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 int (*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);
116 /* map_meta_equal must be implemented for maps that can be
117 * used as an inner map. It is a runtime check to ensure
118 * an inner map can be inserted to an outer map.
120 * Some properties of the inner map has been used during the
121 * verification time. When inserting an inner map at the runtime,
122 * map_meta_equal has to ensure the inserting map has the same
123 * properties that the verifier has used earlier.
125 bool (*map_meta_equal)(const struct bpf_map *meta0,
126 const struct bpf_map *meta1);
128 /* BTF name and id of struct allocated by map_alloc */
129 const char * const map_btf_name;
132 /* bpf_iter info used to open a seq_file */
133 const struct bpf_iter_seq_info *iter_seq_info;
136 struct bpf_map_memory {
138 struct user_struct *user;
142 /* The first two cachelines with read-mostly members of which some
143 * are also accessed in fast-path (e.g. ops, max_entries).
145 const struct bpf_map_ops *ops ____cacheline_aligned;
146 struct bpf_map *inner_map_meta;
147 #ifdef CONFIG_SECURITY
150 enum bpf_map_type map_type;
155 int spin_lock_off; /* >=0 valid offset, <0 error */
159 u32 btf_value_type_id;
161 struct bpf_map_memory memory;
162 char name[BPF_OBJ_NAME_LEN];
163 u32 btf_vmlinux_value_type_id;
165 bool frozen; /* write-once; write-protected by freeze_mutex */
168 /* The 3rd and 4th cacheline with misc members to avoid false sharing
169 * particularly with refcounting.
171 atomic64_t refcnt ____cacheline_aligned;
173 struct work_struct work;
174 struct mutex freeze_mutex;
175 u64 writecnt; /* writable mmap cnt; protected by freeze_mutex */
178 static inline bool map_value_has_spin_lock(const struct bpf_map *map)
180 return map->spin_lock_off >= 0;
183 static inline void check_and_init_map_lock(struct bpf_map *map, void *dst)
185 if (likely(!map_value_has_spin_lock(map)))
187 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) =
188 (struct bpf_spin_lock){};
191 /* copy everything but bpf_spin_lock */
192 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
194 if (unlikely(map_value_has_spin_lock(map))) {
195 u32 off = map->spin_lock_off;
197 memcpy(dst, src, off);
198 memcpy(dst + off + sizeof(struct bpf_spin_lock),
199 src + off + sizeof(struct bpf_spin_lock),
200 map->value_size - off - sizeof(struct bpf_spin_lock));
202 memcpy(dst, src, map->value_size);
205 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
207 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
209 struct bpf_offload_dev;
210 struct bpf_offloaded_map;
212 struct bpf_map_dev_ops {
213 int (*map_get_next_key)(struct bpf_offloaded_map *map,
214 void *key, void *next_key);
215 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
216 void *key, void *value);
217 int (*map_update_elem)(struct bpf_offloaded_map *map,
218 void *key, void *value, u64 flags);
219 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
222 struct bpf_offloaded_map {
224 struct net_device *netdev;
225 const struct bpf_map_dev_ops *dev_ops;
227 struct list_head offloads;
230 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
232 return container_of(map, struct bpf_offloaded_map, map);
235 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
237 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
240 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
242 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
243 map->ops->map_seq_show_elem;
246 int map_check_no_btf(const struct bpf_map *map,
247 const struct btf *btf,
248 const struct btf_type *key_type,
249 const struct btf_type *value_type);
251 bool bpf_map_meta_equal(const struct bpf_map *meta0,
252 const struct bpf_map *meta1);
254 extern const struct bpf_map_ops bpf_map_offload_ops;
256 /* function argument constraints */
258 ARG_DONTCARE = 0, /* unused argument in helper function */
260 /* the following constraints used to prototype
261 * bpf_map_lookup/update/delete_elem() functions
263 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
264 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
265 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
266 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */
267 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */
269 /* the following constraints used to prototype bpf_memcmp() and other
270 * functions that access data on eBPF program stack
272 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
273 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
274 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
275 * helper function must fill all bytes or clear
276 * them in error case.
279 ARG_CONST_SIZE, /* number of bytes accessed from memory */
280 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
282 ARG_PTR_TO_CTX, /* pointer to context */
283 ARG_PTR_TO_CTX_OR_NULL, /* pointer to context or NULL */
284 ARG_ANYTHING, /* any (initialized) argument is ok */
285 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
286 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
287 ARG_PTR_TO_INT, /* pointer to int */
288 ARG_PTR_TO_LONG, /* pointer to long */
289 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
290 ARG_PTR_TO_SOCKET_OR_NULL, /* pointer to bpf_sock (fullsock) or NULL */
291 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
292 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */
293 ARG_PTR_TO_ALLOC_MEM_OR_NULL, /* pointer to dynamically allocated memory or NULL */
294 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
295 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
296 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */
300 /* type of values returned from helper functions */
301 enum bpf_return_type {
302 RET_INTEGER, /* function returns integer */
303 RET_VOID, /* function doesn't return anything */
304 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
305 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
306 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */
307 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */
308 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */
309 RET_PTR_TO_ALLOC_MEM_OR_NULL, /* returns a pointer to dynamically allocated memory or NULL */
310 RET_PTR_TO_BTF_ID_OR_NULL, /* returns a pointer to a btf_id or NULL */
311 RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL, /* returns a pointer to a valid memory or a btf_id or NULL */
312 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */
315 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
316 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
317 * instructions after verifying
319 struct bpf_func_proto {
320 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
323 enum bpf_return_type ret_type;
326 enum bpf_arg_type arg1_type;
327 enum bpf_arg_type arg2_type;
328 enum bpf_arg_type arg3_type;
329 enum bpf_arg_type arg4_type;
330 enum bpf_arg_type arg5_type;
332 enum bpf_arg_type arg_type[5];
344 int *ret_btf_id; /* return value btf_id */
345 bool (*allowed)(const struct bpf_prog *prog);
348 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
349 * the first argument to eBPF programs.
350 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
354 enum bpf_access_type {
359 /* types of values stored in eBPF registers */
360 /* Pointer types represent:
363 * pointer + (u16) var
364 * pointer + (u16) var + imm
365 * if (range > 0) then [ptr, ptr + range - off) is safe to access
366 * if (id > 0) means that some 'var' was added
367 * if (off > 0) means that 'imm' was added
370 NOT_INIT = 0, /* nothing was written into register */
371 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
372 PTR_TO_CTX, /* reg points to bpf_context */
373 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
374 PTR_TO_MAP_VALUE, /* reg points to map element value */
375 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
376 PTR_TO_STACK, /* reg == frame_pointer + offset */
377 PTR_TO_PACKET_META, /* skb->data - meta_len */
378 PTR_TO_PACKET, /* reg points to skb->data */
379 PTR_TO_PACKET_END, /* skb->data + headlen */
380 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
381 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
382 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */
383 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
384 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */
385 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
386 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */
387 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
388 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
389 /* PTR_TO_BTF_ID points to a kernel struct that does not need
390 * to be null checked by the BPF program. This does not imply the
391 * pointer is _not_ null and in practice this can easily be a null
392 * pointer when reading pointer chains. The assumption is program
393 * context will handle null pointer dereference typically via fault
394 * handling. The verifier must keep this in mind and can make no
395 * assumptions about null or non-null when doing branch analysis.
396 * Further, when passed into helpers the helpers can not, without
397 * additional context, assume the value is non-null.
400 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
401 * been checked for null. Used primarily to inform the verifier
402 * an explicit null check is required for this struct.
404 PTR_TO_BTF_ID_OR_NULL,
405 PTR_TO_MEM, /* reg points to valid memory region */
406 PTR_TO_MEM_OR_NULL, /* reg points to valid memory region or NULL */
407 PTR_TO_RDONLY_BUF, /* reg points to a readonly buffer */
408 PTR_TO_RDONLY_BUF_OR_NULL, /* reg points to a readonly buffer or NULL */
409 PTR_TO_RDWR_BUF, /* reg points to a read/write buffer */
410 PTR_TO_RDWR_BUF_OR_NULL, /* reg points to a read/write buffer or NULL */
411 PTR_TO_PERCPU_BTF_ID, /* reg points to a percpu kernel variable */
414 /* The information passed from prog-specific *_is_valid_access
415 * back to the verifier.
417 struct bpf_insn_access_aux {
418 enum bpf_reg_type reg_type;
423 struct bpf_verifier_log *log; /* for verbose logs */
427 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
429 aux->ctx_field_size = size;
432 struct bpf_prog_ops {
433 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
434 union bpf_attr __user *uattr);
437 struct bpf_verifier_ops {
438 /* return eBPF function prototype for verification */
439 const struct bpf_func_proto *
440 (*get_func_proto)(enum bpf_func_id func_id,
441 const struct bpf_prog *prog);
443 /* return true if 'size' wide access at offset 'off' within bpf_context
444 * with 'type' (read or write) is allowed
446 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
447 const struct bpf_prog *prog,
448 struct bpf_insn_access_aux *info);
449 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
450 const struct bpf_prog *prog);
451 int (*gen_ld_abs)(const struct bpf_insn *orig,
452 struct bpf_insn *insn_buf);
453 u32 (*convert_ctx_access)(enum bpf_access_type type,
454 const struct bpf_insn *src,
455 struct bpf_insn *dst,
456 struct bpf_prog *prog, u32 *target_size);
457 int (*btf_struct_access)(struct bpf_verifier_log *log,
458 const struct btf_type *t, int off, int size,
459 enum bpf_access_type atype,
463 struct bpf_prog_offload_ops {
464 /* verifier basic callbacks */
465 int (*insn_hook)(struct bpf_verifier_env *env,
466 int insn_idx, int prev_insn_idx);
467 int (*finalize)(struct bpf_verifier_env *env);
468 /* verifier optimization callbacks (called after .finalize) */
469 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
470 struct bpf_insn *insn);
471 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
472 /* program management callbacks */
473 int (*prepare)(struct bpf_prog *prog);
474 int (*translate)(struct bpf_prog *prog);
475 void (*destroy)(struct bpf_prog *prog);
478 struct bpf_prog_offload {
479 struct bpf_prog *prog;
480 struct net_device *netdev;
481 struct bpf_offload_dev *offdev;
483 struct list_head offloads;
490 enum bpf_cgroup_storage_type {
491 BPF_CGROUP_STORAGE_SHARED,
492 BPF_CGROUP_STORAGE_PERCPU,
493 __BPF_CGROUP_STORAGE_MAX
496 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
498 /* The longest tracepoint has 12 args.
499 * See include/trace/bpf_probe.h
501 #define MAX_BPF_FUNC_ARGS 12
503 struct bpf_prog_stats {
506 struct u64_stats_sync syncp;
507 } __aligned(2 * sizeof(u64));
509 struct btf_func_model {
512 u8 arg_size[MAX_BPF_FUNC_ARGS];
515 /* Restore arguments before returning from trampoline to let original function
516 * continue executing. This flag is used for fentry progs when there are no
519 #define BPF_TRAMP_F_RESTORE_REGS BIT(0)
520 /* Call original function after fentry progs, but before fexit progs.
521 * Makes sense for fentry/fexit, normal calls and indirect calls.
523 #define BPF_TRAMP_F_CALL_ORIG BIT(1)
524 /* Skip current frame and return to parent. Makes sense for fentry/fexit
525 * programs only. Should not be used with normal calls and indirect calls.
527 #define BPF_TRAMP_F_SKIP_FRAME BIT(2)
529 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
530 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2
532 #define BPF_MAX_TRAMP_PROGS 40
534 struct bpf_tramp_progs {
535 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS];
539 /* Different use cases for BPF trampoline:
540 * 1. replace nop at the function entry (kprobe equivalent)
541 * flags = BPF_TRAMP_F_RESTORE_REGS
542 * fentry = a set of programs to run before returning from trampoline
544 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
545 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
546 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
547 * fentry = a set of program to run before calling original function
548 * fexit = a set of program to run after original function
550 * 3. replace direct call instruction anywhere in the function body
551 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
553 * fentry = a set of programs to run before returning from trampoline
554 * With flags = BPF_TRAMP_F_CALL_ORIG
555 * orig_call = original callback addr or direct function addr
556 * fentry = a set of program to run before calling original function
557 * fexit = a set of program to run after original function
559 int arch_prepare_bpf_trampoline(void *image, void *image_end,
560 const struct btf_func_model *m, u32 flags,
561 struct bpf_tramp_progs *tprogs,
563 /* these two functions are called from generated trampoline */
564 u64 notrace __bpf_prog_enter(void);
565 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start);
566 void notrace __bpf_prog_enter_sleepable(void);
567 void notrace __bpf_prog_exit_sleepable(void);
572 char name[KSYM_NAME_LEN];
573 struct list_head lnode;
574 struct latch_tree_node tnode;
578 enum bpf_tramp_prog_type {
581 BPF_TRAMP_MODIFY_RETURN,
583 BPF_TRAMP_REPLACE, /* more than MAX */
586 struct bpf_trampoline {
587 /* hlist for trampoline_table */
588 struct hlist_node hlist;
589 /* serializes access to fields of this trampoline */
594 struct btf_func_model model;
598 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
599 * program by replacing one of its functions. func.addr is the address
600 * of the function it replaced.
602 struct bpf_prog *extension_prog;
603 /* list of BPF programs using this trampoline */
604 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
605 /* Number of attached programs. A counter per kind. */
606 int progs_cnt[BPF_TRAMP_MAX];
607 /* Executable image of trampoline */
610 struct bpf_ksym ksym;
613 struct bpf_attach_target_info {
614 struct btf_func_model fmodel;
616 const char *tgt_name;
617 const struct btf_type *tgt_type;
620 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
622 struct bpf_dispatcher_prog {
623 struct bpf_prog *prog;
627 struct bpf_dispatcher {
628 /* dispatcher mutex */
631 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
635 struct bpf_ksym ksym;
638 static __always_inline unsigned int bpf_dispatcher_nop_func(
640 const struct bpf_insn *insnsi,
641 unsigned int (*bpf_func)(const void *,
642 const struct bpf_insn *))
644 return bpf_func(ctx, insnsi);
646 #ifdef CONFIG_BPF_JIT
647 int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr);
648 int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr);
649 struct bpf_trampoline *bpf_trampoline_get(u64 key,
650 struct bpf_attach_target_info *tgt_info);
651 void bpf_trampoline_put(struct bpf_trampoline *tr);
652 #define BPF_DISPATCHER_INIT(_name) { \
653 .mutex = __MUTEX_INITIALIZER(_name.mutex), \
654 .func = &_name##_func, \
661 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
665 #define DEFINE_BPF_DISPATCHER(name) \
666 noinline unsigned int bpf_dispatcher_##name##_func( \
668 const struct bpf_insn *insnsi, \
669 unsigned int (*bpf_func)(const void *, \
670 const struct bpf_insn *)) \
672 return bpf_func(ctx, insnsi); \
674 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
675 struct bpf_dispatcher bpf_dispatcher_##name = \
676 BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
677 #define DECLARE_BPF_DISPATCHER(name) \
678 unsigned int bpf_dispatcher_##name##_func( \
680 const struct bpf_insn *insnsi, \
681 unsigned int (*bpf_func)(const void *, \
682 const struct bpf_insn *)); \
683 extern struct bpf_dispatcher bpf_dispatcher_##name;
684 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
685 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
686 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
687 struct bpf_prog *to);
688 /* Called only from JIT-enabled code, so there's no need for stubs. */
689 void *bpf_jit_alloc_exec_page(void);
690 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym);
691 void bpf_image_ksym_del(struct bpf_ksym *ksym);
692 void bpf_ksym_add(struct bpf_ksym *ksym);
693 void bpf_ksym_del(struct bpf_ksym *ksym);
695 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog,
696 struct bpf_trampoline *tr)
700 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog,
701 struct bpf_trampoline *tr)
705 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
706 struct bpf_attach_target_info *tgt_info)
708 return ERR_PTR(-EOPNOTSUPP);
710 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
711 #define DEFINE_BPF_DISPATCHER(name)
712 #define DECLARE_BPF_DISPATCHER(name)
713 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
714 #define BPF_DISPATCHER_PTR(name) NULL
715 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
716 struct bpf_prog *from,
717 struct bpf_prog *to) {}
718 static inline bool is_bpf_image_address(unsigned long address)
724 struct bpf_func_info_aux {
729 enum bpf_jit_poke_reason {
730 BPF_POKE_REASON_TAIL_CALL,
733 /* Descriptor of pokes pointing /into/ the JITed image. */
734 struct bpf_jit_poke_descriptor {
735 void *tailcall_target;
736 void *tailcall_bypass;
744 bool tailcall_target_stable;
750 /* reg_type info for ctx arguments */
751 struct bpf_ctx_arg_aux {
753 enum bpf_reg_type reg_type;
757 struct bpf_prog_aux {
765 u32 func_cnt; /* used by non-func prog as the number of func progs */
766 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
767 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
768 u32 ctx_arg_info_size;
769 u32 max_rdonly_access;
771 const struct bpf_ctx_arg_aux *ctx_arg_info;
772 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
773 struct bpf_prog *dst_prog;
774 struct bpf_trampoline *dst_trampoline;
775 enum bpf_prog_type saved_dst_prog_type;
776 enum bpf_attach_type saved_dst_attach_type;
777 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
778 bool offload_requested;
779 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
780 bool func_proto_unreliable;
782 bool tail_call_reachable;
783 enum bpf_tramp_prog_type trampoline_prog_type;
784 struct hlist_node tramp_hlist;
785 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
786 const struct btf_type *attach_func_proto;
787 /* function name for valid attach_btf_id */
788 const char *attach_func_name;
789 struct bpf_prog **func;
790 void *jit_data; /* JIT specific data. arch dependent */
791 struct bpf_jit_poke_descriptor *poke_tab;
793 struct bpf_ksym ksym;
794 const struct bpf_prog_ops *ops;
795 struct bpf_map **used_maps;
796 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
797 struct bpf_prog *prog;
798 struct user_struct *user;
799 u64 load_time; /* ns since boottime */
800 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
801 char name[BPF_OBJ_NAME_LEN];
802 #ifdef CONFIG_SECURITY
805 struct bpf_prog_offload *offload;
807 struct bpf_func_info *func_info;
808 struct bpf_func_info_aux *func_info_aux;
809 /* bpf_line_info loaded from userspace. linfo->insn_off
810 * has the xlated insn offset.
811 * Both the main and sub prog share the same linfo.
812 * The subprog can access its first linfo by
813 * using the linfo_idx.
815 struct bpf_line_info *linfo;
816 /* jited_linfo is the jited addr of the linfo. It has a
817 * one to one mapping to linfo:
818 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
819 * Both the main and sub prog share the same jited_linfo.
820 * The subprog can access its first jited_linfo by
821 * using the linfo_idx.
826 /* subprog can use linfo_idx to access its first linfo and
828 * main prog always has linfo_idx == 0
832 struct exception_table_entry *extable;
833 struct bpf_prog_stats __percpu *stats;
835 struct work_struct work;
840 struct bpf_array_aux {
841 /* 'Ownership' of prog array is claimed by the first program that
842 * is going to use this map or by the first program which FD is
843 * stored in the map to make sure that all callers and callees have
844 * the same prog type and JITed flag.
846 enum bpf_prog_type type;
848 /* Programs with direct jumps into programs part of this array. */
849 struct list_head poke_progs;
851 struct mutex poke_mutex;
852 struct work_struct work;
858 enum bpf_link_type type;
859 const struct bpf_link_ops *ops;
860 struct bpf_prog *prog;
861 struct work_struct work;
864 struct bpf_link_ops {
865 void (*release)(struct bpf_link *link);
866 void (*dealloc)(struct bpf_link *link);
867 int (*detach)(struct bpf_link *link);
868 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
869 struct bpf_prog *old_prog);
870 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
871 int (*fill_link_info)(const struct bpf_link *link,
872 struct bpf_link_info *info);
875 struct bpf_link_primer {
876 struct bpf_link *link;
882 struct bpf_struct_ops_value;
886 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
887 struct bpf_struct_ops {
888 const struct bpf_verifier_ops *verifier_ops;
889 int (*init)(struct btf *btf);
890 int (*check_member)(const struct btf_type *t,
891 const struct btf_member *member);
892 int (*init_member)(const struct btf_type *t,
893 const struct btf_member *member,
894 void *kdata, const void *udata);
895 int (*reg)(void *kdata);
896 void (*unreg)(void *kdata);
897 const struct btf_type *type;
898 const struct btf_type *value_type;
900 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
905 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
906 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
907 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id);
908 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log);
909 bool bpf_struct_ops_get(const void *kdata);
910 void bpf_struct_ops_put(const void *kdata);
911 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
913 static inline bool bpf_try_module_get(const void *data, struct module *owner)
915 if (owner == BPF_MODULE_OWNER)
916 return bpf_struct_ops_get(data);
918 return try_module_get(owner);
920 static inline void bpf_module_put(const void *data, struct module *owner)
922 if (owner == BPF_MODULE_OWNER)
923 bpf_struct_ops_put(data);
928 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id)
932 static inline void bpf_struct_ops_init(struct btf *btf,
933 struct bpf_verifier_log *log)
936 static inline bool bpf_try_module_get(const void *data, struct module *owner)
938 return try_module_get(owner);
940 static inline void bpf_module_put(const void *data, struct module *owner)
944 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
956 struct bpf_array_aux *aux;
958 char value[0] __aligned(8);
959 void *ptrs[0] __aligned(8);
960 void __percpu *pptrs[0] __aligned(8);
964 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
965 #define MAX_TAIL_CALL_CNT 32
967 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
968 BPF_F_RDONLY_PROG | \
972 #define BPF_MAP_CAN_READ BIT(0)
973 #define BPF_MAP_CAN_WRITE BIT(1)
975 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
977 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
979 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
982 if (access_flags & BPF_F_RDONLY_PROG)
983 return BPF_MAP_CAN_READ;
984 else if (access_flags & BPF_F_WRONLY_PROG)
985 return BPF_MAP_CAN_WRITE;
987 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
990 static inline bool bpf_map_flags_access_ok(u32 access_flags)
992 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
993 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
996 struct bpf_event_entry {
997 struct perf_event *event;
998 struct file *perf_file;
999 struct file *map_file;
1000 struct rcu_head rcu;
1003 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
1004 int bpf_prog_calc_tag(struct bpf_prog *fp);
1005 const char *kernel_type_name(u32 btf_type_id);
1007 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
1009 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
1010 unsigned long off, unsigned long len);
1011 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
1012 const struct bpf_insn *src,
1013 struct bpf_insn *dst,
1014 struct bpf_prog *prog,
1017 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
1018 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
1020 /* an array of programs to be executed under rcu_lock.
1023 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
1025 * the structure returned by bpf_prog_array_alloc() should be populated
1026 * with program pointers and the last pointer must be NULL.
1027 * The user has to keep refcnt on the program and make sure the program
1028 * is removed from the array before bpf_prog_put().
1029 * The 'struct bpf_prog_array *' should only be replaced with xchg()
1030 * since other cpus are walking the array of pointers in parallel.
1032 struct bpf_prog_array_item {
1033 struct bpf_prog *prog;
1034 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1037 struct bpf_prog_array {
1038 struct rcu_head rcu;
1039 struct bpf_prog_array_item items[];
1042 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
1043 void bpf_prog_array_free(struct bpf_prog_array *progs);
1044 int bpf_prog_array_length(struct bpf_prog_array *progs);
1045 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
1046 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
1047 __u32 __user *prog_ids, u32 cnt);
1049 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
1050 struct bpf_prog *old_prog);
1051 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
1052 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
1053 struct bpf_prog *prog);
1054 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
1055 u32 *prog_ids, u32 request_cnt,
1057 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
1058 struct bpf_prog *exclude_prog,
1059 struct bpf_prog *include_prog,
1060 struct bpf_prog_array **new_array);
1062 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
1064 struct bpf_prog_array_item *_item; \
1065 struct bpf_prog *_prog; \
1066 struct bpf_prog_array *_array; \
1068 migrate_disable(); \
1070 _array = rcu_dereference(array); \
1071 if (unlikely(check_non_null && !_array))\
1073 _item = &_array->items[0]; \
1074 while ((_prog = READ_ONCE(_item->prog))) { \
1075 bpf_cgroup_storage_set(_item->cgroup_storage); \
1076 _ret &= func(_prog, ctx); \
1080 rcu_read_unlock(); \
1085 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs
1086 * so BPF programs can request cwr for TCP packets.
1088 * Current cgroup skb programs can only return 0 or 1 (0 to drop the
1089 * packet. This macro changes the behavior so the low order bit
1090 * indicates whether the packet should be dropped (0) or not (1)
1091 * and the next bit is a congestion notification bit. This could be
1092 * used by TCP to call tcp_enter_cwr()
1094 * Hence, new allowed return values of CGROUP EGRESS BPF programs are:
1097 * 2: drop packet and cn
1098 * 3: keep packet and cn
1100 * This macro then converts it to one of the NET_XMIT or an error
1101 * code that is then interpreted as drop packet (and no cn):
1102 * 0: NET_XMIT_SUCCESS skb should be transmitted
1103 * 1: NET_XMIT_DROP skb should be dropped and cn
1104 * 2: NET_XMIT_CN skb should be transmitted and cn
1105 * 3: -EPERM skb should be dropped
1107 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \
1109 struct bpf_prog_array_item *_item; \
1110 struct bpf_prog *_prog; \
1111 struct bpf_prog_array *_array; \
1115 migrate_disable(); \
1117 _array = rcu_dereference(array); \
1118 _item = &_array->items[0]; \
1119 while ((_prog = READ_ONCE(_item->prog))) { \
1120 bpf_cgroup_storage_set(_item->cgroup_storage); \
1121 ret = func(_prog, ctx); \
1122 _ret &= (ret & 1); \
1126 rcu_read_unlock(); \
1129 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \
1131 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \
1135 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
1136 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
1138 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
1139 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
1141 #ifdef CONFIG_BPF_SYSCALL
1142 DECLARE_PER_CPU(int, bpf_prog_active);
1143 extern struct mutex bpf_stats_enabled_mutex;
1146 * Block execution of BPF programs attached to instrumentation (perf,
1147 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
1148 * these events can happen inside a region which holds a map bucket lock
1149 * and can deadlock on it.
1151 * Use the preemption safe inc/dec variants on RT because migrate disable
1152 * is preemptible on RT and preemption in the middle of the RMW operation
1153 * might lead to inconsistent state. Use the raw variants for non RT
1154 * kernels as migrate_disable() maps to preempt_disable() so the slightly
1155 * more expensive save operation can be avoided.
1157 static inline void bpf_disable_instrumentation(void)
1160 if (IS_ENABLED(CONFIG_PREEMPT_RT))
1161 this_cpu_inc(bpf_prog_active);
1163 __this_cpu_inc(bpf_prog_active);
1166 static inline void bpf_enable_instrumentation(void)
1168 if (IS_ENABLED(CONFIG_PREEMPT_RT))
1169 this_cpu_dec(bpf_prog_active);
1171 __this_cpu_dec(bpf_prog_active);
1175 extern const struct file_operations bpf_map_fops;
1176 extern const struct file_operations bpf_prog_fops;
1177 extern const struct file_operations bpf_iter_fops;
1179 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
1180 extern const struct bpf_prog_ops _name ## _prog_ops; \
1181 extern const struct bpf_verifier_ops _name ## _verifier_ops;
1182 #define BPF_MAP_TYPE(_id, _ops) \
1183 extern const struct bpf_map_ops _ops;
1184 #define BPF_LINK_TYPE(_id, _name)
1185 #include <linux/bpf_types.h>
1186 #undef BPF_PROG_TYPE
1188 #undef BPF_LINK_TYPE
1190 extern const struct bpf_prog_ops bpf_offload_prog_ops;
1191 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
1192 extern const struct bpf_verifier_ops xdp_analyzer_ops;
1194 struct bpf_prog *bpf_prog_get(u32 ufd);
1195 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
1197 void bpf_prog_add(struct bpf_prog *prog, int i);
1198 void bpf_prog_sub(struct bpf_prog *prog, int i);
1199 void bpf_prog_inc(struct bpf_prog *prog);
1200 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
1201 void bpf_prog_put(struct bpf_prog *prog);
1202 int __bpf_prog_charge(struct user_struct *user, u32 pages);
1203 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
1204 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
1205 struct bpf_map **used_maps, u32 len);
1207 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
1208 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
1210 struct bpf_map *bpf_map_get(u32 ufd);
1211 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
1212 struct bpf_map *__bpf_map_get(struct fd f);
1213 void bpf_map_inc(struct bpf_map *map);
1214 void bpf_map_inc_with_uref(struct bpf_map *map);
1215 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
1216 void bpf_map_put_with_uref(struct bpf_map *map);
1217 void bpf_map_put(struct bpf_map *map);
1218 int bpf_map_charge_memlock(struct bpf_map *map, u32 pages);
1219 void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages);
1220 int bpf_map_charge_init(struct bpf_map_memory *mem, u64 size);
1221 void bpf_map_charge_finish(struct bpf_map_memory *mem);
1222 void bpf_map_charge_move(struct bpf_map_memory *dst,
1223 struct bpf_map_memory *src);
1224 void *bpf_map_area_alloc(u64 size, int numa_node);
1225 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
1226 void bpf_map_area_free(void *base);
1227 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
1228 int generic_map_lookup_batch(struct bpf_map *map,
1229 const union bpf_attr *attr,
1230 union bpf_attr __user *uattr);
1231 int generic_map_update_batch(struct bpf_map *map,
1232 const union bpf_attr *attr,
1233 union bpf_attr __user *uattr);
1234 int generic_map_delete_batch(struct bpf_map *map,
1235 const union bpf_attr *attr,
1236 union bpf_attr __user *uattr);
1237 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
1238 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
1240 extern int sysctl_unprivileged_bpf_disabled;
1242 static inline bool bpf_allow_ptr_leaks(void)
1244 return perfmon_capable();
1247 static inline bool bpf_allow_ptr_to_map_access(void)
1249 return perfmon_capable();
1252 static inline bool bpf_bypass_spec_v1(void)
1254 return perfmon_capable();
1257 static inline bool bpf_bypass_spec_v4(void)
1259 return perfmon_capable();
1262 int bpf_map_new_fd(struct bpf_map *map, int flags);
1263 int bpf_prog_new_fd(struct bpf_prog *prog);
1265 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1266 const struct bpf_link_ops *ops, struct bpf_prog *prog);
1267 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
1268 int bpf_link_settle(struct bpf_link_primer *primer);
1269 void bpf_link_cleanup(struct bpf_link_primer *primer);
1270 void bpf_link_inc(struct bpf_link *link);
1271 void bpf_link_put(struct bpf_link *link);
1272 int bpf_link_new_fd(struct bpf_link *link);
1273 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd);
1274 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
1276 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
1277 int bpf_obj_get_user(const char __user *pathname, int flags);
1279 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
1280 #define DEFINE_BPF_ITER_FUNC(target, args...) \
1281 extern int bpf_iter_ ## target(args); \
1282 int __init bpf_iter_ ## target(args) { return 0; }
1284 struct bpf_iter_aux_info {
1285 struct bpf_map *map;
1288 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
1289 union bpf_iter_link_info *linfo,
1290 struct bpf_iter_aux_info *aux);
1291 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
1292 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
1293 struct seq_file *seq);
1294 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
1295 struct bpf_link_info *info);
1297 #define BPF_ITER_CTX_ARG_MAX 2
1298 struct bpf_iter_reg {
1300 bpf_iter_attach_target_t attach_target;
1301 bpf_iter_detach_target_t detach_target;
1302 bpf_iter_show_fdinfo_t show_fdinfo;
1303 bpf_iter_fill_link_info_t fill_link_info;
1304 u32 ctx_arg_info_size;
1305 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
1306 const struct bpf_iter_seq_info *seq_info;
1309 struct bpf_iter_meta {
1310 __bpf_md_ptr(struct seq_file *, seq);
1315 struct bpf_iter__bpf_map_elem {
1316 __bpf_md_ptr(struct bpf_iter_meta *, meta);
1317 __bpf_md_ptr(struct bpf_map *, map);
1318 __bpf_md_ptr(void *, key);
1319 __bpf_md_ptr(void *, value);
1322 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
1323 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
1324 bool bpf_iter_prog_supported(struct bpf_prog *prog);
1325 int bpf_iter_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
1326 int bpf_iter_new_fd(struct bpf_link *link);
1327 bool bpf_link_is_iter(struct bpf_link *link);
1328 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
1329 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
1330 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
1331 struct seq_file *seq);
1332 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
1333 struct bpf_link_info *info);
1335 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
1336 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
1337 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
1339 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
1342 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
1344 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
1345 void *key, void *value, u64 map_flags);
1346 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1347 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
1348 void *key, void *value, u64 map_flags);
1349 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1351 int bpf_get_file_flag(int flags);
1352 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size,
1353 size_t actual_size);
1355 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
1356 * forced to use 'long' read/writes to try to atomically copy long counters.
1357 * Best-effort only. No barriers here, since it _will_ race with concurrent
1358 * updates from BPF programs. Called from bpf syscall and mostly used with
1359 * size 8 or 16 bytes, so ask compiler to inline it.
1361 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
1363 const long *lsrc = src;
1366 size /= sizeof(long);
1371 /* verify correctness of eBPF program */
1372 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr,
1373 union bpf_attr __user *uattr);
1374 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
1376 struct btf *bpf_get_btf_vmlinux(void);
1382 struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
1383 struct bpf_dtab_netdev *__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key);
1384 void __dev_flush(void);
1385 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp,
1386 struct net_device *dev_rx);
1387 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
1388 struct net_device *dev_rx);
1389 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
1390 struct bpf_prog *xdp_prog);
1391 bool dev_map_can_have_prog(struct bpf_map *map);
1393 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
1394 void __cpu_map_flush(void);
1395 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
1396 struct net_device *dev_rx);
1397 bool cpu_map_prog_allowed(struct bpf_map *map);
1399 /* Return map's numa specified by userspace */
1400 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
1402 return (attr->map_flags & BPF_F_NUMA_NODE) ?
1403 attr->numa_node : NUMA_NO_NODE;
1406 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
1407 int array_map_alloc_check(union bpf_attr *attr);
1409 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
1410 union bpf_attr __user *uattr);
1411 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
1412 union bpf_attr __user *uattr);
1413 int bpf_prog_test_run_tracing(struct bpf_prog *prog,
1414 const union bpf_attr *kattr,
1415 union bpf_attr __user *uattr);
1416 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1417 const union bpf_attr *kattr,
1418 union bpf_attr __user *uattr);
1419 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
1420 const union bpf_attr *kattr,
1421 union bpf_attr __user *uattr);
1422 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
1423 const struct bpf_prog *prog,
1424 struct bpf_insn_access_aux *info);
1425 int btf_struct_access(struct bpf_verifier_log *log,
1426 const struct btf_type *t, int off, int size,
1427 enum bpf_access_type atype,
1429 bool btf_struct_ids_match(struct bpf_verifier_log *log,
1430 int off, u32 id, u32 need_type_id);
1432 int btf_distill_func_proto(struct bpf_verifier_log *log,
1434 const struct btf_type *func_proto,
1435 const char *func_name,
1436 struct btf_func_model *m);
1438 struct bpf_reg_state;
1439 int btf_check_func_arg_match(struct bpf_verifier_env *env, int subprog,
1440 struct bpf_reg_state *regs);
1441 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog,
1442 struct bpf_reg_state *reg);
1443 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
1444 struct btf *btf, const struct btf_type *t);
1446 struct bpf_prog *bpf_prog_by_id(u32 id);
1447 struct bpf_link *bpf_link_by_id(u32 id);
1449 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
1450 #else /* !CONFIG_BPF_SYSCALL */
1451 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
1453 return ERR_PTR(-EOPNOTSUPP);
1456 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
1457 enum bpf_prog_type type,
1460 return ERR_PTR(-EOPNOTSUPP);
1463 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
1467 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
1471 static inline void bpf_prog_put(struct bpf_prog *prog)
1475 static inline void bpf_prog_inc(struct bpf_prog *prog)
1479 static inline struct bpf_prog *__must_check
1480 bpf_prog_inc_not_zero(struct bpf_prog *prog)
1482 return ERR_PTR(-EOPNOTSUPP);
1485 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
1490 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
1494 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1495 const struct bpf_link_ops *ops,
1496 struct bpf_prog *prog)
1500 static inline int bpf_link_prime(struct bpf_link *link,
1501 struct bpf_link_primer *primer)
1506 static inline int bpf_link_settle(struct bpf_link_primer *primer)
1511 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
1515 static inline void bpf_link_inc(struct bpf_link *link)
1519 static inline void bpf_link_put(struct bpf_link *link)
1523 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
1528 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
1534 static inline struct net_device *__dev_map_hash_lookup_elem(struct bpf_map *map,
1539 static inline bool dev_map_can_have_prog(struct bpf_map *map)
1544 static inline void __dev_flush(void)
1549 struct bpf_dtab_netdev;
1552 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp,
1553 struct net_device *dev_rx)
1559 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
1560 struct net_device *dev_rx)
1567 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
1568 struct sk_buff *skb,
1569 struct bpf_prog *xdp_prog)
1575 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
1580 static inline void __cpu_map_flush(void)
1584 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
1585 struct xdp_buff *xdp,
1586 struct net_device *dev_rx)
1591 static inline bool cpu_map_prog_allowed(struct bpf_map *map)
1596 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
1597 enum bpf_prog_type type)
1599 return ERR_PTR(-EOPNOTSUPP);
1602 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
1603 const union bpf_attr *kattr,
1604 union bpf_attr __user *uattr)
1609 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
1610 const union bpf_attr *kattr,
1611 union bpf_attr __user *uattr)
1616 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
1617 const union bpf_attr *kattr,
1618 union bpf_attr __user *uattr)
1623 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1624 const union bpf_attr *kattr,
1625 union bpf_attr __user *uattr)
1630 static inline void bpf_map_put(struct bpf_map *map)
1634 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
1636 return ERR_PTR(-ENOTSUPP);
1639 static inline const struct bpf_func_proto *
1640 bpf_base_func_proto(enum bpf_func_id func_id)
1644 #endif /* CONFIG_BPF_SYSCALL */
1646 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
1647 enum bpf_prog_type type)
1649 return bpf_prog_get_type_dev(ufd, type, false);
1652 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
1654 int bpf_prog_offload_compile(struct bpf_prog *prog);
1655 void bpf_prog_offload_destroy(struct bpf_prog *prog);
1656 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
1657 struct bpf_prog *prog);
1659 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
1661 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
1662 int bpf_map_offload_update_elem(struct bpf_map *map,
1663 void *key, void *value, u64 flags);
1664 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
1665 int bpf_map_offload_get_next_key(struct bpf_map *map,
1666 void *key, void *next_key);
1668 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
1670 struct bpf_offload_dev *
1671 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
1672 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
1673 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
1674 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
1675 struct net_device *netdev);
1676 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
1677 struct net_device *netdev);
1678 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
1680 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
1681 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
1683 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
1685 return aux->offload_requested;
1688 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1690 return unlikely(map->ops == &bpf_map_offload_ops);
1693 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
1694 void bpf_map_offload_map_free(struct bpf_map *map);
1696 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
1697 union bpf_attr *attr)
1702 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
1707 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1712 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
1714 return ERR_PTR(-EOPNOTSUPP);
1717 static inline void bpf_map_offload_map_free(struct bpf_map *map)
1720 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
1722 #if defined(CONFIG_BPF_STREAM_PARSER)
1723 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
1724 struct bpf_prog *old, u32 which);
1725 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
1726 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
1727 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
1728 void sock_map_unhash(struct sock *sk);
1729 void sock_map_close(struct sock *sk, long timeout);
1731 static inline int sock_map_prog_update(struct bpf_map *map,
1732 struct bpf_prog *prog,
1733 struct bpf_prog *old, u32 which)
1738 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
1739 struct bpf_prog *prog)
1744 static inline int sock_map_prog_detach(const union bpf_attr *attr,
1745 enum bpf_prog_type ptype)
1750 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
1755 #endif /* CONFIG_BPF_STREAM_PARSER */
1757 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
1758 void bpf_sk_reuseport_detach(struct sock *sk);
1759 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
1761 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
1762 void *value, u64 map_flags);
1764 static inline void bpf_sk_reuseport_detach(struct sock *sk)
1768 #ifdef CONFIG_BPF_SYSCALL
1769 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
1770 void *key, void *value)
1775 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
1776 void *key, void *value,
1781 #endif /* CONFIG_BPF_SYSCALL */
1782 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
1784 /* verifier prototypes for helper functions called from eBPF programs */
1785 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
1786 extern const struct bpf_func_proto bpf_map_update_elem_proto;
1787 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
1788 extern const struct bpf_func_proto bpf_map_push_elem_proto;
1789 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
1790 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
1792 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
1793 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
1794 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
1795 extern const struct bpf_func_proto bpf_tail_call_proto;
1796 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
1797 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
1798 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
1799 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
1800 extern const struct bpf_func_proto bpf_get_current_comm_proto;
1801 extern const struct bpf_func_proto bpf_get_stackid_proto;
1802 extern const struct bpf_func_proto bpf_get_stack_proto;
1803 extern const struct bpf_func_proto bpf_get_task_stack_proto;
1804 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
1805 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
1806 extern const struct bpf_func_proto bpf_sock_map_update_proto;
1807 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
1808 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
1809 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
1810 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
1811 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
1812 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
1813 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
1814 extern const struct bpf_func_proto bpf_spin_lock_proto;
1815 extern const struct bpf_func_proto bpf_spin_unlock_proto;
1816 extern const struct bpf_func_proto bpf_get_local_storage_proto;
1817 extern const struct bpf_func_proto bpf_strtol_proto;
1818 extern const struct bpf_func_proto bpf_strtoul_proto;
1819 extern const struct bpf_func_proto bpf_tcp_sock_proto;
1820 extern const struct bpf_func_proto bpf_jiffies64_proto;
1821 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
1822 extern const struct bpf_func_proto bpf_event_output_data_proto;
1823 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
1824 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
1825 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
1826 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
1827 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
1828 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
1829 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
1830 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
1831 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
1832 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
1833 extern const struct bpf_func_proto bpf_copy_from_user_proto;
1834 extern const struct bpf_func_proto bpf_snprintf_btf_proto;
1835 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
1836 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
1838 const struct bpf_func_proto *bpf_tracing_func_proto(
1839 enum bpf_func_id func_id, const struct bpf_prog *prog);
1841 const struct bpf_func_proto *tracing_prog_func_proto(
1842 enum bpf_func_id func_id, const struct bpf_prog *prog);
1844 /* Shared helpers among cBPF and eBPF. */
1845 void bpf_user_rnd_init_once(void);
1846 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1847 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1849 #if defined(CONFIG_NET)
1850 bool bpf_sock_common_is_valid_access(int off, int size,
1851 enum bpf_access_type type,
1852 struct bpf_insn_access_aux *info);
1853 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1854 struct bpf_insn_access_aux *info);
1855 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
1856 const struct bpf_insn *si,
1857 struct bpf_insn *insn_buf,
1858 struct bpf_prog *prog,
1861 static inline bool bpf_sock_common_is_valid_access(int off, int size,
1862 enum bpf_access_type type,
1863 struct bpf_insn_access_aux *info)
1867 static inline bool bpf_sock_is_valid_access(int off, int size,
1868 enum bpf_access_type type,
1869 struct bpf_insn_access_aux *info)
1873 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
1874 const struct bpf_insn *si,
1875 struct bpf_insn *insn_buf,
1876 struct bpf_prog *prog,
1884 struct sk_reuseport_kern {
1885 struct sk_buff *skb;
1887 struct sock *selected_sk;
1893 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1894 struct bpf_insn_access_aux *info);
1896 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
1897 const struct bpf_insn *si,
1898 struct bpf_insn *insn_buf,
1899 struct bpf_prog *prog,
1902 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1903 struct bpf_insn_access_aux *info);
1905 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
1906 const struct bpf_insn *si,
1907 struct bpf_insn *insn_buf,
1908 struct bpf_prog *prog,
1911 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
1912 enum bpf_access_type type,
1913 struct bpf_insn_access_aux *info)
1918 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
1919 const struct bpf_insn *si,
1920 struct bpf_insn *insn_buf,
1921 struct bpf_prog *prog,
1926 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
1927 enum bpf_access_type type,
1928 struct bpf_insn_access_aux *info)
1933 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
1934 const struct bpf_insn *si,
1935 struct bpf_insn *insn_buf,
1936 struct bpf_prog *prog,
1941 #endif /* CONFIG_INET */
1943 enum bpf_text_poke_type {
1948 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
1949 void *addr1, void *addr2);
1952 bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
1954 #endif /* _LINUX_BPF_H */