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