else could have |
93bb0ceb JDB |
708 | * confirmed us. |
709 | */ | |
9fb9cbb1 | 710 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); |
0d53778e | 711 | pr_debug("Confirming conntrack %p\n", ct); |
8ca3f5e9 PNA |
712 | /* We have to check the DYING flag after unlink to prevent |
713 | * a race against nf_ct_get_next_corpse() possibly called from | |
714 | * user context, else we insert an already 'dead' hash, blocking | |
715 | * further use of that particular connection -JM. | |
716 | */ | |
717 | nf_ct_del_from_dying_or_unconfirmed_list(ct); | |
718 | ||
71d8c47f PNA |
719 | if (unlikely(nf_ct_is_dying(ct))) { |
720 | nf_ct_add_to_dying_list(ct); | |
721 | goto dying; | |
722 | } | |
fc350777 | 723 | |
9fb9cbb1 YK |
724 | /* See if there's one in the list already, including reverse: |
725 | NAT could have grabbed it without realizing, since we're | |
726 | not in the hash. If there is, we lost race. */ | |
56d52d48 | 727 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[hash], hnnode) |
86804348 | 728 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
e0c7d472 | 729 | zone, net)) |
df0933dc | 730 | goto out; |
86804348 | 731 | |
56d52d48 | 732 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[reply_hash], hnnode) |
86804348 | 733 | if (nf_ct_key_equal(h, &ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
e0c7d472 | 734 | zone, net)) |
df0933dc | 735 | goto out; |
9fb9cbb1 | 736 | |
df0933dc PM |
737 | /* Timer relative to confirmation time, not original |
738 | setting time, otherwise we'd get timer wrap in | |
739 | weird delay cases. */ | |
740 | ct->timeout.expires += jiffies; | |
741 | add_timer(&ct->timeout); | |
742 | atomic_inc(&ct->ct_general.use); | |
45eec341 | 743 | ct->status |= IPS_CONFIRMED; |
5c8ec910 | 744 | |
a992ca2a PNA |
745 | /* set conntrack timestamp, if enabled. */ |
746 | tstamp = nf_conn_tstamp_find(ct); | |
747 | if (tstamp) { | |
748 | if (skb->tstamp.tv64 == 0) | |
e3192690 | 749 | __net_timestamp(skb); |
a992ca2a PNA |
750 | |
751 | tstamp->start = ktime_to_ns(skb->tstamp); | |
752 | } | |
5c8ec910 PM |
753 | /* Since the lookup is lockless, hash insertion must be done after |
754 | * starting the timer and setting the CONFIRMED bit. The RCU barriers | |
755 | * guarantee that no other CPU can find the conntrack before the above | |
756 | * stores are visible. | |
757 | */ | |
b476b72a | 758 | __nf_conntrack_hash_insert(ct, hash, reply_hash); |
93bb0ceb | 759 | nf_conntrack_double_unlock(hash, reply_hash); |
0d55af87 | 760 | NF_CT_STAT_INC(net, insert); |
93bb0ceb | 761 | local_bh_enable(); |
5c8ec910 | 762 | |
df0933dc PM |
763 | help = nfct_help(ct); |
764 | if (help && help->helper) | |
a71996fc | 765 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
17e6e4ea | 766 | |
df0933dc | 767 | nf_conntrack_event_cache(master_ct(ct) ? |
a71996fc | 768 | IPCT_RELATED : IPCT_NEW, ct); |
df0933dc | 769 | return NF_ACCEPT; |
9fb9cbb1 | 770 | |
df0933dc | 771 | out: |
8ca3f5e9 | 772 | nf_ct_add_to_dying_list(ct); |
71d8c47f PNA |
773 | ret = nf_ct_resolve_clash(net, skb, ctinfo, h); |
774 | dying: | |
93bb0ceb | 775 | nf_conntrack_double_unlock(hash, reply_hash); |
0d55af87 | 776 | NF_CT_STAT_INC(net, insert_failed); |
93bb0ceb | 777 | local_bh_enable(); |
71d8c47f | 778 | return ret; |
9fb9cbb1 | 779 | } |
13b18339 | 780 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
9fb9cbb1 YK |
781 | |
782 | /* Returns true if a connection correspondings to the tuple (required | |
783 | for NAT). */ | |
784 | int | |
785 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | |
786 | const struct nf_conn *ignored_conntrack) | |
787 | { | |
400dad39 | 788 | struct net *net = nf_ct_net(ignored_conntrack); |
308ac914 | 789 | const struct nf_conntrack_zone *zone; |
9fb9cbb1 | 790 | struct nf_conntrack_tuple_hash *h; |
5e3c61f9 FW |
791 | struct hlist_nulls_head *ct_hash; |
792 | unsigned int hash, sequence; | |
ea781f19 | 793 | struct hlist_nulls_node *n; |
5d0aa2cc | 794 | struct nf_conn *ct; |
308ac914 DB |
795 | |
796 | zone = nf_ct_zone(ignored_conntrack); | |
9fb9cbb1 | 797 | |
2cf12348 | 798 | rcu_read_lock(); |
5e3c61f9 FW |
799 | do { |
800 | sequence = read_seqcount_begin(&nf_conntrack_generation); | |
801 | hash = hash_conntrack(net, tuple); | |
56d52d48 | 802 | ct_hash = nf_conntrack_hash; |
5e3c61f9 FW |
803 | } while (read_seqcount_retry(&nf_conntrack_generation, sequence)); |
804 | ||
805 | hlist_nulls_for_each_entry_rcu(h, n, &ct_hash[hash], hnnode) { | |
5d0aa2cc PM |
806 | ct = nf_ct_tuplehash_to_ctrack(h); |
807 | if (ct != ignored_conntrack && | |
e0c7d472 | 808 | nf_ct_key_equal(h, tuple, zone, net)) { |
2cf12348 FW |
809 | NF_CT_STAT_INC_ATOMIC(net, found); |
810 | rcu_read_unlock(); | |
ba419aff PM |
811 | return 1; |
812 | } | |
2cf12348 | 813 | NF_CT_STAT_INC_ATOMIC(net, searched); |
ba419aff | 814 | } |
2cf12348 | 815 | rcu_read_unlock(); |
9fb9cbb1 | 816 | |
ba419aff | 817 | return 0; |
9fb9cbb1 | 818 | } |
13b18339 | 819 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
9fb9cbb1 | 820 | |
7ae7730f PM |
821 | #define NF_CT_EVICTION_RANGE 8 |
822 | ||
9fb9cbb1 YK |
823 | /* There's a small race here where we may free a just-assured |
824 | connection. Too bad: we're in trouble anyway. */ | |
93bb0ceb | 825 | static noinline int early_drop(struct net *net, unsigned int _hash) |
9fb9cbb1 | 826 | { |
f205c5e0 | 827 | /* Use oldest entry, which is roughly LRU */ |
9fb9cbb1 | 828 | struct nf_conntrack_tuple_hash *h; |
3e86638e | 829 | struct nf_conn *tmp; |
ea781f19 | 830 | struct hlist_nulls_node *n; |
3e86638e FW |
831 | unsigned int i, hash, sequence; |
832 | struct nf_conn *ct = NULL; | |
93bb0ceb | 833 | spinlock_t *lockp; |
3e86638e FW |
834 | bool ret = false; |
835 | ||
836 | i = 0; | |
9fb9cbb1 | 837 | |
93bb0ceb JDB |
838 | local_bh_disable(); |
839 | restart: | |
a3efd812 | 840 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
3e86638e FW |
841 | for (; i < NF_CT_EVICTION_RANGE; i++) { |
842 | hash = scale_hash(_hash++); | |
93bb0ceb | 843 | lockp = &nf_conntrack_locks[hash % CONNTRACK_LOCKS]; |
b16c2919 | 844 | nf_conntrack_lock(lockp); |
a3efd812 | 845 | if (read_seqcount_retry(&nf_conntrack_generation, sequence)) { |
93bb0ceb JDB |
846 | spin_unlock(lockp); |
847 | goto restart; | |
848 | } | |
56d52d48 FW |
849 | hlist_nulls_for_each_entry_rcu(h, n, &nf_conntrack_hash[hash], |
850 | hnnode) { | |
7ae7730f | 851 | tmp = nf_ct_tuplehash_to_ctrack(h); |
3e86638e FW |
852 | |
853 | if (test_bit(IPS_ASSURED_BIT, &tmp->status) || | |
854 | !net_eq(nf_ct_net(tmp), net) || | |
855 | nf_ct_is_dying(tmp)) | |
856 | continue; | |
857 | ||
858 | if (atomic_inc_not_zero(&tmp->ct_general.use)) { | |
7ae7730f | 859 | ct = tmp; |
93bb0ceb JDB |
860 | break; |
861 | } | |
7ae7730f | 862 | } |
76507f69 | 863 | |
93bb0ceb | 864 | spin_unlock(lockp); |
3e86638e | 865 | if (ct) |
7ae7730f | 866 | break; |
9fb9cbb1 | 867 | } |
3e86638e | 868 | |
93bb0ceb | 869 | local_bh_enable(); |
9fb9cbb1 YK |
870 | |
871 | if (!ct) | |
3e86638e | 872 | return false; |
9fb9cbb1 | 873 | |
3e86638e FW |
874 | /* kill only if in same netns -- might have moved due to |
875 | * SLAB_DESTROY_BY_RCU rules | |
876 | */ | |
877 | if (net_eq(nf_ct_net(ct), net) && del_timer(&ct->timeout)) { | |
02982c27 | 878 | if (nf_ct_delete(ct, 0, 0)) { |
74138511 | 879 | NF_CT_STAT_INC_ATOMIC(net, early_drop); |
3e86638e | 880 | ret = true; |
74138511 | 881 | } |
9fb9cbb1 | 882 | } |
3e86638e | 883 | |
9fb9cbb1 | 884 | nf_ct_put(ct); |
3e86638e | 885 | return ret; |
9fb9cbb1 YK |
886 | } |
887 | ||
99f07e91 | 888 | static struct nf_conn * |
308ac914 DB |
889 | __nf_conntrack_alloc(struct net *net, |
890 | const struct nf_conntrack_zone *zone, | |
99f07e91 CG |
891 | const struct nf_conntrack_tuple *orig, |
892 | const struct nf_conntrack_tuple *repl, | |
893 | gfp_t gfp, u32 hash) | |
9fb9cbb1 | 894 | { |
cd7fcbf1 | 895 | struct nf_conn *ct; |
9fb9cbb1 | 896 | |
5251e2d2 | 897 | /* We don't want any race condition at early drop stage */ |
49ac8713 | 898 | atomic_inc(&net->ct.count); |
5251e2d2 | 899 | |
76eb9460 | 900 | if (nf_conntrack_max && |
49ac8713 | 901 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
93bb0ceb | 902 | if (!early_drop(net, hash)) { |
49ac8713 | 903 | atomic_dec(&net->ct.count); |
e87cc472 | 904 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); |
9fb9cbb1 YK |
905 | return ERR_PTR(-ENOMEM); |
906 | } | |
907 | } | |
908 | ||
941297f4 ED |
909 | /* |
910 | * Do not use kmem_cache_zalloc(), as this cache uses | |
911 | * SLAB_DESTROY_BY_RCU. | |
912 | */ | |
5b3501fa | 913 | ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp); |
5e8018fc DB |
914 | if (ct == NULL) |
915 | goto out; | |
916 | ||
440f0d58 | 917 | spin_lock_init(&ct->lock); |
c88130bc | 918 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; |
941297f4 | 919 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; |
c88130bc | 920 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; |
99f07e91 CG |
921 | /* save hash for reusing when confirming */ |
922 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; | |
c41884ce | 923 | ct->status = 0; |
9fb9cbb1 | 924 | /* Don't set timer yet: wait for confirmation */ |
c88130bc | 925 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
c2d9ba9b | 926 | write_pnet(&ct->ct_net, net); |
c41884ce FW |
927 | memset(&ct->__nfct_init_offset[0], 0, |
928 | offsetof(struct nf_conn, proto) - | |
929 | offsetof(struct nf_conn, __nfct_init_offset[0])); | |
5e8018fc DB |
930 | |
931 | if (zone && nf_ct_zone_add(ct, GFP_ATOMIC, zone) < 0) | |
932 | goto out_free; | |
933 | ||
e53376be PNA |
934 | /* Because we use RCU lookups, we set ct_general.use to zero before |
935 | * this is inserted in any list. | |
941297f4 | 936 | */ |
e53376be | 937 | atomic_set(&ct->ct_general.use, 0); |
c88130bc | 938 | return ct; |
5d0aa2cc PM |
939 | out_free: |
940 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); | |
5e8018fc DB |
941 | out: |
942 | atomic_dec(&net->ct.count); | |
5d0aa2cc | 943 | return ERR_PTR(-ENOMEM); |
9fb9cbb1 | 944 | } |
99f07e91 | 945 | |
308ac914 DB |
946 | struct nf_conn *nf_conntrack_alloc(struct net *net, |
947 | const struct nf_conntrack_zone *zone, | |
99f07e91 CG |
948 | const struct nf_conntrack_tuple *orig, |
949 | const struct nf_conntrack_tuple *repl, | |
950 | gfp_t gfp) | |
951 | { | |
952 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); | |
953 | } | |
13b18339 | 954 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
9fb9cbb1 | 955 | |
c88130bc | 956 | void nf_conntrack_free(struct nf_conn *ct) |
76507f69 | 957 | { |
1d45209d ED |
958 | struct net *net = nf_ct_net(ct); |
959 | ||
e53376be PNA |
960 | /* A freed object has refcnt == 0, that's |
961 | * the golden rule for SLAB_DESTROY_BY_RCU | |
962 | */ | |
963 | NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 0); | |
964 | ||
ceeff754 | 965 | nf_ct_ext_destroy(ct); |
ea781f19 | 966 | nf_ct_ext_free(ct); |
5b3501fa | 967 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
4e857c58 | 968 | smp_mb__before_atomic(); |
0c3c6c00 | 969 | atomic_dec(&net->ct.count); |
76507f69 | 970 | } |
13b18339 | 971 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
9fb9cbb1 | 972 | |
c539f017 | 973 | |
9fb9cbb1 YK |
974 | /* Allocate a new conntrack: we return -ENOMEM if classification |
975 | failed due to stress. Otherwise it really is unclassifiable. */ | |
976 | static struct nf_conntrack_tuple_hash * | |
b2a15a60 | 977 | init_conntrack(struct net *net, struct nf_conn *tmpl, |
5a1fb391 | 978 | const struct nf_conntrack_tuple *tuple, |
9fb9cbb1 | 979 | struct nf_conntrack_l3proto *l3proto, |
605dcad6 | 980 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 981 | struct sk_buff *skb, |
60b5f8f7 | 982 | unsigned int dataoff, u32 hash) |
9fb9cbb1 | 983 | { |
c88130bc | 984 | struct nf_conn *ct; |
3c158f7f | 985 | struct nf_conn_help *help; |
9fb9cbb1 | 986 | struct nf_conntrack_tuple repl_tuple; |
b2a15a60 | 987 | struct nf_conntrack_ecache *ecache; |
ca7433df | 988 | struct nf_conntrack_expect *exp = NULL; |
308ac914 | 989 | const struct nf_conntrack_zone *zone; |
60b5f8f7 | 990 | struct nf_conn_timeout *timeout_ext; |
5e8018fc | 991 | struct nf_conntrack_zone tmp; |
60b5f8f7 | 992 | unsigned int *timeouts; |
9fb9cbb1 | 993 | |
605dcad6 | 994 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
0d53778e | 995 | pr_debug("Can't invert tuple.\n"); |
9fb9cbb1 YK |
996 | return NULL; |
997 | } | |
998 | ||
5e8018fc | 999 | zone = nf_ct_zone_tmpl(tmpl, skb, &tmp); |
99f07e91 CG |
1000 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, |
1001 | hash); | |
0a9ee813 | 1002 | if (IS_ERR(ct)) |
c88130bc | 1003 | return (struct nf_conntrack_tuple_hash *)ct; |
9fb9cbb1 | 1004 | |
48b1de4c PM |
1005 | if (tmpl && nfct_synproxy(tmpl)) { |
1006 | nfct_seqadj_ext_add(ct); | |
1007 | nfct_synproxy_ext_add(ct); | |
1008 | } | |
1009 | ||
60b5f8f7 | 1010 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; |
ae2d708e PNA |
1011 | if (timeout_ext) { |
1012 | timeouts = nf_ct_timeout_data(timeout_ext); | |
1013 | if (unlikely(!timeouts)) | |
1014 | timeouts = l4proto->get_timeouts(net); | |
1015 | } else { | |
60b5f8f7 | 1016 | timeouts = l4proto->get_timeouts(net); |
ae2d708e | 1017 | } |
60b5f8f7 | 1018 | |
2c8503f5 | 1019 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { |
c88130bc | 1020 | nf_conntrack_free(ct); |
ccd63c20 | 1021 | pr_debug("can't track with proto module\n"); |
9fb9cbb1 YK |
1022 | return NULL; |
1023 | } | |
1024 | ||
60b5f8f7 | 1025 | if (timeout_ext) |
ae2d708e PNA |
1026 | nf_ct_timeout_ext_add(ct, rcu_dereference(timeout_ext->timeout), |
1027 | GFP_ATOMIC); | |
60b5f8f7 | 1028 | |
58401572 | 1029 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
a992ca2a | 1030 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); |
c539f017 | 1031 | nf_ct_labels_ext_add(ct); |
b2a15a60 PM |
1032 | |
1033 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; | |
1034 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, | |
1035 | ecache ? ecache->expmask : 0, | |
1036 | GFP_ATOMIC); | |
58401572 | 1037 | |
ca7433df JDB |
1038 | local_bh_disable(); |
1039 | if (net->ct.expect_count) { | |
1040 | spin_lock(&nf_conntrack_expect_lock); | |
1041 | exp = nf_ct_find_expectation(net, zone, tuple); | |
1042 | if (exp) { | |
ccd63c20 | 1043 | pr_debug("expectation arrives ct=%p exp=%p\n", |
ca7433df JDB |
1044 | ct, exp); |
1045 | /* Welcome, Mr. Bond. We've been expecting you... */ | |
1046 | __set_bit(IPS_EXPECTED_BIT, &ct->status); | |
1047 | /* exp->master safe, refcnt bumped in nf_ct_find_expectation */ | |
1048 | ct->master = exp->master; | |
1049 | if (exp->helper) { | |
1050 | help = nf_ct_helper_ext_add(ct, exp->helper, | |
1051 | GFP_ATOMIC); | |
1052 | if (help) | |
1053 | rcu_assign_pointer(help->helper, exp->helper); | |
1054 | } | |
ceceae1b | 1055 | |
9fb9cbb1 | 1056 | #ifdef CONFIG_NF_CONNTRACK_MARK |
ca7433df | 1057 | ct->mark = exp->master->mark; |
7c9728c3 JM |
1058 | #endif |
1059 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | |
ca7433df | 1060 | ct->secmark = exp->master->secmark; |
9fb9cbb1 | 1061 | #endif |
ca7433df JDB |
1062 | NF_CT_STAT_INC(net, expect_new); |
1063 | } | |
1064 | spin_unlock(&nf_conntrack_expect_lock); | |
1065 | } | |
1066 | if (!exp) { | |
b2a15a60 | 1067 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); |
0d55af87 | 1068 | NF_CT_STAT_INC(net, new); |
22e7410b | 1069 | } |
9fb9cbb1 | 1070 | |
e53376be PNA |
1071 | /* Now it is inserted into the unconfirmed list, bump refcount */ |
1072 | nf_conntrack_get(&ct->ct_general); | |
b7779d06 | 1073 | nf_ct_add_to_unconfirmed_list(ct); |
9fb9cbb1 | 1074 | |
ca7433df | 1075 | local_bh_enable(); |
9fb9cbb1 YK |
1076 | |
1077 | if (exp) { | |
1078 | if (exp->expectfn) | |
c88130bc | 1079 | exp->expectfn(ct, exp); |
6823645d | 1080 | nf_ct_expect_put(exp); |
9fb9cbb1 YK |
1081 | } |
1082 | ||
c88130bc | 1083 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
9fb9cbb1 YK |
1084 | } |
1085 | ||
1086 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | |
1087 | static inline struct nf_conn * | |
b2a15a60 | 1088 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, |
a702a65f | 1089 | struct sk_buff *skb, |
9fb9cbb1 YK |
1090 | unsigned int dataoff, |
1091 | u_int16_t l3num, | |
1092 | u_int8_t protonum, | |
1093 | struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 1094 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 1095 | int *set_reply, |
60b5f8f7 | 1096 | enum ip_conntrack_info *ctinfo) |
9fb9cbb1 | 1097 | { |
308ac914 | 1098 | const struct nf_conntrack_zone *zone; |
9fb9cbb1 YK |
1099 | struct nf_conntrack_tuple tuple; |
1100 | struct nf_conntrack_tuple_hash *h; | |
5e8018fc | 1101 | struct nf_conntrack_zone tmp; |
9fb9cbb1 | 1102 | struct nf_conn *ct; |
99f07e91 | 1103 | u32 hash; |
9fb9cbb1 | 1104 | |
bbe735e4 | 1105 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
a31f1adc | 1106 | dataoff, l3num, protonum, net, &tuple, l3proto, |
605dcad6 | 1107 | l4proto)) { |
ccd63c20 | 1108 | pr_debug("Can't get tuple\n"); |
9fb9cbb1 YK |
1109 | return NULL; |
1110 | } | |
1111 | ||
1112 | /* look for tuple match */ | |
5e8018fc | 1113 | zone = nf_ct_zone_tmpl(tmpl, skb, &tmp); |
1b8c8a9f | 1114 | hash = hash_conntrack_raw(&tuple, net); |
99f07e91 | 1115 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); |
9fb9cbb1 | 1116 | if (!h) { |
b2a15a60 | 1117 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, |
60b5f8f7 | 1118 | skb, dataoff, hash); |
9fb9cbb1 YK |
1119 | if (!h) |
1120 | return NULL; | |
1121 | if (IS_ERR(h)) | |
1122 | return (void *)h; | |
1123 | } | |
1124 | ct = nf_ct_tuplehash_to_ctrack(h); | |
1125 | ||
1126 | /* It exists; we have (non-exclusive) reference. */ | |
1127 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | |
fb048833 | 1128 | *ctinfo = IP_CT_ESTABLISHED_REPLY; |
9fb9cbb1 YK |
1129 | /* Please set reply bit if this packet OK */ |
1130 | *set_reply = 1; | |
1131 | } else { | |
1132 | /* Once we've had two way comms, always ESTABLISHED. */ | |
1133 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | |
ccd63c20 | 1134 | pr_debug("normal packet for %p\n", ct); |
9fb9cbb1 YK |
1135 | *ctinfo = IP_CT_ESTABLISHED; |
1136 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | |
ccd63c20 | 1137 | pr_debug("related packet for %p\n", ct); |
9fb9cbb1 YK |
1138 | *ctinfo = IP_CT_RELATED; |
1139 | } else { | |
ccd63c20 | 1140 | pr_debug("new packet for %p\n", ct); |
9fb9cbb1 YK |
1141 | *ctinfo = IP_CT_NEW; |
1142 | } | |
1143 | *set_reply = 0; | |
1144 | } | |
1145 | skb->nfct = &ct->ct_general; | |
1146 | skb->nfctinfo = *ctinfo; | |
1147 | return ct; | |
1148 | } | |
1149 | ||
1150 | unsigned int | |
a702a65f AD |
1151 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
1152 | struct sk_buff *skb) | |
9fb9cbb1 | 1153 | { |
b2a15a60 | 1154 | struct nf_conn *ct, *tmpl = NULL; |
9fb9cbb1 YK |
1155 | enum ip_conntrack_info ctinfo; |
1156 | struct nf_conntrack_l3proto *l3proto; | |
605dcad6 | 1157 | struct nf_conntrack_l4proto *l4proto; |
2c8503f5 | 1158 | unsigned int *timeouts; |
9fb9cbb1 YK |
1159 | unsigned int dataoff; |
1160 | u_int8_t protonum; | |
1161 | int set_reply = 0; | |
1162 | int ret; | |
1163 | ||
3db05fea | 1164 | if (skb->nfct) { |
b2a15a60 PM |
1165 | /* Previously seen (loopback or untracked)? Ignore. */ |
1166 | tmpl = (struct nf_conn *)skb->nfct; | |
1167 | if (!nf_ct_is_template(tmpl)) { | |
1168 | NF_CT_STAT_INC_ATOMIC(net, ignore); | |
1169 | return NF_ACCEPT; | |
1170 | } | |
1171 | skb->nfct = NULL; | |
9fb9cbb1 YK |
1172 | } |
1173 | ||
923f4902 | 1174 | /* rcu_read_lock()ed by nf_hook_slow */ |
76108cea | 1175 | l3proto = __nf_ct_l3proto_find(pf); |
3db05fea | 1176 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
ffc30690 YK |
1177 | &dataoff, &protonum); |
1178 | if (ret <= 0) { | |
25985edc | 1179 | pr_debug("not prepared to track yet or error occurred\n"); |
0d55af87 AD |
1180 | NF_CT_STAT_INC_ATOMIC(net, error); |
1181 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
1182 | ret = -ret; |
1183 | goto out; | |
9fb9cbb1 YK |
1184 | } |
1185 | ||
76108cea | 1186 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
9fb9cbb1 YK |
1187 | |
1188 | /* It may be an special packet, error, unclean... | |
1189 | * inverse of the return code tells to the netfilter | |
1190 | * core what to do with the packet. */ | |
74c51a14 | 1191 | if (l4proto->error != NULL) { |
8fea97ec PM |
1192 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, |
1193 | pf, hooknum); | |
74c51a14 | 1194 | if (ret <= 0) { |
0d55af87 AD |
1195 | NF_CT_STAT_INC_ATOMIC(net, error); |
1196 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
1197 | ret = -ret; |
1198 | goto out; | |
74c51a14 | 1199 | } |
88ed01d1 PNA |
1200 | /* ICMP[v6] protocol trackers may assign one conntrack. */ |
1201 | if (skb->nfct) | |
1202 | goto out; | |
9fb9cbb1 YK |
1203 | } |
1204 | ||
b2a15a60 | 1205 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, |
60b5f8f7 | 1206 | l3proto, l4proto, &set_reply, &ctinfo); |
9fb9cbb1 YK |
1207 | if (!ct) { |
1208 | /* Not valid part of a connection */ | |
0d55af87 | 1209 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
b2a15a60 PM |
1210 | ret = NF_ACCEPT; |
1211 | goto out; | |
9fb9cbb1 YK |
1212 | } |
1213 | ||
1214 | if (IS_ERR(ct)) { | |
1215 | /* Too stressed to deal. */ | |
0d55af87 | 1216 | NF_CT_STAT_INC_ATOMIC(net, drop); |
b2a15a60 PM |
1217 | ret = NF_DROP; |
1218 | goto out; | |
9fb9cbb1 YK |
1219 | } |
1220 | ||
3db05fea | 1221 | NF_CT_ASSERT(skb->nfct); |
9fb9cbb1 | 1222 | |
60b5f8f7 | 1223 | /* Decide what timeout policy we want to apply to this flow. */ |
84b5ee93 | 1224 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); |
60b5f8f7 | 1225 | |
2c8503f5 | 1226 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); |
ec8d5409 | 1227 | if (ret <= 0) { |
9fb9cbb1 YK |
1228 | /* Invalid: inverse of the return code tells |
1229 | * the netfilter core what to do */ | |
0d53778e | 1230 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
3db05fea HX |
1231 | nf_conntrack_put(skb->nfct); |
1232 | skb->nfct = NULL; | |
0d55af87 | 1233 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
7d1e0459 PNA |
1234 | if (ret == -NF_DROP) |
1235 | NF_CT_STAT_INC_ATOMIC(net, drop); | |
b2a15a60 PM |
1236 | ret = -ret; |
1237 | goto out; | |
9fb9cbb1 YK |
1238 | } |
1239 | ||
1240 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
858b3133 | 1241 | nf_conntrack_event_cache(IPCT_REPLY, ct); |
b2a15a60 | 1242 | out: |
c3174286 PNA |
1243 | if (tmpl) { |
1244 | /* Special case: we have to repeat this hook, assign the | |
1245 | * template again to this packet. We assume that this packet | |
1246 | * has no conntrack assigned. This is used by nf_ct_tcp. */ | |
1247 | if (ret == NF_REPEAT) | |
1248 | skb->nfct = (struct nf_conntrack *)tmpl; | |
1249 | else | |
1250 | nf_ct_put(tmpl); | |
1251 | } | |
9fb9cbb1 YK |
1252 | |
1253 | return ret; | |
1254 | } | |
13b18339 | 1255 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
9fb9cbb1 | 1256 | |
5f2b4c90 JE |
1257 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
1258 | const struct nf_conntrack_tuple *orig) | |
9fb9cbb1 | 1259 | { |
5f2b4c90 | 1260 | bool ret; |
923f4902 PM |
1261 | |
1262 | rcu_read_lock(); | |
1263 | ret = nf_ct_invert_tuple(inverse, orig, | |
1264 | __nf_ct_l3proto_find(orig->src.l3num), | |
1265 | __nf_ct_l4proto_find(orig->src.l3num, | |
1266 | orig->dst.protonum)); | |
1267 | rcu_read_unlock(); | |
1268 | return ret; | |
9fb9cbb1 | 1269 | } |
13b18339 | 1270 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
9fb9cbb1 | 1271 | |
5b1158e9 JK |
1272 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
1273 | implicitly racy: see __nf_conntrack_confirm */ | |
1274 | void nf_conntrack_alter_reply(struct nf_conn *ct, | |
1275 | const struct nf_conntrack_tuple *newreply) | |
1276 | { | |
1277 | struct nf_conn_help *help = nfct_help(ct); | |
1278 | ||
5b1158e9 JK |
1279 | /* Should be unconfirmed, so not in hash table yet */ |
1280 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
1281 | ||
0d53778e | 1282 | pr_debug("Altering reply tuple of %p to ", ct); |
3c9fba65 | 1283 | nf_ct_dump_tuple(newreply); |
5b1158e9 JK |
1284 | |
1285 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | |
ef1a5a50 | 1286 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
c52fbb41 | 1287 | return; |
ceceae1b | 1288 | |
c52fbb41 | 1289 | rcu_read_lock(); |
b2a15a60 | 1290 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); |
c52fbb41 | 1291 | rcu_read_unlock(); |
5b1158e9 | 1292 | } |
13b18339 | 1293 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
5b1158e9 | 1294 | |
9fb9cbb1 YK |
1295 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
1296 | void __nf_ct_refresh_acct(struct nf_conn *ct, | |
1297 | enum ip_conntrack_info ctinfo, | |
1298 | const struct sk_buff *skb, | |
1299 | unsigned long extra_jiffies, | |
1300 | int do_acct) | |
1301 | { | |
9fb9cbb1 YK |
1302 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
1303 | NF_CT_ASSERT(skb); | |
1304 | ||
997ae831 | 1305 | /* Only update if this is not a fixed timeout */ |
47d95045 PM |
1306 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
1307 | goto acct; | |
997ae831 | 1308 | |
9fb9cbb1 YK |
1309 | /* If not in hash table, timer will not be active yet */ |
1310 | if (!nf_ct_is_confirmed(ct)) { | |
1311 | ct->timeout.expires = extra_jiffies; | |
9fb9cbb1 | 1312 | } else { |
be00c8e4 MJ |
1313 | unsigned long newtime = jiffies + extra_jiffies; |
1314 | ||
1315 | /* Only update the timeout if the new timeout is at least | |
1316 | HZ jiffies from the old timeout. Need del_timer for race | |
1317 | avoidance (may already be dying). */ | |
65cb9fda PM |
1318 | if (newtime - ct->timeout.expires >= HZ) |
1319 | mod_timer_pending(&ct->timeout, newtime); | |
9fb9cbb1 YK |
1320 | } |
1321 | ||
47d95045 | 1322 | acct: |
ba76738c PNA |
1323 | if (do_acct) |
1324 | nf_ct_acct_update(ct, ctinfo, skb->len); | |
9fb9cbb1 | 1325 | } |
13b18339 | 1326 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
9fb9cbb1 | 1327 | |
4c889498 DM |
1328 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
1329 | enum ip_conntrack_info ctinfo, | |
1330 | const struct sk_buff *skb, | |
1331 | int do_acct) | |
51091764 | 1332 | { |
ba76738c PNA |
1333 | if (do_acct) |
1334 | nf_ct_acct_update(ct, ctinfo, skb->len); | |
58401572 | 1335 | |
4c889498 | 1336 | if (del_timer(&ct->timeout)) { |
51091764 | 1337 | ct->timeout.function((unsigned long)ct); |
4c889498 DM |
1338 | return true; |
1339 | } | |
1340 | return false; | |
51091764 | 1341 | } |
718d4ad9 | 1342 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
51091764 | 1343 | |
5d0aa2cc PM |
1344 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1345 | static struct nf_ct_ext_type nf_ct_zone_extend __read_mostly = { | |
1346 | .len = sizeof(struct nf_conntrack_zone), | |
1347 | .align = __alignof__(struct nf_conntrack_zone), | |
1348 | .id = NF_CT_EXT_ZONE, | |
1349 | }; | |
1350 | #endif | |
1351 | ||
c0cd1156 | 1352 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
c1d10adb PNA |
1353 | |
1354 | #include <linux/netfilter/nfnetlink.h> | |
1355 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
57b47a53 IM |
1356 | #include <linux/mutex.h> |
1357 | ||
c1d10adb PNA |
1358 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
1359 | * in ip_conntrack_core, since we don't want the protocols to autoload | |
1360 | * or depend on ctnetlink */ | |
fdf70832 | 1361 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
c1d10adb PNA |
1362 | const struct nf_conntrack_tuple *tuple) |
1363 | { | |
bae65be8 DM |
1364 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || |
1365 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) | |
1366 | goto nla_put_failure; | |
c1d10adb PNA |
1367 | return 0; |
1368 | ||
df6fb868 | 1369 | nla_put_failure: |
c1d10adb PNA |
1370 | return -1; |
1371 | } | |
fdf70832 | 1372 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
c1d10adb | 1373 | |
f73e924c PM |
1374 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
1375 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | |
1376 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | |
c1d10adb | 1377 | }; |
f73e924c | 1378 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
c1d10adb | 1379 | |
fdf70832 | 1380 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
c1d10adb PNA |
1381 | struct nf_conntrack_tuple *t) |
1382 | { | |
df6fb868 | 1383 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
c1d10adb PNA |
1384 | return -EINVAL; |
1385 | ||
77236b6e PM |
1386 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
1387 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | |
c1d10adb PNA |
1388 | |
1389 | return 0; | |
1390 | } | |
fdf70832 | 1391 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
5c0de29d HE |
1392 | |
1393 | int nf_ct_port_nlattr_tuple_size(void) | |
1394 | { | |
1395 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
1396 | } | |
1397 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | |
c1d10adb PNA |
1398 | #endif |
1399 | ||
9fb9cbb1 | 1400 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
312a0c16 | 1401 | static void nf_conntrack_attach(struct sk_buff *nskb, const struct sk_buff *skb) |
9fb9cbb1 YK |
1402 | { |
1403 | struct nf_conn *ct; | |
1404 | enum ip_conntrack_info ctinfo; | |
1405 | ||
1406 | /* This ICMP is in reverse direction to the packet which caused it */ | |
1407 | ct = nf_ct_get(skb, &ctinfo); | |
1408 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | |
fb048833 | 1409 | ctinfo = IP_CT_RELATED_REPLY; |
9fb9cbb1 YK |
1410 | else |
1411 | ctinfo = IP_CT_RELATED; | |
1412 | ||
1413 | /* Attach to new skbuff, and increment count */ | |
1414 | nskb->nfct = &ct->ct_general; | |
1415 | nskb->nfctinfo = ctinfo; | |
1416 | nf_conntrack_get(nskb->nfct); | |
1417 | } | |
1418 | ||
9fb9cbb1 | 1419 | /* Bring out ya dead! */ |
df0933dc | 1420 | static struct nf_conn * |
400dad39 | 1421 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
9fb9cbb1 YK |
1422 | void *data, unsigned int *bucket) |
1423 | { | |
df0933dc PM |
1424 | struct nf_conntrack_tuple_hash *h; |
1425 | struct nf_conn *ct; | |
ea781f19 | 1426 | struct hlist_nulls_node *n; |
b7779d06 | 1427 | int cpu; |
93bb0ceb | 1428 | spinlock_t *lockp; |
9fb9cbb1 | 1429 | |
56d52d48 | 1430 | for (; *bucket < nf_conntrack_htable_size; (*bucket)++) { |
93bb0ceb JDB |
1431 | lockp = &nf_conntrack_locks[*bucket % CONNTRACK_LOCKS]; |
1432 | local_bh_disable(); | |
b16c2919 | 1433 | nf_conntrack_lock(lockp); |
56d52d48 FW |
1434 | if (*bucket < nf_conntrack_htable_size) { |
1435 | hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[*bucket], hnnode) { | |
93bb0ceb JDB |
1436 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) |
1437 | continue; | |
1438 | ct = nf_ct_tuplehash_to_ctrack(h); | |
e0c7d472 FW |
1439 | if (net_eq(nf_ct_net(ct), net) && |
1440 | iter(ct, data)) | |
93bb0ceb JDB |
1441 | goto found; |
1442 | } | |
df0933dc | 1443 | } |
93bb0ceb JDB |
1444 | spin_unlock(lockp); |
1445 | local_bh_enable(); | |
d93c6258 | 1446 | cond_resched(); |
601e68e1 | 1447 | } |
b7779d06 JDB |
1448 | |
1449 | for_each_possible_cpu(cpu) { | |
1450 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | |
1451 | ||
1452 | spin_lock_bh(&pcpu->lock); | |
1453 | hlist_nulls_for_each_entry(h, n, &pcpu->unconfirmed, hnnode) { | |
1454 | ct = nf_ct_tuplehash_to_ctrack(h); | |
1455 | if (iter(ct, data)) | |
1456 | set_bit(IPS_DYING_BIT, &ct->status); | |
1457 | } | |
1458 | spin_unlock_bh(&pcpu->lock); | |
d93c6258 | 1459 | cond_resched(); |
b7779d06 | 1460 | } |
df0933dc PM |
1461 | return NULL; |
1462 | found: | |
c073e3fa | 1463 | atomic_inc(&ct->ct_general.use); |
93bb0ceb JDB |
1464 | spin_unlock(lockp); |
1465 | local_bh_enable(); | |
df0933dc | 1466 | return ct; |
9fb9cbb1 YK |
1467 | } |
1468 | ||
400dad39 AD |
1469 | void nf_ct_iterate_cleanup(struct net *net, |
1470 | int (*iter)(struct nf_conn *i, void *data), | |
c655bc68 | 1471 | void *data, u32 portid, int report) |
9fb9cbb1 | 1472 | { |
df0933dc | 1473 | struct nf_conn *ct; |
9fb9cbb1 YK |
1474 | unsigned int bucket = 0; |
1475 | ||
d93c6258 FW |
1476 | might_sleep(); |
1477 | ||
88b68bc5 FW |
1478 | if (atomic_read(&net->ct.count) == 0) |
1479 | return; | |
1480 | ||
400dad39 | 1481 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
9fb9cbb1 YK |
1482 | /* Time to push up daises... */ |
1483 | if (del_timer(&ct->timeout)) | |
c655bc68 | 1484 | nf_ct_delete(ct, portid, report); |
02982c27 | 1485 | |
9fb9cbb1 YK |
1486 | /* ... else the timer will get him soon. */ |
1487 | ||
1488 | nf_ct_put(ct); | |
d93c6258 | 1489 | cond_resched(); |
9fb9cbb1 YK |
1490 | } |
1491 | } | |
13b18339 | 1492 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
9fb9cbb1 | 1493 | |
274d383b PNA |
1494 | static int kill_all(struct nf_conn *i, void *data) |
1495 | { | |
1496 | return 1; | |
1497 | } | |
1498 | ||
d862a662 | 1499 | void nf_ct_free_hashtable(void *hash, unsigned int size) |
9fb9cbb1 | 1500 | { |
d862a662 | 1501 | if (is_vmalloc_addr(hash)) |
9fb9cbb1 YK |
1502 | vfree(hash); |
1503 | else | |
601e68e1 | 1504 | free_pages((unsigned long)hash, |
f205c5e0 | 1505 | get_order(sizeof(struct hlist_head) * size)); |
9fb9cbb1 | 1506 | } |
ac565e5f | 1507 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
9fb9cbb1 | 1508 | |
b3c5163f ED |
1509 | static int untrack_refs(void) |
1510 | { | |
1511 | int cnt = 0, cpu; | |
1512 | ||
1513 | for_each_possible_cpu(cpu) { | |
1514 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
1515 | ||
1516 | cnt += atomic_read(&ct->ct_general.use) - 1; | |
1517 | } | |
1518 | return cnt; | |
1519 | } | |
1520 | ||
f94161c1 | 1521 | void nf_conntrack_cleanup_start(void) |
9fb9cbb1 | 1522 | { |
f94161c1 G |
1523 | RCU_INIT_POINTER(ip_ct_attach, NULL); |
1524 | } | |
1525 | ||
1526 | void nf_conntrack_cleanup_end(void) | |
1527 | { | |
1528 | RCU_INIT_POINTER(nf_ct_destroy, NULL); | |
b3c5163f | 1529 | while (untrack_refs() > 0) |
9edd7ca0 PM |
1530 | schedule(); |
1531 | ||
56d52d48 FW |
1532 | nf_ct_free_hashtable(nf_conntrack_hash, nf_conntrack_htable_size); |
1533 | ||
5d0aa2cc PM |
1534 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1535 | nf_ct_extend_unregister(&nf_ct_zone_extend); | |
1536 | #endif | |
04d87001 | 1537 | nf_conntrack_proto_fini(); |
41d73ec0 | 1538 | nf_conntrack_seqadj_fini(); |
5f69b8f5 | 1539 | nf_conntrack_labels_fini(); |
5e615b22 | 1540 | nf_conntrack_helper_fini(); |
8684094c | 1541 | nf_conntrack_timeout_fini(); |
3fe0f943 | 1542 | nf_conntrack_ecache_fini(); |
73f4001a | 1543 | nf_conntrack_tstamp_fini(); |
b7ff3a1f | 1544 | nf_conntrack_acct_fini(); |
83b4dbe1 | 1545 | nf_conntrack_expect_fini(); |
08f6547d | 1546 | } |
9fb9cbb1 | 1547 | |
f94161c1 G |
1548 | /* |
1549 | * Mishearing the voices in his head, our hero wonders how he's | |
1550 | * supposed to kill the mall. | |
1551 | */ | |
1552 | void nf_conntrack_cleanup_net(struct net *net) | |
08f6547d | 1553 | { |
dece40e8 VD |
1554 | LIST_HEAD(single); |
1555 | ||
1556 | list_add(&net->exit_list, &single); | |
1557 | nf_conntrack_cleanup_net_list(&single); | |
1558 | } | |
1559 | ||
1560 | void nf_conntrack_cleanup_net_list(struct list_head *net_exit_list) | |
1561 | { | |
1562 | int busy; | |
1563 | struct net *net; | |
1564 | ||
f94161c1 G |
1565 | /* |
1566 | * This makes sure all current packets have passed through | |
1567 | * netfilter framework. Roll on, two-stage module | |
1568 | * delete... | |
1569 | */ | |
1570 | synchronize_net(); | |
dece40e8 VD |
1571 | i_see_dead_people: |
1572 | busy = 0; | |
1573 | list_for_each_entry(net, net_exit_list, exit_list) { | |
c655bc68 | 1574 | nf_ct_iterate_cleanup(net, kill_all, NULL, 0, 0); |
dece40e8 VD |
1575 | if (atomic_read(&net->ct.count) != 0) |
1576 | busy = 1; | |
1577 | } | |
1578 | if (busy) { | |
9fb9cbb1 YK |
1579 | schedule(); |
1580 | goto i_see_dead_people; | |
1581 | } | |
1582 | ||
dece40e8 | 1583 | list_for_each_entry(net, net_exit_list, exit_list) { |
dece40e8 VD |
1584 | nf_conntrack_proto_pernet_fini(net); |
1585 | nf_conntrack_helper_pernet_fini(net); | |
1586 | nf_conntrack_ecache_pernet_fini(net); | |
1587 | nf_conntrack_tstamp_pernet_fini(net); | |
1588 | nf_conntrack_acct_pernet_fini(net); | |
1589 | nf_conntrack_expect_pernet_fini(net); | |
1590 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); | |
1591 | kfree(net->ct.slabname); | |
1592 | free_percpu(net->ct.stat); | |
b7779d06 | 1593 | free_percpu(net->ct.pcpu_lists); |
dece40e8 | 1594 | } |
08f6547d AD |
1595 | } |
1596 | ||
d862a662 | 1597 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) |
9fb9cbb1 | 1598 | { |
ea781f19 ED |
1599 | struct hlist_nulls_head *hash; |
1600 | unsigned int nr_slots, i; | |
1601 | size_t sz; | |
9fb9cbb1 | 1602 | |
ea781f19 ED |
1603 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1604 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | |
1605 | sz = nr_slots * sizeof(struct hlist_nulls_head); | |
1606 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | |
1607 | get_order(sz)); | |
f0ad4621 | 1608 | if (!hash) |
966567b7 | 1609 | hash = vzalloc(sz); |
9fb9cbb1 | 1610 | |
ea781f19 ED |
1611 | if (hash && nulls) |
1612 | for (i = 0; i < nr_slots; i++) | |
1613 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | |
9fb9cbb1 YK |
1614 | |
1615 | return hash; | |
1616 | } | |
ac565e5f | 1617 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
9fb9cbb1 | 1618 | |
fae718dd | 1619 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
9fb9cbb1 | 1620 | { |
4b5511eb | 1621 | int i, bucket, rc; |
96eb24d7 | 1622 | unsigned int hashsize, old_size; |
ea781f19 | 1623 | struct hlist_nulls_head *hash, *old_hash; |
9fb9cbb1 | 1624 | struct nf_conntrack_tuple_hash *h; |
5d0aa2cc | 1625 | struct nf_conn *ct; |
9fb9cbb1 | 1626 | |
d696c7bd PM |
1627 | if (current->nsproxy->net_ns != &init_net) |
1628 | return -EOPNOTSUPP; | |
1629 | ||
9fb9cbb1 YK |
1630 | /* On boot, we can set this without any fancy locking. */ |
1631 | if (!nf_conntrack_htable_size) | |
1632 | return param_set_uint(val, kp); | |
1633 | ||
4b5511eb AP |
1634 | rc = kstrtouint(val, 0, &hashsize); |
1635 | if (rc) | |
1636 | return rc; | |
9fb9cbb1 YK |
1637 | if (!hashsize) |
1638 | return -EINVAL; | |
1639 | ||
d862a662 | 1640 | hash = nf_ct_alloc_hashtable(&hashsize, 1); |
9fb9cbb1 YK |
1641 | if (!hash) |
1642 | return -ENOMEM; | |
1643 | ||
93bb0ceb JDB |
1644 | local_bh_disable(); |
1645 | nf_conntrack_all_lock(); | |
a3efd812 | 1646 | write_seqcount_begin(&nf_conntrack_generation); |
93bb0ceb | 1647 | |
76507f69 PM |
1648 | /* Lookups in the old hash might happen in parallel, which means we |
1649 | * might get false negatives during connection lookup. New connections | |
1650 | * created because of a false negative won't make it into the hash | |
93bb0ceb | 1651 | * though since that required taking the locks. |
76507f69 | 1652 | */ |
93bb0ceb | 1653 | |
56d52d48 FW |
1654 | for (i = 0; i < nf_conntrack_htable_size; i++) { |
1655 | while (!hlist_nulls_empty(&nf_conntrack_hash[i])) { | |
1656 | h = hlist_nulls_entry(nf_conntrack_hash[i].first, | |
1657 | struct nf_conntrack_tuple_hash, hnnode); | |
5d0aa2cc | 1658 | ct = nf_ct_tuplehash_to_ctrack(h); |
ea781f19 | 1659 | hlist_nulls_del_rcu(&h->hnnode); |
1b8c8a9f FW |
1660 | bucket = __hash_conntrack(nf_ct_net(ct), |
1661 | &h->tuple, hashsize); | |
ea781f19 | 1662 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
9fb9cbb1 YK |
1663 | } |
1664 | } | |
56d52d48 FW |
1665 | old_size = nf_conntrack_htable_size; |
1666 | old_hash = nf_conntrack_hash; | |
9fb9cbb1 | 1667 | |
56d52d48 FW |
1668 | nf_conntrack_hash = hash; |
1669 | nf_conntrack_htable_size = hashsize; | |
93bb0ceb | 1670 | |
a3efd812 | 1671 | write_seqcount_end(&nf_conntrack_generation); |
93bb0ceb JDB |
1672 | nf_conntrack_all_unlock(); |
1673 | local_bh_enable(); | |
9fb9cbb1 | 1674 | |
5e3c61f9 | 1675 | synchronize_net(); |
d862a662 | 1676 | nf_ct_free_hashtable(old_hash, old_size); |
9fb9cbb1 YK |
1677 | return 0; |
1678 | } | |
fae718dd | 1679 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
9fb9cbb1 | 1680 | |
fae718dd | 1681 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
9fb9cbb1 YK |
1682 | &nf_conntrack_htable_size, 0600); |
1683 | ||
5bfddbd4 ED |
1684 | void nf_ct_untracked_status_or(unsigned long bits) |
1685 | { | |
b3c5163f ED |
1686 | int cpu; |
1687 | ||
1688 | for_each_possible_cpu(cpu) | |
1689 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; | |
5bfddbd4 ED |
1690 | } |
1691 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); | |
1692 | ||
f94161c1 | 1693 | int nf_conntrack_init_start(void) |
9fb9cbb1 | 1694 | { |
f205c5e0 | 1695 | int max_factor = 8; |
93bb0ceb JDB |
1696 | int i, ret, cpu; |
1697 | ||
a3efd812 FW |
1698 | seqcount_init(&nf_conntrack_generation); |
1699 | ||
d5d20912 | 1700 | for (i = 0; i < CONNTRACK_LOCKS; i++) |
93bb0ceb | 1701 | spin_lock_init(&nf_conntrack_locks[i]); |
9fb9cbb1 | 1702 | |
9fb9cbb1 | 1703 | if (!nf_conntrack_htable_size) { |
88eab472 ML |
1704 | /* Idea from tcp.c: use 1/16384 of memory. |
1705 | * On i386: 32MB machine has 512 buckets. | |
1706 | * >= 1GB machines have 16384 buckets. | |
1707 | * >= 4GB machines have 65536 buckets. | |
1708 | */ | |
9fb9cbb1 | 1709 | nf_conntrack_htable_size |
4481374c | 1710 | = (((totalram_pages << PAGE_SHIFT) / 16384) |
f205c5e0 | 1711 | / sizeof(struct hlist_head)); |
88eab472 ML |
1712 | if (totalram_pages > (4 * (1024 * 1024 * 1024 / PAGE_SIZE))) |
1713 | nf_conntrack_htable_size = 65536; | |
1714 | else if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) | |
f205c5e0 PM |
1715 | nf_conntrack_htable_size = 16384; |
1716 | if (nf_conntrack_htable_size < 32) | |
1717 | nf_conntrack_htable_size = 32; | |
1718 | ||
1719 | /* Use a max. factor of four by default to get the same max as | |
1720 | * with the old struct list_heads. When a table size is given | |
1721 | * we use the old value of 8 to avoid reducing the max. | |
1722 | * entries. */ | |
1723 | max_factor = 4; | |
9fb9cbb1 | 1724 | } |
56d52d48 FW |
1725 | |
1726 | nf_conntrack_hash = nf_ct_alloc_hashtable(&nf_conntrack_htable_size, 1); | |
1727 | if (!nf_conntrack_hash) | |
1728 | return -ENOMEM; | |
1729 | ||
f205c5e0 | 1730 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
8e5105a0 | 1731 | |
654d0fbd | 1732 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", |
8e5105a0 PM |
1733 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, |
1734 | nf_conntrack_max); | |
83b4dbe1 G |
1735 | |
1736 | ret = nf_conntrack_expect_init(); | |
1737 | if (ret < 0) | |
1738 | goto err_expect; | |
1739 | ||
b7ff3a1f G |
1740 | ret = nf_conntrack_acct_init(); |
1741 | if (ret < 0) | |
1742 | goto err_acct; | |
1743 | ||
73f4001a G |
1744 | ret = nf_conntrack_tstamp_init(); |
1745 | if (ret < 0) | |
1746 | goto err_tstamp; | |
1747 | ||
3fe0f943 G |
1748 | ret = nf_conntrack_ecache_init(); |
1749 | if (ret < 0) | |
1750 | goto err_ecache; | |
1751 | ||
8684094c G |
1752 | ret = nf_conntrack_timeout_init(); |
1753 | if (ret < 0) | |
1754 | goto err_timeout; | |
1755 | ||
5e615b22 G |
1756 | ret = nf_conntrack_helper_init(); |
1757 | if (ret < 0) | |
1758 | goto err_helper; | |
1759 | ||
5f69b8f5 G |
1760 | ret = nf_conntrack_labels_init(); |
1761 | if (ret < 0) | |
1762 | goto err_labels; | |
1763 | ||
41d73ec0 PM |
1764 | ret = nf_conntrack_seqadj_init(); |
1765 | if (ret < 0) | |
1766 | goto err_seqadj; | |
1767 | ||
5d0aa2cc PM |
1768 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1769 | ret = nf_ct_extend_register(&nf_ct_zone_extend); | |
1770 | if (ret < 0) | |
1771 | goto err_extend; | |
1772 | #endif | |
04d87001 G |
1773 | ret = nf_conntrack_proto_init(); |
1774 | if (ret < 0) | |
1775 | goto err_proto; | |
1776 | ||
9edd7ca0 | 1777 | /* Set up fake conntrack: to never be deleted, not in any hashes */ |
b3c5163f ED |
1778 | for_each_possible_cpu(cpu) { |
1779 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
b3c5163f ED |
1780 | write_pnet(&ct->ct_net, &init_net); |
1781 | atomic_set(&ct->ct_general.use, 1); | |
1782 | } | |
9edd7ca0 | 1783 | /* - and look it like as a confirmed connection */ |
5bfddbd4 | 1784 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); |
08f6547d AD |
1785 | return 0; |
1786 | ||
04d87001 | 1787 | err_proto: |
5d0aa2cc | 1788 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
04d87001 | 1789 | nf_ct_extend_unregister(&nf_ct_zone_extend); |
5d0aa2cc | 1790 | err_extend: |
a9006892 | 1791 | #endif |
41d73ec0 PM |
1792 | nf_conntrack_seqadj_fini(); |
1793 | err_seqadj: | |
04d87001 | 1794 | nf_conntrack_labels_fini(); |
5f69b8f5 G |
1795 | err_labels: |
1796 | nf_conntrack_helper_fini(); | |
5e615b22 G |
1797 | err_helper: |
1798 | nf_conntrack_timeout_fini(); | |
8684094c G |
1799 | err_timeout: |
1800 | nf_conntrack_ecache_fini(); | |
3fe0f943 G |
1801 | err_ecache: |
1802 | nf_conntrack_tstamp_fini(); | |
73f4001a G |
1803 | err_tstamp: |
1804 | nf_conntrack_acct_fini(); | |
b7ff3a1f G |
1805 | err_acct: |
1806 | nf_conntrack_expect_fini(); | |
83b4dbe1 | 1807 | err_expect: |
56d52d48 | 1808 | nf_ct_free_hashtable(nf_conntrack_hash, nf_conntrack_htable_size); |
08f6547d AD |
1809 | return ret; |
1810 | } | |
1811 | ||
f94161c1 G |
1812 | void nf_conntrack_init_end(void) |
1813 | { | |
1814 | /* For use by REJECT target */ | |
1815 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); | |
1816 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); | |
f94161c1 G |
1817 | } |
1818 | ||
8cc20198 ED |
1819 | /* |
1820 | * We need to use special "null" values, not used in hash table | |
1821 | */ | |
1822 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) | |
1823 | #define DYING_NULLS_VAL ((1<<30)+1) | |
252b3e8c | 1824 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) |
8cc20198 | 1825 | |
f94161c1 | 1826 | int nf_conntrack_init_net(struct net *net) |
08f6547d | 1827 | { |
b7779d06 JDB |
1828 | int ret = -ENOMEM; |
1829 | int cpu; | |
ceceae1b | 1830 | |
08f6547d | 1831 | atomic_set(&net->ct.count, 0); |
b7779d06 JDB |
1832 | |
1833 | net->ct.pcpu_lists = alloc_percpu(struct ct_pcpu); | |
1834 | if (!net->ct.pcpu_lists) | |
08f6547d | 1835 | goto err_stat; |
b7779d06 JDB |
1836 | |
1837 | for_each_possible_cpu(cpu) { | |
1838 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | |
1839 | ||
1840 | spin_lock_init(&pcpu->lock); | |
1841 | INIT_HLIST_NULLS_HEAD(&pcpu->unconfirmed, UNCONFIRMED_NULLS_VAL); | |
1842 | INIT_HLIST_NULLS_HEAD(&pcpu->dying, DYING_NULLS_VAL); | |
08f6547d | 1843 | } |
5b3501fa | 1844 | |
b7779d06 JDB |
1845 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1846 | if (!net->ct.stat) | |
1847 | goto err_pcpu_lists; | |
1848 | ||
5b3501fa | 1849 | net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net); |
b7779d06 | 1850 | if (!net->ct.slabname) |
5b3501fa | 1851 | goto err_slabname; |
5b3501fa ED |
1852 | |
1853 | net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname, | |
1854 | sizeof(struct nf_conn), 0, | |
1855 | SLAB_DESTROY_BY_RCU, NULL); | |
1856 | if (!net->ct.nf_conntrack_cachep) { | |
1857 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); | |
5b3501fa ED |
1858 | goto err_cache; |
1859 | } | |
d696c7bd | 1860 | |
83b4dbe1 | 1861 | ret = nf_conntrack_expect_pernet_init(net); |
08f6547d AD |
1862 | if (ret < 0) |
1863 | goto err_expect; | |
b7ff3a1f | 1864 | ret = nf_conntrack_acct_pernet_init(net); |
58401572 | 1865 | if (ret < 0) |
08f6547d | 1866 | goto err_acct; |
73f4001a | 1867 | ret = nf_conntrack_tstamp_pernet_init(net); |
a992ca2a PNA |
1868 | if (ret < 0) |
1869 | goto err_tstamp; | |
3fe0f943 | 1870 | ret = nf_conntrack_ecache_pernet_init(net); |
a0891aa6 PNA |
1871 | if (ret < 0) |
1872 | goto err_ecache; | |
5e615b22 | 1873 | ret = nf_conntrack_helper_pernet_init(net); |
a9006892 EL |
1874 | if (ret < 0) |
1875 | goto err_helper; | |
04d87001 | 1876 | ret = nf_conntrack_proto_pernet_init(net); |
f94161c1 G |
1877 | if (ret < 0) |
1878 | goto err_proto; | |
08f6547d | 1879 | return 0; |
c539f017 | 1880 | |
f94161c1 | 1881 | err_proto: |
5e615b22 | 1882 | nf_conntrack_helper_pernet_fini(net); |
a9006892 | 1883 | err_helper: |
3fe0f943 | 1884 | nf_conntrack_ecache_pernet_fini(net); |
a0891aa6 | 1885 | err_ecache: |
73f4001a | 1886 | nf_conntrack_tstamp_pernet_fini(net); |
a992ca2a | 1887 | err_tstamp: |
b7ff3a1f | 1888 | nf_conntrack_acct_pernet_fini(net); |
08f6547d | 1889 | err_acct: |
83b4dbe1 | 1890 | nf_conntrack_expect_pernet_fini(net); |
08f6547d | 1891 | err_expect: |
5b3501fa ED |
1892 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
1893 | err_cache: | |
1894 | kfree(net->ct.slabname); | |
1895 | err_slabname: | |
0d55af87 | 1896 | free_percpu(net->ct.stat); |
b7779d06 JDB |
1897 | err_pcpu_lists: |
1898 | free_percpu(net->ct.pcpu_lists); | |
0d55af87 | 1899 | err_stat: |
08f6547d AD |
1900 | return ret; |
1901 | } |