include/net/netfilter/nf_conntrack.h

   1 /*
   2  * Connection state tracking for netfilter.  This is separated from,
   3  * but required by, the (future) NAT layer; it can also be used by an iptables
   4  * extension.
   5  *
   6  * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
   7  *      - generalize L3 protocol dependent part.
   8  *
   9  * Derived from include/linux/netfiter_ipv4/ip_conntrack.h
  10  */
  11
  12 #ifndef _NF_CONNTRACK_H
  13 #define _NF_CONNTRACK_H
  14
  15 #include <linux/netfilter/nf_conntrack_common.h>
  16
  17 #include <linux/bitops.h>
  18 #include <linux/compiler.h>
  19 #include <linux/atomic.h>
  20
  21 #include <linux/netfilter/nf_conntrack_tcp.h>
  22 #include <linux/netfilter/nf_conntrack_dccp.h>
  23 #include <linux/netfilter/nf_conntrack_sctp.h>
  24 #include <linux/netfilter/nf_conntrack_proto_gre.h>
  25 #include <net/netfilter/ipv6/nf_conntrack_icmpv6.h>
  26
  27 #include <net/netfilter/nf_conntrack_tuple.h>
  28
  29 /* per conntrack: protocol private data */
  30 union nf_conntrack_proto {
  31         /* insert conntrack proto private data here */
  32         struct nf_ct_dccp dccp;
  33         struct ip_ct_sctp sctp;
  34         struct ip_ct_tcp tcp;
  35         struct nf_ct_gre gre;
  36         unsigned int tmpl_padto;
  37 };
  38
  39 union nf_conntrack_expect_proto {
  40         /* insert expect proto private data here */
  41 };
  42
  43 #include <linux/types.h>
  44 #include <linux/skbuff.h>
  45
  46 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
  47 #include <net/netfilter/ipv6/nf_conntrack_ipv6.h>
  48
  49 struct nf_conn {
  50         /* Usage count in here is 1 for hash table, 1 per skb,
  51          * plus 1 for any connection(s) we are `master' for
  52          *
  53          * Hint, SKB address this struct and refcnt via skb->_nfct and
  54          * helpers nf_conntrack_get() and nf_conntrack_put().
  55          * Helper nf_ct_put() equals nf_conntrack_put() by dec refcnt,
  56          * beware nf_ct_get() is different and don't inc refcnt.
  57          */
  58         struct nf_conntrack ct_general;
  59
  60         spinlock_t      lock;
  61         u16             cpu;
  62
  63 #ifdef CONFIG_NF_CONNTRACK_ZONES
  64         struct nf_conntrack_zone zone;
  65 #endif
  66         /* XXX should I move this to the tail ? - Y.K */
  67         /* These are my tuples; original and reply */
  68         struct nf_conntrack_tuple_hash tuplehash[IP_CT_DIR_MAX];
  69
  70         /* Have we seen traffic both ways yet? (bitset) */
  71         unsigned long status;
  72
  73         /* jiffies32 when this ct is considered dead */
  74         u32 timeout;
  75
  76         possible_net_t ct_net;
  77
  78 #if IS_ENABLED(CONFIG_NF_NAT)
  79         struct hlist_node       nat_bysource;
  80 #endif
  81         /* all members below initialized via memset */
  82         u8 __nfct_init_offset[0];
  83
  84         /* If we were expected by an expectation, this will be it */
  85         struct nf_conn *master;
  86
  87 #if defined(CONFIG_NF_CONNTRACK_MARK)
  88         u_int32_t mark;
  89 #endif
  90
  91 #ifdef CONFIG_NF_CONNTRACK_SECMARK
  92         u_int32_t secmark;
  93 #endif
  94
  95         /* Extensions */
  96         struct nf_ct_ext *ext;
  97
  98         /* Storage reserved for other modules, must be the last member */
  99         union nf_conntrack_proto proto;
 100 };
 101
 102 static inline struct nf_conn *
 103 nf_ct_tuplehash_to_ctrack(const struct nf_conntrack_tuple_hash *hash)
 104 {
 105         return container_of(hash, struct nf_conn,
 106                             tuplehash[hash->tuple.dst.dir]);
 107 }
 108
 109 static inline u_int16_t nf_ct_l3num(const struct nf_conn *ct)
 110 {
 111         return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.l3num;
 112 }
 113
 114 static inline u_int8_t nf_ct_protonum(const struct nf_conn *ct)
 115 {
 116         return ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 117 }
 118
 119 #define nf_ct_tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 120
 121 /* get master conntrack via master expectation */
 122 #define master_ct(conntr) (conntr->master)
 123
 124 extern struct net init_net;
 125
 126 static inline struct net *nf_ct_net(const struct nf_conn *ct)
 127 {
 128         return read_pnet(&ct->ct_net);
 129 }
 130
 131 /* Alter reply tuple (maybe alter helper). */
 132 void nf_conntrack_alter_reply(struct nf_conn *ct,
 133                               const struct nf_conntrack_tuple *newreply);
 134
 135 /* Is this tuple taken? (ignoring any belonging to the given
 136    conntrack). */
 137 int nf_conntrack_tuple_taken(const struct nf_conntrack_tuple *tuple,
 138                              const struct nf_conn *ignored_conntrack);
 139
 140 #define NFCT_INFOMASK   7UL
 141 #define NFCT_PTRMASK    ~(NFCT_INFOMASK)
 142
 143 /* Return conntrack_info and tuple hash for given skb. */
 144 static inline struct nf_conn *
 145 nf_ct_get(const struct sk_buff *skb, enum ip_conntrack_info *ctinfo)
 146 {
 147         *ctinfo = skb->_nfct & NFCT_INFOMASK;
 148
 149         return (struct nf_conn *)(skb->_nfct & NFCT_PTRMASK);
 150 }
 151
 152 /* decrement reference count on a conntrack */
 153 static inline void nf_ct_put(struct nf_conn *ct)
 154 {
 155         WARN_ON(!ct);
 156         nf_conntrack_put(&ct->ct_general);
 157 }
 158
 159 /* Protocol module loading */
 160 int nf_ct_l3proto_try_module_get(unsigned short l3proto);
 161 void nf_ct_l3proto_module_put(unsigned short l3proto);
 162
 163 /* load module; enable/disable conntrack in this namespace */
 164 int nf_ct_netns_get(struct net *net, u8 nfproto);
 165 void nf_ct_netns_put(struct net *net, u8 nfproto);
 166
 167 /*
 168  * Allocate a hashtable of hlist_head (if nulls == 0),
 169  * or hlist_nulls_head (if nulls == 1)
 170  */
 171 void *nf_ct_alloc_hashtable(unsigned int *sizep, int nulls);
 172
 173 void nf_ct_free_hashtable(void *hash, unsigned int size);
 174
 175 int nf_conntrack_hash_check_insert(struct nf_conn *ct);
 176 bool nf_ct_delete(struct nf_conn *ct, u32 pid, int report);
 177
 178 bool nf_ct_get_tuplepr(const struct sk_buff *skb, unsigned int nhoff,
 179                        u_int16_t l3num, struct net *net,
 180                        struct nf_conntrack_tuple *tuple);
 181 bool nf_ct_invert_tuplepr(struct nf_conntrack_tuple *inverse,
 182                           const struct nf_conntrack_tuple *orig);
 183
 184 void __nf_ct_refresh_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
 185                           const struct sk_buff *skb,
 186                           unsigned long extra_jiffies, int do_acct);
 187
 188 /* Refresh conntrack for this many jiffies and do accounting */
 189 static inline void nf_ct_refresh_acct(struct nf_conn *ct,
 190                                       enum ip_conntrack_info ctinfo,
 191                                       const struct sk_buff *skb,
 192                                       unsigned long extra_jiffies)
 193 {
 194         __nf_ct_refresh_acct(ct, ctinfo, skb, extra_jiffies, 1);
 195 }
 196
 197 /* Refresh conntrack for this many jiffies */
 198 static inline void nf_ct_refresh(struct nf_conn *ct,
 199                                  const struct sk_buff *skb,
 200                                  unsigned long extra_jiffies)
 201 {
 202         __nf_ct_refresh_acct(ct, 0, skb, extra_jiffies, 0);
 203 }
 204
 205 /* kill conntrack and do accounting */
 206 bool nf_ct_kill_acct(struct nf_conn *ct, enum ip_conntrack_info ctinfo,
 207                      const struct sk_buff *skb);
 208
 209 /* kill conntrack without accounting */
 210 static inline bool nf_ct_kill(struct nf_conn *ct)
 211 {
 212         return nf_ct_delete(ct, 0, 0);
 213 }
 214
 215 /* These are for NAT.  Icky. */
 216 extern s32 (*nf_ct_nat_offset)(const struct nf_conn *ct,
 217                                enum ip_conntrack_dir dir,
 218                                u32 seq);
 219
 220 /* Set all unconfirmed conntrack as dying */
 221 void nf_ct_unconfirmed_destroy(struct net *);
 222
 223 /* Iterate over all conntracks: if iter returns true, it's deleted. */
 224 void nf_ct_iterate_cleanup_net(struct net *net,
 225                                int (*iter)(struct nf_conn *i, void *data),
 226                                void *data, u32 portid, int report);
 227
 228 /* also set unconfirmed conntracks as dying. Only use in module exit path. */
 229 void nf_ct_iterate_destroy(int (*iter)(struct nf_conn *i, void *data),
 230                            void *data);
 231
 232 struct nf_conntrack_zone;
 233
 234 void nf_conntrack_free(struct nf_conn *ct);
 235 struct nf_conn *nf_conntrack_alloc(struct net *net,
 236                                    const struct nf_conntrack_zone *zone,
 237                                    const struct nf_conntrack_tuple *orig,
 238                                    const struct nf_conntrack_tuple *repl,
 239                                    gfp_t gfp);
 240
 241 static inline int nf_ct_is_template(const struct nf_conn *ct)
 242 {
 243         return test_bit(IPS_TEMPLATE_BIT, &ct->status);
 244 }
 245
 246 /* It's confirmed if it is, or has been in the hash table. */
 247 static inline int nf_ct_is_confirmed(const struct nf_conn *ct)
 248 {
 249         return test_bit(IPS_CONFIRMED_BIT, &ct->status);
 250 }
 251
 252 static inline int nf_ct_is_dying(const struct nf_conn *ct)
 253 {
 254         return test_bit(IPS_DYING_BIT, &ct->status);
 255 }
 256
 257 /* Packet is received from loopback */
 258 static inline bool nf_is_loopback_packet(const struct sk_buff *skb)
 259 {
 260         return skb->dev && skb->skb_iif && skb->dev->flags & IFF_LOOPBACK;
 261 }
 262
 263 #define nfct_time_stamp ((u32)(jiffies))
 264
 265 /* jiffies until ct expires, 0 if already expired */
 266 static inline unsigned long nf_ct_expires(const struct nf_conn *ct)
 267 {
 268         s32 timeout = ct->timeout - nfct_time_stamp;
 269
 270         return timeout > 0 ? timeout : 0;
 271 }
 272
 273 static inline bool nf_ct_is_expired(const struct nf_conn *ct)
 274 {
 275         return (__s32)(ct->timeout - nfct_time_stamp) <= 0;
 276 }
 277
 278 /* use after obtaining a reference count */
 279 static inline bool nf_ct_should_gc(const struct nf_conn *ct)
 280 {
 281         return nf_ct_is_expired(ct) && nf_ct_is_confirmed(ct) &&
 282                !nf_ct_is_dying(ct);
 283 }
 284
 285 struct kernel_param;
 286
 287 int nf_conntrack_set_hashsize(const char *val, struct kernel_param *kp);
 288 int nf_conntrack_hash_resize(unsigned int hashsize);
 289
 290 extern struct hlist_nulls_head *nf_conntrack_hash;
 291 extern unsigned int nf_conntrack_htable_size;
 292 extern seqcount_t nf_conntrack_generation;
 293 extern unsigned int nf_conntrack_max;
 294
 295 /* must be called with rcu read lock held */
 296 static inline void
 297 nf_conntrack_get_ht(struct hlist_nulls_head **hash, unsigned int *hsize)
 298 {
 299         struct hlist_nulls_head *hptr;
 300         unsigned int sequence, hsz;
 301
 302         do {
 303                 sequence = read_seqcount_begin(&nf_conntrack_generation);
 304                 hsz = nf_conntrack_htable_size;
 305                 hptr = nf_conntrack_hash;
 306         } while (read_seqcount_retry(&nf_conntrack_generation, sequence));
 307
 308         *hash = hptr;
 309         *hsize = hsz;
 310 }
 311
 312 struct nf_conn *nf_ct_tmpl_alloc(struct net *net,
 313                                  const struct nf_conntrack_zone *zone,
 314                                  gfp_t flags);
 315 void nf_ct_tmpl_free(struct nf_conn *tmpl);
 316
 317 static inline void
 318 nf_ct_set(struct sk_buff *skb, struct nf_conn *ct, enum ip_conntrack_info info)
 319 {
 320         skb->_nfct = (unsigned long)ct | info;
 321 }
 322
 323 #define NF_CT_STAT_INC(net, count)        __this_cpu_inc((net)->ct.stat->count)
 324 #define NF_CT_STAT_INC_ATOMIC(net, count) this_cpu_inc((net)->ct.stat->count)
 325 #define NF_CT_STAT_ADD_ATOMIC(net, count, v) this_cpu_add((net)->ct.stat->count, (v))
 326
 327 #define MODULE_ALIAS_NFCT_HELPER(helper) \
 328         MODULE_ALIAS("nfct-helper-" helper)
 329
 330 #endif /* _NF_CONNTRACK_H */