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 struct ftrace_ops; 51 struct cgroup; 52 53 extern struct idr btf_idr; 54 extern spinlock_t btf_idr_lock; 55 extern struct kobject *btf_kobj; 56 57 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64); 58 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, 59 struct bpf_iter_aux_info *aux); 60 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); 61 typedef unsigned int (*bpf_func_t)(const void *, 62 const struct bpf_insn *); 63 struct bpf_iter_seq_info { 64 const struct seq_operations *seq_ops; 65 bpf_iter_init_seq_priv_t init_seq_private; 66 bpf_iter_fini_seq_priv_t fini_seq_private; 67 u32 seq_priv_size; 68 }; 69 70 /* map is generic key/value storage optionally accessible by eBPF programs */ 71 struct bpf_map_ops { 72 /* funcs callable from userspace (via syscall) */ 73 int (*map_alloc_check)(union bpf_attr *attr); 74 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 75 void (*map_release)(struct bpf_map *map, struct file *map_file); 76 void (*map_free)(struct bpf_map *map); 77 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 78 void (*map_release_uref)(struct bpf_map *map); 79 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 80 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, 81 union bpf_attr __user *uattr); 82 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key, 83 void *value, u64 flags); 84 int (*map_lookup_and_delete_batch)(struct bpf_map *map, 85 const union bpf_attr *attr, 86 union bpf_attr __user *uattr); 87 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr, 88 union bpf_attr __user *uattr); 89 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, 90 union bpf_attr __user *uattr); 91 92 /* funcs callable from userspace and from eBPF programs */ 93 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 94 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 95 int (*map_delete_elem)(struct bpf_map *map, void *key); 96 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 97 int (*map_pop_elem)(struct bpf_map *map, void *value); 98 int (*map_peek_elem)(struct bpf_map *map, void *value); 99 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu); 100 101 /* funcs called by prog_array and perf_event_array map */ 102 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 103 int fd); 104 void (*map_fd_put_ptr)(void *ptr); 105 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 106 u32 (*map_fd_sys_lookup_elem)(void *ptr); 107 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 108 struct seq_file *m); 109 int (*map_check_btf)(const struct bpf_map *map, 110 const struct btf *btf, 111 const struct btf_type *key_type, 112 const struct btf_type *value_type); 113 114 /* Prog poke tracking helpers. */ 115 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux); 116 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux); 117 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, 118 struct bpf_prog *new); 119 120 /* Direct value access helpers. */ 121 int (*map_direct_value_addr)(const struct bpf_map *map, 122 u64 *imm, u32 off); 123 int (*map_direct_value_meta)(const struct bpf_map *map, 124 u64 imm, u32 *off); 125 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma); 126 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, 127 struct poll_table_struct *pts); 128 129 /* Functions called by bpf_local_storage maps */ 130 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, 131 void *owner, u32 size); 132 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, 133 void *owner, u32 size); 134 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner); 135 136 /* Misc helpers.*/ 137 int (*map_redirect)(struct bpf_map *map, u32 ifindex, u64 flags); 138 139 /* map_meta_equal must be implemented for maps that can be 140 * used as an inner map. It is a runtime check to ensure 141 * an inner map can be inserted to an outer map. 142 * 143 * Some properties of the inner map has been used during the 144 * verification time. When inserting an inner map at the runtime, 145 * map_meta_equal has to ensure the inserting map has the same 146 * properties that the verifier has used earlier. 147 */ 148 bool (*map_meta_equal)(const struct bpf_map *meta0, 149 const struct bpf_map *meta1); 150 151 152 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env, 153 struct bpf_func_state *caller, 154 struct bpf_func_state *callee); 155 int (*map_for_each_callback)(struct bpf_map *map, 156 bpf_callback_t callback_fn, 157 void *callback_ctx, u64 flags); 158 159 /* BTF id of struct allocated by map_alloc */ 160 int *map_btf_id; 161 162 /* bpf_iter info used to open a seq_file */ 163 const struct bpf_iter_seq_info *iter_seq_info; 164 }; 165 166 enum { 167 /* Support at most 8 pointers in a BPF map value */ 168 BPF_MAP_VALUE_OFF_MAX = 8, 169 BPF_MAP_OFF_ARR_MAX = BPF_MAP_VALUE_OFF_MAX + 170 1 + /* for bpf_spin_lock */ 171 1, /* for bpf_timer */ 172 }; 173 174 enum bpf_kptr_type { 175 BPF_KPTR_UNREF, 176 BPF_KPTR_REF, 177 }; 178 179 struct bpf_map_value_off_desc { 180 u32 offset; 181 enum bpf_kptr_type type; 182 struct { 183 struct btf *btf; 184 struct module *module; 185 btf_dtor_kfunc_t dtor; 186 u32 btf_id; 187 } kptr; 188 }; 189 190 struct bpf_map_value_off { 191 u32 nr_off; 192 struct bpf_map_value_off_desc off[]; 193 }; 194 195 struct bpf_map_off_arr { 196 u32 cnt; 197 u32 field_off[BPF_MAP_OFF_ARR_MAX]; 198 u8 field_sz[BPF_MAP_OFF_ARR_MAX]; 199 }; 200 201 struct bpf_map { 202 /* The first two cachelines with read-mostly members of which some 203 * are also accessed in fast-path (e.g. ops, max_entries). 204 */ 205 const struct bpf_map_ops *ops ____cacheline_aligned; 206 struct bpf_map *inner_map_meta; 207 #ifdef CONFIG_SECURITY 208 void *security; 209 #endif 210 enum bpf_map_type map_type; 211 u32 key_size; 212 u32 value_size; 213 u32 max_entries; 214 u64 map_extra; /* any per-map-type extra fields */ 215 u32 map_flags; 216 int spin_lock_off; /* >=0 valid offset, <0 error */ 217 struct bpf_map_value_off *kptr_off_tab; 218 int timer_off; /* >=0 valid offset, <0 error */ 219 u32 id; 220 int numa_node; 221 u32 btf_key_type_id; 222 u32 btf_value_type_id; 223 u32 btf_vmlinux_value_type_id; 224 struct btf *btf; 225 #ifdef CONFIG_MEMCG_KMEM 226 struct obj_cgroup *objcg; 227 #endif 228 char name[BPF_OBJ_NAME_LEN]; 229 struct bpf_map_off_arr *off_arr; 230 /* The 3rd and 4th cacheline with misc members to avoid false sharing 231 * particularly with refcounting. 232 */ 233 atomic64_t refcnt ____cacheline_aligned; 234 atomic64_t usercnt; 235 struct work_struct work; 236 struct mutex freeze_mutex; 237 atomic64_t writecnt; 238 /* 'Ownership' of program-containing map is claimed by the first program 239 * that is going to use this map or by the first program which FD is 240 * stored in the map to make sure that all callers and callees have the 241 * same prog type, JITed flag and xdp_has_frags flag. 242 */ 243 struct { 244 spinlock_t lock; 245 enum bpf_prog_type type; 246 bool jited; 247 bool xdp_has_frags; 248 } owner; 249 bool bypass_spec_v1; 250 bool frozen; /* write-once; write-protected by freeze_mutex */ 251 }; 252 253 static inline bool map_value_has_spin_lock(const struct bpf_map *map) 254 { 255 return map->spin_lock_off >= 0; 256 } 257 258 static inline bool map_value_has_timer(const struct bpf_map *map) 259 { 260 return map->timer_off >= 0; 261 } 262 263 static inline bool map_value_has_kptrs(const struct bpf_map *map) 264 { 265 return !IS_ERR_OR_NULL(map->kptr_off_tab); 266 } 267 268 static inline void check_and_init_map_value(struct bpf_map *map, void *dst) 269 { 270 if (unlikely(map_value_has_spin_lock(map))) 271 memset(dst + map->spin_lock_off, 0, sizeof(struct bpf_spin_lock)); 272 if (unlikely(map_value_has_timer(map))) 273 memset(dst + map->timer_off, 0, sizeof(struct bpf_timer)); 274 if (unlikely(map_value_has_kptrs(map))) { 275 struct bpf_map_value_off *tab = map->kptr_off_tab; 276 int i; 277 278 for (i = 0; i < tab->nr_off; i++) 279 *(u64 *)(dst + tab->off[i].offset) = 0; 280 } 281 } 282 283 /* copy everything but bpf_spin_lock and bpf_timer. There could be one of each. */ 284 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 285 { 286 u32 curr_off = 0; 287 int i; 288 289 if (likely(!map->off_arr)) { 290 memcpy(dst, src, map->value_size); 291 return; 292 } 293 294 for (i = 0; i < map->off_arr->cnt; i++) { 295 u32 next_off = map->off_arr->field_off[i]; 296 297 memcpy(dst + curr_off, src + curr_off, next_off - curr_off); 298 curr_off += map->off_arr->field_sz[i]; 299 } 300 memcpy(dst + curr_off, src + curr_off, map->value_size - curr_off); 301 } 302 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 303 bool lock_src); 304 void bpf_timer_cancel_and_free(void *timer); 305 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size); 306 307 struct bpf_offload_dev; 308 struct bpf_offloaded_map; 309 310 struct bpf_map_dev_ops { 311 int (*map_get_next_key)(struct bpf_offloaded_map *map, 312 void *key, void *next_key); 313 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 314 void *key, void *value); 315 int (*map_update_elem)(struct bpf_offloaded_map *map, 316 void *key, void *value, u64 flags); 317 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 318 }; 319 320 struct bpf_offloaded_map { 321 struct bpf_map map; 322 struct net_device *netdev; 323 const struct bpf_map_dev_ops *dev_ops; 324 void *dev_priv; 325 struct list_head offloads; 326 }; 327 328 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 329 { 330 return container_of(map, struct bpf_offloaded_map, map); 331 } 332 333 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 334 { 335 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 336 } 337 338 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 339 { 340 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && 341 map->ops->map_seq_show_elem; 342 } 343 344 int map_check_no_btf(const struct bpf_map *map, 345 const struct btf *btf, 346 const struct btf_type *key_type, 347 const struct btf_type *value_type); 348 349 bool bpf_map_meta_equal(const struct bpf_map *meta0, 350 const struct bpf_map *meta1); 351 352 extern const struct bpf_map_ops bpf_map_offload_ops; 353 354 /* bpf_type_flag contains a set of flags that are applicable to the values of 355 * arg_type, ret_type and reg_type. For example, a pointer value may be null, 356 * or a memory is read-only. We classify types into two categories: base types 357 * and extended types. Extended types are base types combined with a type flag. 358 * 359 * Currently there are no more than 32 base types in arg_type, ret_type and 360 * reg_types. 361 */ 362 #define BPF_BASE_TYPE_BITS 8 363 364 enum bpf_type_flag { 365 /* PTR may be NULL. */ 366 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS), 367 368 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is 369 * compatible with both mutable and immutable memory. 370 */ 371 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS), 372 373 /* MEM was "allocated" from a different helper, and cannot be mixed 374 * with regular non-MEM_ALLOC'ed MEM types. 375 */ 376 MEM_ALLOC = BIT(2 + BPF_BASE_TYPE_BITS), 377 378 /* MEM is in user address space. */ 379 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS), 380 381 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged 382 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In 383 * order to drop this tag, it must be passed into bpf_per_cpu_ptr() 384 * or bpf_this_cpu_ptr(), which will return the pointer corresponding 385 * to the specified cpu. 386 */ 387 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS), 388 389 /* Indicates that the argument will be released. */ 390 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS), 391 392 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark 393 * unreferenced and referenced kptr loaded from map value using a load 394 * instruction, so that they can only be dereferenced but not escape the 395 * BPF program into the kernel (i.e. cannot be passed as arguments to 396 * kfunc or bpf helpers). 397 */ 398 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS), 399 400 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS), 401 402 /* DYNPTR points to memory local to the bpf program. */ 403 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS), 404 405 /* DYNPTR points to a ringbuf record. */ 406 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS), 407 408 /* Size is known at compile time. */ 409 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS), 410 411 __BPF_TYPE_FLAG_MAX, 412 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1, 413 }; 414 415 #define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF) 416 417 /* Max number of base types. */ 418 #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS) 419 420 /* Max number of all types. */ 421 #define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1)) 422 423 /* function argument constraints */ 424 enum bpf_arg_type { 425 ARG_DONTCARE = 0, /* unused argument in helper function */ 426 427 /* the following constraints used to prototype 428 * bpf_map_lookup/update/delete_elem() functions 429 */ 430 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 431 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 432 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 433 434 /* Used to prototype bpf_memcmp() and other functions that access data 435 * on eBPF program stack 436 */ 437 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 438 439 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 440 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 441 442 ARG_PTR_TO_CTX, /* pointer to context */ 443 ARG_ANYTHING, /* any (initialized) argument is ok */ 444 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 445 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 446 ARG_PTR_TO_INT, /* pointer to int */ 447 ARG_PTR_TO_LONG, /* pointer to long */ 448 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 449 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */ 450 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */ 451 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ 452 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ 453 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ 454 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */ 455 ARG_PTR_TO_STACK, /* pointer to stack */ 456 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */ 457 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */ 458 ARG_PTR_TO_KPTR, /* pointer to referenced kptr */ 459 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */ 460 __BPF_ARG_TYPE_MAX, 461 462 /* Extended arg_types. */ 463 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE, 464 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM, 465 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX, 466 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET, 467 ARG_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_ALLOC_MEM, 468 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK, 469 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID, 470 /* pointer to memory does not need to be initialized, helper function must fill 471 * all bytes or clear them in error case. 472 */ 473 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | ARG_PTR_TO_MEM, 474 /* Pointer to valid memory of size known at compile time. */ 475 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM, 476 477 /* This must be the last entry. Its purpose is to ensure the enum is 478 * wide enough to hold the higher bits reserved for bpf_type_flag. 479 */ 480 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT, 481 }; 482 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 483 484 /* type of values returned from helper functions */ 485 enum bpf_return_type { 486 RET_INTEGER, /* function returns integer */ 487 RET_VOID, /* function doesn't return anything */ 488 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 489 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */ 490 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */ 491 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */ 492 RET_PTR_TO_ALLOC_MEM, /* returns a pointer to dynamically allocated memory */ 493 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ 494 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */ 495 __BPF_RET_TYPE_MAX, 496 497 /* Extended ret_types. */ 498 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE, 499 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET, 500 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK, 501 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON, 502 RET_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_ALLOC | RET_PTR_TO_ALLOC_MEM, 503 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_ALLOC_MEM, 504 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID, 505 506 /* This must be the last entry. Its purpose is to ensure the enum is 507 * wide enough to hold the higher bits reserved for bpf_type_flag. 508 */ 509 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT, 510 }; 511 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 512 513 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 514 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 515 * instructions after verifying 516 */ 517 struct bpf_func_proto { 518 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 519 bool gpl_only; 520 bool pkt_access; 521 enum bpf_return_type ret_type; 522 union { 523 struct { 524 enum bpf_arg_type arg1_type; 525 enum bpf_arg_type arg2_type; 526 enum bpf_arg_type arg3_type; 527 enum bpf_arg_type arg4_type; 528 enum bpf_arg_type arg5_type; 529 }; 530 enum bpf_arg_type arg_type[5]; 531 }; 532 union { 533 struct { 534 u32 *arg1_btf_id; 535 u32 *arg2_btf_id; 536 u32 *arg3_btf_id; 537 u32 *arg4_btf_id; 538 u32 *arg5_btf_id; 539 }; 540 u32 *arg_btf_id[5]; 541 struct { 542 size_t arg1_size; 543 size_t arg2_size; 544 size_t arg3_size; 545 size_t arg4_size; 546 size_t arg5_size; 547 }; 548 size_t arg_size[5]; 549 }; 550 int *ret_btf_id; /* return value btf_id */ 551 bool (*allowed)(const struct bpf_prog *prog); 552 }; 553 554 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 555 * the first argument to eBPF programs. 556 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 557 */ 558 struct bpf_context; 559 560 enum bpf_access_type { 561 BPF_READ = 1, 562 BPF_WRITE = 2 563 }; 564 565 /* types of values stored in eBPF registers */ 566 /* Pointer types represent: 567 * pointer 568 * pointer + imm 569 * pointer + (u16) var 570 * pointer + (u16) var + imm 571 * if (range > 0) then [ptr, ptr + range - off) is safe to access 572 * if (id > 0) means that some 'var' was added 573 * if (off > 0) means that 'imm' was added 574 */ 575 enum bpf_reg_type { 576 NOT_INIT = 0, /* nothing was written into register */ 577 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 578 PTR_TO_CTX, /* reg points to bpf_context */ 579 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 580 PTR_TO_MAP_VALUE, /* reg points to map element value */ 581 PTR_TO_MAP_KEY, /* reg points to a map element key */ 582 PTR_TO_STACK, /* reg == frame_pointer + offset */ 583 PTR_TO_PACKET_META, /* skb->data - meta_len */ 584 PTR_TO_PACKET, /* reg points to skb->data */ 585 PTR_TO_PACKET_END, /* skb->data + headlen */ 586 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 587 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 588 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 589 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 590 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 591 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 592 /* PTR_TO_BTF_ID points to a kernel struct that does not need 593 * to be null checked by the BPF program. This does not imply the 594 * pointer is _not_ null and in practice this can easily be a null 595 * pointer when reading pointer chains. The assumption is program 596 * context will handle null pointer dereference typically via fault 597 * handling. The verifier must keep this in mind and can make no 598 * assumptions about null or non-null when doing branch analysis. 599 * Further, when passed into helpers the helpers can not, without 600 * additional context, assume the value is non-null. 601 */ 602 PTR_TO_BTF_ID, 603 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not 604 * been checked for null. Used primarily to inform the verifier 605 * an explicit null check is required for this struct. 606 */ 607 PTR_TO_MEM, /* reg points to valid memory region */ 608 PTR_TO_BUF, /* reg points to a read/write buffer */ 609 PTR_TO_FUNC, /* reg points to a bpf program function */ 610 __BPF_REG_TYPE_MAX, 611 612 /* Extended reg_types. */ 613 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE, 614 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET, 615 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON, 616 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK, 617 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID, 618 619 /* This must be the last entry. Its purpose is to ensure the enum is 620 * wide enough to hold the higher bits reserved for bpf_type_flag. 621 */ 622 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT, 623 }; 624 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 625 626 /* The information passed from prog-specific *_is_valid_access 627 * back to the verifier. 628 */ 629 struct bpf_insn_access_aux { 630 enum bpf_reg_type reg_type; 631 union { 632 int ctx_field_size; 633 struct { 634 struct btf *btf; 635 u32 btf_id; 636 }; 637 }; 638 struct bpf_verifier_log *log; /* for verbose logs */ 639 }; 640 641 static inline void 642 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 643 { 644 aux->ctx_field_size = size; 645 } 646 647 static inline bool bpf_pseudo_func(const struct bpf_insn *insn) 648 { 649 return insn->code == (BPF_LD | BPF_IMM | BPF_DW) && 650 insn->src_reg == BPF_PSEUDO_FUNC; 651 } 652 653 struct bpf_prog_ops { 654 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 655 union bpf_attr __user *uattr); 656 }; 657 658 struct bpf_verifier_ops { 659 /* return eBPF function prototype for verification */ 660 const struct bpf_func_proto * 661 (*get_func_proto)(enum bpf_func_id func_id, 662 const struct bpf_prog *prog); 663 664 /* return true if 'size' wide access at offset 'off' within bpf_context 665 * with 'type' (read or write) is allowed 666 */ 667 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 668 const struct bpf_prog *prog, 669 struct bpf_insn_access_aux *info); 670 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 671 const struct bpf_prog *prog); 672 int (*gen_ld_abs)(const struct bpf_insn *orig, 673 struct bpf_insn *insn_buf); 674 u32 (*convert_ctx_access)(enum bpf_access_type type, 675 const struct bpf_insn *src, 676 struct bpf_insn *dst, 677 struct bpf_prog *prog, u32 *target_size); 678 int (*btf_struct_access)(struct bpf_verifier_log *log, 679 const struct btf *btf, 680 const struct btf_type *t, int off, int size, 681 enum bpf_access_type atype, 682 u32 *next_btf_id, enum bpf_type_flag *flag); 683 }; 684 685 struct bpf_prog_offload_ops { 686 /* verifier basic callbacks */ 687 int (*insn_hook)(struct bpf_verifier_env *env, 688 int insn_idx, int prev_insn_idx); 689 int (*finalize)(struct bpf_verifier_env *env); 690 /* verifier optimization callbacks (called after .finalize) */ 691 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 692 struct bpf_insn *insn); 693 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 694 /* program management callbacks */ 695 int (*prepare)(struct bpf_prog *prog); 696 int (*translate)(struct bpf_prog *prog); 697 void (*destroy)(struct bpf_prog *prog); 698 }; 699 700 struct bpf_prog_offload { 701 struct bpf_prog *prog; 702 struct net_device *netdev; 703 struct bpf_offload_dev *offdev; 704 void *dev_priv; 705 struct list_head offloads; 706 bool dev_state; 707 bool opt_failed; 708 void *jited_image; 709 u32 jited_len; 710 }; 711 712 enum bpf_cgroup_storage_type { 713 BPF_CGROUP_STORAGE_SHARED, 714 BPF_CGROUP_STORAGE_PERCPU, 715 __BPF_CGROUP_STORAGE_MAX 716 }; 717 718 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 719 720 /* The longest tracepoint has 12 args. 721 * See include/trace/bpf_probe.h 722 */ 723 #define MAX_BPF_FUNC_ARGS 12 724 725 /* The maximum number of arguments passed through registers 726 * a single function may have. 727 */ 728 #define MAX_BPF_FUNC_REG_ARGS 5 729 730 struct btf_func_model { 731 u8 ret_size; 732 u8 nr_args; 733 u8 arg_size[MAX_BPF_FUNC_ARGS]; 734 }; 735 736 /* Restore arguments before returning from trampoline to let original function 737 * continue executing. This flag is used for fentry progs when there are no 738 * fexit progs. 739 */ 740 #define BPF_TRAMP_F_RESTORE_REGS BIT(0) 741 /* Call original function after fentry progs, but before fexit progs. 742 * Makes sense for fentry/fexit, normal calls and indirect calls. 743 */ 744 #define BPF_TRAMP_F_CALL_ORIG BIT(1) 745 /* Skip current frame and return to parent. Makes sense for fentry/fexit 746 * programs only. Should not be used with normal calls and indirect calls. 747 */ 748 #define BPF_TRAMP_F_SKIP_FRAME BIT(2) 749 /* Store IP address of the caller on the trampoline stack, 750 * so it's available for trampoline's programs. 751 */ 752 #define BPF_TRAMP_F_IP_ARG BIT(3) 753 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */ 754 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4) 755 756 /* Get original function from stack instead of from provided direct address. 757 * Makes sense for trampolines with fexit or fmod_ret programs. 758 */ 759 #define BPF_TRAMP_F_ORIG_STACK BIT(5) 760 761 /* This trampoline is on a function with another ftrace_ops with IPMODIFY, 762 * e.g., a live patch. This flag is set and cleared by ftrace call backs, 763 */ 764 #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6) 765 766 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 767 * bytes on x86. 768 */ 769 #define BPF_MAX_TRAMP_LINKS 38 770 771 struct bpf_tramp_links { 772 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS]; 773 int nr_links; 774 }; 775 776 struct bpf_tramp_run_ctx; 777 778 /* Different use cases for BPF trampoline: 779 * 1. replace nop at the function entry (kprobe equivalent) 780 * flags = BPF_TRAMP_F_RESTORE_REGS 781 * fentry = a set of programs to run before returning from trampoline 782 * 783 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 784 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 785 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 786 * fentry = a set of program to run before calling original function 787 * fexit = a set of program to run after original function 788 * 789 * 3. replace direct call instruction anywhere in the function body 790 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 791 * With flags = 0 792 * fentry = a set of programs to run before returning from trampoline 793 * With flags = BPF_TRAMP_F_CALL_ORIG 794 * orig_call = original callback addr or direct function addr 795 * fentry = a set of program to run before calling original function 796 * fexit = a set of program to run after original function 797 */ 798 struct bpf_tramp_image; 799 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end, 800 const struct btf_func_model *m, u32 flags, 801 struct bpf_tramp_links *tlinks, 802 void *orig_call); 803 /* these two functions are called from generated trampoline */ 804 u64 notrace __bpf_prog_enter(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx); 805 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start, struct bpf_tramp_run_ctx *run_ctx); 806 u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx); 807 void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start, 808 struct bpf_tramp_run_ctx *run_ctx); 809 u64 notrace __bpf_prog_enter_lsm_cgroup(struct bpf_prog *prog, 810 struct bpf_tramp_run_ctx *run_ctx); 811 void notrace __bpf_prog_exit_lsm_cgroup(struct bpf_prog *prog, u64 start, 812 struct bpf_tramp_run_ctx *run_ctx); 813 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr); 814 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr); 815 816 struct bpf_ksym { 817 unsigned long start; 818 unsigned long end; 819 char name[KSYM_NAME_LEN]; 820 struct list_head lnode; 821 struct latch_tree_node tnode; 822 bool prog; 823 }; 824 825 enum bpf_tramp_prog_type { 826 BPF_TRAMP_FENTRY, 827 BPF_TRAMP_FEXIT, 828 BPF_TRAMP_MODIFY_RETURN, 829 BPF_TRAMP_MAX, 830 BPF_TRAMP_REPLACE, /* more than MAX */ 831 }; 832 833 struct bpf_tramp_image { 834 void *image; 835 struct bpf_ksym ksym; 836 struct percpu_ref pcref; 837 void *ip_after_call; 838 void *ip_epilogue; 839 union { 840 struct rcu_head rcu; 841 struct work_struct work; 842 }; 843 }; 844 845 struct bpf_trampoline { 846 /* hlist for trampoline_table */ 847 struct hlist_node hlist; 848 struct ftrace_ops *fops; 849 /* serializes access to fields of this trampoline */ 850 struct mutex mutex; 851 refcount_t refcnt; 852 u32 flags; 853 u64 key; 854 struct { 855 struct btf_func_model model; 856 void *addr; 857 bool ftrace_managed; 858 } func; 859 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 860 * program by replacing one of its functions. func.addr is the address 861 * of the function it replaced. 862 */ 863 struct bpf_prog *extension_prog; 864 /* list of BPF programs using this trampoline */ 865 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 866 /* Number of attached programs. A counter per kind. */ 867 int progs_cnt[BPF_TRAMP_MAX]; 868 /* Executable image of trampoline */ 869 struct bpf_tramp_image *cur_image; 870 u64 selector; 871 struct module *mod; 872 }; 873 874 struct bpf_attach_target_info { 875 struct btf_func_model fmodel; 876 long tgt_addr; 877 const char *tgt_name; 878 const struct btf_type *tgt_type; 879 }; 880 881 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 882 883 struct bpf_dispatcher_prog { 884 struct bpf_prog *prog; 885 refcount_t users; 886 }; 887 888 struct bpf_dispatcher { 889 /* dispatcher mutex */ 890 struct mutex mutex; 891 void *func; 892 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 893 int num_progs; 894 void *image; 895 u32 image_off; 896 struct bpf_ksym ksym; 897 }; 898 899 static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func( 900 const void *ctx, 901 const struct bpf_insn *insnsi, 902 bpf_func_t bpf_func) 903 { 904 return bpf_func(ctx, insnsi); 905 } 906 907 #ifdef CONFIG_BPF_JIT 908 int bpf_trampoline_link_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr); 909 int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr); 910 struct bpf_trampoline *bpf_trampoline_get(u64 key, 911 struct bpf_attach_target_info *tgt_info); 912 void bpf_trampoline_put(struct bpf_trampoline *tr); 913 int arch_prepare_bpf_dispatcher(void *image, s64 *funcs, int num_funcs); 914 #define BPF_DISPATCHER_INIT(_name) { \ 915 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 916 .func = &_name##_func, \ 917 .progs = {}, \ 918 .num_progs = 0, \ 919 .image = NULL, \ 920 .image_off = 0, \ 921 .ksym = { \ 922 .name = #_name, \ 923 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 924 }, \ 925 } 926 927 #define DEFINE_BPF_DISPATCHER(name) \ 928 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \ 929 const void *ctx, \ 930 const struct bpf_insn *insnsi, \ 931 bpf_func_t bpf_func) \ 932 { \ 933 return bpf_func(ctx, insnsi); \ 934 } \ 935 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 936 struct bpf_dispatcher bpf_dispatcher_##name = \ 937 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 938 #define DECLARE_BPF_DISPATCHER(name) \ 939 unsigned int bpf_dispatcher_##name##_func( \ 940 const void *ctx, \ 941 const struct bpf_insn *insnsi, \ 942 bpf_func_t bpf_func); \ 943 extern struct bpf_dispatcher bpf_dispatcher_##name; 944 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 945 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 946 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 947 struct bpf_prog *to); 948 /* Called only from JIT-enabled code, so there's no need for stubs. */ 949 void *bpf_jit_alloc_exec_page(void); 950 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym); 951 void bpf_image_ksym_del(struct bpf_ksym *ksym); 952 void bpf_ksym_add(struct bpf_ksym *ksym); 953 void bpf_ksym_del(struct bpf_ksym *ksym); 954 int bpf_jit_charge_modmem(u32 size); 955 void bpf_jit_uncharge_modmem(u32 size); 956 bool bpf_prog_has_trampoline(const struct bpf_prog *prog); 957 #else 958 static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link, 959 struct bpf_trampoline *tr) 960 { 961 return -ENOTSUPP; 962 } 963 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, 964 struct bpf_trampoline *tr) 965 { 966 return -ENOTSUPP; 967 } 968 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 969 struct bpf_attach_target_info *tgt_info) 970 { 971 return ERR_PTR(-EOPNOTSUPP); 972 } 973 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 974 #define DEFINE_BPF_DISPATCHER(name) 975 #define DECLARE_BPF_DISPATCHER(name) 976 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 977 #define BPF_DISPATCHER_PTR(name) NULL 978 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 979 struct bpf_prog *from, 980 struct bpf_prog *to) {} 981 static inline bool is_bpf_image_address(unsigned long address) 982 { 983 return false; 984 } 985 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog) 986 { 987 return false; 988 } 989 #endif 990 991 struct bpf_func_info_aux { 992 u16 linkage; 993 bool unreliable; 994 }; 995 996 enum bpf_jit_poke_reason { 997 BPF_POKE_REASON_TAIL_CALL, 998 }; 999 1000 /* Descriptor of pokes pointing /into/ the JITed image. */ 1001 struct bpf_jit_poke_descriptor { 1002 void *tailcall_target; 1003 void *tailcall_bypass; 1004 void *bypass_addr; 1005 void *aux; 1006 union { 1007 struct { 1008 struct bpf_map *map; 1009 u32 key; 1010 } tail_call; 1011 }; 1012 bool tailcall_target_stable; 1013 u8 adj_off; 1014 u16 reason; 1015 u32 insn_idx; 1016 }; 1017 1018 /* reg_type info for ctx arguments */ 1019 struct bpf_ctx_arg_aux { 1020 u32 offset; 1021 enum bpf_reg_type reg_type; 1022 u32 btf_id; 1023 }; 1024 1025 struct btf_mod_pair { 1026 struct btf *btf; 1027 struct module *module; 1028 }; 1029 1030 struct bpf_kfunc_desc_tab; 1031 1032 struct bpf_prog_aux { 1033 atomic64_t refcnt; 1034 u32 used_map_cnt; 1035 u32 used_btf_cnt; 1036 u32 max_ctx_offset; 1037 u32 max_pkt_offset; 1038 u32 max_tp_access; 1039 u32 stack_depth; 1040 u32 id; 1041 u32 func_cnt; /* used by non-func prog as the number of func progs */ 1042 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 1043 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1044 u32 ctx_arg_info_size; 1045 u32 max_rdonly_access; 1046 u32 max_rdwr_access; 1047 struct btf *attach_btf; 1048 const struct bpf_ctx_arg_aux *ctx_arg_info; 1049 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 1050 struct bpf_prog *dst_prog; 1051 struct bpf_trampoline *dst_trampoline; 1052 enum bpf_prog_type saved_dst_prog_type; 1053 enum bpf_attach_type saved_dst_attach_type; 1054 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 1055 bool offload_requested; 1056 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 1057 bool func_proto_unreliable; 1058 bool sleepable; 1059 bool tail_call_reachable; 1060 bool xdp_has_frags; 1061 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 1062 const struct btf_type *attach_func_proto; 1063 /* function name for valid attach_btf_id */ 1064 const char *attach_func_name; 1065 struct bpf_prog **func; 1066 void *jit_data; /* JIT specific data. arch dependent */ 1067 struct bpf_jit_poke_descriptor *poke_tab; 1068 struct bpf_kfunc_desc_tab *kfunc_tab; 1069 struct bpf_kfunc_btf_tab *kfunc_btf_tab; 1070 u32 size_poke_tab; 1071 struct bpf_ksym ksym; 1072 const struct bpf_prog_ops *ops; 1073 struct bpf_map **used_maps; 1074 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 1075 struct btf_mod_pair *used_btfs; 1076 struct bpf_prog *prog; 1077 struct user_struct *user; 1078 u64 load_time; /* ns since boottime */ 1079 u32 verified_insns; 1080 int cgroup_atype; /* enum cgroup_bpf_attach_type */ 1081 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1082 char name[BPF_OBJ_NAME_LEN]; 1083 #ifdef CONFIG_SECURITY 1084 void *security; 1085 #endif 1086 struct bpf_prog_offload *offload; 1087 struct btf *btf; 1088 struct bpf_func_info *func_info; 1089 struct bpf_func_info_aux *func_info_aux; 1090 /* bpf_line_info loaded from userspace. linfo->insn_off 1091 * has the xlated insn offset. 1092 * Both the main and sub prog share the same linfo. 1093 * The subprog can access its first linfo by 1094 * using the linfo_idx. 1095 */ 1096 struct bpf_line_info *linfo; 1097 /* jited_linfo is the jited addr of the linfo. It has a 1098 * one to one mapping to linfo: 1099 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 1100 * Both the main and sub prog share the same jited_linfo. 1101 * The subprog can access its first jited_linfo by 1102 * using the linfo_idx. 1103 */ 1104 void **jited_linfo; 1105 u32 func_info_cnt; 1106 u32 nr_linfo; 1107 /* subprog can use linfo_idx to access its first linfo and 1108 * jited_linfo. 1109 * main prog always has linfo_idx == 0 1110 */ 1111 u32 linfo_idx; 1112 u32 num_exentries; 1113 struct exception_table_entry *extable; 1114 union { 1115 struct work_struct work; 1116 struct rcu_head rcu; 1117 }; 1118 }; 1119 1120 struct bpf_prog { 1121 u16 pages; /* Number of allocated pages */ 1122 u16 jited:1, /* Is our filter JIT'ed? */ 1123 jit_requested:1,/* archs need to JIT the prog */ 1124 gpl_compatible:1, /* Is filter GPL compatible? */ 1125 cb_access:1, /* Is control block accessed? */ 1126 dst_needed:1, /* Do we need dst entry? */ 1127 blinding_requested:1, /* needs constant blinding */ 1128 blinded:1, /* Was blinded */ 1129 is_func:1, /* program is a bpf function */ 1130 kprobe_override:1, /* Do we override a kprobe? */ 1131 has_callchain_buf:1, /* callchain buffer allocated? */ 1132 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */ 1133 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */ 1134 call_get_func_ip:1, /* Do we call get_func_ip() */ 1135 tstamp_type_access:1; /* Accessed __sk_buff->tstamp_type */ 1136 enum bpf_prog_type type; /* Type of BPF program */ 1137 enum bpf_attach_type expected_attach_type; /* For some prog types */ 1138 u32 len; /* Number of filter blocks */ 1139 u32 jited_len; /* Size of jited insns in bytes */ 1140 u8 tag[BPF_TAG_SIZE]; 1141 struct bpf_prog_stats __percpu *stats; 1142 int __percpu *active; 1143 unsigned int (*bpf_func)(const void *ctx, 1144 const struct bpf_insn *insn); 1145 struct bpf_prog_aux *aux; /* Auxiliary fields */ 1146 struct sock_fprog_kern *orig_prog; /* Original BPF program */ 1147 /* Instructions for interpreter */ 1148 union { 1149 DECLARE_FLEX_ARRAY(struct sock_filter, insns); 1150 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi); 1151 }; 1152 }; 1153 1154 struct bpf_array_aux { 1155 /* Programs with direct jumps into programs part of this array. */ 1156 struct list_head poke_progs; 1157 struct bpf_map *map; 1158 struct mutex poke_mutex; 1159 struct work_struct work; 1160 }; 1161 1162 struct bpf_link { 1163 atomic64_t refcnt; 1164 u32 id; 1165 enum bpf_link_type type; 1166 const struct bpf_link_ops *ops; 1167 struct bpf_prog *prog; 1168 struct work_struct work; 1169 }; 1170 1171 struct bpf_link_ops { 1172 void (*release)(struct bpf_link *link); 1173 void (*dealloc)(struct bpf_link *link); 1174 int (*detach)(struct bpf_link *link); 1175 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 1176 struct bpf_prog *old_prog); 1177 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 1178 int (*fill_link_info)(const struct bpf_link *link, 1179 struct bpf_link_info *info); 1180 }; 1181 1182 struct bpf_tramp_link { 1183 struct bpf_link link; 1184 struct hlist_node tramp_hlist; 1185 u64 cookie; 1186 }; 1187 1188 struct bpf_shim_tramp_link { 1189 struct bpf_tramp_link link; 1190 struct bpf_trampoline *trampoline; 1191 }; 1192 1193 struct bpf_tracing_link { 1194 struct bpf_tramp_link link; 1195 enum bpf_attach_type attach_type; 1196 struct bpf_trampoline *trampoline; 1197 struct bpf_prog *tgt_prog; 1198 }; 1199 1200 struct bpf_link_primer { 1201 struct bpf_link *link; 1202 struct file *file; 1203 int fd; 1204 u32 id; 1205 }; 1206 1207 struct bpf_struct_ops_value; 1208 struct btf_member; 1209 1210 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 1211 struct bpf_struct_ops { 1212 const struct bpf_verifier_ops *verifier_ops; 1213 int (*init)(struct btf *btf); 1214 int (*check_member)(const struct btf_type *t, 1215 const struct btf_member *member); 1216 int (*init_member)(const struct btf_type *t, 1217 const struct btf_member *member, 1218 void *kdata, const void *udata); 1219 int (*reg)(void *kdata); 1220 void (*unreg)(void *kdata); 1221 const struct btf_type *type; 1222 const struct btf_type *value_type; 1223 const char *name; 1224 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 1225 u32 type_id; 1226 u32 value_id; 1227 }; 1228 1229 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 1230 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 1231 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id); 1232 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log); 1233 bool bpf_struct_ops_get(const void *kdata); 1234 void bpf_struct_ops_put(const void *kdata); 1235 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 1236 void *value); 1237 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks, 1238 struct bpf_tramp_link *link, 1239 const struct btf_func_model *model, 1240 void *image, void *image_end); 1241 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1242 { 1243 if (owner == BPF_MODULE_OWNER) 1244 return bpf_struct_ops_get(data); 1245 else 1246 return try_module_get(owner); 1247 } 1248 static inline void bpf_module_put(const void *data, struct module *owner) 1249 { 1250 if (owner == BPF_MODULE_OWNER) 1251 bpf_struct_ops_put(data); 1252 else 1253 module_put(owner); 1254 } 1255 1256 #ifdef CONFIG_NET 1257 /* Define it here to avoid the use of forward declaration */ 1258 struct bpf_dummy_ops_state { 1259 int val; 1260 }; 1261 1262 struct bpf_dummy_ops { 1263 int (*test_1)(struct bpf_dummy_ops_state *cb); 1264 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2, 1265 char a3, unsigned long a4); 1266 }; 1267 1268 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr, 1269 union bpf_attr __user *uattr); 1270 #endif 1271 #else 1272 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id) 1273 { 1274 return NULL; 1275 } 1276 static inline void bpf_struct_ops_init(struct btf *btf, 1277 struct bpf_verifier_log *log) 1278 { 1279 } 1280 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1281 { 1282 return try_module_get(owner); 1283 } 1284 static inline void bpf_module_put(const void *data, struct module *owner) 1285 { 1286 module_put(owner); 1287 } 1288 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 1289 void *key, 1290 void *value) 1291 { 1292 return -EINVAL; 1293 } 1294 #endif 1295 1296 #if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM) 1297 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 1298 int cgroup_atype); 1299 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog); 1300 #else 1301 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 1302 int cgroup_atype) 1303 { 1304 return -EOPNOTSUPP; 1305 } 1306 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog) 1307 { 1308 } 1309 #endif 1310 1311 struct bpf_array { 1312 struct bpf_map map; 1313 u32 elem_size; 1314 u32 index_mask; 1315 struct bpf_array_aux *aux; 1316 union { 1317 char value[0] __aligned(8); 1318 void *ptrs[0] __aligned(8); 1319 void __percpu *pptrs[0] __aligned(8); 1320 }; 1321 }; 1322 1323 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 1324 #define MAX_TAIL_CALL_CNT 33 1325 1326 /* Maximum number of loops for bpf_loop */ 1327 #define BPF_MAX_LOOPS BIT(23) 1328 1329 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 1330 BPF_F_RDONLY_PROG | \ 1331 BPF_F_WRONLY | \ 1332 BPF_F_WRONLY_PROG) 1333 1334 #define BPF_MAP_CAN_READ BIT(0) 1335 #define BPF_MAP_CAN_WRITE BIT(1) 1336 1337 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 1338 { 1339 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1340 1341 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 1342 * not possible. 1343 */ 1344 if (access_flags & BPF_F_RDONLY_PROG) 1345 return BPF_MAP_CAN_READ; 1346 else if (access_flags & BPF_F_WRONLY_PROG) 1347 return BPF_MAP_CAN_WRITE; 1348 else 1349 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 1350 } 1351 1352 static inline bool bpf_map_flags_access_ok(u32 access_flags) 1353 { 1354 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 1355 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1356 } 1357 1358 struct bpf_event_entry { 1359 struct perf_event *event; 1360 struct file *perf_file; 1361 struct file *map_file; 1362 struct rcu_head rcu; 1363 }; 1364 1365 static inline bool map_type_contains_progs(struct bpf_map *map) 1366 { 1367 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY || 1368 map->map_type == BPF_MAP_TYPE_DEVMAP || 1369 map->map_type == BPF_MAP_TYPE_CPUMAP; 1370 } 1371 1372 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp); 1373 int bpf_prog_calc_tag(struct bpf_prog *fp); 1374 1375 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 1376 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void); 1377 1378 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 1379 unsigned long off, unsigned long len); 1380 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 1381 const struct bpf_insn *src, 1382 struct bpf_insn *dst, 1383 struct bpf_prog *prog, 1384 u32 *target_size); 1385 1386 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 1387 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 1388 1389 /* an array of programs to be executed under rcu_lock. 1390 * 1391 * Typical usage: 1392 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run); 1393 * 1394 * the structure returned by bpf_prog_array_alloc() should be populated 1395 * with program pointers and the last pointer must be NULL. 1396 * The user has to keep refcnt on the program and make sure the program 1397 * is removed from the array before bpf_prog_put(). 1398 * The 'struct bpf_prog_array *' should only be replaced with xchg() 1399 * since other cpus are walking the array of pointers in parallel. 1400 */ 1401 struct bpf_prog_array_item { 1402 struct bpf_prog *prog; 1403 union { 1404 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1405 u64 bpf_cookie; 1406 }; 1407 }; 1408 1409 struct bpf_prog_array { 1410 struct rcu_head rcu; 1411 struct bpf_prog_array_item items[]; 1412 }; 1413 1414 struct bpf_empty_prog_array { 1415 struct bpf_prog_array hdr; 1416 struct bpf_prog *null_prog; 1417 }; 1418 1419 /* to avoid allocating empty bpf_prog_array for cgroups that 1420 * don't have bpf program attached use one global 'bpf_empty_prog_array' 1421 * It will not be modified the caller of bpf_prog_array_alloc() 1422 * (since caller requested prog_cnt == 0) 1423 * that pointer should be 'freed' by bpf_prog_array_free() 1424 */ 1425 extern struct bpf_empty_prog_array bpf_empty_prog_array; 1426 1427 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 1428 void bpf_prog_array_free(struct bpf_prog_array *progs); 1429 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */ 1430 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs); 1431 int bpf_prog_array_length(struct bpf_prog_array *progs); 1432 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 1433 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 1434 __u32 __user *prog_ids, u32 cnt); 1435 1436 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 1437 struct bpf_prog *old_prog); 1438 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 1439 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 1440 struct bpf_prog *prog); 1441 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 1442 u32 *prog_ids, u32 request_cnt, 1443 u32 *prog_cnt); 1444 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 1445 struct bpf_prog *exclude_prog, 1446 struct bpf_prog *include_prog, 1447 u64 bpf_cookie, 1448 struct bpf_prog_array **new_array); 1449 1450 struct bpf_run_ctx {}; 1451 1452 struct bpf_cg_run_ctx { 1453 struct bpf_run_ctx run_ctx; 1454 const struct bpf_prog_array_item *prog_item; 1455 int retval; 1456 }; 1457 1458 struct bpf_trace_run_ctx { 1459 struct bpf_run_ctx run_ctx; 1460 u64 bpf_cookie; 1461 }; 1462 1463 struct bpf_tramp_run_ctx { 1464 struct bpf_run_ctx run_ctx; 1465 u64 bpf_cookie; 1466 struct bpf_run_ctx *saved_run_ctx; 1467 }; 1468 1469 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 1470 { 1471 struct bpf_run_ctx *old_ctx = NULL; 1472 1473 #ifdef CONFIG_BPF_SYSCALL 1474 old_ctx = current->bpf_ctx; 1475 current->bpf_ctx = new_ctx; 1476 #endif 1477 return old_ctx; 1478 } 1479 1480 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 1481 { 1482 #ifdef CONFIG_BPF_SYSCALL 1483 current->bpf_ctx = old_ctx; 1484 #endif 1485 } 1486 1487 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 1488 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 1489 /* BPF program asks to set CN on the packet. */ 1490 #define BPF_RET_SET_CN (1 << 0) 1491 1492 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 1493 1494 static __always_inline u32 1495 bpf_prog_run_array(const struct bpf_prog_array *array, 1496 const void *ctx, bpf_prog_run_fn run_prog) 1497 { 1498 const struct bpf_prog_array_item *item; 1499 const struct bpf_prog *prog; 1500 struct bpf_run_ctx *old_run_ctx; 1501 struct bpf_trace_run_ctx run_ctx; 1502 u32 ret = 1; 1503 1504 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held"); 1505 1506 if (unlikely(!array)) 1507 return ret; 1508 1509 migrate_disable(); 1510 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1511 item = &array->items[0]; 1512 while ((prog = READ_ONCE(item->prog))) { 1513 run_ctx.bpf_cookie = item->bpf_cookie; 1514 ret &= run_prog(prog, ctx); 1515 item++; 1516 } 1517 bpf_reset_run_ctx(old_run_ctx); 1518 migrate_enable(); 1519 return ret; 1520 } 1521 1522 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs: 1523 * 1524 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array 1525 * overall. As a result, we must use the bpf_prog_array_free_sleepable 1526 * in order to use the tasks_trace rcu grace period. 1527 * 1528 * When a non-sleepable program is inside the array, we take the rcu read 1529 * section and disable preemption for that program alone, so it can access 1530 * rcu-protected dynamically sized maps. 1531 */ 1532 static __always_inline u32 1533 bpf_prog_run_array_sleepable(const struct bpf_prog_array __rcu *array_rcu, 1534 const void *ctx, bpf_prog_run_fn run_prog) 1535 { 1536 const struct bpf_prog_array_item *item; 1537 const struct bpf_prog *prog; 1538 const struct bpf_prog_array *array; 1539 struct bpf_run_ctx *old_run_ctx; 1540 struct bpf_trace_run_ctx run_ctx; 1541 u32 ret = 1; 1542 1543 might_fault(); 1544 1545 rcu_read_lock_trace(); 1546 migrate_disable(); 1547 1548 array = rcu_dereference_check(array_rcu, rcu_read_lock_trace_held()); 1549 if (unlikely(!array)) 1550 goto out; 1551 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1552 item = &array->items[0]; 1553 while ((prog = READ_ONCE(item->prog))) { 1554 if (!prog->aux->sleepable) 1555 rcu_read_lock(); 1556 1557 run_ctx.bpf_cookie = item->bpf_cookie; 1558 ret &= run_prog(prog, ctx); 1559 item++; 1560 1561 if (!prog->aux->sleepable) 1562 rcu_read_unlock(); 1563 } 1564 bpf_reset_run_ctx(old_run_ctx); 1565 out: 1566 migrate_enable(); 1567 rcu_read_unlock_trace(); 1568 return ret; 1569 } 1570 1571 #ifdef CONFIG_BPF_SYSCALL 1572 DECLARE_PER_CPU(int, bpf_prog_active); 1573 extern struct mutex bpf_stats_enabled_mutex; 1574 1575 /* 1576 * Block execution of BPF programs attached to instrumentation (perf, 1577 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 1578 * these events can happen inside a region which holds a map bucket lock 1579 * and can deadlock on it. 1580 */ 1581 static inline void bpf_disable_instrumentation(void) 1582 { 1583 migrate_disable(); 1584 this_cpu_inc(bpf_prog_active); 1585 } 1586 1587 static inline void bpf_enable_instrumentation(void) 1588 { 1589 this_cpu_dec(bpf_prog_active); 1590 migrate_enable(); 1591 } 1592 1593 extern const struct file_operations bpf_map_fops; 1594 extern const struct file_operations bpf_prog_fops; 1595 extern const struct file_operations bpf_iter_fops; 1596 1597 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 1598 extern const struct bpf_prog_ops _name ## _prog_ops; \ 1599 extern const struct bpf_verifier_ops _name ## _verifier_ops; 1600 #define BPF_MAP_TYPE(_id, _ops) \ 1601 extern const struct bpf_map_ops _ops; 1602 #define BPF_LINK_TYPE(_id, _name) 1603 #include <linux/bpf_types.h> 1604 #undef BPF_PROG_TYPE 1605 #undef BPF_MAP_TYPE 1606 #undef BPF_LINK_TYPE 1607 1608 extern const struct bpf_prog_ops bpf_offload_prog_ops; 1609 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 1610 extern const struct bpf_verifier_ops xdp_analyzer_ops; 1611 1612 struct bpf_prog *bpf_prog_get(u32 ufd); 1613 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 1614 bool attach_drv); 1615 void bpf_prog_add(struct bpf_prog *prog, int i); 1616 void bpf_prog_sub(struct bpf_prog *prog, int i); 1617 void bpf_prog_inc(struct bpf_prog *prog); 1618 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 1619 void bpf_prog_put(struct bpf_prog *prog); 1620 1621 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 1622 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 1623 1624 struct bpf_map_value_off_desc *bpf_map_kptr_off_contains(struct bpf_map *map, u32 offset); 1625 void bpf_map_free_kptr_off_tab(struct bpf_map *map); 1626 struct bpf_map_value_off *bpf_map_copy_kptr_off_tab(const struct bpf_map *map); 1627 bool bpf_map_equal_kptr_off_tab(const struct bpf_map *map_a, const struct bpf_map *map_b); 1628 void bpf_map_free_kptrs(struct bpf_map *map, void *map_value); 1629 1630 struct bpf_map *bpf_map_get(u32 ufd); 1631 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 1632 struct bpf_map *__bpf_map_get(struct fd f); 1633 void bpf_map_inc(struct bpf_map *map); 1634 void bpf_map_inc_with_uref(struct bpf_map *map); 1635 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 1636 void bpf_map_put_with_uref(struct bpf_map *map); 1637 void bpf_map_put(struct bpf_map *map); 1638 void *bpf_map_area_alloc(u64 size, int numa_node); 1639 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 1640 void bpf_map_area_free(void *base); 1641 bool bpf_map_write_active(const struct bpf_map *map); 1642 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 1643 int generic_map_lookup_batch(struct bpf_map *map, 1644 const union bpf_attr *attr, 1645 union bpf_attr __user *uattr); 1646 int generic_map_update_batch(struct bpf_map *map, 1647 const union bpf_attr *attr, 1648 union bpf_attr __user *uattr); 1649 int generic_map_delete_batch(struct bpf_map *map, 1650 const union bpf_attr *attr, 1651 union bpf_attr __user *uattr); 1652 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 1653 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 1654 1655 #ifdef CONFIG_MEMCG_KMEM 1656 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1657 int node); 1658 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 1659 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 1660 size_t align, gfp_t flags); 1661 #else 1662 static inline void * 1663 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1664 int node) 1665 { 1666 return kmalloc_node(size, flags, node); 1667 } 1668 1669 static inline void * 1670 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags) 1671 { 1672 return kzalloc(size, flags); 1673 } 1674 1675 static inline void __percpu * 1676 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align, 1677 gfp_t flags) 1678 { 1679 return __alloc_percpu_gfp(size, align, flags); 1680 } 1681 #endif 1682 1683 extern int sysctl_unprivileged_bpf_disabled; 1684 1685 static inline bool bpf_allow_ptr_leaks(void) 1686 { 1687 return perfmon_capable(); 1688 } 1689 1690 static inline bool bpf_allow_uninit_stack(void) 1691 { 1692 return perfmon_capable(); 1693 } 1694 1695 static inline bool bpf_allow_ptr_to_map_access(void) 1696 { 1697 return perfmon_capable(); 1698 } 1699 1700 static inline bool bpf_bypass_spec_v1(void) 1701 { 1702 return perfmon_capable(); 1703 } 1704 1705 static inline bool bpf_bypass_spec_v4(void) 1706 { 1707 return perfmon_capable(); 1708 } 1709 1710 int bpf_map_new_fd(struct bpf_map *map, int flags); 1711 int bpf_prog_new_fd(struct bpf_prog *prog); 1712 1713 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1714 const struct bpf_link_ops *ops, struct bpf_prog *prog); 1715 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 1716 int bpf_link_settle(struct bpf_link_primer *primer); 1717 void bpf_link_cleanup(struct bpf_link_primer *primer); 1718 void bpf_link_inc(struct bpf_link *link); 1719 void bpf_link_put(struct bpf_link *link); 1720 int bpf_link_new_fd(struct bpf_link *link); 1721 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd); 1722 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 1723 struct bpf_link *bpf_link_get_curr_or_next(u32 *id); 1724 1725 int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 1726 int bpf_obj_get_user(const char __user *pathname, int flags); 1727 1728 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 1729 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 1730 extern int bpf_iter_ ## target(args); \ 1731 int __init bpf_iter_ ## target(args) { return 0; } 1732 1733 struct bpf_iter_aux_info { 1734 /* for map_elem iter */ 1735 struct bpf_map *map; 1736 1737 /* for cgroup iter */ 1738 struct { 1739 struct cgroup *start; /* starting cgroup */ 1740 enum bpf_cgroup_iter_order order; 1741 } cgroup; 1742 }; 1743 1744 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 1745 union bpf_iter_link_info *linfo, 1746 struct bpf_iter_aux_info *aux); 1747 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 1748 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 1749 struct seq_file *seq); 1750 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 1751 struct bpf_link_info *info); 1752 typedef const struct bpf_func_proto * 1753 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 1754 const struct bpf_prog *prog); 1755 1756 enum bpf_iter_feature { 1757 BPF_ITER_RESCHED = BIT(0), 1758 }; 1759 1760 #define BPF_ITER_CTX_ARG_MAX 2 1761 struct bpf_iter_reg { 1762 const char *target; 1763 bpf_iter_attach_target_t attach_target; 1764 bpf_iter_detach_target_t detach_target; 1765 bpf_iter_show_fdinfo_t show_fdinfo; 1766 bpf_iter_fill_link_info_t fill_link_info; 1767 bpf_iter_get_func_proto_t get_func_proto; 1768 u32 ctx_arg_info_size; 1769 u32 feature; 1770 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 1771 const struct bpf_iter_seq_info *seq_info; 1772 }; 1773 1774 struct bpf_iter_meta { 1775 __bpf_md_ptr(struct seq_file *, seq); 1776 u64 session_id; 1777 u64 seq_num; 1778 }; 1779 1780 struct bpf_iter__bpf_map_elem { 1781 __bpf_md_ptr(struct bpf_iter_meta *, meta); 1782 __bpf_md_ptr(struct bpf_map *, map); 1783 __bpf_md_ptr(void *, key); 1784 __bpf_md_ptr(void *, value); 1785 }; 1786 1787 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 1788 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 1789 bool bpf_iter_prog_supported(struct bpf_prog *prog); 1790 const struct bpf_func_proto * 1791 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 1792 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 1793 int bpf_iter_new_fd(struct bpf_link *link); 1794 bool bpf_link_is_iter(struct bpf_link *link); 1795 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 1796 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 1797 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 1798 struct seq_file *seq); 1799 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 1800 struct bpf_link_info *info); 1801 1802 int map_set_for_each_callback_args(struct bpf_verifier_env *env, 1803 struct bpf_func_state *caller, 1804 struct bpf_func_state *callee); 1805 1806 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 1807 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 1808 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 1809 u64 flags); 1810 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 1811 u64 flags); 1812 1813 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 1814 1815 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 1816 void *key, void *value, u64 map_flags); 1817 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1818 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 1819 void *key, void *value, u64 map_flags); 1820 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1821 1822 int bpf_get_file_flag(int flags); 1823 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 1824 size_t actual_size); 1825 1826 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 1827 * forced to use 'long' read/writes to try to atomically copy long counters. 1828 * Best-effort only. No barriers here, since it _will_ race with concurrent 1829 * updates from BPF programs. Called from bpf syscall and mostly used with 1830 * size 8 or 16 bytes, so ask compiler to inline it. 1831 */ 1832 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 1833 { 1834 const long *lsrc = src; 1835 long *ldst = dst; 1836 1837 size /= sizeof(long); 1838 while (size--) 1839 *ldst++ = *lsrc++; 1840 } 1841 1842 /* verify correctness of eBPF program */ 1843 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr); 1844 1845 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 1846 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 1847 #endif 1848 1849 struct btf *bpf_get_btf_vmlinux(void); 1850 1851 /* Map specifics */ 1852 struct xdp_frame; 1853 struct sk_buff; 1854 struct bpf_dtab_netdev; 1855 struct bpf_cpu_map_entry; 1856 1857 void __dev_flush(void); 1858 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 1859 struct net_device *dev_rx); 1860 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 1861 struct net_device *dev_rx); 1862 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 1863 struct bpf_map *map, bool exclude_ingress); 1864 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 1865 struct bpf_prog *xdp_prog); 1866 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1867 struct bpf_prog *xdp_prog, struct bpf_map *map, 1868 bool exclude_ingress); 1869 1870 void __cpu_map_flush(void); 1871 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf, 1872 struct net_device *dev_rx); 1873 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1874 struct sk_buff *skb); 1875 1876 /* Return map's numa specified by userspace */ 1877 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 1878 { 1879 return (attr->map_flags & BPF_F_NUMA_NODE) ? 1880 attr->numa_node : NUMA_NO_NODE; 1881 } 1882 1883 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 1884 int array_map_alloc_check(union bpf_attr *attr); 1885 1886 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 1887 union bpf_attr __user *uattr); 1888 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 1889 union bpf_attr __user *uattr); 1890 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 1891 const union bpf_attr *kattr, 1892 union bpf_attr __user *uattr); 1893 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 1894 const union bpf_attr *kattr, 1895 union bpf_attr __user *uattr); 1896 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 1897 const union bpf_attr *kattr, 1898 union bpf_attr __user *uattr); 1899 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 1900 const union bpf_attr *kattr, 1901 union bpf_attr __user *uattr); 1902 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 1903 const struct bpf_prog *prog, 1904 struct bpf_insn_access_aux *info); 1905 1906 static inline bool bpf_tracing_ctx_access(int off, int size, 1907 enum bpf_access_type type) 1908 { 1909 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS) 1910 return false; 1911 if (type != BPF_READ) 1912 return false; 1913 if (off % size != 0) 1914 return false; 1915 return true; 1916 } 1917 1918 static inline bool bpf_tracing_btf_ctx_access(int off, int size, 1919 enum bpf_access_type type, 1920 const struct bpf_prog *prog, 1921 struct bpf_insn_access_aux *info) 1922 { 1923 if (!bpf_tracing_ctx_access(off, size, type)) 1924 return false; 1925 return btf_ctx_access(off, size, type, prog, info); 1926 } 1927 1928 int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf, 1929 const struct btf_type *t, int off, int size, 1930 enum bpf_access_type atype, 1931 u32 *next_btf_id, enum bpf_type_flag *flag); 1932 bool btf_struct_ids_match(struct bpf_verifier_log *log, 1933 const struct btf *btf, u32 id, int off, 1934 const struct btf *need_btf, u32 need_type_id, 1935 bool strict); 1936 1937 int btf_distill_func_proto(struct bpf_verifier_log *log, 1938 struct btf *btf, 1939 const struct btf_type *func_proto, 1940 const char *func_name, 1941 struct btf_func_model *m); 1942 1943 struct bpf_reg_state; 1944 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog, 1945 struct bpf_reg_state *regs); 1946 int btf_check_kfunc_arg_match(struct bpf_verifier_env *env, 1947 const struct btf *btf, u32 func_id, 1948 struct bpf_reg_state *regs, 1949 u32 kfunc_flags); 1950 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, 1951 struct bpf_reg_state *reg); 1952 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 1953 struct btf *btf, const struct btf_type *t); 1954 1955 struct bpf_prog *bpf_prog_by_id(u32 id); 1956 struct bpf_link *bpf_link_by_id(u32 id); 1957 1958 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id); 1959 void bpf_task_storage_free(struct task_struct *task); 1960 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog); 1961 const struct btf_func_model * 1962 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 1963 const struct bpf_insn *insn); 1964 struct bpf_core_ctx { 1965 struct bpf_verifier_log *log; 1966 const struct btf *btf; 1967 }; 1968 1969 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo, 1970 int relo_idx, void *insn); 1971 1972 static inline bool unprivileged_ebpf_enabled(void) 1973 { 1974 return !sysctl_unprivileged_bpf_disabled; 1975 } 1976 1977 /* Not all bpf prog type has the bpf_ctx. 1978 * For the bpf prog type that has initialized the bpf_ctx, 1979 * this function can be used to decide if a kernel function 1980 * is called by a bpf program. 1981 */ 1982 static inline bool has_current_bpf_ctx(void) 1983 { 1984 return !!current->bpf_ctx; 1985 } 1986 #else /* !CONFIG_BPF_SYSCALL */ 1987 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 1988 { 1989 return ERR_PTR(-EOPNOTSUPP); 1990 } 1991 1992 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 1993 enum bpf_prog_type type, 1994 bool attach_drv) 1995 { 1996 return ERR_PTR(-EOPNOTSUPP); 1997 } 1998 1999 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 2000 { 2001 } 2002 2003 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 2004 { 2005 } 2006 2007 static inline void bpf_prog_put(struct bpf_prog *prog) 2008 { 2009 } 2010 2011 static inline void bpf_prog_inc(struct bpf_prog *prog) 2012 { 2013 } 2014 2015 static inline struct bpf_prog *__must_check 2016 bpf_prog_inc_not_zero(struct bpf_prog *prog) 2017 { 2018 return ERR_PTR(-EOPNOTSUPP); 2019 } 2020 2021 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 2022 const struct bpf_link_ops *ops, 2023 struct bpf_prog *prog) 2024 { 2025 } 2026 2027 static inline int bpf_link_prime(struct bpf_link *link, 2028 struct bpf_link_primer *primer) 2029 { 2030 return -EOPNOTSUPP; 2031 } 2032 2033 static inline int bpf_link_settle(struct bpf_link_primer *primer) 2034 { 2035 return -EOPNOTSUPP; 2036 } 2037 2038 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 2039 { 2040 } 2041 2042 static inline void bpf_link_inc(struct bpf_link *link) 2043 { 2044 } 2045 2046 static inline void bpf_link_put(struct bpf_link *link) 2047 { 2048 } 2049 2050 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 2051 { 2052 return -EOPNOTSUPP; 2053 } 2054 2055 static inline void __dev_flush(void) 2056 { 2057 } 2058 2059 struct xdp_frame; 2060 struct bpf_dtab_netdev; 2061 struct bpf_cpu_map_entry; 2062 2063 static inline 2064 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 2065 struct net_device *dev_rx) 2066 { 2067 return 0; 2068 } 2069 2070 static inline 2071 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 2072 struct net_device *dev_rx) 2073 { 2074 return 0; 2075 } 2076 2077 static inline 2078 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 2079 struct bpf_map *map, bool exclude_ingress) 2080 { 2081 return 0; 2082 } 2083 2084 struct sk_buff; 2085 2086 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 2087 struct sk_buff *skb, 2088 struct bpf_prog *xdp_prog) 2089 { 2090 return 0; 2091 } 2092 2093 static inline 2094 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 2095 struct bpf_prog *xdp_prog, struct bpf_map *map, 2096 bool exclude_ingress) 2097 { 2098 return 0; 2099 } 2100 2101 static inline void __cpu_map_flush(void) 2102 { 2103 } 2104 2105 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 2106 struct xdp_frame *xdpf, 2107 struct net_device *dev_rx) 2108 { 2109 return 0; 2110 } 2111 2112 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 2113 struct sk_buff *skb) 2114 { 2115 return -EOPNOTSUPP; 2116 } 2117 2118 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 2119 enum bpf_prog_type type) 2120 { 2121 return ERR_PTR(-EOPNOTSUPP); 2122 } 2123 2124 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 2125 const union bpf_attr *kattr, 2126 union bpf_attr __user *uattr) 2127 { 2128 return -ENOTSUPP; 2129 } 2130 2131 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 2132 const union bpf_attr *kattr, 2133 union bpf_attr __user *uattr) 2134 { 2135 return -ENOTSUPP; 2136 } 2137 2138 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 2139 const union bpf_attr *kattr, 2140 union bpf_attr __user *uattr) 2141 { 2142 return -ENOTSUPP; 2143 } 2144 2145 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 2146 const union bpf_attr *kattr, 2147 union bpf_attr __user *uattr) 2148 { 2149 return -ENOTSUPP; 2150 } 2151 2152 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 2153 const union bpf_attr *kattr, 2154 union bpf_attr __user *uattr) 2155 { 2156 return -ENOTSUPP; 2157 } 2158 2159 static inline void bpf_map_put(struct bpf_map *map) 2160 { 2161 } 2162 2163 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 2164 { 2165 return ERR_PTR(-ENOTSUPP); 2166 } 2167 2168 static inline const struct bpf_func_proto * 2169 bpf_base_func_proto(enum bpf_func_id func_id) 2170 { 2171 return NULL; 2172 } 2173 2174 static inline void bpf_task_storage_free(struct task_struct *task) 2175 { 2176 } 2177 2178 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog) 2179 { 2180 return false; 2181 } 2182 2183 static inline const struct btf_func_model * 2184 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 2185 const struct bpf_insn *insn) 2186 { 2187 return NULL; 2188 } 2189 2190 static inline bool unprivileged_ebpf_enabled(void) 2191 { 2192 return false; 2193 } 2194 2195 static inline bool has_current_bpf_ctx(void) 2196 { 2197 return false; 2198 } 2199 #endif /* CONFIG_BPF_SYSCALL */ 2200 2201 void __bpf_free_used_btfs(struct bpf_prog_aux *aux, 2202 struct btf_mod_pair *used_btfs, u32 len); 2203 2204 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 2205 enum bpf_prog_type type) 2206 { 2207 return bpf_prog_get_type_dev(ufd, type, false); 2208 } 2209 2210 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 2211 struct bpf_map **used_maps, u32 len); 2212 2213 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 2214 2215 int bpf_prog_offload_compile(struct bpf_prog *prog); 2216 void bpf_prog_offload_destroy(struct bpf_prog *prog); 2217 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 2218 struct bpf_prog *prog); 2219 2220 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 2221 2222 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 2223 int bpf_map_offload_update_elem(struct bpf_map *map, 2224 void *key, void *value, u64 flags); 2225 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 2226 int bpf_map_offload_get_next_key(struct bpf_map *map, 2227 void *key, void *next_key); 2228 2229 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 2230 2231 struct bpf_offload_dev * 2232 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 2233 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 2234 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 2235 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 2236 struct net_device *netdev); 2237 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 2238 struct net_device *netdev); 2239 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 2240 2241 void unpriv_ebpf_notify(int new_state); 2242 2243 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 2244 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 2245 2246 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 2247 { 2248 return aux->offload_requested; 2249 } 2250 2251 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 2252 { 2253 return unlikely(map->ops == &bpf_map_offload_ops); 2254 } 2255 2256 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 2257 void bpf_map_offload_map_free(struct bpf_map *map); 2258 int bpf_prog_test_run_syscall(struct bpf_prog *prog, 2259 const union bpf_attr *kattr, 2260 union bpf_attr __user *uattr); 2261 2262 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 2263 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 2264 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 2265 int sock_map_bpf_prog_query(const union bpf_attr *attr, 2266 union bpf_attr __user *uattr); 2267 2268 void sock_map_unhash(struct sock *sk); 2269 void sock_map_destroy(struct sock *sk); 2270 void sock_map_close(struct sock *sk, long timeout); 2271 #else 2272 static inline int bpf_prog_offload_init(struct bpf_prog *prog, 2273 union bpf_attr *attr) 2274 { 2275 return -EOPNOTSUPP; 2276 } 2277 2278 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 2279 { 2280 return false; 2281 } 2282 2283 static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 2284 { 2285 return false; 2286 } 2287 2288 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 2289 { 2290 return ERR_PTR(-EOPNOTSUPP); 2291 } 2292 2293 static inline void bpf_map_offload_map_free(struct bpf_map *map) 2294 { 2295 } 2296 2297 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 2298 const union bpf_attr *kattr, 2299 union bpf_attr __user *uattr) 2300 { 2301 return -ENOTSUPP; 2302 } 2303 2304 #ifdef CONFIG_BPF_SYSCALL 2305 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 2306 struct bpf_prog *prog) 2307 { 2308 return -EINVAL; 2309 } 2310 2311 static inline int sock_map_prog_detach(const union bpf_attr *attr, 2312 enum bpf_prog_type ptype) 2313 { 2314 return -EOPNOTSUPP; 2315 } 2316 2317 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 2318 u64 flags) 2319 { 2320 return -EOPNOTSUPP; 2321 } 2322 2323 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr, 2324 union bpf_attr __user *uattr) 2325 { 2326 return -EINVAL; 2327 } 2328 #endif /* CONFIG_BPF_SYSCALL */ 2329 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 2330 2331 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 2332 void bpf_sk_reuseport_detach(struct sock *sk); 2333 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 2334 void *value); 2335 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 2336 void *value, u64 map_flags); 2337 #else 2338 static inline void bpf_sk_reuseport_detach(struct sock *sk) 2339 { 2340 } 2341 2342 #ifdef CONFIG_BPF_SYSCALL 2343 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 2344 void *key, void *value) 2345 { 2346 return -EOPNOTSUPP; 2347 } 2348 2349 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 2350 void *key, void *value, 2351 u64 map_flags) 2352 { 2353 return -EOPNOTSUPP; 2354 } 2355 #endif /* CONFIG_BPF_SYSCALL */ 2356 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 2357 2358 /* verifier prototypes for helper functions called from eBPF programs */ 2359 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 2360 extern const struct bpf_func_proto bpf_map_update_elem_proto; 2361 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 2362 extern const struct bpf_func_proto bpf_map_push_elem_proto; 2363 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 2364 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 2365 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto; 2366 2367 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 2368 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 2369 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 2370 extern const struct bpf_func_proto bpf_tail_call_proto; 2371 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 2372 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 2373 extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto; 2374 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 2375 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 2376 extern const struct bpf_func_proto bpf_get_current_comm_proto; 2377 extern const struct bpf_func_proto bpf_get_stackid_proto; 2378 extern const struct bpf_func_proto bpf_get_stack_proto; 2379 extern const struct bpf_func_proto bpf_get_task_stack_proto; 2380 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 2381 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 2382 extern const struct bpf_func_proto bpf_sock_map_update_proto; 2383 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 2384 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 2385 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 2386 extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto; 2387 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 2388 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 2389 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 2390 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 2391 extern const struct bpf_func_proto bpf_spin_lock_proto; 2392 extern const struct bpf_func_proto bpf_spin_unlock_proto; 2393 extern const struct bpf_func_proto bpf_get_local_storage_proto; 2394 extern const struct bpf_func_proto bpf_strtol_proto; 2395 extern const struct bpf_func_proto bpf_strtoul_proto; 2396 extern const struct bpf_func_proto bpf_tcp_sock_proto; 2397 extern const struct bpf_func_proto bpf_jiffies64_proto; 2398 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 2399 extern const struct bpf_func_proto bpf_event_output_data_proto; 2400 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 2401 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 2402 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 2403 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 2404 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 2405 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto; 2406 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto; 2407 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto; 2408 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 2409 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 2410 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 2411 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 2412 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 2413 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto; 2414 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto; 2415 extern const struct bpf_func_proto bpf_copy_from_user_proto; 2416 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 2417 extern const struct bpf_func_proto bpf_snprintf_proto; 2418 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 2419 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 2420 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 2421 extern const struct bpf_func_proto bpf_sock_from_file_proto; 2422 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 2423 extern const struct bpf_func_proto bpf_task_storage_get_proto; 2424 extern const struct bpf_func_proto bpf_task_storage_delete_proto; 2425 extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 2426 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 2427 extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 2428 extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 2429 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto; 2430 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto; 2431 extern const struct bpf_func_proto bpf_find_vma_proto; 2432 extern const struct bpf_func_proto bpf_loop_proto; 2433 extern const struct bpf_func_proto bpf_copy_from_user_task_proto; 2434 extern const struct bpf_func_proto bpf_set_retval_proto; 2435 extern const struct bpf_func_proto bpf_get_retval_proto; 2436 2437 const struct bpf_func_proto *tracing_prog_func_proto( 2438 enum bpf_func_id func_id, const struct bpf_prog *prog); 2439 2440 /* Shared helpers among cBPF and eBPF. */ 2441 void bpf_user_rnd_init_once(void); 2442 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2443 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2444 2445 #if defined(CONFIG_NET) 2446 bool bpf_sock_common_is_valid_access(int off, int size, 2447 enum bpf_access_type type, 2448 struct bpf_insn_access_aux *info); 2449 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2450 struct bpf_insn_access_aux *info); 2451 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2452 const struct bpf_insn *si, 2453 struct bpf_insn *insn_buf, 2454 struct bpf_prog *prog, 2455 u32 *target_size); 2456 #else 2457 static inline bool bpf_sock_common_is_valid_access(int off, int size, 2458 enum bpf_access_type type, 2459 struct bpf_insn_access_aux *info) 2460 { 2461 return false; 2462 } 2463 static inline bool bpf_sock_is_valid_access(int off, int size, 2464 enum bpf_access_type type, 2465 struct bpf_insn_access_aux *info) 2466 { 2467 return false; 2468 } 2469 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2470 const struct bpf_insn *si, 2471 struct bpf_insn *insn_buf, 2472 struct bpf_prog *prog, 2473 u32 *target_size) 2474 { 2475 return 0; 2476 } 2477 #endif 2478 2479 #ifdef CONFIG_INET 2480 struct sk_reuseport_kern { 2481 struct sk_buff *skb; 2482 struct sock *sk; 2483 struct sock *selected_sk; 2484 struct sock *migrating_sk; 2485 void *data_end; 2486 u32 hash; 2487 u32 reuseport_id; 2488 bool bind_inany; 2489 }; 2490 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2491 struct bpf_insn_access_aux *info); 2492 2493 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2494 const struct bpf_insn *si, 2495 struct bpf_insn *insn_buf, 2496 struct bpf_prog *prog, 2497 u32 *target_size); 2498 2499 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2500 struct bpf_insn_access_aux *info); 2501 2502 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2503 const struct bpf_insn *si, 2504 struct bpf_insn *insn_buf, 2505 struct bpf_prog *prog, 2506 u32 *target_size); 2507 #else 2508 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 2509 enum bpf_access_type type, 2510 struct bpf_insn_access_aux *info) 2511 { 2512 return false; 2513 } 2514 2515 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2516 const struct bpf_insn *si, 2517 struct bpf_insn *insn_buf, 2518 struct bpf_prog *prog, 2519 u32 *target_size) 2520 { 2521 return 0; 2522 } 2523 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 2524 enum bpf_access_type type, 2525 struct bpf_insn_access_aux *info) 2526 { 2527 return false; 2528 } 2529 2530 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2531 const struct bpf_insn *si, 2532 struct bpf_insn *insn_buf, 2533 struct bpf_prog *prog, 2534 u32 *target_size) 2535 { 2536 return 0; 2537 } 2538 #endif /* CONFIG_INET */ 2539 2540 enum bpf_text_poke_type { 2541 BPF_MOD_CALL, 2542 BPF_MOD_JUMP, 2543 }; 2544 2545 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 2546 void *addr1, void *addr2); 2547 2548 void *bpf_arch_text_copy(void *dst, void *src, size_t len); 2549 int bpf_arch_text_invalidate(void *dst, size_t len); 2550 2551 struct btf_id_set; 2552 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 2553 2554 #define MAX_BPRINTF_VARARGS 12 2555 2556 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args, 2557 u32 **bin_buf, u32 num_args); 2558 void bpf_bprintf_cleanup(void); 2559 2560 /* the implementation of the opaque uapi struct bpf_dynptr */ 2561 struct bpf_dynptr_kern { 2562 void *data; 2563 /* Size represents the number of usable bytes of dynptr data. 2564 * If for example the offset is at 4 for a local dynptr whose data is 2565 * of type u64, the number of usable bytes is 4. 2566 * 2567 * The upper 8 bits are reserved. It is as follows: 2568 * Bits 0 - 23 = size 2569 * Bits 24 - 30 = dynptr type 2570 * Bit 31 = whether dynptr is read-only 2571 */ 2572 u32 size; 2573 u32 offset; 2574 } __aligned(8); 2575 2576 enum bpf_dynptr_type { 2577 BPF_DYNPTR_TYPE_INVALID, 2578 /* Points to memory that is local to the bpf program */ 2579 BPF_DYNPTR_TYPE_LOCAL, 2580 /* Underlying data is a ringbuf record */ 2581 BPF_DYNPTR_TYPE_RINGBUF, 2582 }; 2583 2584 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 2585 enum bpf_dynptr_type type, u32 offset, u32 size); 2586 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr); 2587 int bpf_dynptr_check_size(u32 size); 2588 2589 #ifdef CONFIG_BPF_LSM 2590 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype); 2591 void bpf_cgroup_atype_put(int cgroup_atype); 2592 #else 2593 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {} 2594 static inline void bpf_cgroup_atype_put(int cgroup_atype) {} 2595 #endif /* CONFIG_BPF_LSM */ 2596 2597 #endif /* _LINUX_BPF_H */ 2598