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 #include <uapi/linux/filter.h> 9 10 #include <linux/workqueue.h> 11 #include <linux/file.h> 12 #include <linux/percpu.h> 13 #include <linux/err.h> 14 #include <linux/rbtree_latch.h> 15 #include <linux/numa.h> 16 #include <linux/mm_types.h> 17 #include <linux/wait.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 #include <linux/percpu-refcount.h> 26 #include <linux/stddef.h> 27 #include <linux/bpfptr.h> 28 #include <linux/btf.h> 29 #include <linux/rcupdate_trace.h> 30 31 struct bpf_verifier_env; 32 struct bpf_verifier_log; 33 struct perf_event; 34 struct bpf_prog; 35 struct bpf_prog_aux; 36 struct bpf_map; 37 struct sock; 38 struct seq_file; 39 struct btf; 40 struct btf_type; 41 struct exception_table_entry; 42 struct seq_operations; 43 struct bpf_iter_aux_info; 44 struct bpf_local_storage; 45 struct bpf_local_storage_map; 46 struct kobject; 47 struct mem_cgroup; 48 struct module; 49 struct bpf_func_state; 50 51 extern struct idr btf_idr; 52 extern spinlock_t btf_idr_lock; 53 extern struct kobject *btf_kobj; 54 55 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64); 56 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, 57 struct bpf_iter_aux_info *aux); 58 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); 59 struct bpf_iter_seq_info { 60 const struct seq_operations *seq_ops; 61 bpf_iter_init_seq_priv_t init_seq_private; 62 bpf_iter_fini_seq_priv_t fini_seq_private; 63 u32 seq_priv_size; 64 }; 65 66 /* map is generic key/value storage optionally accessible by eBPF programs */ 67 struct bpf_map_ops { 68 /* funcs callable from userspace (via syscall) */ 69 int (*map_alloc_check)(union bpf_attr *attr); 70 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 71 void (*map_release)(struct bpf_map *map, struct file *map_file); 72 void (*map_free)(struct bpf_map *map); 73 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 74 void (*map_release_uref)(struct bpf_map *map); 75 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 76 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, 77 union bpf_attr __user *uattr); 78 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key, 79 void *value, u64 flags); 80 int (*map_lookup_and_delete_batch)(struct bpf_map *map, 81 const union bpf_attr *attr, 82 union bpf_attr __user *uattr); 83 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr, 84 union bpf_attr __user *uattr); 85 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, 86 union bpf_attr __user *uattr); 87 88 /* funcs callable from userspace and from eBPF programs */ 89 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 90 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 91 int (*map_delete_elem)(struct bpf_map *map, void *key); 92 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 93 int (*map_pop_elem)(struct bpf_map *map, void *value); 94 int (*map_peek_elem)(struct bpf_map *map, void *value); 95 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu); 96 97 /* funcs called by prog_array and perf_event_array map */ 98 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 99 int fd); 100 void (*map_fd_put_ptr)(void *ptr); 101 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 102 u32 (*map_fd_sys_lookup_elem)(void *ptr); 103 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 104 struct seq_file *m); 105 int (*map_check_btf)(const struct bpf_map *map, 106 const struct btf *btf, 107 const struct btf_type *key_type, 108 const struct btf_type *value_type); 109 110 /* Prog poke tracking helpers. */ 111 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux); 112 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux); 113 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, 114 struct bpf_prog *new); 115 116 /* Direct value access helpers. */ 117 int (*map_direct_value_addr)(const struct bpf_map *map, 118 u64 *imm, u32 off); 119 int (*map_direct_value_meta)(const struct bpf_map *map, 120 u64 imm, u32 *off); 121 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma); 122 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, 123 struct poll_table_struct *pts); 124 125 /* Functions called by bpf_local_storage maps */ 126 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, 127 void *owner, u32 size); 128 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, 129 void *owner, u32 size); 130 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner); 131 132 /* Misc helpers.*/ 133 int (*map_redirect)(struct bpf_map *map, u32 ifindex, u64 flags); 134 135 /* map_meta_equal must be implemented for maps that can be 136 * used as an inner map. It is a runtime check to ensure 137 * an inner map can be inserted to an outer map. 138 * 139 * Some properties of the inner map has been used during the 140 * verification time. When inserting an inner map at the runtime, 141 * map_meta_equal has to ensure the inserting map has the same 142 * properties that the verifier has used earlier. 143 */ 144 bool (*map_meta_equal)(const struct bpf_map *meta0, 145 const struct bpf_map *meta1); 146 147 148 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env, 149 struct bpf_func_state *caller, 150 struct bpf_func_state *callee); 151 int (*map_for_each_callback)(struct bpf_map *map, 152 bpf_callback_t callback_fn, 153 void *callback_ctx, u64 flags); 154 155 /* BTF id of struct allocated by map_alloc */ 156 int *map_btf_id; 157 158 /* bpf_iter info used to open a seq_file */ 159 const struct bpf_iter_seq_info *iter_seq_info; 160 }; 161 162 enum { 163 /* Support at most 8 pointers in a BPF map value */ 164 BPF_MAP_VALUE_OFF_MAX = 8, 165 BPF_MAP_OFF_ARR_MAX = BPF_MAP_VALUE_OFF_MAX + 166 1 + /* for bpf_spin_lock */ 167 1, /* for bpf_timer */ 168 }; 169 170 enum bpf_kptr_type { 171 BPF_KPTR_UNREF, 172 BPF_KPTR_REF, 173 }; 174 175 struct bpf_map_value_off_desc { 176 u32 offset; 177 enum bpf_kptr_type type; 178 struct { 179 struct btf *btf; 180 struct module *module; 181 btf_dtor_kfunc_t dtor; 182 u32 btf_id; 183 } kptr; 184 }; 185 186 struct bpf_map_value_off { 187 u32 nr_off; 188 struct bpf_map_value_off_desc off[]; 189 }; 190 191 struct bpf_map_off_arr { 192 u32 cnt; 193 u32 field_off[BPF_MAP_OFF_ARR_MAX]; 194 u8 field_sz[BPF_MAP_OFF_ARR_MAX]; 195 }; 196 197 struct bpf_map { 198 /* The first two cachelines with read-mostly members of which some 199 * are also accessed in fast-path (e.g. ops, max_entries). 200 */ 201 const struct bpf_map_ops *ops ____cacheline_aligned; 202 struct bpf_map *inner_map_meta; 203 #ifdef CONFIG_SECURITY 204 void *security; 205 #endif 206 enum bpf_map_type map_type; 207 u32 key_size; 208 u32 value_size; 209 u32 max_entries; 210 u64 map_extra; /* any per-map-type extra fields */ 211 u32 map_flags; 212 int spin_lock_off; /* >=0 valid offset, <0 error */ 213 struct bpf_map_value_off *kptr_off_tab; 214 int timer_off; /* >=0 valid offset, <0 error */ 215 u32 id; 216 int numa_node; 217 u32 btf_key_type_id; 218 u32 btf_value_type_id; 219 u32 btf_vmlinux_value_type_id; 220 struct btf *btf; 221 #ifdef CONFIG_MEMCG_KMEM 222 struct mem_cgroup *memcg; 223 #endif 224 char name[BPF_OBJ_NAME_LEN]; 225 struct bpf_map_off_arr *off_arr; 226 /* The 3rd and 4th cacheline with misc members to avoid false sharing 227 * particularly with refcounting. 228 */ 229 atomic64_t refcnt ____cacheline_aligned; 230 atomic64_t usercnt; 231 struct work_struct work; 232 struct mutex freeze_mutex; 233 atomic64_t writecnt; 234 /* 'Ownership' of program-containing map is claimed by the first program 235 * that is going to use this map or by the first program which FD is 236 * stored in the map to make sure that all callers and callees have the 237 * same prog type, JITed flag and xdp_has_frags flag. 238 */ 239 struct { 240 spinlock_t lock; 241 enum bpf_prog_type type; 242 bool jited; 243 bool xdp_has_frags; 244 } owner; 245 bool bypass_spec_v1; 246 bool frozen; /* write-once; write-protected by freeze_mutex */ 247 }; 248 249 static inline bool map_value_has_spin_lock(const struct bpf_map *map) 250 { 251 return map->spin_lock_off >= 0; 252 } 253 254 static inline bool map_value_has_timer(const struct bpf_map *map) 255 { 256 return map->timer_off >= 0; 257 } 258 259 static inline bool map_value_has_kptrs(const struct bpf_map *map) 260 { 261 return !IS_ERR_OR_NULL(map->kptr_off_tab); 262 } 263 264 static inline void check_and_init_map_value(struct bpf_map *map, void *dst) 265 { 266 if (unlikely(map_value_has_spin_lock(map))) 267 memset(dst + map->spin_lock_off, 0, sizeof(struct bpf_spin_lock)); 268 if (unlikely(map_value_has_timer(map))) 269 memset(dst + map->timer_off, 0, sizeof(struct bpf_timer)); 270 if (unlikely(map_value_has_kptrs(map))) { 271 struct bpf_map_value_off *tab = map->kptr_off_tab; 272 int i; 273 274 for (i = 0; i < tab->nr_off; i++) 275 *(u64 *)(dst + tab->off[i].offset) = 0; 276 } 277 } 278 279 /* copy everything but bpf_spin_lock and bpf_timer. There could be one of each. */ 280 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 281 { 282 u32 curr_off = 0; 283 int i; 284 285 if (likely(!map->off_arr)) { 286 memcpy(dst, src, map->value_size); 287 return; 288 } 289 290 for (i = 0; i < map->off_arr->cnt; i++) { 291 u32 next_off = map->off_arr->field_off[i]; 292 293 memcpy(dst + curr_off, src + curr_off, next_off - curr_off); 294 curr_off += map->off_arr->field_sz[i]; 295 } 296 memcpy(dst + curr_off, src + curr_off, map->value_size - curr_off); 297 } 298 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 299 bool lock_src); 300 void bpf_timer_cancel_and_free(void *timer); 301 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size); 302 303 struct bpf_offload_dev; 304 struct bpf_offloaded_map; 305 306 struct bpf_map_dev_ops { 307 int (*map_get_next_key)(struct bpf_offloaded_map *map, 308 void *key, void *next_key); 309 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 310 void *key, void *value); 311 int (*map_update_elem)(struct bpf_offloaded_map *map, 312 void *key, void *value, u64 flags); 313 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 314 }; 315 316 struct bpf_offloaded_map { 317 struct bpf_map map; 318 struct net_device *netdev; 319 const struct bpf_map_dev_ops *dev_ops; 320 void *dev_priv; 321 struct list_head offloads; 322 }; 323 324 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 325 { 326 return container_of(map, struct bpf_offloaded_map, map); 327 } 328 329 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 330 { 331 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 332 } 333 334 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 335 { 336 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && 337 map->ops->map_seq_show_elem; 338 } 339 340 int map_check_no_btf(const struct bpf_map *map, 341 const struct btf *btf, 342 const struct btf_type *key_type, 343 const struct btf_type *value_type); 344 345 bool bpf_map_meta_equal(const struct bpf_map *meta0, 346 const struct bpf_map *meta1); 347 348 extern const struct bpf_map_ops bpf_map_offload_ops; 349 350 /* bpf_type_flag contains a set of flags that are applicable to the values of 351 * arg_type, ret_type and reg_type. For example, a pointer value may be null, 352 * or a memory is read-only. We classify types into two categories: base types 353 * and extended types. Extended types are base types combined with a type flag. 354 * 355 * Currently there are no more than 32 base types in arg_type, ret_type and 356 * reg_types. 357 */ 358 #define BPF_BASE_TYPE_BITS 8 359 360 enum bpf_type_flag { 361 /* PTR may be NULL. */ 362 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS), 363 364 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is 365 * compatible with both mutable and immutable memory. 366 */ 367 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS), 368 369 /* MEM was "allocated" from a different helper, and cannot be mixed 370 * with regular non-MEM_ALLOC'ed MEM types. 371 */ 372 MEM_ALLOC = BIT(2 + BPF_BASE_TYPE_BITS), 373 374 /* MEM is in user address space. */ 375 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS), 376 377 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged 378 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In 379 * order to drop this tag, it must be passed into bpf_per_cpu_ptr() 380 * or bpf_this_cpu_ptr(), which will return the pointer corresponding 381 * to the specified cpu. 382 */ 383 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS), 384 385 /* Indicates that the argument will be released. */ 386 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS), 387 388 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark 389 * unreferenced and referenced kptr loaded from map value using a load 390 * instruction, so that they can only be dereferenced but not escape the 391 * BPF program into the kernel (i.e. cannot be passed as arguments to 392 * kfunc or bpf helpers). 393 */ 394 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS), 395 396 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS), 397 398 /* DYNPTR points to memory local to the bpf program. */ 399 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS), 400 401 /* DYNPTR points to a ringbuf record. */ 402 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS), 403 404 /* Size is known at compile time. */ 405 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS), 406 407 __BPF_TYPE_FLAG_MAX, 408 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1, 409 }; 410 411 #define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF) 412 413 /* Max number of base types. */ 414 #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS) 415 416 /* Max number of all types. */ 417 #define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1)) 418 419 /* function argument constraints */ 420 enum bpf_arg_type { 421 ARG_DONTCARE = 0, /* unused argument in helper function */ 422 423 /* the following constraints used to prototype 424 * bpf_map_lookup/update/delete_elem() functions 425 */ 426 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 427 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 428 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 429 430 /* Used to prototype bpf_memcmp() and other functions that access data 431 * on eBPF program stack 432 */ 433 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 434 435 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 436 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 437 438 ARG_PTR_TO_CTX, /* pointer to context */ 439 ARG_ANYTHING, /* any (initialized) argument is ok */ 440 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 441 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 442 ARG_PTR_TO_INT, /* pointer to int */ 443 ARG_PTR_TO_LONG, /* pointer to long */ 444 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 445 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */ 446 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */ 447 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ 448 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ 449 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ 450 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */ 451 ARG_PTR_TO_STACK, /* pointer to stack */ 452 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */ 453 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */ 454 ARG_PTR_TO_KPTR, /* pointer to referenced kptr */ 455 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */ 456 __BPF_ARG_TYPE_MAX, 457 458 /* Extended arg_types. */ 459 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE, 460 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM, 461 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX, 462 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET, 463 ARG_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_ALLOC_MEM, 464 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK, 465 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID, 466 /* pointer to memory does not need to be initialized, helper function must fill 467 * all bytes or clear them in error case. 468 */ 469 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | ARG_PTR_TO_MEM, 470 /* Pointer to valid memory of size known at compile time. */ 471 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM, 472 473 /* This must be the last entry. Its purpose is to ensure the enum is 474 * wide enough to hold the higher bits reserved for bpf_type_flag. 475 */ 476 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT, 477 }; 478 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 479 480 /* type of values returned from helper functions */ 481 enum bpf_return_type { 482 RET_INTEGER, /* function returns integer */ 483 RET_VOID, /* function doesn't return anything */ 484 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 485 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */ 486 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */ 487 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */ 488 RET_PTR_TO_ALLOC_MEM, /* returns a pointer to dynamically allocated memory */ 489 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ 490 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */ 491 __BPF_RET_TYPE_MAX, 492 493 /* Extended ret_types. */ 494 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE, 495 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET, 496 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK, 497 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON, 498 RET_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_ALLOC | RET_PTR_TO_ALLOC_MEM, 499 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_ALLOC_MEM, 500 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID, 501 502 /* This must be the last entry. Its purpose is to ensure the enum is 503 * wide enough to hold the higher bits reserved for bpf_type_flag. 504 */ 505 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT, 506 }; 507 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 508 509 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 510 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 511 * instructions after verifying 512 */ 513 struct bpf_func_proto { 514 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 515 bool gpl_only; 516 bool pkt_access; 517 enum bpf_return_type ret_type; 518 union { 519 struct { 520 enum bpf_arg_type arg1_type; 521 enum bpf_arg_type arg2_type; 522 enum bpf_arg_type arg3_type; 523 enum bpf_arg_type arg4_type; 524 enum bpf_arg_type arg5_type; 525 }; 526 enum bpf_arg_type arg_type[5]; 527 }; 528 union { 529 struct { 530 u32 *arg1_btf_id; 531 u32 *arg2_btf_id; 532 u32 *arg3_btf_id; 533 u32 *arg4_btf_id; 534 u32 *arg5_btf_id; 535 }; 536 u32 *arg_btf_id[5]; 537 struct { 538 size_t arg1_size; 539 size_t arg2_size; 540 size_t arg3_size; 541 size_t arg4_size; 542 size_t arg5_size; 543 }; 544 size_t arg_size[5]; 545 }; 546 int *ret_btf_id; /* return value btf_id */ 547 bool (*allowed)(const struct bpf_prog *prog); 548 }; 549 550 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 551 * the first argument to eBPF programs. 552 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 553 */ 554 struct bpf_context; 555 556 enum bpf_access_type { 557 BPF_READ = 1, 558 BPF_WRITE = 2 559 }; 560 561 /* types of values stored in eBPF registers */ 562 /* Pointer types represent: 563 * pointer 564 * pointer + imm 565 * pointer + (u16) var 566 * pointer + (u16) var + imm 567 * if (range > 0) then [ptr, ptr + range - off) is safe to access 568 * if (id > 0) means that some 'var' was added 569 * if (off > 0) means that 'imm' was added 570 */ 571 enum bpf_reg_type { 572 NOT_INIT = 0, /* nothing was written into register */ 573 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 574 PTR_TO_CTX, /* reg points to bpf_context */ 575 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 576 PTR_TO_MAP_VALUE, /* reg points to map element value */ 577 PTR_TO_MAP_KEY, /* reg points to a map element key */ 578 PTR_TO_STACK, /* reg == frame_pointer + offset */ 579 PTR_TO_PACKET_META, /* skb->data - meta_len */ 580 PTR_TO_PACKET, /* reg points to skb->data */ 581 PTR_TO_PACKET_END, /* skb->data + headlen */ 582 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 583 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 584 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 585 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 586 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 587 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 588 /* PTR_TO_BTF_ID points to a kernel struct that does not need 589 * to be null checked by the BPF program. This does not imply the 590 * pointer is _not_ null and in practice this can easily be a null 591 * pointer when reading pointer chains. The assumption is program 592 * context will handle null pointer dereference typically via fault 593 * handling. The verifier must keep this in mind and can make no 594 * assumptions about null or non-null when doing branch analysis. 595 * Further, when passed into helpers the helpers can not, without 596 * additional context, assume the value is non-null. 597 */ 598 PTR_TO_BTF_ID, 599 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not 600 * been checked for null. Used primarily to inform the verifier 601 * an explicit null check is required for this struct. 602 */ 603 PTR_TO_MEM, /* reg points to valid memory region */ 604 PTR_TO_BUF, /* reg points to a read/write buffer */ 605 PTR_TO_FUNC, /* reg points to a bpf program function */ 606 __BPF_REG_TYPE_MAX, 607 608 /* Extended reg_types. */ 609 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE, 610 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET, 611 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON, 612 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK, 613 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID, 614 615 /* This must be the last entry. Its purpose is to ensure the enum is 616 * wide enough to hold the higher bits reserved for bpf_type_flag. 617 */ 618 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT, 619 }; 620 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 621 622 /* The information passed from prog-specific *_is_valid_access 623 * back to the verifier. 624 */ 625 struct bpf_insn_access_aux { 626 enum bpf_reg_type reg_type; 627 union { 628 int ctx_field_size; 629 struct { 630 struct btf *btf; 631 u32 btf_id; 632 }; 633 }; 634 struct bpf_verifier_log *log; /* for verbose logs */ 635 }; 636 637 static inline void 638 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 639 { 640 aux->ctx_field_size = size; 641 } 642 643 static inline bool bpf_pseudo_func(const struct bpf_insn *insn) 644 { 645 return insn->code == (BPF_LD | BPF_IMM | BPF_DW) && 646 insn->src_reg == BPF_PSEUDO_FUNC; 647 } 648 649 struct bpf_prog_ops { 650 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 651 union bpf_attr __user *uattr); 652 }; 653 654 struct bpf_verifier_ops { 655 /* return eBPF function prototype for verification */ 656 const struct bpf_func_proto * 657 (*get_func_proto)(enum bpf_func_id func_id, 658 const struct bpf_prog *prog); 659 660 /* return true if 'size' wide access at offset 'off' within bpf_context 661 * with 'type' (read or write) is allowed 662 */ 663 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 664 const struct bpf_prog *prog, 665 struct bpf_insn_access_aux *info); 666 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 667 const struct bpf_prog *prog); 668 int (*gen_ld_abs)(const struct bpf_insn *orig, 669 struct bpf_insn *insn_buf); 670 u32 (*convert_ctx_access)(enum bpf_access_type type, 671 const struct bpf_insn *src, 672 struct bpf_insn *dst, 673 struct bpf_prog *prog, u32 *target_size); 674 int (*btf_struct_access)(struct bpf_verifier_log *log, 675 const struct btf *btf, 676 const struct btf_type *t, int off, int size, 677 enum bpf_access_type atype, 678 u32 *next_btf_id, enum bpf_type_flag *flag); 679 }; 680 681 struct bpf_prog_offload_ops { 682 /* verifier basic callbacks */ 683 int (*insn_hook)(struct bpf_verifier_env *env, 684 int insn_idx, int prev_insn_idx); 685 int (*finalize)(struct bpf_verifier_env *env); 686 /* verifier optimization callbacks (called after .finalize) */ 687 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 688 struct bpf_insn *insn); 689 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 690 /* program management callbacks */ 691 int (*prepare)(struct bpf_prog *prog); 692 int (*translate)(struct bpf_prog *prog); 693 void (*destroy)(struct bpf_prog *prog); 694 }; 695 696 struct bpf_prog_offload { 697 struct bpf_prog *prog; 698 struct net_device *netdev; 699 struct bpf_offload_dev *offdev; 700 void *dev_priv; 701 struct list_head offloads; 702 bool dev_state; 703 bool opt_failed; 704 void *jited_image; 705 u32 jited_len; 706 }; 707 708 enum bpf_cgroup_storage_type { 709 BPF_CGROUP_STORAGE_SHARED, 710 BPF_CGROUP_STORAGE_PERCPU, 711 __BPF_CGROUP_STORAGE_MAX 712 }; 713 714 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 715 716 /* The longest tracepoint has 12 args. 717 * See include/trace/bpf_probe.h 718 */ 719 #define MAX_BPF_FUNC_ARGS 12 720 721 /* The maximum number of arguments passed through registers 722 * a single function may have. 723 */ 724 #define MAX_BPF_FUNC_REG_ARGS 5 725 726 struct btf_func_model { 727 u8 ret_size; 728 u8 nr_args; 729 u8 arg_size[MAX_BPF_FUNC_ARGS]; 730 }; 731 732 /* Restore arguments before returning from trampoline to let original function 733 * continue executing. This flag is used for fentry progs when there are no 734 * fexit progs. 735 */ 736 #define BPF_TRAMP_F_RESTORE_REGS BIT(0) 737 /* Call original function after fentry progs, but before fexit progs. 738 * Makes sense for fentry/fexit, normal calls and indirect calls. 739 */ 740 #define BPF_TRAMP_F_CALL_ORIG BIT(1) 741 /* Skip current frame and return to parent. Makes sense for fentry/fexit 742 * programs only. Should not be used with normal calls and indirect calls. 743 */ 744 #define BPF_TRAMP_F_SKIP_FRAME BIT(2) 745 /* Store IP address of the caller on the trampoline stack, 746 * so it's available for trampoline's programs. 747 */ 748 #define BPF_TRAMP_F_IP_ARG BIT(3) 749 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */ 750 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4) 751 752 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 753 * bytes on x86. 754 */ 755 #define BPF_MAX_TRAMP_LINKS 38 756 757 struct bpf_tramp_links { 758 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS]; 759 int nr_links; 760 }; 761 762 struct bpf_tramp_run_ctx; 763 764 /* Different use cases for BPF trampoline: 765 * 1. replace nop at the function entry (kprobe equivalent) 766 * flags = BPF_TRAMP_F_RESTORE_REGS 767 * fentry = a set of programs to run before returning from trampoline 768 * 769 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 770 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 771 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 772 * fentry = a set of program to run before calling original function 773 * fexit = a set of program to run after original function 774 * 775 * 3. replace direct call instruction anywhere in the function body 776 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 777 * With flags = 0 778 * fentry = a set of programs to run before returning from trampoline 779 * With flags = BPF_TRAMP_F_CALL_ORIG 780 * orig_call = original callback addr or direct function addr 781 * fentry = a set of program to run before calling original function 782 * fexit = a set of program to run after original function 783 */ 784 struct bpf_tramp_image; 785 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end, 786 const struct btf_func_model *m, u32 flags, 787 struct bpf_tramp_links *tlinks, 788 void *orig_call); 789 /* these two functions are called from generated trampoline */ 790 u64 notrace __bpf_prog_enter(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx); 791 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start, struct bpf_tramp_run_ctx *run_ctx); 792 u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx); 793 void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start, 794 struct bpf_tramp_run_ctx *run_ctx); 795 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr); 796 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr); 797 798 struct bpf_ksym { 799 unsigned long start; 800 unsigned long end; 801 char name[KSYM_NAME_LEN]; 802 struct list_head lnode; 803 struct latch_tree_node tnode; 804 bool prog; 805 }; 806 807 enum bpf_tramp_prog_type { 808 BPF_TRAMP_FENTRY, 809 BPF_TRAMP_FEXIT, 810 BPF_TRAMP_MODIFY_RETURN, 811 BPF_TRAMP_MAX, 812 BPF_TRAMP_REPLACE, /* more than MAX */ 813 }; 814 815 struct bpf_tramp_image { 816 void *image; 817 struct bpf_ksym ksym; 818 struct percpu_ref pcref; 819 void *ip_after_call; 820 void *ip_epilogue; 821 union { 822 struct rcu_head rcu; 823 struct work_struct work; 824 }; 825 }; 826 827 struct bpf_trampoline { 828 /* hlist for trampoline_table */ 829 struct hlist_node hlist; 830 /* serializes access to fields of this trampoline */ 831 struct mutex mutex; 832 refcount_t refcnt; 833 u64 key; 834 struct { 835 struct btf_func_model model; 836 void *addr; 837 bool ftrace_managed; 838 } func; 839 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 840 * program by replacing one of its functions. func.addr is the address 841 * of the function it replaced. 842 */ 843 struct bpf_prog *extension_prog; 844 /* list of BPF programs using this trampoline */ 845 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 846 /* Number of attached programs. A counter per kind. */ 847 int progs_cnt[BPF_TRAMP_MAX]; 848 /* Executable image of trampoline */ 849 struct bpf_tramp_image *cur_image; 850 u64 selector; 851 struct module *mod; 852 }; 853 854 struct bpf_attach_target_info { 855 struct btf_func_model fmodel; 856 long tgt_addr; 857 const char *tgt_name; 858 const struct btf_type *tgt_type; 859 }; 860 861 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 862 863 struct bpf_dispatcher_prog { 864 struct bpf_prog *prog; 865 refcount_t users; 866 }; 867 868 struct bpf_dispatcher { 869 /* dispatcher mutex */ 870 struct mutex mutex; 871 void *func; 872 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 873 int num_progs; 874 void *image; 875 u32 image_off; 876 struct bpf_ksym ksym; 877 }; 878 879 static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func( 880 const void *ctx, 881 const struct bpf_insn *insnsi, 882 unsigned int (*bpf_func)(const void *, 883 const struct bpf_insn *)) 884 { 885 return bpf_func(ctx, insnsi); 886 } 887 888 #ifdef CONFIG_BPF_JIT 889 int bpf_trampoline_link_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr); 890 int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr); 891 struct bpf_trampoline *bpf_trampoline_get(u64 key, 892 struct bpf_attach_target_info *tgt_info); 893 void bpf_trampoline_put(struct bpf_trampoline *tr); 894 int arch_prepare_bpf_dispatcher(void *image, s64 *funcs, int num_funcs); 895 #define BPF_DISPATCHER_INIT(_name) { \ 896 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 897 .func = &_name##_func, \ 898 .progs = {}, \ 899 .num_progs = 0, \ 900 .image = NULL, \ 901 .image_off = 0, \ 902 .ksym = { \ 903 .name = #_name, \ 904 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 905 }, \ 906 } 907 908 #define DEFINE_BPF_DISPATCHER(name) \ 909 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \ 910 const void *ctx, \ 911 const struct bpf_insn *insnsi, \ 912 unsigned int (*bpf_func)(const void *, \ 913 const struct bpf_insn *)) \ 914 { \ 915 return bpf_func(ctx, insnsi); \ 916 } \ 917 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 918 struct bpf_dispatcher bpf_dispatcher_##name = \ 919 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 920 #define DECLARE_BPF_DISPATCHER(name) \ 921 unsigned int bpf_dispatcher_##name##_func( \ 922 const void *ctx, \ 923 const struct bpf_insn *insnsi, \ 924 unsigned int (*bpf_func)(const void *, \ 925 const struct bpf_insn *)); \ 926 extern struct bpf_dispatcher bpf_dispatcher_##name; 927 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 928 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 929 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 930 struct bpf_prog *to); 931 /* Called only from JIT-enabled code, so there's no need for stubs. */ 932 void *bpf_jit_alloc_exec_page(void); 933 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym); 934 void bpf_image_ksym_del(struct bpf_ksym *ksym); 935 void bpf_ksym_add(struct bpf_ksym *ksym); 936 void bpf_ksym_del(struct bpf_ksym *ksym); 937 int bpf_jit_charge_modmem(u32 size); 938 void bpf_jit_uncharge_modmem(u32 size); 939 bool bpf_prog_has_trampoline(const struct bpf_prog *prog); 940 #else 941 static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link, 942 struct bpf_trampoline *tr) 943 { 944 return -ENOTSUPP; 945 } 946 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, 947 struct bpf_trampoline *tr) 948 { 949 return -ENOTSUPP; 950 } 951 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 952 struct bpf_attach_target_info *tgt_info) 953 { 954 return ERR_PTR(-EOPNOTSUPP); 955 } 956 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 957 #define DEFINE_BPF_DISPATCHER(name) 958 #define DECLARE_BPF_DISPATCHER(name) 959 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 960 #define BPF_DISPATCHER_PTR(name) NULL 961 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 962 struct bpf_prog *from, 963 struct bpf_prog *to) {} 964 static inline bool is_bpf_image_address(unsigned long address) 965 { 966 return false; 967 } 968 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog) 969 { 970 return false; 971 } 972 #endif 973 974 struct bpf_func_info_aux { 975 u16 linkage; 976 bool unreliable; 977 }; 978 979 enum bpf_jit_poke_reason { 980 BPF_POKE_REASON_TAIL_CALL, 981 }; 982 983 /* Descriptor of pokes pointing /into/ the JITed image. */ 984 struct bpf_jit_poke_descriptor { 985 void *tailcall_target; 986 void *tailcall_bypass; 987 void *bypass_addr; 988 void *aux; 989 union { 990 struct { 991 struct bpf_map *map; 992 u32 key; 993 } tail_call; 994 }; 995 bool tailcall_target_stable; 996 u8 adj_off; 997 u16 reason; 998 u32 insn_idx; 999 }; 1000 1001 /* reg_type info for ctx arguments */ 1002 struct bpf_ctx_arg_aux { 1003 u32 offset; 1004 enum bpf_reg_type reg_type; 1005 u32 btf_id; 1006 }; 1007 1008 struct btf_mod_pair { 1009 struct btf *btf; 1010 struct module *module; 1011 }; 1012 1013 struct bpf_kfunc_desc_tab; 1014 1015 struct bpf_prog_aux { 1016 atomic64_t refcnt; 1017 u32 used_map_cnt; 1018 u32 used_btf_cnt; 1019 u32 max_ctx_offset; 1020 u32 max_pkt_offset; 1021 u32 max_tp_access; 1022 u32 stack_depth; 1023 u32 id; 1024 u32 func_cnt; /* used by non-func prog as the number of func progs */ 1025 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 1026 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1027 u32 ctx_arg_info_size; 1028 u32 max_rdonly_access; 1029 u32 max_rdwr_access; 1030 struct btf *attach_btf; 1031 const struct bpf_ctx_arg_aux *ctx_arg_info; 1032 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 1033 struct bpf_prog *dst_prog; 1034 struct bpf_trampoline *dst_trampoline; 1035 enum bpf_prog_type saved_dst_prog_type; 1036 enum bpf_attach_type saved_dst_attach_type; 1037 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 1038 bool offload_requested; 1039 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 1040 bool func_proto_unreliable; 1041 bool sleepable; 1042 bool tail_call_reachable; 1043 bool xdp_has_frags; 1044 bool use_bpf_prog_pack; 1045 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 1046 const struct btf_type *attach_func_proto; 1047 /* function name for valid attach_btf_id */ 1048 const char *attach_func_name; 1049 struct bpf_prog **func; 1050 void *jit_data; /* JIT specific data. arch dependent */ 1051 struct bpf_jit_poke_descriptor *poke_tab; 1052 struct bpf_kfunc_desc_tab *kfunc_tab; 1053 struct bpf_kfunc_btf_tab *kfunc_btf_tab; 1054 u32 size_poke_tab; 1055 struct bpf_ksym ksym; 1056 const struct bpf_prog_ops *ops; 1057 struct bpf_map **used_maps; 1058 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 1059 struct btf_mod_pair *used_btfs; 1060 struct bpf_prog *prog; 1061 struct user_struct *user; 1062 u64 load_time; /* ns since boottime */ 1063 u32 verified_insns; 1064 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1065 char name[BPF_OBJ_NAME_LEN]; 1066 #ifdef CONFIG_SECURITY 1067 void *security; 1068 #endif 1069 struct bpf_prog_offload *offload; 1070 struct btf *btf; 1071 struct bpf_func_info *func_info; 1072 struct bpf_func_info_aux *func_info_aux; 1073 /* bpf_line_info loaded from userspace. linfo->insn_off 1074 * has the xlated insn offset. 1075 * Both the main and sub prog share the same linfo. 1076 * The subprog can access its first linfo by 1077 * using the linfo_idx. 1078 */ 1079 struct bpf_line_info *linfo; 1080 /* jited_linfo is the jited addr of the linfo. It has a 1081 * one to one mapping to linfo: 1082 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 1083 * Both the main and sub prog share the same jited_linfo. 1084 * The subprog can access its first jited_linfo by 1085 * using the linfo_idx. 1086 */ 1087 void **jited_linfo; 1088 u32 func_info_cnt; 1089 u32 nr_linfo; 1090 /* subprog can use linfo_idx to access its first linfo and 1091 * jited_linfo. 1092 * main prog always has linfo_idx == 0 1093 */ 1094 u32 linfo_idx; 1095 u32 num_exentries; 1096 struct exception_table_entry *extable; 1097 union { 1098 struct work_struct work; 1099 struct rcu_head rcu; 1100 }; 1101 }; 1102 1103 struct bpf_prog { 1104 u16 pages; /* Number of allocated pages */ 1105 u16 jited:1, /* Is our filter JIT'ed? */ 1106 jit_requested:1,/* archs need to JIT the prog */ 1107 gpl_compatible:1, /* Is filter GPL compatible? */ 1108 cb_access:1, /* Is control block accessed? */ 1109 dst_needed:1, /* Do we need dst entry? */ 1110 blinding_requested:1, /* needs constant blinding */ 1111 blinded:1, /* Was blinded */ 1112 is_func:1, /* program is a bpf function */ 1113 kprobe_override:1, /* Do we override a kprobe? */ 1114 has_callchain_buf:1, /* callchain buffer allocated? */ 1115 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */ 1116 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */ 1117 call_get_func_ip:1, /* Do we call get_func_ip() */ 1118 tstamp_type_access:1; /* Accessed __sk_buff->tstamp_type */ 1119 enum bpf_prog_type type; /* Type of BPF program */ 1120 enum bpf_attach_type expected_attach_type; /* For some prog types */ 1121 u32 len; /* Number of filter blocks */ 1122 u32 jited_len; /* Size of jited insns in bytes */ 1123 u8 tag[BPF_TAG_SIZE]; 1124 struct bpf_prog_stats __percpu *stats; 1125 int __percpu *active; 1126 unsigned int (*bpf_func)(const void *ctx, 1127 const struct bpf_insn *insn); 1128 struct bpf_prog_aux *aux; /* Auxiliary fields */ 1129 struct sock_fprog_kern *orig_prog; /* Original BPF program */ 1130 /* Instructions for interpreter */ 1131 union { 1132 DECLARE_FLEX_ARRAY(struct sock_filter, insns); 1133 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi); 1134 }; 1135 }; 1136 1137 struct bpf_array_aux { 1138 /* Programs with direct jumps into programs part of this array. */ 1139 struct list_head poke_progs; 1140 struct bpf_map *map; 1141 struct mutex poke_mutex; 1142 struct work_struct work; 1143 }; 1144 1145 struct bpf_link { 1146 atomic64_t refcnt; 1147 u32 id; 1148 enum bpf_link_type type; 1149 const struct bpf_link_ops *ops; 1150 struct bpf_prog *prog; 1151 struct work_struct work; 1152 }; 1153 1154 struct bpf_link_ops { 1155 void (*release)(struct bpf_link *link); 1156 void (*dealloc)(struct bpf_link *link); 1157 int (*detach)(struct bpf_link *link); 1158 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 1159 struct bpf_prog *old_prog); 1160 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 1161 int (*fill_link_info)(const struct bpf_link *link, 1162 struct bpf_link_info *info); 1163 }; 1164 1165 struct bpf_tramp_link { 1166 struct bpf_link link; 1167 struct hlist_node tramp_hlist; 1168 u64 cookie; 1169 }; 1170 1171 struct bpf_tracing_link { 1172 struct bpf_tramp_link link; 1173 enum bpf_attach_type attach_type; 1174 struct bpf_trampoline *trampoline; 1175 struct bpf_prog *tgt_prog; 1176 }; 1177 1178 struct bpf_link_primer { 1179 struct bpf_link *link; 1180 struct file *file; 1181 int fd; 1182 u32 id; 1183 }; 1184 1185 struct bpf_struct_ops_value; 1186 struct btf_member; 1187 1188 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 1189 struct bpf_struct_ops { 1190 const struct bpf_verifier_ops *verifier_ops; 1191 int (*init)(struct btf *btf); 1192 int (*check_member)(const struct btf_type *t, 1193 const struct btf_member *member); 1194 int (*init_member)(const struct btf_type *t, 1195 const struct btf_member *member, 1196 void *kdata, const void *udata); 1197 int (*reg)(void *kdata); 1198 void (*unreg)(void *kdata); 1199 const struct btf_type *type; 1200 const struct btf_type *value_type; 1201 const char *name; 1202 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 1203 u32 type_id; 1204 u32 value_id; 1205 }; 1206 1207 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 1208 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 1209 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id); 1210 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log); 1211 bool bpf_struct_ops_get(const void *kdata); 1212 void bpf_struct_ops_put(const void *kdata); 1213 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 1214 void *value); 1215 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks, 1216 struct bpf_tramp_link *link, 1217 const struct btf_func_model *model, 1218 void *image, void *image_end); 1219 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1220 { 1221 if (owner == BPF_MODULE_OWNER) 1222 return bpf_struct_ops_get(data); 1223 else 1224 return try_module_get(owner); 1225 } 1226 static inline void bpf_module_put(const void *data, struct module *owner) 1227 { 1228 if (owner == BPF_MODULE_OWNER) 1229 bpf_struct_ops_put(data); 1230 else 1231 module_put(owner); 1232 } 1233 1234 #ifdef CONFIG_NET 1235 /* Define it here to avoid the use of forward declaration */ 1236 struct bpf_dummy_ops_state { 1237 int val; 1238 }; 1239 1240 struct bpf_dummy_ops { 1241 int (*test_1)(struct bpf_dummy_ops_state *cb); 1242 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2, 1243 char a3, unsigned long a4); 1244 }; 1245 1246 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr, 1247 union bpf_attr __user *uattr); 1248 #endif 1249 #else 1250 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id) 1251 { 1252 return NULL; 1253 } 1254 static inline void bpf_struct_ops_init(struct btf *btf, 1255 struct bpf_verifier_log *log) 1256 { 1257 } 1258 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1259 { 1260 return try_module_get(owner); 1261 } 1262 static inline void bpf_module_put(const void *data, struct module *owner) 1263 { 1264 module_put(owner); 1265 } 1266 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 1267 void *key, 1268 void *value) 1269 { 1270 return -EINVAL; 1271 } 1272 #endif 1273 1274 struct bpf_array { 1275 struct bpf_map map; 1276 u32 elem_size; 1277 u32 index_mask; 1278 struct bpf_array_aux *aux; 1279 union { 1280 char value[0] __aligned(8); 1281 void *ptrs[0] __aligned(8); 1282 void __percpu *pptrs[0] __aligned(8); 1283 }; 1284 }; 1285 1286 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 1287 #define MAX_TAIL_CALL_CNT 33 1288 1289 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 1290 BPF_F_RDONLY_PROG | \ 1291 BPF_F_WRONLY | \ 1292 BPF_F_WRONLY_PROG) 1293 1294 #define BPF_MAP_CAN_READ BIT(0) 1295 #define BPF_MAP_CAN_WRITE BIT(1) 1296 1297 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 1298 { 1299 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1300 1301 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 1302 * not possible. 1303 */ 1304 if (access_flags & BPF_F_RDONLY_PROG) 1305 return BPF_MAP_CAN_READ; 1306 else if (access_flags & BPF_F_WRONLY_PROG) 1307 return BPF_MAP_CAN_WRITE; 1308 else 1309 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 1310 } 1311 1312 static inline bool bpf_map_flags_access_ok(u32 access_flags) 1313 { 1314 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 1315 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1316 } 1317 1318 struct bpf_event_entry { 1319 struct perf_event *event; 1320 struct file *perf_file; 1321 struct file *map_file; 1322 struct rcu_head rcu; 1323 }; 1324 1325 static inline bool map_type_contains_progs(struct bpf_map *map) 1326 { 1327 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY || 1328 map->map_type == BPF_MAP_TYPE_DEVMAP || 1329 map->map_type == BPF_MAP_TYPE_CPUMAP; 1330 } 1331 1332 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp); 1333 int bpf_prog_calc_tag(struct bpf_prog *fp); 1334 1335 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 1336 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void); 1337 1338 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 1339 unsigned long off, unsigned long len); 1340 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 1341 const struct bpf_insn *src, 1342 struct bpf_insn *dst, 1343 struct bpf_prog *prog, 1344 u32 *target_size); 1345 1346 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 1347 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 1348 1349 /* an array of programs to be executed under rcu_lock. 1350 * 1351 * Typical usage: 1352 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run); 1353 * 1354 * the structure returned by bpf_prog_array_alloc() should be populated 1355 * with program pointers and the last pointer must be NULL. 1356 * The user has to keep refcnt on the program and make sure the program 1357 * is removed from the array before bpf_prog_put(). 1358 * The 'struct bpf_prog_array *' should only be replaced with xchg() 1359 * since other cpus are walking the array of pointers in parallel. 1360 */ 1361 struct bpf_prog_array_item { 1362 struct bpf_prog *prog; 1363 union { 1364 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1365 u64 bpf_cookie; 1366 }; 1367 }; 1368 1369 struct bpf_prog_array { 1370 struct rcu_head rcu; 1371 struct bpf_prog_array_item items[]; 1372 }; 1373 1374 struct bpf_empty_prog_array { 1375 struct bpf_prog_array hdr; 1376 struct bpf_prog *null_prog; 1377 }; 1378 1379 /* to avoid allocating empty bpf_prog_array for cgroups that 1380 * don't have bpf program attached use one global 'bpf_empty_prog_array' 1381 * It will not be modified the caller of bpf_prog_array_alloc() 1382 * (since caller requested prog_cnt == 0) 1383 * that pointer should be 'freed' by bpf_prog_array_free() 1384 */ 1385 extern struct bpf_empty_prog_array bpf_empty_prog_array; 1386 1387 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 1388 void bpf_prog_array_free(struct bpf_prog_array *progs); 1389 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */ 1390 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs); 1391 int bpf_prog_array_length(struct bpf_prog_array *progs); 1392 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 1393 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 1394 __u32 __user *prog_ids, u32 cnt); 1395 1396 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 1397 struct bpf_prog *old_prog); 1398 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 1399 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 1400 struct bpf_prog *prog); 1401 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 1402 u32 *prog_ids, u32 request_cnt, 1403 u32 *prog_cnt); 1404 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 1405 struct bpf_prog *exclude_prog, 1406 struct bpf_prog *include_prog, 1407 u64 bpf_cookie, 1408 struct bpf_prog_array **new_array); 1409 1410 struct bpf_run_ctx {}; 1411 1412 struct bpf_cg_run_ctx { 1413 struct bpf_run_ctx run_ctx; 1414 const struct bpf_prog_array_item *prog_item; 1415 int retval; 1416 }; 1417 1418 struct bpf_trace_run_ctx { 1419 struct bpf_run_ctx run_ctx; 1420 u64 bpf_cookie; 1421 }; 1422 1423 struct bpf_tramp_run_ctx { 1424 struct bpf_run_ctx run_ctx; 1425 u64 bpf_cookie; 1426 struct bpf_run_ctx *saved_run_ctx; 1427 }; 1428 1429 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 1430 { 1431 struct bpf_run_ctx *old_ctx = NULL; 1432 1433 #ifdef CONFIG_BPF_SYSCALL 1434 old_ctx = current->bpf_ctx; 1435 current->bpf_ctx = new_ctx; 1436 #endif 1437 return old_ctx; 1438 } 1439 1440 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 1441 { 1442 #ifdef CONFIG_BPF_SYSCALL 1443 current->bpf_ctx = old_ctx; 1444 #endif 1445 } 1446 1447 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 1448 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 1449 /* BPF program asks to set CN on the packet. */ 1450 #define BPF_RET_SET_CN (1 << 0) 1451 1452 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 1453 1454 static __always_inline u32 1455 bpf_prog_run_array(const struct bpf_prog_array *array, 1456 const void *ctx, bpf_prog_run_fn run_prog) 1457 { 1458 const struct bpf_prog_array_item *item; 1459 const struct bpf_prog *prog; 1460 struct bpf_run_ctx *old_run_ctx; 1461 struct bpf_trace_run_ctx run_ctx; 1462 u32 ret = 1; 1463 1464 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held"); 1465 1466 if (unlikely(!array)) 1467 return ret; 1468 1469 migrate_disable(); 1470 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1471 item = &array->items[0]; 1472 while ((prog = READ_ONCE(item->prog))) { 1473 run_ctx.bpf_cookie = item->bpf_cookie; 1474 ret &= run_prog(prog, ctx); 1475 item++; 1476 } 1477 bpf_reset_run_ctx(old_run_ctx); 1478 migrate_enable(); 1479 return ret; 1480 } 1481 1482 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs: 1483 * 1484 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array 1485 * overall. As a result, we must use the bpf_prog_array_free_sleepable 1486 * in order to use the tasks_trace rcu grace period. 1487 * 1488 * When a non-sleepable program is inside the array, we take the rcu read 1489 * section and disable preemption for that program alone, so it can access 1490 * rcu-protected dynamically sized maps. 1491 */ 1492 static __always_inline u32 1493 bpf_prog_run_array_sleepable(const struct bpf_prog_array __rcu *array_rcu, 1494 const void *ctx, bpf_prog_run_fn run_prog) 1495 { 1496 const struct bpf_prog_array_item *item; 1497 const struct bpf_prog *prog; 1498 const struct bpf_prog_array *array; 1499 struct bpf_run_ctx *old_run_ctx; 1500 struct bpf_trace_run_ctx run_ctx; 1501 u32 ret = 1; 1502 1503 might_fault(); 1504 1505 rcu_read_lock_trace(); 1506 migrate_disable(); 1507 1508 array = rcu_dereference_check(array_rcu, rcu_read_lock_trace_held()); 1509 if (unlikely(!array)) 1510 goto out; 1511 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1512 item = &array->items[0]; 1513 while ((prog = READ_ONCE(item->prog))) { 1514 if (!prog->aux->sleepable) 1515 rcu_read_lock(); 1516 1517 run_ctx.bpf_cookie = item->bpf_cookie; 1518 ret &= run_prog(prog, ctx); 1519 item++; 1520 1521 if (!prog->aux->sleepable) 1522 rcu_read_unlock(); 1523 } 1524 bpf_reset_run_ctx(old_run_ctx); 1525 out: 1526 migrate_enable(); 1527 rcu_read_unlock_trace(); 1528 return ret; 1529 } 1530 1531 #ifdef CONFIG_BPF_SYSCALL 1532 DECLARE_PER_CPU(int, bpf_prog_active); 1533 extern struct mutex bpf_stats_enabled_mutex; 1534 1535 /* 1536 * Block execution of BPF programs attached to instrumentation (perf, 1537 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 1538 * these events can happen inside a region which holds a map bucket lock 1539 * and can deadlock on it. 1540 */ 1541 static inline void bpf_disable_instrumentation(void) 1542 { 1543 migrate_disable(); 1544 this_cpu_inc(bpf_prog_active); 1545 } 1546 1547 static inline void bpf_enable_instrumentation(void) 1548 { 1549 this_cpu_dec(bpf_prog_active); 1550 migrate_enable(); 1551 } 1552 1553 extern const struct file_operations bpf_map_fops; 1554 extern const struct file_operations bpf_prog_fops; 1555 extern const struct file_operations bpf_iter_fops; 1556 1557 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 1558 extern const struct bpf_prog_ops _name ## _prog_ops; \ 1559 extern const struct bpf_verifier_ops _name ## _verifier_ops; 1560 #define BPF_MAP_TYPE(_id, _ops) \ 1561 extern const struct bpf_map_ops _ops; 1562 #define BPF_LINK_TYPE(_id, _name) 1563 #include <linux/bpf_types.h> 1564 #undef BPF_PROG_TYPE 1565 #undef BPF_MAP_TYPE 1566 #undef BPF_LINK_TYPE 1567 1568 extern const struct bpf_prog_ops bpf_offload_prog_ops; 1569 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 1570 extern const struct bpf_verifier_ops xdp_analyzer_ops; 1571 1572 struct bpf_prog *bpf_prog_get(u32 ufd); 1573 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 1574 bool attach_drv); 1575 void bpf_prog_add(struct bpf_prog *prog, int i); 1576 void bpf_prog_sub(struct bpf_prog *prog, int i); 1577 void bpf_prog_inc(struct bpf_prog *prog); 1578 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 1579 void bpf_prog_put(struct bpf_prog *prog); 1580 1581 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 1582 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 1583 1584 struct bpf_map_value_off_desc *bpf_map_kptr_off_contains(struct bpf_map *map, u32 offset); 1585 void bpf_map_free_kptr_off_tab(struct bpf_map *map); 1586 struct bpf_map_value_off *bpf_map_copy_kptr_off_tab(const struct bpf_map *map); 1587 bool bpf_map_equal_kptr_off_tab(const struct bpf_map *map_a, const struct bpf_map *map_b); 1588 void bpf_map_free_kptrs(struct bpf_map *map, void *map_value); 1589 1590 struct bpf_map *bpf_map_get(u32 ufd); 1591 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 1592 struct bpf_map *__bpf_map_get(struct fd f); 1593 void bpf_map_inc(struct bpf_map *map); 1594 void bpf_map_inc_with_uref(struct bpf_map *map); 1595 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 1596 void bpf_map_put_with_uref(struct bpf_map *map); 1597 void bpf_map_put(struct bpf_map *map); 1598 void *bpf_map_area_alloc(u64 size, int numa_node); 1599 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 1600 void bpf_map_area_free(void *base); 1601 bool bpf_map_write_active(const struct bpf_map *map); 1602 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 1603 int generic_map_lookup_batch(struct bpf_map *map, 1604 const union bpf_attr *attr, 1605 union bpf_attr __user *uattr); 1606 int generic_map_update_batch(struct bpf_map *map, 1607 const union bpf_attr *attr, 1608 union bpf_attr __user *uattr); 1609 int generic_map_delete_batch(struct bpf_map *map, 1610 const union bpf_attr *attr, 1611 union bpf_attr __user *uattr); 1612 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 1613 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 1614 1615 #ifdef CONFIG_MEMCG_KMEM 1616 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1617 int node); 1618 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 1619 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 1620 size_t align, gfp_t flags); 1621 #else 1622 static inline void * 1623 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1624 int node) 1625 { 1626 return kmalloc_node(size, flags, node); 1627 } 1628 1629 static inline void * 1630 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags) 1631 { 1632 return kzalloc(size, flags); 1633 } 1634 1635 static inline void __percpu * 1636 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align, 1637 gfp_t flags) 1638 { 1639 return __alloc_percpu_gfp(size, align, flags); 1640 } 1641 #endif 1642 1643 extern int sysctl_unprivileged_bpf_disabled; 1644 1645 static inline bool bpf_allow_ptr_leaks(void) 1646 { 1647 return perfmon_capable(); 1648 } 1649 1650 static inline bool bpf_allow_uninit_stack(void) 1651 { 1652 return perfmon_capable(); 1653 } 1654 1655 static inline bool bpf_allow_ptr_to_map_access(void) 1656 { 1657 return perfmon_capable(); 1658 } 1659 1660 static inline bool bpf_bypass_spec_v1(void) 1661 { 1662 return perfmon_capable(); 1663 } 1664 1665 static inline bool bpf_bypass_spec_v4(void) 1666 { 1667 return perfmon_capable(); 1668 } 1669 1670 int bpf_map_new_fd(struct bpf_map *map, int flags); 1671 int bpf_prog_new_fd(struct bpf_prog *prog); 1672 1673 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1674 const struct bpf_link_ops *ops, struct bpf_prog *prog); 1675 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 1676 int bpf_link_settle(struct bpf_link_primer *primer); 1677 void bpf_link_cleanup(struct bpf_link_primer *primer); 1678 void bpf_link_inc(struct bpf_link *link); 1679 void bpf_link_put(struct bpf_link *link); 1680 int bpf_link_new_fd(struct bpf_link *link); 1681 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd); 1682 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 1683 struct bpf_link *bpf_link_get_curr_or_next(u32 *id); 1684 1685 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 1686 int bpf_obj_get_user(const char __user *pathname, int flags); 1687 1688 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 1689 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 1690 extern int bpf_iter_ ## target(args); \ 1691 int __init bpf_iter_ ## target(args) { return 0; } 1692 1693 struct bpf_iter_aux_info { 1694 struct bpf_map *map; 1695 }; 1696 1697 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 1698 union bpf_iter_link_info *linfo, 1699 struct bpf_iter_aux_info *aux); 1700 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 1701 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 1702 struct seq_file *seq); 1703 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 1704 struct bpf_link_info *info); 1705 typedef const struct bpf_func_proto * 1706 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 1707 const struct bpf_prog *prog); 1708 1709 enum bpf_iter_feature { 1710 BPF_ITER_RESCHED = BIT(0), 1711 }; 1712 1713 #define BPF_ITER_CTX_ARG_MAX 2 1714 struct bpf_iter_reg { 1715 const char *target; 1716 bpf_iter_attach_target_t attach_target; 1717 bpf_iter_detach_target_t detach_target; 1718 bpf_iter_show_fdinfo_t show_fdinfo; 1719 bpf_iter_fill_link_info_t fill_link_info; 1720 bpf_iter_get_func_proto_t get_func_proto; 1721 u32 ctx_arg_info_size; 1722 u32 feature; 1723 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 1724 const struct bpf_iter_seq_info *seq_info; 1725 }; 1726 1727 struct bpf_iter_meta { 1728 __bpf_md_ptr(struct seq_file *, seq); 1729 u64 session_id; 1730 u64 seq_num; 1731 }; 1732 1733 struct bpf_iter__bpf_map_elem { 1734 __bpf_md_ptr(struct bpf_iter_meta *, meta); 1735 __bpf_md_ptr(struct bpf_map *, map); 1736 __bpf_md_ptr(void *, key); 1737 __bpf_md_ptr(void *, value); 1738 }; 1739 1740 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 1741 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 1742 bool bpf_iter_prog_supported(struct bpf_prog *prog); 1743 const struct bpf_func_proto * 1744 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 1745 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 1746 int bpf_iter_new_fd(struct bpf_link *link); 1747 bool bpf_link_is_iter(struct bpf_link *link); 1748 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 1749 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 1750 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 1751 struct seq_file *seq); 1752 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 1753 struct bpf_link_info *info); 1754 1755 int map_set_for_each_callback_args(struct bpf_verifier_env *env, 1756 struct bpf_func_state *caller, 1757 struct bpf_func_state *callee); 1758 1759 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 1760 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 1761 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 1762 u64 flags); 1763 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 1764 u64 flags); 1765 1766 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 1767 1768 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 1769 void *key, void *value, u64 map_flags); 1770 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1771 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 1772 void *key, void *value, u64 map_flags); 1773 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1774 1775 int bpf_get_file_flag(int flags); 1776 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 1777 size_t actual_size); 1778 1779 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 1780 * forced to use 'long' read/writes to try to atomically copy long counters. 1781 * Best-effort only. No barriers here, since it _will_ race with concurrent 1782 * updates from BPF programs. Called from bpf syscall and mostly used with 1783 * size 8 or 16 bytes, so ask compiler to inline it. 1784 */ 1785 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 1786 { 1787 const long *lsrc = src; 1788 long *ldst = dst; 1789 1790 size /= sizeof(long); 1791 while (size--) 1792 *ldst++ = *lsrc++; 1793 } 1794 1795 /* verify correctness of eBPF program */ 1796 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr); 1797 1798 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 1799 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 1800 #endif 1801 1802 struct btf *bpf_get_btf_vmlinux(void); 1803 1804 /* Map specifics */ 1805 struct xdp_frame; 1806 struct sk_buff; 1807 struct bpf_dtab_netdev; 1808 struct bpf_cpu_map_entry; 1809 1810 void __dev_flush(void); 1811 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 1812 struct net_device *dev_rx); 1813 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 1814 struct net_device *dev_rx); 1815 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 1816 struct bpf_map *map, bool exclude_ingress); 1817 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 1818 struct bpf_prog *xdp_prog); 1819 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1820 struct bpf_prog *xdp_prog, struct bpf_map *map, 1821 bool exclude_ingress); 1822 1823 void __cpu_map_flush(void); 1824 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf, 1825 struct net_device *dev_rx); 1826 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1827 struct sk_buff *skb); 1828 1829 /* Return map's numa specified by userspace */ 1830 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 1831 { 1832 return (attr->map_flags & BPF_F_NUMA_NODE) ? 1833 attr->numa_node : NUMA_NO_NODE; 1834 } 1835 1836 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 1837 int array_map_alloc_check(union bpf_attr *attr); 1838 1839 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 1840 union bpf_attr __user *uattr); 1841 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 1842 union bpf_attr __user *uattr); 1843 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1844 const union bpf_attr *kattr, 1845 union bpf_attr __user *uattr); 1846 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1847 const union bpf_attr *kattr, 1848 union bpf_attr __user *uattr); 1849 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 1850 const union bpf_attr *kattr, 1851 union bpf_attr __user *uattr); 1852 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 1853 const union bpf_attr *kattr, 1854 union bpf_attr __user *uattr); 1855 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 1856 const struct bpf_prog *prog, 1857 struct bpf_insn_access_aux *info); 1858 1859 static inline bool bpf_tracing_ctx_access(int off, int size, 1860 enum bpf_access_type type) 1861 { 1862 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS) 1863 return false; 1864 if (type != BPF_READ) 1865 return false; 1866 if (off % size != 0) 1867 return false; 1868 return true; 1869 } 1870 1871 static inline bool bpf_tracing_btf_ctx_access(int off, int size, 1872 enum bpf_access_type type, 1873 const struct bpf_prog *prog, 1874 struct bpf_insn_access_aux *info) 1875 { 1876 if (!bpf_tracing_ctx_access(off, size, type)) 1877 return false; 1878 return btf_ctx_access(off, size, type, prog, info); 1879 } 1880 1881 int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf, 1882 const struct btf_type *t, int off, int size, 1883 enum bpf_access_type atype, 1884 u32 *next_btf_id, enum bpf_type_flag *flag); 1885 bool btf_struct_ids_match(struct bpf_verifier_log *log, 1886 const struct btf *btf, u32 id, int off, 1887 const struct btf *need_btf, u32 need_type_id, 1888 bool strict); 1889 1890 int btf_distill_func_proto(struct bpf_verifier_log *log, 1891 struct btf *btf, 1892 const struct btf_type *func_proto, 1893 const char *func_name, 1894 struct btf_func_model *m); 1895 1896 struct bpf_reg_state; 1897 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog, 1898 struct bpf_reg_state *regs); 1899 int btf_check_kfunc_arg_match(struct bpf_verifier_env *env, 1900 const struct btf *btf, u32 func_id, 1901 struct bpf_reg_state *regs); 1902 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, 1903 struct bpf_reg_state *reg); 1904 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 1905 struct btf *btf, const struct btf_type *t); 1906 1907 struct bpf_prog *bpf_prog_by_id(u32 id); 1908 struct bpf_link *bpf_link_by_id(u32 id); 1909 1910 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id); 1911 void bpf_task_storage_free(struct task_struct *task); 1912 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog); 1913 const struct btf_func_model * 1914 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 1915 const struct bpf_insn *insn); 1916 struct bpf_core_ctx { 1917 struct bpf_verifier_log *log; 1918 const struct btf *btf; 1919 }; 1920 1921 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo, 1922 int relo_idx, void *insn); 1923 1924 static inline bool unprivileged_ebpf_enabled(void) 1925 { 1926 return !sysctl_unprivileged_bpf_disabled; 1927 } 1928 1929 #else /* !CONFIG_BPF_SYSCALL */ 1930 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 1931 { 1932 return ERR_PTR(-EOPNOTSUPP); 1933 } 1934 1935 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 1936 enum bpf_prog_type type, 1937 bool attach_drv) 1938 { 1939 return ERR_PTR(-EOPNOTSUPP); 1940 } 1941 1942 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 1943 { 1944 } 1945 1946 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 1947 { 1948 } 1949 1950 static inline void bpf_prog_put(struct bpf_prog *prog) 1951 { 1952 } 1953 1954 static inline void bpf_prog_inc(struct bpf_prog *prog) 1955 { 1956 } 1957 1958 static inline struct bpf_prog *__must_check 1959 bpf_prog_inc_not_zero(struct bpf_prog *prog) 1960 { 1961 return ERR_PTR(-EOPNOTSUPP); 1962 } 1963 1964 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1965 const struct bpf_link_ops *ops, 1966 struct bpf_prog *prog) 1967 { 1968 } 1969 1970 static inline int bpf_link_prime(struct bpf_link *link, 1971 struct bpf_link_primer *primer) 1972 { 1973 return -EOPNOTSUPP; 1974 } 1975 1976 static inline int bpf_link_settle(struct bpf_link_primer *primer) 1977 { 1978 return -EOPNOTSUPP; 1979 } 1980 1981 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 1982 { 1983 } 1984 1985 static inline void bpf_link_inc(struct bpf_link *link) 1986 { 1987 } 1988 1989 static inline void bpf_link_put(struct bpf_link *link) 1990 { 1991 } 1992 1993 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 1994 { 1995 return -EOPNOTSUPP; 1996 } 1997 1998 static inline void __dev_flush(void) 1999 { 2000 } 2001 2002 struct xdp_frame; 2003 struct bpf_dtab_netdev; 2004 struct bpf_cpu_map_entry; 2005 2006 static inline 2007 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 2008 struct net_device *dev_rx) 2009 { 2010 return 0; 2011 } 2012 2013 static inline 2014 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 2015 struct net_device *dev_rx) 2016 { 2017 return 0; 2018 } 2019 2020 static inline 2021 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 2022 struct bpf_map *map, bool exclude_ingress) 2023 { 2024 return 0; 2025 } 2026 2027 struct sk_buff; 2028 2029 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 2030 struct sk_buff *skb, 2031 struct bpf_prog *xdp_prog) 2032 { 2033 return 0; 2034 } 2035 2036 static inline 2037 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 2038 struct bpf_prog *xdp_prog, struct bpf_map *map, 2039 bool exclude_ingress) 2040 { 2041 return 0; 2042 } 2043 2044 static inline void __cpu_map_flush(void) 2045 { 2046 } 2047 2048 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 2049 struct xdp_frame *xdpf, 2050 struct net_device *dev_rx) 2051 { 2052 return 0; 2053 } 2054 2055 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 2056 struct sk_buff *skb) 2057 { 2058 return -EOPNOTSUPP; 2059 } 2060 2061 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 2062 enum bpf_prog_type type) 2063 { 2064 return ERR_PTR(-EOPNOTSUPP); 2065 } 2066 2067 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 2068 const union bpf_attr *kattr, 2069 union bpf_attr __user *uattr) 2070 { 2071 return -ENOTSUPP; 2072 } 2073 2074 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 2075 const union bpf_attr *kattr, 2076 union bpf_attr __user *uattr) 2077 { 2078 return -ENOTSUPP; 2079 } 2080 2081 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 2082 const union bpf_attr *kattr, 2083 union bpf_attr __user *uattr) 2084 { 2085 return -ENOTSUPP; 2086 } 2087 2088 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 2089 const union bpf_attr *kattr, 2090 union bpf_attr __user *uattr) 2091 { 2092 return -ENOTSUPP; 2093 } 2094 2095 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 2096 const union bpf_attr *kattr, 2097 union bpf_attr __user *uattr) 2098 { 2099 return -ENOTSUPP; 2100 } 2101 2102 static inline void bpf_map_put(struct bpf_map *map) 2103 { 2104 } 2105 2106 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 2107 { 2108 return ERR_PTR(-ENOTSUPP); 2109 } 2110 2111 static inline const struct bpf_func_proto * 2112 bpf_base_func_proto(enum bpf_func_id func_id) 2113 { 2114 return NULL; 2115 } 2116 2117 static inline void bpf_task_storage_free(struct task_struct *task) 2118 { 2119 } 2120 2121 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog) 2122 { 2123 return false; 2124 } 2125 2126 static inline const struct btf_func_model * 2127 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 2128 const struct bpf_insn *insn) 2129 { 2130 return NULL; 2131 } 2132 2133 static inline bool unprivileged_ebpf_enabled(void) 2134 { 2135 return false; 2136 } 2137 2138 #endif /* CONFIG_BPF_SYSCALL */ 2139 2140 void __bpf_free_used_btfs(struct bpf_prog_aux *aux, 2141 struct btf_mod_pair *used_btfs, u32 len); 2142 2143 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 2144 enum bpf_prog_type type) 2145 { 2146 return bpf_prog_get_type_dev(ufd, type, false); 2147 } 2148 2149 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 2150 struct bpf_map **used_maps, u32 len); 2151 2152 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 2153 2154 int bpf_prog_offload_compile(struct bpf_prog *prog); 2155 void bpf_prog_offload_destroy(struct bpf_prog *prog); 2156 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 2157 struct bpf_prog *prog); 2158 2159 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 2160 2161 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 2162 int bpf_map_offload_update_elem(struct bpf_map *map, 2163 void *key, void *value, u64 flags); 2164 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 2165 int bpf_map_offload_get_next_key(struct bpf_map *map, 2166 void *key, void *next_key); 2167 2168 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 2169 2170 struct bpf_offload_dev * 2171 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 2172 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 2173 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 2174 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 2175 struct net_device *netdev); 2176 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 2177 struct net_device *netdev); 2178 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 2179 2180 void unpriv_ebpf_notify(int new_state); 2181 2182 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 2183 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 2184 2185 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 2186 { 2187 return aux->offload_requested; 2188 } 2189 2190 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 2191 { 2192 return unlikely(map->ops == &bpf_map_offload_ops); 2193 } 2194 2195 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 2196 void bpf_map_offload_map_free(struct bpf_map *map); 2197 int bpf_prog_test_run_syscall(struct bpf_prog *prog, 2198 const union bpf_attr *kattr, 2199 union bpf_attr __user *uattr); 2200 2201 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 2202 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 2203 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 2204 int sock_map_bpf_prog_query(const union bpf_attr *attr, 2205 union bpf_attr __user *uattr); 2206 2207 void sock_map_unhash(struct sock *sk); 2208 void sock_map_destroy(struct sock *sk); 2209 void sock_map_close(struct sock *sk, long timeout); 2210 #else 2211 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 2212 union bpf_attr *attr) 2213 { 2214 return -EOPNOTSUPP; 2215 } 2216 2217 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 2218 { 2219 return false; 2220 } 2221 2222 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 2223 { 2224 return false; 2225 } 2226 2227 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 2228 { 2229 return ERR_PTR(-EOPNOTSUPP); 2230 } 2231 2232 static inline void bpf_map_offload_map_free(struct bpf_map *map) 2233 { 2234 } 2235 2236 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 2237 const union bpf_attr *kattr, 2238 union bpf_attr __user *uattr) 2239 { 2240 return -ENOTSUPP; 2241 } 2242 2243 #ifdef CONFIG_BPF_SYSCALL 2244 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 2245 struct bpf_prog *prog) 2246 { 2247 return -EINVAL; 2248 } 2249 2250 static inline int sock_map_prog_detach(const union bpf_attr *attr, 2251 enum bpf_prog_type ptype) 2252 { 2253 return -EOPNOTSUPP; 2254 } 2255 2256 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 2257 u64 flags) 2258 { 2259 return -EOPNOTSUPP; 2260 } 2261 2262 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr, 2263 union bpf_attr __user *uattr) 2264 { 2265 return -EINVAL; 2266 } 2267 #endif /* CONFIG_BPF_SYSCALL */ 2268 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 2269 2270 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 2271 void bpf_sk_reuseport_detach(struct sock *sk); 2272 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 2273 void *value); 2274 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 2275 void *value, u64 map_flags); 2276 #else 2277 static inline void bpf_sk_reuseport_detach(struct sock *sk) 2278 { 2279 } 2280 2281 #ifdef CONFIG_BPF_SYSCALL 2282 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 2283 void *key, void *value) 2284 { 2285 return -EOPNOTSUPP; 2286 } 2287 2288 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 2289 void *key, void *value, 2290 u64 map_flags) 2291 { 2292 return -EOPNOTSUPP; 2293 } 2294 #endif /* CONFIG_BPF_SYSCALL */ 2295 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 2296 2297 /* verifier prototypes for helper functions called from eBPF programs */ 2298 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 2299 extern const struct bpf_func_proto bpf_map_update_elem_proto; 2300 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 2301 extern const struct bpf_func_proto bpf_map_push_elem_proto; 2302 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 2303 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 2304 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto; 2305 2306 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 2307 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 2308 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 2309 extern const struct bpf_func_proto bpf_tail_call_proto; 2310 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 2311 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 2312 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 2313 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 2314 extern const struct bpf_func_proto bpf_get_current_comm_proto; 2315 extern const struct bpf_func_proto bpf_get_stackid_proto; 2316 extern const struct bpf_func_proto bpf_get_stack_proto; 2317 extern const struct bpf_func_proto bpf_get_task_stack_proto; 2318 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 2319 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 2320 extern const struct bpf_func_proto bpf_sock_map_update_proto; 2321 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 2322 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 2323 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 2324 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 2325 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 2326 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 2327 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 2328 extern const struct bpf_func_proto bpf_spin_lock_proto; 2329 extern const struct bpf_func_proto bpf_spin_unlock_proto; 2330 extern const struct bpf_func_proto bpf_get_local_storage_proto; 2331 extern const struct bpf_func_proto bpf_strtol_proto; 2332 extern const struct bpf_func_proto bpf_strtoul_proto; 2333 extern const struct bpf_func_proto bpf_tcp_sock_proto; 2334 extern const struct bpf_func_proto bpf_jiffies64_proto; 2335 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 2336 extern const struct bpf_func_proto bpf_event_output_data_proto; 2337 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 2338 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 2339 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 2340 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 2341 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 2342 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto; 2343 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto; 2344 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto; 2345 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 2346 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 2347 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 2348 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 2349 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 2350 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto; 2351 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto; 2352 extern const struct bpf_func_proto bpf_copy_from_user_proto; 2353 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 2354 extern const struct bpf_func_proto bpf_snprintf_proto; 2355 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 2356 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 2357 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 2358 extern const struct bpf_func_proto bpf_sock_from_file_proto; 2359 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 2360 extern const struct bpf_func_proto bpf_task_storage_get_proto; 2361 extern const struct bpf_func_proto bpf_task_storage_delete_proto; 2362 extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 2363 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 2364 extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 2365 extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 2366 extern const struct bpf_func_proto bpf_find_vma_proto; 2367 extern const struct bpf_func_proto bpf_loop_proto; 2368 extern const struct bpf_func_proto bpf_copy_from_user_task_proto; 2369 2370 const struct bpf_func_proto *tracing_prog_func_proto( 2371 enum bpf_func_id func_id, const struct bpf_prog *prog); 2372 2373 /* Shared helpers among cBPF and eBPF. */ 2374 void bpf_user_rnd_init_once(void); 2375 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2376 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2377 2378 #if defined(CONFIG_NET) 2379 bool bpf_sock_common_is_valid_access(int off, int size, 2380 enum bpf_access_type type, 2381 struct bpf_insn_access_aux *info); 2382 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2383 struct bpf_insn_access_aux *info); 2384 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2385 const struct bpf_insn *si, 2386 struct bpf_insn *insn_buf, 2387 struct bpf_prog *prog, 2388 u32 *target_size); 2389 #else 2390 static inline bool bpf_sock_common_is_valid_access(int off, int size, 2391 enum bpf_access_type type, 2392 struct bpf_insn_access_aux *info) 2393 { 2394 return false; 2395 } 2396 static inline bool bpf_sock_is_valid_access(int off, int size, 2397 enum bpf_access_type type, 2398 struct bpf_insn_access_aux *info) 2399 { 2400 return false; 2401 } 2402 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2403 const struct bpf_insn *si, 2404 struct bpf_insn *insn_buf, 2405 struct bpf_prog *prog, 2406 u32 *target_size) 2407 { 2408 return 0; 2409 } 2410 #endif 2411 2412 #ifdef CONFIG_INET 2413 struct sk_reuseport_kern { 2414 struct sk_buff *skb; 2415 struct sock *sk; 2416 struct sock *selected_sk; 2417 struct sock *migrating_sk; 2418 void *data_end; 2419 u32 hash; 2420 u32 reuseport_id; 2421 bool bind_inany; 2422 }; 2423 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2424 struct bpf_insn_access_aux *info); 2425 2426 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2427 const struct bpf_insn *si, 2428 struct bpf_insn *insn_buf, 2429 struct bpf_prog *prog, 2430 u32 *target_size); 2431 2432 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2433 struct bpf_insn_access_aux *info); 2434 2435 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2436 const struct bpf_insn *si, 2437 struct bpf_insn *insn_buf, 2438 struct bpf_prog *prog, 2439 u32 *target_size); 2440 #else 2441 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 2442 enum bpf_access_type type, 2443 struct bpf_insn_access_aux *info) 2444 { 2445 return false; 2446 } 2447 2448 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2449 const struct bpf_insn *si, 2450 struct bpf_insn *insn_buf, 2451 struct bpf_prog *prog, 2452 u32 *target_size) 2453 { 2454 return 0; 2455 } 2456 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 2457 enum bpf_access_type type, 2458 struct bpf_insn_access_aux *info) 2459 { 2460 return false; 2461 } 2462 2463 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2464 const struct bpf_insn *si, 2465 struct bpf_insn *insn_buf, 2466 struct bpf_prog *prog, 2467 u32 *target_size) 2468 { 2469 return 0; 2470 } 2471 #endif /* CONFIG_INET */ 2472 2473 enum bpf_text_poke_type { 2474 BPF_MOD_CALL, 2475 BPF_MOD_JUMP, 2476 }; 2477 2478 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 2479 void *addr1, void *addr2); 2480 2481 void *bpf_arch_text_copy(void *dst, void *src, size_t len); 2482 int bpf_arch_text_invalidate(void *dst, size_t len); 2483 2484 struct btf_id_set; 2485 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 2486 2487 #define MAX_BPRINTF_VARARGS 12 2488 2489 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args, 2490 u32 **bin_buf, u32 num_args); 2491 void bpf_bprintf_cleanup(void); 2492 2493 /* the implementation of the opaque uapi struct bpf_dynptr */ 2494 struct bpf_dynptr_kern { 2495 void *data; 2496 /* Size represents the number of usable bytes of dynptr data. 2497 * If for example the offset is at 4 for a local dynptr whose data is 2498 * of type u64, the number of usable bytes is 4. 2499 * 2500 * The upper 8 bits are reserved. It is as follows: 2501 * Bits 0 - 23 = size 2502 * Bits 24 - 30 = dynptr type 2503 * Bit 31 = whether dynptr is read-only 2504 */ 2505 u32 size; 2506 u32 offset; 2507 } __aligned(8); 2508 2509 enum bpf_dynptr_type { 2510 BPF_DYNPTR_TYPE_INVALID, 2511 /* Points to memory that is local to the bpf program */ 2512 BPF_DYNPTR_TYPE_LOCAL, 2513 /* Underlying data is a ringbuf record */ 2514 BPF_DYNPTR_TYPE_RINGBUF, 2515 }; 2516 2517 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 2518 enum bpf_dynptr_type type, u32 offset, u32 size); 2519 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr); 2520 int bpf_dynptr_check_size(u32 size); 2521 2522 #endif /* _LINUX_BPF_H */ 2523