1 /* SPDX-License-Identifier: GPL-2.0-only */ 2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3 */ 4 #ifndef _LINUX_BPF_H 5 #define _LINUX_BPF_H 1 6 7 #include <uapi/linux/bpf.h> 8 9 #include <linux/workqueue.h> 10 #include <linux/file.h> 11 #include <linux/percpu.h> 12 #include <linux/err.h> 13 #include <linux/rbtree_latch.h> 14 #include <linux/numa.h> 15 #include <linux/wait.h> 16 #include <linux/u64_stats_sync.h> 17 18 struct bpf_verifier_env; 19 struct perf_event; 20 struct bpf_prog; 21 struct bpf_map; 22 struct sock; 23 struct seq_file; 24 struct btf; 25 struct btf_type; 26 27 /* map is generic key/value storage optionally accesible by eBPF programs */ 28 struct bpf_map_ops { 29 /* funcs callable from userspace (via syscall) */ 30 int (*map_alloc_check)(union bpf_attr *attr); 31 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 32 void (*map_release)(struct bpf_map *map, struct file *map_file); 33 void (*map_free)(struct bpf_map *map); 34 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 35 void (*map_release_uref)(struct bpf_map *map); 36 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 37 38 /* funcs callable from userspace and from eBPF programs */ 39 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 40 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 41 int (*map_delete_elem)(struct bpf_map *map, void *key); 42 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 43 int (*map_pop_elem)(struct bpf_map *map, void *value); 44 int (*map_peek_elem)(struct bpf_map *map, void *value); 45 46 /* funcs called by prog_array and perf_event_array map */ 47 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 48 int fd); 49 void (*map_fd_put_ptr)(void *ptr); 50 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 51 u32 (*map_fd_sys_lookup_elem)(void *ptr); 52 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 53 struct seq_file *m); 54 int (*map_check_btf)(const struct bpf_map *map, 55 const struct btf *btf, 56 const struct btf_type *key_type, 57 const struct btf_type *value_type); 58 59 /* Direct value access helpers. */ 60 int (*map_direct_value_addr)(const struct bpf_map *map, 61 u64 *imm, u32 off); 62 int (*map_direct_value_meta)(const struct bpf_map *map, 63 u64 imm, u32 *off); 64 }; 65 66 struct bpf_map_memory { 67 u32 pages; 68 struct user_struct *user; 69 }; 70 71 struct bpf_map { 72 /* The first two cachelines with read-mostly members of which some 73 * are also accessed in fast-path (e.g. ops, max_entries). 74 */ 75 const struct bpf_map_ops *ops ____cacheline_aligned; 76 struct bpf_map *inner_map_meta; 77 #ifdef CONFIG_SECURITY 78 void *security; 79 #endif 80 enum bpf_map_type map_type; 81 u32 key_size; 82 u32 value_size; 83 u32 max_entries; 84 u32 map_flags; 85 int spin_lock_off; /* >=0 valid offset, <0 error */ 86 u32 id; 87 int numa_node; 88 u32 btf_key_type_id; 89 u32 btf_value_type_id; 90 struct btf *btf; 91 struct bpf_map_memory memory; 92 bool unpriv_array; 93 bool frozen; /* write-once */ 94 /* 48 bytes hole */ 95 96 /* The 3rd and 4th cacheline with misc members to avoid false sharing 97 * particularly with refcounting. 98 */ 99 atomic_t refcnt ____cacheline_aligned; 100 atomic_t usercnt; 101 struct work_struct work; 102 char name[BPF_OBJ_NAME_LEN]; 103 }; 104 105 static inline bool map_value_has_spin_lock(const struct bpf_map *map) 106 { 107 return map->spin_lock_off >= 0; 108 } 109 110 static inline void check_and_init_map_lock(struct bpf_map *map, void *dst) 111 { 112 if (likely(!map_value_has_spin_lock(map))) 113 return; 114 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) = 115 (struct bpf_spin_lock){}; 116 } 117 118 /* copy everything but bpf_spin_lock */ 119 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 120 { 121 if (unlikely(map_value_has_spin_lock(map))) { 122 u32 off = map->spin_lock_off; 123 124 memcpy(dst, src, off); 125 memcpy(dst + off + sizeof(struct bpf_spin_lock), 126 src + off + sizeof(struct bpf_spin_lock), 127 map->value_size - off - sizeof(struct bpf_spin_lock)); 128 } else { 129 memcpy(dst, src, map->value_size); 130 } 131 } 132 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 133 bool lock_src); 134 135 struct bpf_offload_dev; 136 struct bpf_offloaded_map; 137 138 struct bpf_map_dev_ops { 139 int (*map_get_next_key)(struct bpf_offloaded_map *map, 140 void *key, void *next_key); 141 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 142 void *key, void *value); 143 int (*map_update_elem)(struct bpf_offloaded_map *map, 144 void *key, void *value, u64 flags); 145 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 146 }; 147 148 struct bpf_offloaded_map { 149 struct bpf_map map; 150 struct net_device *netdev; 151 const struct bpf_map_dev_ops *dev_ops; 152 void *dev_priv; 153 struct list_head offloads; 154 }; 155 156 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 157 { 158 return container_of(map, struct bpf_offloaded_map, map); 159 } 160 161 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 162 { 163 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 164 } 165 166 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 167 { 168 return map->btf && map->ops->map_seq_show_elem; 169 } 170 171 int map_check_no_btf(const struct bpf_map *map, 172 const struct btf *btf, 173 const struct btf_type *key_type, 174 const struct btf_type *value_type); 175 176 extern const struct bpf_map_ops bpf_map_offload_ops; 177 178 /* function argument constraints */ 179 enum bpf_arg_type { 180 ARG_DONTCARE = 0, /* unused argument in helper function */ 181 182 /* the following constraints used to prototype 183 * bpf_map_lookup/update/delete_elem() functions 184 */ 185 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 186 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 187 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 188 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */ 189 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */ 190 191 /* the following constraints used to prototype bpf_memcmp() and other 192 * functions that access data on eBPF program stack 193 */ 194 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 195 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */ 196 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized, 197 * helper function must fill all bytes or clear 198 * them in error case. 199 */ 200 201 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 202 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 203 204 ARG_PTR_TO_CTX, /* pointer to context */ 205 ARG_ANYTHING, /* any (initialized) argument is ok */ 206 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 207 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 208 ARG_PTR_TO_INT, /* pointer to int */ 209 ARG_PTR_TO_LONG, /* pointer to long */ 210 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 211 }; 212 213 /* type of values returned from helper functions */ 214 enum bpf_return_type { 215 RET_INTEGER, /* function returns integer */ 216 RET_VOID, /* function doesn't return anything */ 217 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 218 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */ 219 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */ 220 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */ 221 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */ 222 }; 223 224 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 225 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 226 * instructions after verifying 227 */ 228 struct bpf_func_proto { 229 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 230 bool gpl_only; 231 bool pkt_access; 232 enum bpf_return_type ret_type; 233 enum bpf_arg_type arg1_type; 234 enum bpf_arg_type arg2_type; 235 enum bpf_arg_type arg3_type; 236 enum bpf_arg_type arg4_type; 237 enum bpf_arg_type arg5_type; 238 }; 239 240 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 241 * the first argument to eBPF programs. 242 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 243 */ 244 struct bpf_context; 245 246 enum bpf_access_type { 247 BPF_READ = 1, 248 BPF_WRITE = 2 249 }; 250 251 /* types of values stored in eBPF registers */ 252 /* Pointer types represent: 253 * pointer 254 * pointer + imm 255 * pointer + (u16) var 256 * pointer + (u16) var + imm 257 * if (range > 0) then [ptr, ptr + range - off) is safe to access 258 * if (id > 0) means that some 'var' was added 259 * if (off > 0) means that 'imm' was added 260 */ 261 enum bpf_reg_type { 262 NOT_INIT = 0, /* nothing was written into register */ 263 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 264 PTR_TO_CTX, /* reg points to bpf_context */ 265 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 266 PTR_TO_MAP_VALUE, /* reg points to map element value */ 267 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */ 268 PTR_TO_STACK, /* reg == frame_pointer + offset */ 269 PTR_TO_PACKET_META, /* skb->data - meta_len */ 270 PTR_TO_PACKET, /* reg points to skb->data */ 271 PTR_TO_PACKET_END, /* skb->data + headlen */ 272 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 273 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 274 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */ 275 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 276 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */ 277 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 278 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */ 279 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 280 }; 281 282 /* The information passed from prog-specific *_is_valid_access 283 * back to the verifier. 284 */ 285 struct bpf_insn_access_aux { 286 enum bpf_reg_type reg_type; 287 int ctx_field_size; 288 }; 289 290 static inline void 291 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 292 { 293 aux->ctx_field_size = size; 294 } 295 296 struct bpf_prog_ops { 297 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 298 union bpf_attr __user *uattr); 299 }; 300 301 struct bpf_verifier_ops { 302 /* return eBPF function prototype for verification */ 303 const struct bpf_func_proto * 304 (*get_func_proto)(enum bpf_func_id func_id, 305 const struct bpf_prog *prog); 306 307 /* return true if 'size' wide access at offset 'off' within bpf_context 308 * with 'type' (read or write) is allowed 309 */ 310 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 311 const struct bpf_prog *prog, 312 struct bpf_insn_access_aux *info); 313 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 314 const struct bpf_prog *prog); 315 int (*gen_ld_abs)(const struct bpf_insn *orig, 316 struct bpf_insn *insn_buf); 317 u32 (*convert_ctx_access)(enum bpf_access_type type, 318 const struct bpf_insn *src, 319 struct bpf_insn *dst, 320 struct bpf_prog *prog, u32 *target_size); 321 }; 322 323 struct bpf_prog_offload_ops { 324 /* verifier basic callbacks */ 325 int (*insn_hook)(struct bpf_verifier_env *env, 326 int insn_idx, int prev_insn_idx); 327 int (*finalize)(struct bpf_verifier_env *env); 328 /* verifier optimization callbacks (called after .finalize) */ 329 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 330 struct bpf_insn *insn); 331 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 332 /* program management callbacks */ 333 int (*prepare)(struct bpf_prog *prog); 334 int (*translate)(struct bpf_prog *prog); 335 void (*destroy)(struct bpf_prog *prog); 336 }; 337 338 struct bpf_prog_offload { 339 struct bpf_prog *prog; 340 struct net_device *netdev; 341 struct bpf_offload_dev *offdev; 342 void *dev_priv; 343 struct list_head offloads; 344 bool dev_state; 345 bool opt_failed; 346 void *jited_image; 347 u32 jited_len; 348 }; 349 350 enum bpf_cgroup_storage_type { 351 BPF_CGROUP_STORAGE_SHARED, 352 BPF_CGROUP_STORAGE_PERCPU, 353 __BPF_CGROUP_STORAGE_MAX 354 }; 355 356 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 357 358 struct bpf_prog_stats { 359 u64 cnt; 360 u64 nsecs; 361 struct u64_stats_sync syncp; 362 }; 363 364 struct bpf_prog_aux { 365 atomic_t refcnt; 366 u32 used_map_cnt; 367 u32 max_ctx_offset; 368 u32 max_pkt_offset; 369 u32 max_tp_access; 370 u32 stack_depth; 371 u32 id; 372 u32 func_cnt; /* used by non-func prog as the number of func progs */ 373 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 374 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 375 bool offload_requested; 376 struct bpf_prog **func; 377 void *jit_data; /* JIT specific data. arch dependent */ 378 struct latch_tree_node ksym_tnode; 379 struct list_head ksym_lnode; 380 const struct bpf_prog_ops *ops; 381 struct bpf_map **used_maps; 382 struct bpf_prog *prog; 383 struct user_struct *user; 384 u64 load_time; /* ns since boottime */ 385 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 386 char name[BPF_OBJ_NAME_LEN]; 387 #ifdef CONFIG_SECURITY 388 void *security; 389 #endif 390 struct bpf_prog_offload *offload; 391 struct btf *btf; 392 struct bpf_func_info *func_info; 393 /* bpf_line_info loaded from userspace. linfo->insn_off 394 * has the xlated insn offset. 395 * Both the main and sub prog share the same linfo. 396 * The subprog can access its first linfo by 397 * using the linfo_idx. 398 */ 399 struct bpf_line_info *linfo; 400 /* jited_linfo is the jited addr of the linfo. It has a 401 * one to one mapping to linfo: 402 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 403 * Both the main and sub prog share the same jited_linfo. 404 * The subprog can access its first jited_linfo by 405 * using the linfo_idx. 406 */ 407 void **jited_linfo; 408 u32 func_info_cnt; 409 u32 nr_linfo; 410 /* subprog can use linfo_idx to access its first linfo and 411 * jited_linfo. 412 * main prog always has linfo_idx == 0 413 */ 414 u32 linfo_idx; 415 struct bpf_prog_stats __percpu *stats; 416 union { 417 struct work_struct work; 418 struct rcu_head rcu; 419 }; 420 }; 421 422 struct bpf_array { 423 struct bpf_map map; 424 u32 elem_size; 425 u32 index_mask; 426 /* 'ownership' of prog_array is claimed by the first program that 427 * is going to use this map or by the first program which FD is stored 428 * in the map to make sure that all callers and callees have the same 429 * prog_type and JITed flag 430 */ 431 enum bpf_prog_type owner_prog_type; 432 bool owner_jited; 433 union { 434 char value[0] __aligned(8); 435 void *ptrs[0] __aligned(8); 436 void __percpu *pptrs[0] __aligned(8); 437 }; 438 }; 439 440 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 441 #define MAX_TAIL_CALL_CNT 32 442 443 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 444 BPF_F_RDONLY_PROG | \ 445 BPF_F_WRONLY | \ 446 BPF_F_WRONLY_PROG) 447 448 #define BPF_MAP_CAN_READ BIT(0) 449 #define BPF_MAP_CAN_WRITE BIT(1) 450 451 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 452 { 453 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 454 455 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 456 * not possible. 457 */ 458 if (access_flags & BPF_F_RDONLY_PROG) 459 return BPF_MAP_CAN_READ; 460 else if (access_flags & BPF_F_WRONLY_PROG) 461 return BPF_MAP_CAN_WRITE; 462 else 463 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 464 } 465 466 static inline bool bpf_map_flags_access_ok(u32 access_flags) 467 { 468 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 469 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 470 } 471 472 struct bpf_event_entry { 473 struct perf_event *event; 474 struct file *perf_file; 475 struct file *map_file; 476 struct rcu_head rcu; 477 }; 478 479 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); 480 int bpf_prog_calc_tag(struct bpf_prog *fp); 481 482 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 483 484 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 485 unsigned long off, unsigned long len); 486 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 487 const struct bpf_insn *src, 488 struct bpf_insn *dst, 489 struct bpf_prog *prog, 490 u32 *target_size); 491 492 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 493 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 494 495 /* an array of programs to be executed under rcu_lock. 496 * 497 * Typical usage: 498 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN); 499 * 500 * the structure returned by bpf_prog_array_alloc() should be populated 501 * with program pointers and the last pointer must be NULL. 502 * The user has to keep refcnt on the program and make sure the program 503 * is removed from the array before bpf_prog_put(). 504 * The 'struct bpf_prog_array *' should only be replaced with xchg() 505 * since other cpus are walking the array of pointers in parallel. 506 */ 507 struct bpf_prog_array_item { 508 struct bpf_prog *prog; 509 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 510 }; 511 512 struct bpf_prog_array { 513 struct rcu_head rcu; 514 struct bpf_prog_array_item items[0]; 515 }; 516 517 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 518 void bpf_prog_array_free(struct bpf_prog_array *progs); 519 int bpf_prog_array_length(struct bpf_prog_array *progs); 520 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 521 __u32 __user *prog_ids, u32 cnt); 522 523 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 524 struct bpf_prog *old_prog); 525 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 526 u32 *prog_ids, u32 request_cnt, 527 u32 *prog_cnt); 528 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 529 struct bpf_prog *exclude_prog, 530 struct bpf_prog *include_prog, 531 struct bpf_prog_array **new_array); 532 533 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \ 534 ({ \ 535 struct bpf_prog_array_item *_item; \ 536 struct bpf_prog *_prog; \ 537 struct bpf_prog_array *_array; \ 538 u32 _ret = 1; \ 539 preempt_disable(); \ 540 rcu_read_lock(); \ 541 _array = rcu_dereference(array); \ 542 if (unlikely(check_non_null && !_array))\ 543 goto _out; \ 544 _item = &_array->items[0]; \ 545 while ((_prog = READ_ONCE(_item->prog))) { \ 546 bpf_cgroup_storage_set(_item->cgroup_storage); \ 547 _ret &= func(_prog, ctx); \ 548 _item++; \ 549 } \ 550 _out: \ 551 rcu_read_unlock(); \ 552 preempt_enable(); \ 553 _ret; \ 554 }) 555 556 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs 557 * so BPF programs can request cwr for TCP packets. 558 * 559 * Current cgroup skb programs can only return 0 or 1 (0 to drop the 560 * packet. This macro changes the behavior so the low order bit 561 * indicates whether the packet should be dropped (0) or not (1) 562 * and the next bit is a congestion notification bit. This could be 563 * used by TCP to call tcp_enter_cwr() 564 * 565 * Hence, new allowed return values of CGROUP EGRESS BPF programs are: 566 * 0: drop packet 567 * 1: keep packet 568 * 2: drop packet and cn 569 * 3: keep packet and cn 570 * 571 * This macro then converts it to one of the NET_XMIT or an error 572 * code that is then interpreted as drop packet (and no cn): 573 * 0: NET_XMIT_SUCCESS skb should be transmitted 574 * 1: NET_XMIT_DROP skb should be dropped and cn 575 * 2: NET_XMIT_CN skb should be transmitted and cn 576 * 3: -EPERM skb should be dropped 577 */ 578 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \ 579 ({ \ 580 struct bpf_prog_array_item *_item; \ 581 struct bpf_prog *_prog; \ 582 struct bpf_prog_array *_array; \ 583 u32 ret; \ 584 u32 _ret = 1; \ 585 u32 _cn = 0; \ 586 preempt_disable(); \ 587 rcu_read_lock(); \ 588 _array = rcu_dereference(array); \ 589 _item = &_array->items[0]; \ 590 while ((_prog = READ_ONCE(_item->prog))) { \ 591 bpf_cgroup_storage_set(_item->cgroup_storage); \ 592 ret = func(_prog, ctx); \ 593 _ret &= (ret & 1); \ 594 _cn |= (ret & 2); \ 595 _item++; \ 596 } \ 597 rcu_read_unlock(); \ 598 preempt_enable(); \ 599 if (_ret) \ 600 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \ 601 else \ 602 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \ 603 _ret; \ 604 }) 605 606 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \ 607 __BPF_PROG_RUN_ARRAY(array, ctx, func, false) 608 609 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \ 610 __BPF_PROG_RUN_ARRAY(array, ctx, func, true) 611 612 #ifdef CONFIG_BPF_SYSCALL 613 DECLARE_PER_CPU(int, bpf_prog_active); 614 615 extern const struct file_operations bpf_map_fops; 616 extern const struct file_operations bpf_prog_fops; 617 618 #define BPF_PROG_TYPE(_id, _name) \ 619 extern const struct bpf_prog_ops _name ## _prog_ops; \ 620 extern const struct bpf_verifier_ops _name ## _verifier_ops; 621 #define BPF_MAP_TYPE(_id, _ops) \ 622 extern const struct bpf_map_ops _ops; 623 #include <linux/bpf_types.h> 624 #undef BPF_PROG_TYPE 625 #undef BPF_MAP_TYPE 626 627 extern const struct bpf_prog_ops bpf_offload_prog_ops; 628 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 629 extern const struct bpf_verifier_ops xdp_analyzer_ops; 630 631 struct bpf_prog *bpf_prog_get(u32 ufd); 632 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 633 bool attach_drv); 634 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i); 635 void bpf_prog_sub(struct bpf_prog *prog, int i); 636 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog); 637 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 638 void bpf_prog_put(struct bpf_prog *prog); 639 int __bpf_prog_charge(struct user_struct *user, u32 pages); 640 void __bpf_prog_uncharge(struct user_struct *user, u32 pages); 641 642 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 643 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 644 645 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 646 struct bpf_map *__bpf_map_get(struct fd f); 647 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref); 648 void bpf_map_put_with_uref(struct bpf_map *map); 649 void bpf_map_put(struct bpf_map *map); 650 int bpf_map_charge_memlock(struct bpf_map *map, u32 pages); 651 void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages); 652 int bpf_map_charge_init(struct bpf_map_memory *mem, size_t size); 653 void bpf_map_charge_finish(struct bpf_map_memory *mem); 654 void bpf_map_charge_move(struct bpf_map_memory *dst, 655 struct bpf_map_memory *src); 656 void *bpf_map_area_alloc(size_t size, int numa_node); 657 void bpf_map_area_free(void *base); 658 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 659 660 extern int sysctl_unprivileged_bpf_disabled; 661 extern int sysctl_bpf_stats_enabled; 662 663 int bpf_map_new_fd(struct bpf_map *map, int flags); 664 int bpf_prog_new_fd(struct bpf_prog *prog); 665 666 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 667 int bpf_obj_get_user(const char __user *pathname, int flags); 668 669 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 670 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 671 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 672 u64 flags); 673 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 674 u64 flags); 675 676 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 677 678 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 679 void *key, void *value, u64 map_flags); 680 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 681 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 682 void *key, void *value, u64 map_flags); 683 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 684 685 int bpf_get_file_flag(int flags); 686 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size, 687 size_t actual_size); 688 689 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 690 * forced to use 'long' read/writes to try to atomically copy long counters. 691 * Best-effort only. No barriers here, since it _will_ race with concurrent 692 * updates from BPF programs. Called from bpf syscall and mostly used with 693 * size 8 or 16 bytes, so ask compiler to inline it. 694 */ 695 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 696 { 697 const long *lsrc = src; 698 long *ldst = dst; 699 700 size /= sizeof(long); 701 while (size--) 702 *ldst++ = *lsrc++; 703 } 704 705 /* verify correctness of eBPF program */ 706 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, 707 union bpf_attr __user *uattr); 708 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 709 710 /* Map specifics */ 711 struct xdp_buff; 712 struct sk_buff; 713 714 struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key); 715 void __dev_map_insert_ctx(struct bpf_map *map, u32 index); 716 void __dev_map_flush(struct bpf_map *map); 717 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 718 struct net_device *dev_rx); 719 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 720 struct bpf_prog *xdp_prog); 721 722 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key); 723 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index); 724 void __cpu_map_flush(struct bpf_map *map); 725 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, 726 struct net_device *dev_rx); 727 728 /* Return map's numa specified by userspace */ 729 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 730 { 731 return (attr->map_flags & BPF_F_NUMA_NODE) ? 732 attr->numa_node : NUMA_NO_NODE; 733 } 734 735 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 736 int array_map_alloc_check(union bpf_attr *attr); 737 738 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 739 union bpf_attr __user *uattr); 740 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 741 union bpf_attr __user *uattr); 742 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 743 const union bpf_attr *kattr, 744 union bpf_attr __user *uattr); 745 #else /* !CONFIG_BPF_SYSCALL */ 746 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 747 { 748 return ERR_PTR(-EOPNOTSUPP); 749 } 750 751 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 752 enum bpf_prog_type type, 753 bool attach_drv) 754 { 755 return ERR_PTR(-EOPNOTSUPP); 756 } 757 758 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, 759 int i) 760 { 761 return ERR_PTR(-EOPNOTSUPP); 762 } 763 764 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 765 { 766 } 767 768 static inline void bpf_prog_put(struct bpf_prog *prog) 769 { 770 } 771 772 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog) 773 { 774 return ERR_PTR(-EOPNOTSUPP); 775 } 776 777 static inline struct bpf_prog *__must_check 778 bpf_prog_inc_not_zero(struct bpf_prog *prog) 779 { 780 return ERR_PTR(-EOPNOTSUPP); 781 } 782 783 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages) 784 { 785 return 0; 786 } 787 788 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages) 789 { 790 } 791 792 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 793 { 794 return -EOPNOTSUPP; 795 } 796 797 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map, 798 u32 key) 799 { 800 return NULL; 801 } 802 803 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index) 804 { 805 } 806 807 static inline void __dev_map_flush(struct bpf_map *map) 808 { 809 } 810 811 struct xdp_buff; 812 struct bpf_dtab_netdev; 813 814 static inline 815 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 816 struct net_device *dev_rx) 817 { 818 return 0; 819 } 820 821 struct sk_buff; 822 823 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 824 struct sk_buff *skb, 825 struct bpf_prog *xdp_prog) 826 { 827 return 0; 828 } 829 830 static inline 831 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key) 832 { 833 return NULL; 834 } 835 836 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index) 837 { 838 } 839 840 static inline void __cpu_map_flush(struct bpf_map *map) 841 { 842 } 843 844 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 845 struct xdp_buff *xdp, 846 struct net_device *dev_rx) 847 { 848 return 0; 849 } 850 851 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 852 enum bpf_prog_type type) 853 { 854 return ERR_PTR(-EOPNOTSUPP); 855 } 856 857 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 858 const union bpf_attr *kattr, 859 union bpf_attr __user *uattr) 860 { 861 return -ENOTSUPP; 862 } 863 864 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 865 const union bpf_attr *kattr, 866 union bpf_attr __user *uattr) 867 { 868 return -ENOTSUPP; 869 } 870 871 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 872 const union bpf_attr *kattr, 873 union bpf_attr __user *uattr) 874 { 875 return -ENOTSUPP; 876 } 877 #endif /* CONFIG_BPF_SYSCALL */ 878 879 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 880 enum bpf_prog_type type) 881 { 882 return bpf_prog_get_type_dev(ufd, type, false); 883 } 884 885 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 886 887 int bpf_prog_offload_compile(struct bpf_prog *prog); 888 void bpf_prog_offload_destroy(struct bpf_prog *prog); 889 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 890 struct bpf_prog *prog); 891 892 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 893 894 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 895 int bpf_map_offload_update_elem(struct bpf_map *map, 896 void *key, void *value, u64 flags); 897 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 898 int bpf_map_offload_get_next_key(struct bpf_map *map, 899 void *key, void *next_key); 900 901 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 902 903 struct bpf_offload_dev * 904 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 905 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 906 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 907 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 908 struct net_device *netdev); 909 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 910 struct net_device *netdev); 911 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 912 913 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 914 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 915 916 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 917 { 918 return aux->offload_requested; 919 } 920 921 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 922 { 923 return unlikely(map->ops == &bpf_map_offload_ops); 924 } 925 926 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 927 void bpf_map_offload_map_free(struct bpf_map *map); 928 #else 929 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 930 union bpf_attr *attr) 931 { 932 return -EOPNOTSUPP; 933 } 934 935 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 936 { 937 return false; 938 } 939 940 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 941 { 942 return false; 943 } 944 945 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 946 { 947 return ERR_PTR(-EOPNOTSUPP); 948 } 949 950 static inline void bpf_map_offload_map_free(struct bpf_map *map) 951 { 952 } 953 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 954 955 #if defined(CONFIG_BPF_STREAM_PARSER) 956 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, u32 which); 957 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 958 #else 959 static inline int sock_map_prog_update(struct bpf_map *map, 960 struct bpf_prog *prog, u32 which) 961 { 962 return -EOPNOTSUPP; 963 } 964 965 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 966 struct bpf_prog *prog) 967 { 968 return -EINVAL; 969 } 970 #endif 971 972 #if defined(CONFIG_XDP_SOCKETS) 973 struct xdp_sock; 974 struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map, u32 key); 975 int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp, 976 struct xdp_sock *xs); 977 void __xsk_map_flush(struct bpf_map *map); 978 #else 979 struct xdp_sock; 980 static inline struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map, 981 u32 key) 982 { 983 return NULL; 984 } 985 986 static inline int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp, 987 struct xdp_sock *xs) 988 { 989 return -EOPNOTSUPP; 990 } 991 992 static inline void __xsk_map_flush(struct bpf_map *map) 993 { 994 } 995 #endif 996 997 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 998 void bpf_sk_reuseport_detach(struct sock *sk); 999 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 1000 void *value); 1001 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 1002 void *value, u64 map_flags); 1003 #else 1004 static inline void bpf_sk_reuseport_detach(struct sock *sk) 1005 { 1006 } 1007 1008 #ifdef CONFIG_BPF_SYSCALL 1009 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 1010 void *key, void *value) 1011 { 1012 return -EOPNOTSUPP; 1013 } 1014 1015 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 1016 void *key, void *value, 1017 u64 map_flags) 1018 { 1019 return -EOPNOTSUPP; 1020 } 1021 #endif /* CONFIG_BPF_SYSCALL */ 1022 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 1023 1024 /* verifier prototypes for helper functions called from eBPF programs */ 1025 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 1026 extern const struct bpf_func_proto bpf_map_update_elem_proto; 1027 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 1028 extern const struct bpf_func_proto bpf_map_push_elem_proto; 1029 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 1030 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 1031 1032 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 1033 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 1034 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 1035 extern const struct bpf_func_proto bpf_tail_call_proto; 1036 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 1037 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 1038 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 1039 extern const struct bpf_func_proto bpf_get_current_comm_proto; 1040 extern const struct bpf_func_proto bpf_get_stackid_proto; 1041 extern const struct bpf_func_proto bpf_get_stack_proto; 1042 extern const struct bpf_func_proto bpf_sock_map_update_proto; 1043 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 1044 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 1045 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 1046 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 1047 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 1048 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 1049 extern const struct bpf_func_proto bpf_spin_lock_proto; 1050 extern const struct bpf_func_proto bpf_spin_unlock_proto; 1051 extern const struct bpf_func_proto bpf_get_local_storage_proto; 1052 extern const struct bpf_func_proto bpf_strtol_proto; 1053 extern const struct bpf_func_proto bpf_strtoul_proto; 1054 1055 /* Shared helpers among cBPF and eBPF. */ 1056 void bpf_user_rnd_init_once(void); 1057 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 1058 1059 #if defined(CONFIG_NET) 1060 bool bpf_sock_common_is_valid_access(int off, int size, 1061 enum bpf_access_type type, 1062 struct bpf_insn_access_aux *info); 1063 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1064 struct bpf_insn_access_aux *info); 1065 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 1066 const struct bpf_insn *si, 1067 struct bpf_insn *insn_buf, 1068 struct bpf_prog *prog, 1069 u32 *target_size); 1070 #else 1071 static inline bool bpf_sock_common_is_valid_access(int off, int size, 1072 enum bpf_access_type type, 1073 struct bpf_insn_access_aux *info) 1074 { 1075 return false; 1076 } 1077 static inline bool bpf_sock_is_valid_access(int off, int size, 1078 enum bpf_access_type type, 1079 struct bpf_insn_access_aux *info) 1080 { 1081 return false; 1082 } 1083 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 1084 const struct bpf_insn *si, 1085 struct bpf_insn *insn_buf, 1086 struct bpf_prog *prog, 1087 u32 *target_size) 1088 { 1089 return 0; 1090 } 1091 #endif 1092 1093 #ifdef CONFIG_INET 1094 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1095 struct bpf_insn_access_aux *info); 1096 1097 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 1098 const struct bpf_insn *si, 1099 struct bpf_insn *insn_buf, 1100 struct bpf_prog *prog, 1101 u32 *target_size); 1102 #else 1103 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 1104 enum bpf_access_type type, 1105 struct bpf_insn_access_aux *info) 1106 { 1107 return false; 1108 } 1109 1110 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 1111 const struct bpf_insn *si, 1112 struct bpf_insn *insn_buf, 1113 struct bpf_prog *prog, 1114 u32 *target_size) 1115 { 1116 return 0; 1117 } 1118 #endif /* CONFIG_INET */ 1119 1120 #endif /* _LINUX_BPF_H */ 1121