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 #include <linux/bpf_defs.h> 10 11 #include <crypto/sha2.h> 12 #include <linux/workqueue.h> 13 #include <linux/file.h> 14 #include <linux/percpu.h> 15 #include <linux/err.h> 16 #include <linux/rbtree_latch.h> 17 #include <linux/numa.h> 18 #include <linux/mm_types.h> 19 #include <linux/wait.h> 20 #include <linux/irq_work_types.h> 21 #include <linux/refcount.h> 22 #include <linux/mutex.h> 23 #include <linux/module.h> 24 #include <linux/kallsyms.h> 25 #include <linux/capability.h> 26 #include <linux/sched/mm.h> 27 #include <linux/slab.h> 28 #include <linux/percpu-refcount.h> 29 #include <linux/stddef.h> 30 #include <linux/bpfptr.h> 31 #include <linux/btf.h> 32 #include <linux/rcupdate_trace.h> 33 #include <linux/static_call.h> 34 #include <linux/memcontrol.h> 35 #include <linux/cfi.h> 36 #include <linux/xattr.h> 37 #include <linux/key.h> 38 #include <linux/ftrace.h> 39 #include <asm/rqspinlock.h> 40 41 struct bpf_verifier_env; 42 struct bpf_verifier_log; 43 struct perf_event; 44 struct bpf_prog; 45 struct bpf_prog_aux; 46 struct bpf_map; 47 struct bpf_arena; 48 struct sock; 49 struct seq_file; 50 struct btf; 51 struct btf_type; 52 struct exception_table_entry; 53 struct seq_operations; 54 struct bpf_iter_aux_info; 55 struct bpf_local_storage; 56 struct bpf_local_storage_map; 57 struct kobject; 58 struct mem_cgroup; 59 struct module; 60 struct bpf_func_state; 61 struct ftrace_ops; 62 struct cgroup; 63 struct bpf_token; 64 struct user_namespace; 65 struct super_block; 66 struct inode; 67 68 extern struct idr btf_idr; 69 extern spinlock_t btf_idr_lock; 70 extern struct kobject *btf_kobj; 71 extern struct bpf_mem_alloc bpf_global_ma, bpf_global_percpu_ma; 72 extern bool bpf_global_ma_set; 73 extern struct workqueue_struct *bpf_prog_free_wq __ro_after_init; 74 75 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64); 76 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, 77 struct bpf_iter_aux_info *aux); 78 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data); 79 typedef unsigned int (*bpf_func_t)(const void *, 80 const struct bpf_insn *); 81 struct bpf_iter_seq_info { 82 const struct seq_operations *seq_ops; 83 bpf_iter_init_seq_priv_t init_seq_private; 84 bpf_iter_fini_seq_priv_t fini_seq_private; 85 u32 seq_priv_size; 86 }; 87 88 /* map is generic key/value storage optionally accessible by eBPF programs */ 89 struct bpf_map_ops { 90 /* funcs callable from userspace (via syscall) */ 91 int (*map_alloc_check)(union bpf_attr *attr); 92 struct bpf_map *(*map_alloc)(union bpf_attr *attr); 93 void (*map_release)(struct bpf_map *map, struct file *map_file); 94 void (*map_free)(struct bpf_map *map); 95 void (*map_free_pre_rcu)(struct bpf_map *map); 96 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); 97 void (*map_release_uref)(struct bpf_map *map); 98 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key); 99 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, 100 union bpf_attr __user *uattr); 101 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key, 102 void *value, u64 flags); 103 int (*map_lookup_and_delete_batch)(struct bpf_map *map, 104 const union bpf_attr *attr, 105 union bpf_attr __user *uattr); 106 int (*map_update_batch)(struct bpf_map *map, struct file *map_file, 107 const union bpf_attr *attr, 108 union bpf_attr __user *uattr); 109 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, 110 union bpf_attr __user *uattr); 111 112 /* funcs callable from userspace and from eBPF programs */ 113 void *(*map_lookup_elem)(struct bpf_map *map, void *key); 114 long (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); 115 long (*map_delete_elem)(struct bpf_map *map, void *key); 116 long (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); 117 long (*map_pop_elem)(struct bpf_map *map, void *value); 118 long (*map_peek_elem)(struct bpf_map *map, void *value); 119 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu); 120 int (*map_get_hash)(struct bpf_map *map); 121 122 /* funcs called by prog_array and perf_event_array map */ 123 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, 124 int fd); 125 /* If need_defer is true, the implementation should guarantee that 126 * the to-be-put element is still alive before the bpf program, which 127 * may manipulate it, exists. 128 */ 129 void (*map_fd_put_ptr)(struct bpf_map *map, void *ptr, bool need_defer); 130 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); 131 u32 (*map_fd_sys_lookup_elem)(void *ptr); 132 void (*map_seq_show_elem)(struct bpf_map *map, void *key, 133 struct seq_file *m); 134 int (*map_check_btf)(struct bpf_map *map, 135 const struct btf *btf, 136 const struct btf_type *key_type, 137 const struct btf_type *value_type); 138 139 /* Prog poke tracking helpers. */ 140 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux); 141 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux); 142 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, 143 struct bpf_prog *new); 144 145 /* Direct value access helpers. */ 146 int (*map_direct_value_addr)(const struct bpf_map *map, 147 u64 *imm, u32 off); 148 int (*map_direct_value_meta)(const struct bpf_map *map, 149 u64 imm, u32 *off); 150 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma); 151 vm_fault_t (*map_mmap_fault)(struct bpf_map *map, struct vm_fault *vmf); 152 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, 153 struct poll_table_struct *pts); 154 unsigned long (*map_get_unmapped_area)(struct file *filep, unsigned long addr, 155 unsigned long len, unsigned long pgoff, 156 unsigned long flags); 157 158 /* Functions called by bpf_local_storage maps */ 159 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, 160 void *owner, u32 size); 161 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, 162 void *owner, u32 size); 163 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner); 164 165 /* Misc helpers.*/ 166 long (*map_redirect)(struct bpf_map *map, u64 key, u64 flags); 167 168 /* map_meta_equal must be implemented for maps that can be 169 * used as an inner map. It is a runtime check to ensure 170 * an inner map can be inserted to an outer map. 171 * 172 * Some properties of the inner map has been used during the 173 * verification time. When inserting an inner map at the runtime, 174 * map_meta_equal has to ensure the inserting map has the same 175 * properties that the verifier has used earlier. 176 */ 177 bool (*map_meta_equal)(const struct bpf_map *meta0, 178 const struct bpf_map *meta1); 179 180 181 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env, 182 struct bpf_func_state *caller, 183 struct bpf_func_state *callee); 184 long (*map_for_each_callback)(struct bpf_map *map, 185 bpf_callback_t callback_fn, 186 void *callback_ctx, u64 flags); 187 188 u64 (*map_mem_usage)(const struct bpf_map *map); 189 190 /* BTF id of struct allocated by map_alloc */ 191 int *map_btf_id; 192 193 /* bpf_iter info used to open a seq_file */ 194 const struct bpf_iter_seq_info *iter_seq_info; 195 }; 196 197 enum { 198 /* Support at most 11 fields in a BTF type */ 199 BTF_FIELDS_MAX = 11, 200 }; 201 202 enum btf_field_type { 203 BPF_SPIN_LOCK = (1 << 0), 204 BPF_TIMER = (1 << 1), 205 BPF_KPTR_UNREF = (1 << 2), 206 BPF_KPTR_REF = (1 << 3), 207 BPF_KPTR_PERCPU = (1 << 4), 208 BPF_KPTR = BPF_KPTR_UNREF | BPF_KPTR_REF | BPF_KPTR_PERCPU, 209 BPF_LIST_HEAD = (1 << 5), 210 BPF_LIST_NODE = (1 << 6), 211 BPF_RB_ROOT = (1 << 7), 212 BPF_RB_NODE = (1 << 8), 213 BPF_GRAPH_NODE = BPF_RB_NODE | BPF_LIST_NODE, 214 BPF_GRAPH_ROOT = BPF_RB_ROOT | BPF_LIST_HEAD, 215 BPF_REFCOUNT = (1 << 9), 216 BPF_WORKQUEUE = (1 << 10), 217 BPF_UPTR = (1 << 11), 218 BPF_RES_SPIN_LOCK = (1 << 12), 219 BPF_TASK_WORK = (1 << 13), 220 BPF_RCU_HEAD = (1 << 14), 221 }; 222 223 enum bpf_cgroup_storage_type { 224 BPF_CGROUP_STORAGE_SHARED, 225 BPF_CGROUP_STORAGE_PERCPU, 226 __BPF_CGROUP_STORAGE_MAX 227 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX 228 }; 229 230 #ifdef CONFIG_CGROUP_BPF 231 # define for_each_cgroup_storage_type(stype) \ 232 for (stype = 0; stype < MAX_BPF_CGROUP_STORAGE_TYPE; stype++) 233 #else 234 # define for_each_cgroup_storage_type(stype) for (; false; ) 235 #endif /* CONFIG_CGROUP_BPF */ 236 237 typedef void (*btf_dtor_kfunc_t)(void *); 238 239 struct btf_field_kptr { 240 struct btf *btf; 241 struct module *module; 242 /* dtor used if btf_is_kernel(btf), otherwise the type is 243 * program-allocated, dtor is NULL, and __bpf_obj_drop_impl is used 244 */ 245 btf_dtor_kfunc_t dtor; 246 u32 btf_id; 247 }; 248 249 struct btf_field_graph_root { 250 struct btf *btf; 251 u32 value_btf_id; 252 u32 node_offset; 253 struct btf_record *value_rec; 254 }; 255 256 struct btf_field { 257 u32 offset; 258 u32 size; 259 enum btf_field_type type; 260 union { 261 struct btf_field_kptr kptr; 262 struct btf_field_graph_root graph_root; 263 }; 264 }; 265 266 struct btf_record { 267 u32 cnt; 268 u32 field_mask; 269 int spin_lock_off; 270 int res_spin_lock_off; 271 int timer_off; 272 int wq_off; 273 int refcount_off; 274 int task_work_off; 275 int rcu_head_off; 276 struct btf_field fields[]; 277 }; 278 279 /* Non-opaque version of bpf_rb_node in uapi/linux/bpf.h */ 280 struct bpf_rb_node_kern { 281 struct rb_node rb_node; 282 void *owner; 283 } __attribute__((aligned(8))); 284 285 /* Non-opaque version of bpf_list_node in uapi/linux/bpf.h */ 286 struct bpf_list_node_kern { 287 struct list_head list_head; 288 void *owner; 289 } __attribute__((aligned(8))); 290 291 /* 'Ownership' of program-containing map is claimed by the first program 292 * that is going to use this map or by the first program which FD is 293 * stored in the map to make sure that all callers and callees have the 294 * same prog type, JITed flag and xdp_has_frags flag. 295 */ 296 struct bpf_map_owner { 297 enum bpf_prog_type type; 298 bool jited; 299 bool xdp_has_frags; 300 bool sleepable; 301 u64 storage_cookie[MAX_BPF_CGROUP_STORAGE_TYPE]; 302 const struct btf_type *attach_func_proto; 303 enum bpf_attach_type expected_attach_type; 304 }; 305 306 struct bpf_map { 307 u8 sha[SHA256_DIGEST_SIZE]; 308 const struct bpf_map_ops *ops; 309 struct bpf_map *inner_map_meta; 310 #ifdef CONFIG_SECURITY 311 void *security; 312 #endif 313 enum bpf_map_type map_type; 314 u32 key_size; 315 u32 value_size; 316 u32 max_entries; 317 u64 map_extra; /* any per-map-type extra fields */ 318 u32 map_flags; 319 u32 id; 320 struct btf_record *record; 321 int numa_node; 322 u32 btf_key_type_id; 323 u32 btf_value_type_id; 324 u32 btf_vmlinux_value_type_id; 325 struct btf *btf; 326 #ifdef CONFIG_MEMCG 327 struct obj_cgroup *objcg; 328 #endif 329 char name[BPF_OBJ_NAME_LEN]; 330 struct mutex freeze_mutex; 331 atomic64_t refcnt; 332 atomic64_t usercnt; 333 /* rcu is used before freeing and work is only used during freeing */ 334 union { 335 struct work_struct work; 336 struct rcu_head rcu; 337 }; 338 atomic64_t writecnt; 339 spinlock_t owner_lock; 340 struct bpf_map_owner *owner; 341 bool bypass_spec_v1; 342 bool frozen; /* write-once; write-protected by freeze_mutex */ 343 bool free_after_mult_rcu_gp; 344 bool free_after_rcu_gp; 345 bool was_inner_map; 346 atomic64_t sleepable_refcnt; 347 s64 __percpu *elem_count; 348 u64 cookie; /* write-once */ 349 char *excl_prog_sha; 350 /* 351 * Which programs use the map, see bpf_map_claim(): 0 - none so far, 352 * aux of the program - only that one, the same with BPF_MAP_USER_PATCHED 353 * set - only that one and it stored the addresses of its functions into 354 * the map, BPF_MAP_USER_MANY - more than one. 355 */ 356 unsigned long user; 357 }; 358 359 #define BPF_MAP_USER_MANY 1UL 360 #define BPF_MAP_USER_PATCHED 1UL 361 362 static inline const char *btf_field_type_name(enum btf_field_type type) 363 { 364 switch (type) { 365 case BPF_SPIN_LOCK: 366 return "bpf_spin_lock"; 367 case BPF_RES_SPIN_LOCK: 368 return "bpf_res_spin_lock"; 369 case BPF_TIMER: 370 return "bpf_timer"; 371 case BPF_WORKQUEUE: 372 return "bpf_wq"; 373 case BPF_KPTR_UNREF: 374 case BPF_KPTR_REF: 375 return "kptr"; 376 case BPF_KPTR_PERCPU: 377 return "percpu_kptr"; 378 case BPF_UPTR: 379 return "uptr"; 380 case BPF_LIST_HEAD: 381 return "bpf_list_head"; 382 case BPF_LIST_NODE: 383 return "bpf_list_node"; 384 case BPF_RB_ROOT: 385 return "bpf_rb_root"; 386 case BPF_RB_NODE: 387 return "bpf_rb_node"; 388 case BPF_REFCOUNT: 389 return "bpf_refcount"; 390 case BPF_TASK_WORK: 391 return "bpf_task_work"; 392 case BPF_RCU_HEAD: 393 return "bpf_rcu_head"; 394 default: 395 WARN_ON_ONCE(1); 396 return "unknown"; 397 } 398 } 399 400 #if IS_ENABLED(CONFIG_DEBUG_KERNEL) 401 #define BPF_WARN_ONCE(cond, format...) WARN_ONCE(cond, format) 402 #else 403 #define BPF_WARN_ONCE(cond, format...) BUILD_BUG_ON_INVALID(cond) 404 #endif 405 406 static inline u32 btf_field_type_size(enum btf_field_type type) 407 { 408 switch (type) { 409 case BPF_SPIN_LOCK: 410 return sizeof(struct bpf_spin_lock); 411 case BPF_RES_SPIN_LOCK: 412 return sizeof(struct bpf_res_spin_lock); 413 case BPF_TIMER: 414 return sizeof(struct bpf_timer); 415 case BPF_WORKQUEUE: 416 return sizeof(struct bpf_wq); 417 case BPF_KPTR_UNREF: 418 case BPF_KPTR_REF: 419 case BPF_KPTR_PERCPU: 420 case BPF_UPTR: 421 return sizeof(u64); 422 case BPF_LIST_HEAD: 423 return sizeof(struct bpf_list_head); 424 case BPF_LIST_NODE: 425 return sizeof(struct bpf_list_node); 426 case BPF_RB_ROOT: 427 return sizeof(struct bpf_rb_root); 428 case BPF_RB_NODE: 429 return sizeof(struct bpf_rb_node); 430 case BPF_REFCOUNT: 431 return sizeof(struct bpf_refcount); 432 case BPF_TASK_WORK: 433 return sizeof(struct bpf_task_work); 434 case BPF_RCU_HEAD: 435 return sizeof(struct bpf_rcu_head); 436 default: 437 WARN_ON_ONCE(1); 438 return 0; 439 } 440 } 441 442 static inline u32 btf_field_type_align(enum btf_field_type type) 443 { 444 switch (type) { 445 case BPF_SPIN_LOCK: 446 return __alignof__(struct bpf_spin_lock); 447 case BPF_RES_SPIN_LOCK: 448 return __alignof__(struct bpf_res_spin_lock); 449 case BPF_TIMER: 450 return __alignof__(struct bpf_timer); 451 case BPF_WORKQUEUE: 452 return __alignof__(struct bpf_wq); 453 case BPF_KPTR_UNREF: 454 case BPF_KPTR_REF: 455 case BPF_KPTR_PERCPU: 456 case BPF_UPTR: 457 return __alignof__(u64); 458 case BPF_LIST_HEAD: 459 return __alignof__(struct bpf_list_head); 460 case BPF_LIST_NODE: 461 return __alignof__(struct bpf_list_node); 462 case BPF_RB_ROOT: 463 return __alignof__(struct bpf_rb_root); 464 case BPF_RB_NODE: 465 return __alignof__(struct bpf_rb_node); 466 case BPF_REFCOUNT: 467 return __alignof__(struct bpf_refcount); 468 case BPF_TASK_WORK: 469 return __alignof__(struct bpf_task_work); 470 case BPF_RCU_HEAD: 471 return __alignof__(struct bpf_rcu_head); 472 default: 473 WARN_ON_ONCE(1); 474 return 0; 475 } 476 } 477 478 static inline void bpf_obj_init_field(const struct btf_field *field, void *addr) 479 { 480 memset(addr, 0, field->size); 481 482 switch (field->type) { 483 case BPF_REFCOUNT: 484 refcount_set((refcount_t *)addr, 1); 485 break; 486 case BPF_RB_NODE: 487 RB_CLEAR_NODE((struct rb_node *)addr); 488 break; 489 case BPF_LIST_HEAD: 490 case BPF_LIST_NODE: 491 INIT_LIST_HEAD((struct list_head *)addr); 492 break; 493 case BPF_RB_ROOT: 494 /* RB_ROOT_CACHED 0-inits, no need to do anything after memset */ 495 case BPF_SPIN_LOCK: 496 case BPF_RES_SPIN_LOCK: 497 case BPF_TIMER: 498 case BPF_WORKQUEUE: 499 case BPF_KPTR_UNREF: 500 case BPF_KPTR_REF: 501 case BPF_KPTR_PERCPU: 502 case BPF_UPTR: 503 case BPF_TASK_WORK: 504 case BPF_RCU_HEAD: 505 break; 506 default: 507 WARN_ON_ONCE(1); 508 return; 509 } 510 } 511 512 static inline bool btf_record_has_field(const struct btf_record *rec, enum btf_field_type type) 513 { 514 if (IS_ERR_OR_NULL(rec)) 515 return false; 516 return rec->field_mask & type; 517 } 518 519 static inline bool btf_field_is_nmi_safe(enum btf_field_type type) 520 { 521 switch (type) { 522 case BPF_SPIN_LOCK: 523 case BPF_RES_SPIN_LOCK: 524 case BPF_TIMER: 525 case BPF_WORKQUEUE: 526 case BPF_TASK_WORK: 527 case BPF_KPTR_UNREF: 528 case BPF_REFCOUNT: 529 return true; 530 default: 531 return false; 532 } 533 } 534 535 static inline bool btf_record_has_nmi_unsafe_fields(const struct btf_record *rec) 536 { 537 int i; 538 539 if (IS_ERR_OR_NULL(rec)) 540 return false; 541 for (i = 0; i < rec->cnt; i++) { 542 if (!btf_field_is_nmi_safe(rec->fields[i].type)) 543 return true; 544 } 545 return false; 546 } 547 548 static inline void bpf_obj_init(const struct btf_record *rec, void *obj) 549 { 550 int i; 551 552 if (IS_ERR_OR_NULL(rec)) 553 return; 554 for (i = 0; i < rec->cnt; i++) 555 bpf_obj_init_field(&rec->fields[i], obj + rec->fields[i].offset); 556 } 557 558 /* 'dst' must be a temporary buffer and should not point to memory that is being 559 * used in parallel by a bpf program or bpf syscall, otherwise the access from 560 * the bpf program or bpf syscall may be corrupted by the reinitialization, 561 * leading to weird problems. Even 'dst' is newly-allocated from bpf memory 562 * allocator, it is still possible for 'dst' to be used in parallel by a bpf 563 * program or bpf syscall. 564 */ 565 static inline void check_and_init_map_value(struct bpf_map *map, void *dst) 566 { 567 bpf_obj_init(map->record, dst); 568 } 569 570 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and 571 * forced to use 'long' read/writes to try to atomically copy long counters. 572 * Best-effort only. No barriers here, since it _will_ race with concurrent 573 * updates from BPF programs. Called from bpf syscall and mostly used with 574 * size 8 or 16 bytes, so ask compiler to inline it. 575 */ 576 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) 577 { 578 const long *lsrc = src; 579 long *ldst = dst; 580 581 size /= sizeof(long); 582 while (size--) 583 data_race(*ldst++ = *lsrc++); 584 } 585 586 /* copy everything but bpf_spin_lock, bpf_timer, and kptrs. There could be one of each. */ 587 static inline void bpf_obj_memcpy(struct btf_record *rec, 588 void *dst, void *src, u32 size, 589 bool long_memcpy) 590 { 591 u32 curr_off = 0; 592 int i; 593 594 if (IS_ERR_OR_NULL(rec)) { 595 if (long_memcpy) 596 bpf_long_memcpy(dst, src, size); 597 else 598 data_race(memcpy(dst, src, size)); 599 return; 600 } 601 602 for (i = 0; i < rec->cnt; i++) { 603 u32 next_off = rec->fields[i].offset; 604 u32 sz = next_off - curr_off; 605 606 data_race(memcpy(dst + curr_off, src + curr_off, sz)); 607 curr_off += rec->fields[i].size + sz; 608 } 609 data_race(memcpy(dst + curr_off, src + curr_off, size - curr_off)); 610 } 611 612 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) 613 { 614 bpf_obj_memcpy(map->record, dst, src, map->value_size, false); 615 } 616 617 static inline void copy_map_value_long(struct bpf_map *map, void *dst, void *src) 618 { 619 bpf_obj_memcpy(map->record, dst, src, round_up(map->value_size, 8), true); 620 } 621 622 static inline void bpf_obj_swap_uptrs(const struct btf_record *rec, void *dst, void *src) 623 { 624 unsigned long *src_uptr, *dst_uptr; 625 const struct btf_field *field; 626 int i; 627 628 if (!btf_record_has_field(rec, BPF_UPTR)) 629 return; 630 631 for (i = 0, field = rec->fields; i < rec->cnt; i++, field++) { 632 if (field->type != BPF_UPTR) 633 continue; 634 635 src_uptr = src + field->offset; 636 dst_uptr = dst + field->offset; 637 swap(*src_uptr, *dst_uptr); 638 } 639 } 640 641 static inline void bpf_obj_memzero(struct btf_record *rec, void *dst, u32 size) 642 { 643 u32 curr_off = 0; 644 int i; 645 646 if (IS_ERR_OR_NULL(rec)) { 647 memset(dst, 0, size); 648 return; 649 } 650 651 for (i = 0; i < rec->cnt; i++) { 652 u32 next_off = rec->fields[i].offset; 653 u32 sz = next_off - curr_off; 654 655 memset(dst + curr_off, 0, sz); 656 curr_off += rec->fields[i].size + sz; 657 } 658 memset(dst + curr_off, 0, size - curr_off); 659 } 660 661 static inline void zero_map_value(struct bpf_map *map, void *dst) 662 { 663 bpf_obj_memzero(map->record, dst, map->value_size); 664 } 665 666 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, 667 bool lock_src); 668 void bpf_timer_cancel_and_free(void *timer); 669 void bpf_wq_cancel_and_free(void *timer); 670 void bpf_task_work_cancel_and_free(void *timer); 671 void bpf_list_head_free(const struct btf_field *field, void *list_head, 672 struct bpf_spin_lock *spin_lock); 673 void bpf_rb_root_free(const struct btf_field *field, void *rb_root, 674 struct bpf_spin_lock *spin_lock); 675 u64 bpf_arena_get_kern_vm_start(struct bpf_arena *arena); 676 u64 bpf_arena_get_user_vm_start(struct bpf_arena *arena); 677 u64 bpf_arena_map_kern_vm_start(struct bpf_map *map); 678 struct bpf_map *bpf_prog_arena(struct bpf_prog *prog); 679 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size); 680 681 struct bpf_offload_dev; 682 struct bpf_offloaded_map; 683 684 struct bpf_map_dev_ops { 685 int (*map_get_next_key)(struct bpf_offloaded_map *map, 686 void *key, void *next_key); 687 int (*map_lookup_elem)(struct bpf_offloaded_map *map, 688 void *key, void *value); 689 int (*map_update_elem)(struct bpf_offloaded_map *map, 690 void *key, void *value, u64 flags); 691 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); 692 }; 693 694 struct bpf_offloaded_map { 695 struct bpf_map map; 696 struct net_device *netdev; 697 const struct bpf_map_dev_ops *dev_ops; 698 void *dev_priv; 699 struct list_head offloads; 700 }; 701 702 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) 703 { 704 return container_of(map, struct bpf_offloaded_map, map); 705 } 706 707 static inline bool bpf_map_offload_neutral(const struct bpf_map *map) 708 { 709 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; 710 } 711 712 static inline bool bpf_map_support_seq_show(const struct bpf_map *map) 713 { 714 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && 715 map->ops->map_seq_show_elem; 716 } 717 718 int map_check_no_btf(struct bpf_map *map, 719 const struct btf *btf, 720 const struct btf_type *key_type, 721 const struct btf_type *value_type); 722 723 bool bpf_map_meta_equal(const struct bpf_map *meta0, 724 const struct bpf_map *meta1); 725 726 static inline bool bpf_map_has_internal_structs(struct bpf_map *map) 727 { 728 return btf_record_has_field(map->record, BPF_TIMER | BPF_WORKQUEUE | BPF_TASK_WORK); 729 } 730 731 void bpf_map_free_internal_structs(struct bpf_map *map, void *obj); 732 733 int bpf_dynptr_from_file_sleepable(struct file *file, u32 flags, 734 struct bpf_dynptr *ptr__uninit); 735 736 #if defined(CONFIG_MMU) && defined(CONFIG_64BIT) 737 void *bpf_arena_alloc_pages_non_sleepable(void *p__map, void *addr__ign, u32 page_cnt, int node_id, 738 u64 flags); 739 void bpf_arena_free_pages_non_sleepable(void *p__map, void *ptr__ign, u32 page_cnt); 740 void *bpf_arena_alloc_pages_sleepable(void *p__map, void *addr__ign, u32 page_cnt, int node_id, 741 u64 flags); 742 #else 743 static inline void *bpf_arena_alloc_pages_non_sleepable(void *p__map, void *addr__ign, u32 page_cnt, 744 int node_id, u64 flags) 745 { 746 return NULL; 747 } 748 749 static inline void bpf_arena_free_pages_non_sleepable(void *p__map, void *ptr__ign, u32 page_cnt) 750 { 751 } 752 753 static inline void *bpf_arena_alloc_pages_sleepable(void *p__map, void *addr__ign, u32 page_cnt, 754 int node_id, u64 flags) 755 { 756 return NULL; 757 } 758 #endif 759 760 extern const struct bpf_map_ops bpf_map_offload_ops; 761 762 /* bpf_type_flag contains a set of flags that are applicable to the values of 763 * arg_type, ret_type and reg_type. For example, a pointer value may be null, 764 * or a memory is read-only. We classify types into two categories: base types 765 * and extended types. Extended types are base types combined with a type flag. 766 * 767 * Currently there are no more than 32 base types in arg_type, ret_type and 768 * reg_types. 769 */ 770 #define BPF_BASE_TYPE_BITS 8 771 772 enum bpf_type_flag { 773 /* PTR may be NULL. */ 774 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS), 775 776 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is 777 * compatible with both mutable and immutable memory. 778 */ 779 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS), 780 781 /* MEM points to BPF ring buffer reservation. */ 782 MEM_RINGBUF = BIT(2 + BPF_BASE_TYPE_BITS), 783 784 /* MEM is in user address space. */ 785 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS), 786 787 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged 788 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In 789 * order to drop this tag, it must be passed into bpf_per_cpu_ptr() 790 * or bpf_this_cpu_ptr(), which will return the pointer corresponding 791 * to the specified cpu. 792 */ 793 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS), 794 795 /* Indicates that the argument will be released. */ 796 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS), 797 798 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark 799 * unreferenced and referenced kptr loaded from map value using a load 800 * instruction, so that they can only be dereferenced but not escape the 801 * BPF program into the kernel (i.e. cannot be passed as arguments to 802 * kfunc or bpf helpers). 803 */ 804 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS), 805 806 /* 807 * MEM can be uninitialized. Generic memory outputs need not be fully 808 * initialized by the callee. 809 */ 810 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS), 811 812 /* DYNPTR points to memory local to the bpf program. */ 813 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS), 814 815 /* DYNPTR points to a kernel-produced ringbuf record. */ 816 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS), 817 818 /* Size is known at compile time. */ 819 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS), 820 821 /* MEM is of an allocated object of type in program BTF. This is used to 822 * tag PTR_TO_BTF_ID allocated using bpf_obj_new. 823 */ 824 MEM_ALLOC = BIT(11 + BPF_BASE_TYPE_BITS), 825 826 /* PTR was passed from the kernel in a trusted context, and may be 827 * passed to kfuncs or BPF helper functions. 828 * Confusingly, this is _not_ the opposite of PTR_UNTRUSTED above. 829 * PTR_UNTRUSTED refers to a kptr that was read directly from a map 830 * without invoking bpf_kptr_xchg(). What we really need to know is 831 * whether a pointer is safe to pass to a kfunc or BPF helper function. 832 * While PTR_UNTRUSTED pointers are unsafe to pass to kfuncs and BPF 833 * helpers, they do not cover all possible instances of unsafe 834 * pointers. For example, a pointer that was obtained from walking a 835 * struct will _not_ get the PTR_UNTRUSTED type modifier, despite the 836 * fact that it may be NULL, invalid, etc. This is due to backwards 837 * compatibility requirements, as this was the behavior that was first 838 * introduced when kptrs were added. The behavior is now considered 839 * deprecated, and PTR_UNTRUSTED will eventually be removed. 840 * 841 * PTR_TRUSTED, on the other hand, is a pointer that the kernel 842 * guarantees to be valid and safe to pass to kfuncs and BPF helpers. 843 * For example, pointers passed to tracepoint arguments are considered 844 * PTR_TRUSTED, as are pointers that are passed to struct_ops 845 * callbacks. As alluded to above, pointers that are obtained from 846 * walking PTR_TRUSTED pointers are _not_ trusted. For example, if a 847 * struct task_struct *task is PTR_TRUSTED, then accessing 848 * task->last_wakee will lose the PTR_TRUSTED modifier when it's stored 849 * in a BPF register. Similarly, pointers passed to certain programs 850 * types such as kretprobes are not guaranteed to be valid, as they may 851 * for example contain an object that was recently freed. 852 */ 853 PTR_TRUSTED = BIT(12 + BPF_BASE_TYPE_BITS), 854 855 /* MEM is tagged with rcu and memory access needs rcu_read_lock protection. */ 856 MEM_RCU = BIT(13 + BPF_BASE_TYPE_BITS), 857 858 /* Used to tag PTR_TO_BTF_ID | MEM_ALLOC references which are non-owning. 859 * Currently only valid for linked-list and rbtree nodes. If the nodes 860 * have a bpf_refcount_field, they must be tagged MEM_RCU as well. 861 */ 862 NON_OWN_REF = BIT(14 + BPF_BASE_TYPE_BITS), 863 864 /* DYNPTR points to sk_buff */ 865 DYNPTR_TYPE_SKB = BIT(15 + BPF_BASE_TYPE_BITS), 866 867 /* DYNPTR points to xdp_buff */ 868 DYNPTR_TYPE_XDP = BIT(16 + BPF_BASE_TYPE_BITS), 869 870 /* Memory must be aligned on some architectures, used in combination with 871 * MEM_FIXED_SIZE. 872 */ 873 MEM_ALIGNED = BIT(17 + BPF_BASE_TYPE_BITS), 874 875 /* MEM is being written to, often combined with MEM_UNINIT. Non-presence 876 * of MEM_WRITE means that MEM is only being read. MEM_WRITE without the 877 * MEM_UNINIT means that memory needs to be initialized since it is also 878 * read. 879 */ 880 MEM_WRITE = BIT(18 + BPF_BASE_TYPE_BITS), 881 882 /* DYNPTR points to skb_metadata_end()-skb_metadata_len() */ 883 DYNPTR_TYPE_SKB_META = BIT(19 + BPF_BASE_TYPE_BITS), 884 885 /* DYNPTR points to file */ 886 DYNPTR_TYPE_FILE = BIT(20 + BPF_BASE_TYPE_BITS), 887 888 __BPF_TYPE_FLAG_MAX, 889 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1, 890 }; 891 892 #define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF | DYNPTR_TYPE_SKB \ 893 | DYNPTR_TYPE_XDP | DYNPTR_TYPE_SKB_META | DYNPTR_TYPE_FILE) 894 895 /* Max number of base types. */ 896 #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS) 897 898 /* Max number of all types. */ 899 #define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1)) 900 901 /* function argument constraints */ 902 enum bpf_arg_type { 903 ARG_UNUSED = 0, /* unused argument; terminates argument iteration */ 904 905 /* the following constraints used to prototype 906 * bpf_map_lookup/update/delete_elem() functions 907 */ 908 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ 909 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ 910 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ 911 912 /* Used to prototype bpf_memcmp() and other functions that access data 913 * on eBPF program stack 914 */ 915 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ 916 ARG_PTR_TO_ARENA, 917 918 ARG_MEM_SIZE, /* number of bytes accessed from memory */ 919 ARG_MEM_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ 920 921 ARG_PTR_TO_CTX, /* pointer to context */ 922 ARG_ANYTHING, /* any (initialized) argument is ok */ 923 ARG_SCALAR, /* scalar argument */ 924 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ 925 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ 926 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */ 927 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */ 928 ARG_PTR_TO_RINGBUF_MEM, /* pointer to dynamically reserved ringbuf memory */ 929 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */ 930 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */ 931 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */ 932 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */ 933 ARG_PTR_TO_STACK, /* pointer to stack */ 934 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */ 935 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */ 936 ARG_KPTR_XCHG_DEST, /* pointer to destination that kptrs are bpf_kptr_xchg'd into */ 937 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */ 938 939 ARG_CONST_SCALAR, /* scalar known at verification time */ 940 ARG_CONST_MEM_SIZE, /* ARG_MEM_SIZE that must be constant */ 941 ARG_PTR_TO_ALLOC_BTF_ID, /* pointer to an allocated object */ 942 ARG_PTR_TO_REFCOUNTED_KPTR, /* pointer to a refcounted local kptr */ 943 ARG_PTR_TO_ITER, /* pointer to an iterator */ 944 ARG_PTR_TO_LIST_HEAD, /* pointer to bpf_list_head */ 945 ARG_PTR_TO_LIST_NODE, /* pointer to bpf_list_node */ 946 ARG_PTR_TO_RB_ROOT, /* pointer to bpf_rb_root */ 947 ARG_PTR_TO_RB_NODE, /* pointer to bpf_rb_node */ 948 ARG_PTR_TO_WORKQUEUE, /* pointer to bpf_wq */ 949 ARG_PTR_TO_TASK_WORK, /* pointer to bpf_task_work */ 950 ARG_PTR_TO_RCU_HEAD, /* pointer to bpf_rcu_head */ 951 ARG_PTR_TO_IRQ_FLAG, /* pointer to saved IRQ flags on the stack */ 952 ARG_PTR_TO_RES_SPIN_LOCK, /* pointer to bpf_res_spin_lock */ 953 ARG_PTR_TO_CTX_OUT, /* hook output argument passed through from ctx */ 954 ARG_PTR_TO_PROG_AUX, /* pointer to the caller's bpf_prog_aux */ 955 ARG_IGNORE, /* argument the verifier does not check at all */ 956 __BPF_ARG_TYPE_MAX, 957 958 /* Extended arg_types. */ 959 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE, 960 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM, 961 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX, 962 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET, 963 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK, 964 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID, 965 /* Pointer to memory does not need to be initialized, since helper function 966 * fills all bytes or clears them in error case. 967 */ 968 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | MEM_WRITE | ARG_PTR_TO_MEM, 969 /* Pointer to valid memory of size known at compile time. */ 970 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM, 971 972 /* This must be the last entry. Its purpose is to ensure the enum is 973 * wide enough to hold the higher bits reserved for bpf_type_flag. 974 */ 975 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT, 976 }; 977 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 978 979 /* type of values returned from helper functions */ 980 enum bpf_return_type { 981 RET_INTEGER, /* function returns integer */ 982 RET_VOID, /* function doesn't return anything */ 983 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ 984 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */ 985 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */ 986 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */ 987 RET_PTR_TO_MEM, /* returns a pointer to memory */ 988 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */ 989 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */ 990 __BPF_RET_TYPE_MAX, 991 992 /* Extended ret_types. */ 993 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE, 994 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET, 995 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK, 996 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON, 997 RET_PTR_TO_RINGBUF_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_RINGBUF | RET_PTR_TO_MEM, 998 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MEM, 999 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID, 1000 RET_PTR_TO_BTF_ID_TRUSTED = PTR_TRUSTED | RET_PTR_TO_BTF_ID, 1001 1002 /* This must be the last entry. Its purpose is to ensure the enum is 1003 * wide enough to hold the higher bits reserved for bpf_type_flag. 1004 */ 1005 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT, 1006 }; 1007 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 1008 1009 /* The longest tracepoint has 12 args. 1010 * See include/trace/bpf_probe.h 1011 * 1012 * Also reuse this macro for maximum number of arguments a BPF function 1013 * or a kfunc can have. Args 1-5 are passed in registers, args 6-12 via 1014 * stack arg slots. The JIT may map some stack arg slots to registers based 1015 * on the native calling convention (e.g., arg 6 to R9 on x86-64). 1016 */ 1017 #define MAX_BPF_FUNC_ARGS 12 1018 1019 /* The maximum number of arguments passed through registers 1020 * a single function may have. 1021 */ 1022 #define MAX_BPF_FUNC_REG_ARGS 5 1023 1024 /* A by-value argument takes two eightbytes at most, so the maximum number of 1025 * argument slots of any function is 2 * MAX_BPF_FUNC_ARGS. A local array may 1026 * need that size for processing, although eventually the maximum slots will 1027 * be capped at MAX_BPF_FUNC_ARGS. 1028 */ 1029 #define MAX_BPF_FUNC_ARG_SLOTS (2 * MAX_BPF_FUNC_ARGS) 1030 1031 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs 1032 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL 1033 * instructions after verifying 1034 */ 1035 struct bpf_func_proto { 1036 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 1037 bool gpl_only; 1038 bool pkt_access; 1039 bool might_sleep; 1040 /* set to true if helper follows contract for llvm 1041 * attribute bpf_fastcall: 1042 * - void functions do not scratch r0 1043 * - functions taking N arguments scratch only registers r1-rN 1044 */ 1045 bool allow_fastcall; 1046 enum bpf_return_type ret_type; 1047 union { 1048 struct { 1049 enum bpf_arg_type arg1_type; 1050 enum bpf_arg_type arg2_type; 1051 enum bpf_arg_type arg3_type; 1052 enum bpf_arg_type arg4_type; 1053 enum bpf_arg_type arg5_type; 1054 }; 1055 enum bpf_arg_type arg_type[MAX_BPF_FUNC_ARGS]; 1056 }; 1057 union { 1058 struct { 1059 const u32 *arg1_btf_id; 1060 const u32 *arg2_btf_id; 1061 const u32 *arg3_btf_id; 1062 const u32 *arg4_btf_id; 1063 const u32 *arg5_btf_id; 1064 }; 1065 const u32 *arg_btf_id[MAX_BPF_FUNC_ARGS]; 1066 struct { 1067 size_t arg1_size; 1068 size_t arg2_size; 1069 size_t arg3_size; 1070 size_t arg4_size; 1071 size_t arg5_size; 1072 }; 1073 size_t arg_size[MAX_BPF_FUNC_ARGS]; 1074 }; 1075 int *ret_btf_id; /* return value btf_id */ 1076 bool (*allowed)(const struct bpf_prog *prog); 1077 }; 1078 1079 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is 1080 * the first argument to eBPF programs. 1081 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' 1082 */ 1083 struct bpf_context; 1084 1085 enum bpf_access_type { 1086 BPF_READ = 1, 1087 BPF_WRITE = 2 1088 }; 1089 1090 /* types of values stored in eBPF registers */ 1091 /* Pointer types represent: 1092 * pointer 1093 * pointer + imm 1094 * pointer + (u16) var 1095 * pointer + (u16) var + imm 1096 * if (range > 0) then [ptr, ptr + range - off) is safe to access 1097 * if (id > 0) means that some 'var' was added 1098 * if (off > 0) means that 'imm' was added 1099 */ 1100 enum bpf_reg_type { 1101 NOT_INIT = 0, /* nothing was written into register */ 1102 SCALAR_VALUE, /* reg doesn't contain a valid pointer */ 1103 PTR_TO_CTX, /* reg points to bpf_context */ 1104 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ 1105 PTR_TO_MAP_VALUE, /* reg points to map element value */ 1106 PTR_TO_MAP_KEY, /* reg points to a map element key */ 1107 PTR_TO_STACK, /* reg == frame_pointer + offset */ 1108 PTR_TO_PACKET_META, /* skb->data - meta_len */ 1109 PTR_TO_PACKET, /* reg points to skb->data */ 1110 PTR_TO_PACKET_END, /* skb->data + headlen */ 1111 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ 1112 PTR_TO_SOCKET, /* reg points to struct bpf_sock */ 1113 PTR_TO_SOCK_COMMON, /* reg points to sock_common */ 1114 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ 1115 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */ 1116 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */ 1117 /* PTR_TO_BTF_ID points to a kernel struct that does not need 1118 * to be null checked by the BPF program. This does not imply the 1119 * pointer is _not_ null and in practice this can easily be a null 1120 * pointer when reading pointer chains. The assumption is program 1121 * context will handle null pointer dereference typically via fault 1122 * handling. The verifier must keep this in mind and can make no 1123 * assumptions about null or non-null when doing branch analysis. 1124 * Further, when passed into helpers the helpers can not, without 1125 * additional context, assume the value is non-null. 1126 */ 1127 PTR_TO_BTF_ID, 1128 PTR_TO_MEM, /* reg points to valid memory region */ 1129 PTR_TO_ARENA, 1130 PTR_TO_BUF, /* reg points to a read/write buffer */ 1131 PTR_TO_FUNC, /* reg points to a bpf program function */ 1132 PTR_TO_INSN, /* reg points to a bpf program instruction */ 1133 CONST_PTR_TO_DYNPTR, /* reg points to a const struct bpf_dynptr */ 1134 __BPF_REG_TYPE_MAX, 1135 1136 /* Extended reg_types. */ 1137 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE, 1138 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET, 1139 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON, 1140 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK, 1141 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not 1142 * been checked for null. Used primarily to inform the verifier 1143 * an explicit null check is required for this struct. 1144 */ 1145 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID, 1146 1147 /* This must be the last entry. Its purpose is to ensure the enum is 1148 * wide enough to hold the higher bits reserved for bpf_type_flag. 1149 */ 1150 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT, 1151 }; 1152 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT); 1153 1154 /* The information passed from prog-specific *_is_valid_access 1155 * back to the verifier. 1156 */ 1157 struct bpf_insn_access_aux { 1158 enum bpf_reg_type reg_type; 1159 bool is_ldsx; 1160 union { 1161 int ctx_field_size; 1162 struct { 1163 struct btf *btf; 1164 u32 btf_id; 1165 u32 ref_id; 1166 }; 1167 }; 1168 u32 mem_size; 1169 struct bpf_verifier_log *log; /* for verbose logs */ 1170 bool is_retval; /* is accessing function return value ? */ 1171 }; 1172 1173 static inline void 1174 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) 1175 { 1176 aux->ctx_field_size = size; 1177 } 1178 1179 static bool bpf_is_ldimm64(const struct bpf_insn *insn) 1180 { 1181 return insn->code == (BPF_LD | BPF_IMM | BPF_DW); 1182 } 1183 1184 static inline bool bpf_pseudo_func(const struct bpf_insn *insn) 1185 { 1186 return bpf_is_ldimm64(insn) && insn->src_reg == BPF_PSEUDO_FUNC; 1187 } 1188 1189 struct bpf_prog_ops { 1190 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, 1191 union bpf_attr __user *uattr); 1192 }; 1193 1194 struct bpf_reg_state; 1195 struct bpf_verifier_ops { 1196 /* return eBPF function prototype for verification */ 1197 const struct bpf_func_proto * 1198 (*get_func_proto)(enum bpf_func_id func_id, 1199 const struct bpf_prog *prog); 1200 1201 /* return true if 'size' wide access at offset 'off' within bpf_context 1202 * with 'type' (read or write) is allowed 1203 */ 1204 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, 1205 const struct bpf_prog *prog, 1206 struct bpf_insn_access_aux *info); 1207 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, 1208 const struct bpf_prog *prog); 1209 int (*gen_epilogue)(struct bpf_insn *insn, const struct bpf_prog *prog, 1210 s16 ctx_stack_off); 1211 int (*gen_ld_abs)(const struct bpf_insn *orig, 1212 struct bpf_insn *insn_buf); 1213 u32 (*convert_ctx_access)(enum bpf_access_type type, 1214 const struct bpf_insn *src, 1215 struct bpf_insn *dst, 1216 struct bpf_prog *prog, u32 *target_size); 1217 int (*btf_struct_access)(struct bpf_verifier_log *log, 1218 const struct bpf_reg_state *reg, 1219 int off, int size); 1220 }; 1221 1222 struct bpf_prog_offload_ops { 1223 /* verifier basic callbacks */ 1224 int (*insn_hook)(struct bpf_verifier_env *env, 1225 int insn_idx, int prev_insn_idx); 1226 int (*finalize)(struct bpf_verifier_env *env); 1227 /* verifier optimization callbacks (called after .finalize) */ 1228 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, 1229 struct bpf_insn *insn); 1230 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); 1231 /* program management callbacks */ 1232 int (*prepare)(struct bpf_prog *prog); 1233 int (*translate)(struct bpf_prog *prog); 1234 void (*destroy)(struct bpf_prog *prog); 1235 }; 1236 1237 struct bpf_prog_offload { 1238 struct bpf_prog *prog; 1239 struct net_device *netdev; 1240 struct bpf_offload_dev *offdev; 1241 void *dev_priv; 1242 struct list_head offloads; 1243 bool dev_state; 1244 bool opt_failed; 1245 void *jited_image; 1246 u32 jited_len; 1247 }; 1248 1249 /* The argument is aligned to 16 bytes. */ 1250 #define BTF_FMODEL_ALIGN16_ARG BIT(0) 1251 1252 /* The argument is signed. */ 1253 #define BTF_FMODEL_SIGNED_ARG BIT(1) 1254 1255 /* The argument is an arena pointer. */ 1256 #define BTF_FMODEL_ARENA_ARG BIT(2) 1257 1258 /* The argument is nullable. */ 1259 #define BTF_FMODEL_NULLABLE_ARG BIT(3) 1260 1261 struct btf_func_model { 1262 u8 ret_size; 1263 u8 ret_flags; 1264 u8 nr_args; 1265 u8 arg_size[MAX_BPF_FUNC_ARGS]; 1266 u8 arg_flags[MAX_BPF_FUNC_ARGS]; 1267 }; 1268 1269 static inline u32 btf_func_model_arg_slots(const struct btf_func_model *m, u32 arg) 1270 { 1271 return (m->arg_size[arg] + sizeof(u64) - 1) / sizeof(u64); 1272 } 1273 1274 /* Restore arguments before returning from trampoline to let original function 1275 * continue executing. This flag is used for fentry progs when there are no 1276 * fexit progs. 1277 */ 1278 #define BPF_TRAMP_F_RESTORE_REGS BIT(0) 1279 /* Call original function after fentry progs, but before fexit progs. 1280 * Makes sense for fentry/fexit, normal calls and indirect calls. 1281 */ 1282 #define BPF_TRAMP_F_CALL_ORIG BIT(1) 1283 /* Skip current frame and return to parent. Makes sense for fentry/fexit 1284 * programs only. Should not be used with normal calls and indirect calls. 1285 */ 1286 #define BPF_TRAMP_F_SKIP_FRAME BIT(2) 1287 /* Store IP address of the caller on the trampoline stack, 1288 * so it's available for trampoline's programs. 1289 */ 1290 #define BPF_TRAMP_F_IP_ARG BIT(3) 1291 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */ 1292 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4) 1293 1294 /* Get original function from stack instead of from provided direct address. 1295 * Makes sense for trampolines with fexit or fmod_ret programs. 1296 */ 1297 #define BPF_TRAMP_F_ORIG_STACK BIT(5) 1298 1299 /* This trampoline is on a function with another ftrace_ops with IPMODIFY, 1300 * e.g., a live patch. This flag is set and cleared by ftrace call backs, 1301 */ 1302 #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6) 1303 1304 /* Indicate that current trampoline is in a tail call context. Then, it has to 1305 * cache and restore tail_call_cnt to avoid infinite tail call loop. 1306 */ 1307 #define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7) 1308 1309 /* 1310 * Indicate the trampoline should be suitable to receive indirect calls; 1311 * without this indirectly calling the generated code can result in #UD/#CP, 1312 * depending on the CFI options. 1313 * 1314 * Used by bpf_struct_ops. 1315 * 1316 * Incompatible with FENTRY usage, overloads @func_addr argument. 1317 */ 1318 #define BPF_TRAMP_F_INDIRECT BIT(8) 1319 1320 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50 1321 * bytes on x86. The trampoline image has to fit in PAGE_SIZE. 1322 */ 1323 enum { 1324 #if defined(__s390x__) || defined(__powerpc64__) 1325 BPF_MAX_TRAMP_LINKS = 27, 1326 #elif defined(__x86_64__) 1327 BPF_MAX_TRAMP_LINKS = 36, 1328 #elif defined(__aarch64__) 1329 BPF_MAX_TRAMP_LINKS = 37, 1330 #else 1331 BPF_MAX_TRAMP_LINKS = 38, 1332 #endif 1333 }; 1334 1335 #define BPF_TRAMP_COOKIE_INDEX_SHIFT 8 1336 #define BPF_TRAMP_IS_RETURN_SHIFT 63 1337 1338 struct bpf_tramp_nodes { 1339 struct bpf_tramp_node *nodes[BPF_MAX_TRAMP_LINKS]; 1340 int nr_nodes; 1341 }; 1342 1343 /* 1344 * The arena base against which a struct_ops trampoline converts the 1345 * arguments marked with BTF_FMODEL_ARENA_ARG while saving them into the BPF 1346 * ctx, ctx[arg] = (u32)(kaddr - kern_vm_start). Zero when the trampoline 1347 * converts nothing. 1348 */ 1349 u64 bpf_tramp_arena_base(const struct btf_func_model *m, 1350 struct bpf_tramp_nodes *tnodes, u32 flags); 1351 1352 struct bpf_tramp_run_ctx; 1353 1354 /* Different use cases for BPF trampoline: 1355 * 1. replace nop at the function entry (kprobe equivalent) 1356 * flags = BPF_TRAMP_F_RESTORE_REGS 1357 * fentry = a set of programs to run before returning from trampoline 1358 * 1359 * 2. replace nop at the function entry (kprobe + kretprobe equivalent) 1360 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME 1361 * orig_call = fentry_ip + MCOUNT_INSN_SIZE 1362 * fentry = a set of program to run before calling original function 1363 * fexit = a set of program to run after original function 1364 * 1365 * 3. replace direct call instruction anywhere in the function body 1366 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid) 1367 * With flags = 0 1368 * fentry = a set of programs to run before returning from trampoline 1369 * With flags = BPF_TRAMP_F_CALL_ORIG 1370 * orig_call = original callback addr or direct function addr 1371 * fentry = a set of program to run before calling original function 1372 * fexit = a set of program to run after original function 1373 */ 1374 struct bpf_tramp_image; 1375 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *image_end, 1376 const struct btf_func_model *m, u32 flags, 1377 struct bpf_tramp_nodes *tnodes, 1378 void *func_addr); 1379 void *arch_alloc_bpf_trampoline(unsigned int size); 1380 void arch_free_bpf_trampoline(void *image, unsigned int size); 1381 int __must_check arch_protect_bpf_trampoline(void *image, unsigned int size); 1382 int arch_bpf_trampoline_skip(void *nop, void *target); 1383 int arch_bpf_trampoline_size(const struct btf_func_model *m, u32 flags, 1384 struct bpf_tramp_nodes *tnodes, void *func_addr); 1385 1386 u64 notrace __bpf_prog_enter_sleepable_recur(struct bpf_prog *prog, 1387 struct bpf_tramp_run_ctx *run_ctx); 1388 void notrace __bpf_prog_exit_sleepable_recur(struct bpf_prog *prog, u64 start, 1389 struct bpf_tramp_run_ctx *run_ctx); 1390 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr); 1391 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr); 1392 typedef u64 (*bpf_trampoline_enter_t)(struct bpf_prog *prog, 1393 struct bpf_tramp_run_ctx *run_ctx); 1394 typedef void (*bpf_trampoline_exit_t)(struct bpf_prog *prog, u64 start, 1395 struct bpf_tramp_run_ctx *run_ctx); 1396 bpf_trampoline_enter_t bpf_trampoline_enter(const struct bpf_prog *prog); 1397 bpf_trampoline_exit_t bpf_trampoline_exit(const struct bpf_prog *prog); 1398 1399 #ifdef CONFIG_DYNAMIC_FTRACE_WITH_JMP 1400 static inline bool bpf_trampoline_use_jmp(u64 flags) 1401 { 1402 return flags & BPF_TRAMP_F_CALL_ORIG && !(flags & BPF_TRAMP_F_SKIP_FRAME); 1403 } 1404 #else 1405 static inline bool bpf_trampoline_use_jmp(u64 flags) 1406 { 1407 return false; 1408 } 1409 #endif 1410 1411 struct bpf_ksym { 1412 unsigned long start; 1413 unsigned long end; 1414 char name[KSYM_NAME_LEN]; 1415 struct list_head lnode; 1416 struct latch_tree_node tnode; 1417 bool prog; 1418 u32 fp_start; 1419 u32 fp_end; 1420 }; 1421 1422 enum bpf_tramp_prog_type { 1423 BPF_TRAMP_FENTRY, 1424 BPF_TRAMP_FEXIT, 1425 BPF_TRAMP_MODIFY_RETURN, 1426 BPF_TRAMP_MAX, 1427 BPF_TRAMP_REPLACE, /* more than MAX */ 1428 BPF_TRAMP_FSESSION, 1429 }; 1430 1431 /* 1432 * Each prog call in a trampoline image is preceded by a nop. When the prog is 1433 * detached, the nop is patched to a jump to target, right after the call, so 1434 * that tasks still running in the image skip the prog. 1435 */ 1436 struct bpf_tramp_skip { 1437 struct bpf_prog *prog; 1438 void *nop; 1439 void *target; 1440 }; 1441 1442 struct bpf_tramp_image { 1443 void *image; 1444 int size; 1445 struct bpf_ksym ksym; 1446 struct percpu_ref pcref; 1447 bool call_orig; 1448 /* entry in tr->images, the image holds a reference on tr */ 1449 struct bpf_trampoline *tr; 1450 struct list_head list; 1451 int nr_skips; 1452 struct bpf_tramp_skip *skips; 1453 union { 1454 struct rcu_head rcu; 1455 struct work_struct work; 1456 }; 1457 }; 1458 1459 static inline void bpf_tramp_image_add_skip(struct bpf_tramp_image *im, struct bpf_prog *prog, 1460 void *nop, void *target) 1461 { 1462 struct bpf_tramp_skip *skip; 1463 1464 /* struct_ops trampolines and arch_bpf_trampoline_size() have no image */ 1465 if (!im || !im->skips) 1466 return; 1467 skip = &im->skips[im->nr_skips++]; 1468 skip->prog = prog; 1469 skip->nop = nop; 1470 skip->target = target; 1471 } 1472 1473 struct bpf_trampoline { 1474 /* hlist for trampoline_key_table */ 1475 struct hlist_node hlist_key; 1476 /* hlist for trampoline_ip_table */ 1477 struct hlist_node hlist_ip; 1478 struct ftrace_ops *fops; 1479 refcount_t refcnt; 1480 u32 flags; 1481 u64 key; 1482 unsigned long ip; 1483 struct { 1484 struct btf_func_model model; 1485 void *addr; 1486 bool ftrace_managed; 1487 } func; 1488 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF 1489 * program by replacing one of its functions. func.addr is the address 1490 * of the function it replaced. 1491 */ 1492 struct bpf_prog *extension_prog; 1493 /* list of BPF programs using this trampoline */ 1494 struct hlist_head progs_hlist[BPF_TRAMP_MAX]; 1495 /* Number of attached programs. A counter per kind. */ 1496 int progs_cnt[BPF_TRAMP_MAX]; 1497 /* Executable image of trampoline */ 1498 struct bpf_tramp_image *cur_image; 1499 /* Images not freed yet, cur_image and older ones still in use */ 1500 struct list_head images; 1501 /* Used as temporary old image storage for multi_attach */ 1502 struct { 1503 struct bpf_tramp_image *old_image; 1504 u32 old_flags; 1505 } multi_attach; 1506 }; 1507 1508 struct bpf_attach_target_info { 1509 struct btf_func_model fmodel; 1510 long tgt_addr; 1511 struct module *tgt_mod; 1512 const char *tgt_name; 1513 const struct btf_type *tgt_type; 1514 }; 1515 1516 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */ 1517 1518 struct bpf_dispatcher_prog { 1519 struct bpf_prog *prog; 1520 refcount_t users; 1521 }; 1522 1523 struct bpf_dispatcher { 1524 /* dispatcher mutex */ 1525 struct mutex mutex; 1526 void *func; 1527 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX]; 1528 int num_progs; 1529 void *image; 1530 void *rw_image; 1531 u32 image_off; 1532 struct bpf_ksym ksym; 1533 #ifdef CONFIG_HAVE_STATIC_CALL 1534 struct static_call_key *sc_key; 1535 void *sc_tramp; 1536 #endif 1537 }; 1538 1539 #ifndef __bpfcall 1540 #define __bpfcall __nocfi 1541 #endif 1542 1543 static __always_inline __bpfcall unsigned int bpf_dispatcher_nop_func( 1544 const void *ctx, 1545 const struct bpf_insn *insnsi, 1546 bpf_func_t bpf_func) 1547 { 1548 return bpf_func(ctx, insnsi); 1549 } 1550 1551 /* the implementation of the opaque uapi struct bpf_dynptr */ 1552 struct bpf_dynptr_kern { 1553 void *data; 1554 /* Size represents the number of usable bytes of dynptr data. 1555 * If for example the offset is at 4 for a local dynptr whose data is 1556 * of type u64, the number of usable bytes is 4. 1557 * 1558 * The upper 8 bits are reserved. It is as follows: 1559 * Bits 0 - 23 = size 1560 * Bits 24 - 30 = dynptr type 1561 * Bit 31 = whether dynptr is read-only 1562 */ 1563 u32 size; 1564 u32 offset; 1565 } __aligned(8); 1566 1567 enum bpf_dynptr_type { 1568 BPF_DYNPTR_TYPE_INVALID, 1569 /* Points to memory that is local to the bpf program */ 1570 BPF_DYNPTR_TYPE_LOCAL, 1571 /* Underlying data is a ringbuf record */ 1572 BPF_DYNPTR_TYPE_RINGBUF, 1573 /* Underlying data is a sk_buff */ 1574 BPF_DYNPTR_TYPE_SKB, 1575 /* Underlying data is a xdp_buff */ 1576 BPF_DYNPTR_TYPE_XDP, 1577 /* Points to skb_metadata_end()-skb_metadata_len() */ 1578 BPF_DYNPTR_TYPE_SKB_META, 1579 /* Underlying data is a file */ 1580 BPF_DYNPTR_TYPE_FILE, 1581 }; 1582 1583 int bpf_dynptr_check_size(u64 size); 1584 u64 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr); 1585 const void *__bpf_dynptr_data(const struct bpf_dynptr_kern *ptr, u64 len); 1586 void *__bpf_dynptr_data_rw(const struct bpf_dynptr_kern *ptr, u64 len); 1587 bool __bpf_dynptr_is_rdonly(const struct bpf_dynptr_kern *ptr); 1588 int __bpf_dynptr_write(const struct bpf_dynptr_kern *dst, u64 offset, 1589 void *src, u64 len, u64 flags); 1590 void *bpf_dynptr_slice_rdwr(const struct bpf_dynptr *p, u64 offset, 1591 void *buffer__nullable, u64 buffer__szk); 1592 1593 static inline int bpf_dynptr_check_off_len(const struct bpf_dynptr_kern *ptr, u64 offset, u64 len) 1594 { 1595 u64 size = __bpf_dynptr_size(ptr); 1596 1597 if (len > size || offset > size - len) 1598 return -E2BIG; 1599 1600 return 0; 1601 } 1602 1603 struct bpf_tracing_multi_link; 1604 1605 #ifdef CONFIG_BPF_JIT 1606 int bpf_trampoline_link_prog(struct bpf_tramp_node *node, 1607 struct bpf_trampoline *tr, 1608 struct bpf_prog *tgt_prog); 1609 int bpf_trampoline_unlink_prog(struct bpf_tramp_node *node, 1610 struct bpf_trampoline *tr, 1611 struct bpf_prog *tgt_prog); 1612 struct bpf_trampoline *bpf_trampoline_get(u64 key, 1613 struct bpf_attach_target_info *tgt_info); 1614 void bpf_trampoline_put(struct bpf_trampoline *tr); 1615 int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_funcs); 1616 1617 int bpf_trampoline_multi_attach(struct bpf_prog *prog, u32 *ids, 1618 struct bpf_tracing_multi_link *link); 1619 void bpf_trampoline_multi_detach(struct bpf_prog *prog, 1620 struct bpf_tracing_multi_link *link); 1621 void bpf_trampoline_set_flags(struct bpf_trampoline *tr, u32 flags); 1622 1623 /* 1624 * When the architecture supports STATIC_CALL replace the bpf_dispatcher_fn 1625 * indirection with a direct call to the bpf program. If the architecture does 1626 * not have STATIC_CALL, avoid a double-indirection. 1627 */ 1628 #ifdef CONFIG_HAVE_STATIC_CALL 1629 1630 #define __BPF_DISPATCHER_SC_INIT(_name) \ 1631 .sc_key = &STATIC_CALL_KEY(_name), \ 1632 .sc_tramp = STATIC_CALL_TRAMP_ADDR(_name), 1633 1634 #define __BPF_DISPATCHER_SC(name) \ 1635 DEFINE_STATIC_CALL(bpf_dispatcher_##name##_call, bpf_dispatcher_nop_func) 1636 1637 #define __BPF_DISPATCHER_CALL(name) \ 1638 static_call(bpf_dispatcher_##name##_call)(ctx, insnsi, bpf_func) 1639 1640 #define __BPF_DISPATCHER_UPDATE(_d, _new) \ 1641 __static_call_update((_d)->sc_key, (_d)->sc_tramp, (_new)) 1642 1643 #else 1644 #define __BPF_DISPATCHER_SC_INIT(name) 1645 #define __BPF_DISPATCHER_SC(name) 1646 #define __BPF_DISPATCHER_CALL(name) bpf_func(ctx, insnsi) 1647 #define __BPF_DISPATCHER_UPDATE(_d, _new) 1648 #endif 1649 1650 #define BPF_DISPATCHER_INIT(_name) { \ 1651 .mutex = __MUTEX_INITIALIZER(_name.mutex), \ 1652 .func = &_name##_func, \ 1653 .progs = {}, \ 1654 .num_progs = 0, \ 1655 .image = NULL, \ 1656 .image_off = 0, \ 1657 .ksym = { \ 1658 .name = #_name, \ 1659 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \ 1660 }, \ 1661 __BPF_DISPATCHER_SC_INIT(_name##_call) \ 1662 } 1663 1664 #define DEFINE_BPF_DISPATCHER(name) \ 1665 __BPF_DISPATCHER_SC(name); \ 1666 noinline __bpfcall unsigned int bpf_dispatcher_##name##_func( \ 1667 const void *ctx, \ 1668 const struct bpf_insn *insnsi, \ 1669 bpf_func_t bpf_func) \ 1670 { \ 1671 return __BPF_DISPATCHER_CALL(name); \ 1672 } \ 1673 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \ 1674 struct bpf_dispatcher bpf_dispatcher_##name = \ 1675 BPF_DISPATCHER_INIT(bpf_dispatcher_##name); 1676 1677 #define DECLARE_BPF_DISPATCHER(name) \ 1678 unsigned int bpf_dispatcher_##name##_func( \ 1679 const void *ctx, \ 1680 const struct bpf_insn *insnsi, \ 1681 bpf_func_t bpf_func); \ 1682 extern struct bpf_dispatcher bpf_dispatcher_##name; 1683 1684 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func 1685 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name) 1686 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, 1687 struct bpf_prog *to); 1688 /* Called only from JIT-enabled code, so there's no need for stubs. */ 1689 void bpf_image_ksym_init(void *data, unsigned int size, struct bpf_ksym *ksym); 1690 void bpf_image_ksym_add(struct bpf_ksym *ksym); 1691 void bpf_image_ksym_del(struct bpf_ksym *ksym); 1692 void bpf_ksym_add(struct bpf_ksym *ksym); 1693 void bpf_ksym_del(struct bpf_ksym *ksym); 1694 bool bpf_has_frame_pointer(unsigned long ip); 1695 int bpf_jit_charge_modmem(u32 size); 1696 void bpf_jit_uncharge_modmem(u32 size); 1697 bool bpf_prog_has_trampoline(const struct bpf_prog *prog); 1698 bool bpf_insn_is_indirect_target(const struct bpf_verifier_env *env, const struct bpf_prog *prog, 1699 int insn_idx); 1700 u16 bpf_out_stack_arg_cnt(const struct bpf_verifier_env *env, const struct bpf_prog *prog); 1701 #else 1702 static inline int bpf_trampoline_link_prog(struct bpf_tramp_node *node, 1703 struct bpf_trampoline *tr, 1704 struct bpf_prog *tgt_prog) 1705 { 1706 return -ENOTSUPP; 1707 } 1708 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_node *node, 1709 struct bpf_trampoline *tr, 1710 struct bpf_prog *tgt_prog) 1711 { 1712 return -ENOTSUPP; 1713 } 1714 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, 1715 struct bpf_attach_target_info *tgt_info) 1716 { 1717 return NULL; 1718 } 1719 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {} 1720 #define DEFINE_BPF_DISPATCHER(name) 1721 #define DECLARE_BPF_DISPATCHER(name) 1722 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func 1723 #define BPF_DISPATCHER_PTR(name) NULL 1724 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, 1725 struct bpf_prog *from, 1726 struct bpf_prog *to) {} 1727 static inline bool is_bpf_image_address(unsigned long address) 1728 { 1729 return false; 1730 } 1731 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog) 1732 { 1733 return false; 1734 } 1735 static inline int bpf_trampoline_multi_attach(struct bpf_prog *prog, u32 *ids, 1736 struct bpf_tracing_multi_link *link) 1737 { 1738 return -ENOTSUPP; 1739 } 1740 static inline void bpf_trampoline_multi_detach(struct bpf_prog *prog, 1741 struct bpf_tracing_multi_link *link) 1742 { 1743 } 1744 static inline void bpf_trampoline_set_flags(struct bpf_trampoline *tr, u32 flags) {} 1745 #endif 1746 1747 /* The contexts that a global function is called in, and is verified for. */ 1748 enum bpf_func_context { 1749 BPF_FUNC_CTX_NON_SLEEPABLE, 1750 BPF_FUNC_CTX_SLEEPABLE, 1751 /* Sleepable code has disabled preemption or interrupts: no RCU read lock. */ 1752 BPF_FUNC_CTX_NO_RCU_LOCK, 1753 BPF_FUNC_CTX_CNT, 1754 }; 1755 1756 struct bpf_subprog_arg_info { 1757 enum bpf_arg_type arg_type; 1758 union { 1759 u32 mem_size; 1760 u32 btf_id; 1761 }; 1762 }; 1763 1764 struct bpf_func_info_aux { 1765 u16 linkage; 1766 bool unreliable; 1767 bool called[BPF_FUNC_CTX_CNT]; 1768 bool verified[BPF_FUNC_CTX_CNT]; 1769 bool called_with_pkt_ptrs; 1770 /* what the callers of a global function are checked against */ 1771 struct bpf_subprog_arg_info args[MAX_BPF_FUNC_REG_ARGS]; 1772 }; 1773 1774 enum bpf_jit_poke_reason { 1775 BPF_POKE_REASON_TAIL_CALL, 1776 }; 1777 1778 /* Descriptor of pokes pointing /into/ the JITed image. */ 1779 struct bpf_jit_poke_descriptor { 1780 void *tailcall_target; 1781 void *tailcall_bypass; 1782 void *bypass_addr; 1783 void *aux; 1784 union { 1785 struct { 1786 struct bpf_map *map; 1787 u32 key; 1788 } tail_call; 1789 }; 1790 bool tailcall_target_stable; 1791 u8 adj_off; 1792 u16 reason; 1793 u32 insn_idx; 1794 }; 1795 1796 /* reg_type info for ctx arguments */ 1797 struct bpf_ctx_arg_aux { 1798 u32 offset; 1799 enum bpf_reg_type reg_type; 1800 struct btf *btf; 1801 u32 btf_id; 1802 u32 ref_id; 1803 u32 mem_size; 1804 bool refcounted; 1805 }; 1806 1807 struct btf_mod_pair { 1808 struct btf *btf; 1809 struct module *module; 1810 }; 1811 1812 struct bpf_kfunc_desc_tab; 1813 1814 enum bpf_stream_id { 1815 BPF_STDOUT = 1, 1816 BPF_STDERR = 2, 1817 }; 1818 1819 struct bpf_stream_elem { 1820 struct llist_node node; 1821 int total_len; 1822 int consumed_len; 1823 char str[]; 1824 }; 1825 1826 enum { 1827 /* 100k bytes */ 1828 BPF_STREAM_MAX_CAPACITY = 100000ULL, 1829 }; 1830 1831 struct bpf_stream { 1832 refcount_t refcnt; 1833 atomic_t capacity; /* bytes reserved against the stream limit */ 1834 atomic_t readable; /* published bytes available to readers */ 1835 struct llist_head log; /* list of in-flight stream elements in LIFO order */ 1836 1837 struct mutex lock; /* lock protecting backlog_{head,tail} */ 1838 struct llist_node *backlog_head; /* list of in-flight stream elements in FIFO order */ 1839 struct llist_node *backlog_tail; /* tail of the list above */ 1840 wait_queue_head_t waitq; 1841 struct irq_work notify_work; 1842 bool notify_used; /* notify_work was queued at least once */ 1843 bool dead; 1844 }; 1845 1846 struct bpf_stream_stage { 1847 struct llist_head log; 1848 int len; 1849 }; 1850 1851 enum bpf_sig_verdict { 1852 BPF_SIG_UNSIGNED = 0, 1853 BPF_SIG_VERIFIED, 1854 }; 1855 1856 enum bpf_sig_keyring { 1857 BPF_SIG_KEYRING_NONE = 0, 1858 BPF_SIG_KEYRING_BUILTIN, 1859 BPF_SIG_KEYRING_SECONDARY, 1860 BPF_SIG_KEYRING_PLATFORM, 1861 BPF_SIG_KEYRING_USER, 1862 BPF_SIG_KEYRING_BPF, 1863 }; 1864 1865 struct bpf_prog_aux { 1866 atomic64_t refcnt; 1867 u32 used_map_cnt; 1868 u32 used_btf_cnt; 1869 u32 max_ctx_offset; 1870 u32 max_pkt_offset; 1871 u32 max_tp_access; 1872 u32 stack_depth; 1873 u32 id; 1874 u32 func_cnt; /* used by non-func prog as the number of func progs */ 1875 u32 real_func_cnt; /* includes hidden progs, only used for JIT and freeing progs */ 1876 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ 1877 u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 1878 u32 attach_st_ops_member_off; 1879 u32 ctx_arg_info_size; 1880 u32 max_rdonly_access; 1881 u32 max_rdwr_access; 1882 /* [min, max): the bytes of a rdwr buffer that may be written */ 1883 u32 min_rdwr_write; 1884 u32 max_rdwr_write; 1885 u32 subprog_start; 1886 struct btf *attach_btf; 1887 struct bpf_ctx_arg_aux *ctx_arg_info; 1888 void __percpu *priv_stack_ptr; 1889 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */ 1890 struct bpf_prog *dst_prog; 1891 struct bpf_trampoline *dst_trampoline; 1892 enum bpf_prog_type saved_dst_prog_type; 1893 enum bpf_attach_type saved_dst_attach_type; 1894 bool verifier_zext; /* Zero extensions has been inserted by verifier. */ 1895 bool dev_bound; /* Program is bound to the netdev. */ 1896 bool offload_requested; /* Program is bound and offloaded to the netdev. */ 1897 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */ 1898 bool attach_tracing_prog; /* true if tracing another tracing program */ 1899 bool tramp_linked; /* true if it was ever linked to a trampoline */ 1900 bool sleepable_async_cb; /* has a bpf_wq or bpf_task_work callback */ 1901 bool func_proto_unreliable; 1902 bool tail_call_reachable; 1903 bool xdp_has_frags; 1904 bool exception_cb; 1905 bool exception_boundary; 1906 bool jits_use_priv_stack; 1907 bool priv_stack_requested; 1908 bool changes_pkt_data; 1909 bool has_prologue; 1910 bool might_sleep; 1911 bool can_unwind; 1912 bool kprobe_write_ctx; 1913 /* has insns that work with the arena base, see bpf_prog_has_arena_insns() */ 1914 bool uses_arena_insns; 1915 struct { 1916 s32 keyring_serial; 1917 u8 keyring_type; 1918 u8 verdict; 1919 } sig; 1920 u64 prog_array_member_cnt; /* counts how many times as member of prog_array */ 1921 struct mutex ext_mutex; /* mutex for freplace_link_cnt and prog_array_member_cnt */ 1922 struct bpf_arena *arena; 1923 void (*recursion_detected)(struct bpf_prog *prog); /* callback if recursion is detected */ 1924 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */ 1925 const struct btf_type *attach_func_proto; 1926 /* function name for valid attach_btf_id */ 1927 const char *attach_func_name; 1928 struct bpf_prog **func; 1929 struct bpf_prog_aux *main_prog_aux; 1930 /* main prog: there when the prog has timed may_goto */ 1931 struct bpf_timed_may_goto __percpu *timed_may_goto; 1932 void *jit_data; /* JIT specific data. arch dependent */ 1933 struct bpf_jit_poke_descriptor *poke_tab; 1934 struct bpf_kfunc_desc_tab *kfunc_tab; 1935 struct bpf_kfunc_btf_tab *kfunc_btf_tab; 1936 u32 size_poke_tab; 1937 #ifdef CONFIG_FINEIBT 1938 struct bpf_ksym ksym_prefix; 1939 #endif 1940 struct bpf_ksym ksym; 1941 const struct bpf_prog_ops *ops; 1942 const struct bpf_struct_ops *st_ops; 1943 struct bpf_map **used_maps; 1944 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */ 1945 struct btf_mod_pair *used_btfs; 1946 struct bpf_prog *prog; 1947 struct user_struct *user; 1948 u64 load_time; /* ns since boottime */ 1949 u32 verified_insns; 1950 int cgroup_atype; /* enum cgroup_bpf_attach_type */ 1951 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 1952 char name[BPF_OBJ_NAME_LEN]; 1953 u64 (*bpf_exception_cb)(u64 cookie, u64 sp, u64 bp, u64, u64); 1954 u16 stack_arg_sp_adjust; 1955 u16 freplace_link_cnt; /* counts freplace links extending this prog */ 1956 #ifdef CONFIG_SECURITY 1957 void *security; 1958 #endif 1959 struct bpf_token *token; 1960 struct bpf_prog_offload *offload; 1961 struct btf *btf; 1962 struct bpf_func_info *func_info; 1963 struct bpf_func_info_aux *func_info_aux; 1964 /* bpf_line_info loaded from userspace. linfo->insn_off 1965 * has the xlated insn offset. 1966 * Both the main and sub prog share the same linfo. 1967 * The subprog can access its first linfo by 1968 * using the linfo_idx. 1969 */ 1970 struct bpf_line_info *linfo; 1971 /* jited_linfo is the jited addr of the linfo. It has a 1972 * one to one mapping to linfo: 1973 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. 1974 * Both the main and sub prog share the same jited_linfo. 1975 * The subprog can access its first jited_linfo by 1976 * using the linfo_idx. 1977 */ 1978 void **jited_linfo; 1979 u32 func_info_cnt; 1980 u32 nr_linfo; 1981 /* subprog can use linfo_idx to access its first linfo and 1982 * jited_linfo. 1983 * main prog always has linfo_idx == 0 1984 */ 1985 u32 linfo_idx; 1986 struct module *mod; 1987 u32 num_exentries; 1988 struct exception_table_entry *extable; 1989 union { 1990 struct work_struct work; 1991 struct rcu_head rcu; 1992 struct rcu_work rwork; 1993 }; 1994 struct bpf_stream *stream[2]; 1995 spinlock_t st_ops_assoc_lock; /* protects st_ops_assoc */ 1996 struct bpf_map __rcu *st_ops_assoc; 1997 }; 1998 1999 #define BPF_NR_CONTEXTS 4 /* normal, softirq, hardirq, NMI */ 2000 2001 struct bpf_prog { 2002 u16 pages; /* Number of allocated pages */ 2003 u32 jited:1, /* Is our filter JIT'ed? */ 2004 jit_requested:1,/* archs need to JIT the prog */ 2005 jit_required:1, /* program strictly requires JIT compiler */ 2006 gpl_compatible:1, /* Is filter GPL compatible? */ 2007 cb_access:1, /* Is control block accessed? */ 2008 dst_needed:1, /* Do we need dst entry? */ 2009 blinding_requested:1, /* needs constant blinding */ 2010 blinded:1, /* Was blinded */ 2011 is_func:1, /* program is a bpf function */ 2012 kprobe_override:1, /* Do we override a kprobe? */ 2013 has_callchain_buf:1, /* callchain buffer allocated? */ 2014 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */ 2015 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */ 2016 call_get_func_ip:1, /* Do we call get_func_ip() */ 2017 call_session_cookie:1, /* Do we call bpf_session_cookie() */ 2018 tstamp_type_access:1, /* Accessed __sk_buff->tstamp_type */ 2019 sleepable:1; /* BPF program is sleepable */ 2020 enum bpf_prog_type type; /* Type of BPF program */ 2021 enum bpf_attach_type expected_attach_type; /* For some prog types */ 2022 u32 len; /* Number of filter blocks */ 2023 u32 jited_len; /* Size of jited insns in bytes */ 2024 union { 2025 u8 digest[SHA256_DIGEST_SIZE]; 2026 u8 tag[BPF_TAG_SIZE]; 2027 }; 2028 struct bpf_prog_stats __percpu *stats; 2029 u8 __percpu *active; /* u8[BPF_NR_CONTEXTS] for recursion protection */ 2030 unsigned int (*bpf_func)(const void *ctx, 2031 const struct bpf_insn *insn); 2032 struct bpf_prog_aux *aux; /* Auxiliary fields */ 2033 struct sock_fprog_kern *orig_prog; /* Original BPF program */ 2034 /* Instructions for interpreter */ 2035 union { 2036 DECLARE_FLEX_ARRAY(struct sock_filter, insns); 2037 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi); 2038 }; 2039 }; 2040 2041 /* 2042 * Does the program, or a callback of it, run under RCU Tasks Trace protection 2043 * only, outside of any RCU read side critical section? 2044 */ 2045 static inline bool bpf_prog_runs_sleepable(const struct bpf_prog *prog) 2046 { 2047 return prog->sleepable || prog->aux->sleepable_async_cb; 2048 } 2049 2050 struct bpf_array_aux { 2051 /* Programs with direct jumps into programs part of this array. */ 2052 struct list_head poke_progs; 2053 struct bpf_map *map; 2054 struct mutex poke_mutex; 2055 struct work_struct work; 2056 }; 2057 2058 struct bpf_link { 2059 atomic64_t refcnt; 2060 u32 id; 2061 enum bpf_link_type type; 2062 const struct bpf_link_ops *ops; 2063 struct bpf_prog *prog; 2064 2065 u32 flags; 2066 enum bpf_attach_type attach_type; 2067 2068 /* rcu is used before freeing, work can be used to schedule that 2069 * RCU-based freeing before that, so they never overlap 2070 */ 2071 union { 2072 struct rcu_head rcu; 2073 struct work_struct work; 2074 }; 2075 /* whether BPF link itself has "sleepable" semantics, which can differ 2076 * from underlying BPF program having a "sleepable" semantics, as BPF 2077 * link's semantics is determined by target attach hook 2078 */ 2079 bool sleepable; 2080 }; 2081 2082 struct bpf_link_ops { 2083 void (*release)(struct bpf_link *link); 2084 /* deallocate link resources callback, called without RCU grace period 2085 * waiting 2086 */ 2087 void (*dealloc)(struct bpf_link *link); 2088 /* deallocate link resources callback, called after RCU grace period; 2089 * if either the underlying BPF program is sleepable or BPF link's 2090 * target hook is sleepable, we'll go through tasks trace RCU GP and 2091 * then "classic" RCU GP; this need for chaining tasks trace and 2092 * classic RCU GPs is designated by setting bpf_link->sleepable flag 2093 * 2094 * For non-sleepable tracepoint links we go through SRCU gp instead, 2095 * since RCU is not used in that case. Sleepable tracepoints still 2096 * follow the scheme above. 2097 */ 2098 void (*dealloc_deferred)(struct bpf_link *link); 2099 int (*detach)(struct bpf_link *link); 2100 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, 2101 struct bpf_prog *old_prog); 2102 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq); 2103 int (*fill_link_info)(const struct bpf_link *link, 2104 struct bpf_link_info *info); 2105 int (*update_map)(struct bpf_link *link, struct bpf_map *new_map, 2106 struct bpf_map *old_map); 2107 __poll_t (*poll)(struct file *file, struct poll_table_struct *pts); 2108 }; 2109 2110 struct bpf_tramp_node { 2111 struct bpf_link *link; 2112 struct hlist_node tramp_hlist; 2113 u64 cookie; 2114 }; 2115 2116 struct bpf_tramp_link { 2117 struct bpf_link link; 2118 struct bpf_tramp_node node; 2119 }; 2120 2121 struct bpf_shim_tramp_link { 2122 struct bpf_tramp_link link; 2123 struct bpf_trampoline *trampoline; 2124 }; 2125 2126 struct bpf_tracing_link { 2127 struct bpf_tramp_link link; 2128 struct bpf_tramp_node fexit; 2129 struct bpf_trampoline *trampoline; 2130 struct bpf_prog *tgt_prog; 2131 }; 2132 2133 struct bpf_tracing_multi_node { 2134 struct bpf_tramp_node node; 2135 struct bpf_trampoline *trampoline; 2136 struct ftrace_func_entry entry; 2137 }; 2138 2139 struct bpf_tracing_multi_data { 2140 struct ftrace_hash *unreg; 2141 struct ftrace_hash *modify; 2142 struct ftrace_hash *reg; 2143 struct ftrace_func_entry *entry; 2144 }; 2145 2146 struct bpf_tracing_multi_link { 2147 struct bpf_link link; 2148 struct bpf_tracing_multi_data data; 2149 u64 *cookies; 2150 struct bpf_tramp_node *fexits; 2151 int nodes_cnt; 2152 struct bpf_tracing_multi_node nodes[] __counted_by(nodes_cnt); 2153 }; 2154 2155 struct bpf_raw_tp_link { 2156 struct bpf_link link; 2157 struct bpf_raw_event_map *btp; 2158 u64 cookie; 2159 }; 2160 2161 struct bpf_link_primer { 2162 struct bpf_link *link; 2163 struct file *file; 2164 int fd; 2165 u32 id; 2166 }; 2167 2168 struct bpf_mount_opts { 2169 kuid_t uid; 2170 kgid_t gid; 2171 umode_t mode; 2172 2173 /* BPF token-related delegation options */ 2174 u64 delegate_cmds; 2175 u64 delegate_maps; 2176 u64 delegate_progs; 2177 u64 delegate_attachs; 2178 2179 struct simple_xattr_cache xa_cache; 2180 }; 2181 2182 struct bpf_token { 2183 struct work_struct work; 2184 atomic64_t refcnt; 2185 struct user_namespace *userns; 2186 u64 allowed_cmds; 2187 u64 allowed_maps; 2188 u64 allowed_progs; 2189 u64 allowed_attachs; 2190 #ifdef CONFIG_SECURITY 2191 void *security; 2192 #endif 2193 }; 2194 2195 struct bpf_struct_ops_value; 2196 struct btf_member; 2197 2198 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64 2199 /** 2200 * struct bpf_struct_ops - A structure of callbacks allowing a subsystem to 2201 * define a BPF_MAP_TYPE_STRUCT_OPS map type composed 2202 * of BPF_PROG_TYPE_STRUCT_OPS progs. 2203 * @verifier_ops: A structure of callbacks that are invoked by the verifier 2204 * when determining whether the struct_ops progs in the 2205 * struct_ops map are valid. 2206 * @init: A callback that is invoked a single time, and before any other 2207 * callback, to initialize the structure. A nonzero return value means 2208 * the subsystem could not be initialized. 2209 * @check_member: When defined, a callback invoked by the verifier to allow 2210 * the subsystem to determine if an entry in the struct_ops map 2211 * is valid. A nonzero return value means that the map is 2212 * invalid and should be rejected by the verifier. 2213 * @init_member: A callback that is invoked for each member of the struct_ops 2214 * map to allow the subsystem to initialize the member. A nonzero 2215 * value means the member could not be initialized. This callback 2216 * is exclusive with the @type, @type_id, @value_type, and 2217 * @value_id fields. 2218 * @reg: A callback that is invoked when the struct_ops map has been 2219 * initialized and is being attached to. Zero means the struct_ops map 2220 * has been successfully registered and is live. A nonzero return value 2221 * means the struct_ops map could not be registered. 2222 * @unreg: A callback that is invoked when the struct_ops map should be 2223 * unregistered. 2224 * @update: A callback that is invoked when the live struct_ops map is being 2225 * updated to contain new values. This callback is only invoked when 2226 * the struct_ops map is loaded with BPF_F_LINK. If not defined, the 2227 * it is assumed that the struct_ops map cannot be updated. 2228 * @validate: A callback that is invoked after all of the members have been 2229 * initialized. This callback should perform static checks on the 2230 * map, meaning that it should either fail or succeed 2231 * deterministically. A struct_ops map that has been validated may 2232 * not necessarily succeed in being registered if the call to @reg 2233 * fails. For example, a valid struct_ops map may be loaded, but 2234 * then fail to be registered due to there being another active 2235 * struct_ops map on the system in the subsystem already. For this 2236 * reason, if this callback is not defined, the check is skipped as 2237 * the struct_ops map will have final verification performed in 2238 * @reg. 2239 * @cfi_stubs: Pointer to a structure of stub functions for CFI. These stubs 2240 * provide the correct Control Flow Integrity hashes for the 2241 * trampolines generated by BPF struct_ops. 2242 * @owner: The module that owns this struct_ops. Used for module reference 2243 * counting to ensure the module providing the struct_ops cannot be 2244 * unloaded while in use. 2245 * @name: The name of the struct bpf_struct_ops object. 2246 * @func_models: Func models 2247 * @cgroup_atype: A value in enum cgroup_bpf_attach_type for cgroup attachment. 2248 * 0 means the struct_ops type does not support cgroup attachment. 2249 * If cgroup_atype is non-zero, the @reg and @unreg must be NULL 2250 * because the attachment/detachment will be handled by the bpf core. 2251 * @free_after_tasks_rcu_gp: Set to true if it needs the bpf core to wait for 2252 * a tasks_rcu gp before freeing the struct_ops map 2253 * and its progs. It is unnecessary if the @unreg 2254 * has waited for the correct rcu gp or the @unreg 2255 * has ensured all struct_ops prog has finished running. 2256 * @free_after_mult_rcu_gp: Set to true if the cgroup prog array of this 2257 * struct_ops needs a tasks_trace_rcu gp, besides a 2258 * regular rcu gp, before it is freed. It is usually 2259 * needed if the struct_ops has sleepable prog. A 2260 * struct_ops map always waits for both. 2261 */ 2262 struct bpf_struct_ops { 2263 const struct bpf_verifier_ops *verifier_ops; 2264 int (*init)(struct btf *btf); 2265 int (*check_member)(const struct btf_type *t, 2266 const struct btf_member *member, 2267 const struct bpf_prog *prog); 2268 int (*init_member)(const struct btf_type *t, 2269 const struct btf_member *member, 2270 void *kdata, const void *udata); 2271 int (*reg)(void *kdata, struct bpf_link *link); 2272 void (*unreg)(void *kdata, struct bpf_link *link); 2273 int (*update)(void *kdata, void *old_kdata, struct bpf_link *link); 2274 int (*validate)(void *kdata); 2275 void *cfi_stubs; 2276 struct module *owner; 2277 const char *name; 2278 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS]; 2279 int cgroup_atype; 2280 bool free_after_tasks_rcu_gp; 2281 bool free_after_mult_rcu_gp; 2282 }; 2283 2284 /* Every member of a struct_ops type has an instance even a member is not 2285 * an operator (function pointer). The "info" field will be assigned to 2286 * prog->aux->ctx_arg_info of BPF struct_ops programs to provide the 2287 * argument information required by the verifier to verify the program. 2288 * 2289 * btf_ctx_access() will lookup prog->aux->ctx_arg_info to find the 2290 * corresponding entry for an given argument. 2291 */ 2292 struct bpf_struct_ops_arg_info { 2293 struct bpf_ctx_arg_aux *info; 2294 u32 cnt; 2295 }; 2296 2297 struct bpf_struct_ops_desc { 2298 struct bpf_struct_ops *st_ops; 2299 2300 const struct btf_type *type; 2301 const struct btf_type *value_type; 2302 u32 type_id; 2303 u32 value_id; 2304 2305 /* Collection of argument information for each member */ 2306 struct bpf_struct_ops_arg_info *arg_info; 2307 }; 2308 2309 enum bpf_struct_ops_state { 2310 BPF_STRUCT_OPS_STATE_INIT, 2311 BPF_STRUCT_OPS_STATE_INUSE, 2312 BPF_STRUCT_OPS_STATE_TOBEFREE, 2313 BPF_STRUCT_OPS_STATE_READY, 2314 }; 2315 2316 struct bpf_struct_ops_common_value { 2317 refcount_t refcnt; 2318 enum bpf_struct_ops_state state; 2319 }; 2320 2321 static inline bool bpf_prog_get_recursion_context(struct bpf_prog *prog) 2322 { 2323 #ifdef CONFIG_ARM64 2324 u8 rctx = interrupt_context_level(); 2325 u8 *active = this_cpu_ptr(prog->active); 2326 u32 val; 2327 2328 preempt_disable(); 2329 active[rctx]++; 2330 val = le32_to_cpu(*(__le32 *)active); 2331 preempt_enable(); 2332 if (val != BIT(rctx * 8)) 2333 return false; 2334 2335 return true; 2336 #else 2337 return this_cpu_inc_return(*(int __percpu *)(prog->active)) == 1; 2338 #endif 2339 } 2340 2341 static inline void bpf_prog_put_recursion_context(struct bpf_prog *prog) 2342 { 2343 #ifdef CONFIG_ARM64 2344 u8 rctx = interrupt_context_level(); 2345 u8 *active = this_cpu_ptr(prog->active); 2346 2347 preempt_disable(); 2348 active[rctx]--; 2349 preempt_enable(); 2350 #else 2351 this_cpu_dec(*(int __percpu *)(prog->active)); 2352 #endif 2353 } 2354 2355 static inline bool is_tracing_multi(enum bpf_attach_type type) 2356 { 2357 return type == BPF_TRACE_FENTRY_MULTI || type == BPF_TRACE_FEXIT_MULTI || 2358 type == BPF_TRACE_FSESSION_MULTI; 2359 } 2360 2361 static inline bool is_struct_ops_tramp(const struct bpf_tramp_nodes *fentry_nodes) 2362 { 2363 return fentry_nodes->nr_nodes == 1 && 2364 fentry_nodes->nodes[0]->link->type == BPF_LINK_TYPE_STRUCT_OPS; 2365 } 2366 2367 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL) 2368 /* This macro helps developer to register a struct_ops type and generate 2369 * type information correctly. Developers should use this macro to register 2370 * a struct_ops type instead of calling __register_bpf_struct_ops() directly. 2371 */ 2372 #define register_bpf_struct_ops(st_ops, type) \ 2373 ({ \ 2374 struct bpf_struct_ops_##type { \ 2375 struct bpf_struct_ops_common_value common; \ 2376 struct type data ____cacheline_aligned_in_smp; \ 2377 }; \ 2378 BTF_TYPE_EMIT(struct bpf_struct_ops_##type); \ 2379 __register_bpf_struct_ops(st_ops); \ 2380 }) 2381 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA)) 2382 bool bpf_struct_ops_get(const void *kdata); 2383 void bpf_struct_ops_put(const void *kdata); 2384 int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff); 2385 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, 2386 void *value); 2387 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_nodes *tnodes, 2388 struct bpf_tramp_node *node, 2389 const struct btf_func_model *model, 2390 void *stub_func, 2391 void **image, u32 *image_off, 2392 bool allow_alloc); 2393 void bpf_struct_ops_image_free(void *image); 2394 static inline bool bpf_try_module_get(const void *data, struct module *owner) 2395 { 2396 if (owner == BPF_MODULE_OWNER) 2397 return bpf_struct_ops_get(data); 2398 else 2399 return try_module_get(owner); 2400 } 2401 static inline void bpf_module_put(const void *data, struct module *owner) 2402 { 2403 if (owner == BPF_MODULE_OWNER) 2404 bpf_struct_ops_put(data); 2405 else 2406 module_put(owner); 2407 } 2408 int bpf_struct_ops_link_create(union bpf_attr *attr); 2409 int bpf_prog_assoc_struct_ops(struct bpf_prog *prog, struct bpf_map *map); 2410 void bpf_prog_disassoc_struct_ops(struct bpf_prog *prog); 2411 void *bpf_prog_get_assoc_struct_ops(const struct bpf_prog_aux *aux); 2412 u32 bpf_struct_ops_id(const void *kdata); 2413 int bpf_struct_ops_for_each_prog(const void *kdata, 2414 int (*cb)(struct bpf_prog *prog, void *data), 2415 void *data); 2416 void *bpf_struct_ops_map_kdata(struct bpf_map *map); 2417 void *bpf_struct_ops_map_cfi_stubs(struct bpf_map *map); 2418 bool bpf_struct_ops_valid_to_reg(struct bpf_map *map); 2419 int bpf_struct_ops_link_update_check(struct bpf_map *new_map, struct bpf_map *old_map, 2420 struct bpf_map *expected_old_map); 2421 int bpf_struct_ops_map_cgroup_atype(struct bpf_map *map); 2422 2423 #ifdef CONFIG_NET 2424 /* Define it here to avoid the use of forward declaration */ 2425 struct bpf_dummy_ops_state { 2426 int val; 2427 }; 2428 2429 struct bpf_dummy_ops { 2430 int (*test_1)(struct bpf_dummy_ops_state *cb); 2431 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2, 2432 char a3, unsigned long a4); 2433 int (*test_sleepable)(struct bpf_dummy_ops_state *cb); 2434 }; 2435 2436 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr, 2437 union bpf_attr __user *uattr); 2438 #endif 2439 int bpf_struct_ops_desc_init(struct bpf_struct_ops_desc *st_ops_desc, 2440 struct btf *btf, 2441 struct bpf_verifier_log *log); 2442 void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map); 2443 void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc); 2444 #else 2445 #define register_bpf_struct_ops(st_ops, type) ({ (void *)(st_ops); 0; }) 2446 static inline bool bpf_try_module_get(const void *data, struct module *owner) 2447 { 2448 return try_module_get(owner); 2449 } 2450 static inline void bpf_module_put(const void *data, struct module *owner) 2451 { 2452 module_put(owner); 2453 } 2454 static inline int bpf_struct_ops_supported(const struct bpf_struct_ops *st_ops, u32 moff) 2455 { 2456 return -ENOTSUPP; 2457 } 2458 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, 2459 void *key, 2460 void *value) 2461 { 2462 return -EINVAL; 2463 } 2464 static inline int bpf_struct_ops_link_create(union bpf_attr *attr) 2465 { 2466 return -EOPNOTSUPP; 2467 } 2468 static inline int bpf_prog_assoc_struct_ops(struct bpf_prog *prog, struct bpf_map *map) 2469 { 2470 return -EOPNOTSUPP; 2471 } 2472 static inline void bpf_prog_disassoc_struct_ops(struct bpf_prog *prog) 2473 { 2474 } 2475 static inline void *bpf_prog_get_assoc_struct_ops(const struct bpf_prog_aux *aux) 2476 { 2477 return NULL; 2478 } 2479 static inline void bpf_map_struct_ops_info_fill(struct bpf_map_info *info, struct bpf_map *map) 2480 { 2481 } 2482 2483 static inline void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc) 2484 { 2485 } 2486 static inline u32 bpf_struct_ops_id(void *kdata) 2487 { 2488 return 0; 2489 } 2490 static inline void *bpf_struct_ops_map_kdata(struct bpf_map *map) 2491 { 2492 return NULL; 2493 } 2494 static inline int bpf_struct_ops_map_cgroup_atype(struct bpf_map *map) 2495 { 2496 return 0; 2497 } 2498 static inline void *bpf_struct_ops_map_cfi_stubs(struct bpf_map *map) 2499 { 2500 return NULL; 2501 } 2502 static inline bool bpf_struct_ops_valid_to_reg(struct bpf_map *map) 2503 { 2504 return false; 2505 } 2506 static inline int bpf_struct_ops_link_update_check(struct bpf_map *new_map, 2507 struct bpf_map *old_map, 2508 struct bpf_map *expected_old_map) 2509 { 2510 return -EOPNOTSUPP; 2511 } 2512 2513 #endif 2514 2515 static inline int bpf_fsession_cnt(struct bpf_tramp_nodes *nodes) 2516 { 2517 struct bpf_tramp_nodes fentries = nodes[BPF_TRAMP_FENTRY]; 2518 int cnt = 0; 2519 2520 for (int i = 0; i < nodes[BPF_TRAMP_FENTRY].nr_nodes; i++) { 2521 if (fentries.nodes[i]->link->prog->expected_attach_type == BPF_TRACE_FSESSION) 2522 cnt++; 2523 if (fentries.nodes[i]->link->prog->expected_attach_type == BPF_TRACE_FSESSION_MULTI) 2524 cnt++; 2525 } 2526 2527 return cnt; 2528 } 2529 2530 static inline bool bpf_prog_calls_session_cookie(struct bpf_tramp_node *node) 2531 { 2532 return node->link->prog->call_session_cookie; 2533 } 2534 2535 static inline int bpf_fsession_cookie_cnt(struct bpf_tramp_nodes *nodes) 2536 { 2537 struct bpf_tramp_nodes fentries = nodes[BPF_TRAMP_FENTRY]; 2538 int cnt = 0; 2539 2540 for (int i = 0; i < nodes[BPF_TRAMP_FENTRY].nr_nodes; i++) { 2541 if (bpf_prog_calls_session_cookie(fentries.nodes[i])) 2542 cnt++; 2543 } 2544 2545 return cnt; 2546 } 2547 2548 int bpf_prog_ctx_arg_info_init(struct bpf_prog *prog, 2549 const struct bpf_ctx_arg_aux *info, u32 cnt); 2550 2551 #if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM) 2552 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 2553 int cgroup_atype, 2554 enum bpf_attach_type attach_type); 2555 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog); 2556 #else 2557 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog, 2558 int cgroup_atype, 2559 enum bpf_attach_type attach_type) 2560 { 2561 return -EOPNOTSUPP; 2562 } 2563 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog) 2564 { 2565 } 2566 #endif 2567 2568 struct bpf_array { 2569 struct bpf_map map; 2570 u32 elem_size; 2571 u32 index_mask; 2572 struct bpf_array_aux *aux; 2573 union { 2574 DECLARE_FLEX_ARRAY(char, value) __aligned(8); 2575 DECLARE_FLEX_ARRAY(void *, ptrs) __aligned(8); 2576 DECLARE_FLEX_ARRAY(void __percpu *, pptrs) __aligned(8); 2577 }; 2578 }; 2579 2580 /* 2581 * The bpf_array_get_next_key() function may be used for all array-like 2582 * maps, i.e., maps with u32 keys with range [0 ,..., max_entries) 2583 */ 2584 int bpf_array_get_next_key(struct bpf_map *map, void *key, void *next_key); 2585 2586 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */ 2587 #define MAX_TAIL_CALL_CNT 33 2588 2589 /* Maximum number of loops for bpf_loop and bpf_iter_num. 2590 * It's enum to expose it (and thus make it discoverable) through BTF. 2591 */ 2592 enum { 2593 BPF_MAX_LOOPS = 8 * 1024 * 1024, 2594 BPF_MAX_TIMED_LOOPS = 0xffff, 2595 }; 2596 2597 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \ 2598 BPF_F_RDONLY_PROG | \ 2599 BPF_F_WRONLY | \ 2600 BPF_F_WRONLY_PROG) 2601 2602 #define BPF_MAP_CAN_READ BIT(0) 2603 #define BPF_MAP_CAN_WRITE BIT(1) 2604 2605 /* Maximum number of user-producer ring buffer samples that can be drained in 2606 * a call to bpf_user_ringbuf_drain(). 2607 */ 2608 #define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024) 2609 2610 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map) 2611 { 2612 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 2613 2614 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is 2615 * not possible. 2616 */ 2617 if (access_flags & BPF_F_RDONLY_PROG) 2618 return BPF_MAP_CAN_READ; 2619 else if (access_flags & BPF_F_WRONLY_PROG) 2620 return BPF_MAP_CAN_WRITE; 2621 else 2622 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE; 2623 } 2624 2625 static inline bool bpf_map_flags_access_ok(u32 access_flags) 2626 { 2627 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != 2628 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG); 2629 } 2630 2631 static inline struct bpf_map_owner *bpf_map_owner_alloc(struct bpf_map *map) 2632 { 2633 return kzalloc_obj(*map->owner, GFP_ATOMIC); 2634 } 2635 2636 static inline void bpf_map_owner_free(struct bpf_map *map) 2637 { 2638 kfree(map->owner); 2639 } 2640 2641 struct bpf_event_entry { 2642 struct perf_event *event; 2643 struct file *perf_file; 2644 struct file *map_file; 2645 struct rcu_head rcu; 2646 }; 2647 2648 static inline bool map_type_contains_progs(struct bpf_map *map) 2649 { 2650 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY || 2651 map->map_type == BPF_MAP_TYPE_DEVMAP || 2652 map->map_type == BPF_MAP_TYPE_CPUMAP; 2653 } 2654 2655 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp); 2656 int bpf_prog_calc_tag(struct bpf_prog *fp); 2657 2658 const struct bpf_func_proto *bpf_get_trace_printk_proto(void); 2659 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void); 2660 2661 const struct bpf_func_proto *bpf_get_perf_event_read_value_proto(void); 2662 2663 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, 2664 unsigned long off, unsigned long len); 2665 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, 2666 const struct bpf_insn *src, 2667 struct bpf_insn *dst, 2668 struct bpf_prog *prog, 2669 u32 *target_size); 2670 2671 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, 2672 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); 2673 2674 /* an array of programs to be executed under rcu_lock. 2675 * 2676 * Typical usage: 2677 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run); 2678 * 2679 * the structure returned by bpf_prog_array_alloc() should be populated 2680 * with program pointers and the last pointer must be NULL. 2681 * The user has to keep refcnt on the program and make sure the program 2682 * is removed from the array before bpf_prog_put(). 2683 * The 'struct bpf_prog_array *' should only be replaced with xchg() 2684 * since other cpus are walking the array of pointers in parallel. 2685 */ 2686 struct bpf_prog_array_item { 2687 union { 2688 struct bpf_prog *prog; 2689 void *kdata; 2690 }; 2691 union { 2692 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; 2693 u64 bpf_cookie; 2694 }; 2695 }; 2696 2697 struct bpf_prog_array { 2698 struct rcu_head rcu; 2699 struct bpf_prog_array_item items[]; 2700 }; 2701 2702 /* to avoid allocating empty bpf_prog_array for cgroups that 2703 * don't have bpf program attached use one global 'bpf_empty_prog_array' 2704 * It will not be modified the caller of bpf_prog_array_alloc() 2705 * (since caller requested prog_cnt == 0) 2706 * that pointer should be 'freed' by bpf_prog_array_free() 2707 */ 2708 extern struct bpf_prog_array bpf_empty_prog_array; 2709 2710 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); 2711 void bpf_prog_array_free(struct bpf_prog_array *progs); 2712 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */ 2713 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs); 2714 int bpf_prog_array_length(struct bpf_prog_array *progs); 2715 bool bpf_prog_array_is_empty(struct bpf_prog_array *array); 2716 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, 2717 __u32 __user *prog_ids, u32 cnt); 2718 2719 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, 2720 struct bpf_prog *old_prog); 2721 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index); 2722 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, 2723 struct bpf_prog *prog); 2724 int bpf_prog_array_copy_info(struct bpf_prog_array *array, 2725 u32 *prog_ids, u32 request_cnt, 2726 u32 *prog_cnt); 2727 int bpf_prog_array_copy(struct bpf_prog_array *old_array, 2728 struct bpf_prog *exclude_prog, 2729 struct bpf_prog *include_prog, 2730 u64 bpf_cookie, 2731 struct bpf_prog_array **new_array); 2732 struct bpf_prog *bpf_prog_dummy(void); 2733 2734 struct bpf_run_ctx {}; 2735 2736 struct bpf_cg_run_ctx { 2737 struct bpf_run_ctx run_ctx; 2738 const struct bpf_prog_array_item *prog_item; 2739 int retval; 2740 }; 2741 2742 struct bpf_trace_run_ctx { 2743 struct bpf_run_ctx run_ctx; 2744 u64 bpf_cookie; 2745 bool is_uprobe; 2746 }; 2747 2748 struct bpf_tramp_run_ctx { 2749 struct bpf_run_ctx run_ctx; 2750 u64 bpf_cookie; 2751 int retval; 2752 struct bpf_run_ctx *saved_run_ctx; 2753 }; 2754 2755 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx) 2756 { 2757 struct bpf_run_ctx *old_ctx = NULL; 2758 2759 #ifdef CONFIG_BPF_SYSCALL 2760 old_ctx = current->bpf_ctx; 2761 current->bpf_ctx = new_ctx; 2762 #endif 2763 return old_ctx; 2764 } 2765 2766 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx) 2767 { 2768 #ifdef CONFIG_BPF_SYSCALL 2769 current->bpf_ctx = old_ctx; 2770 #endif 2771 } 2772 2773 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */ 2774 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0) 2775 /* BPF program asks to set CN on the packet. */ 2776 #define BPF_RET_SET_CN (1 << 0) 2777 2778 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx); 2779 2780 static __always_inline u32 2781 bpf_prog_run_array(const struct bpf_prog_array *array, 2782 const void *ctx, bpf_prog_run_fn run_prog) 2783 { 2784 const struct bpf_prog_array_item *item; 2785 const struct bpf_prog *prog; 2786 struct bpf_run_ctx *old_run_ctx; 2787 struct bpf_trace_run_ctx run_ctx; 2788 u32 ret = 1; 2789 2790 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held"); 2791 2792 if (unlikely(!array)) 2793 return ret; 2794 2795 run_ctx.is_uprobe = false; 2796 2797 migrate_disable(); 2798 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 2799 item = &array->items[0]; 2800 while ((prog = READ_ONCE(item->prog))) { 2801 run_ctx.bpf_cookie = item->bpf_cookie; 2802 ret &= run_prog(prog, ctx); 2803 item++; 2804 } 2805 bpf_reset_run_ctx(old_run_ctx); 2806 migrate_enable(); 2807 return ret; 2808 } 2809 2810 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs: 2811 * 2812 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array 2813 * overall. As a result, we must use the bpf_prog_array_free_sleepable 2814 * in order to use the tasks_trace rcu grace period. 2815 * 2816 * When a non-sleepable program is inside the array, we take the rcu read 2817 * section and disable preemption for that program alone, so it can access 2818 * rcu-protected dynamically sized maps. 2819 */ 2820 static __always_inline u32 2821 bpf_prog_run_array_uprobe(const struct bpf_prog_array *array, 2822 const void *ctx, bpf_prog_run_fn run_prog) 2823 { 2824 const struct bpf_prog_array_item *item; 2825 const struct bpf_prog *prog; 2826 struct bpf_run_ctx *old_run_ctx; 2827 struct bpf_trace_run_ctx run_ctx; 2828 u32 ret = 1; 2829 2830 might_fault(); 2831 RCU_LOCKDEP_WARN(!rcu_read_lock_trace_held(), "no rcu lock held"); 2832 2833 if (unlikely(!array)) 2834 return ret; 2835 2836 migrate_disable(); 2837 2838 run_ctx.is_uprobe = true; 2839 2840 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 2841 item = &array->items[0]; 2842 while ((prog = READ_ONCE(item->prog))) { 2843 if (!prog->sleepable) 2844 rcu_read_lock(); 2845 2846 run_ctx.bpf_cookie = item->bpf_cookie; 2847 ret &= run_prog(prog, ctx); 2848 item++; 2849 2850 if (!prog->sleepable) 2851 rcu_read_unlock(); 2852 } 2853 bpf_reset_run_ctx(old_run_ctx); 2854 migrate_enable(); 2855 return ret; 2856 } 2857 2858 bool bpf_jit_bypass_spec_v1(void); 2859 bool bpf_jit_bypass_spec_v4(void); 2860 2861 #define bpf_rcu_lock_held() \ 2862 (rcu_read_lock_held() || rcu_read_lock_trace_held() || rcu_read_lock_bh_held()) 2863 2864 #ifdef CONFIG_BPF_SYSCALL 2865 DECLARE_PER_CPU(int, bpf_prog_active); 2866 extern struct mutex bpf_stats_enabled_mutex; 2867 2868 /* 2869 * Block execution of BPF programs attached to instrumentation (perf, 2870 * kprobes, tracepoints) to prevent deadlocks on map operations as any of 2871 * these events can happen inside a region which holds a map bucket lock 2872 * and can deadlock on it. 2873 */ 2874 static inline void bpf_disable_instrumentation(void) 2875 { 2876 migrate_disable(); 2877 this_cpu_inc(bpf_prog_active); 2878 } 2879 2880 static inline void bpf_enable_instrumentation(void) 2881 { 2882 this_cpu_dec(bpf_prog_active); 2883 migrate_enable(); 2884 } 2885 2886 extern const struct super_operations bpf_super_ops; 2887 extern const struct file_operations bpf_map_fops; 2888 extern const struct file_operations bpf_prog_fops; 2889 extern const struct file_operations bpf_iter_fops; 2890 extern const struct file_operations bpf_token_fops; 2891 2892 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \ 2893 extern const struct bpf_prog_ops _name ## _prog_ops; \ 2894 extern const struct bpf_verifier_ops _name ## _verifier_ops; 2895 #define BPF_MAP_TYPE(_id, _ops) \ 2896 extern const struct bpf_map_ops _ops; 2897 #define BPF_LINK_TYPE(_id, _name) 2898 #include <linux/bpf_types.h> 2899 #undef BPF_PROG_TYPE 2900 #undef BPF_MAP_TYPE 2901 #undef BPF_LINK_TYPE 2902 2903 extern const struct bpf_prog_ops bpf_offload_prog_ops; 2904 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; 2905 extern const struct bpf_verifier_ops xdp_analyzer_ops; 2906 2907 struct bpf_prog *bpf_prog_get(u32 ufd); 2908 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, 2909 bool attach_drv); 2910 void bpf_prog_add(struct bpf_prog *prog, int i); 2911 void bpf_prog_sub(struct bpf_prog *prog, int i); 2912 void bpf_prog_inc(struct bpf_prog *prog); 2913 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); 2914 void bpf_prog_put(struct bpf_prog *prog); 2915 2916 void bpf_prog_free_id(struct bpf_prog *prog); 2917 void bpf_map_free_id(struct bpf_map *map); 2918 2919 struct btf_field *btf_record_find(const struct btf_record *rec, 2920 u32 offset, u32 field_mask); 2921 void btf_record_free(struct btf_record *rec); 2922 void bpf_map_free_record(struct bpf_map *map); 2923 struct btf_record *btf_record_dup(const struct btf_record *rec); 2924 bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b); 2925 void bpf_obj_free_timer(const struct btf_record *rec, void *obj); 2926 void bpf_obj_free_workqueue(const struct btf_record *rec, void *obj); 2927 void bpf_obj_free_task_work(const struct btf_record *rec, void *obj); 2928 void bpf_obj_cancel_fields(struct bpf_map *map, void *obj); 2929 void bpf_obj_free_fields(const struct btf_record *rec, void *obj); 2930 void __bpf_obj_drop_impl(void *p, const struct btf_record *rec, bool percpu); 2931 2932 struct bpf_map *bpf_map_get(u32 ufd); 2933 struct bpf_map *bpf_map_get_with_uref(u32 ufd); 2934 2935 /* 2936 * The __bpf_map_get() and __btf_get_by_fd() functions parse a file 2937 * descriptor and return a corresponding map or btf object. 2938 * Their names are double underscored to emphasize the fact that they 2939 * do not increase refcnt. To also increase refcnt use corresponding 2940 * bpf_map_get() and btf_get_by_fd() functions. 2941 */ 2942 2943 static inline struct bpf_map *__bpf_map_get(struct fd f) 2944 { 2945 if (fd_empty(f)) 2946 return ERR_PTR(-EBADF); 2947 if (unlikely(fd_file(f)->f_op != &bpf_map_fops)) 2948 return ERR_PTR(-EINVAL); 2949 return fd_file(f)->private_data; 2950 } 2951 2952 static inline struct btf *__btf_get_by_fd(struct fd f) 2953 { 2954 if (fd_empty(f)) 2955 return ERR_PTR(-EBADF); 2956 if (unlikely(fd_file(f)->f_op != &btf_fops)) 2957 return ERR_PTR(-EINVAL); 2958 return fd_file(f)->private_data; 2959 } 2960 2961 void bpf_map_inc(struct bpf_map *map); 2962 void bpf_map_inc_with_uref(struct bpf_map *map); 2963 struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref); 2964 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map); 2965 void bpf_map_put_with_uref(struct bpf_map *map); 2966 void bpf_map_put(struct bpf_map *map); 2967 void *bpf_map_area_alloc(u64 size, int numa_node); 2968 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node); 2969 void bpf_map_area_free(void *base); 2970 bool bpf_map_write_active(const struct bpf_map *map); 2971 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); 2972 int generic_map_lookup_batch(struct bpf_map *map, 2973 const union bpf_attr *attr, 2974 union bpf_attr __user *uattr); 2975 int generic_map_update_batch(struct bpf_map *map, struct file *map_file, 2976 const union bpf_attr *attr, 2977 union bpf_attr __user *uattr); 2978 int generic_map_delete_batch(struct bpf_map *map, 2979 const union bpf_attr *attr, 2980 union bpf_attr __user *uattr); 2981 struct bpf_map *bpf_map_get_curr_or_next(u32 *id); 2982 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id); 2983 2984 2985 struct page *bpf_alloc_page(int nid, bool sleepable); 2986 int bpf_alloc_pages(int nid, unsigned long nr_pages, 2987 struct llist_head *pages, bool sleepable); 2988 void bpf_free_pages(struct llist_head *pages); 2989 #ifdef CONFIG_MEMCG 2990 void bpf_map_memcg_enter(const struct bpf_map *map, struct mem_cgroup **old_memcg, 2991 struct mem_cgroup **new_memcg); 2992 void bpf_map_memcg_exit(struct mem_cgroup *old_memcg, 2993 struct mem_cgroup *memcg); 2994 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags, 2995 int node); 2996 void *bpf_map_kmalloc_nolock(const struct bpf_map *map, size_t size, gfp_t flags, 2997 int node); 2998 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags); 2999 void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size, 3000 gfp_t flags); 3001 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, 3002 size_t align, gfp_t flags); 3003 #else 3004 /* 3005 * These specialized allocators have to be macros for their allocations to be 3006 * accounted separately (to have separate alloc_tag). 3007 */ 3008 #define bpf_map_kmalloc_node(_map, _size, _flags, _node) \ 3009 kmalloc_node(_size, _flags, _node) 3010 #define bpf_map_kmalloc_nolock(_map, _size, _flags, _node) \ 3011 kmalloc_nolock(_size, _flags, _node) 3012 #define bpf_map_kzalloc(_map, _size, _flags) \ 3013 kzalloc(_size, _flags) 3014 #define bpf_map_kvcalloc(_map, _n, _size, _flags) \ 3015 kvcalloc(_n, _size, _flags) 3016 #define bpf_map_alloc_percpu(_map, _size, _align, _flags) \ 3017 __alloc_percpu_gfp(_size, _align, _flags) 3018 static inline void bpf_map_memcg_enter(const struct bpf_map *map, struct mem_cgroup **old_memcg, 3019 struct mem_cgroup **new_memcg) 3020 { 3021 *new_memcg = NULL; 3022 *old_memcg = NULL; 3023 } 3024 3025 static inline void bpf_map_memcg_exit(struct mem_cgroup *old_memcg, 3026 struct mem_cgroup *memcg) 3027 { 3028 } 3029 #endif 3030 3031 static inline int 3032 bpf_map_init_elem_count(struct bpf_map *map) 3033 { 3034 size_t size = sizeof(*map->elem_count), align = size; 3035 gfp_t flags = GFP_USER | __GFP_NOWARN; 3036 3037 map->elem_count = bpf_map_alloc_percpu(map, size, align, flags); 3038 if (!map->elem_count) 3039 return -ENOMEM; 3040 3041 return 0; 3042 } 3043 3044 static inline void 3045 bpf_map_free_elem_count(struct bpf_map *map) 3046 { 3047 free_percpu(map->elem_count); 3048 } 3049 3050 static inline void bpf_map_inc_elem_count(struct bpf_map *map) 3051 { 3052 this_cpu_inc(*map->elem_count); 3053 } 3054 3055 static inline void bpf_map_dec_elem_count(struct bpf_map *map) 3056 { 3057 this_cpu_dec(*map->elem_count); 3058 } 3059 3060 s64 bpf_map_sum_elem_count(const struct bpf_map *map); 3061 3062 extern int sysctl_unprivileged_bpf_disabled; 3063 3064 bool bpf_token_capable(const struct bpf_token *token, int cap); 3065 3066 static inline bool bpf_allow_ptr_leaks(const struct bpf_token *token) 3067 { 3068 return bpf_token_capable(token, CAP_PERFMON); 3069 } 3070 3071 static inline bool bpf_allow_uninit_stack(const struct bpf_token *token) 3072 { 3073 return bpf_token_capable(token, CAP_PERFMON); 3074 } 3075 3076 static inline bool bpf_bypass_spec_v1(const struct bpf_token *token) 3077 { 3078 return bpf_jit_bypass_spec_v1() || 3079 cpu_mitigations_off() || 3080 bpf_token_capable(token, CAP_PERFMON); 3081 } 3082 3083 static inline bool bpf_bypass_spec_v4(const struct bpf_token *token) 3084 { 3085 return bpf_jit_bypass_spec_v4() || 3086 cpu_mitigations_off() || 3087 bpf_token_capable(token, CAP_PERFMON); 3088 } 3089 3090 int bpf_map_new_fd(struct bpf_map *map, int flags); 3091 int bpf_prog_new_fd(struct bpf_prog *prog); 3092 3093 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 3094 const struct bpf_link_ops *ops, struct bpf_prog *prog, 3095 enum bpf_attach_type attach_type); 3096 void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type, 3097 const struct bpf_link_ops *ops, struct bpf_prog *prog, 3098 enum bpf_attach_type attach_type, bool sleepable); 3099 void bpf_tramp_link_init(struct bpf_tramp_link *link, enum bpf_link_type type, 3100 const struct bpf_link_ops *ops, struct bpf_prog *prog, 3101 enum bpf_attach_type attach_type, u64 cookie); 3102 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer); 3103 int bpf_link_settle(struct bpf_link_primer *primer); 3104 void bpf_link_cleanup(struct bpf_link_primer *primer); 3105 void bpf_link_inc(struct bpf_link *link); 3106 struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link); 3107 void bpf_link_put(struct bpf_link *link); 3108 int bpf_link_new_fd(struct bpf_link *link); 3109 struct bpf_link *bpf_link_get_from_fd(u32 ufd); 3110 struct bpf_link *bpf_link_get_curr_or_next(u32 *id); 3111 3112 void bpf_token_inc(struct bpf_token *token); 3113 void bpf_token_put(struct bpf_token *token); 3114 int bpf_token_create(union bpf_attr *attr); 3115 struct bpf_token *bpf_token_get_from_fd(u32 ufd); 3116 int bpf_token_get_info_by_fd(struct bpf_token *token, 3117 const union bpf_attr *attr, 3118 union bpf_attr __user *uattr); 3119 3120 bool bpf_token_allow_cmd(const struct bpf_token *token, enum bpf_cmd cmd); 3121 bool bpf_token_allow_map_type(const struct bpf_token *token, enum bpf_map_type type); 3122 bool bpf_token_allow_prog_type(const struct bpf_token *token, 3123 enum bpf_prog_type prog_type, 3124 enum bpf_attach_type attach_type); 3125 3126 int bpf_obj_pin_user(u32 ufd, int path_fd, const char __user *pathname); 3127 int bpf_obj_get_user(int path_fd, const char __user *pathname, int flags); 3128 struct inode *bpf_get_inode(struct super_block *sb, const struct inode *dir, 3129 umode_t mode); 3130 3131 #define BPF_ITER_FUNC_PREFIX "bpf_iter_" 3132 #define DEFINE_BPF_ITER_FUNC(target, args...) \ 3133 extern int bpf_iter_ ## target(args); \ 3134 int __init bpf_iter_ ## target(args) { return 0; } 3135 3136 /* 3137 * The task type of iterators. 3138 * 3139 * For BPF task iterators, they can be parameterized with various 3140 * parameters to visit only some of tasks. 3141 * 3142 * BPF_TASK_ITER_ALL (default) 3143 * Iterate over resources of every task. 3144 * 3145 * BPF_TASK_ITER_TID 3146 * Iterate over resources of a task/tid. 3147 * 3148 * BPF_TASK_ITER_TGID 3149 * Iterate over resources of every task of a process / task group. 3150 */ 3151 enum bpf_iter_task_type { 3152 BPF_TASK_ITER_ALL = 0, 3153 BPF_TASK_ITER_TID, 3154 BPF_TASK_ITER_TGID, 3155 }; 3156 3157 struct bpf_iter_aux_info { 3158 /* for map_elem iter */ 3159 struct bpf_map *map; 3160 3161 /* for cgroup iter */ 3162 struct { 3163 struct cgroup *start; /* starting cgroup */ 3164 enum bpf_cgroup_iter_order order; 3165 } cgroup; 3166 struct { 3167 enum bpf_iter_task_type type; 3168 u32 pid; 3169 } task; 3170 }; 3171 3172 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, 3173 union bpf_iter_link_info *linfo, 3174 struct bpf_iter_aux_info *aux); 3175 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux); 3176 typedef int (*bpf_iter_check_prog_t)(const struct bpf_prog *prog, 3177 const struct bpf_iter_aux_info *aux); 3178 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux, 3179 struct seq_file *seq); 3180 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, 3181 struct bpf_link_info *info); 3182 typedef const struct bpf_func_proto * 3183 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id, 3184 const struct bpf_prog *prog); 3185 3186 enum bpf_iter_feature { 3187 BPF_ITER_RESCHED = BIT(0), /* may cond_resched() between objects */ 3188 BPF_ITER_NOSLEEP = BIT(1), /* ->show() can't run sleepable progs */ 3189 }; 3190 3191 #define BPF_ITER_CTX_ARG_MAX 2 3192 struct bpf_iter_reg { 3193 const char *target; 3194 bpf_iter_attach_target_t attach_target; 3195 bpf_iter_detach_target_t detach_target; 3196 /* may the program run on what the link is attached to? */ 3197 bpf_iter_check_prog_t check_prog; 3198 bpf_iter_show_fdinfo_t show_fdinfo; 3199 bpf_iter_fill_link_info_t fill_link_info; 3200 bpf_iter_get_func_proto_t get_func_proto; 3201 u32 ctx_arg_info_size; 3202 u32 feature; 3203 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX]; 3204 const struct bpf_iter_seq_info *seq_info; 3205 }; 3206 3207 struct bpf_iter_meta { 3208 __bpf_md_ptr(struct seq_file *, seq); 3209 u64 session_id; 3210 u64 seq_num; 3211 }; 3212 3213 struct bpf_iter__bpf_map_elem { 3214 __bpf_md_ptr(struct bpf_iter_meta *, meta); 3215 __bpf_md_ptr(struct bpf_map *, map); 3216 __bpf_md_ptr(void *, key); 3217 __bpf_md_ptr(void *, value); 3218 }; 3219 3220 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info); 3221 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info); 3222 int bpf_iter_prog_supported(struct bpf_prog *prog); 3223 const struct bpf_func_proto * 3224 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog); 3225 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog); 3226 int bpf_iter_new_fd(struct bpf_link *link); 3227 bool bpf_link_is_iter(struct bpf_link *link); 3228 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop); 3229 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx); 3230 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, 3231 struct seq_file *seq); 3232 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, 3233 struct bpf_link_info *info); 3234 int bpf_iter_map_check_value_write(const struct bpf_prog *prog, 3235 const struct bpf_map *map); 3236 3237 int map_set_for_each_callback_args(struct bpf_verifier_env *env, 3238 struct bpf_func_state *caller, 3239 struct bpf_func_state *callee); 3240 3241 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value, u64 flags); 3242 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value, u64 flags); 3243 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, 3244 u64 flags); 3245 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, 3246 u64 flags); 3247 3248 int bpf_stackmap_extract(struct bpf_map *map, void *key, void *value, bool delete); 3249 3250 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, 3251 void *key, void *value, u64 map_flags); 3252 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 3253 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, 3254 void *key, void *value, u64 map_flags); 3255 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); 3256 3257 int bpf_get_file_flag(int flags); 3258 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size, 3259 size_t actual_size); 3260 3261 /* verify correctness of eBPF program */ 3262 struct bpf_log_attr; 3263 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr, 3264 struct bpf_log_attr *attr_log); 3265 3266 #ifndef CONFIG_BPF_JIT_ALWAYS_ON 3267 int bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); 3268 s32 bpf_call_args_imm(s16 idx); 3269 #else 3270 static inline s32 bpf_call_args_imm(s16 idx) 3271 { 3272 return 0; 3273 } 3274 #endif 3275 3276 struct btf *bpf_get_btf_vmlinux(void); 3277 3278 /* Map specifics */ 3279 struct xdp_frame; 3280 struct sk_buff; 3281 struct bpf_dtab_netdev; 3282 struct bpf_cpu_map_entry; 3283 3284 void __dev_flush(struct list_head *flush_list); 3285 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 3286 struct net_device *dev_rx); 3287 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 3288 struct net_device *dev_rx); 3289 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 3290 struct bpf_map *map, bool exclude_ingress); 3291 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, 3292 const struct bpf_prog *xdp_prog); 3293 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 3294 const struct bpf_prog *xdp_prog, 3295 struct bpf_map *map, bool exclude_ingress); 3296 3297 void __cpu_map_flush(struct list_head *flush_list); 3298 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf, 3299 struct net_device *dev_rx); 3300 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 3301 struct sk_buff *skb); 3302 3303 /* Return map's numa specified by userspace */ 3304 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) 3305 { 3306 return (attr->map_flags & BPF_F_NUMA_NODE) ? 3307 attr->numa_node : NUMA_NO_NODE; 3308 } 3309 3310 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); 3311 int array_map_alloc_check(union bpf_attr *attr); 3312 3313 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, 3314 union bpf_attr __user *uattr); 3315 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, 3316 union bpf_attr __user *uattr); 3317 int bpf_prog_test_run_tracing(struct bpf_prog *prog, 3318 const union bpf_attr *kattr, 3319 union bpf_attr __user *uattr); 3320 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 3321 const union bpf_attr *kattr, 3322 union bpf_attr __user *uattr); 3323 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, 3324 const union bpf_attr *kattr, 3325 union bpf_attr __user *uattr); 3326 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 3327 const union bpf_attr *kattr, 3328 union bpf_attr __user *uattr); 3329 int bpf_prog_test_run_nf(struct bpf_prog *prog, 3330 const union bpf_attr *kattr, 3331 union bpf_attr __user *uattr); 3332 bool btf_ctx_access(int off, int size, enum bpf_access_type type, 3333 const struct bpf_prog *prog, 3334 struct bpf_insn_access_aux *info); 3335 3336 static inline bool bpf_tracing_ctx_access(int off, int size, 3337 enum bpf_access_type type) 3338 { 3339 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS) 3340 return false; 3341 if (type != BPF_READ) 3342 return false; 3343 if (off % size != 0) 3344 return false; 3345 return true; 3346 } 3347 3348 static inline bool bpf_tracing_btf_ctx_access(int off, int size, 3349 enum bpf_access_type type, 3350 const struct bpf_prog *prog, 3351 struct bpf_insn_access_aux *info) 3352 { 3353 if (!bpf_tracing_ctx_access(off, size, type)) 3354 return false; 3355 return btf_ctx_access(off, size, type, prog, info); 3356 } 3357 3358 int btf_struct_access(struct bpf_verifier_log *log, 3359 const struct bpf_reg_state *reg, 3360 int off, int size, enum bpf_access_type atype, 3361 u32 *next_btf_id, enum bpf_type_flag *flag, const char **field_name); 3362 bool btf_struct_ids_match(struct bpf_verifier_log *log, 3363 const struct btf *btf, u32 id, int off, 3364 const struct btf *need_btf, u32 need_type_id, 3365 bool strict, bool walk_flex_arrays); 3366 3367 int btf_distill_func_proto(struct bpf_verifier_log *log, 3368 struct btf *btf, 3369 const struct btf_type *func_proto, 3370 const char *func_name, 3371 struct btf_func_model *m); 3372 3373 struct bpf_reg_state; 3374 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog); 3375 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, 3376 struct btf *btf, const struct btf_type *t); 3377 const char *btf_find_decl_tag_value(const struct btf *btf, const struct btf_type *pt, 3378 int comp_idx, const char *tag_key); 3379 int btf_find_next_decl_tag(const struct btf *btf, const struct btf_type *pt, 3380 int comp_idx, const char *tag_key, int last_id); 3381 3382 struct bpf_prog *bpf_prog_by_id(u32 id); 3383 struct bpf_link *bpf_link_by_id(u32 id); 3384 3385 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id, 3386 const struct bpf_prog *prog); 3387 void bpf_task_storage_free(struct task_struct *task); 3388 void bpf_cgrp_storage_free(struct cgroup *cgroup); 3389 const struct btf_func_model * 3390 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 3391 const struct bpf_insn *insn); 3392 int bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id, 3393 u16 desc_idx, u8 **func_addr); 3394 3395 struct bpf_core_ctx { 3396 struct bpf_verifier_log *log; 3397 const struct btf *btf; 3398 }; 3399 3400 bool btf_nested_type_is_trusted(struct bpf_verifier_log *log, 3401 const struct bpf_reg_state *reg, 3402 const char *field_name, u32 btf_id, const char *suffix); 3403 3404 bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log, 3405 const struct btf *reg_btf, u32 reg_id, 3406 const struct btf *arg_btf, u32 arg_id); 3407 3408 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo, 3409 int relo_idx, void *insn); 3410 3411 static inline bool unprivileged_ebpf_enabled(void) 3412 { 3413 return !sysctl_unprivileged_bpf_disabled; 3414 } 3415 3416 /* Not all bpf prog type has the bpf_ctx. 3417 * For the bpf prog type that has initialized the bpf_ctx, 3418 * this function can be used to decide if a kernel function 3419 * is called by a bpf program. 3420 */ 3421 static inline bool has_current_bpf_ctx(void) 3422 { 3423 return !!current->bpf_ctx; 3424 } 3425 3426 void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog); 3427 3428 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 3429 enum bpf_dynptr_type type, u32 offset, u32 size); 3430 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr); 3431 void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr); 3432 void bpf_prog_report_arena_violation(bool write, unsigned long addr, unsigned long fault_ip); 3433 3434 static __always_inline u32 3435 bpf_prog_run_array_sleepable(const struct bpf_prog_array *array, 3436 const void *ctx, bpf_prog_run_fn run_prog) 3437 { 3438 const struct bpf_prog_array_item *item; 3439 struct bpf_prog *prog; 3440 struct bpf_run_ctx *old_run_ctx; 3441 struct bpf_trace_run_ctx run_ctx; 3442 u32 ret = 1; 3443 3444 if (unlikely(!array)) 3445 return ret; 3446 3447 migrate_disable(); 3448 3449 run_ctx.is_uprobe = false; 3450 3451 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); 3452 item = &array->items[0]; 3453 while ((prog = READ_ONCE(item->prog))) { 3454 /* Skip dummy_bpf_prog placeholder (len == 0) */ 3455 if (unlikely(!prog->len)) { 3456 item++; 3457 continue; 3458 } 3459 3460 if (unlikely(!bpf_prog_get_recursion_context(prog))) { 3461 bpf_prog_inc_misses_counter(prog); 3462 bpf_prog_put_recursion_context(prog); 3463 item++; 3464 continue; 3465 } 3466 3467 run_ctx.bpf_cookie = item->bpf_cookie; 3468 3469 if (!prog->sleepable) { 3470 guard(rcu)(); 3471 ret &= run_prog(prog, ctx); 3472 } else { 3473 ret &= run_prog(prog, ctx); 3474 } 3475 3476 bpf_prog_put_recursion_context(prog); 3477 item++; 3478 } 3479 bpf_reset_run_ctx(old_run_ctx); 3480 migrate_enable(); 3481 return ret; 3482 } 3483 3484 #else /* !CONFIG_BPF_SYSCALL */ 3485 static inline struct bpf_prog *bpf_prog_get(u32 ufd) 3486 { 3487 return ERR_PTR(-EOPNOTSUPP); 3488 } 3489 3490 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, 3491 enum bpf_prog_type type, 3492 bool attach_drv) 3493 { 3494 return ERR_PTR(-EOPNOTSUPP); 3495 } 3496 3497 static inline void bpf_prog_add(struct bpf_prog *prog, int i) 3498 { 3499 } 3500 3501 static inline void bpf_prog_sub(struct bpf_prog *prog, int i) 3502 { 3503 } 3504 3505 static inline void bpf_prog_put(struct bpf_prog *prog) 3506 { 3507 } 3508 3509 static inline void bpf_prog_inc(struct bpf_prog *prog) 3510 { 3511 } 3512 3513 static inline struct bpf_prog *__must_check 3514 bpf_prog_inc_not_zero(struct bpf_prog *prog) 3515 { 3516 return ERR_PTR(-EOPNOTSUPP); 3517 } 3518 3519 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, 3520 const struct bpf_link_ops *ops, 3521 struct bpf_prog *prog, enum bpf_attach_type attach_type) 3522 { 3523 } 3524 3525 static inline void bpf_link_init_sleepable(struct bpf_link *link, enum bpf_link_type type, 3526 const struct bpf_link_ops *ops, struct bpf_prog *prog, 3527 enum bpf_attach_type attach_type, bool sleepable) 3528 { 3529 } 3530 3531 static inline void bpf_tramp_link_init(struct bpf_tramp_link *link, enum bpf_link_type type, 3532 const struct bpf_link_ops *ops, struct bpf_prog *prog, 3533 enum bpf_attach_type attach_type, u64 cookie) 3534 { 3535 } 3536 3537 static inline int bpf_link_prime(struct bpf_link *link, 3538 struct bpf_link_primer *primer) 3539 { 3540 return -EOPNOTSUPP; 3541 } 3542 3543 static inline int bpf_link_settle(struct bpf_link_primer *primer) 3544 { 3545 return -EOPNOTSUPP; 3546 } 3547 3548 static inline void bpf_link_cleanup(struct bpf_link_primer *primer) 3549 { 3550 } 3551 3552 static inline void bpf_link_inc(struct bpf_link *link) 3553 { 3554 } 3555 3556 static inline struct bpf_link *bpf_link_inc_not_zero(struct bpf_link *link) 3557 { 3558 return NULL; 3559 } 3560 3561 static inline void bpf_link_put(struct bpf_link *link) 3562 { 3563 } 3564 3565 static inline int bpf_obj_get_user(const char __user *pathname, int flags) 3566 { 3567 return -EOPNOTSUPP; 3568 } 3569 3570 static inline bool bpf_token_capable(const struct bpf_token *token, int cap) 3571 { 3572 return capable(cap) || (cap != CAP_SYS_ADMIN && capable(CAP_SYS_ADMIN)); 3573 } 3574 3575 static inline void bpf_token_inc(struct bpf_token *token) 3576 { 3577 } 3578 3579 static inline void bpf_token_put(struct bpf_token *token) 3580 { 3581 } 3582 3583 static inline struct bpf_token *bpf_token_get_from_fd(u32 ufd) 3584 { 3585 return ERR_PTR(-EOPNOTSUPP); 3586 } 3587 3588 static inline int bpf_token_get_info_by_fd(struct bpf_token *token, 3589 const union bpf_attr *attr, 3590 union bpf_attr __user *uattr) 3591 { 3592 return -EOPNOTSUPP; 3593 } 3594 3595 static inline void __dev_flush(struct list_head *flush_list) 3596 { 3597 } 3598 3599 struct xdp_frame; 3600 struct bpf_dtab_netdev; 3601 struct bpf_cpu_map_entry; 3602 3603 static inline 3604 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf, 3605 struct net_device *dev_rx) 3606 { 3607 return 0; 3608 } 3609 3610 static inline 3611 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf, 3612 struct net_device *dev_rx) 3613 { 3614 return 0; 3615 } 3616 3617 static inline 3618 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx, 3619 struct bpf_map *map, bool exclude_ingress) 3620 { 3621 return 0; 3622 } 3623 3624 struct sk_buff; 3625 3626 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, 3627 struct sk_buff *skb, 3628 const struct bpf_prog *xdp_prog) 3629 { 3630 return 0; 3631 } 3632 3633 static inline 3634 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb, 3635 const struct bpf_prog *xdp_prog, 3636 struct bpf_map *map, bool exclude_ingress) 3637 { 3638 return 0; 3639 } 3640 3641 static inline void __cpu_map_flush(struct list_head *flush_list) 3642 { 3643 } 3644 3645 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, 3646 struct xdp_frame *xdpf, 3647 struct net_device *dev_rx) 3648 { 3649 return 0; 3650 } 3651 3652 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu, 3653 struct sk_buff *skb) 3654 { 3655 return -EOPNOTSUPP; 3656 } 3657 3658 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, 3659 enum bpf_prog_type type) 3660 { 3661 return ERR_PTR(-EOPNOTSUPP); 3662 } 3663 3664 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, 3665 const union bpf_attr *kattr, 3666 union bpf_attr __user *uattr) 3667 { 3668 return -ENOTSUPP; 3669 } 3670 3671 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, 3672 const union bpf_attr *kattr, 3673 union bpf_attr __user *uattr) 3674 { 3675 return -ENOTSUPP; 3676 } 3677 3678 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, 3679 const union bpf_attr *kattr, 3680 union bpf_attr __user *uattr) 3681 { 3682 return -ENOTSUPP; 3683 } 3684 3685 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, 3686 const union bpf_attr *kattr, 3687 union bpf_attr __user *uattr) 3688 { 3689 return -ENOTSUPP; 3690 } 3691 3692 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog, 3693 const union bpf_attr *kattr, 3694 union bpf_attr __user *uattr) 3695 { 3696 return -ENOTSUPP; 3697 } 3698 3699 static inline void bpf_map_put(struct bpf_map *map) 3700 { 3701 } 3702 3703 static inline struct bpf_prog *bpf_prog_by_id(u32 id) 3704 { 3705 return ERR_PTR(-ENOTSUPP); 3706 } 3707 3708 static inline int btf_struct_access(struct bpf_verifier_log *log, 3709 const struct bpf_reg_state *reg, 3710 int off, int size, enum bpf_access_type atype, 3711 u32 *next_btf_id, enum bpf_type_flag *flag, 3712 const char **field_name) 3713 { 3714 return -EACCES; 3715 } 3716 3717 static inline const struct bpf_func_proto * 3718 bpf_base_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog) 3719 { 3720 return NULL; 3721 } 3722 3723 static inline void bpf_task_storage_free(struct task_struct *task) 3724 { 3725 } 3726 3727 static inline const struct btf_func_model * 3728 bpf_jit_find_kfunc_model(const struct bpf_prog *prog, 3729 const struct bpf_insn *insn) 3730 { 3731 return NULL; 3732 } 3733 3734 static inline int 3735 bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id, 3736 u16 desc_idx, u8 **func_addr) 3737 { 3738 return -ENOTSUPP; 3739 } 3740 3741 static inline bool unprivileged_ebpf_enabled(void) 3742 { 3743 return false; 3744 } 3745 3746 static inline bool has_current_bpf_ctx(void) 3747 { 3748 return false; 3749 } 3750 3751 static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog) 3752 { 3753 } 3754 3755 static inline void bpf_cgrp_storage_free(struct cgroup *cgroup) 3756 { 3757 } 3758 3759 static inline void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data, 3760 enum bpf_dynptr_type type, u32 offset, u32 size) 3761 { 3762 } 3763 3764 static inline void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr) 3765 { 3766 } 3767 3768 static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr) 3769 { 3770 } 3771 3772 static inline void bpf_prog_report_arena_violation(bool write, unsigned long addr, 3773 unsigned long fault_ip) 3774 { 3775 } 3776 #endif /* CONFIG_BPF_SYSCALL */ 3777 3778 static inline bool bpf_net_capable(void) 3779 { 3780 return capable(CAP_NET_ADMIN) || capable(CAP_SYS_ADMIN); 3781 } 3782 3783 static __always_inline int 3784 bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr) 3785 { 3786 int ret = -EFAULT; 3787 3788 if (IS_ENABLED(CONFIG_BPF_EVENTS)) 3789 ret = copy_from_kernel_nofault(dst, unsafe_ptr, size); 3790 if (unlikely(ret < 0)) 3791 memset(dst, 0, size); 3792 return ret; 3793 } 3794 3795 void __bpf_free_used_btfs(struct btf_mod_pair *used_btfs, u32 len); 3796 3797 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, 3798 enum bpf_prog_type type) 3799 { 3800 return bpf_prog_get_type_dev(ufd, type, false); 3801 } 3802 3803 void __bpf_free_used_maps(struct bpf_prog_aux *aux, 3804 struct bpf_map **used_maps, u32 len); 3805 3806 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); 3807 3808 int bpf_prog_offload_compile(struct bpf_prog *prog); 3809 void bpf_prog_dev_bound_destroy(struct bpf_prog *prog); 3810 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, 3811 struct bpf_prog *prog); 3812 3813 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); 3814 3815 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); 3816 int bpf_map_offload_update_elem(struct bpf_map *map, 3817 void *key, void *value, u64 flags); 3818 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); 3819 int bpf_map_offload_get_next_key(struct bpf_map *map, 3820 void *key, void *next_key); 3821 3822 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); 3823 3824 struct bpf_offload_dev * 3825 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); 3826 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); 3827 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); 3828 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, 3829 struct net_device *netdev); 3830 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, 3831 struct net_device *netdev); 3832 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); 3833 3834 void unpriv_ebpf_notify(int new_state); 3835 3836 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) 3837 int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log, 3838 struct bpf_prog_aux *prog_aux); 3839 void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, u32 func_id); 3840 int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr); 3841 int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, struct bpf_prog *old_prog); 3842 void bpf_prog_dev_bound_moved(struct bpf_prog *prog); 3843 void bpf_dev_bound_netdev_unregister(struct net_device *dev); 3844 3845 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 3846 { 3847 return aux->dev_bound; 3848 } 3849 3850 static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux) 3851 { 3852 return aux->offload_requested; 3853 } 3854 3855 bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs); 3856 3857 static inline bool bpf_map_is_offloaded(struct bpf_map *map) 3858 { 3859 return unlikely(map->ops == &bpf_map_offload_ops); 3860 } 3861 3862 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); 3863 void bpf_map_offload_map_free(struct bpf_map *map); 3864 u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map); 3865 int bpf_prog_test_run_syscall(struct bpf_prog *prog, 3866 const union bpf_attr *kattr, 3867 union bpf_attr __user *uattr); 3868 3869 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); 3870 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype); 3871 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags); 3872 int sock_map_bpf_prog_query(const union bpf_attr *attr, 3873 union bpf_attr __user *uattr); 3874 int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog); 3875 3876 void sock_map_unhash(struct sock *sk); 3877 void sock_map_destroy(struct sock *sk); 3878 void sock_map_close(struct sock *sk, long timeout); 3879 #else 3880 static inline int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log, 3881 struct bpf_prog_aux *prog_aux) 3882 { 3883 return -EOPNOTSUPP; 3884 } 3885 3886 static inline void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, 3887 u32 func_id) 3888 { 3889 return NULL; 3890 } 3891 3892 static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog, 3893 union bpf_attr *attr) 3894 { 3895 return -EOPNOTSUPP; 3896 } 3897 3898 static inline int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, 3899 struct bpf_prog *old_prog) 3900 { 3901 return -EOPNOTSUPP; 3902 } 3903 3904 static inline void bpf_prog_dev_bound_moved(struct bpf_prog *prog) 3905 { 3906 } 3907 3908 static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev) 3909 { 3910 } 3911 3912 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) 3913 { 3914 return false; 3915 } 3916 3917 static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux) 3918 { 3919 return false; 3920 } 3921 3922 static inline bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs) 3923 { 3924 return false; 3925 } 3926 3927 static inline bool bpf_map_is_offloaded(struct bpf_map *map) 3928 { 3929 return false; 3930 } 3931 3932 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) 3933 { 3934 return ERR_PTR(-EOPNOTSUPP); 3935 } 3936 3937 static inline void bpf_map_offload_map_free(struct bpf_map *map) 3938 { 3939 } 3940 3941 static inline u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map) 3942 { 3943 return 0; 3944 } 3945 3946 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog, 3947 const union bpf_attr *kattr, 3948 union bpf_attr __user *uattr) 3949 { 3950 return -ENOTSUPP; 3951 } 3952 3953 #ifdef CONFIG_BPF_SYSCALL 3954 static inline int sock_map_get_from_fd(const union bpf_attr *attr, 3955 struct bpf_prog *prog) 3956 { 3957 return -EINVAL; 3958 } 3959 3960 static inline int sock_map_prog_detach(const union bpf_attr *attr, 3961 enum bpf_prog_type ptype) 3962 { 3963 return -EOPNOTSUPP; 3964 } 3965 3966 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, 3967 u64 flags) 3968 { 3969 return -EOPNOTSUPP; 3970 } 3971 3972 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr, 3973 union bpf_attr __user *uattr) 3974 { 3975 return -EINVAL; 3976 } 3977 3978 static inline int sock_map_link_create(const union bpf_attr *attr, struct bpf_prog *prog) 3979 { 3980 return -EOPNOTSUPP; 3981 } 3982 #endif /* CONFIG_BPF_SYSCALL */ 3983 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ 3984 3985 static __always_inline void 3986 bpf_prog_inc_misses_counters(const struct bpf_prog_array *array) 3987 { 3988 const struct bpf_prog_array_item *item; 3989 struct bpf_prog *prog; 3990 3991 if (unlikely(!array)) 3992 return; 3993 3994 item = &array->items[0]; 3995 while ((prog = READ_ONCE(item->prog))) { 3996 bpf_prog_inc_misses_counter(prog); 3997 item++; 3998 } 3999 } 4000 4001 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) 4002 void bpf_sk_reuseport_detach(struct sock *sk); 4003 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, 4004 void *value); 4005 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, 4006 void *value, u64 map_flags); 4007 #else 4008 static inline void bpf_sk_reuseport_detach(struct sock *sk) 4009 { 4010 } 4011 4012 #ifdef CONFIG_BPF_SYSCALL 4013 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, 4014 void *key, void *value) 4015 { 4016 return -EOPNOTSUPP; 4017 } 4018 4019 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, 4020 void *key, void *value, 4021 u64 map_flags) 4022 { 4023 return -EOPNOTSUPP; 4024 } 4025 #endif /* CONFIG_BPF_SYSCALL */ 4026 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ 4027 4028 #ifdef CONFIG_KEYS 4029 struct bpf_key { 4030 struct key *key; 4031 bool has_ref; 4032 }; 4033 #endif /* CONFIG_KEYS */ 4034 4035 #if defined(CONFIG_KEYS) && defined(CONFIG_BPF_SYSCALL) 4036 struct bpf_key *bpf_lookup_user_key(s32 serial, u64 flags); 4037 struct bpf_key *bpf_lookup_system_key(u64 id); 4038 struct bpf_key *bpf_lookup_keyring(void); 4039 bool bpf_keyring_enforced(void); 4040 void bpf_key_put(struct bpf_key *bkey); 4041 int bpf_verify_pkcs7_signature(const struct bpf_dynptr *data_p, 4042 const struct bpf_dynptr *sig_p, 4043 struct bpf_key *trusted_keyring); 4044 4045 static inline s32 bpf_key_serial(const struct bpf_key *key) 4046 { 4047 return key->has_ref ? key->key->serial : 0; 4048 } 4049 #else 4050 static inline struct bpf_key *bpf_lookup_user_key(u32 serial, u64 flags) 4051 { 4052 return NULL; 4053 } 4054 4055 static inline struct bpf_key *bpf_lookup_system_key(u64 id) 4056 { 4057 return NULL; 4058 } 4059 4060 static inline struct bpf_key *bpf_lookup_keyring(void) 4061 { 4062 return NULL; 4063 } 4064 4065 static inline bool bpf_keyring_enforced(void) 4066 { 4067 return false; 4068 } 4069 4070 static inline void bpf_key_put(struct bpf_key *bkey) 4071 { 4072 } 4073 4074 static inline int bpf_verify_pkcs7_signature(const struct bpf_dynptr *data_p, 4075 const struct bpf_dynptr *sig_p, 4076 struct bpf_key *trusted_keyring) 4077 { 4078 return -EOPNOTSUPP; 4079 } 4080 4081 static inline s32 bpf_key_serial(const struct bpf_key *key) 4082 { 4083 return 0; 4084 } 4085 #endif /* defined(CONFIG_KEYS) && defined(CONFIG_BPF_SYSCALL) */ 4086 4087 /* verifier prototypes for helper functions called from eBPF programs */ 4088 extern const struct bpf_func_proto bpf_map_lookup_elem_proto; 4089 extern const struct bpf_func_proto bpf_map_update_elem_proto; 4090 extern const struct bpf_func_proto bpf_map_delete_elem_proto; 4091 extern const struct bpf_func_proto bpf_map_push_elem_proto; 4092 extern const struct bpf_func_proto bpf_map_pop_elem_proto; 4093 extern const struct bpf_func_proto bpf_map_peek_elem_proto; 4094 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto; 4095 4096 extern const struct bpf_func_proto bpf_get_prandom_u32_proto; 4097 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; 4098 extern const struct bpf_func_proto bpf_get_numa_node_id_proto; 4099 extern const struct bpf_func_proto bpf_tail_call_proto; 4100 extern const struct bpf_func_proto bpf_ktime_get_ns_proto; 4101 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto; 4102 extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto; 4103 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; 4104 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; 4105 extern const struct bpf_func_proto bpf_get_current_comm_proto; 4106 extern const struct bpf_func_proto bpf_get_stackid_proto; 4107 extern const struct bpf_func_proto bpf_get_stack_proto; 4108 extern const struct bpf_func_proto bpf_get_stack_sleepable_proto; 4109 extern const struct bpf_func_proto bpf_get_task_stack_proto; 4110 extern const struct bpf_func_proto bpf_get_task_stack_sleepable_proto; 4111 extern const struct bpf_func_proto bpf_get_stackid_proto_pe; 4112 extern const struct bpf_func_proto bpf_get_stack_proto_pe; 4113 extern const struct bpf_func_proto bpf_sock_map_update_proto; 4114 extern const struct bpf_func_proto bpf_sock_hash_update_proto; 4115 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; 4116 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto; 4117 extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto; 4118 extern const struct bpf_func_proto bpf_current_task_under_cgroup_proto; 4119 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; 4120 extern const struct bpf_func_proto bpf_msg_redirect_map_proto; 4121 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; 4122 extern const struct bpf_func_proto bpf_sk_redirect_map_proto; 4123 extern const struct bpf_func_proto bpf_spin_lock_proto; 4124 extern const struct bpf_func_proto bpf_spin_unlock_proto; 4125 extern const struct bpf_func_proto bpf_get_local_storage_proto; 4126 extern const struct bpf_func_proto bpf_strtol_proto; 4127 extern const struct bpf_func_proto bpf_strtoul_proto; 4128 extern const struct bpf_func_proto bpf_tcp_sock_proto; 4129 extern const struct bpf_func_proto bpf_jiffies64_proto; 4130 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto; 4131 extern const struct bpf_func_proto bpf_event_output_data_proto; 4132 extern const struct bpf_func_proto bpf_ringbuf_output_proto; 4133 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto; 4134 extern const struct bpf_func_proto bpf_ringbuf_submit_proto; 4135 extern const struct bpf_func_proto bpf_ringbuf_discard_proto; 4136 extern const struct bpf_func_proto bpf_ringbuf_query_proto; 4137 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto; 4138 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto; 4139 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto; 4140 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto; 4141 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto; 4142 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto; 4143 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto; 4144 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto; 4145 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto; 4146 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto; 4147 extern const struct bpf_func_proto bpf_copy_from_user_proto; 4148 extern const struct bpf_func_proto bpf_snprintf_btf_proto; 4149 extern const struct bpf_func_proto bpf_snprintf_proto; 4150 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto; 4151 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto; 4152 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto; 4153 extern const struct bpf_func_proto bpf_sock_from_file_proto; 4154 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto; 4155 extern const struct bpf_func_proto bpf_task_storage_get_recur_proto; 4156 extern const struct bpf_func_proto bpf_task_storage_get_proto; 4157 extern const struct bpf_func_proto bpf_task_storage_delete_recur_proto; 4158 extern const struct bpf_func_proto bpf_task_storage_delete_proto; 4159 extern const struct bpf_func_proto bpf_for_each_map_elem_proto; 4160 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto; 4161 extern const struct bpf_func_proto bpf_sk_setsockopt_proto; 4162 extern const struct bpf_func_proto bpf_sk_getsockopt_proto; 4163 extern const struct bpf_func_proto bpf_sk_setsockopt_nodelay_proto; 4164 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto; 4165 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto; 4166 extern const struct bpf_func_proto bpf_find_vma_proto; 4167 extern const struct bpf_func_proto bpf_loop_proto; 4168 extern const struct bpf_func_proto bpf_copy_from_user_task_proto; 4169 extern const struct bpf_func_proto bpf_set_retval_proto; 4170 extern const struct bpf_func_proto bpf_get_retval_proto; 4171 extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto; 4172 extern const struct bpf_func_proto bpf_cgrp_storage_get_proto; 4173 extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto; 4174 4175 const struct bpf_func_proto *tracing_prog_func_proto( 4176 enum bpf_func_id func_id, const struct bpf_prog *prog); 4177 4178 /* Shared helpers among cBPF and eBPF. */ 4179 void bpf_user_rnd_init_once(void); 4180 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 4181 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); 4182 4183 #if defined(CONFIG_NET) 4184 bool bpf_sock_common_is_valid_access(int off, int size, 4185 enum bpf_access_type type, 4186 struct bpf_insn_access_aux *info); 4187 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, 4188 struct bpf_insn_access_aux *info); 4189 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 4190 const struct bpf_insn *si, 4191 struct bpf_insn *insn_buf, 4192 struct bpf_prog *prog, 4193 u32 *target_size); 4194 int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags, 4195 struct bpf_dynptr *ptr); 4196 #else 4197 static inline bool bpf_sock_common_is_valid_access(int off, int size, 4198 enum bpf_access_type type, 4199 struct bpf_insn_access_aux *info) 4200 { 4201 return false; 4202 } 4203 static inline bool bpf_sock_is_valid_access(int off, int size, 4204 enum bpf_access_type type, 4205 struct bpf_insn_access_aux *info) 4206 { 4207 return false; 4208 } 4209 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, 4210 const struct bpf_insn *si, 4211 struct bpf_insn *insn_buf, 4212 struct bpf_prog *prog, 4213 u32 *target_size) 4214 { 4215 return 0; 4216 } 4217 static inline int bpf_dynptr_from_skb_rdonly(struct __sk_buff *skb, u64 flags, 4218 struct bpf_dynptr *ptr) 4219 { 4220 return -EOPNOTSUPP; 4221 } 4222 #endif 4223 4224 #ifdef CONFIG_INET 4225 struct sk_reuseport_kern { 4226 struct sk_buff *skb; 4227 struct sock *sk; 4228 struct sock *selected_sk; 4229 struct sock *migrating_sk; 4230 void *data_end; 4231 u32 hash; 4232 u32 reuseport_id; 4233 bool bind_inany; 4234 }; 4235 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 4236 struct bpf_insn_access_aux *info); 4237 4238 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 4239 const struct bpf_insn *si, 4240 struct bpf_insn *insn_buf, 4241 struct bpf_prog *prog, 4242 u32 *target_size); 4243 4244 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, 4245 struct bpf_insn_access_aux *info); 4246 4247 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 4248 const struct bpf_insn *si, 4249 struct bpf_insn *insn_buf, 4250 struct bpf_prog *prog, 4251 u32 *target_size); 4252 #else 4253 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, 4254 enum bpf_access_type type, 4255 struct bpf_insn_access_aux *info) 4256 { 4257 return false; 4258 } 4259 4260 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, 4261 const struct bpf_insn *si, 4262 struct bpf_insn *insn_buf, 4263 struct bpf_prog *prog, 4264 u32 *target_size) 4265 { 4266 return 0; 4267 } 4268 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, 4269 enum bpf_access_type type, 4270 struct bpf_insn_access_aux *info) 4271 { 4272 return false; 4273 } 4274 4275 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, 4276 const struct bpf_insn *si, 4277 struct bpf_insn *insn_buf, 4278 struct bpf_prog *prog, 4279 u32 *target_size) 4280 { 4281 return 0; 4282 } 4283 #endif /* CONFIG_INET */ 4284 4285 enum bpf_text_poke_type { 4286 BPF_MOD_NOP, 4287 BPF_MOD_CALL, 4288 BPF_MOD_JUMP, 4289 }; 4290 4291 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type old_t, 4292 enum bpf_text_poke_type new_t, void *old_addr, 4293 void *new_addr); 4294 4295 void bpf_arch_poke_desc_update(struct bpf_jit_poke_descriptor *poke, 4296 struct bpf_prog *new, struct bpf_prog *old); 4297 4298 void *bpf_arch_text_copy(void *dst, void *src, size_t len); 4299 int bpf_arch_text_invalidate(void *dst, size_t len); 4300 4301 struct btf_id_set; 4302 bool btf_id_set_contains(const struct btf_id_set *set, u32 id); 4303 4304 #define MAX_BPRINTF_VARARGS 12 4305 #define MAX_BPRINTF_BUF 1024 4306 4307 /* Per-cpu temp buffers used by printf-like helpers to store the bprintf binary 4308 * arguments representation. 4309 */ 4310 #define MAX_BPRINTF_BIN_ARGS 512 4311 4312 struct bpf_bprintf_buffers { 4313 char bin_args[MAX_BPRINTF_BIN_ARGS]; 4314 char buf[MAX_BPRINTF_BUF]; 4315 }; 4316 4317 struct bpf_bprintf_data { 4318 u32 *bin_args; 4319 char *buf; 4320 bool get_bin_args; 4321 bool get_buf; 4322 }; 4323 4324 int bpf_bprintf_prepare(const char *fmt, u32 fmt_size, const u64 *raw_args, 4325 u32 num_args, struct bpf_bprintf_data *data); 4326 void bpf_bprintf_cleanup(struct bpf_bprintf_data *data); 4327 int bpf_try_get_buffers(struct bpf_bprintf_buffers **bufs); 4328 void bpf_put_buffers(void); 4329 4330 int bpf_prog_stream_init(struct bpf_prog *prog, gfp_t gfp_extra_flags); 4331 void bpf_prog_stream_free(struct bpf_prog *prog); 4332 int bpf_prog_stream_read(struct bpf_prog *prog, enum bpf_stream_id stream_id, void __user *buf, u32 len); 4333 int bpf_prog_stream_new_fd(struct bpf_prog *prog, enum bpf_stream_id stream_id, u32 flags); 4334 void bpf_stream_stage_init(struct bpf_stream_stage *ss); 4335 void bpf_stream_stage_free(struct bpf_stream_stage *ss); 4336 __printf(2, 3) 4337 int bpf_stream_stage_printk(struct bpf_stream_stage *ss, const char *fmt, ...); 4338 int bpf_stream_stage_commit(struct bpf_stream_stage *ss, struct bpf_prog *prog, 4339 enum bpf_stream_id stream_id); 4340 int bpf_stream_stage_dump_stack(struct bpf_stream_stage *ss); 4341 4342 #define bpf_stream_printk(ss, ...) bpf_stream_stage_printk(&ss, __VA_ARGS__) 4343 #define bpf_stream_dump_stack(ss) bpf_stream_stage_dump_stack(&ss) 4344 4345 #define bpf_stream_stage(ss, prog, stream_id, expr) \ 4346 ({ \ 4347 bpf_stream_stage_init(&ss); \ 4348 (expr); \ 4349 bpf_stream_stage_commit(&ss, prog, stream_id); \ 4350 bpf_stream_stage_free(&ss); \ 4351 }) 4352 4353 #ifdef CONFIG_BPF_LSM 4354 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype); 4355 void bpf_cgroup_atype_put(int cgroup_atype); 4356 #else 4357 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {} 4358 static inline void bpf_cgroup_atype_put(int cgroup_atype) {} 4359 #endif /* CONFIG_BPF_LSM */ 4360 4361 static inline bool type_is_alloc(u32 type) 4362 { 4363 return type & MEM_ALLOC; 4364 } 4365 4366 static inline gfp_t bpf_memcg_flags(gfp_t flags) 4367 { 4368 if (memcg_bpf_enabled()) 4369 return flags | __GFP_ACCOUNT; 4370 return flags; 4371 } 4372 4373 static inline bool bpf_is_subprog(const struct bpf_prog *prog) 4374 { 4375 return prog->aux->func_idx != 0; 4376 } 4377 4378 const struct bpf_line_info *bpf_find_linfo(const struct bpf_prog *prog, u32 insn_off); 4379 struct bpf_linfo_source { 4380 const char *file; 4381 const char *line; 4382 u32 file_name_off; 4383 int line_num; 4384 int line_col; 4385 }; 4386 4387 void bpf_get_linfo_source(struct btf *btf, const struct bpf_line_info *linfo, 4388 struct bpf_linfo_source *src); 4389 int bpf_prog_get_file_line(struct bpf_prog *prog, unsigned long ip, const char **filep, 4390 const char **linep, int *nump); 4391 struct bpf_prog *bpf_prog_find_from_stack(void); 4392 4393 int bpf_insn_array_init(struct bpf_map *map, const struct bpf_prog *prog); 4394 int bpf_insn_array_ready(struct bpf_map *map); 4395 void bpf_insn_array_release(struct bpf_map *map); 4396 void bpf_insn_array_adjust(struct bpf_map *map, u32 first, u32 len); 4397 void bpf_insn_array_adjust_after_remove(struct bpf_map *map, u32 off, u32 len); 4398 4399 #ifdef CONFIG_BPF_SYSCALL 4400 void bpf_prog_update_insn_ptrs(struct bpf_prog *prog, u32 *offsets, void *image); 4401 #else 4402 static inline void 4403 bpf_prog_update_insn_ptrs(struct bpf_prog *prog, u32 *offsets, void *image) 4404 { 4405 } 4406 #endif 4407 4408 static inline bool bpf_map_is_percpu_map(enum bpf_map_type map_type) 4409 { 4410 switch (map_type) { 4411 case BPF_MAP_TYPE_PERCPU_ARRAY: 4412 case BPF_MAP_TYPE_PERCPU_HASH: 4413 case BPF_MAP_TYPE_LRU_PERCPU_HASH: 4414 case BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE: 4415 return true; 4416 default: 4417 return false; 4418 } 4419 } 4420 4421 #define BPF_EXIST_FLAGS (BPF_EXIST | BPF_NOEXIST) 4422 #define BPF_CPU_FLAGS (BPF_F_CPU | BPF_F_ALL_CPUS) 4423 4424 static inline int bpf_map_check_op_flags(struct bpf_map *map, u64 flags, u64 allowed_flags) 4425 { 4426 u32 cpu; 4427 4428 if ((u32)flags & ~allowed_flags) 4429 return -EINVAL; 4430 4431 if ((flags & BPF_F_LOCK) && !btf_record_has_field(map->record, BPF_SPIN_LOCK)) 4432 return -EINVAL; 4433 4434 if (!(flags & BPF_F_CPU) && flags >> 32) 4435 return -EINVAL; 4436 4437 if (flags & (BPF_F_CPU | BPF_F_ALL_CPUS)) { 4438 if (!bpf_map_is_percpu_map(map->map_type)) 4439 return -EINVAL; 4440 if ((flags & BPF_F_CPU) && (flags & BPF_F_ALL_CPUS)) 4441 return -EINVAL; 4442 4443 cpu = flags >> 32; 4444 if ((flags & BPF_F_CPU) && (cpu >= nr_cpu_ids || !cpu_possible(cpu))) 4445 return -ERANGE; 4446 } 4447 4448 return 0; 4449 } 4450 4451 #endif /* _LINUX_BPF_H */ 4452