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