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