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 /* Store IP address of the caller on the trampoline stack, 582 * so it's available for trampoline's programs. 583 */ 584 #define BPF_TRAMP_F_IP_ARG BIT(3) 585 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */ 586 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4) 587 588 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 589 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2 590 */ 591 #define BPF_MAX_TRAMP_PROGS 38 592 593 struct bpf_tramp_progs { 594 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS]; 595 int nr_progs; 596 }; 597 598 /* Different use cases for BPF trampoline: 599 * 1. replace nop at the function entry (kprobe equivalent) 600 * flags = BPF_TRAMP_F_RESTORE_REGS 601 * fentry = a set of programs to run before returning from trampoline 602 * 603 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 604 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 605 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 606 * fentry = a set of program to run before calling original function 607 * fexit = a set of program to run after original function 608 * 609 * 3. replace direct call instruction anywhere in the function body 610 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 611 * With flags = 0 612 * fentry = a set of programs to run before returning from trampoline 613 * With flags = BPF_TRAMP_F_CALL_ORIG 614 * orig_call = original callback addr or direct function addr 615 * fentry = a set of program to run before calling original function 616 * fexit = a set of program to run after original function 617 */ 618 struct bpf_tramp_image; 619 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end, 620 const struct btf_func_model *m, u32 flags, 621 struct bpf_tramp_progs *tprogs, 622 void *orig_call); 623 /* these two functions are called from generated trampoline */ 624 u64 notrace __bpf_prog_enter(struct bpf_prog *prog); 625 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start); 626 u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog); 627 void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start); 628 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr); 629 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr); 630 631 struct bpf_ksym { 632 unsigned long start; 633 unsigned long end; 634 char name[KSYM_NAME_LEN]; 635 struct list_head lnode; 636 struct latch_tree_node tnode; 637 bool prog; 638 }; 639 640 enum bpf_tramp_prog_type { 641 BPF_TRAMP_FENTRY, 642 BPF_TRAMP_FEXIT, 643 BPF_TRAMP_MODIFY_RETURN, 644 BPF_TRAMP_MAX, 645 BPF_TRAMP_REPLACE, /* more than MAX */ 646 }; 647 648 struct bpf_tramp_image { 649 void *image; 650 struct bpf_ksym ksym; 651 struct percpu_ref pcref; 652 void *ip_after_call; 653 void *ip_epilogue; 654 union { 655 struct rcu_head rcu; 656 struct work_struct work; 657 }; 658 }; 659 660 struct bpf_trampoline { 661 /* hlist for trampoline_table */ 662 struct hlist_node hlist; 663 /* serializes access to fields of this trampoline */ 664 struct mutex mutex; 665 refcount_t refcnt; 666 u64 key; 667 struct { 668 struct btf_func_model model; 669 void *addr; 670 bool ftrace_managed; 671 } func; 672 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 673 * program by replacing one of its functions. func.addr is the address 674 * of the function it replaced. 675 */ 676 struct bpf_prog *extension_prog; 677 /* list of BPF programs using this trampoline */ 678 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 679 /* Number of attached programs. A counter per kind. */ 680 int progs_cnt[BPF_TRAMP_MAX]; 681 /* Executable image of trampoline */ 682 struct bpf_tramp_image *cur_image; 683 u64 selector; 684 struct module *mod; 685 }; 686 687 struct bpf_attach_target_info { 688 struct btf_func_model fmodel; 689 long tgt_addr; 690 const char *tgt_name; 691 const struct btf_type *tgt_type; 692 }; 693 694 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 695 696 struct bpf_dispatcher_prog { 697 struct bpf_prog *prog; 698 refcount_t users; 699 }; 700 701 struct bpf_dispatcher { 702 /* dispatcher mutex */ 703 struct mutex mutex; 704 void *func; 705 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 706 int num_progs; 707 void *image; 708 u32 image_off; 709 struct bpf_ksym ksym; 710 }; 711 712 static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func( 713 const void *ctx, 714 const struct bpf_insn *insnsi, 715 unsigned int (*bpf_func)(const void *, 716 const struct bpf_insn *)) 717 { 718 return bpf_func(ctx, insnsi); 719 } 720 #ifdef CONFIG_BPF_JIT 721 int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 722 int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 723 struct bpf_trampoline *bpf_trampoline_get(u64 key, 724 struct bpf_attach_target_info *tgt_info); 725 void bpf_trampoline_put(struct bpf_trampoline *tr); 726 #define BPF_DISPATCHER_INIT(_name) { \ 727 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 728 .func = &_name##_func, \ 729 .progs = {}, \ 730 .num_progs = 0, \ 731 .image = NULL, \ 732 .image_off = 0, \ 733 .ksym = { \ 734 .name = #_name, \ 735 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 736 }, \ 737 } 738 739 #define DEFINE_BPF_DISPATCHER(name) \ 740 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \ 741 const void *ctx, \ 742 const struct bpf_insn *insnsi, \ 743 unsigned int (*bpf_func)(const void *, \ 744 const struct bpf_insn *)) \ 745 { \ 746 return bpf_func(ctx, insnsi); \ 747 } \ 748 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 749 struct bpf_dispatcher bpf_dispatcher_##name = \ 750 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 751 #define DECLARE_BPF_DISPATCHER(name) \ 752 unsigned int bpf_dispatcher_##name##_func( \ 753 const void *ctx, \ 754 const struct bpf_insn *insnsi, \ 755 unsigned int (*bpf_func)(const void *, \ 756 const struct bpf_insn *)); \ 757 extern struct bpf_dispatcher bpf_dispatcher_##name; 758 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 759 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 760 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 761 struct bpf_prog *to); 762 /* Called only from JIT-enabled code, so there's no need for stubs. */ 763 void *bpf_jit_alloc_exec_page(void); 764 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym); 765 void bpf_image_ksym_del(struct bpf_ksym *ksym); 766 void bpf_ksym_add(struct bpf_ksym *ksym); 767 void bpf_ksym_del(struct bpf_ksym *ksym); 768 int bpf_jit_charge_modmem(u32 pages); 769 void bpf_jit_uncharge_modmem(u32 pages); 770 #else 771 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog, 772 struct bpf_trampoline *tr) 773 { 774 return -ENOTSUPP; 775 } 776 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog, 777 struct bpf_trampoline *tr) 778 { 779 return -ENOTSUPP; 780 } 781 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 782 struct bpf_attach_target_info *tgt_info) 783 { 784 return ERR_PTR(-EOPNOTSUPP); 785 } 786 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 787 #define DEFINE_BPF_DISPATCHER(name) 788 #define DECLARE_BPF_DISPATCHER(name) 789 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 790 #define BPF_DISPATCHER_PTR(name) NULL 791 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 792 struct bpf_prog *from, 793 struct bpf_prog *to) {} 794 static inline bool is_bpf_image_address(unsigned long address) 795 { 796 return false; 797 } 798 #endif 799 800 struct bpf_func_info_aux { 801 u16 linkage; 802 bool unreliable; 803 }; 804 805 enum bpf_jit_poke_reason { 806 BPF_POKE_REASON_TAIL_CALL, 807 }; 808 809 /* Descriptor of pokes pointing /into/ the JITed image. */ 810 struct bpf_jit_poke_descriptor { 811 void *tailcall_target; 812 void *tailcall_bypass; 813 void *bypass_addr; 814 void *aux; 815 union { 816 struct { 817 struct bpf_map *map; 818 u32 key; 819 } tail_call; 820 }; 821 bool tailcall_target_stable; 822 u8 adj_off; 823 u16 reason; 824 u32 insn_idx; 825 }; 826 827 /* reg_type info for ctx arguments */ 828 struct bpf_ctx_arg_aux { 829 u32 offset; 830 enum bpf_reg_type reg_type; 831 u32 btf_id; 832 }; 833 834 struct btf_mod_pair { 835 struct btf *btf; 836 struct module *module; 837 }; 838 839 struct bpf_kfunc_desc_tab; 840 841 struct bpf_prog_aux { 842 atomic64_t refcnt; 843 u32 used_map_cnt; 844 u32 used_btf_cnt; 845 u32 max_ctx_offset; 846 u32 max_pkt_offset; 847 u32 max_tp_access; 848 u32 stack_depth; 849 u32 id; 850 u32 func_cnt; /* used by non-func prog as the number of func progs */ 851 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 852 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 853 u32 ctx_arg_info_size; 854 u32 max_rdonly_access; 855 u32 max_rdwr_access; 856 struct btf *attach_btf; 857 const struct bpf_ctx_arg_aux *ctx_arg_info; 858 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 859 struct bpf_prog *dst_prog; 860 struct bpf_trampoline *dst_trampoline; 861 enum bpf_prog_type saved_dst_prog_type; 862 enum bpf_attach_type saved_dst_attach_type; 863 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 864 bool offload_requested; 865 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 866 bool func_proto_unreliable; 867 bool sleepable; 868 bool tail_call_reachable; 869 struct hlist_node tramp_hlist; 870 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 871 const struct btf_type *attach_func_proto; 872 /* function name for valid attach_btf_id */ 873 const char *attach_func_name; 874 struct bpf_prog **func; 875 void *jit_data; /* JIT specific data. arch dependent */ 876 struct bpf_jit_poke_descriptor *poke_tab; 877 struct bpf_kfunc_desc_tab *kfunc_tab; 878 u32 size_poke_tab; 879 struct bpf_ksym ksym; 880 const struct bpf_prog_ops *ops; 881 struct bpf_map **used_maps; 882 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 883 struct btf_mod_pair *used_btfs; 884 struct bpf_prog *prog; 885 struct user_struct *user; 886 u64 load_time; /* ns since boottime */ 887 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 888 char name[BPF_OBJ_NAME_LEN]; 889 #ifdef CONFIG_SECURITY 890 void *security; 891 #endif 892 struct bpf_prog_offload *offload; 893 struct btf *btf; 894 struct bpf_func_info *func_info; 895 struct bpf_func_info_aux *func_info_aux; 896 /* bpf_line_info loaded from userspace. linfo->insn_off 897 * has the xlated insn offset. 898 * Both the main and sub prog share the same linfo. 899 * The subprog can access its first linfo by 900 * using the linfo_idx. 901 */ 902 struct bpf_line_info *linfo; 903 /* jited_linfo is the jited addr of the linfo. It has a 904 * one to one mapping to linfo: 905 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 906 * Both the main and sub prog share the same jited_linfo. 907 * The subprog can access its first jited_linfo by 908 * using the linfo_idx. 909 */ 910 void **jited_linfo; 911 u32 func_info_cnt; 912 u32 nr_linfo; 913 /* subprog can use linfo_idx to access its first linfo and 914 * jited_linfo. 915 * main prog always has linfo_idx == 0 916 */ 917 u32 linfo_idx; 918 u32 num_exentries; 919 struct exception_table_entry *extable; 920 union { 921 struct work_struct work; 922 struct rcu_head rcu; 923 }; 924 }; 925 926 struct bpf_array_aux { 927 /* 'Ownership' of prog array is claimed by the first program that 928 * is going to use this map or by the first program which FD is 929 * stored in the map to make sure that all callers and callees have 930 * the same prog type and JITed flag. 931 */ 932 enum bpf_prog_type type; 933 bool jited; 934 /* Programs with direct jumps into programs part of this array. */ 935 struct list_head poke_progs; 936 struct bpf_map *map; 937 struct mutex poke_mutex; 938 struct work_struct work; 939 }; 940 941 struct bpf_link { 942 atomic64_t refcnt; 943 u32 id; 944 enum bpf_link_type type; 945 const struct bpf_link_ops *ops; 946 struct bpf_prog *prog; 947 struct work_struct work; 948 }; 949 950 struct bpf_link_ops { 951 void (*release)(struct bpf_link *link); 952 void (*dealloc)(struct bpf_link *link); 953 int (*detach)(struct bpf_link *link); 954 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 955 struct bpf_prog *old_prog); 956 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 957 int (*fill_link_info)(const struct bpf_link *link, 958 struct bpf_link_info *info); 959 }; 960 961 struct bpf_link_primer { 962 struct bpf_link *link; 963 struct file *file; 964 int fd; 965 u32 id; 966 }; 967 968 struct bpf_struct_ops_value; 969 struct btf_member; 970 971 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 972 struct bpf_struct_ops { 973 const struct bpf_verifier_ops *verifier_ops; 974 int (*init)(struct btf *btf); 975 int (*check_member)(const struct btf_type *t, 976 const struct btf_member *member); 977 int (*init_member)(const struct btf_type *t, 978 const struct btf_member *member, 979 void *kdata, const void *udata); 980 int (*reg)(void *kdata); 981 void (*unreg)(void *kdata); 982 const struct btf_type *type; 983 const struct btf_type *value_type; 984 const char *name; 985 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 986 u32 type_id; 987 u32 value_id; 988 }; 989 990 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 991 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 992 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id); 993 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log); 994 bool bpf_struct_ops_get(const void *kdata); 995 void bpf_struct_ops_put(const void *kdata); 996 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 997 void *value); 998 static inline bool bpf_try_module_get(const void *data, struct module *owner) 999 { 1000 if (owner == BPF_MODULE_OWNER) 1001 return bpf_struct_ops_get(data); 1002 else 1003 return try_module_get(owner); 1004 } 1005 static inline void bpf_module_put(const void *data, struct module *owner) 1006 { 1007 if (owner == BPF_MODULE_OWNER) 1008 bpf_struct_ops_put(data); 1009 else 1010 module_put(owner); 1011 } 1012 #else 1013 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id) 1014 { 1015 return NULL; 1016 } 1017 static inline void bpf_struct_ops_init(struct btf *btf, 1018 struct bpf_verifier_log *log) 1019 { 1020 } 1021 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1022 { 1023 return try_module_get(owner); 1024 } 1025 static inline void bpf_module_put(const void *data, struct module *owner) 1026 { 1027 module_put(owner); 1028 } 1029 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 1030 void *key, 1031 void *value) 1032 { 1033 return -EINVAL; 1034 } 1035 #endif 1036 1037 struct bpf_array { 1038 struct bpf_map map; 1039 u32 elem_size; 1040 u32 index_mask; 1041 struct bpf_array_aux *aux; 1042 union { 1043 char value[0] __aligned(8); 1044 void *ptrs[0] __aligned(8); 1045 void __percpu *pptrs[0] __aligned(8); 1046 }; 1047 }; 1048 1049 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 1050 #define MAX_TAIL_CALL_CNT 32 1051 1052 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 1053 BPF_F_RDONLY_PROG | \ 1054 BPF_F_WRONLY | \ 1055 BPF_F_WRONLY_PROG) 1056 1057 #define BPF_MAP_CAN_READ BIT(0) 1058 #define BPF_MAP_CAN_WRITE BIT(1) 1059 1060 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 1061 { 1062 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1063 1064 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 1065 * not possible. 1066 */ 1067 if (access_flags & BPF_F_RDONLY_PROG) 1068 return BPF_MAP_CAN_READ; 1069 else if (access_flags & BPF_F_WRONLY_PROG) 1070 return BPF_MAP_CAN_WRITE; 1071 else 1072 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 1073 } 1074 1075 static inline bool bpf_map_flags_access_ok(u32 access_flags) 1076 { 1077 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 1078 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1079 } 1080 1081 struct bpf_event_entry { 1082 struct perf_event *event; 1083 struct file *perf_file; 1084 struct file *map_file; 1085 struct rcu_head rcu; 1086 }; 1087 1088 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); 1089 int bpf_prog_calc_tag(struct bpf_prog *fp); 1090 1091 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 1092 1093 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 1094 unsigned long off, unsigned long len); 1095 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 1096 const struct bpf_insn *src, 1097 struct bpf_insn *dst, 1098 struct bpf_prog *prog, 1099 u32 *target_size); 1100 1101 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 1102 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 1103 1104 /* an array of programs to be executed under rcu_lock. 1105 * 1106 * Typical usage: 1107 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, bpf_prog_run); 1108 * 1109 * the structure returned by bpf_prog_array_alloc() should be populated 1110 * with program pointers and the last pointer must be NULL. 1111 * The user has to keep refcnt on the program and make sure the program 1112 * is removed from the array before bpf_prog_put(). 1113 * The 'struct bpf_prog_array *' should only be replaced with xchg() 1114 * since other cpus are walking the array of pointers in parallel. 1115 */ 1116 struct bpf_prog_array_item { 1117 struct bpf_prog *prog; 1118 union { 1119 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1120 u64 bpf_cookie; 1121 }; 1122 }; 1123 1124 struct bpf_prog_array { 1125 struct rcu_head rcu; 1126 struct bpf_prog_array_item items[]; 1127 }; 1128 1129 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 1130 void bpf_prog_array_free(struct bpf_prog_array *progs); 1131 int bpf_prog_array_length(struct bpf_prog_array *progs); 1132 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 1133 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 1134 __u32 __user *prog_ids, u32 cnt); 1135 1136 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 1137 struct bpf_prog *old_prog); 1138 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 1139 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 1140 struct bpf_prog *prog); 1141 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 1142 u32 *prog_ids, u32 request_cnt, 1143 u32 *prog_cnt); 1144 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 1145 struct bpf_prog *exclude_prog, 1146 struct bpf_prog *include_prog, 1147 u64 bpf_cookie, 1148 struct bpf_prog_array **new_array); 1149 1150 struct bpf_run_ctx {}; 1151 1152 struct bpf_cg_run_ctx { 1153 struct bpf_run_ctx run_ctx; 1154 const struct bpf_prog_array_item *prog_item; 1155 }; 1156 1157 struct bpf_trace_run_ctx { 1158 struct bpf_run_ctx run_ctx; 1159 u64 bpf_cookie; 1160 }; 1161 1162 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 1163 { 1164 struct bpf_run_ctx *old_ctx = NULL; 1165 1166 #ifdef CONFIG_BPF_SYSCALL 1167 old_ctx = current->bpf_ctx; 1168 current->bpf_ctx = new_ctx; 1169 #endif 1170 return old_ctx; 1171 } 1172 1173 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 1174 { 1175 #ifdef CONFIG_BPF_SYSCALL 1176 current->bpf_ctx = old_ctx; 1177 #endif 1178 } 1179 1180 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 1181 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 1182 /* BPF program asks to set CN on the packet. */ 1183 #define BPF_RET_SET_CN (1 << 0) 1184 1185 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 1186 1187 static __always_inline u32 1188 BPF_PROG_RUN_ARRAY_CG_FLAGS(const struct bpf_prog_array __rcu *array_rcu, 1189 const void *ctx, bpf_prog_run_fn run_prog, 1190 u32 *ret_flags) 1191 { 1192 const struct bpf_prog_array_item *item; 1193 const struct bpf_prog *prog; 1194 const struct bpf_prog_array *array; 1195 struct bpf_run_ctx *old_run_ctx; 1196 struct bpf_cg_run_ctx run_ctx; 1197 u32 ret = 1; 1198 u32 func_ret; 1199 1200 migrate_disable(); 1201 rcu_read_lock(); 1202 array = rcu_dereference(array_rcu); 1203 item = &array->items[0]; 1204 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1205 while ((prog = READ_ONCE(item->prog))) { 1206 run_ctx.prog_item = item; 1207 func_ret = run_prog(prog, ctx); 1208 ret &= (func_ret & 1); 1209 *(ret_flags) |= (func_ret >> 1); 1210 item++; 1211 } 1212 bpf_reset_run_ctx(old_run_ctx); 1213 rcu_read_unlock(); 1214 migrate_enable(); 1215 return ret; 1216 } 1217 1218 static __always_inline u32 1219 BPF_PROG_RUN_ARRAY_CG(const struct bpf_prog_array __rcu *array_rcu, 1220 const void *ctx, bpf_prog_run_fn run_prog) 1221 { 1222 const struct bpf_prog_array_item *item; 1223 const struct bpf_prog *prog; 1224 const struct bpf_prog_array *array; 1225 struct bpf_run_ctx *old_run_ctx; 1226 struct bpf_cg_run_ctx run_ctx; 1227 u32 ret = 1; 1228 1229 migrate_disable(); 1230 rcu_read_lock(); 1231 array = rcu_dereference(array_rcu); 1232 item = &array->items[0]; 1233 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1234 while ((prog = READ_ONCE(item->prog))) { 1235 run_ctx.prog_item = item; 1236 ret &= run_prog(prog, ctx); 1237 item++; 1238 } 1239 bpf_reset_run_ctx(old_run_ctx); 1240 rcu_read_unlock(); 1241 migrate_enable(); 1242 return ret; 1243 } 1244 1245 static __always_inline u32 1246 BPF_PROG_RUN_ARRAY(const struct bpf_prog_array __rcu *array_rcu, 1247 const void *ctx, bpf_prog_run_fn run_prog) 1248 { 1249 const struct bpf_prog_array_item *item; 1250 const struct bpf_prog *prog; 1251 const struct bpf_prog_array *array; 1252 struct bpf_run_ctx *old_run_ctx; 1253 struct bpf_trace_run_ctx run_ctx; 1254 u32 ret = 1; 1255 1256 migrate_disable(); 1257 rcu_read_lock(); 1258 array = rcu_dereference(array_rcu); 1259 if (unlikely(!array)) 1260 goto out; 1261 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1262 item = &array->items[0]; 1263 while ((prog = READ_ONCE(item->prog))) { 1264 run_ctx.bpf_cookie = item->bpf_cookie; 1265 ret &= run_prog(prog, ctx); 1266 item++; 1267 } 1268 bpf_reset_run_ctx(old_run_ctx); 1269 out: 1270 rcu_read_unlock(); 1271 migrate_enable(); 1272 return ret; 1273 } 1274 1275 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs 1276 * so BPF programs can request cwr for TCP packets. 1277 * 1278 * Current cgroup skb programs can only return 0 or 1 (0 to drop the 1279 * packet. This macro changes the behavior so the low order bit 1280 * indicates whether the packet should be dropped (0) or not (1) 1281 * and the next bit is a congestion notification bit. This could be 1282 * used by TCP to call tcp_enter_cwr() 1283 * 1284 * Hence, new allowed return values of CGROUP EGRESS BPF programs are: 1285 * 0: drop packet 1286 * 1: keep packet 1287 * 2: drop packet and cn 1288 * 3: keep packet and cn 1289 * 1290 * This macro then converts it to one of the NET_XMIT or an error 1291 * code that is then interpreted as drop packet (and no cn): 1292 * 0: NET_XMIT_SUCCESS skb should be transmitted 1293 * 1: NET_XMIT_DROP skb should be dropped and cn 1294 * 2: NET_XMIT_CN skb should be transmitted and cn 1295 * 3: -EPERM skb should be dropped 1296 */ 1297 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \ 1298 ({ \ 1299 u32 _flags = 0; \ 1300 bool _cn; \ 1301 u32 _ret; \ 1302 _ret = BPF_PROG_RUN_ARRAY_CG_FLAGS(array, ctx, func, &_flags); \ 1303 _cn = _flags & BPF_RET_SET_CN; \ 1304 if (_ret) \ 1305 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \ 1306 else \ 1307 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \ 1308 _ret; \ 1309 }) 1310 1311 #ifdef CONFIG_BPF_SYSCALL 1312 DECLARE_PER_CPU(int, bpf_prog_active); 1313 extern struct mutex bpf_stats_enabled_mutex; 1314 1315 /* 1316 * Block execution of BPF programs attached to instrumentation (perf, 1317 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 1318 * these events can happen inside a region which holds a map bucket lock 1319 * and can deadlock on it. 1320 * 1321 * Use the preemption safe inc/dec variants on RT because migrate disable 1322 * is preemptible on RT and preemption in the middle of the RMW operation 1323 * might lead to inconsistent state. Use the raw variants for non RT 1324 * kernels as migrate_disable() maps to preempt_disable() so the slightly 1325 * more expensive save operation can be avoided. 1326 */ 1327 static inline void bpf_disable_instrumentation(void) 1328 { 1329 migrate_disable(); 1330 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1331 this_cpu_inc(bpf_prog_active); 1332 else 1333 __this_cpu_inc(bpf_prog_active); 1334 } 1335 1336 static inline void bpf_enable_instrumentation(void) 1337 { 1338 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1339 this_cpu_dec(bpf_prog_active); 1340 else 1341 __this_cpu_dec(bpf_prog_active); 1342 migrate_enable(); 1343 } 1344 1345 extern const struct file_operations bpf_map_fops; 1346 extern const struct file_operations bpf_prog_fops; 1347 extern const struct file_operations bpf_iter_fops; 1348 1349 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 1350 extern const struct bpf_prog_ops _name ## _prog_ops; \ 1351 extern const struct bpf_verifier_ops _name ## _verifier_ops; 1352 #define BPF_MAP_TYPE(_id, _ops) \ 1353 extern const struct bpf_map_ops _ops; 1354 #define BPF_LINK_TYPE(_id, _name) 1355 #include <linux/bpf_types.h> 1356 #undef BPF_PROG_TYPE 1357 #undef BPF_MAP_TYPE 1358 #undef BPF_LINK_TYPE 1359 1360 extern const struct bpf_prog_ops bpf_offload_prog_ops; 1361 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 1362 extern const struct bpf_verifier_ops xdp_analyzer_ops; 1363 1364 struct bpf_prog *bpf_prog_get(u32 ufd); 1365 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 1366 bool attach_drv); 1367 void bpf_prog_add(struct bpf_prog *prog, int i); 1368 void bpf_prog_sub(struct bpf_prog *prog, int i); 1369 void bpf_prog_inc(struct bpf_prog *prog); 1370 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 1371 void bpf_prog_put(struct bpf_prog *prog); 1372 1373 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 1374 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 1375 1376 struct bpf_map *bpf_map_get(u32 ufd); 1377 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 1378 struct bpf_map *__bpf_map_get(struct fd f); 1379 void bpf_map_inc(struct bpf_map *map); 1380 void bpf_map_inc_with_uref(struct bpf_map *map); 1381 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 1382 void bpf_map_put_with_uref(struct bpf_map *map); 1383 void bpf_map_put(struct bpf_map *map); 1384 void *bpf_map_area_alloc(u64 size, int numa_node); 1385 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 1386 void bpf_map_area_free(void *base); 1387 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 1388 int generic_map_lookup_batch(struct bpf_map *map, 1389 const union bpf_attr *attr, 1390 union bpf_attr __user *uattr); 1391 int generic_map_update_batch(struct bpf_map *map, 1392 const union bpf_attr *attr, 1393 union bpf_attr __user *uattr); 1394 int generic_map_delete_batch(struct bpf_map *map, 1395 const union bpf_attr *attr, 1396 union bpf_attr __user *uattr); 1397 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 1398 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 1399 1400 #ifdef CONFIG_MEMCG_KMEM 1401 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1402 int node); 1403 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 1404 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 1405 size_t align, gfp_t flags); 1406 #else 1407 static inline void * 1408 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1409 int node) 1410 { 1411 return kmalloc_node(size, flags, node); 1412 } 1413 1414 static inline void * 1415 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags) 1416 { 1417 return kzalloc(size, flags); 1418 } 1419 1420 static inline void __percpu * 1421 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align, 1422 gfp_t flags) 1423 { 1424 return __alloc_percpu_gfp(size, align, flags); 1425 } 1426 #endif 1427 1428 extern int sysctl_unprivileged_bpf_disabled; 1429 1430 static inline bool bpf_allow_ptr_leaks(void) 1431 { 1432 return perfmon_capable(); 1433 } 1434 1435 static inline bool bpf_allow_uninit_stack(void) 1436 { 1437 return perfmon_capable(); 1438 } 1439 1440 static inline bool bpf_allow_ptr_to_map_access(void) 1441 { 1442 return perfmon_capable(); 1443 } 1444 1445 static inline bool bpf_bypass_spec_v1(void) 1446 { 1447 return perfmon_capable(); 1448 } 1449 1450 static inline bool bpf_bypass_spec_v4(void) 1451 { 1452 return perfmon_capable(); 1453 } 1454 1455 int bpf_map_new_fd(struct bpf_map *map, int flags); 1456 int bpf_prog_new_fd(struct bpf_prog *prog); 1457 1458 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1459 const struct bpf_link_ops *ops, struct bpf_prog *prog); 1460 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 1461 int bpf_link_settle(struct bpf_link_primer *primer); 1462 void bpf_link_cleanup(struct bpf_link_primer *primer); 1463 void bpf_link_inc(struct bpf_link *link); 1464 void bpf_link_put(struct bpf_link *link); 1465 int bpf_link_new_fd(struct bpf_link *link); 1466 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd); 1467 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 1468 1469 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 1470 int bpf_obj_get_user(const char __user *pathname, int flags); 1471 1472 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 1473 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 1474 extern int bpf_iter_ ## target(args); \ 1475 int __init bpf_iter_ ## target(args) { return 0; } 1476 1477 struct bpf_iter_aux_info { 1478 struct bpf_map *map; 1479 }; 1480 1481 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 1482 union bpf_iter_link_info *linfo, 1483 struct bpf_iter_aux_info *aux); 1484 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 1485 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 1486 struct seq_file *seq); 1487 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 1488 struct bpf_link_info *info); 1489 typedef const struct bpf_func_proto * 1490 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 1491 const struct bpf_prog *prog); 1492 1493 enum bpf_iter_feature { 1494 BPF_ITER_RESCHED = BIT(0), 1495 }; 1496 1497 #define BPF_ITER_CTX_ARG_MAX 2 1498 struct bpf_iter_reg { 1499 const char *target; 1500 bpf_iter_attach_target_t attach_target; 1501 bpf_iter_detach_target_t detach_target; 1502 bpf_iter_show_fdinfo_t show_fdinfo; 1503 bpf_iter_fill_link_info_t fill_link_info; 1504 bpf_iter_get_func_proto_t get_func_proto; 1505 u32 ctx_arg_info_size; 1506 u32 feature; 1507 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 1508 const struct bpf_iter_seq_info *seq_info; 1509 }; 1510 1511 struct bpf_iter_meta { 1512 __bpf_md_ptr(struct seq_file *, seq); 1513 u64 session_id; 1514 u64 seq_num; 1515 }; 1516 1517 struct bpf_iter__bpf_map_elem { 1518 __bpf_md_ptr(struct bpf_iter_meta *, meta); 1519 __bpf_md_ptr(struct bpf_map *, map); 1520 __bpf_md_ptr(void *, key); 1521 __bpf_md_ptr(void *, value); 1522 }; 1523 1524 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 1525 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 1526 bool bpf_iter_prog_supported(struct bpf_prog *prog); 1527 const struct bpf_func_proto * 1528 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 1529 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 1530 int bpf_iter_new_fd(struct bpf_link *link); 1531 bool bpf_link_is_iter(struct bpf_link *link); 1532 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 1533 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 1534 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 1535 struct seq_file *seq); 1536 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 1537 struct bpf_link_info *info); 1538 1539 int map_set_for_each_callback_args(struct bpf_verifier_env *env, 1540 struct bpf_func_state *caller, 1541 struct bpf_func_state *callee); 1542 1543 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 1544 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 1545 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 1546 u64 flags); 1547 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 1548 u64 flags); 1549 1550 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 1551 1552 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 1553 void *key, void *value, u64 map_flags); 1554 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1555 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 1556 void *key, void *value, u64 map_flags); 1557 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1558 1559 int bpf_get_file_flag(int flags); 1560 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 1561 size_t actual_size); 1562 1563 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 1564 * forced to use 'long' read/writes to try to atomically copy long counters. 1565 * Best-effort only. No barriers here, since it _will_ race with concurrent 1566 * updates from BPF programs. Called from bpf syscall and mostly used with 1567 * size 8 or 16 bytes, so ask compiler to inline it. 1568 */ 1569 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 1570 { 1571 const long *lsrc = src; 1572 long *ldst = dst; 1573 1574 size /= sizeof(long); 1575 while (size--) 1576 *ldst++ = *lsrc++; 1577 } 1578 1579 /* verify correctness of eBPF program */ 1580 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr); 1581 1582 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 1583 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 1584 #endif 1585 1586 struct btf *bpf_get_btf_vmlinux(void); 1587 1588 /* Map specifics */ 1589 struct xdp_buff; 1590 struct sk_buff; 1591 struct bpf_dtab_netdev; 1592 struct bpf_cpu_map_entry; 1593 1594 void __dev_flush(void); 1595 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1596 struct net_device *dev_rx); 1597 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1598 struct net_device *dev_rx); 1599 int dev_map_enqueue_multi(struct xdp_buff *xdp, struct net_device *dev_rx, 1600 struct bpf_map *map, bool exclude_ingress); 1601 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 1602 struct bpf_prog *xdp_prog); 1603 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1604 struct bpf_prog *xdp_prog, struct bpf_map *map, 1605 bool exclude_ingress); 1606 1607 void __cpu_map_flush(void); 1608 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, 1609 struct net_device *dev_rx); 1610 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1611 struct sk_buff *skb); 1612 1613 /* Return map's numa specified by userspace */ 1614 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 1615 { 1616 return (attr->map_flags & BPF_F_NUMA_NODE) ? 1617 attr->numa_node : NUMA_NO_NODE; 1618 } 1619 1620 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 1621 int array_map_alloc_check(union bpf_attr *attr); 1622 1623 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 1624 union bpf_attr __user *uattr); 1625 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 1626 union bpf_attr __user *uattr); 1627 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1628 const union bpf_attr *kattr, 1629 union bpf_attr __user *uattr); 1630 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1631 const union bpf_attr *kattr, 1632 union bpf_attr __user *uattr); 1633 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 1634 const union bpf_attr *kattr, 1635 union bpf_attr __user *uattr); 1636 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 1637 const union bpf_attr *kattr, 1638 union bpf_attr __user *uattr); 1639 bool bpf_prog_test_check_kfunc_call(u32 kfunc_id); 1640 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 1641 const struct bpf_prog *prog, 1642 struct bpf_insn_access_aux *info); 1643 int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf, 1644 const struct btf_type *t, int off, int size, 1645 enum bpf_access_type atype, 1646 u32 *next_btf_id); 1647 bool btf_struct_ids_match(struct bpf_verifier_log *log, 1648 const struct btf *btf, u32 id, int off, 1649 const struct btf *need_btf, u32 need_type_id); 1650 1651 int btf_distill_func_proto(struct bpf_verifier_log *log, 1652 struct btf *btf, 1653 const struct btf_type *func_proto, 1654 const char *func_name, 1655 struct btf_func_model *m); 1656 1657 struct bpf_reg_state; 1658 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog, 1659 struct bpf_reg_state *regs); 1660 int btf_check_kfunc_arg_match(struct bpf_verifier_env *env, 1661 const struct btf *btf, u32 func_id, 1662 struct bpf_reg_state *regs); 1663 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, 1664 struct bpf_reg_state *reg); 1665 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 1666 struct btf *btf, const struct btf_type *t); 1667 1668 struct bpf_prog *bpf_prog_by_id(u32 id); 1669 struct bpf_link *bpf_link_by_id(u32 id); 1670 1671 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id); 1672 void bpf_task_storage_free(struct task_struct *task); 1673 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog); 1674 const struct btf_func_model * 1675 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 1676 const struct bpf_insn *insn); 1677 #else /* !CONFIG_BPF_SYSCALL */ 1678 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 1679 { 1680 return ERR_PTR(-EOPNOTSUPP); 1681 } 1682 1683 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 1684 enum bpf_prog_type type, 1685 bool attach_drv) 1686 { 1687 return ERR_PTR(-EOPNOTSUPP); 1688 } 1689 1690 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 1691 { 1692 } 1693 1694 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 1695 { 1696 } 1697 1698 static inline void bpf_prog_put(struct bpf_prog *prog) 1699 { 1700 } 1701 1702 static inline void bpf_prog_inc(struct bpf_prog *prog) 1703 { 1704 } 1705 1706 static inline struct bpf_prog *__must_check 1707 bpf_prog_inc_not_zero(struct bpf_prog *prog) 1708 { 1709 return ERR_PTR(-EOPNOTSUPP); 1710 } 1711 1712 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1713 const struct bpf_link_ops *ops, 1714 struct bpf_prog *prog) 1715 { 1716 } 1717 1718 static inline int bpf_link_prime(struct bpf_link *link, 1719 struct bpf_link_primer *primer) 1720 { 1721 return -EOPNOTSUPP; 1722 } 1723 1724 static inline int bpf_link_settle(struct bpf_link_primer *primer) 1725 { 1726 return -EOPNOTSUPP; 1727 } 1728 1729 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 1730 { 1731 } 1732 1733 static inline void bpf_link_inc(struct bpf_link *link) 1734 { 1735 } 1736 1737 static inline void bpf_link_put(struct bpf_link *link) 1738 { 1739 } 1740 1741 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 1742 { 1743 return -EOPNOTSUPP; 1744 } 1745 1746 static inline bool dev_map_can_have_prog(struct bpf_map *map) 1747 { 1748 return false; 1749 } 1750 1751 static inline void __dev_flush(void) 1752 { 1753 } 1754 1755 struct xdp_buff; 1756 struct bpf_dtab_netdev; 1757 struct bpf_cpu_map_entry; 1758 1759 static inline 1760 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1761 struct net_device *dev_rx) 1762 { 1763 return 0; 1764 } 1765 1766 static inline 1767 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1768 struct net_device *dev_rx) 1769 { 1770 return 0; 1771 } 1772 1773 static inline 1774 int dev_map_enqueue_multi(struct xdp_buff *xdp, struct net_device *dev_rx, 1775 struct bpf_map *map, bool exclude_ingress) 1776 { 1777 return 0; 1778 } 1779 1780 struct sk_buff; 1781 1782 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 1783 struct sk_buff *skb, 1784 struct bpf_prog *xdp_prog) 1785 { 1786 return 0; 1787 } 1788 1789 static inline 1790 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1791 struct bpf_prog *xdp_prog, struct bpf_map *map, 1792 bool exclude_ingress) 1793 { 1794 return 0; 1795 } 1796 1797 static inline void __cpu_map_flush(void) 1798 { 1799 } 1800 1801 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 1802 struct xdp_buff *xdp, 1803 struct net_device *dev_rx) 1804 { 1805 return 0; 1806 } 1807 1808 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1809 struct sk_buff *skb) 1810 { 1811 return -EOPNOTSUPP; 1812 } 1813 1814 static inline bool cpu_map_prog_allowed(struct bpf_map *map) 1815 { 1816 return false; 1817 } 1818 1819 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 1820 enum bpf_prog_type type) 1821 { 1822 return ERR_PTR(-EOPNOTSUPP); 1823 } 1824 1825 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 1826 const union bpf_attr *kattr, 1827 union bpf_attr __user *uattr) 1828 { 1829 return -ENOTSUPP; 1830 } 1831 1832 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 1833 const union bpf_attr *kattr, 1834 union bpf_attr __user *uattr) 1835 { 1836 return -ENOTSUPP; 1837 } 1838 1839 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1840 const union bpf_attr *kattr, 1841 union bpf_attr __user *uattr) 1842 { 1843 return -ENOTSUPP; 1844 } 1845 1846 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1847 const union bpf_attr *kattr, 1848 union bpf_attr __user *uattr) 1849 { 1850 return -ENOTSUPP; 1851 } 1852 1853 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 1854 const union bpf_attr *kattr, 1855 union bpf_attr __user *uattr) 1856 { 1857 return -ENOTSUPP; 1858 } 1859 1860 static inline bool bpf_prog_test_check_kfunc_call(u32 kfunc_id) 1861 { 1862 return false; 1863 } 1864 1865 static inline void bpf_map_put(struct bpf_map *map) 1866 { 1867 } 1868 1869 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 1870 { 1871 return ERR_PTR(-ENOTSUPP); 1872 } 1873 1874 static inline const struct bpf_func_proto * 1875 bpf_base_func_proto(enum bpf_func_id func_id) 1876 { 1877 return NULL; 1878 } 1879 1880 static inline void bpf_task_storage_free(struct task_struct *task) 1881 { 1882 } 1883 1884 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog) 1885 { 1886 return false; 1887 } 1888 1889 static inline const struct btf_func_model * 1890 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 1891 const struct bpf_insn *insn) 1892 { 1893 return NULL; 1894 } 1895 #endif /* CONFIG_BPF_SYSCALL */ 1896 1897 void __bpf_free_used_btfs(struct bpf_prog_aux *aux, 1898 struct btf_mod_pair *used_btfs, u32 len); 1899 1900 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 1901 enum bpf_prog_type type) 1902 { 1903 return bpf_prog_get_type_dev(ufd, type, false); 1904 } 1905 1906 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 1907 struct bpf_map **used_maps, u32 len); 1908 1909 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 1910 1911 int bpf_prog_offload_compile(struct bpf_prog *prog); 1912 void bpf_prog_offload_destroy(struct bpf_prog *prog); 1913 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 1914 struct bpf_prog *prog); 1915 1916 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 1917 1918 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 1919 int bpf_map_offload_update_elem(struct bpf_map *map, 1920 void *key, void *value, u64 flags); 1921 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 1922 int bpf_map_offload_get_next_key(struct bpf_map *map, 1923 void *key, void *next_key); 1924 1925 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 1926 1927 struct bpf_offload_dev * 1928 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 1929 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 1930 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 1931 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 1932 struct net_device *netdev); 1933 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 1934 struct net_device *netdev); 1935 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 1936 1937 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 1938 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 1939 1940 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 1941 { 1942 return aux->offload_requested; 1943 } 1944 1945 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1946 { 1947 return unlikely(map->ops == &bpf_map_offload_ops); 1948 } 1949 1950 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 1951 void bpf_map_offload_map_free(struct bpf_map *map); 1952 int bpf_prog_test_run_syscall(struct bpf_prog *prog, 1953 const union bpf_attr *kattr, 1954 union bpf_attr __user *uattr); 1955 1956 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 1957 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 1958 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 1959 void sock_map_unhash(struct sock *sk); 1960 void sock_map_close(struct sock *sk, long timeout); 1961 #else 1962 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 1963 union bpf_attr *attr) 1964 { 1965 return -EOPNOTSUPP; 1966 } 1967 1968 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 1969 { 1970 return false; 1971 } 1972 1973 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1974 { 1975 return false; 1976 } 1977 1978 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 1979 { 1980 return ERR_PTR(-EOPNOTSUPP); 1981 } 1982 1983 static inline void bpf_map_offload_map_free(struct bpf_map *map) 1984 { 1985 } 1986 1987 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 1988 const union bpf_attr *kattr, 1989 union bpf_attr __user *uattr) 1990 { 1991 return -ENOTSUPP; 1992 } 1993 1994 #ifdef CONFIG_BPF_SYSCALL 1995 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 1996 struct bpf_prog *prog) 1997 { 1998 return -EINVAL; 1999 } 2000 2001 static inline int sock_map_prog_detach(const union bpf_attr *attr, 2002 enum bpf_prog_type ptype) 2003 { 2004 return -EOPNOTSUPP; 2005 } 2006 2007 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 2008 u64 flags) 2009 { 2010 return -EOPNOTSUPP; 2011 } 2012 #endif /* CONFIG_BPF_SYSCALL */ 2013 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 2014 2015 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 2016 void bpf_sk_reuseport_detach(struct sock *sk); 2017 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 2018 void *value); 2019 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 2020 void *value, u64 map_flags); 2021 #else 2022 static inline void bpf_sk_reuseport_detach(struct sock *sk) 2023 { 2024 } 2025 2026 #ifdef CONFIG_BPF_SYSCALL 2027 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 2028 void *key, void *value) 2029 { 2030 return -EOPNOTSUPP; 2031 } 2032 2033 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 2034 void *key, void *value, 2035 u64 map_flags) 2036 { 2037 return -EOPNOTSUPP; 2038 } 2039 #endif /* CONFIG_BPF_SYSCALL */ 2040 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 2041 2042 /* verifier prototypes for helper functions called from eBPF programs */ 2043 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 2044 extern const struct bpf_func_proto bpf_map_update_elem_proto; 2045 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 2046 extern const struct bpf_func_proto bpf_map_push_elem_proto; 2047 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 2048 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 2049 2050 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 2051 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 2052 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 2053 extern const struct bpf_func_proto bpf_tail_call_proto; 2054 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 2055 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 2056 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 2057 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 2058 extern const struct bpf_func_proto bpf_get_current_comm_proto; 2059 extern const struct bpf_func_proto bpf_get_stackid_proto; 2060 extern const struct bpf_func_proto bpf_get_stack_proto; 2061 extern const struct bpf_func_proto bpf_get_task_stack_proto; 2062 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 2063 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 2064 extern const struct bpf_func_proto bpf_sock_map_update_proto; 2065 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 2066 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 2067 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 2068 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 2069 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 2070 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 2071 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 2072 extern const struct bpf_func_proto bpf_spin_lock_proto; 2073 extern const struct bpf_func_proto bpf_spin_unlock_proto; 2074 extern const struct bpf_func_proto bpf_get_local_storage_proto; 2075 extern const struct bpf_func_proto bpf_strtol_proto; 2076 extern const struct bpf_func_proto bpf_strtoul_proto; 2077 extern const struct bpf_func_proto bpf_tcp_sock_proto; 2078 extern const struct bpf_func_proto bpf_jiffies64_proto; 2079 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 2080 extern const struct bpf_func_proto bpf_event_output_data_proto; 2081 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 2082 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 2083 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 2084 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 2085 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 2086 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 2087 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 2088 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 2089 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 2090 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 2091 extern const struct bpf_func_proto bpf_copy_from_user_proto; 2092 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 2093 extern const struct bpf_func_proto bpf_snprintf_proto; 2094 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 2095 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 2096 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 2097 extern const struct bpf_func_proto bpf_sock_from_file_proto; 2098 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 2099 extern const struct bpf_func_proto bpf_task_storage_get_proto; 2100 extern const struct bpf_func_proto bpf_task_storage_delete_proto; 2101 extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 2102 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 2103 extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 2104 extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 2105 2106 const struct bpf_func_proto *tracing_prog_func_proto( 2107 enum bpf_func_id func_id, const struct bpf_prog *prog); 2108 2109 /* Shared helpers among cBPF and eBPF. */ 2110 void bpf_user_rnd_init_once(void); 2111 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2112 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2113 2114 #if defined(CONFIG_NET) 2115 bool bpf_sock_common_is_valid_access(int off, int size, 2116 enum bpf_access_type type, 2117 struct bpf_insn_access_aux *info); 2118 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2119 struct bpf_insn_access_aux *info); 2120 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2121 const struct bpf_insn *si, 2122 struct bpf_insn *insn_buf, 2123 struct bpf_prog *prog, 2124 u32 *target_size); 2125 #else 2126 static inline bool bpf_sock_common_is_valid_access(int off, int size, 2127 enum bpf_access_type type, 2128 struct bpf_insn_access_aux *info) 2129 { 2130 return false; 2131 } 2132 static inline bool bpf_sock_is_valid_access(int off, int size, 2133 enum bpf_access_type type, 2134 struct bpf_insn_access_aux *info) 2135 { 2136 return false; 2137 } 2138 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2139 const struct bpf_insn *si, 2140 struct bpf_insn *insn_buf, 2141 struct bpf_prog *prog, 2142 u32 *target_size) 2143 { 2144 return 0; 2145 } 2146 #endif 2147 2148 #ifdef CONFIG_INET 2149 struct sk_reuseport_kern { 2150 struct sk_buff *skb; 2151 struct sock *sk; 2152 struct sock *selected_sk; 2153 struct sock *migrating_sk; 2154 void *data_end; 2155 u32 hash; 2156 u32 reuseport_id; 2157 bool bind_inany; 2158 }; 2159 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2160 struct bpf_insn_access_aux *info); 2161 2162 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2163 const struct bpf_insn *si, 2164 struct bpf_insn *insn_buf, 2165 struct bpf_prog *prog, 2166 u32 *target_size); 2167 2168 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2169 struct bpf_insn_access_aux *info); 2170 2171 u32 bpf_xdp_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 #else 2177 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 2178 enum bpf_access_type type, 2179 struct bpf_insn_access_aux *info) 2180 { 2181 return false; 2182 } 2183 2184 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2185 const struct bpf_insn *si, 2186 struct bpf_insn *insn_buf, 2187 struct bpf_prog *prog, 2188 u32 *target_size) 2189 { 2190 return 0; 2191 } 2192 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 2193 enum bpf_access_type type, 2194 struct bpf_insn_access_aux *info) 2195 { 2196 return false; 2197 } 2198 2199 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2200 const struct bpf_insn *si, 2201 struct bpf_insn *insn_buf, 2202 struct bpf_prog *prog, 2203 u32 *target_size) 2204 { 2205 return 0; 2206 } 2207 #endif /* CONFIG_INET */ 2208 2209 enum bpf_text_poke_type { 2210 BPF_MOD_CALL, 2211 BPF_MOD_JUMP, 2212 }; 2213 2214 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 2215 void *addr1, void *addr2); 2216 2217 struct btf_id_set; 2218 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 2219 2220 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args, 2221 u32 **bin_buf, u32 num_args); 2222 void bpf_bprintf_cleanup(void); 2223 2224 #endif /* _LINUX_BPF_H */ 2225