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