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