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/module.h> 15 #include <linux/netfilter_defs.h> 16 #include <linux/netdevice.h> 17 #include <linux/sockptr.h> 18 #include <net/net_namespace.h> 19 20 static inline int NF_DROP_GETERR(int verdict) 21 { 22 return -(verdict >> NF_VERDICT_QBITS); 23 } 24 25 static inline int nf_inet_addr_cmp(const union nf_inet_addr *a1, 26 const union nf_inet_addr *a2) 27 { 28 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 29 const unsigned long *ul1 = (const unsigned long *)a1; 30 const unsigned long *ul2 = (const unsigned long *)a2; 31 32 return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL; 33 #else 34 return a1->all[0] == a2->all[0] && 35 a1->all[1] == a2->all[1] && 36 a1->all[2] == a2->all[2] && 37 a1->all[3] == a2->all[3]; 38 #endif 39 } 40 41 static inline void nf_inet_addr_mask(const union nf_inet_addr *a1, 42 union nf_inet_addr *result, 43 const union nf_inet_addr *mask) 44 { 45 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 46 const unsigned long *ua = (const unsigned long *)a1; 47 unsigned long *ur = (unsigned long *)result; 48 const unsigned long *um = (const unsigned long *)mask; 49 50 ur[0] = ua[0] & um[0]; 51 ur[1] = ua[1] & um[1]; 52 #else 53 result->all[0] = a1->all[0] & mask->all[0]; 54 result->all[1] = a1->all[1] & mask->all[1]; 55 result->all[2] = a1->all[2] & mask->all[2]; 56 result->all[3] = a1->all[3] & mask->all[3]; 57 #endif 58 } 59 60 int netfilter_init(void); 61 62 struct sk_buff; 63 64 struct nf_hook_ops; 65 66 struct sock; 67 68 struct nf_hook_state { 69 u8 hook; 70 u8 pf; 71 struct net_device *in; 72 struct net_device *out; 73 struct sock *sk; 74 struct net *net; 75 int (*okfn)(struct net *, struct sock *, struct sk_buff *); 76 }; 77 78 typedef unsigned int nf_hookfn(void *priv, 79 struct sk_buff *skb, 80 const struct nf_hook_state *state); 81 enum nf_hook_ops_type { 82 NF_HOOK_OP_UNDEFINED, 83 NF_HOOK_OP_NF_TABLES, 84 NF_HOOK_OP_BPF, 85 }; 86 87 struct nf_hook_ops { 88 /* User fills in from here down. */ 89 nf_hookfn *hook; 90 struct net_device *dev; 91 void *priv; 92 u8 pf; 93 enum nf_hook_ops_type hook_ops_type:8; 94 unsigned int hooknum; 95 /* Hooks are ordered in ascending priority. */ 96 int priority; 97 }; 98 99 struct nf_hook_entry { 100 nf_hookfn *hook; 101 void *priv; 102 }; 103 104 struct nf_hook_entries_rcu_head { 105 struct rcu_head head; 106 void *allocation; 107 }; 108 109 struct nf_hook_entries { 110 u16 num_hook_entries; 111 /* padding */ 112 struct nf_hook_entry hooks[]; 113 114 /* trailer: pointers to original orig_ops of each hook, 115 * followed by rcu_head and scratch space used for freeing 116 * the structure via call_rcu. 117 * 118 * This is not part of struct nf_hook_entry since its only 119 * needed in slow path (hook register/unregister): 120 * const struct nf_hook_ops *orig_ops[] 121 * 122 * For the same reason, we store this at end -- its 123 * only needed when a hook is deleted, not during 124 * packet path processing: 125 * struct nf_hook_entries_rcu_head head 126 */ 127 }; 128 129 #ifdef CONFIG_NETFILTER 130 static inline struct nf_hook_ops **nf_hook_entries_get_hook_ops(const struct nf_hook_entries *e) 131 { 132 unsigned int n = e->num_hook_entries; 133 const void *hook_end; 134 135 hook_end = &e->hooks[n]; /* this is *past* ->hooks[]! */ 136 137 return (struct nf_hook_ops **)hook_end; 138 } 139 140 static inline int 141 nf_hook_entry_hookfn(const struct nf_hook_entry *entry, struct sk_buff *skb, 142 struct nf_hook_state *state) 143 { 144 return entry->hook(entry->priv, skb, state); 145 } 146 147 static inline void nf_hook_state_init(struct nf_hook_state *p, 148 unsigned int hook, 149 u_int8_t pf, 150 struct net_device *indev, 151 struct net_device *outdev, 152 struct sock *sk, 153 struct net *net, 154 int (*okfn)(struct net *, struct sock *, struct sk_buff *)) 155 { 156 p->hook = hook; 157 p->pf = pf; 158 p->in = indev; 159 p->out = outdev; 160 p->sk = sk; 161 p->net = net; 162 p->okfn = okfn; 163 } 164 165 166 167 struct nf_sockopt_ops { 168 struct list_head list; 169 170 u_int8_t pf; 171 172 /* Non-inclusive ranges: use 0/0/NULL to never get called. */ 173 int set_optmin; 174 int set_optmax; 175 int (*set)(struct sock *sk, int optval, sockptr_t arg, 176 unsigned int len); 177 int get_optmin; 178 int get_optmax; 179 int (*get)(struct sock *sk, int optval, void __user *user, int *len); 180 /* Use the module struct to lock set/get code in place */ 181 struct module *owner; 182 }; 183 184 /* Function to register/unregister hook points. */ 185 int nf_register_net_hook(struct net *net, const struct nf_hook_ops *ops); 186 void nf_unregister_net_hook(struct net *net, const struct nf_hook_ops *ops); 187 int nf_register_net_hooks(struct net *net, const struct nf_hook_ops *reg, 188 unsigned int n); 189 void nf_unregister_net_hooks(struct net *net, const struct nf_hook_ops *reg, 190 unsigned int n); 191 192 /* Functions to register get/setsockopt ranges (non-inclusive). You 193 need to check permissions yourself! */ 194 int nf_register_sockopt(struct nf_sockopt_ops *reg); 195 void nf_unregister_sockopt(struct nf_sockopt_ops *reg); 196 197 #ifdef CONFIG_JUMP_LABEL 198 extern struct static_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS]; 199 #endif 200 201 int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state, 202 const struct nf_hook_entries *e, unsigned int i); 203 204 void nf_hook_slow_list(struct list_head *head, struct nf_hook_state *state, 205 const struct nf_hook_entries *e); 206 /** 207 * nf_hook - call a netfilter hook 208 * 209 * Returns 1 if the hook has allowed the packet to pass. The function 210 * okfn must be invoked by the caller in this case. Any other return 211 * value indicates the packet has been consumed by the hook. 212 */ 213 static inline int nf_hook(u_int8_t pf, unsigned int hook, struct net *net, 214 struct sock *sk, struct sk_buff *skb, 215 struct net_device *indev, struct net_device *outdev, 216 int (*okfn)(struct net *, struct sock *, struct sk_buff *)) 217 { 218 struct nf_hook_entries *hook_head = NULL; 219 int ret = 1; 220 221 #ifdef CONFIG_JUMP_LABEL 222 if (__builtin_constant_p(pf) && 223 __builtin_constant_p(hook) && 224 !static_key_false(&nf_hooks_needed[pf][hook])) 225 return 1; 226 #endif 227 228 rcu_read_lock(); 229 switch (pf) { 230 case NFPROTO_IPV4: 231 hook_head = rcu_dereference(net->nf.hooks_ipv4[hook]); 232 break; 233 case NFPROTO_IPV6: 234 hook_head = rcu_dereference(net->nf.hooks_ipv6[hook]); 235 break; 236 case NFPROTO_ARP: 237 #ifdef CONFIG_NETFILTER_FAMILY_ARP 238 if (WARN_ON_ONCE(hook >= ARRAY_SIZE(net->nf.hooks_arp))) 239 break; 240 hook_head = rcu_dereference(net->nf.hooks_arp[hook]); 241 #endif 242 break; 243 case NFPROTO_BRIDGE: 244 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE 245 hook_head = rcu_dereference(net->nf.hooks_bridge[hook]); 246 #endif 247 break; 248 default: 249 WARN_ON_ONCE(1); 250 break; 251 } 252 253 if (hook_head) { 254 struct nf_hook_state state; 255 256 nf_hook_state_init(&state, hook, pf, indev, outdev, 257 sk, net, okfn); 258 259 ret = nf_hook_slow(skb, &state, hook_head, 0); 260 } 261 rcu_read_unlock(); 262 263 return ret; 264 } 265 266 /* Activate hook; either okfn or kfree_skb called, unless a hook 267 returns NF_STOLEN (in which case, it's up to the hook to deal with 268 the consequences). 269 270 Returns -ERRNO if packet dropped. Zero means queued, stolen or 271 accepted. 272 */ 273 274 /* RR: 275 > I don't want nf_hook to return anything because people might forget 276 > about async and trust the return value to mean "packet was ok". 277 278 AK: 279 Just document it clearly, then you can expect some sense from kernel 280 coders :) 281 */ 282 283 static inline int 284 NF_HOOK_COND(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk, 285 struct sk_buff *skb, struct net_device *in, struct net_device *out, 286 int (*okfn)(struct net *, struct sock *, struct sk_buff *), 287 bool cond) 288 { 289 int ret; 290 291 if (!cond || 292 ((ret = nf_hook(pf, hook, net, sk, skb, in, out, okfn)) == 1)) 293 ret = okfn(net, sk, skb); 294 return ret; 295 } 296 297 static inline int 298 NF_HOOK(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk, struct sk_buff *skb, 299 struct net_device *in, struct net_device *out, 300 int (*okfn)(struct net *, struct sock *, struct sk_buff *)) 301 { 302 int ret = nf_hook(pf, hook, net, sk, skb, in, out, okfn); 303 if (ret == 1) 304 ret = okfn(net, sk, skb); 305 return ret; 306 } 307 308 static inline void 309 NF_HOOK_LIST(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk, 310 struct list_head *head, struct net_device *in, struct net_device *out, 311 int (*okfn)(struct net *, struct sock *, struct sk_buff *)) 312 { 313 struct nf_hook_entries *hook_head = NULL; 314 315 #ifdef CONFIG_JUMP_LABEL 316 if (__builtin_constant_p(pf) && 317 __builtin_constant_p(hook) && 318 !static_key_false(&nf_hooks_needed[pf][hook])) 319 return; 320 #endif 321 322 rcu_read_lock(); 323 switch (pf) { 324 case NFPROTO_IPV4: 325 hook_head = rcu_dereference(net->nf.hooks_ipv4[hook]); 326 break; 327 case NFPROTO_IPV6: 328 hook_head = rcu_dereference(net->nf.hooks_ipv6[hook]); 329 break; 330 default: 331 WARN_ON_ONCE(1); 332 break; 333 } 334 335 if (hook_head) { 336 struct nf_hook_state state; 337 338 nf_hook_state_init(&state, hook, pf, in, out, sk, net, okfn); 339 340 nf_hook_slow_list(head, &state, hook_head); 341 } 342 rcu_read_unlock(); 343 } 344 345 /* Call setsockopt() */ 346 int nf_setsockopt(struct sock *sk, u_int8_t pf, int optval, sockptr_t opt, 347 unsigned int len); 348 int nf_getsockopt(struct sock *sk, u_int8_t pf, int optval, char __user *opt, 349 int *len); 350 351 struct flowi; 352 struct nf_queue_entry; 353 354 __sum16 nf_checksum(struct sk_buff *skb, unsigned int hook, 355 unsigned int dataoff, u_int8_t protocol, 356 unsigned short family); 357 358 __sum16 nf_checksum_partial(struct sk_buff *skb, unsigned int hook, 359 unsigned int dataoff, unsigned int len, 360 u_int8_t protocol, unsigned short family); 361 int nf_route(struct net *net, struct dst_entry **dst, struct flowi *fl, 362 bool strict, unsigned short family); 363 int nf_reroute(struct sk_buff *skb, struct nf_queue_entry *entry); 364 365 #include <net/flow.h> 366 367 struct nf_conn; 368 enum nf_nat_manip_type; 369 struct nlattr; 370 enum ip_conntrack_dir; 371 372 struct nf_nat_hook { 373 int (*parse_nat_setup)(struct nf_conn *ct, enum nf_nat_manip_type manip, 374 const struct nlattr *attr); 375 void (*decode_session)(struct sk_buff *skb, struct flowi *fl); 376 unsigned int (*manip_pkt)(struct sk_buff *skb, struct nf_conn *ct, 377 enum nf_nat_manip_type mtype, 378 enum ip_conntrack_dir dir); 379 void (*remove_nat_bysrc)(struct nf_conn *ct); 380 }; 381 382 extern const struct nf_nat_hook __rcu *nf_nat_hook; 383 384 static inline void 385 nf_nat_decode_session(struct sk_buff *skb, struct flowi *fl, u_int8_t family) 386 { 387 #if IS_ENABLED(CONFIG_NF_NAT) 388 const struct nf_nat_hook *nat_hook; 389 390 rcu_read_lock(); 391 nat_hook = rcu_dereference(nf_nat_hook); 392 if (nat_hook && nat_hook->decode_session) 393 nat_hook->decode_session(skb, fl); 394 rcu_read_unlock(); 395 #endif 396 } 397 398 #else /* !CONFIG_NETFILTER */ 399 static inline int 400 NF_HOOK_COND(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk, 401 struct sk_buff *skb, struct net_device *in, struct net_device *out, 402 int (*okfn)(struct net *, struct sock *, struct sk_buff *), 403 bool cond) 404 { 405 return okfn(net, sk, skb); 406 } 407 408 static inline int 409 NF_HOOK(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk, 410 struct sk_buff *skb, struct net_device *in, struct net_device *out, 411 int (*okfn)(struct net *, struct sock *, struct sk_buff *)) 412 { 413 return okfn(net, sk, skb); 414 } 415 416 static inline void 417 NF_HOOK_LIST(uint8_t pf, unsigned int hook, struct net *net, struct sock *sk, 418 struct list_head *head, struct net_device *in, struct net_device *out, 419 int (*okfn)(struct net *, struct sock *, struct sk_buff *)) 420 { 421 /* nothing to do */ 422 } 423 424 static inline int nf_hook(u_int8_t pf, unsigned int hook, struct net *net, 425 struct sock *sk, struct sk_buff *skb, 426 struct net_device *indev, struct net_device *outdev, 427 int (*okfn)(struct net *, struct sock *, struct sk_buff *)) 428 { 429 return 1; 430 } 431 struct flowi; 432 static inline void 433 nf_nat_decode_session(struct sk_buff *skb, struct flowi *fl, u_int8_t family) 434 { 435 } 436 #endif /*CONFIG_NETFILTER*/ 437 438 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 439 #include <linux/netfilter/nf_conntrack_zones_common.h> 440 441 void nf_ct_attach(struct sk_buff *, const struct sk_buff *); 442 void nf_ct_set_closing(struct nf_conntrack *nfct); 443 struct nf_conntrack_tuple; 444 bool nf_ct_get_tuple_skb(struct nf_conntrack_tuple *dst_tuple, 445 const struct sk_buff *skb); 446 #else 447 static inline void nf_ct_attach(struct sk_buff *new, struct sk_buff *skb) {} 448 static inline void nf_ct_set_closing(struct nf_conntrack *nfct) {} 449 struct nf_conntrack_tuple; 450 static inline bool nf_ct_get_tuple_skb(struct nf_conntrack_tuple *dst_tuple, 451 const struct sk_buff *skb) 452 { 453 return false; 454 } 455 #endif 456 457 struct nf_conn; 458 enum ip_conntrack_info; 459 460 struct nf_ct_hook { 461 int (*update)(struct net *net, struct sk_buff *skb); 462 void (*destroy)(struct nf_conntrack *); 463 bool (*get_tuple_skb)(struct nf_conntrack_tuple *, 464 const struct sk_buff *); 465 void (*attach)(struct sk_buff *nskb, const struct sk_buff *skb); 466 void (*set_closing)(struct nf_conntrack *nfct); 467 }; 468 extern const struct nf_ct_hook __rcu *nf_ct_hook; 469 470 struct nlattr; 471 472 struct nfnl_ct_hook { 473 size_t (*build_size)(const struct nf_conn *ct); 474 int (*build)(struct sk_buff *skb, struct nf_conn *ct, 475 enum ip_conntrack_info ctinfo, 476 u_int16_t ct_attr, u_int16_t ct_info_attr); 477 int (*parse)(const struct nlattr *attr, struct nf_conn *ct); 478 int (*attach_expect)(const struct nlattr *attr, struct nf_conn *ct, 479 u32 portid, u32 report); 480 void (*seq_adjust)(struct sk_buff *skb, struct nf_conn *ct, 481 enum ip_conntrack_info ctinfo, s32 off); 482 }; 483 extern const struct nfnl_ct_hook __rcu *nfnl_ct_hook; 484 485 struct nf_defrag_hook { 486 struct module *owner; 487 int (*enable)(struct net *net); 488 void (*disable)(struct net *net); 489 }; 490 491 extern const struct nf_defrag_hook __rcu *nf_defrag_v4_hook; 492 extern const struct nf_defrag_hook __rcu *nf_defrag_v6_hook; 493 494 /* 495 * nf_skb_duplicated - TEE target has sent a packet 496 * 497 * When a xtables target sends a packet, the OUTPUT and POSTROUTING 498 * hooks are traversed again, i.e. nft and xtables are invoked recursively. 499 * 500 * This is used by xtables TEE target to prevent the duplicated skb from 501 * being duplicated again. 502 */ 503 DECLARE_PER_CPU(bool, nf_skb_duplicated); 504 505 /* 506 * Contains bitmask of ctnetlink event subscribers, if any. 507 * Can't be pernet due to NETLINK_LISTEN_ALL_NSID setsockopt flag. 508 */ 509 extern u8 nf_ctnetlink_has_listener; 510 #endif /*__LINUX_NETFILTER_H*/ 511