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 #include <linux/memcontrol.h> 32 #include <linux/cfi.h> 33 #include <asm/rqspinlock.h> 34 35 struct bpf_verifier_env; 36 struct bpf_verifier_log; 37 struct perf_event; 38 struct bpf_prog; 39 struct bpf_prog_aux; 40 struct bpf_map; 41 struct bpf_arena; 42 struct sock; 43 struct seq_file; 44 struct btf; 45 struct btf_type; 46 struct exception_table_entry; 47 struct seq_operations; 48 struct bpf_iter_aux_info; 49 struct bpf_local_storage; 50 struct bpf_local_storage_map; 51 struct kobject; 52 struct mem_cgroup; 53 struct module; 54 struct bpf_func_state; 55 struct ftrace_ops; 56 struct cgroup; 57 struct bpf_token; 58 struct user_namespace; 59 struct super_block; 60 struct inode; 61 62 extern struct idr btf_idr; 63 extern spinlock_t btf_idr_lock; 64 extern struct kobject *btf_kobj; 65 extern struct bpf_mem_alloc bpf_global_ma, bpf_global_percpu_ma; 66 extern bool bpf_global_ma_set; 67 68 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64); 69 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, 70 struct bpf_iter_aux_info *aux); 71 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); 72 typedef unsigned int (*bpf_func_t)(const void *, 73 const struct bpf_insn *); 74 struct bpf_iter_seq_info { 75 const struct seq_operations *seq_ops; 76 bpf_iter_init_seq_priv_t init_seq_private; 77 bpf_iter_fini_seq_priv_t fini_seq_private; 78 u32 seq_priv_size; 79 }; 80 81 /* map is generic key/value storage optionally accessible by eBPF programs */ 82 struct bpf_map_ops { 83 /* funcs callable from userspace (via syscall) */ 84 int (*map_alloc_check)(union bpf_attr *attr); 85 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 86 void (*map_release)(struct bpf_map *map, struct file *map_file); 87 void (*map_free)(struct bpf_map *map); 88 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 89 void (*map_release_uref)(struct bpf_map *map); 90 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 91 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, 92 union bpf_attr __user *uattr); 93 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key, 94 void *value, u64 flags); 95 int (*map_lookup_and_delete_batch)(struct bpf_map *map, 96 const union bpf_attr *attr, 97 union bpf_attr __user *uattr); 98 int (*map_update_batch)(struct bpf_map *map, struct file *map_file, 99 const union bpf_attr *attr, 100 union bpf_attr __user *uattr); 101 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, 102 union bpf_attr __user *uattr); 103 104 /* funcs callable from userspace and from eBPF programs */ 105 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 106 long (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 107 long (*map_delete_elem)(struct bpf_map *map, void *key); 108 long (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 109 long (*map_pop_elem)(struct bpf_map *map, void *value); 110 long (*map_peek_elem)(struct bpf_map *map, void *value); 111 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu); 112 113 /* funcs called by prog_array and perf_event_array map */ 114 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 115 int fd); 116 /* If need_defer is true, the implementation should guarantee that 117 * the to-be-put element is still alive before the bpf program, which 118 * may manipulate it, exists. 119 */ 120 void (*map_fd_put_ptr)(struct bpf_map *map, void *ptr, bool need_defer); 121 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 122 u32 (*map_fd_sys_lookup_elem)(void *ptr); 123 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 124 struct seq_file *m); 125 int (*map_check_btf)(const struct bpf_map *map, 126 const struct btf *btf, 127 const struct btf_type *key_type, 128 const struct btf_type *value_type); 129 130 /* Prog poke tracking helpers. */ 131 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux); 132 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux); 133 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, 134 struct bpf_prog *new); 135 136 /* Direct value access helpers. */ 137 int (*map_direct_value_addr)(const struct bpf_map *map, 138 u64 *imm, u32 off); 139 int (*map_direct_value_meta)(const struct bpf_map *map, 140 u64 imm, u32 *off); 141 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma); 142 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, 143 struct poll_table_struct *pts); 144 unsigned long (*map_get_unmapped_area)(struct file *filep, unsigned long addr, 145 unsigned long len, unsigned long pgoff, 146 unsigned long flags); 147 148 /* Functions called by bpf_local_storage maps */ 149 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, 150 void *owner, u32 size); 151 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, 152 void *owner, u32 size); 153 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner); 154 155 /* Misc helpers.*/ 156 long (*map_redirect)(struct bpf_map *map, u64 key, u64 flags); 157 158 /* map_meta_equal must be implemented for maps that can be 159 * used as an inner map. It is a runtime check to ensure 160 * an inner map can be inserted to an outer map. 161 * 162 * Some properties of the inner map has been used during the 163 * verification time. When inserting an inner map at the runtime, 164 * map_meta_equal has to ensure the inserting map has the same 165 * properties that the verifier has used earlier. 166 */ 167 bool (*map_meta_equal)(const struct bpf_map *meta0, 168 const struct bpf_map *meta1); 169 170 171 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env, 172 struct bpf_func_state *caller, 173 struct bpf_func_state *callee); 174 long (*map_for_each_callback)(struct bpf_map *map, 175 bpf_callback_t callback_fn, 176 void *callback_ctx, u64 flags); 177 178 u64 (*map_mem_usage)(const struct bpf_map *map); 179 180 /* BTF id of struct allocated by map_alloc */ 181 int *map_btf_id; 182 183 /* bpf_iter info used to open a seq_file */ 184 const struct bpf_iter_seq_info *iter_seq_info; 185 }; 186 187 enum { 188 /* Support at most 11 fields in a BTF type */ 189 BTF_FIELDS_MAX = 11, 190 }; 191 192 enum btf_field_type { 193 BPF_SPIN_LOCK = (1 << 0), 194 BPF_TIMER = (1 << 1), 195 BPF_KPTR_UNREF = (1 << 2), 196 BPF_KPTR_REF = (1 << 3), 197 BPF_KPTR_PERCPU = (1 << 4), 198 BPF_KPTR = BPF_KPTR_UNREF | BPF_KPTR_REF | BPF_KPTR_PERCPU, 199 BPF_LIST_HEAD = (1 << 5), 200 BPF_LIST_NODE = (1 << 6), 201 BPF_RB_ROOT = (1 << 7), 202 BPF_RB_NODE = (1 << 8), 203 BPF_GRAPH_NODE = BPF_RB_NODE | BPF_LIST_NODE, 204 BPF_GRAPH_ROOT = BPF_RB_ROOT | BPF_LIST_HEAD, 205 BPF_REFCOUNT = (1 << 9), 206 BPF_WORKQUEUE = (1 << 10), 207 BPF_UPTR = (1 << 11), 208 BPF_RES_SPIN_LOCK = (1 << 12), 209 }; 210 211 typedef void (*btf_dtor_kfunc_t)(void *); 212 213 struct btf_field_kptr { 214 struct btf *btf; 215 struct module *module; 216 /* dtor used if btf_is_kernel(btf), otherwise the type is 217 * program-allocated, dtor is NULL, and __bpf_obj_drop_impl is used 218 */ 219 btf_dtor_kfunc_t dtor; 220 u32 btf_id; 221 }; 222 223 struct btf_field_graph_root { 224 struct btf *btf; 225 u32 value_btf_id; 226 u32 node_offset; 227 struct btf_record *value_rec; 228 }; 229 230 struct btf_field { 231 u32 offset; 232 u32 size; 233 enum btf_field_type type; 234 union { 235 struct btf_field_kptr kptr; 236 struct btf_field_graph_root graph_root; 237 }; 238 }; 239 240 struct btf_record { 241 u32 cnt; 242 u32 field_mask; 243 int spin_lock_off; 244 int res_spin_lock_off; 245 int timer_off; 246 int wq_off; 247 int refcount_off; 248 struct btf_field fields[]; 249 }; 250 251 /* Non-opaque version of bpf_rb_node in uapi/linux/bpf.h */ 252 struct bpf_rb_node_kern { 253 struct rb_node rb_node; 254 void *owner; 255 } __attribute__((aligned(8))); 256 257 /* Non-opaque version of bpf_list_node in uapi/linux/bpf.h */ 258 struct bpf_list_node_kern { 259 struct list_head list_head; 260 void *owner; 261 } __attribute__((aligned(8))); 262 263 struct bpf_map { 264 const struct bpf_map_ops *ops; 265 struct bpf_map *inner_map_meta; 266 #ifdef CONFIG_SECURITY 267 void *security; 268 #endif 269 enum bpf_map_type map_type; 270 u32 key_size; 271 u32 value_size; 272 u32 max_entries; 273 u64 map_extra; /* any per-map-type extra fields */ 274 u32 map_flags; 275 u32 id; 276 struct btf_record *record; 277 int numa_node; 278 u32 btf_key_type_id; 279 u32 btf_value_type_id; 280 u32 btf_vmlinux_value_type_id; 281 struct btf *btf; 282 #ifdef CONFIG_MEMCG 283 struct obj_cgroup *objcg; 284 #endif 285 char name[BPF_OBJ_NAME_LEN]; 286 struct mutex freeze_mutex; 287 atomic64_t refcnt; 288 atomic64_t usercnt; 289 /* rcu is used before freeing and work is only used during freeing */ 290 union { 291 struct work_struct work; 292 struct rcu_head rcu; 293 }; 294 atomic64_t writecnt; 295 /* 'Ownership' of program-containing map is claimed by the first program 296 * that is going to use this map or by the first program which FD is 297 * stored in the map to make sure that all callers and callees have the 298 * same prog type, JITed flag and xdp_has_frags flag. 299 */ 300 struct { 301 const struct btf_type *attach_func_proto; 302 spinlock_t lock; 303 enum bpf_prog_type type; 304 bool jited; 305 bool xdp_has_frags; 306 } owner; 307 bool bypass_spec_v1; 308 bool frozen; /* write-once; write-protected by freeze_mutex */ 309 bool free_after_mult_rcu_gp; 310 bool free_after_rcu_gp; 311 atomic64_t sleepable_refcnt; 312 s64 __percpu *elem_count; 313 }; 314 315 static inline const char *btf_field_type_name(enum btf_field_type type) 316 { 317 switch (type) { 318 case BPF_SPIN_LOCK: 319 return "bpf_spin_lock"; 320 case BPF_RES_SPIN_LOCK: 321 return "bpf_res_spin_lock"; 322 case BPF_TIMER: 323 return "bpf_timer"; 324 case BPF_WORKQUEUE: 325 return "bpf_wq"; 326 case BPF_KPTR_UNREF: 327 case BPF_KPTR_REF: 328 return "kptr"; 329 case BPF_KPTR_PERCPU: 330 return "percpu_kptr"; 331 case BPF_UPTR: 332 return "uptr"; 333 case BPF_LIST_HEAD: 334 return "bpf_list_head"; 335 case BPF_LIST_NODE: 336 return "bpf_list_node"; 337 case BPF_RB_ROOT: 338 return "bpf_rb_root"; 339 case BPF_RB_NODE: 340 return "bpf_rb_node"; 341 case BPF_REFCOUNT: 342 return "bpf_refcount"; 343 default: 344 WARN_ON_ONCE(1); 345 return "unknown"; 346 } 347 } 348 349 static inline u32 btf_field_type_size(enum btf_field_type type) 350 { 351 switch (type) { 352 case BPF_SPIN_LOCK: 353 return sizeof(struct bpf_spin_lock); 354 case BPF_RES_SPIN_LOCK: 355 return sizeof(struct bpf_res_spin_lock); 356 case BPF_TIMER: 357 return sizeof(struct bpf_timer); 358 case BPF_WORKQUEUE: 359 return sizeof(struct bpf_wq); 360 case BPF_KPTR_UNREF: 361 case BPF_KPTR_REF: 362 case BPF_KPTR_PERCPU: 363 case BPF_UPTR: 364 return sizeof(u64); 365 case BPF_LIST_HEAD: 366 return sizeof(struct bpf_list_head); 367 case BPF_LIST_NODE: 368 return sizeof(struct bpf_list_node); 369 case BPF_RB_ROOT: 370 return sizeof(struct bpf_rb_root); 371 case BPF_RB_NODE: 372 return sizeof(struct bpf_rb_node); 373 case BPF_REFCOUNT: 374 return sizeof(struct bpf_refcount); 375 default: 376 WARN_ON_ONCE(1); 377 return 0; 378 } 379 } 380 381 static inline u32 btf_field_type_align(enum btf_field_type type) 382 { 383 switch (type) { 384 case BPF_SPIN_LOCK: 385 return __alignof__(struct bpf_spin_lock); 386 case BPF_RES_SPIN_LOCK: 387 return __alignof__(struct bpf_res_spin_lock); 388 case BPF_TIMER: 389 return __alignof__(struct bpf_timer); 390 case BPF_WORKQUEUE: 391 return __alignof__(struct bpf_wq); 392 case BPF_KPTR_UNREF: 393 case BPF_KPTR_REF: 394 case BPF_KPTR_PERCPU: 395 case BPF_UPTR: 396 return __alignof__(u64); 397 case BPF_LIST_HEAD: 398 return __alignof__(struct bpf_list_head); 399 case BPF_LIST_NODE: 400 return __alignof__(struct bpf_list_node); 401 case BPF_RB_ROOT: 402 return __alignof__(struct bpf_rb_root); 403 case BPF_RB_NODE: 404 return __alignof__(struct bpf_rb_node); 405 case BPF_REFCOUNT: 406 return __alignof__(struct bpf_refcount); 407 default: 408 WARN_ON_ONCE(1); 409 return 0; 410 } 411 } 412 413 static inline void bpf_obj_init_field(const struct btf_field *field, void *addr) 414 { 415 memset(addr, 0, field->size); 416 417 switch (field->type) { 418 case BPF_REFCOUNT: 419 refcount_set((refcount_t *)addr, 1); 420 break; 421 case BPF_RB_NODE: 422 RB_CLEAR_NODE((struct rb_node *)addr); 423 break; 424 case BPF_LIST_HEAD: 425 case BPF_LIST_NODE: 426 INIT_LIST_HEAD((struct list_head *)addr); 427 break; 428 case BPF_RB_ROOT: 429 /* RB_ROOT_CACHED 0-inits, no need to do anything after memset */ 430 case BPF_SPIN_LOCK: 431 case BPF_RES_SPIN_LOCK: 432 case BPF_TIMER: 433 case BPF_WORKQUEUE: 434 case BPF_KPTR_UNREF: 435 case BPF_KPTR_REF: 436 case BPF_KPTR_PERCPU: 437 case BPF_UPTR: 438 break; 439 default: 440 WARN_ON_ONCE(1); 441 return; 442 } 443 } 444 445 static inline bool btf_record_has_field(const struct btf_record *rec, enum btf_field_type type) 446 { 447 if (IS_ERR_OR_NULL(rec)) 448 return false; 449 return rec->field_mask & type; 450 } 451 452 static inline void bpf_obj_init(const struct btf_record *rec, void *obj) 453 { 454 int i; 455 456 if (IS_ERR_OR_NULL(rec)) 457 return; 458 for (i = 0; i < rec->cnt; i++) 459 bpf_obj_init_field(&rec->fields[i], obj + rec->fields[i].offset); 460 } 461 462 /* 'dst' must be a temporary buffer and should not point to memory that is being 463 * used in parallel by a bpf program or bpf syscall, otherwise the access from 464 * the bpf program or bpf syscall may be corrupted by the reinitialization, 465 * leading to weird problems. Even 'dst' is newly-allocated from bpf memory 466 * allocator, it is still possible for 'dst' to be used in parallel by a bpf 467 * program or bpf syscall. 468 */ 469 static inline void check_and_init_map_value(struct bpf_map *map, void *dst) 470 { 471 bpf_obj_init(map->record, dst); 472 } 473 474 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 475 * forced to use 'long' read/writes to try to atomically copy long counters. 476 * Best-effort only. No barriers here, since it _will_ race with concurrent 477 * updates from BPF programs. Called from bpf syscall and mostly used with 478 * size 8 or 16 bytes, so ask compiler to inline it. 479 */ 480 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 481 { 482 const long *lsrc = src; 483 long *ldst = dst; 484 485 size /= sizeof(long); 486 while (size--) 487 data_race(*ldst++ = *lsrc++); 488 } 489 490 /* copy everything but bpf_spin_lock, bpf_timer, and kptrs. There could be one of each. */ 491 static inline void bpf_obj_memcpy(struct btf_record *rec, 492 void *dst, void *src, u32 size, 493 bool long_memcpy) 494 { 495 u32 curr_off = 0; 496 int i; 497 498 if (IS_ERR_OR_NULL(rec)) { 499 if (long_memcpy) 500 bpf_long_memcpy(dst, src, round_up(size, 8)); 501 else 502 memcpy(dst, src, size); 503 return; 504 } 505 506 for (i = 0; i < rec->cnt; i++) { 507 u32 next_off = rec->fields[i].offset; 508 u32 sz = next_off - curr_off; 509 510 memcpy(dst + curr_off, src + curr_off, sz); 511 curr_off += rec->fields[i].size + sz; 512 } 513 memcpy(dst + curr_off, src + curr_off, size - curr_off); 514 } 515 516 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 517 { 518 bpf_obj_memcpy(map->record, dst, src, map->value_size, false); 519 } 520 521 static inline void copy_map_value_long(struct bpf_map *map, void *dst, void *src) 522 { 523 bpf_obj_memcpy(map->record, dst, src, map->value_size, true); 524 } 525 526 static inline void bpf_obj_swap_uptrs(const struct btf_record *rec, void *dst, void *src) 527 { 528 unsigned long *src_uptr, *dst_uptr; 529 const struct btf_field *field; 530 int i; 531 532 if (!btf_record_has_field(rec, BPF_UPTR)) 533 return; 534 535 for (i = 0, field = rec->fields; i < rec->cnt; i++, field++) { 536 if (field->type != BPF_UPTR) 537 continue; 538 539 src_uptr = src + field->offset; 540 dst_uptr = dst + field->offset; 541 swap(*src_uptr, *dst_uptr); 542 } 543 } 544 545 static inline void bpf_obj_memzero(struct btf_record *rec, void *dst, u32 size) 546 { 547 u32 curr_off = 0; 548 int i; 549 550 if (IS_ERR_OR_NULL(rec)) { 551 memset(dst, 0, size); 552 return; 553 } 554 555 for (i = 0; i < rec->cnt; i++) { 556 u32 next_off = rec->fields[i].offset; 557 u32 sz = next_off - curr_off; 558 559 memset(dst + curr_off, 0, sz); 560 curr_off += rec->fields[i].size + sz; 561 } 562 memset(dst + curr_off, 0, size - curr_off); 563 } 564 565 static inline void zero_map_value(struct bpf_map *map, void *dst) 566 { 567 bpf_obj_memzero(map->record, dst, map->value_size); 568 } 569 570 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 571 bool lock_src); 572 void bpf_timer_cancel_and_free(void *timer); 573 void bpf_wq_cancel_and_free(void *timer); 574 void bpf_list_head_free(const struct btf_field *field, void *list_head, 575 struct bpf_spin_lock *spin_lock); 576 void bpf_rb_root_free(const struct btf_field *field, void *rb_root, 577 struct bpf_spin_lock *spin_lock); 578 u64 bpf_arena_get_kern_vm_start(struct bpf_arena *arena); 579 u64 bpf_arena_get_user_vm_start(struct bpf_arena *arena); 580 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size); 581 582 struct bpf_offload_dev; 583 struct bpf_offloaded_map; 584 585 struct bpf_map_dev_ops { 586 int (*map_get_next_key)(struct bpf_offloaded_map *map, 587 void *key, void *next_key); 588 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 589 void *key, void *value); 590 int (*map_update_elem)(struct bpf_offloaded_map *map, 591 void *key, void *value, u64 flags); 592 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 593 }; 594 595 struct bpf_offloaded_map { 596 struct bpf_map map; 597 struct net_device *netdev; 598 const struct bpf_map_dev_ops *dev_ops; 599 void *dev_priv; 600 struct list_head offloads; 601 }; 602 603 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 604 { 605 return container_of(map, struct bpf_offloaded_map, map); 606 } 607 608 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 609 { 610 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 611 } 612 613 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 614 { 615 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && 616 map->ops->map_seq_show_elem; 617 } 618 619 int map_check_no_btf(const struct bpf_map *map, 620 const struct btf *btf, 621 const struct btf_type *key_type, 622 const struct btf_type *value_type); 623 624 bool bpf_map_meta_equal(const struct bpf_map *meta0, 625 const struct bpf_map *meta1); 626 627 extern const struct bpf_map_ops bpf_map_offload_ops; 628 629 /* bpf_type_flag contains a set of flags that are applicable to the values of 630 * arg_type, ret_type and reg_type. For example, a pointer value may be null, 631 * or a memory is read-only. We classify types into two categories: base types 632 * and extended types. Extended types are base types combined with a type flag. 633 * 634 * Currently there are no more than 32 base types in arg_type, ret_type and 635 * reg_types. 636 */ 637 #define BPF_BASE_TYPE_BITS 8 638 639 enum bpf_type_flag { 640 /* PTR may be NULL. */ 641 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS), 642 643 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is 644 * compatible with both mutable and immutable memory. 645 */ 646 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS), 647 648 /* MEM points to BPF ring buffer reservation. */ 649 MEM_RINGBUF = BIT(2 + BPF_BASE_TYPE_BITS), 650 651 /* MEM is in user address space. */ 652 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS), 653 654 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged 655 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In 656 * order to drop this tag, it must be passed into bpf_per_cpu_ptr() 657 * or bpf_this_cpu_ptr(), which will return the pointer corresponding 658 * to the specified cpu. 659 */ 660 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS), 661 662 /* Indicates that the argument will be released. */ 663 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS), 664 665 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark 666 * unreferenced and referenced kptr loaded from map value using a load 667 * instruction, so that they can only be dereferenced but not escape the 668 * BPF program into the kernel (i.e. cannot be passed as arguments to 669 * kfunc or bpf helpers). 670 */ 671 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS), 672 673 /* MEM can be uninitialized. */ 674 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS), 675 676 /* DYNPTR points to memory local to the bpf program. */ 677 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS), 678 679 /* DYNPTR points to a kernel-produced ringbuf record. */ 680 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS), 681 682 /* Size is known at compile time. */ 683 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS), 684 685 /* MEM is of an allocated object of type in program BTF. This is used to 686 * tag PTR_TO_BTF_ID allocated using bpf_obj_new. 687 */ 688 MEM_ALLOC = BIT(11 + BPF_BASE_TYPE_BITS), 689 690 /* PTR was passed from the kernel in a trusted context, and may be 691 * passed to KF_TRUSTED_ARGS kfuncs or BPF helper functions. 692 * Confusingly, this is _not_ the opposite of PTR_UNTRUSTED above. 693 * PTR_UNTRUSTED refers to a kptr that was read directly from a map 694 * without invoking bpf_kptr_xchg(). What we really need to know is 695 * whether a pointer is safe to pass to a kfunc or BPF helper function. 696 * While PTR_UNTRUSTED pointers are unsafe to pass to kfuncs and BPF 697 * helpers, they do not cover all possible instances of unsafe 698 * pointers. For example, a pointer that was obtained from walking a 699 * struct will _not_ get the PTR_UNTRUSTED type modifier, despite the 700 * fact that it may be NULL, invalid, etc. This is due to backwards 701 * compatibility requirements, as this was the behavior that was first 702 * introduced when kptrs were added. The behavior is now considered 703 * deprecated, and PTR_UNTRUSTED will eventually be removed. 704 * 705 * PTR_TRUSTED, on the other hand, is a pointer that the kernel 706 * guarantees to be valid and safe to pass to kfuncs and BPF helpers. 707 * For example, pointers passed to tracepoint arguments are considered 708 * PTR_TRUSTED, as are pointers that are passed to struct_ops 709 * callbacks. As alluded to above, pointers that are obtained from 710 * walking PTR_TRUSTED pointers are _not_ trusted. For example, if a 711 * struct task_struct *task is PTR_TRUSTED, then accessing 712 * task->last_wakee will lose the PTR_TRUSTED modifier when it's stored 713 * in a BPF register. Similarly, pointers passed to certain programs 714 * types such as kretprobes are not guaranteed to be valid, as they may 715 * for example contain an object that was recently freed. 716 */ 717 PTR_TRUSTED = BIT(12 + BPF_BASE_TYPE_BITS), 718 719 /* MEM is tagged with rcu and memory access needs rcu_read_lock protection. */ 720 MEM_RCU = BIT(13 + BPF_BASE_TYPE_BITS), 721 722 /* Used to tag PTR_TO_BTF_ID | MEM_ALLOC references which are non-owning. 723 * Currently only valid for linked-list and rbtree nodes. If the nodes 724 * have a bpf_refcount_field, they must be tagged MEM_RCU as well. 725 */ 726 NON_OWN_REF = BIT(14 + BPF_BASE_TYPE_BITS), 727 728 /* DYNPTR points to sk_buff */ 729 DYNPTR_TYPE_SKB = BIT(15 + BPF_BASE_TYPE_BITS), 730 731 /* DYNPTR points to xdp_buff */ 732 DYNPTR_TYPE_XDP = BIT(16 + BPF_BASE_TYPE_BITS), 733 734 /* Memory must be aligned on some architectures, used in combination with 735 * MEM_FIXED_SIZE. 736 */ 737 MEM_ALIGNED = BIT(17 + BPF_BASE_TYPE_BITS), 738 739 /* MEM is being written to, often combined with MEM_UNINIT. Non-presence 740 * of MEM_WRITE means that MEM is only being read. MEM_WRITE without the 741 * MEM_UNINIT means that memory needs to be initialized since it is also 742 * read. 743 */ 744 MEM_WRITE = BIT(18 + BPF_BASE_TYPE_BITS), 745 746 __BPF_TYPE_FLAG_MAX, 747 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1, 748 }; 749 750 #define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF | DYNPTR_TYPE_SKB \ 751 | DYNPTR_TYPE_XDP) 752 753 /* Max number of base types. */ 754 #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS) 755 756 /* Max number of all types. */ 757 #define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1)) 758 759 /* function argument constraints */ 760 enum bpf_arg_type { 761 ARG_DONTCARE = 0, /* unused argument in helper function */ 762 763 /* the following constraints used to prototype 764 * bpf_map_lookup/update/delete_elem() functions 765 */ 766 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 767 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 768 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 769 770 /* Used to prototype bpf_memcmp() and other functions that access data 771 * on eBPF program stack 772 */ 773 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 774 ARG_PTR_TO_ARENA, 775 776 ARG_CONST_SIZE, /* number of bytes accessed from memory */ 777 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 778 779 ARG_PTR_TO_CTX, /* pointer to context */ 780 ARG_ANYTHING, /* any (initialized) argument is ok */ 781 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 782 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 783 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 784 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */ 785 ARG_PTR_TO_RINGBUF_MEM, /* pointer to dynamically reserved ringbuf memory */ 786 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ 787 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ 788 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ 789 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */ 790 ARG_PTR_TO_STACK, /* pointer to stack */ 791 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */ 792 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */ 793 ARG_KPTR_XCHG_DEST, /* pointer to destination that kptrs are bpf_kptr_xchg'd into */ 794 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */ 795 __BPF_ARG_TYPE_MAX, 796 797 /* Extended arg_types. */ 798 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE, 799 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM, 800 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX, 801 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET, 802 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK, 803 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID, 804 /* Pointer to memory does not need to be initialized, since helper function 805 * fills all bytes or clears them in error case. 806 */ 807 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | MEM_WRITE | ARG_PTR_TO_MEM, 808 /* Pointer to valid memory of size known at compile time. */ 809 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM, 810 811 /* This must be the last entry. Its purpose is to ensure the enum is 812 * wide enough to hold the higher bits reserved for bpf_type_flag. 813 */ 814 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT, 815 }; 816 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 817 818 /* type of values returned from helper functions */ 819 enum bpf_return_type { 820 RET_INTEGER, /* function returns integer */ 821 RET_VOID, /* function doesn't return anything */ 822 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 823 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */ 824 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */ 825 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */ 826 RET_PTR_TO_MEM, /* returns a pointer to memory */ 827 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ 828 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */ 829 __BPF_RET_TYPE_MAX, 830 831 /* Extended ret_types. */ 832 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE, 833 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET, 834 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK, 835 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON, 836 RET_PTR_TO_RINGBUF_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_RINGBUF | RET_PTR_TO_MEM, 837 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MEM, 838 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID, 839 RET_PTR_TO_BTF_ID_TRUSTED = PTR_TRUSTED | RET_PTR_TO_BTF_ID, 840 841 /* This must be the last entry. Its purpose is to ensure the enum is 842 * wide enough to hold the higher bits reserved for bpf_type_flag. 843 */ 844 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT, 845 }; 846 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 847 848 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 849 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 850 * instructions after verifying 851 */ 852 struct bpf_func_proto { 853 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 854 bool gpl_only; 855 bool pkt_access; 856 bool might_sleep; 857 /* set to true if helper follows contract for llvm 858 * attribute bpf_fastcall: 859 * - void functions do not scratch r0 860 * - functions taking N arguments scratch only registers r1-rN 861 */ 862 bool allow_fastcall; 863 enum bpf_return_type ret_type; 864 union { 865 struct { 866 enum bpf_arg_type arg1_type; 867 enum bpf_arg_type arg2_type; 868 enum bpf_arg_type arg3_type; 869 enum bpf_arg_type arg4_type; 870 enum bpf_arg_type arg5_type; 871 }; 872 enum bpf_arg_type arg_type[5]; 873 }; 874 union { 875 struct { 876 u32 *arg1_btf_id; 877 u32 *arg2_btf_id; 878 u32 *arg3_btf_id; 879 u32 *arg4_btf_id; 880 u32 *arg5_btf_id; 881 }; 882 u32 *arg_btf_id[5]; 883 struct { 884 size_t arg1_size; 885 size_t arg2_size; 886 size_t arg3_size; 887 size_t arg4_size; 888 size_t arg5_size; 889 }; 890 size_t arg_size[5]; 891 }; 892 int *ret_btf_id; /* return value btf_id */ 893 bool (*allowed)(const struct bpf_prog *prog); 894 }; 895 896 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 897 * the first argument to eBPF programs. 898 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 899 */ 900 struct bpf_context; 901 902 enum bpf_access_type { 903 BPF_READ = 1, 904 BPF_WRITE = 2 905 }; 906 907 /* types of values stored in eBPF registers */ 908 /* Pointer types represent: 909 * pointer 910 * pointer + imm 911 * pointer + (u16) var 912 * pointer + (u16) var + imm 913 * if (range > 0) then [ptr, ptr + range - off) is safe to access 914 * if (id > 0) means that some 'var' was added 915 * if (off > 0) means that 'imm' was added 916 */ 917 enum bpf_reg_type { 918 NOT_INIT = 0, /* nothing was written into register */ 919 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 920 PTR_TO_CTX, /* reg points to bpf_context */ 921 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 922 PTR_TO_MAP_VALUE, /* reg points to map element value */ 923 PTR_TO_MAP_KEY, /* reg points to a map element key */ 924 PTR_TO_STACK, /* reg == frame_pointer + offset */ 925 PTR_TO_PACKET_META, /* skb->data - meta_len */ 926 PTR_TO_PACKET, /* reg points to skb->data */ 927 PTR_TO_PACKET_END, /* skb->data + headlen */ 928 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 929 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 930 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 931 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 932 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 933 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 934 /* PTR_TO_BTF_ID points to a kernel struct that does not need 935 * to be null checked by the BPF program. This does not imply the 936 * pointer is _not_ null and in practice this can easily be a null 937 * pointer when reading pointer chains. The assumption is program 938 * context will handle null pointer dereference typically via fault 939 * handling. The verifier must keep this in mind and can make no 940 * assumptions about null or non-null when doing branch analysis. 941 * Further, when passed into helpers the helpers can not, without 942 * additional context, assume the value is non-null. 943 */ 944 PTR_TO_BTF_ID, 945 PTR_TO_MEM, /* reg points to valid memory region */ 946 PTR_TO_ARENA, 947 PTR_TO_BUF, /* reg points to a read/write buffer */ 948 PTR_TO_FUNC, /* reg points to a bpf program function */ 949 CONST_PTR_TO_DYNPTR, /* reg points to a const struct bpf_dynptr */ 950 __BPF_REG_TYPE_MAX, 951 952 /* Extended reg_types. */ 953 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE, 954 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET, 955 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON, 956 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK, 957 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not 958 * been checked for null. Used primarily to inform the verifier 959 * an explicit null check is required for this struct. 960 */ 961 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID, 962 963 /* This must be the last entry. Its purpose is to ensure the enum is 964 * wide enough to hold the higher bits reserved for bpf_type_flag. 965 */ 966 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT, 967 }; 968 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 969 970 /* The information passed from prog-specific *_is_valid_access 971 * back to the verifier. 972 */ 973 struct bpf_insn_access_aux { 974 enum bpf_reg_type reg_type; 975 bool is_ldsx; 976 union { 977 int ctx_field_size; 978 struct { 979 struct btf *btf; 980 u32 btf_id; 981 u32 ref_obj_id; 982 }; 983 }; 984 struct bpf_verifier_log *log; /* for verbose logs */ 985 bool is_retval; /* is accessing function return value ? */ 986 }; 987 988 static inline void 989 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 990 { 991 aux->ctx_field_size = size; 992 } 993 994 static bool bpf_is_ldimm64(const struct bpf_insn *insn) 995 { 996 return insn->code == (BPF_LD | BPF_IMM | BPF_DW); 997 } 998 999 static inline bool bpf_pseudo_func(const struct bpf_insn *insn) 1000 { 1001 return bpf_is_ldimm64(insn) && insn->src_reg == BPF_PSEUDO_FUNC; 1002 } 1003 1004 /* Given a BPF_ATOMIC instruction @atomic_insn, return true if it is an 1005 * atomic load or store, and false if it is a read-modify-write instruction. 1006 */ 1007 static inline bool 1008 bpf_atomic_is_load_store(const struct bpf_insn *atomic_insn) 1009 { 1010 switch (atomic_insn->imm) { 1011 case BPF_LOAD_ACQ: 1012 case BPF_STORE_REL: 1013 return true; 1014 default: 1015 return false; 1016 } 1017 } 1018 1019 struct bpf_prog_ops { 1020 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 1021 union bpf_attr __user *uattr); 1022 }; 1023 1024 struct bpf_reg_state; 1025 struct bpf_verifier_ops { 1026 /* return eBPF function prototype for verification */ 1027 const struct bpf_func_proto * 1028 (*get_func_proto)(enum bpf_func_id func_id, 1029 const struct bpf_prog *prog); 1030 1031 /* return true if 'size' wide access at offset 'off' within bpf_context 1032 * with 'type' (read or write) is allowed 1033 */ 1034 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 1035 const struct bpf_prog *prog, 1036 struct bpf_insn_access_aux *info); 1037 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 1038 const struct bpf_prog *prog); 1039 int (*gen_epilogue)(struct bpf_insn *insn, const struct bpf_prog *prog, 1040 s16 ctx_stack_off); 1041 int (*gen_ld_abs)(const struct bpf_insn *orig, 1042 struct bpf_insn *insn_buf); 1043 u32 (*convert_ctx_access)(enum bpf_access_type type, 1044 const struct bpf_insn *src, 1045 struct bpf_insn *dst, 1046 struct bpf_prog *prog, u32 *target_size); 1047 int (*btf_struct_access)(struct bpf_verifier_log *log, 1048 const struct bpf_reg_state *reg, 1049 int off, int size); 1050 }; 1051 1052 struct bpf_prog_offload_ops { 1053 /* verifier basic callbacks */ 1054 int (*insn_hook)(struct bpf_verifier_env *env, 1055 int insn_idx, int prev_insn_idx); 1056 int (*finalize)(struct bpf_verifier_env *env); 1057 /* verifier optimization callbacks (called after .finalize) */ 1058 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 1059 struct bpf_insn *insn); 1060 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 1061 /* program management callbacks */ 1062 int (*prepare)(struct bpf_prog *prog); 1063 int (*translate)(struct bpf_prog *prog); 1064 void (*destroy)(struct bpf_prog *prog); 1065 }; 1066 1067 struct bpf_prog_offload { 1068 struct bpf_prog *prog; 1069 struct net_device *netdev; 1070 struct bpf_offload_dev *offdev; 1071 void *dev_priv; 1072 struct list_head offloads; 1073 bool dev_state; 1074 bool opt_failed; 1075 void *jited_image; 1076 u32 jited_len; 1077 }; 1078 1079 enum bpf_cgroup_storage_type { 1080 BPF_CGROUP_STORAGE_SHARED, 1081 BPF_CGROUP_STORAGE_PERCPU, 1082 __BPF_CGROUP_STORAGE_MAX 1083 }; 1084 1085 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 1086 1087 /* The longest tracepoint has 12 args. 1088 * See include/trace/bpf_probe.h 1089 */ 1090 #define MAX_BPF_FUNC_ARGS 12 1091 1092 /* The maximum number of arguments passed through registers 1093 * a single function may have. 1094 */ 1095 #define MAX_BPF_FUNC_REG_ARGS 5 1096 1097 /* The argument is a structure. */ 1098 #define BTF_FMODEL_STRUCT_ARG BIT(0) 1099 1100 /* The argument is signed. */ 1101 #define BTF_FMODEL_SIGNED_ARG BIT(1) 1102 1103 struct btf_func_model { 1104 u8 ret_size; 1105 u8 ret_flags; 1106 u8 nr_args; 1107 u8 arg_size[MAX_BPF_FUNC_ARGS]; 1108 u8 arg_flags[MAX_BPF_FUNC_ARGS]; 1109 }; 1110 1111 /* Restore arguments before returning from trampoline to let original function 1112 * continue executing. This flag is used for fentry progs when there are no 1113 * fexit progs. 1114 */ 1115 #define BPF_TRAMP_F_RESTORE_REGS BIT(0) 1116 /* Call original function after fentry progs, but before fexit progs. 1117 * Makes sense for fentry/fexit, normal calls and indirect calls. 1118 */ 1119 #define BPF_TRAMP_F_CALL_ORIG BIT(1) 1120 /* Skip current frame and return to parent. Makes sense for fentry/fexit 1121 * programs only. Should not be used with normal calls and indirect calls. 1122 */ 1123 #define BPF_TRAMP_F_SKIP_FRAME BIT(2) 1124 /* Store IP address of the caller on the trampoline stack, 1125 * so it's available for trampoline's programs. 1126 */ 1127 #define BPF_TRAMP_F_IP_ARG BIT(3) 1128 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */ 1129 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4) 1130 1131 /* Get original function from stack instead of from provided direct address. 1132 * Makes sense for trampolines with fexit or fmod_ret programs. 1133 */ 1134 #define BPF_TRAMP_F_ORIG_STACK BIT(5) 1135 1136 /* This trampoline is on a function with another ftrace_ops with IPMODIFY, 1137 * e.g., a live patch. This flag is set and cleared by ftrace call backs, 1138 */ 1139 #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6) 1140 1141 /* Indicate that current trampoline is in a tail call context. Then, it has to 1142 * cache and restore tail_call_cnt to avoid infinite tail call loop. 1143 */ 1144 #define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7) 1145 1146 /* 1147 * Indicate the trampoline should be suitable to receive indirect calls; 1148 * without this indirectly calling the generated code can result in #UD/#CP, 1149 * depending on the CFI options. 1150 * 1151 * Used by bpf_struct_ops. 1152 * 1153 * Incompatible with FENTRY usage, overloads @func_addr argument. 1154 */ 1155 #define BPF_TRAMP_F_INDIRECT BIT(8) 1156 1157 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 1158 * bytes on x86. 1159 */ 1160 enum { 1161 #if defined(__s390x__) 1162 BPF_MAX_TRAMP_LINKS = 27, 1163 #else 1164 BPF_MAX_TRAMP_LINKS = 38, 1165 #endif 1166 }; 1167 1168 struct bpf_tramp_links { 1169 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS]; 1170 int nr_links; 1171 }; 1172 1173 struct bpf_tramp_run_ctx; 1174 1175 /* Different use cases for BPF trampoline: 1176 * 1. replace nop at the function entry (kprobe equivalent) 1177 * flags = BPF_TRAMP_F_RESTORE_REGS 1178 * fentry = a set of programs to run before returning from trampoline 1179 * 1180 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 1181 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 1182 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 1183 * fentry = a set of program to run before calling original function 1184 * fexit = a set of program to run after original function 1185 * 1186 * 3. replace direct call instruction anywhere in the function body 1187 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 1188 * With flags = 0 1189 * fentry = a set of programs to run before returning from trampoline 1190 * With flags = BPF_TRAMP_F_CALL_ORIG 1191 * orig_call = original callback addr or direct function addr 1192 * fentry = a set of program to run before calling original function 1193 * fexit = a set of program to run after original function 1194 */ 1195 struct bpf_tramp_image; 1196 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *image_end, 1197 const struct btf_func_model *m, u32 flags, 1198 struct bpf_tramp_links *tlinks, 1199 void *func_addr); 1200 void *arch_alloc_bpf_trampoline(unsigned int size); 1201 void arch_free_bpf_trampoline(void *image, unsigned int size); 1202 int __must_check arch_protect_bpf_trampoline(void *image, unsigned int size); 1203 int arch_bpf_trampoline_size(const struct btf_func_model *m, u32 flags, 1204 struct bpf_tramp_links *tlinks, void *func_addr); 1205 1206 u64 notrace __bpf_prog_enter_sleepable_recur(struct bpf_prog *prog, 1207 struct bpf_tramp_run_ctx *run_ctx); 1208 void notrace __bpf_prog_exit_sleepable_recur(struct bpf_prog *prog, u64 start, 1209 struct bpf_tramp_run_ctx *run_ctx); 1210 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr); 1211 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr); 1212 typedef u64 (*bpf_trampoline_enter_t)(struct bpf_prog *prog, 1213 struct bpf_tramp_run_ctx *run_ctx); 1214 typedef void (*bpf_trampoline_exit_t)(struct bpf_prog *prog, u64 start, 1215 struct bpf_tramp_run_ctx *run_ctx); 1216 bpf_trampoline_enter_t bpf_trampoline_enter(const struct bpf_prog *prog); 1217 bpf_trampoline_exit_t bpf_trampoline_exit(const struct bpf_prog *prog); 1218 1219 struct bpf_ksym { 1220 unsigned long start; 1221 unsigned long end; 1222 char name[KSYM_NAME_LEN]; 1223 struct list_head lnode; 1224 struct latch_tree_node tnode; 1225 bool prog; 1226 }; 1227 1228 enum bpf_tramp_prog_type { 1229 BPF_TRAMP_FENTRY, 1230 BPF_TRAMP_FEXIT, 1231 BPF_TRAMP_MODIFY_RETURN, 1232 BPF_TRAMP_MAX, 1233 BPF_TRAMP_REPLACE, /* more than MAX */ 1234 }; 1235 1236 struct bpf_tramp_image { 1237 void *image; 1238 int size; 1239 struct bpf_ksym ksym; 1240 struct percpu_ref pcref; 1241 void *ip_after_call; 1242 void *ip_epilogue; 1243 union { 1244 struct rcu_head rcu; 1245 struct work_struct work; 1246 }; 1247 }; 1248 1249 struct bpf_trampoline { 1250 /* hlist for trampoline_table */ 1251 struct hlist_node hlist; 1252 struct ftrace_ops *fops; 1253 /* serializes access to fields of this trampoline */ 1254 struct mutex mutex; 1255 refcount_t refcnt; 1256 u32 flags; 1257 u64 key; 1258 struct { 1259 struct btf_func_model model; 1260 void *addr; 1261 bool ftrace_managed; 1262 } func; 1263 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 1264 * program by replacing one of its functions. func.addr is the address 1265 * of the function it replaced. 1266 */ 1267 struct bpf_prog *extension_prog; 1268 /* list of BPF programs using this trampoline */ 1269 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 1270 /* Number of attached programs. A counter per kind. */ 1271 int progs_cnt[BPF_TRAMP_MAX]; 1272 /* Executable image of trampoline */ 1273 struct bpf_tramp_image *cur_image; 1274 }; 1275 1276 struct bpf_attach_target_info { 1277 struct btf_func_model fmodel; 1278 long tgt_addr; 1279 struct module *tgt_mod; 1280 const char *tgt_name; 1281 const struct btf_type *tgt_type; 1282 }; 1283 1284 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 1285 1286 struct bpf_dispatcher_prog { 1287 struct bpf_prog *prog; 1288 refcount_t users; 1289 }; 1290 1291 struct bpf_dispatcher { 1292 /* dispatcher mutex */ 1293 struct mutex mutex; 1294 void *func; 1295 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 1296 int num_progs; 1297 void *image; 1298 void *rw_image; 1299 u32 image_off; 1300 struct bpf_ksym ksym; 1301 #ifdef CONFIG_HAVE_STATIC_CALL 1302 struct static_call_key *sc_key; 1303 void *sc_tramp; 1304 #endif 1305 }; 1306 1307 #ifndef __bpfcall 1308 #define __bpfcall __nocfi 1309 #endif 1310 1311 static __always_inline __bpfcall unsigned int bpf_dispatcher_nop_func( 1312 const void *ctx, 1313 const struct bpf_insn *insnsi, 1314 bpf_func_t bpf_func) 1315 { 1316 return bpf_func(ctx, insnsi); 1317 } 1318 1319 /* the implementation of the opaque uapi struct bpf_dynptr */ 1320 struct bpf_dynptr_kern { 1321 void *data; 1322 /* Size represents the number of usable bytes of dynptr data. 1323 * If for example the offset is at 4 for a local dynptr whose data is 1324 * of type u64, the number of usable bytes is 4. 1325 * 1326 * The upper 8 bits are reserved. It is as follows: 1327 * Bits 0 - 23 = size 1328 * Bits 24 - 30 = dynptr type 1329 * Bit 31 = whether dynptr is read-only 1330 */ 1331 u32 size; 1332 u32 offset; 1333 } __aligned(8); 1334 1335 enum bpf_dynptr_type { 1336 BPF_DYNPTR_TYPE_INVALID, 1337 /* Points to memory that is local to the bpf program */ 1338 BPF_DYNPTR_TYPE_LOCAL, 1339 /* Underlying data is a ringbuf record */ 1340 BPF_DYNPTR_TYPE_RINGBUF, 1341 /* Underlying data is a sk_buff */ 1342 BPF_DYNPTR_TYPE_SKB, 1343 /* Underlying data is a xdp_buff */ 1344 BPF_DYNPTR_TYPE_XDP, 1345 }; 1346 1347 int bpf_dynptr_check_size(u32 size); 1348 u32 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr); 1349 const void *__bpf_dynptr_data(const struct bpf_dynptr_kern *ptr, u32 len); 1350 void *__bpf_dynptr_data_rw(const struct bpf_dynptr_kern *ptr, u32 len); 1351 bool __bpf_dynptr_is_rdonly(const struct bpf_dynptr_kern *ptr); 1352 1353 #ifdef CONFIG_BPF_JIT 1354 int bpf_trampoline_link_prog(struct bpf_tramp_link *link, 1355 struct bpf_trampoline *tr, 1356 struct bpf_prog *tgt_prog); 1357 int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, 1358 struct bpf_trampoline *tr, 1359 struct bpf_prog *tgt_prog); 1360 struct bpf_trampoline *bpf_trampoline_get(u64 key, 1361 struct bpf_attach_target_info *tgt_info); 1362 void bpf_trampoline_put(struct bpf_trampoline *tr); 1363 int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_funcs); 1364 1365 /* 1366 * When the architecture supports STATIC_CALL replace the bpf_dispatcher_fn 1367 * indirection with a direct call to the bpf program. If the architecture does 1368 * not have STATIC_CALL, avoid a double-indirection. 1369 */ 1370 #ifdef CONFIG_HAVE_STATIC_CALL 1371 1372 #define __BPF_DISPATCHER_SC_INIT(_name) \ 1373 .sc_key = &STATIC_CALL_KEY(_name), \ 1374 .sc_tramp = STATIC_CALL_TRAMP_ADDR(_name), 1375 1376 #define __BPF_DISPATCHER_SC(name) \ 1377 DEFINE_STATIC_CALL(bpf_dispatcher_##name##_call, bpf_dispatcher_nop_func) 1378 1379 #define __BPF_DISPATCHER_CALL(name) \ 1380 static_call(bpf_dispatcher_##name##_call)(ctx, insnsi, bpf_func) 1381 1382 #define __BPF_DISPATCHER_UPDATE(_d, _new) \ 1383 __static_call_update((_d)->sc_key, (_d)->sc_tramp, (_new)) 1384 1385 #else 1386 #define __BPF_DISPATCHER_SC_INIT(name) 1387 #define __BPF_DISPATCHER_SC(name) 1388 #define __BPF_DISPATCHER_CALL(name) bpf_func(ctx, insnsi) 1389 #define __BPF_DISPATCHER_UPDATE(_d, _new) 1390 #endif 1391 1392 #define BPF_DISPATCHER_INIT(_name) { \ 1393 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 1394 .func = &_name##_func, \ 1395 .progs = {}, \ 1396 .num_progs = 0, \ 1397 .image = NULL, \ 1398 .image_off = 0, \ 1399 .ksym = { \ 1400 .name = #_name, \ 1401 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 1402 }, \ 1403 __BPF_DISPATCHER_SC_INIT(_name##_call) \ 1404 } 1405 1406 #define DEFINE_BPF_DISPATCHER(name) \ 1407 __BPF_DISPATCHER_SC(name); \ 1408 noinline __bpfcall unsigned int bpf_dispatcher_##name##_func( \ 1409 const void *ctx, \ 1410 const struct bpf_insn *insnsi, \ 1411 bpf_func_t bpf_func) \ 1412 { \ 1413 return __BPF_DISPATCHER_CALL(name); \ 1414 } \ 1415 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 1416 struct bpf_dispatcher bpf_dispatcher_##name = \ 1417 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 1418 1419 #define DECLARE_BPF_DISPATCHER(name) \ 1420 unsigned int bpf_dispatcher_##name##_func( \ 1421 const void *ctx, \ 1422 const struct bpf_insn *insnsi, \ 1423 bpf_func_t bpf_func); \ 1424 extern struct bpf_dispatcher bpf_dispatcher_##name; 1425 1426 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 1427 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 1428 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 1429 struct bpf_prog *to); 1430 /* Called only from JIT-enabled code, so there's no need for stubs. */ 1431 void bpf_image_ksym_init(void *data, unsigned int size, struct bpf_ksym *ksym); 1432 void bpf_image_ksym_add(struct bpf_ksym *ksym); 1433 void bpf_image_ksym_del(struct bpf_ksym *ksym); 1434 void bpf_ksym_add(struct bpf_ksym *ksym); 1435 void bpf_ksym_del(struct bpf_ksym *ksym); 1436 int bpf_jit_charge_modmem(u32 size); 1437 void bpf_jit_uncharge_modmem(u32 size); 1438 bool bpf_prog_has_trampoline(const struct bpf_prog *prog); 1439 #else 1440 static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link, 1441 struct bpf_trampoline *tr, 1442 struct bpf_prog *tgt_prog) 1443 { 1444 return -ENOTSUPP; 1445 } 1446 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, 1447 struct bpf_trampoline *tr, 1448 struct bpf_prog *tgt_prog) 1449 { 1450 return -ENOTSUPP; 1451 } 1452 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 1453 struct bpf_attach_target_info *tgt_info) 1454 { 1455 return NULL; 1456 } 1457 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 1458 #define DEFINE_BPF_DISPATCHER(name) 1459 #define DECLARE_BPF_DISPATCHER(name) 1460 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 1461 #define BPF_DISPATCHER_PTR(name) NULL 1462 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 1463 struct bpf_prog *from, 1464 struct bpf_prog *to) {} 1465 static inline bool is_bpf_image_address(unsigned long address) 1466 { 1467 return false; 1468 } 1469 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog) 1470 { 1471 return false; 1472 } 1473 #endif 1474 1475 struct bpf_func_info_aux { 1476 u16 linkage; 1477 bool unreliable; 1478 bool called : 1; 1479 bool verified : 1; 1480 }; 1481 1482 enum bpf_jit_poke_reason { 1483 BPF_POKE_REASON_TAIL_CALL, 1484 }; 1485 1486 /* Descriptor of pokes pointing /into/ the JITed image. */ 1487 struct bpf_jit_poke_descriptor { 1488 void *tailcall_target; 1489 void *tailcall_bypass; 1490 void *bypass_addr; 1491 void *aux; 1492 union { 1493 struct { 1494 struct bpf_map *map; 1495 u32 key; 1496 } tail_call; 1497 }; 1498 bool tailcall_target_stable; 1499 u8 adj_off; 1500 u16 reason; 1501 u32 insn_idx; 1502 }; 1503 1504 /* reg_type info for ctx arguments */ 1505 struct bpf_ctx_arg_aux { 1506 u32 offset; 1507 enum bpf_reg_type reg_type; 1508 struct btf *btf; 1509 u32 btf_id; 1510 u32 ref_obj_id; 1511 bool refcounted; 1512 }; 1513 1514 struct btf_mod_pair { 1515 struct btf *btf; 1516 struct module *module; 1517 }; 1518 1519 struct bpf_kfunc_desc_tab; 1520 1521 struct bpf_prog_aux { 1522 atomic64_t refcnt; 1523 u32 used_map_cnt; 1524 u32 used_btf_cnt; 1525 u32 max_ctx_offset; 1526 u32 max_pkt_offset; 1527 u32 max_tp_access; 1528 u32 stack_depth; 1529 u32 id; 1530 u32 func_cnt; /* used by non-func prog as the number of func progs */ 1531 u32 real_func_cnt; /* includes hidden progs, only used for JIT and freeing progs */ 1532 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 1533 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1534 u32 attach_st_ops_member_off; 1535 u32 ctx_arg_info_size; 1536 u32 max_rdonly_access; 1537 u32 max_rdwr_access; 1538 struct btf *attach_btf; 1539 struct bpf_ctx_arg_aux *ctx_arg_info; 1540 void __percpu *priv_stack_ptr; 1541 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 1542 struct bpf_prog *dst_prog; 1543 struct bpf_trampoline *dst_trampoline; 1544 enum bpf_prog_type saved_dst_prog_type; 1545 enum bpf_attach_type saved_dst_attach_type; 1546 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 1547 bool dev_bound; /* Program is bound to the netdev. */ 1548 bool offload_requested; /* Program is bound and offloaded to the netdev. */ 1549 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 1550 bool attach_tracing_prog; /* true if tracing another tracing program */ 1551 bool func_proto_unreliable; 1552 bool tail_call_reachable; 1553 bool xdp_has_frags; 1554 bool exception_cb; 1555 bool exception_boundary; 1556 bool is_extended; /* true if extended by freplace program */ 1557 bool jits_use_priv_stack; 1558 bool priv_stack_requested; 1559 bool changes_pkt_data; 1560 bool might_sleep; 1561 u64 prog_array_member_cnt; /* counts how many times as member of prog_array */ 1562 struct mutex ext_mutex; /* mutex for is_extended and prog_array_member_cnt */ 1563 struct bpf_arena *arena; 1564 void (*recursion_detected)(struct bpf_prog *prog); /* callback if recursion is detected */ 1565 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 1566 const struct btf_type *attach_func_proto; 1567 /* function name for valid attach_btf_id */ 1568 const char *attach_func_name; 1569 struct bpf_prog **func; 1570 void *jit_data; /* JIT specific data. arch dependent */ 1571 struct bpf_jit_poke_descriptor *poke_tab; 1572 struct bpf_kfunc_desc_tab *kfunc_tab; 1573 struct bpf_kfunc_btf_tab *kfunc_btf_tab; 1574 u32 size_poke_tab; 1575 #ifdef CONFIG_FINEIBT 1576 struct bpf_ksym ksym_prefix; 1577 #endif 1578 struct bpf_ksym ksym; 1579 const struct bpf_prog_ops *ops; 1580 const struct bpf_struct_ops *st_ops; 1581 struct bpf_map **used_maps; 1582 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 1583 struct btf_mod_pair *used_btfs; 1584 struct bpf_prog *prog; 1585 struct user_struct *user; 1586 u64 load_time; /* ns since boottime */ 1587 u32 verified_insns; 1588 int cgroup_atype; /* enum cgroup_bpf_attach_type */ 1589 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1590 char name[BPF_OBJ_NAME_LEN]; 1591 u64 (*bpf_exception_cb)(u64 cookie, u64 sp, u64 bp, u64, u64); 1592 #ifdef CONFIG_SECURITY 1593 void *security; 1594 #endif 1595 struct bpf_token *token; 1596 struct bpf_prog_offload *offload; 1597 struct btf *btf; 1598 struct bpf_func_info *func_info; 1599 struct bpf_func_info_aux *func_info_aux; 1600 /* bpf_line_info loaded from userspace. linfo->insn_off 1601 * has the xlated insn offset. 1602 * Both the main and sub prog share the same linfo. 1603 * The subprog can access its first linfo by 1604 * using the linfo_idx. 1605 */ 1606 struct bpf_line_info *linfo; 1607 /* jited_linfo is the jited addr of the linfo. It has a 1608 * one to one mapping to linfo: 1609 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 1610 * Both the main and sub prog share the same jited_linfo. 1611 * The subprog can access its first jited_linfo by 1612 * using the linfo_idx. 1613 */ 1614 void **jited_linfo; 1615 u32 func_info_cnt; 1616 u32 nr_linfo; 1617 /* subprog can use linfo_idx to access its first linfo and 1618 * jited_linfo. 1619 * main prog always has linfo_idx == 0 1620 */ 1621 u32 linfo_idx; 1622 struct module *mod; 1623 u32 num_exentries; 1624 struct exception_table_entry *extable; 1625 union { 1626 struct work_struct work; 1627 struct rcu_head rcu; 1628 }; 1629 }; 1630 1631 struct bpf_prog { 1632 u16 pages; /* Number of allocated pages */ 1633 u16 jited:1, /* Is our filter JIT'ed? */ 1634 jit_requested:1,/* archs need to JIT the prog */ 1635 gpl_compatible:1, /* Is filter GPL compatible? */ 1636 cb_access:1, /* Is control block accessed? */ 1637 dst_needed:1, /* Do we need dst entry? */ 1638 blinding_requested:1, /* needs constant blinding */ 1639 blinded:1, /* Was blinded */ 1640 is_func:1, /* program is a bpf function */ 1641 kprobe_override:1, /* Do we override a kprobe? */ 1642 has_callchain_buf:1, /* callchain buffer allocated? */ 1643 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */ 1644 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */ 1645 call_get_func_ip:1, /* Do we call get_func_ip() */ 1646 tstamp_type_access:1, /* Accessed __sk_buff->tstamp_type */ 1647 sleepable:1; /* BPF program is sleepable */ 1648 enum bpf_prog_type type; /* Type of BPF program */ 1649 enum bpf_attach_type expected_attach_type; /* For some prog types */ 1650 u32 len; /* Number of filter blocks */ 1651 u32 jited_len; /* Size of jited insns in bytes */ 1652 u8 tag[BPF_TAG_SIZE]; 1653 struct bpf_prog_stats __percpu *stats; 1654 int __percpu *active; 1655 unsigned int (*bpf_func)(const void *ctx, 1656 const struct bpf_insn *insn); 1657 struct bpf_prog_aux *aux; /* Auxiliary fields */ 1658 struct sock_fprog_kern *orig_prog; /* Original BPF program */ 1659 /* Instructions for interpreter */ 1660 union { 1661 DECLARE_FLEX_ARRAY(struct sock_filter, insns); 1662 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi); 1663 }; 1664 }; 1665 1666 struct bpf_array_aux { 1667 /* Programs with direct jumps into programs part of this array. */ 1668 struct list_head poke_progs; 1669 struct bpf_map *map; 1670 struct mutex poke_mutex; 1671 struct work_struct work; 1672 }; 1673 1674 struct bpf_link { 1675 atomic64_t refcnt; 1676 u32 id; 1677 enum bpf_link_type type; 1678 const struct bpf_link_ops *ops; 1679 struct bpf_prog *prog; 1680 /* whether BPF link itself has "sleepable" semantics, which can differ 1681 * from underlying BPF program having a "sleepable" semantics, as BPF 1682 * link's semantics is determined by target attach hook 1683 */ 1684 bool sleepable; 1685 /* rcu is used before freeing, work can be used to schedule that 1686 * RCU-based freeing before that, so they never overlap 1687 */ 1688 union { 1689 struct rcu_head rcu; 1690 struct work_struct work; 1691 }; 1692 }; 1693 1694 struct bpf_link_ops { 1695 void (*release)(struct bpf_link *link); 1696 /* deallocate link resources callback, called without RCU grace period 1697 * waiting 1698 */ 1699 void (*dealloc)(struct bpf_link *link); 1700 /* deallocate link resources callback, called after RCU grace period; 1701 * if either the underlying BPF program is sleepable or BPF link's 1702 * target hook is sleepable, we'll go through tasks trace RCU GP and 1703 * then "classic" RCU GP; this need for chaining tasks trace and 1704 * classic RCU GPs is designated by setting bpf_link->sleepable flag 1705 */ 1706 void (*dealloc_deferred)(struct bpf_link *link); 1707 int (*detach)(struct bpf_link *link); 1708 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 1709 struct bpf_prog *old_prog); 1710 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 1711 int (*fill_link_info)(const struct bpf_link *link, 1712 struct bpf_link_info *info); 1713 int (*update_map)(struct bpf_link *link, struct bpf_map *new_map, 1714 struct bpf_map *old_map); 1715 __poll_t (*poll)(struct file *file, struct poll_table_struct *pts); 1716 }; 1717 1718 struct bpf_tramp_link { 1719 struct bpf_link link; 1720 struct hlist_node tramp_hlist; 1721 u64 cookie; 1722 }; 1723 1724 struct bpf_shim_tramp_link { 1725 struct bpf_tramp_link link; 1726 struct bpf_trampoline *trampoline; 1727 }; 1728 1729 struct bpf_tracing_link { 1730 struct bpf_tramp_link link; 1731 enum bpf_attach_type attach_type; 1732 struct bpf_trampoline *trampoline; 1733 struct bpf_prog *tgt_prog; 1734 }; 1735 1736 struct bpf_raw_tp_link { 1737 struct bpf_link link; 1738 struct bpf_raw_event_map *btp; 1739 u64 cookie; 1740 }; 1741 1742 struct bpf_link_primer { 1743 struct bpf_link *link; 1744 struct file *file; 1745 int fd; 1746 u32 id; 1747 }; 1748 1749 struct bpf_mount_opts { 1750 kuid_t uid; 1751 kgid_t gid; 1752 umode_t mode; 1753 1754 /* BPF token-related delegation options */ 1755 u64 delegate_cmds; 1756 u64 delegate_maps; 1757 u64 delegate_progs; 1758 u64 delegate_attachs; 1759 }; 1760 1761 struct bpf_token { 1762 struct work_struct work; 1763 atomic64_t refcnt; 1764 struct user_namespace *userns; 1765 u64 allowed_cmds; 1766 u64 allowed_maps; 1767 u64 allowed_progs; 1768 u64 allowed_attachs; 1769 #ifdef CONFIG_SECURITY 1770 void *security; 1771 #endif 1772 }; 1773 1774 struct bpf_struct_ops_value; 1775 struct btf_member; 1776 1777 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 1778 /** 1779 * struct bpf_struct_ops - A structure of callbacks allowing a subsystem to 1780 * define a BPF_MAP_TYPE_STRUCT_OPS map type composed 1781 * of BPF_PROG_TYPE_STRUCT_OPS progs. 1782 * @verifier_ops: A structure of callbacks that are invoked by the verifier 1783 * when determining whether the struct_ops progs in the 1784 * struct_ops map are valid. 1785 * @init: A callback that is invoked a single time, and before any other 1786 * callback, to initialize the structure. A nonzero return value means 1787 * the subsystem could not be initialized. 1788 * @check_member: When defined, a callback invoked by the verifier to allow 1789 * the subsystem to determine if an entry in the struct_ops map 1790 * is valid. A nonzero return value means that the map is 1791 * invalid and should be rejected by the verifier. 1792 * @init_member: A callback that is invoked for each member of the struct_ops 1793 * map to allow the subsystem to initialize the member. A nonzero 1794 * value means the member could not be initialized. This callback 1795 * is exclusive with the @type, @type_id, @value_type, and 1796 * @value_id fields. 1797 * @reg: A callback that is invoked when the struct_ops map has been 1798 * initialized and is being attached to. Zero means the struct_ops map 1799 * has been successfully registered and is live. A nonzero return value 1800 * means the struct_ops map could not be registered. 1801 * @unreg: A callback that is invoked when the struct_ops map should be 1802 * unregistered. 1803 * @update: A callback that is invoked when the live struct_ops map is being 1804 * updated to contain new values. This callback is only invoked when 1805 * the struct_ops map is loaded with BPF_F_LINK. If not defined, the 1806 * it is assumed that the struct_ops map cannot be updated. 1807 * @validate: A callback that is invoked after all of the members have been 1808 * initialized. This callback should perform static checks on the 1809 * map, meaning that it should either fail or succeed 1810 * deterministically. A struct_ops map that has been validated may 1811 * not necessarily succeed in being registered if the call to @reg 1812 * fails. For example, a valid struct_ops map may be loaded, but 1813 * then fail to be registered due to there being another active 1814 * struct_ops map on the system in the subsystem already. For this 1815 * reason, if this callback is not defined, the check is skipped as 1816 * the struct_ops map will have final verification performed in 1817 * @reg. 1818 * @type: BTF type. 1819 * @value_type: Value type. 1820 * @name: The name of the struct bpf_struct_ops object. 1821 * @func_models: Func models 1822 * @type_id: BTF type id. 1823 * @value_id: BTF value id. 1824 */ 1825 struct bpf_struct_ops { 1826 const struct bpf_verifier_ops *verifier_ops; 1827 int (*init)(struct btf *btf); 1828 int (*check_member)(const struct btf_type *t, 1829 const struct btf_member *member, 1830 const struct bpf_prog *prog); 1831 int (*init_member)(const struct btf_type *t, 1832 const struct btf_member *member, 1833 void *kdata, const void *udata); 1834 int (*reg)(void *kdata, struct bpf_link *link); 1835 void (*unreg)(void *kdata, struct bpf_link *link); 1836 int (*update)(void *kdata, void *old_kdata, struct bpf_link *link); 1837 int (*validate)(void *kdata); 1838 void *cfi_stubs; 1839 struct module *owner; 1840 const char *name; 1841 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 1842 }; 1843 1844 /* Every member of a struct_ops type has an instance even a member is not 1845 * an operator (function pointer). The "info" field will be assigned to 1846 * prog->aux->ctx_arg_info of BPF struct_ops programs to provide the 1847 * argument information required by the verifier to verify the program. 1848 * 1849 * btf_ctx_access() will lookup prog->aux->ctx_arg_info to find the 1850 * corresponding entry for an given argument. 1851 */ 1852 struct bpf_struct_ops_arg_info { 1853 struct bpf_ctx_arg_aux *info; 1854 u32 cnt; 1855 }; 1856 1857 struct bpf_struct_ops_desc { 1858 struct bpf_struct_ops *st_ops; 1859 1860 const struct btf_type *type; 1861 const struct btf_type *value_type; 1862 u32 type_id; 1863 u32 value_id; 1864 1865 /* Collection of argument information for each member */ 1866 struct bpf_struct_ops_arg_info *arg_info; 1867 }; 1868 1869 enum bpf_struct_ops_state { 1870 BPF_STRUCT_OPS_STATE_INIT, 1871 BPF_STRUCT_OPS_STATE_INUSE, 1872 BPF_STRUCT_OPS_STATE_TOBEFREE, 1873 BPF_STRUCT_OPS_STATE_READY, 1874 }; 1875 1876 struct bpf_struct_ops_common_value { 1877 refcount_t refcnt; 1878 enum bpf_struct_ops_state state; 1879 }; 1880 1881 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 1882 /* This macro helps developer to register a struct_ops type and generate 1883 * type information correctly. Developers should use this macro to register 1884 * a struct_ops type instead of calling __register_bpf_struct_ops() directly. 1885 */ 1886 #define register_bpf_struct_ops(st_ops, type) \ 1887 ({ \ 1888 struct bpf_struct_ops_##type { \ 1889 struct bpf_struct_ops_common_value common; \ 1890 struct type data ____cacheline_aligned_in_smp; \ 1891 }; \ 1892 BTF_TYPE_EMIT(struct bpf_struct_ops_##type); \ 1893 __register_bpf_struct_ops(st_ops); \ 1894 }) 1895 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 1896 bool bpf_struct_ops_get(const void *kdata); 1897 void bpf_struct_ops_put(const void *kdata); 1898 int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff); 1899 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 1900 void *value); 1901 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks, 1902 struct bpf_tramp_link *link, 1903 const struct btf_func_model *model, 1904 void *stub_func, 1905 void **image, u32 *image_off, 1906 bool allow_alloc); 1907 void bpf_struct_ops_image_free(void *image); 1908 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1909 { 1910 if (owner == BPF_MODULE_OWNER) 1911 return bpf_struct_ops_get(data); 1912 else 1913 return try_module_get(owner); 1914 } 1915 static inline void bpf_module_put(const void *data, struct module *owner) 1916 { 1917 if (owner == BPF_MODULE_OWNER) 1918 bpf_struct_ops_put(data); 1919 else 1920 module_put(owner); 1921 } 1922 int bpf_struct_ops_link_create(union bpf_attr *attr); 1923 1924 #ifdef CONFIG_NET 1925 /* Define it here to avoid the use of forward declaration */ 1926 struct bpf_dummy_ops_state { 1927 int val; 1928 }; 1929 1930 struct bpf_dummy_ops { 1931 int (*test_1)(struct bpf_dummy_ops_state *cb); 1932 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2, 1933 char a3, unsigned long a4); 1934 int (*test_sleepable)(struct bpf_dummy_ops_state *cb); 1935 }; 1936 1937 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr, 1938 union bpf_attr __user *uattr); 1939 #endif 1940 int bpf_struct_ops_desc_init(struct bpf_struct_ops_desc *st_ops_desc, 1941 struct btf *btf, 1942 struct bpf_verifier_log *log); 1943 void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map); 1944 void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc); 1945 #else 1946 #define register_bpf_struct_ops(st_ops, type) ({ (void *)(st_ops); 0; }) 1947 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1948 { 1949 return try_module_get(owner); 1950 } 1951 static inline void bpf_module_put(const void *data, struct module *owner) 1952 { 1953 module_put(owner); 1954 } 1955 static inline int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff) 1956 { 1957 return -ENOTSUPP; 1958 } 1959 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 1960 void *key, 1961 void *value) 1962 { 1963 return -EINVAL; 1964 } 1965 static inline int bpf_struct_ops_link_create(union bpf_attr *attr) 1966 { 1967 return -EOPNOTSUPP; 1968 } 1969 static inline void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map) 1970 { 1971 } 1972 1973 static inline void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc) 1974 { 1975 } 1976 1977 #endif 1978 1979 int bpf_prog_ctx_arg_info_init(struct bpf_prog *prog, 1980 const struct bpf_ctx_arg_aux *info, u32 cnt); 1981 1982 #if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM) 1983 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 1984 int cgroup_atype); 1985 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog); 1986 #else 1987 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 1988 int cgroup_atype) 1989 { 1990 return -EOPNOTSUPP; 1991 } 1992 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog) 1993 { 1994 } 1995 #endif 1996 1997 struct bpf_array { 1998 struct bpf_map map; 1999 u32 elem_size; 2000 u32 index_mask; 2001 struct bpf_array_aux *aux; 2002 union { 2003 DECLARE_FLEX_ARRAY(char, value) __aligned(8); 2004 DECLARE_FLEX_ARRAY(void *, ptrs) __aligned(8); 2005 DECLARE_FLEX_ARRAY(void __percpu *, pptrs) __aligned(8); 2006 }; 2007 }; 2008 2009 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 2010 #define MAX_TAIL_CALL_CNT 33 2011 2012 /* Maximum number of loops for bpf_loop and bpf_iter_num. 2013 * It's enum to expose it (and thus make it discoverable) through BTF. 2014 */ 2015 enum { 2016 BPF_MAX_LOOPS = 8 * 1024 * 1024, 2017 BPF_MAX_TIMED_LOOPS = 0xffff, 2018 }; 2019 2020 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 2021 BPF_F_RDONLY_PROG | \ 2022 BPF_F_WRONLY | \ 2023 BPF_F_WRONLY_PROG) 2024 2025 #define BPF_MAP_CAN_READ BIT(0) 2026 #define BPF_MAP_CAN_WRITE BIT(1) 2027 2028 /* Maximum number of user-producer ring buffer samples that can be drained in 2029 * a call to bpf_user_ringbuf_drain(). 2030 */ 2031 #define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024) 2032 2033 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 2034 { 2035 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 2036 2037 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 2038 * not possible. 2039 */ 2040 if (access_flags & BPF_F_RDONLY_PROG) 2041 return BPF_MAP_CAN_READ; 2042 else if (access_flags & BPF_F_WRONLY_PROG) 2043 return BPF_MAP_CAN_WRITE; 2044 else 2045 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 2046 } 2047 2048 static inline bool bpf_map_flags_access_ok(u32 access_flags) 2049 { 2050 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 2051 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 2052 } 2053 2054 struct bpf_event_entry { 2055 struct perf_event *event; 2056 struct file *perf_file; 2057 struct file *map_file; 2058 struct rcu_head rcu; 2059 }; 2060 2061 static inline bool map_type_contains_progs(struct bpf_map *map) 2062 { 2063 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY || 2064 map->map_type == BPF_MAP_TYPE_DEVMAP || 2065 map->map_type == BPF_MAP_TYPE_CPUMAP; 2066 } 2067 2068 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp); 2069 int bpf_prog_calc_tag(struct bpf_prog *fp); 2070 2071 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 2072 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void); 2073 2074 const struct bpf_func_proto *bpf_get_perf_event_read_value_proto(void); 2075 2076 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 2077 unsigned long off, unsigned long len); 2078 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 2079 const struct bpf_insn *src, 2080 struct bpf_insn *dst, 2081 struct bpf_prog *prog, 2082 u32 *target_size); 2083 2084 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 2085 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 2086 2087 /* an array of programs to be executed under rcu_lock. 2088 * 2089 * Typical usage: 2090 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run); 2091 * 2092 * the structure returned by bpf_prog_array_alloc() should be populated 2093 * with program pointers and the last pointer must be NULL. 2094 * The user has to keep refcnt on the program and make sure the program 2095 * is removed from the array before bpf_prog_put(). 2096 * The 'struct bpf_prog_array *' should only be replaced with xchg() 2097 * since other cpus are walking the array of pointers in parallel. 2098 */ 2099 struct bpf_prog_array_item { 2100 struct bpf_prog *prog; 2101 union { 2102 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 2103 u64 bpf_cookie; 2104 }; 2105 }; 2106 2107 struct bpf_prog_array { 2108 struct rcu_head rcu; 2109 struct bpf_prog_array_item items[]; 2110 }; 2111 2112 struct bpf_empty_prog_array { 2113 struct bpf_prog_array hdr; 2114 struct bpf_prog *null_prog; 2115 }; 2116 2117 /* to avoid allocating empty bpf_prog_array for cgroups that 2118 * don't have bpf program attached use one global 'bpf_empty_prog_array' 2119 * It will not be modified the caller of bpf_prog_array_alloc() 2120 * (since caller requested prog_cnt == 0) 2121 * that pointer should be 'freed' by bpf_prog_array_free() 2122 */ 2123 extern struct bpf_empty_prog_array bpf_empty_prog_array; 2124 2125 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 2126 void bpf_prog_array_free(struct bpf_prog_array *progs); 2127 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */ 2128 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs); 2129 int bpf_prog_array_length(struct bpf_prog_array *progs); 2130 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 2131 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 2132 __u32 __user *prog_ids, u32 cnt); 2133 2134 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 2135 struct bpf_prog *old_prog); 2136 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 2137 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 2138 struct bpf_prog *prog); 2139 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 2140 u32 *prog_ids, u32 request_cnt, 2141 u32 *prog_cnt); 2142 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 2143 struct bpf_prog *exclude_prog, 2144 struct bpf_prog *include_prog, 2145 u64 bpf_cookie, 2146 struct bpf_prog_array **new_array); 2147 2148 struct bpf_run_ctx {}; 2149 2150 struct bpf_cg_run_ctx { 2151 struct bpf_run_ctx run_ctx; 2152 const struct bpf_prog_array_item *prog_item; 2153 int retval; 2154 }; 2155 2156 struct bpf_trace_run_ctx { 2157 struct bpf_run_ctx run_ctx; 2158 u64 bpf_cookie; 2159 bool is_uprobe; 2160 }; 2161 2162 struct bpf_tramp_run_ctx { 2163 struct bpf_run_ctx run_ctx; 2164 u64 bpf_cookie; 2165 struct bpf_run_ctx *saved_run_ctx; 2166 }; 2167 2168 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 2169 { 2170 struct bpf_run_ctx *old_ctx = NULL; 2171 2172 #ifdef CONFIG_BPF_SYSCALL 2173 old_ctx = current->bpf_ctx; 2174 current->bpf_ctx = new_ctx; 2175 #endif 2176 return old_ctx; 2177 } 2178 2179 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 2180 { 2181 #ifdef CONFIG_BPF_SYSCALL 2182 current->bpf_ctx = old_ctx; 2183 #endif 2184 } 2185 2186 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 2187 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 2188 /* BPF program asks to set CN on the packet. */ 2189 #define BPF_RET_SET_CN (1 << 0) 2190 2191 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 2192 2193 static __always_inline u32 2194 bpf_prog_run_array(const struct bpf_prog_array *array, 2195 const void *ctx, bpf_prog_run_fn run_prog) 2196 { 2197 const struct bpf_prog_array_item *item; 2198 const struct bpf_prog *prog; 2199 struct bpf_run_ctx *old_run_ctx; 2200 struct bpf_trace_run_ctx run_ctx; 2201 u32 ret = 1; 2202 2203 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held"); 2204 2205 if (unlikely(!array)) 2206 return ret; 2207 2208 run_ctx.is_uprobe = false; 2209 2210 migrate_disable(); 2211 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 2212 item = &array->items[0]; 2213 while ((prog = READ_ONCE(item->prog))) { 2214 run_ctx.bpf_cookie = item->bpf_cookie; 2215 ret &= run_prog(prog, ctx); 2216 item++; 2217 } 2218 bpf_reset_run_ctx(old_run_ctx); 2219 migrate_enable(); 2220 return ret; 2221 } 2222 2223 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs: 2224 * 2225 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array 2226 * overall. As a result, we must use the bpf_prog_array_free_sleepable 2227 * in order to use the tasks_trace rcu grace period. 2228 * 2229 * When a non-sleepable program is inside the array, we take the rcu read 2230 * section and disable preemption for that program alone, so it can access 2231 * rcu-protected dynamically sized maps. 2232 */ 2233 static __always_inline u32 2234 bpf_prog_run_array_uprobe(const struct bpf_prog_array *array, 2235 const void *ctx, bpf_prog_run_fn run_prog) 2236 { 2237 const struct bpf_prog_array_item *item; 2238 const struct bpf_prog *prog; 2239 struct bpf_run_ctx *old_run_ctx; 2240 struct bpf_trace_run_ctx run_ctx; 2241 u32 ret = 1; 2242 2243 might_fault(); 2244 RCU_LOCKDEP_WARN(!rcu_read_lock_trace_held(), "no rcu lock held"); 2245 2246 if (unlikely(!array)) 2247 return ret; 2248 2249 migrate_disable(); 2250 2251 run_ctx.is_uprobe = true; 2252 2253 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 2254 item = &array->items[0]; 2255 while ((prog = READ_ONCE(item->prog))) { 2256 if (!prog->sleepable) 2257 rcu_read_lock(); 2258 2259 run_ctx.bpf_cookie = item->bpf_cookie; 2260 ret &= run_prog(prog, ctx); 2261 item++; 2262 2263 if (!prog->sleepable) 2264 rcu_read_unlock(); 2265 } 2266 bpf_reset_run_ctx(old_run_ctx); 2267 migrate_enable(); 2268 return ret; 2269 } 2270 2271 #ifdef CONFIG_BPF_SYSCALL 2272 DECLARE_PER_CPU(int, bpf_prog_active); 2273 extern struct mutex bpf_stats_enabled_mutex; 2274 2275 /* 2276 * Block execution of BPF programs attached to instrumentation (perf, 2277 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 2278 * these events can happen inside a region which holds a map bucket lock 2279 * and can deadlock on it. 2280 */ 2281 static inline void bpf_disable_instrumentation(void) 2282 { 2283 migrate_disable(); 2284 this_cpu_inc(bpf_prog_active); 2285 } 2286 2287 static inline void bpf_enable_instrumentation(void) 2288 { 2289 this_cpu_dec(bpf_prog_active); 2290 migrate_enable(); 2291 } 2292 2293 extern const struct super_operations bpf_super_ops; 2294 extern const struct file_operations bpf_map_fops; 2295 extern const struct file_operations bpf_prog_fops; 2296 extern const struct file_operations bpf_iter_fops; 2297 2298 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 2299 extern const struct bpf_prog_ops _name ## _prog_ops; \ 2300 extern const struct bpf_verifier_ops _name ## _verifier_ops; 2301 #define BPF_MAP_TYPE(_id, _ops) \ 2302 extern const struct bpf_map_ops _ops; 2303 #define BPF_LINK_TYPE(_id, _name) 2304 #include <linux/bpf_types.h> 2305 #undef BPF_PROG_TYPE 2306 #undef BPF_MAP_TYPE 2307 #undef BPF_LINK_TYPE 2308 2309 extern const struct bpf_prog_ops bpf_offload_prog_ops; 2310 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 2311 extern const struct bpf_verifier_ops xdp_analyzer_ops; 2312 2313 struct bpf_prog *bpf_prog_get(u32 ufd); 2314 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 2315 bool attach_drv); 2316 void bpf_prog_add(struct bpf_prog *prog, int i); 2317 void bpf_prog_sub(struct bpf_prog *prog, int i); 2318 void bpf_prog_inc(struct bpf_prog *prog); 2319 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 2320 void bpf_prog_put(struct bpf_prog *prog); 2321 2322 void bpf_prog_free_id(struct bpf_prog *prog); 2323 void bpf_map_free_id(struct bpf_map *map); 2324 2325 struct btf_field *btf_record_find(const struct btf_record *rec, 2326 u32 offset, u32 field_mask); 2327 void btf_record_free(struct btf_record *rec); 2328 void bpf_map_free_record(struct bpf_map *map); 2329 struct btf_record *btf_record_dup(const struct btf_record *rec); 2330 bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b); 2331 void bpf_obj_free_timer(const struct btf_record *rec, void *obj); 2332 void bpf_obj_free_workqueue(const struct btf_record *rec, void *obj); 2333 void bpf_obj_free_fields(const struct btf_record *rec, void *obj); 2334 void __bpf_obj_drop_impl(void *p, const struct btf_record *rec, bool percpu); 2335 2336 struct bpf_map *bpf_map_get(u32 ufd); 2337 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 2338 2339 /* 2340 * The __bpf_map_get() and __btf_get_by_fd() functions parse a file 2341 * descriptor and return a corresponding map or btf object. 2342 * Their names are double underscored to emphasize the fact that they 2343 * do not increase refcnt. To also increase refcnt use corresponding 2344 * bpf_map_get() and btf_get_by_fd() functions. 2345 */ 2346 2347 static inline struct bpf_map *__bpf_map_get(struct fd f) 2348 { 2349 if (fd_empty(f)) 2350 return ERR_PTR(-EBADF); 2351 if (unlikely(fd_file(f)->f_op != &bpf_map_fops)) 2352 return ERR_PTR(-EINVAL); 2353 return fd_file(f)->private_data; 2354 } 2355 2356 static inline struct btf *__btf_get_by_fd(struct fd f) 2357 { 2358 if (fd_empty(f)) 2359 return ERR_PTR(-EBADF); 2360 if (unlikely(fd_file(f)->f_op != &btf_fops)) 2361 return ERR_PTR(-EINVAL); 2362 return fd_file(f)->private_data; 2363 } 2364 2365 void bpf_map_inc(struct bpf_map *map); 2366 void bpf_map_inc_with_uref(struct bpf_map *map); 2367 struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref); 2368 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 2369 void bpf_map_put_with_uref(struct bpf_map *map); 2370 void bpf_map_put(struct bpf_map *map); 2371 void *bpf_map_area_alloc(u64 size, int numa_node); 2372 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 2373 void bpf_map_area_free(void *base); 2374 bool bpf_map_write_active(const struct bpf_map *map); 2375 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 2376 int generic_map_lookup_batch(struct bpf_map *map, 2377 const union bpf_attr *attr, 2378 union bpf_attr __user *uattr); 2379 int generic_map_update_batch(struct bpf_map *map, struct file *map_file, 2380 const union bpf_attr *attr, 2381 union bpf_attr __user *uattr); 2382 int generic_map_delete_batch(struct bpf_map *map, 2383 const union bpf_attr *attr, 2384 union bpf_attr __user *uattr); 2385 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 2386 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 2387 2388 int bpf_map_alloc_pages(const struct bpf_map *map, int nid, 2389 unsigned long nr_pages, struct page **page_array); 2390 #ifdef CONFIG_MEMCG 2391 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 2392 int node); 2393 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 2394 void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size, 2395 gfp_t flags); 2396 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 2397 size_t align, gfp_t flags); 2398 #else 2399 /* 2400 * These specialized allocators have to be macros for their allocations to be 2401 * accounted separately (to have separate alloc_tag). 2402 */ 2403 #define bpf_map_kmalloc_node(_map, _size, _flags, _node) \ 2404 kmalloc_node(_size, _flags, _node) 2405 #define bpf_map_kzalloc(_map, _size, _flags) \ 2406 kzalloc(_size, _flags) 2407 #define bpf_map_kvcalloc(_map, _n, _size, _flags) \ 2408 kvcalloc(_n, _size, _flags) 2409 #define bpf_map_alloc_percpu(_map, _size, _align, _flags) \ 2410 __alloc_percpu_gfp(_size, _align, _flags) 2411 #endif 2412 2413 static inline int 2414 bpf_map_init_elem_count(struct bpf_map *map) 2415 { 2416 size_t size = sizeof(*map->elem_count), align = size; 2417 gfp_t flags = GFP_USER | __GFP_NOWARN; 2418 2419 map->elem_count = bpf_map_alloc_percpu(map, size, align, flags); 2420 if (!map->elem_count) 2421 return -ENOMEM; 2422 2423 return 0; 2424 } 2425 2426 static inline void 2427 bpf_map_free_elem_count(struct bpf_map *map) 2428 { 2429 free_percpu(map->elem_count); 2430 } 2431 2432 static inline void bpf_map_inc_elem_count(struct bpf_map *map) 2433 { 2434 this_cpu_inc(*map->elem_count); 2435 } 2436 2437 static inline void bpf_map_dec_elem_count(struct bpf_map *map) 2438 { 2439 this_cpu_dec(*map->elem_count); 2440 } 2441 2442 extern int sysctl_unprivileged_bpf_disabled; 2443 2444 bool bpf_token_capable(const struct bpf_token *token, int cap); 2445 2446 static inline bool bpf_allow_ptr_leaks(const struct bpf_token *token) 2447 { 2448 return bpf_token_capable(token, CAP_PERFMON); 2449 } 2450 2451 static inline bool bpf_allow_uninit_stack(const struct bpf_token *token) 2452 { 2453 return bpf_token_capable(token, CAP_PERFMON); 2454 } 2455 2456 static inline bool bpf_bypass_spec_v1(const struct bpf_token *token) 2457 { 2458 return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON); 2459 } 2460 2461 static inline bool bpf_bypass_spec_v4(const struct bpf_token *token) 2462 { 2463 return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON); 2464 } 2465 2466 int bpf_map_new_fd(struct bpf_map *map, int flags); 2467 int bpf_prog_new_fd(struct bpf_prog *prog); 2468 2469 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 2470 const struct bpf_link_ops *ops, struct bpf_prog *prog); 2471 void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type, 2472 const struct bpf_link_ops *ops, struct bpf_prog *prog, 2473 bool sleepable); 2474 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 2475 int bpf_link_settle(struct bpf_link_primer *primer); 2476 void bpf_link_cleanup(struct bpf_link_primer *primer); 2477 void bpf_link_inc(struct bpf_link *link); 2478 struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link); 2479 void bpf_link_put(struct bpf_link *link); 2480 int bpf_link_new_fd(struct bpf_link *link); 2481 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 2482 struct bpf_link *bpf_link_get_curr_or_next(u32 *id); 2483 2484 void bpf_token_inc(struct bpf_token *token); 2485 void bpf_token_put(struct bpf_token *token); 2486 int bpf_token_create(union bpf_attr *attr); 2487 struct bpf_token *bpf_token_get_from_fd(u32 ufd); 2488 2489 bool bpf_token_allow_cmd(const struct bpf_token *token, enum bpf_cmd cmd); 2490 bool bpf_token_allow_map_type(const struct bpf_token *token, enum bpf_map_type type); 2491 bool bpf_token_allow_prog_type(const struct bpf_token *token, 2492 enum bpf_prog_type prog_type, 2493 enum bpf_attach_type attach_type); 2494 2495 int bpf_obj_pin_user(u32 ufd, int path_fd, const char __user *pathname); 2496 int bpf_obj_get_user(int path_fd, const char __user *pathname, int flags); 2497 struct inode *bpf_get_inode(struct super_block *sb, const struct inode *dir, 2498 umode_t mode); 2499 2500 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 2501 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 2502 extern int bpf_iter_ ## target(args); \ 2503 int __init bpf_iter_ ## target(args) { return 0; } 2504 2505 /* 2506 * The task type of iterators. 2507 * 2508 * For BPF task iterators, they can be parameterized with various 2509 * parameters to visit only some of tasks. 2510 * 2511 * BPF_TASK_ITER_ALL (default) 2512 * Iterate over resources of every task. 2513 * 2514 * BPF_TASK_ITER_TID 2515 * Iterate over resources of a task/tid. 2516 * 2517 * BPF_TASK_ITER_TGID 2518 * Iterate over resources of every task of a process / task group. 2519 */ 2520 enum bpf_iter_task_type { 2521 BPF_TASK_ITER_ALL = 0, 2522 BPF_TASK_ITER_TID, 2523 BPF_TASK_ITER_TGID, 2524 }; 2525 2526 struct bpf_iter_aux_info { 2527 /* for map_elem iter */ 2528 struct bpf_map *map; 2529 2530 /* for cgroup iter */ 2531 struct { 2532 struct cgroup *start; /* starting cgroup */ 2533 enum bpf_cgroup_iter_order order; 2534 } cgroup; 2535 struct { 2536 enum bpf_iter_task_type type; 2537 u32 pid; 2538 } task; 2539 }; 2540 2541 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 2542 union bpf_iter_link_info *linfo, 2543 struct bpf_iter_aux_info *aux); 2544 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 2545 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 2546 struct seq_file *seq); 2547 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 2548 struct bpf_link_info *info); 2549 typedef const struct bpf_func_proto * 2550 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 2551 const struct bpf_prog *prog); 2552 2553 enum bpf_iter_feature { 2554 BPF_ITER_RESCHED = BIT(0), 2555 }; 2556 2557 #define BPF_ITER_CTX_ARG_MAX 2 2558 struct bpf_iter_reg { 2559 const char *target; 2560 bpf_iter_attach_target_t attach_target; 2561 bpf_iter_detach_target_t detach_target; 2562 bpf_iter_show_fdinfo_t show_fdinfo; 2563 bpf_iter_fill_link_info_t fill_link_info; 2564 bpf_iter_get_func_proto_t get_func_proto; 2565 u32 ctx_arg_info_size; 2566 u32 feature; 2567 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 2568 const struct bpf_iter_seq_info *seq_info; 2569 }; 2570 2571 struct bpf_iter_meta { 2572 __bpf_md_ptr(struct seq_file *, seq); 2573 u64 session_id; 2574 u64 seq_num; 2575 }; 2576 2577 struct bpf_iter__bpf_map_elem { 2578 __bpf_md_ptr(struct bpf_iter_meta *, meta); 2579 __bpf_md_ptr(struct bpf_map *, map); 2580 __bpf_md_ptr(void *, key); 2581 __bpf_md_ptr(void *, value); 2582 }; 2583 2584 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 2585 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 2586 int bpf_iter_prog_supported(struct bpf_prog *prog); 2587 const struct bpf_func_proto * 2588 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 2589 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 2590 int bpf_iter_new_fd(struct bpf_link *link); 2591 bool bpf_link_is_iter(struct bpf_link *link); 2592 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 2593 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 2594 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 2595 struct seq_file *seq); 2596 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 2597 struct bpf_link_info *info); 2598 2599 int map_set_for_each_callback_args(struct bpf_verifier_env *env, 2600 struct bpf_func_state *caller, 2601 struct bpf_func_state *callee); 2602 2603 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 2604 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 2605 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 2606 u64 flags); 2607 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 2608 u64 flags); 2609 2610 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 2611 2612 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 2613 void *key, void *value, u64 map_flags); 2614 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 2615 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 2616 void *key, void *value, u64 map_flags); 2617 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 2618 2619 int bpf_get_file_flag(int flags); 2620 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 2621 size_t actual_size); 2622 2623 /* verify correctness of eBPF program */ 2624 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size); 2625 2626 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 2627 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 2628 #endif 2629 2630 struct btf *bpf_get_btf_vmlinux(void); 2631 2632 /* Map specifics */ 2633 struct xdp_frame; 2634 struct sk_buff; 2635 struct bpf_dtab_netdev; 2636 struct bpf_cpu_map_entry; 2637 2638 void __dev_flush(struct list_head *flush_list); 2639 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 2640 struct net_device *dev_rx); 2641 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 2642 struct net_device *dev_rx); 2643 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 2644 struct bpf_map *map, bool exclude_ingress); 2645 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 2646 const struct bpf_prog *xdp_prog); 2647 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 2648 const struct bpf_prog *xdp_prog, 2649 struct bpf_map *map, bool exclude_ingress); 2650 2651 void __cpu_map_flush(struct list_head *flush_list); 2652 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf, 2653 struct net_device *dev_rx); 2654 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 2655 struct sk_buff *skb); 2656 2657 /* Return map's numa specified by userspace */ 2658 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 2659 { 2660 return (attr->map_flags & BPF_F_NUMA_NODE) ? 2661 attr->numa_node : NUMA_NO_NODE; 2662 } 2663 2664 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 2665 int array_map_alloc_check(union bpf_attr *attr); 2666 2667 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 2668 union bpf_attr __user *uattr); 2669 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 2670 union bpf_attr __user *uattr); 2671 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 2672 const union bpf_attr *kattr, 2673 union bpf_attr __user *uattr); 2674 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 2675 const union bpf_attr *kattr, 2676 union bpf_attr __user *uattr); 2677 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 2678 const union bpf_attr *kattr, 2679 union bpf_attr __user *uattr); 2680 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 2681 const union bpf_attr *kattr, 2682 union bpf_attr __user *uattr); 2683 int bpf_prog_test_run_nf(struct bpf_prog *prog, 2684 const union bpf_attr *kattr, 2685 union bpf_attr __user *uattr); 2686 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 2687 const struct bpf_prog *prog, 2688 struct bpf_insn_access_aux *info); 2689 2690 static inline bool bpf_tracing_ctx_access(int off, int size, 2691 enum bpf_access_type type) 2692 { 2693 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS) 2694 return false; 2695 if (type != BPF_READ) 2696 return false; 2697 if (off % size != 0) 2698 return false; 2699 return true; 2700 } 2701 2702 static inline bool bpf_tracing_btf_ctx_access(int off, int size, 2703 enum bpf_access_type type, 2704 const struct bpf_prog *prog, 2705 struct bpf_insn_access_aux *info) 2706 { 2707 if (!bpf_tracing_ctx_access(off, size, type)) 2708 return false; 2709 return btf_ctx_access(off, size, type, prog, info); 2710 } 2711 2712 int btf_struct_access(struct bpf_verifier_log *log, 2713 const struct bpf_reg_state *reg, 2714 int off, int size, enum bpf_access_type atype, 2715 u32 *next_btf_id, enum bpf_type_flag *flag, const char **field_name); 2716 bool btf_struct_ids_match(struct bpf_verifier_log *log, 2717 const struct btf *btf, u32 id, int off, 2718 const struct btf *need_btf, u32 need_type_id, 2719 bool strict); 2720 2721 int btf_distill_func_proto(struct bpf_verifier_log *log, 2722 struct btf *btf, 2723 const struct btf_type *func_proto, 2724 const char *func_name, 2725 struct btf_func_model *m); 2726 2727 struct bpf_reg_state; 2728 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog); 2729 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 2730 struct btf *btf, const struct btf_type *t); 2731 const char *btf_find_decl_tag_value(const struct btf *btf, const struct btf_type *pt, 2732 int comp_idx, const char *tag_key); 2733 int btf_find_next_decl_tag(const struct btf *btf, const struct btf_type *pt, 2734 int comp_idx, const char *tag_key, int last_id); 2735 2736 struct bpf_prog *bpf_prog_by_id(u32 id); 2737 struct bpf_link *bpf_link_by_id(u32 id); 2738 2739 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id, 2740 const struct bpf_prog *prog); 2741 void bpf_task_storage_free(struct task_struct *task); 2742 void bpf_cgrp_storage_free(struct cgroup *cgroup); 2743 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog); 2744 const struct btf_func_model * 2745 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 2746 const struct bpf_insn *insn); 2747 int bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id, 2748 u16 btf_fd_idx, u8 **func_addr); 2749 2750 struct bpf_core_ctx { 2751 struct bpf_verifier_log *log; 2752 const struct btf *btf; 2753 }; 2754 2755 bool btf_nested_type_is_trusted(struct bpf_verifier_log *log, 2756 const struct bpf_reg_state *reg, 2757 const char *field_name, u32 btf_id, const char *suffix); 2758 2759 bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log, 2760 const struct btf *reg_btf, u32 reg_id, 2761 const struct btf *arg_btf, u32 arg_id); 2762 2763 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo, 2764 int relo_idx, void *insn); 2765 2766 static inline bool unprivileged_ebpf_enabled(void) 2767 { 2768 return !sysctl_unprivileged_bpf_disabled; 2769 } 2770 2771 /* Not all bpf prog type has the bpf_ctx. 2772 * For the bpf prog type that has initialized the bpf_ctx, 2773 * this function can be used to decide if a kernel function 2774 * is called by a bpf program. 2775 */ 2776 static inline bool has_current_bpf_ctx(void) 2777 { 2778 return !!current->bpf_ctx; 2779 } 2780 2781 void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog); 2782 2783 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 2784 enum bpf_dynptr_type type, u32 offset, u32 size); 2785 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr); 2786 void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr); 2787 2788 #else /* !CONFIG_BPF_SYSCALL */ 2789 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 2790 { 2791 return ERR_PTR(-EOPNOTSUPP); 2792 } 2793 2794 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 2795 enum bpf_prog_type type, 2796 bool attach_drv) 2797 { 2798 return ERR_PTR(-EOPNOTSUPP); 2799 } 2800 2801 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 2802 { 2803 } 2804 2805 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 2806 { 2807 } 2808 2809 static inline void bpf_prog_put(struct bpf_prog *prog) 2810 { 2811 } 2812 2813 static inline void bpf_prog_inc(struct bpf_prog *prog) 2814 { 2815 } 2816 2817 static inline struct bpf_prog *__must_check 2818 bpf_prog_inc_not_zero(struct bpf_prog *prog) 2819 { 2820 return ERR_PTR(-EOPNOTSUPP); 2821 } 2822 2823 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 2824 const struct bpf_link_ops *ops, 2825 struct bpf_prog *prog) 2826 { 2827 } 2828 2829 static inline void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type, 2830 const struct bpf_link_ops *ops, struct bpf_prog *prog, 2831 bool sleepable) 2832 { 2833 } 2834 2835 static inline int bpf_link_prime(struct bpf_link *link, 2836 struct bpf_link_primer *primer) 2837 { 2838 return -EOPNOTSUPP; 2839 } 2840 2841 static inline int bpf_link_settle(struct bpf_link_primer *primer) 2842 { 2843 return -EOPNOTSUPP; 2844 } 2845 2846 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 2847 { 2848 } 2849 2850 static inline void bpf_link_inc(struct bpf_link *link) 2851 { 2852 } 2853 2854 static inline struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link) 2855 { 2856 return NULL; 2857 } 2858 2859 static inline void bpf_link_put(struct bpf_link *link) 2860 { 2861 } 2862 2863 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 2864 { 2865 return -EOPNOTSUPP; 2866 } 2867 2868 static inline bool bpf_token_capable(const struct bpf_token *token, int cap) 2869 { 2870 return capable(cap) || (cap != CAP_SYS_ADMIN && capable(CAP_SYS_ADMIN)); 2871 } 2872 2873 static inline void bpf_token_inc(struct bpf_token *token) 2874 { 2875 } 2876 2877 static inline void bpf_token_put(struct bpf_token *token) 2878 { 2879 } 2880 2881 static inline struct bpf_token *bpf_token_get_from_fd(u32 ufd) 2882 { 2883 return ERR_PTR(-EOPNOTSUPP); 2884 } 2885 2886 static inline void __dev_flush(struct list_head *flush_list) 2887 { 2888 } 2889 2890 struct xdp_frame; 2891 struct bpf_dtab_netdev; 2892 struct bpf_cpu_map_entry; 2893 2894 static inline 2895 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 2896 struct net_device *dev_rx) 2897 { 2898 return 0; 2899 } 2900 2901 static inline 2902 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 2903 struct net_device *dev_rx) 2904 { 2905 return 0; 2906 } 2907 2908 static inline 2909 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 2910 struct bpf_map *map, bool exclude_ingress) 2911 { 2912 return 0; 2913 } 2914 2915 struct sk_buff; 2916 2917 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 2918 struct sk_buff *skb, 2919 const struct bpf_prog *xdp_prog) 2920 { 2921 return 0; 2922 } 2923 2924 static inline 2925 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 2926 const struct bpf_prog *xdp_prog, 2927 struct bpf_map *map, bool exclude_ingress) 2928 { 2929 return 0; 2930 } 2931 2932 static inline void __cpu_map_flush(struct list_head *flush_list) 2933 { 2934 } 2935 2936 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 2937 struct xdp_frame *xdpf, 2938 struct net_device *dev_rx) 2939 { 2940 return 0; 2941 } 2942 2943 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 2944 struct sk_buff *skb) 2945 { 2946 return -EOPNOTSUPP; 2947 } 2948 2949 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 2950 enum bpf_prog_type type) 2951 { 2952 return ERR_PTR(-EOPNOTSUPP); 2953 } 2954 2955 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 2956 const union bpf_attr *kattr, 2957 union bpf_attr __user *uattr) 2958 { 2959 return -ENOTSUPP; 2960 } 2961 2962 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 2963 const union bpf_attr *kattr, 2964 union bpf_attr __user *uattr) 2965 { 2966 return -ENOTSUPP; 2967 } 2968 2969 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 2970 const union bpf_attr *kattr, 2971 union bpf_attr __user *uattr) 2972 { 2973 return -ENOTSUPP; 2974 } 2975 2976 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 2977 const union bpf_attr *kattr, 2978 union bpf_attr __user *uattr) 2979 { 2980 return -ENOTSUPP; 2981 } 2982 2983 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 2984 const union bpf_attr *kattr, 2985 union bpf_attr __user *uattr) 2986 { 2987 return -ENOTSUPP; 2988 } 2989 2990 static inline void bpf_map_put(struct bpf_map *map) 2991 { 2992 } 2993 2994 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 2995 { 2996 return ERR_PTR(-ENOTSUPP); 2997 } 2998 2999 static inline int btf_struct_access(struct bpf_verifier_log *log, 3000 const struct bpf_reg_state *reg, 3001 int off, int size, enum bpf_access_type atype, 3002 u32 *next_btf_id, enum bpf_type_flag *flag, 3003 const char **field_name) 3004 { 3005 return -EACCES; 3006 } 3007 3008 static inline const struct bpf_func_proto * 3009 bpf_base_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) 3010 { 3011 return NULL; 3012 } 3013 3014 static inline void bpf_task_storage_free(struct task_struct *task) 3015 { 3016 } 3017 3018 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog) 3019 { 3020 return false; 3021 } 3022 3023 static inline const struct btf_func_model * 3024 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 3025 const struct bpf_insn *insn) 3026 { 3027 return NULL; 3028 } 3029 3030 static inline int 3031 bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id, 3032 u16 btf_fd_idx, u8 **func_addr) 3033 { 3034 return -ENOTSUPP; 3035 } 3036 3037 static inline bool unprivileged_ebpf_enabled(void) 3038 { 3039 return false; 3040 } 3041 3042 static inline bool has_current_bpf_ctx(void) 3043 { 3044 return false; 3045 } 3046 3047 static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog) 3048 { 3049 } 3050 3051 static inline void bpf_cgrp_storage_free(struct cgroup *cgroup) 3052 { 3053 } 3054 3055 static inline void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 3056 enum bpf_dynptr_type type, u32 offset, u32 size) 3057 { 3058 } 3059 3060 static inline void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr) 3061 { 3062 } 3063 3064 static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr) 3065 { 3066 } 3067 #endif /* CONFIG_BPF_SYSCALL */ 3068 3069 static __always_inline int 3070 bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr) 3071 { 3072 int ret = -EFAULT; 3073 3074 if (IS_ENABLED(CONFIG_BPF_EVENTS)) 3075 ret = copy_from_kernel_nofault(dst, unsafe_ptr, size); 3076 if (unlikely(ret < 0)) 3077 memset(dst, 0, size); 3078 return ret; 3079 } 3080 3081 void __bpf_free_used_btfs(struct btf_mod_pair *used_btfs, u32 len); 3082 3083 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 3084 enum bpf_prog_type type) 3085 { 3086 return bpf_prog_get_type_dev(ufd, type, false); 3087 } 3088 3089 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 3090 struct bpf_map **used_maps, u32 len); 3091 3092 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 3093 3094 int bpf_prog_offload_compile(struct bpf_prog *prog); 3095 void bpf_prog_dev_bound_destroy(struct bpf_prog *prog); 3096 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 3097 struct bpf_prog *prog); 3098 3099 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 3100 3101 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 3102 int bpf_map_offload_update_elem(struct bpf_map *map, 3103 void *key, void *value, u64 flags); 3104 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 3105 int bpf_map_offload_get_next_key(struct bpf_map *map, 3106 void *key, void *next_key); 3107 3108 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 3109 3110 struct bpf_offload_dev * 3111 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 3112 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 3113 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 3114 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 3115 struct net_device *netdev); 3116 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 3117 struct net_device *netdev); 3118 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 3119 3120 void unpriv_ebpf_notify(int new_state); 3121 3122 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 3123 int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log, 3124 struct bpf_prog_aux *prog_aux); 3125 void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, u32 func_id); 3126 int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr); 3127 int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, struct bpf_prog *old_prog); 3128 void bpf_dev_bound_netdev_unregister(struct net_device *dev); 3129 3130 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 3131 { 3132 return aux->dev_bound; 3133 } 3134 3135 static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux) 3136 { 3137 return aux->offload_requested; 3138 } 3139 3140 bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs); 3141 3142 static inline bool bpf_map_is_offloaded(struct bpf_map *map) 3143 { 3144 return unlikely(map->ops == &bpf_map_offload_ops); 3145 } 3146 3147 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 3148 void bpf_map_offload_map_free(struct bpf_map *map); 3149 u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map); 3150 int bpf_prog_test_run_syscall(struct bpf_prog *prog, 3151 const union bpf_attr *kattr, 3152 union bpf_attr __user *uattr); 3153 3154 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 3155 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 3156 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 3157 int sock_map_bpf_prog_query(const union bpf_attr *attr, 3158 union bpf_attr __user *uattr); 3159 int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog); 3160 3161 void sock_map_unhash(struct sock *sk); 3162 void sock_map_destroy(struct sock *sk); 3163 void sock_map_close(struct sock *sk, long timeout); 3164 #else 3165 static inline int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log, 3166 struct bpf_prog_aux *prog_aux) 3167 { 3168 return -EOPNOTSUPP; 3169 } 3170 3171 static inline void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, 3172 u32 func_id) 3173 { 3174 return NULL; 3175 } 3176 3177 static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog, 3178 union bpf_attr *attr) 3179 { 3180 return -EOPNOTSUPP; 3181 } 3182 3183 static inline int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, 3184 struct bpf_prog *old_prog) 3185 { 3186 return -EOPNOTSUPP; 3187 } 3188 3189 static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev) 3190 { 3191 } 3192 3193 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 3194 { 3195 return false; 3196 } 3197 3198 static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux) 3199 { 3200 return false; 3201 } 3202 3203 static inline bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs) 3204 { 3205 return false; 3206 } 3207 3208 static inline bool bpf_map_is_offloaded(struct bpf_map *map) 3209 { 3210 return false; 3211 } 3212 3213 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 3214 { 3215 return ERR_PTR(-EOPNOTSUPP); 3216 } 3217 3218 static inline void bpf_map_offload_map_free(struct bpf_map *map) 3219 { 3220 } 3221 3222 static inline u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map) 3223 { 3224 return 0; 3225 } 3226 3227 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 3228 const union bpf_attr *kattr, 3229 union bpf_attr __user *uattr) 3230 { 3231 return -ENOTSUPP; 3232 } 3233 3234 #ifdef CONFIG_BPF_SYSCALL 3235 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 3236 struct bpf_prog *prog) 3237 { 3238 return -EINVAL; 3239 } 3240 3241 static inline int sock_map_prog_detach(const union bpf_attr *attr, 3242 enum bpf_prog_type ptype) 3243 { 3244 return -EOPNOTSUPP; 3245 } 3246 3247 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 3248 u64 flags) 3249 { 3250 return -EOPNOTSUPP; 3251 } 3252 3253 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr, 3254 union bpf_attr __user *uattr) 3255 { 3256 return -EINVAL; 3257 } 3258 3259 static inline int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog) 3260 { 3261 return -EOPNOTSUPP; 3262 } 3263 #endif /* CONFIG_BPF_SYSCALL */ 3264 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 3265 3266 static __always_inline void 3267 bpf_prog_inc_misses_counters(const struct bpf_prog_array *array) 3268 { 3269 const struct bpf_prog_array_item *item; 3270 struct bpf_prog *prog; 3271 3272 if (unlikely(!array)) 3273 return; 3274 3275 item = &array->items[0]; 3276 while ((prog = READ_ONCE(item->prog))) { 3277 bpf_prog_inc_misses_counter(prog); 3278 item++; 3279 } 3280 } 3281 3282 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 3283 void bpf_sk_reuseport_detach(struct sock *sk); 3284 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 3285 void *value); 3286 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 3287 void *value, u64 map_flags); 3288 #else 3289 static inline void bpf_sk_reuseport_detach(struct sock *sk) 3290 { 3291 } 3292 3293 #ifdef CONFIG_BPF_SYSCALL 3294 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 3295 void *key, void *value) 3296 { 3297 return -EOPNOTSUPP; 3298 } 3299 3300 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 3301 void *key, void *value, 3302 u64 map_flags) 3303 { 3304 return -EOPNOTSUPP; 3305 } 3306 #endif /* CONFIG_BPF_SYSCALL */ 3307 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 3308 3309 /* verifier prototypes for helper functions called from eBPF programs */ 3310 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 3311 extern const struct bpf_func_proto bpf_map_update_elem_proto; 3312 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 3313 extern const struct bpf_func_proto bpf_map_push_elem_proto; 3314 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 3315 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 3316 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto; 3317 3318 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 3319 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 3320 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 3321 extern const struct bpf_func_proto bpf_tail_call_proto; 3322 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 3323 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 3324 extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto; 3325 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 3326 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 3327 extern const struct bpf_func_proto bpf_get_current_comm_proto; 3328 extern const struct bpf_func_proto bpf_get_stackid_proto; 3329 extern const struct bpf_func_proto bpf_get_stack_proto; 3330 extern const struct bpf_func_proto bpf_get_stack_sleepable_proto; 3331 extern const struct bpf_func_proto bpf_get_task_stack_proto; 3332 extern const struct bpf_func_proto bpf_get_task_stack_sleepable_proto; 3333 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 3334 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 3335 extern const struct bpf_func_proto bpf_sock_map_update_proto; 3336 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 3337 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 3338 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 3339 extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto; 3340 extern const struct bpf_func_proto bpf_current_task_under_cgroup_proto; 3341 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 3342 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 3343 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 3344 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 3345 extern const struct bpf_func_proto bpf_spin_lock_proto; 3346 extern const struct bpf_func_proto bpf_spin_unlock_proto; 3347 extern const struct bpf_func_proto bpf_get_local_storage_proto; 3348 extern const struct bpf_func_proto bpf_strtol_proto; 3349 extern const struct bpf_func_proto bpf_strtoul_proto; 3350 extern const struct bpf_func_proto bpf_tcp_sock_proto; 3351 extern const struct bpf_func_proto bpf_jiffies64_proto; 3352 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 3353 extern const struct bpf_func_proto bpf_event_output_data_proto; 3354 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 3355 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 3356 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 3357 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 3358 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 3359 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto; 3360 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto; 3361 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto; 3362 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 3363 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 3364 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 3365 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 3366 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 3367 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto; 3368 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto; 3369 extern const struct bpf_func_proto bpf_copy_from_user_proto; 3370 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 3371 extern const struct bpf_func_proto bpf_snprintf_proto; 3372 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 3373 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 3374 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 3375 extern const struct bpf_func_proto bpf_sock_from_file_proto; 3376 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 3377 extern const struct bpf_func_proto bpf_task_storage_get_recur_proto; 3378 extern const struct bpf_func_proto bpf_task_storage_get_proto; 3379 extern const struct bpf_func_proto bpf_task_storage_delete_recur_proto; 3380 extern const struct bpf_func_proto bpf_task_storage_delete_proto; 3381 extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 3382 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 3383 extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 3384 extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 3385 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto; 3386 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto; 3387 extern const struct bpf_func_proto bpf_find_vma_proto; 3388 extern const struct bpf_func_proto bpf_loop_proto; 3389 extern const struct bpf_func_proto bpf_copy_from_user_task_proto; 3390 extern const struct bpf_func_proto bpf_set_retval_proto; 3391 extern const struct bpf_func_proto bpf_get_retval_proto; 3392 extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto; 3393 extern const struct bpf_func_proto bpf_cgrp_storage_get_proto; 3394 extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto; 3395 3396 const struct bpf_func_proto *tracing_prog_func_proto( 3397 enum bpf_func_id func_id, const struct bpf_prog *prog); 3398 3399 /* Shared helpers among cBPF and eBPF. */ 3400 void bpf_user_rnd_init_once(void); 3401 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 3402 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 3403 3404 #if defined(CONFIG_NET) 3405 bool bpf_sock_common_is_valid_access(int off, int size, 3406 enum bpf_access_type type, 3407 struct bpf_insn_access_aux *info); 3408 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 3409 struct bpf_insn_access_aux *info); 3410 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 3411 const struct bpf_insn *si, 3412 struct bpf_insn *insn_buf, 3413 struct bpf_prog *prog, 3414 u32 *target_size); 3415 int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags, 3416 struct bpf_dynptr *ptr); 3417 #else 3418 static inline bool bpf_sock_common_is_valid_access(int off, int size, 3419 enum bpf_access_type type, 3420 struct bpf_insn_access_aux *info) 3421 { 3422 return false; 3423 } 3424 static inline bool bpf_sock_is_valid_access(int off, int size, 3425 enum bpf_access_type type, 3426 struct bpf_insn_access_aux *info) 3427 { 3428 return false; 3429 } 3430 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 3431 const struct bpf_insn *si, 3432 struct bpf_insn *insn_buf, 3433 struct bpf_prog *prog, 3434 u32 *target_size) 3435 { 3436 return 0; 3437 } 3438 static inline int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags, 3439 struct bpf_dynptr *ptr) 3440 { 3441 return -EOPNOTSUPP; 3442 } 3443 #endif 3444 3445 #ifdef CONFIG_INET 3446 struct sk_reuseport_kern { 3447 struct sk_buff *skb; 3448 struct sock *sk; 3449 struct sock *selected_sk; 3450 struct sock *migrating_sk; 3451 void *data_end; 3452 u32 hash; 3453 u32 reuseport_id; 3454 bool bind_inany; 3455 }; 3456 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 3457 struct bpf_insn_access_aux *info); 3458 3459 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 3460 const struct bpf_insn *si, 3461 struct bpf_insn *insn_buf, 3462 struct bpf_prog *prog, 3463 u32 *target_size); 3464 3465 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 3466 struct bpf_insn_access_aux *info); 3467 3468 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 3469 const struct bpf_insn *si, 3470 struct bpf_insn *insn_buf, 3471 struct bpf_prog *prog, 3472 u32 *target_size); 3473 #else 3474 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 3475 enum bpf_access_type type, 3476 struct bpf_insn_access_aux *info) 3477 { 3478 return false; 3479 } 3480 3481 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 3482 const struct bpf_insn *si, 3483 struct bpf_insn *insn_buf, 3484 struct bpf_prog *prog, 3485 u32 *target_size) 3486 { 3487 return 0; 3488 } 3489 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 3490 enum bpf_access_type type, 3491 struct bpf_insn_access_aux *info) 3492 { 3493 return false; 3494 } 3495 3496 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 3497 const struct bpf_insn *si, 3498 struct bpf_insn *insn_buf, 3499 struct bpf_prog *prog, 3500 u32 *target_size) 3501 { 3502 return 0; 3503 } 3504 #endif /* CONFIG_INET */ 3505 3506 enum bpf_text_poke_type { 3507 BPF_MOD_CALL, 3508 BPF_MOD_JUMP, 3509 }; 3510 3511 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 3512 void *addr1, void *addr2); 3513 3514 void bpf_arch_poke_desc_update(struct bpf_jit_poke_descriptor *poke, 3515 struct bpf_prog *new, struct bpf_prog *old); 3516 3517 void *bpf_arch_text_copy(void *dst, void *src, size_t len); 3518 int bpf_arch_text_invalidate(void *dst, size_t len); 3519 3520 struct btf_id_set; 3521 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 3522 3523 #define MAX_BPRINTF_VARARGS 12 3524 #define MAX_BPRINTF_BUF 1024 3525 3526 struct bpf_bprintf_data { 3527 u32 *bin_args; 3528 char *buf; 3529 bool get_bin_args; 3530 bool get_buf; 3531 }; 3532 3533 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args, 3534 u32 num_args, struct bpf_bprintf_data *data); 3535 void bpf_bprintf_cleanup(struct bpf_bprintf_data *data); 3536 3537 #ifdef CONFIG_BPF_LSM 3538 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype); 3539 void bpf_cgroup_atype_put(int cgroup_atype); 3540 #else 3541 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {} 3542 static inline void bpf_cgroup_atype_put(int cgroup_atype) {} 3543 #endif /* CONFIG_BPF_LSM */ 3544 3545 struct key; 3546 3547 #ifdef CONFIG_KEYS 3548 struct bpf_key { 3549 struct key *key; 3550 bool has_ref; 3551 }; 3552 #endif /* CONFIG_KEYS */ 3553 3554 static inline bool type_is_alloc(u32 type) 3555 { 3556 return type & MEM_ALLOC; 3557 } 3558 3559 static inline gfp_t bpf_memcg_flags(gfp_t flags) 3560 { 3561 if (memcg_bpf_enabled()) 3562 return flags | __GFP_ACCOUNT; 3563 return flags; 3564 } 3565 3566 static inline bool bpf_is_subprog(const struct bpf_prog *prog) 3567 { 3568 return prog->aux->func_idx != 0; 3569 } 3570 3571 #endif /* _LINUX_BPF_H */ 3572