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