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 #include <linux/sched/mm.h> 24 #include <linux/slab.h> 25 26 struct bpf_verifier_env; 27 struct bpf_verifier_log; 28 struct perf_event; 29 struct bpf_prog; 30 struct bpf_prog_aux; 31 struct bpf_map; 32 struct sock; 33 struct seq_file; 34 struct btf; 35 struct btf_type; 36 struct exception_table_entry; 37 struct seq_operations; 38 struct bpf_iter_aux_info; 39 struct bpf_local_storage; 40 struct bpf_local_storage_map; 41 struct kobject; 42 struct mem_cgroup; 43 44 extern struct idr btf_idr; 45 extern spinlock_t btf_idr_lock; 46 extern struct kobject *btf_kobj; 47 48 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, 49 struct bpf_iter_aux_info *aux); 50 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); 51 struct bpf_iter_seq_info { 52 const struct seq_operations *seq_ops; 53 bpf_iter_init_seq_priv_t init_seq_private; 54 bpf_iter_fini_seq_priv_t fini_seq_private; 55 u32 seq_priv_size; 56 }; 57 58 /* map is generic key/value storage optionally accesible by eBPF programs */ 59 struct bpf_map_ops { 60 /* funcs callable from userspace (via syscall) */ 61 int (*map_alloc_check)(union bpf_attr *attr); 62 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 63 void (*map_release)(struct bpf_map *map, struct file *map_file); 64 void (*map_free)(struct bpf_map *map); 65 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 66 void (*map_release_uref)(struct bpf_map *map); 67 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 68 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, 69 union bpf_attr __user *uattr); 70 int (*map_lookup_and_delete_batch)(struct bpf_map *map, 71 const union bpf_attr *attr, 72 union bpf_attr __user *uattr); 73 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr, 74 union bpf_attr __user *uattr); 75 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, 76 union bpf_attr __user *uattr); 77 78 /* funcs callable from userspace and from eBPF programs */ 79 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 80 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 81 int (*map_delete_elem)(struct bpf_map *map, void *key); 82 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 83 int (*map_pop_elem)(struct bpf_map *map, void *value); 84 int (*map_peek_elem)(struct bpf_map *map, void *value); 85 86 /* funcs called by prog_array and perf_event_array map */ 87 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 88 int fd); 89 void (*map_fd_put_ptr)(void *ptr); 90 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 91 u32 (*map_fd_sys_lookup_elem)(void *ptr); 92 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 93 struct seq_file *m); 94 int (*map_check_btf)(const struct bpf_map *map, 95 const struct btf *btf, 96 const struct btf_type *key_type, 97 const struct btf_type *value_type); 98 99 /* Prog poke tracking helpers. */ 100 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux); 101 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux); 102 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, 103 struct bpf_prog *new); 104 105 /* Direct value access helpers. */ 106 int (*map_direct_value_addr)(const struct bpf_map *map, 107 u64 *imm, u32 off); 108 int (*map_direct_value_meta)(const struct bpf_map *map, 109 u64 imm, u32 *off); 110 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma); 111 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, 112 struct poll_table_struct *pts); 113 114 /* Functions called by bpf_local_storage maps */ 115 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, 116 void *owner, u32 size); 117 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, 118 void *owner, u32 size); 119 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner); 120 121 /* map_meta_equal must be implemented for maps that can be 122 * used as an inner map. It is a runtime check to ensure 123 * an inner map can be inserted to an outer map. 124 * 125 * Some properties of the inner map has been used during the 126 * verification time. When inserting an inner map at the runtime, 127 * map_meta_equal has to ensure the inserting map has the same 128 * properties that the verifier has used earlier. 129 */ 130 bool (*map_meta_equal)(const struct bpf_map *meta0, 131 const struct bpf_map *meta1); 132 133 /* BTF name and id of struct allocated by map_alloc */ 134 const char * const map_btf_name; 135 int *map_btf_id; 136 137 /* bpf_iter info used to open a seq_file */ 138 const struct bpf_iter_seq_info *iter_seq_info; 139 }; 140 141 struct bpf_map { 142 /* The first two cachelines with read-mostly members of which some 143 * are also accessed in fast-path (e.g. ops, max_entries). 144 */ 145 const struct bpf_map_ops *ops ____cacheline_aligned; 146 struct bpf_map *inner_map_meta; 147 #ifdef CONFIG_SECURITY 148 void *security; 149 #endif 150 enum bpf_map_type map_type; 151 u32 key_size; 152 u32 value_size; 153 u32 max_entries; 154 u32 map_flags; 155 int spin_lock_off; /* >=0 valid offset, <0 error */ 156 u32 id; 157 int numa_node; 158 u32 btf_key_type_id; 159 u32 btf_value_type_id; 160 struct btf *btf; 161 #ifdef CONFIG_MEMCG_KMEM 162 struct mem_cgroup *memcg; 163 #endif 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 struct { 425 struct btf *btf; 426 u32 btf_id; 427 }; 428 }; 429 struct bpf_verifier_log *log; /* for verbose logs */ 430 }; 431 432 static inline void 433 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 434 { 435 aux->ctx_field_size = size; 436 } 437 438 struct bpf_prog_ops { 439 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 440 union bpf_attr __user *uattr); 441 }; 442 443 struct bpf_verifier_ops { 444 /* return eBPF function prototype for verification */ 445 const struct bpf_func_proto * 446 (*get_func_proto)(enum bpf_func_id func_id, 447 const struct bpf_prog *prog); 448 449 /* return true if 'size' wide access at offset 'off' within bpf_context 450 * with 'type' (read or write) is allowed 451 */ 452 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 453 const struct bpf_prog *prog, 454 struct bpf_insn_access_aux *info); 455 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 456 const struct bpf_prog *prog); 457 int (*gen_ld_abs)(const struct bpf_insn *orig, 458 struct bpf_insn *insn_buf); 459 u32 (*convert_ctx_access)(enum bpf_access_type type, 460 const struct bpf_insn *src, 461 struct bpf_insn *dst, 462 struct bpf_prog *prog, u32 *target_size); 463 int (*btf_struct_access)(struct bpf_verifier_log *log, 464 const struct btf *btf, 465 const struct btf_type *t, int off, int size, 466 enum bpf_access_type atype, 467 u32 *next_btf_id); 468 }; 469 470 struct bpf_prog_offload_ops { 471 /* verifier basic callbacks */ 472 int (*insn_hook)(struct bpf_verifier_env *env, 473 int insn_idx, int prev_insn_idx); 474 int (*finalize)(struct bpf_verifier_env *env); 475 /* verifier optimization callbacks (called after .finalize) */ 476 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 477 struct bpf_insn *insn); 478 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 479 /* program management callbacks */ 480 int (*prepare)(struct bpf_prog *prog); 481 int (*translate)(struct bpf_prog *prog); 482 void (*destroy)(struct bpf_prog *prog); 483 }; 484 485 struct bpf_prog_offload { 486 struct bpf_prog *prog; 487 struct net_device *netdev; 488 struct bpf_offload_dev *offdev; 489 void *dev_priv; 490 struct list_head offloads; 491 bool dev_state; 492 bool opt_failed; 493 void *jited_image; 494 u32 jited_len; 495 }; 496 497 enum bpf_cgroup_storage_type { 498 BPF_CGROUP_STORAGE_SHARED, 499 BPF_CGROUP_STORAGE_PERCPU, 500 __BPF_CGROUP_STORAGE_MAX 501 }; 502 503 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 504 505 /* The longest tracepoint has 12 args. 506 * See include/trace/bpf_probe.h 507 */ 508 #define MAX_BPF_FUNC_ARGS 12 509 510 struct bpf_prog_stats { 511 u64 cnt; 512 u64 nsecs; 513 struct u64_stats_sync syncp; 514 } __aligned(2 * sizeof(u64)); 515 516 struct btf_func_model { 517 u8 ret_size; 518 u8 nr_args; 519 u8 arg_size[MAX_BPF_FUNC_ARGS]; 520 }; 521 522 /* Restore arguments before returning from trampoline to let original function 523 * continue executing. This flag is used for fentry progs when there are no 524 * fexit progs. 525 */ 526 #define BPF_TRAMP_F_RESTORE_REGS BIT(0) 527 /* Call original function after fentry progs, but before fexit progs. 528 * Makes sense for fentry/fexit, normal calls and indirect calls. 529 */ 530 #define BPF_TRAMP_F_CALL_ORIG BIT(1) 531 /* Skip current frame and return to parent. Makes sense for fentry/fexit 532 * programs only. Should not be used with normal calls and indirect calls. 533 */ 534 #define BPF_TRAMP_F_SKIP_FRAME BIT(2) 535 536 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 537 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2 538 */ 539 #define BPF_MAX_TRAMP_PROGS 40 540 541 struct bpf_tramp_progs { 542 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS]; 543 int nr_progs; 544 }; 545 546 /* Different use cases for BPF trampoline: 547 * 1. replace nop at the function entry (kprobe equivalent) 548 * flags = BPF_TRAMP_F_RESTORE_REGS 549 * fentry = a set of programs to run before returning from trampoline 550 * 551 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 552 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 553 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 554 * fentry = a set of program to run before calling original function 555 * fexit = a set of program to run after original function 556 * 557 * 3. replace direct call instruction anywhere in the function body 558 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 559 * With flags = 0 560 * fentry = a set of programs to run before returning from trampoline 561 * With flags = BPF_TRAMP_F_CALL_ORIG 562 * orig_call = original callback addr or direct function addr 563 * fentry = a set of program to run before calling original function 564 * fexit = a set of program to run after original function 565 */ 566 int arch_prepare_bpf_trampoline(void *image, void *image_end, 567 const struct btf_func_model *m, u32 flags, 568 struct bpf_tramp_progs *tprogs, 569 void *orig_call); 570 /* these two functions are called from generated trampoline */ 571 u64 notrace __bpf_prog_enter(void); 572 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start); 573 void notrace __bpf_prog_enter_sleepable(void); 574 void notrace __bpf_prog_exit_sleepable(void); 575 576 struct bpf_ksym { 577 unsigned long start; 578 unsigned long end; 579 char name[KSYM_NAME_LEN]; 580 struct list_head lnode; 581 struct latch_tree_node tnode; 582 bool prog; 583 }; 584 585 enum bpf_tramp_prog_type { 586 BPF_TRAMP_FENTRY, 587 BPF_TRAMP_FEXIT, 588 BPF_TRAMP_MODIFY_RETURN, 589 BPF_TRAMP_MAX, 590 BPF_TRAMP_REPLACE, /* more than MAX */ 591 }; 592 593 struct bpf_trampoline { 594 /* hlist for trampoline_table */ 595 struct hlist_node hlist; 596 /* serializes access to fields of this trampoline */ 597 struct mutex mutex; 598 refcount_t refcnt; 599 u64 key; 600 struct { 601 struct btf_func_model model; 602 void *addr; 603 bool ftrace_managed; 604 } func; 605 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 606 * program by replacing one of its functions. func.addr is the address 607 * of the function it replaced. 608 */ 609 struct bpf_prog *extension_prog; 610 /* list of BPF programs using this trampoline */ 611 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 612 /* Number of attached programs. A counter per kind. */ 613 int progs_cnt[BPF_TRAMP_MAX]; 614 /* Executable image of trampoline */ 615 void *image; 616 u64 selector; 617 struct bpf_ksym ksym; 618 }; 619 620 struct bpf_attach_target_info { 621 struct btf_func_model fmodel; 622 long tgt_addr; 623 const char *tgt_name; 624 const struct btf_type *tgt_type; 625 }; 626 627 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 628 629 struct bpf_dispatcher_prog { 630 struct bpf_prog *prog; 631 refcount_t users; 632 }; 633 634 struct bpf_dispatcher { 635 /* dispatcher mutex */ 636 struct mutex mutex; 637 void *func; 638 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 639 int num_progs; 640 void *image; 641 u32 image_off; 642 struct bpf_ksym ksym; 643 }; 644 645 static __always_inline unsigned int bpf_dispatcher_nop_func( 646 const void *ctx, 647 const struct bpf_insn *insnsi, 648 unsigned int (*bpf_func)(const void *, 649 const struct bpf_insn *)) 650 { 651 return bpf_func(ctx, insnsi); 652 } 653 #ifdef CONFIG_BPF_JIT 654 int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 655 int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr); 656 struct bpf_trampoline *bpf_trampoline_get(u64 key, 657 struct bpf_attach_target_info *tgt_info); 658 void bpf_trampoline_put(struct bpf_trampoline *tr); 659 #define BPF_DISPATCHER_INIT(_name) { \ 660 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 661 .func = &_name##_func, \ 662 .progs = {}, \ 663 .num_progs = 0, \ 664 .image = NULL, \ 665 .image_off = 0, \ 666 .ksym = { \ 667 .name = #_name, \ 668 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 669 }, \ 670 } 671 672 #define DEFINE_BPF_DISPATCHER(name) \ 673 noinline unsigned int bpf_dispatcher_##name##_func( \ 674 const void *ctx, \ 675 const struct bpf_insn *insnsi, \ 676 unsigned int (*bpf_func)(const void *, \ 677 const struct bpf_insn *)) \ 678 { \ 679 return bpf_func(ctx, insnsi); \ 680 } \ 681 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 682 struct bpf_dispatcher bpf_dispatcher_##name = \ 683 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 684 #define DECLARE_BPF_DISPATCHER(name) \ 685 unsigned int bpf_dispatcher_##name##_func( \ 686 const void *ctx, \ 687 const struct bpf_insn *insnsi, \ 688 unsigned int (*bpf_func)(const void *, \ 689 const struct bpf_insn *)); \ 690 extern struct bpf_dispatcher bpf_dispatcher_##name; 691 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 692 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 693 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 694 struct bpf_prog *to); 695 /* Called only from JIT-enabled code, so there's no need for stubs. */ 696 void *bpf_jit_alloc_exec_page(void); 697 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym); 698 void bpf_image_ksym_del(struct bpf_ksym *ksym); 699 void bpf_ksym_add(struct bpf_ksym *ksym); 700 void bpf_ksym_del(struct bpf_ksym *ksym); 701 #else 702 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog, 703 struct bpf_trampoline *tr) 704 { 705 return -ENOTSUPP; 706 } 707 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog, 708 struct bpf_trampoline *tr) 709 { 710 return -ENOTSUPP; 711 } 712 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 713 struct bpf_attach_target_info *tgt_info) 714 { 715 return ERR_PTR(-EOPNOTSUPP); 716 } 717 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 718 #define DEFINE_BPF_DISPATCHER(name) 719 #define DECLARE_BPF_DISPATCHER(name) 720 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 721 #define BPF_DISPATCHER_PTR(name) NULL 722 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 723 struct bpf_prog *from, 724 struct bpf_prog *to) {} 725 static inline bool is_bpf_image_address(unsigned long address) 726 { 727 return false; 728 } 729 #endif 730 731 struct bpf_func_info_aux { 732 u16 linkage; 733 bool unreliable; 734 }; 735 736 enum bpf_jit_poke_reason { 737 BPF_POKE_REASON_TAIL_CALL, 738 }; 739 740 /* Descriptor of pokes pointing /into/ the JITed image. */ 741 struct bpf_jit_poke_descriptor { 742 void *tailcall_target; 743 void *tailcall_bypass; 744 void *bypass_addr; 745 union { 746 struct { 747 struct bpf_map *map; 748 u32 key; 749 } tail_call; 750 }; 751 bool tailcall_target_stable; 752 u8 adj_off; 753 u16 reason; 754 u32 insn_idx; 755 }; 756 757 /* reg_type info for ctx arguments */ 758 struct bpf_ctx_arg_aux { 759 u32 offset; 760 enum bpf_reg_type reg_type; 761 u32 btf_id; 762 }; 763 764 struct btf_mod_pair { 765 struct btf *btf; 766 struct module *module; 767 }; 768 769 struct bpf_prog_aux { 770 atomic64_t refcnt; 771 u32 used_map_cnt; 772 u32 used_btf_cnt; 773 u32 max_ctx_offset; 774 u32 max_pkt_offset; 775 u32 max_tp_access; 776 u32 stack_depth; 777 u32 id; 778 u32 func_cnt; /* used by non-func prog as the number of func progs */ 779 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 780 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 781 u32 ctx_arg_info_size; 782 u32 max_rdonly_access; 783 u32 max_rdwr_access; 784 struct btf *attach_btf; 785 const struct bpf_ctx_arg_aux *ctx_arg_info; 786 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 787 struct bpf_prog *dst_prog; 788 struct bpf_trampoline *dst_trampoline; 789 enum bpf_prog_type saved_dst_prog_type; 790 enum bpf_attach_type saved_dst_attach_type; 791 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 792 bool offload_requested; 793 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 794 bool func_proto_unreliable; 795 bool sleepable; 796 bool tail_call_reachable; 797 enum bpf_tramp_prog_type trampoline_prog_type; 798 struct hlist_node tramp_hlist; 799 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 800 const struct btf_type *attach_func_proto; 801 /* function name for valid attach_btf_id */ 802 const char *attach_func_name; 803 struct bpf_prog **func; 804 void *jit_data; /* JIT specific data. arch dependent */ 805 struct bpf_jit_poke_descriptor *poke_tab; 806 u32 size_poke_tab; 807 struct bpf_ksym ksym; 808 const struct bpf_prog_ops *ops; 809 struct bpf_map **used_maps; 810 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 811 struct btf_mod_pair *used_btfs; 812 struct bpf_prog *prog; 813 struct user_struct *user; 814 u64 load_time; /* ns since boottime */ 815 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 816 char name[BPF_OBJ_NAME_LEN]; 817 #ifdef CONFIG_SECURITY 818 void *security; 819 #endif 820 struct bpf_prog_offload *offload; 821 struct btf *btf; 822 struct bpf_func_info *func_info; 823 struct bpf_func_info_aux *func_info_aux; 824 /* bpf_line_info loaded from userspace. linfo->insn_off 825 * has the xlated insn offset. 826 * Both the main and sub prog share the same linfo. 827 * The subprog can access its first linfo by 828 * using the linfo_idx. 829 */ 830 struct bpf_line_info *linfo; 831 /* jited_linfo is the jited addr of the linfo. It has a 832 * one to one mapping to linfo: 833 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 834 * Both the main and sub prog share the same jited_linfo. 835 * The subprog can access its first jited_linfo by 836 * using the linfo_idx. 837 */ 838 void **jited_linfo; 839 u32 func_info_cnt; 840 u32 nr_linfo; 841 /* subprog can use linfo_idx to access its first linfo and 842 * jited_linfo. 843 * main prog always has linfo_idx == 0 844 */ 845 u32 linfo_idx; 846 u32 num_exentries; 847 struct exception_table_entry *extable; 848 struct bpf_prog_stats __percpu *stats; 849 union { 850 struct work_struct work; 851 struct rcu_head rcu; 852 }; 853 }; 854 855 struct bpf_array_aux { 856 /* 'Ownership' of prog array is claimed by the first program that 857 * is going to use this map or by the first program which FD is 858 * stored in the map to make sure that all callers and callees have 859 * the same prog type and JITed flag. 860 */ 861 enum bpf_prog_type type; 862 bool jited; 863 /* Programs with direct jumps into programs part of this array. */ 864 struct list_head poke_progs; 865 struct bpf_map *map; 866 struct mutex poke_mutex; 867 struct work_struct work; 868 }; 869 870 struct bpf_link { 871 atomic64_t refcnt; 872 u32 id; 873 enum bpf_link_type type; 874 const struct bpf_link_ops *ops; 875 struct bpf_prog *prog; 876 struct work_struct work; 877 }; 878 879 struct bpf_link_ops { 880 void (*release)(struct bpf_link *link); 881 void (*dealloc)(struct bpf_link *link); 882 int (*detach)(struct bpf_link *link); 883 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 884 struct bpf_prog *old_prog); 885 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 886 int (*fill_link_info)(const struct bpf_link *link, 887 struct bpf_link_info *info); 888 }; 889 890 struct bpf_link_primer { 891 struct bpf_link *link; 892 struct file *file; 893 int fd; 894 u32 id; 895 }; 896 897 struct bpf_struct_ops_value; 898 struct btf_type; 899 struct btf_member; 900 901 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 902 struct bpf_struct_ops { 903 const struct bpf_verifier_ops *verifier_ops; 904 int (*init)(struct btf *btf); 905 int (*check_member)(const struct btf_type *t, 906 const struct btf_member *member); 907 int (*init_member)(const struct btf_type *t, 908 const struct btf_member *member, 909 void *kdata, const void *udata); 910 int (*reg)(void *kdata); 911 void (*unreg)(void *kdata); 912 const struct btf_type *type; 913 const struct btf_type *value_type; 914 const char *name; 915 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 916 u32 type_id; 917 u32 value_id; 918 }; 919 920 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 921 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 922 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id); 923 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log); 924 bool bpf_struct_ops_get(const void *kdata); 925 void bpf_struct_ops_put(const void *kdata); 926 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 927 void *value); 928 static inline bool bpf_try_module_get(const void *data, struct module *owner) 929 { 930 if (owner == BPF_MODULE_OWNER) 931 return bpf_struct_ops_get(data); 932 else 933 return try_module_get(owner); 934 } 935 static inline void bpf_module_put(const void *data, struct module *owner) 936 { 937 if (owner == BPF_MODULE_OWNER) 938 bpf_struct_ops_put(data); 939 else 940 module_put(owner); 941 } 942 #else 943 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id) 944 { 945 return NULL; 946 } 947 static inline void bpf_struct_ops_init(struct btf *btf, 948 struct bpf_verifier_log *log) 949 { 950 } 951 static inline bool bpf_try_module_get(const void *data, struct module *owner) 952 { 953 return try_module_get(owner); 954 } 955 static inline void bpf_module_put(const void *data, struct module *owner) 956 { 957 module_put(owner); 958 } 959 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 960 void *key, 961 void *value) 962 { 963 return -EINVAL; 964 } 965 #endif 966 967 struct bpf_array { 968 struct bpf_map map; 969 u32 elem_size; 970 u32 index_mask; 971 struct bpf_array_aux *aux; 972 union { 973 char value[0] __aligned(8); 974 void *ptrs[0] __aligned(8); 975 void __percpu *pptrs[0] __aligned(8); 976 }; 977 }; 978 979 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 980 #define MAX_TAIL_CALL_CNT 32 981 982 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 983 BPF_F_RDONLY_PROG | \ 984 BPF_F_WRONLY | \ 985 BPF_F_WRONLY_PROG) 986 987 #define BPF_MAP_CAN_READ BIT(0) 988 #define BPF_MAP_CAN_WRITE BIT(1) 989 990 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 991 { 992 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 993 994 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 995 * not possible. 996 */ 997 if (access_flags & BPF_F_RDONLY_PROG) 998 return BPF_MAP_CAN_READ; 999 else if (access_flags & BPF_F_WRONLY_PROG) 1000 return BPF_MAP_CAN_WRITE; 1001 else 1002 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 1003 } 1004 1005 static inline bool bpf_map_flags_access_ok(u32 access_flags) 1006 { 1007 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 1008 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1009 } 1010 1011 struct bpf_event_entry { 1012 struct perf_event *event; 1013 struct file *perf_file; 1014 struct file *map_file; 1015 struct rcu_head rcu; 1016 }; 1017 1018 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); 1019 int bpf_prog_calc_tag(struct bpf_prog *fp); 1020 1021 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 1022 1023 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 1024 unsigned long off, unsigned long len); 1025 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 1026 const struct bpf_insn *src, 1027 struct bpf_insn *dst, 1028 struct bpf_prog *prog, 1029 u32 *target_size); 1030 1031 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 1032 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 1033 1034 /* an array of programs to be executed under rcu_lock. 1035 * 1036 * Typical usage: 1037 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN); 1038 * 1039 * the structure returned by bpf_prog_array_alloc() should be populated 1040 * with program pointers and the last pointer must be NULL. 1041 * The user has to keep refcnt on the program and make sure the program 1042 * is removed from the array before bpf_prog_put(). 1043 * The 'struct bpf_prog_array *' should only be replaced with xchg() 1044 * since other cpus are walking the array of pointers in parallel. 1045 */ 1046 struct bpf_prog_array_item { 1047 struct bpf_prog *prog; 1048 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1049 }; 1050 1051 struct bpf_prog_array { 1052 struct rcu_head rcu; 1053 struct bpf_prog_array_item items[]; 1054 }; 1055 1056 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 1057 void bpf_prog_array_free(struct bpf_prog_array *progs); 1058 int bpf_prog_array_length(struct bpf_prog_array *progs); 1059 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 1060 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 1061 __u32 __user *prog_ids, u32 cnt); 1062 1063 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 1064 struct bpf_prog *old_prog); 1065 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 1066 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 1067 struct bpf_prog *prog); 1068 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 1069 u32 *prog_ids, u32 request_cnt, 1070 u32 *prog_cnt); 1071 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 1072 struct bpf_prog *exclude_prog, 1073 struct bpf_prog *include_prog, 1074 struct bpf_prog_array **new_array); 1075 1076 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \ 1077 ({ \ 1078 struct bpf_prog_array_item *_item; \ 1079 struct bpf_prog *_prog; \ 1080 struct bpf_prog_array *_array; \ 1081 u32 _ret = 1; \ 1082 migrate_disable(); \ 1083 rcu_read_lock(); \ 1084 _array = rcu_dereference(array); \ 1085 if (unlikely(check_non_null && !_array))\ 1086 goto _out; \ 1087 _item = &_array->items[0]; \ 1088 while ((_prog = READ_ONCE(_item->prog))) { \ 1089 bpf_cgroup_storage_set(_item->cgroup_storage); \ 1090 _ret &= func(_prog, ctx); \ 1091 _item++; \ 1092 } \ 1093 _out: \ 1094 rcu_read_unlock(); \ 1095 migrate_enable(); \ 1096 _ret; \ 1097 }) 1098 1099 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs 1100 * so BPF programs can request cwr for TCP packets. 1101 * 1102 * Current cgroup skb programs can only return 0 or 1 (0 to drop the 1103 * packet. This macro changes the behavior so the low order bit 1104 * indicates whether the packet should be dropped (0) or not (1) 1105 * and the next bit is a congestion notification bit. This could be 1106 * used by TCP to call tcp_enter_cwr() 1107 * 1108 * Hence, new allowed return values of CGROUP EGRESS BPF programs are: 1109 * 0: drop packet 1110 * 1: keep packet 1111 * 2: drop packet and cn 1112 * 3: keep packet and cn 1113 * 1114 * This macro then converts it to one of the NET_XMIT or an error 1115 * code that is then interpreted as drop packet (and no cn): 1116 * 0: NET_XMIT_SUCCESS skb should be transmitted 1117 * 1: NET_XMIT_DROP skb should be dropped and cn 1118 * 2: NET_XMIT_CN skb should be transmitted and cn 1119 * 3: -EPERM skb should be dropped 1120 */ 1121 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \ 1122 ({ \ 1123 struct bpf_prog_array_item *_item; \ 1124 struct bpf_prog *_prog; \ 1125 struct bpf_prog_array *_array; \ 1126 u32 ret; \ 1127 u32 _ret = 1; \ 1128 u32 _cn = 0; \ 1129 migrate_disable(); \ 1130 rcu_read_lock(); \ 1131 _array = rcu_dereference(array); \ 1132 _item = &_array->items[0]; \ 1133 while ((_prog = READ_ONCE(_item->prog))) { \ 1134 bpf_cgroup_storage_set(_item->cgroup_storage); \ 1135 ret = func(_prog, ctx); \ 1136 _ret &= (ret & 1); \ 1137 _cn |= (ret & 2); \ 1138 _item++; \ 1139 } \ 1140 rcu_read_unlock(); \ 1141 migrate_enable(); \ 1142 if (_ret) \ 1143 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \ 1144 else \ 1145 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \ 1146 _ret; \ 1147 }) 1148 1149 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \ 1150 __BPF_PROG_RUN_ARRAY(array, ctx, func, false) 1151 1152 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \ 1153 __BPF_PROG_RUN_ARRAY(array, ctx, func, true) 1154 1155 #ifdef CONFIG_BPF_SYSCALL 1156 DECLARE_PER_CPU(int, bpf_prog_active); 1157 extern struct mutex bpf_stats_enabled_mutex; 1158 1159 /* 1160 * Block execution of BPF programs attached to instrumentation (perf, 1161 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 1162 * these events can happen inside a region which holds a map bucket lock 1163 * and can deadlock on it. 1164 * 1165 * Use the preemption safe inc/dec variants on RT because migrate disable 1166 * is preemptible on RT and preemption in the middle of the RMW operation 1167 * might lead to inconsistent state. Use the raw variants for non RT 1168 * kernels as migrate_disable() maps to preempt_disable() so the slightly 1169 * more expensive save operation can be avoided. 1170 */ 1171 static inline void bpf_disable_instrumentation(void) 1172 { 1173 migrate_disable(); 1174 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1175 this_cpu_inc(bpf_prog_active); 1176 else 1177 __this_cpu_inc(bpf_prog_active); 1178 } 1179 1180 static inline void bpf_enable_instrumentation(void) 1181 { 1182 if (IS_ENABLED(CONFIG_PREEMPT_RT)) 1183 this_cpu_dec(bpf_prog_active); 1184 else 1185 __this_cpu_dec(bpf_prog_active); 1186 migrate_enable(); 1187 } 1188 1189 extern const struct file_operations bpf_map_fops; 1190 extern const struct file_operations bpf_prog_fops; 1191 extern const struct file_operations bpf_iter_fops; 1192 1193 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 1194 extern const struct bpf_prog_ops _name ## _prog_ops; \ 1195 extern const struct bpf_verifier_ops _name ## _verifier_ops; 1196 #define BPF_MAP_TYPE(_id, _ops) \ 1197 extern const struct bpf_map_ops _ops; 1198 #define BPF_LINK_TYPE(_id, _name) 1199 #include <linux/bpf_types.h> 1200 #undef BPF_PROG_TYPE 1201 #undef BPF_MAP_TYPE 1202 #undef BPF_LINK_TYPE 1203 1204 extern const struct bpf_prog_ops bpf_offload_prog_ops; 1205 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 1206 extern const struct bpf_verifier_ops xdp_analyzer_ops; 1207 1208 struct bpf_prog *bpf_prog_get(u32 ufd); 1209 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 1210 bool attach_drv); 1211 void bpf_prog_add(struct bpf_prog *prog, int i); 1212 void bpf_prog_sub(struct bpf_prog *prog, int i); 1213 void bpf_prog_inc(struct bpf_prog *prog); 1214 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 1215 void bpf_prog_put(struct bpf_prog *prog); 1216 1217 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 1218 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 1219 1220 struct bpf_map *bpf_map_get(u32 ufd); 1221 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 1222 struct bpf_map *__bpf_map_get(struct fd f); 1223 void bpf_map_inc(struct bpf_map *map); 1224 void bpf_map_inc_with_uref(struct bpf_map *map); 1225 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 1226 void bpf_map_put_with_uref(struct bpf_map *map); 1227 void bpf_map_put(struct bpf_map *map); 1228 void *bpf_map_area_alloc(u64 size, int numa_node); 1229 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 1230 void bpf_map_area_free(void *base); 1231 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 1232 int generic_map_lookup_batch(struct bpf_map *map, 1233 const union bpf_attr *attr, 1234 union bpf_attr __user *uattr); 1235 int generic_map_update_batch(struct bpf_map *map, 1236 const union bpf_attr *attr, 1237 union bpf_attr __user *uattr); 1238 int generic_map_delete_batch(struct bpf_map *map, 1239 const union bpf_attr *attr, 1240 union bpf_attr __user *uattr); 1241 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 1242 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 1243 1244 #ifdef CONFIG_MEMCG_KMEM 1245 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1246 int node); 1247 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 1248 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 1249 size_t align, gfp_t flags); 1250 #else 1251 static inline void * 1252 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1253 int node) 1254 { 1255 return kmalloc_node(size, flags, node); 1256 } 1257 1258 static inline void * 1259 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags) 1260 { 1261 return kzalloc(size, flags); 1262 } 1263 1264 static inline void __percpu * 1265 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align, 1266 gfp_t flags) 1267 { 1268 return __alloc_percpu_gfp(size, align, flags); 1269 } 1270 #endif 1271 1272 extern int sysctl_unprivileged_bpf_disabled; 1273 1274 static inline bool bpf_allow_ptr_leaks(void) 1275 { 1276 return perfmon_capable(); 1277 } 1278 1279 static inline bool bpf_allow_ptr_to_map_access(void) 1280 { 1281 return perfmon_capable(); 1282 } 1283 1284 static inline bool bpf_bypass_spec_v1(void) 1285 { 1286 return perfmon_capable(); 1287 } 1288 1289 static inline bool bpf_bypass_spec_v4(void) 1290 { 1291 return perfmon_capable(); 1292 } 1293 1294 int bpf_map_new_fd(struct bpf_map *map, int flags); 1295 int bpf_prog_new_fd(struct bpf_prog *prog); 1296 1297 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1298 const struct bpf_link_ops *ops, struct bpf_prog *prog); 1299 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 1300 int bpf_link_settle(struct bpf_link_primer *primer); 1301 void bpf_link_cleanup(struct bpf_link_primer *primer); 1302 void bpf_link_inc(struct bpf_link *link); 1303 void bpf_link_put(struct bpf_link *link); 1304 int bpf_link_new_fd(struct bpf_link *link); 1305 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd); 1306 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 1307 1308 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 1309 int bpf_obj_get_user(const char __user *pathname, int flags); 1310 1311 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 1312 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 1313 extern int bpf_iter_ ## target(args); \ 1314 int __init bpf_iter_ ## target(args) { return 0; } 1315 1316 struct bpf_iter_aux_info { 1317 struct bpf_map *map; 1318 }; 1319 1320 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 1321 union bpf_iter_link_info *linfo, 1322 struct bpf_iter_aux_info *aux); 1323 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 1324 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 1325 struct seq_file *seq); 1326 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 1327 struct bpf_link_info *info); 1328 1329 enum bpf_iter_feature { 1330 BPF_ITER_RESCHED = BIT(0), 1331 }; 1332 1333 #define BPF_ITER_CTX_ARG_MAX 2 1334 struct bpf_iter_reg { 1335 const char *target; 1336 bpf_iter_attach_target_t attach_target; 1337 bpf_iter_detach_target_t detach_target; 1338 bpf_iter_show_fdinfo_t show_fdinfo; 1339 bpf_iter_fill_link_info_t fill_link_info; 1340 u32 ctx_arg_info_size; 1341 u32 feature; 1342 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 1343 const struct bpf_iter_seq_info *seq_info; 1344 }; 1345 1346 struct bpf_iter_meta { 1347 __bpf_md_ptr(struct seq_file *, seq); 1348 u64 session_id; 1349 u64 seq_num; 1350 }; 1351 1352 struct bpf_iter__bpf_map_elem { 1353 __bpf_md_ptr(struct bpf_iter_meta *, meta); 1354 __bpf_md_ptr(struct bpf_map *, map); 1355 __bpf_md_ptr(void *, key); 1356 __bpf_md_ptr(void *, value); 1357 }; 1358 1359 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 1360 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 1361 bool bpf_iter_prog_supported(struct bpf_prog *prog); 1362 int bpf_iter_link_attach(const union bpf_attr *attr, struct bpf_prog *prog); 1363 int bpf_iter_new_fd(struct bpf_link *link); 1364 bool bpf_link_is_iter(struct bpf_link *link); 1365 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 1366 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 1367 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 1368 struct seq_file *seq); 1369 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 1370 struct bpf_link_info *info); 1371 1372 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 1373 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 1374 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 1375 u64 flags); 1376 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 1377 u64 flags); 1378 1379 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 1380 1381 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 1382 void *key, void *value, u64 map_flags); 1383 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1384 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 1385 void *key, void *value, u64 map_flags); 1386 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1387 1388 int bpf_get_file_flag(int flags); 1389 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size, 1390 size_t actual_size); 1391 1392 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 1393 * forced to use 'long' read/writes to try to atomically copy long counters. 1394 * Best-effort only. No barriers here, since it _will_ race with concurrent 1395 * updates from BPF programs. Called from bpf syscall and mostly used with 1396 * size 8 or 16 bytes, so ask compiler to inline it. 1397 */ 1398 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 1399 { 1400 const long *lsrc = src; 1401 long *ldst = dst; 1402 1403 size /= sizeof(long); 1404 while (size--) 1405 *ldst++ = *lsrc++; 1406 } 1407 1408 /* verify correctness of eBPF program */ 1409 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, 1410 union bpf_attr __user *uattr); 1411 1412 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 1413 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 1414 #endif 1415 1416 struct btf *bpf_get_btf_vmlinux(void); 1417 1418 /* Map specifics */ 1419 struct xdp_buff; 1420 struct sk_buff; 1421 1422 struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key); 1423 struct bpf_dtab_netdev *__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key); 1424 void __dev_flush(void); 1425 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1426 struct net_device *dev_rx); 1427 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1428 struct net_device *dev_rx); 1429 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 1430 struct bpf_prog *xdp_prog); 1431 bool dev_map_can_have_prog(struct bpf_map *map); 1432 1433 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key); 1434 void __cpu_map_flush(void); 1435 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, 1436 struct net_device *dev_rx); 1437 bool cpu_map_prog_allowed(struct bpf_map *map); 1438 1439 /* Return map's numa specified by userspace */ 1440 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 1441 { 1442 return (attr->map_flags & BPF_F_NUMA_NODE) ? 1443 attr->numa_node : NUMA_NO_NODE; 1444 } 1445 1446 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 1447 int array_map_alloc_check(union bpf_attr *attr); 1448 1449 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 1450 union bpf_attr __user *uattr); 1451 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 1452 union bpf_attr __user *uattr); 1453 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1454 const union bpf_attr *kattr, 1455 union bpf_attr __user *uattr); 1456 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1457 const union bpf_attr *kattr, 1458 union bpf_attr __user *uattr); 1459 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 1460 const union bpf_attr *kattr, 1461 union bpf_attr __user *uattr); 1462 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 1463 const struct bpf_prog *prog, 1464 struct bpf_insn_access_aux *info); 1465 int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf, 1466 const struct btf_type *t, int off, int size, 1467 enum bpf_access_type atype, 1468 u32 *next_btf_id); 1469 bool btf_struct_ids_match(struct bpf_verifier_log *log, 1470 const struct btf *btf, u32 id, int off, 1471 const struct btf *need_btf, u32 need_type_id); 1472 1473 int btf_distill_func_proto(struct bpf_verifier_log *log, 1474 struct btf *btf, 1475 const struct btf_type *func_proto, 1476 const char *func_name, 1477 struct btf_func_model *m); 1478 1479 struct bpf_reg_state; 1480 int btf_check_func_arg_match(struct bpf_verifier_env *env, int subprog, 1481 struct bpf_reg_state *regs); 1482 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, 1483 struct bpf_reg_state *reg); 1484 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 1485 struct btf *btf, const struct btf_type *t); 1486 1487 struct bpf_prog *bpf_prog_by_id(u32 id); 1488 struct bpf_link *bpf_link_by_id(u32 id); 1489 1490 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id); 1491 #else /* !CONFIG_BPF_SYSCALL */ 1492 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 1493 { 1494 return ERR_PTR(-EOPNOTSUPP); 1495 } 1496 1497 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 1498 enum bpf_prog_type type, 1499 bool attach_drv) 1500 { 1501 return ERR_PTR(-EOPNOTSUPP); 1502 } 1503 1504 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 1505 { 1506 } 1507 1508 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 1509 { 1510 } 1511 1512 static inline void bpf_prog_put(struct bpf_prog *prog) 1513 { 1514 } 1515 1516 static inline void bpf_prog_inc(struct bpf_prog *prog) 1517 { 1518 } 1519 1520 static inline struct bpf_prog *__must_check 1521 bpf_prog_inc_not_zero(struct bpf_prog *prog) 1522 { 1523 return ERR_PTR(-EOPNOTSUPP); 1524 } 1525 1526 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1527 const struct bpf_link_ops *ops, 1528 struct bpf_prog *prog) 1529 { 1530 } 1531 1532 static inline int bpf_link_prime(struct bpf_link *link, 1533 struct bpf_link_primer *primer) 1534 { 1535 return -EOPNOTSUPP; 1536 } 1537 1538 static inline int bpf_link_settle(struct bpf_link_primer *primer) 1539 { 1540 return -EOPNOTSUPP; 1541 } 1542 1543 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 1544 { 1545 } 1546 1547 static inline void bpf_link_inc(struct bpf_link *link) 1548 { 1549 } 1550 1551 static inline void bpf_link_put(struct bpf_link *link) 1552 { 1553 } 1554 1555 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 1556 { 1557 return -EOPNOTSUPP; 1558 } 1559 1560 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map, 1561 u32 key) 1562 { 1563 return NULL; 1564 } 1565 1566 static inline struct net_device *__dev_map_hash_lookup_elem(struct bpf_map *map, 1567 u32 key) 1568 { 1569 return NULL; 1570 } 1571 static inline bool dev_map_can_have_prog(struct bpf_map *map) 1572 { 1573 return false; 1574 } 1575 1576 static inline void __dev_flush(void) 1577 { 1578 } 1579 1580 struct xdp_buff; 1581 struct bpf_dtab_netdev; 1582 1583 static inline 1584 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, 1585 struct net_device *dev_rx) 1586 { 1587 return 0; 1588 } 1589 1590 static inline 1591 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, 1592 struct net_device *dev_rx) 1593 { 1594 return 0; 1595 } 1596 1597 struct sk_buff; 1598 1599 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 1600 struct sk_buff *skb, 1601 struct bpf_prog *xdp_prog) 1602 { 1603 return 0; 1604 } 1605 1606 static inline 1607 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key) 1608 { 1609 return NULL; 1610 } 1611 1612 static inline void __cpu_map_flush(void) 1613 { 1614 } 1615 1616 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 1617 struct xdp_buff *xdp, 1618 struct net_device *dev_rx) 1619 { 1620 return 0; 1621 } 1622 1623 static inline bool cpu_map_prog_allowed(struct bpf_map *map) 1624 { 1625 return false; 1626 } 1627 1628 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 1629 enum bpf_prog_type type) 1630 { 1631 return ERR_PTR(-EOPNOTSUPP); 1632 } 1633 1634 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 1635 const union bpf_attr *kattr, 1636 union bpf_attr __user *uattr) 1637 { 1638 return -ENOTSUPP; 1639 } 1640 1641 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 1642 const union bpf_attr *kattr, 1643 union bpf_attr __user *uattr) 1644 { 1645 return -ENOTSUPP; 1646 } 1647 1648 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1649 const union bpf_attr *kattr, 1650 union bpf_attr __user *uattr) 1651 { 1652 return -ENOTSUPP; 1653 } 1654 1655 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1656 const union bpf_attr *kattr, 1657 union bpf_attr __user *uattr) 1658 { 1659 return -ENOTSUPP; 1660 } 1661 1662 static inline void bpf_map_put(struct bpf_map *map) 1663 { 1664 } 1665 1666 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 1667 { 1668 return ERR_PTR(-ENOTSUPP); 1669 } 1670 1671 static inline const struct bpf_func_proto * 1672 bpf_base_func_proto(enum bpf_func_id func_id) 1673 { 1674 return NULL; 1675 } 1676 #endif /* CONFIG_BPF_SYSCALL */ 1677 1678 void __bpf_free_used_btfs(struct bpf_prog_aux *aux, 1679 struct btf_mod_pair *used_btfs, u32 len); 1680 1681 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 1682 enum bpf_prog_type type) 1683 { 1684 return bpf_prog_get_type_dev(ufd, type, false); 1685 } 1686 1687 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 1688 struct bpf_map **used_maps, u32 len); 1689 1690 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 1691 1692 int bpf_prog_offload_compile(struct bpf_prog *prog); 1693 void bpf_prog_offload_destroy(struct bpf_prog *prog); 1694 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 1695 struct bpf_prog *prog); 1696 1697 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 1698 1699 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 1700 int bpf_map_offload_update_elem(struct bpf_map *map, 1701 void *key, void *value, u64 flags); 1702 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 1703 int bpf_map_offload_get_next_key(struct bpf_map *map, 1704 void *key, void *next_key); 1705 1706 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 1707 1708 struct bpf_offload_dev * 1709 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 1710 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 1711 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 1712 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 1713 struct net_device *netdev); 1714 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 1715 struct net_device *netdev); 1716 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 1717 1718 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 1719 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 1720 1721 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 1722 { 1723 return aux->offload_requested; 1724 } 1725 1726 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1727 { 1728 return unlikely(map->ops == &bpf_map_offload_ops); 1729 } 1730 1731 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 1732 void bpf_map_offload_map_free(struct bpf_map *map); 1733 #else 1734 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 1735 union bpf_attr *attr) 1736 { 1737 return -EOPNOTSUPP; 1738 } 1739 1740 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 1741 { 1742 return false; 1743 } 1744 1745 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 1746 { 1747 return false; 1748 } 1749 1750 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 1751 { 1752 return ERR_PTR(-EOPNOTSUPP); 1753 } 1754 1755 static inline void bpf_map_offload_map_free(struct bpf_map *map) 1756 { 1757 } 1758 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 1759 1760 #if defined(CONFIG_BPF_STREAM_PARSER) 1761 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, 1762 struct bpf_prog *old, u32 which); 1763 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 1764 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 1765 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 1766 void sock_map_unhash(struct sock *sk); 1767 void sock_map_close(struct sock *sk, long timeout); 1768 #else 1769 static inline int sock_map_prog_update(struct bpf_map *map, 1770 struct bpf_prog *prog, 1771 struct bpf_prog *old, u32 which) 1772 { 1773 return -EOPNOTSUPP; 1774 } 1775 1776 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 1777 struct bpf_prog *prog) 1778 { 1779 return -EINVAL; 1780 } 1781 1782 static inline int sock_map_prog_detach(const union bpf_attr *attr, 1783 enum bpf_prog_type ptype) 1784 { 1785 return -EOPNOTSUPP; 1786 } 1787 1788 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 1789 u64 flags) 1790 { 1791 return -EOPNOTSUPP; 1792 } 1793 #endif /* CONFIG_BPF_STREAM_PARSER */ 1794 1795 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 1796 void bpf_sk_reuseport_detach(struct sock *sk); 1797 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 1798 void *value); 1799 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 1800 void *value, u64 map_flags); 1801 #else 1802 static inline void bpf_sk_reuseport_detach(struct sock *sk) 1803 { 1804 } 1805 1806 #ifdef CONFIG_BPF_SYSCALL 1807 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 1808 void *key, void *value) 1809 { 1810 return -EOPNOTSUPP; 1811 } 1812 1813 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 1814 void *key, void *value, 1815 u64 map_flags) 1816 { 1817 return -EOPNOTSUPP; 1818 } 1819 #endif /* CONFIG_BPF_SYSCALL */ 1820 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 1821 1822 /* verifier prototypes for helper functions called from eBPF programs */ 1823 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 1824 extern const struct bpf_func_proto bpf_map_update_elem_proto; 1825 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 1826 extern const struct bpf_func_proto bpf_map_push_elem_proto; 1827 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 1828 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 1829 1830 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 1831 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 1832 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 1833 extern const struct bpf_func_proto bpf_tail_call_proto; 1834 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 1835 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 1836 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 1837 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 1838 extern const struct bpf_func_proto bpf_get_current_comm_proto; 1839 extern const struct bpf_func_proto bpf_get_stackid_proto; 1840 extern const struct bpf_func_proto bpf_get_stack_proto; 1841 extern const struct bpf_func_proto bpf_get_task_stack_proto; 1842 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 1843 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 1844 extern const struct bpf_func_proto bpf_sock_map_update_proto; 1845 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 1846 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 1847 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 1848 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 1849 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 1850 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 1851 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 1852 extern const struct bpf_func_proto bpf_spin_lock_proto; 1853 extern const struct bpf_func_proto bpf_spin_unlock_proto; 1854 extern const struct bpf_func_proto bpf_get_local_storage_proto; 1855 extern const struct bpf_func_proto bpf_strtol_proto; 1856 extern const struct bpf_func_proto bpf_strtoul_proto; 1857 extern const struct bpf_func_proto bpf_tcp_sock_proto; 1858 extern const struct bpf_func_proto bpf_jiffies64_proto; 1859 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 1860 extern const struct bpf_func_proto bpf_event_output_data_proto; 1861 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 1862 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 1863 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 1864 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 1865 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 1866 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 1867 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 1868 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 1869 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 1870 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 1871 extern const struct bpf_func_proto bpf_copy_from_user_proto; 1872 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 1873 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 1874 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 1875 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 1876 extern const struct bpf_func_proto bpf_sock_from_file_proto; 1877 1878 const struct bpf_func_proto *bpf_tracing_func_proto( 1879 enum bpf_func_id func_id, const struct bpf_prog *prog); 1880 1881 const struct bpf_func_proto *tracing_prog_func_proto( 1882 enum bpf_func_id func_id, const struct bpf_prog *prog); 1883 1884 /* Shared helpers among cBPF and eBPF. */ 1885 void bpf_user_rnd_init_once(void); 1886 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 1887 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 1888 1889 #if defined(CONFIG_NET) 1890 bool bpf_sock_common_is_valid_access(int off, int size, 1891 enum bpf_access_type type, 1892 struct bpf_insn_access_aux *info); 1893 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1894 struct bpf_insn_access_aux *info); 1895 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 1896 const struct bpf_insn *si, 1897 struct bpf_insn *insn_buf, 1898 struct bpf_prog *prog, 1899 u32 *target_size); 1900 #else 1901 static inline bool bpf_sock_common_is_valid_access(int off, int size, 1902 enum bpf_access_type type, 1903 struct bpf_insn_access_aux *info) 1904 { 1905 return false; 1906 } 1907 static inline bool bpf_sock_is_valid_access(int off, int size, 1908 enum bpf_access_type type, 1909 struct bpf_insn_access_aux *info) 1910 { 1911 return false; 1912 } 1913 static inline u32 bpf_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 { 1919 return 0; 1920 } 1921 #endif 1922 1923 #ifdef CONFIG_INET 1924 struct sk_reuseport_kern { 1925 struct sk_buff *skb; 1926 struct sock *sk; 1927 struct sock *selected_sk; 1928 void *data_end; 1929 u32 hash; 1930 u32 reuseport_id; 1931 bool bind_inany; 1932 }; 1933 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1934 struct bpf_insn_access_aux *info); 1935 1936 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 1937 const struct bpf_insn *si, 1938 struct bpf_insn *insn_buf, 1939 struct bpf_prog *prog, 1940 u32 *target_size); 1941 1942 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 1943 struct bpf_insn_access_aux *info); 1944 1945 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 1946 const struct bpf_insn *si, 1947 struct bpf_insn *insn_buf, 1948 struct bpf_prog *prog, 1949 u32 *target_size); 1950 #else 1951 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 1952 enum bpf_access_type type, 1953 struct bpf_insn_access_aux *info) 1954 { 1955 return false; 1956 } 1957 1958 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 1959 const struct bpf_insn *si, 1960 struct bpf_insn *insn_buf, 1961 struct bpf_prog *prog, 1962 u32 *target_size) 1963 { 1964 return 0; 1965 } 1966 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 1967 enum bpf_access_type type, 1968 struct bpf_insn_access_aux *info) 1969 { 1970 return false; 1971 } 1972 1973 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 1974 const struct bpf_insn *si, 1975 struct bpf_insn *insn_buf, 1976 struct bpf_prog *prog, 1977 u32 *target_size) 1978 { 1979 return 0; 1980 } 1981 #endif /* CONFIG_INET */ 1982 1983 enum bpf_text_poke_type { 1984 BPF_MOD_CALL, 1985 BPF_MOD_JUMP, 1986 }; 1987 1988 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 1989 void *addr1, void *addr2); 1990 1991 struct btf_id_set; 1992 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 1993 1994 #endif /* _LINUX_BPF_H */ 1995