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/mm_types.h> 16 #include <linux/wait.h> 17 #include <linux/refcount.h> 18 #include <linux/mutex.h> 19 #include <linux/module.h> 20 #include <linux/kallsyms.h> 21 #include <linux/capability.h> 22 #include <linux/sched/mm.h> 23 #include <linux/slab.h> 24 #include <linux/percpu-refcount.h> 25 #include <linux/bpfptr.h> 26 27 struct bpf_verifier_env; 28 struct bpf_verifier_log; 29 struct perf_event; 30 struct bpf_prog; 31 struct bpf_prog_aux; 32 struct bpf_map; 33 struct sock; 34 struct seq_file; 35 struct btf; 36 struct btf_type; 37 struct exception_table_entry; 38 struct seq_operations; 39 struct bpf_iter_aux_info; 40 struct bpf_local_storage; 41 struct bpf_local_storage_map; 42 struct kobject; 43 struct mem_cgroup; 44 struct module; 45 struct bpf_func_state; 46 47 extern struct idr btf_idr; 48 extern spinlock_t btf_idr_lock; 49 extern struct kobject *btf_kobj; 50 51 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64); 52 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, 53 struct bpf_iter_aux_info *aux); 54 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); 55 struct bpf_iter_seq_info { 56 const struct seq_operations *seq_ops; 57 bpf_iter_init_seq_priv_t init_seq_private; 58 bpf_iter_fini_seq_priv_t fini_seq_private; 59 u32 seq_priv_size; 60 }; 61 62 /* map is generic key/value storage optionally accessible by eBPF programs */ 63 struct bpf_map_ops { 64 /* funcs callable from userspace (via syscall) */ 65 int (*map_alloc_check)(union bpf_attr *attr); 66 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 67 void (*map_release)(struct bpf_map *map, struct file *map_file); 68 void (*map_free)(struct bpf_map *map); 69 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 70 void (*map_release_uref)(struct bpf_map *map); 71 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 72 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, 73 union bpf_attr __user *uattr); 74 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key, 75 void *value, u64 flags); 76 int (*map_lookup_and_delete_batch)(struct bpf_map *map, 77 const union bpf_attr *attr, 78 union bpf_attr __user *uattr); 79 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr, 80 union bpf_attr __user *uattr); 81 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, 82 union bpf_attr __user *uattr); 83 84 /* funcs callable from userspace and from eBPF programs */ 85 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 86 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 87 int (*map_delete_elem)(struct bpf_map *map, void *key); 88 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 89 int (*map_pop_elem)(struct bpf_map *map, void *value); 90 int (*map_peek_elem)(struct bpf_map *map, void *value); 91 92 /* funcs called by prog_array and perf_event_array map */ 93 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 94 int fd); 95 void (*map_fd_put_ptr)(void *ptr); 96 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 97 u32 (*map_fd_sys_lookup_elem)(void *ptr); 98 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 99 struct seq_file *m); 100 int (*map_check_btf)(const struct bpf_map *map, 101 const struct btf *btf, 102 const struct btf_type *key_type, 103 const struct btf_type *value_type); 104 105 /* Prog poke tracking helpers. */ 106 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux); 107 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux); 108 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, 109 struct bpf_prog *new); 110 111 /* Direct value access helpers. */ 112 int (*map_direct_value_addr)(const struct bpf_map *map, 113 u64 *imm, u32 off); 114 int (*map_direct_value_meta)(const struct bpf_map *map, 115 u64 imm, u32 *off); 116 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma); 117 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, 118 struct poll_table_struct *pts); 119 120 /* Functions called by bpf_local_storage maps */ 121 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, 122 void *owner, u32 size); 123 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, 124 void *owner, u32 size); 125 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner); 126 127 /* Misc helpers.*/ 128 int (*map_redirect)(struct bpf_map *map, u32 ifindex, u64 flags); 129 130 /* map_meta_equal must be implemented for maps that can be 131 * used as an inner map. It is a runtime check to ensure 132 * an inner map can be inserted to an outer map. 133 * 134 * Some properties of the inner map has been used during the 135 * verification time. When inserting an inner map at the runtime, 136 * map_meta_equal has to ensure the inserting map has the same 137 * properties that the verifier has used earlier. 138 */ 139 bool (*map_meta_equal)(const struct bpf_map *meta0, 140 const struct bpf_map *meta1); 141 142 143 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env, 144 struct bpf_func_state *caller, 145 struct bpf_func_state *callee); 146 int (*map_for_each_callback)(struct bpf_map *map, 147 bpf_callback_t callback_fn, 148 void *callback_ctx, u64 flags); 149 150 /* BTF name and id of struct allocated by map_alloc */ 151 const char * const map_btf_name; 152 int *map_btf_id; 153 154 /* bpf_iter info used to open a seq_file */ 155 const struct bpf_iter_seq_info *iter_seq_info; 156 }; 157 158 struct bpf_map { 159 /* The first two cachelines with read-mostly members of which some 160 * are also accessed in fast-path (e.g. ops, max_entries). 161 */ 162 const struct bpf_map_ops *ops ____cacheline_aligned; 163 struct bpf_map *inner_map_meta; 164 #ifdef CONFIG_SECURITY 165 void *security; 166 #endif 167 enum bpf_map_type map_type; 168 u32 key_size; 169 u32 value_size; 170 u32 max_entries; 171 u32 map_flags; 172 int spin_lock_off; /* >=0 valid offset, <0 error */ 173 int timer_off; /* >=0 valid offset, <0 error */ 174 u32 id; 175 int numa_node; 176 u32 btf_key_type_id; 177 u32 btf_value_type_id; 178 struct btf *btf; 179 #ifdef CONFIG_MEMCG_KMEM 180 struct mem_cgroup *memcg; 181 #endif 182 char name[BPF_OBJ_NAME_LEN]; 183 u32 btf_vmlinux_value_type_id; 184 bool bypass_spec_v1; 185 bool frozen; /* write-once; write-protected by freeze_mutex */ 186 /* 22 bytes hole */ 187 188 /* The 3rd and 4th cacheline with misc members to avoid false sharing 189 * particularly with refcounting. 190 */ 191 atomic64_t refcnt ____cacheline_aligned; 192 atomic64_t usercnt; 193 struct work_struct work; 194 struct mutex freeze_mutex; 195 u64 writecnt; /* writable mmap cnt; protected by freeze_mutex */ 196 }; 197 198 static inline bool map_value_has_spin_lock(const struct bpf_map *map) 199 { 200 return map->spin_lock_off >= 0; 201 } 202 203 static inline bool map_value_has_timer(const struct bpf_map *map) 204 { 205 return map->timer_off >= 0; 206 } 207 208 static inline void check_and_init_map_value(struct bpf_map *map, void *dst) 209 { 210 if (unlikely(map_value_has_spin_lock(map))) 211 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) = 212 (struct bpf_spin_lock){}; 213 if (unlikely(map_value_has_timer(map))) 214 *(struct bpf_timer *)(dst + map->timer_off) = 215 (struct bpf_timer){}; 216 } 217 218 /* copy everything but bpf_spin_lock and bpf_timer. There could be one of each. */ 219 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 220 { 221 u32 s_off = 0, s_sz = 0, t_off = 0, t_sz = 0; 222 223 if (unlikely(map_value_has_spin_lock(map))) { 224 s_off = map->spin_lock_off; 225 s_sz = sizeof(struct bpf_spin_lock); 226 } else if (unlikely(map_value_has_timer(map))) { 227 t_off = map->timer_off; 228 t_sz = sizeof(struct bpf_timer); 229 } 230 231 if (unlikely(s_sz || t_sz)) { 232 if (s_off < t_off || !s_sz) { 233 swap(s_off, t_off); 234 swap(s_sz, t_sz); 235 } 236 memcpy(dst, src, t_off); 237 memcpy(dst + t_off + t_sz, 238 src + t_off + t_sz, 239 s_off - t_off - t_sz); 240 memcpy(dst + s_off + s_sz, 241 src + s_off + s_sz, 242 map->value_size - s_off - s_sz); 243 } else { 244 memcpy(dst, src, map->value_size); 245 } 246 } 247 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 248 bool lock_src); 249 void bpf_timer_cancel_and_free(void *timer); 250 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size); 251 252 struct bpf_offload_dev; 253 struct bpf_offloaded_map; 254 255 struct bpf_map_dev_ops { 256 int (*map_get_next_key)(struct bpf_offloaded_map *map, 257 void *key, void *next_key); 258 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 259 void *key, void *value); 260 int (*map_update_elem)(struct bpf_offloaded_map *map, 261 void *key, void *value, u64 flags); 262 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 263 }; 264 265 struct bpf_offloaded_map { 266 struct bpf_map map; 267 struct net_device *netdev; 268 const struct bpf_map_dev_ops *dev_ops; 269 void *dev_priv; 270 struct list_head offloads; 271 }; 272 273 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 274 { 275 return container_of(map, struct bpf_offloaded_map, map); 276 } 277 278 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 279 { 280 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 281 } 282 283 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 284 { 285 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && 286 map->ops->map_seq_show_elem; 287 } 288 289 int map_check_no_btf(const struct bpf_map *map, 290 const struct btf *btf, 291 const struct btf_type *key_type, 292 const struct btf_type *value_type); 293 294 bool bpf_map_meta_equal(const struct bpf_map *meta0, 295 const struct bpf_map *meta1); 296 297 extern const struct bpf_map_ops bpf_map_offload_ops; 298 299 /* function argument constraints */ 300 enum bpf_arg_type { 301 ARG_DONTCARE = 0, /* unused argument in helper function */ 302 303 /* the following constraints used to prototype 304 * bpf_map_lookup/update/delete_elem() functions 305 */ 306 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 307 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 308 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 309 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */ 310 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */ 311 312 /* the following constraints used to prototype bpf_memcmp() and other 313 * functions that access data on eBPF program stack 314 */ 315 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 316 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */ 317 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized, 318 * helper function must fill all bytes or clear 319 * them in error case. 320 */ 321 322 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 323 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 324 325 ARG_PTR_TO_CTX, /* pointer to context */ 326 ARG_PTR_TO_CTX_OR_NULL, /* pointer to context or NULL */ 327 ARG_ANYTHING, /* any (initialized) argument is ok */ 328 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 329 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 330 ARG_PTR_TO_INT, /* pointer to int */ 331 ARG_PTR_TO_LONG, /* pointer to long */ 332 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 333 ARG_PTR_TO_SOCKET_OR_NULL, /* pointer to bpf_sock (fullsock) or NULL */ 334 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */ 335 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */ 336 ARG_PTR_TO_ALLOC_MEM_OR_NULL, /* pointer to dynamically allocated memory or NULL */ 337 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ 338 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ 339 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ 340 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */ 341 ARG_PTR_TO_STACK_OR_NULL, /* pointer to stack or NULL */ 342 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */ 343 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */ 344 __BPF_ARG_TYPE_MAX, 345 }; 346 347 /* type of values returned from helper functions */ 348 enum bpf_return_type { 349 RET_INTEGER, /* function returns integer */ 350 RET_VOID, /* function doesn't return anything */ 351 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 352 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */ 353 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */ 354 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */ 355 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */ 356 RET_PTR_TO_ALLOC_MEM_OR_NULL, /* returns a pointer to dynamically allocated memory or NULL */ 357 RET_PTR_TO_BTF_ID_OR_NULL, /* returns a pointer to a btf_id or NULL */ 358 RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL, /* returns a pointer to a valid memory or a btf_id or NULL */ 359 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ 360 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */ 361 }; 362 363 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 364 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 365 * instructions after verifying 366 */ 367 struct bpf_func_proto { 368 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 369 bool gpl_only; 370 bool pkt_access; 371 enum bpf_return_type ret_type; 372 union { 373 struct { 374 enum bpf_arg_type arg1_type; 375 enum bpf_arg_type arg2_type; 376 enum bpf_arg_type arg3_type; 377 enum bpf_arg_type arg4_type; 378 enum bpf_arg_type arg5_type; 379 }; 380 enum bpf_arg_type arg_type[5]; 381 }; 382 union { 383 struct { 384 u32 *arg1_btf_id; 385 u32 *arg2_btf_id; 386 u32 *arg3_btf_id; 387 u32 *arg4_btf_id; 388 u32 *arg5_btf_id; 389 }; 390 u32 *arg_btf_id[5]; 391 }; 392 int *ret_btf_id; /* return value btf_id */ 393 bool (*allowed)(const struct bpf_prog *prog); 394 }; 395 396 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 397 * the first argument to eBPF programs. 398 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 399 */ 400 struct bpf_context; 401 402 enum bpf_access_type { 403 BPF_READ = 1, 404 BPF_WRITE = 2 405 }; 406 407 /* types of values stored in eBPF registers */ 408 /* Pointer types represent: 409 * pointer 410 * pointer + imm 411 * pointer + (u16) var 412 * pointer + (u16) var + imm 413 * if (range > 0) then [ptr, ptr + range - off) is safe to access 414 * if (id > 0) means that some 'var' was added 415 * if (off > 0) means that 'imm' was added 416 */ 417 enum bpf_reg_type { 418 NOT_INIT = 0, /* nothing was written into register */ 419 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 420 PTR_TO_CTX, /* reg points to bpf_context */ 421 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 422 PTR_TO_MAP_VALUE, /* reg points to map element value */ 423 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */ 424 PTR_TO_STACK, /* reg == frame_pointer + offset */ 425 PTR_TO_PACKET_META, /* skb->data - meta_len */ 426 PTR_TO_PACKET, /* reg points to skb->data */ 427 PTR_TO_PACKET_END, /* skb->data + headlen */ 428 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 429 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 430 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */ 431 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 432 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */ 433 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 434 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */ 435 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 436 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 437 /* PTR_TO_BTF_ID points to a kernel struct that does not need 438 * to be null checked by the BPF program. This does not imply the 439 * pointer is _not_ null and in practice this can easily be a null 440 * pointer when reading pointer chains. The assumption is program 441 * context will handle null pointer dereference typically via fault 442 * handling. The verifier must keep this in mind and can make no 443 * assumptions about null or non-null when doing branch analysis. 444 * Further, when passed into helpers the helpers can not, without 445 * additional context, assume the value is non-null. 446 */ 447 PTR_TO_BTF_ID, 448 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not 449 * been checked for null. Used primarily to inform the verifier 450 * an explicit null check is required for this struct. 451 */ 452 PTR_TO_BTF_ID_OR_NULL, 453 PTR_TO_MEM, /* reg points to valid memory region */ 454 PTR_TO_MEM_OR_NULL, /* reg points to valid memory region or NULL */ 455 PTR_TO_RDONLY_BUF, /* reg points to a readonly buffer */ 456 PTR_TO_RDONLY_BUF_OR_NULL, /* reg points to a readonly buffer or NULL */ 457 PTR_TO_RDWR_BUF, /* reg points to a read/write buffer */ 458 PTR_TO_RDWR_BUF_OR_NULL, /* reg points to a read/write buffer or NULL */ 459 PTR_TO_PERCPU_BTF_ID, /* reg points to a percpu kernel variable */ 460 PTR_TO_FUNC, /* reg points to a bpf program function */ 461 PTR_TO_MAP_KEY, /* reg points to a map element key */ 462 __BPF_REG_TYPE_MAX, 463 }; 464 465 /* The information passed from prog-specific *_is_valid_access 466 * back to the verifier. 467 */ 468 struct bpf_insn_access_aux { 469 enum bpf_reg_type reg_type; 470 union { 471 int ctx_field_size; 472 struct { 473 struct btf *btf; 474 u32 btf_id; 475 }; 476 }; 477 struct bpf_verifier_log *log; /* for verbose logs */ 478 }; 479 480 static inline void 481 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 482 { 483 aux->ctx_field_size = size; 484 } 485 486 struct bpf_prog_ops { 487 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 488 union bpf_attr __user *uattr); 489 }; 490 491 struct bpf_verifier_ops { 492 /* return eBPF function prototype for verification */ 493 const struct bpf_func_proto * 494 (*get_func_proto)(enum bpf_func_id func_id, 495 const struct bpf_prog *prog); 496 497 /* return true if 'size' wide access at offset 'off' within bpf_context 498 * with 'type' (read or write) is allowed 499 */ 500 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 501 const struct bpf_prog *prog, 502 struct bpf_insn_access_aux *info); 503 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 504 const struct bpf_prog *prog); 505 int (*gen_ld_abs)(const struct bpf_insn *orig, 506 struct bpf_insn *insn_buf); 507 u32 (*convert_ctx_access)(enum bpf_access_type type, 508 const struct bpf_insn *src, 509 struct bpf_insn *dst, 510 struct bpf_prog *prog, u32 *target_size); 511 int (*btf_struct_access)(struct bpf_verifier_log *log, 512 const struct btf *btf, 513 const struct btf_type *t, int off, int size, 514 enum bpf_access_type atype, 515 u32 *next_btf_id); 516 bool (*check_kfunc_call)(u32 kfunc_btf_id, struct module *owner); 517 }; 518 519 struct bpf_prog_offload_ops { 520 /* verifier basic callbacks */ 521 int (*insn_hook)(struct bpf_verifier_env *env, 522 int insn_idx, int prev_insn_idx); 523 int (*finalize)(struct bpf_verifier_env *env); 524 /* verifier optimization callbacks (called after .finalize) */ 525 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 526 struct bpf_insn *insn); 527 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 528 /* program management callbacks */ 529 int (*prepare)(struct bpf_prog *prog); 530 int (*translate)(struct bpf_prog *prog); 531 void (*destroy)(struct bpf_prog *prog); 532 }; 533 534 struct bpf_prog_offload { 535 struct bpf_prog *prog; 536 struct net_device *netdev; 537 struct bpf_offload_dev *offdev; 538 void *dev_priv; 539 struct list_head offloads; 540 bool dev_state; 541 bool opt_failed; 542 void *jited_image; 543 u32 jited_len; 544 }; 545 546 enum bpf_cgroup_storage_type { 547 BPF_CGROUP_STORAGE_SHARED, 548 BPF_CGROUP_STORAGE_PERCPU, 549 __BPF_CGROUP_STORAGE_MAX 550 }; 551 552 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 553 554 /* The longest tracepoint has 12 args. 555 * See include/trace/bpf_probe.h 556 */ 557 #define MAX_BPF_FUNC_ARGS 12 558 559 /* The maximum number of arguments passed through registers 560 * a single function may have. 561 */ 562 #define MAX_BPF_FUNC_REG_ARGS 5 563 564 struct btf_func_model { 565 u8 ret_size; 566 u8 nr_args; 567 u8 arg_size[MAX_BPF_FUNC_ARGS]; 568 }; 569 570 /* Restore arguments before returning from trampoline to let original function 571 * continue executing. This flag is used for fentry progs when there are no 572 * fexit progs. 573 */ 574 #define BPF_TRAMP_F_RESTORE_REGS BIT(0) 575 /* Call original function after fentry progs, but before fexit progs. 576 * Makes sense for fentry/fexit, normal calls and indirect calls. 577 */ 578 #define BPF_TRAMP_F_CALL_ORIG BIT(1) 579 /* Skip current frame and return to parent. Makes sense for fentry/fexit 580 * programs only. Should not be used with normal calls and indirect calls. 581 */ 582 #define BPF_TRAMP_F_SKIP_FRAME BIT(2) 583 /* Store IP address of the caller on the trampoline stack, 584 * so it's available for trampoline's programs. 585 */ 586 #define BPF_TRAMP_F_IP_ARG BIT(3) 587 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */ 588 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4) 589 590 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 591 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2 592 */ 593 #define BPF_MAX_TRAMP_PROGS 38 594 595 struct bpf_tramp_progs { 596 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS]; 597 int nr_progs; 598 }; 599 600 /* Different use cases for BPF trampoline: 601 * 1. replace nop at the function entry (kprobe equivalent) 602 * flags = BPF_TRAMP_F_RESTORE_REGS 603 * fentry = a set of programs to run before returning from trampoline 604 * 605 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 606 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 607 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 608 * fentry = a set of program to run before calling original function 609 * fexit = a set of program to run after original function 610 * 611 * 3. replace direct call instruction anywhere in the function body 612 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 613 * With flags = 0 614 * fentry = a set of programs to run before returning from trampoline 615 * With flags = BPF_TRAMP_F_CALL_ORIG 616 * orig_call = original callback addr or direct function addr 617 * fentry = a set of program to run before calling original function 618 * fexit = a set of program to run after original function 619 */ 620 struct bpf_tramp_image; 621 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end, 622 const struct btf_func_model *m, u32 flags, 623 struct bpf_tramp_progs *tprogs, 624 void *orig_call); 625 /* these two functions are called from generated trampoline */ 626 u64 notrace __bpf_prog_enter(struct bpf_prog *prog); 627 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start); 628 u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog); 629 void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start); 630 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr); 631 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr); 632 633 struct bpf_ksym { 634 unsigned long start; 635 unsigned long end; 636 char name[KSYM_NAME_LEN]; 637 struct list_head lnode; 638 struct latch_tree_node tnode; 639 bool prog; 640 }; 641 642 enum bpf_tramp_prog_type { 643 BPF_TRAMP_FENTRY, 644 BPF_TRAMP_FEXIT, 645 BPF_TRAMP_MODIFY_RETURN, 646 BPF_TRAMP_MAX, 647 BPF_TRAMP_REPLACE, /* more than MAX */ 648 }; 649 650 struct bpf_tramp_image { 651 void *image; 652 struct bpf_ksym ksym; 653 struct percpu_ref pcref; 654 void *ip_after_call; 655 void *ip_epilogue; 656 union { 657 struct rcu_head rcu; 658 struct work_struct work; 659 }; 660 }; 661 662 struct bpf_trampoline { 663 /* hlist for trampoline_table */ 664 struct hlist_node hlist; 665 /* serializes access to fields of this trampoline */ 666 struct mutex mutex; 667 refcount_t refcnt; 668 u64 key; 669 struct { 670 struct btf_func_model model; 671 void *addr; 672 bool ftrace_managed; 673 } func; 674 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 675 * program by replacing one of its functions. func.addr is the address 676 * of the function it replaced. 677 */ 678 struct bpf_prog *extension_prog; 679 /* list of BPF programs using this trampoline */ 680 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 681 /* Number of attached programs. A counter per kind. */ 682 int progs_cnt[BPF_TRAMP_MAX]; 683 /* Executable image of trampoline */ 684 struct bpf_tramp_image *cur_image; 685 u64 selector; 686 struct module *mod; 687 }; 688 689 struct bpf_attach_target_info { 690 struct btf_func_model fmodel; 691 long tgt_addr; 692 const char *tgt_name; 693 const struct btf_type *tgt_type; 694 }; 695 696 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 697 698 struct bpf_dispatcher_prog { 699 struct bpf_prog *prog; 700 refcount_t users; 701 }; 702 703 struct bpf_dispatcher { 704 /* dispatcher mutex */ 705 struct mutex mutex; 706 void *func; 707 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 708 int num_progs; 709 void *image; 710 u32 image_off; 711 struct bpf_ksym ksym; 712 }; 713 714 static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func( 715 const void *ctx, 716 const struct bpf_insn *insnsi, 717 unsigned int (*bpf_func)(const void *, 718 const struct bpf_insn *)) 719 { 720 return bpf_func(ctx, insnsi); 721 } 722 #ifdef CONFIG_BPF_JIT 723 int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 724 int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 725 struct bpf_trampoline *bpf_trampoline_get(u64 key, 726 struct bpf_attach_target_info *tgt_info); 727 void bpf_trampoline_put(struct bpf_trampoline *tr); 728 #define BPF_DISPATCHER_INIT(_name) { \ 729 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 730 .func = &_name##_func, \ 731 .progs = {}, \ 732 .num_progs = 0, \ 733 .image = NULL, \ 734 .image_off = 0, \ 735 .ksym = { \ 736 .name = #_name, \ 737 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 738 }, \ 739 } 740 741 #define DEFINE_BPF_DISPATCHER(name) \ 742 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \ 743 const void *ctx, \ 744 const struct bpf_insn *insnsi, \ 745 unsigned int (*bpf_func)(const void *, \ 746 const struct bpf_insn *)) \ 747 { \ 748 return bpf_func(ctx, insnsi); \ 749 } \ 750 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 751 struct bpf_dispatcher bpf_dispatcher_##name = \ 752 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 753 #define DECLARE_BPF_DISPATCHER(name) \ 754 unsigned int bpf_dispatcher_##name##_func( \ 755 const void *ctx, \ 756 const struct bpf_insn *insnsi, \ 757 unsigned int (*bpf_func)(const void *, \ 758 const struct bpf_insn *)); \ 759 extern struct bpf_dispatcher bpf_dispatcher_##name; 760 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 761 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 762 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 763 struct bpf_prog *to); 764 /* Called only from JIT-enabled code, so there's no need for stubs. */ 765 void *bpf_jit_alloc_exec_page(void); 766 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym); 767 void bpf_image_ksym_del(struct bpf_ksym *ksym); 768 void bpf_ksym_add(struct bpf_ksym *ksym); 769 void bpf_ksym_del(struct bpf_ksym *ksym); 770 int bpf_jit_charge_modmem(u32 pages); 771 void bpf_jit_uncharge_modmem(u32 pages); 772 #else 773 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog, 774 struct bpf_trampoline *tr) 775 { 776 return -ENOTSUPP; 777 } 778 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog, 779 struct bpf_trampoline *tr) 780 { 781 return -ENOTSUPP; 782 } 783 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 784 struct bpf_attach_target_info *tgt_info) 785 { 786 return ERR_PTR(-EOPNOTSUPP); 787 } 788 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 789 #define DEFINE_BPF_DISPATCHER(name) 790 #define DECLARE_BPF_DISPATCHER(name) 791 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 792 #define BPF_DISPATCHER_PTR(name) NULL 793 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 794 struct bpf_prog *from, 795 struct bpf_prog *to) {} 796 static inline bool is_bpf_image_address(unsigned long address) 797 { 798 return false; 799 } 800 #endif 801 802 struct bpf_func_info_aux { 803 u16 linkage; 804 bool unreliable; 805 }; 806 807 enum bpf_jit_poke_reason { 808 BPF_POKE_REASON_TAIL_CALL, 809 }; 810 811 /* Descriptor of pokes pointing /into/ the JITed image. */ 812 struct bpf_jit_poke_descriptor { 813 void *tailcall_target; 814 void *tailcall_bypass; 815 void *bypass_addr; 816 void *aux; 817 union { 818 struct { 819 struct bpf_map *map; 820 u32 key; 821 } tail_call; 822 }; 823 bool tailcall_target_stable; 824 u8 adj_off; 825 u16 reason; 826 u32 insn_idx; 827 }; 828 829 /* reg_type info for ctx arguments */ 830 struct bpf_ctx_arg_aux { 831 u32 offset; 832 enum bpf_reg_type reg_type; 833 u32 btf_id; 834 }; 835 836 struct btf_mod_pair { 837 struct btf *btf; 838 struct module *module; 839 }; 840 841 struct bpf_kfunc_desc_tab; 842 843 struct bpf_prog_aux { 844 atomic64_t refcnt; 845 u32 used_map_cnt; 846 u32 used_btf_cnt; 847 u32 max_ctx_offset; 848 u32 max_pkt_offset; 849 u32 max_tp_access; 850 u32 stack_depth; 851 u32 id; 852 u32 func_cnt; /* used by non-func prog as the number of func progs */ 853 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 854 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 855 u32 ctx_arg_info_size; 856 u32 max_rdonly_access; 857 u32 max_rdwr_access; 858 struct btf *attach_btf; 859 const struct bpf_ctx_arg_aux *ctx_arg_info; 860 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 861 struct bpf_prog *dst_prog; 862 struct bpf_trampoline *dst_trampoline; 863 enum bpf_prog_type saved_dst_prog_type; 864 enum bpf_attach_type saved_dst_attach_type; 865 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 866 bool offload_requested; 867 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 868 bool func_proto_unreliable; 869 bool sleepable; 870 bool tail_call_reachable; 871 struct hlist_node tramp_hlist; 872 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 873 const struct btf_type *attach_func_proto; 874 /* function name for valid attach_btf_id */ 875 const char *attach_func_name; 876 struct bpf_prog **func; 877 void *jit_data; /* JIT specific data. arch dependent */ 878 struct bpf_jit_poke_descriptor *poke_tab; 879 struct bpf_kfunc_desc_tab *kfunc_tab; 880 struct bpf_kfunc_btf_tab *kfunc_btf_tab; 881 u32 size_poke_tab; 882 struct bpf_ksym ksym; 883 const struct bpf_prog_ops *ops; 884 struct bpf_map **used_maps; 885 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 886 struct btf_mod_pair *used_btfs; 887 struct bpf_prog *prog; 888 struct user_struct *user; 889 u64 load_time; /* ns since boottime */ 890 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 891 char name[BPF_OBJ_NAME_LEN]; 892 #ifdef CONFIG_SECURITY 893 void *security; 894 #endif 895 struct bpf_prog_offload *offload; 896 struct btf *btf; 897 struct bpf_func_info *func_info; 898 struct bpf_func_info_aux *func_info_aux; 899 /* bpf_line_info loaded from userspace. linfo->insn_off 900 * has the xlated insn offset. 901 * Both the main and sub prog share the same linfo. 902 * The subprog can access its first linfo by 903 * using the linfo_idx. 904 */ 905 struct bpf_line_info *linfo; 906 /* jited_linfo is the jited addr of the linfo. It has a 907 * one to one mapping to linfo: 908 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 909 * Both the main and sub prog share the same jited_linfo. 910 * The subprog can access its first jited_linfo by 911 * using the linfo_idx. 912 */ 913 void **jited_linfo; 914 u32 func_info_cnt; 915 u32 nr_linfo; 916 /* subprog can use linfo_idx to access its first linfo and 917 * jited_linfo. 918 * main prog always has linfo_idx == 0 919 */ 920 u32 linfo_idx; 921 u32 num_exentries; 922 struct exception_table_entry *extable; 923 union { 924 struct work_struct work; 925 struct rcu_head rcu; 926 }; 927 }; 928 929 struct bpf_array_aux { 930 /* 'Ownership' of prog array is claimed by the first program that 931 * is going to use this map or by the first program which FD is 932 * stored in the map to make sure that all callers and callees have 933 * the same prog type and JITed flag. 934 */ 935 enum bpf_prog_type type; 936 bool jited; 937 /* Programs with direct jumps into programs part of this array. */ 938 struct list_head poke_progs; 939 struct bpf_map *map; 940 struct mutex poke_mutex; 941 struct work_struct work; 942 }; 943 944 struct bpf_link { 945 atomic64_t refcnt; 946 u32 id; 947 enum bpf_link_type type; 948 const struct bpf_link_ops *ops; 949 struct bpf_prog *prog; 950 struct work_struct work; 951 }; 952 953 struct bpf_link_ops { 954 void (*release)(struct bpf_link *link); 955 void (*dealloc)(struct bpf_link *link); 956 int (*detach)(struct bpf_link *link); 957 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 958 struct bpf_prog *old_prog); 959 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 960 int (*fill_link_info)(const struct bpf_link *link, 961 struct bpf_link_info *info); 962 }; 963 964 struct bpf_link_primer { 965 struct bpf_link *link; 966 struct file *file; 967 int fd; 968 u32 id; 969 }; 970 971 struct bpf_struct_ops_value; 972 struct btf_member; 973 974 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 975 struct bpf_struct_ops { 976 const struct bpf_verifier_ops *verifier_ops; 977 int (*init)(struct btf *btf); 978 int (*check_member)(const struct btf_type *t, 979 const struct btf_member *member); 980 int (*init_member)(const struct btf_type *t, 981 const struct btf_member *member, 982 void *kdata, const void *udata); 983 int (*reg)(void *kdata); 984 void (*unreg)(void *kdata); 985 const struct btf_type *type; 986 const struct btf_type *value_type; 987 const char *name; 988 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 989 u32 type_id; 990 u32 value_id; 991 }; 992 993 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 994 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 995 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id); 996 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log); 997 bool bpf_struct_ops_get(const void *kdata); 998 void bpf_struct_ops_put(const void *kdata); 999 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 1000 void *value); 1001 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1002 { 1003 if (owner == BPF_MODULE_OWNER) 1004 return bpf_struct_ops_get(data); 1005 else 1006 return try_module_get(owner); 1007 } 1008 static inline void bpf_module_put(const void *data, struct module *owner) 1009 { 1010 if (owner == BPF_MODULE_OWNER) 1011 bpf_struct_ops_put(data); 1012 else 1013 module_put(owner); 1014 } 1015 #else 1016 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id) 1017 { 1018 return NULL; 1019 } 1020 static inline void bpf_struct_ops_init(struct btf *btf, 1021 struct bpf_verifier_log *log) 1022 { 1023 } 1024 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1025 { 1026 return try_module_get(owner); 1027 } 1028 static inline void bpf_module_put(const void *data, struct module *owner) 1029 { 1030 module_put(owner); 1031 } 1032 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 1033 void *key, 1034 void *value) 1035 { 1036 return -EINVAL; 1037 } 1038 #endif 1039 1040 struct bpf_array { 1041 struct bpf_map map; 1042 u32 elem_size; 1043 u32 index_mask; 1044 struct bpf_array_aux *aux; 1045 union { 1046 char value[0] __aligned(8); 1047 void *ptrs[0] __aligned(8); 1048 void __percpu *pptrs[0] __aligned(8); 1049 }; 1050 }; 1051 1052 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 1053 #define MAX_TAIL_CALL_CNT 32 1054 1055 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 1056 BPF_F_RDONLY_PROG | \ 1057 BPF_F_WRONLY | \ 1058 BPF_F_WRONLY_PROG) 1059 1060 #define BPF_MAP_CAN_READ BIT(0) 1061 #define BPF_MAP_CAN_WRITE BIT(1) 1062 1063 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 1064 { 1065 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1066 1067 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 1068 * not possible. 1069 */ 1070 if (access_flags & BPF_F_RDONLY_PROG) 1071 return BPF_MAP_CAN_READ; 1072 else if (access_flags & BPF_F_WRONLY_PROG) 1073 return BPF_MAP_CAN_WRITE; 1074 else 1075 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 1076 } 1077 1078 static inline bool bpf_map_flags_access_ok(u32 access_flags) 1079 { 1080 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 1081 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1082 } 1083 1084 struct bpf_event_entry { 1085 struct perf_event *event; 1086 struct file *perf_file; 1087 struct file *map_file; 1088 struct rcu_head rcu; 1089 }; 1090 1091 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); 1092 int bpf_prog_calc_tag(struct bpf_prog *fp); 1093 1094 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 1095 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void); 1096 1097 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 1098 unsigned long off, unsigned long len); 1099 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 1100 const struct bpf_insn *src, 1101 struct bpf_insn *dst, 1102 struct bpf_prog *prog, 1103 u32 *target_size); 1104 1105 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 1106 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 1107 1108 /* an array of programs to be executed under rcu_lock. 1109 * 1110 * Typical usage: 1111 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, bpf_prog_run); 1112 * 1113 * the structure returned by bpf_prog_array_alloc() should be populated 1114 * with program pointers and the last pointer must be NULL. 1115 * The user has to keep refcnt on the program and make sure the program 1116 * is removed from the array before bpf_prog_put(). 1117 * The 'struct bpf_prog_array *' should only be replaced with xchg() 1118 * since other cpus are walking the array of pointers in parallel. 1119 */ 1120 struct bpf_prog_array_item { 1121 struct bpf_prog *prog; 1122 union { 1123 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1124 u64 bpf_cookie; 1125 }; 1126 }; 1127 1128 struct bpf_prog_array { 1129 struct rcu_head rcu; 1130 struct bpf_prog_array_item items[]; 1131 }; 1132 1133 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 1134 void bpf_prog_array_free(struct bpf_prog_array *progs); 1135 int bpf_prog_array_length(struct bpf_prog_array *progs); 1136 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 1137 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 1138 __u32 __user *prog_ids, u32 cnt); 1139 1140 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 1141 struct bpf_prog *old_prog); 1142 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 1143 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 1144 struct bpf_prog *prog); 1145 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 1146 u32 *prog_ids, u32 request_cnt, 1147 u32 *prog_cnt); 1148 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 1149 struct bpf_prog *exclude_prog, 1150 struct bpf_prog *include_prog, 1151 u64 bpf_cookie, 1152 struct bpf_prog_array **new_array); 1153 1154 struct bpf_run_ctx {}; 1155 1156 struct bpf_cg_run_ctx { 1157 struct bpf_run_ctx run_ctx; 1158 const struct bpf_prog_array_item *prog_item; 1159 }; 1160 1161 struct bpf_trace_run_ctx { 1162 struct bpf_run_ctx run_ctx; 1163 u64 bpf_cookie; 1164 }; 1165 1166 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 1167 { 1168 struct bpf_run_ctx *old_ctx = NULL; 1169 1170 #ifdef CONFIG_BPF_SYSCALL 1171 old_ctx = current->bpf_ctx; 1172 current->bpf_ctx = new_ctx; 1173 #endif 1174 return old_ctx; 1175 } 1176 1177 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 1178 { 1179 #ifdef CONFIG_BPF_SYSCALL 1180 current->bpf_ctx = old_ctx; 1181 #endif 1182 } 1183 1184 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 1185 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 1186 /* BPF program asks to set CN on the packet. */ 1187 #define BPF_RET_SET_CN (1 << 0) 1188 1189 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 1190 1191 static __always_inline u32 1192 BPF_PROG_RUN_ARRAY_CG_FLAGS(const struct bpf_prog_array __rcu *array_rcu, 1193 const void *ctx, bpf_prog_run_fn run_prog, 1194 u32 *ret_flags) 1195 { 1196 const struct bpf_prog_array_item *item; 1197 const struct bpf_prog *prog; 1198 const struct bpf_prog_array *array; 1199 struct bpf_run_ctx *old_run_ctx; 1200 struct bpf_cg_run_ctx run_ctx; 1201 u32 ret = 1; 1202 u32 func_ret; 1203 1204 migrate_disable(); 1205 rcu_read_lock(); 1206 array = rcu_dereference(array_rcu); 1207 item = &array->items[0]; 1208 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1209 while ((prog = READ_ONCE(item->prog))) { 1210 run_ctx.prog_item = item; 1211 func_ret = run_prog(prog, ctx); 1212 ret &= (func_ret & 1); 1213 *(ret_flags) |= (func_ret >> 1); 1214 item++; 1215 } 1216 bpf_reset_run_ctx(old_run_ctx); 1217 rcu_read_unlock(); 1218 migrate_enable(); 1219 return ret; 1220 } 1221 1222 static __always_inline u32 1223 BPF_PROG_RUN_ARRAY_CG(const struct bpf_prog_array __rcu *array_rcu, 1224 const void *ctx, bpf_prog_run_fn run_prog) 1225 { 1226 const struct bpf_prog_array_item *item; 1227 const struct bpf_prog *prog; 1228 const struct bpf_prog_array *array; 1229 struct bpf_run_ctx *old_run_ctx; 1230 struct bpf_cg_run_ctx run_ctx; 1231 u32 ret = 1; 1232 1233 migrate_disable(); 1234 rcu_read_lock(); 1235 array = rcu_dereference(array_rcu); 1236 item = &array->items[0]; 1237 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1238 while ((prog = READ_ONCE(item->prog))) { 1239 run_ctx.prog_item = item; 1240 ret &= run_prog(prog, ctx); 1241 item++; 1242 } 1243 bpf_reset_run_ctx(old_run_ctx); 1244 rcu_read_unlock(); 1245 migrate_enable(); 1246 return ret; 1247 } 1248 1249 static __always_inline u32 1250 BPF_PROG_RUN_ARRAY(const struct bpf_prog_array __rcu *array_rcu, 1251 const void *ctx, bpf_prog_run_fn run_prog) 1252 { 1253 const struct bpf_prog_array_item *item; 1254 const struct bpf_prog *prog; 1255 const struct bpf_prog_array *array; 1256 struct bpf_run_ctx *old_run_ctx; 1257 struct bpf_trace_run_ctx run_ctx; 1258 u32 ret = 1; 1259 1260 migrate_disable(); 1261 rcu_read_lock(); 1262 array = rcu_dereference(array_rcu); 1263 if (unlikely(!array)) 1264 goto out; 1265 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1266 item = &array->items[0]; 1267 while ((prog = READ_ONCE(item->prog))) { 1268 run_ctx.bpf_cookie = item->bpf_cookie; 1269 ret &= run_prog(prog, ctx); 1270 item++; 1271 } 1272 bpf_reset_run_ctx(old_run_ctx); 1273 out: 1274 rcu_read_unlock(); 1275 migrate_enable(); 1276 return ret; 1277 } 1278 1279 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs 1280 * so BPF programs can request cwr for TCP packets. 1281 * 1282 * Current cgroup skb programs can only return 0 or 1 (0 to drop the 1283 * packet. This macro changes the behavior so the low order bit 1284 * indicates whether the packet should be dropped (0) or not (1) 1285 * and the next bit is a congestion notification bit. This could be 1286 * used by TCP to call tcp_enter_cwr() 1287 * 1288 * Hence, new allowed return values of CGROUP EGRESS BPF programs are: 1289 * 0: drop packet 1290 * 1: keep packet 1291 * 2: drop packet and cn 1292 * 3: keep packet and cn 1293 * 1294 * This macro then converts it to one of the NET_XMIT or an error 1295 * code that is then interpreted as drop packet (and no cn): 1296 * 0: NET_XMIT_SUCCESS skb should be transmitted 1297 * 1: NET_XMIT_DROP skb should be dropped and cn 1298 * 2: NET_XMIT_CN skb should be transmitted and cn 1299 * 3: -EPERM skb should be dropped 1300 */ 1301 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \ 1302 ({ \ 1303 u32 _flags = 0; \ 1304 bool _cn; \ 1305 u32 _ret; \ 1306 _ret = BPF_PROG_RUN_ARRAY_CG_FLAGS(array, ctx, func, &_flags); \ 1307 _cn = _flags & BPF_RET_SET_CN; \ 1308 if (_ret) \ 1309 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \ 1310 else \ 1311 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \ 1312 _ret; \ 1313 }) 1314 1315 #ifdef CONFIG_BPF_SYSCALL 1316 DECLARE_PER_CPU(int, bpf_prog_active); 1317 extern struct mutex bpf_stats_enabled_mutex; 1318 1319 /* 1320 * Block execution of BPF programs attached to instrumentation (perf, 1321 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 1322 * these events can happen inside a region which holds a map bucket lock 1323 * and can deadlock on it. 1324 * 1325 * Use the preemption safe inc/dec variants on RT because migrate disable 1326 * is preemptible on RT and preemption in the middle of the RMW operation 1327 * might lead to inconsistent state. Use the raw variants for non RT 1328 * kernels as migrate_disable() maps to preempt_disable() so the slightly 1329 * more expensive save operation can be avoided. 1330 */ 1331 static inline void bpf_disable_instrumentation(void) 1332 { 1333 migrate_disable(); 1334 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1335 this_cpu_inc(bpf_prog_active); 1336 else 1337 __this_cpu_inc(bpf_prog_active); 1338 } 1339 1340 static inline void bpf_enable_instrumentation(void) 1341 { 1342 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1343 this_cpu_dec(bpf_prog_active); 1344 else 1345 __this_cpu_dec(bpf_prog_active); 1346 migrate_enable(); 1347 } 1348 1349 extern const struct file_operations bpf_map_fops; 1350 extern const struct file_operations bpf_prog_fops; 1351 extern const struct file_operations bpf_iter_fops; 1352 1353 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 1354 extern const struct bpf_prog_ops _name ## _prog_ops; \ 1355 extern const struct bpf_verifier_ops _name ## _verifier_ops; 1356 #define BPF_MAP_TYPE(_id, _ops) \ 1357 extern const struct bpf_map_ops _ops; 1358 #define BPF_LINK_TYPE(_id, _name) 1359 #include <linux/bpf_types.h> 1360 #undef BPF_PROG_TYPE 1361 #undef BPF_MAP_TYPE 1362 #undef BPF_LINK_TYPE 1363 1364 extern const struct bpf_prog_ops bpf_offload_prog_ops; 1365 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 1366 extern const struct bpf_verifier_ops xdp_analyzer_ops; 1367 1368 struct bpf_prog *bpf_prog_get(u32 ufd); 1369 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 1370 bool attach_drv); 1371 void bpf_prog_add(struct bpf_prog *prog, int i); 1372 void bpf_prog_sub(struct bpf_prog *prog, int i); 1373 void bpf_prog_inc(struct bpf_prog *prog); 1374 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 1375 void bpf_prog_put(struct bpf_prog *prog); 1376 1377 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 1378 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 1379 1380 struct bpf_map *bpf_map_get(u32 ufd); 1381 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 1382 struct bpf_map *__bpf_map_get(struct fd f); 1383 void bpf_map_inc(struct bpf_map *map); 1384 void bpf_map_inc_with_uref(struct bpf_map *map); 1385 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 1386 void bpf_map_put_with_uref(struct bpf_map *map); 1387 void bpf_map_put(struct bpf_map *map); 1388 void *bpf_map_area_alloc(u64 size, int numa_node); 1389 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 1390 void bpf_map_area_free(void *base); 1391 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 1392 int generic_map_lookup_batch(struct bpf_map *map, 1393 const union bpf_attr *attr, 1394 union bpf_attr __user *uattr); 1395 int generic_map_update_batch(struct bpf_map *map, 1396 const union bpf_attr *attr, 1397 union bpf_attr __user *uattr); 1398 int generic_map_delete_batch(struct bpf_map *map, 1399 const union bpf_attr *attr, 1400 union bpf_attr __user *uattr); 1401 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 1402 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 1403 1404 #ifdef CONFIG_MEMCG_KMEM 1405 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1406 int node); 1407 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 1408 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 1409 size_t align, gfp_t flags); 1410 #else 1411 static inline void * 1412 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1413 int node) 1414 { 1415 return kmalloc_node(size, flags, node); 1416 } 1417 1418 static inline void * 1419 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags) 1420 { 1421 return kzalloc(size, flags); 1422 } 1423 1424 static inline void __percpu * 1425 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align, 1426 gfp_t flags) 1427 { 1428 return __alloc_percpu_gfp(size, align, flags); 1429 } 1430 #endif 1431 1432 extern int sysctl_unprivileged_bpf_disabled; 1433 1434 static inline bool bpf_allow_ptr_leaks(void) 1435 { 1436 return perfmon_capable(); 1437 } 1438 1439 static inline bool bpf_allow_uninit_stack(void) 1440 { 1441 return perfmon_capable(); 1442 } 1443 1444 static inline bool bpf_allow_ptr_to_map_access(void) 1445 { 1446 return perfmon_capable(); 1447 } 1448 1449 static inline bool bpf_bypass_spec_v1(void) 1450 { 1451 return perfmon_capable(); 1452 } 1453 1454 static inline bool bpf_bypass_spec_v4(void) 1455 { 1456 return perfmon_capable(); 1457 } 1458 1459 int bpf_map_new_fd(struct bpf_map *map, int flags); 1460 int bpf_prog_new_fd(struct bpf_prog *prog); 1461 1462 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1463 const struct bpf_link_ops *ops, struct bpf_prog *prog); 1464 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 1465 int bpf_link_settle(struct bpf_link_primer *primer); 1466 void bpf_link_cleanup(struct bpf_link_primer *primer); 1467 void bpf_link_inc(struct bpf_link *link); 1468 void bpf_link_put(struct bpf_link *link); 1469 int bpf_link_new_fd(struct bpf_link *link); 1470 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd); 1471 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 1472 1473 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 1474 int bpf_obj_get_user(const char __user *pathname, int flags); 1475 1476 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 1477 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 1478 extern int bpf_iter_ ## target(args); \ 1479 int __init bpf_iter_ ## target(args) { return 0; } 1480 1481 struct bpf_iter_aux_info { 1482 struct bpf_map *map; 1483 }; 1484 1485 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 1486 union bpf_iter_link_info *linfo, 1487 struct bpf_iter_aux_info *aux); 1488 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 1489 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 1490 struct seq_file *seq); 1491 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 1492 struct bpf_link_info *info); 1493 typedef const struct bpf_func_proto * 1494 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 1495 const struct bpf_prog *prog); 1496 1497 enum bpf_iter_feature { 1498 BPF_ITER_RESCHED = BIT(0), 1499 }; 1500 1501 #define BPF_ITER_CTX_ARG_MAX 2 1502 struct bpf_iter_reg { 1503 const char *target; 1504 bpf_iter_attach_target_t attach_target; 1505 bpf_iter_detach_target_t detach_target; 1506 bpf_iter_show_fdinfo_t show_fdinfo; 1507 bpf_iter_fill_link_info_t fill_link_info; 1508 bpf_iter_get_func_proto_t get_func_proto; 1509 u32 ctx_arg_info_size; 1510 u32 feature; 1511 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 1512 const struct bpf_iter_seq_info *seq_info; 1513 }; 1514 1515 struct bpf_iter_meta { 1516 __bpf_md_ptr(struct seq_file *, seq); 1517 u64 session_id; 1518 u64 seq_num; 1519 }; 1520 1521 struct bpf_iter__bpf_map_elem { 1522 __bpf_md_ptr(struct bpf_iter_meta *, meta); 1523 __bpf_md_ptr(struct bpf_map *, map); 1524 __bpf_md_ptr(void *, key); 1525 __bpf_md_ptr(void *, value); 1526 }; 1527 1528 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 1529 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 1530 bool bpf_iter_prog_supported(struct bpf_prog *prog); 1531 const struct bpf_func_proto * 1532 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 1533 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 1534 int bpf_iter_new_fd(struct bpf_link *link); 1535 bool bpf_link_is_iter(struct bpf_link *link); 1536 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 1537 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 1538 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 1539 struct seq_file *seq); 1540 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 1541 struct bpf_link_info *info); 1542 1543 int map_set_for_each_callback_args(struct bpf_verifier_env *env, 1544 struct bpf_func_state *caller, 1545 struct bpf_func_state *callee); 1546 1547 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 1548 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 1549 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 1550 u64 flags); 1551 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 1552 u64 flags); 1553 1554 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 1555 1556 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 1557 void *key, void *value, u64 map_flags); 1558 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1559 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 1560 void *key, void *value, u64 map_flags); 1561 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1562 1563 int bpf_get_file_flag(int flags); 1564 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 1565 size_t actual_size); 1566 1567 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 1568 * forced to use 'long' read/writes to try to atomically copy long counters. 1569 * Best-effort only. No barriers here, since it _will_ race with concurrent 1570 * updates from BPF programs. Called from bpf syscall and mostly used with 1571 * size 8 or 16 bytes, so ask compiler to inline it. 1572 */ 1573 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 1574 { 1575 const long *lsrc = src; 1576 long *ldst = dst; 1577 1578 size /= sizeof(long); 1579 while (size--) 1580 *ldst++ = *lsrc++; 1581 } 1582 1583 /* verify correctness of eBPF program */ 1584 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr); 1585 1586 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 1587 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 1588 #endif 1589 1590 struct btf *bpf_get_btf_vmlinux(void); 1591 1592 /* Map specifics */ 1593 struct xdp_buff; 1594 struct sk_buff; 1595 struct bpf_dtab_netdev; 1596 struct bpf_cpu_map_entry; 1597 1598 void __dev_flush(void); 1599 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1600 struct net_device *dev_rx); 1601 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1602 struct net_device *dev_rx); 1603 int dev_map_enqueue_multi(struct xdp_buff *xdp, struct net_device *dev_rx, 1604 struct bpf_map *map, bool exclude_ingress); 1605 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 1606 struct bpf_prog *xdp_prog); 1607 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1608 struct bpf_prog *xdp_prog, struct bpf_map *map, 1609 bool exclude_ingress); 1610 1611 void __cpu_map_flush(void); 1612 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, 1613 struct net_device *dev_rx); 1614 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1615 struct sk_buff *skb); 1616 1617 /* Return map's numa specified by userspace */ 1618 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 1619 { 1620 return (attr->map_flags & BPF_F_NUMA_NODE) ? 1621 attr->numa_node : NUMA_NO_NODE; 1622 } 1623 1624 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 1625 int array_map_alloc_check(union bpf_attr *attr); 1626 1627 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 1628 union bpf_attr __user *uattr); 1629 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 1630 union bpf_attr __user *uattr); 1631 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1632 const union bpf_attr *kattr, 1633 union bpf_attr __user *uattr); 1634 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1635 const union bpf_attr *kattr, 1636 union bpf_attr __user *uattr); 1637 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 1638 const union bpf_attr *kattr, 1639 union bpf_attr __user *uattr); 1640 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 1641 const union bpf_attr *kattr, 1642 union bpf_attr __user *uattr); 1643 bool bpf_prog_test_check_kfunc_call(u32 kfunc_id, struct module *owner); 1644 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 1645 const struct bpf_prog *prog, 1646 struct bpf_insn_access_aux *info); 1647 int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf, 1648 const struct btf_type *t, int off, int size, 1649 enum bpf_access_type atype, 1650 u32 *next_btf_id); 1651 bool btf_struct_ids_match(struct bpf_verifier_log *log, 1652 const struct btf *btf, u32 id, int off, 1653 const struct btf *need_btf, u32 need_type_id); 1654 1655 int btf_distill_func_proto(struct bpf_verifier_log *log, 1656 struct btf *btf, 1657 const struct btf_type *func_proto, 1658 const char *func_name, 1659 struct btf_func_model *m); 1660 1661 struct bpf_reg_state; 1662 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog, 1663 struct bpf_reg_state *regs); 1664 int btf_check_kfunc_arg_match(struct bpf_verifier_env *env, 1665 const struct btf *btf, u32 func_id, 1666 struct bpf_reg_state *regs); 1667 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, 1668 struct bpf_reg_state *reg); 1669 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 1670 struct btf *btf, const struct btf_type *t); 1671 1672 struct bpf_prog *bpf_prog_by_id(u32 id); 1673 struct bpf_link *bpf_link_by_id(u32 id); 1674 1675 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id); 1676 void bpf_task_storage_free(struct task_struct *task); 1677 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog); 1678 const struct btf_func_model * 1679 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 1680 const struct bpf_insn *insn); 1681 #else /* !CONFIG_BPF_SYSCALL */ 1682 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 1683 { 1684 return ERR_PTR(-EOPNOTSUPP); 1685 } 1686 1687 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 1688 enum bpf_prog_type type, 1689 bool attach_drv) 1690 { 1691 return ERR_PTR(-EOPNOTSUPP); 1692 } 1693 1694 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 1695 { 1696 } 1697 1698 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 1699 { 1700 } 1701 1702 static inline void bpf_prog_put(struct bpf_prog *prog) 1703 { 1704 } 1705 1706 static inline void bpf_prog_inc(struct bpf_prog *prog) 1707 { 1708 } 1709 1710 static inline struct bpf_prog *__must_check 1711 bpf_prog_inc_not_zero(struct bpf_prog *prog) 1712 { 1713 return ERR_PTR(-EOPNOTSUPP); 1714 } 1715 1716 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1717 const struct bpf_link_ops *ops, 1718 struct bpf_prog *prog) 1719 { 1720 } 1721 1722 static inline int bpf_link_prime(struct bpf_link *link, 1723 struct bpf_link_primer *primer) 1724 { 1725 return -EOPNOTSUPP; 1726 } 1727 1728 static inline int bpf_link_settle(struct bpf_link_primer *primer) 1729 { 1730 return -EOPNOTSUPP; 1731 } 1732 1733 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 1734 { 1735 } 1736 1737 static inline void bpf_link_inc(struct bpf_link *link) 1738 { 1739 } 1740 1741 static inline void bpf_link_put(struct bpf_link *link) 1742 { 1743 } 1744 1745 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 1746 { 1747 return -EOPNOTSUPP; 1748 } 1749 1750 static inline bool dev_map_can_have_prog(struct bpf_map *map) 1751 { 1752 return false; 1753 } 1754 1755 static inline void __dev_flush(void) 1756 { 1757 } 1758 1759 struct xdp_buff; 1760 struct bpf_dtab_netdev; 1761 struct bpf_cpu_map_entry; 1762 1763 static inline 1764 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1765 struct net_device *dev_rx) 1766 { 1767 return 0; 1768 } 1769 1770 static inline 1771 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1772 struct net_device *dev_rx) 1773 { 1774 return 0; 1775 } 1776 1777 static inline 1778 int dev_map_enqueue_multi(struct xdp_buff *xdp, struct net_device *dev_rx, 1779 struct bpf_map *map, bool exclude_ingress) 1780 { 1781 return 0; 1782 } 1783 1784 struct sk_buff; 1785 1786 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 1787 struct sk_buff *skb, 1788 struct bpf_prog *xdp_prog) 1789 { 1790 return 0; 1791 } 1792 1793 static inline 1794 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1795 struct bpf_prog *xdp_prog, struct bpf_map *map, 1796 bool exclude_ingress) 1797 { 1798 return 0; 1799 } 1800 1801 static inline void __cpu_map_flush(void) 1802 { 1803 } 1804 1805 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 1806 struct xdp_buff *xdp, 1807 struct net_device *dev_rx) 1808 { 1809 return 0; 1810 } 1811 1812 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1813 struct sk_buff *skb) 1814 { 1815 return -EOPNOTSUPP; 1816 } 1817 1818 static inline bool cpu_map_prog_allowed(struct bpf_map *map) 1819 { 1820 return false; 1821 } 1822 1823 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 1824 enum bpf_prog_type type) 1825 { 1826 return ERR_PTR(-EOPNOTSUPP); 1827 } 1828 1829 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 1830 const union bpf_attr *kattr, 1831 union bpf_attr __user *uattr) 1832 { 1833 return -ENOTSUPP; 1834 } 1835 1836 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 1837 const union bpf_attr *kattr, 1838 union bpf_attr __user *uattr) 1839 { 1840 return -ENOTSUPP; 1841 } 1842 1843 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1844 const union bpf_attr *kattr, 1845 union bpf_attr __user *uattr) 1846 { 1847 return -ENOTSUPP; 1848 } 1849 1850 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1851 const union bpf_attr *kattr, 1852 union bpf_attr __user *uattr) 1853 { 1854 return -ENOTSUPP; 1855 } 1856 1857 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 1858 const union bpf_attr *kattr, 1859 union bpf_attr __user *uattr) 1860 { 1861 return -ENOTSUPP; 1862 } 1863 1864 static inline bool bpf_prog_test_check_kfunc_call(u32 kfunc_id, 1865 struct module *owner) 1866 { 1867 return false; 1868 } 1869 1870 static inline void bpf_map_put(struct bpf_map *map) 1871 { 1872 } 1873 1874 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 1875 { 1876 return ERR_PTR(-ENOTSUPP); 1877 } 1878 1879 static inline const struct bpf_func_proto * 1880 bpf_base_func_proto(enum bpf_func_id func_id) 1881 { 1882 return NULL; 1883 } 1884 1885 static inline void bpf_task_storage_free(struct task_struct *task) 1886 { 1887 } 1888 1889 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog) 1890 { 1891 return false; 1892 } 1893 1894 static inline const struct btf_func_model * 1895 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 1896 const struct bpf_insn *insn) 1897 { 1898 return NULL; 1899 } 1900 #endif /* CONFIG_BPF_SYSCALL */ 1901 1902 void __bpf_free_used_btfs(struct bpf_prog_aux *aux, 1903 struct btf_mod_pair *used_btfs, u32 len); 1904 1905 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 1906 enum bpf_prog_type type) 1907 { 1908 return bpf_prog_get_type_dev(ufd, type, false); 1909 } 1910 1911 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 1912 struct bpf_map **used_maps, u32 len); 1913 1914 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 1915 1916 int bpf_prog_offload_compile(struct bpf_prog *prog); 1917 void bpf_prog_offload_destroy(struct bpf_prog *prog); 1918 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 1919 struct bpf_prog *prog); 1920 1921 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 1922 1923 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 1924 int bpf_map_offload_update_elem(struct bpf_map *map, 1925 void *key, void *value, u64 flags); 1926 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 1927 int bpf_map_offload_get_next_key(struct bpf_map *map, 1928 void *key, void *next_key); 1929 1930 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 1931 1932 struct bpf_offload_dev * 1933 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 1934 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 1935 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 1936 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 1937 struct net_device *netdev); 1938 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 1939 struct net_device *netdev); 1940 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 1941 1942 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 1943 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 1944 1945 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 1946 { 1947 return aux->offload_requested; 1948 } 1949 1950 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1951 { 1952 return unlikely(map->ops == &bpf_map_offload_ops); 1953 } 1954 1955 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 1956 void bpf_map_offload_map_free(struct bpf_map *map); 1957 int bpf_prog_test_run_syscall(struct bpf_prog *prog, 1958 const union bpf_attr *kattr, 1959 union bpf_attr __user *uattr); 1960 1961 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 1962 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 1963 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 1964 void sock_map_unhash(struct sock *sk); 1965 void sock_map_close(struct sock *sk, long timeout); 1966 #else 1967 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 1968 union bpf_attr *attr) 1969 { 1970 return -EOPNOTSUPP; 1971 } 1972 1973 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 1974 { 1975 return false; 1976 } 1977 1978 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1979 { 1980 return false; 1981 } 1982 1983 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 1984 { 1985 return ERR_PTR(-EOPNOTSUPP); 1986 } 1987 1988 static inline void bpf_map_offload_map_free(struct bpf_map *map) 1989 { 1990 } 1991 1992 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 1993 const union bpf_attr *kattr, 1994 union bpf_attr __user *uattr) 1995 { 1996 return -ENOTSUPP; 1997 } 1998 1999 #ifdef CONFIG_BPF_SYSCALL 2000 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 2001 struct bpf_prog *prog) 2002 { 2003 return -EINVAL; 2004 } 2005 2006 static inline int sock_map_prog_detach(const union bpf_attr *attr, 2007 enum bpf_prog_type ptype) 2008 { 2009 return -EOPNOTSUPP; 2010 } 2011 2012 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 2013 u64 flags) 2014 { 2015 return -EOPNOTSUPP; 2016 } 2017 #endif /* CONFIG_BPF_SYSCALL */ 2018 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 2019 2020 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 2021 void bpf_sk_reuseport_detach(struct sock *sk); 2022 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 2023 void *value); 2024 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 2025 void *value, u64 map_flags); 2026 #else 2027 static inline void bpf_sk_reuseport_detach(struct sock *sk) 2028 { 2029 } 2030 2031 #ifdef CONFIG_BPF_SYSCALL 2032 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 2033 void *key, void *value) 2034 { 2035 return -EOPNOTSUPP; 2036 } 2037 2038 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 2039 void *key, void *value, 2040 u64 map_flags) 2041 { 2042 return -EOPNOTSUPP; 2043 } 2044 #endif /* CONFIG_BPF_SYSCALL */ 2045 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 2046 2047 /* verifier prototypes for helper functions called from eBPF programs */ 2048 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 2049 extern const struct bpf_func_proto bpf_map_update_elem_proto; 2050 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 2051 extern const struct bpf_func_proto bpf_map_push_elem_proto; 2052 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 2053 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 2054 2055 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 2056 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 2057 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 2058 extern const struct bpf_func_proto bpf_tail_call_proto; 2059 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 2060 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 2061 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 2062 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 2063 extern const struct bpf_func_proto bpf_get_current_comm_proto; 2064 extern const struct bpf_func_proto bpf_get_stackid_proto; 2065 extern const struct bpf_func_proto bpf_get_stack_proto; 2066 extern const struct bpf_func_proto bpf_get_task_stack_proto; 2067 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 2068 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 2069 extern const struct bpf_func_proto bpf_sock_map_update_proto; 2070 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 2071 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 2072 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 2073 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 2074 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 2075 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 2076 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 2077 extern const struct bpf_func_proto bpf_spin_lock_proto; 2078 extern const struct bpf_func_proto bpf_spin_unlock_proto; 2079 extern const struct bpf_func_proto bpf_get_local_storage_proto; 2080 extern const struct bpf_func_proto bpf_strtol_proto; 2081 extern const struct bpf_func_proto bpf_strtoul_proto; 2082 extern const struct bpf_func_proto bpf_tcp_sock_proto; 2083 extern const struct bpf_func_proto bpf_jiffies64_proto; 2084 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 2085 extern const struct bpf_func_proto bpf_event_output_data_proto; 2086 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 2087 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 2088 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 2089 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 2090 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 2091 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 2092 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 2093 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 2094 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 2095 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 2096 extern const struct bpf_func_proto bpf_copy_from_user_proto; 2097 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 2098 extern const struct bpf_func_proto bpf_snprintf_proto; 2099 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 2100 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 2101 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 2102 extern const struct bpf_func_proto bpf_sock_from_file_proto; 2103 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 2104 extern const struct bpf_func_proto bpf_task_storage_get_proto; 2105 extern const struct bpf_func_proto bpf_task_storage_delete_proto; 2106 extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 2107 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 2108 extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 2109 extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 2110 2111 const struct bpf_func_proto *tracing_prog_func_proto( 2112 enum bpf_func_id func_id, const struct bpf_prog *prog); 2113 2114 /* Shared helpers among cBPF and eBPF. */ 2115 void bpf_user_rnd_init_once(void); 2116 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2117 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2118 2119 #if defined(CONFIG_NET) 2120 bool bpf_sock_common_is_valid_access(int off, int size, 2121 enum bpf_access_type type, 2122 struct bpf_insn_access_aux *info); 2123 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2124 struct bpf_insn_access_aux *info); 2125 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2126 const struct bpf_insn *si, 2127 struct bpf_insn *insn_buf, 2128 struct bpf_prog *prog, 2129 u32 *target_size); 2130 #else 2131 static inline bool bpf_sock_common_is_valid_access(int off, int size, 2132 enum bpf_access_type type, 2133 struct bpf_insn_access_aux *info) 2134 { 2135 return false; 2136 } 2137 static inline bool bpf_sock_is_valid_access(int off, int size, 2138 enum bpf_access_type type, 2139 struct bpf_insn_access_aux *info) 2140 { 2141 return false; 2142 } 2143 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2144 const struct bpf_insn *si, 2145 struct bpf_insn *insn_buf, 2146 struct bpf_prog *prog, 2147 u32 *target_size) 2148 { 2149 return 0; 2150 } 2151 #endif 2152 2153 #ifdef CONFIG_INET 2154 struct sk_reuseport_kern { 2155 struct sk_buff *skb; 2156 struct sock *sk; 2157 struct sock *selected_sk; 2158 struct sock *migrating_sk; 2159 void *data_end; 2160 u32 hash; 2161 u32 reuseport_id; 2162 bool bind_inany; 2163 }; 2164 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2165 struct bpf_insn_access_aux *info); 2166 2167 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2168 const struct bpf_insn *si, 2169 struct bpf_insn *insn_buf, 2170 struct bpf_prog *prog, 2171 u32 *target_size); 2172 2173 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2174 struct bpf_insn_access_aux *info); 2175 2176 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2177 const struct bpf_insn *si, 2178 struct bpf_insn *insn_buf, 2179 struct bpf_prog *prog, 2180 u32 *target_size); 2181 #else 2182 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 2183 enum bpf_access_type type, 2184 struct bpf_insn_access_aux *info) 2185 { 2186 return false; 2187 } 2188 2189 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2190 const struct bpf_insn *si, 2191 struct bpf_insn *insn_buf, 2192 struct bpf_prog *prog, 2193 u32 *target_size) 2194 { 2195 return 0; 2196 } 2197 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 2198 enum bpf_access_type type, 2199 struct bpf_insn_access_aux *info) 2200 { 2201 return false; 2202 } 2203 2204 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2205 const struct bpf_insn *si, 2206 struct bpf_insn *insn_buf, 2207 struct bpf_prog *prog, 2208 u32 *target_size) 2209 { 2210 return 0; 2211 } 2212 #endif /* CONFIG_INET */ 2213 2214 enum bpf_text_poke_type { 2215 BPF_MOD_CALL, 2216 BPF_MOD_JUMP, 2217 }; 2218 2219 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 2220 void *addr1, void *addr2); 2221 2222 struct btf_id_set; 2223 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 2224 2225 #define MAX_BPRINTF_VARARGS 12 2226 2227 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args, 2228 u32 **bin_buf, u32 num_args); 2229 void bpf_bprintf_cleanup(void); 2230 2231 #endif /* _LINUX_BPF_H */ 2232