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 <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>
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 nf_ct_dccp dccp;
35 	struct ip_ct_sctp sctp;
36 	struct ip_ct_tcp tcp;
37 	struct nf_ct_udp udp;
38 	struct nf_ct_gre gre;
39 	unsigned int tmpl_padto;
40 };
41 
42 union nf_conntrack_expect_proto {
43 	/* insert expect proto private data here */
44 };
45 
46 struct nf_conntrack_net {
47 	/* only used when new connection is allocated: */
48 	atomic_t count;
49 	unsigned int expect_count;
50 	u8 sysctl_auto_assign_helper;
51 	bool auto_assign_helper_warned;
52 
53 	/* only used from work queues, configuration plane, and so on: */
54 	unsigned int users4;
55 	unsigned int users6;
56 	unsigned int users_bridge;
57 #ifdef CONFIG_SYSCTL
58 	struct ctl_table_header	*sysctl_header;
59 #endif
60 #ifdef CONFIG_NF_CONNTRACK_EVENTS
61 	struct delayed_work ecache_dwork;
62 	struct netns_ct *ct_net;
63 #endif
64 };
65 
66 #include <linux/types.h>
67 #include <linux/skbuff.h>
68 
69 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
70 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
71 
72 struct nf_conn {
73 	/* Usage count in here is 1 for hash table, 1 per skb,
74 	 * plus 1 for any connection(s) we are `master' for
75 	 *
76 	 * Hint, SKB address this struct and refcnt via skb->_nfct and
77 	 * helpers nf_conntrack_get() and nf_conntrack_put().
78 	 * Helper nf_ct_put() equals nf_conntrack_put() by dec refcnt,
79 	 * beware nf_ct_get() is different and don't inc refcnt.
80 	 */
81 	struct nf_conntrack ct_general;
82 
83 	spinlock_t	lock;
84 	/* jiffies32 when this ct is considered dead */
85 	u32 timeout;
86 
87 #ifdef CONFIG_NF_CONNTRACK_ZONES
88 	struct nf_conntrack_zone zone;
89 #endif
90 	/* XXX should I move this to the tail ? - Y.K */
91 	/* These are my tuples; original and reply */
92 	struct nf_conntrack_tuple_hash tuplehash[IP_CT_DIR_MAX];
93 
94 	/* Have we seen traffic both ways yet? (bitset) */
95 	unsigned long status;
96 
97 	u16		cpu;
98 	possible_net_t ct_net;
99 
100 #if IS_ENABLED(CONFIG_NF_NAT)
101 	struct hlist_node	nat_bysource;
102 #endif
103 	/* all members below initialized via memset */
104 	struct { } __nfct_init_offset;
105 
106 	/* If we were expected by an expectation, this will be it */
107 	struct nf_conn *master;
108 
109 #if defined(CONFIG_NF_CONNTRACK_MARK)
110 	u_int32_t mark;
111 #endif
112 
113 #ifdef CONFIG_NF_CONNTRACK_SECMARK
114 	u_int32_t secmark;
115 #endif
116 
117 	/* Extensions */
118 	struct nf_ct_ext *ext;
119 
120 	/* Storage reserved for other modules, must be the last member */
121 	union nf_conntrack_proto proto;
122 };
123 
124 static inline struct nf_conn *
nf_ct_tuplehash_to_ctrack(const struct nf_conntrack_tuple_hash * hash)125 nf_ct_tuplehash_to_ctrack(const struct nf_conntrack_tuple_hash *hash)
126 {
127 	return container_of(hash, struct nf_conn,
128 			    tuplehash[hash->tuple.dst.dir]);
129 }
130 
nf_ct_l3num(const struct nf_conn * ct)131 static inline u_int16_t nf_ct_l3num(const struct nf_conn *ct)
132 {
133 	return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
134 }
135 
nf_ct_protonum(const struct nf_conn * ct)136 static inline u_int8_t nf_ct_protonum(const struct nf_conn *ct)
137 {
138 	return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
139 }
140 
141 #define nf_ct_tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
142 
143 /* get master conntrack via master expectation */
144 #define master_ct(conntr) (conntr->master)
145 
146 extern struct net init_net;
147 
nf_ct_net(const struct nf_conn * ct)148 static inline struct net *nf_ct_net(const struct nf_conn *ct)
149 {
150 	return read_pnet(&ct->ct_net);
151 }
152 
153 /* Alter reply tuple (maybe alter helper). */
154 void nf_conntrack_alter_reply(struct nf_conn *ct,
155 			      const struct nf_conntrack_tuple *newreply);
156 
157 /* Is this tuple taken? (ignoring any belonging to the given
158    conntrack). */
159 int nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple,
160 			     const struct nf_conn *ignored_conntrack);
161 
162 /* Return conntrack_info and tuple hash for given skb. */
163 static inline struct nf_conn *
nf_ct_get(const struct sk_buff * skb,enum ip_conntrack_info * ctinfo)164 nf_ct_get(const struct sk_buff *skb, enum ip_conntrack_info *ctinfo)
165 {
166 	unsigned long nfct = skb_get_nfct(skb);
167 
168 	*ctinfo = nfct & NFCT_INFOMASK;
169 	return (struct nf_conn *)(nfct & NFCT_PTRMASK);
170 }
171 
172 /* decrement reference count on a conntrack */
nf_ct_put(struct nf_conn * ct)173 static inline void nf_ct_put(struct nf_conn *ct)
174 {
175 	WARN_ON(!ct);
176 	nf_conntrack_put(&ct->ct_general);
177 }
178 
179 /* Protocol module loading */
180 int nf_ct_l3proto_try_module_get(unsigned short l3proto);
181 void nf_ct_l3proto_module_put(unsigned short l3proto);
182 
183 /* load module; enable/disable conntrack in this namespace */
184 int nf_ct_netns_get(struct net *net, u8 nfproto);
185 void nf_ct_netns_put(struct net *net, u8 nfproto);
186 
187 /*
188  * Allocate a hashtable of hlist_head (if nulls == 0),
189  * or hlist_nulls_head (if nulls == 1)
190  */
191 void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls);
192 
193 int nf_conntrack_hash_check_insert(struct nf_conn *ct);
194 bool nf_ct_delete(struct nf_conn *ct, u32 pid, int report);
195 
196 bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff,
197 		       u_int16_t l3num, struct net *net,
198 		       struct nf_conntrack_tuple *tuple);
199 
200 void __nf_ct_refresh_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
201 			  const struct sk_buff *skb,
202 			  u32 extra_jiffies, bool do_acct);
203 
204 /* Refresh conntrack for this many jiffies and do accounting */
nf_ct_refresh_acct(struct nf_conn * ct,enum ip_conntrack_info ctinfo,const struct sk_buff * skb,u32 extra_jiffies)205 static inline void nf_ct_refresh_acct(struct nf_conn *ct,
206 				      enum ip_conntrack_info ctinfo,
207 				      const struct sk_buff *skb,
208 				      u32 extra_jiffies)
209 {
210 	__nf_ct_refresh_acct(ct, ctinfo, skb, extra_jiffies, true);
211 }
212 
213 /* Refresh conntrack for this many jiffies */
nf_ct_refresh(struct nf_conn * ct,const struct sk_buff * skb,u32 extra_jiffies)214 static inline void nf_ct_refresh(struct nf_conn *ct,
215 				 const struct sk_buff *skb,
216 				 u32 extra_jiffies)
217 {
218 	__nf_ct_refresh_acct(ct, 0, skb, extra_jiffies, false);
219 }
220 
221 /* kill conntrack and do accounting */
222 bool nf_ct_kill_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
223 		     const struct sk_buff *skb);
224 
225 /* kill conntrack without accounting */
nf_ct_kill(struct nf_conn * ct)226 static inline bool nf_ct_kill(struct nf_conn *ct)
227 {
228 	return nf_ct_delete(ct, 0, 0);
229 }
230 
231 /* Set all unconfirmed conntrack as dying */
232 void nf_ct_unconfirmed_destroy(struct net *);
233 
234 /* Iterate over all conntracks: if iter returns true, it's deleted. */
235 void nf_ct_iterate_cleanup_net(struct net *net,
236 			       int (*iter)(struct nf_conn *i, void *data),
237 			       void *data, u32 portid, int report);
238 
239 /* also set unconfirmed conntracks as dying. Only use in module exit path. */
240 void nf_ct_iterate_destroy(int (*iter)(struct nf_conn *i, void *data),
241 			   void *data);
242 
243 struct nf_conntrack_zone;
244 
245 void nf_conntrack_free(struct nf_conn *ct);
246 struct nf_conn *nf_conntrack_alloc(struct net *net,
247 				   const struct nf_conntrack_zone *zone,
248 				   const struct nf_conntrack_tuple *orig,
249 				   const struct nf_conntrack_tuple *repl,
250 				   gfp_t gfp);
251 
nf_ct_is_template(const struct nf_conn * ct)252 static inline int nf_ct_is_template(const struct nf_conn *ct)
253 {
254 	return test_bit(IPS_TEMPLATE_BIT, &ct->status);
255 }
256 
257 /* It's confirmed if it is, or has been in the hash table. */
nf_ct_is_confirmed(const struct nf_conn * ct)258 static inline int nf_ct_is_confirmed(const struct nf_conn *ct)
259 {
260 	return test_bit(IPS_CONFIRMED_BIT, &ct->status);
261 }
262 
nf_ct_is_dying(const struct nf_conn * ct)263 static inline int nf_ct_is_dying(const struct nf_conn *ct)
264 {
265 	return test_bit(IPS_DYING_BIT, &ct->status);
266 }
267 
268 /* Packet is received from loopback */
nf_is_loopback_packet(const struct sk_buff * skb)269 static inline bool nf_is_loopback_packet(const struct sk_buff *skb)
270 {
271 	return skb->dev && skb->skb_iif && skb->dev->flags & IFF_LOOPBACK;
272 }
273 
274 #define nfct_time_stamp ((u32)(jiffies))
275 
276 /* jiffies until ct expires, 0 if already expired */
nf_ct_expires(const struct nf_conn * ct)277 static inline unsigned long nf_ct_expires(const struct nf_conn *ct)
278 {
279 	s32 timeout = READ_ONCE(ct->timeout) - nfct_time_stamp;
280 
281 	return timeout > 0 ? timeout : 0;
282 }
283 
nf_ct_is_expired(const struct nf_conn * ct)284 static inline bool nf_ct_is_expired(const struct nf_conn *ct)
285 {
286 	return (__s32)(READ_ONCE(ct->timeout) - nfct_time_stamp) <= 0;
287 }
288 
289 /* use after obtaining a reference count */
nf_ct_should_gc(const struct nf_conn * ct)290 static inline bool nf_ct_should_gc(const struct nf_conn *ct)
291 {
292 	return nf_ct_is_expired(ct) && nf_ct_is_confirmed(ct) &&
293 	       !nf_ct_is_dying(ct);
294 }
295 
296 #define	NF_CT_DAY	(86400 * HZ)
297 
298 /* Set an arbitrary timeout large enough not to ever expire, this save
299  * us a check for the IPS_OFFLOAD_BIT from the packet path via
300  * nf_ct_is_expired().
301  */
nf_ct_offload_timeout(struct nf_conn * ct)302 static inline void nf_ct_offload_timeout(struct nf_conn *ct)
303 {
304 	if (nf_ct_expires(ct) < NF_CT_DAY / 2)
305 		WRITE_ONCE(ct->timeout, nfct_time_stamp + NF_CT_DAY);
306 }
307 
308 struct kernel_param;
309 
310 int nf_conntrack_set_hashsize(const char *val, const struct kernel_param *kp);
311 int nf_conntrack_hash_resize(unsigned int hashsize);
312 
313 extern struct hlist_nulls_head *nf_conntrack_hash;
314 extern unsigned int nf_conntrack_htable_size;
315 extern seqcount_spinlock_t nf_conntrack_generation;
316 extern unsigned int nf_conntrack_max;
317 
318 /* must be called with rcu read lock held */
319 static inline void
nf_conntrack_get_ht(struct hlist_nulls_head ** hash,unsigned int * hsize)320 nf_conntrack_get_ht(struct hlist_nulls_head **hash, unsigned int *hsize)
321 {
322 	struct hlist_nulls_head *hptr;
323 	unsigned int sequence, hsz;
324 
325 	do {
326 		sequence = read_seqcount_begin(&nf_conntrack_generation);
327 		hsz = nf_conntrack_htable_size;
328 		hptr = nf_conntrack_hash;
329 	} while (read_seqcount_retry(&nf_conntrack_generation, sequence));
330 
331 	*hash = hptr;
332 	*hsize = hsz;
333 }
334 
335 struct nf_conn *nf_ct_tmpl_alloc(struct net *net,
336 				 const struct nf_conntrack_zone *zone,
337 				 gfp_t flags);
338 void nf_ct_tmpl_free(struct nf_conn *tmpl);
339 
340 u32 nf_ct_get_id(const struct nf_conn *ct);
341 u32 nf_conntrack_count(const struct net *net);
342 
343 static inline void
nf_ct_set(struct sk_buff * skb,struct nf_conn * ct,enum ip_conntrack_info info)344 nf_ct_set(struct sk_buff *skb, struct nf_conn *ct, enum ip_conntrack_info info)
345 {
346 	skb_set_nfct(skb, (unsigned long)ct | info);
347 }
348 
349 extern unsigned int nf_conntrack_net_id;
350 
nf_ct_pernet(const struct net * net)351 static inline struct nf_conntrack_net *nf_ct_pernet(const struct net *net)
352 {
353 	return net_generic(net, nf_conntrack_net_id);
354 }
355 
356 #define NF_CT_STAT_INC(net, count)	  __this_cpu_inc((net)->ct.stat->count)
357 #define NF_CT_STAT_INC_ATOMIC(net, count) this_cpu_inc((net)->ct.stat->count)
358 #define NF_CT_STAT_ADD_ATOMIC(net, count, v) this_cpu_add((net)->ct.stat->count, (v))
359 
360 #define MODULE_ALIAS_NFCT_HELPER(helper) \
361         MODULE_ALIAS("nfct-helper-" helper)
362 
363 #endif /* _NF_CONNTRACK_H */
364