1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
7 #include <uapi/linux/bpf.h>
9 #include <linux/workqueue.h>
10 #include <linux/file.h>
11 #include <linux/percpu.h>
12 #include <linux/err.h>
13 #include <linux/rbtree_latch.h>
14 #include <linux/numa.h>
15 #include <linux/mm_types.h>
16 #include <linux/wait.h>
17 #include <linux/u64_stats_sync.h>
18 #include <linux/refcount.h>
19 #include <linux/mutex.h>
20 #include <linux/module.h>
21 #include <linux/kallsyms.h>
22 #include <linux/capability.h>
24 struct bpf_verifier_env;
25 struct bpf_verifier_log;
34 struct exception_table_entry;
35 struct seq_operations;
36 struct bpf_iter_aux_info;
37 struct bpf_local_storage;
38 struct bpf_local_storage_map;
40 extern struct idr btf_idr;
41 extern spinlock_t btf_idr_lock;
43 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data,
44 struct bpf_iter_aux_info *aux);
45 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
46 struct bpf_iter_seq_info {
47 const struct seq_operations *seq_ops;
48 bpf_iter_init_seq_priv_t init_seq_private;
49 bpf_iter_fini_seq_priv_t fini_seq_private;
53 /* map is generic key/value storage optionally accesible by eBPF programs */
55 /* funcs callable from userspace (via syscall) */
56 int (*map_alloc_check)(union bpf_attr *attr);
57 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
58 void (*map_release)(struct bpf_map *map, struct file *map_file);
59 void (*map_free)(struct bpf_map *map);
60 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
61 void (*map_release_uref)(struct bpf_map *map);
62 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
63 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr,
64 union bpf_attr __user *uattr);
65 int (*map_lookup_and_delete_batch)(struct bpf_map *map,
66 const union bpf_attr *attr,
67 union bpf_attr __user *uattr);
68 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr,
69 union bpf_attr __user *uattr);
70 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr,
71 union bpf_attr __user *uattr);
73 /* funcs callable from userspace and from eBPF programs */
74 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
75 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
76 int (*map_delete_elem)(struct bpf_map *map, void *key);
77 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
78 int (*map_pop_elem)(struct bpf_map *map, void *value);
79 int (*map_peek_elem)(struct bpf_map *map, void *value);
81 /* funcs called by prog_array and perf_event_array map */
82 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
84 void (*map_fd_put_ptr)(void *ptr);
85 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
86 u32 (*map_fd_sys_lookup_elem)(void *ptr);
87 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
89 int (*map_check_btf)(const struct bpf_map *map,
90 const struct btf *btf,
91 const struct btf_type *key_type,
92 const struct btf_type *value_type);
94 /* Prog poke tracking helpers. */
95 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
96 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
97 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
98 struct bpf_prog *new);
100 /* Direct value access helpers. */
101 int (*map_direct_value_addr)(const struct bpf_map *map,
103 int (*map_direct_value_meta)(const struct bpf_map *map,
105 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
106 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp,
107 struct poll_table_struct *pts);
109 /* Functions called by bpf_local_storage maps */
110 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap,
111 void *owner, u32 size);
112 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap,
113 void *owner, u32 size);
114 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner);
115 /* BTF name and id of struct allocated by map_alloc */
116 const char * const map_btf_name;
119 /* bpf_iter info used to open a seq_file */
120 const struct bpf_iter_seq_info *iter_seq_info;
123 struct bpf_map_memory {
125 struct user_struct *user;
129 /* The first two cachelines with read-mostly members of which some
130 * are also accessed in fast-path (e.g. ops, max_entries).
132 const struct bpf_map_ops *ops ____cacheline_aligned;
133 struct bpf_map *inner_map_meta;
134 #ifdef CONFIG_SECURITY
137 enum bpf_map_type map_type;
142 int spin_lock_off; /* >=0 valid offset, <0 error */
146 u32 btf_value_type_id;
148 struct bpf_map_memory memory;
149 char name[BPF_OBJ_NAME_LEN];
150 u32 btf_vmlinux_value_type_id;
152 bool frozen; /* write-once; write-protected by freeze_mutex */
155 /* The 3rd and 4th cacheline with misc members to avoid false sharing
156 * particularly with refcounting.
158 atomic64_t refcnt ____cacheline_aligned;
160 struct work_struct work;
161 struct mutex freeze_mutex;
162 u64 writecnt; /* writable mmap cnt; protected by freeze_mutex */
165 static inline bool map_value_has_spin_lock(const struct bpf_map *map)
167 return map->spin_lock_off >= 0;
170 static inline void check_and_init_map_lock(struct bpf_map *map, void *dst)
172 if (likely(!map_value_has_spin_lock(map)))
174 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) =
175 (struct bpf_spin_lock){};
178 /* copy everything but bpf_spin_lock */
179 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
181 if (unlikely(map_value_has_spin_lock(map))) {
182 u32 off = map->spin_lock_off;
184 memcpy(dst, src, off);
185 memcpy(dst + off + sizeof(struct bpf_spin_lock),
186 src + off + sizeof(struct bpf_spin_lock),
187 map->value_size - off - sizeof(struct bpf_spin_lock));
189 memcpy(dst, src, map->value_size);
192 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
194 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
196 struct bpf_offload_dev;
197 struct bpf_offloaded_map;
199 struct bpf_map_dev_ops {
200 int (*map_get_next_key)(struct bpf_offloaded_map *map,
201 void *key, void *next_key);
202 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
203 void *key, void *value);
204 int (*map_update_elem)(struct bpf_offloaded_map *map,
205 void *key, void *value, u64 flags);
206 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
209 struct bpf_offloaded_map {
211 struct net_device *netdev;
212 const struct bpf_map_dev_ops *dev_ops;
214 struct list_head offloads;
217 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
219 return container_of(map, struct bpf_offloaded_map, map);
222 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
224 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
227 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
229 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
230 map->ops->map_seq_show_elem;
233 int map_check_no_btf(const struct bpf_map *map,
234 const struct btf *btf,
235 const struct btf_type *key_type,
236 const struct btf_type *value_type);
238 extern const struct bpf_map_ops bpf_map_offload_ops;
240 /* function argument constraints */
242 ARG_DONTCARE = 0, /* unused argument in helper function */
244 /* the following constraints used to prototype
245 * bpf_map_lookup/update/delete_elem() functions
247 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
248 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
249 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
250 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */
251 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */
253 /* the following constraints used to prototype bpf_memcmp() and other
254 * functions that access data on eBPF program stack
256 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
257 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
258 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
259 * helper function must fill all bytes or clear
260 * them in error case.
263 ARG_CONST_SIZE, /* number of bytes accessed from memory */
264 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
266 ARG_PTR_TO_CTX, /* pointer to context */
267 ARG_PTR_TO_CTX_OR_NULL, /* pointer to context or NULL */
268 ARG_ANYTHING, /* any (initialized) argument is ok */
269 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
270 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
271 ARG_PTR_TO_INT, /* pointer to int */
272 ARG_PTR_TO_LONG, /* pointer to long */
273 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
274 ARG_PTR_TO_SOCKET_OR_NULL, /* pointer to bpf_sock (fullsock) or NULL */
275 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
276 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */
277 ARG_PTR_TO_ALLOC_MEM_OR_NULL, /* pointer to dynamically allocated memory or NULL */
278 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
281 /* type of values returned from helper functions */
282 enum bpf_return_type {
283 RET_INTEGER, /* function returns integer */
284 RET_VOID, /* function doesn't return anything */
285 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
286 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
287 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */
288 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */
289 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */
290 RET_PTR_TO_ALLOC_MEM_OR_NULL, /* returns a pointer to dynamically allocated memory or NULL */
291 RET_PTR_TO_BTF_ID_OR_NULL, /* returns a pointer to a btf_id or NULL */
294 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
295 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
296 * instructions after verifying
298 struct bpf_func_proto {
299 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
302 enum bpf_return_type ret_type;
305 enum bpf_arg_type arg1_type;
306 enum bpf_arg_type arg2_type;
307 enum bpf_arg_type arg3_type;
308 enum bpf_arg_type arg4_type;
309 enum bpf_arg_type arg5_type;
311 enum bpf_arg_type arg_type[5];
313 int *btf_id; /* BTF ids of arguments */
314 bool (*check_btf_id)(u32 btf_id, u32 arg); /* if the argument btf_id is
315 * valid. Often used if more
316 * than one btf id is permitted
319 int *ret_btf_id; /* return value btf_id */
320 bool (*allowed)(const struct bpf_prog *prog);
323 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
324 * the first argument to eBPF programs.
325 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
329 enum bpf_access_type {
334 /* types of values stored in eBPF registers */
335 /* Pointer types represent:
338 * pointer + (u16) var
339 * pointer + (u16) var + imm
340 * if (range > 0) then [ptr, ptr + range - off) is safe to access
341 * if (id > 0) means that some 'var' was added
342 * if (off > 0) means that 'imm' was added
345 NOT_INIT = 0, /* nothing was written into register */
346 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
347 PTR_TO_CTX, /* reg points to bpf_context */
348 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
349 PTR_TO_MAP_VALUE, /* reg points to map element value */
350 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
351 PTR_TO_STACK, /* reg == frame_pointer + offset */
352 PTR_TO_PACKET_META, /* skb->data - meta_len */
353 PTR_TO_PACKET, /* reg points to skb->data */
354 PTR_TO_PACKET_END, /* skb->data + headlen */
355 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
356 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
357 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */
358 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
359 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */
360 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
361 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */
362 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
363 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
364 PTR_TO_BTF_ID, /* reg points to kernel struct */
365 PTR_TO_BTF_ID_OR_NULL, /* reg points to kernel struct or NULL */
366 PTR_TO_MEM, /* reg points to valid memory region */
367 PTR_TO_MEM_OR_NULL, /* reg points to valid memory region or NULL */
368 PTR_TO_RDONLY_BUF, /* reg points to a readonly buffer */
369 PTR_TO_RDONLY_BUF_OR_NULL, /* reg points to a readonly buffer or NULL */
370 PTR_TO_RDWR_BUF, /* reg points to a read/write buffer */
371 PTR_TO_RDWR_BUF_OR_NULL, /* reg points to a read/write buffer or NULL */
374 /* The information passed from prog-specific *_is_valid_access
375 * back to the verifier.
377 struct bpf_insn_access_aux {
378 enum bpf_reg_type reg_type;
383 struct bpf_verifier_log *log; /* for verbose logs */
387 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
389 aux->ctx_field_size = size;
392 struct bpf_prog_ops {
393 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
394 union bpf_attr __user *uattr);
397 struct bpf_verifier_ops {
398 /* return eBPF function prototype for verification */
399 const struct bpf_func_proto *
400 (*get_func_proto)(enum bpf_func_id func_id,
401 const struct bpf_prog *prog);
403 /* return true if 'size' wide access at offset 'off' within bpf_context
404 * with 'type' (read or write) is allowed
406 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
407 const struct bpf_prog *prog,
408 struct bpf_insn_access_aux *info);
409 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
410 const struct bpf_prog *prog);
411 int (*gen_ld_abs)(const struct bpf_insn *orig,
412 struct bpf_insn *insn_buf);
413 u32 (*convert_ctx_access)(enum bpf_access_type type,
414 const struct bpf_insn *src,
415 struct bpf_insn *dst,
416 struct bpf_prog *prog, u32 *target_size);
417 int (*btf_struct_access)(struct bpf_verifier_log *log,
418 const struct btf_type *t, int off, int size,
419 enum bpf_access_type atype,
423 struct bpf_prog_offload_ops {
424 /* verifier basic callbacks */
425 int (*insn_hook)(struct bpf_verifier_env *env,
426 int insn_idx, int prev_insn_idx);
427 int (*finalize)(struct bpf_verifier_env *env);
428 /* verifier optimization callbacks (called after .finalize) */
429 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
430 struct bpf_insn *insn);
431 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
432 /* program management callbacks */
433 int (*prepare)(struct bpf_prog *prog);
434 int (*translate)(struct bpf_prog *prog);
435 void (*destroy)(struct bpf_prog *prog);
438 struct bpf_prog_offload {
439 struct bpf_prog *prog;
440 struct net_device *netdev;
441 struct bpf_offload_dev *offdev;
443 struct list_head offloads;
450 enum bpf_cgroup_storage_type {
451 BPF_CGROUP_STORAGE_SHARED,
452 BPF_CGROUP_STORAGE_PERCPU,
453 __BPF_CGROUP_STORAGE_MAX
456 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
458 /* The longest tracepoint has 12 args.
459 * See include/trace/bpf_probe.h
461 #define MAX_BPF_FUNC_ARGS 12
463 struct bpf_prog_stats {
466 struct u64_stats_sync syncp;
467 } __aligned(2 * sizeof(u64));
469 struct btf_func_model {
472 u8 arg_size[MAX_BPF_FUNC_ARGS];
475 /* Restore arguments before returning from trampoline to let original function
476 * continue executing. This flag is used for fentry progs when there are no
479 #define BPF_TRAMP_F_RESTORE_REGS BIT(0)
480 /* Call original function after fentry progs, but before fexit progs.
481 * Makes sense for fentry/fexit, normal calls and indirect calls.
483 #define BPF_TRAMP_F_CALL_ORIG BIT(1)
484 /* Skip current frame and return to parent. Makes sense for fentry/fexit
485 * programs only. Should not be used with normal calls and indirect calls.
487 #define BPF_TRAMP_F_SKIP_FRAME BIT(2)
489 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
490 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2
492 #define BPF_MAX_TRAMP_PROGS 40
494 struct bpf_tramp_progs {
495 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS];
499 /* Different use cases for BPF trampoline:
500 * 1. replace nop at the function entry (kprobe equivalent)
501 * flags = BPF_TRAMP_F_RESTORE_REGS
502 * fentry = a set of programs to run before returning from trampoline
504 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
505 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
506 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
507 * fentry = a set of program to run before calling original function
508 * fexit = a set of program to run after original function
510 * 3. replace direct call instruction anywhere in the function body
511 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
513 * fentry = a set of programs to run before returning from trampoline
514 * With flags = BPF_TRAMP_F_CALL_ORIG
515 * orig_call = original callback addr or direct function addr
516 * fentry = a set of program to run before calling original function
517 * fexit = a set of program to run after original function
519 int arch_prepare_bpf_trampoline(void *image, void *image_end,
520 const struct btf_func_model *m, u32 flags,
521 struct bpf_tramp_progs *tprogs,
523 /* these two functions are called from generated trampoline */
524 u64 notrace __bpf_prog_enter(void);
525 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start);
530 char name[KSYM_NAME_LEN];
531 struct list_head lnode;
532 struct latch_tree_node tnode;
536 enum bpf_tramp_prog_type {
539 BPF_TRAMP_MODIFY_RETURN,
541 BPF_TRAMP_REPLACE, /* more than MAX */
544 struct bpf_trampoline {
545 /* hlist for trampoline_table */
546 struct hlist_node hlist;
547 /* serializes access to fields of this trampoline */
552 struct btf_func_model model;
556 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
557 * program by replacing one of its functions. func.addr is the address
558 * of the function it replaced.
560 struct bpf_prog *extension_prog;
561 /* list of BPF programs using this trampoline */
562 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
563 /* Number of attached programs. A counter per kind. */
564 int progs_cnt[BPF_TRAMP_MAX];
565 /* Executable image of trampoline */
568 struct bpf_ksym ksym;
571 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
573 struct bpf_dispatcher_prog {
574 struct bpf_prog *prog;
578 struct bpf_dispatcher {
579 /* dispatcher mutex */
582 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
586 struct bpf_ksym ksym;
589 static __always_inline unsigned int bpf_dispatcher_nop_func(
591 const struct bpf_insn *insnsi,
592 unsigned int (*bpf_func)(const void *,
593 const struct bpf_insn *))
595 return bpf_func(ctx, insnsi);
597 #ifdef CONFIG_BPF_JIT
598 struct bpf_trampoline *bpf_trampoline_lookup(u64 key);
599 int bpf_trampoline_link_prog(struct bpf_prog *prog);
600 int bpf_trampoline_unlink_prog(struct bpf_prog *prog);
601 void bpf_trampoline_put(struct bpf_trampoline *tr);
602 #define BPF_DISPATCHER_INIT(_name) { \
603 .mutex = __MUTEX_INITIALIZER(_name.mutex), \
604 .func = &_name##_func, \
611 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
615 #define DEFINE_BPF_DISPATCHER(name) \
616 noinline unsigned int bpf_dispatcher_##name##_func( \
618 const struct bpf_insn *insnsi, \
619 unsigned int (*bpf_func)(const void *, \
620 const struct bpf_insn *)) \
622 return bpf_func(ctx, insnsi); \
624 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
625 struct bpf_dispatcher bpf_dispatcher_##name = \
626 BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
627 #define DECLARE_BPF_DISPATCHER(name) \
628 unsigned int bpf_dispatcher_##name##_func( \
630 const struct bpf_insn *insnsi, \
631 unsigned int (*bpf_func)(const void *, \
632 const struct bpf_insn *)); \
633 extern struct bpf_dispatcher bpf_dispatcher_##name;
634 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
635 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
636 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
637 struct bpf_prog *to);
638 /* Called only from JIT-enabled code, so there's no need for stubs. */
639 void *bpf_jit_alloc_exec_page(void);
640 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym);
641 void bpf_image_ksym_del(struct bpf_ksym *ksym);
642 void bpf_ksym_add(struct bpf_ksym *ksym);
643 void bpf_ksym_del(struct bpf_ksym *ksym);
645 static inline struct bpf_trampoline *bpf_trampoline_lookup(u64 key)
649 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog)
653 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog)
657 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
658 #define DEFINE_BPF_DISPATCHER(name)
659 #define DECLARE_BPF_DISPATCHER(name)
660 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
661 #define BPF_DISPATCHER_PTR(name) NULL
662 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
663 struct bpf_prog *from,
664 struct bpf_prog *to) {}
665 static inline bool is_bpf_image_address(unsigned long address)
671 struct bpf_func_info_aux {
676 enum bpf_jit_poke_reason {
677 BPF_POKE_REASON_TAIL_CALL,
680 /* Descriptor of pokes pointing /into/ the JITed image. */
681 struct bpf_jit_poke_descriptor {
694 /* reg_type info for ctx arguments */
695 struct bpf_ctx_arg_aux {
697 enum bpf_reg_type reg_type;
701 struct bpf_prog_aux {
709 u32 func_cnt; /* used by non-func prog as the number of func progs */
710 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
711 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
712 u32 ctx_arg_info_size;
713 u32 max_rdonly_access;
715 const struct bpf_ctx_arg_aux *ctx_arg_info;
716 struct bpf_prog *linked_prog;
717 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
718 bool offload_requested;
719 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
720 bool func_proto_unreliable;
721 enum bpf_tramp_prog_type trampoline_prog_type;
722 struct bpf_trampoline *trampoline;
723 struct hlist_node tramp_hlist;
724 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
725 const struct btf_type *attach_func_proto;
726 /* function name for valid attach_btf_id */
727 const char *attach_func_name;
728 struct bpf_prog **func;
729 void *jit_data; /* JIT specific data. arch dependent */
730 struct bpf_jit_poke_descriptor *poke_tab;
732 struct bpf_ksym ksym;
733 const struct bpf_prog_ops *ops;
734 struct bpf_map **used_maps;
735 struct bpf_prog *prog;
736 struct user_struct *user;
737 u64 load_time; /* ns since boottime */
738 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
739 char name[BPF_OBJ_NAME_LEN];
740 #ifdef CONFIG_SECURITY
743 struct bpf_prog_offload *offload;
745 struct bpf_func_info *func_info;
746 struct bpf_func_info_aux *func_info_aux;
747 /* bpf_line_info loaded from userspace. linfo->insn_off
748 * has the xlated insn offset.
749 * Both the main and sub prog share the same linfo.
750 * The subprog can access its first linfo by
751 * using the linfo_idx.
753 struct bpf_line_info *linfo;
754 /* jited_linfo is the jited addr of the linfo. It has a
755 * one to one mapping to linfo:
756 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
757 * Both the main and sub prog share the same jited_linfo.
758 * The subprog can access its first jited_linfo by
759 * using the linfo_idx.
764 /* subprog can use linfo_idx to access its first linfo and
766 * main prog always has linfo_idx == 0
770 struct exception_table_entry *extable;
771 struct bpf_prog_stats __percpu *stats;
773 struct work_struct work;
778 struct bpf_array_aux {
779 /* 'Ownership' of prog array is claimed by the first program that
780 * is going to use this map or by the first program which FD is
781 * stored in the map to make sure that all callers and callees have
782 * the same prog type and JITed flag.
784 enum bpf_prog_type type;
786 /* Programs with direct jumps into programs part of this array. */
787 struct list_head poke_progs;
789 struct mutex poke_mutex;
790 struct work_struct work;
796 enum bpf_link_type type;
797 const struct bpf_link_ops *ops;
798 struct bpf_prog *prog;
799 struct work_struct work;
802 struct bpf_link_ops {
803 void (*release)(struct bpf_link *link);
804 void (*dealloc)(struct bpf_link *link);
805 int (*detach)(struct bpf_link *link);
806 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
807 struct bpf_prog *old_prog);
808 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
809 int (*fill_link_info)(const struct bpf_link *link,
810 struct bpf_link_info *info);
813 struct bpf_link_primer {
814 struct bpf_link *link;
820 struct bpf_struct_ops_value;
824 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
825 struct bpf_struct_ops {
826 const struct bpf_verifier_ops *verifier_ops;
827 int (*init)(struct btf *btf);
828 int (*check_member)(const struct btf_type *t,
829 const struct btf_member *member);
830 int (*init_member)(const struct btf_type *t,
831 const struct btf_member *member,
832 void *kdata, const void *udata);
833 int (*reg)(void *kdata);
834 void (*unreg)(void *kdata);
835 const struct btf_type *type;
836 const struct btf_type *value_type;
838 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
843 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
844 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
845 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id);
846 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log);
847 bool bpf_struct_ops_get(const void *kdata);
848 void bpf_struct_ops_put(const void *kdata);
849 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
851 static inline bool bpf_try_module_get(const void *data, struct module *owner)
853 if (owner == BPF_MODULE_OWNER)
854 return bpf_struct_ops_get(data);
856 return try_module_get(owner);
858 static inline void bpf_module_put(const void *data, struct module *owner)
860 if (owner == BPF_MODULE_OWNER)
861 bpf_struct_ops_put(data);
866 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id)
870 static inline void bpf_struct_ops_init(struct btf *btf,
871 struct bpf_verifier_log *log)
874 static inline bool bpf_try_module_get(const void *data, struct module *owner)
876 return try_module_get(owner);
878 static inline void bpf_module_put(const void *data, struct module *owner)
882 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
894 struct bpf_array_aux *aux;
896 char value[0] __aligned(8);
897 void *ptrs[0] __aligned(8);
898 void __percpu *pptrs[0] __aligned(8);
902 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
903 #define MAX_TAIL_CALL_CNT 32
905 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
906 BPF_F_RDONLY_PROG | \
910 #define BPF_MAP_CAN_READ BIT(0)
911 #define BPF_MAP_CAN_WRITE BIT(1)
913 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
915 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
917 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
920 if (access_flags & BPF_F_RDONLY_PROG)
921 return BPF_MAP_CAN_READ;
922 else if (access_flags & BPF_F_WRONLY_PROG)
923 return BPF_MAP_CAN_WRITE;
925 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
928 static inline bool bpf_map_flags_access_ok(u32 access_flags)
930 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
931 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
934 struct bpf_event_entry {
935 struct perf_event *event;
936 struct file *perf_file;
937 struct file *map_file;
941 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
942 int bpf_prog_calc_tag(struct bpf_prog *fp);
943 const char *kernel_type_name(u32 btf_type_id);
945 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
947 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
948 unsigned long off, unsigned long len);
949 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
950 const struct bpf_insn *src,
951 struct bpf_insn *dst,
952 struct bpf_prog *prog,
955 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
956 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
958 /* an array of programs to be executed under rcu_lock.
961 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
963 * the structure returned by bpf_prog_array_alloc() should be populated
964 * with program pointers and the last pointer must be NULL.
965 * The user has to keep refcnt on the program and make sure the program
966 * is removed from the array before bpf_prog_put().
967 * The 'struct bpf_prog_array *' should only be replaced with xchg()
968 * since other cpus are walking the array of pointers in parallel.
970 struct bpf_prog_array_item {
971 struct bpf_prog *prog;
972 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
975 struct bpf_prog_array {
977 struct bpf_prog_array_item items[];
980 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
981 void bpf_prog_array_free(struct bpf_prog_array *progs);
982 int bpf_prog_array_length(struct bpf_prog_array *progs);
983 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
984 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
985 __u32 __user *prog_ids, u32 cnt);
987 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
988 struct bpf_prog *old_prog);
989 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
990 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
991 struct bpf_prog *prog);
992 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
993 u32 *prog_ids, u32 request_cnt,
995 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
996 struct bpf_prog *exclude_prog,
997 struct bpf_prog *include_prog,
998 struct bpf_prog_array **new_array);
1000 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
1002 struct bpf_prog_array_item *_item; \
1003 struct bpf_prog *_prog; \
1004 struct bpf_prog_array *_array; \
1006 migrate_disable(); \
1008 _array = rcu_dereference(array); \
1009 if (unlikely(check_non_null && !_array))\
1011 _item = &_array->items[0]; \
1012 while ((_prog = READ_ONCE(_item->prog))) { \
1013 bpf_cgroup_storage_set(_item->cgroup_storage); \
1014 _ret &= func(_prog, ctx); \
1018 rcu_read_unlock(); \
1023 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs
1024 * so BPF programs can request cwr for TCP packets.
1026 * Current cgroup skb programs can only return 0 or 1 (0 to drop the
1027 * packet. This macro changes the behavior so the low order bit
1028 * indicates whether the packet should be dropped (0) or not (1)
1029 * and the next bit is a congestion notification bit. This could be
1030 * used by TCP to call tcp_enter_cwr()
1032 * Hence, new allowed return values of CGROUP EGRESS BPF programs are:
1035 * 2: drop packet and cn
1036 * 3: keep packet and cn
1038 * This macro then converts it to one of the NET_XMIT or an error
1039 * code that is then interpreted as drop packet (and no cn):
1040 * 0: NET_XMIT_SUCCESS skb should be transmitted
1041 * 1: NET_XMIT_DROP skb should be dropped and cn
1042 * 2: NET_XMIT_CN skb should be transmitted and cn
1043 * 3: -EPERM skb should be dropped
1045 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \
1047 struct bpf_prog_array_item *_item; \
1048 struct bpf_prog *_prog; \
1049 struct bpf_prog_array *_array; \
1053 migrate_disable(); \
1055 _array = rcu_dereference(array); \
1056 _item = &_array->items[0]; \
1057 while ((_prog = READ_ONCE(_item->prog))) { \
1058 bpf_cgroup_storage_set(_item->cgroup_storage); \
1059 ret = func(_prog, ctx); \
1060 _ret &= (ret & 1); \
1064 rcu_read_unlock(); \
1067 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \
1069 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \
1073 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
1074 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
1076 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
1077 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
1079 #ifdef CONFIG_BPF_SYSCALL
1080 DECLARE_PER_CPU(int, bpf_prog_active);
1081 extern struct mutex bpf_stats_enabled_mutex;
1084 * Block execution of BPF programs attached to instrumentation (perf,
1085 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
1086 * these events can happen inside a region which holds a map bucket lock
1087 * and can deadlock on it.
1089 * Use the preemption safe inc/dec variants on RT because migrate disable
1090 * is preemptible on RT and preemption in the middle of the RMW operation
1091 * might lead to inconsistent state. Use the raw variants for non RT
1092 * kernels as migrate_disable() maps to preempt_disable() so the slightly
1093 * more expensive save operation can be avoided.
1095 static inline void bpf_disable_instrumentation(void)
1098 if (IS_ENABLED(CONFIG_PREEMPT_RT))
1099 this_cpu_inc(bpf_prog_active);
1101 __this_cpu_inc(bpf_prog_active);
1104 static inline void bpf_enable_instrumentation(void)
1106 if (IS_ENABLED(CONFIG_PREEMPT_RT))
1107 this_cpu_dec(bpf_prog_active);
1109 __this_cpu_dec(bpf_prog_active);
1113 extern const struct file_operations bpf_map_fops;
1114 extern const struct file_operations bpf_prog_fops;
1115 extern const struct file_operations bpf_iter_fops;
1117 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
1118 extern const struct bpf_prog_ops _name ## _prog_ops; \
1119 extern const struct bpf_verifier_ops _name ## _verifier_ops;
1120 #define BPF_MAP_TYPE(_id, _ops) \
1121 extern const struct bpf_map_ops _ops;
1122 #define BPF_LINK_TYPE(_id, _name)
1123 #include <linux/bpf_types.h>
1124 #undef BPF_PROG_TYPE
1126 #undef BPF_LINK_TYPE
1128 extern const struct bpf_prog_ops bpf_offload_prog_ops;
1129 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
1130 extern const struct bpf_verifier_ops xdp_analyzer_ops;
1132 struct bpf_prog *bpf_prog_get(u32 ufd);
1133 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
1135 void bpf_prog_add(struct bpf_prog *prog, int i);
1136 void bpf_prog_sub(struct bpf_prog *prog, int i);
1137 void bpf_prog_inc(struct bpf_prog *prog);
1138 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
1139 void bpf_prog_put(struct bpf_prog *prog);
1140 int __bpf_prog_charge(struct user_struct *user, u32 pages);
1141 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
1142 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
1143 struct bpf_map **used_maps, u32 len);
1145 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
1146 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
1148 struct bpf_map *bpf_map_get(u32 ufd);
1149 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
1150 struct bpf_map *__bpf_map_get(struct fd f);
1151 void bpf_map_inc(struct bpf_map *map);
1152 void bpf_map_inc_with_uref(struct bpf_map *map);
1153 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
1154 void bpf_map_put_with_uref(struct bpf_map *map);
1155 void bpf_map_put(struct bpf_map *map);
1156 int bpf_map_charge_memlock(struct bpf_map *map, u32 pages);
1157 void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages);
1158 int bpf_map_charge_init(struct bpf_map_memory *mem, u64 size);
1159 void bpf_map_charge_finish(struct bpf_map_memory *mem);
1160 void bpf_map_charge_move(struct bpf_map_memory *dst,
1161 struct bpf_map_memory *src);
1162 void *bpf_map_area_alloc(u64 size, int numa_node);
1163 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
1164 void bpf_map_area_free(void *base);
1165 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
1166 int generic_map_lookup_batch(struct bpf_map *map,
1167 const union bpf_attr *attr,
1168 union bpf_attr __user *uattr);
1169 int generic_map_update_batch(struct bpf_map *map,
1170 const union bpf_attr *attr,
1171 union bpf_attr __user *uattr);
1172 int generic_map_delete_batch(struct bpf_map *map,
1173 const union bpf_attr *attr,
1174 union bpf_attr __user *uattr);
1175 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
1176 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
1178 extern int sysctl_unprivileged_bpf_disabled;
1180 static inline bool bpf_allow_ptr_leaks(void)
1182 return perfmon_capable();
1185 static inline bool bpf_allow_ptr_to_map_access(void)
1187 return perfmon_capable();
1190 static inline bool bpf_bypass_spec_v1(void)
1192 return perfmon_capable();
1195 static inline bool bpf_bypass_spec_v4(void)
1197 return perfmon_capable();
1200 int bpf_map_new_fd(struct bpf_map *map, int flags);
1201 int bpf_prog_new_fd(struct bpf_prog *prog);
1203 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1204 const struct bpf_link_ops *ops, struct bpf_prog *prog);
1205 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
1206 int bpf_link_settle(struct bpf_link_primer *primer);
1207 void bpf_link_cleanup(struct bpf_link_primer *primer);
1208 void bpf_link_inc(struct bpf_link *link);
1209 void bpf_link_put(struct bpf_link *link);
1210 int bpf_link_new_fd(struct bpf_link *link);
1211 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd);
1212 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
1214 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
1215 int bpf_obj_get_user(const char __user *pathname, int flags);
1217 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
1218 #define DEFINE_BPF_ITER_FUNC(target, args...) \
1219 extern int bpf_iter_ ## target(args); \
1220 int __init bpf_iter_ ## target(args) { return 0; }
1222 struct bpf_iter_aux_info {
1223 struct bpf_map *map;
1226 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
1227 union bpf_iter_link_info *linfo,
1228 struct bpf_iter_aux_info *aux);
1229 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
1230 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
1231 struct seq_file *seq);
1232 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
1233 struct bpf_link_info *info);
1235 #define BPF_ITER_CTX_ARG_MAX 2
1236 struct bpf_iter_reg {
1238 bpf_iter_attach_target_t attach_target;
1239 bpf_iter_detach_target_t detach_target;
1240 bpf_iter_show_fdinfo_t show_fdinfo;
1241 bpf_iter_fill_link_info_t fill_link_info;
1242 u32 ctx_arg_info_size;
1243 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
1244 const struct bpf_iter_seq_info *seq_info;
1247 struct bpf_iter_meta {
1248 __bpf_md_ptr(struct seq_file *, seq);
1253 struct bpf_iter__bpf_map_elem {
1254 __bpf_md_ptr(struct bpf_iter_meta *, meta);
1255 __bpf_md_ptr(struct bpf_map *, map);
1256 __bpf_md_ptr(void *, key);
1257 __bpf_md_ptr(void *, value);
1260 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
1261 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
1262 bool bpf_iter_prog_supported(struct bpf_prog *prog);
1263 int bpf_iter_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
1264 int bpf_iter_new_fd(struct bpf_link *link);
1265 bool bpf_link_is_iter(struct bpf_link *link);
1266 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
1267 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
1268 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
1269 struct seq_file *seq);
1270 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
1271 struct bpf_link_info *info);
1273 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
1274 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
1275 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
1277 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
1280 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
1282 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
1283 void *key, void *value, u64 map_flags);
1284 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1285 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
1286 void *key, void *value, u64 map_flags);
1287 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1289 int bpf_get_file_flag(int flags);
1290 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size,
1291 size_t actual_size);
1293 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
1294 * forced to use 'long' read/writes to try to atomically copy long counters.
1295 * Best-effort only. No barriers here, since it _will_ race with concurrent
1296 * updates from BPF programs. Called from bpf syscall and mostly used with
1297 * size 8 or 16 bytes, so ask compiler to inline it.
1299 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
1301 const long *lsrc = src;
1304 size /= sizeof(long);
1309 /* verify correctness of eBPF program */
1310 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr,
1311 union bpf_attr __user *uattr);
1312 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
1318 struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
1319 struct bpf_dtab_netdev *__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key);
1320 void __dev_flush(void);
1321 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp,
1322 struct net_device *dev_rx);
1323 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
1324 struct net_device *dev_rx);
1325 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
1326 struct bpf_prog *xdp_prog);
1327 bool dev_map_can_have_prog(struct bpf_map *map);
1329 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
1330 void __cpu_map_flush(void);
1331 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
1332 struct net_device *dev_rx);
1333 bool cpu_map_prog_allowed(struct bpf_map *map);
1335 /* Return map's numa specified by userspace */
1336 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
1338 return (attr->map_flags & BPF_F_NUMA_NODE) ?
1339 attr->numa_node : NUMA_NO_NODE;
1342 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
1343 int array_map_alloc_check(union bpf_attr *attr);
1345 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
1346 union bpf_attr __user *uattr);
1347 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
1348 union bpf_attr __user *uattr);
1349 int bpf_prog_test_run_tracing(struct bpf_prog *prog,
1350 const union bpf_attr *kattr,
1351 union bpf_attr __user *uattr);
1352 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1353 const union bpf_attr *kattr,
1354 union bpf_attr __user *uattr);
1355 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
1356 const struct bpf_prog *prog,
1357 struct bpf_insn_access_aux *info);
1358 int btf_struct_access(struct bpf_verifier_log *log,
1359 const struct btf_type *t, int off, int size,
1360 enum bpf_access_type atype,
1362 bool btf_struct_ids_match(struct bpf_verifier_log *log,
1363 int off, u32 id, u32 need_type_id);
1364 int btf_resolve_helper_id(struct bpf_verifier_log *log,
1365 const struct bpf_func_proto *fn, int);
1367 int btf_distill_func_proto(struct bpf_verifier_log *log,
1369 const struct btf_type *func_proto,
1370 const char *func_name,
1371 struct btf_func_model *m);
1373 struct bpf_reg_state;
1374 int btf_check_func_arg_match(struct bpf_verifier_env *env, int subprog,
1375 struct bpf_reg_state *regs);
1376 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog,
1377 struct bpf_reg_state *reg);
1378 int btf_check_type_match(struct bpf_verifier_env *env, struct bpf_prog *prog,
1379 struct btf *btf, const struct btf_type *t);
1381 struct bpf_prog *bpf_prog_by_id(u32 id);
1382 struct bpf_link *bpf_link_by_id(u32 id);
1384 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
1385 #else /* !CONFIG_BPF_SYSCALL */
1386 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
1388 return ERR_PTR(-EOPNOTSUPP);
1391 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
1392 enum bpf_prog_type type,
1395 return ERR_PTR(-EOPNOTSUPP);
1398 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
1402 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
1406 static inline void bpf_prog_put(struct bpf_prog *prog)
1410 static inline void bpf_prog_inc(struct bpf_prog *prog)
1414 static inline struct bpf_prog *__must_check
1415 bpf_prog_inc_not_zero(struct bpf_prog *prog)
1417 return ERR_PTR(-EOPNOTSUPP);
1420 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
1425 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
1429 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1430 const struct bpf_link_ops *ops,
1431 struct bpf_prog *prog)
1435 static inline int bpf_link_prime(struct bpf_link *link,
1436 struct bpf_link_primer *primer)
1441 static inline int bpf_link_settle(struct bpf_link_primer *primer)
1446 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
1450 static inline void bpf_link_inc(struct bpf_link *link)
1454 static inline void bpf_link_put(struct bpf_link *link)
1458 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
1463 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
1469 static inline struct net_device *__dev_map_hash_lookup_elem(struct bpf_map *map,
1474 static inline bool dev_map_can_have_prog(struct bpf_map *map)
1479 static inline void __dev_flush(void)
1484 struct bpf_dtab_netdev;
1487 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp,
1488 struct net_device *dev_rx)
1494 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
1495 struct net_device *dev_rx)
1502 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
1503 struct sk_buff *skb,
1504 struct bpf_prog *xdp_prog)
1510 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
1515 static inline void __cpu_map_flush(void)
1519 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
1520 struct xdp_buff *xdp,
1521 struct net_device *dev_rx)
1526 static inline bool cpu_map_prog_allowed(struct bpf_map *map)
1531 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
1532 enum bpf_prog_type type)
1534 return ERR_PTR(-EOPNOTSUPP);
1537 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
1538 const union bpf_attr *kattr,
1539 union bpf_attr __user *uattr)
1544 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
1545 const union bpf_attr *kattr,
1546 union bpf_attr __user *uattr)
1551 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
1552 const union bpf_attr *kattr,
1553 union bpf_attr __user *uattr)
1558 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1559 const union bpf_attr *kattr,
1560 union bpf_attr __user *uattr)
1565 static inline void bpf_map_put(struct bpf_map *map)
1569 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
1571 return ERR_PTR(-ENOTSUPP);
1574 static inline const struct bpf_func_proto *
1575 bpf_base_func_proto(enum bpf_func_id func_id)
1579 #endif /* CONFIG_BPF_SYSCALL */
1581 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
1582 enum bpf_prog_type type)
1584 return bpf_prog_get_type_dev(ufd, type, false);
1587 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
1589 int bpf_prog_offload_compile(struct bpf_prog *prog);
1590 void bpf_prog_offload_destroy(struct bpf_prog *prog);
1591 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
1592 struct bpf_prog *prog);
1594 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
1596 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
1597 int bpf_map_offload_update_elem(struct bpf_map *map,
1598 void *key, void *value, u64 flags);
1599 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
1600 int bpf_map_offload_get_next_key(struct bpf_map *map,
1601 void *key, void *next_key);
1603 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
1605 struct bpf_offload_dev *
1606 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
1607 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
1608 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
1609 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
1610 struct net_device *netdev);
1611 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
1612 struct net_device *netdev);
1613 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
1615 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
1616 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
1618 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
1620 return aux->offload_requested;
1623 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1625 return unlikely(map->ops == &bpf_map_offload_ops);
1628 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
1629 void bpf_map_offload_map_free(struct bpf_map *map);
1631 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
1632 union bpf_attr *attr)
1637 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
1642 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1647 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
1649 return ERR_PTR(-EOPNOTSUPP);
1652 static inline void bpf_map_offload_map_free(struct bpf_map *map)
1655 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
1657 #if defined(CONFIG_BPF_STREAM_PARSER)
1658 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
1659 struct bpf_prog *old, u32 which);
1660 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
1661 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
1662 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
1663 void sock_map_unhash(struct sock *sk);
1664 void sock_map_close(struct sock *sk, long timeout);
1666 static inline int sock_map_prog_update(struct bpf_map *map,
1667 struct bpf_prog *prog,
1668 struct bpf_prog *old, u32 which)
1673 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
1674 struct bpf_prog *prog)
1679 static inline int sock_map_prog_detach(const union bpf_attr *attr,
1680 enum bpf_prog_type ptype)
1685 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
1690 #endif /* CONFIG_BPF_STREAM_PARSER */
1692 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
1693 void bpf_sk_reuseport_detach(struct sock *sk);
1694 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
1696 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
1697 void *value, u64 map_flags);
1699 static inline void bpf_sk_reuseport_detach(struct sock *sk)
1703 #ifdef CONFIG_BPF_SYSCALL
1704 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
1705 void *key, void *value)
1710 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
1711 void *key, void *value,
1716 #endif /* CONFIG_BPF_SYSCALL */
1717 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
1719 /* verifier prototypes for helper functions called from eBPF programs */
1720 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
1721 extern const struct bpf_func_proto bpf_map_update_elem_proto;
1722 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
1723 extern const struct bpf_func_proto bpf_map_push_elem_proto;
1724 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
1725 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
1727 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
1728 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
1729 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
1730 extern const struct bpf_func_proto bpf_tail_call_proto;
1731 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
1732 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
1733 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
1734 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
1735 extern const struct bpf_func_proto bpf_get_current_comm_proto;
1736 extern const struct bpf_func_proto bpf_get_stackid_proto;
1737 extern const struct bpf_func_proto bpf_get_stack_proto;
1738 extern const struct bpf_func_proto bpf_get_task_stack_proto;
1739 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
1740 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
1741 extern const struct bpf_func_proto bpf_sock_map_update_proto;
1742 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
1743 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
1744 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
1745 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
1746 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
1747 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
1748 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
1749 extern const struct bpf_func_proto bpf_spin_lock_proto;
1750 extern const struct bpf_func_proto bpf_spin_unlock_proto;
1751 extern const struct bpf_func_proto bpf_get_local_storage_proto;
1752 extern const struct bpf_func_proto bpf_strtol_proto;
1753 extern const struct bpf_func_proto bpf_strtoul_proto;
1754 extern const struct bpf_func_proto bpf_tcp_sock_proto;
1755 extern const struct bpf_func_proto bpf_jiffies64_proto;
1756 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
1757 extern const struct bpf_func_proto bpf_event_output_data_proto;
1758 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
1759 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
1760 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
1761 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
1762 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
1763 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
1764 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
1765 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
1766 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
1767 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
1769 const struct bpf_func_proto *bpf_tracing_func_proto(
1770 enum bpf_func_id func_id, const struct bpf_prog *prog);
1772 const struct bpf_func_proto *tracing_prog_func_proto(
1773 enum bpf_func_id func_id, const struct bpf_prog *prog);
1775 /* Shared helpers among cBPF and eBPF. */
1776 void bpf_user_rnd_init_once(void);
1777 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1778 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1780 #if defined(CONFIG_NET)
1781 bool bpf_sock_common_is_valid_access(int off, int size,
1782 enum bpf_access_type type,
1783 struct bpf_insn_access_aux *info);
1784 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1785 struct bpf_insn_access_aux *info);
1786 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
1787 const struct bpf_insn *si,
1788 struct bpf_insn *insn_buf,
1789 struct bpf_prog *prog,
1792 static inline bool bpf_sock_common_is_valid_access(int off, int size,
1793 enum bpf_access_type type,
1794 struct bpf_insn_access_aux *info)
1798 static inline bool bpf_sock_is_valid_access(int off, int size,
1799 enum bpf_access_type type,
1800 struct bpf_insn_access_aux *info)
1804 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
1805 const struct bpf_insn *si,
1806 struct bpf_insn *insn_buf,
1807 struct bpf_prog *prog,
1815 struct sk_reuseport_kern {
1816 struct sk_buff *skb;
1818 struct sock *selected_sk;
1824 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1825 struct bpf_insn_access_aux *info);
1827 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
1828 const struct bpf_insn *si,
1829 struct bpf_insn *insn_buf,
1830 struct bpf_prog *prog,
1833 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1834 struct bpf_insn_access_aux *info);
1836 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
1837 const struct bpf_insn *si,
1838 struct bpf_insn *insn_buf,
1839 struct bpf_prog *prog,
1842 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
1843 enum bpf_access_type type,
1844 struct bpf_insn_access_aux *info)
1849 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
1850 const struct bpf_insn *si,
1851 struct bpf_insn *insn_buf,
1852 struct bpf_prog *prog,
1857 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
1858 enum bpf_access_type type,
1859 struct bpf_insn_access_aux *info)
1864 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
1865 const struct bpf_insn *si,
1866 struct bpf_insn *insn_buf,
1867 struct bpf_prog *prog,
1872 #endif /* CONFIG_INET */
1874 enum bpf_text_poke_type {
1879 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
1880 void *addr1, void *addr2);
1883 bool btf_id_set_contains(struct btf_id_set *set, u32 id);
1885 #endif /* _LINUX_BPF_H */