at v6.1 87 kB view raw
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#include <linux/static_call.h> 31 32struct bpf_verifier_env; 33struct bpf_verifier_log; 34struct perf_event; 35struct bpf_prog; 36struct bpf_prog_aux; 37struct bpf_map; 38struct sock; 39struct seq_file; 40struct btf; 41struct btf_type; 42struct exception_table_entry; 43struct seq_operations; 44struct bpf_iter_aux_info; 45struct bpf_local_storage; 46struct bpf_local_storage_map; 47struct kobject; 48struct mem_cgroup; 49struct module; 50struct bpf_func_state; 51struct ftrace_ops; 52struct cgroup; 53 54extern struct idr btf_idr; 55extern spinlock_t btf_idr_lock; 56extern struct kobject *btf_kobj; 57 58typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64); 59typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, 60 struct bpf_iter_aux_info *aux); 61typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); 62typedef unsigned int (*bpf_func_t)(const void *, 63 const struct bpf_insn *); 64struct bpf_iter_seq_info { 65 const struct seq_operations *seq_ops; 66 bpf_iter_init_seq_priv_t init_seq_private; 67 bpf_iter_fini_seq_priv_t fini_seq_private; 68 u32 seq_priv_size; 69}; 70 71/* map is generic key/value storage optionally accessible by eBPF programs */ 72struct bpf_map_ops { 73 /* funcs callable from userspace (via syscall) */ 74 int (*map_alloc_check)(union bpf_attr *attr); 75 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 76 void (*map_release)(struct bpf_map *map, struct file *map_file); 77 void (*map_free)(struct bpf_map *map); 78 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 79 void (*map_release_uref)(struct bpf_map *map); 80 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 81 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, 82 union bpf_attr __user *uattr); 83 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key, 84 void *value, u64 flags); 85 int (*map_lookup_and_delete_batch)(struct bpf_map *map, 86 const union bpf_attr *attr, 87 union bpf_attr __user *uattr); 88 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr, 89 union bpf_attr __user *uattr); 90 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, 91 union bpf_attr __user *uattr); 92 93 /* funcs callable from userspace and from eBPF programs */ 94 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 95 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 96 int (*map_delete_elem)(struct bpf_map *map, void *key); 97 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 98 int (*map_pop_elem)(struct bpf_map *map, void *value); 99 int (*map_peek_elem)(struct bpf_map *map, void *value); 100 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu); 101 102 /* funcs called by prog_array and perf_event_array map */ 103 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 104 int fd); 105 void (*map_fd_put_ptr)(void *ptr); 106 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 107 u32 (*map_fd_sys_lookup_elem)(void *ptr); 108 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 109 struct seq_file *m); 110 int (*map_check_btf)(const struct bpf_map *map, 111 const struct btf *btf, 112 const struct btf_type *key_type, 113 const struct btf_type *value_type); 114 115 /* Prog poke tracking helpers. */ 116 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux); 117 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux); 118 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, 119 struct bpf_prog *new); 120 121 /* Direct value access helpers. */ 122 int (*map_direct_value_addr)(const struct bpf_map *map, 123 u64 *imm, u32 off); 124 int (*map_direct_value_meta)(const struct bpf_map *map, 125 u64 imm, u32 *off); 126 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma); 127 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, 128 struct poll_table_struct *pts); 129 130 /* Functions called by bpf_local_storage maps */ 131 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, 132 void *owner, u32 size); 133 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, 134 void *owner, u32 size); 135 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner); 136 137 /* Misc helpers.*/ 138 int (*map_redirect)(struct bpf_map *map, u32 ifindex, u64 flags); 139 140 /* map_meta_equal must be implemented for maps that can be 141 * used as an inner map. It is a runtime check to ensure 142 * an inner map can be inserted to an outer map. 143 * 144 * Some properties of the inner map has been used during the 145 * verification time. When inserting an inner map at the runtime, 146 * map_meta_equal has to ensure the inserting map has the same 147 * properties that the verifier has used earlier. 148 */ 149 bool (*map_meta_equal)(const struct bpf_map *meta0, 150 const struct bpf_map *meta1); 151 152 153 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env, 154 struct bpf_func_state *caller, 155 struct bpf_func_state *callee); 156 int (*map_for_each_callback)(struct bpf_map *map, 157 bpf_callback_t callback_fn, 158 void *callback_ctx, u64 flags); 159 160 /* BTF id of struct allocated by map_alloc */ 161 int *map_btf_id; 162 163 /* bpf_iter info used to open a seq_file */ 164 const struct bpf_iter_seq_info *iter_seq_info; 165}; 166 167enum { 168 /* Support at most 8 pointers in a BPF map value */ 169 BPF_MAP_VALUE_OFF_MAX = 8, 170 BPF_MAP_OFF_ARR_MAX = BPF_MAP_VALUE_OFF_MAX + 171 1 + /* for bpf_spin_lock */ 172 1, /* for bpf_timer */ 173}; 174 175enum bpf_kptr_type { 176 BPF_KPTR_UNREF, 177 BPF_KPTR_REF, 178}; 179 180struct bpf_map_value_off_desc { 181 u32 offset; 182 enum bpf_kptr_type type; 183 struct { 184 struct btf *btf; 185 struct module *module; 186 btf_dtor_kfunc_t dtor; 187 u32 btf_id; 188 } kptr; 189}; 190 191struct bpf_map_value_off { 192 u32 nr_off; 193 struct bpf_map_value_off_desc off[]; 194}; 195 196struct bpf_map_off_arr { 197 u32 cnt; 198 u32 field_off[BPF_MAP_OFF_ARR_MAX]; 199 u8 field_sz[BPF_MAP_OFF_ARR_MAX]; 200}; 201 202struct bpf_map { 203 /* The first two cachelines with read-mostly members of which some 204 * are also accessed in fast-path (e.g. ops, max_entries). 205 */ 206 const struct bpf_map_ops *ops ____cacheline_aligned; 207 struct bpf_map *inner_map_meta; 208#ifdef CONFIG_SECURITY 209 void *security; 210#endif 211 enum bpf_map_type map_type; 212 u32 key_size; 213 u32 value_size; 214 u32 max_entries; 215 u64 map_extra; /* any per-map-type extra fields */ 216 u32 map_flags; 217 int spin_lock_off; /* >=0 valid offset, <0 error */ 218 struct bpf_map_value_off *kptr_off_tab; 219 int timer_off; /* >=0 valid offset, <0 error */ 220 u32 id; 221 int numa_node; 222 u32 btf_key_type_id; 223 u32 btf_value_type_id; 224 u32 btf_vmlinux_value_type_id; 225 struct btf *btf; 226#ifdef CONFIG_MEMCG_KMEM 227 struct obj_cgroup *objcg; 228#endif 229 char name[BPF_OBJ_NAME_LEN]; 230 struct bpf_map_off_arr *off_arr; 231 /* The 3rd and 4th cacheline with misc members to avoid false sharing 232 * particularly with refcounting. 233 */ 234 atomic64_t refcnt ____cacheline_aligned; 235 atomic64_t usercnt; 236 struct work_struct work; 237 struct mutex freeze_mutex; 238 atomic64_t writecnt; 239 /* 'Ownership' of program-containing map is claimed by the first program 240 * that is going to use this map or by the first program which FD is 241 * stored in the map to make sure that all callers and callees have the 242 * same prog type, JITed flag and xdp_has_frags flag. 243 */ 244 struct { 245 spinlock_t lock; 246 enum bpf_prog_type type; 247 bool jited; 248 bool xdp_has_frags; 249 } owner; 250 bool bypass_spec_v1; 251 bool frozen; /* write-once; write-protected by freeze_mutex */ 252}; 253 254static inline bool map_value_has_spin_lock(const struct bpf_map *map) 255{ 256 return map->spin_lock_off >= 0; 257} 258 259static inline bool map_value_has_timer(const struct bpf_map *map) 260{ 261 return map->timer_off >= 0; 262} 263 264static inline bool map_value_has_kptrs(const struct bpf_map *map) 265{ 266 return !IS_ERR_OR_NULL(map->kptr_off_tab); 267} 268 269static inline void check_and_init_map_value(struct bpf_map *map, void *dst) 270{ 271 if (unlikely(map_value_has_spin_lock(map))) 272 memset(dst + map->spin_lock_off, 0, sizeof(struct bpf_spin_lock)); 273 if (unlikely(map_value_has_timer(map))) 274 memset(dst + map->timer_off, 0, sizeof(struct bpf_timer)); 275 if (unlikely(map_value_has_kptrs(map))) { 276 struct bpf_map_value_off *tab = map->kptr_off_tab; 277 int i; 278 279 for (i = 0; i < tab->nr_off; i++) 280 *(u64 *)(dst + tab->off[i].offset) = 0; 281 } 282} 283 284/* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 285 * forced to use 'long' read/writes to try to atomically copy long counters. 286 * Best-effort only. No barriers here, since it _will_ race with concurrent 287 * updates from BPF programs. Called from bpf syscall and mostly used with 288 * size 8 or 16 bytes, so ask compiler to inline it. 289 */ 290static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 291{ 292 const long *lsrc = src; 293 long *ldst = dst; 294 295 size /= sizeof(long); 296 while (size--) 297 *ldst++ = *lsrc++; 298} 299 300/* copy everything but bpf_spin_lock, bpf_timer, and kptrs. There could be one of each. */ 301static inline void __copy_map_value(struct bpf_map *map, void *dst, void *src, bool long_memcpy) 302{ 303 u32 curr_off = 0; 304 int i; 305 306 if (likely(!map->off_arr)) { 307 if (long_memcpy) 308 bpf_long_memcpy(dst, src, round_up(map->value_size, 8)); 309 else 310 memcpy(dst, src, map->value_size); 311 return; 312 } 313 314 for (i = 0; i < map->off_arr->cnt; i++) { 315 u32 next_off = map->off_arr->field_off[i]; 316 317 memcpy(dst + curr_off, src + curr_off, next_off - curr_off); 318 curr_off = next_off + map->off_arr->field_sz[i]; 319 } 320 memcpy(dst + curr_off, src + curr_off, map->value_size - curr_off); 321} 322 323static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 324{ 325 __copy_map_value(map, dst, src, false); 326} 327 328static inline void copy_map_value_long(struct bpf_map *map, void *dst, void *src) 329{ 330 __copy_map_value(map, dst, src, true); 331} 332 333static inline void zero_map_value(struct bpf_map *map, void *dst) 334{ 335 u32 curr_off = 0; 336 int i; 337 338 if (likely(!map->off_arr)) { 339 memset(dst, 0, map->value_size); 340 return; 341 } 342 343 for (i = 0; i < map->off_arr->cnt; i++) { 344 u32 next_off = map->off_arr->field_off[i]; 345 346 memset(dst + curr_off, 0, next_off - curr_off); 347 curr_off = next_off + map->off_arr->field_sz[i]; 348 } 349 memset(dst + curr_off, 0, map->value_size - curr_off); 350} 351 352void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 353 bool lock_src); 354void bpf_timer_cancel_and_free(void *timer); 355int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size); 356 357struct bpf_offload_dev; 358struct bpf_offloaded_map; 359 360struct bpf_map_dev_ops { 361 int (*map_get_next_key)(struct bpf_offloaded_map *map, 362 void *key, void *next_key); 363 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 364 void *key, void *value); 365 int (*map_update_elem)(struct bpf_offloaded_map *map, 366 void *key, void *value, u64 flags); 367 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 368}; 369 370struct bpf_offloaded_map { 371 struct bpf_map map; 372 struct net_device *netdev; 373 const struct bpf_map_dev_ops *dev_ops; 374 void *dev_priv; 375 struct list_head offloads; 376}; 377 378static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 379{ 380 return container_of(map, struct bpf_offloaded_map, map); 381} 382 383static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 384{ 385 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 386} 387 388static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 389{ 390 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && 391 map->ops->map_seq_show_elem; 392} 393 394int map_check_no_btf(const struct bpf_map *map, 395 const struct btf *btf, 396 const struct btf_type *key_type, 397 const struct btf_type *value_type); 398 399bool bpf_map_meta_equal(const struct bpf_map *meta0, 400 const struct bpf_map *meta1); 401 402extern const struct bpf_map_ops bpf_map_offload_ops; 403 404/* bpf_type_flag contains a set of flags that are applicable to the values of 405 * arg_type, ret_type and reg_type. For example, a pointer value may be null, 406 * or a memory is read-only. We classify types into two categories: base types 407 * and extended types. Extended types are base types combined with a type flag. 408 * 409 * Currently there are no more than 32 base types in arg_type, ret_type and 410 * reg_types. 411 */ 412#define BPF_BASE_TYPE_BITS 8 413 414enum bpf_type_flag { 415 /* PTR may be NULL. */ 416 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS), 417 418 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is 419 * compatible with both mutable and immutable memory. 420 */ 421 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS), 422 423 /* MEM was "allocated" from a different helper, and cannot be mixed 424 * with regular non-MEM_ALLOC'ed MEM types. 425 */ 426 MEM_ALLOC = BIT(2 + BPF_BASE_TYPE_BITS), 427 428 /* MEM is in user address space. */ 429 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS), 430 431 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged 432 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In 433 * order to drop this tag, it must be passed into bpf_per_cpu_ptr() 434 * or bpf_this_cpu_ptr(), which will return the pointer corresponding 435 * to the specified cpu. 436 */ 437 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS), 438 439 /* Indicates that the argument will be released. */ 440 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS), 441 442 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark 443 * unreferenced and referenced kptr loaded from map value using a load 444 * instruction, so that they can only be dereferenced but not escape the 445 * BPF program into the kernel (i.e. cannot be passed as arguments to 446 * kfunc or bpf helpers). 447 */ 448 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS), 449 450 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS), 451 452 /* DYNPTR points to memory local to the bpf program. */ 453 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS), 454 455 /* DYNPTR points to a kernel-produced ringbuf record. */ 456 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS), 457 458 /* Size is known at compile time. */ 459 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS), 460 461 __BPF_TYPE_FLAG_MAX, 462 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1, 463}; 464 465#define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF) 466 467/* Max number of base types. */ 468#define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS) 469 470/* Max number of all types. */ 471#define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1)) 472 473/* function argument constraints */ 474enum bpf_arg_type { 475 ARG_DONTCARE = 0, /* unused argument in helper function */ 476 477 /* the following constraints used to prototype 478 * bpf_map_lookup/update/delete_elem() functions 479 */ 480 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 481 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 482 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 483 484 /* Used to prototype bpf_memcmp() and other functions that access data 485 * on eBPF program stack 486 */ 487 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 488 489 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 490 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 491 492 ARG_PTR_TO_CTX, /* pointer to context */ 493 ARG_ANYTHING, /* any (initialized) argument is ok */ 494 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 495 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 496 ARG_PTR_TO_INT, /* pointer to int */ 497 ARG_PTR_TO_LONG, /* pointer to long */ 498 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 499 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */ 500 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */ 501 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ 502 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ 503 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ 504 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */ 505 ARG_PTR_TO_STACK, /* pointer to stack */ 506 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */ 507 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */ 508 ARG_PTR_TO_KPTR, /* pointer to referenced kptr */ 509 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */ 510 __BPF_ARG_TYPE_MAX, 511 512 /* Extended arg_types. */ 513 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE, 514 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM, 515 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX, 516 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET, 517 ARG_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_ALLOC_MEM, 518 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK, 519 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID, 520 /* pointer to memory does not need to be initialized, helper function must fill 521 * all bytes or clear them in error case. 522 */ 523 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | ARG_PTR_TO_MEM, 524 /* Pointer to valid memory of size known at compile time. */ 525 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM, 526 527 /* This must be the last entry. Its purpose is to ensure the enum is 528 * wide enough to hold the higher bits reserved for bpf_type_flag. 529 */ 530 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT, 531}; 532static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 533 534/* type of values returned from helper functions */ 535enum bpf_return_type { 536 RET_INTEGER, /* function returns integer */ 537 RET_VOID, /* function doesn't return anything */ 538 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 539 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */ 540 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */ 541 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */ 542 RET_PTR_TO_ALLOC_MEM, /* returns a pointer to dynamically allocated memory */ 543 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ 544 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */ 545 __BPF_RET_TYPE_MAX, 546 547 /* Extended ret_types. */ 548 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE, 549 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET, 550 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK, 551 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON, 552 RET_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_ALLOC | RET_PTR_TO_ALLOC_MEM, 553 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_ALLOC_MEM, 554 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID, 555 556 /* This must be the last entry. Its purpose is to ensure the enum is 557 * wide enough to hold the higher bits reserved for bpf_type_flag. 558 */ 559 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT, 560}; 561static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 562 563/* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 564 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 565 * instructions after verifying 566 */ 567struct bpf_func_proto { 568 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 569 bool gpl_only; 570 bool pkt_access; 571 enum bpf_return_type ret_type; 572 union { 573 struct { 574 enum bpf_arg_type arg1_type; 575 enum bpf_arg_type arg2_type; 576 enum bpf_arg_type arg3_type; 577 enum bpf_arg_type arg4_type; 578 enum bpf_arg_type arg5_type; 579 }; 580 enum bpf_arg_type arg_type[5]; 581 }; 582 union { 583 struct { 584 u32 *arg1_btf_id; 585 u32 *arg2_btf_id; 586 u32 *arg3_btf_id; 587 u32 *arg4_btf_id; 588 u32 *arg5_btf_id; 589 }; 590 u32 *arg_btf_id[5]; 591 struct { 592 size_t arg1_size; 593 size_t arg2_size; 594 size_t arg3_size; 595 size_t arg4_size; 596 size_t arg5_size; 597 }; 598 size_t arg_size[5]; 599 }; 600 int *ret_btf_id; /* return value btf_id */ 601 bool (*allowed)(const struct bpf_prog *prog); 602}; 603 604/* bpf_context is intentionally undefined structure. Pointer to bpf_context is 605 * the first argument to eBPF programs. 606 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 607 */ 608struct bpf_context; 609 610enum bpf_access_type { 611 BPF_READ = 1, 612 BPF_WRITE = 2 613}; 614 615/* types of values stored in eBPF registers */ 616/* Pointer types represent: 617 * pointer 618 * pointer + imm 619 * pointer + (u16) var 620 * pointer + (u16) var + imm 621 * if (range > 0) then [ptr, ptr + range - off) is safe to access 622 * if (id > 0) means that some 'var' was added 623 * if (off > 0) means that 'imm' was added 624 */ 625enum bpf_reg_type { 626 NOT_INIT = 0, /* nothing was written into register */ 627 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 628 PTR_TO_CTX, /* reg points to bpf_context */ 629 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 630 PTR_TO_MAP_VALUE, /* reg points to map element value */ 631 PTR_TO_MAP_KEY, /* reg points to a map element key */ 632 PTR_TO_STACK, /* reg == frame_pointer + offset */ 633 PTR_TO_PACKET_META, /* skb->data - meta_len */ 634 PTR_TO_PACKET, /* reg points to skb->data */ 635 PTR_TO_PACKET_END, /* skb->data + headlen */ 636 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 637 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 638 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 639 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 640 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 641 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 642 /* PTR_TO_BTF_ID points to a kernel struct that does not need 643 * to be null checked by the BPF program. This does not imply the 644 * pointer is _not_ null and in practice this can easily be a null 645 * pointer when reading pointer chains. The assumption is program 646 * context will handle null pointer dereference typically via fault 647 * handling. The verifier must keep this in mind and can make no 648 * assumptions about null or non-null when doing branch analysis. 649 * Further, when passed into helpers the helpers can not, without 650 * additional context, assume the value is non-null. 651 */ 652 PTR_TO_BTF_ID, 653 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not 654 * been checked for null. Used primarily to inform the verifier 655 * an explicit null check is required for this struct. 656 */ 657 PTR_TO_MEM, /* reg points to valid memory region */ 658 PTR_TO_BUF, /* reg points to a read/write buffer */ 659 PTR_TO_FUNC, /* reg points to a bpf program function */ 660 PTR_TO_DYNPTR, /* reg points to a dynptr */ 661 __BPF_REG_TYPE_MAX, 662 663 /* Extended reg_types. */ 664 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE, 665 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET, 666 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON, 667 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK, 668 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID, 669 670 /* This must be the last entry. Its purpose is to ensure the enum is 671 * wide enough to hold the higher bits reserved for bpf_type_flag. 672 */ 673 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT, 674}; 675static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 676 677/* The information passed from prog-specific *_is_valid_access 678 * back to the verifier. 679 */ 680struct bpf_insn_access_aux { 681 enum bpf_reg_type reg_type; 682 union { 683 int ctx_field_size; 684 struct { 685 struct btf *btf; 686 u32 btf_id; 687 }; 688 }; 689 struct bpf_verifier_log *log; /* for verbose logs */ 690}; 691 692static inline void 693bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 694{ 695 aux->ctx_field_size = size; 696} 697 698static inline bool bpf_pseudo_func(const struct bpf_insn *insn) 699{ 700 return insn->code == (BPF_LD | BPF_IMM | BPF_DW) && 701 insn->src_reg == BPF_PSEUDO_FUNC; 702} 703 704struct bpf_prog_ops { 705 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 706 union bpf_attr __user *uattr); 707}; 708 709struct bpf_verifier_ops { 710 /* return eBPF function prototype for verification */ 711 const struct bpf_func_proto * 712 (*get_func_proto)(enum bpf_func_id func_id, 713 const struct bpf_prog *prog); 714 715 /* return true if 'size' wide access at offset 'off' within bpf_context 716 * with 'type' (read or write) is allowed 717 */ 718 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 719 const struct bpf_prog *prog, 720 struct bpf_insn_access_aux *info); 721 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 722 const struct bpf_prog *prog); 723 int (*gen_ld_abs)(const struct bpf_insn *orig, 724 struct bpf_insn *insn_buf); 725 u32 (*convert_ctx_access)(enum bpf_access_type type, 726 const struct bpf_insn *src, 727 struct bpf_insn *dst, 728 struct bpf_prog *prog, u32 *target_size); 729 int (*btf_struct_access)(struct bpf_verifier_log *log, 730 const struct btf *btf, 731 const struct btf_type *t, int off, int size, 732 enum bpf_access_type atype, 733 u32 *next_btf_id, enum bpf_type_flag *flag); 734}; 735 736struct bpf_prog_offload_ops { 737 /* verifier basic callbacks */ 738 int (*insn_hook)(struct bpf_verifier_env *env, 739 int insn_idx, int prev_insn_idx); 740 int (*finalize)(struct bpf_verifier_env *env); 741 /* verifier optimization callbacks (called after .finalize) */ 742 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 743 struct bpf_insn *insn); 744 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 745 /* program management callbacks */ 746 int (*prepare)(struct bpf_prog *prog); 747 int (*translate)(struct bpf_prog *prog); 748 void (*destroy)(struct bpf_prog *prog); 749}; 750 751struct bpf_prog_offload { 752 struct bpf_prog *prog; 753 struct net_device *netdev; 754 struct bpf_offload_dev *offdev; 755 void *dev_priv; 756 struct list_head offloads; 757 bool dev_state; 758 bool opt_failed; 759 void *jited_image; 760 u32 jited_len; 761}; 762 763enum bpf_cgroup_storage_type { 764 BPF_CGROUP_STORAGE_SHARED, 765 BPF_CGROUP_STORAGE_PERCPU, 766 __BPF_CGROUP_STORAGE_MAX 767}; 768 769#define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 770 771/* The longest tracepoint has 12 args. 772 * See include/trace/bpf_probe.h 773 */ 774#define MAX_BPF_FUNC_ARGS 12 775 776/* The maximum number of arguments passed through registers 777 * a single function may have. 778 */ 779#define MAX_BPF_FUNC_REG_ARGS 5 780 781/* The argument is a structure. */ 782#define BTF_FMODEL_STRUCT_ARG BIT(0) 783 784struct btf_func_model { 785 u8 ret_size; 786 u8 nr_args; 787 u8 arg_size[MAX_BPF_FUNC_ARGS]; 788 u8 arg_flags[MAX_BPF_FUNC_ARGS]; 789}; 790 791/* Restore arguments before returning from trampoline to let original function 792 * continue executing. This flag is used for fentry progs when there are no 793 * fexit progs. 794 */ 795#define BPF_TRAMP_F_RESTORE_REGS BIT(0) 796/* Call original function after fentry progs, but before fexit progs. 797 * Makes sense for fentry/fexit, normal calls and indirect calls. 798 */ 799#define BPF_TRAMP_F_CALL_ORIG BIT(1) 800/* Skip current frame and return to parent. Makes sense for fentry/fexit 801 * programs only. Should not be used with normal calls and indirect calls. 802 */ 803#define BPF_TRAMP_F_SKIP_FRAME BIT(2) 804/* Store IP address of the caller on the trampoline stack, 805 * so it's available for trampoline's programs. 806 */ 807#define BPF_TRAMP_F_IP_ARG BIT(3) 808/* Return the return value of fentry prog. Only used by bpf_struct_ops. */ 809#define BPF_TRAMP_F_RET_FENTRY_RET BIT(4) 810 811/* Get original function from stack instead of from provided direct address. 812 * Makes sense for trampolines with fexit or fmod_ret programs. 813 */ 814#define BPF_TRAMP_F_ORIG_STACK BIT(5) 815 816/* This trampoline is on a function with another ftrace_ops with IPMODIFY, 817 * e.g., a live patch. This flag is set and cleared by ftrace call backs, 818 */ 819#define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6) 820 821/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 822 * bytes on x86. 823 */ 824#define BPF_MAX_TRAMP_LINKS 38 825 826struct bpf_tramp_links { 827 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS]; 828 int nr_links; 829}; 830 831struct bpf_tramp_run_ctx; 832 833/* Different use cases for BPF trampoline: 834 * 1. replace nop at the function entry (kprobe equivalent) 835 * flags = BPF_TRAMP_F_RESTORE_REGS 836 * fentry = a set of programs to run before returning from trampoline 837 * 838 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 839 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 840 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 841 * fentry = a set of program to run before calling original function 842 * fexit = a set of program to run after original function 843 * 844 * 3. replace direct call instruction anywhere in the function body 845 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 846 * With flags = 0 847 * fentry = a set of programs to run before returning from trampoline 848 * With flags = BPF_TRAMP_F_CALL_ORIG 849 * orig_call = original callback addr or direct function addr 850 * fentry = a set of program to run before calling original function 851 * fexit = a set of program to run after original function 852 */ 853struct bpf_tramp_image; 854int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end, 855 const struct btf_func_model *m, u32 flags, 856 struct bpf_tramp_links *tlinks, 857 void *orig_call); 858/* these two functions are called from generated trampoline */ 859u64 notrace __bpf_prog_enter(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx); 860void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start, struct bpf_tramp_run_ctx *run_ctx); 861u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx); 862void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start, 863 struct bpf_tramp_run_ctx *run_ctx); 864u64 notrace __bpf_prog_enter_lsm_cgroup(struct bpf_prog *prog, 865 struct bpf_tramp_run_ctx *run_ctx); 866void notrace __bpf_prog_exit_lsm_cgroup(struct bpf_prog *prog, u64 start, 867 struct bpf_tramp_run_ctx *run_ctx); 868u64 notrace __bpf_prog_enter_struct_ops(struct bpf_prog *prog, 869 struct bpf_tramp_run_ctx *run_ctx); 870void notrace __bpf_prog_exit_struct_ops(struct bpf_prog *prog, u64 start, 871 struct bpf_tramp_run_ctx *run_ctx); 872void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr); 873void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr); 874 875struct bpf_ksym { 876 unsigned long start; 877 unsigned long end; 878 char name[KSYM_NAME_LEN]; 879 struct list_head lnode; 880 struct latch_tree_node tnode; 881 bool prog; 882}; 883 884enum bpf_tramp_prog_type { 885 BPF_TRAMP_FENTRY, 886 BPF_TRAMP_FEXIT, 887 BPF_TRAMP_MODIFY_RETURN, 888 BPF_TRAMP_MAX, 889 BPF_TRAMP_REPLACE, /* more than MAX */ 890}; 891 892struct bpf_tramp_image { 893 void *image; 894 struct bpf_ksym ksym; 895 struct percpu_ref pcref; 896 void *ip_after_call; 897 void *ip_epilogue; 898 union { 899 struct rcu_head rcu; 900 struct work_struct work; 901 }; 902}; 903 904struct bpf_trampoline { 905 /* hlist for trampoline_table */ 906 struct hlist_node hlist; 907 struct ftrace_ops *fops; 908 /* serializes access to fields of this trampoline */ 909 struct mutex mutex; 910 refcount_t refcnt; 911 u32 flags; 912 u64 key; 913 struct { 914 struct btf_func_model model; 915 void *addr; 916 bool ftrace_managed; 917 } func; 918 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 919 * program by replacing one of its functions. func.addr is the address 920 * of the function it replaced. 921 */ 922 struct bpf_prog *extension_prog; 923 /* list of BPF programs using this trampoline */ 924 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 925 /* Number of attached programs. A counter per kind. */ 926 int progs_cnt[BPF_TRAMP_MAX]; 927 /* Executable image of trampoline */ 928 struct bpf_tramp_image *cur_image; 929 u64 selector; 930 struct module *mod; 931}; 932 933struct bpf_attach_target_info { 934 struct btf_func_model fmodel; 935 long tgt_addr; 936 const char *tgt_name; 937 const struct btf_type *tgt_type; 938}; 939 940#define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 941 942struct bpf_dispatcher_prog { 943 struct bpf_prog *prog; 944 refcount_t users; 945}; 946 947struct bpf_dispatcher { 948 /* dispatcher mutex */ 949 struct mutex mutex; 950 void *func; 951 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 952 int num_progs; 953 void *image; 954 void *rw_image; 955 u32 image_off; 956 struct bpf_ksym ksym; 957#ifdef CONFIG_HAVE_STATIC_CALL 958 struct static_call_key *sc_key; 959 void *sc_tramp; 960#endif 961}; 962 963static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func( 964 const void *ctx, 965 const struct bpf_insn *insnsi, 966 bpf_func_t bpf_func) 967{ 968 return bpf_func(ctx, insnsi); 969} 970 971#ifdef CONFIG_BPF_JIT 972int bpf_trampoline_link_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr); 973int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr); 974struct bpf_trampoline *bpf_trampoline_get(u64 key, 975 struct bpf_attach_target_info *tgt_info); 976void bpf_trampoline_put(struct bpf_trampoline *tr); 977int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_funcs); 978 979/* 980 * When the architecture supports STATIC_CALL replace the bpf_dispatcher_fn 981 * indirection with a direct call to the bpf program. If the architecture does 982 * not have STATIC_CALL, avoid a double-indirection. 983 */ 984#ifdef CONFIG_HAVE_STATIC_CALL 985 986#define __BPF_DISPATCHER_SC_INIT(_name) \ 987 .sc_key = &STATIC_CALL_KEY(_name), \ 988 .sc_tramp = STATIC_CALL_TRAMP_ADDR(_name), 989 990#define __BPF_DISPATCHER_SC(name) \ 991 DEFINE_STATIC_CALL(bpf_dispatcher_##name##_call, bpf_dispatcher_nop_func) 992 993#define __BPF_DISPATCHER_CALL(name) \ 994 static_call(bpf_dispatcher_##name##_call)(ctx, insnsi, bpf_func) 995 996#define __BPF_DISPATCHER_UPDATE(_d, _new) \ 997 __static_call_update((_d)->sc_key, (_d)->sc_tramp, (_new)) 998 999#else 1000#define __BPF_DISPATCHER_SC_INIT(name) 1001#define __BPF_DISPATCHER_SC(name) 1002#define __BPF_DISPATCHER_CALL(name) bpf_func(ctx, insnsi) 1003#define __BPF_DISPATCHER_UPDATE(_d, _new) 1004#endif 1005 1006#define BPF_DISPATCHER_INIT(_name) { \ 1007 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 1008 .func = &_name##_func, \ 1009 .progs = {}, \ 1010 .num_progs = 0, \ 1011 .image = NULL, \ 1012 .image_off = 0, \ 1013 .ksym = { \ 1014 .name = #_name, \ 1015 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 1016 }, \ 1017 __BPF_DISPATCHER_SC_INIT(_name##_call) \ 1018} 1019 1020#define DEFINE_BPF_DISPATCHER(name) \ 1021 __BPF_DISPATCHER_SC(name); \ 1022 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \ 1023 const void *ctx, \ 1024 const struct bpf_insn *insnsi, \ 1025 bpf_func_t bpf_func) \ 1026 { \ 1027 return __BPF_DISPATCHER_CALL(name); \ 1028 } \ 1029 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 1030 struct bpf_dispatcher bpf_dispatcher_##name = \ 1031 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 1032 1033#define DECLARE_BPF_DISPATCHER(name) \ 1034 unsigned int bpf_dispatcher_##name##_func( \ 1035 const void *ctx, \ 1036 const struct bpf_insn *insnsi, \ 1037 bpf_func_t bpf_func); \ 1038 extern struct bpf_dispatcher bpf_dispatcher_##name; 1039 1040#define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 1041#define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 1042void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 1043 struct bpf_prog *to); 1044/* Called only from JIT-enabled code, so there's no need for stubs. */ 1045void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym); 1046void bpf_image_ksym_del(struct bpf_ksym *ksym); 1047void bpf_ksym_add(struct bpf_ksym *ksym); 1048void bpf_ksym_del(struct bpf_ksym *ksym); 1049int bpf_jit_charge_modmem(u32 size); 1050void bpf_jit_uncharge_modmem(u32 size); 1051bool bpf_prog_has_trampoline(const struct bpf_prog *prog); 1052#else 1053static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link, 1054 struct bpf_trampoline *tr) 1055{ 1056 return -ENOTSUPP; 1057} 1058static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, 1059 struct bpf_trampoline *tr) 1060{ 1061 return -ENOTSUPP; 1062} 1063static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 1064 struct bpf_attach_target_info *tgt_info) 1065{ 1066 return ERR_PTR(-EOPNOTSUPP); 1067} 1068static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 1069#define DEFINE_BPF_DISPATCHER(name) 1070#define DECLARE_BPF_DISPATCHER(name) 1071#define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 1072#define BPF_DISPATCHER_PTR(name) NULL 1073static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 1074 struct bpf_prog *from, 1075 struct bpf_prog *to) {} 1076static inline bool is_bpf_image_address(unsigned long address) 1077{ 1078 return false; 1079} 1080static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog) 1081{ 1082 return false; 1083} 1084#endif 1085 1086struct bpf_func_info_aux { 1087 u16 linkage; 1088 bool unreliable; 1089}; 1090 1091enum bpf_jit_poke_reason { 1092 BPF_POKE_REASON_TAIL_CALL, 1093}; 1094 1095/* Descriptor of pokes pointing /into/ the JITed image. */ 1096struct bpf_jit_poke_descriptor { 1097 void *tailcall_target; 1098 void *tailcall_bypass; 1099 void *bypass_addr; 1100 void *aux; 1101 union { 1102 struct { 1103 struct bpf_map *map; 1104 u32 key; 1105 } tail_call; 1106 }; 1107 bool tailcall_target_stable; 1108 u8 adj_off; 1109 u16 reason; 1110 u32 insn_idx; 1111}; 1112 1113/* reg_type info for ctx arguments */ 1114struct bpf_ctx_arg_aux { 1115 u32 offset; 1116 enum bpf_reg_type reg_type; 1117 u32 btf_id; 1118}; 1119 1120struct btf_mod_pair { 1121 struct btf *btf; 1122 struct module *module; 1123}; 1124 1125struct bpf_kfunc_desc_tab; 1126 1127struct bpf_prog_aux { 1128 atomic64_t refcnt; 1129 u32 used_map_cnt; 1130 u32 used_btf_cnt; 1131 u32 max_ctx_offset; 1132 u32 max_pkt_offset; 1133 u32 max_tp_access; 1134 u32 stack_depth; 1135 u32 id; 1136 u32 func_cnt; /* used by non-func prog as the number of func progs */ 1137 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 1138 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1139 u32 ctx_arg_info_size; 1140 u32 max_rdonly_access; 1141 u32 max_rdwr_access; 1142 struct btf *attach_btf; 1143 const struct bpf_ctx_arg_aux *ctx_arg_info; 1144 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 1145 struct bpf_prog *dst_prog; 1146 struct bpf_trampoline *dst_trampoline; 1147 enum bpf_prog_type saved_dst_prog_type; 1148 enum bpf_attach_type saved_dst_attach_type; 1149 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 1150 bool offload_requested; 1151 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 1152 bool func_proto_unreliable; 1153 bool sleepable; 1154 bool tail_call_reachable; 1155 bool xdp_has_frags; 1156 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 1157 const struct btf_type *attach_func_proto; 1158 /* function name for valid attach_btf_id */ 1159 const char *attach_func_name; 1160 struct bpf_prog **func; 1161 void *jit_data; /* JIT specific data. arch dependent */ 1162 struct bpf_jit_poke_descriptor *poke_tab; 1163 struct bpf_kfunc_desc_tab *kfunc_tab; 1164 struct bpf_kfunc_btf_tab *kfunc_btf_tab; 1165 u32 size_poke_tab; 1166 struct bpf_ksym ksym; 1167 const struct bpf_prog_ops *ops; 1168 struct bpf_map **used_maps; 1169 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 1170 struct btf_mod_pair *used_btfs; 1171 struct bpf_prog *prog; 1172 struct user_struct *user; 1173 u64 load_time; /* ns since boottime */ 1174 u32 verified_insns; 1175 int cgroup_atype; /* enum cgroup_bpf_attach_type */ 1176 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1177 char name[BPF_OBJ_NAME_LEN]; 1178#ifdef CONFIG_SECURITY 1179 void *security; 1180#endif 1181 struct bpf_prog_offload *offload; 1182 struct btf *btf; 1183 struct bpf_func_info *func_info; 1184 struct bpf_func_info_aux *func_info_aux; 1185 /* bpf_line_info loaded from userspace. linfo->insn_off 1186 * has the xlated insn offset. 1187 * Both the main and sub prog share the same linfo. 1188 * The subprog can access its first linfo by 1189 * using the linfo_idx. 1190 */ 1191 struct bpf_line_info *linfo; 1192 /* jited_linfo is the jited addr of the linfo. It has a 1193 * one to one mapping to linfo: 1194 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 1195 * Both the main and sub prog share the same jited_linfo. 1196 * The subprog can access its first jited_linfo by 1197 * using the linfo_idx. 1198 */ 1199 void **jited_linfo; 1200 u32 func_info_cnt; 1201 u32 nr_linfo; 1202 /* subprog can use linfo_idx to access its first linfo and 1203 * jited_linfo. 1204 * main prog always has linfo_idx == 0 1205 */ 1206 u32 linfo_idx; 1207 u32 num_exentries; 1208 struct exception_table_entry *extable; 1209 union { 1210 struct work_struct work; 1211 struct rcu_head rcu; 1212 }; 1213}; 1214 1215struct bpf_prog { 1216 u16 pages; /* Number of allocated pages */ 1217 u16 jited:1, /* Is our filter JIT'ed? */ 1218 jit_requested:1,/* archs need to JIT the prog */ 1219 gpl_compatible:1, /* Is filter GPL compatible? */ 1220 cb_access:1, /* Is control block accessed? */ 1221 dst_needed:1, /* Do we need dst entry? */ 1222 blinding_requested:1, /* needs constant blinding */ 1223 blinded:1, /* Was blinded */ 1224 is_func:1, /* program is a bpf function */ 1225 kprobe_override:1, /* Do we override a kprobe? */ 1226 has_callchain_buf:1, /* callchain buffer allocated? */ 1227 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */ 1228 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */ 1229 call_get_func_ip:1, /* Do we call get_func_ip() */ 1230 tstamp_type_access:1; /* Accessed __sk_buff->tstamp_type */ 1231 enum bpf_prog_type type; /* Type of BPF program */ 1232 enum bpf_attach_type expected_attach_type; /* For some prog types */ 1233 u32 len; /* Number of filter blocks */ 1234 u32 jited_len; /* Size of jited insns in bytes */ 1235 u8 tag[BPF_TAG_SIZE]; 1236 struct bpf_prog_stats __percpu *stats; 1237 int __percpu *active; 1238 unsigned int (*bpf_func)(const void *ctx, 1239 const struct bpf_insn *insn); 1240 struct bpf_prog_aux *aux; /* Auxiliary fields */ 1241 struct sock_fprog_kern *orig_prog; /* Original BPF program */ 1242 /* Instructions for interpreter */ 1243 union { 1244 DECLARE_FLEX_ARRAY(struct sock_filter, insns); 1245 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi); 1246 }; 1247}; 1248 1249struct bpf_array_aux { 1250 /* Programs with direct jumps into programs part of this array. */ 1251 struct list_head poke_progs; 1252 struct bpf_map *map; 1253 struct mutex poke_mutex; 1254 struct work_struct work; 1255}; 1256 1257struct bpf_link { 1258 atomic64_t refcnt; 1259 u32 id; 1260 enum bpf_link_type type; 1261 const struct bpf_link_ops *ops; 1262 struct bpf_prog *prog; 1263 struct work_struct work; 1264}; 1265 1266struct bpf_link_ops { 1267 void (*release)(struct bpf_link *link); 1268 void (*dealloc)(struct bpf_link *link); 1269 int (*detach)(struct bpf_link *link); 1270 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 1271 struct bpf_prog *old_prog); 1272 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 1273 int (*fill_link_info)(const struct bpf_link *link, 1274 struct bpf_link_info *info); 1275}; 1276 1277struct bpf_tramp_link { 1278 struct bpf_link link; 1279 struct hlist_node tramp_hlist; 1280 u64 cookie; 1281}; 1282 1283struct bpf_shim_tramp_link { 1284 struct bpf_tramp_link link; 1285 struct bpf_trampoline *trampoline; 1286}; 1287 1288struct bpf_tracing_link { 1289 struct bpf_tramp_link link; 1290 enum bpf_attach_type attach_type; 1291 struct bpf_trampoline *trampoline; 1292 struct bpf_prog *tgt_prog; 1293}; 1294 1295struct bpf_link_primer { 1296 struct bpf_link *link; 1297 struct file *file; 1298 int fd; 1299 u32 id; 1300}; 1301 1302struct bpf_struct_ops_value; 1303struct btf_member; 1304 1305#define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 1306struct bpf_struct_ops { 1307 const struct bpf_verifier_ops *verifier_ops; 1308 int (*init)(struct btf *btf); 1309 int (*check_member)(const struct btf_type *t, 1310 const struct btf_member *member); 1311 int (*init_member)(const struct btf_type *t, 1312 const struct btf_member *member, 1313 void *kdata, const void *udata); 1314 int (*reg)(void *kdata); 1315 void (*unreg)(void *kdata); 1316 const struct btf_type *type; 1317 const struct btf_type *value_type; 1318 const char *name; 1319 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 1320 u32 type_id; 1321 u32 value_id; 1322}; 1323 1324#if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 1325#define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 1326const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id); 1327void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log); 1328bool bpf_struct_ops_get(const void *kdata); 1329void bpf_struct_ops_put(const void *kdata); 1330int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 1331 void *value); 1332int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks, 1333 struct bpf_tramp_link *link, 1334 const struct btf_func_model *model, 1335 void *image, void *image_end); 1336static inline bool bpf_try_module_get(const void *data, struct module *owner) 1337{ 1338 if (owner == BPF_MODULE_OWNER) 1339 return bpf_struct_ops_get(data); 1340 else 1341 return try_module_get(owner); 1342} 1343static inline void bpf_module_put(const void *data, struct module *owner) 1344{ 1345 if (owner == BPF_MODULE_OWNER) 1346 bpf_struct_ops_put(data); 1347 else 1348 module_put(owner); 1349} 1350 1351#ifdef CONFIG_NET 1352/* Define it here to avoid the use of forward declaration */ 1353struct bpf_dummy_ops_state { 1354 int val; 1355}; 1356 1357struct bpf_dummy_ops { 1358 int (*test_1)(struct bpf_dummy_ops_state *cb); 1359 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2, 1360 char a3, unsigned long a4); 1361}; 1362 1363int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr, 1364 union bpf_attr __user *uattr); 1365#endif 1366#else 1367static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id) 1368{ 1369 return NULL; 1370} 1371static inline void bpf_struct_ops_init(struct btf *btf, 1372 struct bpf_verifier_log *log) 1373{ 1374} 1375static inline bool bpf_try_module_get(const void *data, struct module *owner) 1376{ 1377 return try_module_get(owner); 1378} 1379static inline void bpf_module_put(const void *data, struct module *owner) 1380{ 1381 module_put(owner); 1382} 1383static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 1384 void *key, 1385 void *value) 1386{ 1387 return -EINVAL; 1388} 1389#endif 1390 1391#if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM) 1392int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 1393 int cgroup_atype); 1394void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog); 1395#else 1396static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 1397 int cgroup_atype) 1398{ 1399 return -EOPNOTSUPP; 1400} 1401static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog) 1402{ 1403} 1404#endif 1405 1406struct bpf_array { 1407 struct bpf_map map; 1408 u32 elem_size; 1409 u32 index_mask; 1410 struct bpf_array_aux *aux; 1411 union { 1412 char value[0] __aligned(8); 1413 void *ptrs[0] __aligned(8); 1414 void __percpu *pptrs[0] __aligned(8); 1415 }; 1416}; 1417 1418#define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 1419#define MAX_TAIL_CALL_CNT 33 1420 1421/* Maximum number of loops for bpf_loop */ 1422#define BPF_MAX_LOOPS BIT(23) 1423 1424#define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 1425 BPF_F_RDONLY_PROG | \ 1426 BPF_F_WRONLY | \ 1427 BPF_F_WRONLY_PROG) 1428 1429#define BPF_MAP_CAN_READ BIT(0) 1430#define BPF_MAP_CAN_WRITE BIT(1) 1431 1432/* Maximum number of user-producer ring buffer samples that can be drained in 1433 * a call to bpf_user_ringbuf_drain(). 1434 */ 1435#define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024) 1436 1437static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 1438{ 1439 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1440 1441 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 1442 * not possible. 1443 */ 1444 if (access_flags & BPF_F_RDONLY_PROG) 1445 return BPF_MAP_CAN_READ; 1446 else if (access_flags & BPF_F_WRONLY_PROG) 1447 return BPF_MAP_CAN_WRITE; 1448 else 1449 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 1450} 1451 1452static inline bool bpf_map_flags_access_ok(u32 access_flags) 1453{ 1454 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 1455 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 1456} 1457 1458struct bpf_event_entry { 1459 struct perf_event *event; 1460 struct file *perf_file; 1461 struct file *map_file; 1462 struct rcu_head rcu; 1463}; 1464 1465static inline bool map_type_contains_progs(struct bpf_map *map) 1466{ 1467 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY || 1468 map->map_type == BPF_MAP_TYPE_DEVMAP || 1469 map->map_type == BPF_MAP_TYPE_CPUMAP; 1470} 1471 1472bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp); 1473int bpf_prog_calc_tag(struct bpf_prog *fp); 1474 1475const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 1476const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void); 1477 1478typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 1479 unsigned long off, unsigned long len); 1480typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 1481 const struct bpf_insn *src, 1482 struct bpf_insn *dst, 1483 struct bpf_prog *prog, 1484 u32 *target_size); 1485 1486u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 1487 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 1488 1489/* an array of programs to be executed under rcu_lock. 1490 * 1491 * Typical usage: 1492 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run); 1493 * 1494 * the structure returned by bpf_prog_array_alloc() should be populated 1495 * with program pointers and the last pointer must be NULL. 1496 * The user has to keep refcnt on the program and make sure the program 1497 * is removed from the array before bpf_prog_put(). 1498 * The 'struct bpf_prog_array *' should only be replaced with xchg() 1499 * since other cpus are walking the array of pointers in parallel. 1500 */ 1501struct bpf_prog_array_item { 1502 struct bpf_prog *prog; 1503 union { 1504 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1505 u64 bpf_cookie; 1506 }; 1507}; 1508 1509struct bpf_prog_array { 1510 struct rcu_head rcu; 1511 struct bpf_prog_array_item items[]; 1512}; 1513 1514struct bpf_empty_prog_array { 1515 struct bpf_prog_array hdr; 1516 struct bpf_prog *null_prog; 1517}; 1518 1519/* to avoid allocating empty bpf_prog_array for cgroups that 1520 * don't have bpf program attached use one global 'bpf_empty_prog_array' 1521 * It will not be modified the caller of bpf_prog_array_alloc() 1522 * (since caller requested prog_cnt == 0) 1523 * that pointer should be 'freed' by bpf_prog_array_free() 1524 */ 1525extern struct bpf_empty_prog_array bpf_empty_prog_array; 1526 1527struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 1528void bpf_prog_array_free(struct bpf_prog_array *progs); 1529/* Use when traversal over the bpf_prog_array uses tasks_trace rcu */ 1530void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs); 1531int bpf_prog_array_length(struct bpf_prog_array *progs); 1532bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 1533int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 1534 __u32 __user *prog_ids, u32 cnt); 1535 1536void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 1537 struct bpf_prog *old_prog); 1538int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 1539int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 1540 struct bpf_prog *prog); 1541int bpf_prog_array_copy_info(struct bpf_prog_array *array, 1542 u32 *prog_ids, u32 request_cnt, 1543 u32 *prog_cnt); 1544int bpf_prog_array_copy(struct bpf_prog_array *old_array, 1545 struct bpf_prog *exclude_prog, 1546 struct bpf_prog *include_prog, 1547 u64 bpf_cookie, 1548 struct bpf_prog_array **new_array); 1549 1550struct bpf_run_ctx {}; 1551 1552struct bpf_cg_run_ctx { 1553 struct bpf_run_ctx run_ctx; 1554 const struct bpf_prog_array_item *prog_item; 1555 int retval; 1556}; 1557 1558struct bpf_trace_run_ctx { 1559 struct bpf_run_ctx run_ctx; 1560 u64 bpf_cookie; 1561}; 1562 1563struct bpf_tramp_run_ctx { 1564 struct bpf_run_ctx run_ctx; 1565 u64 bpf_cookie; 1566 struct bpf_run_ctx *saved_run_ctx; 1567}; 1568 1569static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 1570{ 1571 struct bpf_run_ctx *old_ctx = NULL; 1572 1573#ifdef CONFIG_BPF_SYSCALL 1574 old_ctx = current->bpf_ctx; 1575 current->bpf_ctx = new_ctx; 1576#endif 1577 return old_ctx; 1578} 1579 1580static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 1581{ 1582#ifdef CONFIG_BPF_SYSCALL 1583 current->bpf_ctx = old_ctx; 1584#endif 1585} 1586 1587/* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 1588#define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 1589/* BPF program asks to set CN on the packet. */ 1590#define BPF_RET_SET_CN (1 << 0) 1591 1592typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 1593 1594static __always_inline u32 1595bpf_prog_run_array(const struct bpf_prog_array *array, 1596 const void *ctx, bpf_prog_run_fn run_prog) 1597{ 1598 const struct bpf_prog_array_item *item; 1599 const struct bpf_prog *prog; 1600 struct bpf_run_ctx *old_run_ctx; 1601 struct bpf_trace_run_ctx run_ctx; 1602 u32 ret = 1; 1603 1604 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held"); 1605 1606 if (unlikely(!array)) 1607 return ret; 1608 1609 migrate_disable(); 1610 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1611 item = &array->items[0]; 1612 while ((prog = READ_ONCE(item->prog))) { 1613 run_ctx.bpf_cookie = item->bpf_cookie; 1614 ret &= run_prog(prog, ctx); 1615 item++; 1616 } 1617 bpf_reset_run_ctx(old_run_ctx); 1618 migrate_enable(); 1619 return ret; 1620} 1621 1622/* Notes on RCU design for bpf_prog_arrays containing sleepable programs: 1623 * 1624 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array 1625 * overall. As a result, we must use the bpf_prog_array_free_sleepable 1626 * in order to use the tasks_trace rcu grace period. 1627 * 1628 * When a non-sleepable program is inside the array, we take the rcu read 1629 * section and disable preemption for that program alone, so it can access 1630 * rcu-protected dynamically sized maps. 1631 */ 1632static __always_inline u32 1633bpf_prog_run_array_sleepable(const struct bpf_prog_array __rcu *array_rcu, 1634 const void *ctx, bpf_prog_run_fn run_prog) 1635{ 1636 const struct bpf_prog_array_item *item; 1637 const struct bpf_prog *prog; 1638 const struct bpf_prog_array *array; 1639 struct bpf_run_ctx *old_run_ctx; 1640 struct bpf_trace_run_ctx run_ctx; 1641 u32 ret = 1; 1642 1643 might_fault(); 1644 1645 rcu_read_lock_trace(); 1646 migrate_disable(); 1647 1648 array = rcu_dereference_check(array_rcu, rcu_read_lock_trace_held()); 1649 if (unlikely(!array)) 1650 goto out; 1651 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 1652 item = &array->items[0]; 1653 while ((prog = READ_ONCE(item->prog))) { 1654 if (!prog->aux->sleepable) 1655 rcu_read_lock(); 1656 1657 run_ctx.bpf_cookie = item->bpf_cookie; 1658 ret &= run_prog(prog, ctx); 1659 item++; 1660 1661 if (!prog->aux->sleepable) 1662 rcu_read_unlock(); 1663 } 1664 bpf_reset_run_ctx(old_run_ctx); 1665out: 1666 migrate_enable(); 1667 rcu_read_unlock_trace(); 1668 return ret; 1669} 1670 1671#ifdef CONFIG_BPF_SYSCALL 1672DECLARE_PER_CPU(int, bpf_prog_active); 1673extern struct mutex bpf_stats_enabled_mutex; 1674 1675/* 1676 * Block execution of BPF programs attached to instrumentation (perf, 1677 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 1678 * these events can happen inside a region which holds a map bucket lock 1679 * and can deadlock on it. 1680 */ 1681static inline void bpf_disable_instrumentation(void) 1682{ 1683 migrate_disable(); 1684 this_cpu_inc(bpf_prog_active); 1685} 1686 1687static inline void bpf_enable_instrumentation(void) 1688{ 1689 this_cpu_dec(bpf_prog_active); 1690 migrate_enable(); 1691} 1692 1693extern const struct file_operations bpf_map_fops; 1694extern const struct file_operations bpf_prog_fops; 1695extern const struct file_operations bpf_iter_fops; 1696 1697#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 1698 extern const struct bpf_prog_ops _name ## _prog_ops; \ 1699 extern const struct bpf_verifier_ops _name ## _verifier_ops; 1700#define BPF_MAP_TYPE(_id, _ops) \ 1701 extern const struct bpf_map_ops _ops; 1702#define BPF_LINK_TYPE(_id, _name) 1703#include <linux/bpf_types.h> 1704#undef BPF_PROG_TYPE 1705#undef BPF_MAP_TYPE 1706#undef BPF_LINK_TYPE 1707 1708extern const struct bpf_prog_ops bpf_offload_prog_ops; 1709extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 1710extern const struct bpf_verifier_ops xdp_analyzer_ops; 1711 1712struct bpf_prog *bpf_prog_get(u32 ufd); 1713struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 1714 bool attach_drv); 1715void bpf_prog_add(struct bpf_prog *prog, int i); 1716void bpf_prog_sub(struct bpf_prog *prog, int i); 1717void bpf_prog_inc(struct bpf_prog *prog); 1718struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 1719void bpf_prog_put(struct bpf_prog *prog); 1720 1721void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); 1722void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); 1723 1724struct bpf_map_value_off_desc *bpf_map_kptr_off_contains(struct bpf_map *map, u32 offset); 1725void bpf_map_free_kptr_off_tab(struct bpf_map *map); 1726struct bpf_map_value_off *bpf_map_copy_kptr_off_tab(const struct bpf_map *map); 1727bool bpf_map_equal_kptr_off_tab(const struct bpf_map *map_a, const struct bpf_map *map_b); 1728void bpf_map_free_kptrs(struct bpf_map *map, void *map_value); 1729 1730struct bpf_map *bpf_map_get(u32 ufd); 1731struct bpf_map *bpf_map_get_with_uref(u32 ufd); 1732struct bpf_map *__bpf_map_get(struct fd f); 1733void bpf_map_inc(struct bpf_map *map); 1734void bpf_map_inc_with_uref(struct bpf_map *map); 1735struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 1736void bpf_map_put_with_uref(struct bpf_map *map); 1737void bpf_map_put(struct bpf_map *map); 1738void *bpf_map_area_alloc(u64 size, int numa_node); 1739void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 1740void bpf_map_area_free(void *base); 1741bool bpf_map_write_active(const struct bpf_map *map); 1742void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 1743int generic_map_lookup_batch(struct bpf_map *map, 1744 const union bpf_attr *attr, 1745 union bpf_attr __user *uattr); 1746int generic_map_update_batch(struct bpf_map *map, 1747 const union bpf_attr *attr, 1748 union bpf_attr __user *uattr); 1749int generic_map_delete_batch(struct bpf_map *map, 1750 const union bpf_attr *attr, 1751 union bpf_attr __user *uattr); 1752struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 1753struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 1754 1755#ifdef CONFIG_MEMCG_KMEM 1756void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1757 int node); 1758void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 1759void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 1760 size_t align, gfp_t flags); 1761#else 1762static inline void * 1763bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 1764 int node) 1765{ 1766 return kmalloc_node(size, flags, node); 1767} 1768 1769static inline void * 1770bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags) 1771{ 1772 return kzalloc(size, flags); 1773} 1774 1775static inline void __percpu * 1776bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align, 1777 gfp_t flags) 1778{ 1779 return __alloc_percpu_gfp(size, align, flags); 1780} 1781#endif 1782 1783extern int sysctl_unprivileged_bpf_disabled; 1784 1785static inline bool bpf_allow_ptr_leaks(void) 1786{ 1787 return perfmon_capable(); 1788} 1789 1790static inline bool bpf_allow_uninit_stack(void) 1791{ 1792 return perfmon_capable(); 1793} 1794 1795static inline bool bpf_allow_ptr_to_map_access(void) 1796{ 1797 return perfmon_capable(); 1798} 1799 1800static inline bool bpf_bypass_spec_v1(void) 1801{ 1802 return perfmon_capable(); 1803} 1804 1805static inline bool bpf_bypass_spec_v4(void) 1806{ 1807 return perfmon_capable(); 1808} 1809 1810int bpf_map_new_fd(struct bpf_map *map, int flags); 1811int bpf_prog_new_fd(struct bpf_prog *prog); 1812 1813void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 1814 const struct bpf_link_ops *ops, struct bpf_prog *prog); 1815int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 1816int bpf_link_settle(struct bpf_link_primer *primer); 1817void bpf_link_cleanup(struct bpf_link_primer *primer); 1818void bpf_link_inc(struct bpf_link *link); 1819void bpf_link_put(struct bpf_link *link); 1820int bpf_link_new_fd(struct bpf_link *link); 1821struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd); 1822struct bpf_link *bpf_link_get_from_fd(u32 ufd); 1823struct bpf_link *bpf_link_get_curr_or_next(u32 *id); 1824 1825int bpf_obj_pin_user(u32 ufd, const char __user *pathname); 1826int bpf_obj_get_user(const char __user *pathname, int flags); 1827 1828#define BPF_ITER_FUNC_PREFIX "bpf_iter_" 1829#define DEFINE_BPF_ITER_FUNC(target, args...) \ 1830 extern int bpf_iter_ ## target(args); \ 1831 int __init bpf_iter_ ## target(args) { return 0; } 1832 1833/* 1834 * The task type of iterators. 1835 * 1836 * For BPF task iterators, they can be parameterized with various 1837 * parameters to visit only some of tasks. 1838 * 1839 * BPF_TASK_ITER_ALL (default) 1840 * Iterate over resources of every task. 1841 * 1842 * BPF_TASK_ITER_TID 1843 * Iterate over resources of a task/tid. 1844 * 1845 * BPF_TASK_ITER_TGID 1846 * Iterate over resources of every task of a process / task group. 1847 */ 1848enum bpf_iter_task_type { 1849 BPF_TASK_ITER_ALL = 0, 1850 BPF_TASK_ITER_TID, 1851 BPF_TASK_ITER_TGID, 1852}; 1853 1854struct bpf_iter_aux_info { 1855 /* for map_elem iter */ 1856 struct bpf_map *map; 1857 1858 /* for cgroup iter */ 1859 struct { 1860 struct cgroup *start; /* starting cgroup */ 1861 enum bpf_cgroup_iter_order order; 1862 } cgroup; 1863 struct { 1864 enum bpf_iter_task_type type; 1865 u32 pid; 1866 } task; 1867}; 1868 1869typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 1870 union bpf_iter_link_info *linfo, 1871 struct bpf_iter_aux_info *aux); 1872typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 1873typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 1874 struct seq_file *seq); 1875typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 1876 struct bpf_link_info *info); 1877typedef const struct bpf_func_proto * 1878(*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 1879 const struct bpf_prog *prog); 1880 1881enum bpf_iter_feature { 1882 BPF_ITER_RESCHED = BIT(0), 1883}; 1884 1885#define BPF_ITER_CTX_ARG_MAX 2 1886struct bpf_iter_reg { 1887 const char *target; 1888 bpf_iter_attach_target_t attach_target; 1889 bpf_iter_detach_target_t detach_target; 1890 bpf_iter_show_fdinfo_t show_fdinfo; 1891 bpf_iter_fill_link_info_t fill_link_info; 1892 bpf_iter_get_func_proto_t get_func_proto; 1893 u32 ctx_arg_info_size; 1894 u32 feature; 1895 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 1896 const struct bpf_iter_seq_info *seq_info; 1897}; 1898 1899struct bpf_iter_meta { 1900 __bpf_md_ptr(struct seq_file *, seq); 1901 u64 session_id; 1902 u64 seq_num; 1903}; 1904 1905struct bpf_iter__bpf_map_elem { 1906 __bpf_md_ptr(struct bpf_iter_meta *, meta); 1907 __bpf_md_ptr(struct bpf_map *, map); 1908 __bpf_md_ptr(void *, key); 1909 __bpf_md_ptr(void *, value); 1910}; 1911 1912int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 1913void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 1914bool bpf_iter_prog_supported(struct bpf_prog *prog); 1915const struct bpf_func_proto * 1916bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 1917int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 1918int bpf_iter_new_fd(struct bpf_link *link); 1919bool bpf_link_is_iter(struct bpf_link *link); 1920struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 1921int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 1922void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 1923 struct seq_file *seq); 1924int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 1925 struct bpf_link_info *info); 1926 1927int map_set_for_each_callback_args(struct bpf_verifier_env *env, 1928 struct bpf_func_state *caller, 1929 struct bpf_func_state *callee); 1930 1931int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 1932int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 1933int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 1934 u64 flags); 1935int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 1936 u64 flags); 1937 1938int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 1939 1940int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 1941 void *key, void *value, u64 map_flags); 1942int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1943int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 1944 void *key, void *value, u64 map_flags); 1945int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 1946 1947int bpf_get_file_flag(int flags); 1948int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 1949 size_t actual_size); 1950 1951/* verify correctness of eBPF program */ 1952int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr); 1953 1954#ifndef CONFIG_BPF_JIT_ALWAYS_ON 1955void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 1956#endif 1957 1958struct btf *bpf_get_btf_vmlinux(void); 1959 1960/* Map specifics */ 1961struct xdp_frame; 1962struct sk_buff; 1963struct bpf_dtab_netdev; 1964struct bpf_cpu_map_entry; 1965 1966void __dev_flush(void); 1967int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 1968 struct net_device *dev_rx); 1969int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 1970 struct net_device *dev_rx); 1971int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 1972 struct bpf_map *map, bool exclude_ingress); 1973int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 1974 struct bpf_prog *xdp_prog); 1975int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 1976 struct bpf_prog *xdp_prog, struct bpf_map *map, 1977 bool exclude_ingress); 1978 1979void __cpu_map_flush(void); 1980int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf, 1981 struct net_device *dev_rx); 1982int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 1983 struct sk_buff *skb); 1984 1985/* Return map's numa specified by userspace */ 1986static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 1987{ 1988 return (attr->map_flags & BPF_F_NUMA_NODE) ? 1989 attr->numa_node : NUMA_NO_NODE; 1990} 1991 1992struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 1993int array_map_alloc_check(union bpf_attr *attr); 1994 1995int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 1996 union bpf_attr __user *uattr); 1997int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 1998 union bpf_attr __user *uattr); 1999int bpf_prog_test_run_tracing(struct bpf_prog *prog, 2000 const union bpf_attr *kattr, 2001 union bpf_attr __user *uattr); 2002int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 2003 const union bpf_attr *kattr, 2004 union bpf_attr __user *uattr); 2005int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 2006 const union bpf_attr *kattr, 2007 union bpf_attr __user *uattr); 2008int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 2009 const union bpf_attr *kattr, 2010 union bpf_attr __user *uattr); 2011bool btf_ctx_access(int off, int size, enum bpf_access_type type, 2012 const struct bpf_prog *prog, 2013 struct bpf_insn_access_aux *info); 2014 2015static inline bool bpf_tracing_ctx_access(int off, int size, 2016 enum bpf_access_type type) 2017{ 2018 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS) 2019 return false; 2020 if (type != BPF_READ) 2021 return false; 2022 if (off % size != 0) 2023 return false; 2024 return true; 2025} 2026 2027static inline bool bpf_tracing_btf_ctx_access(int off, int size, 2028 enum bpf_access_type type, 2029 const struct bpf_prog *prog, 2030 struct bpf_insn_access_aux *info) 2031{ 2032 if (!bpf_tracing_ctx_access(off, size, type)) 2033 return false; 2034 return btf_ctx_access(off, size, type, prog, info); 2035} 2036 2037int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf, 2038 const struct btf_type *t, int off, int size, 2039 enum bpf_access_type atype, 2040 u32 *next_btf_id, enum bpf_type_flag *flag); 2041bool btf_struct_ids_match(struct bpf_verifier_log *log, 2042 const struct btf *btf, u32 id, int off, 2043 const struct btf *need_btf, u32 need_type_id, 2044 bool strict); 2045 2046int btf_distill_func_proto(struct bpf_verifier_log *log, 2047 struct btf *btf, 2048 const struct btf_type *func_proto, 2049 const char *func_name, 2050 struct btf_func_model *m); 2051 2052struct bpf_kfunc_arg_meta { 2053 u64 r0_size; 2054 bool r0_rdonly; 2055 int ref_obj_id; 2056 u32 flags; 2057}; 2058 2059struct bpf_reg_state; 2060int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog, 2061 struct bpf_reg_state *regs); 2062int btf_check_subprog_call(struct bpf_verifier_env *env, int subprog, 2063 struct bpf_reg_state *regs); 2064int btf_check_kfunc_arg_match(struct bpf_verifier_env *env, 2065 const struct btf *btf, u32 func_id, 2066 struct bpf_reg_state *regs, 2067 struct bpf_kfunc_arg_meta *meta); 2068int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, 2069 struct bpf_reg_state *reg); 2070int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 2071 struct btf *btf, const struct btf_type *t); 2072 2073struct bpf_prog *bpf_prog_by_id(u32 id); 2074struct bpf_link *bpf_link_by_id(u32 id); 2075 2076const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id); 2077void bpf_task_storage_free(struct task_struct *task); 2078bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog); 2079const struct btf_func_model * 2080bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 2081 const struct bpf_insn *insn); 2082struct bpf_core_ctx { 2083 struct bpf_verifier_log *log; 2084 const struct btf *btf; 2085}; 2086 2087int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo, 2088 int relo_idx, void *insn); 2089 2090static inline bool unprivileged_ebpf_enabled(void) 2091{ 2092 return !sysctl_unprivileged_bpf_disabled; 2093} 2094 2095/* Not all bpf prog type has the bpf_ctx. 2096 * For the bpf prog type that has initialized the bpf_ctx, 2097 * this function can be used to decide if a kernel function 2098 * is called by a bpf program. 2099 */ 2100static inline bool has_current_bpf_ctx(void) 2101{ 2102 return !!current->bpf_ctx; 2103} 2104 2105void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog); 2106#else /* !CONFIG_BPF_SYSCALL */ 2107static inline struct bpf_prog *bpf_prog_get(u32 ufd) 2108{ 2109 return ERR_PTR(-EOPNOTSUPP); 2110} 2111 2112static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 2113 enum bpf_prog_type type, 2114 bool attach_drv) 2115{ 2116 return ERR_PTR(-EOPNOTSUPP); 2117} 2118 2119static inline void bpf_prog_add(struct bpf_prog *prog, int i) 2120{ 2121} 2122 2123static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 2124{ 2125} 2126 2127static inline void bpf_prog_put(struct bpf_prog *prog) 2128{ 2129} 2130 2131static inline void bpf_prog_inc(struct bpf_prog *prog) 2132{ 2133} 2134 2135static inline struct bpf_prog *__must_check 2136bpf_prog_inc_not_zero(struct bpf_prog *prog) 2137{ 2138 return ERR_PTR(-EOPNOTSUPP); 2139} 2140 2141static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 2142 const struct bpf_link_ops *ops, 2143 struct bpf_prog *prog) 2144{ 2145} 2146 2147static inline int bpf_link_prime(struct bpf_link *link, 2148 struct bpf_link_primer *primer) 2149{ 2150 return -EOPNOTSUPP; 2151} 2152 2153static inline int bpf_link_settle(struct bpf_link_primer *primer) 2154{ 2155 return -EOPNOTSUPP; 2156} 2157 2158static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 2159{ 2160} 2161 2162static inline void bpf_link_inc(struct bpf_link *link) 2163{ 2164} 2165 2166static inline void bpf_link_put(struct bpf_link *link) 2167{ 2168} 2169 2170static inline int bpf_obj_get_user(const char __user *pathname, int flags) 2171{ 2172 return -EOPNOTSUPP; 2173} 2174 2175static inline void __dev_flush(void) 2176{ 2177} 2178 2179struct xdp_frame; 2180struct bpf_dtab_netdev; 2181struct bpf_cpu_map_entry; 2182 2183static inline 2184int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 2185 struct net_device *dev_rx) 2186{ 2187 return 0; 2188} 2189 2190static inline 2191int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 2192 struct net_device *dev_rx) 2193{ 2194 return 0; 2195} 2196 2197static inline 2198int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 2199 struct bpf_map *map, bool exclude_ingress) 2200{ 2201 return 0; 2202} 2203 2204struct sk_buff; 2205 2206static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 2207 struct sk_buff *skb, 2208 struct bpf_prog *xdp_prog) 2209{ 2210 return 0; 2211} 2212 2213static inline 2214int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 2215 struct bpf_prog *xdp_prog, struct bpf_map *map, 2216 bool exclude_ingress) 2217{ 2218 return 0; 2219} 2220 2221static inline void __cpu_map_flush(void) 2222{ 2223} 2224 2225static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 2226 struct xdp_frame *xdpf, 2227 struct net_device *dev_rx) 2228{ 2229 return 0; 2230} 2231 2232static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 2233 struct sk_buff *skb) 2234{ 2235 return -EOPNOTSUPP; 2236} 2237 2238static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 2239 enum bpf_prog_type type) 2240{ 2241 return ERR_PTR(-EOPNOTSUPP); 2242} 2243 2244static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 2245 const union bpf_attr *kattr, 2246 union bpf_attr __user *uattr) 2247{ 2248 return -ENOTSUPP; 2249} 2250 2251static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 2252 const union bpf_attr *kattr, 2253 union bpf_attr __user *uattr) 2254{ 2255 return -ENOTSUPP; 2256} 2257 2258static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 2259 const union bpf_attr *kattr, 2260 union bpf_attr __user *uattr) 2261{ 2262 return -ENOTSUPP; 2263} 2264 2265static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 2266 const union bpf_attr *kattr, 2267 union bpf_attr __user *uattr) 2268{ 2269 return -ENOTSUPP; 2270} 2271 2272static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 2273 const union bpf_attr *kattr, 2274 union bpf_attr __user *uattr) 2275{ 2276 return -ENOTSUPP; 2277} 2278 2279static inline void bpf_map_put(struct bpf_map *map) 2280{ 2281} 2282 2283static inline struct bpf_prog *bpf_prog_by_id(u32 id) 2284{ 2285 return ERR_PTR(-ENOTSUPP); 2286} 2287 2288static inline int btf_struct_access(struct bpf_verifier_log *log, 2289 const struct btf *btf, 2290 const struct btf_type *t, int off, int size, 2291 enum bpf_access_type atype, 2292 u32 *next_btf_id, enum bpf_type_flag *flag) 2293{ 2294 return -EACCES; 2295} 2296 2297static inline const struct bpf_func_proto * 2298bpf_base_func_proto(enum bpf_func_id func_id) 2299{ 2300 return NULL; 2301} 2302 2303static inline void bpf_task_storage_free(struct task_struct *task) 2304{ 2305} 2306 2307static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog) 2308{ 2309 return false; 2310} 2311 2312static inline const struct btf_func_model * 2313bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 2314 const struct bpf_insn *insn) 2315{ 2316 return NULL; 2317} 2318 2319static inline bool unprivileged_ebpf_enabled(void) 2320{ 2321 return false; 2322} 2323 2324static inline bool has_current_bpf_ctx(void) 2325{ 2326 return false; 2327} 2328 2329static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog) 2330{ 2331} 2332#endif /* CONFIG_BPF_SYSCALL */ 2333 2334void __bpf_free_used_btfs(struct bpf_prog_aux *aux, 2335 struct btf_mod_pair *used_btfs, u32 len); 2336 2337static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 2338 enum bpf_prog_type type) 2339{ 2340 return bpf_prog_get_type_dev(ufd, type, false); 2341} 2342 2343void __bpf_free_used_maps(struct bpf_prog_aux *aux, 2344 struct bpf_map **used_maps, u32 len); 2345 2346bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 2347 2348int bpf_prog_offload_compile(struct bpf_prog *prog); 2349void bpf_prog_offload_destroy(struct bpf_prog *prog); 2350int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 2351 struct bpf_prog *prog); 2352 2353int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 2354 2355int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 2356int bpf_map_offload_update_elem(struct bpf_map *map, 2357 void *key, void *value, u64 flags); 2358int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 2359int bpf_map_offload_get_next_key(struct bpf_map *map, 2360 void *key, void *next_key); 2361 2362bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 2363 2364struct bpf_offload_dev * 2365bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 2366void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 2367void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 2368int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 2369 struct net_device *netdev); 2370void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 2371 struct net_device *netdev); 2372bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 2373 2374void unpriv_ebpf_notify(int new_state); 2375 2376#if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 2377int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); 2378 2379static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 2380{ 2381 return aux->offload_requested; 2382} 2383 2384static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 2385{ 2386 return unlikely(map->ops == &bpf_map_offload_ops); 2387} 2388 2389struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 2390void bpf_map_offload_map_free(struct bpf_map *map); 2391int bpf_prog_test_run_syscall(struct bpf_prog *prog, 2392 const union bpf_attr *kattr, 2393 union bpf_attr __user *uattr); 2394 2395int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 2396int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 2397int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 2398int sock_map_bpf_prog_query(const union bpf_attr *attr, 2399 union bpf_attr __user *uattr); 2400 2401void sock_map_unhash(struct sock *sk); 2402void sock_map_destroy(struct sock *sk); 2403void sock_map_close(struct sock *sk, long timeout); 2404#else 2405static inline int bpf_prog_offload_init(struct bpf_prog *prog, 2406 union bpf_attr *attr) 2407{ 2408 return -EOPNOTSUPP; 2409} 2410 2411static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) 2412{ 2413 return false; 2414} 2415 2416static inline bool bpf_map_is_dev_bound(struct bpf_map *map) 2417{ 2418 return false; 2419} 2420 2421static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 2422{ 2423 return ERR_PTR(-EOPNOTSUPP); 2424} 2425 2426static inline void bpf_map_offload_map_free(struct bpf_map *map) 2427{ 2428} 2429 2430static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 2431 const union bpf_attr *kattr, 2432 union bpf_attr __user *uattr) 2433{ 2434 return -ENOTSUPP; 2435} 2436 2437#ifdef CONFIG_BPF_SYSCALL 2438static inline int sock_map_get_from_fd(const union bpf_attr *attr, 2439 struct bpf_prog *prog) 2440{ 2441 return -EINVAL; 2442} 2443 2444static inline int sock_map_prog_detach(const union bpf_attr *attr, 2445 enum bpf_prog_type ptype) 2446{ 2447 return -EOPNOTSUPP; 2448} 2449 2450static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 2451 u64 flags) 2452{ 2453 return -EOPNOTSUPP; 2454} 2455 2456static inline int sock_map_bpf_prog_query(const union bpf_attr *attr, 2457 union bpf_attr __user *uattr) 2458{ 2459 return -EINVAL; 2460} 2461#endif /* CONFIG_BPF_SYSCALL */ 2462#endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 2463 2464#if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 2465void bpf_sk_reuseport_detach(struct sock *sk); 2466int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 2467 void *value); 2468int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 2469 void *value, u64 map_flags); 2470#else 2471static inline void bpf_sk_reuseport_detach(struct sock *sk) 2472{ 2473} 2474 2475#ifdef CONFIG_BPF_SYSCALL 2476static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 2477 void *key, void *value) 2478{ 2479 return -EOPNOTSUPP; 2480} 2481 2482static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 2483 void *key, void *value, 2484 u64 map_flags) 2485{ 2486 return -EOPNOTSUPP; 2487} 2488#endif /* CONFIG_BPF_SYSCALL */ 2489#endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 2490 2491/* verifier prototypes for helper functions called from eBPF programs */ 2492extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 2493extern const struct bpf_func_proto bpf_map_update_elem_proto; 2494extern const struct bpf_func_proto bpf_map_delete_elem_proto; 2495extern const struct bpf_func_proto bpf_map_push_elem_proto; 2496extern const struct bpf_func_proto bpf_map_pop_elem_proto; 2497extern const struct bpf_func_proto bpf_map_peek_elem_proto; 2498extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto; 2499 2500extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 2501extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 2502extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 2503extern const struct bpf_func_proto bpf_tail_call_proto; 2504extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 2505extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 2506extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto; 2507extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 2508extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 2509extern const struct bpf_func_proto bpf_get_current_comm_proto; 2510extern const struct bpf_func_proto bpf_get_stackid_proto; 2511extern const struct bpf_func_proto bpf_get_stack_proto; 2512extern const struct bpf_func_proto bpf_get_task_stack_proto; 2513extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 2514extern const struct bpf_func_proto bpf_get_stack_proto_pe; 2515extern const struct bpf_func_proto bpf_sock_map_update_proto; 2516extern const struct bpf_func_proto bpf_sock_hash_update_proto; 2517extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 2518extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 2519extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto; 2520extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 2521extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 2522extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 2523extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 2524extern const struct bpf_func_proto bpf_spin_lock_proto; 2525extern const struct bpf_func_proto bpf_spin_unlock_proto; 2526extern const struct bpf_func_proto bpf_get_local_storage_proto; 2527extern const struct bpf_func_proto bpf_strtol_proto; 2528extern const struct bpf_func_proto bpf_strtoul_proto; 2529extern const struct bpf_func_proto bpf_tcp_sock_proto; 2530extern const struct bpf_func_proto bpf_jiffies64_proto; 2531extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 2532extern const struct bpf_func_proto bpf_event_output_data_proto; 2533extern const struct bpf_func_proto bpf_ringbuf_output_proto; 2534extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 2535extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 2536extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 2537extern const struct bpf_func_proto bpf_ringbuf_query_proto; 2538extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto; 2539extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto; 2540extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto; 2541extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 2542extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 2543extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 2544extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 2545extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 2546extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto; 2547extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto; 2548extern const struct bpf_func_proto bpf_copy_from_user_proto; 2549extern const struct bpf_func_proto bpf_snprintf_btf_proto; 2550extern const struct bpf_func_proto bpf_snprintf_proto; 2551extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 2552extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 2553extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 2554extern const struct bpf_func_proto bpf_sock_from_file_proto; 2555extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 2556extern const struct bpf_func_proto bpf_task_storage_get_proto; 2557extern const struct bpf_func_proto bpf_task_storage_delete_proto; 2558extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 2559extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 2560extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 2561extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 2562extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto; 2563extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto; 2564extern const struct bpf_func_proto bpf_find_vma_proto; 2565extern const struct bpf_func_proto bpf_loop_proto; 2566extern const struct bpf_func_proto bpf_copy_from_user_task_proto; 2567extern const struct bpf_func_proto bpf_set_retval_proto; 2568extern const struct bpf_func_proto bpf_get_retval_proto; 2569extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto; 2570 2571const struct bpf_func_proto *tracing_prog_func_proto( 2572 enum bpf_func_id func_id, const struct bpf_prog *prog); 2573 2574/* Shared helpers among cBPF and eBPF. */ 2575void bpf_user_rnd_init_once(void); 2576u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2577u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 2578 2579#if defined(CONFIG_NET) 2580bool bpf_sock_common_is_valid_access(int off, int size, 2581 enum bpf_access_type type, 2582 struct bpf_insn_access_aux *info); 2583bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2584 struct bpf_insn_access_aux *info); 2585u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2586 const struct bpf_insn *si, 2587 struct bpf_insn *insn_buf, 2588 struct bpf_prog *prog, 2589 u32 *target_size); 2590#else 2591static inline bool bpf_sock_common_is_valid_access(int off, int size, 2592 enum bpf_access_type type, 2593 struct bpf_insn_access_aux *info) 2594{ 2595 return false; 2596} 2597static inline bool bpf_sock_is_valid_access(int off, int size, 2598 enum bpf_access_type type, 2599 struct bpf_insn_access_aux *info) 2600{ 2601 return false; 2602} 2603static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 2604 const struct bpf_insn *si, 2605 struct bpf_insn *insn_buf, 2606 struct bpf_prog *prog, 2607 u32 *target_size) 2608{ 2609 return 0; 2610} 2611#endif 2612 2613#ifdef CONFIG_INET 2614struct sk_reuseport_kern { 2615 struct sk_buff *skb; 2616 struct sock *sk; 2617 struct sock *selected_sk; 2618 struct sock *migrating_sk; 2619 void *data_end; 2620 u32 hash; 2621 u32 reuseport_id; 2622 bool bind_inany; 2623}; 2624bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2625 struct bpf_insn_access_aux *info); 2626 2627u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2628 const struct bpf_insn *si, 2629 struct bpf_insn *insn_buf, 2630 struct bpf_prog *prog, 2631 u32 *target_size); 2632 2633bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 2634 struct bpf_insn_access_aux *info); 2635 2636u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2637 const struct bpf_insn *si, 2638 struct bpf_insn *insn_buf, 2639 struct bpf_prog *prog, 2640 u32 *target_size); 2641#else 2642static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 2643 enum bpf_access_type type, 2644 struct bpf_insn_access_aux *info) 2645{ 2646 return false; 2647} 2648 2649static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 2650 const struct bpf_insn *si, 2651 struct bpf_insn *insn_buf, 2652 struct bpf_prog *prog, 2653 u32 *target_size) 2654{ 2655 return 0; 2656} 2657static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 2658 enum bpf_access_type type, 2659 struct bpf_insn_access_aux *info) 2660{ 2661 return false; 2662} 2663 2664static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 2665 const struct bpf_insn *si, 2666 struct bpf_insn *insn_buf, 2667 struct bpf_prog *prog, 2668 u32 *target_size) 2669{ 2670 return 0; 2671} 2672#endif /* CONFIG_INET */ 2673 2674enum bpf_text_poke_type { 2675 BPF_MOD_CALL, 2676 BPF_MOD_JUMP, 2677}; 2678 2679int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 2680 void *addr1, void *addr2); 2681 2682void *bpf_arch_text_copy(void *dst, void *src, size_t len); 2683int bpf_arch_text_invalidate(void *dst, size_t len); 2684 2685struct btf_id_set; 2686bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 2687 2688#define MAX_BPRINTF_VARARGS 12 2689 2690int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args, 2691 u32 **bin_buf, u32 num_args); 2692void bpf_bprintf_cleanup(void); 2693 2694/* the implementation of the opaque uapi struct bpf_dynptr */ 2695struct bpf_dynptr_kern { 2696 void *data; 2697 /* Size represents the number of usable bytes of dynptr data. 2698 * If for example the offset is at 4 for a local dynptr whose data is 2699 * of type u64, the number of usable bytes is 4. 2700 * 2701 * The upper 8 bits are reserved. It is as follows: 2702 * Bits 0 - 23 = size 2703 * Bits 24 - 30 = dynptr type 2704 * Bit 31 = whether dynptr is read-only 2705 */ 2706 u32 size; 2707 u32 offset; 2708} __aligned(8); 2709 2710enum bpf_dynptr_type { 2711 BPF_DYNPTR_TYPE_INVALID, 2712 /* Points to memory that is local to the bpf program */ 2713 BPF_DYNPTR_TYPE_LOCAL, 2714 /* Underlying data is a kernel-produced ringbuf record */ 2715 BPF_DYNPTR_TYPE_RINGBUF, 2716}; 2717 2718void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 2719 enum bpf_dynptr_type type, u32 offset, u32 size); 2720void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr); 2721int bpf_dynptr_check_size(u32 size); 2722u32 bpf_dynptr_get_size(struct bpf_dynptr_kern *ptr); 2723 2724#ifdef CONFIG_BPF_LSM 2725void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype); 2726void bpf_cgroup_atype_put(int cgroup_atype); 2727#else 2728static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {} 2729static inline void bpf_cgroup_atype_put(int cgroup_atype) {} 2730#endif /* CONFIG_BPF_LSM */ 2731 2732struct key; 2733 2734#ifdef CONFIG_KEYS 2735struct bpf_key { 2736 struct key *key; 2737 bool has_ref; 2738}; 2739#endif /* CONFIG_KEYS */ 2740#endif /* _LINUX_BPF_H */