1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
7 #include <uapi/linux/bpf.h>
8 #include <uapi/linux/filter.h>
10 #include <linux/workqueue.h>
11 #include <linux/file.h>
12 #include <linux/percpu.h>
13 #include <linux/err.h>
14 #include <linux/rbtree_latch.h>
15 #include <linux/numa.h>
16 #include <linux/mm_types.h>
17 #include <linux/wait.h>
18 #include <linux/refcount.h>
19 #include <linux/mutex.h>
20 #include <linux/module.h>
21 #include <linux/kallsyms.h>
22 #include <linux/capability.h>
23 #include <linux/sched/mm.h>
24 #include <linux/slab.h>
25 #include <linux/percpu-refcount.h>
26 #include <linux/stddef.h>
27 #include <linux/bpfptr.h>
28 #include <linux/btf.h>
29 #include <linux/rcupdate_trace.h>
30 #include <linux/static_call.h>
31 #include <linux/memcontrol.h>
33 struct bpf_verifier_env;
34 struct bpf_verifier_log;
43 struct exception_table_entry;
44 struct seq_operations;
45 struct bpf_iter_aux_info;
46 struct bpf_local_storage;
47 struct bpf_local_storage_map;
51 struct bpf_func_state;
55 extern struct idr btf_idr;
56 extern spinlock_t btf_idr_lock;
57 extern struct kobject *btf_kobj;
58 extern struct bpf_mem_alloc bpf_global_ma;
59 extern bool bpf_global_ma_set;
61 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64);
62 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data,
63 struct bpf_iter_aux_info *aux);
64 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
65 typedef unsigned int (*bpf_func_t)(const void *,
66 const struct bpf_insn *);
67 struct bpf_iter_seq_info {
68 const struct seq_operations *seq_ops;
69 bpf_iter_init_seq_priv_t init_seq_private;
70 bpf_iter_fini_seq_priv_t fini_seq_private;
74 /* map is generic key/value storage optionally accessible by eBPF programs */
76 /* funcs callable from userspace (via syscall) */
77 int (*map_alloc_check)(union bpf_attr *attr);
78 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
79 void (*map_release)(struct bpf_map *map, struct file *map_file);
80 void (*map_free)(struct bpf_map *map);
81 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
82 void (*map_release_uref)(struct bpf_map *map);
83 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
84 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr,
85 union bpf_attr __user *uattr);
86 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key,
87 void *value, u64 flags);
88 int (*map_lookup_and_delete_batch)(struct bpf_map *map,
89 const union bpf_attr *attr,
90 union bpf_attr __user *uattr);
91 int (*map_update_batch)(struct bpf_map *map, struct file *map_file,
92 const union bpf_attr *attr,
93 union bpf_attr __user *uattr);
94 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr,
95 union bpf_attr __user *uattr);
97 /* funcs callable from userspace and from eBPF programs */
98 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
99 long (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
100 long (*map_delete_elem)(struct bpf_map *map, void *key);
101 long (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
102 long (*map_pop_elem)(struct bpf_map *map, void *value);
103 long (*map_peek_elem)(struct bpf_map *map, void *value);
104 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu);
106 /* funcs called by prog_array and perf_event_array map */
107 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
109 void (*map_fd_put_ptr)(void *ptr);
110 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
111 u32 (*map_fd_sys_lookup_elem)(void *ptr);
112 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
114 int (*map_check_btf)(const struct bpf_map *map,
115 const struct btf *btf,
116 const struct btf_type *key_type,
117 const struct btf_type *value_type);
119 /* Prog poke tracking helpers. */
120 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
121 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
122 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
123 struct bpf_prog *new);
125 /* Direct value access helpers. */
126 int (*map_direct_value_addr)(const struct bpf_map *map,
128 int (*map_direct_value_meta)(const struct bpf_map *map,
130 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
131 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp,
132 struct poll_table_struct *pts);
134 /* Functions called by bpf_local_storage maps */
135 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap,
136 void *owner, u32 size);
137 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap,
138 void *owner, u32 size);
139 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner);
142 long (*map_redirect)(struct bpf_map *map, u64 key, u64 flags);
144 /* map_meta_equal must be implemented for maps that can be
145 * used as an inner map. It is a runtime check to ensure
146 * an inner map can be inserted to an outer map.
148 * Some properties of the inner map has been used during the
149 * verification time. When inserting an inner map at the runtime,
150 * map_meta_equal has to ensure the inserting map has the same
151 * properties that the verifier has used earlier.
153 bool (*map_meta_equal)(const struct bpf_map *meta0,
154 const struct bpf_map *meta1);
157 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env,
158 struct bpf_func_state *caller,
159 struct bpf_func_state *callee);
160 long (*map_for_each_callback)(struct bpf_map *map,
161 bpf_callback_t callback_fn,
162 void *callback_ctx, u64 flags);
164 u64 (*map_mem_usage)(const struct bpf_map *map);
166 /* BTF id of struct allocated by map_alloc */
169 /* bpf_iter info used to open a seq_file */
170 const struct bpf_iter_seq_info *iter_seq_info;
174 /* Support at most 10 fields in a BTF type */
178 enum btf_field_type {
179 BPF_SPIN_LOCK = (1 << 0),
180 BPF_TIMER = (1 << 1),
181 BPF_KPTR_UNREF = (1 << 2),
182 BPF_KPTR_REF = (1 << 3),
183 BPF_KPTR = BPF_KPTR_UNREF | BPF_KPTR_REF,
184 BPF_LIST_HEAD = (1 << 4),
185 BPF_LIST_NODE = (1 << 5),
186 BPF_RB_ROOT = (1 << 6),
187 BPF_RB_NODE = (1 << 7),
188 BPF_GRAPH_NODE_OR_ROOT = BPF_LIST_NODE | BPF_LIST_HEAD |
189 BPF_RB_NODE | BPF_RB_ROOT,
190 BPF_REFCOUNT = (1 << 8),
193 typedef void (*btf_dtor_kfunc_t)(void *);
195 struct btf_field_kptr {
197 struct module *module;
198 /* dtor used if btf_is_kernel(btf), otherwise the type is
199 * program-allocated, dtor is NULL, and __bpf_obj_drop_impl is used
201 btf_dtor_kfunc_t dtor;
205 struct btf_field_graph_root {
209 struct btf_record *value_rec;
215 enum btf_field_type type;
217 struct btf_field_kptr kptr;
218 struct btf_field_graph_root graph_root;
228 struct btf_field fields[];
231 /* Non-opaque version of bpf_rb_node in uapi/linux/bpf.h */
232 struct bpf_rb_node_kern {
233 struct rb_node rb_node;
235 } __attribute__((aligned(8)));
237 /* Non-opaque version of bpf_list_node in uapi/linux/bpf.h */
238 struct bpf_list_node_kern {
239 struct list_head list_head;
241 } __attribute__((aligned(8)));
244 /* The first two cachelines with read-mostly members of which some
245 * are also accessed in fast-path (e.g. ops, max_entries).
247 const struct bpf_map_ops *ops ____cacheline_aligned;
248 struct bpf_map *inner_map_meta;
249 #ifdef CONFIG_SECURITY
252 enum bpf_map_type map_type;
256 u64 map_extra; /* any per-map-type extra fields */
259 struct btf_record *record;
262 u32 btf_value_type_id;
263 u32 btf_vmlinux_value_type_id;
265 #ifdef CONFIG_MEMCG_KMEM
266 struct obj_cgroup *objcg;
268 char name[BPF_OBJ_NAME_LEN];
269 /* The 3rd and 4th cacheline with misc members to avoid false sharing
270 * particularly with refcounting.
272 atomic64_t refcnt ____cacheline_aligned;
274 struct work_struct work;
275 struct mutex freeze_mutex;
277 /* 'Ownership' of program-containing map is claimed by the first program
278 * that is going to use this map or by the first program which FD is
279 * stored in the map to make sure that all callers and callees have the
280 * same prog type, JITed flag and xdp_has_frags flag.
284 enum bpf_prog_type type;
289 bool frozen; /* write-once; write-protected by freeze_mutex */
290 s64 __percpu *elem_count;
293 static inline const char *btf_field_type_name(enum btf_field_type type)
297 return "bpf_spin_lock";
304 return "bpf_list_head";
306 return "bpf_list_node";
308 return "bpf_rb_root";
310 return "bpf_rb_node";
312 return "bpf_refcount";
319 static inline u32 btf_field_type_size(enum btf_field_type type)
323 return sizeof(struct bpf_spin_lock);
325 return sizeof(struct bpf_timer);
330 return sizeof(struct bpf_list_head);
332 return sizeof(struct bpf_list_node);
334 return sizeof(struct bpf_rb_root);
336 return sizeof(struct bpf_rb_node);
338 return sizeof(struct bpf_refcount);
345 static inline u32 btf_field_type_align(enum btf_field_type type)
349 return __alignof__(struct bpf_spin_lock);
351 return __alignof__(struct bpf_timer);
354 return __alignof__(u64);
356 return __alignof__(struct bpf_list_head);
358 return __alignof__(struct bpf_list_node);
360 return __alignof__(struct bpf_rb_root);
362 return __alignof__(struct bpf_rb_node);
364 return __alignof__(struct bpf_refcount);
371 static inline void bpf_obj_init_field(const struct btf_field *field, void *addr)
373 memset(addr, 0, field->size);
375 switch (field->type) {
377 refcount_set((refcount_t *)addr, 1);
380 RB_CLEAR_NODE((struct rb_node *)addr);
384 INIT_LIST_HEAD((struct list_head *)addr);
387 /* RB_ROOT_CACHED 0-inits, no need to do anything after memset */
399 static inline bool btf_record_has_field(const struct btf_record *rec, enum btf_field_type type)
401 if (IS_ERR_OR_NULL(rec))
403 return rec->field_mask & type;
406 static inline void bpf_obj_init(const struct btf_record *rec, void *obj)
410 if (IS_ERR_OR_NULL(rec))
412 for (i = 0; i < rec->cnt; i++)
413 bpf_obj_init_field(&rec->fields[i], obj + rec->fields[i].offset);
416 /* 'dst' must be a temporary buffer and should not point to memory that is being
417 * used in parallel by a bpf program or bpf syscall, otherwise the access from
418 * the bpf program or bpf syscall may be corrupted by the reinitialization,
419 * leading to weird problems. Even 'dst' is newly-allocated from bpf memory
420 * allocator, it is still possible for 'dst' to be used in parallel by a bpf
421 * program or bpf syscall.
423 static inline void check_and_init_map_value(struct bpf_map *map, void *dst)
425 bpf_obj_init(map->record, dst);
428 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
429 * forced to use 'long' read/writes to try to atomically copy long counters.
430 * Best-effort only. No barriers here, since it _will_ race with concurrent
431 * updates from BPF programs. Called from bpf syscall and mostly used with
432 * size 8 or 16 bytes, so ask compiler to inline it.
434 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
436 const long *lsrc = src;
439 size /= sizeof(long);
441 data_race(*ldst++ = *lsrc++);
444 /* copy everything but bpf_spin_lock, bpf_timer, and kptrs. There could be one of each. */
445 static inline void bpf_obj_memcpy(struct btf_record *rec,
446 void *dst, void *src, u32 size,
452 if (IS_ERR_OR_NULL(rec)) {
454 bpf_long_memcpy(dst, src, round_up(size, 8));
456 memcpy(dst, src, size);
460 for (i = 0; i < rec->cnt; i++) {
461 u32 next_off = rec->fields[i].offset;
462 u32 sz = next_off - curr_off;
464 memcpy(dst + curr_off, src + curr_off, sz);
465 curr_off += rec->fields[i].size + sz;
467 memcpy(dst + curr_off, src + curr_off, size - curr_off);
470 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
472 bpf_obj_memcpy(map->record, dst, src, map->value_size, false);
475 static inline void copy_map_value_long(struct bpf_map *map, void *dst, void *src)
477 bpf_obj_memcpy(map->record, dst, src, map->value_size, true);
480 static inline void bpf_obj_memzero(struct btf_record *rec, void *dst, u32 size)
485 if (IS_ERR_OR_NULL(rec)) {
486 memset(dst, 0, size);
490 for (i = 0; i < rec->cnt; i++) {
491 u32 next_off = rec->fields[i].offset;
492 u32 sz = next_off - curr_off;
494 memset(dst + curr_off, 0, sz);
495 curr_off += rec->fields[i].size + sz;
497 memset(dst + curr_off, 0, size - curr_off);
500 static inline void zero_map_value(struct bpf_map *map, void *dst)
502 bpf_obj_memzero(map->record, dst, map->value_size);
505 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
507 void bpf_timer_cancel_and_free(void *timer);
508 void bpf_list_head_free(const struct btf_field *field, void *list_head,
509 struct bpf_spin_lock *spin_lock);
510 void bpf_rb_root_free(const struct btf_field *field, void *rb_root,
511 struct bpf_spin_lock *spin_lock);
514 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
516 struct bpf_offload_dev;
517 struct bpf_offloaded_map;
519 struct bpf_map_dev_ops {
520 int (*map_get_next_key)(struct bpf_offloaded_map *map,
521 void *key, void *next_key);
522 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
523 void *key, void *value);
524 int (*map_update_elem)(struct bpf_offloaded_map *map,
525 void *key, void *value, u64 flags);
526 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
529 struct bpf_offloaded_map {
531 struct net_device *netdev;
532 const struct bpf_map_dev_ops *dev_ops;
534 struct list_head offloads;
537 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
539 return container_of(map, struct bpf_offloaded_map, map);
542 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
544 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
547 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
549 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
550 map->ops->map_seq_show_elem;
553 int map_check_no_btf(const struct bpf_map *map,
554 const struct btf *btf,
555 const struct btf_type *key_type,
556 const struct btf_type *value_type);
558 bool bpf_map_meta_equal(const struct bpf_map *meta0,
559 const struct bpf_map *meta1);
561 extern const struct bpf_map_ops bpf_map_offload_ops;
563 /* bpf_type_flag contains a set of flags that are applicable to the values of
564 * arg_type, ret_type and reg_type. For example, a pointer value may be null,
565 * or a memory is read-only. We classify types into two categories: base types
566 * and extended types. Extended types are base types combined with a type flag.
568 * Currently there are no more than 32 base types in arg_type, ret_type and
571 #define BPF_BASE_TYPE_BITS 8
574 /* PTR may be NULL. */
575 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS),
577 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is
578 * compatible with both mutable and immutable memory.
580 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS),
582 /* MEM points to BPF ring buffer reservation. */
583 MEM_RINGBUF = BIT(2 + BPF_BASE_TYPE_BITS),
585 /* MEM is in user address space. */
586 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS),
588 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged
589 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In
590 * order to drop this tag, it must be passed into bpf_per_cpu_ptr()
591 * or bpf_this_cpu_ptr(), which will return the pointer corresponding
592 * to the specified cpu.
594 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS),
596 /* Indicates that the argument will be released. */
597 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS),
599 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark
600 * unreferenced and referenced kptr loaded from map value using a load
601 * instruction, so that they can only be dereferenced but not escape the
602 * BPF program into the kernel (i.e. cannot be passed as arguments to
603 * kfunc or bpf helpers).
605 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS),
607 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS),
609 /* DYNPTR points to memory local to the bpf program. */
610 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS),
612 /* DYNPTR points to a kernel-produced ringbuf record. */
613 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS),
615 /* Size is known at compile time. */
616 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS),
618 /* MEM is of an allocated object of type in program BTF. This is used to
619 * tag PTR_TO_BTF_ID allocated using bpf_obj_new.
621 MEM_ALLOC = BIT(11 + BPF_BASE_TYPE_BITS),
623 /* PTR was passed from the kernel in a trusted context, and may be
624 * passed to KF_TRUSTED_ARGS kfuncs or BPF helper functions.
625 * Confusingly, this is _not_ the opposite of PTR_UNTRUSTED above.
626 * PTR_UNTRUSTED refers to a kptr that was read directly from a map
627 * without invoking bpf_kptr_xchg(). What we really need to know is
628 * whether a pointer is safe to pass to a kfunc or BPF helper function.
629 * While PTR_UNTRUSTED pointers are unsafe to pass to kfuncs and BPF
630 * helpers, they do not cover all possible instances of unsafe
631 * pointers. For example, a pointer that was obtained from walking a
632 * struct will _not_ get the PTR_UNTRUSTED type modifier, despite the
633 * fact that it may be NULL, invalid, etc. This is due to backwards
634 * compatibility requirements, as this was the behavior that was first
635 * introduced when kptrs were added. The behavior is now considered
636 * deprecated, and PTR_UNTRUSTED will eventually be removed.
638 * PTR_TRUSTED, on the other hand, is a pointer that the kernel
639 * guarantees to be valid and safe to pass to kfuncs and BPF helpers.
640 * For example, pointers passed to tracepoint arguments are considered
641 * PTR_TRUSTED, as are pointers that are passed to struct_ops
642 * callbacks. As alluded to above, pointers that are obtained from
643 * walking PTR_TRUSTED pointers are _not_ trusted. For example, if a
644 * struct task_struct *task is PTR_TRUSTED, then accessing
645 * task->last_wakee will lose the PTR_TRUSTED modifier when it's stored
646 * in a BPF register. Similarly, pointers passed to certain programs
647 * types such as kretprobes are not guaranteed to be valid, as they may
648 * for example contain an object that was recently freed.
650 PTR_TRUSTED = BIT(12 + BPF_BASE_TYPE_BITS),
652 /* MEM is tagged with rcu and memory access needs rcu_read_lock protection. */
653 MEM_RCU = BIT(13 + BPF_BASE_TYPE_BITS),
655 /* Used to tag PTR_TO_BTF_ID | MEM_ALLOC references which are non-owning.
656 * Currently only valid for linked-list and rbtree nodes. If the nodes
657 * have a bpf_refcount_field, they must be tagged MEM_RCU as well.
659 NON_OWN_REF = BIT(14 + BPF_BASE_TYPE_BITS),
661 /* DYNPTR points to sk_buff */
662 DYNPTR_TYPE_SKB = BIT(15 + BPF_BASE_TYPE_BITS),
664 /* DYNPTR points to xdp_buff */
665 DYNPTR_TYPE_XDP = BIT(16 + BPF_BASE_TYPE_BITS),
668 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1,
671 #define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF | DYNPTR_TYPE_SKB \
674 /* Max number of base types. */
675 #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS)
677 /* Max number of all types. */
678 #define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1))
680 /* function argument constraints */
682 ARG_DONTCARE = 0, /* unused argument in helper function */
684 /* the following constraints used to prototype
685 * bpf_map_lookup/update/delete_elem() functions
687 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
688 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
689 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
691 /* Used to prototype bpf_memcmp() and other functions that access data
692 * on eBPF program stack
694 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
696 ARG_CONST_SIZE, /* number of bytes accessed from memory */
697 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
699 ARG_PTR_TO_CTX, /* pointer to context */
700 ARG_ANYTHING, /* any (initialized) argument is ok */
701 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
702 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
703 ARG_PTR_TO_INT, /* pointer to int */
704 ARG_PTR_TO_LONG, /* pointer to long */
705 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
706 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
707 ARG_PTR_TO_RINGBUF_MEM, /* pointer to dynamically reserved ringbuf memory */
708 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
709 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
710 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */
711 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */
712 ARG_PTR_TO_STACK, /* pointer to stack */
713 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */
714 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */
715 ARG_PTR_TO_KPTR, /* pointer to referenced kptr */
716 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */
719 /* Extended arg_types. */
720 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE,
721 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM,
722 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX,
723 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET,
724 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK,
725 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID,
726 /* pointer to memory does not need to be initialized, helper function must fill
727 * all bytes or clear them in error case.
729 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | ARG_PTR_TO_MEM,
730 /* Pointer to valid memory of size known at compile time. */
731 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM,
733 /* This must be the last entry. Its purpose is to ensure the enum is
734 * wide enough to hold the higher bits reserved for bpf_type_flag.
736 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT,
738 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
740 /* type of values returned from helper functions */
741 enum bpf_return_type {
742 RET_INTEGER, /* function returns integer */
743 RET_VOID, /* function doesn't return anything */
744 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
745 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */
746 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */
747 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */
748 RET_PTR_TO_MEM, /* returns a pointer to memory */
749 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */
750 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */
753 /* Extended ret_types. */
754 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE,
755 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET,
756 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK,
757 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON,
758 RET_PTR_TO_RINGBUF_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_RINGBUF | RET_PTR_TO_MEM,
759 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MEM,
760 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID,
761 RET_PTR_TO_BTF_ID_TRUSTED = PTR_TRUSTED | RET_PTR_TO_BTF_ID,
763 /* This must be the last entry. Its purpose is to ensure the enum is
764 * wide enough to hold the higher bits reserved for bpf_type_flag.
766 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT,
768 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
770 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
771 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
772 * instructions after verifying
774 struct bpf_func_proto {
775 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
779 enum bpf_return_type ret_type;
782 enum bpf_arg_type arg1_type;
783 enum bpf_arg_type arg2_type;
784 enum bpf_arg_type arg3_type;
785 enum bpf_arg_type arg4_type;
786 enum bpf_arg_type arg5_type;
788 enum bpf_arg_type arg_type[5];
808 int *ret_btf_id; /* return value btf_id */
809 bool (*allowed)(const struct bpf_prog *prog);
812 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
813 * the first argument to eBPF programs.
814 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
818 enum bpf_access_type {
823 /* types of values stored in eBPF registers */
824 /* Pointer types represent:
827 * pointer + (u16) var
828 * pointer + (u16) var + imm
829 * if (range > 0) then [ptr, ptr + range - off) is safe to access
830 * if (id > 0) means that some 'var' was added
831 * if (off > 0) means that 'imm' was added
834 NOT_INIT = 0, /* nothing was written into register */
835 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
836 PTR_TO_CTX, /* reg points to bpf_context */
837 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
838 PTR_TO_MAP_VALUE, /* reg points to map element value */
839 PTR_TO_MAP_KEY, /* reg points to a map element key */
840 PTR_TO_STACK, /* reg == frame_pointer + offset */
841 PTR_TO_PACKET_META, /* skb->data - meta_len */
842 PTR_TO_PACKET, /* reg points to skb->data */
843 PTR_TO_PACKET_END, /* skb->data + headlen */
844 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
845 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
846 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
847 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
848 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
849 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
850 /* PTR_TO_BTF_ID points to a kernel struct that does not need
851 * to be null checked by the BPF program. This does not imply the
852 * pointer is _not_ null and in practice this can easily be a null
853 * pointer when reading pointer chains. The assumption is program
854 * context will handle null pointer dereference typically via fault
855 * handling. The verifier must keep this in mind and can make no
856 * assumptions about null or non-null when doing branch analysis.
857 * Further, when passed into helpers the helpers can not, without
858 * additional context, assume the value is non-null.
861 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
862 * been checked for null. Used primarily to inform the verifier
863 * an explicit null check is required for this struct.
865 PTR_TO_MEM, /* reg points to valid memory region */
866 PTR_TO_BUF, /* reg points to a read/write buffer */
867 PTR_TO_FUNC, /* reg points to a bpf program function */
868 CONST_PTR_TO_DYNPTR, /* reg points to a const struct bpf_dynptr */
871 /* Extended reg_types. */
872 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE,
873 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET,
874 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON,
875 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK,
876 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID,
878 /* This must be the last entry. Its purpose is to ensure the enum is
879 * wide enough to hold the higher bits reserved for bpf_type_flag.
881 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT,
883 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
885 /* The information passed from prog-specific *_is_valid_access
886 * back to the verifier.
888 struct bpf_insn_access_aux {
889 enum bpf_reg_type reg_type;
897 struct bpf_verifier_log *log; /* for verbose logs */
901 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
903 aux->ctx_field_size = size;
906 static bool bpf_is_ldimm64(const struct bpf_insn *insn)
908 return insn->code == (BPF_LD | BPF_IMM | BPF_DW);
911 static inline bool bpf_pseudo_func(const struct bpf_insn *insn)
913 return bpf_is_ldimm64(insn) && insn->src_reg == BPF_PSEUDO_FUNC;
916 struct bpf_prog_ops {
917 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
918 union bpf_attr __user *uattr);
921 struct bpf_reg_state;
922 struct bpf_verifier_ops {
923 /* return eBPF function prototype for verification */
924 const struct bpf_func_proto *
925 (*get_func_proto)(enum bpf_func_id func_id,
926 const struct bpf_prog *prog);
928 /* return true if 'size' wide access at offset 'off' within bpf_context
929 * with 'type' (read or write) is allowed
931 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
932 const struct bpf_prog *prog,
933 struct bpf_insn_access_aux *info);
934 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
935 const struct bpf_prog *prog);
936 int (*gen_ld_abs)(const struct bpf_insn *orig,
937 struct bpf_insn *insn_buf);
938 u32 (*convert_ctx_access)(enum bpf_access_type type,
939 const struct bpf_insn *src,
940 struct bpf_insn *dst,
941 struct bpf_prog *prog, u32 *target_size);
942 int (*btf_struct_access)(struct bpf_verifier_log *log,
943 const struct bpf_reg_state *reg,
947 struct bpf_prog_offload_ops {
948 /* verifier basic callbacks */
949 int (*insn_hook)(struct bpf_verifier_env *env,
950 int insn_idx, int prev_insn_idx);
951 int (*finalize)(struct bpf_verifier_env *env);
952 /* verifier optimization callbacks (called after .finalize) */
953 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
954 struct bpf_insn *insn);
955 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
956 /* program management callbacks */
957 int (*prepare)(struct bpf_prog *prog);
958 int (*translate)(struct bpf_prog *prog);
959 void (*destroy)(struct bpf_prog *prog);
962 struct bpf_prog_offload {
963 struct bpf_prog *prog;
964 struct net_device *netdev;
965 struct bpf_offload_dev *offdev;
967 struct list_head offloads;
974 enum bpf_cgroup_storage_type {
975 BPF_CGROUP_STORAGE_SHARED,
976 BPF_CGROUP_STORAGE_PERCPU,
977 __BPF_CGROUP_STORAGE_MAX
980 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
982 /* The longest tracepoint has 12 args.
983 * See include/trace/bpf_probe.h
985 #define MAX_BPF_FUNC_ARGS 12
987 /* The maximum number of arguments passed through registers
988 * a single function may have.
990 #define MAX_BPF_FUNC_REG_ARGS 5
992 /* The argument is a structure. */
993 #define BTF_FMODEL_STRUCT_ARG BIT(0)
995 /* The argument is signed. */
996 #define BTF_FMODEL_SIGNED_ARG BIT(1)
998 struct btf_func_model {
1002 u8 arg_size[MAX_BPF_FUNC_ARGS];
1003 u8 arg_flags[MAX_BPF_FUNC_ARGS];
1006 /* Restore arguments before returning from trampoline to let original function
1007 * continue executing. This flag is used for fentry progs when there are no
1010 #define BPF_TRAMP_F_RESTORE_REGS BIT(0)
1011 /* Call original function after fentry progs, but before fexit progs.
1012 * Makes sense for fentry/fexit, normal calls and indirect calls.
1014 #define BPF_TRAMP_F_CALL_ORIG BIT(1)
1015 /* Skip current frame and return to parent. Makes sense for fentry/fexit
1016 * programs only. Should not be used with normal calls and indirect calls.
1018 #define BPF_TRAMP_F_SKIP_FRAME BIT(2)
1019 /* Store IP address of the caller on the trampoline stack,
1020 * so it's available for trampoline's programs.
1022 #define BPF_TRAMP_F_IP_ARG BIT(3)
1023 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */
1024 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4)
1026 /* Get original function from stack instead of from provided direct address.
1027 * Makes sense for trampolines with fexit or fmod_ret programs.
1029 #define BPF_TRAMP_F_ORIG_STACK BIT(5)
1031 /* This trampoline is on a function with another ftrace_ops with IPMODIFY,
1032 * e.g., a live patch. This flag is set and cleared by ftrace call backs,
1034 #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6)
1036 /* Indicate that current trampoline is in a tail call context. Then, it has to
1037 * cache and restore tail_call_cnt to avoid infinite tail call loop.
1039 #define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7)
1041 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
1045 #if defined(__s390x__)
1046 BPF_MAX_TRAMP_LINKS = 27,
1048 BPF_MAX_TRAMP_LINKS = 38,
1052 struct bpf_tramp_links {
1053 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS];
1057 struct bpf_tramp_run_ctx;
1059 /* Different use cases for BPF trampoline:
1060 * 1. replace nop at the function entry (kprobe equivalent)
1061 * flags = BPF_TRAMP_F_RESTORE_REGS
1062 * fentry = a set of programs to run before returning from trampoline
1064 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
1065 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
1066 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
1067 * fentry = a set of program to run before calling original function
1068 * fexit = a set of program to run after original function
1070 * 3. replace direct call instruction anywhere in the function body
1071 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
1073 * fentry = a set of programs to run before returning from trampoline
1074 * With flags = BPF_TRAMP_F_CALL_ORIG
1075 * orig_call = original callback addr or direct function addr
1076 * fentry = a set of program to run before calling original function
1077 * fexit = a set of program to run after original function
1079 struct bpf_tramp_image;
1080 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end,
1081 const struct btf_func_model *m, u32 flags,
1082 struct bpf_tramp_links *tlinks,
1084 u64 notrace __bpf_prog_enter_sleepable_recur(struct bpf_prog *prog,
1085 struct bpf_tramp_run_ctx *run_ctx);
1086 void notrace __bpf_prog_exit_sleepable_recur(struct bpf_prog *prog, u64 start,
1087 struct bpf_tramp_run_ctx *run_ctx);
1088 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
1089 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
1090 typedef u64 (*bpf_trampoline_enter_t)(struct bpf_prog *prog,
1091 struct bpf_tramp_run_ctx *run_ctx);
1092 typedef void (*bpf_trampoline_exit_t)(struct bpf_prog *prog, u64 start,
1093 struct bpf_tramp_run_ctx *run_ctx);
1094 bpf_trampoline_enter_t bpf_trampoline_enter(const struct bpf_prog *prog);
1095 bpf_trampoline_exit_t bpf_trampoline_exit(const struct bpf_prog *prog);
1098 unsigned long start;
1100 char name[KSYM_NAME_LEN];
1101 struct list_head lnode;
1102 struct latch_tree_node tnode;
1106 enum bpf_tramp_prog_type {
1109 BPF_TRAMP_MODIFY_RETURN,
1111 BPF_TRAMP_REPLACE, /* more than MAX */
1114 struct bpf_tramp_image {
1116 struct bpf_ksym ksym;
1117 struct percpu_ref pcref;
1118 void *ip_after_call;
1121 struct rcu_head rcu;
1122 struct work_struct work;
1126 struct bpf_trampoline {
1127 /* hlist for trampoline_table */
1128 struct hlist_node hlist;
1129 struct ftrace_ops *fops;
1130 /* serializes access to fields of this trampoline */
1136 struct btf_func_model model;
1138 bool ftrace_managed;
1140 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
1141 * program by replacing one of its functions. func.addr is the address
1142 * of the function it replaced.
1144 struct bpf_prog *extension_prog;
1145 /* list of BPF programs using this trampoline */
1146 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
1147 /* Number of attached programs. A counter per kind. */
1148 int progs_cnt[BPF_TRAMP_MAX];
1149 /* Executable image of trampoline */
1150 struct bpf_tramp_image *cur_image;
1154 struct bpf_attach_target_info {
1155 struct btf_func_model fmodel;
1157 struct module *tgt_mod;
1158 const char *tgt_name;
1159 const struct btf_type *tgt_type;
1162 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
1164 struct bpf_dispatcher_prog {
1165 struct bpf_prog *prog;
1169 struct bpf_dispatcher {
1170 /* dispatcher mutex */
1173 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
1178 struct bpf_ksym ksym;
1179 #ifdef CONFIG_HAVE_STATIC_CALL
1180 struct static_call_key *sc_key;
1185 static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func(
1187 const struct bpf_insn *insnsi,
1188 bpf_func_t bpf_func)
1190 return bpf_func(ctx, insnsi);
1193 /* the implementation of the opaque uapi struct bpf_dynptr */
1194 struct bpf_dynptr_kern {
1196 /* Size represents the number of usable bytes of dynptr data.
1197 * If for example the offset is at 4 for a local dynptr whose data is
1198 * of type u64, the number of usable bytes is 4.
1200 * The upper 8 bits are reserved. It is as follows:
1201 * Bits 0 - 23 = size
1202 * Bits 24 - 30 = dynptr type
1203 * Bit 31 = whether dynptr is read-only
1209 enum bpf_dynptr_type {
1210 BPF_DYNPTR_TYPE_INVALID,
1211 /* Points to memory that is local to the bpf program */
1212 BPF_DYNPTR_TYPE_LOCAL,
1213 /* Underlying data is a ringbuf record */
1214 BPF_DYNPTR_TYPE_RINGBUF,
1215 /* Underlying data is a sk_buff */
1216 BPF_DYNPTR_TYPE_SKB,
1217 /* Underlying data is a xdp_buff */
1218 BPF_DYNPTR_TYPE_XDP,
1221 int bpf_dynptr_check_size(u32 size);
1222 u32 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr);
1224 #ifdef CONFIG_BPF_JIT
1225 int bpf_trampoline_link_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
1226 int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
1227 struct bpf_trampoline *bpf_trampoline_get(u64 key,
1228 struct bpf_attach_target_info *tgt_info);
1229 void bpf_trampoline_put(struct bpf_trampoline *tr);
1230 int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_funcs);
1233 * When the architecture supports STATIC_CALL replace the bpf_dispatcher_fn
1234 * indirection with a direct call to the bpf program. If the architecture does
1235 * not have STATIC_CALL, avoid a double-indirection.
1237 #ifdef CONFIG_HAVE_STATIC_CALL
1239 #define __BPF_DISPATCHER_SC_INIT(_name) \
1240 .sc_key = &STATIC_CALL_KEY(_name), \
1241 .sc_tramp = STATIC_CALL_TRAMP_ADDR(_name),
1243 #define __BPF_DISPATCHER_SC(name) \
1244 DEFINE_STATIC_CALL(bpf_dispatcher_##name##_call, bpf_dispatcher_nop_func)
1246 #define __BPF_DISPATCHER_CALL(name) \
1247 static_call(bpf_dispatcher_##name##_call)(ctx, insnsi, bpf_func)
1249 #define __BPF_DISPATCHER_UPDATE(_d, _new) \
1250 __static_call_update((_d)->sc_key, (_d)->sc_tramp, (_new))
1253 #define __BPF_DISPATCHER_SC_INIT(name)
1254 #define __BPF_DISPATCHER_SC(name)
1255 #define __BPF_DISPATCHER_CALL(name) bpf_func(ctx, insnsi)
1256 #define __BPF_DISPATCHER_UPDATE(_d, _new)
1259 #define BPF_DISPATCHER_INIT(_name) { \
1260 .mutex = __MUTEX_INITIALIZER(_name.mutex), \
1261 .func = &_name##_func, \
1268 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
1270 __BPF_DISPATCHER_SC_INIT(_name##_call) \
1273 #define DEFINE_BPF_DISPATCHER(name) \
1274 __BPF_DISPATCHER_SC(name); \
1275 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \
1277 const struct bpf_insn *insnsi, \
1278 bpf_func_t bpf_func) \
1280 return __BPF_DISPATCHER_CALL(name); \
1282 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
1283 struct bpf_dispatcher bpf_dispatcher_##name = \
1284 BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
1286 #define DECLARE_BPF_DISPATCHER(name) \
1287 unsigned int bpf_dispatcher_##name##_func( \
1289 const struct bpf_insn *insnsi, \
1290 bpf_func_t bpf_func); \
1291 extern struct bpf_dispatcher bpf_dispatcher_##name;
1293 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
1294 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
1295 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
1296 struct bpf_prog *to);
1297 /* Called only from JIT-enabled code, so there's no need for stubs. */
1298 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym);
1299 void bpf_image_ksym_del(struct bpf_ksym *ksym);
1300 void bpf_ksym_add(struct bpf_ksym *ksym);
1301 void bpf_ksym_del(struct bpf_ksym *ksym);
1302 int bpf_jit_charge_modmem(u32 size);
1303 void bpf_jit_uncharge_modmem(u32 size);
1304 bool bpf_prog_has_trampoline(const struct bpf_prog *prog);
1306 static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
1307 struct bpf_trampoline *tr)
1311 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
1312 struct bpf_trampoline *tr)
1316 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
1317 struct bpf_attach_target_info *tgt_info)
1321 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
1322 #define DEFINE_BPF_DISPATCHER(name)
1323 #define DECLARE_BPF_DISPATCHER(name)
1324 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
1325 #define BPF_DISPATCHER_PTR(name) NULL
1326 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
1327 struct bpf_prog *from,
1328 struct bpf_prog *to) {}
1329 static inline bool is_bpf_image_address(unsigned long address)
1333 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog)
1339 struct bpf_func_info_aux {
1344 enum bpf_jit_poke_reason {
1345 BPF_POKE_REASON_TAIL_CALL,
1348 /* Descriptor of pokes pointing /into/ the JITed image. */
1349 struct bpf_jit_poke_descriptor {
1350 void *tailcall_target;
1351 void *tailcall_bypass;
1356 struct bpf_map *map;
1360 bool tailcall_target_stable;
1366 /* reg_type info for ctx arguments */
1367 struct bpf_ctx_arg_aux {
1369 enum bpf_reg_type reg_type;
1373 struct btf_mod_pair {
1375 struct module *module;
1378 struct bpf_kfunc_desc_tab;
1380 struct bpf_prog_aux {
1389 u32 func_cnt; /* used by non-func prog as the number of func progs */
1390 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
1391 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1392 u32 ctx_arg_info_size;
1393 u32 max_rdonly_access;
1394 u32 max_rdwr_access;
1395 struct btf *attach_btf;
1396 const struct bpf_ctx_arg_aux *ctx_arg_info;
1397 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
1398 struct bpf_prog *dst_prog;
1399 struct bpf_trampoline *dst_trampoline;
1400 enum bpf_prog_type saved_dst_prog_type;
1401 enum bpf_attach_type saved_dst_attach_type;
1402 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
1403 bool dev_bound; /* Program is bound to the netdev. */
1404 bool offload_requested; /* Program is bound and offloaded to the netdev. */
1405 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
1406 bool func_proto_unreliable;
1408 bool tail_call_reachable;
1410 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
1411 const struct btf_type *attach_func_proto;
1412 /* function name for valid attach_btf_id */
1413 const char *attach_func_name;
1414 struct bpf_prog **func;
1415 void *jit_data; /* JIT specific data. arch dependent */
1416 struct bpf_jit_poke_descriptor *poke_tab;
1417 struct bpf_kfunc_desc_tab *kfunc_tab;
1418 struct bpf_kfunc_btf_tab *kfunc_btf_tab;
1420 struct bpf_ksym ksym;
1421 const struct bpf_prog_ops *ops;
1422 struct bpf_map **used_maps;
1423 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
1424 struct btf_mod_pair *used_btfs;
1425 struct bpf_prog *prog;
1426 struct user_struct *user;
1427 u64 load_time; /* ns since boottime */
1429 int cgroup_atype; /* enum cgroup_bpf_attach_type */
1430 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1431 char name[BPF_OBJ_NAME_LEN];
1432 #ifdef CONFIG_SECURITY
1435 struct bpf_prog_offload *offload;
1437 struct bpf_func_info *func_info;
1438 struct bpf_func_info_aux *func_info_aux;
1439 /* bpf_line_info loaded from userspace. linfo->insn_off
1440 * has the xlated insn offset.
1441 * Both the main and sub prog share the same linfo.
1442 * The subprog can access its first linfo by
1443 * using the linfo_idx.
1445 struct bpf_line_info *linfo;
1446 /* jited_linfo is the jited addr of the linfo. It has a
1447 * one to one mapping to linfo:
1448 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
1449 * Both the main and sub prog share the same jited_linfo.
1450 * The subprog can access its first jited_linfo by
1451 * using the linfo_idx.
1456 /* subprog can use linfo_idx to access its first linfo and
1458 * main prog always has linfo_idx == 0
1463 struct exception_table_entry *extable;
1465 struct work_struct work;
1466 struct rcu_head rcu;
1471 u16 pages; /* Number of allocated pages */
1472 u16 jited:1, /* Is our filter JIT'ed? */
1473 jit_requested:1,/* archs need to JIT the prog */
1474 gpl_compatible:1, /* Is filter GPL compatible? */
1475 cb_access:1, /* Is control block accessed? */
1476 dst_needed:1, /* Do we need dst entry? */
1477 blinding_requested:1, /* needs constant blinding */
1478 blinded:1, /* Was blinded */
1479 is_func:1, /* program is a bpf function */
1480 kprobe_override:1, /* Do we override a kprobe? */
1481 has_callchain_buf:1, /* callchain buffer allocated? */
1482 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */
1483 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */
1484 call_get_func_ip:1, /* Do we call get_func_ip() */
1485 tstamp_type_access:1; /* Accessed __sk_buff->tstamp_type */
1486 enum bpf_prog_type type; /* Type of BPF program */
1487 enum bpf_attach_type expected_attach_type; /* For some prog types */
1488 u32 len; /* Number of filter blocks */
1489 u32 jited_len; /* Size of jited insns in bytes */
1490 u8 tag[BPF_TAG_SIZE];
1491 struct bpf_prog_stats __percpu *stats;
1492 int __percpu *active;
1493 unsigned int (*bpf_func)(const void *ctx,
1494 const struct bpf_insn *insn);
1495 struct bpf_prog_aux *aux; /* Auxiliary fields */
1496 struct sock_fprog_kern *orig_prog; /* Original BPF program */
1497 /* Instructions for interpreter */
1499 DECLARE_FLEX_ARRAY(struct sock_filter, insns);
1500 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi);
1504 struct bpf_array_aux {
1505 /* Programs with direct jumps into programs part of this array. */
1506 struct list_head poke_progs;
1507 struct bpf_map *map;
1508 struct mutex poke_mutex;
1509 struct work_struct work;
1515 enum bpf_link_type type;
1516 const struct bpf_link_ops *ops;
1517 struct bpf_prog *prog;
1518 struct work_struct work;
1521 struct bpf_link_ops {
1522 void (*release)(struct bpf_link *link);
1523 void (*dealloc)(struct bpf_link *link);
1524 int (*detach)(struct bpf_link *link);
1525 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
1526 struct bpf_prog *old_prog);
1527 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
1528 int (*fill_link_info)(const struct bpf_link *link,
1529 struct bpf_link_info *info);
1530 int (*update_map)(struct bpf_link *link, struct bpf_map *new_map,
1531 struct bpf_map *old_map);
1534 struct bpf_tramp_link {
1535 struct bpf_link link;
1536 struct hlist_node tramp_hlist;
1540 struct bpf_shim_tramp_link {
1541 struct bpf_tramp_link link;
1542 struct bpf_trampoline *trampoline;
1545 struct bpf_tracing_link {
1546 struct bpf_tramp_link link;
1547 enum bpf_attach_type attach_type;
1548 struct bpf_trampoline *trampoline;
1549 struct bpf_prog *tgt_prog;
1552 struct bpf_link_primer {
1553 struct bpf_link *link;
1559 struct bpf_struct_ops_value;
1562 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
1564 * struct bpf_struct_ops - A structure of callbacks allowing a subsystem to
1565 * define a BPF_MAP_TYPE_STRUCT_OPS map type composed
1566 * of BPF_PROG_TYPE_STRUCT_OPS progs.
1567 * @verifier_ops: A structure of callbacks that are invoked by the verifier
1568 * when determining whether the struct_ops progs in the
1569 * struct_ops map are valid.
1570 * @init: A callback that is invoked a single time, and before any other
1571 * callback, to initialize the structure. A nonzero return value means
1572 * the subsystem could not be initialized.
1573 * @check_member: When defined, a callback invoked by the verifier to allow
1574 * the subsystem to determine if an entry in the struct_ops map
1575 * is valid. A nonzero return value means that the map is
1576 * invalid and should be rejected by the verifier.
1577 * @init_member: A callback that is invoked for each member of the struct_ops
1578 * map to allow the subsystem to initialize the member. A nonzero
1579 * value means the member could not be initialized. This callback
1580 * is exclusive with the @type, @type_id, @value_type, and
1582 * @reg: A callback that is invoked when the struct_ops map has been
1583 * initialized and is being attached to. Zero means the struct_ops map
1584 * has been successfully registered and is live. A nonzero return value
1585 * means the struct_ops map could not be registered.
1586 * @unreg: A callback that is invoked when the struct_ops map should be
1588 * @update: A callback that is invoked when the live struct_ops map is being
1589 * updated to contain new values. This callback is only invoked when
1590 * the struct_ops map is loaded with BPF_F_LINK. If not defined, the
1591 * it is assumed that the struct_ops map cannot be updated.
1592 * @validate: A callback that is invoked after all of the members have been
1593 * initialized. This callback should perform static checks on the
1594 * map, meaning that it should either fail or succeed
1595 * deterministically. A struct_ops map that has been validated may
1596 * not necessarily succeed in being registered if the call to @reg
1597 * fails. For example, a valid struct_ops map may be loaded, but
1598 * then fail to be registered due to there being another active
1599 * struct_ops map on the system in the subsystem already. For this
1600 * reason, if this callback is not defined, the check is skipped as
1601 * the struct_ops map will have final verification performed in
1604 * @value_type: Value type.
1605 * @name: The name of the struct bpf_struct_ops object.
1606 * @func_models: Func models
1607 * @type_id: BTF type id.
1608 * @value_id: BTF value id.
1610 struct bpf_struct_ops {
1611 const struct bpf_verifier_ops *verifier_ops;
1612 int (*init)(struct btf *btf);
1613 int (*check_member)(const struct btf_type *t,
1614 const struct btf_member *member,
1615 const struct bpf_prog *prog);
1616 int (*init_member)(const struct btf_type *t,
1617 const struct btf_member *member,
1618 void *kdata, const void *udata);
1619 int (*reg)(void *kdata);
1620 void (*unreg)(void *kdata);
1621 int (*update)(void *kdata, void *old_kdata);
1622 int (*validate)(void *kdata);
1623 const struct btf_type *type;
1624 const struct btf_type *value_type;
1626 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
1631 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
1632 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
1633 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id);
1634 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log);
1635 bool bpf_struct_ops_get(const void *kdata);
1636 void bpf_struct_ops_put(const void *kdata);
1637 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
1639 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks,
1640 struct bpf_tramp_link *link,
1641 const struct btf_func_model *model,
1642 void *image, void *image_end);
1643 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1645 if (owner == BPF_MODULE_OWNER)
1646 return bpf_struct_ops_get(data);
1648 return try_module_get(owner);
1650 static inline void bpf_module_put(const void *data, struct module *owner)
1652 if (owner == BPF_MODULE_OWNER)
1653 bpf_struct_ops_put(data);
1657 int bpf_struct_ops_link_create(union bpf_attr *attr);
1660 /* Define it here to avoid the use of forward declaration */
1661 struct bpf_dummy_ops_state {
1665 struct bpf_dummy_ops {
1666 int (*test_1)(struct bpf_dummy_ops_state *cb);
1667 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2,
1668 char a3, unsigned long a4);
1669 int (*test_sleepable)(struct bpf_dummy_ops_state *cb);
1672 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr,
1673 union bpf_attr __user *uattr);
1676 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id)
1680 static inline void bpf_struct_ops_init(struct btf *btf,
1681 struct bpf_verifier_log *log)
1684 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1686 return try_module_get(owner);
1688 static inline void bpf_module_put(const void *data, struct module *owner)
1692 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
1698 static inline int bpf_struct_ops_link_create(union bpf_attr *attr)
1705 #if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM)
1706 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1708 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog);
1710 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1715 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog)
1724 struct bpf_array_aux *aux;
1726 DECLARE_FLEX_ARRAY(char, value) __aligned(8);
1727 DECLARE_FLEX_ARRAY(void *, ptrs) __aligned(8);
1728 DECLARE_FLEX_ARRAY(void __percpu *, pptrs) __aligned(8);
1732 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
1733 #define MAX_TAIL_CALL_CNT 33
1735 /* Maximum number of loops for bpf_loop and bpf_iter_num.
1736 * It's enum to expose it (and thus make it discoverable) through BTF.
1739 BPF_MAX_LOOPS = 8 * 1024 * 1024,
1742 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
1743 BPF_F_RDONLY_PROG | \
1747 #define BPF_MAP_CAN_READ BIT(0)
1748 #define BPF_MAP_CAN_WRITE BIT(1)
1750 /* Maximum number of user-producer ring buffer samples that can be drained in
1751 * a call to bpf_user_ringbuf_drain().
1753 #define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024)
1755 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
1757 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1759 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
1762 if (access_flags & BPF_F_RDONLY_PROG)
1763 return BPF_MAP_CAN_READ;
1764 else if (access_flags & BPF_F_WRONLY_PROG)
1765 return BPF_MAP_CAN_WRITE;
1767 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
1770 static inline bool bpf_map_flags_access_ok(u32 access_flags)
1772 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
1773 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1776 struct bpf_event_entry {
1777 struct perf_event *event;
1778 struct file *perf_file;
1779 struct file *map_file;
1780 struct rcu_head rcu;
1783 static inline bool map_type_contains_progs(struct bpf_map *map)
1785 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY ||
1786 map->map_type == BPF_MAP_TYPE_DEVMAP ||
1787 map->map_type == BPF_MAP_TYPE_CPUMAP;
1790 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp);
1791 int bpf_prog_calc_tag(struct bpf_prog *fp);
1793 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
1794 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void);
1796 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
1797 unsigned long off, unsigned long len);
1798 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
1799 const struct bpf_insn *src,
1800 struct bpf_insn *dst,
1801 struct bpf_prog *prog,
1804 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
1805 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
1807 /* an array of programs to be executed under rcu_lock.
1810 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run);
1812 * the structure returned by bpf_prog_array_alloc() should be populated
1813 * with program pointers and the last pointer must be NULL.
1814 * The user has to keep refcnt on the program and make sure the program
1815 * is removed from the array before bpf_prog_put().
1816 * The 'struct bpf_prog_array *' should only be replaced with xchg()
1817 * since other cpus are walking the array of pointers in parallel.
1819 struct bpf_prog_array_item {
1820 struct bpf_prog *prog;
1822 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1827 struct bpf_prog_array {
1828 struct rcu_head rcu;
1829 struct bpf_prog_array_item items[];
1832 struct bpf_empty_prog_array {
1833 struct bpf_prog_array hdr;
1834 struct bpf_prog *null_prog;
1837 /* to avoid allocating empty bpf_prog_array for cgroups that
1838 * don't have bpf program attached use one global 'bpf_empty_prog_array'
1839 * It will not be modified the caller of bpf_prog_array_alloc()
1840 * (since caller requested prog_cnt == 0)
1841 * that pointer should be 'freed' by bpf_prog_array_free()
1843 extern struct bpf_empty_prog_array bpf_empty_prog_array;
1845 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
1846 void bpf_prog_array_free(struct bpf_prog_array *progs);
1847 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */
1848 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs);
1849 int bpf_prog_array_length(struct bpf_prog_array *progs);
1850 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
1851 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
1852 __u32 __user *prog_ids, u32 cnt);
1854 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
1855 struct bpf_prog *old_prog);
1856 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
1857 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
1858 struct bpf_prog *prog);
1859 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
1860 u32 *prog_ids, u32 request_cnt,
1862 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
1863 struct bpf_prog *exclude_prog,
1864 struct bpf_prog *include_prog,
1866 struct bpf_prog_array **new_array);
1868 struct bpf_run_ctx {};
1870 struct bpf_cg_run_ctx {
1871 struct bpf_run_ctx run_ctx;
1872 const struct bpf_prog_array_item *prog_item;
1876 struct bpf_trace_run_ctx {
1877 struct bpf_run_ctx run_ctx;
1882 struct bpf_tramp_run_ctx {
1883 struct bpf_run_ctx run_ctx;
1885 struct bpf_run_ctx *saved_run_ctx;
1888 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)
1890 struct bpf_run_ctx *old_ctx = NULL;
1892 #ifdef CONFIG_BPF_SYSCALL
1893 old_ctx = current->bpf_ctx;
1894 current->bpf_ctx = new_ctx;
1899 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
1901 #ifdef CONFIG_BPF_SYSCALL
1902 current->bpf_ctx = old_ctx;
1906 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
1907 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0)
1908 /* BPF program asks to set CN on the packet. */
1909 #define BPF_RET_SET_CN (1 << 0)
1911 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx);
1913 static __always_inline u32
1914 bpf_prog_run_array(const struct bpf_prog_array *array,
1915 const void *ctx, bpf_prog_run_fn run_prog)
1917 const struct bpf_prog_array_item *item;
1918 const struct bpf_prog *prog;
1919 struct bpf_run_ctx *old_run_ctx;
1920 struct bpf_trace_run_ctx run_ctx;
1923 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held");
1925 if (unlikely(!array))
1928 run_ctx.is_uprobe = false;
1931 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
1932 item = &array->items[0];
1933 while ((prog = READ_ONCE(item->prog))) {
1934 run_ctx.bpf_cookie = item->bpf_cookie;
1935 ret &= run_prog(prog, ctx);
1938 bpf_reset_run_ctx(old_run_ctx);
1943 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs:
1945 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array
1946 * overall. As a result, we must use the bpf_prog_array_free_sleepable
1947 * in order to use the tasks_trace rcu grace period.
1949 * When a non-sleepable program is inside the array, we take the rcu read
1950 * section and disable preemption for that program alone, so it can access
1951 * rcu-protected dynamically sized maps.
1953 static __always_inline u32
1954 bpf_prog_run_array_uprobe(const struct bpf_prog_array __rcu *array_rcu,
1955 const void *ctx, bpf_prog_run_fn run_prog)
1957 const struct bpf_prog_array_item *item;
1958 const struct bpf_prog *prog;
1959 const struct bpf_prog_array *array;
1960 struct bpf_run_ctx *old_run_ctx;
1961 struct bpf_trace_run_ctx run_ctx;
1966 rcu_read_lock_trace();
1969 run_ctx.is_uprobe = true;
1971 array = rcu_dereference_check(array_rcu, rcu_read_lock_trace_held());
1972 if (unlikely(!array))
1974 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
1975 item = &array->items[0];
1976 while ((prog = READ_ONCE(item->prog))) {
1977 if (!prog->aux->sleepable)
1980 run_ctx.bpf_cookie = item->bpf_cookie;
1981 ret &= run_prog(prog, ctx);
1984 if (!prog->aux->sleepable)
1987 bpf_reset_run_ctx(old_run_ctx);
1990 rcu_read_unlock_trace();
1994 #ifdef CONFIG_BPF_SYSCALL
1995 DECLARE_PER_CPU(int, bpf_prog_active);
1996 extern struct mutex bpf_stats_enabled_mutex;
1999 * Block execution of BPF programs attached to instrumentation (perf,
2000 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
2001 * these events can happen inside a region which holds a map bucket lock
2002 * and can deadlock on it.
2004 static inline void bpf_disable_instrumentation(void)
2007 this_cpu_inc(bpf_prog_active);
2010 static inline void bpf_enable_instrumentation(void)
2012 this_cpu_dec(bpf_prog_active);
2016 extern const struct file_operations bpf_map_fops;
2017 extern const struct file_operations bpf_prog_fops;
2018 extern const struct file_operations bpf_iter_fops;
2020 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
2021 extern const struct bpf_prog_ops _name ## _prog_ops; \
2022 extern const struct bpf_verifier_ops _name ## _verifier_ops;
2023 #define BPF_MAP_TYPE(_id, _ops) \
2024 extern const struct bpf_map_ops _ops;
2025 #define BPF_LINK_TYPE(_id, _name)
2026 #include <linux/bpf_types.h>
2027 #undef BPF_PROG_TYPE
2029 #undef BPF_LINK_TYPE
2031 extern const struct bpf_prog_ops bpf_offload_prog_ops;
2032 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
2033 extern const struct bpf_verifier_ops xdp_analyzer_ops;
2035 struct bpf_prog *bpf_prog_get(u32 ufd);
2036 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
2038 void bpf_prog_add(struct bpf_prog *prog, int i);
2039 void bpf_prog_sub(struct bpf_prog *prog, int i);
2040 void bpf_prog_inc(struct bpf_prog *prog);
2041 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
2042 void bpf_prog_put(struct bpf_prog *prog);
2044 void bpf_prog_free_id(struct bpf_prog *prog);
2045 void bpf_map_free_id(struct bpf_map *map);
2047 struct btf_field *btf_record_find(const struct btf_record *rec,
2048 u32 offset, u32 field_mask);
2049 void btf_record_free(struct btf_record *rec);
2050 void bpf_map_free_record(struct bpf_map *map);
2051 struct btf_record *btf_record_dup(const struct btf_record *rec);
2052 bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b);
2053 void bpf_obj_free_timer(const struct btf_record *rec, void *obj);
2054 void bpf_obj_free_fields(const struct btf_record *rec, void *obj);
2056 struct bpf_map *bpf_map_get(u32 ufd);
2057 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
2058 struct bpf_map *__bpf_map_get(struct fd f);
2059 void bpf_map_inc(struct bpf_map *map);
2060 void bpf_map_inc_with_uref(struct bpf_map *map);
2061 struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref);
2062 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
2063 void bpf_map_put_with_uref(struct bpf_map *map);
2064 void bpf_map_put(struct bpf_map *map);
2065 void *bpf_map_area_alloc(u64 size, int numa_node);
2066 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
2067 void bpf_map_area_free(void *base);
2068 bool bpf_map_write_active(const struct bpf_map *map);
2069 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
2070 int generic_map_lookup_batch(struct bpf_map *map,
2071 const union bpf_attr *attr,
2072 union bpf_attr __user *uattr);
2073 int generic_map_update_batch(struct bpf_map *map, struct file *map_file,
2074 const union bpf_attr *attr,
2075 union bpf_attr __user *uattr);
2076 int generic_map_delete_batch(struct bpf_map *map,
2077 const union bpf_attr *attr,
2078 union bpf_attr __user *uattr);
2079 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
2080 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
2082 #ifdef CONFIG_MEMCG_KMEM
2083 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
2085 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags);
2086 void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size,
2088 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
2089 size_t align, gfp_t flags);
2091 static inline void *
2092 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
2095 return kmalloc_node(size, flags, node);
2098 static inline void *
2099 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags)
2101 return kzalloc(size, flags);
2104 static inline void *
2105 bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size, gfp_t flags)
2107 return kvcalloc(n, size, flags);
2110 static inline void __percpu *
2111 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align,
2114 return __alloc_percpu_gfp(size, align, flags);
2119 bpf_map_init_elem_count(struct bpf_map *map)
2121 size_t size = sizeof(*map->elem_count), align = size;
2122 gfp_t flags = GFP_USER | __GFP_NOWARN;
2124 map->elem_count = bpf_map_alloc_percpu(map, size, align, flags);
2125 if (!map->elem_count)
2132 bpf_map_free_elem_count(struct bpf_map *map)
2134 free_percpu(map->elem_count);
2137 static inline void bpf_map_inc_elem_count(struct bpf_map *map)
2139 this_cpu_inc(*map->elem_count);
2142 static inline void bpf_map_dec_elem_count(struct bpf_map *map)
2144 this_cpu_dec(*map->elem_count);
2147 extern int sysctl_unprivileged_bpf_disabled;
2149 static inline bool bpf_allow_ptr_leaks(void)
2151 return perfmon_capable();
2154 static inline bool bpf_allow_uninit_stack(void)
2156 return perfmon_capable();
2159 static inline bool bpf_bypass_spec_v1(void)
2161 return perfmon_capable();
2164 static inline bool bpf_bypass_spec_v4(void)
2166 return perfmon_capable();
2169 int bpf_map_new_fd(struct bpf_map *map, int flags);
2170 int bpf_prog_new_fd(struct bpf_prog *prog);
2172 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2173 const struct bpf_link_ops *ops, struct bpf_prog *prog);
2174 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
2175 int bpf_link_settle(struct bpf_link_primer *primer);
2176 void bpf_link_cleanup(struct bpf_link_primer *primer);
2177 void bpf_link_inc(struct bpf_link *link);
2178 void bpf_link_put(struct bpf_link *link);
2179 int bpf_link_new_fd(struct bpf_link *link);
2180 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
2181 struct bpf_link *bpf_link_get_curr_or_next(u32 *id);
2183 int bpf_obj_pin_user(u32 ufd, int path_fd, const char __user *pathname);
2184 int bpf_obj_get_user(int path_fd, const char __user *pathname, int flags);
2186 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
2187 #define DEFINE_BPF_ITER_FUNC(target, args...) \
2188 extern int bpf_iter_ ## target(args); \
2189 int __init bpf_iter_ ## target(args) { return 0; }
2192 * The task type of iterators.
2194 * For BPF task iterators, they can be parameterized with various
2195 * parameters to visit only some of tasks.
2197 * BPF_TASK_ITER_ALL (default)
2198 * Iterate over resources of every task.
2201 * Iterate over resources of a task/tid.
2203 * BPF_TASK_ITER_TGID
2204 * Iterate over resources of every task of a process / task group.
2206 enum bpf_iter_task_type {
2207 BPF_TASK_ITER_ALL = 0,
2212 struct bpf_iter_aux_info {
2213 /* for map_elem iter */
2214 struct bpf_map *map;
2216 /* for cgroup iter */
2218 struct cgroup *start; /* starting cgroup */
2219 enum bpf_cgroup_iter_order order;
2222 enum bpf_iter_task_type type;
2227 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
2228 union bpf_iter_link_info *linfo,
2229 struct bpf_iter_aux_info *aux);
2230 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
2231 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
2232 struct seq_file *seq);
2233 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
2234 struct bpf_link_info *info);
2235 typedef const struct bpf_func_proto *
2236 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id,
2237 const struct bpf_prog *prog);
2239 enum bpf_iter_feature {
2240 BPF_ITER_RESCHED = BIT(0),
2243 #define BPF_ITER_CTX_ARG_MAX 2
2244 struct bpf_iter_reg {
2246 bpf_iter_attach_target_t attach_target;
2247 bpf_iter_detach_target_t detach_target;
2248 bpf_iter_show_fdinfo_t show_fdinfo;
2249 bpf_iter_fill_link_info_t fill_link_info;
2250 bpf_iter_get_func_proto_t get_func_proto;
2251 u32 ctx_arg_info_size;
2253 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
2254 const struct bpf_iter_seq_info *seq_info;
2257 struct bpf_iter_meta {
2258 __bpf_md_ptr(struct seq_file *, seq);
2263 struct bpf_iter__bpf_map_elem {
2264 __bpf_md_ptr(struct bpf_iter_meta *, meta);
2265 __bpf_md_ptr(struct bpf_map *, map);
2266 __bpf_md_ptr(void *, key);
2267 __bpf_md_ptr(void *, value);
2270 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
2271 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
2272 bool bpf_iter_prog_supported(struct bpf_prog *prog);
2273 const struct bpf_func_proto *
2274 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
2275 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog);
2276 int bpf_iter_new_fd(struct bpf_link *link);
2277 bool bpf_link_is_iter(struct bpf_link *link);
2278 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
2279 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
2280 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
2281 struct seq_file *seq);
2282 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
2283 struct bpf_link_info *info);
2285 int map_set_for_each_callback_args(struct bpf_verifier_env *env,
2286 struct bpf_func_state *caller,
2287 struct bpf_func_state *callee);
2289 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
2290 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
2291 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
2293 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
2296 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
2298 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
2299 void *key, void *value, u64 map_flags);
2300 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2301 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
2302 void *key, void *value, u64 map_flags);
2303 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2305 int bpf_get_file_flag(int flags);
2306 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size,
2307 size_t actual_size);
2309 /* verify correctness of eBPF program */
2310 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size);
2312 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
2313 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
2316 struct btf *bpf_get_btf_vmlinux(void);
2321 struct bpf_dtab_netdev;
2322 struct bpf_cpu_map_entry;
2324 void __dev_flush(void);
2325 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2326 struct net_device *dev_rx);
2327 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2328 struct net_device *dev_rx);
2329 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2330 struct bpf_map *map, bool exclude_ingress);
2331 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
2332 struct bpf_prog *xdp_prog);
2333 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2334 struct bpf_prog *xdp_prog, struct bpf_map *map,
2335 bool exclude_ingress);
2337 void __cpu_map_flush(void);
2338 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf,
2339 struct net_device *dev_rx);
2340 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2341 struct sk_buff *skb);
2343 /* Return map's numa specified by userspace */
2344 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
2346 return (attr->map_flags & BPF_F_NUMA_NODE) ?
2347 attr->numa_node : NUMA_NO_NODE;
2350 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
2351 int array_map_alloc_check(union bpf_attr *attr);
2353 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
2354 union bpf_attr __user *uattr);
2355 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
2356 union bpf_attr __user *uattr);
2357 int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2358 const union bpf_attr *kattr,
2359 union bpf_attr __user *uattr);
2360 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2361 const union bpf_attr *kattr,
2362 union bpf_attr __user *uattr);
2363 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
2364 const union bpf_attr *kattr,
2365 union bpf_attr __user *uattr);
2366 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2367 const union bpf_attr *kattr,
2368 union bpf_attr __user *uattr);
2369 int bpf_prog_test_run_nf(struct bpf_prog *prog,
2370 const union bpf_attr *kattr,
2371 union bpf_attr __user *uattr);
2372 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
2373 const struct bpf_prog *prog,
2374 struct bpf_insn_access_aux *info);
2376 static inline bool bpf_tracing_ctx_access(int off, int size,
2377 enum bpf_access_type type)
2379 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
2381 if (type != BPF_READ)
2383 if (off % size != 0)
2388 static inline bool bpf_tracing_btf_ctx_access(int off, int size,
2389 enum bpf_access_type type,
2390 const struct bpf_prog *prog,
2391 struct bpf_insn_access_aux *info)
2393 if (!bpf_tracing_ctx_access(off, size, type))
2395 return btf_ctx_access(off, size, type, prog, info);
2398 int btf_struct_access(struct bpf_verifier_log *log,
2399 const struct bpf_reg_state *reg,
2400 int off, int size, enum bpf_access_type atype,
2401 u32 *next_btf_id, enum bpf_type_flag *flag, const char **field_name);
2402 bool btf_struct_ids_match(struct bpf_verifier_log *log,
2403 const struct btf *btf, u32 id, int off,
2404 const struct btf *need_btf, u32 need_type_id,
2407 int btf_distill_func_proto(struct bpf_verifier_log *log,
2409 const struct btf_type *func_proto,
2410 const char *func_name,
2411 struct btf_func_model *m);
2413 struct bpf_reg_state;
2414 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog,
2415 struct bpf_reg_state *regs);
2416 int btf_check_subprog_call(struct bpf_verifier_env *env, int subprog,
2417 struct bpf_reg_state *regs);
2418 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog,
2419 struct bpf_reg_state *reg);
2420 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
2421 struct btf *btf, const struct btf_type *t);
2423 struct bpf_prog *bpf_prog_by_id(u32 id);
2424 struct bpf_link *bpf_link_by_id(u32 id);
2426 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
2427 void bpf_task_storage_free(struct task_struct *task);
2428 void bpf_cgrp_storage_free(struct cgroup *cgroup);
2429 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
2430 const struct btf_func_model *
2431 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2432 const struct bpf_insn *insn);
2433 int bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
2434 u16 btf_fd_idx, u8 **func_addr);
2436 struct bpf_core_ctx {
2437 struct bpf_verifier_log *log;
2438 const struct btf *btf;
2441 bool btf_nested_type_is_trusted(struct bpf_verifier_log *log,
2442 const struct bpf_reg_state *reg,
2443 const char *field_name, u32 btf_id, const char *suffix);
2445 bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log,
2446 const struct btf *reg_btf, u32 reg_id,
2447 const struct btf *arg_btf, u32 arg_id);
2449 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo,
2450 int relo_idx, void *insn);
2452 static inline bool unprivileged_ebpf_enabled(void)
2454 return !sysctl_unprivileged_bpf_disabled;
2457 /* Not all bpf prog type has the bpf_ctx.
2458 * For the bpf prog type that has initialized the bpf_ctx,
2459 * this function can be used to decide if a kernel function
2460 * is called by a bpf program.
2462 static inline bool has_current_bpf_ctx(void)
2464 return !!current->bpf_ctx;
2467 void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog);
2469 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2470 enum bpf_dynptr_type type, u32 offset, u32 size);
2471 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr);
2472 void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr);
2473 #else /* !CONFIG_BPF_SYSCALL */
2474 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
2476 return ERR_PTR(-EOPNOTSUPP);
2479 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
2480 enum bpf_prog_type type,
2483 return ERR_PTR(-EOPNOTSUPP);
2486 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
2490 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
2494 static inline void bpf_prog_put(struct bpf_prog *prog)
2498 static inline void bpf_prog_inc(struct bpf_prog *prog)
2502 static inline struct bpf_prog *__must_check
2503 bpf_prog_inc_not_zero(struct bpf_prog *prog)
2505 return ERR_PTR(-EOPNOTSUPP);
2508 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2509 const struct bpf_link_ops *ops,
2510 struct bpf_prog *prog)
2514 static inline int bpf_link_prime(struct bpf_link *link,
2515 struct bpf_link_primer *primer)
2520 static inline int bpf_link_settle(struct bpf_link_primer *primer)
2525 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
2529 static inline void bpf_link_inc(struct bpf_link *link)
2533 static inline void bpf_link_put(struct bpf_link *link)
2537 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
2542 static inline void __dev_flush(void)
2547 struct bpf_dtab_netdev;
2548 struct bpf_cpu_map_entry;
2551 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2552 struct net_device *dev_rx)
2558 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2559 struct net_device *dev_rx)
2565 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2566 struct bpf_map *map, bool exclude_ingress)
2573 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
2574 struct sk_buff *skb,
2575 struct bpf_prog *xdp_prog)
2581 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2582 struct bpf_prog *xdp_prog, struct bpf_map *map,
2583 bool exclude_ingress)
2588 static inline void __cpu_map_flush(void)
2592 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
2593 struct xdp_frame *xdpf,
2594 struct net_device *dev_rx)
2599 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2600 struct sk_buff *skb)
2605 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
2606 enum bpf_prog_type type)
2608 return ERR_PTR(-EOPNOTSUPP);
2611 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
2612 const union bpf_attr *kattr,
2613 union bpf_attr __user *uattr)
2618 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
2619 const union bpf_attr *kattr,
2620 union bpf_attr __user *uattr)
2625 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2626 const union bpf_attr *kattr,
2627 union bpf_attr __user *uattr)
2632 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2633 const union bpf_attr *kattr,
2634 union bpf_attr __user *uattr)
2639 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2640 const union bpf_attr *kattr,
2641 union bpf_attr __user *uattr)
2646 static inline void bpf_map_put(struct bpf_map *map)
2650 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
2652 return ERR_PTR(-ENOTSUPP);
2655 static inline int btf_struct_access(struct bpf_verifier_log *log,
2656 const struct bpf_reg_state *reg,
2657 int off, int size, enum bpf_access_type atype,
2658 u32 *next_btf_id, enum bpf_type_flag *flag,
2659 const char **field_name)
2664 static inline const struct bpf_func_proto *
2665 bpf_base_func_proto(enum bpf_func_id func_id)
2670 static inline void bpf_task_storage_free(struct task_struct *task)
2674 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog)
2679 static inline const struct btf_func_model *
2680 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2681 const struct bpf_insn *insn)
2687 bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
2688 u16 btf_fd_idx, u8 **func_addr)
2693 static inline bool unprivileged_ebpf_enabled(void)
2698 static inline bool has_current_bpf_ctx(void)
2703 static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog)
2707 static inline void bpf_cgrp_storage_free(struct cgroup *cgroup)
2711 static inline void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2712 enum bpf_dynptr_type type, u32 offset, u32 size)
2716 static inline void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr)
2720 static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr)
2723 #endif /* CONFIG_BPF_SYSCALL */
2725 static __always_inline int
2726 bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
2730 if (IS_ENABLED(CONFIG_BPF_EVENTS))
2731 ret = copy_from_kernel_nofault(dst, unsafe_ptr, size);
2732 if (unlikely(ret < 0))
2733 memset(dst, 0, size);
2737 void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
2738 struct btf_mod_pair *used_btfs, u32 len);
2740 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
2741 enum bpf_prog_type type)
2743 return bpf_prog_get_type_dev(ufd, type, false);
2746 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
2747 struct bpf_map **used_maps, u32 len);
2749 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
2751 int bpf_prog_offload_compile(struct bpf_prog *prog);
2752 void bpf_prog_dev_bound_destroy(struct bpf_prog *prog);
2753 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
2754 struct bpf_prog *prog);
2756 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
2758 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
2759 int bpf_map_offload_update_elem(struct bpf_map *map,
2760 void *key, void *value, u64 flags);
2761 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
2762 int bpf_map_offload_get_next_key(struct bpf_map *map,
2763 void *key, void *next_key);
2765 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
2767 struct bpf_offload_dev *
2768 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
2769 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
2770 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
2771 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
2772 struct net_device *netdev);
2773 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
2774 struct net_device *netdev);
2775 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
2777 void unpriv_ebpf_notify(int new_state);
2779 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
2780 int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
2781 struct bpf_prog_aux *prog_aux);
2782 void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, u32 func_id);
2783 int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr);
2784 int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, struct bpf_prog *old_prog);
2785 void bpf_dev_bound_netdev_unregister(struct net_device *dev);
2787 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
2789 return aux->dev_bound;
2792 static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux)
2794 return aux->offload_requested;
2797 bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs);
2799 static inline bool bpf_map_is_offloaded(struct bpf_map *map)
2801 return unlikely(map->ops == &bpf_map_offload_ops);
2804 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
2805 void bpf_map_offload_map_free(struct bpf_map *map);
2806 u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map);
2807 int bpf_prog_test_run_syscall(struct bpf_prog *prog,
2808 const union bpf_attr *kattr,
2809 union bpf_attr __user *uattr);
2811 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
2812 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
2813 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
2814 int sock_map_bpf_prog_query(const union bpf_attr *attr,
2815 union bpf_attr __user *uattr);
2817 void sock_map_unhash(struct sock *sk);
2818 void sock_map_destroy(struct sock *sk);
2819 void sock_map_close(struct sock *sk, long timeout);
2821 static inline int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
2822 struct bpf_prog_aux *prog_aux)
2827 static inline void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog,
2833 static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog,
2834 union bpf_attr *attr)
2839 static inline int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog,
2840 struct bpf_prog *old_prog)
2845 static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev)
2849 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
2854 static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux)
2859 static inline bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs)
2864 static inline bool bpf_map_is_offloaded(struct bpf_map *map)
2869 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
2871 return ERR_PTR(-EOPNOTSUPP);
2874 static inline void bpf_map_offload_map_free(struct bpf_map *map)
2878 static inline u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map)
2883 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog,
2884 const union bpf_attr *kattr,
2885 union bpf_attr __user *uattr)
2890 #ifdef CONFIG_BPF_SYSCALL
2891 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
2892 struct bpf_prog *prog)
2897 static inline int sock_map_prog_detach(const union bpf_attr *attr,
2898 enum bpf_prog_type ptype)
2903 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
2909 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr,
2910 union bpf_attr __user *uattr)
2914 #endif /* CONFIG_BPF_SYSCALL */
2915 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
2917 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
2918 void bpf_sk_reuseport_detach(struct sock *sk);
2919 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
2921 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
2922 void *value, u64 map_flags);
2924 static inline void bpf_sk_reuseport_detach(struct sock *sk)
2928 #ifdef CONFIG_BPF_SYSCALL
2929 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
2930 void *key, void *value)
2935 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
2936 void *key, void *value,
2941 #endif /* CONFIG_BPF_SYSCALL */
2942 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
2944 /* verifier prototypes for helper functions called from eBPF programs */
2945 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
2946 extern const struct bpf_func_proto bpf_map_update_elem_proto;
2947 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
2948 extern const struct bpf_func_proto bpf_map_push_elem_proto;
2949 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
2950 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
2951 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto;
2953 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
2954 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
2955 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
2956 extern const struct bpf_func_proto bpf_tail_call_proto;
2957 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
2958 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
2959 extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto;
2960 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
2961 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
2962 extern const struct bpf_func_proto bpf_get_current_comm_proto;
2963 extern const struct bpf_func_proto bpf_get_stackid_proto;
2964 extern const struct bpf_func_proto bpf_get_stack_proto;
2965 extern const struct bpf_func_proto bpf_get_task_stack_proto;
2966 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
2967 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
2968 extern const struct bpf_func_proto bpf_sock_map_update_proto;
2969 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
2970 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
2971 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
2972 extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto;
2973 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
2974 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
2975 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
2976 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
2977 extern const struct bpf_func_proto bpf_spin_lock_proto;
2978 extern const struct bpf_func_proto bpf_spin_unlock_proto;
2979 extern const struct bpf_func_proto bpf_get_local_storage_proto;
2980 extern const struct bpf_func_proto bpf_strtol_proto;
2981 extern const struct bpf_func_proto bpf_strtoul_proto;
2982 extern const struct bpf_func_proto bpf_tcp_sock_proto;
2983 extern const struct bpf_func_proto bpf_jiffies64_proto;
2984 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
2985 extern const struct bpf_func_proto bpf_event_output_data_proto;
2986 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
2987 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
2988 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
2989 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
2990 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
2991 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto;
2992 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto;
2993 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto;
2994 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
2995 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
2996 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
2997 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
2998 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
2999 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto;
3000 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto;
3001 extern const struct bpf_func_proto bpf_copy_from_user_proto;
3002 extern const struct bpf_func_proto bpf_snprintf_btf_proto;
3003 extern const struct bpf_func_proto bpf_snprintf_proto;
3004 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
3005 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
3006 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto;
3007 extern const struct bpf_func_proto bpf_sock_from_file_proto;
3008 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto;
3009 extern const struct bpf_func_proto bpf_task_storage_get_recur_proto;
3010 extern const struct bpf_func_proto bpf_task_storage_get_proto;
3011 extern const struct bpf_func_proto bpf_task_storage_delete_recur_proto;
3012 extern const struct bpf_func_proto bpf_task_storage_delete_proto;
3013 extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
3014 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto;
3015 extern const struct bpf_func_proto bpf_sk_setsockopt_proto;
3016 extern const struct bpf_func_proto bpf_sk_getsockopt_proto;
3017 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto;
3018 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto;
3019 extern const struct bpf_func_proto bpf_find_vma_proto;
3020 extern const struct bpf_func_proto bpf_loop_proto;
3021 extern const struct bpf_func_proto bpf_copy_from_user_task_proto;
3022 extern const struct bpf_func_proto bpf_set_retval_proto;
3023 extern const struct bpf_func_proto bpf_get_retval_proto;
3024 extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto;
3025 extern const struct bpf_func_proto bpf_cgrp_storage_get_proto;
3026 extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto;
3028 const struct bpf_func_proto *tracing_prog_func_proto(
3029 enum bpf_func_id func_id, const struct bpf_prog *prog);
3031 /* Shared helpers among cBPF and eBPF. */
3032 void bpf_user_rnd_init_once(void);
3033 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3034 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3036 #if defined(CONFIG_NET)
3037 bool bpf_sock_common_is_valid_access(int off, int size,
3038 enum bpf_access_type type,
3039 struct bpf_insn_access_aux *info);
3040 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3041 struct bpf_insn_access_aux *info);
3042 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3043 const struct bpf_insn *si,
3044 struct bpf_insn *insn_buf,
3045 struct bpf_prog *prog,
3047 int bpf_dynptr_from_skb_rdonly(struct sk_buff *skb, u64 flags,
3048 struct bpf_dynptr_kern *ptr);
3050 static inline bool bpf_sock_common_is_valid_access(int off, int size,
3051 enum bpf_access_type type,
3052 struct bpf_insn_access_aux *info)
3056 static inline bool bpf_sock_is_valid_access(int off, int size,
3057 enum bpf_access_type type,
3058 struct bpf_insn_access_aux *info)
3062 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3063 const struct bpf_insn *si,
3064 struct bpf_insn *insn_buf,
3065 struct bpf_prog *prog,
3070 static inline int bpf_dynptr_from_skb_rdonly(struct sk_buff *skb, u64 flags,
3071 struct bpf_dynptr_kern *ptr)
3078 struct sk_reuseport_kern {
3079 struct sk_buff *skb;
3081 struct sock *selected_sk;
3082 struct sock *migrating_sk;
3088 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3089 struct bpf_insn_access_aux *info);
3091 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3092 const struct bpf_insn *si,
3093 struct bpf_insn *insn_buf,
3094 struct bpf_prog *prog,
3097 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3098 struct bpf_insn_access_aux *info);
3100 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3101 const struct bpf_insn *si,
3102 struct bpf_insn *insn_buf,
3103 struct bpf_prog *prog,
3106 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
3107 enum bpf_access_type type,
3108 struct bpf_insn_access_aux *info)
3113 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3114 const struct bpf_insn *si,
3115 struct bpf_insn *insn_buf,
3116 struct bpf_prog *prog,
3121 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
3122 enum bpf_access_type type,
3123 struct bpf_insn_access_aux *info)
3128 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3129 const struct bpf_insn *si,
3130 struct bpf_insn *insn_buf,
3131 struct bpf_prog *prog,
3136 #endif /* CONFIG_INET */
3138 enum bpf_text_poke_type {
3143 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
3144 void *addr1, void *addr2);
3146 void bpf_arch_poke_desc_update(struct bpf_jit_poke_descriptor *poke,
3147 struct bpf_prog *new, struct bpf_prog *old);
3149 void *bpf_arch_text_copy(void *dst, void *src, size_t len);
3150 int bpf_arch_text_invalidate(void *dst, size_t len);
3153 bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
3155 #define MAX_BPRINTF_VARARGS 12
3156 #define MAX_BPRINTF_BUF 1024
3158 struct bpf_bprintf_data {
3165 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
3166 u32 num_args, struct bpf_bprintf_data *data);
3167 void bpf_bprintf_cleanup(struct bpf_bprintf_data *data);
3169 #ifdef CONFIG_BPF_LSM
3170 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype);
3171 void bpf_cgroup_atype_put(int cgroup_atype);
3173 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {}
3174 static inline void bpf_cgroup_atype_put(int cgroup_atype) {}
3175 #endif /* CONFIG_BPF_LSM */
3184 #endif /* CONFIG_KEYS */
3186 static inline bool type_is_alloc(u32 type)
3188 return type & MEM_ALLOC;
3191 static inline gfp_t bpf_memcg_flags(gfp_t flags)
3193 if (memcg_bpf_enabled())
3194 return flags | __GFP_ACCOUNT;
3198 #endif /* _LINUX_BPF_H */