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_ecache { 47 struct delayed_work dwork; 48 spinlock_t dying_lock; 49 struct hlist_nulls_head dying_list; 50 }; 51 52 struct nf_conntrack_net { 53 /* only used when new connection is allocated: */ 54 atomic_t count; 55 unsigned int expect_count; 56 u8 sysctl_auto_assign_helper; 57 bool auto_assign_helper_warned; 58 59 /* only used from work queues, configuration plane, and so on: */ 60 unsigned int users4; 61 unsigned int users6; 62 unsigned int users_bridge; 63 #ifdef CONFIG_SYSCTL 64 struct ctl_table_header *sysctl_header; 65 #endif 66 #ifdef CONFIG_NF_CONNTRACK_EVENTS 67 struct nf_conntrack_net_ecache ecache; 68 #endif 69 }; 70 71 #include <linux/types.h> 72 #include <linux/skbuff.h> 73 74 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h> 75 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h> 76 77 struct nf_conn { 78 /* Usage count in here is 1 for hash table, 1 per skb, 79 * plus 1 for any connection(s) we are `master' for 80 * 81 * Hint, SKB address this struct and refcnt via skb->_nfct and 82 * helpers nf_conntrack_get() and nf_conntrack_put(). 83 * Helper nf_ct_put() equals nf_conntrack_put() by dec refcnt, 84 * except that the latter uses internal indirection and does not 85 * result in a conntrack module dependency. 86 * beware nf_ct_get() is different and don't inc refcnt. 87 */ 88 struct nf_conntrack ct_general; 89 90 spinlock_t lock; 91 /* jiffies32 when this ct is considered dead */ 92 u32 timeout; 93 94 #ifdef CONFIG_NF_CONNTRACK_ZONES 95 struct nf_conntrack_zone zone; 96 #endif 97 /* XXX should I move this to the tail ? - Y.K */ 98 /* These are my tuples; original and reply */ 99 struct nf_conntrack_tuple_hash tuplehash[IP_CT_DIR_MAX]; 100 101 /* Have we seen traffic both ways yet? (bitset) */ 102 unsigned long status; 103 104 possible_net_t ct_net; 105 106 #if IS_ENABLED(CONFIG_NF_NAT) 107 struct hlist_node nat_bysource; 108 #endif 109 /* all members below initialized via memset */ 110 struct { } __nfct_init_offset; 111 112 /* If we were expected by an expectation, this will be it */ 113 struct nf_conn *master; 114 115 #if defined(CONFIG_NF_CONNTRACK_MARK) 116 u_int32_t mark; 117 #endif 118 119 #ifdef CONFIG_NF_CONNTRACK_SECMARK 120 u_int32_t secmark; 121 #endif 122 123 /* Extensions */ 124 struct nf_ct_ext *ext; 125 126 /* Storage reserved for other modules, must be the last member */ 127 union nf_conntrack_proto proto; 128 }; 129 130 static inline struct nf_conn * 131 nf_ct_tuplehash_to_ctrack(const struct nf_conntrack_tuple_hash *hash) 132 { 133 return container_of(hash, struct nf_conn, 134 tuplehash[hash->tuple.dst.dir]); 135 } 136 137 static inline u_int16_t nf_ct_l3num(const struct nf_conn *ct) 138 { 139 return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num; 140 } 141 142 static inline u_int8_t nf_ct_protonum(const struct nf_conn *ct) 143 { 144 return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum; 145 } 146 147 #define nf_ct_tuple(ct, dir) (&(ct)->tuplehash[dir].tuple) 148 149 /* get master conntrack via master expectation */ 150 #define master_ct(conntr) (conntr->master) 151 152 extern struct net init_net; 153 154 static inline struct net *nf_ct_net(const struct nf_conn *ct) 155 { 156 return read_pnet(&ct->ct_net); 157 } 158 159 /* Alter reply tuple (maybe alter helper). */ 160 void nf_conntrack_alter_reply(struct nf_conn *ct, 161 const struct nf_conntrack_tuple *newreply); 162 163 /* Is this tuple taken? (ignoring any belonging to the given 164 conntrack). */ 165 int nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, 166 const struct nf_conn *ignored_conntrack); 167 168 /* Return conntrack_info and tuple hash for given skb. */ 169 static inline struct nf_conn * 170 nf_ct_get(const struct sk_buff *skb, enum ip_conntrack_info *ctinfo) 171 { 172 unsigned long nfct = skb_get_nfct(skb); 173 174 *ctinfo = nfct & NFCT_INFOMASK; 175 return (struct nf_conn *)(nfct & NFCT_PTRMASK); 176 } 177 178 void nf_ct_destroy(struct nf_conntrack *nfct); 179 180 /* decrement reference count on a conntrack */ 181 static inline void nf_ct_put(struct nf_conn *ct) 182 { 183 if (ct && refcount_dec_and_test(&ct->ct_general.use)) 184 nf_ct_destroy(&ct->ct_general); 185 } 186 187 /* Protocol module loading */ 188 int nf_ct_l3proto_try_module_get(unsigned short l3proto); 189 void nf_ct_l3proto_module_put(unsigned short l3proto); 190 191 /* load module; enable/disable conntrack in this namespace */ 192 int nf_ct_netns_get(struct net *net, u8 nfproto); 193 void nf_ct_netns_put(struct net *net, u8 nfproto); 194 195 /* 196 * Allocate a hashtable of hlist_head (if nulls == 0), 197 * or hlist_nulls_head (if nulls == 1) 198 */ 199 void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls); 200 201 int nf_conntrack_hash_check_insert(struct nf_conn *ct); 202 bool nf_ct_delete(struct nf_conn *ct, u32 pid, int report); 203 204 bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, 205 u_int16_t l3num, struct net *net, 206 struct nf_conntrack_tuple *tuple); 207 208 void __nf_ct_refresh_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo, 209 const struct sk_buff *skb, 210 u32 extra_jiffies, bool do_acct); 211 212 /* Refresh conntrack for this many jiffies and do accounting */ 213 static inline void nf_ct_refresh_acct(struct nf_conn *ct, 214 enum ip_conntrack_info ctinfo, 215 const struct sk_buff *skb, 216 u32 extra_jiffies) 217 { 218 __nf_ct_refresh_acct(ct, ctinfo, skb, extra_jiffies, true); 219 } 220 221 /* Refresh conntrack for this many jiffies */ 222 static inline void nf_ct_refresh(struct nf_conn *ct, 223 const struct sk_buff *skb, 224 u32 extra_jiffies) 225 { 226 __nf_ct_refresh_acct(ct, 0, skb, extra_jiffies, false); 227 } 228 229 /* kill conntrack and do accounting */ 230 bool nf_ct_kill_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo, 231 const struct sk_buff *skb); 232 233 /* kill conntrack without accounting */ 234 static inline bool nf_ct_kill(struct nf_conn *ct) 235 { 236 return nf_ct_delete(ct, 0, 0); 237 } 238 239 struct nf_ct_iter_data { 240 struct net *net; 241 void *data; 242 u32 portid; 243 int report; 244 }; 245 246 /* Iterate over all conntracks: if iter returns true, it's deleted. */ 247 void nf_ct_iterate_cleanup_net(int (*iter)(struct nf_conn *i, void *data), 248 const struct nf_ct_iter_data *iter_data); 249 250 /* also set unconfirmed conntracks as dying. Only use in module exit path. */ 251 void nf_ct_iterate_destroy(int (*iter)(struct nf_conn *i, void *data), 252 void *data); 253 254 struct nf_conntrack_zone; 255 256 void nf_conntrack_free(struct nf_conn *ct); 257 struct nf_conn *nf_conntrack_alloc(struct net *net, 258 const struct nf_conntrack_zone *zone, 259 const struct nf_conntrack_tuple *orig, 260 const struct nf_conntrack_tuple *repl, 261 gfp_t gfp); 262 263 static inline int nf_ct_is_template(const struct nf_conn *ct) 264 { 265 return test_bit(IPS_TEMPLATE_BIT, &ct->status); 266 } 267 268 /* It's confirmed if it is, or has been in the hash table. */ 269 static inline int nf_ct_is_confirmed(const struct nf_conn *ct) 270 { 271 return test_bit(IPS_CONFIRMED_BIT, &ct->status); 272 } 273 274 static inline int nf_ct_is_dying(const struct nf_conn *ct) 275 { 276 return test_bit(IPS_DYING_BIT, &ct->status); 277 } 278 279 /* Packet is received from loopback */ 280 static inline bool nf_is_loopback_packet(const struct sk_buff *skb) 281 { 282 return skb->dev && skb->skb_iif && skb->dev->flags & IFF_LOOPBACK; 283 } 284 285 #define nfct_time_stamp ((u32)(jiffies)) 286 287 /* jiffies until ct expires, 0 if already expired */ 288 static inline unsigned long nf_ct_expires(const struct nf_conn *ct) 289 { 290 s32 timeout = READ_ONCE(ct->timeout) - nfct_time_stamp; 291 292 return max(timeout, 0); 293 } 294 295 static inline bool nf_ct_is_expired(const struct nf_conn *ct) 296 { 297 return (__s32)(READ_ONCE(ct->timeout) - nfct_time_stamp) <= 0; 298 } 299 300 /* use after obtaining a reference count */ 301 static inline bool nf_ct_should_gc(const struct nf_conn *ct) 302 { 303 return nf_ct_is_expired(ct) && nf_ct_is_confirmed(ct) && 304 !nf_ct_is_dying(ct); 305 } 306 307 #define NF_CT_DAY (86400 * HZ) 308 309 /* Set an arbitrary timeout large enough not to ever expire, this save 310 * us a check for the IPS_OFFLOAD_BIT from the packet path via 311 * nf_ct_is_expired(). 312 */ 313 static inline void nf_ct_offload_timeout(struct nf_conn *ct) 314 { 315 if (nf_ct_expires(ct) < NF_CT_DAY / 2) 316 WRITE_ONCE(ct->timeout, nfct_time_stamp + NF_CT_DAY); 317 } 318 319 struct kernel_param; 320 321 int nf_conntrack_set_hashsize(const char *val, const struct kernel_param *kp); 322 int nf_conntrack_hash_resize(unsigned int hashsize); 323 324 extern struct hlist_nulls_head *nf_conntrack_hash; 325 extern unsigned int nf_conntrack_htable_size; 326 extern seqcount_spinlock_t nf_conntrack_generation; 327 extern unsigned int nf_conntrack_max; 328 329 /* must be called with rcu read lock held */ 330 static inline void 331 nf_conntrack_get_ht(struct hlist_nulls_head **hash, unsigned int *hsize) 332 { 333 struct hlist_nulls_head *hptr; 334 unsigned int sequence, hsz; 335 336 do { 337 sequence = read_seqcount_begin(&nf_conntrack_generation); 338 hsz = nf_conntrack_htable_size; 339 hptr = nf_conntrack_hash; 340 } while (read_seqcount_retry(&nf_conntrack_generation, sequence)); 341 342 *hash = hptr; 343 *hsize = hsz; 344 } 345 346 struct nf_conn *nf_ct_tmpl_alloc(struct net *net, 347 const struct nf_conntrack_zone *zone, 348 gfp_t flags); 349 void nf_ct_tmpl_free(struct nf_conn *tmpl); 350 351 u32 nf_ct_get_id(const struct nf_conn *ct); 352 u32 nf_conntrack_count(const struct net *net); 353 354 static inline void 355 nf_ct_set(struct sk_buff *skb, struct nf_conn *ct, enum ip_conntrack_info info) 356 { 357 skb_set_nfct(skb, (unsigned long)ct | info); 358 } 359 360 extern unsigned int nf_conntrack_net_id; 361 362 static inline struct nf_conntrack_net *nf_ct_pernet(const struct net *net) 363 { 364 return net_generic(net, nf_conntrack_net_id); 365 } 366 367 #define NF_CT_STAT_INC(net, count) __this_cpu_inc((net)->ct.stat->count) 368 #define NF_CT_STAT_INC_ATOMIC(net, count) this_cpu_inc((net)->ct.stat->count) 369 #define NF_CT_STAT_ADD_ATOMIC(net, count, v) this_cpu_add((net)->ct.stat->count, (v)) 370 371 #define MODULE_ALIAS_NFCT_HELPER(helper) \ 372 MODULE_ALIAS("nfct-helper-" helper) 373 374 #endif /* _NF_CONNTRACK_H */ 375