Commit | Line | Data |
---|---|---|
9fb9cbb1 YK |
1 | /* Connection state tracking for netfilter. This is separated from, |
2 | but required by, the NAT layer; it can also be used by an iptables | |
3 | extension. */ | |
4 | ||
5 | /* (C) 1999-2001 Paul `Rusty' Russell | |
dc808fe2 | 6 | * (C) 2002-2006 Netfilter Core Team <coreteam@netfilter.org> |
9fb9cbb1 | 7 | * (C) 2003,2004 USAGI/WIDE Project <http://www.linux-ipv6.org> |
f229f6ce | 8 | * (C) 2005-2012 Patrick McHardy <kaber@trash.net> |
9fb9cbb1 YK |
9 | * |
10 | * This program is free software; you can redistribute it and/or modify | |
11 | * it under the terms of the GNU General Public License version 2 as | |
12 | * published by the Free Software Foundation. | |
9fb9cbb1 YK |
13 | */ |
14 | ||
9fb9cbb1 YK |
15 | #include <linux/types.h> |
16 | #include <linux/netfilter.h> | |
17 | #include <linux/module.h> | |
d43c36dc | 18 | #include <linux/sched.h> |
9fb9cbb1 YK |
19 | #include <linux/skbuff.h> |
20 | #include <linux/proc_fs.h> | |
21 | #include <linux/vmalloc.h> | |
22 | #include <linux/stddef.h> | |
23 | #include <linux/slab.h> | |
24 | #include <linux/random.h> | |
25 | #include <linux/jhash.h> | |
26 | #include <linux/err.h> | |
27 | #include <linux/percpu.h> | |
28 | #include <linux/moduleparam.h> | |
29 | #include <linux/notifier.h> | |
30 | #include <linux/kernel.h> | |
31 | #include <linux/netdevice.h> | |
32 | #include <linux/socket.h> | |
d7fe0f24 | 33 | #include <linux/mm.h> |
d696c7bd | 34 | #include <linux/nsproxy.h> |
ea781f19 | 35 | #include <linux/rculist_nulls.h> |
9fb9cbb1 | 36 | |
9fb9cbb1 YK |
37 | #include <net/netfilter/nf_conntrack.h> |
38 | #include <net/netfilter/nf_conntrack_l3proto.h> | |
605dcad6 | 39 | #include <net/netfilter/nf_conntrack_l4proto.h> |
77ab9cff | 40 | #include <net/netfilter/nf_conntrack_expect.h> |
9fb9cbb1 | 41 | #include <net/netfilter/nf_conntrack_helper.h> |
41d73ec0 | 42 | #include <net/netfilter/nf_conntrack_seqadj.h> |
9fb9cbb1 | 43 | #include <net/netfilter/nf_conntrack_core.h> |
ecfab2c9 | 44 | #include <net/netfilter/nf_conntrack_extend.h> |
58401572 | 45 | #include <net/netfilter/nf_conntrack_acct.h> |
a0891aa6 | 46 | #include <net/netfilter/nf_conntrack_ecache.h> |
5d0aa2cc | 47 | #include <net/netfilter/nf_conntrack_zones.h> |
a992ca2a | 48 | #include <net/netfilter/nf_conntrack_timestamp.h> |
dd705072 | 49 | #include <net/netfilter/nf_conntrack_timeout.h> |
c539f017 | 50 | #include <net/netfilter/nf_conntrack_labels.h> |
48b1de4c | 51 | #include <net/netfilter/nf_conntrack_synproxy.h> |
e6a7d3c0 | 52 | #include <net/netfilter/nf_nat.h> |
e17b666a | 53 | #include <net/netfilter/nf_nat_core.h> |
49376368 | 54 | #include <net/netfilter/nf_nat_helper.h> |
9fb9cbb1 | 55 | |
dc808fe2 | 56 | #define NF_CONNTRACK_VERSION "0.5.0" |
9fb9cbb1 | 57 | |
e17b666a PM |
58 | int (*nfnetlink_parse_nat_setup_hook)(struct nf_conn *ct, |
59 | enum nf_nat_manip_type manip, | |
39938324 | 60 | const struct nlattr *attr) __read_mostly; |
e6a7d3c0 PNA |
61 | EXPORT_SYMBOL_GPL(nfnetlink_parse_nat_setup_hook); |
62 | ||
f8ba1aff | 63 | DEFINE_SPINLOCK(nf_conntrack_lock); |
13b18339 | 64 | EXPORT_SYMBOL_GPL(nf_conntrack_lock); |
9fb9cbb1 | 65 | |
e2b7606c | 66 | unsigned int nf_conntrack_htable_size __read_mostly; |
13b18339 PM |
67 | EXPORT_SYMBOL_GPL(nf_conntrack_htable_size); |
68 | ||
e478075c | 69 | unsigned int nf_conntrack_max __read_mostly; |
a999e683 | 70 | EXPORT_SYMBOL_GPL(nf_conntrack_max); |
13b18339 | 71 | |
b3c5163f ED |
72 | DEFINE_PER_CPU(struct nf_conn, nf_conntrack_untracked); |
73 | EXPORT_PER_CPU_SYMBOL(nf_conntrack_untracked); | |
13b18339 | 74 | |
f682cefa | 75 | unsigned int nf_conntrack_hash_rnd __read_mostly; |
4d4e61c6 | 76 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_rnd); |
9fb9cbb1 | 77 | |
99f07e91 | 78 | static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple, u16 zone) |
9fb9cbb1 | 79 | { |
0794935e | 80 | unsigned int n; |
0794935e PM |
81 | |
82 | /* The direction must be ignored, so we hash everything up to the | |
83 | * destination ports (which is a multiple of 4) and treat the last | |
84 | * three bytes manually. | |
85 | */ | |
86 | n = (sizeof(tuple->src) + sizeof(tuple->dst.u3)) / sizeof(u32); | |
99f07e91 CG |
87 | return jhash2((u32 *)tuple, n, zone ^ nf_conntrack_hash_rnd ^ |
88 | (((__force __u16)tuple->dst.u.all << 16) | | |
89 | tuple->dst.protonum)); | |
90 | } | |
91 | ||
92 | static u32 __hash_bucket(u32 hash, unsigned int size) | |
93 | { | |
94 | return ((u64)hash * size) >> 32; | |
95 | } | |
96 | ||
97 | static u32 hash_bucket(u32 hash, const struct net *net) | |
98 | { | |
99 | return __hash_bucket(hash, net->ct.htable_size); | |
100 | } | |
0794935e | 101 | |
99f07e91 CG |
102 | static u_int32_t __hash_conntrack(const struct nf_conntrack_tuple *tuple, |
103 | u16 zone, unsigned int size) | |
104 | { | |
105 | return __hash_bucket(hash_conntrack_raw(tuple, zone), size); | |
9fb9cbb1 YK |
106 | } |
107 | ||
5d0aa2cc | 108 | static inline u_int32_t hash_conntrack(const struct net *net, u16 zone, |
d696c7bd | 109 | const struct nf_conntrack_tuple *tuple) |
9fb9cbb1 | 110 | { |
99f07e91 | 111 | return __hash_conntrack(tuple, zone, net->ct.htable_size); |
9fb9cbb1 YK |
112 | } |
113 | ||
5f2b4c90 | 114 | bool |
9fb9cbb1 YK |
115 | nf_ct_get_tuple(const struct sk_buff *skb, |
116 | unsigned int nhoff, | |
117 | unsigned int dataoff, | |
118 | u_int16_t l3num, | |
119 | u_int8_t protonum, | |
120 | struct nf_conntrack_tuple *tuple, | |
121 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 122 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 123 | { |
443a70d5 | 124 | memset(tuple, 0, sizeof(*tuple)); |
9fb9cbb1 YK |
125 | |
126 | tuple->src.l3num = l3num; | |
127 | if (l3proto->pkt_to_tuple(skb, nhoff, tuple) == 0) | |
5f2b4c90 | 128 | return false; |
9fb9cbb1 YK |
129 | |
130 | tuple->dst.protonum = protonum; | |
131 | tuple->dst.dir = IP_CT_DIR_ORIGINAL; | |
132 | ||
605dcad6 | 133 | return l4proto->pkt_to_tuple(skb, dataoff, tuple); |
9fb9cbb1 | 134 | } |
13b18339 | 135 | EXPORT_SYMBOL_GPL(nf_ct_get_tuple); |
9fb9cbb1 | 136 | |
5f2b4c90 JE |
137 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
138 | u_int16_t l3num, struct nf_conntrack_tuple *tuple) | |
e2a3123f YK |
139 | { |
140 | struct nf_conntrack_l3proto *l3proto; | |
141 | struct nf_conntrack_l4proto *l4proto; | |
142 | unsigned int protoff; | |
143 | u_int8_t protonum; | |
144 | int ret; | |
145 | ||
146 | rcu_read_lock(); | |
147 | ||
148 | l3proto = __nf_ct_l3proto_find(l3num); | |
149 | ret = l3proto->get_l4proto(skb, nhoff, &protoff, &protonum); | |
150 | if (ret != NF_ACCEPT) { | |
151 | rcu_read_unlock(); | |
5f2b4c90 | 152 | return false; |
e2a3123f YK |
153 | } |
154 | ||
155 | l4proto = __nf_ct_l4proto_find(l3num, protonum); | |
156 | ||
157 | ret = nf_ct_get_tuple(skb, nhoff, protoff, l3num, protonum, tuple, | |
158 | l3proto, l4proto); | |
159 | ||
160 | rcu_read_unlock(); | |
161 | return ret; | |
162 | } | |
163 | EXPORT_SYMBOL_GPL(nf_ct_get_tuplepr); | |
164 | ||
5f2b4c90 | 165 | bool |
9fb9cbb1 YK |
166 | nf_ct_invert_tuple(struct nf_conntrack_tuple *inverse, |
167 | const struct nf_conntrack_tuple *orig, | |
168 | const struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 169 | const struct nf_conntrack_l4proto *l4proto) |
9fb9cbb1 | 170 | { |
443a70d5 | 171 | memset(inverse, 0, sizeof(*inverse)); |
9fb9cbb1 YK |
172 | |
173 | inverse->src.l3num = orig->src.l3num; | |
174 | if (l3proto->invert_tuple(inverse, orig) == 0) | |
5f2b4c90 | 175 | return false; |
9fb9cbb1 YK |
176 | |
177 | inverse->dst.dir = !orig->dst.dir; | |
178 | ||
179 | inverse->dst.protonum = orig->dst.protonum; | |
605dcad6 | 180 | return l4proto->invert_tuple(inverse, orig); |
9fb9cbb1 | 181 | } |
13b18339 | 182 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuple); |
9fb9cbb1 | 183 | |
9fb9cbb1 YK |
184 | static void |
185 | clean_from_lists(struct nf_conn *ct) | |
186 | { | |
0d53778e | 187 | pr_debug("clean_from_lists(%p)\n", ct); |
ea781f19 ED |
188 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); |
189 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode); | |
9fb9cbb1 YK |
190 | |
191 | /* Destroy all pending expectations */ | |
c1d10adb | 192 | nf_ct_remove_expectations(ct); |
9fb9cbb1 YK |
193 | } |
194 | ||
b7779d06 JDB |
195 | /* must be called with local_bh_disable */ |
196 | static void nf_ct_add_to_dying_list(struct nf_conn *ct) | |
197 | { | |
198 | struct ct_pcpu *pcpu; | |
199 | ||
200 | /* add this conntrack to the (per cpu) dying list */ | |
201 | ct->cpu = smp_processor_id(); | |
202 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
203 | ||
204 | spin_lock(&pcpu->lock); | |
205 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
206 | &pcpu->dying); | |
207 | spin_unlock(&pcpu->lock); | |
208 | } | |
209 | ||
210 | /* must be called with local_bh_disable */ | |
211 | static void nf_ct_add_to_unconfirmed_list(struct nf_conn *ct) | |
212 | { | |
213 | struct ct_pcpu *pcpu; | |
214 | ||
215 | /* add this conntrack to the (per cpu) unconfirmed list */ | |
216 | ct->cpu = smp_processor_id(); | |
217 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
218 | ||
219 | spin_lock(&pcpu->lock); | |
220 | hlist_nulls_add_head(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
221 | &pcpu->unconfirmed); | |
222 | spin_unlock(&pcpu->lock); | |
223 | } | |
224 | ||
225 | /* must be called with local_bh_disable */ | |
226 | static void nf_ct_del_from_dying_or_unconfirmed_list(struct nf_conn *ct) | |
227 | { | |
228 | struct ct_pcpu *pcpu; | |
229 | ||
230 | /* We overload first tuple to link into unconfirmed or dying list.*/ | |
231 | pcpu = per_cpu_ptr(nf_ct_net(ct)->ct.pcpu_lists, ct->cpu); | |
232 | ||
233 | spin_lock(&pcpu->lock); | |
234 | BUG_ON(hlist_nulls_unhashed(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode)); | |
235 | hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); | |
236 | spin_unlock(&pcpu->lock); | |
237 | } | |
238 | ||
9fb9cbb1 YK |
239 | static void |
240 | destroy_conntrack(struct nf_conntrack *nfct) | |
241 | { | |
242 | struct nf_conn *ct = (struct nf_conn *)nfct; | |
0d55af87 | 243 | struct net *net = nf_ct_net(ct); |
605dcad6 | 244 | struct nf_conntrack_l4proto *l4proto; |
9fb9cbb1 | 245 | |
0d53778e | 246 | pr_debug("destroy_conntrack(%p)\n", ct); |
9fb9cbb1 YK |
247 | NF_CT_ASSERT(atomic_read(&nfct->use) == 0); |
248 | NF_CT_ASSERT(!timer_pending(&ct->timeout)); | |
249 | ||
9fb9cbb1 YK |
250 | /* To make sure we don't get any weird locking issues here: |
251 | * destroy_conntrack() MUST NOT be called with a write lock | |
252 | * to nf_conntrack_lock!!! -HW */ | |
923f4902 | 253 | rcu_read_lock(); |
5e8fbe2a | 254 | l4proto = __nf_ct_l4proto_find(nf_ct_l3num(ct), nf_ct_protonum(ct)); |
605dcad6 MJ |
255 | if (l4proto && l4proto->destroy) |
256 | l4proto->destroy(ct); | |
9fb9cbb1 | 257 | |
982d9a9c | 258 | rcu_read_unlock(); |
9fb9cbb1 | 259 | |
f8ba1aff | 260 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
261 | /* Expectations will have been removed in clean_from_lists, |
262 | * except TFTP can create an expectation on the first packet, | |
263 | * before connection is in the list, so we need to clean here, | |
264 | * too. */ | |
c1d10adb | 265 | nf_ct_remove_expectations(ct); |
9fb9cbb1 | 266 | |
b7779d06 | 267 | nf_ct_del_from_dying_or_unconfirmed_list(ct); |
9fb9cbb1 | 268 | |
0d55af87 | 269 | NF_CT_STAT_INC(net, delete); |
f8ba1aff | 270 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
271 | |
272 | if (ct->master) | |
273 | nf_ct_put(ct->master); | |
274 | ||
0d53778e | 275 | pr_debug("destroy_conntrack: returning ct=%p to slab\n", ct); |
9fb9cbb1 YK |
276 | nf_conntrack_free(ct); |
277 | } | |
278 | ||
02982c27 | 279 | static void nf_ct_delete_from_lists(struct nf_conn *ct) |
9fb9cbb1 | 280 | { |
0d55af87 | 281 | struct net *net = nf_ct_net(ct); |
9fb9cbb1 | 282 | |
9858a3ae | 283 | nf_ct_helper_destroy(ct); |
f8ba1aff | 284 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
285 | /* Inside lock so preempt is disabled on module removal path. |
286 | * Otherwise we can get spurious warnings. */ | |
0d55af87 | 287 | NF_CT_STAT_INC(net, delete_list); |
9fb9cbb1 | 288 | clean_from_lists(ct); |
b7779d06 | 289 | nf_ct_add_to_dying_list(ct); |
f8ba1aff | 290 | spin_unlock_bh(&nf_conntrack_lock); |
dd7669a9 | 291 | } |
dd7669a9 PNA |
292 | |
293 | static void death_by_event(unsigned long ul_conntrack) | |
294 | { | |
295 | struct nf_conn *ct = (void *)ul_conntrack; | |
296 | struct net *net = nf_ct_net(ct); | |
5b423f6a PNA |
297 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); |
298 | ||
299 | BUG_ON(ecache == NULL); | |
dd7669a9 PNA |
300 | |
301 | if (nf_conntrack_event(IPCT_DESTROY, ct) < 0) { | |
302 | /* bad luck, let's retry again */ | |
5b423f6a | 303 | ecache->timeout.expires = jiffies + |
ca3d41a5 | 304 | (prandom_u32() % net->ct.sysctl_events_retry_timeout); |
5b423f6a | 305 | add_timer(&ecache->timeout); |
dd7669a9 PNA |
306 | return; |
307 | } | |
308 | /* we've got the event delivered, now it's dying */ | |
309 | set_bit(IPS_DYING_BIT, &ct->status); | |
dd7669a9 PNA |
310 | nf_ct_put(ct); |
311 | } | |
312 | ||
02982c27 | 313 | static void nf_ct_dying_timeout(struct nf_conn *ct) |
dd7669a9 PNA |
314 | { |
315 | struct net *net = nf_ct_net(ct); | |
5b423f6a PNA |
316 | struct nf_conntrack_ecache *ecache = nf_ct_ecache_find(ct); |
317 | ||
318 | BUG_ON(ecache == NULL); | |
dd7669a9 | 319 | |
dd7669a9 | 320 | /* set a new timer to retry event delivery */ |
5b423f6a PNA |
321 | setup_timer(&ecache->timeout, death_by_event, (unsigned long)ct); |
322 | ecache->timeout.expires = jiffies + | |
ca3d41a5 | 323 | (prandom_u32() % net->ct.sysctl_events_retry_timeout); |
5b423f6a | 324 | add_timer(&ecache->timeout); |
dd7669a9 | 325 | } |
dd7669a9 | 326 | |
02982c27 | 327 | bool nf_ct_delete(struct nf_conn *ct, u32 portid, int report) |
dd7669a9 | 328 | { |
a992ca2a PNA |
329 | struct nf_conn_tstamp *tstamp; |
330 | ||
331 | tstamp = nf_conn_tstamp_find(ct); | |
332 | if (tstamp && tstamp->stop == 0) | |
333 | tstamp->stop = ktime_to_ns(ktime_get_real()); | |
dd7669a9 | 334 | |
02982c27 FW |
335 | if (!nf_ct_is_dying(ct) && |
336 | unlikely(nf_conntrack_event_report(IPCT_DESTROY, ct, | |
337 | portid, report) < 0)) { | |
dd7669a9 PNA |
338 | /* destroy event was not delivered */ |
339 | nf_ct_delete_from_lists(ct); | |
04dac011 | 340 | nf_ct_dying_timeout(ct); |
02982c27 | 341 | return false; |
dd7669a9 PNA |
342 | } |
343 | set_bit(IPS_DYING_BIT, &ct->status); | |
344 | nf_ct_delete_from_lists(ct); | |
9fb9cbb1 | 345 | nf_ct_put(ct); |
02982c27 FW |
346 | return true; |
347 | } | |
348 | EXPORT_SYMBOL_GPL(nf_ct_delete); | |
349 | ||
350 | static void death_by_timeout(unsigned long ul_conntrack) | |
351 | { | |
352 | nf_ct_delete((struct nf_conn *)ul_conntrack, 0, 0); | |
9fb9cbb1 YK |
353 | } |
354 | ||
c6825c09 AV |
355 | static inline bool |
356 | nf_ct_key_equal(struct nf_conntrack_tuple_hash *h, | |
357 | const struct nf_conntrack_tuple *tuple, | |
358 | u16 zone) | |
359 | { | |
360 | struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h); | |
361 | ||
362 | /* A conntrack can be recreated with the equal tuple, | |
363 | * so we need to check that the conntrack is confirmed | |
364 | */ | |
365 | return nf_ct_tuple_equal(tuple, &h->tuple) && | |
366 | nf_ct_zone(ct) == zone && | |
367 | nf_ct_is_confirmed(ct); | |
368 | } | |
369 | ||
ea781f19 ED |
370 | /* |
371 | * Warning : | |
372 | * - Caller must take a reference on returned object | |
373 | * and recheck nf_ct_tuple_equal(tuple, &h->tuple) | |
374 | * OR | |
375 | * - Caller must lock nf_conntrack_lock before calling this function | |
376 | */ | |
99f07e91 CG |
377 | static struct nf_conntrack_tuple_hash * |
378 | ____nf_conntrack_find(struct net *net, u16 zone, | |
379 | const struct nf_conntrack_tuple *tuple, u32 hash) | |
9fb9cbb1 YK |
380 | { |
381 | struct nf_conntrack_tuple_hash *h; | |
ea781f19 | 382 | struct hlist_nulls_node *n; |
99f07e91 | 383 | unsigned int bucket = hash_bucket(hash, net); |
9fb9cbb1 | 384 | |
4e29e9ec PM |
385 | /* Disable BHs the entire time since we normally need to disable them |
386 | * at least once for the stats anyway. | |
387 | */ | |
388 | local_bh_disable(); | |
ea781f19 | 389 | begin: |
99f07e91 | 390 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[bucket], hnnode) { |
c6825c09 | 391 | if (nf_ct_key_equal(h, tuple, zone)) { |
0d55af87 | 392 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 393 | local_bh_enable(); |
9fb9cbb1 YK |
394 | return h; |
395 | } | |
0d55af87 | 396 | NF_CT_STAT_INC(net, searched); |
9fb9cbb1 | 397 | } |
ea781f19 ED |
398 | /* |
399 | * if the nulls value we got at the end of this lookup is | |
400 | * not the expected one, we must restart lookup. | |
401 | * We probably met an item that was moved to another chain. | |
402 | */ | |
99f07e91 | 403 | if (get_nulls_value(n) != bucket) { |
af740b2c | 404 | NF_CT_STAT_INC(net, search_restart); |
ea781f19 | 405 | goto begin; |
af740b2c | 406 | } |
4e29e9ec | 407 | local_bh_enable(); |
9fb9cbb1 YK |
408 | |
409 | return NULL; | |
410 | } | |
99f07e91 | 411 | |
9fb9cbb1 | 412 | /* Find a connection corresponding to a tuple. */ |
99f07e91 CG |
413 | static struct nf_conntrack_tuple_hash * |
414 | __nf_conntrack_find_get(struct net *net, u16 zone, | |
415 | const struct nf_conntrack_tuple *tuple, u32 hash) | |
9fb9cbb1 YK |
416 | { |
417 | struct nf_conntrack_tuple_hash *h; | |
76507f69 | 418 | struct nf_conn *ct; |
9fb9cbb1 | 419 | |
76507f69 | 420 | rcu_read_lock(); |
ea781f19 | 421 | begin: |
99f07e91 | 422 | h = ____nf_conntrack_find(net, zone, tuple, hash); |
76507f69 PM |
423 | if (h) { |
424 | ct = nf_ct_tuplehash_to_ctrack(h); | |
8d8890b7 PM |
425 | if (unlikely(nf_ct_is_dying(ct) || |
426 | !atomic_inc_not_zero(&ct->ct_general.use))) | |
76507f69 | 427 | h = NULL; |
ea781f19 | 428 | else { |
c6825c09 | 429 | if (unlikely(!nf_ct_key_equal(h, tuple, zone))) { |
ea781f19 ED |
430 | nf_ct_put(ct); |
431 | goto begin; | |
432 | } | |
433 | } | |
76507f69 PM |
434 | } |
435 | rcu_read_unlock(); | |
9fb9cbb1 YK |
436 | |
437 | return h; | |
438 | } | |
99f07e91 CG |
439 | |
440 | struct nf_conntrack_tuple_hash * | |
441 | nf_conntrack_find_get(struct net *net, u16 zone, | |
442 | const struct nf_conntrack_tuple *tuple) | |
443 | { | |
444 | return __nf_conntrack_find_get(net, zone, tuple, | |
445 | hash_conntrack_raw(tuple, zone)); | |
446 | } | |
13b18339 | 447 | EXPORT_SYMBOL_GPL(nf_conntrack_find_get); |
9fb9cbb1 | 448 | |
c1d10adb PNA |
449 | static void __nf_conntrack_hash_insert(struct nf_conn *ct, |
450 | unsigned int hash, | |
b476b72a | 451 | unsigned int reply_hash) |
c1d10adb | 452 | { |
400dad39 AD |
453 | struct net *net = nf_ct_net(ct); |
454 | ||
ea781f19 | 455 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, |
400dad39 | 456 | &net->ct.hash[hash]); |
ea781f19 | 457 | hlist_nulls_add_head_rcu(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode, |
b476b72a | 458 | &net->ct.hash[reply_hash]); |
c1d10adb PNA |
459 | } |
460 | ||
7d367e06 JK |
461 | int |
462 | nf_conntrack_hash_check_insert(struct nf_conn *ct) | |
c1d10adb | 463 | { |
d696c7bd | 464 | struct net *net = nf_ct_net(ct); |
b476b72a | 465 | unsigned int hash, reply_hash; |
7d367e06 JK |
466 | struct nf_conntrack_tuple_hash *h; |
467 | struct hlist_nulls_node *n; | |
5d0aa2cc | 468 | u16 zone; |
c1d10adb | 469 | |
5d0aa2cc | 470 | zone = nf_ct_zone(ct); |
7d367e06 JK |
471 | hash = hash_conntrack(net, zone, |
472 | &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple); | |
b476b72a | 473 | reply_hash = hash_conntrack(net, zone, |
7d367e06 JK |
474 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
475 | ||
476 | spin_lock_bh(&nf_conntrack_lock); | |
477 | ||
478 | /* See if there's one in the list already, including reverse */ | |
479 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) | |
480 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, | |
481 | &h->tuple) && | |
482 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
483 | goto out; | |
b476b72a | 484 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[reply_hash], hnnode) |
7d367e06 JK |
485 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
486 | &h->tuple) && | |
487 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
488 | goto out; | |
c1d10adb | 489 | |
7d367e06 | 490 | add_timer(&ct->timeout); |
e53376be PNA |
491 | smp_wmb(); |
492 | /* The caller holds a reference to this object */ | |
493 | atomic_set(&ct->ct_general.use, 2); | |
b476b72a | 494 | __nf_conntrack_hash_insert(ct, hash, reply_hash); |
7d367e06 JK |
495 | NF_CT_STAT_INC(net, insert); |
496 | spin_unlock_bh(&nf_conntrack_lock); | |
497 | ||
498 | return 0; | |
499 | ||
500 | out: | |
501 | NF_CT_STAT_INC(net, insert_failed); | |
502 | spin_unlock_bh(&nf_conntrack_lock); | |
503 | return -EEXIST; | |
c1d10adb | 504 | } |
7d367e06 | 505 | EXPORT_SYMBOL_GPL(nf_conntrack_hash_check_insert); |
c1d10adb | 506 | |
e53376be PNA |
507 | /* deletion from this larval template list happens via nf_ct_put() */ |
508 | void nf_conntrack_tmpl_insert(struct net *net, struct nf_conn *tmpl) | |
509 | { | |
b7779d06 JDB |
510 | struct ct_pcpu *pcpu; |
511 | ||
e53376be PNA |
512 | __set_bit(IPS_TEMPLATE_BIT, &tmpl->status); |
513 | __set_bit(IPS_CONFIRMED_BIT, &tmpl->status); | |
514 | nf_conntrack_get(&tmpl->ct_general); | |
515 | ||
b7779d06 JDB |
516 | /* add this conntrack to the (per cpu) tmpl list */ |
517 | local_bh_disable(); | |
518 | tmpl->cpu = smp_processor_id(); | |
519 | pcpu = per_cpu_ptr(nf_ct_net(tmpl)->ct.pcpu_lists, tmpl->cpu); | |
520 | ||
521 | spin_lock(&pcpu->lock); | |
e53376be PNA |
522 | /* Overload tuple linked list to put us in template list. */ |
523 | hlist_nulls_add_head_rcu(&tmpl->tuplehash[IP_CT_DIR_ORIGINAL].hnnode, | |
b7779d06 JDB |
524 | &pcpu->tmpl); |
525 | spin_unlock_bh(&pcpu->lock); | |
e53376be PNA |
526 | } |
527 | EXPORT_SYMBOL_GPL(nf_conntrack_tmpl_insert); | |
528 | ||
9fb9cbb1 YK |
529 | /* Confirm a connection given skb; places it in hash table */ |
530 | int | |
3db05fea | 531 | __nf_conntrack_confirm(struct sk_buff *skb) |
9fb9cbb1 | 532 | { |
b476b72a | 533 | unsigned int hash, reply_hash; |
df0933dc | 534 | struct nf_conntrack_tuple_hash *h; |
9fb9cbb1 | 535 | struct nf_conn *ct; |
df0933dc | 536 | struct nf_conn_help *help; |
a992ca2a | 537 | struct nf_conn_tstamp *tstamp; |
ea781f19 | 538 | struct hlist_nulls_node *n; |
9fb9cbb1 | 539 | enum ip_conntrack_info ctinfo; |
400dad39 | 540 | struct net *net; |
5d0aa2cc | 541 | u16 zone; |
9fb9cbb1 | 542 | |
3db05fea | 543 | ct = nf_ct_get(skb, &ctinfo); |
400dad39 | 544 | net = nf_ct_net(ct); |
9fb9cbb1 YK |
545 | |
546 | /* ipt_REJECT uses nf_conntrack_attach to attach related | |
547 | ICMP/TCP RST packets in other direction. Actual packet | |
548 | which created connection will be IP_CT_NEW or for an | |
549 | expected connection, IP_CT_RELATED. */ | |
550 | if (CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) | |
551 | return NF_ACCEPT; | |
552 | ||
5d0aa2cc | 553 | zone = nf_ct_zone(ct); |
99f07e91 CG |
554 | /* reuse the hash saved before */ |
555 | hash = *(unsigned long *)&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev; | |
556 | hash = hash_bucket(hash, net); | |
b476b72a | 557 | reply_hash = hash_conntrack(net, zone, |
99f07e91 | 558 | &ct->tuplehash[IP_CT_DIR_REPLY].tuple); |
9fb9cbb1 YK |
559 | |
560 | /* We're not in hash table, and we refuse to set up related | |
561 | connections for unconfirmed conns. But packet copies and | |
562 | REJECT will give spurious warnings here. */ | |
563 | /* NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 1); */ | |
564 | ||
25985edc | 565 | /* No external references means no one else could have |
9fb9cbb1 YK |
566 | confirmed us. */ |
567 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
0d53778e | 568 | pr_debug("Confirming conntrack %p\n", ct); |
9fb9cbb1 | 569 | |
f8ba1aff | 570 | spin_lock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 571 | |
fc350777 JM |
572 | /* We have to check the DYING flag inside the lock to prevent |
573 | a race against nf_ct_get_next_corpse() possibly called from | |
574 | user context, else we insert an already 'dead' hash, blocking | |
575 | further use of that particular connection -JM */ | |
576 | ||
577 | if (unlikely(nf_ct_is_dying(ct))) { | |
578 | spin_unlock_bh(&nf_conntrack_lock); | |
579 | return NF_ACCEPT; | |
580 | } | |
581 | ||
9fb9cbb1 YK |
582 | /* See if there's one in the list already, including reverse: |
583 | NAT could have grabbed it without realizing, since we're | |
584 | not in the hash. If there is, we lost race. */ | |
ea781f19 | 585 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[hash], hnnode) |
df0933dc | 586 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple, |
5d0aa2cc PM |
587 | &h->tuple) && |
588 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
df0933dc | 589 | goto out; |
b476b72a | 590 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[reply_hash], hnnode) |
df0933dc | 591 | if (nf_ct_tuple_equal(&ct->tuplehash[IP_CT_DIR_REPLY].tuple, |
5d0aa2cc PM |
592 | &h->tuple) && |
593 | zone == nf_ct_zone(nf_ct_tuplehash_to_ctrack(h))) | |
df0933dc | 594 | goto out; |
9fb9cbb1 | 595 | |
b7779d06 | 596 | nf_ct_del_from_dying_or_unconfirmed_list(ct); |
df0933dc | 597 | |
df0933dc PM |
598 | /* Timer relative to confirmation time, not original |
599 | setting time, otherwise we'd get timer wrap in | |
600 | weird delay cases. */ | |
601 | ct->timeout.expires += jiffies; | |
602 | add_timer(&ct->timeout); | |
603 | atomic_inc(&ct->ct_general.use); | |
45eec341 | 604 | ct->status |= IPS_CONFIRMED; |
5c8ec910 | 605 | |
a992ca2a PNA |
606 | /* set conntrack timestamp, if enabled. */ |
607 | tstamp = nf_conn_tstamp_find(ct); | |
608 | if (tstamp) { | |
609 | if (skb->tstamp.tv64 == 0) | |
e3192690 | 610 | __net_timestamp(skb); |
a992ca2a PNA |
611 | |
612 | tstamp->start = ktime_to_ns(skb->tstamp); | |
613 | } | |
5c8ec910 PM |
614 | /* Since the lookup is lockless, hash insertion must be done after |
615 | * starting the timer and setting the CONFIRMED bit. The RCU barriers | |
616 | * guarantee that no other CPU can find the conntrack before the above | |
617 | * stores are visible. | |
618 | */ | |
b476b72a | 619 | __nf_conntrack_hash_insert(ct, hash, reply_hash); |
0d55af87 | 620 | NF_CT_STAT_INC(net, insert); |
f8ba1aff | 621 | spin_unlock_bh(&nf_conntrack_lock); |
5c8ec910 | 622 | |
df0933dc PM |
623 | help = nfct_help(ct); |
624 | if (help && help->helper) | |
a71996fc | 625 | nf_conntrack_event_cache(IPCT_HELPER, ct); |
17e6e4ea | 626 | |
df0933dc | 627 | nf_conntrack_event_cache(master_ct(ct) ? |
a71996fc | 628 | IPCT_RELATED : IPCT_NEW, ct); |
df0933dc | 629 | return NF_ACCEPT; |
9fb9cbb1 | 630 | |
df0933dc | 631 | out: |
0d55af87 | 632 | NF_CT_STAT_INC(net, insert_failed); |
f8ba1aff | 633 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
634 | return NF_DROP; |
635 | } | |
13b18339 | 636 | EXPORT_SYMBOL_GPL(__nf_conntrack_confirm); |
9fb9cbb1 YK |
637 | |
638 | /* Returns true if a connection correspondings to the tuple (required | |
639 | for NAT). */ | |
640 | int | |
641 | nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, | |
642 | const struct nf_conn *ignored_conntrack) | |
643 | { | |
400dad39 | 644 | struct net *net = nf_ct_net(ignored_conntrack); |
9fb9cbb1 | 645 | struct nf_conntrack_tuple_hash *h; |
ea781f19 | 646 | struct hlist_nulls_node *n; |
5d0aa2cc PM |
647 | struct nf_conn *ct; |
648 | u16 zone = nf_ct_zone(ignored_conntrack); | |
649 | unsigned int hash = hash_conntrack(net, zone, tuple); | |
9fb9cbb1 | 650 | |
4e29e9ec PM |
651 | /* Disable BHs the entire time since we need to disable them at |
652 | * least once for the stats anyway. | |
653 | */ | |
654 | rcu_read_lock_bh(); | |
ea781f19 | 655 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], hnnode) { |
5d0aa2cc PM |
656 | ct = nf_ct_tuplehash_to_ctrack(h); |
657 | if (ct != ignored_conntrack && | |
658 | nf_ct_tuple_equal(tuple, &h->tuple) && | |
659 | nf_ct_zone(ct) == zone) { | |
0d55af87 | 660 | NF_CT_STAT_INC(net, found); |
4e29e9ec | 661 | rcu_read_unlock_bh(); |
ba419aff PM |
662 | return 1; |
663 | } | |
0d55af87 | 664 | NF_CT_STAT_INC(net, searched); |
ba419aff | 665 | } |
4e29e9ec | 666 | rcu_read_unlock_bh(); |
9fb9cbb1 | 667 | |
ba419aff | 668 | return 0; |
9fb9cbb1 | 669 | } |
13b18339 | 670 | EXPORT_SYMBOL_GPL(nf_conntrack_tuple_taken); |
9fb9cbb1 | 671 | |
7ae7730f PM |
672 | #define NF_CT_EVICTION_RANGE 8 |
673 | ||
9fb9cbb1 YK |
674 | /* There's a small race here where we may free a just-assured |
675 | connection. Too bad: we're in trouble anyway. */ | |
400dad39 | 676 | static noinline int early_drop(struct net *net, unsigned int hash) |
9fb9cbb1 | 677 | { |
f205c5e0 | 678 | /* Use oldest entry, which is roughly LRU */ |
9fb9cbb1 | 679 | struct nf_conntrack_tuple_hash *h; |
df0933dc | 680 | struct nf_conn *ct = NULL, *tmp; |
ea781f19 | 681 | struct hlist_nulls_node *n; |
7ae7730f | 682 | unsigned int i, cnt = 0; |
9fb9cbb1 YK |
683 | int dropped = 0; |
684 | ||
76507f69 | 685 | rcu_read_lock(); |
d696c7bd | 686 | for (i = 0; i < net->ct.htable_size; i++) { |
ea781f19 ED |
687 | hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash], |
688 | hnnode) { | |
7ae7730f PM |
689 | tmp = nf_ct_tuplehash_to_ctrack(h); |
690 | if (!test_bit(IPS_ASSURED_BIT, &tmp->status)) | |
691 | ct = tmp; | |
692 | cnt++; | |
693 | } | |
76507f69 | 694 | |
5ae27aa2 CG |
695 | if (ct != NULL) { |
696 | if (likely(!nf_ct_is_dying(ct) && | |
697 | atomic_inc_not_zero(&ct->ct_general.use))) | |
698 | break; | |
699 | else | |
700 | ct = NULL; | |
701 | } | |
702 | ||
703 | if (cnt >= NF_CT_EVICTION_RANGE) | |
7ae7730f | 704 | break; |
5ae27aa2 | 705 | |
d696c7bd | 706 | hash = (hash + 1) % net->ct.htable_size; |
9fb9cbb1 | 707 | } |
76507f69 | 708 | rcu_read_unlock(); |
9fb9cbb1 YK |
709 | |
710 | if (!ct) | |
711 | return dropped; | |
712 | ||
713 | if (del_timer(&ct->timeout)) { | |
02982c27 | 714 | if (nf_ct_delete(ct, 0, 0)) { |
74138511 PNA |
715 | dropped = 1; |
716 | NF_CT_STAT_INC_ATOMIC(net, early_drop); | |
717 | } | |
9fb9cbb1 YK |
718 | } |
719 | nf_ct_put(ct); | |
720 | return dropped; | |
721 | } | |
722 | ||
f682cefa CG |
723 | void init_nf_conntrack_hash_rnd(void) |
724 | { | |
725 | unsigned int rand; | |
726 | ||
727 | /* | |
728 | * Why not initialize nf_conntrack_rnd in a "init()" function ? | |
729 | * Because there isn't enough entropy when system initializing, | |
730 | * and we initialize it as late as possible. | |
731 | */ | |
732 | do { | |
733 | get_random_bytes(&rand, sizeof(rand)); | |
734 | } while (!rand); | |
735 | cmpxchg(&nf_conntrack_hash_rnd, 0, rand); | |
736 | } | |
737 | ||
99f07e91 CG |
738 | static struct nf_conn * |
739 | __nf_conntrack_alloc(struct net *net, u16 zone, | |
740 | const struct nf_conntrack_tuple *orig, | |
741 | const struct nf_conntrack_tuple *repl, | |
742 | gfp_t gfp, u32 hash) | |
9fb9cbb1 | 743 | { |
cd7fcbf1 | 744 | struct nf_conn *ct; |
9fb9cbb1 | 745 | |
b2390969 | 746 | if (unlikely(!nf_conntrack_hash_rnd)) { |
f682cefa | 747 | init_nf_conntrack_hash_rnd(); |
99f07e91 CG |
748 | /* recompute the hash as nf_conntrack_hash_rnd is initialized */ |
749 | hash = hash_conntrack_raw(orig, zone); | |
9fb9cbb1 YK |
750 | } |
751 | ||
5251e2d2 | 752 | /* We don't want any race condition at early drop stage */ |
49ac8713 | 753 | atomic_inc(&net->ct.count); |
5251e2d2 | 754 | |
76eb9460 | 755 | if (nf_conntrack_max && |
49ac8713 | 756 | unlikely(atomic_read(&net->ct.count) > nf_conntrack_max)) { |
99f07e91 | 757 | if (!early_drop(net, hash_bucket(hash, net))) { |
49ac8713 | 758 | atomic_dec(&net->ct.count); |
e87cc472 | 759 | net_warn_ratelimited("nf_conntrack: table full, dropping packet\n"); |
9fb9cbb1 YK |
760 | return ERR_PTR(-ENOMEM); |
761 | } | |
762 | } | |
763 | ||
941297f4 ED |
764 | /* |
765 | * Do not use kmem_cache_zalloc(), as this cache uses | |
766 | * SLAB_DESTROY_BY_RCU. | |
767 | */ | |
5b3501fa | 768 | ct = kmem_cache_alloc(net->ct.nf_conntrack_cachep, gfp); |
c88130bc | 769 | if (ct == NULL) { |
49ac8713 | 770 | atomic_dec(&net->ct.count); |
dacd2a1a | 771 | return ERR_PTR(-ENOMEM); |
9fb9cbb1 | 772 | } |
941297f4 ED |
773 | /* |
774 | * Let ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.next | |
775 | * and ct->tuplehash[IP_CT_DIR_REPLY].hnnode.next unchanged. | |
776 | */ | |
777 | memset(&ct->tuplehash[IP_CT_DIR_MAX], 0, | |
e5fc9e7a CG |
778 | offsetof(struct nf_conn, proto) - |
779 | offsetof(struct nf_conn, tuplehash[IP_CT_DIR_MAX])); | |
440f0d58 | 780 | spin_lock_init(&ct->lock); |
c88130bc | 781 | ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple = *orig; |
941297f4 | 782 | ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode.pprev = NULL; |
c88130bc | 783 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *repl; |
99f07e91 CG |
784 | /* save hash for reusing when confirming */ |
785 | *(unsigned long *)(&ct->tuplehash[IP_CT_DIR_REPLY].hnnode.pprev) = hash; | |
9fb9cbb1 | 786 | /* Don't set timer yet: wait for confirmation */ |
c88130bc | 787 | setup_timer(&ct->timeout, death_by_timeout, (unsigned long)ct); |
c2d9ba9b | 788 | write_pnet(&ct->ct_net, net); |
5d0aa2cc PM |
789 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
790 | if (zone) { | |
791 | struct nf_conntrack_zone *nf_ct_zone; | |
792 | ||
793 | nf_ct_zone = nf_ct_ext_add(ct, NF_CT_EXT_ZONE, GFP_ATOMIC); | |
794 | if (!nf_ct_zone) | |
795 | goto out_free; | |
796 | nf_ct_zone->id = zone; | |
797 | } | |
798 | #endif | |
e53376be PNA |
799 | /* Because we use RCU lookups, we set ct_general.use to zero before |
800 | * this is inserted in any list. | |
941297f4 | 801 | */ |
e53376be | 802 | atomic_set(&ct->ct_general.use, 0); |
c88130bc | 803 | return ct; |
5d0aa2cc PM |
804 | |
805 | #ifdef CONFIG_NF_CONNTRACK_ZONES | |
806 | out_free: | |
d96fc659 | 807 | atomic_dec(&net->ct.count); |
5d0aa2cc PM |
808 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
809 | return ERR_PTR(-ENOMEM); | |
810 | #endif | |
9fb9cbb1 | 811 | } |
99f07e91 CG |
812 | |
813 | struct nf_conn *nf_conntrack_alloc(struct net *net, u16 zone, | |
814 | const struct nf_conntrack_tuple *orig, | |
815 | const struct nf_conntrack_tuple *repl, | |
816 | gfp_t gfp) | |
817 | { | |
818 | return __nf_conntrack_alloc(net, zone, orig, repl, gfp, 0); | |
819 | } | |
13b18339 | 820 | EXPORT_SYMBOL_GPL(nf_conntrack_alloc); |
9fb9cbb1 | 821 | |
c88130bc | 822 | void nf_conntrack_free(struct nf_conn *ct) |
76507f69 | 823 | { |
1d45209d ED |
824 | struct net *net = nf_ct_net(ct); |
825 | ||
e53376be PNA |
826 | /* A freed object has refcnt == 0, that's |
827 | * the golden rule for SLAB_DESTROY_BY_RCU | |
828 | */ | |
829 | NF_CT_ASSERT(atomic_read(&ct->ct_general.use) == 0); | |
830 | ||
ceeff754 | 831 | nf_ct_ext_destroy(ct); |
ea781f19 | 832 | nf_ct_ext_free(ct); |
5b3501fa | 833 | kmem_cache_free(net->ct.nf_conntrack_cachep, ct); |
0c3c6c00 PNA |
834 | smp_mb__before_atomic_dec(); |
835 | atomic_dec(&net->ct.count); | |
76507f69 | 836 | } |
13b18339 | 837 | EXPORT_SYMBOL_GPL(nf_conntrack_free); |
9fb9cbb1 | 838 | |
c539f017 | 839 | |
9fb9cbb1 YK |
840 | /* Allocate a new conntrack: we return -ENOMEM if classification |
841 | failed due to stress. Otherwise it really is unclassifiable. */ | |
842 | static struct nf_conntrack_tuple_hash * | |
b2a15a60 | 843 | init_conntrack(struct net *net, struct nf_conn *tmpl, |
5a1fb391 | 844 | const struct nf_conntrack_tuple *tuple, |
9fb9cbb1 | 845 | struct nf_conntrack_l3proto *l3proto, |
605dcad6 | 846 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 847 | struct sk_buff *skb, |
60b5f8f7 | 848 | unsigned int dataoff, u32 hash) |
9fb9cbb1 | 849 | { |
c88130bc | 850 | struct nf_conn *ct; |
3c158f7f | 851 | struct nf_conn_help *help; |
9fb9cbb1 | 852 | struct nf_conntrack_tuple repl_tuple; |
b2a15a60 | 853 | struct nf_conntrack_ecache *ecache; |
9fb9cbb1 | 854 | struct nf_conntrack_expect *exp; |
5d0aa2cc | 855 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
60b5f8f7 PNA |
856 | struct nf_conn_timeout *timeout_ext; |
857 | unsigned int *timeouts; | |
9fb9cbb1 | 858 | |
605dcad6 | 859 | if (!nf_ct_invert_tuple(&repl_tuple, tuple, l3proto, l4proto)) { |
0d53778e | 860 | pr_debug("Can't invert tuple.\n"); |
9fb9cbb1 YK |
861 | return NULL; |
862 | } | |
863 | ||
99f07e91 CG |
864 | ct = __nf_conntrack_alloc(net, zone, tuple, &repl_tuple, GFP_ATOMIC, |
865 | hash); | |
0a9ee813 | 866 | if (IS_ERR(ct)) |
c88130bc | 867 | return (struct nf_conntrack_tuple_hash *)ct; |
9fb9cbb1 | 868 | |
48b1de4c PM |
869 | if (tmpl && nfct_synproxy(tmpl)) { |
870 | nfct_seqadj_ext_add(ct); | |
871 | nfct_synproxy_ext_add(ct); | |
872 | } | |
873 | ||
60b5f8f7 PNA |
874 | timeout_ext = tmpl ? nf_ct_timeout_find(tmpl) : NULL; |
875 | if (timeout_ext) | |
876 | timeouts = NF_CT_TIMEOUT_EXT_DATA(timeout_ext); | |
877 | else | |
878 | timeouts = l4proto->get_timeouts(net); | |
879 | ||
2c8503f5 | 880 | if (!l4proto->new(ct, skb, dataoff, timeouts)) { |
c88130bc | 881 | nf_conntrack_free(ct); |
0d53778e | 882 | pr_debug("init conntrack: can't track with proto module\n"); |
9fb9cbb1 YK |
883 | return NULL; |
884 | } | |
885 | ||
60b5f8f7 PNA |
886 | if (timeout_ext) |
887 | nf_ct_timeout_ext_add(ct, timeout_ext->timeout, GFP_ATOMIC); | |
888 | ||
58401572 | 889 | nf_ct_acct_ext_add(ct, GFP_ATOMIC); |
a992ca2a | 890 | nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); |
c539f017 | 891 | nf_ct_labels_ext_add(ct); |
b2a15a60 PM |
892 | |
893 | ecache = tmpl ? nf_ct_ecache_find(tmpl) : NULL; | |
894 | nf_ct_ecache_ext_add(ct, ecache ? ecache->ctmask : 0, | |
895 | ecache ? ecache->expmask : 0, | |
896 | GFP_ATOMIC); | |
58401572 | 897 | |
f8ba1aff | 898 | spin_lock_bh(&nf_conntrack_lock); |
5d0aa2cc | 899 | exp = nf_ct_find_expectation(net, zone, tuple); |
9fb9cbb1 | 900 | if (exp) { |
0d53778e | 901 | pr_debug("conntrack: expectation arrives ct=%p exp=%p\n", |
c88130bc | 902 | ct, exp); |
9fb9cbb1 | 903 | /* Welcome, Mr. Bond. We've been expecting you... */ |
c88130bc | 904 | __set_bit(IPS_EXPECTED_BIT, &ct->status); |
e1b207da | 905 | /* exp->master safe, refcnt bumped in nf_ct_find_expectation */ |
c88130bc | 906 | ct->master = exp->master; |
ceceae1b | 907 | if (exp->helper) { |
1afc5679 PNA |
908 | help = nf_ct_helper_ext_add(ct, exp->helper, |
909 | GFP_ATOMIC); | |
ceceae1b | 910 | if (help) |
cf778b00 | 911 | rcu_assign_pointer(help->helper, exp->helper); |
ceceae1b YK |
912 | } |
913 | ||
9fb9cbb1 | 914 | #ifdef CONFIG_NF_CONNTRACK_MARK |
c88130bc | 915 | ct->mark = exp->master->mark; |
7c9728c3 JM |
916 | #endif |
917 | #ifdef CONFIG_NF_CONNTRACK_SECMARK | |
c88130bc | 918 | ct->secmark = exp->master->secmark; |
9fb9cbb1 | 919 | #endif |
0d55af87 | 920 | NF_CT_STAT_INC(net, expect_new); |
22e7410b | 921 | } else { |
b2a15a60 | 922 | __nf_ct_try_assign_helper(ct, tmpl, GFP_ATOMIC); |
0d55af87 | 923 | NF_CT_STAT_INC(net, new); |
22e7410b | 924 | } |
9fb9cbb1 | 925 | |
e53376be PNA |
926 | /* Now it is inserted into the unconfirmed list, bump refcount */ |
927 | nf_conntrack_get(&ct->ct_general); | |
b7779d06 | 928 | nf_ct_add_to_unconfirmed_list(ct); |
9fb9cbb1 | 929 | |
f8ba1aff | 930 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 YK |
931 | |
932 | if (exp) { | |
933 | if (exp->expectfn) | |
c88130bc | 934 | exp->expectfn(ct, exp); |
6823645d | 935 | nf_ct_expect_put(exp); |
9fb9cbb1 YK |
936 | } |
937 | ||
c88130bc | 938 | return &ct->tuplehash[IP_CT_DIR_ORIGINAL]; |
9fb9cbb1 YK |
939 | } |
940 | ||
941 | /* On success, returns conntrack ptr, sets skb->nfct and ctinfo */ | |
942 | static inline struct nf_conn * | |
b2a15a60 | 943 | resolve_normal_ct(struct net *net, struct nf_conn *tmpl, |
a702a65f | 944 | struct sk_buff *skb, |
9fb9cbb1 YK |
945 | unsigned int dataoff, |
946 | u_int16_t l3num, | |
947 | u_int8_t protonum, | |
948 | struct nf_conntrack_l3proto *l3proto, | |
605dcad6 | 949 | struct nf_conntrack_l4proto *l4proto, |
9fb9cbb1 | 950 | int *set_reply, |
60b5f8f7 | 951 | enum ip_conntrack_info *ctinfo) |
9fb9cbb1 YK |
952 | { |
953 | struct nf_conntrack_tuple tuple; | |
954 | struct nf_conntrack_tuple_hash *h; | |
955 | struct nf_conn *ct; | |
5d0aa2cc | 956 | u16 zone = tmpl ? nf_ct_zone(tmpl) : NF_CT_DEFAULT_ZONE; |
99f07e91 | 957 | u32 hash; |
9fb9cbb1 | 958 | |
bbe735e4 | 959 | if (!nf_ct_get_tuple(skb, skb_network_offset(skb), |
9fb9cbb1 | 960 | dataoff, l3num, protonum, &tuple, l3proto, |
605dcad6 | 961 | l4proto)) { |
0d53778e | 962 | pr_debug("resolve_normal_ct: Can't get tuple\n"); |
9fb9cbb1 YK |
963 | return NULL; |
964 | } | |
965 | ||
966 | /* look for tuple match */ | |
99f07e91 CG |
967 | hash = hash_conntrack_raw(&tuple, zone); |
968 | h = __nf_conntrack_find_get(net, zone, &tuple, hash); | |
9fb9cbb1 | 969 | if (!h) { |
b2a15a60 | 970 | h = init_conntrack(net, tmpl, &tuple, l3proto, l4proto, |
60b5f8f7 | 971 | skb, dataoff, hash); |
9fb9cbb1 YK |
972 | if (!h) |
973 | return NULL; | |
974 | if (IS_ERR(h)) | |
975 | return (void *)h; | |
976 | } | |
977 | ct = nf_ct_tuplehash_to_ctrack(h); | |
978 | ||
979 | /* It exists; we have (non-exclusive) reference. */ | |
980 | if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY) { | |
fb048833 | 981 | *ctinfo = IP_CT_ESTABLISHED_REPLY; |
9fb9cbb1 YK |
982 | /* Please set reply bit if this packet OK */ |
983 | *set_reply = 1; | |
984 | } else { | |
985 | /* Once we've had two way comms, always ESTABLISHED. */ | |
986 | if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status)) { | |
0d53778e | 987 | pr_debug("nf_conntrack_in: normal packet for %p\n", ct); |
9fb9cbb1 YK |
988 | *ctinfo = IP_CT_ESTABLISHED; |
989 | } else if (test_bit(IPS_EXPECTED_BIT, &ct->status)) { | |
0d53778e PM |
990 | pr_debug("nf_conntrack_in: related packet for %p\n", |
991 | ct); | |
9fb9cbb1 YK |
992 | *ctinfo = IP_CT_RELATED; |
993 | } else { | |
0d53778e | 994 | pr_debug("nf_conntrack_in: new packet for %p\n", ct); |
9fb9cbb1 YK |
995 | *ctinfo = IP_CT_NEW; |
996 | } | |
997 | *set_reply = 0; | |
998 | } | |
999 | skb->nfct = &ct->ct_general; | |
1000 | skb->nfctinfo = *ctinfo; | |
1001 | return ct; | |
1002 | } | |
1003 | ||
1004 | unsigned int | |
a702a65f AD |
1005 | nf_conntrack_in(struct net *net, u_int8_t pf, unsigned int hooknum, |
1006 | struct sk_buff *skb) | |
9fb9cbb1 | 1007 | { |
b2a15a60 | 1008 | struct nf_conn *ct, *tmpl = NULL; |
9fb9cbb1 YK |
1009 | enum ip_conntrack_info ctinfo; |
1010 | struct nf_conntrack_l3proto *l3proto; | |
605dcad6 | 1011 | struct nf_conntrack_l4proto *l4proto; |
2c8503f5 | 1012 | unsigned int *timeouts; |
9fb9cbb1 YK |
1013 | unsigned int dataoff; |
1014 | u_int8_t protonum; | |
1015 | int set_reply = 0; | |
1016 | int ret; | |
1017 | ||
3db05fea | 1018 | if (skb->nfct) { |
b2a15a60 PM |
1019 | /* Previously seen (loopback or untracked)? Ignore. */ |
1020 | tmpl = (struct nf_conn *)skb->nfct; | |
1021 | if (!nf_ct_is_template(tmpl)) { | |
1022 | NF_CT_STAT_INC_ATOMIC(net, ignore); | |
1023 | return NF_ACCEPT; | |
1024 | } | |
1025 | skb->nfct = NULL; | |
9fb9cbb1 YK |
1026 | } |
1027 | ||
923f4902 | 1028 | /* rcu_read_lock()ed by nf_hook_slow */ |
76108cea | 1029 | l3proto = __nf_ct_l3proto_find(pf); |
3db05fea | 1030 | ret = l3proto->get_l4proto(skb, skb_network_offset(skb), |
ffc30690 YK |
1031 | &dataoff, &protonum); |
1032 | if (ret <= 0) { | |
25985edc | 1033 | pr_debug("not prepared to track yet or error occurred\n"); |
0d55af87 AD |
1034 | NF_CT_STAT_INC_ATOMIC(net, error); |
1035 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
1036 | ret = -ret; |
1037 | goto out; | |
9fb9cbb1 YK |
1038 | } |
1039 | ||
76108cea | 1040 | l4proto = __nf_ct_l4proto_find(pf, protonum); |
9fb9cbb1 YK |
1041 | |
1042 | /* It may be an special packet, error, unclean... | |
1043 | * inverse of the return code tells to the netfilter | |
1044 | * core what to do with the packet. */ | |
74c51a14 | 1045 | if (l4proto->error != NULL) { |
8fea97ec PM |
1046 | ret = l4proto->error(net, tmpl, skb, dataoff, &ctinfo, |
1047 | pf, hooknum); | |
74c51a14 | 1048 | if (ret <= 0) { |
0d55af87 AD |
1049 | NF_CT_STAT_INC_ATOMIC(net, error); |
1050 | NF_CT_STAT_INC_ATOMIC(net, invalid); | |
b2a15a60 PM |
1051 | ret = -ret; |
1052 | goto out; | |
74c51a14 | 1053 | } |
88ed01d1 PNA |
1054 | /* ICMP[v6] protocol trackers may assign one conntrack. */ |
1055 | if (skb->nfct) | |
1056 | goto out; | |
9fb9cbb1 YK |
1057 | } |
1058 | ||
b2a15a60 | 1059 | ct = resolve_normal_ct(net, tmpl, skb, dataoff, pf, protonum, |
60b5f8f7 | 1060 | l3proto, l4proto, &set_reply, &ctinfo); |
9fb9cbb1 YK |
1061 | if (!ct) { |
1062 | /* Not valid part of a connection */ | |
0d55af87 | 1063 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
b2a15a60 PM |
1064 | ret = NF_ACCEPT; |
1065 | goto out; | |
9fb9cbb1 YK |
1066 | } |
1067 | ||
1068 | if (IS_ERR(ct)) { | |
1069 | /* Too stressed to deal. */ | |
0d55af87 | 1070 | NF_CT_STAT_INC_ATOMIC(net, drop); |
b2a15a60 PM |
1071 | ret = NF_DROP; |
1072 | goto out; | |
9fb9cbb1 YK |
1073 | } |
1074 | ||
3db05fea | 1075 | NF_CT_ASSERT(skb->nfct); |
9fb9cbb1 | 1076 | |
60b5f8f7 | 1077 | /* Decide what timeout policy we want to apply to this flow. */ |
84b5ee93 | 1078 | timeouts = nf_ct_timeout_lookup(net, ct, l4proto); |
60b5f8f7 | 1079 | |
2c8503f5 | 1080 | ret = l4proto->packet(ct, skb, dataoff, ctinfo, pf, hooknum, timeouts); |
ec8d5409 | 1081 | if (ret <= 0) { |
9fb9cbb1 YK |
1082 | /* Invalid: inverse of the return code tells |
1083 | * the netfilter core what to do */ | |
0d53778e | 1084 | pr_debug("nf_conntrack_in: Can't track with proto module\n"); |
3db05fea HX |
1085 | nf_conntrack_put(skb->nfct); |
1086 | skb->nfct = NULL; | |
0d55af87 | 1087 | NF_CT_STAT_INC_ATOMIC(net, invalid); |
7d1e0459 PNA |
1088 | if (ret == -NF_DROP) |
1089 | NF_CT_STAT_INC_ATOMIC(net, drop); | |
b2a15a60 PM |
1090 | ret = -ret; |
1091 | goto out; | |
9fb9cbb1 YK |
1092 | } |
1093 | ||
1094 | if (set_reply && !test_and_set_bit(IPS_SEEN_REPLY_BIT, &ct->status)) | |
858b3133 | 1095 | nf_conntrack_event_cache(IPCT_REPLY, ct); |
b2a15a60 | 1096 | out: |
c3174286 PNA |
1097 | if (tmpl) { |
1098 | /* Special case: we have to repeat this hook, assign the | |
1099 | * template again to this packet. We assume that this packet | |
1100 | * has no conntrack assigned. This is used by nf_ct_tcp. */ | |
1101 | if (ret == NF_REPEAT) | |
1102 | skb->nfct = (struct nf_conntrack *)tmpl; | |
1103 | else | |
1104 | nf_ct_put(tmpl); | |
1105 | } | |
9fb9cbb1 YK |
1106 | |
1107 | return ret; | |
1108 | } | |
13b18339 | 1109 | EXPORT_SYMBOL_GPL(nf_conntrack_in); |
9fb9cbb1 | 1110 | |
5f2b4c90 JE |
1111 | bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse, |
1112 | const struct nf_conntrack_tuple *orig) | |
9fb9cbb1 | 1113 | { |
5f2b4c90 | 1114 | bool ret; |
923f4902 PM |
1115 | |
1116 | rcu_read_lock(); | |
1117 | ret = nf_ct_invert_tuple(inverse, orig, | |
1118 | __nf_ct_l3proto_find(orig->src.l3num), | |
1119 | __nf_ct_l4proto_find(orig->src.l3num, | |
1120 | orig->dst.protonum)); | |
1121 | rcu_read_unlock(); | |
1122 | return ret; | |
9fb9cbb1 | 1123 | } |
13b18339 | 1124 | EXPORT_SYMBOL_GPL(nf_ct_invert_tuplepr); |
9fb9cbb1 | 1125 | |
5b1158e9 JK |
1126 | /* Alter reply tuple (maybe alter helper). This is for NAT, and is |
1127 | implicitly racy: see __nf_conntrack_confirm */ | |
1128 | void nf_conntrack_alter_reply(struct nf_conn *ct, | |
1129 | const struct nf_conntrack_tuple *newreply) | |
1130 | { | |
1131 | struct nf_conn_help *help = nfct_help(ct); | |
1132 | ||
5b1158e9 JK |
1133 | /* Should be unconfirmed, so not in hash table yet */ |
1134 | NF_CT_ASSERT(!nf_ct_is_confirmed(ct)); | |
1135 | ||
0d53778e | 1136 | pr_debug("Altering reply tuple of %p to ", ct); |
3c9fba65 | 1137 | nf_ct_dump_tuple(newreply); |
5b1158e9 JK |
1138 | |
1139 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; | |
ef1a5a50 | 1140 | if (ct->master || (help && !hlist_empty(&help->expectations))) |
c52fbb41 | 1141 | return; |
ceceae1b | 1142 | |
c52fbb41 | 1143 | rcu_read_lock(); |
b2a15a60 | 1144 | __nf_ct_try_assign_helper(ct, NULL, GFP_ATOMIC); |
c52fbb41 | 1145 | rcu_read_unlock(); |
5b1158e9 | 1146 | } |
13b18339 | 1147 | EXPORT_SYMBOL_GPL(nf_conntrack_alter_reply); |
5b1158e9 | 1148 | |
9fb9cbb1 YK |
1149 | /* Refresh conntrack for this many jiffies and do accounting if do_acct is 1 */ |
1150 | void __nf_ct_refresh_acct(struct nf_conn *ct, | |
1151 | enum ip_conntrack_info ctinfo, | |
1152 | const struct sk_buff *skb, | |
1153 | unsigned long extra_jiffies, | |
1154 | int do_acct) | |
1155 | { | |
9fb9cbb1 YK |
1156 | NF_CT_ASSERT(ct->timeout.data == (unsigned long)ct); |
1157 | NF_CT_ASSERT(skb); | |
1158 | ||
997ae831 | 1159 | /* Only update if this is not a fixed timeout */ |
47d95045 PM |
1160 | if (test_bit(IPS_FIXED_TIMEOUT_BIT, &ct->status)) |
1161 | goto acct; | |
997ae831 | 1162 | |
9fb9cbb1 YK |
1163 | /* If not in hash table, timer will not be active yet */ |
1164 | if (!nf_ct_is_confirmed(ct)) { | |
1165 | ct->timeout.expires = extra_jiffies; | |
9fb9cbb1 | 1166 | } else { |
be00c8e4 MJ |
1167 | unsigned long newtime = jiffies + extra_jiffies; |
1168 | ||
1169 | /* Only update the timeout if the new timeout is at least | |
1170 | HZ jiffies from the old timeout. Need del_timer for race | |
1171 | avoidance (may already be dying). */ | |
65cb9fda PM |
1172 | if (newtime - ct->timeout.expires >= HZ) |
1173 | mod_timer_pending(&ct->timeout, newtime); | |
9fb9cbb1 YK |
1174 | } |
1175 | ||
47d95045 | 1176 | acct: |
9fb9cbb1 | 1177 | if (do_acct) { |
f7b13e43 | 1178 | struct nf_conn_acct *acct; |
3ffd5eeb | 1179 | |
58401572 KPO |
1180 | acct = nf_conn_acct_find(ct); |
1181 | if (acct) { | |
f7b13e43 HE |
1182 | struct nf_conn_counter *counter = acct->counter; |
1183 | ||
1184 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); | |
1185 | atomic64_add(skb->len, &counter[CTINFO2DIR(ctinfo)].bytes); | |
58401572 | 1186 | } |
9fb9cbb1 | 1187 | } |
9fb9cbb1 | 1188 | } |
13b18339 | 1189 | EXPORT_SYMBOL_GPL(__nf_ct_refresh_acct); |
9fb9cbb1 | 1190 | |
4c889498 DM |
1191 | bool __nf_ct_kill_acct(struct nf_conn *ct, |
1192 | enum ip_conntrack_info ctinfo, | |
1193 | const struct sk_buff *skb, | |
1194 | int do_acct) | |
51091764 | 1195 | { |
718d4ad9 | 1196 | if (do_acct) { |
f7b13e43 | 1197 | struct nf_conn_acct *acct; |
58401572 | 1198 | |
58401572 KPO |
1199 | acct = nf_conn_acct_find(ct); |
1200 | if (acct) { | |
f7b13e43 HE |
1201 | struct nf_conn_counter *counter = acct->counter; |
1202 | ||
1203 | atomic64_inc(&counter[CTINFO2DIR(ctinfo)].packets); | |
b3e0bfa7 | 1204 | atomic64_add(skb->len - skb_network_offset(skb), |
f7b13e43 | 1205 | &counter[CTINFO2DIR(ctinfo)].bytes); |
58401572 | 1206 | } |
718d4ad9 | 1207 | } |
58401572 | 1208 | |
4c889498 | 1209 | if (del_timer(&ct->timeout)) { |
51091764 | 1210 | ct->timeout.function((unsigned long)ct); |
4c889498 DM |
1211 | return true; |
1212 | } | |
1213 | return false; | |
51091764 | 1214 | } |
718d4ad9 | 1215 | EXPORT_SYMBOL_GPL(__nf_ct_kill_acct); |
51091764 | 1216 | |
5d0aa2cc PM |
1217 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1218 | static struct nf_ct_ext_type nf_ct_zone_extend __read_mostly = { | |
1219 | .len = sizeof(struct nf_conntrack_zone), | |
1220 | .align = __alignof__(struct nf_conntrack_zone), | |
1221 | .id = NF_CT_EXT_ZONE, | |
1222 | }; | |
1223 | #endif | |
1224 | ||
c0cd1156 | 1225 | #if IS_ENABLED(CONFIG_NF_CT_NETLINK) |
c1d10adb PNA |
1226 | |
1227 | #include <linux/netfilter/nfnetlink.h> | |
1228 | #include <linux/netfilter/nfnetlink_conntrack.h> | |
57b47a53 IM |
1229 | #include <linux/mutex.h> |
1230 | ||
c1d10adb PNA |
1231 | /* Generic function for tcp/udp/sctp/dccp and alike. This needs to be |
1232 | * in ip_conntrack_core, since we don't want the protocols to autoload | |
1233 | * or depend on ctnetlink */ | |
fdf70832 | 1234 | int nf_ct_port_tuple_to_nlattr(struct sk_buff *skb, |
c1d10adb PNA |
1235 | const struct nf_conntrack_tuple *tuple) |
1236 | { | |
bae65be8 DM |
1237 | if (nla_put_be16(skb, CTA_PROTO_SRC_PORT, tuple->src.u.tcp.port) || |
1238 | nla_put_be16(skb, CTA_PROTO_DST_PORT, tuple->dst.u.tcp.port)) | |
1239 | goto nla_put_failure; | |
c1d10adb PNA |
1240 | return 0; |
1241 | ||
df6fb868 | 1242 | nla_put_failure: |
c1d10adb PNA |
1243 | return -1; |
1244 | } | |
fdf70832 | 1245 | EXPORT_SYMBOL_GPL(nf_ct_port_tuple_to_nlattr); |
c1d10adb | 1246 | |
f73e924c PM |
1247 | const struct nla_policy nf_ct_port_nla_policy[CTA_PROTO_MAX+1] = { |
1248 | [CTA_PROTO_SRC_PORT] = { .type = NLA_U16 }, | |
1249 | [CTA_PROTO_DST_PORT] = { .type = NLA_U16 }, | |
c1d10adb | 1250 | }; |
f73e924c | 1251 | EXPORT_SYMBOL_GPL(nf_ct_port_nla_policy); |
c1d10adb | 1252 | |
fdf70832 | 1253 | int nf_ct_port_nlattr_to_tuple(struct nlattr *tb[], |
c1d10adb PNA |
1254 | struct nf_conntrack_tuple *t) |
1255 | { | |
df6fb868 | 1256 | if (!tb[CTA_PROTO_SRC_PORT] || !tb[CTA_PROTO_DST_PORT]) |
c1d10adb PNA |
1257 | return -EINVAL; |
1258 | ||
77236b6e PM |
1259 | t->src.u.tcp.port = nla_get_be16(tb[CTA_PROTO_SRC_PORT]); |
1260 | t->dst.u.tcp.port = nla_get_be16(tb[CTA_PROTO_DST_PORT]); | |
c1d10adb PNA |
1261 | |
1262 | return 0; | |
1263 | } | |
fdf70832 | 1264 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_to_tuple); |
5c0de29d HE |
1265 | |
1266 | int nf_ct_port_nlattr_tuple_size(void) | |
1267 | { | |
1268 | return nla_policy_len(nf_ct_port_nla_policy, CTA_PROTO_MAX + 1); | |
1269 | } | |
1270 | EXPORT_SYMBOL_GPL(nf_ct_port_nlattr_tuple_size); | |
c1d10adb PNA |
1271 | #endif |
1272 | ||
9fb9cbb1 | 1273 | /* Used by ipt_REJECT and ip6t_REJECT. */ |
312a0c16 | 1274 | static void nf_conntrack_attach(struct sk_buff *nskb, const struct sk_buff *skb) |
9fb9cbb1 YK |
1275 | { |
1276 | struct nf_conn *ct; | |
1277 | enum ip_conntrack_info ctinfo; | |
1278 | ||
1279 | /* This ICMP is in reverse direction to the packet which caused it */ | |
1280 | ct = nf_ct_get(skb, &ctinfo); | |
1281 | if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) | |
fb048833 | 1282 | ctinfo = IP_CT_RELATED_REPLY; |
9fb9cbb1 YK |
1283 | else |
1284 | ctinfo = IP_CT_RELATED; | |
1285 | ||
1286 | /* Attach to new skbuff, and increment count */ | |
1287 | nskb->nfct = &ct->ct_general; | |
1288 | nskb->nfctinfo = ctinfo; | |
1289 | nf_conntrack_get(nskb->nfct); | |
1290 | } | |
1291 | ||
9fb9cbb1 | 1292 | /* Bring out ya dead! */ |
df0933dc | 1293 | static struct nf_conn * |
400dad39 | 1294 | get_next_corpse(struct net *net, int (*iter)(struct nf_conn *i, void *data), |
9fb9cbb1 YK |
1295 | void *data, unsigned int *bucket) |
1296 | { | |
df0933dc PM |
1297 | struct nf_conntrack_tuple_hash *h; |
1298 | struct nf_conn *ct; | |
ea781f19 | 1299 | struct hlist_nulls_node *n; |
b7779d06 | 1300 | int cpu; |
9fb9cbb1 | 1301 | |
f8ba1aff | 1302 | spin_lock_bh(&nf_conntrack_lock); |
d696c7bd | 1303 | for (; *bucket < net->ct.htable_size; (*bucket)++) { |
ea781f19 | 1304 | hlist_nulls_for_each_entry(h, n, &net->ct.hash[*bucket], hnnode) { |
b0cdb1d9 PM |
1305 | if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) |
1306 | continue; | |
df0933dc PM |
1307 | ct = nf_ct_tuplehash_to_ctrack(h); |
1308 | if (iter(ct, data)) | |
1309 | goto found; | |
1310 | } | |
601e68e1 | 1311 | } |
f8ba1aff | 1312 | spin_unlock_bh(&nf_conntrack_lock); |
b7779d06 JDB |
1313 | |
1314 | for_each_possible_cpu(cpu) { | |
1315 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | |
1316 | ||
1317 | spin_lock_bh(&pcpu->lock); | |
1318 | hlist_nulls_for_each_entry(h, n, &pcpu->unconfirmed, hnnode) { | |
1319 | ct = nf_ct_tuplehash_to_ctrack(h); | |
1320 | if (iter(ct, data)) | |
1321 | set_bit(IPS_DYING_BIT, &ct->status); | |
1322 | } | |
1323 | spin_unlock_bh(&pcpu->lock); | |
1324 | } | |
df0933dc PM |
1325 | return NULL; |
1326 | found: | |
c073e3fa | 1327 | atomic_inc(&ct->ct_general.use); |
f8ba1aff | 1328 | spin_unlock_bh(&nf_conntrack_lock); |
df0933dc | 1329 | return ct; |
9fb9cbb1 YK |
1330 | } |
1331 | ||
400dad39 AD |
1332 | void nf_ct_iterate_cleanup(struct net *net, |
1333 | int (*iter)(struct nf_conn *i, void *data), | |
c655bc68 | 1334 | void *data, u32 portid, int report) |
9fb9cbb1 | 1335 | { |
df0933dc | 1336 | struct nf_conn *ct; |
9fb9cbb1 YK |
1337 | unsigned int bucket = 0; |
1338 | ||
400dad39 | 1339 | while ((ct = get_next_corpse(net, iter, data, &bucket)) != NULL) { |
9fb9cbb1 YK |
1340 | /* Time to push up daises... */ |
1341 | if (del_timer(&ct->timeout)) | |
c655bc68 | 1342 | nf_ct_delete(ct, portid, report); |
02982c27 | 1343 | |
9fb9cbb1 YK |
1344 | /* ... else the timer will get him soon. */ |
1345 | ||
1346 | nf_ct_put(ct); | |
1347 | } | |
1348 | } | |
13b18339 | 1349 | EXPORT_SYMBOL_GPL(nf_ct_iterate_cleanup); |
9fb9cbb1 | 1350 | |
274d383b PNA |
1351 | static int kill_all(struct nf_conn *i, void *data) |
1352 | { | |
1353 | return 1; | |
1354 | } | |
1355 | ||
d862a662 | 1356 | void nf_ct_free_hashtable(void *hash, unsigned int size) |
9fb9cbb1 | 1357 | { |
d862a662 | 1358 | if (is_vmalloc_addr(hash)) |
9fb9cbb1 YK |
1359 | vfree(hash); |
1360 | else | |
601e68e1 | 1361 | free_pages((unsigned long)hash, |
f205c5e0 | 1362 | get_order(sizeof(struct hlist_head) * size)); |
9fb9cbb1 | 1363 | } |
ac565e5f | 1364 | EXPORT_SYMBOL_GPL(nf_ct_free_hashtable); |
9fb9cbb1 | 1365 | |
ec464e5d | 1366 | void nf_conntrack_flush_report(struct net *net, u32 portid, int report) |
c1d10adb | 1367 | { |
c655bc68 | 1368 | nf_ct_iterate_cleanup(net, kill_all, NULL, portid, report); |
c1d10adb | 1369 | } |
274d383b | 1370 | EXPORT_SYMBOL_GPL(nf_conntrack_flush_report); |
c1d10adb | 1371 | |
ee254fa4 | 1372 | static void nf_ct_release_dying_list(struct net *net) |
dd7669a9 PNA |
1373 | { |
1374 | struct nf_conntrack_tuple_hash *h; | |
1375 | struct nf_conn *ct; | |
1376 | struct hlist_nulls_node *n; | |
b7779d06 | 1377 | int cpu; |
dd7669a9 | 1378 | |
b7779d06 JDB |
1379 | for_each_possible_cpu(cpu) { |
1380 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | |
1381 | ||
1382 | spin_lock_bh(&pcpu->lock); | |
1383 | hlist_nulls_for_each_entry(h, n, &pcpu->dying, hnnode) { | |
1384 | ct = nf_ct_tuplehash_to_ctrack(h); | |
1385 | /* never fails to remove them, no listeners at this point */ | |
1386 | nf_ct_kill(ct); | |
1387 | } | |
1388 | spin_unlock_bh(&pcpu->lock); | |
dd7669a9 | 1389 | } |
dd7669a9 PNA |
1390 | } |
1391 | ||
b3c5163f ED |
1392 | static int untrack_refs(void) |
1393 | { | |
1394 | int cnt = 0, cpu; | |
1395 | ||
1396 | for_each_possible_cpu(cpu) { | |
1397 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
1398 | ||
1399 | cnt += atomic_read(&ct->ct_general.use) - 1; | |
1400 | } | |
1401 | return cnt; | |
1402 | } | |
1403 | ||
f94161c1 | 1404 | void nf_conntrack_cleanup_start(void) |
9fb9cbb1 | 1405 | { |
f94161c1 G |
1406 | RCU_INIT_POINTER(ip_ct_attach, NULL); |
1407 | } | |
1408 | ||
1409 | void nf_conntrack_cleanup_end(void) | |
1410 | { | |
1411 | RCU_INIT_POINTER(nf_ct_destroy, NULL); | |
b3c5163f | 1412 | while (untrack_refs() > 0) |
9edd7ca0 PM |
1413 | schedule(); |
1414 | ||
5d0aa2cc PM |
1415 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1416 | nf_ct_extend_unregister(&nf_ct_zone_extend); | |
1417 | #endif | |
04d87001 | 1418 | nf_conntrack_proto_fini(); |
41d73ec0 | 1419 | nf_conntrack_seqadj_fini(); |
5f69b8f5 | 1420 | nf_conntrack_labels_fini(); |
5e615b22 | 1421 | nf_conntrack_helper_fini(); |
8684094c | 1422 | nf_conntrack_timeout_fini(); |
3fe0f943 | 1423 | nf_conntrack_ecache_fini(); |
73f4001a | 1424 | nf_conntrack_tstamp_fini(); |
b7ff3a1f | 1425 | nf_conntrack_acct_fini(); |
83b4dbe1 | 1426 | nf_conntrack_expect_fini(); |
08f6547d | 1427 | } |
9fb9cbb1 | 1428 | |
f94161c1 G |
1429 | /* |
1430 | * Mishearing the voices in his head, our hero wonders how he's | |
1431 | * supposed to kill the mall. | |
1432 | */ | |
1433 | void nf_conntrack_cleanup_net(struct net *net) | |
08f6547d | 1434 | { |
dece40e8 VD |
1435 | LIST_HEAD(single); |
1436 | ||
1437 | list_add(&net->exit_list, &single); | |
1438 | nf_conntrack_cleanup_net_list(&single); | |
1439 | } | |
1440 | ||
1441 | void nf_conntrack_cleanup_net_list(struct list_head *net_exit_list) | |
1442 | { | |
1443 | int busy; | |
1444 | struct net *net; | |
1445 | ||
f94161c1 G |
1446 | /* |
1447 | * This makes sure all current packets have passed through | |
1448 | * netfilter framework. Roll on, two-stage module | |
1449 | * delete... | |
1450 | */ | |
1451 | synchronize_net(); | |
dece40e8 VD |
1452 | i_see_dead_people: |
1453 | busy = 0; | |
1454 | list_for_each_entry(net, net_exit_list, exit_list) { | |
c655bc68 | 1455 | nf_ct_iterate_cleanup(net, kill_all, NULL, 0, 0); |
dece40e8 VD |
1456 | nf_ct_release_dying_list(net); |
1457 | if (atomic_read(&net->ct.count) != 0) | |
1458 | busy = 1; | |
1459 | } | |
1460 | if (busy) { | |
9fb9cbb1 YK |
1461 | schedule(); |
1462 | goto i_see_dead_people; | |
1463 | } | |
1464 | ||
dece40e8 VD |
1465 | list_for_each_entry(net, net_exit_list, exit_list) { |
1466 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); | |
1467 | nf_conntrack_proto_pernet_fini(net); | |
1468 | nf_conntrack_helper_pernet_fini(net); | |
1469 | nf_conntrack_ecache_pernet_fini(net); | |
1470 | nf_conntrack_tstamp_pernet_fini(net); | |
1471 | nf_conntrack_acct_pernet_fini(net); | |
1472 | nf_conntrack_expect_pernet_fini(net); | |
1473 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); | |
1474 | kfree(net->ct.slabname); | |
1475 | free_percpu(net->ct.stat); | |
b7779d06 | 1476 | free_percpu(net->ct.pcpu_lists); |
dece40e8 | 1477 | } |
08f6547d AD |
1478 | } |
1479 | ||
d862a662 | 1480 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls) |
9fb9cbb1 | 1481 | { |
ea781f19 ED |
1482 | struct hlist_nulls_head *hash; |
1483 | unsigned int nr_slots, i; | |
1484 | size_t sz; | |
9fb9cbb1 | 1485 | |
ea781f19 ED |
1486 | BUILD_BUG_ON(sizeof(struct hlist_nulls_head) != sizeof(struct hlist_head)); |
1487 | nr_slots = *sizep = roundup(*sizep, PAGE_SIZE / sizeof(struct hlist_nulls_head)); | |
1488 | sz = nr_slots * sizeof(struct hlist_nulls_head); | |
1489 | hash = (void *)__get_free_pages(GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO, | |
1490 | get_order(sz)); | |
601e68e1 | 1491 | if (!hash) { |
9fb9cbb1 | 1492 | printk(KERN_WARNING "nf_conntrack: falling back to vmalloc.\n"); |
966567b7 | 1493 | hash = vzalloc(sz); |
9fb9cbb1 YK |
1494 | } |
1495 | ||
ea781f19 ED |
1496 | if (hash && nulls) |
1497 | for (i = 0; i < nr_slots; i++) | |
1498 | INIT_HLIST_NULLS_HEAD(&hash[i], i); | |
9fb9cbb1 YK |
1499 | |
1500 | return hash; | |
1501 | } | |
ac565e5f | 1502 | EXPORT_SYMBOL_GPL(nf_ct_alloc_hashtable); |
9fb9cbb1 | 1503 | |
fae718dd | 1504 | int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp) |
9fb9cbb1 | 1505 | { |
4b5511eb | 1506 | int i, bucket, rc; |
96eb24d7 | 1507 | unsigned int hashsize, old_size; |
ea781f19 | 1508 | struct hlist_nulls_head *hash, *old_hash; |
9fb9cbb1 | 1509 | struct nf_conntrack_tuple_hash *h; |
5d0aa2cc | 1510 | struct nf_conn *ct; |
9fb9cbb1 | 1511 | |
d696c7bd PM |
1512 | if (current->nsproxy->net_ns != &init_net) |
1513 | return -EOPNOTSUPP; | |
1514 | ||
9fb9cbb1 YK |
1515 | /* On boot, we can set this without any fancy locking. */ |
1516 | if (!nf_conntrack_htable_size) | |
1517 | return param_set_uint(val, kp); | |
1518 | ||
4b5511eb AP |
1519 | rc = kstrtouint(val, 0, &hashsize); |
1520 | if (rc) | |
1521 | return rc; | |
9fb9cbb1 YK |
1522 | if (!hashsize) |
1523 | return -EINVAL; | |
1524 | ||
d862a662 | 1525 | hash = nf_ct_alloc_hashtable(&hashsize, 1); |
9fb9cbb1 YK |
1526 | if (!hash) |
1527 | return -ENOMEM; | |
1528 | ||
76507f69 PM |
1529 | /* Lookups in the old hash might happen in parallel, which means we |
1530 | * might get false negatives during connection lookup. New connections | |
1531 | * created because of a false negative won't make it into the hash | |
1532 | * though since that required taking the lock. | |
1533 | */ | |
f8ba1aff | 1534 | spin_lock_bh(&nf_conntrack_lock); |
d696c7bd | 1535 | for (i = 0; i < init_net.ct.htable_size; i++) { |
ea781f19 ED |
1536 | while (!hlist_nulls_empty(&init_net.ct.hash[i])) { |
1537 | h = hlist_nulls_entry(init_net.ct.hash[i].first, | |
1538 | struct nf_conntrack_tuple_hash, hnnode); | |
5d0aa2cc | 1539 | ct = nf_ct_tuplehash_to_ctrack(h); |
ea781f19 | 1540 | hlist_nulls_del_rcu(&h->hnnode); |
5d0aa2cc | 1541 | bucket = __hash_conntrack(&h->tuple, nf_ct_zone(ct), |
99f07e91 | 1542 | hashsize); |
ea781f19 | 1543 | hlist_nulls_add_head_rcu(&h->hnnode, &hash[bucket]); |
9fb9cbb1 YK |
1544 | } |
1545 | } | |
d696c7bd | 1546 | old_size = init_net.ct.htable_size; |
400dad39 | 1547 | old_hash = init_net.ct.hash; |
9fb9cbb1 | 1548 | |
d696c7bd | 1549 | init_net.ct.htable_size = nf_conntrack_htable_size = hashsize; |
400dad39 | 1550 | init_net.ct.hash = hash; |
f8ba1aff | 1551 | spin_unlock_bh(&nf_conntrack_lock); |
9fb9cbb1 | 1552 | |
d862a662 | 1553 | nf_ct_free_hashtable(old_hash, old_size); |
9fb9cbb1 YK |
1554 | return 0; |
1555 | } | |
fae718dd | 1556 | EXPORT_SYMBOL_GPL(nf_conntrack_set_hashsize); |
9fb9cbb1 | 1557 | |
fae718dd | 1558 | module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint, |
9fb9cbb1 YK |
1559 | &nf_conntrack_htable_size, 0600); |
1560 | ||
5bfddbd4 ED |
1561 | void nf_ct_untracked_status_or(unsigned long bits) |
1562 | { | |
b3c5163f ED |
1563 | int cpu; |
1564 | ||
1565 | for_each_possible_cpu(cpu) | |
1566 | per_cpu(nf_conntrack_untracked, cpu).status |= bits; | |
5bfddbd4 ED |
1567 | } |
1568 | EXPORT_SYMBOL_GPL(nf_ct_untracked_status_or); | |
1569 | ||
f94161c1 | 1570 | int nf_conntrack_init_start(void) |
9fb9cbb1 | 1571 | { |
f205c5e0 | 1572 | int max_factor = 8; |
b3c5163f | 1573 | int ret, cpu; |
9fb9cbb1 YK |
1574 | |
1575 | /* Idea from tcp.c: use 1/16384 of memory. On i386: 32MB | |
f205c5e0 | 1576 | * machine has 512 buckets. >= 1GB machines have 16384 buckets. */ |
9fb9cbb1 YK |
1577 | if (!nf_conntrack_htable_size) { |
1578 | nf_conntrack_htable_size | |
4481374c | 1579 | = (((totalram_pages << PAGE_SHIFT) / 16384) |
f205c5e0 | 1580 | / sizeof(struct hlist_head)); |
4481374c | 1581 | if (totalram_pages > (1024 * 1024 * 1024 / PAGE_SIZE)) |
f205c5e0 PM |
1582 | nf_conntrack_htable_size = 16384; |
1583 | if (nf_conntrack_htable_size < 32) | |
1584 | nf_conntrack_htable_size = 32; | |
1585 | ||
1586 | /* Use a max. factor of four by default to get the same max as | |
1587 | * with the old struct list_heads. When a table size is given | |
1588 | * we use the old value of 8 to avoid reducing the max. | |
1589 | * entries. */ | |
1590 | max_factor = 4; | |
9fb9cbb1 | 1591 | } |
f205c5e0 | 1592 | nf_conntrack_max = max_factor * nf_conntrack_htable_size; |
8e5105a0 | 1593 | |
654d0fbd | 1594 | printk(KERN_INFO "nf_conntrack version %s (%u buckets, %d max)\n", |
8e5105a0 PM |
1595 | NF_CONNTRACK_VERSION, nf_conntrack_htable_size, |
1596 | nf_conntrack_max); | |
83b4dbe1 G |
1597 | |
1598 | ret = nf_conntrack_expect_init(); | |
1599 | if (ret < 0) | |
1600 | goto err_expect; | |
1601 | ||
b7ff3a1f G |
1602 | ret = nf_conntrack_acct_init(); |
1603 | if (ret < 0) | |
1604 | goto err_acct; | |
1605 | ||
73f4001a G |
1606 | ret = nf_conntrack_tstamp_init(); |
1607 | if (ret < 0) | |
1608 | goto err_tstamp; | |
1609 | ||
3fe0f943 G |
1610 | ret = nf_conntrack_ecache_init(); |
1611 | if (ret < 0) | |
1612 | goto err_ecache; | |
1613 | ||
8684094c G |
1614 | ret = nf_conntrack_timeout_init(); |
1615 | if (ret < 0) | |
1616 | goto err_timeout; | |
1617 | ||
5e615b22 G |
1618 | ret = nf_conntrack_helper_init(); |
1619 | if (ret < 0) | |
1620 | goto err_helper; | |
1621 | ||
5f69b8f5 G |
1622 | ret = nf_conntrack_labels_init(); |
1623 | if (ret < 0) | |
1624 | goto err_labels; | |
1625 | ||
41d73ec0 PM |
1626 | ret = nf_conntrack_seqadj_init(); |
1627 | if (ret < 0) | |
1628 | goto err_seqadj; | |
1629 | ||
5d0aa2cc PM |
1630 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
1631 | ret = nf_ct_extend_register(&nf_ct_zone_extend); | |
1632 | if (ret < 0) | |
1633 | goto err_extend; | |
1634 | #endif | |
04d87001 G |
1635 | ret = nf_conntrack_proto_init(); |
1636 | if (ret < 0) | |
1637 | goto err_proto; | |
1638 | ||
9edd7ca0 | 1639 | /* Set up fake conntrack: to never be deleted, not in any hashes */ |
b3c5163f ED |
1640 | for_each_possible_cpu(cpu) { |
1641 | struct nf_conn *ct = &per_cpu(nf_conntrack_untracked, cpu); | |
b3c5163f ED |
1642 | write_pnet(&ct->ct_net, &init_net); |
1643 | atomic_set(&ct->ct_general.use, 1); | |
1644 | } | |
9edd7ca0 | 1645 | /* - and look it like as a confirmed connection */ |
5bfddbd4 | 1646 | nf_ct_untracked_status_or(IPS_CONFIRMED | IPS_UNTRACKED); |
08f6547d AD |
1647 | return 0; |
1648 | ||
04d87001 | 1649 | err_proto: |
5d0aa2cc | 1650 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
04d87001 | 1651 | nf_ct_extend_unregister(&nf_ct_zone_extend); |
5d0aa2cc | 1652 | err_extend: |
a9006892 | 1653 | #endif |
41d73ec0 PM |
1654 | nf_conntrack_seqadj_fini(); |
1655 | err_seqadj: | |
04d87001 | 1656 | nf_conntrack_labels_fini(); |
5f69b8f5 G |
1657 | err_labels: |
1658 | nf_conntrack_helper_fini(); | |
5e615b22 G |
1659 | err_helper: |
1660 | nf_conntrack_timeout_fini(); | |
8684094c G |
1661 | err_timeout: |
1662 | nf_conntrack_ecache_fini(); | |
3fe0f943 G |
1663 | err_ecache: |
1664 | nf_conntrack_tstamp_fini(); | |
73f4001a G |
1665 | err_tstamp: |
1666 | nf_conntrack_acct_fini(); | |
b7ff3a1f G |
1667 | err_acct: |
1668 | nf_conntrack_expect_fini(); | |
83b4dbe1 | 1669 | err_expect: |
08f6547d AD |
1670 | return ret; |
1671 | } | |
1672 | ||
f94161c1 G |
1673 | void nf_conntrack_init_end(void) |
1674 | { | |
1675 | /* For use by REJECT target */ | |
1676 | RCU_INIT_POINTER(ip_ct_attach, nf_conntrack_attach); | |
1677 | RCU_INIT_POINTER(nf_ct_destroy, destroy_conntrack); | |
f94161c1 G |
1678 | } |
1679 | ||
8cc20198 ED |
1680 | /* |
1681 | * We need to use special "null" values, not used in hash table | |
1682 | */ | |
1683 | #define UNCONFIRMED_NULLS_VAL ((1<<30)+0) | |
1684 | #define DYING_NULLS_VAL ((1<<30)+1) | |
252b3e8c | 1685 | #define TEMPLATE_NULLS_VAL ((1<<30)+2) |
8cc20198 | 1686 | |
f94161c1 | 1687 | int nf_conntrack_init_net(struct net *net) |
08f6547d | 1688 | { |
b7779d06 JDB |
1689 | int ret = -ENOMEM; |
1690 | int cpu; | |
ceceae1b | 1691 | |
08f6547d | 1692 | atomic_set(&net->ct.count, 0); |
b7779d06 JDB |
1693 | |
1694 | net->ct.pcpu_lists = alloc_percpu(struct ct_pcpu); | |
1695 | if (!net->ct.pcpu_lists) | |
08f6547d | 1696 | goto err_stat; |
b7779d06 JDB |
1697 | |
1698 | for_each_possible_cpu(cpu) { | |
1699 | struct ct_pcpu *pcpu = per_cpu_ptr(net->ct.pcpu_lists, cpu); | |
1700 | ||
1701 | spin_lock_init(&pcpu->lock); | |
1702 | INIT_HLIST_NULLS_HEAD(&pcpu->unconfirmed, UNCONFIRMED_NULLS_VAL); | |
1703 | INIT_HLIST_NULLS_HEAD(&pcpu->dying, DYING_NULLS_VAL); | |
1704 | INIT_HLIST_NULLS_HEAD(&pcpu->tmpl, TEMPLATE_NULLS_VAL); | |
08f6547d | 1705 | } |
5b3501fa | 1706 | |
b7779d06 JDB |
1707 | net->ct.stat = alloc_percpu(struct ip_conntrack_stat); |
1708 | if (!net->ct.stat) | |
1709 | goto err_pcpu_lists; | |
1710 | ||
5b3501fa | 1711 | net->ct.slabname = kasprintf(GFP_KERNEL, "nf_conntrack_%p", net); |
b7779d06 | 1712 | if (!net->ct.slabname) |
5b3501fa | 1713 | goto err_slabname; |
5b3501fa ED |
1714 | |
1715 | net->ct.nf_conntrack_cachep = kmem_cache_create(net->ct.slabname, | |
1716 | sizeof(struct nf_conn), 0, | |
1717 | SLAB_DESTROY_BY_RCU, NULL); | |
1718 | if (!net->ct.nf_conntrack_cachep) { | |
1719 | printk(KERN_ERR "Unable to create nf_conn slab cache\n"); | |
5b3501fa ED |
1720 | goto err_cache; |
1721 | } | |
d696c7bd PM |
1722 | |
1723 | net->ct.htable_size = nf_conntrack_htable_size; | |
d862a662 | 1724 | net->ct.hash = nf_ct_alloc_hashtable(&net->ct.htable_size, 1); |
08f6547d | 1725 | if (!net->ct.hash) { |
08f6547d AD |
1726 | printk(KERN_ERR "Unable to create nf_conntrack_hash\n"); |
1727 | goto err_hash; | |
1728 | } | |
83b4dbe1 | 1729 | ret = nf_conntrack_expect_pernet_init(net); |
08f6547d AD |
1730 | if (ret < 0) |
1731 | goto err_expect; | |
b7ff3a1f | 1732 | ret = nf_conntrack_acct_pernet_init(net); |
58401572 | 1733 | if (ret < 0) |
08f6547d | 1734 | goto err_acct; |
73f4001a | 1735 | ret = nf_conntrack_tstamp_pernet_init(net); |
a992ca2a PNA |
1736 | if (ret < 0) |
1737 | goto err_tstamp; | |
3fe0f943 | 1738 | ret = nf_conntrack_ecache_pernet_init(net); |
a0891aa6 PNA |
1739 | if (ret < 0) |
1740 | goto err_ecache; | |
5e615b22 | 1741 | ret = nf_conntrack_helper_pernet_init(net); |
a9006892 EL |
1742 | if (ret < 0) |
1743 | goto err_helper; | |
04d87001 | 1744 | ret = nf_conntrack_proto_pernet_init(net); |
f94161c1 G |
1745 | if (ret < 0) |
1746 | goto err_proto; | |
08f6547d | 1747 | return 0; |
c539f017 | 1748 | |
f94161c1 | 1749 | err_proto: |
5e615b22 | 1750 | nf_conntrack_helper_pernet_fini(net); |
a9006892 | 1751 | err_helper: |
3fe0f943 | 1752 | nf_conntrack_ecache_pernet_fini(net); |
a0891aa6 | 1753 | err_ecache: |
73f4001a | 1754 | nf_conntrack_tstamp_pernet_fini(net); |
a992ca2a | 1755 | err_tstamp: |
b7ff3a1f | 1756 | nf_conntrack_acct_pernet_fini(net); |
08f6547d | 1757 | err_acct: |
83b4dbe1 | 1758 | nf_conntrack_expect_pernet_fini(net); |
08f6547d | 1759 | err_expect: |
d862a662 | 1760 | nf_ct_free_hashtable(net->ct.hash, net->ct.htable_size); |
6058fa6b | 1761 | err_hash: |
5b3501fa ED |
1762 | kmem_cache_destroy(net->ct.nf_conntrack_cachep); |
1763 | err_cache: | |
1764 | kfree(net->ct.slabname); | |
1765 | err_slabname: | |
0d55af87 | 1766 | free_percpu(net->ct.stat); |
b7779d06 JDB |
1767 | err_pcpu_lists: |
1768 | free_percpu(net->ct.pcpu_lists); | |
0d55af87 | 1769 | err_stat: |
08f6547d AD |
1770 | return ret; |
1771 | } |