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 extern struct idr btf_idr; 28 extern spinlock_t btf_idr_lock; 29 30 /* map is generic key/value storage optionally accesible by eBPF programs */ 31 struct bpf_map_ops { 32 /* funcs callable from userspace (via syscall) */ 33 int (*map_alloc_check)(union bpf_attr *attr); 34 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 35 void (*map_release)(struct bpf_map *map, struct file *map_file); 36 void (*map_free)(struct bpf_map *map); 37 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 38 void (*map_release_uref)(struct bpf_map *map); 39 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 40 41 /* funcs callable from userspace and from eBPF programs */ 42 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 43 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 44 int (*map_delete_elem)(struct bpf_map *map, void *key); 45 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 46 int (*map_pop_elem)(struct bpf_map *map, void *value); 47 int (*map_peek_elem)(struct bpf_map *map, void *value); 48 49 /* funcs called by prog_array and perf_event_array map */ 50 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 51 int fd); 52 void (*map_fd_put_ptr)(void *ptr); 53 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 54 u32 (*map_fd_sys_lookup_elem)(void *ptr); 55 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 56 struct seq_file *m); 57 int (*map_check_btf)(const struct bpf_map *map, 58 const struct btf *btf, 59 const struct btf_type *key_type, 60 const struct btf_type *value_type); 61 62 /* Direct value access helpers. */ 63 int (*map_direct_value_addr)(const struct bpf_map *map, 64 u64 *imm, u32 off); 65 int (*map_direct_value_meta)(const struct bpf_map *map, 66 u64 imm, u32 *off); 67 }; 68 69 struct bpf_map_memory { 70 u32 pages; 71 struct user_struct *user; 72 }; 73 74 struct bpf_map { 75 /* The first two cachelines with read-mostly members of which some 76 * are also accessed in fast-path (e.g. ops, max_entries). 77 */ 78 const struct bpf_map_ops *ops ____cacheline_aligned; 79 struct bpf_map *inner_map_meta; 80 #ifdef CONFIG_SECURITY 81 void *security; 82 #endif 83 enum bpf_map_type map_type; 84 u32 key_size; 85 u32 value_size; 86 u32 max_entries; 87 u32 map_flags; 88 int spin_lock_off; /* >=0 valid offset, <0 error */ 89 u32 id; 90 int numa_node; 91 u32 btf_key_type_id; 92 u32 btf_value_type_id; 93 struct btf *btf; 94 struct bpf_map_memory memory; 95 bool unpriv_array; 96 bool frozen; /* write-once */ 97 /* 48 bytes hole */ 98 99 /* The 3rd and 4th cacheline with misc members to avoid false sharing 100 * particularly with refcounting. 101 */ 102 atomic_t refcnt ____cacheline_aligned; 103 atomic_t usercnt; 104 struct work_struct work; 105 char name[BPF_OBJ_NAME_LEN]; 106 }; 107 108 static inline bool map_value_has_spin_lock(const struct bpf_map *map) 109 { 110 return map->spin_lock_off >= 0; 111 } 112 113 static inline void check_and_init_map_lock(struct bpf_map *map, void *dst) 114 { 115 if (likely(!map_value_has_spin_lock(map))) 116 return; 117 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) = 118 (struct bpf_spin_lock){}; 119 } 120 121 /* copy everything but bpf_spin_lock */ 122 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 123 { 124 if (unlikely(map_value_has_spin_lock(map))) { 125 u32 off = map->spin_lock_off; 126 127 memcpy(dst, src, off); 128 memcpy(dst + off + sizeof(struct bpf_spin_lock), 129 src + off + sizeof(struct bpf_spin_lock), 130 map->value_size - off - sizeof(struct bpf_spin_lock)); 131 } else { 132 memcpy(dst, src, map->value_size); 133 } 134 } 135 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 136 bool lock_src); 137 138 struct bpf_offload_dev; 139 struct bpf_offloaded_map; 140 141 struct bpf_map_dev_ops { 142 int (*map_get_next_key)(struct bpf_offloaded_map *map, 143 void *key, void *next_key); 144 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 145 void *key, void *value); 146 int (*map_update_elem)(struct bpf_offloaded_map *map, 147 void *key, void *value, u64 flags); 148 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 149 }; 150 151 struct bpf_offloaded_map { 152 struct bpf_map map; 153 struct net_device *netdev; 154 const struct bpf_map_dev_ops *dev_ops; 155 void *dev_priv; 156 struct list_head offloads; 157 }; 158 159 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 160 { 161 return container_of(map, struct bpf_offloaded_map, map); 162 } 163 164 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 165 { 166 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 167 } 168 169 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 170 { 171 return map->btf && map->ops->map_seq_show_elem; 172 } 173 174 int map_check_no_btf(const struct bpf_map *map, 175 const struct btf *btf, 176 const struct btf_type *key_type, 177 const struct btf_type *value_type); 178 179 extern const struct bpf_map_ops bpf_map_offload_ops; 180 181 /* function argument constraints */ 182 enum bpf_arg_type { 183 ARG_DONTCARE = 0, /* unused argument in helper function */ 184 185 /* the following constraints used to prototype 186 * bpf_map_lookup/update/delete_elem() functions 187 */ 188 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 189 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 190 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 191 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */ 192 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */ 193 194 /* the following constraints used to prototype bpf_memcmp() and other 195 * functions that access data on eBPF program stack 196 */ 197 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 198 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */ 199 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized, 200 * helper function must fill all bytes or clear 201 * them in error case. 202 */ 203 204 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 205 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 206 207 ARG_PTR_TO_CTX, /* pointer to context */ 208 ARG_ANYTHING, /* any (initialized) argument is ok */ 209 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 210 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 211 ARG_PTR_TO_INT, /* pointer to int */ 212 ARG_PTR_TO_LONG, /* pointer to long */ 213 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 214 }; 215 216 /* type of values returned from helper functions */ 217 enum bpf_return_type { 218 RET_INTEGER, /* function returns integer */ 219 RET_VOID, /* function doesn't return anything */ 220 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 221 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */ 222 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */ 223 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */ 224 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */ 225 }; 226 227 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 228 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 229 * instructions after verifying 230 */ 231 struct bpf_func_proto { 232 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 233 bool gpl_only; 234 bool pkt_access; 235 enum bpf_return_type ret_type; 236 enum bpf_arg_type arg1_type; 237 enum bpf_arg_type arg2_type; 238 enum bpf_arg_type arg3_type; 239 enum bpf_arg_type arg4_type; 240 enum bpf_arg_type arg5_type; 241 }; 242 243 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 244 * the first argument to eBPF programs. 245 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 246 */ 247 struct bpf_context; 248 249 enum bpf_access_type { 250 BPF_READ = 1, 251 BPF_WRITE = 2 252 }; 253 254 /* types of values stored in eBPF registers */ 255 /* Pointer types represent: 256 * pointer 257 * pointer + imm 258 * pointer + (u16) var 259 * pointer + (u16) var + imm 260 * if (range > 0) then [ptr, ptr + range - off) is safe to access 261 * if (id > 0) means that some 'var' was added 262 * if (off > 0) means that 'imm' was added 263 */ 264 enum bpf_reg_type { 265 NOT_INIT = 0, /* nothing was written into register */ 266 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 267 PTR_TO_CTX, /* reg points to bpf_context */ 268 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 269 PTR_TO_MAP_VALUE, /* reg points to map element value */ 270 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */ 271 PTR_TO_STACK, /* reg == frame_pointer + offset */ 272 PTR_TO_PACKET_META, /* skb->data - meta_len */ 273 PTR_TO_PACKET, /* reg points to skb->data */ 274 PTR_TO_PACKET_END, /* skb->data + headlen */ 275 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 276 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 277 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */ 278 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 279 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */ 280 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 281 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */ 282 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 283 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 284 }; 285 286 /* The information passed from prog-specific *_is_valid_access 287 * back to the verifier. 288 */ 289 struct bpf_insn_access_aux { 290 enum bpf_reg_type reg_type; 291 int ctx_field_size; 292 }; 293 294 static inline void 295 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 296 { 297 aux->ctx_field_size = size; 298 } 299 300 struct bpf_prog_ops { 301 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 302 union bpf_attr __user *uattr); 303 }; 304 305 struct bpf_verifier_ops { 306 /* return eBPF function prototype for verification */ 307 const struct bpf_func_proto * 308 (*get_func_proto)(enum bpf_func_id func_id, 309 const struct bpf_prog *prog); 310 311 /* return true if 'size' wide access at offset 'off' within bpf_context 312 * with 'type' (read or write) is allowed 313 */ 314 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 315 const struct bpf_prog *prog, 316 struct bpf_insn_access_aux *info); 317 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 318 const struct bpf_prog *prog); 319 int (*gen_ld_abs)(const struct bpf_insn *orig, 320 struct bpf_insn *insn_buf); 321 u32 (*convert_ctx_access)(enum bpf_access_type type, 322 const struct bpf_insn *src, 323 struct bpf_insn *dst, 324 struct bpf_prog *prog, u32 *target_size); 325 }; 326 327 struct bpf_prog_offload_ops { 328 /* verifier basic callbacks */ 329 int (*insn_hook)(struct bpf_verifier_env *env, 330 int insn_idx, int prev_insn_idx); 331 int (*finalize)(struct bpf_verifier_env *env); 332 /* verifier optimization callbacks (called after .finalize) */ 333 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 334 struct bpf_insn *insn); 335 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 336 /* program management callbacks */ 337 int (*prepare)(struct bpf_prog *prog); 338 int (*translate)(struct bpf_prog *prog); 339 void (*destroy)(struct bpf_prog *prog); 340 }; 341 342 struct bpf_prog_offload { 343 struct bpf_prog *prog; 344 struct net_device *netdev; 345 struct bpf_offload_dev *offdev; 346 void *dev_priv; 347 struct list_head offloads; 348 bool dev_state; 349 bool opt_failed; 350 void *jited_image; 351 u32 jited_len; 352 }; 353 354 enum bpf_cgroup_storage_type { 355 BPF_CGROUP_STORAGE_SHARED, 356 BPF_CGROUP_STORAGE_PERCPU, 357 __BPF_CGROUP_STORAGE_MAX 358 }; 359 360 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 361 362 struct bpf_prog_stats { 363 u64 cnt; 364 u64 nsecs; 365 struct u64_stats_sync syncp; 366 }; 367 368 struct bpf_prog_aux { 369 atomic_t refcnt; 370 u32 used_map_cnt; 371 u32 max_ctx_offset; 372 u32 max_pkt_offset; 373 u32 max_tp_access; 374 u32 stack_depth; 375 u32 id; 376 u32 func_cnt; /* used by non-func prog as the number of func progs */ 377 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 378 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 379 bool offload_requested; 380 struct bpf_prog **func; 381 void *jit_data; /* JIT specific data. arch dependent */ 382 struct latch_tree_node ksym_tnode; 383 struct list_head ksym_lnode; 384 const struct bpf_prog_ops *ops; 385 struct bpf_map **used_maps; 386 struct bpf_prog *prog; 387 struct user_struct *user; 388 u64 load_time; /* ns since boottime */ 389 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 390 char name[BPF_OBJ_NAME_LEN]; 391 #ifdef CONFIG_SECURITY 392 void *security; 393 #endif 394 struct bpf_prog_offload *offload; 395 struct btf *btf; 396 struct bpf_func_info *func_info; 397 /* bpf_line_info loaded from userspace. linfo->insn_off 398 * has the xlated insn offset. 399 * Both the main and sub prog share the same linfo. 400 * The subprog can access its first linfo by 401 * using the linfo_idx. 402 */ 403 struct bpf_line_info *linfo; 404 /* jited_linfo is the jited addr of the linfo. It has a 405 * one to one mapping to linfo: 406 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 407 * Both the main and sub prog share the same jited_linfo. 408 * The subprog can access its first jited_linfo by 409 * using the linfo_idx. 410 */ 411 void **jited_linfo; 412 u32 func_info_cnt; 413 u32 nr_linfo; 414 /* subprog can use linfo_idx to access its first linfo and 415 * jited_linfo. 416 * main prog always has linfo_idx == 0 417 */ 418 u32 linfo_idx; 419 struct bpf_prog_stats __percpu *stats; 420 union { 421 struct work_struct work; 422 struct rcu_head rcu; 423 }; 424 }; 425 426 struct bpf_array { 427 struct bpf_map map; 428 u32 elem_size; 429 u32 index_mask; 430 /* 'ownership' of prog_array is claimed by the first program that 431 * is going to use this map or by the first program which FD is stored 432 * in the map to make sure that all callers and callees have the same 433 * prog_type and JITed flag 434 */ 435 enum bpf_prog_type owner_prog_type; 436 bool owner_jited; 437 union { 438 char value[0] __aligned(8); 439 void *ptrs[0] __aligned(8); 440 void __percpu *pptrs[0] __aligned(8); 441 }; 442 }; 443 444 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 445 #define MAX_TAIL_CALL_CNT 32 446 447 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 448 BPF_F_RDONLY_PROG | \ 449 BPF_F_WRONLY | \ 450 BPF_F_WRONLY_PROG) 451 452 #define BPF_MAP_CAN_READ BIT(0) 453 #define BPF_MAP_CAN_WRITE BIT(1) 454 455 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 456 { 457 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 458 459 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 460 * not possible. 461 */ 462 if (access_flags & BPF_F_RDONLY_PROG) 463 return BPF_MAP_CAN_READ; 464 else if (access_flags & BPF_F_WRONLY_PROG) 465 return BPF_MAP_CAN_WRITE; 466 else 467 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 468 } 469 470 static inline bool bpf_map_flags_access_ok(u32 access_flags) 471 { 472 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 473 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 474 } 475 476 struct bpf_event_entry { 477 struct perf_event *event; 478 struct file *perf_file; 479 struct file *map_file; 480 struct rcu_head rcu; 481 }; 482 483 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); 484 int bpf_prog_calc_tag(struct bpf_prog *fp); 485 486 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 487 488 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 489 unsigned long off, unsigned long len); 490 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 491 const struct bpf_insn *src, 492 struct bpf_insn *dst, 493 struct bpf_prog *prog, 494 u32 *target_size); 495 496 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 497 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 498 499 /* an array of programs to be executed under rcu_lock. 500 * 501 * Typical usage: 502 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN); 503 * 504 * the structure returned by bpf_prog_array_alloc() should be populated 505 * with program pointers and the last pointer must be NULL. 506 * The user has to keep refcnt on the program and make sure the program 507 * is removed from the array before bpf_prog_put(). 508 * The 'struct bpf_prog_array *' should only be replaced with xchg() 509 * since other cpus are walking the array of pointers in parallel. 510 */ 511 struct bpf_prog_array_item { 512 struct bpf_prog *prog; 513 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 514 }; 515 516 struct bpf_prog_array { 517 struct rcu_head rcu; 518 struct bpf_prog_array_item items[0]; 519 }; 520 521 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 522 void bpf_prog_array_free(struct bpf_prog_array *progs); 523 int bpf_prog_array_length(struct bpf_prog_array *progs); 524 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 525 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 526 __u32 __user *prog_ids, u32 cnt); 527 528 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 529 struct bpf_prog *old_prog); 530 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 531 u32 *prog_ids, u32 request_cnt, 532 u32 *prog_cnt); 533 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 534 struct bpf_prog *exclude_prog, 535 struct bpf_prog *include_prog, 536 struct bpf_prog_array **new_array); 537 538 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \ 539 ({ \ 540 struct bpf_prog_array_item *_item; \ 541 struct bpf_prog *_prog; \ 542 struct bpf_prog_array *_array; \ 543 u32 _ret = 1; \ 544 preempt_disable(); \ 545 rcu_read_lock(); \ 546 _array = rcu_dereference(array); \ 547 if (unlikely(check_non_null && !_array))\ 548 goto _out; \ 549 _item = &_array->items[0]; \ 550 while ((_prog = READ_ONCE(_item->prog))) { \ 551 bpf_cgroup_storage_set(_item->cgroup_storage); \ 552 _ret &= func(_prog, ctx); \ 553 _item++; \ 554 } \ 555 _out: \ 556 rcu_read_unlock(); \ 557 preempt_enable(); \ 558 _ret; \ 559 }) 560 561 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs 562 * so BPF programs can request cwr for TCP packets. 563 * 564 * Current cgroup skb programs can only return 0 or 1 (0 to drop the 565 * packet. This macro changes the behavior so the low order bit 566 * indicates whether the packet should be dropped (0) or not (1) 567 * and the next bit is a congestion notification bit. This could be 568 * used by TCP to call tcp_enter_cwr() 569 * 570 * Hence, new allowed return values of CGROUP EGRESS BPF programs are: 571 * 0: drop packet 572 * 1: keep packet 573 * 2: drop packet and cn 574 * 3: keep packet and cn 575 * 576 * This macro then converts it to one of the NET_XMIT or an error 577 * code that is then interpreted as drop packet (and no cn): 578 * 0: NET_XMIT_SUCCESS skb should be transmitted 579 * 1: NET_XMIT_DROP skb should be dropped and cn 580 * 2: NET_XMIT_CN skb should be transmitted and cn 581 * 3: -EPERM skb should be dropped 582 */ 583 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \ 584 ({ \ 585 struct bpf_prog_array_item *_item; \ 586 struct bpf_prog *_prog; \ 587 struct bpf_prog_array *_array; \ 588 u32 ret; \ 589 u32 _ret = 1; \ 590 u32 _cn = 0; \ 591 preempt_disable(); \ 592 rcu_read_lock(); \ 593 _array = rcu_dereference(array); \ 594 _item = &_array->items[0]; \ 595 while ((_prog = READ_ONCE(_item->prog))) { \ 596 bpf_cgroup_storage_set(_item->cgroup_storage); \ 597 ret = func(_prog, ctx); \ 598 _ret &= (ret & 1); \ 599 _cn |= (ret & 2); \ 600 _item++; \ 601 } \ 602 rcu_read_unlock(); \ 603 preempt_enable(); \ 604 if (_ret) \ 605 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \ 606 else \ 607 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \ 608 _ret; \ 609 }) 610 611 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \ 612 __BPF_PROG_RUN_ARRAY(array, ctx, func, false) 613 614 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \ 615 __BPF_PROG_RUN_ARRAY(array, ctx, func, true) 616 617 #ifdef CONFIG_BPF_SYSCALL 618 DECLARE_PER_CPU(int, bpf_prog_active); 619 620 extern const struct file_operations bpf_map_fops; 621 extern const struct file_operations bpf_prog_fops; 622 623 #define BPF_PROG_TYPE(_id, _name) \ 624 extern const struct bpf_prog_ops _name ## _prog_ops; \ 625 extern const struct bpf_verifier_ops _name ## _verifier_ops; 626 #define BPF_MAP_TYPE(_id, _ops) \ 627 extern const struct bpf_map_ops _ops; 628 #include <linux/bpf_types.h> 629 #undef BPF_PROG_TYPE 630 #undef BPF_MAP_TYPE 631 632 extern const struct bpf_prog_ops bpf_offload_prog_ops; 633 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 634 extern const struct bpf_verifier_ops xdp_analyzer_ops; 635 636 struct bpf_prog *bpf_prog_get(u32 ufd); 637 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 638 bool attach_drv); 639 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i); 640 void bpf_prog_sub(struct bpf_prog *prog, int i); 641 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog); 642 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 643 void bpf_prog_put(struct bpf_prog *prog); 644 int __bpf_prog_charge(struct user_struct *user, u32 pages); 645 void __bpf_prog_uncharge(struct user_struct *user, u32 pages); 646 647 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 648 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 649 650 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 651 struct bpf_map *__bpf_map_get(struct fd f); 652 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref); 653 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map, 654 bool uref); 655 void bpf_map_put_with_uref(struct bpf_map *map); 656 void bpf_map_put(struct bpf_map *map); 657 int bpf_map_charge_memlock(struct bpf_map *map, u32 pages); 658 void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages); 659 int bpf_map_charge_init(struct bpf_map_memory *mem, size_t size); 660 void bpf_map_charge_finish(struct bpf_map_memory *mem); 661 void bpf_map_charge_move(struct bpf_map_memory *dst, 662 struct bpf_map_memory *src); 663 void *bpf_map_area_alloc(size_t size, int numa_node); 664 void bpf_map_area_free(void *base); 665 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 666 667 extern int sysctl_unprivileged_bpf_disabled; 668 669 int bpf_map_new_fd(struct bpf_map *map, int flags); 670 int bpf_prog_new_fd(struct bpf_prog *prog); 671 672 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 673 int bpf_obj_get_user(const char __user *pathname, int flags); 674 675 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 676 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 677 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 678 u64 flags); 679 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 680 u64 flags); 681 682 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 683 684 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 685 void *key, void *value, u64 map_flags); 686 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 687 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 688 void *key, void *value, u64 map_flags); 689 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 690 691 int bpf_get_file_flag(int flags); 692 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size, 693 size_t actual_size); 694 695 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 696 * forced to use 'long' read/writes to try to atomically copy long counters. 697 * Best-effort only. No barriers here, since it _will_ race with concurrent 698 * updates from BPF programs. Called from bpf syscall and mostly used with 699 * size 8 or 16 bytes, so ask compiler to inline it. 700 */ 701 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 702 { 703 const long *lsrc = src; 704 long *ldst = dst; 705 706 size /= sizeof(long); 707 while (size--) 708 *ldst++ = *lsrc++; 709 } 710 711 /* verify correctness of eBPF program */ 712 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, 713 union bpf_attr __user *uattr); 714 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 715 716 /* Map specifics */ 717 struct xdp_buff; 718 struct sk_buff; 719 720 struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key); 721 struct bpf_dtab_netdev *__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key); 722 void __dev_map_flush(struct bpf_map *map); 723 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 724 struct net_device *dev_rx); 725 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 726 struct bpf_prog *xdp_prog); 727 728 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key); 729 void __cpu_map_flush(struct bpf_map *map); 730 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, 731 struct net_device *dev_rx); 732 733 /* Return map's numa specified by userspace */ 734 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 735 { 736 return (attr->map_flags & BPF_F_NUMA_NODE) ? 737 attr->numa_node : NUMA_NO_NODE; 738 } 739 740 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 741 int array_map_alloc_check(union bpf_attr *attr); 742 743 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 744 union bpf_attr __user *uattr); 745 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 746 union bpf_attr __user *uattr); 747 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 748 const union bpf_attr *kattr, 749 union bpf_attr __user *uattr); 750 #else /* !CONFIG_BPF_SYSCALL */ 751 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 752 { 753 return ERR_PTR(-EOPNOTSUPP); 754 } 755 756 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 757 enum bpf_prog_type type, 758 bool attach_drv) 759 { 760 return ERR_PTR(-EOPNOTSUPP); 761 } 762 763 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, 764 int i) 765 { 766 return ERR_PTR(-EOPNOTSUPP); 767 } 768 769 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 770 { 771 } 772 773 static inline void bpf_prog_put(struct bpf_prog *prog) 774 { 775 } 776 777 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog) 778 { 779 return ERR_PTR(-EOPNOTSUPP); 780 } 781 782 static inline struct bpf_prog *__must_check 783 bpf_prog_inc_not_zero(struct bpf_prog *prog) 784 { 785 return ERR_PTR(-EOPNOTSUPP); 786 } 787 788 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages) 789 { 790 return 0; 791 } 792 793 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages) 794 { 795 } 796 797 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 798 { 799 return -EOPNOTSUPP; 800 } 801 802 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map, 803 u32 key) 804 { 805 return NULL; 806 } 807 808 static inline struct net_device *__dev_map_hash_lookup_elem(struct bpf_map *map, 809 u32 key) 810 { 811 return NULL; 812 } 813 814 static inline void __dev_map_flush(struct bpf_map *map) 815 { 816 } 817 818 struct xdp_buff; 819 struct bpf_dtab_netdev; 820 821 static inline 822 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 823 struct net_device *dev_rx) 824 { 825 return 0; 826 } 827 828 struct sk_buff; 829 830 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 831 struct sk_buff *skb, 832 struct bpf_prog *xdp_prog) 833 { 834 return 0; 835 } 836 837 static inline 838 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key) 839 { 840 return NULL; 841 } 842 843 static inline void __cpu_map_flush(struct bpf_map *map) 844 { 845 } 846 847 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 848 struct xdp_buff *xdp, 849 struct net_device *dev_rx) 850 { 851 return 0; 852 } 853 854 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 855 enum bpf_prog_type type) 856 { 857 return ERR_PTR(-EOPNOTSUPP); 858 } 859 860 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 861 const union bpf_attr *kattr, 862 union bpf_attr __user *uattr) 863 { 864 return -ENOTSUPP; 865 } 866 867 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 868 const union bpf_attr *kattr, 869 union bpf_attr __user *uattr) 870 { 871 return -ENOTSUPP; 872 } 873 874 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 875 const union bpf_attr *kattr, 876 union bpf_attr __user *uattr) 877 { 878 return -ENOTSUPP; 879 } 880 #endif /* CONFIG_BPF_SYSCALL */ 881 882 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 883 enum bpf_prog_type type) 884 { 885 return bpf_prog_get_type_dev(ufd, type, false); 886 } 887 888 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 889 890 int bpf_prog_offload_compile(struct bpf_prog *prog); 891 void bpf_prog_offload_destroy(struct bpf_prog *prog); 892 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 893 struct bpf_prog *prog); 894 895 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 896 897 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 898 int bpf_map_offload_update_elem(struct bpf_map *map, 899 void *key, void *value, u64 flags); 900 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 901 int bpf_map_offload_get_next_key(struct bpf_map *map, 902 void *key, void *next_key); 903 904 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 905 906 struct bpf_offload_dev * 907 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 908 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 909 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 910 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 911 struct net_device *netdev); 912 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 913 struct net_device *netdev); 914 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 915 916 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 917 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 918 919 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 920 { 921 return aux->offload_requested; 922 } 923 924 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 925 { 926 return unlikely(map->ops == &bpf_map_offload_ops); 927 } 928 929 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 930 void bpf_map_offload_map_free(struct bpf_map *map); 931 #else 932 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 933 union bpf_attr *attr) 934 { 935 return -EOPNOTSUPP; 936 } 937 938 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 939 { 940 return false; 941 } 942 943 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 944 { 945 return false; 946 } 947 948 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 949 { 950 return ERR_PTR(-EOPNOTSUPP); 951 } 952 953 static inline void bpf_map_offload_map_free(struct bpf_map *map) 954 { 955 } 956 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 957 958 #if defined(CONFIG_BPF_STREAM_PARSER) 959 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, u32 which); 960 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 961 #else 962 static inline int sock_map_prog_update(struct bpf_map *map, 963 struct bpf_prog *prog, u32 which) 964 { 965 return -EOPNOTSUPP; 966 } 967 968 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 969 struct bpf_prog *prog) 970 { 971 return -EINVAL; 972 } 973 #endif 974 975 #if defined(CONFIG_XDP_SOCKETS) 976 struct xdp_sock; 977 struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map, u32 key); 978 int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp, 979 struct xdp_sock *xs); 980 void __xsk_map_flush(struct bpf_map *map); 981 #else 982 struct xdp_sock; 983 static inline struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map, 984 u32 key) 985 { 986 return NULL; 987 } 988 989 static inline int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp, 990 struct xdp_sock *xs) 991 { 992 return -EOPNOTSUPP; 993 } 994 995 static inline void __xsk_map_flush(struct bpf_map *map) 996 { 997 } 998 #endif 999 1000 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 1001 void bpf_sk_reuseport_detach(struct sock *sk); 1002 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 1003 void *value); 1004 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 1005 void *value, u64 map_flags); 1006 #else 1007 static inline void bpf_sk_reuseport_detach(struct sock *sk) 1008 { 1009 } 1010 1011 #ifdef CONFIG_BPF_SYSCALL 1012 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 1013 void *key, void *value) 1014 { 1015 return -EOPNOTSUPP; 1016 } 1017 1018 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 1019 void *key, void *value, 1020 u64 map_flags) 1021 { 1022 return -EOPNOTSUPP; 1023 } 1024 #endif /* CONFIG_BPF_SYSCALL */ 1025 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 1026 1027 /* verifier prototypes for helper functions called from eBPF programs */ 1028 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 1029 extern const struct bpf_func_proto bpf_map_update_elem_proto; 1030 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 1031 extern const struct bpf_func_proto bpf_map_push_elem_proto; 1032 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 1033 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 1034 1035 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 1036 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 1037 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 1038 extern const struct bpf_func_proto bpf_tail_call_proto; 1039 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 1040 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 1041 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 1042 extern const struct bpf_func_proto bpf_get_current_comm_proto; 1043 extern const struct bpf_func_proto bpf_get_stackid_proto; 1044 extern const struct bpf_func_proto bpf_get_stack_proto; 1045 extern const struct bpf_func_proto bpf_sock_map_update_proto; 1046 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 1047 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 1048 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 1049 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 1050 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 1051 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 1052 extern const struct bpf_func_proto bpf_spin_lock_proto; 1053 extern const struct bpf_func_proto bpf_spin_unlock_proto; 1054 extern const struct bpf_func_proto bpf_get_local_storage_proto; 1055 extern const struct bpf_func_proto bpf_strtol_proto; 1056 extern const struct bpf_func_proto bpf_strtoul_proto; 1057 extern const struct bpf_func_proto bpf_tcp_sock_proto; 1058 1059 /* Shared helpers among cBPF and eBPF. */ 1060 void bpf_user_rnd_init_once(void); 1061 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 1062 1063 #if defined(CONFIG_NET) 1064 bool bpf_sock_common_is_valid_access(int off, int size, 1065 enum bpf_access_type type, 1066 struct bpf_insn_access_aux *info); 1067 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1068 struct bpf_insn_access_aux *info); 1069 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 1070 const struct bpf_insn *si, 1071 struct bpf_insn *insn_buf, 1072 struct bpf_prog *prog, 1073 u32 *target_size); 1074 #else 1075 static inline bool bpf_sock_common_is_valid_access(int off, int size, 1076 enum bpf_access_type type, 1077 struct bpf_insn_access_aux *info) 1078 { 1079 return false; 1080 } 1081 static inline bool bpf_sock_is_valid_access(int off, int size, 1082 enum bpf_access_type type, 1083 struct bpf_insn_access_aux *info) 1084 { 1085 return false; 1086 } 1087 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 1088 const struct bpf_insn *si, 1089 struct bpf_insn *insn_buf, 1090 struct bpf_prog *prog, 1091 u32 *target_size) 1092 { 1093 return 0; 1094 } 1095 #endif 1096 1097 #ifdef CONFIG_INET 1098 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1099 struct bpf_insn_access_aux *info); 1100 1101 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 1102 const struct bpf_insn *si, 1103 struct bpf_insn *insn_buf, 1104 struct bpf_prog *prog, 1105 u32 *target_size); 1106 1107 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1108 struct bpf_insn_access_aux *info); 1109 1110 u32 bpf_xdp_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 #else 1116 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 1117 enum bpf_access_type type, 1118 struct bpf_insn_access_aux *info) 1119 { 1120 return false; 1121 } 1122 1123 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 1124 const struct bpf_insn *si, 1125 struct bpf_insn *insn_buf, 1126 struct bpf_prog *prog, 1127 u32 *target_size) 1128 { 1129 return 0; 1130 } 1131 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 1132 enum bpf_access_type type, 1133 struct bpf_insn_access_aux *info) 1134 { 1135 return false; 1136 } 1137 1138 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 1139 const struct bpf_insn *si, 1140 struct bpf_insn *insn_buf, 1141 struct bpf_prog *prog, 1142 u32 *target_size) 1143 { 1144 return 0; 1145 } 1146 #endif /* CONFIG_INET */ 1147 1148 #endif /* _LINUX_BPF_H */ 1149