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 /* If need_defer is true, the implementation should guarantee that
110 * the to-be-put element is still alive before the bpf program, which
111 * may manipulate it, exists.
113 void (*map_fd_put_ptr)(struct bpf_map *map, void *ptr, bool need_defer);
114 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
115 u32 (*map_fd_sys_lookup_elem)(void *ptr);
116 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
118 int (*map_check_btf)(const struct bpf_map *map,
119 const struct btf *btf,
120 const struct btf_type *key_type,
121 const struct btf_type *value_type);
123 /* Prog poke tracking helpers. */
124 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
125 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
126 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
127 struct bpf_prog *new);
129 /* Direct value access helpers. */
130 int (*map_direct_value_addr)(const struct bpf_map *map,
132 int (*map_direct_value_meta)(const struct bpf_map *map,
134 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
135 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp,
136 struct poll_table_struct *pts);
138 /* Functions called by bpf_local_storage maps */
139 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap,
140 void *owner, u32 size);
141 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap,
142 void *owner, u32 size);
143 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner);
146 long (*map_redirect)(struct bpf_map *map, u64 key, u64 flags);
148 /* map_meta_equal must be implemented for maps that can be
149 * used as an inner map. It is a runtime check to ensure
150 * an inner map can be inserted to an outer map.
152 * Some properties of the inner map has been used during the
153 * verification time. When inserting an inner map at the runtime,
154 * map_meta_equal has to ensure the inserting map has the same
155 * properties that the verifier has used earlier.
157 bool (*map_meta_equal)(const struct bpf_map *meta0,
158 const struct bpf_map *meta1);
161 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env,
162 struct bpf_func_state *caller,
163 struct bpf_func_state *callee);
164 long (*map_for_each_callback)(struct bpf_map *map,
165 bpf_callback_t callback_fn,
166 void *callback_ctx, u64 flags);
168 u64 (*map_mem_usage)(const struct bpf_map *map);
170 /* BTF id of struct allocated by map_alloc */
173 /* bpf_iter info used to open a seq_file */
174 const struct bpf_iter_seq_info *iter_seq_info;
178 /* Support at most 10 fields in a BTF type */
182 enum btf_field_type {
183 BPF_SPIN_LOCK = (1 << 0),
184 BPF_TIMER = (1 << 1),
185 BPF_KPTR_UNREF = (1 << 2),
186 BPF_KPTR_REF = (1 << 3),
187 BPF_KPTR = BPF_KPTR_UNREF | BPF_KPTR_REF,
188 BPF_LIST_HEAD = (1 << 4),
189 BPF_LIST_NODE = (1 << 5),
190 BPF_RB_ROOT = (1 << 6),
191 BPF_RB_NODE = (1 << 7),
192 BPF_GRAPH_NODE_OR_ROOT = BPF_LIST_NODE | BPF_LIST_HEAD |
193 BPF_RB_NODE | BPF_RB_ROOT,
194 BPF_REFCOUNT = (1 << 8),
197 typedef void (*btf_dtor_kfunc_t)(void *);
199 struct btf_field_kptr {
201 struct module *module;
202 /* dtor used if btf_is_kernel(btf), otherwise the type is
203 * program-allocated, dtor is NULL, and __bpf_obj_drop_impl is used
205 btf_dtor_kfunc_t dtor;
209 struct btf_field_graph_root {
213 struct btf_record *value_rec;
219 enum btf_field_type type;
221 struct btf_field_kptr kptr;
222 struct btf_field_graph_root graph_root;
232 struct btf_field fields[];
235 /* Non-opaque version of bpf_rb_node in uapi/linux/bpf.h */
236 struct bpf_rb_node_kern {
237 struct rb_node rb_node;
239 } __attribute__((aligned(8)));
241 /* Non-opaque version of bpf_list_node in uapi/linux/bpf.h */
242 struct bpf_list_node_kern {
243 struct list_head list_head;
245 } __attribute__((aligned(8)));
248 /* The first two cachelines with read-mostly members of which some
249 * are also accessed in fast-path (e.g. ops, max_entries).
251 const struct bpf_map_ops *ops ____cacheline_aligned;
252 struct bpf_map *inner_map_meta;
253 #ifdef CONFIG_SECURITY
256 enum bpf_map_type map_type;
260 u64 map_extra; /* any per-map-type extra fields */
263 struct btf_record *record;
266 u32 btf_value_type_id;
267 u32 btf_vmlinux_value_type_id;
269 #ifdef CONFIG_MEMCG_KMEM
270 struct obj_cgroup *objcg;
272 char name[BPF_OBJ_NAME_LEN];
273 /* The 3rd and 4th cacheline with misc members to avoid false sharing
274 * particularly with refcounting.
276 atomic64_t refcnt ____cacheline_aligned;
278 struct work_struct work;
279 struct mutex freeze_mutex;
281 /* 'Ownership' of program-containing map is claimed by the first program
282 * that is going to use this map or by the first program which FD is
283 * stored in the map to make sure that all callers and callees have the
284 * same prog type, JITed flag and xdp_has_frags flag.
288 enum bpf_prog_type type;
293 bool frozen; /* write-once; write-protected by freeze_mutex */
294 s64 __percpu *elem_count;
297 static inline const char *btf_field_type_name(enum btf_field_type type)
301 return "bpf_spin_lock";
308 return "bpf_list_head";
310 return "bpf_list_node";
312 return "bpf_rb_root";
314 return "bpf_rb_node";
316 return "bpf_refcount";
323 static inline u32 btf_field_type_size(enum btf_field_type type)
327 return sizeof(struct bpf_spin_lock);
329 return sizeof(struct bpf_timer);
334 return sizeof(struct bpf_list_head);
336 return sizeof(struct bpf_list_node);
338 return sizeof(struct bpf_rb_root);
340 return sizeof(struct bpf_rb_node);
342 return sizeof(struct bpf_refcount);
349 static inline u32 btf_field_type_align(enum btf_field_type type)
353 return __alignof__(struct bpf_spin_lock);
355 return __alignof__(struct bpf_timer);
358 return __alignof__(u64);
360 return __alignof__(struct bpf_list_head);
362 return __alignof__(struct bpf_list_node);
364 return __alignof__(struct bpf_rb_root);
366 return __alignof__(struct bpf_rb_node);
368 return __alignof__(struct bpf_refcount);
375 static inline void bpf_obj_init_field(const struct btf_field *field, void *addr)
377 memset(addr, 0, field->size);
379 switch (field->type) {
381 refcount_set((refcount_t *)addr, 1);
384 RB_CLEAR_NODE((struct rb_node *)addr);
388 INIT_LIST_HEAD((struct list_head *)addr);
391 /* RB_ROOT_CACHED 0-inits, no need to do anything after memset */
403 static inline bool btf_record_has_field(const struct btf_record *rec, enum btf_field_type type)
405 if (IS_ERR_OR_NULL(rec))
407 return rec->field_mask & type;
410 static inline void bpf_obj_init(const struct btf_record *rec, void *obj)
414 if (IS_ERR_OR_NULL(rec))
416 for (i = 0; i < rec->cnt; i++)
417 bpf_obj_init_field(&rec->fields[i], obj + rec->fields[i].offset);
420 /* 'dst' must be a temporary buffer and should not point to memory that is being
421 * used in parallel by a bpf program or bpf syscall, otherwise the access from
422 * the bpf program or bpf syscall may be corrupted by the reinitialization,
423 * leading to weird problems. Even 'dst' is newly-allocated from bpf memory
424 * allocator, it is still possible for 'dst' to be used in parallel by a bpf
425 * program or bpf syscall.
427 static inline void check_and_init_map_value(struct bpf_map *map, void *dst)
429 bpf_obj_init(map->record, dst);
432 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
433 * forced to use 'long' read/writes to try to atomically copy long counters.
434 * Best-effort only. No barriers here, since it _will_ race with concurrent
435 * updates from BPF programs. Called from bpf syscall and mostly used with
436 * size 8 or 16 bytes, so ask compiler to inline it.
438 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
440 const long *lsrc = src;
443 size /= sizeof(long);
445 data_race(*ldst++ = *lsrc++);
448 /* copy everything but bpf_spin_lock, bpf_timer, and kptrs. There could be one of each. */
449 static inline void bpf_obj_memcpy(struct btf_record *rec,
450 void *dst, void *src, u32 size,
456 if (IS_ERR_OR_NULL(rec)) {
458 bpf_long_memcpy(dst, src, round_up(size, 8));
460 memcpy(dst, src, size);
464 for (i = 0; i < rec->cnt; i++) {
465 u32 next_off = rec->fields[i].offset;
466 u32 sz = next_off - curr_off;
468 memcpy(dst + curr_off, src + curr_off, sz);
469 curr_off += rec->fields[i].size + sz;
471 memcpy(dst + curr_off, src + curr_off, size - curr_off);
474 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
476 bpf_obj_memcpy(map->record, dst, src, map->value_size, false);
479 static inline void copy_map_value_long(struct bpf_map *map, void *dst, void *src)
481 bpf_obj_memcpy(map->record, dst, src, map->value_size, true);
484 static inline void bpf_obj_memzero(struct btf_record *rec, void *dst, u32 size)
489 if (IS_ERR_OR_NULL(rec)) {
490 memset(dst, 0, size);
494 for (i = 0; i < rec->cnt; i++) {
495 u32 next_off = rec->fields[i].offset;
496 u32 sz = next_off - curr_off;
498 memset(dst + curr_off, 0, sz);
499 curr_off += rec->fields[i].size + sz;
501 memset(dst + curr_off, 0, size - curr_off);
504 static inline void zero_map_value(struct bpf_map *map, void *dst)
506 bpf_obj_memzero(map->record, dst, map->value_size);
509 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
511 void bpf_timer_cancel_and_free(void *timer);
512 void bpf_list_head_free(const struct btf_field *field, void *list_head,
513 struct bpf_spin_lock *spin_lock);
514 void bpf_rb_root_free(const struct btf_field *field, void *rb_root,
515 struct bpf_spin_lock *spin_lock);
518 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
520 struct bpf_offload_dev;
521 struct bpf_offloaded_map;
523 struct bpf_map_dev_ops {
524 int (*map_get_next_key)(struct bpf_offloaded_map *map,
525 void *key, void *next_key);
526 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
527 void *key, void *value);
528 int (*map_update_elem)(struct bpf_offloaded_map *map,
529 void *key, void *value, u64 flags);
530 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
533 struct bpf_offloaded_map {
535 struct net_device *netdev;
536 const struct bpf_map_dev_ops *dev_ops;
538 struct list_head offloads;
541 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
543 return container_of(map, struct bpf_offloaded_map, map);
546 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
548 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
551 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
553 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
554 map->ops->map_seq_show_elem;
557 int map_check_no_btf(const struct bpf_map *map,
558 const struct btf *btf,
559 const struct btf_type *key_type,
560 const struct btf_type *value_type);
562 bool bpf_map_meta_equal(const struct bpf_map *meta0,
563 const struct bpf_map *meta1);
565 extern const struct bpf_map_ops bpf_map_offload_ops;
567 /* bpf_type_flag contains a set of flags that are applicable to the values of
568 * arg_type, ret_type and reg_type. For example, a pointer value may be null,
569 * or a memory is read-only. We classify types into two categories: base types
570 * and extended types. Extended types are base types combined with a type flag.
572 * Currently there are no more than 32 base types in arg_type, ret_type and
575 #define BPF_BASE_TYPE_BITS 8
578 /* PTR may be NULL. */
579 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS),
581 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is
582 * compatible with both mutable and immutable memory.
584 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS),
586 /* MEM points to BPF ring buffer reservation. */
587 MEM_RINGBUF = BIT(2 + BPF_BASE_TYPE_BITS),
589 /* MEM is in user address space. */
590 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS),
592 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged
593 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In
594 * order to drop this tag, it must be passed into bpf_per_cpu_ptr()
595 * or bpf_this_cpu_ptr(), which will return the pointer corresponding
596 * to the specified cpu.
598 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS),
600 /* Indicates that the argument will be released. */
601 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS),
603 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark
604 * unreferenced and referenced kptr loaded from map value using a load
605 * instruction, so that they can only be dereferenced but not escape the
606 * BPF program into the kernel (i.e. cannot be passed as arguments to
607 * kfunc or bpf helpers).
609 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS),
611 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS),
613 /* DYNPTR points to memory local to the bpf program. */
614 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS),
616 /* DYNPTR points to a kernel-produced ringbuf record. */
617 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS),
619 /* Size is known at compile time. */
620 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS),
622 /* MEM is of an allocated object of type in program BTF. This is used to
623 * tag PTR_TO_BTF_ID allocated using bpf_obj_new.
625 MEM_ALLOC = BIT(11 + BPF_BASE_TYPE_BITS),
627 /* PTR was passed from the kernel in a trusted context, and may be
628 * passed to KF_TRUSTED_ARGS kfuncs or BPF helper functions.
629 * Confusingly, this is _not_ the opposite of PTR_UNTRUSTED above.
630 * PTR_UNTRUSTED refers to a kptr that was read directly from a map
631 * without invoking bpf_kptr_xchg(). What we really need to know is
632 * whether a pointer is safe to pass to a kfunc or BPF helper function.
633 * While PTR_UNTRUSTED pointers are unsafe to pass to kfuncs and BPF
634 * helpers, they do not cover all possible instances of unsafe
635 * pointers. For example, a pointer that was obtained from walking a
636 * struct will _not_ get the PTR_UNTRUSTED type modifier, despite the
637 * fact that it may be NULL, invalid, etc. This is due to backwards
638 * compatibility requirements, as this was the behavior that was first
639 * introduced when kptrs were added. The behavior is now considered
640 * deprecated, and PTR_UNTRUSTED will eventually be removed.
642 * PTR_TRUSTED, on the other hand, is a pointer that the kernel
643 * guarantees to be valid and safe to pass to kfuncs and BPF helpers.
644 * For example, pointers passed to tracepoint arguments are considered
645 * PTR_TRUSTED, as are pointers that are passed to struct_ops
646 * callbacks. As alluded to above, pointers that are obtained from
647 * walking PTR_TRUSTED pointers are _not_ trusted. For example, if a
648 * struct task_struct *task is PTR_TRUSTED, then accessing
649 * task->last_wakee will lose the PTR_TRUSTED modifier when it's stored
650 * in a BPF register. Similarly, pointers passed to certain programs
651 * types such as kretprobes are not guaranteed to be valid, as they may
652 * for example contain an object that was recently freed.
654 PTR_TRUSTED = BIT(12 + BPF_BASE_TYPE_BITS),
656 /* MEM is tagged with rcu and memory access needs rcu_read_lock protection. */
657 MEM_RCU = BIT(13 + BPF_BASE_TYPE_BITS),
659 /* Used to tag PTR_TO_BTF_ID | MEM_ALLOC references which are non-owning.
660 * Currently only valid for linked-list and rbtree nodes. If the nodes
661 * have a bpf_refcount_field, they must be tagged MEM_RCU as well.
663 NON_OWN_REF = BIT(14 + BPF_BASE_TYPE_BITS),
665 /* DYNPTR points to sk_buff */
666 DYNPTR_TYPE_SKB = BIT(15 + BPF_BASE_TYPE_BITS),
668 /* DYNPTR points to xdp_buff */
669 DYNPTR_TYPE_XDP = BIT(16 + BPF_BASE_TYPE_BITS),
672 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1,
675 #define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF | DYNPTR_TYPE_SKB \
678 /* Max number of base types. */
679 #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS)
681 /* Max number of all types. */
682 #define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1))
684 /* function argument constraints */
686 ARG_DONTCARE = 0, /* unused argument in helper function */
688 /* the following constraints used to prototype
689 * bpf_map_lookup/update/delete_elem() functions
691 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
692 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
693 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
695 /* Used to prototype bpf_memcmp() and other functions that access data
696 * on eBPF program stack
698 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
700 ARG_CONST_SIZE, /* number of bytes accessed from memory */
701 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
703 ARG_PTR_TO_CTX, /* pointer to context */
704 ARG_ANYTHING, /* any (initialized) argument is ok */
705 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
706 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
707 ARG_PTR_TO_INT, /* pointer to int */
708 ARG_PTR_TO_LONG, /* pointer to long */
709 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
710 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
711 ARG_PTR_TO_RINGBUF_MEM, /* pointer to dynamically reserved ringbuf memory */
712 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
713 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
714 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */
715 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */
716 ARG_PTR_TO_STACK, /* pointer to stack */
717 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */
718 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */
719 ARG_PTR_TO_KPTR, /* pointer to referenced kptr */
720 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */
723 /* Extended arg_types. */
724 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE,
725 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM,
726 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX,
727 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET,
728 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK,
729 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID,
730 /* pointer to memory does not need to be initialized, helper function must fill
731 * all bytes or clear them in error case.
733 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | ARG_PTR_TO_MEM,
734 /* Pointer to valid memory of size known at compile time. */
735 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM,
737 /* This must be the last entry. Its purpose is to ensure the enum is
738 * wide enough to hold the higher bits reserved for bpf_type_flag.
740 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT,
742 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
744 /* type of values returned from helper functions */
745 enum bpf_return_type {
746 RET_INTEGER, /* function returns integer */
747 RET_VOID, /* function doesn't return anything */
748 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
749 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */
750 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */
751 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */
752 RET_PTR_TO_MEM, /* returns a pointer to memory */
753 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */
754 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */
757 /* Extended ret_types. */
758 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE,
759 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET,
760 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK,
761 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON,
762 RET_PTR_TO_RINGBUF_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_RINGBUF | RET_PTR_TO_MEM,
763 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MEM,
764 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID,
765 RET_PTR_TO_BTF_ID_TRUSTED = PTR_TRUSTED | RET_PTR_TO_BTF_ID,
767 /* This must be the last entry. Its purpose is to ensure the enum is
768 * wide enough to hold the higher bits reserved for bpf_type_flag.
770 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT,
772 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
774 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
775 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
776 * instructions after verifying
778 struct bpf_func_proto {
779 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
783 enum bpf_return_type ret_type;
786 enum bpf_arg_type arg1_type;
787 enum bpf_arg_type arg2_type;
788 enum bpf_arg_type arg3_type;
789 enum bpf_arg_type arg4_type;
790 enum bpf_arg_type arg5_type;
792 enum bpf_arg_type arg_type[5];
812 int *ret_btf_id; /* return value btf_id */
813 bool (*allowed)(const struct bpf_prog *prog);
816 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
817 * the first argument to eBPF programs.
818 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
822 enum bpf_access_type {
827 /* types of values stored in eBPF registers */
828 /* Pointer types represent:
831 * pointer + (u16) var
832 * pointer + (u16) var + imm
833 * if (range > 0) then [ptr, ptr + range - off) is safe to access
834 * if (id > 0) means that some 'var' was added
835 * if (off > 0) means that 'imm' was added
838 NOT_INIT = 0, /* nothing was written into register */
839 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
840 PTR_TO_CTX, /* reg points to bpf_context */
841 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
842 PTR_TO_MAP_VALUE, /* reg points to map element value */
843 PTR_TO_MAP_KEY, /* reg points to a map element key */
844 PTR_TO_STACK, /* reg == frame_pointer + offset */
845 PTR_TO_PACKET_META, /* skb->data - meta_len */
846 PTR_TO_PACKET, /* reg points to skb->data */
847 PTR_TO_PACKET_END, /* skb->data + headlen */
848 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
849 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
850 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
851 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
852 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
853 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
854 /* PTR_TO_BTF_ID points to a kernel struct that does not need
855 * to be null checked by the BPF program. This does not imply the
856 * pointer is _not_ null and in practice this can easily be a null
857 * pointer when reading pointer chains. The assumption is program
858 * context will handle null pointer dereference typically via fault
859 * handling. The verifier must keep this in mind and can make no
860 * assumptions about null or non-null when doing branch analysis.
861 * Further, when passed into helpers the helpers can not, without
862 * additional context, assume the value is non-null.
865 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
866 * been checked for null. Used primarily to inform the verifier
867 * an explicit null check is required for this struct.
869 PTR_TO_MEM, /* reg points to valid memory region */
870 PTR_TO_BUF, /* reg points to a read/write buffer */
871 PTR_TO_FUNC, /* reg points to a bpf program function */
872 CONST_PTR_TO_DYNPTR, /* reg points to a const struct bpf_dynptr */
875 /* Extended reg_types. */
876 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE,
877 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET,
878 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON,
879 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK,
880 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID,
882 /* This must be the last entry. Its purpose is to ensure the enum is
883 * wide enough to hold the higher bits reserved for bpf_type_flag.
885 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT,
887 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
889 /* The information passed from prog-specific *_is_valid_access
890 * back to the verifier.
892 struct bpf_insn_access_aux {
893 enum bpf_reg_type reg_type;
901 struct bpf_verifier_log *log; /* for verbose logs */
905 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
907 aux->ctx_field_size = size;
910 static bool bpf_is_ldimm64(const struct bpf_insn *insn)
912 return insn->code == (BPF_LD | BPF_IMM | BPF_DW);
915 static inline bool bpf_pseudo_func(const struct bpf_insn *insn)
917 return bpf_is_ldimm64(insn) && insn->src_reg == BPF_PSEUDO_FUNC;
920 struct bpf_prog_ops {
921 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
922 union bpf_attr __user *uattr);
925 struct bpf_reg_state;
926 struct bpf_verifier_ops {
927 /* return eBPF function prototype for verification */
928 const struct bpf_func_proto *
929 (*get_func_proto)(enum bpf_func_id func_id,
930 const struct bpf_prog *prog);
932 /* return true if 'size' wide access at offset 'off' within bpf_context
933 * with 'type' (read or write) is allowed
935 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
936 const struct bpf_prog *prog,
937 struct bpf_insn_access_aux *info);
938 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
939 const struct bpf_prog *prog);
940 int (*gen_ld_abs)(const struct bpf_insn *orig,
941 struct bpf_insn *insn_buf);
942 u32 (*convert_ctx_access)(enum bpf_access_type type,
943 const struct bpf_insn *src,
944 struct bpf_insn *dst,
945 struct bpf_prog *prog, u32 *target_size);
946 int (*btf_struct_access)(struct bpf_verifier_log *log,
947 const struct bpf_reg_state *reg,
951 struct bpf_prog_offload_ops {
952 /* verifier basic callbacks */
953 int (*insn_hook)(struct bpf_verifier_env *env,
954 int insn_idx, int prev_insn_idx);
955 int (*finalize)(struct bpf_verifier_env *env);
956 /* verifier optimization callbacks (called after .finalize) */
957 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
958 struct bpf_insn *insn);
959 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
960 /* program management callbacks */
961 int (*prepare)(struct bpf_prog *prog);
962 int (*translate)(struct bpf_prog *prog);
963 void (*destroy)(struct bpf_prog *prog);
966 struct bpf_prog_offload {
967 struct bpf_prog *prog;
968 struct net_device *netdev;
969 struct bpf_offload_dev *offdev;
971 struct list_head offloads;
978 enum bpf_cgroup_storage_type {
979 BPF_CGROUP_STORAGE_SHARED,
980 BPF_CGROUP_STORAGE_PERCPU,
981 __BPF_CGROUP_STORAGE_MAX
984 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
986 /* The longest tracepoint has 12 args.
987 * See include/trace/bpf_probe.h
989 #define MAX_BPF_FUNC_ARGS 12
991 /* The maximum number of arguments passed through registers
992 * a single function may have.
994 #define MAX_BPF_FUNC_REG_ARGS 5
996 /* The argument is a structure. */
997 #define BTF_FMODEL_STRUCT_ARG BIT(0)
999 /* The argument is signed. */
1000 #define BTF_FMODEL_SIGNED_ARG BIT(1)
1002 struct btf_func_model {
1006 u8 arg_size[MAX_BPF_FUNC_ARGS];
1007 u8 arg_flags[MAX_BPF_FUNC_ARGS];
1010 /* Restore arguments before returning from trampoline to let original function
1011 * continue executing. This flag is used for fentry progs when there are no
1014 #define BPF_TRAMP_F_RESTORE_REGS BIT(0)
1015 /* Call original function after fentry progs, but before fexit progs.
1016 * Makes sense for fentry/fexit, normal calls and indirect calls.
1018 #define BPF_TRAMP_F_CALL_ORIG BIT(1)
1019 /* Skip current frame and return to parent. Makes sense for fentry/fexit
1020 * programs only. Should not be used with normal calls and indirect calls.
1022 #define BPF_TRAMP_F_SKIP_FRAME BIT(2)
1023 /* Store IP address of the caller on the trampoline stack,
1024 * so it's available for trampoline's programs.
1026 #define BPF_TRAMP_F_IP_ARG BIT(3)
1027 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */
1028 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4)
1030 /* Get original function from stack instead of from provided direct address.
1031 * Makes sense for trampolines with fexit or fmod_ret programs.
1033 #define BPF_TRAMP_F_ORIG_STACK BIT(5)
1035 /* This trampoline is on a function with another ftrace_ops with IPMODIFY,
1036 * e.g., a live patch. This flag is set and cleared by ftrace call backs,
1038 #define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6)
1040 /* Indicate that current trampoline is in a tail call context. Then, it has to
1041 * cache and restore tail_call_cnt to avoid infinite tail call loop.
1043 #define BPF_TRAMP_F_TAIL_CALL_CTX BIT(7)
1045 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
1049 #if defined(__s390x__)
1050 BPF_MAX_TRAMP_LINKS = 27,
1052 BPF_MAX_TRAMP_LINKS = 38,
1056 struct bpf_tramp_links {
1057 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS];
1061 struct bpf_tramp_run_ctx;
1063 /* Different use cases for BPF trampoline:
1064 * 1. replace nop at the function entry (kprobe equivalent)
1065 * flags = BPF_TRAMP_F_RESTORE_REGS
1066 * fentry = a set of programs to run before returning from trampoline
1068 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
1069 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
1070 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
1071 * fentry = a set of program to run before calling original function
1072 * fexit = a set of program to run after original function
1074 * 3. replace direct call instruction anywhere in the function body
1075 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
1077 * fentry = a set of programs to run before returning from trampoline
1078 * With flags = BPF_TRAMP_F_CALL_ORIG
1079 * orig_call = original callback addr or direct function addr
1080 * fentry = a set of program to run before calling original function
1081 * fexit = a set of program to run after original function
1083 struct bpf_tramp_image;
1084 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end,
1085 const struct btf_func_model *m, u32 flags,
1086 struct bpf_tramp_links *tlinks,
1088 u64 notrace __bpf_prog_enter_sleepable_recur(struct bpf_prog *prog,
1089 struct bpf_tramp_run_ctx *run_ctx);
1090 void notrace __bpf_prog_exit_sleepable_recur(struct bpf_prog *prog, u64 start,
1091 struct bpf_tramp_run_ctx *run_ctx);
1092 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
1093 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
1094 typedef u64 (*bpf_trampoline_enter_t)(struct bpf_prog *prog,
1095 struct bpf_tramp_run_ctx *run_ctx);
1096 typedef void (*bpf_trampoline_exit_t)(struct bpf_prog *prog, u64 start,
1097 struct bpf_tramp_run_ctx *run_ctx);
1098 bpf_trampoline_enter_t bpf_trampoline_enter(const struct bpf_prog *prog);
1099 bpf_trampoline_exit_t bpf_trampoline_exit(const struct bpf_prog *prog);
1102 unsigned long start;
1104 char name[KSYM_NAME_LEN];
1105 struct list_head lnode;
1106 struct latch_tree_node tnode;
1110 enum bpf_tramp_prog_type {
1113 BPF_TRAMP_MODIFY_RETURN,
1115 BPF_TRAMP_REPLACE, /* more than MAX */
1118 struct bpf_tramp_image {
1120 struct bpf_ksym ksym;
1121 struct percpu_ref pcref;
1122 void *ip_after_call;
1125 struct rcu_head rcu;
1126 struct work_struct work;
1130 struct bpf_trampoline {
1131 /* hlist for trampoline_table */
1132 struct hlist_node hlist;
1133 struct ftrace_ops *fops;
1134 /* serializes access to fields of this trampoline */
1140 struct btf_func_model model;
1142 bool ftrace_managed;
1144 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
1145 * program by replacing one of its functions. func.addr is the address
1146 * of the function it replaced.
1148 struct bpf_prog *extension_prog;
1149 /* list of BPF programs using this trampoline */
1150 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
1151 /* Number of attached programs. A counter per kind. */
1152 int progs_cnt[BPF_TRAMP_MAX];
1153 /* Executable image of trampoline */
1154 struct bpf_tramp_image *cur_image;
1158 struct bpf_attach_target_info {
1159 struct btf_func_model fmodel;
1161 struct module *tgt_mod;
1162 const char *tgt_name;
1163 const struct btf_type *tgt_type;
1166 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
1168 struct bpf_dispatcher_prog {
1169 struct bpf_prog *prog;
1173 struct bpf_dispatcher {
1174 /* dispatcher mutex */
1177 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
1182 struct bpf_ksym ksym;
1183 #ifdef CONFIG_HAVE_STATIC_CALL
1184 struct static_call_key *sc_key;
1189 static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func(
1191 const struct bpf_insn *insnsi,
1192 bpf_func_t bpf_func)
1194 return bpf_func(ctx, insnsi);
1197 /* the implementation of the opaque uapi struct bpf_dynptr */
1198 struct bpf_dynptr_kern {
1200 /* Size represents the number of usable bytes of dynptr data.
1201 * If for example the offset is at 4 for a local dynptr whose data is
1202 * of type u64, the number of usable bytes is 4.
1204 * The upper 8 bits are reserved. It is as follows:
1205 * Bits 0 - 23 = size
1206 * Bits 24 - 30 = dynptr type
1207 * Bit 31 = whether dynptr is read-only
1213 enum bpf_dynptr_type {
1214 BPF_DYNPTR_TYPE_INVALID,
1215 /* Points to memory that is local to the bpf program */
1216 BPF_DYNPTR_TYPE_LOCAL,
1217 /* Underlying data is a ringbuf record */
1218 BPF_DYNPTR_TYPE_RINGBUF,
1219 /* Underlying data is a sk_buff */
1220 BPF_DYNPTR_TYPE_SKB,
1221 /* Underlying data is a xdp_buff */
1222 BPF_DYNPTR_TYPE_XDP,
1225 int bpf_dynptr_check_size(u32 size);
1226 u32 __bpf_dynptr_size(const struct bpf_dynptr_kern *ptr);
1228 #ifdef CONFIG_BPF_JIT
1229 int bpf_trampoline_link_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
1230 int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
1231 struct bpf_trampoline *bpf_trampoline_get(u64 key,
1232 struct bpf_attach_target_info *tgt_info);
1233 void bpf_trampoline_put(struct bpf_trampoline *tr);
1234 int arch_prepare_bpf_dispatcher(void *image, void *buf, s64 *funcs, int num_funcs);
1237 * When the architecture supports STATIC_CALL replace the bpf_dispatcher_fn
1238 * indirection with a direct call to the bpf program. If the architecture does
1239 * not have STATIC_CALL, avoid a double-indirection.
1241 #ifdef CONFIG_HAVE_STATIC_CALL
1243 #define __BPF_DISPATCHER_SC_INIT(_name) \
1244 .sc_key = &STATIC_CALL_KEY(_name), \
1245 .sc_tramp = STATIC_CALL_TRAMP_ADDR(_name),
1247 #define __BPF_DISPATCHER_SC(name) \
1248 DEFINE_STATIC_CALL(bpf_dispatcher_##name##_call, bpf_dispatcher_nop_func)
1250 #define __BPF_DISPATCHER_CALL(name) \
1251 static_call(bpf_dispatcher_##name##_call)(ctx, insnsi, bpf_func)
1253 #define __BPF_DISPATCHER_UPDATE(_d, _new) \
1254 __static_call_update((_d)->sc_key, (_d)->sc_tramp, (_new))
1257 #define __BPF_DISPATCHER_SC_INIT(name)
1258 #define __BPF_DISPATCHER_SC(name)
1259 #define __BPF_DISPATCHER_CALL(name) bpf_func(ctx, insnsi)
1260 #define __BPF_DISPATCHER_UPDATE(_d, _new)
1263 #define BPF_DISPATCHER_INIT(_name) { \
1264 .mutex = __MUTEX_INITIALIZER(_name.mutex), \
1265 .func = &_name##_func, \
1272 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
1274 __BPF_DISPATCHER_SC_INIT(_name##_call) \
1277 #define DEFINE_BPF_DISPATCHER(name) \
1278 __BPF_DISPATCHER_SC(name); \
1279 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \
1281 const struct bpf_insn *insnsi, \
1282 bpf_func_t bpf_func) \
1284 return __BPF_DISPATCHER_CALL(name); \
1286 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
1287 struct bpf_dispatcher bpf_dispatcher_##name = \
1288 BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
1290 #define DECLARE_BPF_DISPATCHER(name) \
1291 unsigned int bpf_dispatcher_##name##_func( \
1293 const struct bpf_insn *insnsi, \
1294 bpf_func_t bpf_func); \
1295 extern struct bpf_dispatcher bpf_dispatcher_##name;
1297 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
1298 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
1299 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
1300 struct bpf_prog *to);
1301 /* Called only from JIT-enabled code, so there's no need for stubs. */
1302 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym);
1303 void bpf_image_ksym_del(struct bpf_ksym *ksym);
1304 void bpf_ksym_add(struct bpf_ksym *ksym);
1305 void bpf_ksym_del(struct bpf_ksym *ksym);
1306 int bpf_jit_charge_modmem(u32 size);
1307 void bpf_jit_uncharge_modmem(u32 size);
1308 bool bpf_prog_has_trampoline(const struct bpf_prog *prog);
1310 static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
1311 struct bpf_trampoline *tr)
1315 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
1316 struct bpf_trampoline *tr)
1320 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
1321 struct bpf_attach_target_info *tgt_info)
1325 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
1326 #define DEFINE_BPF_DISPATCHER(name)
1327 #define DECLARE_BPF_DISPATCHER(name)
1328 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
1329 #define BPF_DISPATCHER_PTR(name) NULL
1330 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
1331 struct bpf_prog *from,
1332 struct bpf_prog *to) {}
1333 static inline bool is_bpf_image_address(unsigned long address)
1337 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog)
1343 struct bpf_func_info_aux {
1348 enum bpf_jit_poke_reason {
1349 BPF_POKE_REASON_TAIL_CALL,
1352 /* Descriptor of pokes pointing /into/ the JITed image. */
1353 struct bpf_jit_poke_descriptor {
1354 void *tailcall_target;
1355 void *tailcall_bypass;
1360 struct bpf_map *map;
1364 bool tailcall_target_stable;
1370 /* reg_type info for ctx arguments */
1371 struct bpf_ctx_arg_aux {
1373 enum bpf_reg_type reg_type;
1377 struct btf_mod_pair {
1379 struct module *module;
1382 struct bpf_kfunc_desc_tab;
1384 struct bpf_prog_aux {
1393 u32 func_cnt; /* used by non-func prog as the number of func progs */
1394 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
1395 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1396 u32 ctx_arg_info_size;
1397 u32 max_rdonly_access;
1398 u32 max_rdwr_access;
1399 struct btf *attach_btf;
1400 const struct bpf_ctx_arg_aux *ctx_arg_info;
1401 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
1402 struct bpf_prog *dst_prog;
1403 struct bpf_trampoline *dst_trampoline;
1404 enum bpf_prog_type saved_dst_prog_type;
1405 enum bpf_attach_type saved_dst_attach_type;
1406 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
1407 bool dev_bound; /* Program is bound to the netdev. */
1408 bool offload_requested; /* Program is bound and offloaded to the netdev. */
1409 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
1410 bool func_proto_unreliable;
1412 bool tail_call_reachable;
1414 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
1415 const struct btf_type *attach_func_proto;
1416 /* function name for valid attach_btf_id */
1417 const char *attach_func_name;
1418 struct bpf_prog **func;
1419 void *jit_data; /* JIT specific data. arch dependent */
1420 struct bpf_jit_poke_descriptor *poke_tab;
1421 struct bpf_kfunc_desc_tab *kfunc_tab;
1422 struct bpf_kfunc_btf_tab *kfunc_btf_tab;
1424 struct bpf_ksym ksym;
1425 const struct bpf_prog_ops *ops;
1426 struct bpf_map **used_maps;
1427 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
1428 struct btf_mod_pair *used_btfs;
1429 struct bpf_prog *prog;
1430 struct user_struct *user;
1431 u64 load_time; /* ns since boottime */
1433 int cgroup_atype; /* enum cgroup_bpf_attach_type */
1434 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1435 char name[BPF_OBJ_NAME_LEN];
1436 #ifdef CONFIG_SECURITY
1439 struct bpf_prog_offload *offload;
1441 struct bpf_func_info *func_info;
1442 struct bpf_func_info_aux *func_info_aux;
1443 /* bpf_line_info loaded from userspace. linfo->insn_off
1444 * has the xlated insn offset.
1445 * Both the main and sub prog share the same linfo.
1446 * The subprog can access its first linfo by
1447 * using the linfo_idx.
1449 struct bpf_line_info *linfo;
1450 /* jited_linfo is the jited addr of the linfo. It has a
1451 * one to one mapping to linfo:
1452 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
1453 * Both the main and sub prog share the same jited_linfo.
1454 * The subprog can access its first jited_linfo by
1455 * using the linfo_idx.
1460 /* subprog can use linfo_idx to access its first linfo and
1462 * main prog always has linfo_idx == 0
1467 struct exception_table_entry *extable;
1469 struct work_struct work;
1470 struct rcu_head rcu;
1475 u16 pages; /* Number of allocated pages */
1476 u16 jited:1, /* Is our filter JIT'ed? */
1477 jit_requested:1,/* archs need to JIT the prog */
1478 gpl_compatible:1, /* Is filter GPL compatible? */
1479 cb_access:1, /* Is control block accessed? */
1480 dst_needed:1, /* Do we need dst entry? */
1481 blinding_requested:1, /* needs constant blinding */
1482 blinded:1, /* Was blinded */
1483 is_func:1, /* program is a bpf function */
1484 kprobe_override:1, /* Do we override a kprobe? */
1485 has_callchain_buf:1, /* callchain buffer allocated? */
1486 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */
1487 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */
1488 call_get_func_ip:1, /* Do we call get_func_ip() */
1489 tstamp_type_access:1; /* Accessed __sk_buff->tstamp_type */
1490 enum bpf_prog_type type; /* Type of BPF program */
1491 enum bpf_attach_type expected_attach_type; /* For some prog types */
1492 u32 len; /* Number of filter blocks */
1493 u32 jited_len; /* Size of jited insns in bytes */
1494 u8 tag[BPF_TAG_SIZE];
1495 struct bpf_prog_stats __percpu *stats;
1496 int __percpu *active;
1497 unsigned int (*bpf_func)(const void *ctx,
1498 const struct bpf_insn *insn);
1499 struct bpf_prog_aux *aux; /* Auxiliary fields */
1500 struct sock_fprog_kern *orig_prog; /* Original BPF program */
1501 /* Instructions for interpreter */
1503 DECLARE_FLEX_ARRAY(struct sock_filter, insns);
1504 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi);
1508 struct bpf_array_aux {
1509 /* Programs with direct jumps into programs part of this array. */
1510 struct list_head poke_progs;
1511 struct bpf_map *map;
1512 struct mutex poke_mutex;
1513 struct work_struct work;
1519 enum bpf_link_type type;
1520 const struct bpf_link_ops *ops;
1521 struct bpf_prog *prog;
1522 struct work_struct work;
1525 struct bpf_link_ops {
1526 void (*release)(struct bpf_link *link);
1527 void (*dealloc)(struct bpf_link *link);
1528 int (*detach)(struct bpf_link *link);
1529 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
1530 struct bpf_prog *old_prog);
1531 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
1532 int (*fill_link_info)(const struct bpf_link *link,
1533 struct bpf_link_info *info);
1534 int (*update_map)(struct bpf_link *link, struct bpf_map *new_map,
1535 struct bpf_map *old_map);
1538 struct bpf_tramp_link {
1539 struct bpf_link link;
1540 struct hlist_node tramp_hlist;
1544 struct bpf_shim_tramp_link {
1545 struct bpf_tramp_link link;
1546 struct bpf_trampoline *trampoline;
1549 struct bpf_tracing_link {
1550 struct bpf_tramp_link link;
1551 enum bpf_attach_type attach_type;
1552 struct bpf_trampoline *trampoline;
1553 struct bpf_prog *tgt_prog;
1556 struct bpf_link_primer {
1557 struct bpf_link *link;
1563 struct bpf_struct_ops_value;
1566 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
1568 * struct bpf_struct_ops - A structure of callbacks allowing a subsystem to
1569 * define a BPF_MAP_TYPE_STRUCT_OPS map type composed
1570 * of BPF_PROG_TYPE_STRUCT_OPS progs.
1571 * @verifier_ops: A structure of callbacks that are invoked by the verifier
1572 * when determining whether the struct_ops progs in the
1573 * struct_ops map are valid.
1574 * @init: A callback that is invoked a single time, and before any other
1575 * callback, to initialize the structure. A nonzero return value means
1576 * the subsystem could not be initialized.
1577 * @check_member: When defined, a callback invoked by the verifier to allow
1578 * the subsystem to determine if an entry in the struct_ops map
1579 * is valid. A nonzero return value means that the map is
1580 * invalid and should be rejected by the verifier.
1581 * @init_member: A callback that is invoked for each member of the struct_ops
1582 * map to allow the subsystem to initialize the member. A nonzero
1583 * value means the member could not be initialized. This callback
1584 * is exclusive with the @type, @type_id, @value_type, and
1586 * @reg: A callback that is invoked when the struct_ops map has been
1587 * initialized and is being attached to. Zero means the struct_ops map
1588 * has been successfully registered and is live. A nonzero return value
1589 * means the struct_ops map could not be registered.
1590 * @unreg: A callback that is invoked when the struct_ops map should be
1592 * @update: A callback that is invoked when the live struct_ops map is being
1593 * updated to contain new values. This callback is only invoked when
1594 * the struct_ops map is loaded with BPF_F_LINK. If not defined, the
1595 * it is assumed that the struct_ops map cannot be updated.
1596 * @validate: A callback that is invoked after all of the members have been
1597 * initialized. This callback should perform static checks on the
1598 * map, meaning that it should either fail or succeed
1599 * deterministically. A struct_ops map that has been validated may
1600 * not necessarily succeed in being registered if the call to @reg
1601 * fails. For example, a valid struct_ops map may be loaded, but
1602 * then fail to be registered due to there being another active
1603 * struct_ops map on the system in the subsystem already. For this
1604 * reason, if this callback is not defined, the check is skipped as
1605 * the struct_ops map will have final verification performed in
1608 * @value_type: Value type.
1609 * @name: The name of the struct bpf_struct_ops object.
1610 * @func_models: Func models
1611 * @type_id: BTF type id.
1612 * @value_id: BTF value id.
1614 struct bpf_struct_ops {
1615 const struct bpf_verifier_ops *verifier_ops;
1616 int (*init)(struct btf *btf);
1617 int (*check_member)(const struct btf_type *t,
1618 const struct btf_member *member,
1619 const struct bpf_prog *prog);
1620 int (*init_member)(const struct btf_type *t,
1621 const struct btf_member *member,
1622 void *kdata, const void *udata);
1623 int (*reg)(void *kdata);
1624 void (*unreg)(void *kdata);
1625 int (*update)(void *kdata, void *old_kdata);
1626 int (*validate)(void *kdata);
1627 const struct btf_type *type;
1628 const struct btf_type *value_type;
1630 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
1635 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
1636 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
1637 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id);
1638 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log);
1639 bool bpf_struct_ops_get(const void *kdata);
1640 void bpf_struct_ops_put(const void *kdata);
1641 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
1643 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks,
1644 struct bpf_tramp_link *link,
1645 const struct btf_func_model *model,
1646 void *image, void *image_end);
1647 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1649 if (owner == BPF_MODULE_OWNER)
1650 return bpf_struct_ops_get(data);
1652 return try_module_get(owner);
1654 static inline void bpf_module_put(const void *data, struct module *owner)
1656 if (owner == BPF_MODULE_OWNER)
1657 bpf_struct_ops_put(data);
1661 int bpf_struct_ops_link_create(union bpf_attr *attr);
1664 /* Define it here to avoid the use of forward declaration */
1665 struct bpf_dummy_ops_state {
1669 struct bpf_dummy_ops {
1670 int (*test_1)(struct bpf_dummy_ops_state *cb);
1671 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2,
1672 char a3, unsigned long a4);
1673 int (*test_sleepable)(struct bpf_dummy_ops_state *cb);
1676 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr,
1677 union bpf_attr __user *uattr);
1680 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id)
1684 static inline void bpf_struct_ops_init(struct btf *btf,
1685 struct bpf_verifier_log *log)
1688 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1690 return try_module_get(owner);
1692 static inline void bpf_module_put(const void *data, struct module *owner)
1696 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
1702 static inline int bpf_struct_ops_link_create(union bpf_attr *attr)
1709 #if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM)
1710 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1712 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog);
1714 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1719 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog)
1728 struct bpf_array_aux *aux;
1730 DECLARE_FLEX_ARRAY(char, value) __aligned(8);
1731 DECLARE_FLEX_ARRAY(void *, ptrs) __aligned(8);
1732 DECLARE_FLEX_ARRAY(void __percpu *, pptrs) __aligned(8);
1736 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
1737 #define MAX_TAIL_CALL_CNT 33
1739 /* Maximum number of loops for bpf_loop and bpf_iter_num.
1740 * It's enum to expose it (and thus make it discoverable) through BTF.
1743 BPF_MAX_LOOPS = 8 * 1024 * 1024,
1746 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
1747 BPF_F_RDONLY_PROG | \
1751 #define BPF_MAP_CAN_READ BIT(0)
1752 #define BPF_MAP_CAN_WRITE BIT(1)
1754 /* Maximum number of user-producer ring buffer samples that can be drained in
1755 * a call to bpf_user_ringbuf_drain().
1757 #define BPF_MAX_USER_RINGBUF_SAMPLES (128 * 1024)
1759 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
1761 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1763 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
1766 if (access_flags & BPF_F_RDONLY_PROG)
1767 return BPF_MAP_CAN_READ;
1768 else if (access_flags & BPF_F_WRONLY_PROG)
1769 return BPF_MAP_CAN_WRITE;
1771 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
1774 static inline bool bpf_map_flags_access_ok(u32 access_flags)
1776 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
1777 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1780 struct bpf_event_entry {
1781 struct perf_event *event;
1782 struct file *perf_file;
1783 struct file *map_file;
1784 struct rcu_head rcu;
1787 static inline bool map_type_contains_progs(struct bpf_map *map)
1789 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY ||
1790 map->map_type == BPF_MAP_TYPE_DEVMAP ||
1791 map->map_type == BPF_MAP_TYPE_CPUMAP;
1794 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp);
1795 int bpf_prog_calc_tag(struct bpf_prog *fp);
1797 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
1798 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void);
1800 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
1801 unsigned long off, unsigned long len);
1802 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
1803 const struct bpf_insn *src,
1804 struct bpf_insn *dst,
1805 struct bpf_prog *prog,
1808 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
1809 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
1811 /* an array of programs to be executed under rcu_lock.
1814 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run);
1816 * the structure returned by bpf_prog_array_alloc() should be populated
1817 * with program pointers and the last pointer must be NULL.
1818 * The user has to keep refcnt on the program and make sure the program
1819 * is removed from the array before bpf_prog_put().
1820 * The 'struct bpf_prog_array *' should only be replaced with xchg()
1821 * since other cpus are walking the array of pointers in parallel.
1823 struct bpf_prog_array_item {
1824 struct bpf_prog *prog;
1826 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1831 struct bpf_prog_array {
1832 struct rcu_head rcu;
1833 struct bpf_prog_array_item items[];
1836 struct bpf_empty_prog_array {
1837 struct bpf_prog_array hdr;
1838 struct bpf_prog *null_prog;
1841 /* to avoid allocating empty bpf_prog_array for cgroups that
1842 * don't have bpf program attached use one global 'bpf_empty_prog_array'
1843 * It will not be modified the caller of bpf_prog_array_alloc()
1844 * (since caller requested prog_cnt == 0)
1845 * that pointer should be 'freed' by bpf_prog_array_free()
1847 extern struct bpf_empty_prog_array bpf_empty_prog_array;
1849 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
1850 void bpf_prog_array_free(struct bpf_prog_array *progs);
1851 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */
1852 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs);
1853 int bpf_prog_array_length(struct bpf_prog_array *progs);
1854 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
1855 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
1856 __u32 __user *prog_ids, u32 cnt);
1858 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
1859 struct bpf_prog *old_prog);
1860 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
1861 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
1862 struct bpf_prog *prog);
1863 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
1864 u32 *prog_ids, u32 request_cnt,
1866 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
1867 struct bpf_prog *exclude_prog,
1868 struct bpf_prog *include_prog,
1870 struct bpf_prog_array **new_array);
1872 struct bpf_run_ctx {};
1874 struct bpf_cg_run_ctx {
1875 struct bpf_run_ctx run_ctx;
1876 const struct bpf_prog_array_item *prog_item;
1880 struct bpf_trace_run_ctx {
1881 struct bpf_run_ctx run_ctx;
1886 struct bpf_tramp_run_ctx {
1887 struct bpf_run_ctx run_ctx;
1889 struct bpf_run_ctx *saved_run_ctx;
1892 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)
1894 struct bpf_run_ctx *old_ctx = NULL;
1896 #ifdef CONFIG_BPF_SYSCALL
1897 old_ctx = current->bpf_ctx;
1898 current->bpf_ctx = new_ctx;
1903 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
1905 #ifdef CONFIG_BPF_SYSCALL
1906 current->bpf_ctx = old_ctx;
1910 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
1911 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0)
1912 /* BPF program asks to set CN on the packet. */
1913 #define BPF_RET_SET_CN (1 << 0)
1915 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx);
1917 static __always_inline u32
1918 bpf_prog_run_array(const struct bpf_prog_array *array,
1919 const void *ctx, bpf_prog_run_fn run_prog)
1921 const struct bpf_prog_array_item *item;
1922 const struct bpf_prog *prog;
1923 struct bpf_run_ctx *old_run_ctx;
1924 struct bpf_trace_run_ctx run_ctx;
1927 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held");
1929 if (unlikely(!array))
1932 run_ctx.is_uprobe = false;
1935 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
1936 item = &array->items[0];
1937 while ((prog = READ_ONCE(item->prog))) {
1938 run_ctx.bpf_cookie = item->bpf_cookie;
1939 ret &= run_prog(prog, ctx);
1942 bpf_reset_run_ctx(old_run_ctx);
1947 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs:
1949 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array
1950 * overall. As a result, we must use the bpf_prog_array_free_sleepable
1951 * in order to use the tasks_trace rcu grace period.
1953 * When a non-sleepable program is inside the array, we take the rcu read
1954 * section and disable preemption for that program alone, so it can access
1955 * rcu-protected dynamically sized maps.
1957 static __always_inline u32
1958 bpf_prog_run_array_uprobe(const struct bpf_prog_array __rcu *array_rcu,
1959 const void *ctx, bpf_prog_run_fn run_prog)
1961 const struct bpf_prog_array_item *item;
1962 const struct bpf_prog *prog;
1963 const struct bpf_prog_array *array;
1964 struct bpf_run_ctx *old_run_ctx;
1965 struct bpf_trace_run_ctx run_ctx;
1970 rcu_read_lock_trace();
1973 run_ctx.is_uprobe = true;
1975 array = rcu_dereference_check(array_rcu, rcu_read_lock_trace_held());
1976 if (unlikely(!array))
1978 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
1979 item = &array->items[0];
1980 while ((prog = READ_ONCE(item->prog))) {
1981 if (!prog->aux->sleepable)
1984 run_ctx.bpf_cookie = item->bpf_cookie;
1985 ret &= run_prog(prog, ctx);
1988 if (!prog->aux->sleepable)
1991 bpf_reset_run_ctx(old_run_ctx);
1994 rcu_read_unlock_trace();
1998 #ifdef CONFIG_BPF_SYSCALL
1999 DECLARE_PER_CPU(int, bpf_prog_active);
2000 extern struct mutex bpf_stats_enabled_mutex;
2003 * Block execution of BPF programs attached to instrumentation (perf,
2004 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
2005 * these events can happen inside a region which holds a map bucket lock
2006 * and can deadlock on it.
2008 static inline void bpf_disable_instrumentation(void)
2011 this_cpu_inc(bpf_prog_active);
2014 static inline void bpf_enable_instrumentation(void)
2016 this_cpu_dec(bpf_prog_active);
2020 extern const struct file_operations bpf_map_fops;
2021 extern const struct file_operations bpf_prog_fops;
2022 extern const struct file_operations bpf_iter_fops;
2024 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
2025 extern const struct bpf_prog_ops _name ## _prog_ops; \
2026 extern const struct bpf_verifier_ops _name ## _verifier_ops;
2027 #define BPF_MAP_TYPE(_id, _ops) \
2028 extern const struct bpf_map_ops _ops;
2029 #define BPF_LINK_TYPE(_id, _name)
2030 #include <linux/bpf_types.h>
2031 #undef BPF_PROG_TYPE
2033 #undef BPF_LINK_TYPE
2035 extern const struct bpf_prog_ops bpf_offload_prog_ops;
2036 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
2037 extern const struct bpf_verifier_ops xdp_analyzer_ops;
2039 struct bpf_prog *bpf_prog_get(u32 ufd);
2040 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
2042 void bpf_prog_add(struct bpf_prog *prog, int i);
2043 void bpf_prog_sub(struct bpf_prog *prog, int i);
2044 void bpf_prog_inc(struct bpf_prog *prog);
2045 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
2046 void bpf_prog_put(struct bpf_prog *prog);
2048 void bpf_prog_free_id(struct bpf_prog *prog);
2049 void bpf_map_free_id(struct bpf_map *map);
2051 struct btf_field *btf_record_find(const struct btf_record *rec,
2052 u32 offset, u32 field_mask);
2053 void btf_record_free(struct btf_record *rec);
2054 void bpf_map_free_record(struct bpf_map *map);
2055 struct btf_record *btf_record_dup(const struct btf_record *rec);
2056 bool btf_record_equal(const struct btf_record *rec_a, const struct btf_record *rec_b);
2057 void bpf_obj_free_timer(const struct btf_record *rec, void *obj);
2058 void bpf_obj_free_fields(const struct btf_record *rec, void *obj);
2060 struct bpf_map *bpf_map_get(u32 ufd);
2061 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
2062 struct bpf_map *__bpf_map_get(struct fd f);
2063 void bpf_map_inc(struct bpf_map *map);
2064 void bpf_map_inc_with_uref(struct bpf_map *map);
2065 struct bpf_map *__bpf_map_inc_not_zero(struct bpf_map *map, bool uref);
2066 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
2067 void bpf_map_put_with_uref(struct bpf_map *map);
2068 void bpf_map_put(struct bpf_map *map);
2069 void *bpf_map_area_alloc(u64 size, int numa_node);
2070 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
2071 void bpf_map_area_free(void *base);
2072 bool bpf_map_write_active(const struct bpf_map *map);
2073 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
2074 int generic_map_lookup_batch(struct bpf_map *map,
2075 const union bpf_attr *attr,
2076 union bpf_attr __user *uattr);
2077 int generic_map_update_batch(struct bpf_map *map, struct file *map_file,
2078 const union bpf_attr *attr,
2079 union bpf_attr __user *uattr);
2080 int generic_map_delete_batch(struct bpf_map *map,
2081 const union bpf_attr *attr,
2082 union bpf_attr __user *uattr);
2083 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
2084 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
2086 #ifdef CONFIG_MEMCG_KMEM
2087 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
2089 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags);
2090 void *bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size,
2092 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
2093 size_t align, gfp_t flags);
2095 static inline void *
2096 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
2099 return kmalloc_node(size, flags, node);
2102 static inline void *
2103 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags)
2105 return kzalloc(size, flags);
2108 static inline void *
2109 bpf_map_kvcalloc(struct bpf_map *map, size_t n, size_t size, gfp_t flags)
2111 return kvcalloc(n, size, flags);
2114 static inline void __percpu *
2115 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align,
2118 return __alloc_percpu_gfp(size, align, flags);
2123 bpf_map_init_elem_count(struct bpf_map *map)
2125 size_t size = sizeof(*map->elem_count), align = size;
2126 gfp_t flags = GFP_USER | __GFP_NOWARN;
2128 map->elem_count = bpf_map_alloc_percpu(map, size, align, flags);
2129 if (!map->elem_count)
2136 bpf_map_free_elem_count(struct bpf_map *map)
2138 free_percpu(map->elem_count);
2141 static inline void bpf_map_inc_elem_count(struct bpf_map *map)
2143 this_cpu_inc(*map->elem_count);
2146 static inline void bpf_map_dec_elem_count(struct bpf_map *map)
2148 this_cpu_dec(*map->elem_count);
2151 extern int sysctl_unprivileged_bpf_disabled;
2153 static inline bool bpf_allow_ptr_leaks(void)
2155 return perfmon_capable();
2158 static inline bool bpf_allow_uninit_stack(void)
2160 return perfmon_capable();
2163 static inline bool bpf_bypass_spec_v1(void)
2165 return perfmon_capable();
2168 static inline bool bpf_bypass_spec_v4(void)
2170 return perfmon_capable();
2173 int bpf_map_new_fd(struct bpf_map *map, int flags);
2174 int bpf_prog_new_fd(struct bpf_prog *prog);
2176 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2177 const struct bpf_link_ops *ops, struct bpf_prog *prog);
2178 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
2179 int bpf_link_settle(struct bpf_link_primer *primer);
2180 void bpf_link_cleanup(struct bpf_link_primer *primer);
2181 void bpf_link_inc(struct bpf_link *link);
2182 void bpf_link_put(struct bpf_link *link);
2183 int bpf_link_new_fd(struct bpf_link *link);
2184 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
2185 struct bpf_link *bpf_link_get_curr_or_next(u32 *id);
2187 int bpf_obj_pin_user(u32 ufd, int path_fd, const char __user *pathname);
2188 int bpf_obj_get_user(int path_fd, const char __user *pathname, int flags);
2190 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
2191 #define DEFINE_BPF_ITER_FUNC(target, args...) \
2192 extern int bpf_iter_ ## target(args); \
2193 int __init bpf_iter_ ## target(args) { return 0; }
2196 * The task type of iterators.
2198 * For BPF task iterators, they can be parameterized with various
2199 * parameters to visit only some of tasks.
2201 * BPF_TASK_ITER_ALL (default)
2202 * Iterate over resources of every task.
2205 * Iterate over resources of a task/tid.
2207 * BPF_TASK_ITER_TGID
2208 * Iterate over resources of every task of a process / task group.
2210 enum bpf_iter_task_type {
2211 BPF_TASK_ITER_ALL = 0,
2216 struct bpf_iter_aux_info {
2217 /* for map_elem iter */
2218 struct bpf_map *map;
2220 /* for cgroup iter */
2222 struct cgroup *start; /* starting cgroup */
2223 enum bpf_cgroup_iter_order order;
2226 enum bpf_iter_task_type type;
2231 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
2232 union bpf_iter_link_info *linfo,
2233 struct bpf_iter_aux_info *aux);
2234 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
2235 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
2236 struct seq_file *seq);
2237 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
2238 struct bpf_link_info *info);
2239 typedef const struct bpf_func_proto *
2240 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id,
2241 const struct bpf_prog *prog);
2243 enum bpf_iter_feature {
2244 BPF_ITER_RESCHED = BIT(0),
2247 #define BPF_ITER_CTX_ARG_MAX 2
2248 struct bpf_iter_reg {
2250 bpf_iter_attach_target_t attach_target;
2251 bpf_iter_detach_target_t detach_target;
2252 bpf_iter_show_fdinfo_t show_fdinfo;
2253 bpf_iter_fill_link_info_t fill_link_info;
2254 bpf_iter_get_func_proto_t get_func_proto;
2255 u32 ctx_arg_info_size;
2257 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
2258 const struct bpf_iter_seq_info *seq_info;
2261 struct bpf_iter_meta {
2262 __bpf_md_ptr(struct seq_file *, seq);
2267 struct bpf_iter__bpf_map_elem {
2268 __bpf_md_ptr(struct bpf_iter_meta *, meta);
2269 __bpf_md_ptr(struct bpf_map *, map);
2270 __bpf_md_ptr(void *, key);
2271 __bpf_md_ptr(void *, value);
2274 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
2275 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
2276 bool bpf_iter_prog_supported(struct bpf_prog *prog);
2277 const struct bpf_func_proto *
2278 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
2279 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog);
2280 int bpf_iter_new_fd(struct bpf_link *link);
2281 bool bpf_link_is_iter(struct bpf_link *link);
2282 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
2283 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
2284 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
2285 struct seq_file *seq);
2286 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
2287 struct bpf_link_info *info);
2289 int map_set_for_each_callback_args(struct bpf_verifier_env *env,
2290 struct bpf_func_state *caller,
2291 struct bpf_func_state *callee);
2293 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
2294 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
2295 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
2297 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
2300 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
2302 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
2303 void *key, void *value, u64 map_flags);
2304 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2305 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
2306 void *key, void *value, u64 map_flags);
2307 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
2309 int bpf_get_file_flag(int flags);
2310 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size,
2311 size_t actual_size);
2313 /* verify correctness of eBPF program */
2314 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr, u32 uattr_size);
2316 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
2317 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
2320 struct btf *bpf_get_btf_vmlinux(void);
2325 struct bpf_dtab_netdev;
2326 struct bpf_cpu_map_entry;
2328 void __dev_flush(void);
2329 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2330 struct net_device *dev_rx);
2331 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2332 struct net_device *dev_rx);
2333 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2334 struct bpf_map *map, bool exclude_ingress);
2335 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
2336 struct bpf_prog *xdp_prog);
2337 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2338 struct bpf_prog *xdp_prog, struct bpf_map *map,
2339 bool exclude_ingress);
2341 void __cpu_map_flush(void);
2342 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf,
2343 struct net_device *dev_rx);
2344 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2345 struct sk_buff *skb);
2347 /* Return map's numa specified by userspace */
2348 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
2350 return (attr->map_flags & BPF_F_NUMA_NODE) ?
2351 attr->numa_node : NUMA_NO_NODE;
2354 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
2355 int array_map_alloc_check(union bpf_attr *attr);
2357 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
2358 union bpf_attr __user *uattr);
2359 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
2360 union bpf_attr __user *uattr);
2361 int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2362 const union bpf_attr *kattr,
2363 union bpf_attr __user *uattr);
2364 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2365 const union bpf_attr *kattr,
2366 union bpf_attr __user *uattr);
2367 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
2368 const union bpf_attr *kattr,
2369 union bpf_attr __user *uattr);
2370 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2371 const union bpf_attr *kattr,
2372 union bpf_attr __user *uattr);
2373 int bpf_prog_test_run_nf(struct bpf_prog *prog,
2374 const union bpf_attr *kattr,
2375 union bpf_attr __user *uattr);
2376 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
2377 const struct bpf_prog *prog,
2378 struct bpf_insn_access_aux *info);
2380 static inline bool bpf_tracing_ctx_access(int off, int size,
2381 enum bpf_access_type type)
2383 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
2385 if (type != BPF_READ)
2387 if (off % size != 0)
2392 static inline bool bpf_tracing_btf_ctx_access(int off, int size,
2393 enum bpf_access_type type,
2394 const struct bpf_prog *prog,
2395 struct bpf_insn_access_aux *info)
2397 if (!bpf_tracing_ctx_access(off, size, type))
2399 return btf_ctx_access(off, size, type, prog, info);
2402 int btf_struct_access(struct bpf_verifier_log *log,
2403 const struct bpf_reg_state *reg,
2404 int off, int size, enum bpf_access_type atype,
2405 u32 *next_btf_id, enum bpf_type_flag *flag, const char **field_name);
2406 bool btf_struct_ids_match(struct bpf_verifier_log *log,
2407 const struct btf *btf, u32 id, int off,
2408 const struct btf *need_btf, u32 need_type_id,
2411 int btf_distill_func_proto(struct bpf_verifier_log *log,
2413 const struct btf_type *func_proto,
2414 const char *func_name,
2415 struct btf_func_model *m);
2417 struct bpf_reg_state;
2418 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog,
2419 struct bpf_reg_state *regs);
2420 int btf_check_subprog_call(struct bpf_verifier_env *env, int subprog,
2421 struct bpf_reg_state *regs);
2422 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog,
2423 struct bpf_reg_state *reg);
2424 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
2425 struct btf *btf, const struct btf_type *t);
2427 struct bpf_prog *bpf_prog_by_id(u32 id);
2428 struct bpf_link *bpf_link_by_id(u32 id);
2430 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
2431 void bpf_task_storage_free(struct task_struct *task);
2432 void bpf_cgrp_storage_free(struct cgroup *cgroup);
2433 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
2434 const struct btf_func_model *
2435 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2436 const struct bpf_insn *insn);
2437 int bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
2438 u16 btf_fd_idx, u8 **func_addr);
2440 struct bpf_core_ctx {
2441 struct bpf_verifier_log *log;
2442 const struct btf *btf;
2445 bool btf_nested_type_is_trusted(struct bpf_verifier_log *log,
2446 const struct bpf_reg_state *reg,
2447 const char *field_name, u32 btf_id, const char *suffix);
2449 bool btf_type_ids_nocast_alias(struct bpf_verifier_log *log,
2450 const struct btf *reg_btf, u32 reg_id,
2451 const struct btf *arg_btf, u32 arg_id);
2453 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo,
2454 int relo_idx, void *insn);
2456 static inline bool unprivileged_ebpf_enabled(void)
2458 return !sysctl_unprivileged_bpf_disabled;
2461 /* Not all bpf prog type has the bpf_ctx.
2462 * For the bpf prog type that has initialized the bpf_ctx,
2463 * this function can be used to decide if a kernel function
2464 * is called by a bpf program.
2466 static inline bool has_current_bpf_ctx(void)
2468 return !!current->bpf_ctx;
2471 void notrace bpf_prog_inc_misses_counter(struct bpf_prog *prog);
2473 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2474 enum bpf_dynptr_type type, u32 offset, u32 size);
2475 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr);
2476 void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr);
2477 #else /* !CONFIG_BPF_SYSCALL */
2478 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
2480 return ERR_PTR(-EOPNOTSUPP);
2483 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
2484 enum bpf_prog_type type,
2487 return ERR_PTR(-EOPNOTSUPP);
2490 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
2494 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
2498 static inline void bpf_prog_put(struct bpf_prog *prog)
2502 static inline void bpf_prog_inc(struct bpf_prog *prog)
2506 static inline struct bpf_prog *__must_check
2507 bpf_prog_inc_not_zero(struct bpf_prog *prog)
2509 return ERR_PTR(-EOPNOTSUPP);
2512 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2513 const struct bpf_link_ops *ops,
2514 struct bpf_prog *prog)
2518 static inline int bpf_link_prime(struct bpf_link *link,
2519 struct bpf_link_primer *primer)
2524 static inline int bpf_link_settle(struct bpf_link_primer *primer)
2529 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
2533 static inline void bpf_link_inc(struct bpf_link *link)
2537 static inline void bpf_link_put(struct bpf_link *link)
2541 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
2546 static inline void __dev_flush(void)
2551 struct bpf_dtab_netdev;
2552 struct bpf_cpu_map_entry;
2555 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2556 struct net_device *dev_rx)
2562 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2563 struct net_device *dev_rx)
2569 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2570 struct bpf_map *map, bool exclude_ingress)
2577 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
2578 struct sk_buff *skb,
2579 struct bpf_prog *xdp_prog)
2585 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2586 struct bpf_prog *xdp_prog, struct bpf_map *map,
2587 bool exclude_ingress)
2592 static inline void __cpu_map_flush(void)
2596 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
2597 struct xdp_frame *xdpf,
2598 struct net_device *dev_rx)
2603 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2604 struct sk_buff *skb)
2609 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
2610 enum bpf_prog_type type)
2612 return ERR_PTR(-EOPNOTSUPP);
2615 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
2616 const union bpf_attr *kattr,
2617 union bpf_attr __user *uattr)
2622 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
2623 const union bpf_attr *kattr,
2624 union bpf_attr __user *uattr)
2629 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2630 const union bpf_attr *kattr,
2631 union bpf_attr __user *uattr)
2636 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2637 const union bpf_attr *kattr,
2638 union bpf_attr __user *uattr)
2643 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2644 const union bpf_attr *kattr,
2645 union bpf_attr __user *uattr)
2650 static inline void bpf_map_put(struct bpf_map *map)
2654 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
2656 return ERR_PTR(-ENOTSUPP);
2659 static inline int btf_struct_access(struct bpf_verifier_log *log,
2660 const struct bpf_reg_state *reg,
2661 int off, int size, enum bpf_access_type atype,
2662 u32 *next_btf_id, enum bpf_type_flag *flag,
2663 const char **field_name)
2668 static inline const struct bpf_func_proto *
2669 bpf_base_func_proto(enum bpf_func_id func_id)
2674 static inline void bpf_task_storage_free(struct task_struct *task)
2678 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog)
2683 static inline const struct btf_func_model *
2684 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2685 const struct bpf_insn *insn)
2691 bpf_get_kfunc_addr(const struct bpf_prog *prog, u32 func_id,
2692 u16 btf_fd_idx, u8 **func_addr)
2697 static inline bool unprivileged_ebpf_enabled(void)
2702 static inline bool has_current_bpf_ctx(void)
2707 static inline void bpf_prog_inc_misses_counter(struct bpf_prog *prog)
2711 static inline void bpf_cgrp_storage_free(struct cgroup *cgroup)
2715 static inline void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2716 enum bpf_dynptr_type type, u32 offset, u32 size)
2720 static inline void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr)
2724 static inline void bpf_dynptr_set_rdonly(struct bpf_dynptr_kern *ptr)
2727 #endif /* CONFIG_BPF_SYSCALL */
2729 static __always_inline int
2730 bpf_probe_read_kernel_common(void *dst, u32 size, const void *unsafe_ptr)
2734 if (IS_ENABLED(CONFIG_BPF_EVENTS))
2735 ret = copy_from_kernel_nofault(dst, unsafe_ptr, size);
2736 if (unlikely(ret < 0))
2737 memset(dst, 0, size);
2741 void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
2742 struct btf_mod_pair *used_btfs, u32 len);
2744 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
2745 enum bpf_prog_type type)
2747 return bpf_prog_get_type_dev(ufd, type, false);
2750 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
2751 struct bpf_map **used_maps, u32 len);
2753 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
2755 int bpf_prog_offload_compile(struct bpf_prog *prog);
2756 void bpf_prog_dev_bound_destroy(struct bpf_prog *prog);
2757 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
2758 struct bpf_prog *prog);
2760 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
2762 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
2763 int bpf_map_offload_update_elem(struct bpf_map *map,
2764 void *key, void *value, u64 flags);
2765 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
2766 int bpf_map_offload_get_next_key(struct bpf_map *map,
2767 void *key, void *next_key);
2769 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
2771 struct bpf_offload_dev *
2772 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
2773 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
2774 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
2775 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
2776 struct net_device *netdev);
2777 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
2778 struct net_device *netdev);
2779 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
2781 void unpriv_ebpf_notify(int new_state);
2783 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
2784 int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
2785 struct bpf_prog_aux *prog_aux);
2786 void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog, u32 func_id);
2787 int bpf_prog_dev_bound_init(struct bpf_prog *prog, union bpf_attr *attr);
2788 int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog, struct bpf_prog *old_prog);
2789 void bpf_dev_bound_netdev_unregister(struct net_device *dev);
2791 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
2793 return aux->dev_bound;
2796 static inline bool bpf_prog_is_offloaded(const struct bpf_prog_aux *aux)
2798 return aux->offload_requested;
2801 bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs);
2803 static inline bool bpf_map_is_offloaded(struct bpf_map *map)
2805 return unlikely(map->ops == &bpf_map_offload_ops);
2808 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
2809 void bpf_map_offload_map_free(struct bpf_map *map);
2810 u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map);
2811 int bpf_prog_test_run_syscall(struct bpf_prog *prog,
2812 const union bpf_attr *kattr,
2813 union bpf_attr __user *uattr);
2815 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
2816 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
2817 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
2818 int sock_map_bpf_prog_query(const union bpf_attr *attr,
2819 union bpf_attr __user *uattr);
2821 void sock_map_unhash(struct sock *sk);
2822 void sock_map_destroy(struct sock *sk);
2823 void sock_map_close(struct sock *sk, long timeout);
2825 static inline int bpf_dev_bound_kfunc_check(struct bpf_verifier_log *log,
2826 struct bpf_prog_aux *prog_aux)
2831 static inline void *bpf_dev_bound_resolve_kfunc(struct bpf_prog *prog,
2837 static inline int bpf_prog_dev_bound_init(struct bpf_prog *prog,
2838 union bpf_attr *attr)
2843 static inline int bpf_prog_dev_bound_inherit(struct bpf_prog *new_prog,
2844 struct bpf_prog *old_prog)
2849 static inline void bpf_dev_bound_netdev_unregister(struct net_device *dev)
2853 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
2858 static inline bool bpf_prog_is_offloaded(struct bpf_prog_aux *aux)
2863 static inline bool bpf_prog_dev_bound_match(const struct bpf_prog *lhs, const struct bpf_prog *rhs)
2868 static inline bool bpf_map_is_offloaded(struct bpf_map *map)
2873 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
2875 return ERR_PTR(-EOPNOTSUPP);
2878 static inline void bpf_map_offload_map_free(struct bpf_map *map)
2882 static inline u64 bpf_map_offload_map_mem_usage(const struct bpf_map *map)
2887 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog,
2888 const union bpf_attr *kattr,
2889 union bpf_attr __user *uattr)
2894 #ifdef CONFIG_BPF_SYSCALL
2895 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
2896 struct bpf_prog *prog)
2901 static inline int sock_map_prog_detach(const union bpf_attr *attr,
2902 enum bpf_prog_type ptype)
2907 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
2913 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr,
2914 union bpf_attr __user *uattr)
2918 #endif /* CONFIG_BPF_SYSCALL */
2919 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
2921 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
2922 void bpf_sk_reuseport_detach(struct sock *sk);
2923 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
2925 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
2926 void *value, u64 map_flags);
2928 static inline void bpf_sk_reuseport_detach(struct sock *sk)
2932 #ifdef CONFIG_BPF_SYSCALL
2933 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
2934 void *key, void *value)
2939 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
2940 void *key, void *value,
2945 #endif /* CONFIG_BPF_SYSCALL */
2946 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
2948 /* verifier prototypes for helper functions called from eBPF programs */
2949 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
2950 extern const struct bpf_func_proto bpf_map_update_elem_proto;
2951 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
2952 extern const struct bpf_func_proto bpf_map_push_elem_proto;
2953 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
2954 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
2955 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto;
2957 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
2958 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
2959 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
2960 extern const struct bpf_func_proto bpf_tail_call_proto;
2961 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
2962 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
2963 extern const struct bpf_func_proto bpf_ktime_get_tai_ns_proto;
2964 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
2965 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
2966 extern const struct bpf_func_proto bpf_get_current_comm_proto;
2967 extern const struct bpf_func_proto bpf_get_stackid_proto;
2968 extern const struct bpf_func_proto bpf_get_stack_proto;
2969 extern const struct bpf_func_proto bpf_get_task_stack_proto;
2970 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
2971 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
2972 extern const struct bpf_func_proto bpf_sock_map_update_proto;
2973 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
2974 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
2975 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
2976 extern const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto;
2977 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
2978 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
2979 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
2980 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
2981 extern const struct bpf_func_proto bpf_spin_lock_proto;
2982 extern const struct bpf_func_proto bpf_spin_unlock_proto;
2983 extern const struct bpf_func_proto bpf_get_local_storage_proto;
2984 extern const struct bpf_func_proto bpf_strtol_proto;
2985 extern const struct bpf_func_proto bpf_strtoul_proto;
2986 extern const struct bpf_func_proto bpf_tcp_sock_proto;
2987 extern const struct bpf_func_proto bpf_jiffies64_proto;
2988 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
2989 extern const struct bpf_func_proto bpf_event_output_data_proto;
2990 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
2991 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
2992 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
2993 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
2994 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
2995 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto;
2996 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto;
2997 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto;
2998 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
2999 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
3000 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
3001 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
3002 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
3003 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto;
3004 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto;
3005 extern const struct bpf_func_proto bpf_copy_from_user_proto;
3006 extern const struct bpf_func_proto bpf_snprintf_btf_proto;
3007 extern const struct bpf_func_proto bpf_snprintf_proto;
3008 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
3009 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
3010 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto;
3011 extern const struct bpf_func_proto bpf_sock_from_file_proto;
3012 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto;
3013 extern const struct bpf_func_proto bpf_task_storage_get_recur_proto;
3014 extern const struct bpf_func_proto bpf_task_storage_get_proto;
3015 extern const struct bpf_func_proto bpf_task_storage_delete_recur_proto;
3016 extern const struct bpf_func_proto bpf_task_storage_delete_proto;
3017 extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
3018 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto;
3019 extern const struct bpf_func_proto bpf_sk_setsockopt_proto;
3020 extern const struct bpf_func_proto bpf_sk_getsockopt_proto;
3021 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto;
3022 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto;
3023 extern const struct bpf_func_proto bpf_find_vma_proto;
3024 extern const struct bpf_func_proto bpf_loop_proto;
3025 extern const struct bpf_func_proto bpf_copy_from_user_task_proto;
3026 extern const struct bpf_func_proto bpf_set_retval_proto;
3027 extern const struct bpf_func_proto bpf_get_retval_proto;
3028 extern const struct bpf_func_proto bpf_user_ringbuf_drain_proto;
3029 extern const struct bpf_func_proto bpf_cgrp_storage_get_proto;
3030 extern const struct bpf_func_proto bpf_cgrp_storage_delete_proto;
3032 const struct bpf_func_proto *tracing_prog_func_proto(
3033 enum bpf_func_id func_id, const struct bpf_prog *prog);
3035 /* Shared helpers among cBPF and eBPF. */
3036 void bpf_user_rnd_init_once(void);
3037 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3038 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
3040 #if defined(CONFIG_NET)
3041 bool bpf_sock_common_is_valid_access(int off, int size,
3042 enum bpf_access_type type,
3043 struct bpf_insn_access_aux *info);
3044 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3045 struct bpf_insn_access_aux *info);
3046 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3047 const struct bpf_insn *si,
3048 struct bpf_insn *insn_buf,
3049 struct bpf_prog *prog,
3051 int bpf_dynptr_from_skb_rdonly(struct sk_buff *skb, u64 flags,
3052 struct bpf_dynptr_kern *ptr);
3054 static inline bool bpf_sock_common_is_valid_access(int off, int size,
3055 enum bpf_access_type type,
3056 struct bpf_insn_access_aux *info)
3060 static inline bool bpf_sock_is_valid_access(int off, int size,
3061 enum bpf_access_type type,
3062 struct bpf_insn_access_aux *info)
3066 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
3067 const struct bpf_insn *si,
3068 struct bpf_insn *insn_buf,
3069 struct bpf_prog *prog,
3074 static inline int bpf_dynptr_from_skb_rdonly(struct sk_buff *skb, u64 flags,
3075 struct bpf_dynptr_kern *ptr)
3082 struct sk_reuseport_kern {
3083 struct sk_buff *skb;
3085 struct sock *selected_sk;
3086 struct sock *migrating_sk;
3092 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3093 struct bpf_insn_access_aux *info);
3095 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3096 const struct bpf_insn *si,
3097 struct bpf_insn *insn_buf,
3098 struct bpf_prog *prog,
3101 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
3102 struct bpf_insn_access_aux *info);
3104 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3105 const struct bpf_insn *si,
3106 struct bpf_insn *insn_buf,
3107 struct bpf_prog *prog,
3110 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
3111 enum bpf_access_type type,
3112 struct bpf_insn_access_aux *info)
3117 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
3118 const struct bpf_insn *si,
3119 struct bpf_insn *insn_buf,
3120 struct bpf_prog *prog,
3125 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
3126 enum bpf_access_type type,
3127 struct bpf_insn_access_aux *info)
3132 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
3133 const struct bpf_insn *si,
3134 struct bpf_insn *insn_buf,
3135 struct bpf_prog *prog,
3140 #endif /* CONFIG_INET */
3142 enum bpf_text_poke_type {
3147 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
3148 void *addr1, void *addr2);
3150 void bpf_arch_poke_desc_update(struct bpf_jit_poke_descriptor *poke,
3151 struct bpf_prog *new, struct bpf_prog *old);
3153 void *bpf_arch_text_copy(void *dst, void *src, size_t len);
3154 int bpf_arch_text_invalidate(void *dst, size_t len);
3157 bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
3159 #define MAX_BPRINTF_VARARGS 12
3160 #define MAX_BPRINTF_BUF 1024
3162 struct bpf_bprintf_data {
3169 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
3170 u32 num_args, struct bpf_bprintf_data *data);
3171 void bpf_bprintf_cleanup(struct bpf_bprintf_data *data);
3173 #ifdef CONFIG_BPF_LSM
3174 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype);
3175 void bpf_cgroup_atype_put(int cgroup_atype);
3177 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {}
3178 static inline void bpf_cgroup_atype_put(int cgroup_atype) {}
3179 #endif /* CONFIG_BPF_LSM */
3188 #endif /* CONFIG_KEYS */
3190 static inline bool type_is_alloc(u32 type)
3192 return type & MEM_ALLOC;
3195 static inline gfp_t bpf_memcg_flags(gfp_t flags)
3197 if (memcg_bpf_enabled())
3198 return flags | __GFP_ACCOUNT;
3202 #endif /* _LINUX_BPF_H */