1 /* SPDX-License-Identifier: GPL-2.0 */
3 * Connection state tracking for netfilter. This is separated from,
4 * but required by, the (future) NAT layer; it can also be used by an iptables
7 * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
8 * - generalize L3 protocol dependent part.
10 * Derived from include/linux/netfiter_ipv4/ip_conntrack.h
13 #ifndef _NF_CONNTRACK_H
14 #define _NF_CONNTRACK_H
16 #include <linux/bitops.h>
17 #include <linux/compiler.h>
19 #include <linux/netfilter/nf_conntrack_common.h>
20 #include <linux/netfilter/nf_conntrack_tcp.h>
21 #include <linux/netfilter/nf_conntrack_dccp.h>
22 #include <linux/netfilter/nf_conntrack_sctp.h>
23 #include <linux/netfilter/nf_conntrack_proto_gre.h>
25 #include <net/netfilter/nf_conntrack_tuple.h>
28 unsigned long stream_ts;
31 /* per conntrack: protocol private data */
32 union nf_conntrack_proto {
33 /* insert conntrack proto private data here */
34 struct nf_ct_dccp dccp;
35 struct ip_ct_sctp sctp;
39 unsigned int tmpl_padto;
42 union nf_conntrack_expect_proto {
43 /* insert expect proto private data here */
46 struct nf_conntrack_net_ecache {
47 struct delayed_work dwork;
48 spinlock_t dying_lock;
49 struct hlist_nulls_head dying_list;
52 struct nf_conntrack_net {
53 /* only used when new connection is allocated: */
55 unsigned int expect_count;
57 /* only used from work queues, configuration plane, and so on: */
60 unsigned int users_bridge;
62 struct ctl_table_header *sysctl_header;
64 #ifdef CONFIG_NF_CONNTRACK_EVENTS
65 struct nf_conntrack_net_ecache ecache;
69 #include <linux/types.h>
70 #include <linux/skbuff.h>
72 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
73 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
76 /* Usage count in here is 1 for hash table, 1 per skb,
77 * plus 1 for any connection(s) we are `master' for
79 * Hint, SKB address this struct and refcnt via skb->_nfct and
80 * helpers nf_conntrack_get() and nf_conntrack_put().
81 * Helper nf_ct_put() equals nf_conntrack_put() by dec refcnt,
82 * except that the latter uses internal indirection and does not
83 * result in a conntrack module dependency.
84 * beware nf_ct_get() is different and don't inc refcnt.
86 struct nf_conntrack ct_general;
89 /* jiffies32 when this ct is considered dead */
92 #ifdef CONFIG_NF_CONNTRACK_ZONES
93 struct nf_conntrack_zone zone;
95 /* XXX should I move this to the tail ? - Y.K */
96 /* These are my tuples; original and reply */
97 struct nf_conntrack_tuple_hash tuplehash[IP_CT_DIR_MAX];
99 /* Have we seen traffic both ways yet? (bitset) */
100 unsigned long status;
102 possible_net_t ct_net;
104 #if IS_ENABLED(CONFIG_NF_NAT)
105 struct hlist_node nat_bysource;
107 /* all members below initialized via memset */
108 struct { } __nfct_init_offset;
110 /* If we were expected by an expectation, this will be it */
111 struct nf_conn *master;
113 #if defined(CONFIG_NF_CONNTRACK_MARK)
117 #ifdef CONFIG_NF_CONNTRACK_SECMARK
122 struct nf_ct_ext *ext;
124 /* Storage reserved for other modules, must be the last member */
125 union nf_conntrack_proto proto;
128 static inline struct nf_conn *
129 nf_ct_tuplehash_to_ctrack(const struct nf_conntrack_tuple_hash *hash)
131 return container_of(hash, struct nf_conn,
132 tuplehash[hash->tuple.dst.dir]);
135 static inline u_int16_t nf_ct_l3num(const struct nf_conn *ct)
137 return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
140 static inline u_int8_t nf_ct_protonum(const struct nf_conn *ct)
142 return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
145 #define nf_ct_tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
147 /* get master conntrack via master expectation */
148 #define master_ct(conntr) (conntr->master)
150 extern struct net init_net;
152 static inline struct net *nf_ct_net(const struct nf_conn *ct)
154 return read_pnet(&ct->ct_net);
157 /* Alter reply tuple (maybe alter helper). */
158 void nf_conntrack_alter_reply(struct nf_conn *ct,
159 const struct nf_conntrack_tuple *newreply);
161 /* Is this tuple taken? (ignoring any belonging to the given
163 int nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple,
164 const struct nf_conn *ignored_conntrack);
166 /* Return conntrack_info and tuple hash for given skb. */
167 static inline struct nf_conn *
168 nf_ct_get(const struct sk_buff *skb, enum ip_conntrack_info *ctinfo)
170 unsigned long nfct = skb_get_nfct(skb);
172 *ctinfo = nfct & NFCT_INFOMASK;
173 return (struct nf_conn *)(nfct & NFCT_PTRMASK);
176 void nf_ct_destroy(struct nf_conntrack *nfct);
178 /* decrement reference count on a conntrack */
179 static inline void nf_ct_put(struct nf_conn *ct)
181 if (ct && refcount_dec_and_test(&ct->ct_general.use))
182 nf_ct_destroy(&ct->ct_general);
185 /* Protocol module loading */
186 int nf_ct_l3proto_try_module_get(unsigned short l3proto);
187 void nf_ct_l3proto_module_put(unsigned short l3proto);
189 /* load module; enable/disable conntrack in this namespace */
190 int nf_ct_netns_get(struct net *net, u8 nfproto);
191 void nf_ct_netns_put(struct net *net, u8 nfproto);
194 * Allocate a hashtable of hlist_head (if nulls == 0),
195 * or hlist_nulls_head (if nulls == 1)
197 void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls);
199 int nf_conntrack_hash_check_insert(struct nf_conn *ct);
200 bool nf_ct_delete(struct nf_conn *ct, u32 pid, int report);
202 bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff,
203 u_int16_t l3num, struct net *net,
204 struct nf_conntrack_tuple *tuple);
206 void __nf_ct_refresh_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
207 const struct sk_buff *skb,
208 u32 extra_jiffies, bool do_acct);
210 /* Refresh conntrack for this many jiffies and do accounting */
211 static inline void nf_ct_refresh_acct(struct nf_conn *ct,
212 enum ip_conntrack_info ctinfo,
213 const struct sk_buff *skb,
216 __nf_ct_refresh_acct(ct, ctinfo, skb, extra_jiffies, true);
219 /* Refresh conntrack for this many jiffies */
220 static inline void nf_ct_refresh(struct nf_conn *ct,
221 const struct sk_buff *skb,
224 __nf_ct_refresh_acct(ct, 0, skb, extra_jiffies, false);
227 /* kill conntrack and do accounting */
228 bool nf_ct_kill_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
229 const struct sk_buff *skb);
231 /* kill conntrack without accounting */
232 static inline bool nf_ct_kill(struct nf_conn *ct)
234 return nf_ct_delete(ct, 0, 0);
237 struct nf_ct_iter_data {
244 /* Iterate over all conntracks: if iter returns true, it's deleted. */
245 void nf_ct_iterate_cleanup_net(int (*iter)(struct nf_conn *i, void *data),
246 const struct nf_ct_iter_data *iter_data);
248 /* also set unconfirmed conntracks as dying. Only use in module exit path. */
249 void nf_ct_iterate_destroy(int (*iter)(struct nf_conn *i, void *data),
252 struct nf_conntrack_zone;
254 void nf_conntrack_free(struct nf_conn *ct);
255 struct nf_conn *nf_conntrack_alloc(struct net *net,
256 const struct nf_conntrack_zone *zone,
257 const struct nf_conntrack_tuple *orig,
258 const struct nf_conntrack_tuple *repl,
261 static inline int nf_ct_is_template(const struct nf_conn *ct)
263 return test_bit(IPS_TEMPLATE_BIT, &ct->status);
266 /* It's confirmed if it is, or has been in the hash table. */
267 static inline int nf_ct_is_confirmed(const struct nf_conn *ct)
269 return test_bit(IPS_CONFIRMED_BIT, &ct->status);
272 static inline int nf_ct_is_dying(const struct nf_conn *ct)
274 return test_bit(IPS_DYING_BIT, &ct->status);
277 /* Packet is received from loopback */
278 static inline bool nf_is_loopback_packet(const struct sk_buff *skb)
280 return skb->dev && skb->skb_iif && skb->dev->flags & IFF_LOOPBACK;
283 #define nfct_time_stamp ((u32)(jiffies))
285 /* jiffies until ct expires, 0 if already expired */
286 static inline unsigned long nf_ct_expires(const struct nf_conn *ct)
288 s32 timeout = READ_ONCE(ct->timeout) - nfct_time_stamp;
290 return max(timeout, 0);
293 static inline bool nf_ct_is_expired(const struct nf_conn *ct)
295 return (__s32)(READ_ONCE(ct->timeout) - nfct_time_stamp) <= 0;
298 /* use after obtaining a reference count */
299 static inline bool nf_ct_should_gc(const struct nf_conn *ct)
301 return nf_ct_is_expired(ct) && nf_ct_is_confirmed(ct) &&
305 #define NF_CT_DAY (86400 * HZ)
307 /* Set an arbitrary timeout large enough not to ever expire, this save
308 * us a check for the IPS_OFFLOAD_BIT from the packet path via
309 * nf_ct_is_expired().
311 static inline void nf_ct_offload_timeout(struct nf_conn *ct)
313 if (nf_ct_expires(ct) < NF_CT_DAY / 2)
314 WRITE_ONCE(ct->timeout, nfct_time_stamp + NF_CT_DAY);
319 int nf_conntrack_set_hashsize(const char *val, const struct kernel_param *kp);
320 int nf_conntrack_hash_resize(unsigned int hashsize);
322 extern struct hlist_nulls_head *nf_conntrack_hash;
323 extern unsigned int nf_conntrack_htable_size;
324 extern seqcount_spinlock_t nf_conntrack_generation;
325 extern unsigned int nf_conntrack_max;
327 /* must be called with rcu read lock held */
329 nf_conntrack_get_ht(struct hlist_nulls_head **hash, unsigned int *hsize)
331 struct hlist_nulls_head *hptr;
332 unsigned int sequence, hsz;
335 sequence = read_seqcount_begin(&nf_conntrack_generation);
336 hsz = nf_conntrack_htable_size;
337 hptr = nf_conntrack_hash;
338 } while (read_seqcount_retry(&nf_conntrack_generation, sequence));
344 struct nf_conn *nf_ct_tmpl_alloc(struct net *net,
345 const struct nf_conntrack_zone *zone,
347 void nf_ct_tmpl_free(struct nf_conn *tmpl);
349 u32 nf_ct_get_id(const struct nf_conn *ct);
350 u32 nf_conntrack_count(const struct net *net);
353 nf_ct_set(struct sk_buff *skb, struct nf_conn *ct, enum ip_conntrack_info info)
355 skb_set_nfct(skb, (unsigned long)ct | info);
358 extern unsigned int nf_conntrack_net_id;
360 static inline struct nf_conntrack_net *nf_ct_pernet(const struct net *net)
362 return net_generic(net, nf_conntrack_net_id);
365 #define NF_CT_STAT_INC(net, count) __this_cpu_inc((net)->ct.stat->count)
366 #define NF_CT_STAT_INC_ATOMIC(net, count) this_cpu_inc((net)->ct.stat->count)
367 #define NF_CT_STAT_ADD_ATOMIC(net, count, v) this_cpu_add((net)->ct.stat->count, (v))
369 #define MODULE_ALIAS_NFCT_HELPER(helper) \
370 MODULE_ALIAS("nfct-helper-" helper)
372 #endif /* _NF_CONNTRACK_H */