include/linux/netfilter.h

   1 /* SPDX-License-Identifier: GPL-2.0 */
   2 #ifndef __LINUX_NETFILTER_H
   3 #define __LINUX_NETFILTER_H
   4
   5 #include <linux/init.h>
   6 #include <linux/skbuff.h>
   7 #include <linux/net.h>
   8 #include <linux/if.h>
   9 #include <linux/in.h>
  10 #include <linux/in6.h>
  11 #include <linux/wait.h>
  12 #include <linux/list.h>
  13 #include <linux/static_key.h>
  14 #include <linux/netfilter_defs.h>
  15 #include <linux/netdevice.h>
  16 #include <net/net_namespace.h>
  17
  18 #ifdef CONFIG_NETFILTER
  19 static inline int NF_DROP_GETERR(int verdict)
  20 {
  21         return -(verdict >> NF_VERDICT_QBITS);
  22 }
  23
  24 static inline int nf_inet_addr_cmp(const union nf_inet_addr *a1,
  25                                    const union nf_inet_addr *a2)
  26 {
  27 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
  28         const unsigned long *ul1 = (const unsigned long *)a1;
  29         const unsigned long *ul2 = (const unsigned long *)a2;
  30
  31         return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL;
  32 #else
  33         return a1->all[0] == a2->all[0] &&
  34                a1->all[1] == a2->all[1] &&
  35                a1->all[2] == a2->all[2] &&
  36                a1->all[3] == a2->all[3];
  37 #endif
  38 }
  39
  40 static inline void nf_inet_addr_mask(const union nf_inet_addr *a1,
  41                                      union nf_inet_addr *result,
  42                                      const union nf_inet_addr *mask)
  43 {
  44 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
  45         const unsigned long *ua = (const unsigned long *)a1;
  46         unsigned long *ur = (unsigned long *)result;
  47         const unsigned long *um = (const unsigned long *)mask;
  48
  49         ur[0] = ua[0] & um[0];
  50         ur[1] = ua[1] & um[1];
  51 #else
  52         result->all[0] = a1->all[0] & mask->all[0];
  53         result->all[1] = a1->all[1] & mask->all[1];
  54         result->all[2] = a1->all[2] & mask->all[2];
  55         result->all[3] = a1->all[3] & mask->all[3];
  56 #endif
  57 }
  58
  59 int netfilter_init(void);
  60
  61 struct sk_buff;
  62
  63 struct nf_hook_ops;
  64
  65 struct sock;
  66
  67 struct nf_hook_state {
  68         unsigned int hook;
  69         u_int8_t pf;
  70         struct net_device *in;
  71         struct net_device *out;
  72         struct sock *sk;
  73         struct net *net;
  74         int (*okfn)(struct net *, struct sock *, struct sk_buff *);
  75 };
  76
  77 typedef unsigned int nf_hookfn(void *priv,
  78                                struct sk_buff *skb,
  79                                const struct nf_hook_state *state);
  80 struct nf_hook_ops {
  81         /* User fills in from here down. */
  82         nf_hookfn               *hook;
  83         struct net_device       *dev;
  84         void                    *priv;
  85         u_int8_t                pf;
  86         unsigned int            hooknum;
  87         /* Hooks are ordered in ascending priority. */
  88         int                     priority;
  89 };
  90
  91 struct nf_hook_entry {
  92         nf_hookfn                       *hook;
  93         void                            *priv;
  94 };
  95
  96 struct nf_hook_entries_rcu_head {
  97         struct rcu_head head;
  98         void    *allocation;
  99 };
 100
 101 struct nf_hook_entries {
 102         u16                             num_hook_entries;
 103         /* padding */
 104         struct nf_hook_entry            hooks[];
 105
 106         /* trailer: pointers to original orig_ops of each hook,
 107          * followed by rcu_head and scratch space used for freeing
 108          * the structure via call_rcu.
 109          *
 110          *   This is not part of struct nf_hook_entry since its only
 111          *   needed in slow path (hook register/unregister):
 112          * const struct nf_hook_ops     *orig_ops[]
 113          *
 114          *   For the same reason, we store this at end -- its
 115          *   only needed when a hook is deleted, not during
 116          *   packet path processing:
 117          * struct nf_hook_entries_rcu_head     head
 118          */
 119 };
 120
 121 static inline struct nf_hook_ops **nf_hook_entries_get_hook_ops(const struct nf_hook_entries *e)
 122 {
 123         unsigned int n = e->num_hook_entries;
 124         const void *hook_end;
 125
 126         hook_end = &e->hooks[n]; /* this is *past* ->hooks[]! */
 127
 128         return (struct nf_hook_ops **)hook_end;
 129 }
 130
 131 static inline int
 132 nf_hook_entry_hookfn(const struct nf_hook_entry *entry, struct sk_buff *skb,
 133                      struct nf_hook_state *state)
 134 {
 135         return entry->hook(entry->priv, skb, state);
 136 }
 137
 138 static inline void nf_hook_state_init(struct nf_hook_state *p,
 139                                       unsigned int hook,
 140                                       u_int8_t pf,
 141                                       struct net_device *indev,
 142                                       struct net_device *outdev,
 143                                       struct sock *sk,
 144                                       struct net *net,
 145                                       int (*okfn)(struct net *, struct sock *, struct sk_buff *))
 146 {
 147         p->hook = hook;
 148         p->pf = pf;
 149         p->in = indev;
 150         p->out = outdev;
 151         p->sk = sk;
 152         p->net = net;
 153         p->okfn = okfn;
 154 }
 155
 156
 157
 158 struct nf_sockopt_ops {
 159         struct list_head list;
 160
 161         u_int8_t pf;
 162
 163         /* Non-inclusive ranges: use 0/0/NULL to never get called. */
 164         int set_optmin;
 165         int set_optmax;
 166         int (*set)(struct sock *sk, int optval, void __user *user, unsigned int len);
 167 #ifdef CONFIG_COMPAT
 168         int (*compat_set)(struct sock *sk, int optval,
 169                         void __user *user, unsigned int len);
 170 #endif
 171         int get_optmin;
 172         int get_optmax;
 173         int (*get)(struct sock *sk, int optval, void __user *user, int *len);
 174 #ifdef CONFIG_COMPAT
 175         int (*compat_get)(struct sock *sk, int optval,
 176                         void __user *user, int *len);
 177 #endif
 178         /* Use the module struct to lock set/get code in place */
 179         struct module *owner;
 180 };
 181
 182 /* Function to register/unregister hook points. */
 183 int nf_register_net_hook(struct net *net, const struct nf_hook_ops *ops);
 184 void nf_unregister_net_hook(struct net *net, const struct nf_hook_ops *ops);
 185 int nf_register_net_hooks(struct net *net, const struct nf_hook_ops *reg,
 186                           unsigned int n);
 187 void nf_unregister_net_hooks(struct net *net, const struct nf_hook_ops *reg,
 188                              unsigned int n);
 189
 190 /* Functions to register get/setsockopt ranges (non-inclusive).  You
 191    need to check permissions yourself! */
 192 int nf_register_sockopt(struct nf_sockopt_ops *reg);
 193 void nf_unregister_sockopt(struct nf_sockopt_ops *reg);
 194
 195 #ifdef CONFIG_JUMP_LABEL
 196 extern struct static_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS];
 197 #endif
 198
 199 int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state,
 200                  const struct nf_hook_entries *e, unsigned int i);
 201
 202 /**
 203  *      nf_hook - call a netfilter hook
 204  *
 205  *      Returns 1 if the hook has allowed the packet to pass.  The function
 206  *      okfn must be invoked by the caller in this case.  Any other return
 207  *      value indicates the packet has been consumed by the hook.
 208  */
 209 static inline int nf_hook(u_int8_t pf, unsigned int hook, struct net *net,
 210                           struct sock *sk, struct sk_buff *skb,
 211                           struct net_device *indev, struct net_device *outdev,
 212                           int (*okfn)(struct net *, struct sock *, struct sk_buff *))
 213 {
 214         struct nf_hook_entries *hook_head = NULL;
 215         int ret = 1;
 216
 217 #ifdef CONFIG_JUMP_LABEL
 218         if (__builtin_constant_p(pf) &&
 219             __builtin_constant_p(hook) &&
 220             !static_key_false(&nf_hooks_needed[pf][hook]))
 221                 return 1;
 222 #endif
 223
 224         rcu_read_lock();
 225         switch (pf) {
 226         case NFPROTO_IPV4:
 227                 hook_head = rcu_dereference(net->nf.hooks_ipv4[hook]);
 228                 break;
 229         case NFPROTO_IPV6:
 230                 hook_head = rcu_dereference(net->nf.hooks_ipv6[hook]);
 231                 break;
 232         case NFPROTO_ARP:
 233 #ifdef CONFIG_NETFILTER_FAMILY_ARP
 234                 if (WARN_ON_ONCE(hook >= ARRAY_SIZE(net->nf.hooks_arp)))
 235                         break;
 236                 hook_head = rcu_dereference(net->nf.hooks_arp[hook]);
 237 #endif
 238                 break;
 239         case NFPROTO_BRIDGE:
 240 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
 241                 hook_head = rcu_dereference(net->nf.hooks_bridge[hook]);
 242 #endif
 243                 break;
 244 #if IS_ENABLED(CONFIG_DECNET)
 245         case NFPROTO_DECNET:
 246                 hook_head = rcu_dereference(net->nf.hooks_decnet[hook]);
 247                 break;
 248 #endif
 249         default:
 250                 WARN_ON_ONCE(1);
 251                 break;
 252         }
 253
 254         if (hook_head) {
 255                 struct nf_hook_state state;
 256
 257                 nf_hook_state_init(&state, hook, pf, indev, outdev,
 258                                    sk, net, okfn);
 259
 260                 ret = nf_hook_slow(skb, &state, hook_head, 0);
 261         }
 262         rcu_read_unlock();
 263
 264         return ret;
 265 }
 266
 267 /* Activate hook; either okfn or kfree_skb called, unless a hook
 268    returns NF_STOLEN (in which case, it's up to the hook to deal with
 269    the consequences).
 270
 271    Returns -ERRNO if packet dropped.  Zero means queued, stolen or
 272    accepted.
 273 */
 274
 275 /* RR:
 276    > I don't want nf_hook to return anything because people might forget
 277    > about async and trust the return value to mean "packet was ok".
 278
 279    AK:
 280    Just document it clearly, then you can expect some sense from kernel
 281    coders :)
 282 */
 283
 284 static inline int
 285 NF_HOOK_COND(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk,
 286              struct sk_buff *skb, struct net_device *in, struct net_device *out,
 287              int (*okfn)(struct net *, struct sock *, struct sk_buff *),
 288              bool cond)
 289 {
 290         int ret;
 291
 292         if (!cond ||
 293             ((ret = nf_hook(pf, hook, net, sk, skb, in, out, okfn)) == 1))
 294                 ret = okfn(net, sk, skb);
 295         return ret;
 296 }
 297
 298 static inline int
 299 NF_HOOK(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk, struct sk_buff *skb,
 300         struct net_device *in, struct net_device *out,
 301         int (*okfn)(struct net *, struct sock *, struct sk_buff *))
 302 {
 303         int ret = nf_hook(pf, hook, net, sk, skb, in, out, okfn);
 304         if (ret == 1)
 305                 ret = okfn(net, sk, skb);
 306         return ret;
 307 }
 308
 309 static inline void
 310 NF_HOOK_LIST(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk,
 311              struct list_head *head, struct net_device *in, struct net_device *out,
 312              int (*okfn)(struct net *, struct sock *, struct sk_buff *))
 313 {
 314         struct sk_buff *skb, *next;
 315         struct list_head sublist;
 316
 317         INIT_LIST_HEAD(&sublist);
 318         list_for_each_entry_safe(skb, next, head, list) {
 319                 list_del(&skb->list);
 320                 if (nf_hook(pf, hook, net, sk, skb, in, out, okfn) == 1)
 321                         list_add_tail(&skb->list, &sublist);
 322         }
 323         /* Put passed packets back on main list */
 324         list_splice(&sublist, head);
 325 }
 326
 327 /* Call setsockopt() */
 328 int nf_setsockopt(struct sock *sk, u_int8_t pf, int optval, char __user *opt,
 329                   unsigned int len);
 330 int nf_getsockopt(struct sock *sk, u_int8_t pf, int optval, char __user *opt,
 331                   int *len);
 332 #ifdef CONFIG_COMPAT
 333 int compat_nf_setsockopt(struct sock *sk, u_int8_t pf, int optval,
 334                 char __user *opt, unsigned int len);
 335 int compat_nf_getsockopt(struct sock *sk, u_int8_t pf, int optval,
 336                 char __user *opt, int *len);
 337 #endif
 338
 339 struct flowi;
 340 struct nf_queue_entry;
 341
 342 __sum16 nf_checksum(struct sk_buff *skb, unsigned int hook,
 343                     unsigned int dataoff, u_int8_t protocol,
 344                     unsigned short family);
 345
 346 __sum16 nf_checksum_partial(struct sk_buff *skb, unsigned int hook,
 347                             unsigned int dataoff, unsigned int len,
 348                             u_int8_t protocol, unsigned short family);
 349 int nf_route(struct net *net, struct dst_entry **dst, struct flowi *fl,
 350              bool strict, unsigned short family);
 351 int nf_reroute(struct sk_buff *skb, struct nf_queue_entry *entry);
 352
 353 #include <net/flow.h>
 354
 355 struct nf_conn;
 356 enum nf_nat_manip_type;
 357 struct nlattr;
 358 enum ip_conntrack_dir;
 359
 360 struct nf_nat_hook {
 361         int (*parse_nat_setup)(struct nf_conn *ct, enum nf_nat_manip_type manip,
 362                                const struct nlattr *attr);
 363         void (*decode_session)(struct sk_buff *skb, struct flowi *fl);
 364         unsigned int (*manip_pkt)(struct sk_buff *skb, struct nf_conn *ct,
 365                                   enum nf_nat_manip_type mtype,
 366                                   enum ip_conntrack_dir dir);
 367 };
 368
 369 extern struct nf_nat_hook __rcu *nf_nat_hook;
 370
 371 static inline void
 372 nf_nat_decode_session(struct sk_buff *skb, struct flowi *fl, u_int8_t family)
 373 {
 374 #if IS_ENABLED(CONFIG_NF_NAT)
 375         struct nf_nat_hook *nat_hook;
 376
 377         rcu_read_lock();
 378         nat_hook = rcu_dereference(nf_nat_hook);
 379         if (nat_hook && nat_hook->decode_session)
 380                 nat_hook->decode_session(skb, fl);
 381         rcu_read_unlock();
 382 #endif
 383 }
 384
 385 #else /* !CONFIG_NETFILTER */
 386 static inline int
 387 NF_HOOK_COND(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk,
 388              struct sk_buff *skb, struct net_device *in, struct net_device *out,
 389              int (*okfn)(struct net *, struct sock *, struct sk_buff *),
 390              bool cond)
 391 {
 392         return okfn(net, sk, skb);
 393 }
 394
 395 static inline int
 396 NF_HOOK(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk,
 397         struct sk_buff *skb, struct net_device *in, struct net_device *out,
 398         int (*okfn)(struct net *, struct sock *, struct sk_buff *))
 399 {
 400         return okfn(net, sk, skb);
 401 }
 402
 403 static inline void
 404 NF_HOOK_LIST(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk,
 405              struct list_head *head, struct net_device *in, struct net_device *out,
 406              int (*okfn)(struct net *, struct sock *, struct sk_buff *))
 407 {
 408         /* nothing to do */
 409 }
 410
 411 static inline int nf_hook(u_int8_t pf, unsigned int hook, struct net *net,
 412                           struct sock *sk, struct sk_buff *skb,
 413                           struct net_device *indev, struct net_device *outdev,
 414                           int (*okfn)(struct net *, struct sock *, struct sk_buff *))
 415 {
 416         return 1;
 417 }
 418 struct flowi;
 419 static inline void
 420 nf_nat_decode_session(struct sk_buff *skb, struct flowi *fl, u_int8_t family)
 421 {
 422 }
 423 #endif /*CONFIG_NETFILTER*/
 424
 425 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 426 #include <linux/netfilter/nf_conntrack_zones_common.h>
 427
 428 extern void (*ip_ct_attach)(struct sk_buff *, const struct sk_buff *) __rcu;
 429 void nf_ct_attach(struct sk_buff *, const struct sk_buff *);
 430 struct nf_conntrack_tuple;
 431 bool nf_ct_get_tuple_skb(struct nf_conntrack_tuple *dst_tuple,
 432                          const struct sk_buff *skb);
 433 #else
 434 static inline void nf_ct_attach(struct sk_buff *new, struct sk_buff *skb) {}
 435 struct nf_conntrack_tuple;
 436 static inline bool nf_ct_get_tuple_skb(struct nf_conntrack_tuple *dst_tuple,
 437                                        const struct sk_buff *skb)
 438 {
 439         return false;
 440 }
 441 #endif
 442
 443 struct nf_conn;
 444 enum ip_conntrack_info;
 445
 446 struct nf_ct_hook {
 447         int (*update)(struct net *net, struct sk_buff *skb);
 448         void (*destroy)(struct nf_conntrack *);
 449         bool (*get_tuple_skb)(struct nf_conntrack_tuple *,
 450                               const struct sk_buff *);
 451 };
 452 extern struct nf_ct_hook __rcu *nf_ct_hook;
 453
 454 struct nlattr;
 455
 456 struct nfnl_ct_hook {
 457         struct nf_conn *(*get_ct)(const struct sk_buff *skb,
 458                                   enum ip_conntrack_info *ctinfo);
 459         size_t (*build_size)(const struct nf_conn *ct);
 460         int (*build)(struct sk_buff *skb, struct nf_conn *ct,
 461                      enum ip_conntrack_info ctinfo,
 462                      u_int16_t ct_attr, u_int16_t ct_info_attr);
 463         int (*parse)(const struct nlattr *attr, struct nf_conn *ct);
 464         int (*attach_expect)(const struct nlattr *attr, struct nf_conn *ct,
 465                              u32 portid, u32 report);
 466         void (*seq_adjust)(struct sk_buff *skb, struct nf_conn *ct,
 467                            enum ip_conntrack_info ctinfo, s32 off);
 468 };
 469 extern struct nfnl_ct_hook __rcu *nfnl_ct_hook;
 470
 471 /**
 472  * nf_skb_duplicated - TEE target has sent a packet
 473  *
 474  * When a xtables target sends a packet, the OUTPUT and POSTROUTING
 475  * hooks are traversed again, i.e. nft and xtables are invoked recursively.
 476  *
 477  * This is used by xtables TEE target to prevent the duplicated skb from
 478  * being duplicated again.
 479  */
 480 DECLARE_PER_CPU(bool, nf_skb_duplicated);
 481
 482 #endif /*__LINUX_NETFILTER_H*/