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 27struct nf_ct_udp { 28 unsigned long stream_ts; 29}; 30 31/* per conntrack: protocol private data */ 32union 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 42union nf_conntrack_expect_proto { 43 /* insert expect proto private data here */ 44}; 45 46struct nf_conntrack_net_ecache { 47 struct delayed_work dwork; 48 spinlock_t dying_lock; 49 struct hlist_nulls_head dying_list; 50}; 51 52struct 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 *sysctl_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 75struct 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 128static inline struct nf_conn * 129nf_ct_to_nf_conn(const struct nf_conntrack *nfct) 130{ 131 return container_of(nfct, struct nf_conn, ct_general); 132} 133 134static inline struct nf_conn * 135nf_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 141static 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 146static 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 156extern struct net init_net; 157 158static inline struct net *nf_ct_net(const struct nf_conn *ct) 159{ 160 return read_pnet(&ct->ct_net); 161} 162 163/* Is this tuple taken? (ignoring any belonging to the given 164 conntrack). */ 165int 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. */ 169static inline struct nf_conn * 170nf_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 178void nf_ct_destroy(struct nf_conntrack *nfct); 179 180void nf_conntrack_tcp_set_closing(struct nf_conn *ct); 181 182/* decrement reference count on a conntrack */ 183static inline void nf_ct_put(struct nf_conn *ct) 184{ 185 if (ct && refcount_dec_and_test(&ct->ct_general.use)) 186 nf_ct_destroy(&ct->ct_general); 187} 188 189/* load module; enable/disable conntrack in this namespace */ 190int nf_ct_netns_get(struct net *net, u8 nfproto); 191void 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 */ 197void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls); 198 199int nf_conntrack_hash_check_insert(struct nf_conn *ct); 200bool nf_ct_delete(struct nf_conn *ct, u32 pid, int report); 201 202bool 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 206void __nf_ct_refresh_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo, 207 const struct sk_buff *skb, 208 u32 extra_jiffies, bool do_acct); 209 210/* Refresh conntrack for this many jiffies and do accounting */ 211static inline void nf_ct_refresh_acct(struct nf_conn *ct, 212 enum ip_conntrack_info ctinfo, 213 const struct sk_buff *skb, 214 u32 extra_jiffies) 215{ 216 __nf_ct_refresh_acct(ct, ctinfo, skb, extra_jiffies, true); 217} 218 219/* Refresh conntrack for this many jiffies */ 220static inline void nf_ct_refresh(struct nf_conn *ct, 221 const struct sk_buff *skb, 222 u32 extra_jiffies) 223{ 224 __nf_ct_refresh_acct(ct, 0, skb, extra_jiffies, false); 225} 226 227/* kill conntrack and do accounting */ 228bool nf_ct_kill_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo, 229 const struct sk_buff *skb); 230 231/* kill conntrack without accounting */ 232static inline bool nf_ct_kill(struct nf_conn *ct) 233{ 234 return nf_ct_delete(ct, 0, 0); 235} 236 237struct nf_ct_iter_data { 238 struct net *net; 239 void *data; 240 u32 portid; 241 int report; 242}; 243 244/* Iterate over all conntracks: if iter returns true, it's deleted. */ 245void nf_ct_iterate_cleanup_net(int (*iter)(struct nf_conn *i, void *data), 246 const struct nf_ct_iter_data *iter_data); 247 248/* also set unconfirmed conntracks as dying. Only use in module exit path. */ 249void nf_ct_iterate_destroy(int (*iter)(struct nf_conn *i, void *data), 250 void *data); 251 252struct nf_conntrack_zone; 253 254void nf_conntrack_free(struct nf_conn *ct); 255struct nf_conn *nf_conntrack_alloc(struct net *net, 256 const struct nf_conntrack_zone *zone, 257 const struct nf_conntrack_tuple *orig, 258 const struct nf_conntrack_tuple *repl, 259 gfp_t gfp); 260 261static inline int nf_ct_is_template(const struct nf_conn *ct) 262{ 263 return test_bit(IPS_TEMPLATE_BIT, &ct->status); 264} 265 266/* It's confirmed if it is, or has been in the hash table. */ 267static inline int nf_ct_is_confirmed(const struct nf_conn *ct) 268{ 269 return test_bit(IPS_CONFIRMED_BIT, &ct->status); 270} 271 272static inline int nf_ct_is_dying(const struct nf_conn *ct) 273{ 274 return test_bit(IPS_DYING_BIT, &ct->status); 275} 276 277/* Packet is received from loopback */ 278static inline bool nf_is_loopback_packet(const struct sk_buff *skb) 279{ 280 return skb->dev && skb->skb_iif && skb->dev->flags & IFF_LOOPBACK; 281} 282 283static inline void nf_conntrack_alter_reply(struct nf_conn *ct, 284 const struct nf_conntrack_tuple *newreply) 285{ 286 /* Must be unconfirmed, so not in hash table yet */ 287 if (WARN_ON(nf_ct_is_confirmed(ct))) 288 return; 289 290 ct->tuplehash[IP_CT_DIR_REPLY].tuple = *newreply; 291} 292 293#define nfct_time_stamp ((u32)(jiffies)) 294 295/* jiffies until ct expires, 0 if already expired */ 296static inline unsigned long nf_ct_expires(const struct nf_conn *ct) 297{ 298 s32 timeout = READ_ONCE(ct->timeout) - nfct_time_stamp; 299 300 return max(timeout, 0); 301} 302 303static inline bool nf_ct_is_expired(const struct nf_conn *ct) 304{ 305 return (__s32)(READ_ONCE(ct->timeout) - nfct_time_stamp) <= 0; 306} 307 308/* use after obtaining a reference count */ 309static inline bool nf_ct_should_gc(const struct nf_conn *ct) 310{ 311 return nf_ct_is_expired(ct) && nf_ct_is_confirmed(ct) && 312 !nf_ct_is_dying(ct); 313} 314 315#define NF_CT_DAY (86400 * HZ) 316 317/* Set an arbitrary timeout large enough not to ever expire, this save 318 * us a check for the IPS_OFFLOAD_BIT from the packet path via 319 * nf_ct_is_expired(). 320 */ 321static inline void nf_ct_offload_timeout(struct nf_conn *ct) 322{ 323 if (nf_ct_expires(ct) < NF_CT_DAY / 2) 324 WRITE_ONCE(ct->timeout, nfct_time_stamp + NF_CT_DAY); 325} 326 327struct kernel_param; 328 329int nf_conntrack_set_hashsize(const char *val, const struct kernel_param *kp); 330int nf_conntrack_hash_resize(unsigned int hashsize); 331 332extern struct hlist_nulls_head *nf_conntrack_hash; 333extern unsigned int nf_conntrack_htable_size; 334extern seqcount_spinlock_t nf_conntrack_generation; 335extern unsigned int nf_conntrack_max; 336 337/* must be called with rcu read lock held */ 338static inline void 339nf_conntrack_get_ht(struct hlist_nulls_head **hash, unsigned int *hsize) 340{ 341 struct hlist_nulls_head *hptr; 342 unsigned int sequence, hsz; 343 344 do { 345 sequence = read_seqcount_begin(&nf_conntrack_generation); 346 hsz = nf_conntrack_htable_size; 347 hptr = nf_conntrack_hash; 348 } while (read_seqcount_retry(&nf_conntrack_generation, sequence)); 349 350 *hash = hptr; 351 *hsize = hsz; 352} 353 354struct nf_conn *nf_ct_tmpl_alloc(struct net *net, 355 const struct nf_conntrack_zone *zone, 356 gfp_t flags); 357void nf_ct_tmpl_free(struct nf_conn *tmpl); 358 359u32 nf_ct_get_id(const struct nf_conn *ct); 360u32 nf_conntrack_count(const struct net *net); 361 362static inline void 363nf_ct_set(struct sk_buff *skb, struct nf_conn *ct, enum ip_conntrack_info info) 364{ 365 skb_set_nfct(skb, (unsigned long)ct | info); 366} 367 368extern unsigned int nf_conntrack_net_id; 369 370static inline struct nf_conntrack_net *nf_ct_pernet(const struct net *net) 371{ 372 return net_generic(net, nf_conntrack_net_id); 373} 374 375int nf_ct_skb_network_trim(struct sk_buff *skb, int family); 376int nf_ct_handle_fragments(struct net *net, struct sk_buff *skb, 377 u16 zone, u8 family, u8 *proto, u16 *mru); 378 379#define NF_CT_STAT_INC(net, count) __this_cpu_inc((net)->ct.stat->count) 380#define NF_CT_STAT_INC_ATOMIC(net, count) this_cpu_inc((net)->ct.stat->count) 381#define NF_CT_STAT_ADD_ATOMIC(net, count, v) this_cpu_add((net)->ct.stat->count, (v)) 382 383#define MODULE_ALIAS_NFCT_HELPER(helper) \ 384 MODULE_ALIAS("nfct-helper-" helper) 385 386#endif /* _NF_CONNTRACK_H */ 387