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