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/u64_stats_sync.h> 18 #include <linux/refcount.h> 19 #include <linux/mutex.h> 20 #include <linux/module.h> 21 #include <linux/kallsyms.h> 22 #include <linux/capability.h> 23 24 struct bpf_verifier_env; 25 struct bpf_verifier_log; 26 struct perf_event; 27 struct bpf_prog; 28 struct bpf_prog_aux; 29 struct bpf_map; 30 struct sock; 31 struct seq_file; 32 struct btf; 33 struct btf_type; 34 struct exception_table_entry; 35 struct seq_operations; 36 struct bpf_iter_aux_info; 37 struct bpf_local_storage; 38 struct bpf_local_storage_map; 39 struct kobject; 40 41 extern struct idr btf_idr; 42 extern spinlock_t btf_idr_lock; 43 extern struct kobject *btf_kobj; 44 45 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, 46 struct bpf_iter_aux_info *aux); 47 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); 48 struct bpf_iter_seq_info { 49 const struct seq_operations *seq_ops; 50 bpf_iter_init_seq_priv_t init_seq_private; 51 bpf_iter_fini_seq_priv_t fini_seq_private; 52 u32 seq_priv_size; 53 }; 54 55 /* map is generic key/value storage optionally accesible by eBPF programs */ 56 struct bpf_map_ops { 57 /* funcs callable from userspace (via syscall) */ 58 int (*map_alloc_check)(union bpf_attr *attr); 59 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 60 void (*map_release)(struct bpf_map *map, struct file *map_file); 61 void (*map_free)(struct bpf_map *map); 62 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 63 void (*map_release_uref)(struct bpf_map *map); 64 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 65 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, 66 union bpf_attr __user *uattr); 67 int (*map_lookup_and_delete_batch)(struct bpf_map *map, 68 const union bpf_attr *attr, 69 union bpf_attr __user *uattr); 70 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr, 71 union bpf_attr __user *uattr); 72 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, 73 union bpf_attr __user *uattr); 74 75 /* funcs callable from userspace and from eBPF programs */ 76 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 77 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 78 int (*map_delete_elem)(struct bpf_map *map, void *key); 79 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 80 int (*map_pop_elem)(struct bpf_map *map, void *value); 81 int (*map_peek_elem)(struct bpf_map *map, void *value); 82 83 /* funcs called by prog_array and perf_event_array map */ 84 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 85 int fd); 86 void (*map_fd_put_ptr)(void *ptr); 87 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 88 u32 (*map_fd_sys_lookup_elem)(void *ptr); 89 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 90 struct seq_file *m); 91 int (*map_check_btf)(const struct bpf_map *map, 92 const struct btf *btf, 93 const struct btf_type *key_type, 94 const struct btf_type *value_type); 95 96 /* Prog poke tracking helpers. */ 97 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux); 98 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux); 99 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, 100 struct bpf_prog *new); 101 102 /* Direct value access helpers. */ 103 int (*map_direct_value_addr)(const struct bpf_map *map, 104 u64 *imm, u32 off); 105 int (*map_direct_value_meta)(const struct bpf_map *map, 106 u64 imm, u32 *off); 107 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma); 108 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, 109 struct poll_table_struct *pts); 110 111 /* Functions called by bpf_local_storage maps */ 112 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, 113 void *owner, u32 size); 114 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, 115 void *owner, u32 size); 116 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner); 117 118 /* map_meta_equal must be implemented for maps that can be 119 * used as an inner map. It is a runtime check to ensure 120 * an inner map can be inserted to an outer map. 121 * 122 * Some properties of the inner map has been used during the 123 * verification time. When inserting an inner map at the runtime, 124 * map_meta_equal has to ensure the inserting map has the same 125 * properties that the verifier has used earlier. 126 */ 127 bool (*map_meta_equal)(const struct bpf_map *meta0, 128 const struct bpf_map *meta1); 129 130 /* BTF name and id of struct allocated by map_alloc */ 131 const char * const map_btf_name; 132 int *map_btf_id; 133 134 /* bpf_iter info used to open a seq_file */ 135 const struct bpf_iter_seq_info *iter_seq_info; 136 }; 137 138 struct bpf_map_memory { 139 u32 pages; 140 struct user_struct *user; 141 }; 142 143 struct bpf_map { 144 /* The first two cachelines with read-mostly members of which some 145 * are also accessed in fast-path (e.g. ops, max_entries). 146 */ 147 const struct bpf_map_ops *ops ____cacheline_aligned; 148 struct bpf_map *inner_map_meta; 149 #ifdef CONFIG_SECURITY 150 void *security; 151 #endif 152 enum bpf_map_type map_type; 153 u32 key_size; 154 u32 value_size; 155 u32 max_entries; 156 u32 map_flags; 157 int spin_lock_off; /* >=0 valid offset, <0 error */ 158 u32 id; 159 int numa_node; 160 u32 btf_key_type_id; 161 u32 btf_value_type_id; 162 struct btf *btf; 163 struct bpf_map_memory memory; 164 char name[BPF_OBJ_NAME_LEN]; 165 u32 btf_vmlinux_value_type_id; 166 bool bypass_spec_v1; 167 bool frozen; /* write-once; write-protected by freeze_mutex */ 168 /* 22 bytes hole */ 169 170 /* The 3rd and 4th cacheline with misc members to avoid false sharing 171 * particularly with refcounting. 172 */ 173 atomic64_t refcnt ____cacheline_aligned; 174 atomic64_t usercnt; 175 struct work_struct work; 176 struct mutex freeze_mutex; 177 u64 writecnt; /* writable mmap cnt; protected by freeze_mutex */ 178 }; 179 180 static inline bool map_value_has_spin_lock(const struct bpf_map *map) 181 { 182 return map->spin_lock_off >= 0; 183 } 184 185 static inline void check_and_init_map_lock(struct bpf_map *map, void *dst) 186 { 187 if (likely(!map_value_has_spin_lock(map))) 188 return; 189 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) = 190 (struct bpf_spin_lock){}; 191 } 192 193 /* copy everything but bpf_spin_lock */ 194 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 195 { 196 if (unlikely(map_value_has_spin_lock(map))) { 197 u32 off = map->spin_lock_off; 198 199 memcpy(dst, src, off); 200 memcpy(dst + off + sizeof(struct bpf_spin_lock), 201 src + off + sizeof(struct bpf_spin_lock), 202 map->value_size - off - sizeof(struct bpf_spin_lock)); 203 } else { 204 memcpy(dst, src, map->value_size); 205 } 206 } 207 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 208 bool lock_src); 209 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size); 210 211 struct bpf_offload_dev; 212 struct bpf_offloaded_map; 213 214 struct bpf_map_dev_ops { 215 int (*map_get_next_key)(struct bpf_offloaded_map *map, 216 void *key, void *next_key); 217 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 218 void *key, void *value); 219 int (*map_update_elem)(struct bpf_offloaded_map *map, 220 void *key, void *value, u64 flags); 221 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 222 }; 223 224 struct bpf_offloaded_map { 225 struct bpf_map map; 226 struct net_device *netdev; 227 const struct bpf_map_dev_ops *dev_ops; 228 void *dev_priv; 229 struct list_head offloads; 230 }; 231 232 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 233 { 234 return container_of(map, struct bpf_offloaded_map, map); 235 } 236 237 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 238 { 239 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 240 } 241 242 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 243 { 244 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && 245 map->ops->map_seq_show_elem; 246 } 247 248 int map_check_no_btf(const struct bpf_map *map, 249 const struct btf *btf, 250 const struct btf_type *key_type, 251 const struct btf_type *value_type); 252 253 bool bpf_map_meta_equal(const struct bpf_map *meta0, 254 const struct bpf_map *meta1); 255 256 extern const struct bpf_map_ops bpf_map_offload_ops; 257 258 /* function argument constraints */ 259 enum bpf_arg_type { 260 ARG_DONTCARE = 0, /* unused argument in helper function */ 261 262 /* the following constraints used to prototype 263 * bpf_map_lookup/update/delete_elem() functions 264 */ 265 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 266 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 267 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 268 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */ 269 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */ 270 271 /* the following constraints used to prototype bpf_memcmp() and other 272 * functions that access data on eBPF program stack 273 */ 274 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 275 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */ 276 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized, 277 * helper function must fill all bytes or clear 278 * them in error case. 279 */ 280 281 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 282 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 283 284 ARG_PTR_TO_CTX, /* pointer to context */ 285 ARG_PTR_TO_CTX_OR_NULL, /* pointer to context or NULL */ 286 ARG_ANYTHING, /* any (initialized) argument is ok */ 287 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 288 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 289 ARG_PTR_TO_INT, /* pointer to int */ 290 ARG_PTR_TO_LONG, /* pointer to long */ 291 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 292 ARG_PTR_TO_SOCKET_OR_NULL, /* pointer to bpf_sock (fullsock) or NULL */ 293 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */ 294 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */ 295 ARG_PTR_TO_ALLOC_MEM_OR_NULL, /* pointer to dynamically allocated memory or NULL */ 296 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ 297 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ 298 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ 299 __BPF_ARG_TYPE_MAX, 300 }; 301 302 /* type of values returned from helper functions */ 303 enum bpf_return_type { 304 RET_INTEGER, /* function returns integer */ 305 RET_VOID, /* function doesn't return anything */ 306 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 307 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */ 308 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */ 309 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */ 310 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */ 311 RET_PTR_TO_ALLOC_MEM_OR_NULL, /* returns a pointer to dynamically allocated memory or NULL */ 312 RET_PTR_TO_BTF_ID_OR_NULL, /* returns a pointer to a btf_id or NULL */ 313 RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL, /* returns a pointer to a valid memory or a btf_id or NULL */ 314 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ 315 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */ 316 }; 317 318 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 319 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 320 * instructions after verifying 321 */ 322 struct bpf_func_proto { 323 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 324 bool gpl_only; 325 bool pkt_access; 326 enum bpf_return_type ret_type; 327 union { 328 struct { 329 enum bpf_arg_type arg1_type; 330 enum bpf_arg_type arg2_type; 331 enum bpf_arg_type arg3_type; 332 enum bpf_arg_type arg4_type; 333 enum bpf_arg_type arg5_type; 334 }; 335 enum bpf_arg_type arg_type[5]; 336 }; 337 union { 338 struct { 339 u32 *arg1_btf_id; 340 u32 *arg2_btf_id; 341 u32 *arg3_btf_id; 342 u32 *arg4_btf_id; 343 u32 *arg5_btf_id; 344 }; 345 u32 *arg_btf_id[5]; 346 }; 347 int *ret_btf_id; /* return value btf_id */ 348 bool (*allowed)(const struct bpf_prog *prog); 349 }; 350 351 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 352 * the first argument to eBPF programs. 353 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 354 */ 355 struct bpf_context; 356 357 enum bpf_access_type { 358 BPF_READ = 1, 359 BPF_WRITE = 2 360 }; 361 362 /* types of values stored in eBPF registers */ 363 /* Pointer types represent: 364 * pointer 365 * pointer + imm 366 * pointer + (u16) var 367 * pointer + (u16) var + imm 368 * if (range > 0) then [ptr, ptr + range - off) is safe to access 369 * if (id > 0) means that some 'var' was added 370 * if (off > 0) means that 'imm' was added 371 */ 372 enum bpf_reg_type { 373 NOT_INIT = 0, /* nothing was written into register */ 374 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 375 PTR_TO_CTX, /* reg points to bpf_context */ 376 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 377 PTR_TO_MAP_VALUE, /* reg points to map element value */ 378 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */ 379 PTR_TO_STACK, /* reg == frame_pointer + offset */ 380 PTR_TO_PACKET_META, /* skb->data - meta_len */ 381 PTR_TO_PACKET, /* reg points to skb->data */ 382 PTR_TO_PACKET_END, /* skb->data + headlen */ 383 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 384 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 385 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */ 386 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 387 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */ 388 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 389 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */ 390 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 391 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 392 /* PTR_TO_BTF_ID points to a kernel struct that does not need 393 * to be null checked by the BPF program. This does not imply the 394 * pointer is _not_ null and in practice this can easily be a null 395 * pointer when reading pointer chains. The assumption is program 396 * context will handle null pointer dereference typically via fault 397 * handling. The verifier must keep this in mind and can make no 398 * assumptions about null or non-null when doing branch analysis. 399 * Further, when passed into helpers the helpers can not, without 400 * additional context, assume the value is non-null. 401 */ 402 PTR_TO_BTF_ID, 403 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not 404 * been checked for null. Used primarily to inform the verifier 405 * an explicit null check is required for this struct. 406 */ 407 PTR_TO_BTF_ID_OR_NULL, 408 PTR_TO_MEM, /* reg points to valid memory region */ 409 PTR_TO_MEM_OR_NULL, /* reg points to valid memory region or NULL */ 410 PTR_TO_RDONLY_BUF, /* reg points to a readonly buffer */ 411 PTR_TO_RDONLY_BUF_OR_NULL, /* reg points to a readonly buffer or NULL */ 412 PTR_TO_RDWR_BUF, /* reg points to a read/write buffer */ 413 PTR_TO_RDWR_BUF_OR_NULL, /* reg points to a read/write buffer or NULL */ 414 PTR_TO_PERCPU_BTF_ID, /* reg points to a percpu kernel variable */ 415 }; 416 417 /* The information passed from prog-specific *_is_valid_access 418 * back to the verifier. 419 */ 420 struct bpf_insn_access_aux { 421 enum bpf_reg_type reg_type; 422 union { 423 int ctx_field_size; 424 u32 btf_id; 425 }; 426 struct bpf_verifier_log *log; /* for verbose logs */ 427 }; 428 429 static inline void 430 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 431 { 432 aux->ctx_field_size = size; 433 } 434 435 struct bpf_prog_ops { 436 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 437 union bpf_attr __user *uattr); 438 }; 439 440 struct bpf_verifier_ops { 441 /* return eBPF function prototype for verification */ 442 const struct bpf_func_proto * 443 (*get_func_proto)(enum bpf_func_id func_id, 444 const struct bpf_prog *prog); 445 446 /* return true if 'size' wide access at offset 'off' within bpf_context 447 * with 'type' (read or write) is allowed 448 */ 449 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 450 const struct bpf_prog *prog, 451 struct bpf_insn_access_aux *info); 452 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 453 const struct bpf_prog *prog); 454 int (*gen_ld_abs)(const struct bpf_insn *orig, 455 struct bpf_insn *insn_buf); 456 u32 (*convert_ctx_access)(enum bpf_access_type type, 457 const struct bpf_insn *src, 458 struct bpf_insn *dst, 459 struct bpf_prog *prog, u32 *target_size); 460 int (*btf_struct_access)(struct bpf_verifier_log *log, 461 const struct btf_type *t, int off, int size, 462 enum bpf_access_type atype, 463 u32 *next_btf_id); 464 }; 465 466 struct bpf_prog_offload_ops { 467 /* verifier basic callbacks */ 468 int (*insn_hook)(struct bpf_verifier_env *env, 469 int insn_idx, int prev_insn_idx); 470 int (*finalize)(struct bpf_verifier_env *env); 471 /* verifier optimization callbacks (called after .finalize) */ 472 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 473 struct bpf_insn *insn); 474 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 475 /* program management callbacks */ 476 int (*prepare)(struct bpf_prog *prog); 477 int (*translate)(struct bpf_prog *prog); 478 void (*destroy)(struct bpf_prog *prog); 479 }; 480 481 struct bpf_prog_offload { 482 struct bpf_prog *prog; 483 struct net_device *netdev; 484 struct bpf_offload_dev *offdev; 485 void *dev_priv; 486 struct list_head offloads; 487 bool dev_state; 488 bool opt_failed; 489 void *jited_image; 490 u32 jited_len; 491 }; 492 493 enum bpf_cgroup_storage_type { 494 BPF_CGROUP_STORAGE_SHARED, 495 BPF_CGROUP_STORAGE_PERCPU, 496 __BPF_CGROUP_STORAGE_MAX 497 }; 498 499 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 500 501 /* The longest tracepoint has 12 args. 502 * See include/trace/bpf_probe.h 503 */ 504 #define MAX_BPF_FUNC_ARGS 12 505 506 struct bpf_prog_stats { 507 u64 cnt; 508 u64 nsecs; 509 struct u64_stats_sync syncp; 510 } __aligned(2 * sizeof(u64)); 511 512 struct btf_func_model { 513 u8 ret_size; 514 u8 nr_args; 515 u8 arg_size[MAX_BPF_FUNC_ARGS]; 516 }; 517 518 /* Restore arguments before returning from trampoline to let original function 519 * continue executing. This flag is used for fentry progs when there are no 520 * fexit progs. 521 */ 522 #define BPF_TRAMP_F_RESTORE_REGS BIT(0) 523 /* Call original function after fentry progs, but before fexit progs. 524 * Makes sense for fentry/fexit, normal calls and indirect calls. 525 */ 526 #define BPF_TRAMP_F_CALL_ORIG BIT(1) 527 /* Skip current frame and return to parent. Makes sense for fentry/fexit 528 * programs only. Should not be used with normal calls and indirect calls. 529 */ 530 #define BPF_TRAMP_F_SKIP_FRAME BIT(2) 531 532 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 533 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2 534 */ 535 #define BPF_MAX_TRAMP_PROGS 40 536 537 struct bpf_tramp_progs { 538 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS]; 539 int nr_progs; 540 }; 541 542 /* Different use cases for BPF trampoline: 543 * 1. replace nop at the function entry (kprobe equivalent) 544 * flags = BPF_TRAMP_F_RESTORE_REGS 545 * fentry = a set of programs to run before returning from trampoline 546 * 547 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 548 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 549 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 550 * fentry = a set of program to run before calling original function 551 * fexit = a set of program to run after original function 552 * 553 * 3. replace direct call instruction anywhere in the function body 554 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 555 * With flags = 0 556 * fentry = a set of programs to run before returning from trampoline 557 * With flags = BPF_TRAMP_F_CALL_ORIG 558 * orig_call = original callback addr or direct function addr 559 * fentry = a set of program to run before calling original function 560 * fexit = a set of program to run after original function 561 */ 562 int arch_prepare_bpf_trampoline(void *image, void *image_end, 563 const struct btf_func_model *m, u32 flags, 564 struct bpf_tramp_progs *tprogs, 565 void *orig_call); 566 /* these two functions are called from generated trampoline */ 567 u64 notrace __bpf_prog_enter(void); 568 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start); 569 void notrace __bpf_prog_enter_sleepable(void); 570 void notrace __bpf_prog_exit_sleepable(void); 571 572 struct bpf_ksym { 573 unsigned long start; 574 unsigned long end; 575 char name[KSYM_NAME_LEN]; 576 struct list_head lnode; 577 struct latch_tree_node tnode; 578 bool prog; 579 }; 580 581 enum bpf_tramp_prog_type { 582 BPF_TRAMP_FENTRY, 583 BPF_TRAMP_FEXIT, 584 BPF_TRAMP_MODIFY_RETURN, 585 BPF_TRAMP_MAX, 586 BPF_TRAMP_REPLACE, /* more than MAX */ 587 }; 588 589 struct bpf_trampoline { 590 /* hlist for trampoline_table */ 591 struct hlist_node hlist; 592 /* serializes access to fields of this trampoline */ 593 struct mutex mutex; 594 refcount_t refcnt; 595 u64 key; 596 struct { 597 struct btf_func_model model; 598 void *addr; 599 bool ftrace_managed; 600 } func; 601 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 602 * program by replacing one of its functions. func.addr is the address 603 * of the function it replaced. 604 */ 605 struct bpf_prog *extension_prog; 606 /* list of BPF programs using this trampoline */ 607 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 608 /* Number of attached programs. A counter per kind. */ 609 int progs_cnt[BPF_TRAMP_MAX]; 610 /* Executable image of trampoline */ 611 void *image; 612 u64 selector; 613 struct bpf_ksym ksym; 614 }; 615 616 struct bpf_attach_target_info { 617 struct btf_func_model fmodel; 618 long tgt_addr; 619 const char *tgt_name; 620 const struct btf_type *tgt_type; 621 }; 622 623 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 624 625 struct bpf_dispatcher_prog { 626 struct bpf_prog *prog; 627 refcount_t users; 628 }; 629 630 struct bpf_dispatcher { 631 /* dispatcher mutex */ 632 struct mutex mutex; 633 void *func; 634 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 635 int num_progs; 636 void *image; 637 u32 image_off; 638 struct bpf_ksym ksym; 639 }; 640 641 static __always_inline unsigned int bpf_dispatcher_nop_func( 642 const void *ctx, 643 const struct bpf_insn *insnsi, 644 unsigned int (*bpf_func)(const void *, 645 const struct bpf_insn *)) 646 { 647 return bpf_func(ctx, insnsi); 648 } 649 #ifdef CONFIG_BPF_JIT 650 int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 651 int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 652 struct bpf_trampoline *bpf_trampoline_get(u64 key, 653 struct bpf_attach_target_info *tgt_info); 654 void bpf_trampoline_put(struct bpf_trampoline *tr); 655 #define BPF_DISPATCHER_INIT(_name) { \ 656 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 657 .func = &_name##_func, \ 658 .progs = {}, \ 659 .num_progs = 0, \ 660 .image = NULL, \ 661 .image_off = 0, \ 662 .ksym = { \ 663 .name = #_name, \ 664 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 665 }, \ 666 } 667 668 #define DEFINE_BPF_DISPATCHER(name) \ 669 noinline unsigned int bpf_dispatcher_##name##_func( \ 670 const void *ctx, \ 671 const struct bpf_insn *insnsi, \ 672 unsigned int (*bpf_func)(const void *, \ 673 const struct bpf_insn *)) \ 674 { \ 675 return bpf_func(ctx, insnsi); \ 676 } \ 677 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 678 struct bpf_dispatcher bpf_dispatcher_##name = \ 679 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 680 #define DECLARE_BPF_DISPATCHER(name) \ 681 unsigned int bpf_dispatcher_##name##_func( \ 682 const void *ctx, \ 683 const struct bpf_insn *insnsi, \ 684 unsigned int (*bpf_func)(const void *, \ 685 const struct bpf_insn *)); \ 686 extern struct bpf_dispatcher bpf_dispatcher_##name; 687 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 688 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 689 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 690 struct bpf_prog *to); 691 /* Called only from JIT-enabled code, so there's no need for stubs. */ 692 void *bpf_jit_alloc_exec_page(void); 693 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym); 694 void bpf_image_ksym_del(struct bpf_ksym *ksym); 695 void bpf_ksym_add(struct bpf_ksym *ksym); 696 void bpf_ksym_del(struct bpf_ksym *ksym); 697 #else 698 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog, 699 struct bpf_trampoline *tr) 700 { 701 return -ENOTSUPP; 702 } 703 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog, 704 struct bpf_trampoline *tr) 705 { 706 return -ENOTSUPP; 707 } 708 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 709 struct bpf_attach_target_info *tgt_info) 710 { 711 return ERR_PTR(-EOPNOTSUPP); 712 } 713 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 714 #define DEFINE_BPF_DISPATCHER(name) 715 #define DECLARE_BPF_DISPATCHER(name) 716 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 717 #define BPF_DISPATCHER_PTR(name) NULL 718 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 719 struct bpf_prog *from, 720 struct bpf_prog *to) {} 721 static inline bool is_bpf_image_address(unsigned long address) 722 { 723 return false; 724 } 725 #endif 726 727 struct bpf_func_info_aux { 728 u16 linkage; 729 bool unreliable; 730 }; 731 732 enum bpf_jit_poke_reason { 733 BPF_POKE_REASON_TAIL_CALL, 734 }; 735 736 /* Descriptor of pokes pointing /into/ the JITed image. */ 737 struct bpf_jit_poke_descriptor { 738 void *tailcall_target; 739 void *tailcall_bypass; 740 void *bypass_addr; 741 union { 742 struct { 743 struct bpf_map *map; 744 u32 key; 745 } tail_call; 746 }; 747 bool tailcall_target_stable; 748 u8 adj_off; 749 u16 reason; 750 u32 insn_idx; 751 }; 752 753 /* reg_type info for ctx arguments */ 754 struct bpf_ctx_arg_aux { 755 u32 offset; 756 enum bpf_reg_type reg_type; 757 u32 btf_id; 758 }; 759 760 struct bpf_prog_aux { 761 atomic64_t refcnt; 762 u32 used_map_cnt; 763 u32 max_ctx_offset; 764 u32 max_pkt_offset; 765 u32 max_tp_access; 766 u32 stack_depth; 767 u32 id; 768 u32 func_cnt; /* used by non-func prog as the number of func progs */ 769 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 770 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 771 u32 ctx_arg_info_size; 772 u32 max_rdonly_access; 773 u32 max_rdwr_access; 774 const struct bpf_ctx_arg_aux *ctx_arg_info; 775 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 776 struct bpf_prog *dst_prog; 777 struct bpf_trampoline *dst_trampoline; 778 enum bpf_prog_type saved_dst_prog_type; 779 enum bpf_attach_type saved_dst_attach_type; 780 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 781 bool offload_requested; 782 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 783 bool func_proto_unreliable; 784 bool sleepable; 785 bool tail_call_reachable; 786 enum bpf_tramp_prog_type trampoline_prog_type; 787 struct hlist_node tramp_hlist; 788 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 789 const struct btf_type *attach_func_proto; 790 /* function name for valid attach_btf_id */ 791 const char *attach_func_name; 792 struct bpf_prog **func; 793 void *jit_data; /* JIT specific data. arch dependent */ 794 struct bpf_jit_poke_descriptor *poke_tab; 795 u32 size_poke_tab; 796 struct bpf_ksym ksym; 797 const struct bpf_prog_ops *ops; 798 struct bpf_map **used_maps; 799 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 800 struct bpf_prog *prog; 801 struct user_struct *user; 802 u64 load_time; /* ns since boottime */ 803 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 804 char name[BPF_OBJ_NAME_LEN]; 805 #ifdef CONFIG_SECURITY 806 void *security; 807 #endif 808 struct bpf_prog_offload *offload; 809 struct btf *btf; 810 struct bpf_func_info *func_info; 811 struct bpf_func_info_aux *func_info_aux; 812 /* bpf_line_info loaded from userspace. linfo->insn_off 813 * has the xlated insn offset. 814 * Both the main and sub prog share the same linfo. 815 * The subprog can access its first linfo by 816 * using the linfo_idx. 817 */ 818 struct bpf_line_info *linfo; 819 /* jited_linfo is the jited addr of the linfo. It has a 820 * one to one mapping to linfo: 821 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 822 * Both the main and sub prog share the same jited_linfo. 823 * The subprog can access its first jited_linfo by 824 * using the linfo_idx. 825 */ 826 void **jited_linfo; 827 u32 func_info_cnt; 828 u32 nr_linfo; 829 /* subprog can use linfo_idx to access its first linfo and 830 * jited_linfo. 831 * main prog always has linfo_idx == 0 832 */ 833 u32 linfo_idx; 834 u32 num_exentries; 835 struct exception_table_entry *extable; 836 struct bpf_prog_stats __percpu *stats; 837 union { 838 struct work_struct work; 839 struct rcu_head rcu; 840 }; 841 }; 842 843 struct bpf_array_aux { 844 /* 'Ownership' of prog array is claimed by the first program that 845 * is going to use this map or by the first program which FD is 846 * stored in the map to make sure that all callers and callees have 847 * the same prog type and JITed flag. 848 */ 849 enum bpf_prog_type type; 850 bool jited; 851 /* Programs with direct jumps into programs part of this array. */ 852 struct list_head poke_progs; 853 struct bpf_map *map; 854 struct mutex poke_mutex; 855 struct work_struct work; 856 }; 857 858 struct bpf_link { 859 atomic64_t refcnt; 860 u32 id; 861 enum bpf_link_type type; 862 const struct bpf_link_ops *ops; 863 struct bpf_prog *prog; 864 struct work_struct work; 865 }; 866 867 struct bpf_link_ops { 868 void (*release)(struct bpf_link *link); 869 void (*dealloc)(struct bpf_link *link); 870 int (*detach)(struct bpf_link *link); 871 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 872 struct bpf_prog *old_prog); 873 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 874 int (*fill_link_info)(const struct bpf_link *link, 875 struct bpf_link_info *info); 876 }; 877 878 struct bpf_link_primer { 879 struct bpf_link *link; 880 struct file *file; 881 int fd; 882 u32 id; 883 }; 884 885 struct bpf_struct_ops_value; 886 struct btf_type; 887 struct btf_member; 888 889 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 890 struct bpf_struct_ops { 891 const struct bpf_verifier_ops *verifier_ops; 892 int (*init)(struct btf *btf); 893 int (*check_member)(const struct btf_type *t, 894 const struct btf_member *member); 895 int (*init_member)(const struct btf_type *t, 896 const struct btf_member *member, 897 void *kdata, const void *udata); 898 int (*reg)(void *kdata); 899 void (*unreg)(void *kdata); 900 const struct btf_type *type; 901 const struct btf_type *value_type; 902 const char *name; 903 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 904 u32 type_id; 905 u32 value_id; 906 }; 907 908 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 909 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 910 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id); 911 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log); 912 bool bpf_struct_ops_get(const void *kdata); 913 void bpf_struct_ops_put(const void *kdata); 914 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 915 void *value); 916 static inline bool bpf_try_module_get(const void *data, struct module *owner) 917 { 918 if (owner == BPF_MODULE_OWNER) 919 return bpf_struct_ops_get(data); 920 else 921 return try_module_get(owner); 922 } 923 static inline void bpf_module_put(const void *data, struct module *owner) 924 { 925 if (owner == BPF_MODULE_OWNER) 926 bpf_struct_ops_put(data); 927 else 928 module_put(owner); 929 } 930 #else 931 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id) 932 { 933 return NULL; 934 } 935 static inline void bpf_struct_ops_init(struct btf *btf, 936 struct bpf_verifier_log *log) 937 { 938 } 939 static inline bool bpf_try_module_get(const void *data, struct module *owner) 940 { 941 return try_module_get(owner); 942 } 943 static inline void bpf_module_put(const void *data, struct module *owner) 944 { 945 module_put(owner); 946 } 947 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 948 void *key, 949 void *value) 950 { 951 return -EINVAL; 952 } 953 #endif 954 955 struct bpf_array { 956 struct bpf_map map; 957 u32 elem_size; 958 u32 index_mask; 959 struct bpf_array_aux *aux; 960 union { 961 char value[0] __aligned(8); 962 void *ptrs[0] __aligned(8); 963 void __percpu *pptrs[0] __aligned(8); 964 }; 965 }; 966 967 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 968 #define MAX_TAIL_CALL_CNT 32 969 970 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 971 BPF_F_RDONLY_PROG | \ 972 BPF_F_WRONLY | \ 973 BPF_F_WRONLY_PROG) 974 975 #define BPF_MAP_CAN_READ BIT(0) 976 #define BPF_MAP_CAN_WRITE BIT(1) 977 978 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 979 { 980 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 981 982 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 983 * not possible. 984 */ 985 if (access_flags & BPF_F_RDONLY_PROG) 986 return BPF_MAP_CAN_READ; 987 else if (access_flags & BPF_F_WRONLY_PROG) 988 return BPF_MAP_CAN_WRITE; 989 else 990 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 991 } 992 993 static inline bool bpf_map_flags_access_ok(u32 access_flags) 994 { 995 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 996 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 997 } 998 999 struct bpf_event_entry { 1000 struct perf_event *event; 1001 struct file *perf_file; 1002 struct file *map_file; 1003 struct rcu_head rcu; 1004 }; 1005 1006 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); 1007 int bpf_prog_calc_tag(struct bpf_prog *fp); 1008 const char *kernel_type_name(u32 btf_type_id); 1009 1010 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 1011 1012 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 1013 unsigned long off, unsigned long len); 1014 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 1015 const struct bpf_insn *src, 1016 struct bpf_insn *dst, 1017 struct bpf_prog *prog, 1018 u32 *target_size); 1019 1020 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 1021 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 1022 1023 /* an array of programs to be executed under rcu_lock. 1024 * 1025 * Typical usage: 1026 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN); 1027 * 1028 * the structure returned by bpf_prog_array_alloc() should be populated 1029 * with program pointers and the last pointer must be NULL. 1030 * The user has to keep refcnt on the program and make sure the program 1031 * is removed from the array before bpf_prog_put(). 1032 * The 'struct bpf_prog_array *' should only be replaced with xchg() 1033 * since other cpus are walking the array of pointers in parallel. 1034 */ 1035 struct bpf_prog_array_item { 1036 struct bpf_prog *prog; 1037 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1038 }; 1039 1040 struct bpf_prog_array { 1041 struct rcu_head rcu; 1042 struct bpf_prog_array_item items[]; 1043 }; 1044 1045 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 1046 void bpf_prog_array_free(struct bpf_prog_array *progs); 1047 int bpf_prog_array_length(struct bpf_prog_array *progs); 1048 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 1049 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 1050 __u32 __user *prog_ids, u32 cnt); 1051 1052 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 1053 struct bpf_prog *old_prog); 1054 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 1055 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 1056 struct bpf_prog *prog); 1057 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 1058 u32 *prog_ids, u32 request_cnt, 1059 u32 *prog_cnt); 1060 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 1061 struct bpf_prog *exclude_prog, 1062 struct bpf_prog *include_prog, 1063 struct bpf_prog_array **new_array); 1064 1065 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \ 1066 ({ \ 1067 struct bpf_prog_array_item *_item; \ 1068 struct bpf_prog *_prog; \ 1069 struct bpf_prog_array *_array; \ 1070 u32 _ret = 1; \ 1071 migrate_disable(); \ 1072 rcu_read_lock(); \ 1073 _array = rcu_dereference(array); \ 1074 if (unlikely(check_non_null && !_array))\ 1075 goto _out; \ 1076 _item = &_array->items[0]; \ 1077 while ((_prog = READ_ONCE(_item->prog))) { \ 1078 bpf_cgroup_storage_set(_item->cgroup_storage); \ 1079 _ret &= func(_prog, ctx); \ 1080 _item++; \ 1081 } \ 1082 _out: \ 1083 rcu_read_unlock(); \ 1084 migrate_enable(); \ 1085 _ret; \ 1086 }) 1087 1088 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs 1089 * so BPF programs can request cwr for TCP packets. 1090 * 1091 * Current cgroup skb programs can only return 0 or 1 (0 to drop the 1092 * packet. This macro changes the behavior so the low order bit 1093 * indicates whether the packet should be dropped (0) or not (1) 1094 * and the next bit is a congestion notification bit. This could be 1095 * used by TCP to call tcp_enter_cwr() 1096 * 1097 * Hence, new allowed return values of CGROUP EGRESS BPF programs are: 1098 * 0: drop packet 1099 * 1: keep packet 1100 * 2: drop packet and cn 1101 * 3: keep packet and cn 1102 * 1103 * This macro then converts it to one of the NET_XMIT or an error 1104 * code that is then interpreted as drop packet (and no cn): 1105 * 0: NET_XMIT_SUCCESS skb should be transmitted 1106 * 1: NET_XMIT_DROP skb should be dropped and cn 1107 * 2: NET_XMIT_CN skb should be transmitted and cn 1108 * 3: -EPERM skb should be dropped 1109 */ 1110 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \ 1111 ({ \ 1112 struct bpf_prog_array_item *_item; \ 1113 struct bpf_prog *_prog; \ 1114 struct bpf_prog_array *_array; \ 1115 u32 ret; \ 1116 u32 _ret = 1; \ 1117 u32 _cn = 0; \ 1118 migrate_disable(); \ 1119 rcu_read_lock(); \ 1120 _array = rcu_dereference(array); \ 1121 _item = &_array->items[0]; \ 1122 while ((_prog = READ_ONCE(_item->prog))) { \ 1123 bpf_cgroup_storage_set(_item->cgroup_storage); \ 1124 ret = func(_prog, ctx); \ 1125 _ret &= (ret & 1); \ 1126 _cn |= (ret & 2); \ 1127 _item++; \ 1128 } \ 1129 rcu_read_unlock(); \ 1130 migrate_enable(); \ 1131 if (_ret) \ 1132 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \ 1133 else \ 1134 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \ 1135 _ret; \ 1136 }) 1137 1138 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \ 1139 __BPF_PROG_RUN_ARRAY(array, ctx, func, false) 1140 1141 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \ 1142 __BPF_PROG_RUN_ARRAY(array, ctx, func, true) 1143 1144 #ifdef CONFIG_BPF_SYSCALL 1145 DECLARE_PER_CPU(int, bpf_prog_active); 1146 extern struct mutex bpf_stats_enabled_mutex; 1147 1148 /* 1149 * Block execution of BPF programs attached to instrumentation (perf, 1150 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 1151 * these events can happen inside a region which holds a map bucket lock 1152 * and can deadlock on it. 1153 * 1154 * Use the preemption safe inc/dec variants on RT because migrate disable 1155 * is preemptible on RT and preemption in the middle of the RMW operation 1156 * might lead to inconsistent state. Use the raw variants for non RT 1157 * kernels as migrate_disable() maps to preempt_disable() so the slightly 1158 * more expensive save operation can be avoided. 1159 */ 1160 static inline void bpf_disable_instrumentation(void) 1161 { 1162 migrate_disable(); 1163 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1164 this_cpu_inc(bpf_prog_active); 1165 else 1166 __this_cpu_inc(bpf_prog_active); 1167 } 1168 1169 static inline void bpf_enable_instrumentation(void) 1170 { 1171 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1172 this_cpu_dec(bpf_prog_active); 1173 else 1174 __this_cpu_dec(bpf_prog_active); 1175 migrate_enable(); 1176 } 1177 1178 extern const struct file_operations bpf_map_fops; 1179 extern const struct file_operations bpf_prog_fops; 1180 extern const struct file_operations bpf_iter_fops; 1181 1182 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 1183 extern const struct bpf_prog_ops _name ## _prog_ops; \ 1184 extern const struct bpf_verifier_ops _name ## _verifier_ops; 1185 #define BPF_MAP_TYPE(_id, _ops) \ 1186 extern const struct bpf_map_ops _ops; 1187 #define BPF_LINK_TYPE(_id, _name) 1188 #include <linux/bpf_types.h> 1189 #undef BPF_PROG_TYPE 1190 #undef BPF_MAP_TYPE 1191 #undef BPF_LINK_TYPE 1192 1193 extern const struct bpf_prog_ops bpf_offload_prog_ops; 1194 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 1195 extern const struct bpf_verifier_ops xdp_analyzer_ops; 1196 1197 struct bpf_prog *bpf_prog_get(u32 ufd); 1198 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 1199 bool attach_drv); 1200 void bpf_prog_add(struct bpf_prog *prog, int i); 1201 void bpf_prog_sub(struct bpf_prog *prog, int i); 1202 void bpf_prog_inc(struct bpf_prog *prog); 1203 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 1204 void bpf_prog_put(struct bpf_prog *prog); 1205 int __bpf_prog_charge(struct user_struct *user, u32 pages); 1206 void __bpf_prog_uncharge(struct user_struct *user, u32 pages); 1207 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 1208 struct bpf_map **used_maps, u32 len); 1209 1210 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 1211 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 1212 1213 struct bpf_map *bpf_map_get(u32 ufd); 1214 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 1215 struct bpf_map *__bpf_map_get(struct fd f); 1216 void bpf_map_inc(struct bpf_map *map); 1217 void bpf_map_inc_with_uref(struct bpf_map *map); 1218 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 1219 void bpf_map_put_with_uref(struct bpf_map *map); 1220 void bpf_map_put(struct bpf_map *map); 1221 int bpf_map_charge_memlock(struct bpf_map *map, u32 pages); 1222 void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages); 1223 int bpf_map_charge_init(struct bpf_map_memory *mem, u64 size); 1224 void bpf_map_charge_finish(struct bpf_map_memory *mem); 1225 void bpf_map_charge_move(struct bpf_map_memory *dst, 1226 struct bpf_map_memory *src); 1227 void *bpf_map_area_alloc(u64 size, int numa_node); 1228 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 1229 void bpf_map_area_free(void *base); 1230 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 1231 int generic_map_lookup_batch(struct bpf_map *map, 1232 const union bpf_attr *attr, 1233 union bpf_attr __user *uattr); 1234 int generic_map_update_batch(struct bpf_map *map, 1235 const union bpf_attr *attr, 1236 union bpf_attr __user *uattr); 1237 int generic_map_delete_batch(struct bpf_map *map, 1238 const union bpf_attr *attr, 1239 union bpf_attr __user *uattr); 1240 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 1241 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 1242 1243 extern int sysctl_unprivileged_bpf_disabled; 1244 1245 static inline bool bpf_allow_ptr_leaks(void) 1246 { 1247 return perfmon_capable(); 1248 } 1249 1250 static inline bool bpf_allow_ptr_to_map_access(void) 1251 { 1252 return perfmon_capable(); 1253 } 1254 1255 static inline bool bpf_bypass_spec_v1(void) 1256 { 1257 return perfmon_capable(); 1258 } 1259 1260 static inline bool bpf_bypass_spec_v4(void) 1261 { 1262 return perfmon_capable(); 1263 } 1264 1265 int bpf_map_new_fd(struct bpf_map *map, int flags); 1266 int bpf_prog_new_fd(struct bpf_prog *prog); 1267 1268 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1269 const struct bpf_link_ops *ops, struct bpf_prog *prog); 1270 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 1271 int bpf_link_settle(struct bpf_link_primer *primer); 1272 void bpf_link_cleanup(struct bpf_link_primer *primer); 1273 void bpf_link_inc(struct bpf_link *link); 1274 void bpf_link_put(struct bpf_link *link); 1275 int bpf_link_new_fd(struct bpf_link *link); 1276 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd); 1277 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 1278 1279 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 1280 int bpf_obj_get_user(const char __user *pathname, int flags); 1281 1282 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 1283 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 1284 extern int bpf_iter_ ## target(args); \ 1285 int __init bpf_iter_ ## target(args) { return 0; } 1286 1287 struct bpf_iter_aux_info { 1288 struct bpf_map *map; 1289 }; 1290 1291 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 1292 union bpf_iter_link_info *linfo, 1293 struct bpf_iter_aux_info *aux); 1294 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 1295 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 1296 struct seq_file *seq); 1297 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 1298 struct bpf_link_info *info); 1299 1300 enum bpf_iter_feature { 1301 BPF_ITER_RESCHED = BIT(0), 1302 }; 1303 1304 #define BPF_ITER_CTX_ARG_MAX 2 1305 struct bpf_iter_reg { 1306 const char *target; 1307 bpf_iter_attach_target_t attach_target; 1308 bpf_iter_detach_target_t detach_target; 1309 bpf_iter_show_fdinfo_t show_fdinfo; 1310 bpf_iter_fill_link_info_t fill_link_info; 1311 u32 ctx_arg_info_size; 1312 u32 feature; 1313 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 1314 const struct bpf_iter_seq_info *seq_info; 1315 }; 1316 1317 struct bpf_iter_meta { 1318 __bpf_md_ptr(struct seq_file *, seq); 1319 u64 session_id; 1320 u64 seq_num; 1321 }; 1322 1323 struct bpf_iter__bpf_map_elem { 1324 __bpf_md_ptr(struct bpf_iter_meta *, meta); 1325 __bpf_md_ptr(struct bpf_map *, map); 1326 __bpf_md_ptr(void *, key); 1327 __bpf_md_ptr(void *, value); 1328 }; 1329 1330 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 1331 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 1332 bool bpf_iter_prog_supported(struct bpf_prog *prog); 1333 int bpf_iter_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 1334 int bpf_iter_new_fd(struct bpf_link *link); 1335 bool bpf_link_is_iter(struct bpf_link *link); 1336 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 1337 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 1338 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 1339 struct seq_file *seq); 1340 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 1341 struct bpf_link_info *info); 1342 1343 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 1344 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 1345 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 1346 u64 flags); 1347 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 1348 u64 flags); 1349 1350 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 1351 1352 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 1353 void *key, void *value, u64 map_flags); 1354 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1355 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 1356 void *key, void *value, u64 map_flags); 1357 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1358 1359 int bpf_get_file_flag(int flags); 1360 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size, 1361 size_t actual_size); 1362 1363 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 1364 * forced to use 'long' read/writes to try to atomically copy long counters. 1365 * Best-effort only. No barriers here, since it _will_ race with concurrent 1366 * updates from BPF programs. Called from bpf syscall and mostly used with 1367 * size 8 or 16 bytes, so ask compiler to inline it. 1368 */ 1369 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 1370 { 1371 const long *lsrc = src; 1372 long *ldst = dst; 1373 1374 size /= sizeof(long); 1375 while (size--) 1376 *ldst++ = *lsrc++; 1377 } 1378 1379 /* verify correctness of eBPF program */ 1380 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, 1381 union bpf_attr __user *uattr); 1382 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 1383 1384 struct btf *bpf_get_btf_vmlinux(void); 1385 1386 /* Map specifics */ 1387 struct xdp_buff; 1388 struct sk_buff; 1389 1390 struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key); 1391 struct bpf_dtab_netdev *__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key); 1392 void __dev_flush(void); 1393 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1394 struct net_device *dev_rx); 1395 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1396 struct net_device *dev_rx); 1397 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 1398 struct bpf_prog *xdp_prog); 1399 bool dev_map_can_have_prog(struct bpf_map *map); 1400 1401 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key); 1402 void __cpu_map_flush(void); 1403 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, 1404 struct net_device *dev_rx); 1405 bool cpu_map_prog_allowed(struct bpf_map *map); 1406 1407 /* Return map's numa specified by userspace */ 1408 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 1409 { 1410 return (attr->map_flags & BPF_F_NUMA_NODE) ? 1411 attr->numa_node : NUMA_NO_NODE; 1412 } 1413 1414 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 1415 int array_map_alloc_check(union bpf_attr *attr); 1416 1417 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 1418 union bpf_attr __user *uattr); 1419 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 1420 union bpf_attr __user *uattr); 1421 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1422 const union bpf_attr *kattr, 1423 union bpf_attr __user *uattr); 1424 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1425 const union bpf_attr *kattr, 1426 union bpf_attr __user *uattr); 1427 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 1428 const union bpf_attr *kattr, 1429 union bpf_attr __user *uattr); 1430 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 1431 const struct bpf_prog *prog, 1432 struct bpf_insn_access_aux *info); 1433 int btf_struct_access(struct bpf_verifier_log *log, 1434 const struct btf_type *t, int off, int size, 1435 enum bpf_access_type atype, 1436 u32 *next_btf_id); 1437 bool btf_struct_ids_match(struct bpf_verifier_log *log, 1438 int off, u32 id, u32 need_type_id); 1439 1440 int btf_distill_func_proto(struct bpf_verifier_log *log, 1441 struct btf *btf, 1442 const struct btf_type *func_proto, 1443 const char *func_name, 1444 struct btf_func_model *m); 1445 1446 struct bpf_reg_state; 1447 int btf_check_func_arg_match(struct bpf_verifier_env *env, int subprog, 1448 struct bpf_reg_state *regs); 1449 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, 1450 struct bpf_reg_state *reg); 1451 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 1452 struct btf *btf, const struct btf_type *t); 1453 1454 struct bpf_prog *bpf_prog_by_id(u32 id); 1455 struct bpf_link *bpf_link_by_id(u32 id); 1456 1457 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id); 1458 #else /* !CONFIG_BPF_SYSCALL */ 1459 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 1460 { 1461 return ERR_PTR(-EOPNOTSUPP); 1462 } 1463 1464 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 1465 enum bpf_prog_type type, 1466 bool attach_drv) 1467 { 1468 return ERR_PTR(-EOPNOTSUPP); 1469 } 1470 1471 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 1472 { 1473 } 1474 1475 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 1476 { 1477 } 1478 1479 static inline void bpf_prog_put(struct bpf_prog *prog) 1480 { 1481 } 1482 1483 static inline void bpf_prog_inc(struct bpf_prog *prog) 1484 { 1485 } 1486 1487 static inline struct bpf_prog *__must_check 1488 bpf_prog_inc_not_zero(struct bpf_prog *prog) 1489 { 1490 return ERR_PTR(-EOPNOTSUPP); 1491 } 1492 1493 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages) 1494 { 1495 return 0; 1496 } 1497 1498 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages) 1499 { 1500 } 1501 1502 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1503 const struct bpf_link_ops *ops, 1504 struct bpf_prog *prog) 1505 { 1506 } 1507 1508 static inline int bpf_link_prime(struct bpf_link *link, 1509 struct bpf_link_primer *primer) 1510 { 1511 return -EOPNOTSUPP; 1512 } 1513 1514 static inline int bpf_link_settle(struct bpf_link_primer *primer) 1515 { 1516 return -EOPNOTSUPP; 1517 } 1518 1519 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 1520 { 1521 } 1522 1523 static inline void bpf_link_inc(struct bpf_link *link) 1524 { 1525 } 1526 1527 static inline void bpf_link_put(struct bpf_link *link) 1528 { 1529 } 1530 1531 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 1532 { 1533 return -EOPNOTSUPP; 1534 } 1535 1536 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map, 1537 u32 key) 1538 { 1539 return NULL; 1540 } 1541 1542 static inline struct net_device *__dev_map_hash_lookup_elem(struct bpf_map *map, 1543 u32 key) 1544 { 1545 return NULL; 1546 } 1547 static inline bool dev_map_can_have_prog(struct bpf_map *map) 1548 { 1549 return false; 1550 } 1551 1552 static inline void __dev_flush(void) 1553 { 1554 } 1555 1556 struct xdp_buff; 1557 struct bpf_dtab_netdev; 1558 1559 static inline 1560 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1561 struct net_device *dev_rx) 1562 { 1563 return 0; 1564 } 1565 1566 static inline 1567 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1568 struct net_device *dev_rx) 1569 { 1570 return 0; 1571 } 1572 1573 struct sk_buff; 1574 1575 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 1576 struct sk_buff *skb, 1577 struct bpf_prog *xdp_prog) 1578 { 1579 return 0; 1580 } 1581 1582 static inline 1583 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key) 1584 { 1585 return NULL; 1586 } 1587 1588 static inline void __cpu_map_flush(void) 1589 { 1590 } 1591 1592 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 1593 struct xdp_buff *xdp, 1594 struct net_device *dev_rx) 1595 { 1596 return 0; 1597 } 1598 1599 static inline bool cpu_map_prog_allowed(struct bpf_map *map) 1600 { 1601 return false; 1602 } 1603 1604 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 1605 enum bpf_prog_type type) 1606 { 1607 return ERR_PTR(-EOPNOTSUPP); 1608 } 1609 1610 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 1611 const union bpf_attr *kattr, 1612 union bpf_attr __user *uattr) 1613 { 1614 return -ENOTSUPP; 1615 } 1616 1617 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 1618 const union bpf_attr *kattr, 1619 union bpf_attr __user *uattr) 1620 { 1621 return -ENOTSUPP; 1622 } 1623 1624 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1625 const union bpf_attr *kattr, 1626 union bpf_attr __user *uattr) 1627 { 1628 return -ENOTSUPP; 1629 } 1630 1631 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1632 const union bpf_attr *kattr, 1633 union bpf_attr __user *uattr) 1634 { 1635 return -ENOTSUPP; 1636 } 1637 1638 static inline void bpf_map_put(struct bpf_map *map) 1639 { 1640 } 1641 1642 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 1643 { 1644 return ERR_PTR(-ENOTSUPP); 1645 } 1646 1647 static inline const struct bpf_func_proto * 1648 bpf_base_func_proto(enum bpf_func_id func_id) 1649 { 1650 return NULL; 1651 } 1652 #endif /* CONFIG_BPF_SYSCALL */ 1653 1654 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 1655 enum bpf_prog_type type) 1656 { 1657 return bpf_prog_get_type_dev(ufd, type, false); 1658 } 1659 1660 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 1661 1662 int bpf_prog_offload_compile(struct bpf_prog *prog); 1663 void bpf_prog_offload_destroy(struct bpf_prog *prog); 1664 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 1665 struct bpf_prog *prog); 1666 1667 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 1668 1669 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 1670 int bpf_map_offload_update_elem(struct bpf_map *map, 1671 void *key, void *value, u64 flags); 1672 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 1673 int bpf_map_offload_get_next_key(struct bpf_map *map, 1674 void *key, void *next_key); 1675 1676 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 1677 1678 struct bpf_offload_dev * 1679 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 1680 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 1681 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 1682 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 1683 struct net_device *netdev); 1684 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 1685 struct net_device *netdev); 1686 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 1687 1688 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 1689 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 1690 1691 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 1692 { 1693 return aux->offload_requested; 1694 } 1695 1696 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1697 { 1698 return unlikely(map->ops == &bpf_map_offload_ops); 1699 } 1700 1701 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 1702 void bpf_map_offload_map_free(struct bpf_map *map); 1703 #else 1704 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 1705 union bpf_attr *attr) 1706 { 1707 return -EOPNOTSUPP; 1708 } 1709 1710 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 1711 { 1712 return false; 1713 } 1714 1715 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1716 { 1717 return false; 1718 } 1719 1720 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 1721 { 1722 return ERR_PTR(-EOPNOTSUPP); 1723 } 1724 1725 static inline void bpf_map_offload_map_free(struct bpf_map *map) 1726 { 1727 } 1728 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 1729 1730 #if defined(CONFIG_BPF_STREAM_PARSER) 1731 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, 1732 struct bpf_prog *old, u32 which); 1733 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 1734 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 1735 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 1736 void sock_map_unhash(struct sock *sk); 1737 void sock_map_close(struct sock *sk, long timeout); 1738 #else 1739 static inline int sock_map_prog_update(struct bpf_map *map, 1740 struct bpf_prog *prog, 1741 struct bpf_prog *old, u32 which) 1742 { 1743 return -EOPNOTSUPP; 1744 } 1745 1746 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 1747 struct bpf_prog *prog) 1748 { 1749 return -EINVAL; 1750 } 1751 1752 static inline int sock_map_prog_detach(const union bpf_attr *attr, 1753 enum bpf_prog_type ptype) 1754 { 1755 return -EOPNOTSUPP; 1756 } 1757 1758 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 1759 u64 flags) 1760 { 1761 return -EOPNOTSUPP; 1762 } 1763 #endif /* CONFIG_BPF_STREAM_PARSER */ 1764 1765 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 1766 void bpf_sk_reuseport_detach(struct sock *sk); 1767 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 1768 void *value); 1769 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 1770 void *value, u64 map_flags); 1771 #else 1772 static inline void bpf_sk_reuseport_detach(struct sock *sk) 1773 { 1774 } 1775 1776 #ifdef CONFIG_BPF_SYSCALL 1777 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 1778 void *key, void *value) 1779 { 1780 return -EOPNOTSUPP; 1781 } 1782 1783 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 1784 void *key, void *value, 1785 u64 map_flags) 1786 { 1787 return -EOPNOTSUPP; 1788 } 1789 #endif /* CONFIG_BPF_SYSCALL */ 1790 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 1791 1792 /* verifier prototypes for helper functions called from eBPF programs */ 1793 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 1794 extern const struct bpf_func_proto bpf_map_update_elem_proto; 1795 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 1796 extern const struct bpf_func_proto bpf_map_push_elem_proto; 1797 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 1798 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 1799 1800 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 1801 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 1802 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 1803 extern const struct bpf_func_proto bpf_tail_call_proto; 1804 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 1805 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 1806 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 1807 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 1808 extern const struct bpf_func_proto bpf_get_current_comm_proto; 1809 extern const struct bpf_func_proto bpf_get_stackid_proto; 1810 extern const struct bpf_func_proto bpf_get_stack_proto; 1811 extern const struct bpf_func_proto bpf_get_task_stack_proto; 1812 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 1813 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 1814 extern const struct bpf_func_proto bpf_sock_map_update_proto; 1815 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 1816 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 1817 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 1818 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 1819 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 1820 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 1821 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 1822 extern const struct bpf_func_proto bpf_spin_lock_proto; 1823 extern const struct bpf_func_proto bpf_spin_unlock_proto; 1824 extern const struct bpf_func_proto bpf_get_local_storage_proto; 1825 extern const struct bpf_func_proto bpf_strtol_proto; 1826 extern const struct bpf_func_proto bpf_strtoul_proto; 1827 extern const struct bpf_func_proto bpf_tcp_sock_proto; 1828 extern const struct bpf_func_proto bpf_jiffies64_proto; 1829 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 1830 extern const struct bpf_func_proto bpf_event_output_data_proto; 1831 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 1832 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 1833 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 1834 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 1835 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 1836 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 1837 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 1838 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 1839 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 1840 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 1841 extern const struct bpf_func_proto bpf_copy_from_user_proto; 1842 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 1843 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 1844 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 1845 1846 const struct bpf_func_proto *bpf_tracing_func_proto( 1847 enum bpf_func_id func_id, const struct bpf_prog *prog); 1848 1849 const struct bpf_func_proto *tracing_prog_func_proto( 1850 enum bpf_func_id func_id, const struct bpf_prog *prog); 1851 1852 /* Shared helpers among cBPF and eBPF. */ 1853 void bpf_user_rnd_init_once(void); 1854 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 1855 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 1856 1857 #if defined(CONFIG_NET) 1858 bool bpf_sock_common_is_valid_access(int off, int size, 1859 enum bpf_access_type type, 1860 struct bpf_insn_access_aux *info); 1861 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1862 struct bpf_insn_access_aux *info); 1863 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 1864 const struct bpf_insn *si, 1865 struct bpf_insn *insn_buf, 1866 struct bpf_prog *prog, 1867 u32 *target_size); 1868 #else 1869 static inline bool bpf_sock_common_is_valid_access(int off, int size, 1870 enum bpf_access_type type, 1871 struct bpf_insn_access_aux *info) 1872 { 1873 return false; 1874 } 1875 static inline bool bpf_sock_is_valid_access(int off, int size, 1876 enum bpf_access_type type, 1877 struct bpf_insn_access_aux *info) 1878 { 1879 return false; 1880 } 1881 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 1882 const struct bpf_insn *si, 1883 struct bpf_insn *insn_buf, 1884 struct bpf_prog *prog, 1885 u32 *target_size) 1886 { 1887 return 0; 1888 } 1889 #endif 1890 1891 #ifdef CONFIG_INET 1892 struct sk_reuseport_kern { 1893 struct sk_buff *skb; 1894 struct sock *sk; 1895 struct sock *selected_sk; 1896 void *data_end; 1897 u32 hash; 1898 u32 reuseport_id; 1899 bool bind_inany; 1900 }; 1901 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1902 struct bpf_insn_access_aux *info); 1903 1904 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 1905 const struct bpf_insn *si, 1906 struct bpf_insn *insn_buf, 1907 struct bpf_prog *prog, 1908 u32 *target_size); 1909 1910 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1911 struct bpf_insn_access_aux *info); 1912 1913 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 1914 const struct bpf_insn *si, 1915 struct bpf_insn *insn_buf, 1916 struct bpf_prog *prog, 1917 u32 *target_size); 1918 #else 1919 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 1920 enum bpf_access_type type, 1921 struct bpf_insn_access_aux *info) 1922 { 1923 return false; 1924 } 1925 1926 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 1927 const struct bpf_insn *si, 1928 struct bpf_insn *insn_buf, 1929 struct bpf_prog *prog, 1930 u32 *target_size) 1931 { 1932 return 0; 1933 } 1934 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 1935 enum bpf_access_type type, 1936 struct bpf_insn_access_aux *info) 1937 { 1938 return false; 1939 } 1940 1941 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 1942 const struct bpf_insn *si, 1943 struct bpf_insn *insn_buf, 1944 struct bpf_prog *prog, 1945 u32 *target_size) 1946 { 1947 return 0; 1948 } 1949 #endif /* CONFIG_INET */ 1950 1951 enum bpf_text_poke_type { 1952 BPF_MOD_CALL, 1953 BPF_MOD_JUMP, 1954 }; 1955 1956 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 1957 void *addr1, void *addr2); 1958 1959 struct btf_id_set; 1960 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 1961 1962 #endif /* _LINUX_BPF_H */ 1963