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 * except that the latter uses internal indirection and does not 80 * result in a conntrack module dependency. 81 * beware nf_ct_get() is different and don't inc refcnt. 82 */ 83 struct nf_conntrack ct_general; 84 85 spinlock_t lock; 86 /* jiffies32 when this ct is considered dead */ 87 u32 timeout; 88 89 #ifdef CONFIG_NF_CONNTRACK_ZONES 90 struct nf_conntrack_zone zone; 91 #endif 92 /* XXX should I move this to the tail ? - Y.K */ 93 /* These are my tuples; original and reply */ 94 struct nf_conntrack_tuple_hash tuplehash[IP_CT_DIR_MAX]; 95 96 /* Have we seen traffic both ways yet? (bitset) */ 97 unsigned long status; 98 99 u16 cpu; 100 possible_net_t ct_net; 101 102 #if IS_ENABLED(CONFIG_NF_NAT) 103 struct hlist_node nat_bysource; 104 #endif 105 /* all members below initialized via memset */ 106 struct { } __nfct_init_offset; 107 108 /* If we were expected by an expectation, this will be it */ 109 struct nf_conn *master; 110 111 #if defined(CONFIG_NF_CONNTRACK_MARK) 112 u_int32_t mark; 113 #endif 114 115 #ifdef CONFIG_NF_CONNTRACK_SECMARK 116 u_int32_t secmark; 117 #endif 118 119 /* Extensions */ 120 struct nf_ct_ext *ext; 121 122 /* Storage reserved for other modules, must be the last member */ 123 union nf_conntrack_proto proto; 124 }; 125 126 static inline struct nf_conn * 127 nf_ct_tuplehash_to_ctrack(const struct nf_conntrack_tuple_hash *hash) 128 { 129 return container_of(hash, struct nf_conn, 130 tuplehash[hash->tuple.dst.dir]); 131 } 132 133 static inline u_int16_t nf_ct_l3num(const struct nf_conn *ct) 134 { 135 return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num; 136 } 137 138 static inline u_int8_t nf_ct_protonum(const struct nf_conn *ct) 139 { 140 return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum; 141 } 142 143 #define nf_ct_tuple(ct, dir) (&(ct)->tuplehash[dir].tuple) 144 145 /* get master conntrack via master expectation */ 146 #define master_ct(conntr) (conntr->master) 147 148 extern struct net init_net; 149 150 static inline struct net *nf_ct_net(const struct nf_conn *ct) 151 { 152 return read_pnet(&ct->ct_net); 153 } 154 155 /* Alter reply tuple (maybe alter helper). */ 156 void nf_conntrack_alter_reply(struct nf_conn *ct, 157 const struct nf_conntrack_tuple *newreply); 158 159 /* Is this tuple taken? (ignoring any belonging to the given 160 conntrack). */ 161 int nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple, 162 const struct nf_conn *ignored_conntrack); 163 164 /* Return conntrack_info and tuple hash for given skb. */ 165 static inline struct nf_conn * 166 nf_ct_get(const struct sk_buff *skb, enum ip_conntrack_info *ctinfo) 167 { 168 unsigned long nfct = skb_get_nfct(skb); 169 170 *ctinfo = nfct & NFCT_INFOMASK; 171 return (struct nf_conn *)(nfct & NFCT_PTRMASK); 172 } 173 174 void nf_ct_destroy(struct nf_conntrack *nfct); 175 176 /* decrement reference count on a conntrack */ 177 static inline void nf_ct_put(struct nf_conn *ct) 178 { 179 if (ct && refcount_dec_and_test(&ct->ct_general.use)) 180 nf_ct_destroy(&ct->ct_general); 181 } 182 183 /* Protocol module loading */ 184 int nf_ct_l3proto_try_module_get(unsigned short l3proto); 185 void nf_ct_l3proto_module_put(unsigned short l3proto); 186 187 /* load module; enable/disable conntrack in this namespace */ 188 int nf_ct_netns_get(struct net *net, u8 nfproto); 189 void nf_ct_netns_put(struct net *net, u8 nfproto); 190 191 /* 192 * Allocate a hashtable of hlist_head (if nulls == 0), 193 * or hlist_nulls_head (if nulls == 1) 194 */ 195 void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls); 196 197 int nf_conntrack_hash_check_insert(struct nf_conn *ct); 198 bool nf_ct_delete(struct nf_conn *ct, u32 pid, int report); 199 200 bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, 201 u_int16_t l3num, struct net *net, 202 struct nf_conntrack_tuple *tuple); 203 204 void __nf_ct_refresh_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo, 205 const struct sk_buff *skb, 206 u32 extra_jiffies, bool do_acct); 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, skb, extra_jiffies, true); 215 } 216 217 /* Refresh conntrack for this many jiffies */ 218 static inline void nf_ct_refresh(struct nf_conn *ct, 219 const struct sk_buff *skb, 220 u32 extra_jiffies) 221 { 222 __nf_ct_refresh_acct(ct, 0, skb, extra_jiffies, false); 223 } 224 225 /* kill conntrack and do accounting */ 226 bool nf_ct_kill_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo, 227 const struct sk_buff *skb); 228 229 /* kill conntrack without accounting */ 230 static inline bool nf_ct_kill(struct nf_conn *ct) 231 { 232 return nf_ct_delete(ct, 0, 0); 233 } 234 235 /* Set all unconfirmed conntrack as dying */ 236 void nf_ct_unconfirmed_destroy(struct net *); 237 238 /* Iterate over all conntracks: if iter returns true, it's deleted. */ 239 void nf_ct_iterate_cleanup_net(struct net *net, 240 int (*iter)(struct nf_conn *i, void *data), 241 void *data, u32 portid, int report); 242 243 /* also set unconfirmed conntracks as dying. Only use in module exit path. */ 244 void nf_ct_iterate_destroy(int (*iter)(struct nf_conn *i, void *data), 245 void *data); 246 247 struct nf_conntrack_zone; 248 249 void nf_conntrack_free(struct nf_conn *ct); 250 struct nf_conn *nf_conntrack_alloc(struct net *net, 251 const struct nf_conntrack_zone *zone, 252 const struct nf_conntrack_tuple *orig, 253 const struct nf_conntrack_tuple *repl, 254 gfp_t gfp); 255 256 static inline int nf_ct_is_template(const struct nf_conn *ct) 257 { 258 return test_bit(IPS_TEMPLATE_BIT, &ct->status); 259 } 260 261 /* It's confirmed if it is, or has been in the hash table. */ 262 static inline int nf_ct_is_confirmed(const struct nf_conn *ct) 263 { 264 return test_bit(IPS_CONFIRMED_BIT, &ct->status); 265 } 266 267 static inline int nf_ct_is_dying(const struct nf_conn *ct) 268 { 269 return test_bit(IPS_DYING_BIT, &ct->status); 270 } 271 272 /* Packet is received from loopback */ 273 static inline bool nf_is_loopback_packet(const struct sk_buff *skb) 274 { 275 return skb->dev && skb->skb_iif && skb->dev->flags & IFF_LOOPBACK; 276 } 277 278 #define nfct_time_stamp ((u32)(jiffies)) 279 280 /* jiffies until ct expires, 0 if already expired */ 281 static inline unsigned long nf_ct_expires(const struct nf_conn *ct) 282 { 283 s32 timeout = READ_ONCE(ct->timeout) - nfct_time_stamp; 284 285 return max(timeout, 0); 286 } 287 288 static inline bool nf_ct_is_expired(const struct nf_conn *ct) 289 { 290 return (__s32)(READ_ONCE(ct->timeout) - nfct_time_stamp) <= 0; 291 } 292 293 /* use after obtaining a reference count */ 294 static inline bool nf_ct_should_gc(const struct nf_conn *ct) 295 { 296 return nf_ct_is_expired(ct) && nf_ct_is_confirmed(ct) && 297 !nf_ct_is_dying(ct); 298 } 299 300 #define NF_CT_DAY (86400 * HZ) 301 302 /* Set an arbitrary timeout large enough not to ever expire, this save 303 * us a check for the IPS_OFFLOAD_BIT from the packet path via 304 * nf_ct_is_expired(). 305 */ 306 static inline void nf_ct_offload_timeout(struct nf_conn *ct) 307 { 308 if (nf_ct_expires(ct) < NF_CT_DAY / 2) 309 WRITE_ONCE(ct->timeout, nfct_time_stamp + NF_CT_DAY); 310 } 311 312 struct kernel_param; 313 314 int nf_conntrack_set_hashsize(const char *val, const struct kernel_param *kp); 315 int nf_conntrack_hash_resize(unsigned int hashsize); 316 317 extern struct hlist_nulls_head *nf_conntrack_hash; 318 extern unsigned int nf_conntrack_htable_size; 319 extern seqcount_spinlock_t nf_conntrack_generation; 320 extern unsigned int nf_conntrack_max; 321 322 /* must be called with rcu read lock held */ 323 static inline void 324 nf_conntrack_get_ht(struct hlist_nulls_head **hash, unsigned int *hsize) 325 { 326 struct hlist_nulls_head *hptr; 327 unsigned int sequence, hsz; 328 329 do { 330 sequence = read_seqcount_begin(&nf_conntrack_generation); 331 hsz = nf_conntrack_htable_size; 332 hptr = nf_conntrack_hash; 333 } while (read_seqcount_retry(&nf_conntrack_generation, sequence)); 334 335 *hash = hptr; 336 *hsize = hsz; 337 } 338 339 struct nf_conn *nf_ct_tmpl_alloc(struct net *net, 340 const struct nf_conntrack_zone *zone, 341 gfp_t flags); 342 void nf_ct_tmpl_free(struct nf_conn *tmpl); 343 344 u32 nf_ct_get_id(const struct nf_conn *ct); 345 u32 nf_conntrack_count(const struct net *net); 346 347 static inline void 348 nf_ct_set(struct sk_buff *skb, struct nf_conn *ct, enum ip_conntrack_info info) 349 { 350 skb_set_nfct(skb, (unsigned long)ct | info); 351 } 352 353 extern unsigned int nf_conntrack_net_id; 354 355 static inline struct nf_conntrack_net *nf_ct_pernet(const struct net *net) 356 { 357 return net_generic(net, nf_conntrack_net_id); 358 } 359 360 #define NF_CT_STAT_INC(net, count) __this_cpu_inc((net)->ct.stat->count) 361 #define NF_CT_STAT_INC_ATOMIC(net, count) this_cpu_inc((net)->ct.stat->count) 362 #define NF_CT_STAT_ADD_ATOMIC(net, count, v) this_cpu_add((net)->ct.stat->count, (v)) 363 364 #define MODULE_ALIAS_NFCT_HELPER(helper) \ 365 MODULE_ALIAS("nfct-helper-" helper) 366 367 #endif /* _NF_CONNTRACK_H */ 368