| 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 | |
| 57 | /* only used from work queues, configuration plane, and so on: */ |
| 58 | unsigned int users4; |
| 59 | unsigned int users6; |
| 60 | unsigned int users_bridge; |
| 61 | #ifdef CONFIG_SYSCTL |
| 62 | struct ctl_table_header *; |
| 63 | #endif |
| 64 | #ifdef CONFIG_NF_CONNTRACK_EVENTS |
| 65 | struct nf_conntrack_net_ecache ecache; |
| 66 | #endif |
| 67 | }; |
| 68 | |
| 69 | #include <linux/types.h> |
| 70 | #include <linux/skbuff.h> |
| 71 | |
| 72 | #include <net/netfilter/ipv4/nf_conntrack_ipv4.h> |
| 73 | #include <net/netfilter/ipv6/nf_conntrack_ipv6.h> |
| 74 | |
| 75 | struct nf_conn { |
| 76 | /* Usage count in here is 1 for hash table, 1 per skb, |
| 77 | * plus 1 for any connection(s) we are `master' for |
| 78 | * |
| 79 | * Hint, SKB address this struct and refcnt via skb->_nfct and |
| 80 | * helpers nf_conntrack_get() and nf_conntrack_put(). |
| 81 | * Helper nf_ct_put() equals nf_conntrack_put() by dec refcnt, |
| 82 | * except that the latter uses internal indirection and does not |
| 83 | * result in a conntrack module dependency. |
| 84 | * beware nf_ct_get() is different and don't inc refcnt. |
| 85 | */ |
| 86 | struct nf_conntrack ct_general; |
| 87 | |
| 88 | spinlock_t lock; |
| 89 | /* jiffies32 when this ct is considered dead */ |
| 90 | u32 timeout; |
| 91 | |
| 92 | #ifdef CONFIG_NF_CONNTRACK_ZONES |
| 93 | struct nf_conntrack_zone zone; |
| 94 | #endif |
| 95 | /* XXX should I move this to the tail ? - Y.K */ |
| 96 | /* These are my tuples; original and reply */ |
| 97 | struct nf_conntrack_tuple_hash tuplehash[IP_CT_DIR_MAX]; |
| 98 | |
| 99 | /* Have we seen traffic both ways yet? (bitset) */ |
| 100 | unsigned long status; |
| 101 | |
| 102 | possible_net_t ct_net; |
| 103 | |
| 104 | #if IS_ENABLED(CONFIG_NF_NAT) |
| 105 | struct hlist_node nat_bysource; |
| 106 | #endif |
| 107 | /* all members below initialized via memset */ |
| 108 | struct { } __nfct_init_offset; |
| 109 | |
| 110 | /* If we were expected by an expectation, this will be it */ |
| 111 | struct nf_conn *master; |
| 112 | |
| 113 | #if defined(CONFIG_NF_CONNTRACK_MARK) |
| 114 | u_int32_t mark; |
| 115 | #endif |
| 116 | |
| 117 | #ifdef CONFIG_NF_CONNTRACK_SECMARK |
| 118 | u_int32_t secmark; |
| 119 | #endif |
| 120 | |
| 121 | /* Extensions */ |
| 122 | struct nf_ct_ext *ext; |
| 123 | |
| 124 | /* Storage reserved for other modules, must be the last member */ |
| 125 | union nf_conntrack_proto proto; |
| 126 | }; |
| 127 | |
| 128 | static inline struct nf_conn * |
| 129 | nf_ct_to_nf_conn(const struct nf_conntrack *nfct) |
| 130 | { |
| 131 | return container_of(nfct, struct nf_conn, ct_general); |
| 132 | } |
| 133 | |
| 134 | static inline struct nf_conn * |
| 135 | nf_ct_tuplehash_to_ctrack(const struct nf_conntrack_tuple_hash *hash) |
| 136 | { |
| 137 | return container_of(hash, struct nf_conn, |
| 138 | tuplehash[hash->tuple.dst.dir]); |
| 139 | } |
| 140 | |
| 141 | static inline u_int16_t nf_ct_l3num(const struct nf_conn *ct) |
| 142 | { |
| 143 | return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num; |
| 144 | } |
| 145 | |
| 146 | static inline u_int8_t nf_ct_protonum(const struct nf_conn *ct) |
| 147 | { |
| 148 | return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum; |
| 149 | } |
| 150 | |
| 151 | #define nf_ct_tuple(ct, dir) (&(ct)->tuplehash[dir].tuple) |
| 152 | |
| 153 | /* get master conntrack via master expectation */ |
| 154 | #define master_ct(conntr) (conntr->master) |
| 155 | |
| 156 | extern struct net init_net; |
| 157 | |
| 158 | static inline struct net *nf_ct_net(const struct nf_conn *ct) |
| 159 | { |
| 160 | return read_pnet(pnet: &ct->ct_net); |
| 161 | } |
| 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 | void nf_conntrack_tcp_set_closing(struct nf_conn *ct); |
| 181 | |
| 182 | /* decrement reference count on a conntrack */ |
| 183 | static inline void nf_ct_put(struct nf_conn *ct) |
| 184 | { |
| 185 | if (ct && refcount_dec_and_test(r: &ct->ct_general.use)) |
| 186 | nf_ct_destroy(nfct: &ct->ct_general); |
| 187 | } |
| 188 | |
| 189 | /* load module; enable/disable conntrack in this namespace */ |
| 190 | int nf_ct_netns_get(struct net *net, u8 nfproto); |
| 191 | void nf_ct_netns_put(struct net *net, u8 nfproto); |
| 192 | |
| 193 | /* |
| 194 | * Allocate a hashtable of hlist_head (if nulls == 0), |
| 195 | * or hlist_nulls_head (if nulls == 1) |
| 196 | */ |
| 197 | void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls); |
| 198 | |
| 199 | int nf_conntrack_hash_check_insert(struct nf_conn *ct); |
| 200 | bool nf_ct_delete(struct nf_conn *ct, u32 pid, int report); |
| 201 | |
| 202 | bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff, |
| 203 | u_int16_t l3num, struct net *net, |
| 204 | struct nf_conntrack_tuple *tuple); |
| 205 | |
| 206 | void __nf_ct_refresh_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo, |
| 207 | u32 , unsigned int bytes); |
| 208 | |
| 209 | /* Refresh conntrack for this many jiffies and do accounting */ |
| 210 | static inline void nf_ct_refresh_acct(struct nf_conn *ct, |
| 211 | enum ip_conntrack_info ctinfo, |
| 212 | const struct sk_buff *skb, |
| 213 | u32 ) |
| 214 | { |
| 215 | __nf_ct_refresh_acct(ct, ctinfo, extra_jiffies, bytes: skb->len); |
| 216 | } |
| 217 | |
| 218 | /* Refresh conntrack for this many jiffies */ |
| 219 | static inline void nf_ct_refresh(struct nf_conn *ct, |
| 220 | u32 ) |
| 221 | { |
| 222 | __nf_ct_refresh_acct(ct, ctinfo: 0, extra_jiffies, bytes: 0); |
| 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, pid: 0, report: 0); |
| 233 | } |
| 234 | |
| 235 | struct nf_ct_iter_data { |
| 236 | struct net *net; |
| 237 | void *data; |
| 238 | u32 portid; |
| 239 | int report; |
| 240 | }; |
| 241 | |
| 242 | /* Iterate over all conntracks: if iter returns true, it's deleted. */ |
| 243 | void nf_ct_iterate_cleanup_net(int (*iter)(struct nf_conn *i, void *data), |
| 244 | const struct nf_ct_iter_data *iter_data); |
| 245 | |
| 246 | /* also set unconfirmed conntracks as dying. Only use in module exit path. */ |
| 247 | void nf_ct_iterate_destroy(int (*iter)(struct nf_conn *i, void *data), |
| 248 | void *data); |
| 249 | |
| 250 | struct nf_conntrack_zone; |
| 251 | |
| 252 | void nf_conntrack_free(struct nf_conn *ct); |
| 253 | struct nf_conn *nf_conntrack_alloc(struct net *net, |
| 254 | const struct nf_conntrack_zone *zone, |
| 255 | const struct nf_conntrack_tuple *orig, |
| 256 | const struct nf_conntrack_tuple *repl, |
| 257 | gfp_t gfp); |
| 258 | |
| 259 | static inline int nf_ct_is_template(const struct nf_conn *ct) |
| 260 | { |
| 261 | return test_bit(IPS_TEMPLATE_BIT, &ct->status); |
| 262 | } |
| 263 | |
| 264 | /* It's confirmed if it is, or has been in the hash table. */ |
| 265 | static inline int nf_ct_is_confirmed(const struct nf_conn *ct) |
| 266 | { |
| 267 | return test_bit(IPS_CONFIRMED_BIT, &ct->status); |
| 268 | } |
| 269 | |
| 270 | static inline int nf_ct_is_dying(const struct nf_conn *ct) |
| 271 | { |
| 272 | return test_bit(IPS_DYING_BIT, &ct->status); |
| 273 | } |
| 274 | |
| 275 | /* Packet is received from loopback */ |
| 276 | static inline bool nf_is_loopback_packet(const struct sk_buff *skb) |
| 277 | { |
| 278 | return skb->dev && skb->skb_iif && skb->dev->flags & IFF_LOOPBACK; |
| 279 | } |
| 280 | |
| 281 | static inline void nf_conntrack_alter_reply(struct nf_conn *ct, |
| 282 | const struct nf_conntrack_tuple *newreply) |
| 283 | { |
| 284 | /* Must be unconfirmed, so not in hash table yet */ |
| 285 | if (WARN_ON(nf_ct_is_confirmed(ct))) |
| 286 | return; |
| 287 | |
| 288 | ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; |
| 289 | } |
| 290 | |
| 291 | #define nfct_time_stamp ((u32)(jiffies)) |
| 292 | |
| 293 | /* jiffies until ct expires, 0 if already expired */ |
| 294 | static inline unsigned long nf_ct_expires(const struct nf_conn *ct) |
| 295 | { |
| 296 | s32 timeout = READ_ONCE(ct->timeout) - nfct_time_stamp; |
| 297 | |
| 298 | return max(timeout, 0); |
| 299 | } |
| 300 | |
| 301 | static inline bool nf_ct_is_expired(const struct nf_conn *ct) |
| 302 | { |
| 303 | return (__s32)(READ_ONCE(ct->timeout) - nfct_time_stamp) <= 0; |
| 304 | } |
| 305 | |
| 306 | /* use after obtaining a reference count */ |
| 307 | static inline bool nf_ct_should_gc(const struct nf_conn *ct) |
| 308 | { |
| 309 | return nf_ct_is_expired(ct) && nf_ct_is_confirmed(ct) && |
| 310 | !nf_ct_is_dying(ct); |
| 311 | } |
| 312 | |
| 313 | #define NF_CT_DAY (86400 * HZ) |
| 314 | |
| 315 | struct kernel_param; |
| 316 | |
| 317 | int nf_conntrack_set_hashsize(const char *val, const struct kernel_param *kp); |
| 318 | int nf_conntrack_hash_resize(unsigned int hashsize); |
| 319 | |
| 320 | extern struct hlist_nulls_head *nf_conntrack_hash; |
| 321 | extern unsigned int nf_conntrack_htable_size; |
| 322 | extern seqcount_spinlock_t nf_conntrack_generation; |
| 323 | extern unsigned int nf_conntrack_max; |
| 324 | |
| 325 | /* must be called with rcu read lock held */ |
| 326 | static inline void |
| 327 | nf_conntrack_get_ht(struct hlist_nulls_head **hash, unsigned int *hsize) |
| 328 | { |
| 329 | struct hlist_nulls_head *hptr; |
| 330 | unsigned int sequence, hsz; |
| 331 | |
| 332 | do { |
| 333 | sequence = read_seqcount_begin(&nf_conntrack_generation); |
| 334 | hsz = nf_conntrack_htable_size; |
| 335 | hptr = nf_conntrack_hash; |
| 336 | } while (read_seqcount_retry(&nf_conntrack_generation, sequence)); |
| 337 | |
| 338 | *hash = hptr; |
| 339 | *hsize = hsz; |
| 340 | } |
| 341 | |
| 342 | struct nf_conn *nf_ct_tmpl_alloc(struct net *net, |
| 343 | const struct nf_conntrack_zone *zone, |
| 344 | gfp_t flags); |
| 345 | void nf_ct_tmpl_free(struct nf_conn *tmpl); |
| 346 | |
| 347 | u32 nf_ct_get_id(const struct nf_conn *ct); |
| 348 | u32 nf_conntrack_count(const struct net *net); |
| 349 | |
| 350 | static inline void |
| 351 | nf_ct_set(struct sk_buff *skb, struct nf_conn *ct, enum ip_conntrack_info info) |
| 352 | { |
| 353 | skb_set_nfct(skb, nfct: (unsigned long)ct | info); |
| 354 | } |
| 355 | |
| 356 | extern unsigned int nf_conntrack_net_id; |
| 357 | |
| 358 | static inline struct nf_conntrack_net *nf_ct_pernet(const struct net *net) |
| 359 | { |
| 360 | return net_generic(net, id: nf_conntrack_net_id); |
| 361 | } |
| 362 | |
| 363 | int nf_ct_skb_network_trim(struct sk_buff *skb, int family); |
| 364 | int nf_ct_handle_fragments(struct net *net, struct sk_buff *skb, |
| 365 | u16 zone, u8 family, u8 *proto, u16 *mru); |
| 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 | |