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