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 u64 map_extra; /* any per-map-type extra fields */ 173 int spin_lock_off; /* >=0 valid offset, <0 error */ 174 int timer_off; /* >=0 valid offset, <0 error */ 175 u32 id; 176 int numa_node; 177 u32 btf_key_type_id; 178 u32 btf_value_type_id; 179 struct btf *btf; 180 #ifdef CONFIG_MEMCG_KMEM 181 struct mem_cgroup *memcg; 182 #endif 183 char name[BPF_OBJ_NAME_LEN]; 184 u32 btf_vmlinux_value_type_id; 185 bool bypass_spec_v1; 186 bool frozen; /* write-once; write-protected by freeze_mutex */ 187 /* 22 bytes hole */ 188 189 /* The 3rd and 4th cacheline with misc members to avoid false sharing 190 * particularly with refcounting. 191 */ 192 atomic64_t refcnt ____cacheline_aligned; 193 atomic64_t usercnt; 194 struct work_struct work; 195 struct mutex freeze_mutex; 196 u64 writecnt; /* writable mmap cnt; protected by freeze_mutex */ 197 }; 198 199 static inline bool map_value_has_spin_lock(const struct bpf_map *map) 200 { 201 return map->spin_lock_off >= 0; 202 } 203 204 static inline bool map_value_has_timer(const struct bpf_map *map) 205 { 206 return map->timer_off >= 0; 207 } 208 209 static inline void check_and_init_map_value(struct bpf_map *map, void *dst) 210 { 211 if (unlikely(map_value_has_spin_lock(map))) 212 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) = 213 (struct bpf_spin_lock){}; 214 if (unlikely(map_value_has_timer(map))) 215 *(struct bpf_timer *)(dst + map->timer_off) = 216 (struct bpf_timer){}; 217 } 218 219 /* copy everything but bpf_spin_lock and bpf_timer. There could be one of each. */ 220 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 221 { 222 u32 s_off = 0, s_sz = 0, t_off = 0, t_sz = 0; 223 224 if (unlikely(map_value_has_spin_lock(map))) { 225 s_off = map->spin_lock_off; 226 s_sz = sizeof(struct bpf_spin_lock); 227 } else if (unlikely(map_value_has_timer(map))) { 228 t_off = map->timer_off; 229 t_sz = sizeof(struct bpf_timer); 230 } 231 232 if (unlikely(s_sz || t_sz)) { 233 if (s_off < t_off || !s_sz) { 234 swap(s_off, t_off); 235 swap(s_sz, t_sz); 236 } 237 memcpy(dst, src, t_off); 238 memcpy(dst + t_off + t_sz, 239 src + t_off + t_sz, 240 s_off - t_off - t_sz); 241 memcpy(dst + s_off + s_sz, 242 src + s_off + s_sz, 243 map->value_size - s_off - s_sz); 244 } else { 245 memcpy(dst, src, map->value_size); 246 } 247 } 248 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 249 bool lock_src); 250 void bpf_timer_cancel_and_free(void *timer); 251 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size); 252 253 struct bpf_offload_dev; 254 struct bpf_offloaded_map; 255 256 struct bpf_map_dev_ops { 257 int (*map_get_next_key)(struct bpf_offloaded_map *map, 258 void *key, void *next_key); 259 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 260 void *key, void *value); 261 int (*map_update_elem)(struct bpf_offloaded_map *map, 262 void *key, void *value, u64 flags); 263 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 264 }; 265 266 struct bpf_offloaded_map { 267 struct bpf_map map; 268 struct net_device *netdev; 269 const struct bpf_map_dev_ops *dev_ops; 270 void *dev_priv; 271 struct list_head offloads; 272 }; 273 274 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 275 { 276 return container_of(map, struct bpf_offloaded_map, map); 277 } 278 279 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 280 { 281 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 282 } 283 284 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 285 { 286 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && 287 map->ops->map_seq_show_elem; 288 } 289 290 int map_check_no_btf(const struct bpf_map *map, 291 const struct btf *btf, 292 const struct btf_type *key_type, 293 const struct btf_type *value_type); 294 295 bool bpf_map_meta_equal(const struct bpf_map *meta0, 296 const struct bpf_map *meta1); 297 298 extern const struct bpf_map_ops bpf_map_offload_ops; 299 300 /* function argument constraints */ 301 enum bpf_arg_type { 302 ARG_DONTCARE = 0, /* unused argument in helper function */ 303 304 /* the following constraints used to prototype 305 * bpf_map_lookup/update/delete_elem() functions 306 */ 307 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 308 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 309 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 310 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */ 311 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */ 312 313 /* the following constraints used to prototype bpf_memcmp() and other 314 * functions that access data on eBPF program stack 315 */ 316 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 317 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */ 318 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized, 319 * helper function must fill all bytes or clear 320 * them in error case. 321 */ 322 323 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 324 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 325 326 ARG_PTR_TO_CTX, /* pointer to context */ 327 ARG_PTR_TO_CTX_OR_NULL, /* pointer to context or NULL */ 328 ARG_ANYTHING, /* any (initialized) argument is ok */ 329 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 330 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 331 ARG_PTR_TO_INT, /* pointer to int */ 332 ARG_PTR_TO_LONG, /* pointer to long */ 333 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 334 ARG_PTR_TO_SOCKET_OR_NULL, /* pointer to bpf_sock (fullsock) or NULL */ 335 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */ 336 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */ 337 ARG_PTR_TO_ALLOC_MEM_OR_NULL, /* pointer to dynamically allocated memory or NULL */ 338 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ 339 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ 340 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ 341 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */ 342 ARG_PTR_TO_STACK_OR_NULL, /* pointer to stack or NULL */ 343 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */ 344 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */ 345 __BPF_ARG_TYPE_MAX, 346 }; 347 348 /* type of values returned from helper functions */ 349 enum bpf_return_type { 350 RET_INTEGER, /* function returns integer */ 351 RET_VOID, /* function doesn't return anything */ 352 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 353 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */ 354 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */ 355 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */ 356 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */ 357 RET_PTR_TO_ALLOC_MEM_OR_NULL, /* returns a pointer to dynamically allocated memory or NULL */ 358 RET_PTR_TO_BTF_ID_OR_NULL, /* returns a pointer to a btf_id or NULL */ 359 RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL, /* returns a pointer to a valid memory or a btf_id or NULL */ 360 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ 361 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */ 362 }; 363 364 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 365 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 366 * instructions after verifying 367 */ 368 struct bpf_func_proto { 369 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 370 bool gpl_only; 371 bool pkt_access; 372 enum bpf_return_type ret_type; 373 union { 374 struct { 375 enum bpf_arg_type arg1_type; 376 enum bpf_arg_type arg2_type; 377 enum bpf_arg_type arg3_type; 378 enum bpf_arg_type arg4_type; 379 enum bpf_arg_type arg5_type; 380 }; 381 enum bpf_arg_type arg_type[5]; 382 }; 383 union { 384 struct { 385 u32 *arg1_btf_id; 386 u32 *arg2_btf_id; 387 u32 *arg3_btf_id; 388 u32 *arg4_btf_id; 389 u32 *arg5_btf_id; 390 }; 391 u32 *arg_btf_id[5]; 392 }; 393 int *ret_btf_id; /* return value btf_id */ 394 bool (*allowed)(const struct bpf_prog *prog); 395 }; 396 397 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 398 * the first argument to eBPF programs. 399 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 400 */ 401 struct bpf_context; 402 403 enum bpf_access_type { 404 BPF_READ = 1, 405 BPF_WRITE = 2 406 }; 407 408 /* types of values stored in eBPF registers */ 409 /* Pointer types represent: 410 * pointer 411 * pointer + imm 412 * pointer + (u16) var 413 * pointer + (u16) var + imm 414 * if (range > 0) then [ptr, ptr + range - off) is safe to access 415 * if (id > 0) means that some 'var' was added 416 * if (off > 0) means that 'imm' was added 417 */ 418 enum bpf_reg_type { 419 NOT_INIT = 0, /* nothing was written into register */ 420 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 421 PTR_TO_CTX, /* reg points to bpf_context */ 422 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 423 PTR_TO_MAP_VALUE, /* reg points to map element value */ 424 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */ 425 PTR_TO_STACK, /* reg == frame_pointer + offset */ 426 PTR_TO_PACKET_META, /* skb->data - meta_len */ 427 PTR_TO_PACKET, /* reg points to skb->data */ 428 PTR_TO_PACKET_END, /* skb->data + headlen */ 429 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 430 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 431 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */ 432 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 433 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */ 434 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 435 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */ 436 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 437 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 438 /* PTR_TO_BTF_ID points to a kernel struct that does not need 439 * to be null checked by the BPF program. This does not imply the 440 * pointer is _not_ null and in practice this can easily be a null 441 * pointer when reading pointer chains. The assumption is program 442 * context will handle null pointer dereference typically via fault 443 * handling. The verifier must keep this in mind and can make no 444 * assumptions about null or non-null when doing branch analysis. 445 * Further, when passed into helpers the helpers can not, without 446 * additional context, assume the value is non-null. 447 */ 448 PTR_TO_BTF_ID, 449 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not 450 * been checked for null. Used primarily to inform the verifier 451 * an explicit null check is required for this struct. 452 */ 453 PTR_TO_BTF_ID_OR_NULL, 454 PTR_TO_MEM, /* reg points to valid memory region */ 455 PTR_TO_MEM_OR_NULL, /* reg points to valid memory region or NULL */ 456 PTR_TO_RDONLY_BUF, /* reg points to a readonly buffer */ 457 PTR_TO_RDONLY_BUF_OR_NULL, /* reg points to a readonly buffer or NULL */ 458 PTR_TO_RDWR_BUF, /* reg points to a read/write buffer */ 459 PTR_TO_RDWR_BUF_OR_NULL, /* reg points to a read/write buffer or NULL */ 460 PTR_TO_PERCPU_BTF_ID, /* reg points to a percpu kernel variable */ 461 PTR_TO_FUNC, /* reg points to a bpf program function */ 462 PTR_TO_MAP_KEY, /* reg points to a map element key */ 463 __BPF_REG_TYPE_MAX, 464 }; 465 466 /* The information passed from prog-specific *_is_valid_access 467 * back to the verifier. 468 */ 469 struct bpf_insn_access_aux { 470 enum bpf_reg_type reg_type; 471 union { 472 int ctx_field_size; 473 struct { 474 struct btf *btf; 475 u32 btf_id; 476 }; 477 }; 478 struct bpf_verifier_log *log; /* for verbose logs */ 479 }; 480 481 static inline void 482 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 483 { 484 aux->ctx_field_size = size; 485 } 486 487 struct bpf_prog_ops { 488 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 489 union bpf_attr __user *uattr); 490 }; 491 492 struct bpf_verifier_ops { 493 /* return eBPF function prototype for verification */ 494 const struct bpf_func_proto * 495 (*get_func_proto)(enum bpf_func_id func_id, 496 const struct bpf_prog *prog); 497 498 /* return true if 'size' wide access at offset 'off' within bpf_context 499 * with 'type' (read or write) is allowed 500 */ 501 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 502 const struct bpf_prog *prog, 503 struct bpf_insn_access_aux *info); 504 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 505 const struct bpf_prog *prog); 506 int (*gen_ld_abs)(const struct bpf_insn *orig, 507 struct bpf_insn *insn_buf); 508 u32 (*convert_ctx_access)(enum bpf_access_type type, 509 const struct bpf_insn *src, 510 struct bpf_insn *dst, 511 struct bpf_prog *prog, u32 *target_size); 512 int (*btf_struct_access)(struct bpf_verifier_log *log, 513 const struct btf *btf, 514 const struct btf_type *t, int off, int size, 515 enum bpf_access_type atype, 516 u32 *next_btf_id); 517 bool (*check_kfunc_call)(u32 kfunc_btf_id, struct module *owner); 518 }; 519 520 struct bpf_prog_offload_ops { 521 /* verifier basic callbacks */ 522 int (*insn_hook)(struct bpf_verifier_env *env, 523 int insn_idx, int prev_insn_idx); 524 int (*finalize)(struct bpf_verifier_env *env); 525 /* verifier optimization callbacks (called after .finalize) */ 526 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 527 struct bpf_insn *insn); 528 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 529 /* program management callbacks */ 530 int (*prepare)(struct bpf_prog *prog); 531 int (*translate)(struct bpf_prog *prog); 532 void (*destroy)(struct bpf_prog *prog); 533 }; 534 535 struct bpf_prog_offload { 536 struct bpf_prog *prog; 537 struct net_device *netdev; 538 struct bpf_offload_dev *offdev; 539 void *dev_priv; 540 struct list_head offloads; 541 bool dev_state; 542 bool opt_failed; 543 void *jited_image; 544 u32 jited_len; 545 }; 546 547 enum bpf_cgroup_storage_type { 548 BPF_CGROUP_STORAGE_SHARED, 549 BPF_CGROUP_STORAGE_PERCPU, 550 __BPF_CGROUP_STORAGE_MAX 551 }; 552 553 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 554 555 /* The longest tracepoint has 12 args. 556 * See include/trace/bpf_probe.h 557 */ 558 #define MAX_BPF_FUNC_ARGS 12 559 560 /* The maximum number of arguments passed through registers 561 * a single function may have. 562 */ 563 #define MAX_BPF_FUNC_REG_ARGS 5 564 565 struct btf_func_model { 566 u8 ret_size; 567 u8 nr_args; 568 u8 arg_size[MAX_BPF_FUNC_ARGS]; 569 }; 570 571 /* Restore arguments before returning from trampoline to let original function 572 * continue executing. This flag is used for fentry progs when there are no 573 * fexit progs. 574 */ 575 #define BPF_TRAMP_F_RESTORE_REGS BIT(0) 576 /* Call original function after fentry progs, but before fexit progs. 577 * Makes sense for fentry/fexit, normal calls and indirect calls. 578 */ 579 #define BPF_TRAMP_F_CALL_ORIG BIT(1) 580 /* Skip current frame and return to parent. Makes sense for fentry/fexit 581 * programs only. Should not be used with normal calls and indirect calls. 582 */ 583 #define BPF_TRAMP_F_SKIP_FRAME BIT(2) 584 /* Store IP address of the caller on the trampoline stack, 585 * so it's available for trampoline's programs. 586 */ 587 #define BPF_TRAMP_F_IP_ARG BIT(3) 588 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */ 589 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4) 590 591 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 592 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2 593 */ 594 #define BPF_MAX_TRAMP_PROGS 38 595 596 struct bpf_tramp_progs { 597 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS]; 598 int nr_progs; 599 }; 600 601 /* Different use cases for BPF trampoline: 602 * 1. replace nop at the function entry (kprobe equivalent) 603 * flags = BPF_TRAMP_F_RESTORE_REGS 604 * fentry = a set of programs to run before returning from trampoline 605 * 606 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 607 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 608 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 609 * fentry = a set of program to run before calling original function 610 * fexit = a set of program to run after original function 611 * 612 * 3. replace direct call instruction anywhere in the function body 613 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 614 * With flags = 0 615 * fentry = a set of programs to run before returning from trampoline 616 * With flags = BPF_TRAMP_F_CALL_ORIG 617 * orig_call = original callback addr or direct function addr 618 * fentry = a set of program to run before calling original function 619 * fexit = a set of program to run after original function 620 */ 621 struct bpf_tramp_image; 622 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end, 623 const struct btf_func_model *m, u32 flags, 624 struct bpf_tramp_progs *tprogs, 625 void *orig_call); 626 /* these two functions are called from generated trampoline */ 627 u64 notrace __bpf_prog_enter(struct bpf_prog *prog); 628 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start); 629 u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog); 630 void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start); 631 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr); 632 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr); 633 634 struct bpf_ksym { 635 unsigned long start; 636 unsigned long end; 637 char name[KSYM_NAME_LEN]; 638 struct list_head lnode; 639 struct latch_tree_node tnode; 640 bool prog; 641 }; 642 643 enum bpf_tramp_prog_type { 644 BPF_TRAMP_FENTRY, 645 BPF_TRAMP_FEXIT, 646 BPF_TRAMP_MODIFY_RETURN, 647 BPF_TRAMP_MAX, 648 BPF_TRAMP_REPLACE, /* more than MAX */ 649 }; 650 651 struct bpf_tramp_image { 652 void *image; 653 struct bpf_ksym ksym; 654 struct percpu_ref pcref; 655 void *ip_after_call; 656 void *ip_epilogue; 657 union { 658 struct rcu_head rcu; 659 struct work_struct work; 660 }; 661 }; 662 663 struct bpf_trampoline { 664 /* hlist for trampoline_table */ 665 struct hlist_node hlist; 666 /* serializes access to fields of this trampoline */ 667 struct mutex mutex; 668 refcount_t refcnt; 669 u64 key; 670 struct { 671 struct btf_func_model model; 672 void *addr; 673 bool ftrace_managed; 674 } func; 675 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 676 * program by replacing one of its functions. func.addr is the address 677 * of the function it replaced. 678 */ 679 struct bpf_prog *extension_prog; 680 /* list of BPF programs using this trampoline */ 681 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 682 /* Number of attached programs. A counter per kind. */ 683 int progs_cnt[BPF_TRAMP_MAX]; 684 /* Executable image of trampoline */ 685 struct bpf_tramp_image *cur_image; 686 u64 selector; 687 struct module *mod; 688 }; 689 690 struct bpf_attach_target_info { 691 struct btf_func_model fmodel; 692 long tgt_addr; 693 const char *tgt_name; 694 const struct btf_type *tgt_type; 695 }; 696 697 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 698 699 struct bpf_dispatcher_prog { 700 struct bpf_prog *prog; 701 refcount_t users; 702 }; 703 704 struct bpf_dispatcher { 705 /* dispatcher mutex */ 706 struct mutex mutex; 707 void *func; 708 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 709 int num_progs; 710 void *image; 711 u32 image_off; 712 struct bpf_ksym ksym; 713 }; 714 715 static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func( 716 const void *ctx, 717 const struct bpf_insn *insnsi, 718 unsigned int (*bpf_func)(const void *, 719 const struct bpf_insn *)) 720 { 721 return bpf_func(ctx, insnsi); 722 } 723 #ifdef CONFIG_BPF_JIT 724 int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 725 int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 726 struct bpf_trampoline *bpf_trampoline_get(u64 key, 727 struct bpf_attach_target_info *tgt_info); 728 void bpf_trampoline_put(struct bpf_trampoline *tr); 729 #define BPF_DISPATCHER_INIT(_name) { \ 730 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 731 .func = &_name##_func, \ 732 .progs = {}, \ 733 .num_progs = 0, \ 734 .image = NULL, \ 735 .image_off = 0, \ 736 .ksym = { \ 737 .name = #_name, \ 738 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 739 }, \ 740 } 741 742 #define DEFINE_BPF_DISPATCHER(name) \ 743 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \ 744 const void *ctx, \ 745 const struct bpf_insn *insnsi, \ 746 unsigned int (*bpf_func)(const void *, \ 747 const struct bpf_insn *)) \ 748 { \ 749 return bpf_func(ctx, insnsi); \ 750 } \ 751 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 752 struct bpf_dispatcher bpf_dispatcher_##name = \ 753 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 754 #define DECLARE_BPF_DISPATCHER(name) \ 755 unsigned int bpf_dispatcher_##name##_func( \ 756 const void *ctx, \ 757 const struct bpf_insn *insnsi, \ 758 unsigned int (*bpf_func)(const void *, \ 759 const struct bpf_insn *)); \ 760 extern struct bpf_dispatcher bpf_dispatcher_##name; 761 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 762 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 763 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 764 struct bpf_prog *to); 765 /* Called only from JIT-enabled code, so there's no need for stubs. */ 766 void *bpf_jit_alloc_exec_page(void); 767 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym); 768 void bpf_image_ksym_del(struct bpf_ksym *ksym); 769 void bpf_ksym_add(struct bpf_ksym *ksym); 770 void bpf_ksym_del(struct bpf_ksym *ksym); 771 int bpf_jit_charge_modmem(u32 pages); 772 void bpf_jit_uncharge_modmem(u32 pages); 773 #else 774 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog, 775 struct bpf_trampoline *tr) 776 { 777 return -ENOTSUPP; 778 } 779 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog, 780 struct bpf_trampoline *tr) 781 { 782 return -ENOTSUPP; 783 } 784 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 785 struct bpf_attach_target_info *tgt_info) 786 { 787 return ERR_PTR(-EOPNOTSUPP); 788 } 789 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 790 #define DEFINE_BPF_DISPATCHER(name) 791 #define DECLARE_BPF_DISPATCHER(name) 792 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 793 #define BPF_DISPATCHER_PTR(name) NULL 794 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 795 struct bpf_prog *from, 796 struct bpf_prog *to) {} 797 static inline bool is_bpf_image_address(unsigned long address) 798 { 799 return false; 800 } 801 #endif 802 803 struct bpf_func_info_aux { 804 u16 linkage; 805 bool unreliable; 806 }; 807 808 enum bpf_jit_poke_reason { 809 BPF_POKE_REASON_TAIL_CALL, 810 }; 811 812 /* Descriptor of pokes pointing /into/ the JITed image. */ 813 struct bpf_jit_poke_descriptor { 814 void *tailcall_target; 815 void *tailcall_bypass; 816 void *bypass_addr; 817 void *aux; 818 union { 819 struct { 820 struct bpf_map *map; 821 u32 key; 822 } tail_call; 823 }; 824 bool tailcall_target_stable; 825 u8 adj_off; 826 u16 reason; 827 u32 insn_idx; 828 }; 829 830 /* reg_type info for ctx arguments */ 831 struct bpf_ctx_arg_aux { 832 u32 offset; 833 enum bpf_reg_type reg_type; 834 u32 btf_id; 835 }; 836 837 struct btf_mod_pair { 838 struct btf *btf; 839 struct module *module; 840 }; 841 842 struct bpf_kfunc_desc_tab; 843 844 struct bpf_prog_aux { 845 atomic64_t refcnt; 846 u32 used_map_cnt; 847 u32 used_btf_cnt; 848 u32 max_ctx_offset; 849 u32 max_pkt_offset; 850 u32 max_tp_access; 851 u32 stack_depth; 852 u32 id; 853 u32 func_cnt; /* used by non-func prog as the number of func progs */ 854 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 855 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 856 u32 ctx_arg_info_size; 857 u32 max_rdonly_access; 858 u32 max_rdwr_access; 859 struct btf *attach_btf; 860 const struct bpf_ctx_arg_aux *ctx_arg_info; 861 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 862 struct bpf_prog *dst_prog; 863 struct bpf_trampoline *dst_trampoline; 864 enum bpf_prog_type saved_dst_prog_type; 865 enum bpf_attach_type saved_dst_attach_type; 866 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 867 bool offload_requested; 868 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 869 bool func_proto_unreliable; 870 bool sleepable; 871 bool tail_call_reachable; 872 struct hlist_node tramp_hlist; 873 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 874 const struct btf_type *attach_func_proto; 875 /* function name for valid attach_btf_id */ 876 const char *attach_func_name; 877 struct bpf_prog **func; 878 void *jit_data; /* JIT specific data. arch dependent */ 879 struct bpf_jit_poke_descriptor *poke_tab; 880 struct bpf_kfunc_desc_tab *kfunc_tab; 881 struct bpf_kfunc_btf_tab *kfunc_btf_tab; 882 u32 size_poke_tab; 883 struct bpf_ksym ksym; 884 const struct bpf_prog_ops *ops; 885 struct bpf_map **used_maps; 886 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 887 struct btf_mod_pair *used_btfs; 888 struct bpf_prog *prog; 889 struct user_struct *user; 890 u64 load_time; /* ns since boottime */ 891 u32 verified_insns; 892 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 893 char name[BPF_OBJ_NAME_LEN]; 894 #ifdef CONFIG_SECURITY 895 void *security; 896 #endif 897 struct bpf_prog_offload *offload; 898 struct btf *btf; 899 struct bpf_func_info *func_info; 900 struct bpf_func_info_aux *func_info_aux; 901 /* bpf_line_info loaded from userspace. linfo->insn_off 902 * has the xlated insn offset. 903 * Both the main and sub prog share the same linfo. 904 * The subprog can access its first linfo by 905 * using the linfo_idx. 906 */ 907 struct bpf_line_info *linfo; 908 /* jited_linfo is the jited addr of the linfo. It has a 909 * one to one mapping to linfo: 910 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 911 * Both the main and sub prog share the same jited_linfo. 912 * The subprog can access its first jited_linfo by 913 * using the linfo_idx. 914 */ 915 void **jited_linfo; 916 u32 func_info_cnt; 917 u32 nr_linfo; 918 /* subprog can use linfo_idx to access its first linfo and 919 * jited_linfo. 920 * main prog always has linfo_idx == 0 921 */ 922 u32 linfo_idx; 923 u32 num_exentries; 924 struct exception_table_entry *extable; 925 union { 926 struct work_struct work; 927 struct rcu_head rcu; 928 }; 929 }; 930 931 struct bpf_array_aux { 932 /* 'Ownership' of prog array is claimed by the first program that 933 * is going to use this map or by the first program which FD is 934 * stored in the map to make sure that all callers and callees have 935 * the same prog type and JITed flag. 936 */ 937 enum bpf_prog_type type; 938 bool jited; 939 /* Programs with direct jumps into programs part of this array. */ 940 struct list_head poke_progs; 941 struct bpf_map *map; 942 struct mutex poke_mutex; 943 struct work_struct work; 944 }; 945 946 struct bpf_link { 947 atomic64_t refcnt; 948 u32 id; 949 enum bpf_link_type type; 950 const struct bpf_link_ops *ops; 951 struct bpf_prog *prog; 952 struct work_struct work; 953 }; 954 955 struct bpf_link_ops { 956 void (*release)(struct bpf_link *link); 957 void (*dealloc)(struct bpf_link *link); 958 int (*detach)(struct bpf_link *link); 959 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 960 struct bpf_prog *old_prog); 961 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 962 int (*fill_link_info)(const struct bpf_link *link, 963 struct bpf_link_info *info); 964 }; 965 966 struct bpf_link_primer { 967 struct bpf_link *link; 968 struct file *file; 969 int fd; 970 u32 id; 971 }; 972 973 struct bpf_struct_ops_value; 974 struct btf_member; 975 976 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 977 struct bpf_struct_ops { 978 const struct bpf_verifier_ops *verifier_ops; 979 int (*init)(struct btf *btf); 980 int (*check_member)(const struct btf_type *t, 981 const struct btf_member *member); 982 int (*init_member)(const struct btf_type *t, 983 const struct btf_member *member, 984 void *kdata, const void *udata); 985 int (*reg)(void *kdata); 986 void (*unreg)(void *kdata); 987 const struct btf_type *type; 988 const struct btf_type *value_type; 989 const char *name; 990 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 991 u32 type_id; 992 u32 value_id; 993 }; 994 995 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 996 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 997 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id); 998 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log); 999 bool bpf_struct_ops_get(const void *kdata); 1000 void bpf_struct_ops_put(const void *kdata); 1001 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 1002 void *value); 1003 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1004 { 1005 if (owner == BPF_MODULE_OWNER) 1006 return bpf_struct_ops_get(data); 1007 else 1008 return try_module_get(owner); 1009 } 1010 static inline void bpf_module_put(const void *data, struct module *owner) 1011 { 1012 if (owner == BPF_MODULE_OWNER) 1013 bpf_struct_ops_put(data); 1014 else 1015 module_put(owner); 1016 } 1017 #else 1018 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id) 1019 { 1020 return NULL; 1021 } 1022 static inline void bpf_struct_ops_init(struct btf *btf, 1023 struct bpf_verifier_log *log) 1024 { 1025 } 1026 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1027 { 1028 return try_module_get(owner); 1029 } 1030 static inline void bpf_module_put(const void *data, struct module *owner) 1031 { 1032 module_put(owner); 1033 } 1034 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 1035 void *key, 1036 void *value) 1037 { 1038 return -EINVAL; 1039 } 1040 #endif 1041 1042 struct bpf_array { 1043 struct bpf_map map; 1044 u32 elem_size; 1045 u32 index_mask; 1046 struct bpf_array_aux *aux; 1047 union { 1048 char value[0] __aligned(8); 1049 void *ptrs[0] __aligned(8); 1050 void __percpu *pptrs[0] __aligned(8); 1051 }; 1052 }; 1053 1054 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 1055 #define MAX_TAIL_CALL_CNT 32 1056 1057 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 1058 BPF_F_RDONLY_PROG | \ 1059 BPF_F_WRONLY | \ 1060 BPF_F_WRONLY_PROG) 1061 1062 #define BPF_MAP_CAN_READ BIT(0) 1063 #define BPF_MAP_CAN_WRITE BIT(1) 1064 1065 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 1066 { 1067 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1068 1069 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 1070 * not possible. 1071 */ 1072 if (access_flags & BPF_F_RDONLY_PROG) 1073 return BPF_MAP_CAN_READ; 1074 else if (access_flags & BPF_F_WRONLY_PROG) 1075 return BPF_MAP_CAN_WRITE; 1076 else 1077 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 1078 } 1079 1080 static inline bool bpf_map_flags_access_ok(u32 access_flags) 1081 { 1082 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 1083 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1084 } 1085 1086 struct bpf_event_entry { 1087 struct perf_event *event; 1088 struct file *perf_file; 1089 struct file *map_file; 1090 struct rcu_head rcu; 1091 }; 1092 1093 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); 1094 int bpf_prog_calc_tag(struct bpf_prog *fp); 1095 1096 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 1097 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void); 1098 1099 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 1100 unsigned long off, unsigned long len); 1101 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 1102 const struct bpf_insn *src, 1103 struct bpf_insn *dst, 1104 struct bpf_prog *prog, 1105 u32 *target_size); 1106 1107 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 1108 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 1109 1110 /* an array of programs to be executed under rcu_lock. 1111 * 1112 * Typical usage: 1113 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, bpf_prog_run); 1114 * 1115 * the structure returned by bpf_prog_array_alloc() should be populated 1116 * with program pointers and the last pointer must be NULL. 1117 * The user has to keep refcnt on the program and make sure the program 1118 * is removed from the array before bpf_prog_put(). 1119 * The 'struct bpf_prog_array *' should only be replaced with xchg() 1120 * since other cpus are walking the array of pointers in parallel. 1121 */ 1122 struct bpf_prog_array_item { 1123 struct bpf_prog *prog; 1124 union { 1125 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1126 u64 bpf_cookie; 1127 }; 1128 }; 1129 1130 struct bpf_prog_array { 1131 struct rcu_head rcu; 1132 struct bpf_prog_array_item items[]; 1133 }; 1134 1135 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 1136 void bpf_prog_array_free(struct bpf_prog_array *progs); 1137 int bpf_prog_array_length(struct bpf_prog_array *progs); 1138 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 1139 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 1140 __u32 __user *prog_ids, u32 cnt); 1141 1142 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 1143 struct bpf_prog *old_prog); 1144 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 1145 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 1146 struct bpf_prog *prog); 1147 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 1148 u32 *prog_ids, u32 request_cnt, 1149 u32 *prog_cnt); 1150 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 1151 struct bpf_prog *exclude_prog, 1152 struct bpf_prog *include_prog, 1153 u64 bpf_cookie, 1154 struct bpf_prog_array **new_array); 1155 1156 struct bpf_run_ctx {}; 1157 1158 struct bpf_cg_run_ctx { 1159 struct bpf_run_ctx run_ctx; 1160 const struct bpf_prog_array_item *prog_item; 1161 }; 1162 1163 struct bpf_trace_run_ctx { 1164 struct bpf_run_ctx run_ctx; 1165 u64 bpf_cookie; 1166 }; 1167 1168 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 1169 { 1170 struct bpf_run_ctx *old_ctx = NULL; 1171 1172 #ifdef CONFIG_BPF_SYSCALL 1173 old_ctx = current->bpf_ctx; 1174 current->bpf_ctx = new_ctx; 1175 #endif 1176 return old_ctx; 1177 } 1178 1179 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 1180 { 1181 #ifdef CONFIG_BPF_SYSCALL 1182 current->bpf_ctx = old_ctx; 1183 #endif 1184 } 1185 1186 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 1187 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 1188 /* BPF program asks to set CN on the packet. */ 1189 #define BPF_RET_SET_CN (1 << 0) 1190 1191 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 1192 1193 static __always_inline u32 1194 BPF_PROG_RUN_ARRAY_CG_FLAGS(const struct bpf_prog_array __rcu *array_rcu, 1195 const void *ctx, bpf_prog_run_fn run_prog, 1196 u32 *ret_flags) 1197 { 1198 const struct bpf_prog_array_item *item; 1199 const struct bpf_prog *prog; 1200 const struct bpf_prog_array *array; 1201 struct bpf_run_ctx *old_run_ctx; 1202 struct bpf_cg_run_ctx run_ctx; 1203 u32 ret = 1; 1204 u32 func_ret; 1205 1206 migrate_disable(); 1207 rcu_read_lock(); 1208 array = rcu_dereference(array_rcu); 1209 item = &array->items[0]; 1210 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1211 while ((prog = READ_ONCE(item->prog))) { 1212 run_ctx.prog_item = item; 1213 func_ret = run_prog(prog, ctx); 1214 ret &= (func_ret & 1); 1215 *(ret_flags) |= (func_ret >> 1); 1216 item++; 1217 } 1218 bpf_reset_run_ctx(old_run_ctx); 1219 rcu_read_unlock(); 1220 migrate_enable(); 1221 return ret; 1222 } 1223 1224 static __always_inline u32 1225 BPF_PROG_RUN_ARRAY_CG(const struct bpf_prog_array __rcu *array_rcu, 1226 const void *ctx, bpf_prog_run_fn run_prog) 1227 { 1228 const struct bpf_prog_array_item *item; 1229 const struct bpf_prog *prog; 1230 const struct bpf_prog_array *array; 1231 struct bpf_run_ctx *old_run_ctx; 1232 struct bpf_cg_run_ctx run_ctx; 1233 u32 ret = 1; 1234 1235 migrate_disable(); 1236 rcu_read_lock(); 1237 array = rcu_dereference(array_rcu); 1238 item = &array->items[0]; 1239 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1240 while ((prog = READ_ONCE(item->prog))) { 1241 run_ctx.prog_item = item; 1242 ret &= run_prog(prog, ctx); 1243 item++; 1244 } 1245 bpf_reset_run_ctx(old_run_ctx); 1246 rcu_read_unlock(); 1247 migrate_enable(); 1248 return ret; 1249 } 1250 1251 static __always_inline u32 1252 BPF_PROG_RUN_ARRAY(const struct bpf_prog_array __rcu *array_rcu, 1253 const void *ctx, bpf_prog_run_fn run_prog) 1254 { 1255 const struct bpf_prog_array_item *item; 1256 const struct bpf_prog *prog; 1257 const struct bpf_prog_array *array; 1258 struct bpf_run_ctx *old_run_ctx; 1259 struct bpf_trace_run_ctx run_ctx; 1260 u32 ret = 1; 1261 1262 migrate_disable(); 1263 rcu_read_lock(); 1264 array = rcu_dereference(array_rcu); 1265 if (unlikely(!array)) 1266 goto out; 1267 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1268 item = &array->items[0]; 1269 while ((prog = READ_ONCE(item->prog))) { 1270 run_ctx.bpf_cookie = item->bpf_cookie; 1271 ret &= run_prog(prog, ctx); 1272 item++; 1273 } 1274 bpf_reset_run_ctx(old_run_ctx); 1275 out: 1276 rcu_read_unlock(); 1277 migrate_enable(); 1278 return ret; 1279 } 1280 1281 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs 1282 * so BPF programs can request cwr for TCP packets. 1283 * 1284 * Current cgroup skb programs can only return 0 or 1 (0 to drop the 1285 * packet. This macro changes the behavior so the low order bit 1286 * indicates whether the packet should be dropped (0) or not (1) 1287 * and the next bit is a congestion notification bit. This could be 1288 * used by TCP to call tcp_enter_cwr() 1289 * 1290 * Hence, new allowed return values of CGROUP EGRESS BPF programs are: 1291 * 0: drop packet 1292 * 1: keep packet 1293 * 2: drop packet and cn 1294 * 3: keep packet and cn 1295 * 1296 * This macro then converts it to one of the NET_XMIT or an error 1297 * code that is then interpreted as drop packet (and no cn): 1298 * 0: NET_XMIT_SUCCESS skb should be transmitted 1299 * 1: NET_XMIT_DROP skb should be dropped and cn 1300 * 2: NET_XMIT_CN skb should be transmitted and cn 1301 * 3: -EPERM skb should be dropped 1302 */ 1303 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \ 1304 ({ \ 1305 u32 _flags = 0; \ 1306 bool _cn; \ 1307 u32 _ret; \ 1308 _ret = BPF_PROG_RUN_ARRAY_CG_FLAGS(array, ctx, func, &_flags); \ 1309 _cn = _flags & BPF_RET_SET_CN; \ 1310 if (_ret) \ 1311 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \ 1312 else \ 1313 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \ 1314 _ret; \ 1315 }) 1316 1317 #ifdef CONFIG_BPF_SYSCALL 1318 DECLARE_PER_CPU(int, bpf_prog_active); 1319 extern struct mutex bpf_stats_enabled_mutex; 1320 1321 /* 1322 * Block execution of BPF programs attached to instrumentation (perf, 1323 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 1324 * these events can happen inside a region which holds a map bucket lock 1325 * and can deadlock on it. 1326 * 1327 * Use the preemption safe inc/dec variants on RT because migrate disable 1328 * is preemptible on RT and preemption in the middle of the RMW operation 1329 * might lead to inconsistent state. Use the raw variants for non RT 1330 * kernels as migrate_disable() maps to preempt_disable() so the slightly 1331 * more expensive save operation can be avoided. 1332 */ 1333 static inline void bpf_disable_instrumentation(void) 1334 { 1335 migrate_disable(); 1336 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1337 this_cpu_inc(bpf_prog_active); 1338 else 1339 __this_cpu_inc(bpf_prog_active); 1340 } 1341 1342 static inline void bpf_enable_instrumentation(void) 1343 { 1344 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1345 this_cpu_dec(bpf_prog_active); 1346 else 1347 __this_cpu_dec(bpf_prog_active); 1348 migrate_enable(); 1349 } 1350 1351 extern const struct file_operations bpf_map_fops; 1352 extern const struct file_operations bpf_prog_fops; 1353 extern const struct file_operations bpf_iter_fops; 1354 1355 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 1356 extern const struct bpf_prog_ops _name ## _prog_ops; \ 1357 extern const struct bpf_verifier_ops _name ## _verifier_ops; 1358 #define BPF_MAP_TYPE(_id, _ops) \ 1359 extern const struct bpf_map_ops _ops; 1360 #define BPF_LINK_TYPE(_id, _name) 1361 #include <linux/bpf_types.h> 1362 #undef BPF_PROG_TYPE 1363 #undef BPF_MAP_TYPE 1364 #undef BPF_LINK_TYPE 1365 1366 extern const struct bpf_prog_ops bpf_offload_prog_ops; 1367 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 1368 extern const struct bpf_verifier_ops xdp_analyzer_ops; 1369 1370 struct bpf_prog *bpf_prog_get(u32 ufd); 1371 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 1372 bool attach_drv); 1373 void bpf_prog_add(struct bpf_prog *prog, int i); 1374 void bpf_prog_sub(struct bpf_prog *prog, int i); 1375 void bpf_prog_inc(struct bpf_prog *prog); 1376 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 1377 void bpf_prog_put(struct bpf_prog *prog); 1378 1379 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 1380 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 1381 1382 struct bpf_map *bpf_map_get(u32 ufd); 1383 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 1384 struct bpf_map *__bpf_map_get(struct fd f); 1385 void bpf_map_inc(struct bpf_map *map); 1386 void bpf_map_inc_with_uref(struct bpf_map *map); 1387 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 1388 void bpf_map_put_with_uref(struct bpf_map *map); 1389 void bpf_map_put(struct bpf_map *map); 1390 void *bpf_map_area_alloc(u64 size, int numa_node); 1391 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 1392 void bpf_map_area_free(void *base); 1393 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 1394 int generic_map_lookup_batch(struct bpf_map *map, 1395 const union bpf_attr *attr, 1396 union bpf_attr __user *uattr); 1397 int generic_map_update_batch(struct bpf_map *map, 1398 const union bpf_attr *attr, 1399 union bpf_attr __user *uattr); 1400 int generic_map_delete_batch(struct bpf_map *map, 1401 const union bpf_attr *attr, 1402 union bpf_attr __user *uattr); 1403 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 1404 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 1405 1406 #ifdef CONFIG_MEMCG_KMEM 1407 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1408 int node); 1409 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 1410 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 1411 size_t align, gfp_t flags); 1412 #else 1413 static inline void * 1414 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1415 int node) 1416 { 1417 return kmalloc_node(size, flags, node); 1418 } 1419 1420 static inline void * 1421 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags) 1422 { 1423 return kzalloc(size, flags); 1424 } 1425 1426 static inline void __percpu * 1427 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align, 1428 gfp_t flags) 1429 { 1430 return __alloc_percpu_gfp(size, align, flags); 1431 } 1432 #endif 1433 1434 extern int sysctl_unprivileged_bpf_disabled; 1435 1436 static inline bool bpf_allow_ptr_leaks(void) 1437 { 1438 return perfmon_capable(); 1439 } 1440 1441 static inline bool bpf_allow_uninit_stack(void) 1442 { 1443 return perfmon_capable(); 1444 } 1445 1446 static inline bool bpf_allow_ptr_to_map_access(void) 1447 { 1448 return perfmon_capable(); 1449 } 1450 1451 static inline bool bpf_bypass_spec_v1(void) 1452 { 1453 return perfmon_capable(); 1454 } 1455 1456 static inline bool bpf_bypass_spec_v4(void) 1457 { 1458 return perfmon_capable(); 1459 } 1460 1461 int bpf_map_new_fd(struct bpf_map *map, int flags); 1462 int bpf_prog_new_fd(struct bpf_prog *prog); 1463 1464 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1465 const struct bpf_link_ops *ops, struct bpf_prog *prog); 1466 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 1467 int bpf_link_settle(struct bpf_link_primer *primer); 1468 void bpf_link_cleanup(struct bpf_link_primer *primer); 1469 void bpf_link_inc(struct bpf_link *link); 1470 void bpf_link_put(struct bpf_link *link); 1471 int bpf_link_new_fd(struct bpf_link *link); 1472 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd); 1473 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 1474 1475 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 1476 int bpf_obj_get_user(const char __user *pathname, int flags); 1477 1478 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 1479 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 1480 extern int bpf_iter_ ## target(args); \ 1481 int __init bpf_iter_ ## target(args) { return 0; } 1482 1483 struct bpf_iter_aux_info { 1484 struct bpf_map *map; 1485 }; 1486 1487 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 1488 union bpf_iter_link_info *linfo, 1489 struct bpf_iter_aux_info *aux); 1490 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 1491 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 1492 struct seq_file *seq); 1493 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 1494 struct bpf_link_info *info); 1495 typedef const struct bpf_func_proto * 1496 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 1497 const struct bpf_prog *prog); 1498 1499 enum bpf_iter_feature { 1500 BPF_ITER_RESCHED = BIT(0), 1501 }; 1502 1503 #define BPF_ITER_CTX_ARG_MAX 2 1504 struct bpf_iter_reg { 1505 const char *target; 1506 bpf_iter_attach_target_t attach_target; 1507 bpf_iter_detach_target_t detach_target; 1508 bpf_iter_show_fdinfo_t show_fdinfo; 1509 bpf_iter_fill_link_info_t fill_link_info; 1510 bpf_iter_get_func_proto_t get_func_proto; 1511 u32 ctx_arg_info_size; 1512 u32 feature; 1513 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 1514 const struct bpf_iter_seq_info *seq_info; 1515 }; 1516 1517 struct bpf_iter_meta { 1518 __bpf_md_ptr(struct seq_file *, seq); 1519 u64 session_id; 1520 u64 seq_num; 1521 }; 1522 1523 struct bpf_iter__bpf_map_elem { 1524 __bpf_md_ptr(struct bpf_iter_meta *, meta); 1525 __bpf_md_ptr(struct bpf_map *, map); 1526 __bpf_md_ptr(void *, key); 1527 __bpf_md_ptr(void *, value); 1528 }; 1529 1530 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 1531 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 1532 bool bpf_iter_prog_supported(struct bpf_prog *prog); 1533 const struct bpf_func_proto * 1534 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 1535 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 1536 int bpf_iter_new_fd(struct bpf_link *link); 1537 bool bpf_link_is_iter(struct bpf_link *link); 1538 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 1539 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 1540 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 1541 struct seq_file *seq); 1542 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 1543 struct bpf_link_info *info); 1544 1545 int map_set_for_each_callback_args(struct bpf_verifier_env *env, 1546 struct bpf_func_state *caller, 1547 struct bpf_func_state *callee); 1548 1549 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 1550 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 1551 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 1552 u64 flags); 1553 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 1554 u64 flags); 1555 1556 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 1557 1558 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 1559 void *key, void *value, u64 map_flags); 1560 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1561 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 1562 void *key, void *value, u64 map_flags); 1563 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1564 1565 int bpf_get_file_flag(int flags); 1566 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 1567 size_t actual_size); 1568 1569 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 1570 * forced to use 'long' read/writes to try to atomically copy long counters. 1571 * Best-effort only. No barriers here, since it _will_ race with concurrent 1572 * updates from BPF programs. Called from bpf syscall and mostly used with 1573 * size 8 or 16 bytes, so ask compiler to inline it. 1574 */ 1575 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 1576 { 1577 const long *lsrc = src; 1578 long *ldst = dst; 1579 1580 size /= sizeof(long); 1581 while (size--) 1582 *ldst++ = *lsrc++; 1583 } 1584 1585 /* verify correctness of eBPF program */ 1586 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr); 1587 1588 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 1589 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 1590 #endif 1591 1592 struct btf *bpf_get_btf_vmlinux(void); 1593 1594 /* Map specifics */ 1595 struct xdp_buff; 1596 struct sk_buff; 1597 struct bpf_dtab_netdev; 1598 struct bpf_cpu_map_entry; 1599 1600 void __dev_flush(void); 1601 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1602 struct net_device *dev_rx); 1603 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1604 struct net_device *dev_rx); 1605 int dev_map_enqueue_multi(struct xdp_buff *xdp, struct net_device *dev_rx, 1606 struct bpf_map *map, bool exclude_ingress); 1607 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 1608 struct bpf_prog *xdp_prog); 1609 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1610 struct bpf_prog *xdp_prog, struct bpf_map *map, 1611 bool exclude_ingress); 1612 1613 void __cpu_map_flush(void); 1614 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, 1615 struct net_device *dev_rx); 1616 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1617 struct sk_buff *skb); 1618 1619 /* Return map's numa specified by userspace */ 1620 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 1621 { 1622 return (attr->map_flags & BPF_F_NUMA_NODE) ? 1623 attr->numa_node : NUMA_NO_NODE; 1624 } 1625 1626 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 1627 int array_map_alloc_check(union bpf_attr *attr); 1628 1629 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 1630 union bpf_attr __user *uattr); 1631 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 1632 union bpf_attr __user *uattr); 1633 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1634 const union bpf_attr *kattr, 1635 union bpf_attr __user *uattr); 1636 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1637 const union bpf_attr *kattr, 1638 union bpf_attr __user *uattr); 1639 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 1640 const union bpf_attr *kattr, 1641 union bpf_attr __user *uattr); 1642 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 1643 const union bpf_attr *kattr, 1644 union bpf_attr __user *uattr); 1645 bool bpf_prog_test_check_kfunc_call(u32 kfunc_id, struct module *owner); 1646 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 1647 const struct bpf_prog *prog, 1648 struct bpf_insn_access_aux *info); 1649 int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf, 1650 const struct btf_type *t, int off, int size, 1651 enum bpf_access_type atype, 1652 u32 *next_btf_id); 1653 bool btf_struct_ids_match(struct bpf_verifier_log *log, 1654 const struct btf *btf, u32 id, int off, 1655 const struct btf *need_btf, u32 need_type_id); 1656 1657 int btf_distill_func_proto(struct bpf_verifier_log *log, 1658 struct btf *btf, 1659 const struct btf_type *func_proto, 1660 const char *func_name, 1661 struct btf_func_model *m); 1662 1663 struct bpf_reg_state; 1664 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog, 1665 struct bpf_reg_state *regs); 1666 int btf_check_kfunc_arg_match(struct bpf_verifier_env *env, 1667 const struct btf *btf, u32 func_id, 1668 struct bpf_reg_state *regs); 1669 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, 1670 struct bpf_reg_state *reg); 1671 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 1672 struct btf *btf, const struct btf_type *t); 1673 1674 struct bpf_prog *bpf_prog_by_id(u32 id); 1675 struct bpf_link *bpf_link_by_id(u32 id); 1676 1677 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id); 1678 void bpf_task_storage_free(struct task_struct *task); 1679 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog); 1680 const struct btf_func_model * 1681 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 1682 const struct bpf_insn *insn); 1683 #else /* !CONFIG_BPF_SYSCALL */ 1684 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 1685 { 1686 return ERR_PTR(-EOPNOTSUPP); 1687 } 1688 1689 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 1690 enum bpf_prog_type type, 1691 bool attach_drv) 1692 { 1693 return ERR_PTR(-EOPNOTSUPP); 1694 } 1695 1696 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 1697 { 1698 } 1699 1700 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 1701 { 1702 } 1703 1704 static inline void bpf_prog_put(struct bpf_prog *prog) 1705 { 1706 } 1707 1708 static inline void bpf_prog_inc(struct bpf_prog *prog) 1709 { 1710 } 1711 1712 static inline struct bpf_prog *__must_check 1713 bpf_prog_inc_not_zero(struct bpf_prog *prog) 1714 { 1715 return ERR_PTR(-EOPNOTSUPP); 1716 } 1717 1718 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1719 const struct bpf_link_ops *ops, 1720 struct bpf_prog *prog) 1721 { 1722 } 1723 1724 static inline int bpf_link_prime(struct bpf_link *link, 1725 struct bpf_link_primer *primer) 1726 { 1727 return -EOPNOTSUPP; 1728 } 1729 1730 static inline int bpf_link_settle(struct bpf_link_primer *primer) 1731 { 1732 return -EOPNOTSUPP; 1733 } 1734 1735 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 1736 { 1737 } 1738 1739 static inline void bpf_link_inc(struct bpf_link *link) 1740 { 1741 } 1742 1743 static inline void bpf_link_put(struct bpf_link *link) 1744 { 1745 } 1746 1747 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 1748 { 1749 return -EOPNOTSUPP; 1750 } 1751 1752 static inline bool dev_map_can_have_prog(struct bpf_map *map) 1753 { 1754 return false; 1755 } 1756 1757 static inline void __dev_flush(void) 1758 { 1759 } 1760 1761 struct xdp_buff; 1762 struct bpf_dtab_netdev; 1763 struct bpf_cpu_map_entry; 1764 1765 static inline 1766 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1767 struct net_device *dev_rx) 1768 { 1769 return 0; 1770 } 1771 1772 static inline 1773 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1774 struct net_device *dev_rx) 1775 { 1776 return 0; 1777 } 1778 1779 static inline 1780 int dev_map_enqueue_multi(struct xdp_buff *xdp, struct net_device *dev_rx, 1781 struct bpf_map *map, bool exclude_ingress) 1782 { 1783 return 0; 1784 } 1785 1786 struct sk_buff; 1787 1788 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 1789 struct sk_buff *skb, 1790 struct bpf_prog *xdp_prog) 1791 { 1792 return 0; 1793 } 1794 1795 static inline 1796 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1797 struct bpf_prog *xdp_prog, struct bpf_map *map, 1798 bool exclude_ingress) 1799 { 1800 return 0; 1801 } 1802 1803 static inline void __cpu_map_flush(void) 1804 { 1805 } 1806 1807 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 1808 struct xdp_buff *xdp, 1809 struct net_device *dev_rx) 1810 { 1811 return 0; 1812 } 1813 1814 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1815 struct sk_buff *skb) 1816 { 1817 return -EOPNOTSUPP; 1818 } 1819 1820 static inline bool cpu_map_prog_allowed(struct bpf_map *map) 1821 { 1822 return false; 1823 } 1824 1825 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 1826 enum bpf_prog_type type) 1827 { 1828 return ERR_PTR(-EOPNOTSUPP); 1829 } 1830 1831 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 1832 const union bpf_attr *kattr, 1833 union bpf_attr __user *uattr) 1834 { 1835 return -ENOTSUPP; 1836 } 1837 1838 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 1839 const union bpf_attr *kattr, 1840 union bpf_attr __user *uattr) 1841 { 1842 return -ENOTSUPP; 1843 } 1844 1845 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1846 const union bpf_attr *kattr, 1847 union bpf_attr __user *uattr) 1848 { 1849 return -ENOTSUPP; 1850 } 1851 1852 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1853 const union bpf_attr *kattr, 1854 union bpf_attr __user *uattr) 1855 { 1856 return -ENOTSUPP; 1857 } 1858 1859 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 1860 const union bpf_attr *kattr, 1861 union bpf_attr __user *uattr) 1862 { 1863 return -ENOTSUPP; 1864 } 1865 1866 static inline bool bpf_prog_test_check_kfunc_call(u32 kfunc_id, 1867 struct module *owner) 1868 { 1869 return false; 1870 } 1871 1872 static inline void bpf_map_put(struct bpf_map *map) 1873 { 1874 } 1875 1876 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 1877 { 1878 return ERR_PTR(-ENOTSUPP); 1879 } 1880 1881 static inline const struct bpf_func_proto * 1882 bpf_base_func_proto(enum bpf_func_id func_id) 1883 { 1884 return NULL; 1885 } 1886 1887 static inline void bpf_task_storage_free(struct task_struct *task) 1888 { 1889 } 1890 1891 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog) 1892 { 1893 return false; 1894 } 1895 1896 static inline const struct btf_func_model * 1897 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 1898 const struct bpf_insn *insn) 1899 { 1900 return NULL; 1901 } 1902 #endif /* CONFIG_BPF_SYSCALL */ 1903 1904 void __bpf_free_used_btfs(struct bpf_prog_aux *aux, 1905 struct btf_mod_pair *used_btfs, u32 len); 1906 1907 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 1908 enum bpf_prog_type type) 1909 { 1910 return bpf_prog_get_type_dev(ufd, type, false); 1911 } 1912 1913 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 1914 struct bpf_map **used_maps, u32 len); 1915 1916 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 1917 1918 int bpf_prog_offload_compile(struct bpf_prog *prog); 1919 void bpf_prog_offload_destroy(struct bpf_prog *prog); 1920 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 1921 struct bpf_prog *prog); 1922 1923 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 1924 1925 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 1926 int bpf_map_offload_update_elem(struct bpf_map *map, 1927 void *key, void *value, u64 flags); 1928 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 1929 int bpf_map_offload_get_next_key(struct bpf_map *map, 1930 void *key, void *next_key); 1931 1932 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 1933 1934 struct bpf_offload_dev * 1935 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 1936 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 1937 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 1938 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 1939 struct net_device *netdev); 1940 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 1941 struct net_device *netdev); 1942 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 1943 1944 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 1945 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 1946 1947 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 1948 { 1949 return aux->offload_requested; 1950 } 1951 1952 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1953 { 1954 return unlikely(map->ops == &bpf_map_offload_ops); 1955 } 1956 1957 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 1958 void bpf_map_offload_map_free(struct bpf_map *map); 1959 int bpf_prog_test_run_syscall(struct bpf_prog *prog, 1960 const union bpf_attr *kattr, 1961 union bpf_attr __user *uattr); 1962 1963 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 1964 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 1965 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 1966 void sock_map_unhash(struct sock *sk); 1967 void sock_map_close(struct sock *sk, long timeout); 1968 #else 1969 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 1970 union bpf_attr *attr) 1971 { 1972 return -EOPNOTSUPP; 1973 } 1974 1975 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 1976 { 1977 return false; 1978 } 1979 1980 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1981 { 1982 return false; 1983 } 1984 1985 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 1986 { 1987 return ERR_PTR(-EOPNOTSUPP); 1988 } 1989 1990 static inline void bpf_map_offload_map_free(struct bpf_map *map) 1991 { 1992 } 1993 1994 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 1995 const union bpf_attr *kattr, 1996 union bpf_attr __user *uattr) 1997 { 1998 return -ENOTSUPP; 1999 } 2000 2001 #ifdef CONFIG_BPF_SYSCALL 2002 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 2003 struct bpf_prog *prog) 2004 { 2005 return -EINVAL; 2006 } 2007 2008 static inline int sock_map_prog_detach(const union bpf_attr *attr, 2009 enum bpf_prog_type ptype) 2010 { 2011 return -EOPNOTSUPP; 2012 } 2013 2014 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 2015 u64 flags) 2016 { 2017 return -EOPNOTSUPP; 2018 } 2019 #endif /* CONFIG_BPF_SYSCALL */ 2020 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 2021 2022 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 2023 void bpf_sk_reuseport_detach(struct sock *sk); 2024 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 2025 void *value); 2026 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 2027 void *value, u64 map_flags); 2028 #else 2029 static inline void bpf_sk_reuseport_detach(struct sock *sk) 2030 { 2031 } 2032 2033 #ifdef CONFIG_BPF_SYSCALL 2034 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 2035 void *key, void *value) 2036 { 2037 return -EOPNOTSUPP; 2038 } 2039 2040 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 2041 void *key, void *value, 2042 u64 map_flags) 2043 { 2044 return -EOPNOTSUPP; 2045 } 2046 #endif /* CONFIG_BPF_SYSCALL */ 2047 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 2048 2049 /* verifier prototypes for helper functions called from eBPF programs */ 2050 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 2051 extern const struct bpf_func_proto bpf_map_update_elem_proto; 2052 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 2053 extern const struct bpf_func_proto bpf_map_push_elem_proto; 2054 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 2055 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 2056 2057 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 2058 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 2059 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 2060 extern const struct bpf_func_proto bpf_tail_call_proto; 2061 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 2062 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 2063 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 2064 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 2065 extern const struct bpf_func_proto bpf_get_current_comm_proto; 2066 extern const struct bpf_func_proto bpf_get_stackid_proto; 2067 extern const struct bpf_func_proto bpf_get_stack_proto; 2068 extern const struct bpf_func_proto bpf_get_task_stack_proto; 2069 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 2070 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 2071 extern const struct bpf_func_proto bpf_sock_map_update_proto; 2072 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 2073 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 2074 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 2075 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 2076 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 2077 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 2078 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 2079 extern const struct bpf_func_proto bpf_spin_lock_proto; 2080 extern const struct bpf_func_proto bpf_spin_unlock_proto; 2081 extern const struct bpf_func_proto bpf_get_local_storage_proto; 2082 extern const struct bpf_func_proto bpf_strtol_proto; 2083 extern const struct bpf_func_proto bpf_strtoul_proto; 2084 extern const struct bpf_func_proto bpf_tcp_sock_proto; 2085 extern const struct bpf_func_proto bpf_jiffies64_proto; 2086 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 2087 extern const struct bpf_func_proto bpf_event_output_data_proto; 2088 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 2089 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 2090 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 2091 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 2092 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 2093 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 2094 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 2095 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 2096 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 2097 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 2098 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto; 2099 extern const struct bpf_func_proto bpf_copy_from_user_proto; 2100 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 2101 extern const struct bpf_func_proto bpf_snprintf_proto; 2102 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 2103 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 2104 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 2105 extern const struct bpf_func_proto bpf_sock_from_file_proto; 2106 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 2107 extern const struct bpf_func_proto bpf_task_storage_get_proto; 2108 extern const struct bpf_func_proto bpf_task_storage_delete_proto; 2109 extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 2110 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 2111 extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 2112 extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 2113 extern const struct bpf_func_proto bpf_kallsyms_lookup_name_proto; 2114 2115 const struct bpf_func_proto *tracing_prog_func_proto( 2116 enum bpf_func_id func_id, const struct bpf_prog *prog); 2117 2118 /* Shared helpers among cBPF and eBPF. */ 2119 void bpf_user_rnd_init_once(void); 2120 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2121 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2122 2123 #if defined(CONFIG_NET) 2124 bool bpf_sock_common_is_valid_access(int off, int size, 2125 enum bpf_access_type type, 2126 struct bpf_insn_access_aux *info); 2127 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2128 struct bpf_insn_access_aux *info); 2129 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2130 const struct bpf_insn *si, 2131 struct bpf_insn *insn_buf, 2132 struct bpf_prog *prog, 2133 u32 *target_size); 2134 #else 2135 static inline bool bpf_sock_common_is_valid_access(int off, int size, 2136 enum bpf_access_type type, 2137 struct bpf_insn_access_aux *info) 2138 { 2139 return false; 2140 } 2141 static inline bool bpf_sock_is_valid_access(int off, int size, 2142 enum bpf_access_type type, 2143 struct bpf_insn_access_aux *info) 2144 { 2145 return false; 2146 } 2147 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2148 const struct bpf_insn *si, 2149 struct bpf_insn *insn_buf, 2150 struct bpf_prog *prog, 2151 u32 *target_size) 2152 { 2153 return 0; 2154 } 2155 #endif 2156 2157 #ifdef CONFIG_INET 2158 struct sk_reuseport_kern { 2159 struct sk_buff *skb; 2160 struct sock *sk; 2161 struct sock *selected_sk; 2162 struct sock *migrating_sk; 2163 void *data_end; 2164 u32 hash; 2165 u32 reuseport_id; 2166 bool bind_inany; 2167 }; 2168 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2169 struct bpf_insn_access_aux *info); 2170 2171 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2172 const struct bpf_insn *si, 2173 struct bpf_insn *insn_buf, 2174 struct bpf_prog *prog, 2175 u32 *target_size); 2176 2177 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2178 struct bpf_insn_access_aux *info); 2179 2180 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2181 const struct bpf_insn *si, 2182 struct bpf_insn *insn_buf, 2183 struct bpf_prog *prog, 2184 u32 *target_size); 2185 #else 2186 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 2187 enum bpf_access_type type, 2188 struct bpf_insn_access_aux *info) 2189 { 2190 return false; 2191 } 2192 2193 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2194 const struct bpf_insn *si, 2195 struct bpf_insn *insn_buf, 2196 struct bpf_prog *prog, 2197 u32 *target_size) 2198 { 2199 return 0; 2200 } 2201 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 2202 enum bpf_access_type type, 2203 struct bpf_insn_access_aux *info) 2204 { 2205 return false; 2206 } 2207 2208 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2209 const struct bpf_insn *si, 2210 struct bpf_insn *insn_buf, 2211 struct bpf_prog *prog, 2212 u32 *target_size) 2213 { 2214 return 0; 2215 } 2216 #endif /* CONFIG_INET */ 2217 2218 enum bpf_text_poke_type { 2219 BPF_MOD_CALL, 2220 BPF_MOD_JUMP, 2221 }; 2222 2223 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 2224 void *addr1, void *addr2); 2225 2226 struct btf_id_set; 2227 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 2228 2229 #define MAX_BPRINTF_VARARGS 12 2230 2231 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args, 2232 u32 **bin_buf, u32 num_args); 2233 void bpf_bprintf_cleanup(void); 2234 2235 #endif /* _LINUX_BPF_H */ 2236