xref: /linux/include/net/netfilter/nf_conntrack.h (revision d00453b6e3a3d2340b88c5292c3c5b5f9c4ece75)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
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
5  * extension.
6  *
7  * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
8  *	- generalize L3 protocol dependent part.
9  *
10  * Derived from include/linux/netfiter_ipv4/ip_conntrack.h
11  */
12 
13 #ifndef _NF_CONNTRACK_H
14 #define _NF_CONNTRACK_H
15 
16 #include <linux/bitops.h>
17 #include <linux/compiler.h>
18 
19 #include <net/netns/generic.h>
20 #include <linux/netfilter/nf_conntrack_common.h>
21 #include <linux/netfilter/nf_conntrack_tcp.h>
22 #include <linux/netfilter/nf_conntrack_sctp.h>
23 #include <linux/netfilter/nf_conntrack_proto_gre.h>
24 
25 #include <net/netfilter/nf_conntrack_tuple.h>
26 
27 struct nf_ct_udp {
28 	unsigned long	stream_ts;
29 };
30 
31 /* per conntrack: protocol private data */
32 union nf_conntrack_proto {
33 	/* insert conntrack proto private data here */
34 	struct ip_ct_sctp sctp;
35 	struct ip_ct_tcp tcp;
36 	struct nf_ct_udp udp;
37 	struct nf_ct_gre gre;
38 	unsigned int tmpl_padto;
39 };
40 
41 union nf_conntrack_expect_proto {
42 	/* insert expect proto private data here */
43 };
44 
45 struct nf_conntrack_net_ecache {
46 	struct delayed_work dwork;
47 	spinlock_t dying_lock;
48 	struct hlist_nulls_head dying_list;
49 };
50 
51 struct nf_conntrack_net {
52 	/* only used when new connection is allocated: */
53 	atomic_t count;
54 	unsigned int expect_count;
55 
56 	/* only used from work queues, configuration plane, and so on: */
57 	unsigned int users4;
58 	unsigned int users6;
59 	unsigned int users_bridge;
60 #ifdef CONFIG_SYSCTL
61 	struct ctl_table_header	*sysctl_header;
62 #endif
63 #ifdef CONFIG_NF_CONNTRACK_EVENTS
64 	struct nf_conntrack_net_ecache ecache;
65 #endif
66 };
67 
68 #include <linux/types.h>
69 #include <linux/skbuff.h>
70 
71 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
72 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
73 
74 struct nf_conn {
75 	/* Usage count in here is 1 for hash table, 1 per skb,
76 	 * plus 1 for any connection(s) we are `master' for
77 	 *
78 	 * Hint, SKB address this struct and refcnt via skb->_nfct and
79 	 * helpers nf_conntrack_get() and nf_conntrack_put().
80 	 * Helper nf_ct_put() equals nf_conntrack_put() by dec refcnt,
81 	 * except that the latter uses internal indirection and does not
82 	 * result in a conntrack module dependency.
83 	 * beware nf_ct_get() is different and don't inc refcnt.
84 	 */
85 	struct nf_conntrack ct_general;
86 
87 	spinlock_t	lock;
88 	/* jiffies32 when this ct is considered dead */
89 	u32 timeout;
90 
91 #ifdef CONFIG_NF_CONNTRACK_ZONES
92 	struct nf_conntrack_zone zone;
93 #endif
94 	/* XXX should I move this to the tail ? - Y.K */
95 	/* These are my tuples; original and reply */
96 	struct nf_conntrack_tuple_hash tuplehash[IP_CT_DIR_MAX];
97 
98 	/* Have we seen traffic both ways yet? (bitset) */
99 	unsigned long status;
100 
101 	possible_net_t ct_net;
102 
103 #if IS_ENABLED(CONFIG_NF_NAT)
104 	struct hlist_node	nat_bysource;
105 #endif
106 	/* all members below initialized via memset */
107 	struct { } __nfct_init_offset;
108 
109 	/* If we were expected by an expectation, this will be it */
110 	struct nf_conn *master;
111 
112 #if defined(CONFIG_NF_CONNTRACK_MARK)
113 	u_int32_t mark;
114 #endif
115 
116 #ifdef CONFIG_NF_CONNTRACK_SECMARK
117 	u_int32_t secmark;
118 #endif
119 
120 	/* Extensions */
121 	struct nf_ct_ext *ext;
122 
123 	/* Storage reserved for other modules, must be the last member */
124 	union nf_conntrack_proto proto;
125 };
126 
127 static inline struct nf_conn *
128 nf_ct_to_nf_conn(const struct nf_conntrack *nfct)
129 {
130 	return container_of(nfct, struct nf_conn, ct_general);
131 }
132 
133 static inline struct nf_conn *
134 nf_ct_tuplehash_to_ctrack(const struct nf_conntrack_tuple_hash *hash)
135 {
136 	return container_of(hash, struct nf_conn,
137 			    tuplehash[hash->tuple.dst.dir]);
138 }
139 
140 static inline u_int16_t nf_ct_l3num(const struct nf_conn *ct)
141 {
142 	return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
143 }
144 
145 static inline u_int8_t nf_ct_protonum(const struct nf_conn *ct)
146 {
147 	return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
148 }
149 
150 #define nf_ct_tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
151 
152 /* get master conntrack via master expectation */
153 #define master_ct(conntr) (conntr->master)
154 
155 extern struct net init_net;
156 
157 static inline struct net *nf_ct_net(const struct nf_conn *ct)
158 {
159 	return read_pnet(&ct->ct_net);
160 }
161 
162 /* Is this tuple taken? (ignoring any belonging to the given
163    conntrack). */
164 int nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple,
165 			     const struct nf_conn *ignored_conntrack);
166 
167 /* Return conntrack_info and tuple hash for given skb. */
168 static inline struct nf_conn *
169 nf_ct_get(const struct sk_buff *skb, enum ip_conntrack_info *ctinfo)
170 {
171 	unsigned long nfct = skb_get_nfct(skb);
172 
173 	*ctinfo = nfct & NFCT_INFOMASK;
174 	return (struct nf_conn *)(nfct & NFCT_PTRMASK);
175 }
176 
177 void nf_ct_destroy(struct nf_conntrack *nfct);
178 
179 void nf_conntrack_tcp_set_closing(struct nf_conn *ct);
180 
181 /* decrement reference count on a conntrack */
182 static inline void nf_ct_put(struct nf_conn *ct)
183 {
184 	if (ct && refcount_dec_and_test(&ct->ct_general.use))
185 		nf_ct_destroy(&ct->ct_general);
186 }
187 
188 /* load module; enable/disable conntrack in this namespace */
189 int nf_ct_netns_get(struct net *net, u8 nfproto);
190 void nf_ct_netns_put(struct net *net, u8 nfproto);
191 
192 /*
193  * Allocate a hashtable of hlist_head (if nulls == 0),
194  * or hlist_nulls_head (if nulls == 1)
195  */
196 void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls);
197 
198 int nf_conntrack_hash_check_insert(struct nf_conn *ct);
199 bool nf_ct_delete(struct nf_conn *ct, u32 pid, int report);
200 
201 bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff,
202 		       u_int16_t l3num, struct net *net,
203 		       struct nf_conntrack_tuple *tuple);
204 
205 void __nf_ct_refresh_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
206 			  u32 extra_jiffies, unsigned int bytes);
207 
208 /* Refresh conntrack for this many jiffies and do accounting */
209 static inline void nf_ct_refresh_acct(struct nf_conn *ct,
210 				      enum ip_conntrack_info ctinfo,
211 				      const struct sk_buff *skb,
212 				      u32 extra_jiffies)
213 {
214 	__nf_ct_refresh_acct(ct, ctinfo, extra_jiffies, skb->len);
215 }
216 
217 /* Refresh conntrack for this many jiffies */
218 static inline void nf_ct_refresh(struct nf_conn *ct,
219 				 u32 extra_jiffies)
220 {
221 	__nf_ct_refresh_acct(ct, 0, extra_jiffies, 0);
222 }
223 
224 /* kill conntrack and do accounting */
225 bool nf_ct_kill_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
226 		     const struct sk_buff *skb);
227 
228 /* kill conntrack without accounting */
229 static inline bool nf_ct_kill(struct nf_conn *ct)
230 {
231 	return nf_ct_delete(ct, 0, 0);
232 }
233 
234 struct nf_ct_iter_data {
235 	struct net *net;
236 	void *data;
237 	u32 portid;
238 	int report;
239 };
240 
241 /* Iterate over all conntracks: if iter returns true, it's deleted. */
242 void nf_ct_iterate_cleanup_net(int (*iter)(struct nf_conn *i, void *data),
243 			       const struct nf_ct_iter_data *iter_data);
244 
245 /* also set unconfirmed conntracks as dying. Only use in module exit path. */
246 void nf_ct_iterate_destroy(int (*iter)(struct nf_conn *i, void *data),
247 			   void *data);
248 
249 struct nf_conntrack_zone;
250 
251 void nf_conntrack_free(struct nf_conn *ct);
252 struct nf_conn *nf_conntrack_alloc(struct net *net,
253 				   const struct nf_conntrack_zone *zone,
254 				   const struct nf_conntrack_tuple *orig,
255 				   const struct nf_conntrack_tuple *repl,
256 				   gfp_t gfp);
257 
258 static inline int nf_ct_is_template(const struct nf_conn *ct)
259 {
260 	return test_bit(IPS_TEMPLATE_BIT, &ct->status);
261 }
262 
263 /* It's confirmed if it is, or has been in the hash table. */
264 static inline int nf_ct_is_confirmed(const struct nf_conn *ct)
265 {
266 	return test_bit(IPS_CONFIRMED_BIT, &ct->status);
267 }
268 
269 static inline int nf_ct_is_dying(const struct nf_conn *ct)
270 {
271 	return test_bit(IPS_DYING_BIT, &ct->status);
272 }
273 
274 /* Packet is received from loopback */
275 static inline bool nf_is_loopback_packet(const struct sk_buff *skb)
276 {
277 	return skb->dev && skb->skb_iif && skb->dev->flags & IFF_LOOPBACK;
278 }
279 
280 static inline void nf_conntrack_alter_reply(struct nf_conn *ct,
281 					    const struct nf_conntrack_tuple *newreply)
282 {
283 	/* Must be unconfirmed, so not in hash table yet */
284 	if (WARN_ON(nf_ct_is_confirmed(ct)))
285 		return;
286 
287 	ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply;
288 }
289 
290 #define nfct_time_stamp ((u32)(jiffies))
291 
292 /* jiffies until ct expires, 0 if already expired */
293 static inline unsigned long nf_ct_expires(const struct nf_conn *ct)
294 {
295 	s32 timeout = READ_ONCE(ct->timeout) - nfct_time_stamp;
296 
297 	return max(timeout, 0);
298 }
299 
300 static inline bool nf_ct_is_expired(const struct nf_conn *ct)
301 {
302 	return (__s32)(READ_ONCE(ct->timeout) - nfct_time_stamp) <= 0;
303 }
304 
305 /* use after obtaining a reference count */
306 static inline bool nf_ct_should_gc(const struct nf_conn *ct)
307 {
308 	if (!nf_ct_is_confirmed(ct))
309 		return false;
310 
311 	/* load ct->timeout after is_confirmed() test.
312 	 * Pairs with __nf_conntrack_confirm() which:
313 	 * 1. Increases ct->timeout value
314 	 * 2. Inserts ct into rcu hlist
315 	 * 3. Sets the confirmed bit
316 	 * 4. Unlocks the hlist lock
317 	 */
318 	smp_acquire__after_ctrl_dep();
319 
320 	return nf_ct_is_expired(ct) && !nf_ct_is_dying(ct);
321 }
322 
323 #define	NF_CT_DAY	(86400 * HZ)
324 
325 struct kernel_param;
326 
327 int nf_conntrack_set_hashsize(const char *val, const struct kernel_param *kp);
328 int nf_conntrack_hash_resize(unsigned int hashsize);
329 
330 extern struct hlist_nulls_head *nf_conntrack_hash;
331 extern unsigned int nf_conntrack_htable_size;
332 extern seqcount_spinlock_t nf_conntrack_generation;
333 extern unsigned int nf_conntrack_max;
334 
335 /* must be called with rcu read lock held */
336 static inline void
337 nf_conntrack_get_ht(struct hlist_nulls_head **hash, unsigned int *hsize)
338 {
339 	struct hlist_nulls_head *hptr;
340 	unsigned int sequence, hsz;
341 
342 	do {
343 		sequence = read_seqcount_begin(&nf_conntrack_generation);
344 		hsz = nf_conntrack_htable_size;
345 		hptr = nf_conntrack_hash;
346 	} while (read_seqcount_retry(&nf_conntrack_generation, sequence));
347 
348 	*hash = hptr;
349 	*hsize = hsz;
350 }
351 
352 struct nf_conn *nf_ct_tmpl_alloc(struct net *net,
353 				 const struct nf_conntrack_zone *zone,
354 				 gfp_t flags);
355 void nf_ct_tmpl_free(struct nf_conn *tmpl);
356 
357 u32 nf_ct_get_id(const struct nf_conn *ct);
358 u32 nf_conntrack_count(const struct net *net);
359 
360 static inline void
361 nf_ct_set(struct sk_buff *skb, struct nf_conn *ct, enum ip_conntrack_info info)
362 {
363 	skb_set_nfct(skb, (unsigned long)ct | info);
364 }
365 
366 extern unsigned int nf_conntrack_net_id;
367 
368 static inline struct nf_conntrack_net *nf_ct_pernet(const struct net *net)
369 {
370 	return net_generic(net, nf_conntrack_net_id);
371 }
372 
373 int nf_ct_skb_network_trim(struct sk_buff *skb, int family);
374 int nf_ct_handle_fragments(struct net *net, struct sk_buff *skb,
375 			   u16 zone, u8 family, u8 *proto, u16 *mru);
376 
377 #define NF_CT_STAT_INC(net, count)	  __this_cpu_inc((net)->ct.stat->count)
378 #define NF_CT_STAT_INC_ATOMIC(net, count) this_cpu_inc((net)->ct.stat->count)
379 #define NF_CT_STAT_ADD_ATOMIC(net, count, v) this_cpu_add((net)->ct.stat->count, (v))
380 
381 #define MODULE_ALIAS_NFCT_HELPER(helper) \
382         MODULE_ALIAS("nfct-helper-" helper)
383 
384 #endif /* _NF_CONNTRACK_H */
385