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