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