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