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 ctx_arg_info_size; 1535 u32 max_rdonly_access; 1536 u32 max_rdwr_access; 1537 struct btf *attach_btf; 1538 struct bpf_ctx_arg_aux *ctx_arg_info; 1539 void __percpu *priv_stack_ptr; 1540 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 1541 struct bpf_prog *dst_prog; 1542 struct bpf_trampoline *dst_trampoline; 1543 enum bpf_prog_type saved_dst_prog_type; 1544 enum bpf_attach_type saved_dst_attach_type; 1545 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 1546 bool dev_bound; /* Program is bound to the netdev. */ 1547 bool offload_requested; /* Program is bound and offloaded to the netdev. */ 1548 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 1549 bool attach_tracing_prog; /* true if tracing another tracing program */ 1550 bool func_proto_unreliable; 1551 bool tail_call_reachable; 1552 bool xdp_has_frags; 1553 bool exception_cb; 1554 bool exception_boundary; 1555 bool is_extended; /* true if extended by freplace program */ 1556 bool jits_use_priv_stack; 1557 bool priv_stack_requested; 1558 bool changes_pkt_data; 1559 bool might_sleep; 1560 u64 prog_array_member_cnt; /* counts how many times as member of prog_array */ 1561 struct mutex ext_mutex; /* mutex for is_extended and prog_array_member_cnt */ 1562 struct bpf_arena *arena; 1563 void (*recursion_detected)(struct bpf_prog *prog); /* callback if recursion is detected */ 1564 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 1565 const struct btf_type *attach_func_proto; 1566 /* function name for valid attach_btf_id */ 1567 const char *attach_func_name; 1568 struct bpf_prog **func; 1569 void *jit_data; /* JIT specific data. arch dependent */ 1570 struct bpf_jit_poke_descriptor *poke_tab; 1571 struct bpf_kfunc_desc_tab *kfunc_tab; 1572 struct bpf_kfunc_btf_tab *kfunc_btf_tab; 1573 u32 size_poke_tab; 1574 #ifdef CONFIG_FINEIBT 1575 struct bpf_ksym ksym_prefix; 1576 #endif 1577 struct bpf_ksym ksym; 1578 const struct bpf_prog_ops *ops; 1579 struct bpf_map **used_maps; 1580 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 1581 struct btf_mod_pair *used_btfs; 1582 struct bpf_prog *prog; 1583 struct user_struct *user; 1584 u64 load_time; /* ns since boottime */ 1585 u32 verified_insns; 1586 int cgroup_atype; /* enum cgroup_bpf_attach_type */ 1587 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1588 char name[BPF_OBJ_NAME_LEN]; 1589 u64 (*bpf_exception_cb)(u64 cookie, u64 sp, u64 bp, u64, u64); 1590 #ifdef CONFIG_SECURITY 1591 void *security; 1592 #endif 1593 struct bpf_token *token; 1594 struct bpf_prog_offload *offload; 1595 struct btf *btf; 1596 struct bpf_func_info *func_info; 1597 struct bpf_func_info_aux *func_info_aux; 1598 /* bpf_line_info loaded from userspace. linfo->insn_off 1599 * has the xlated insn offset. 1600 * Both the main and sub prog share the same linfo. 1601 * The subprog can access its first linfo by 1602 * using the linfo_idx. 1603 */ 1604 struct bpf_line_info *linfo; 1605 /* jited_linfo is the jited addr of the linfo. It has a 1606 * one to one mapping to linfo: 1607 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 1608 * Both the main and sub prog share the same jited_linfo. 1609 * The subprog can access its first jited_linfo by 1610 * using the linfo_idx. 1611 */ 1612 void **jited_linfo; 1613 u32 func_info_cnt; 1614 u32 nr_linfo; 1615 /* subprog can use linfo_idx to access its first linfo and 1616 * jited_linfo. 1617 * main prog always has linfo_idx == 0 1618 */ 1619 u32 linfo_idx; 1620 struct module *mod; 1621 u32 num_exentries; 1622 struct exception_table_entry *extable; 1623 union { 1624 struct work_struct work; 1625 struct rcu_head rcu; 1626 }; 1627 }; 1628 1629 struct bpf_prog { 1630 u16 pages; /* Number of allocated pages */ 1631 u16 jited:1, /* Is our filter JIT'ed? */ 1632 jit_requested:1,/* archs need to JIT the prog */ 1633 gpl_compatible:1, /* Is filter GPL compatible? */ 1634 cb_access:1, /* Is control block accessed? */ 1635 dst_needed:1, /* Do we need dst entry? */ 1636 blinding_requested:1, /* needs constant blinding */ 1637 blinded:1, /* Was blinded */ 1638 is_func:1, /* program is a bpf function */ 1639 kprobe_override:1, /* Do we override a kprobe? */ 1640 has_callchain_buf:1, /* callchain buffer allocated? */ 1641 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */ 1642 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */ 1643 call_get_func_ip:1, /* Do we call get_func_ip() */ 1644 tstamp_type_access:1, /* Accessed __sk_buff->tstamp_type */ 1645 sleepable:1; /* BPF program is sleepable */ 1646 enum bpf_prog_type type; /* Type of BPF program */ 1647 enum bpf_attach_type expected_attach_type; /* For some prog types */ 1648 u32 len; /* Number of filter blocks */ 1649 u32 jited_len; /* Size of jited insns in bytes */ 1650 u8 tag[BPF_TAG_SIZE]; 1651 struct bpf_prog_stats __percpu *stats; 1652 int __percpu *active; 1653 unsigned int (*bpf_func)(const void *ctx, 1654 const struct bpf_insn *insn); 1655 struct bpf_prog_aux *aux; /* Auxiliary fields */ 1656 struct sock_fprog_kern *orig_prog; /* Original BPF program */ 1657 /* Instructions for interpreter */ 1658 union { 1659 DECLARE_FLEX_ARRAY(struct sock_filter, insns); 1660 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi); 1661 }; 1662 }; 1663 1664 struct bpf_array_aux { 1665 /* Programs with direct jumps into programs part of this array. */ 1666 struct list_head poke_progs; 1667 struct bpf_map *map; 1668 struct mutex poke_mutex; 1669 struct work_struct work; 1670 }; 1671 1672 struct bpf_link { 1673 atomic64_t refcnt; 1674 u32 id; 1675 enum bpf_link_type type; 1676 const struct bpf_link_ops *ops; 1677 struct bpf_prog *prog; 1678 /* whether BPF link itself has "sleepable" semantics, which can differ 1679 * from underlying BPF program having a "sleepable" semantics, as BPF 1680 * link's semantics is determined by target attach hook 1681 */ 1682 bool sleepable; 1683 /* rcu is used before freeing, work can be used to schedule that 1684 * RCU-based freeing before that, so they never overlap 1685 */ 1686 union { 1687 struct rcu_head rcu; 1688 struct work_struct work; 1689 }; 1690 }; 1691 1692 struct bpf_link_ops { 1693 void (*release)(struct bpf_link *link); 1694 /* deallocate link resources callback, called without RCU grace period 1695 * waiting 1696 */ 1697 void (*dealloc)(struct bpf_link *link); 1698 /* deallocate link resources callback, called after RCU grace period; 1699 * if either the underlying BPF program is sleepable or BPF link's 1700 * target hook is sleepable, we'll go through tasks trace RCU GP and 1701 * then "classic" RCU GP; this need for chaining tasks trace and 1702 * classic RCU GPs is designated by setting bpf_link->sleepable flag 1703 */ 1704 void (*dealloc_deferred)(struct bpf_link *link); 1705 int (*detach)(struct bpf_link *link); 1706 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 1707 struct bpf_prog *old_prog); 1708 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 1709 int (*fill_link_info)(const struct bpf_link *link, 1710 struct bpf_link_info *info); 1711 int (*update_map)(struct bpf_link *link, struct bpf_map *new_map, 1712 struct bpf_map *old_map); 1713 __poll_t (*poll)(struct file *file, struct poll_table_struct *pts); 1714 }; 1715 1716 struct bpf_tramp_link { 1717 struct bpf_link link; 1718 struct hlist_node tramp_hlist; 1719 u64 cookie; 1720 }; 1721 1722 struct bpf_shim_tramp_link { 1723 struct bpf_tramp_link link; 1724 struct bpf_trampoline *trampoline; 1725 }; 1726 1727 struct bpf_tracing_link { 1728 struct bpf_tramp_link link; 1729 enum bpf_attach_type attach_type; 1730 struct bpf_trampoline *trampoline; 1731 struct bpf_prog *tgt_prog; 1732 }; 1733 1734 struct bpf_raw_tp_link { 1735 struct bpf_link link; 1736 struct bpf_raw_event_map *btp; 1737 u64 cookie; 1738 }; 1739 1740 struct bpf_link_primer { 1741 struct bpf_link *link; 1742 struct file *file; 1743 int fd; 1744 u32 id; 1745 }; 1746 1747 struct bpf_mount_opts { 1748 kuid_t uid; 1749 kgid_t gid; 1750 umode_t mode; 1751 1752 /* BPF token-related delegation options */ 1753 u64 delegate_cmds; 1754 u64 delegate_maps; 1755 u64 delegate_progs; 1756 u64 delegate_attachs; 1757 }; 1758 1759 struct bpf_token { 1760 struct work_struct work; 1761 atomic64_t refcnt; 1762 struct user_namespace *userns; 1763 u64 allowed_cmds; 1764 u64 allowed_maps; 1765 u64 allowed_progs; 1766 u64 allowed_attachs; 1767 #ifdef CONFIG_SECURITY 1768 void *security; 1769 #endif 1770 }; 1771 1772 struct bpf_struct_ops_value; 1773 struct btf_member; 1774 1775 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 1776 /** 1777 * struct bpf_struct_ops - A structure of callbacks allowing a subsystem to 1778 * define a BPF_MAP_TYPE_STRUCT_OPS map type composed 1779 * of BPF_PROG_TYPE_STRUCT_OPS progs. 1780 * @verifier_ops: A structure of callbacks that are invoked by the verifier 1781 * when determining whether the struct_ops progs in the 1782 * struct_ops map are valid. 1783 * @init: A callback that is invoked a single time, and before any other 1784 * callback, to initialize the structure. A nonzero return value means 1785 * the subsystem could not be initialized. 1786 * @check_member: When defined, a callback invoked by the verifier to allow 1787 * the subsystem to determine if an entry in the struct_ops map 1788 * is valid. A nonzero return value means that the map is 1789 * invalid and should be rejected by the verifier. 1790 * @init_member: A callback that is invoked for each member of the struct_ops 1791 * map to allow the subsystem to initialize the member. A nonzero 1792 * value means the member could not be initialized. This callback 1793 * is exclusive with the @type, @type_id, @value_type, and 1794 * @value_id fields. 1795 * @reg: A callback that is invoked when the struct_ops map has been 1796 * initialized and is being attached to. Zero means the struct_ops map 1797 * has been successfully registered and is live. A nonzero return value 1798 * means the struct_ops map could not be registered. 1799 * @unreg: A callback that is invoked when the struct_ops map should be 1800 * unregistered. 1801 * @update: A callback that is invoked when the live struct_ops map is being 1802 * updated to contain new values. This callback is only invoked when 1803 * the struct_ops map is loaded with BPF_F_LINK. If not defined, the 1804 * it is assumed that the struct_ops map cannot be updated. 1805 * @validate: A callback that is invoked after all of the members have been 1806 * initialized. This callback should perform static checks on the 1807 * map, meaning that it should either fail or succeed 1808 * deterministically. A struct_ops map that has been validated may 1809 * not necessarily succeed in being registered if the call to @reg 1810 * fails. For example, a valid struct_ops map may be loaded, but 1811 * then fail to be registered due to there being another active 1812 * struct_ops map on the system in the subsystem already. For this 1813 * reason, if this callback is not defined, the check is skipped as 1814 * the struct_ops map will have final verification performed in 1815 * @reg. 1816 * @type: BTF type. 1817 * @value_type: Value type. 1818 * @name: The name of the struct bpf_struct_ops object. 1819 * @func_models: Func models 1820 * @type_id: BTF type id. 1821 * @value_id: BTF value id. 1822 */ 1823 struct bpf_struct_ops { 1824 const struct bpf_verifier_ops *verifier_ops; 1825 int (*init)(struct btf *btf); 1826 int (*check_member)(const struct btf_type *t, 1827 const struct btf_member *member, 1828 const struct bpf_prog *prog); 1829 int (*init_member)(const struct btf_type *t, 1830 const struct btf_member *member, 1831 void *kdata, const void *udata); 1832 int (*reg)(void *kdata, struct bpf_link *link); 1833 void (*unreg)(void *kdata, struct bpf_link *link); 1834 int (*update)(void *kdata, void *old_kdata, struct bpf_link *link); 1835 int (*validate)(void *kdata); 1836 void *cfi_stubs; 1837 struct module *owner; 1838 const char *name; 1839 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 1840 }; 1841 1842 /* Every member of a struct_ops type has an instance even a member is not 1843 * an operator (function pointer). The "info" field will be assigned to 1844 * prog->aux->ctx_arg_info of BPF struct_ops programs to provide the 1845 * argument information required by the verifier to verify the program. 1846 * 1847 * btf_ctx_access() will lookup prog->aux->ctx_arg_info to find the 1848 * corresponding entry for an given argument. 1849 */ 1850 struct bpf_struct_ops_arg_info { 1851 struct bpf_ctx_arg_aux *info; 1852 u32 cnt; 1853 }; 1854 1855 struct bpf_struct_ops_desc { 1856 struct bpf_struct_ops *st_ops; 1857 1858 const struct btf_type *type; 1859 const struct btf_type *value_type; 1860 u32 type_id; 1861 u32 value_id; 1862 1863 /* Collection of argument information for each member */ 1864 struct bpf_struct_ops_arg_info *arg_info; 1865 }; 1866 1867 enum bpf_struct_ops_state { 1868 BPF_STRUCT_OPS_STATE_INIT, 1869 BPF_STRUCT_OPS_STATE_INUSE, 1870 BPF_STRUCT_OPS_STATE_TOBEFREE, 1871 BPF_STRUCT_OPS_STATE_READY, 1872 }; 1873 1874 struct bpf_struct_ops_common_value { 1875 refcount_t refcnt; 1876 enum bpf_struct_ops_state state; 1877 }; 1878 1879 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 1880 /* This macro helps developer to register a struct_ops type and generate 1881 * type information correctly. Developers should use this macro to register 1882 * a struct_ops type instead of calling __register_bpf_struct_ops() directly. 1883 */ 1884 #define register_bpf_struct_ops(st_ops, type) \ 1885 ({ \ 1886 struct bpf_struct_ops_##type { \ 1887 struct bpf_struct_ops_common_value common; \ 1888 struct type data ____cacheline_aligned_in_smp; \ 1889 }; \ 1890 BTF_TYPE_EMIT(struct bpf_struct_ops_##type); \ 1891 __register_bpf_struct_ops(st_ops); \ 1892 }) 1893 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 1894 bool bpf_struct_ops_get(const void *kdata); 1895 void bpf_struct_ops_put(const void *kdata); 1896 int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff); 1897 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 1898 void *value); 1899 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks, 1900 struct bpf_tramp_link *link, 1901 const struct btf_func_model *model, 1902 void *stub_func, 1903 void **image, u32 *image_off, 1904 bool allow_alloc); 1905 void bpf_struct_ops_image_free(void *image); 1906 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1907 { 1908 if (owner == BPF_MODULE_OWNER) 1909 return bpf_struct_ops_get(data); 1910 else 1911 return try_module_get(owner); 1912 } 1913 static inline void bpf_module_put(const void *data, struct module *owner) 1914 { 1915 if (owner == BPF_MODULE_OWNER) 1916 bpf_struct_ops_put(data); 1917 else 1918 module_put(owner); 1919 } 1920 int bpf_struct_ops_link_create(union bpf_attr *attr); 1921 1922 #ifdef CONFIG_NET 1923 /* Define it here to avoid the use of forward declaration */ 1924 struct bpf_dummy_ops_state { 1925 int val; 1926 }; 1927 1928 struct bpf_dummy_ops { 1929 int (*test_1)(struct bpf_dummy_ops_state *cb); 1930 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2, 1931 char a3, unsigned long a4); 1932 int (*test_sleepable)(struct bpf_dummy_ops_state *cb); 1933 }; 1934 1935 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr, 1936 union bpf_attr __user *uattr); 1937 #endif 1938 int bpf_struct_ops_desc_init(struct bpf_struct_ops_desc *st_ops_desc, 1939 struct btf *btf, 1940 struct bpf_verifier_log *log); 1941 void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map); 1942 void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc); 1943 #else 1944 #define register_bpf_struct_ops(st_ops, type) ({ (void *)(st_ops); 0; }) 1945 static inline bool bpf_try_module_get(const void *data, struct module *owner) 1946 { 1947 return try_module_get(owner); 1948 } 1949 static inline void bpf_module_put(const void *data, struct module *owner) 1950 { 1951 module_put(owner); 1952 } 1953 static inline int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff) 1954 { 1955 return -ENOTSUPP; 1956 } 1957 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 1958 void *key, 1959 void *value) 1960 { 1961 return -EINVAL; 1962 } 1963 static inline int bpf_struct_ops_link_create(union bpf_attr *attr) 1964 { 1965 return -EOPNOTSUPP; 1966 } 1967 static inline void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map) 1968 { 1969 } 1970 1971 static inline void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc) 1972 { 1973 } 1974 1975 #endif 1976 1977 int bpf_prog_ctx_arg_info_init(struct bpf_prog *prog, 1978 const struct bpf_ctx_arg_aux *info, u32 cnt); 1979 1980 #if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM) 1981 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 1982 int cgroup_atype); 1983 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog); 1984 #else 1985 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 1986 int cgroup_atype) 1987 { 1988 return -EOPNOTSUPP; 1989 } 1990 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog) 1991 { 1992 } 1993 #endif 1994 1995 struct bpf_array { 1996 struct bpf_map map; 1997 u32 elem_size; 1998 u32 index_mask; 1999 struct bpf_array_aux *aux; 2000 union { 2001 DECLARE_FLEX_ARRAY(char, value) __aligned(8); 2002 DECLARE_FLEX_ARRAY(void *, ptrs) __aligned(8); 2003 DECLARE_FLEX_ARRAY(void __percpu *, pptrs) __aligned(8); 2004 }; 2005 }; 2006 2007 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 2008 #define MAX_TAIL_CALL_CNT 33 2009 2010 /* Maximum number of loops for bpf_loop and bpf_iter_num. 2011 * It's enum to expose it (and thus make it discoverable) through BTF. 2012 */ 2013 enum { 2014 BPF_MAX_LOOPS = 8 * 1024 * 1024, 2015 BPF_MAX_TIMED_LOOPS = 0xffff, 2016 }; 2017 2018 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 2019 BPF_F_RDONLY_PROG | \ 2020 BPF_F_WRONLY | \ 2021 BPF_F_WRONLY_PROG) 2022 2023 #define BPF_MAP_CAN_READ BIT(0) 2024 #define BPF_MAP_CAN_WRITE BIT(1) 2025 2026 /* Maximum number of user-producer ring buffer samples that can be drained in 2027 * a call to bpf_user_ringbuf_drain(). 2028 */ 2029 #define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024) 2030 2031 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 2032 { 2033 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 2034 2035 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 2036 * not possible. 2037 */ 2038 if (access_flags & BPF_F_RDONLY_PROG) 2039 return BPF_MAP_CAN_READ; 2040 else if (access_flags & BPF_F_WRONLY_PROG) 2041 return BPF_MAP_CAN_WRITE; 2042 else 2043 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 2044 } 2045 2046 static inline bool bpf_map_flags_access_ok(u32 access_flags) 2047 { 2048 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 2049 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 2050 } 2051 2052 struct bpf_event_entry { 2053 struct perf_event *event; 2054 struct file *perf_file; 2055 struct file *map_file; 2056 struct rcu_head rcu; 2057 }; 2058 2059 static inline bool map_type_contains_progs(struct bpf_map *map) 2060 { 2061 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY || 2062 map->map_type == BPF_MAP_TYPE_DEVMAP || 2063 map->map_type == BPF_MAP_TYPE_CPUMAP; 2064 } 2065 2066 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp); 2067 int bpf_prog_calc_tag(struct bpf_prog *fp); 2068 2069 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 2070 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void); 2071 2072 const struct bpf_func_proto *bpf_get_perf_event_read_value_proto(void); 2073 2074 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 2075 unsigned long off, unsigned long len); 2076 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 2077 const struct bpf_insn *src, 2078 struct bpf_insn *dst, 2079 struct bpf_prog *prog, 2080 u32 *target_size); 2081 2082 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 2083 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 2084 2085 /* an array of programs to be executed under rcu_lock. 2086 * 2087 * Typical usage: 2088 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run); 2089 * 2090 * the structure returned by bpf_prog_array_alloc() should be populated 2091 * with program pointers and the last pointer must be NULL. 2092 * The user has to keep refcnt on the program and make sure the program 2093 * is removed from the array before bpf_prog_put(). 2094 * The 'struct bpf_prog_array *' should only be replaced with xchg() 2095 * since other cpus are walking the array of pointers in parallel. 2096 */ 2097 struct bpf_prog_array_item { 2098 struct bpf_prog *prog; 2099 union { 2100 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 2101 u64 bpf_cookie; 2102 }; 2103 }; 2104 2105 struct bpf_prog_array { 2106 struct rcu_head rcu; 2107 struct bpf_prog_array_item items[]; 2108 }; 2109 2110 struct bpf_empty_prog_array { 2111 struct bpf_prog_array hdr; 2112 struct bpf_prog *null_prog; 2113 }; 2114 2115 /* to avoid allocating empty bpf_prog_array for cgroups that 2116 * don't have bpf program attached use one global 'bpf_empty_prog_array' 2117 * It will not be modified the caller of bpf_prog_array_alloc() 2118 * (since caller requested prog_cnt == 0) 2119 * that pointer should be 'freed' by bpf_prog_array_free() 2120 */ 2121 extern struct bpf_empty_prog_array bpf_empty_prog_array; 2122 2123 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 2124 void bpf_prog_array_free(struct bpf_prog_array *progs); 2125 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */ 2126 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs); 2127 int bpf_prog_array_length(struct bpf_prog_array *progs); 2128 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 2129 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 2130 __u32 __user *prog_ids, u32 cnt); 2131 2132 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 2133 struct bpf_prog *old_prog); 2134 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 2135 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 2136 struct bpf_prog *prog); 2137 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 2138 u32 *prog_ids, u32 request_cnt, 2139 u32 *prog_cnt); 2140 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 2141 struct bpf_prog *exclude_prog, 2142 struct bpf_prog *include_prog, 2143 u64 bpf_cookie, 2144 struct bpf_prog_array **new_array); 2145 2146 struct bpf_run_ctx {}; 2147 2148 struct bpf_cg_run_ctx { 2149 struct bpf_run_ctx run_ctx; 2150 const struct bpf_prog_array_item *prog_item; 2151 int retval; 2152 }; 2153 2154 struct bpf_trace_run_ctx { 2155 struct bpf_run_ctx run_ctx; 2156 u64 bpf_cookie; 2157 bool is_uprobe; 2158 }; 2159 2160 struct bpf_tramp_run_ctx { 2161 struct bpf_run_ctx run_ctx; 2162 u64 bpf_cookie; 2163 struct bpf_run_ctx *saved_run_ctx; 2164 }; 2165 2166 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 2167 { 2168 struct bpf_run_ctx *old_ctx = NULL; 2169 2170 #ifdef CONFIG_BPF_SYSCALL 2171 old_ctx = current->bpf_ctx; 2172 current->bpf_ctx = new_ctx; 2173 #endif 2174 return old_ctx; 2175 } 2176 2177 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 2178 { 2179 #ifdef CONFIG_BPF_SYSCALL 2180 current->bpf_ctx = old_ctx; 2181 #endif 2182 } 2183 2184 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 2185 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 2186 /* BPF program asks to set CN on the packet. */ 2187 #define BPF_RET_SET_CN (1 << 0) 2188 2189 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 2190 2191 static __always_inline u32 2192 bpf_prog_run_array(const struct bpf_prog_array *array, 2193 const void *ctx, bpf_prog_run_fn run_prog) 2194 { 2195 const struct bpf_prog_array_item *item; 2196 const struct bpf_prog *prog; 2197 struct bpf_run_ctx *old_run_ctx; 2198 struct bpf_trace_run_ctx run_ctx; 2199 u32 ret = 1; 2200 2201 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held"); 2202 2203 if (unlikely(!array)) 2204 return ret; 2205 2206 run_ctx.is_uprobe = false; 2207 2208 migrate_disable(); 2209 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 2210 item = &array->items[0]; 2211 while ((prog = READ_ONCE(item->prog))) { 2212 run_ctx.bpf_cookie = item->bpf_cookie; 2213 ret &= run_prog(prog, ctx); 2214 item++; 2215 } 2216 bpf_reset_run_ctx(old_run_ctx); 2217 migrate_enable(); 2218 return ret; 2219 } 2220 2221 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs: 2222 * 2223 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array 2224 * overall. As a result, we must use the bpf_prog_array_free_sleepable 2225 * in order to use the tasks_trace rcu grace period. 2226 * 2227 * When a non-sleepable program is inside the array, we take the rcu read 2228 * section and disable preemption for that program alone, so it can access 2229 * rcu-protected dynamically sized maps. 2230 */ 2231 static __always_inline u32 2232 bpf_prog_run_array_uprobe(const struct bpf_prog_array *array, 2233 const void *ctx, bpf_prog_run_fn run_prog) 2234 { 2235 const struct bpf_prog_array_item *item; 2236 const struct bpf_prog *prog; 2237 struct bpf_run_ctx *old_run_ctx; 2238 struct bpf_trace_run_ctx run_ctx; 2239 u32 ret = 1; 2240 2241 might_fault(); 2242 RCU_LOCKDEP_WARN(!rcu_read_lock_trace_held(), "no rcu lock held"); 2243 2244 if (unlikely(!array)) 2245 return ret; 2246 2247 migrate_disable(); 2248 2249 run_ctx.is_uprobe = true; 2250 2251 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 2252 item = &array->items[0]; 2253 while ((prog = READ_ONCE(item->prog))) { 2254 if (!prog->sleepable) 2255 rcu_read_lock(); 2256 2257 run_ctx.bpf_cookie = item->bpf_cookie; 2258 ret &= run_prog(prog, ctx); 2259 item++; 2260 2261 if (!prog->sleepable) 2262 rcu_read_unlock(); 2263 } 2264 bpf_reset_run_ctx(old_run_ctx); 2265 migrate_enable(); 2266 return ret; 2267 } 2268 2269 #ifdef CONFIG_BPF_SYSCALL 2270 DECLARE_PER_CPU(int, bpf_prog_active); 2271 extern struct mutex bpf_stats_enabled_mutex; 2272 2273 /* 2274 * Block execution of BPF programs attached to instrumentation (perf, 2275 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 2276 * these events can happen inside a region which holds a map bucket lock 2277 * and can deadlock on it. 2278 */ 2279 static inline void bpf_disable_instrumentation(void) 2280 { 2281 migrate_disable(); 2282 this_cpu_inc(bpf_prog_active); 2283 } 2284 2285 static inline void bpf_enable_instrumentation(void) 2286 { 2287 this_cpu_dec(bpf_prog_active); 2288 migrate_enable(); 2289 } 2290 2291 extern const struct super_operations bpf_super_ops; 2292 extern const struct file_operations bpf_map_fops; 2293 extern const struct file_operations bpf_prog_fops; 2294 extern const struct file_operations bpf_iter_fops; 2295 2296 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 2297 extern const struct bpf_prog_ops _name ## _prog_ops; \ 2298 extern const struct bpf_verifier_ops _name ## _verifier_ops; 2299 #define BPF_MAP_TYPE(_id, _ops) \ 2300 extern const struct bpf_map_ops _ops; 2301 #define BPF_LINK_TYPE(_id, _name) 2302 #include <linux/bpf_types.h> 2303 #undef BPF_PROG_TYPE 2304 #undef BPF_MAP_TYPE 2305 #undef BPF_LINK_TYPE 2306 2307 extern const struct bpf_prog_ops bpf_offload_prog_ops; 2308 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 2309 extern const struct bpf_verifier_ops xdp_analyzer_ops; 2310 2311 struct bpf_prog *bpf_prog_get(u32 ufd); 2312 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 2313 bool attach_drv); 2314 void bpf_prog_add(struct bpf_prog *prog, int i); 2315 void bpf_prog_sub(struct bpf_prog *prog, int i); 2316 void bpf_prog_inc(struct bpf_prog *prog); 2317 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 2318 void bpf_prog_put(struct bpf_prog *prog); 2319 2320 void bpf_prog_free_id(struct bpf_prog *prog); 2321 void bpf_map_free_id(struct bpf_map *map); 2322 2323 struct btf_field *btf_record_find(const struct btf_record *rec, 2324 u32 offset, u32 field_mask); 2325 void btf_record_free(struct btf_record *rec); 2326 void bpf_map_free_record(struct bpf_map *map); 2327 struct btf_record *btf_record_dup(const struct btf_record *rec); 2328 bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b); 2329 void bpf_obj_free_timer(const struct btf_record *rec, void *obj); 2330 void bpf_obj_free_workqueue(const struct btf_record *rec, void *obj); 2331 void bpf_obj_free_fields(const struct btf_record *rec, void *obj); 2332 void __bpf_obj_drop_impl(void *p, const struct btf_record *rec, bool percpu); 2333 2334 struct bpf_map *bpf_map_get(u32 ufd); 2335 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 2336 2337 /* 2338 * The __bpf_map_get() and __btf_get_by_fd() functions parse a file 2339 * descriptor and return a corresponding map or btf object. 2340 * Their names are double underscored to emphasize the fact that they 2341 * do not increase refcnt. To also increase refcnt use corresponding 2342 * bpf_map_get() and btf_get_by_fd() functions. 2343 */ 2344 2345 static inline struct bpf_map *__bpf_map_get(struct fd f) 2346 { 2347 if (fd_empty(f)) 2348 return ERR_PTR(-EBADF); 2349 if (unlikely(fd_file(f)->f_op != &bpf_map_fops)) 2350 return ERR_PTR(-EINVAL); 2351 return fd_file(f)->private_data; 2352 } 2353 2354 static inline struct btf *__btf_get_by_fd(struct fd f) 2355 { 2356 if (fd_empty(f)) 2357 return ERR_PTR(-EBADF); 2358 if (unlikely(fd_file(f)->f_op != &btf_fops)) 2359 return ERR_PTR(-EINVAL); 2360 return fd_file(f)->private_data; 2361 } 2362 2363 void bpf_map_inc(struct bpf_map *map); 2364 void bpf_map_inc_with_uref(struct bpf_map *map); 2365 struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref); 2366 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 2367 void bpf_map_put_with_uref(struct bpf_map *map); 2368 void bpf_map_put(struct bpf_map *map); 2369 void *bpf_map_area_alloc(u64 size, int numa_node); 2370 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 2371 void bpf_map_area_free(void *base); 2372 bool bpf_map_write_active(const struct bpf_map *map); 2373 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 2374 int generic_map_lookup_batch(struct bpf_map *map, 2375 const union bpf_attr *attr, 2376 union bpf_attr __user *uattr); 2377 int generic_map_update_batch(struct bpf_map *map, struct file *map_file, 2378 const union bpf_attr *attr, 2379 union bpf_attr __user *uattr); 2380 int generic_map_delete_batch(struct bpf_map *map, 2381 const union bpf_attr *attr, 2382 union bpf_attr __user *uattr); 2383 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 2384 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 2385 2386 int bpf_map_alloc_pages(const struct bpf_map *map, gfp_t gfp, int nid, 2387 unsigned long nr_pages, struct page **page_array); 2388 #ifdef CONFIG_MEMCG 2389 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 2390 int node); 2391 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 2392 void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size, 2393 gfp_t flags); 2394 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 2395 size_t align, gfp_t flags); 2396 #else 2397 /* 2398 * These specialized allocators have to be macros for their allocations to be 2399 * accounted separately (to have separate alloc_tag). 2400 */ 2401 #define bpf_map_kmalloc_node(_map, _size, _flags, _node) \ 2402 kmalloc_node(_size, _flags, _node) 2403 #define bpf_map_kzalloc(_map, _size, _flags) \ 2404 kzalloc(_size, _flags) 2405 #define bpf_map_kvcalloc(_map, _n, _size, _flags) \ 2406 kvcalloc(_n, _size, _flags) 2407 #define bpf_map_alloc_percpu(_map, _size, _align, _flags) \ 2408 __alloc_percpu_gfp(_size, _align, _flags) 2409 #endif 2410 2411 static inline int 2412 bpf_map_init_elem_count(struct bpf_map *map) 2413 { 2414 size_t size = sizeof(*map->elem_count), align = size; 2415 gfp_t flags = GFP_USER | __GFP_NOWARN; 2416 2417 map->elem_count = bpf_map_alloc_percpu(map, size, align, flags); 2418 if (!map->elem_count) 2419 return -ENOMEM; 2420 2421 return 0; 2422 } 2423 2424 static inline void 2425 bpf_map_free_elem_count(struct bpf_map *map) 2426 { 2427 free_percpu(map->elem_count); 2428 } 2429 2430 static inline void bpf_map_inc_elem_count(struct bpf_map *map) 2431 { 2432 this_cpu_inc(*map->elem_count); 2433 } 2434 2435 static inline void bpf_map_dec_elem_count(struct bpf_map *map) 2436 { 2437 this_cpu_dec(*map->elem_count); 2438 } 2439 2440 extern int sysctl_unprivileged_bpf_disabled; 2441 2442 bool bpf_token_capable(const struct bpf_token *token, int cap); 2443 2444 static inline bool bpf_allow_ptr_leaks(const struct bpf_token *token) 2445 { 2446 return bpf_token_capable(token, CAP_PERFMON); 2447 } 2448 2449 static inline bool bpf_allow_uninit_stack(const struct bpf_token *token) 2450 { 2451 return bpf_token_capable(token, CAP_PERFMON); 2452 } 2453 2454 static inline bool bpf_bypass_spec_v1(const struct bpf_token *token) 2455 { 2456 return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON); 2457 } 2458 2459 static inline bool bpf_bypass_spec_v4(const struct bpf_token *token) 2460 { 2461 return cpu_mitigations_off() || bpf_token_capable(token, CAP_PERFMON); 2462 } 2463 2464 int bpf_map_new_fd(struct bpf_map *map, int flags); 2465 int bpf_prog_new_fd(struct bpf_prog *prog); 2466 2467 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 2468 const struct bpf_link_ops *ops, struct bpf_prog *prog); 2469 void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type, 2470 const struct bpf_link_ops *ops, struct bpf_prog *prog, 2471 bool sleepable); 2472 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 2473 int bpf_link_settle(struct bpf_link_primer *primer); 2474 void bpf_link_cleanup(struct bpf_link_primer *primer); 2475 void bpf_link_inc(struct bpf_link *link); 2476 struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link); 2477 void bpf_link_put(struct bpf_link *link); 2478 int bpf_link_new_fd(struct bpf_link *link); 2479 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 2480 struct bpf_link *bpf_link_get_curr_or_next(u32 *id); 2481 2482 void bpf_token_inc(struct bpf_token *token); 2483 void bpf_token_put(struct bpf_token *token); 2484 int bpf_token_create(union bpf_attr *attr); 2485 struct bpf_token *bpf_token_get_from_fd(u32 ufd); 2486 2487 bool bpf_token_allow_cmd(const struct bpf_token *token, enum bpf_cmd cmd); 2488 bool bpf_token_allow_map_type(const struct bpf_token *token, enum bpf_map_type type); 2489 bool bpf_token_allow_prog_type(const struct bpf_token *token, 2490 enum bpf_prog_type prog_type, 2491 enum bpf_attach_type attach_type); 2492 2493 int bpf_obj_pin_user(u32 ufd, int path_fd, const char __user *pathname); 2494 int bpf_obj_get_user(int path_fd, const char __user *pathname, int flags); 2495 struct inode *bpf_get_inode(struct super_block *sb, const struct inode *dir, 2496 umode_t mode); 2497 2498 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 2499 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 2500 extern int bpf_iter_ ## target(args); \ 2501 int __init bpf_iter_ ## target(args) { return 0; } 2502 2503 /* 2504 * The task type of iterators. 2505 * 2506 * For BPF task iterators, they can be parameterized with various 2507 * parameters to visit only some of tasks. 2508 * 2509 * BPF_TASK_ITER_ALL (default) 2510 * Iterate over resources of every task. 2511 * 2512 * BPF_TASK_ITER_TID 2513 * Iterate over resources of a task/tid. 2514 * 2515 * BPF_TASK_ITER_TGID 2516 * Iterate over resources of every task of a process / task group. 2517 */ 2518 enum bpf_iter_task_type { 2519 BPF_TASK_ITER_ALL = 0, 2520 BPF_TASK_ITER_TID, 2521 BPF_TASK_ITER_TGID, 2522 }; 2523 2524 struct bpf_iter_aux_info { 2525 /* for map_elem iter */ 2526 struct bpf_map *map; 2527 2528 /* for cgroup iter */ 2529 struct { 2530 struct cgroup *start; /* starting cgroup */ 2531 enum bpf_cgroup_iter_order order; 2532 } cgroup; 2533 struct { 2534 enum bpf_iter_task_type type; 2535 u32 pid; 2536 } task; 2537 }; 2538 2539 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 2540 union bpf_iter_link_info *linfo, 2541 struct bpf_iter_aux_info *aux); 2542 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 2543 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 2544 struct seq_file *seq); 2545 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 2546 struct bpf_link_info *info); 2547 typedef const struct bpf_func_proto * 2548 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 2549 const struct bpf_prog *prog); 2550 2551 enum bpf_iter_feature { 2552 BPF_ITER_RESCHED = BIT(0), 2553 }; 2554 2555 #define BPF_ITER_CTX_ARG_MAX 2 2556 struct bpf_iter_reg { 2557 const char *target; 2558 bpf_iter_attach_target_t attach_target; 2559 bpf_iter_detach_target_t detach_target; 2560 bpf_iter_show_fdinfo_t show_fdinfo; 2561 bpf_iter_fill_link_info_t fill_link_info; 2562 bpf_iter_get_func_proto_t get_func_proto; 2563 u32 ctx_arg_info_size; 2564 u32 feature; 2565 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 2566 const struct bpf_iter_seq_info *seq_info; 2567 }; 2568 2569 struct bpf_iter_meta { 2570 __bpf_md_ptr(struct seq_file *, seq); 2571 u64 session_id; 2572 u64 seq_num; 2573 }; 2574 2575 struct bpf_iter__bpf_map_elem { 2576 __bpf_md_ptr(struct bpf_iter_meta *, meta); 2577 __bpf_md_ptr(struct bpf_map *, map); 2578 __bpf_md_ptr(void *, key); 2579 __bpf_md_ptr(void *, value); 2580 }; 2581 2582 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 2583 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 2584 int bpf_iter_prog_supported(struct bpf_prog *prog); 2585 const struct bpf_func_proto * 2586 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 2587 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 2588 int bpf_iter_new_fd(struct bpf_link *link); 2589 bool bpf_link_is_iter(struct bpf_link *link); 2590 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 2591 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 2592 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 2593 struct seq_file *seq); 2594 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 2595 struct bpf_link_info *info); 2596 2597 int map_set_for_each_callback_args(struct bpf_verifier_env *env, 2598 struct bpf_func_state *caller, 2599 struct bpf_func_state *callee); 2600 2601 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); 2602 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); 2603 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 2604 u64 flags); 2605 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 2606 u64 flags); 2607 2608 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); 2609 2610 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 2611 void *key, void *value, u64 map_flags); 2612 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 2613 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 2614 void *key, void *value, u64 map_flags); 2615 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 2616 2617 int bpf_get_file_flag(int flags); 2618 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 2619 size_t actual_size); 2620 2621 /* verify correctness of eBPF program */ 2622 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size); 2623 2624 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 2625 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 2626 #endif 2627 2628 struct btf *bpf_get_btf_vmlinux(void); 2629 2630 /* Map specifics */ 2631 struct xdp_frame; 2632 struct sk_buff; 2633 struct bpf_dtab_netdev; 2634 struct bpf_cpu_map_entry; 2635 2636 void __dev_flush(struct list_head *flush_list); 2637 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 2638 struct net_device *dev_rx); 2639 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 2640 struct net_device *dev_rx); 2641 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 2642 struct bpf_map *map, bool exclude_ingress); 2643 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 2644 const struct bpf_prog *xdp_prog); 2645 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 2646 const struct bpf_prog *xdp_prog, 2647 struct bpf_map *map, bool exclude_ingress); 2648 2649 void __cpu_map_flush(struct list_head *flush_list); 2650 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf, 2651 struct net_device *dev_rx); 2652 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 2653 struct sk_buff *skb); 2654 2655 /* Return map's numa specified by userspace */ 2656 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 2657 { 2658 return (attr->map_flags & BPF_F_NUMA_NODE) ? 2659 attr->numa_node : NUMA_NO_NODE; 2660 } 2661 2662 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 2663 int array_map_alloc_check(union bpf_attr *attr); 2664 2665 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 2666 union bpf_attr __user *uattr); 2667 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 2668 union bpf_attr __user *uattr); 2669 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 2670 const union bpf_attr *kattr, 2671 union bpf_attr __user *uattr); 2672 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 2673 const union bpf_attr *kattr, 2674 union bpf_attr __user *uattr); 2675 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 2676 const union bpf_attr *kattr, 2677 union bpf_attr __user *uattr); 2678 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 2679 const union bpf_attr *kattr, 2680 union bpf_attr __user *uattr); 2681 int bpf_prog_test_run_nf(struct bpf_prog *prog, 2682 const union bpf_attr *kattr, 2683 union bpf_attr __user *uattr); 2684 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 2685 const struct bpf_prog *prog, 2686 struct bpf_insn_access_aux *info); 2687 2688 static inline bool bpf_tracing_ctx_access(int off, int size, 2689 enum bpf_access_type type) 2690 { 2691 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS) 2692 return false; 2693 if (type != BPF_READ) 2694 return false; 2695 if (off % size != 0) 2696 return false; 2697 return true; 2698 } 2699 2700 static inline bool bpf_tracing_btf_ctx_access(int off, int size, 2701 enum bpf_access_type type, 2702 const struct bpf_prog *prog, 2703 struct bpf_insn_access_aux *info) 2704 { 2705 if (!bpf_tracing_ctx_access(off, size, type)) 2706 return false; 2707 return btf_ctx_access(off, size, type, prog, info); 2708 } 2709 2710 int btf_struct_access(struct bpf_verifier_log *log, 2711 const struct bpf_reg_state *reg, 2712 int off, int size, enum bpf_access_type atype, 2713 u32 *next_btf_id, enum bpf_type_flag *flag, const char **field_name); 2714 bool btf_struct_ids_match(struct bpf_verifier_log *log, 2715 const struct btf *btf, u32 id, int off, 2716 const struct btf *need_btf, u32 need_type_id, 2717 bool strict); 2718 2719 int btf_distill_func_proto(struct bpf_verifier_log *log, 2720 struct btf *btf, 2721 const struct btf_type *func_proto, 2722 const char *func_name, 2723 struct btf_func_model *m); 2724 2725 struct bpf_reg_state; 2726 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog); 2727 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 2728 struct btf *btf, const struct btf_type *t); 2729 const char *btf_find_decl_tag_value(const struct btf *btf, const struct btf_type *pt, 2730 int comp_idx, const char *tag_key); 2731 int btf_find_next_decl_tag(const struct btf *btf, const struct btf_type *pt, 2732 int comp_idx, const char *tag_key, int last_id); 2733 2734 struct bpf_prog *bpf_prog_by_id(u32 id); 2735 struct bpf_link *bpf_link_by_id(u32 id); 2736 2737 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id, 2738 const struct bpf_prog *prog); 2739 void bpf_task_storage_free(struct task_struct *task); 2740 void bpf_cgrp_storage_free(struct cgroup *cgroup); 2741 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog); 2742 const struct btf_func_model * 2743 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 2744 const struct bpf_insn *insn); 2745 int bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id, 2746 u16 btf_fd_idx, u8 **func_addr); 2747 2748 struct bpf_core_ctx { 2749 struct bpf_verifier_log *log; 2750 const struct btf *btf; 2751 }; 2752 2753 bool btf_nested_type_is_trusted(struct bpf_verifier_log *log, 2754 const struct bpf_reg_state *reg, 2755 const char *field_name, u32 btf_id, const char *suffix); 2756 2757 bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log, 2758 const struct btf *reg_btf, u32 reg_id, 2759 const struct btf *arg_btf, u32 arg_id); 2760 2761 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo, 2762 int relo_idx, void *insn); 2763 2764 static inline bool unprivileged_ebpf_enabled(void) 2765 { 2766 return !sysctl_unprivileged_bpf_disabled; 2767 } 2768 2769 /* Not all bpf prog type has the bpf_ctx. 2770 * For the bpf prog type that has initialized the bpf_ctx, 2771 * this function can be used to decide if a kernel function 2772 * is called by a bpf program. 2773 */ 2774 static inline bool has_current_bpf_ctx(void) 2775 { 2776 return !!current->bpf_ctx; 2777 } 2778 2779 void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog); 2780 2781 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 2782 enum bpf_dynptr_type type, u32 offset, u32 size); 2783 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr); 2784 void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr); 2785 2786 #else /* !CONFIG_BPF_SYSCALL */ 2787 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 2788 { 2789 return ERR_PTR(-EOPNOTSUPP); 2790 } 2791 2792 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 2793 enum bpf_prog_type type, 2794 bool attach_drv) 2795 { 2796 return ERR_PTR(-EOPNOTSUPP); 2797 } 2798 2799 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 2800 { 2801 } 2802 2803 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 2804 { 2805 } 2806 2807 static inline void bpf_prog_put(struct bpf_prog *prog) 2808 { 2809 } 2810 2811 static inline void bpf_prog_inc(struct bpf_prog *prog) 2812 { 2813 } 2814 2815 static inline struct bpf_prog *__must_check 2816 bpf_prog_inc_not_zero(struct bpf_prog *prog) 2817 { 2818 return ERR_PTR(-EOPNOTSUPP); 2819 } 2820 2821 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 2822 const struct bpf_link_ops *ops, 2823 struct bpf_prog *prog) 2824 { 2825 } 2826 2827 static inline void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type, 2828 const struct bpf_link_ops *ops, struct bpf_prog *prog, 2829 bool sleepable) 2830 { 2831 } 2832 2833 static inline int bpf_link_prime(struct bpf_link *link, 2834 struct bpf_link_primer *primer) 2835 { 2836 return -EOPNOTSUPP; 2837 } 2838 2839 static inline int bpf_link_settle(struct bpf_link_primer *primer) 2840 { 2841 return -EOPNOTSUPP; 2842 } 2843 2844 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 2845 { 2846 } 2847 2848 static inline void bpf_link_inc(struct bpf_link *link) 2849 { 2850 } 2851 2852 static inline struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link) 2853 { 2854 return NULL; 2855 } 2856 2857 static inline void bpf_link_put(struct bpf_link *link) 2858 { 2859 } 2860 2861 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 2862 { 2863 return -EOPNOTSUPP; 2864 } 2865 2866 static inline bool bpf_token_capable(const struct bpf_token *token, int cap) 2867 { 2868 return capable(cap) || (cap != CAP_SYS_ADMIN && capable(CAP_SYS_ADMIN)); 2869 } 2870 2871 static inline void bpf_token_inc(struct bpf_token *token) 2872 { 2873 } 2874 2875 static inline void bpf_token_put(struct bpf_token *token) 2876 { 2877 } 2878 2879 static inline struct bpf_token *bpf_token_get_from_fd(u32 ufd) 2880 { 2881 return ERR_PTR(-EOPNOTSUPP); 2882 } 2883 2884 static inline void __dev_flush(struct list_head *flush_list) 2885 { 2886 } 2887 2888 struct xdp_frame; 2889 struct bpf_dtab_netdev; 2890 struct bpf_cpu_map_entry; 2891 2892 static inline 2893 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 2894 struct net_device *dev_rx) 2895 { 2896 return 0; 2897 } 2898 2899 static inline 2900 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 2901 struct net_device *dev_rx) 2902 { 2903 return 0; 2904 } 2905 2906 static inline 2907 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 2908 struct bpf_map *map, bool exclude_ingress) 2909 { 2910 return 0; 2911 } 2912 2913 struct sk_buff; 2914 2915 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 2916 struct sk_buff *skb, 2917 const struct bpf_prog *xdp_prog) 2918 { 2919 return 0; 2920 } 2921 2922 static inline 2923 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 2924 const struct bpf_prog *xdp_prog, 2925 struct bpf_map *map, bool exclude_ingress) 2926 { 2927 return 0; 2928 } 2929 2930 static inline void __cpu_map_flush(struct list_head *flush_list) 2931 { 2932 } 2933 2934 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 2935 struct xdp_frame *xdpf, 2936 struct net_device *dev_rx) 2937 { 2938 return 0; 2939 } 2940 2941 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 2942 struct sk_buff *skb) 2943 { 2944 return -EOPNOTSUPP; 2945 } 2946 2947 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 2948 enum bpf_prog_type type) 2949 { 2950 return ERR_PTR(-EOPNOTSUPP); 2951 } 2952 2953 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 2954 const union bpf_attr *kattr, 2955 union bpf_attr __user *uattr) 2956 { 2957 return -ENOTSUPP; 2958 } 2959 2960 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 2961 const union bpf_attr *kattr, 2962 union bpf_attr __user *uattr) 2963 { 2964 return -ENOTSUPP; 2965 } 2966 2967 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 2968 const union bpf_attr *kattr, 2969 union bpf_attr __user *uattr) 2970 { 2971 return -ENOTSUPP; 2972 } 2973 2974 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 2975 const union bpf_attr *kattr, 2976 union bpf_attr __user *uattr) 2977 { 2978 return -ENOTSUPP; 2979 } 2980 2981 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 2982 const union bpf_attr *kattr, 2983 union bpf_attr __user *uattr) 2984 { 2985 return -ENOTSUPP; 2986 } 2987 2988 static inline void bpf_map_put(struct bpf_map *map) 2989 { 2990 } 2991 2992 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 2993 { 2994 return ERR_PTR(-ENOTSUPP); 2995 } 2996 2997 static inline int btf_struct_access(struct bpf_verifier_log *log, 2998 const struct bpf_reg_state *reg, 2999 int off, int size, enum bpf_access_type atype, 3000 u32 *next_btf_id, enum bpf_type_flag *flag, 3001 const char **field_name) 3002 { 3003 return -EACCES; 3004 } 3005 3006 static inline const struct bpf_func_proto * 3007 bpf_base_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) 3008 { 3009 return NULL; 3010 } 3011 3012 static inline void bpf_task_storage_free(struct task_struct *task) 3013 { 3014 } 3015 3016 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog) 3017 { 3018 return false; 3019 } 3020 3021 static inline const struct btf_func_model * 3022 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 3023 const struct bpf_insn *insn) 3024 { 3025 return NULL; 3026 } 3027 3028 static inline int 3029 bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id, 3030 u16 btf_fd_idx, u8 **func_addr) 3031 { 3032 return -ENOTSUPP; 3033 } 3034 3035 static inline bool unprivileged_ebpf_enabled(void) 3036 { 3037 return false; 3038 } 3039 3040 static inline bool has_current_bpf_ctx(void) 3041 { 3042 return false; 3043 } 3044 3045 static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog) 3046 { 3047 } 3048 3049 static inline void bpf_cgrp_storage_free(struct cgroup *cgroup) 3050 { 3051 } 3052 3053 static inline void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 3054 enum bpf_dynptr_type type, u32 offset, u32 size) 3055 { 3056 } 3057 3058 static inline void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr) 3059 { 3060 } 3061 3062 static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr) 3063 { 3064 } 3065 #endif /* CONFIG_BPF_SYSCALL */ 3066 3067 static __always_inline int 3068 bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr) 3069 { 3070 int ret = -EFAULT; 3071 3072 if (IS_ENABLED(CONFIG_BPF_EVENTS)) 3073 ret = copy_from_kernel_nofault(dst, unsafe_ptr, size); 3074 if (unlikely(ret < 0)) 3075 memset(dst, 0, size); 3076 return ret; 3077 } 3078 3079 void __bpf_free_used_btfs(struct btf_mod_pair *used_btfs, u32 len); 3080 3081 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 3082 enum bpf_prog_type type) 3083 { 3084 return bpf_prog_get_type_dev(ufd, type, false); 3085 } 3086 3087 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 3088 struct bpf_map **used_maps, u32 len); 3089 3090 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 3091 3092 int bpf_prog_offload_compile(struct bpf_prog *prog); 3093 void bpf_prog_dev_bound_destroy(struct bpf_prog *prog); 3094 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 3095 struct bpf_prog *prog); 3096 3097 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 3098 3099 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 3100 int bpf_map_offload_update_elem(struct bpf_map *map, 3101 void *key, void *value, u64 flags); 3102 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 3103 int bpf_map_offload_get_next_key(struct bpf_map *map, 3104 void *key, void *next_key); 3105 3106 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 3107 3108 struct bpf_offload_dev * 3109 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 3110 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 3111 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 3112 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 3113 struct net_device *netdev); 3114 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 3115 struct net_device *netdev); 3116 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 3117 3118 void unpriv_ebpf_notify(int new_state); 3119 3120 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 3121 int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log, 3122 struct bpf_prog_aux *prog_aux); 3123 void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, u32 func_id); 3124 int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr); 3125 int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, struct bpf_prog *old_prog); 3126 void bpf_dev_bound_netdev_unregister(struct net_device *dev); 3127 3128 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 3129 { 3130 return aux->dev_bound; 3131 } 3132 3133 static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux) 3134 { 3135 return aux->offload_requested; 3136 } 3137 3138 bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs); 3139 3140 static inline bool bpf_map_is_offloaded(struct bpf_map *map) 3141 { 3142 return unlikely(map->ops == &bpf_map_offload_ops); 3143 } 3144 3145 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 3146 void bpf_map_offload_map_free(struct bpf_map *map); 3147 u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map); 3148 int bpf_prog_test_run_syscall(struct bpf_prog *prog, 3149 const union bpf_attr *kattr, 3150 union bpf_attr __user *uattr); 3151 3152 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 3153 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 3154 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 3155 int sock_map_bpf_prog_query(const union bpf_attr *attr, 3156 union bpf_attr __user *uattr); 3157 int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog); 3158 3159 void sock_map_unhash(struct sock *sk); 3160 void sock_map_destroy(struct sock *sk); 3161 void sock_map_close(struct sock *sk, long timeout); 3162 #else 3163 static inline int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log, 3164 struct bpf_prog_aux *prog_aux) 3165 { 3166 return -EOPNOTSUPP; 3167 } 3168 3169 static inline void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, 3170 u32 func_id) 3171 { 3172 return NULL; 3173 } 3174 3175 static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog, 3176 union bpf_attr *attr) 3177 { 3178 return -EOPNOTSUPP; 3179 } 3180 3181 static inline int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, 3182 struct bpf_prog *old_prog) 3183 { 3184 return -EOPNOTSUPP; 3185 } 3186 3187 static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev) 3188 { 3189 } 3190 3191 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 3192 { 3193 return false; 3194 } 3195 3196 static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux) 3197 { 3198 return false; 3199 } 3200 3201 static inline bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs) 3202 { 3203 return false; 3204 } 3205 3206 static inline bool bpf_map_is_offloaded(struct bpf_map *map) 3207 { 3208 return false; 3209 } 3210 3211 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 3212 { 3213 return ERR_PTR(-EOPNOTSUPP); 3214 } 3215 3216 static inline void bpf_map_offload_map_free(struct bpf_map *map) 3217 { 3218 } 3219 3220 static inline u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map) 3221 { 3222 return 0; 3223 } 3224 3225 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 3226 const union bpf_attr *kattr, 3227 union bpf_attr __user *uattr) 3228 { 3229 return -ENOTSUPP; 3230 } 3231 3232 #ifdef CONFIG_BPF_SYSCALL 3233 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 3234 struct bpf_prog *prog) 3235 { 3236 return -EINVAL; 3237 } 3238 3239 static inline int sock_map_prog_detach(const union bpf_attr *attr, 3240 enum bpf_prog_type ptype) 3241 { 3242 return -EOPNOTSUPP; 3243 } 3244 3245 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 3246 u64 flags) 3247 { 3248 return -EOPNOTSUPP; 3249 } 3250 3251 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr, 3252 union bpf_attr __user *uattr) 3253 { 3254 return -EINVAL; 3255 } 3256 3257 static inline int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog) 3258 { 3259 return -EOPNOTSUPP; 3260 } 3261 #endif /* CONFIG_BPF_SYSCALL */ 3262 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 3263 3264 static __always_inline void 3265 bpf_prog_inc_misses_counters(const struct bpf_prog_array *array) 3266 { 3267 const struct bpf_prog_array_item *item; 3268 struct bpf_prog *prog; 3269 3270 if (unlikely(!array)) 3271 return; 3272 3273 item = &array->items[0]; 3274 while ((prog = READ_ONCE(item->prog))) { 3275 bpf_prog_inc_misses_counter(prog); 3276 item++; 3277 } 3278 } 3279 3280 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 3281 void bpf_sk_reuseport_detach(struct sock *sk); 3282 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 3283 void *value); 3284 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 3285 void *value, u64 map_flags); 3286 #else 3287 static inline void bpf_sk_reuseport_detach(struct sock *sk) 3288 { 3289 } 3290 3291 #ifdef CONFIG_BPF_SYSCALL 3292 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 3293 void *key, void *value) 3294 { 3295 return -EOPNOTSUPP; 3296 } 3297 3298 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 3299 void *key, void *value, 3300 u64 map_flags) 3301 { 3302 return -EOPNOTSUPP; 3303 } 3304 #endif /* CONFIG_BPF_SYSCALL */ 3305 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 3306 3307 /* verifier prototypes for helper functions called from eBPF programs */ 3308 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 3309 extern const struct bpf_func_proto bpf_map_update_elem_proto; 3310 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 3311 extern const struct bpf_func_proto bpf_map_push_elem_proto; 3312 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 3313 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 3314 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto; 3315 3316 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 3317 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 3318 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 3319 extern const struct bpf_func_proto bpf_tail_call_proto; 3320 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 3321 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 3322 extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto; 3323 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 3324 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 3325 extern const struct bpf_func_proto bpf_get_current_comm_proto; 3326 extern const struct bpf_func_proto bpf_get_stackid_proto; 3327 extern const struct bpf_func_proto bpf_get_stack_proto; 3328 extern const struct bpf_func_proto bpf_get_stack_sleepable_proto; 3329 extern const struct bpf_func_proto bpf_get_task_stack_proto; 3330 extern const struct bpf_func_proto bpf_get_task_stack_sleepable_proto; 3331 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 3332 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 3333 extern const struct bpf_func_proto bpf_sock_map_update_proto; 3334 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 3335 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 3336 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 3337 extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto; 3338 extern const struct bpf_func_proto bpf_current_task_under_cgroup_proto; 3339 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 3340 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 3341 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 3342 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 3343 extern const struct bpf_func_proto bpf_spin_lock_proto; 3344 extern const struct bpf_func_proto bpf_spin_unlock_proto; 3345 extern const struct bpf_func_proto bpf_get_local_storage_proto; 3346 extern const struct bpf_func_proto bpf_strtol_proto; 3347 extern const struct bpf_func_proto bpf_strtoul_proto; 3348 extern const struct bpf_func_proto bpf_tcp_sock_proto; 3349 extern const struct bpf_func_proto bpf_jiffies64_proto; 3350 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 3351 extern const struct bpf_func_proto bpf_event_output_data_proto; 3352 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 3353 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 3354 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 3355 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 3356 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 3357 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto; 3358 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto; 3359 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto; 3360 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 3361 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 3362 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 3363 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 3364 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 3365 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto; 3366 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto; 3367 extern const struct bpf_func_proto bpf_copy_from_user_proto; 3368 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 3369 extern const struct bpf_func_proto bpf_snprintf_proto; 3370 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 3371 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 3372 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 3373 extern const struct bpf_func_proto bpf_sock_from_file_proto; 3374 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 3375 extern const struct bpf_func_proto bpf_task_storage_get_recur_proto; 3376 extern const struct bpf_func_proto bpf_task_storage_get_proto; 3377 extern const struct bpf_func_proto bpf_task_storage_delete_recur_proto; 3378 extern const struct bpf_func_proto bpf_task_storage_delete_proto; 3379 extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 3380 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 3381 extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 3382 extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 3383 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto; 3384 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto; 3385 extern const struct bpf_func_proto bpf_find_vma_proto; 3386 extern const struct bpf_func_proto bpf_loop_proto; 3387 extern const struct bpf_func_proto bpf_copy_from_user_task_proto; 3388 extern const struct bpf_func_proto bpf_set_retval_proto; 3389 extern const struct bpf_func_proto bpf_get_retval_proto; 3390 extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto; 3391 extern const struct bpf_func_proto bpf_cgrp_storage_get_proto; 3392 extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto; 3393 3394 const struct bpf_func_proto *tracing_prog_func_proto( 3395 enum bpf_func_id func_id, const struct bpf_prog *prog); 3396 3397 /* Shared helpers among cBPF and eBPF. */ 3398 void bpf_user_rnd_init_once(void); 3399 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 3400 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 3401 3402 #if defined(CONFIG_NET) 3403 bool bpf_sock_common_is_valid_access(int off, int size, 3404 enum bpf_access_type type, 3405 struct bpf_insn_access_aux *info); 3406 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 3407 struct bpf_insn_access_aux *info); 3408 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 3409 const struct bpf_insn *si, 3410 struct bpf_insn *insn_buf, 3411 struct bpf_prog *prog, 3412 u32 *target_size); 3413 int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags, 3414 struct bpf_dynptr *ptr); 3415 #else 3416 static inline bool bpf_sock_common_is_valid_access(int off, int size, 3417 enum bpf_access_type type, 3418 struct bpf_insn_access_aux *info) 3419 { 3420 return false; 3421 } 3422 static inline bool bpf_sock_is_valid_access(int off, int size, 3423 enum bpf_access_type type, 3424 struct bpf_insn_access_aux *info) 3425 { 3426 return false; 3427 } 3428 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 3429 const struct bpf_insn *si, 3430 struct bpf_insn *insn_buf, 3431 struct bpf_prog *prog, 3432 u32 *target_size) 3433 { 3434 return 0; 3435 } 3436 static inline int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags, 3437 struct bpf_dynptr *ptr) 3438 { 3439 return -EOPNOTSUPP; 3440 } 3441 #endif 3442 3443 #ifdef CONFIG_INET 3444 struct sk_reuseport_kern { 3445 struct sk_buff *skb; 3446 struct sock *sk; 3447 struct sock *selected_sk; 3448 struct sock *migrating_sk; 3449 void *data_end; 3450 u32 hash; 3451 u32 reuseport_id; 3452 bool bind_inany; 3453 }; 3454 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 3455 struct bpf_insn_access_aux *info); 3456 3457 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 3458 const struct bpf_insn *si, 3459 struct bpf_insn *insn_buf, 3460 struct bpf_prog *prog, 3461 u32 *target_size); 3462 3463 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 3464 struct bpf_insn_access_aux *info); 3465 3466 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 3467 const struct bpf_insn *si, 3468 struct bpf_insn *insn_buf, 3469 struct bpf_prog *prog, 3470 u32 *target_size); 3471 #else 3472 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 3473 enum bpf_access_type type, 3474 struct bpf_insn_access_aux *info) 3475 { 3476 return false; 3477 } 3478 3479 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 3480 const struct bpf_insn *si, 3481 struct bpf_insn *insn_buf, 3482 struct bpf_prog *prog, 3483 u32 *target_size) 3484 { 3485 return 0; 3486 } 3487 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 3488 enum bpf_access_type type, 3489 struct bpf_insn_access_aux *info) 3490 { 3491 return false; 3492 } 3493 3494 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 3495 const struct bpf_insn *si, 3496 struct bpf_insn *insn_buf, 3497 struct bpf_prog *prog, 3498 u32 *target_size) 3499 { 3500 return 0; 3501 } 3502 #endif /* CONFIG_INET */ 3503 3504 enum bpf_text_poke_type { 3505 BPF_MOD_CALL, 3506 BPF_MOD_JUMP, 3507 }; 3508 3509 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, 3510 void *addr1, void *addr2); 3511 3512 void bpf_arch_poke_desc_update(struct bpf_jit_poke_descriptor *poke, 3513 struct bpf_prog *new, struct bpf_prog *old); 3514 3515 void *bpf_arch_text_copy(void *dst, void *src, size_t len); 3516 int bpf_arch_text_invalidate(void *dst, size_t len); 3517 3518 struct btf_id_set; 3519 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 3520 3521 #define MAX_BPRINTF_VARARGS 12 3522 #define MAX_BPRINTF_BUF 1024 3523 3524 struct bpf_bprintf_data { 3525 u32 *bin_args; 3526 char *buf; 3527 bool get_bin_args; 3528 bool get_buf; 3529 }; 3530 3531 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args, 3532 u32 num_args, struct bpf_bprintf_data *data); 3533 void bpf_bprintf_cleanup(struct bpf_bprintf_data *data); 3534 3535 #ifdef CONFIG_BPF_LSM 3536 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype); 3537 void bpf_cgroup_atype_put(int cgroup_atype); 3538 #else 3539 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {} 3540 static inline void bpf_cgroup_atype_put(int cgroup_atype) {} 3541 #endif /* CONFIG_BPF_LSM */ 3542 3543 struct key; 3544 3545 #ifdef CONFIG_KEYS 3546 struct bpf_key { 3547 struct key *key; 3548 bool has_ref; 3549 }; 3550 #endif /* CONFIG_KEYS */ 3551 3552 static inline bool type_is_alloc(u32 type) 3553 { 3554 return type & MEM_ALLOC; 3555 } 3556 3557 static inline gfp_t bpf_memcg_flags(gfp_t flags) 3558 { 3559 if (memcg_bpf_enabled()) 3560 return flags | __GFP_ACCOUNT; 3561 return flags; 3562 } 3563 3564 static inline bool bpf_is_subprog(const struct bpf_prog *prog) 3565 { 3566 return prog->aux->func_idx != 0; 3567 } 3568 3569 #endif /* _LINUX_BPF_H */ 3570