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>
31 struct bpf_verifier_env;
32 struct bpf_verifier_log;
41 struct exception_table_entry;
42 struct seq_operations;
43 struct bpf_iter_aux_info;
44 struct bpf_local_storage;
45 struct bpf_local_storage_map;
49 struct bpf_func_state;
51 extern struct idr btf_idr;
52 extern spinlock_t btf_idr_lock;
53 extern struct kobject *btf_kobj;
55 typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64);
56 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data,
57 struct bpf_iter_aux_info *aux);
58 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
59 typedef unsigned int (*bpf_func_t)(const void *,
60 const struct bpf_insn *);
61 struct bpf_iter_seq_info {
62 const struct seq_operations *seq_ops;
63 bpf_iter_init_seq_priv_t init_seq_private;
64 bpf_iter_fini_seq_priv_t fini_seq_private;
68 /* map is generic key/value storage optionally accessible by eBPF programs */
70 /* funcs callable from userspace (via syscall) */
71 int (*map_alloc_check)(union bpf_attr *attr);
72 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
73 void (*map_release)(struct bpf_map *map, struct file *map_file);
74 void (*map_free)(struct bpf_map *map);
75 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
76 void (*map_release_uref)(struct bpf_map *map);
77 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
78 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr,
79 union bpf_attr __user *uattr);
80 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key,
81 void *value, u64 flags);
82 int (*map_lookup_and_delete_batch)(struct bpf_map *map,
83 const union bpf_attr *attr,
84 union bpf_attr __user *uattr);
85 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr,
86 union bpf_attr __user *uattr);
87 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr,
88 union bpf_attr __user *uattr);
90 /* funcs callable from userspace and from eBPF programs */
91 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
92 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
93 int (*map_delete_elem)(struct bpf_map *map, void *key);
94 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
95 int (*map_pop_elem)(struct bpf_map *map, void *value);
96 int (*map_peek_elem)(struct bpf_map *map, void *value);
97 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu);
99 /* funcs called by prog_array and perf_event_array map */
100 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
102 void (*map_fd_put_ptr)(void *ptr);
103 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
104 u32 (*map_fd_sys_lookup_elem)(void *ptr);
105 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
107 int (*map_check_btf)(const struct bpf_map *map,
108 const struct btf *btf,
109 const struct btf_type *key_type,
110 const struct btf_type *value_type);
112 /* Prog poke tracking helpers. */
113 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
114 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
115 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
116 struct bpf_prog *new);
118 /* Direct value access helpers. */
119 int (*map_direct_value_addr)(const struct bpf_map *map,
121 int (*map_direct_value_meta)(const struct bpf_map *map,
123 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
124 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp,
125 struct poll_table_struct *pts);
127 /* Functions called by bpf_local_storage maps */
128 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap,
129 void *owner, u32 size);
130 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap,
131 void *owner, u32 size);
132 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner);
135 int (*map_redirect)(struct bpf_map *map, u32 ifindex, u64 flags);
137 /* map_meta_equal must be implemented for maps that can be
138 * used as an inner map. It is a runtime check to ensure
139 * an inner map can be inserted to an outer map.
141 * Some properties of the inner map has been used during the
142 * verification time. When inserting an inner map at the runtime,
143 * map_meta_equal has to ensure the inserting map has the same
144 * properties that the verifier has used earlier.
146 bool (*map_meta_equal)(const struct bpf_map *meta0,
147 const struct bpf_map *meta1);
150 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env,
151 struct bpf_func_state *caller,
152 struct bpf_func_state *callee);
153 int (*map_for_each_callback)(struct bpf_map *map,
154 bpf_callback_t callback_fn,
155 void *callback_ctx, u64 flags);
157 /* BTF id of struct allocated by map_alloc */
160 /* bpf_iter info used to open a seq_file */
161 const struct bpf_iter_seq_info *iter_seq_info;
165 /* Support at most 8 pointers in a BPF map value */
166 BPF_MAP_VALUE_OFF_MAX = 8,
167 BPF_MAP_OFF_ARR_MAX = BPF_MAP_VALUE_OFF_MAX +
168 1 + /* for bpf_spin_lock */
169 1, /* for bpf_timer */
177 struct bpf_map_value_off_desc {
179 enum bpf_kptr_type type;
182 struct module *module;
183 btf_dtor_kfunc_t dtor;
188 struct bpf_map_value_off {
190 struct bpf_map_value_off_desc off[];
193 struct bpf_map_off_arr {
195 u32 field_off[BPF_MAP_OFF_ARR_MAX];
196 u8 field_sz[BPF_MAP_OFF_ARR_MAX];
200 /* The first two cachelines with read-mostly members of which some
201 * are also accessed in fast-path (e.g. ops, max_entries).
203 const struct bpf_map_ops *ops ____cacheline_aligned;
204 struct bpf_map *inner_map_meta;
205 #ifdef CONFIG_SECURITY
208 enum bpf_map_type map_type;
212 u64 map_extra; /* any per-map-type extra fields */
214 int spin_lock_off; /* >=0 valid offset, <0 error */
215 struct bpf_map_value_off *kptr_off_tab;
216 int timer_off; /* >=0 valid offset, <0 error */
220 u32 btf_value_type_id;
221 u32 btf_vmlinux_value_type_id;
223 #ifdef CONFIG_MEMCG_KMEM
224 struct mem_cgroup *memcg;
226 char name[BPF_OBJ_NAME_LEN];
227 struct bpf_map_off_arr *off_arr;
228 /* The 3rd and 4th cacheline with misc members to avoid false sharing
229 * particularly with refcounting.
231 atomic64_t refcnt ____cacheline_aligned;
233 struct work_struct work;
234 struct mutex freeze_mutex;
236 /* 'Ownership' of program-containing map is claimed by the first program
237 * that is going to use this map or by the first program which FD is
238 * stored in the map to make sure that all callers and callees have the
239 * same prog type, JITed flag and xdp_has_frags flag.
243 enum bpf_prog_type type;
248 bool frozen; /* write-once; write-protected by freeze_mutex */
251 static inline bool map_value_has_spin_lock(const struct bpf_map *map)
253 return map->spin_lock_off >= 0;
256 static inline bool map_value_has_timer(const struct bpf_map *map)
258 return map->timer_off >= 0;
261 static inline bool map_value_has_kptrs(const struct bpf_map *map)
263 return !IS_ERR_OR_NULL(map->kptr_off_tab);
266 static inline void check_and_init_map_value(struct bpf_map *map, void *dst)
268 if (unlikely(map_value_has_spin_lock(map)))
269 memset(dst + map->spin_lock_off, 0, sizeof(struct bpf_spin_lock));
270 if (unlikely(map_value_has_timer(map)))
271 memset(dst + map->timer_off, 0, sizeof(struct bpf_timer));
272 if (unlikely(map_value_has_kptrs(map))) {
273 struct bpf_map_value_off *tab = map->kptr_off_tab;
276 for (i = 0; i < tab->nr_off; i++)
277 *(u64 *)(dst + tab->off[i].offset) = 0;
281 /* copy everything but bpf_spin_lock and bpf_timer. There could be one of each. */
282 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
287 if (likely(!map->off_arr)) {
288 memcpy(dst, src, map->value_size);
292 for (i = 0; i < map->off_arr->cnt; i++) {
293 u32 next_off = map->off_arr->field_off[i];
295 memcpy(dst + curr_off, src + curr_off, next_off - curr_off);
296 curr_off += map->off_arr->field_sz[i];
298 memcpy(dst + curr_off, src + curr_off, map->value_size - curr_off);
300 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
302 void bpf_timer_cancel_and_free(void *timer);
303 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
305 struct bpf_offload_dev;
306 struct bpf_offloaded_map;
308 struct bpf_map_dev_ops {
309 int (*map_get_next_key)(struct bpf_offloaded_map *map,
310 void *key, void *next_key);
311 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
312 void *key, void *value);
313 int (*map_update_elem)(struct bpf_offloaded_map *map,
314 void *key, void *value, u64 flags);
315 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
318 struct bpf_offloaded_map {
320 struct net_device *netdev;
321 const struct bpf_map_dev_ops *dev_ops;
323 struct list_head offloads;
326 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
328 return container_of(map, struct bpf_offloaded_map, map);
331 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
333 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
336 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
338 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
339 map->ops->map_seq_show_elem;
342 int map_check_no_btf(const struct bpf_map *map,
343 const struct btf *btf,
344 const struct btf_type *key_type,
345 const struct btf_type *value_type);
347 bool bpf_map_meta_equal(const struct bpf_map *meta0,
348 const struct bpf_map *meta1);
350 extern const struct bpf_map_ops bpf_map_offload_ops;
352 /* bpf_type_flag contains a set of flags that are applicable to the values of
353 * arg_type, ret_type and reg_type. For example, a pointer value may be null,
354 * or a memory is read-only. We classify types into two categories: base types
355 * and extended types. Extended types are base types combined with a type flag.
357 * Currently there are no more than 32 base types in arg_type, ret_type and
360 #define BPF_BASE_TYPE_BITS 8
363 /* PTR may be NULL. */
364 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS),
366 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is
367 * compatible with both mutable and immutable memory.
369 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS),
371 /* MEM was "allocated" from a different helper, and cannot be mixed
372 * with regular non-MEM_ALLOC'ed MEM types.
374 MEM_ALLOC = BIT(2 + BPF_BASE_TYPE_BITS),
376 /* MEM is in user address space. */
377 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS),
379 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged
380 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In
381 * order to drop this tag, it must be passed into bpf_per_cpu_ptr()
382 * or bpf_this_cpu_ptr(), which will return the pointer corresponding
383 * to the specified cpu.
385 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS),
387 /* Indicates that the argument will be released. */
388 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS),
390 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark
391 * unreferenced and referenced kptr loaded from map value using a load
392 * instruction, so that they can only be dereferenced but not escape the
393 * BPF program into the kernel (i.e. cannot be passed as arguments to
394 * kfunc or bpf helpers).
396 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS),
398 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS),
400 /* DYNPTR points to memory local to the bpf program. */
401 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS),
403 /* DYNPTR points to a ringbuf record. */
404 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS),
406 /* Size is known at compile time. */
407 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS),
410 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1,
413 #define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF)
415 /* Max number of base types. */
416 #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS)
418 /* Max number of all types. */
419 #define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1))
421 /* function argument constraints */
423 ARG_DONTCARE = 0, /* unused argument in helper function */
425 /* the following constraints used to prototype
426 * bpf_map_lookup/update/delete_elem() functions
428 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
429 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
430 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
432 /* Used to prototype bpf_memcmp() and other functions that access data
433 * on eBPF program stack
435 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
437 ARG_CONST_SIZE, /* number of bytes accessed from memory */
438 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
440 ARG_PTR_TO_CTX, /* pointer to context */
441 ARG_ANYTHING, /* any (initialized) argument is ok */
442 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
443 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
444 ARG_PTR_TO_INT, /* pointer to int */
445 ARG_PTR_TO_LONG, /* pointer to long */
446 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
447 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
448 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */
449 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
450 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
451 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */
452 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */
453 ARG_PTR_TO_STACK, /* pointer to stack */
454 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */
455 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */
456 ARG_PTR_TO_KPTR, /* pointer to referenced kptr */
457 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */
460 /* Extended arg_types. */
461 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE,
462 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM,
463 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX,
464 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET,
465 ARG_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_ALLOC_MEM,
466 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK,
467 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID,
468 /* pointer to memory does not need to be initialized, helper function must fill
469 * all bytes or clear them in error case.
471 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | ARG_PTR_TO_MEM,
472 /* Pointer to valid memory of size known at compile time. */
473 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM,
475 /* This must be the last entry. Its purpose is to ensure the enum is
476 * wide enough to hold the higher bits reserved for bpf_type_flag.
478 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT,
480 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
482 /* type of values returned from helper functions */
483 enum bpf_return_type {
484 RET_INTEGER, /* function returns integer */
485 RET_VOID, /* function doesn't return anything */
486 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
487 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */
488 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */
489 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */
490 RET_PTR_TO_ALLOC_MEM, /* returns a pointer to dynamically allocated memory */
491 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */
492 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */
495 /* Extended ret_types. */
496 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE,
497 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET,
498 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK,
499 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON,
500 RET_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_ALLOC | RET_PTR_TO_ALLOC_MEM,
501 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_ALLOC_MEM,
502 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID,
504 /* This must be the last entry. Its purpose is to ensure the enum is
505 * wide enough to hold the higher bits reserved for bpf_type_flag.
507 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT,
509 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
511 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
512 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
513 * instructions after verifying
515 struct bpf_func_proto {
516 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
519 enum bpf_return_type ret_type;
522 enum bpf_arg_type arg1_type;
523 enum bpf_arg_type arg2_type;
524 enum bpf_arg_type arg3_type;
525 enum bpf_arg_type arg4_type;
526 enum bpf_arg_type arg5_type;
528 enum bpf_arg_type arg_type[5];
548 int *ret_btf_id; /* return value btf_id */
549 bool (*allowed)(const struct bpf_prog *prog);
552 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
553 * the first argument to eBPF programs.
554 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
558 enum bpf_access_type {
563 /* types of values stored in eBPF registers */
564 /* Pointer types represent:
567 * pointer + (u16) var
568 * pointer + (u16) var + imm
569 * if (range > 0) then [ptr, ptr + range - off) is safe to access
570 * if (id > 0) means that some 'var' was added
571 * if (off > 0) means that 'imm' was added
574 NOT_INIT = 0, /* nothing was written into register */
575 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
576 PTR_TO_CTX, /* reg points to bpf_context */
577 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
578 PTR_TO_MAP_VALUE, /* reg points to map element value */
579 PTR_TO_MAP_KEY, /* reg points to a map element key */
580 PTR_TO_STACK, /* reg == frame_pointer + offset */
581 PTR_TO_PACKET_META, /* skb->data - meta_len */
582 PTR_TO_PACKET, /* reg points to skb->data */
583 PTR_TO_PACKET_END, /* skb->data + headlen */
584 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
585 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
586 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
587 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
588 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
589 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
590 /* PTR_TO_BTF_ID points to a kernel struct that does not need
591 * to be null checked by the BPF program. This does not imply the
592 * pointer is _not_ null and in practice this can easily be a null
593 * pointer when reading pointer chains. The assumption is program
594 * context will handle null pointer dereference typically via fault
595 * handling. The verifier must keep this in mind and can make no
596 * assumptions about null or non-null when doing branch analysis.
597 * Further, when passed into helpers the helpers can not, without
598 * additional context, assume the value is non-null.
601 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
602 * been checked for null. Used primarily to inform the verifier
603 * an explicit null check is required for this struct.
605 PTR_TO_MEM, /* reg points to valid memory region */
606 PTR_TO_BUF, /* reg points to a read/write buffer */
607 PTR_TO_FUNC, /* reg points to a bpf program function */
610 /* Extended reg_types. */
611 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE,
612 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET,
613 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON,
614 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK,
615 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID,
617 /* This must be the last entry. Its purpose is to ensure the enum is
618 * wide enough to hold the higher bits reserved for bpf_type_flag.
620 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT,
622 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
624 /* The information passed from prog-specific *_is_valid_access
625 * back to the verifier.
627 struct bpf_insn_access_aux {
628 enum bpf_reg_type reg_type;
636 struct bpf_verifier_log *log; /* for verbose logs */
640 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
642 aux->ctx_field_size = size;
645 static inline bool bpf_pseudo_func(const struct bpf_insn *insn)
647 return insn->code == (BPF_LD | BPF_IMM | BPF_DW) &&
648 insn->src_reg == BPF_PSEUDO_FUNC;
651 struct bpf_prog_ops {
652 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
653 union bpf_attr __user *uattr);
656 struct bpf_verifier_ops {
657 /* return eBPF function prototype for verification */
658 const struct bpf_func_proto *
659 (*get_func_proto)(enum bpf_func_id func_id,
660 const struct bpf_prog *prog);
662 /* return true if 'size' wide access at offset 'off' within bpf_context
663 * with 'type' (read or write) is allowed
665 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
666 const struct bpf_prog *prog,
667 struct bpf_insn_access_aux *info);
668 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
669 const struct bpf_prog *prog);
670 int (*gen_ld_abs)(const struct bpf_insn *orig,
671 struct bpf_insn *insn_buf);
672 u32 (*convert_ctx_access)(enum bpf_access_type type,
673 const struct bpf_insn *src,
674 struct bpf_insn *dst,
675 struct bpf_prog *prog, u32 *target_size);
676 int (*btf_struct_access)(struct bpf_verifier_log *log,
677 const struct btf *btf,
678 const struct btf_type *t, int off, int size,
679 enum bpf_access_type atype,
680 u32 *next_btf_id, enum bpf_type_flag *flag);
683 struct bpf_prog_offload_ops {
684 /* verifier basic callbacks */
685 int (*insn_hook)(struct bpf_verifier_env *env,
686 int insn_idx, int prev_insn_idx);
687 int (*finalize)(struct bpf_verifier_env *env);
688 /* verifier optimization callbacks (called after .finalize) */
689 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
690 struct bpf_insn *insn);
691 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
692 /* program management callbacks */
693 int (*prepare)(struct bpf_prog *prog);
694 int (*translate)(struct bpf_prog *prog);
695 void (*destroy)(struct bpf_prog *prog);
698 struct bpf_prog_offload {
699 struct bpf_prog *prog;
700 struct net_device *netdev;
701 struct bpf_offload_dev *offdev;
703 struct list_head offloads;
710 enum bpf_cgroup_storage_type {
711 BPF_CGROUP_STORAGE_SHARED,
712 BPF_CGROUP_STORAGE_PERCPU,
713 __BPF_CGROUP_STORAGE_MAX
716 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
718 /* The longest tracepoint has 12 args.
719 * See include/trace/bpf_probe.h
721 #define MAX_BPF_FUNC_ARGS 12
723 /* The maximum number of arguments passed through registers
724 * a single function may have.
726 #define MAX_BPF_FUNC_REG_ARGS 5
728 struct btf_func_model {
731 u8 arg_size[MAX_BPF_FUNC_ARGS];
734 /* Restore arguments before returning from trampoline to let original function
735 * continue executing. This flag is used for fentry progs when there are no
738 #define BPF_TRAMP_F_RESTORE_REGS BIT(0)
739 /* Call original function after fentry progs, but before fexit progs.
740 * Makes sense for fentry/fexit, normal calls and indirect calls.
742 #define BPF_TRAMP_F_CALL_ORIG BIT(1)
743 /* Skip current frame and return to parent. Makes sense for fentry/fexit
744 * programs only. Should not be used with normal calls and indirect calls.
746 #define BPF_TRAMP_F_SKIP_FRAME BIT(2)
747 /* Store IP address of the caller on the trampoline stack,
748 * so it's available for trampoline's programs.
750 #define BPF_TRAMP_F_IP_ARG BIT(3)
751 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */
752 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4)
754 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
757 #define BPF_MAX_TRAMP_LINKS 38
759 struct bpf_tramp_links {
760 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS];
764 struct bpf_tramp_run_ctx;
766 /* Different use cases for BPF trampoline:
767 * 1. replace nop at the function entry (kprobe equivalent)
768 * flags = BPF_TRAMP_F_RESTORE_REGS
769 * fentry = a set of programs to run before returning from trampoline
771 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
772 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
773 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
774 * fentry = a set of program to run before calling original function
775 * fexit = a set of program to run after original function
777 * 3. replace direct call instruction anywhere in the function body
778 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
780 * fentry = a set of programs to run before returning from trampoline
781 * With flags = BPF_TRAMP_F_CALL_ORIG
782 * orig_call = original callback addr or direct function addr
783 * fentry = a set of program to run before calling original function
784 * fexit = a set of program to run after original function
786 struct bpf_tramp_image;
787 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end,
788 const struct btf_func_model *m, u32 flags,
789 struct bpf_tramp_links *tlinks,
791 /* these two functions are called from generated trampoline */
792 u64 notrace __bpf_prog_enter(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx);
793 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start, struct bpf_tramp_run_ctx *run_ctx);
794 u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx);
795 void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start,
796 struct bpf_tramp_run_ctx *run_ctx);
797 u64 notrace __bpf_prog_enter_lsm_cgroup(struct bpf_prog *prog,
798 struct bpf_tramp_run_ctx *run_ctx);
799 void notrace __bpf_prog_exit_lsm_cgroup(struct bpf_prog *prog, u64 start,
800 struct bpf_tramp_run_ctx *run_ctx);
801 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
802 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
807 char name[KSYM_NAME_LEN];
808 struct list_head lnode;
809 struct latch_tree_node tnode;
813 enum bpf_tramp_prog_type {
816 BPF_TRAMP_MODIFY_RETURN,
818 BPF_TRAMP_REPLACE, /* more than MAX */
821 struct bpf_tramp_image {
823 struct bpf_ksym ksym;
824 struct percpu_ref pcref;
829 struct work_struct work;
833 struct bpf_trampoline {
834 /* hlist for trampoline_table */
835 struct hlist_node hlist;
836 /* serializes access to fields of this trampoline */
841 struct btf_func_model model;
845 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
846 * program by replacing one of its functions. func.addr is the address
847 * of the function it replaced.
849 struct bpf_prog *extension_prog;
850 /* list of BPF programs using this trampoline */
851 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
852 /* Number of attached programs. A counter per kind. */
853 int progs_cnt[BPF_TRAMP_MAX];
854 /* Executable image of trampoline */
855 struct bpf_tramp_image *cur_image;
860 struct bpf_attach_target_info {
861 struct btf_func_model fmodel;
863 const char *tgt_name;
864 const struct btf_type *tgt_type;
867 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
869 struct bpf_dispatcher_prog {
870 struct bpf_prog *prog;
874 struct bpf_dispatcher {
875 /* dispatcher mutex */
878 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
882 struct bpf_ksym ksym;
885 static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func(
887 const struct bpf_insn *insnsi,
890 return bpf_func(ctx, insnsi);
893 #ifdef CONFIG_BPF_JIT
894 int bpf_trampoline_link_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
895 int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
896 struct bpf_trampoline *bpf_trampoline_get(u64 key,
897 struct bpf_attach_target_info *tgt_info);
898 void bpf_trampoline_put(struct bpf_trampoline *tr);
899 int arch_prepare_bpf_dispatcher(void *image, s64 *funcs, int num_funcs);
900 #define BPF_DISPATCHER_INIT(_name) { \
901 .mutex = __MUTEX_INITIALIZER(_name.mutex), \
902 .func = &_name##_func, \
909 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
913 #define DEFINE_BPF_DISPATCHER(name) \
914 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \
916 const struct bpf_insn *insnsi, \
917 bpf_func_t bpf_func) \
919 return bpf_func(ctx, insnsi); \
921 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
922 struct bpf_dispatcher bpf_dispatcher_##name = \
923 BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
924 #define DECLARE_BPF_DISPATCHER(name) \
925 unsigned int bpf_dispatcher_##name##_func( \
927 const struct bpf_insn *insnsi, \
928 bpf_func_t bpf_func); \
929 extern struct bpf_dispatcher bpf_dispatcher_##name;
930 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
931 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
932 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
933 struct bpf_prog *to);
934 /* Called only from JIT-enabled code, so there's no need for stubs. */
935 void *bpf_jit_alloc_exec_page(void);
936 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym);
937 void bpf_image_ksym_del(struct bpf_ksym *ksym);
938 void bpf_ksym_add(struct bpf_ksym *ksym);
939 void bpf_ksym_del(struct bpf_ksym *ksym);
940 int bpf_jit_charge_modmem(u32 size);
941 void bpf_jit_uncharge_modmem(u32 size);
942 bool bpf_prog_has_trampoline(const struct bpf_prog *prog);
944 static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
945 struct bpf_trampoline *tr)
949 static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
950 struct bpf_trampoline *tr)
954 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
955 struct bpf_attach_target_info *tgt_info)
957 return ERR_PTR(-EOPNOTSUPP);
959 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
960 #define DEFINE_BPF_DISPATCHER(name)
961 #define DECLARE_BPF_DISPATCHER(name)
962 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
963 #define BPF_DISPATCHER_PTR(name) NULL
964 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
965 struct bpf_prog *from,
966 struct bpf_prog *to) {}
967 static inline bool is_bpf_image_address(unsigned long address)
971 static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog)
977 struct bpf_func_info_aux {
982 enum bpf_jit_poke_reason {
983 BPF_POKE_REASON_TAIL_CALL,
986 /* Descriptor of pokes pointing /into/ the JITed image. */
987 struct bpf_jit_poke_descriptor {
988 void *tailcall_target;
989 void *tailcall_bypass;
998 bool tailcall_target_stable;
1004 /* reg_type info for ctx arguments */
1005 struct bpf_ctx_arg_aux {
1007 enum bpf_reg_type reg_type;
1011 struct btf_mod_pair {
1013 struct module *module;
1016 struct bpf_kfunc_desc_tab;
1018 struct bpf_prog_aux {
1027 u32 func_cnt; /* used by non-func prog as the number of func progs */
1028 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
1029 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1030 u32 ctx_arg_info_size;
1031 u32 max_rdonly_access;
1032 u32 max_rdwr_access;
1033 struct btf *attach_btf;
1034 const struct bpf_ctx_arg_aux *ctx_arg_info;
1035 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
1036 struct bpf_prog *dst_prog;
1037 struct bpf_trampoline *dst_trampoline;
1038 enum bpf_prog_type saved_dst_prog_type;
1039 enum bpf_attach_type saved_dst_attach_type;
1040 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
1041 bool offload_requested;
1042 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
1043 bool func_proto_unreliable;
1045 bool tail_call_reachable;
1047 bool use_bpf_prog_pack;
1048 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
1049 const struct btf_type *attach_func_proto;
1050 /* function name for valid attach_btf_id */
1051 const char *attach_func_name;
1052 struct bpf_prog **func;
1053 void *jit_data; /* JIT specific data. arch dependent */
1054 struct bpf_jit_poke_descriptor *poke_tab;
1055 struct bpf_kfunc_desc_tab *kfunc_tab;
1056 struct bpf_kfunc_btf_tab *kfunc_btf_tab;
1058 struct bpf_ksym ksym;
1059 const struct bpf_prog_ops *ops;
1060 struct bpf_map **used_maps;
1061 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
1062 struct btf_mod_pair *used_btfs;
1063 struct bpf_prog *prog;
1064 struct user_struct *user;
1065 u64 load_time; /* ns since boottime */
1067 int cgroup_atype; /* enum cgroup_bpf_attach_type */
1068 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1069 char name[BPF_OBJ_NAME_LEN];
1070 #ifdef CONFIG_SECURITY
1073 struct bpf_prog_offload *offload;
1075 struct bpf_func_info *func_info;
1076 struct bpf_func_info_aux *func_info_aux;
1077 /* bpf_line_info loaded from userspace. linfo->insn_off
1078 * has the xlated insn offset.
1079 * Both the main and sub prog share the same linfo.
1080 * The subprog can access its first linfo by
1081 * using the linfo_idx.
1083 struct bpf_line_info *linfo;
1084 /* jited_linfo is the jited addr of the linfo. It has a
1085 * one to one mapping to linfo:
1086 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
1087 * Both the main and sub prog share the same jited_linfo.
1088 * The subprog can access its first jited_linfo by
1089 * using the linfo_idx.
1094 /* subprog can use linfo_idx to access its first linfo and
1096 * main prog always has linfo_idx == 0
1100 struct exception_table_entry *extable;
1102 struct work_struct work;
1103 struct rcu_head rcu;
1108 u16 pages; /* Number of allocated pages */
1109 u16 jited:1, /* Is our filter JIT'ed? */
1110 jit_requested:1,/* archs need to JIT the prog */
1111 gpl_compatible:1, /* Is filter GPL compatible? */
1112 cb_access:1, /* Is control block accessed? */
1113 dst_needed:1, /* Do we need dst entry? */
1114 blinding_requested:1, /* needs constant blinding */
1115 blinded:1, /* Was blinded */
1116 is_func:1, /* program is a bpf function */
1117 kprobe_override:1, /* Do we override a kprobe? */
1118 has_callchain_buf:1, /* callchain buffer allocated? */
1119 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */
1120 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */
1121 call_get_func_ip:1, /* Do we call get_func_ip() */
1122 tstamp_type_access:1; /* Accessed __sk_buff->tstamp_type */
1123 enum bpf_prog_type type; /* Type of BPF program */
1124 enum bpf_attach_type expected_attach_type; /* For some prog types */
1125 u32 len; /* Number of filter blocks */
1126 u32 jited_len; /* Size of jited insns in bytes */
1127 u8 tag[BPF_TAG_SIZE];
1128 struct bpf_prog_stats __percpu *stats;
1129 int __percpu *active;
1130 unsigned int (*bpf_func)(const void *ctx,
1131 const struct bpf_insn *insn);
1132 struct bpf_prog_aux *aux; /* Auxiliary fields */
1133 struct sock_fprog_kern *orig_prog; /* Original BPF program */
1134 /* Instructions for interpreter */
1136 DECLARE_FLEX_ARRAY(struct sock_filter, insns);
1137 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi);
1141 struct bpf_array_aux {
1142 /* Programs with direct jumps into programs part of this array. */
1143 struct list_head poke_progs;
1144 struct bpf_map *map;
1145 struct mutex poke_mutex;
1146 struct work_struct work;
1152 enum bpf_link_type type;
1153 const struct bpf_link_ops *ops;
1154 struct bpf_prog *prog;
1155 struct work_struct work;
1158 struct bpf_link_ops {
1159 void (*release)(struct bpf_link *link);
1160 void (*dealloc)(struct bpf_link *link);
1161 int (*detach)(struct bpf_link *link);
1162 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
1163 struct bpf_prog *old_prog);
1164 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
1165 int (*fill_link_info)(const struct bpf_link *link,
1166 struct bpf_link_info *info);
1169 struct bpf_tramp_link {
1170 struct bpf_link link;
1171 struct hlist_node tramp_hlist;
1175 struct bpf_shim_tramp_link {
1176 struct bpf_tramp_link link;
1177 struct bpf_trampoline *trampoline;
1180 struct bpf_tracing_link {
1181 struct bpf_tramp_link link;
1182 enum bpf_attach_type attach_type;
1183 struct bpf_trampoline *trampoline;
1184 struct bpf_prog *tgt_prog;
1187 struct bpf_link_primer {
1188 struct bpf_link *link;
1194 struct bpf_struct_ops_value;
1197 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
1198 struct bpf_struct_ops {
1199 const struct bpf_verifier_ops *verifier_ops;
1200 int (*init)(struct btf *btf);
1201 int (*check_member)(const struct btf_type *t,
1202 const struct btf_member *member);
1203 int (*init_member)(const struct btf_type *t,
1204 const struct btf_member *member,
1205 void *kdata, const void *udata);
1206 int (*reg)(void *kdata);
1207 void (*unreg)(void *kdata);
1208 const struct btf_type *type;
1209 const struct btf_type *value_type;
1211 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
1216 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
1217 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
1218 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id);
1219 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log);
1220 bool bpf_struct_ops_get(const void *kdata);
1221 void bpf_struct_ops_put(const void *kdata);
1222 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
1224 int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks,
1225 struct bpf_tramp_link *link,
1226 const struct btf_func_model *model,
1227 void *image, void *image_end);
1228 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1230 if (owner == BPF_MODULE_OWNER)
1231 return bpf_struct_ops_get(data);
1233 return try_module_get(owner);
1235 static inline void bpf_module_put(const void *data, struct module *owner)
1237 if (owner == BPF_MODULE_OWNER)
1238 bpf_struct_ops_put(data);
1244 /* Define it here to avoid the use of forward declaration */
1245 struct bpf_dummy_ops_state {
1249 struct bpf_dummy_ops {
1250 int (*test_1)(struct bpf_dummy_ops_state *cb);
1251 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2,
1252 char a3, unsigned long a4);
1255 int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr,
1256 union bpf_attr __user *uattr);
1258 int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1260 void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog);
1262 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id)
1266 static inline void bpf_struct_ops_init(struct btf *btf,
1267 struct bpf_verifier_log *log)
1270 static inline bool bpf_try_module_get(const void *data, struct module *owner)
1272 return try_module_get(owner);
1274 static inline void bpf_module_put(const void *data, struct module *owner)
1278 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
1284 static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1289 static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog)
1298 struct bpf_array_aux *aux;
1300 char value[0] __aligned(8);
1301 void *ptrs[0] __aligned(8);
1302 void __percpu *pptrs[0] __aligned(8);
1306 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
1307 #define MAX_TAIL_CALL_CNT 33
1309 /* Maximum number of loops for bpf_loop */
1310 #define BPF_MAX_LOOPS BIT(23)
1312 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
1313 BPF_F_RDONLY_PROG | \
1317 #define BPF_MAP_CAN_READ BIT(0)
1318 #define BPF_MAP_CAN_WRITE BIT(1)
1320 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
1322 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1324 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
1327 if (access_flags & BPF_F_RDONLY_PROG)
1328 return BPF_MAP_CAN_READ;
1329 else if (access_flags & BPF_F_WRONLY_PROG)
1330 return BPF_MAP_CAN_WRITE;
1332 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
1335 static inline bool bpf_map_flags_access_ok(u32 access_flags)
1337 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
1338 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1341 struct bpf_event_entry {
1342 struct perf_event *event;
1343 struct file *perf_file;
1344 struct file *map_file;
1345 struct rcu_head rcu;
1348 static inline bool map_type_contains_progs(struct bpf_map *map)
1350 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY ||
1351 map->map_type == BPF_MAP_TYPE_DEVMAP ||
1352 map->map_type == BPF_MAP_TYPE_CPUMAP;
1355 bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp);
1356 int bpf_prog_calc_tag(struct bpf_prog *fp);
1358 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
1359 const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void);
1361 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
1362 unsigned long off, unsigned long len);
1363 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
1364 const struct bpf_insn *src,
1365 struct bpf_insn *dst,
1366 struct bpf_prog *prog,
1369 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
1370 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
1372 /* an array of programs to be executed under rcu_lock.
1375 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run);
1377 * the structure returned by bpf_prog_array_alloc() should be populated
1378 * with program pointers and the last pointer must be NULL.
1379 * The user has to keep refcnt on the program and make sure the program
1380 * is removed from the array before bpf_prog_put().
1381 * The 'struct bpf_prog_array *' should only be replaced with xchg()
1382 * since other cpus are walking the array of pointers in parallel.
1384 struct bpf_prog_array_item {
1385 struct bpf_prog *prog;
1387 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1392 struct bpf_prog_array {
1393 struct rcu_head rcu;
1394 struct bpf_prog_array_item items[];
1397 struct bpf_empty_prog_array {
1398 struct bpf_prog_array hdr;
1399 struct bpf_prog *null_prog;
1402 /* to avoid allocating empty bpf_prog_array for cgroups that
1403 * don't have bpf program attached use one global 'bpf_empty_prog_array'
1404 * It will not be modified the caller of bpf_prog_array_alloc()
1405 * (since caller requested prog_cnt == 0)
1406 * that pointer should be 'freed' by bpf_prog_array_free()
1408 extern struct bpf_empty_prog_array bpf_empty_prog_array;
1410 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
1411 void bpf_prog_array_free(struct bpf_prog_array *progs);
1412 /* Use when traversal over the bpf_prog_array uses tasks_trace rcu */
1413 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs);
1414 int bpf_prog_array_length(struct bpf_prog_array *progs);
1415 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
1416 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
1417 __u32 __user *prog_ids, u32 cnt);
1419 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
1420 struct bpf_prog *old_prog);
1421 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
1422 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
1423 struct bpf_prog *prog);
1424 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
1425 u32 *prog_ids, u32 request_cnt,
1427 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
1428 struct bpf_prog *exclude_prog,
1429 struct bpf_prog *include_prog,
1431 struct bpf_prog_array **new_array);
1433 struct bpf_run_ctx {};
1435 struct bpf_cg_run_ctx {
1436 struct bpf_run_ctx run_ctx;
1437 const struct bpf_prog_array_item *prog_item;
1441 struct bpf_trace_run_ctx {
1442 struct bpf_run_ctx run_ctx;
1446 struct bpf_tramp_run_ctx {
1447 struct bpf_run_ctx run_ctx;
1449 struct bpf_run_ctx *saved_run_ctx;
1452 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)
1454 struct bpf_run_ctx *old_ctx = NULL;
1456 #ifdef CONFIG_BPF_SYSCALL
1457 old_ctx = current->bpf_ctx;
1458 current->bpf_ctx = new_ctx;
1463 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
1465 #ifdef CONFIG_BPF_SYSCALL
1466 current->bpf_ctx = old_ctx;
1470 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
1471 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0)
1472 /* BPF program asks to set CN on the packet. */
1473 #define BPF_RET_SET_CN (1 << 0)
1475 typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx);
1477 static __always_inline u32
1478 bpf_prog_run_array(const struct bpf_prog_array *array,
1479 const void *ctx, bpf_prog_run_fn run_prog)
1481 const struct bpf_prog_array_item *item;
1482 const struct bpf_prog *prog;
1483 struct bpf_run_ctx *old_run_ctx;
1484 struct bpf_trace_run_ctx run_ctx;
1487 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held");
1489 if (unlikely(!array))
1493 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
1494 item = &array->items[0];
1495 while ((prog = READ_ONCE(item->prog))) {
1496 run_ctx.bpf_cookie = item->bpf_cookie;
1497 ret &= run_prog(prog, ctx);
1500 bpf_reset_run_ctx(old_run_ctx);
1505 /* Notes on RCU design for bpf_prog_arrays containing sleepable programs:
1507 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array
1508 * overall. As a result, we must use the bpf_prog_array_free_sleepable
1509 * in order to use the tasks_trace rcu grace period.
1511 * When a non-sleepable program is inside the array, we take the rcu read
1512 * section and disable preemption for that program alone, so it can access
1513 * rcu-protected dynamically sized maps.
1515 static __always_inline u32
1516 bpf_prog_run_array_sleepable(const struct bpf_prog_array __rcu *array_rcu,
1517 const void *ctx, bpf_prog_run_fn run_prog)
1519 const struct bpf_prog_array_item *item;
1520 const struct bpf_prog *prog;
1521 const struct bpf_prog_array *array;
1522 struct bpf_run_ctx *old_run_ctx;
1523 struct bpf_trace_run_ctx run_ctx;
1528 rcu_read_lock_trace();
1531 array = rcu_dereference_check(array_rcu, rcu_read_lock_trace_held());
1532 if (unlikely(!array))
1534 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
1535 item = &array->items[0];
1536 while ((prog = READ_ONCE(item->prog))) {
1537 if (!prog->aux->sleepable)
1540 run_ctx.bpf_cookie = item->bpf_cookie;
1541 ret &= run_prog(prog, ctx);
1544 if (!prog->aux->sleepable)
1547 bpf_reset_run_ctx(old_run_ctx);
1550 rcu_read_unlock_trace();
1554 #ifdef CONFIG_BPF_SYSCALL
1555 DECLARE_PER_CPU(int, bpf_prog_active);
1556 extern struct mutex bpf_stats_enabled_mutex;
1559 * Block execution of BPF programs attached to instrumentation (perf,
1560 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
1561 * these events can happen inside a region which holds a map bucket lock
1562 * and can deadlock on it.
1564 static inline void bpf_disable_instrumentation(void)
1567 this_cpu_inc(bpf_prog_active);
1570 static inline void bpf_enable_instrumentation(void)
1572 this_cpu_dec(bpf_prog_active);
1576 extern const struct file_operations bpf_map_fops;
1577 extern const struct file_operations bpf_prog_fops;
1578 extern const struct file_operations bpf_iter_fops;
1580 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
1581 extern const struct bpf_prog_ops _name ## _prog_ops; \
1582 extern const struct bpf_verifier_ops _name ## _verifier_ops;
1583 #define BPF_MAP_TYPE(_id, _ops) \
1584 extern const struct bpf_map_ops _ops;
1585 #define BPF_LINK_TYPE(_id, _name)
1586 #include <linux/bpf_types.h>
1587 #undef BPF_PROG_TYPE
1589 #undef BPF_LINK_TYPE
1591 extern const struct bpf_prog_ops bpf_offload_prog_ops;
1592 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
1593 extern const struct bpf_verifier_ops xdp_analyzer_ops;
1595 struct bpf_prog *bpf_prog_get(u32 ufd);
1596 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
1598 void bpf_prog_add(struct bpf_prog *prog, int i);
1599 void bpf_prog_sub(struct bpf_prog *prog, int i);
1600 void bpf_prog_inc(struct bpf_prog *prog);
1601 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
1602 void bpf_prog_put(struct bpf_prog *prog);
1604 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
1605 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
1607 struct bpf_map_value_off_desc *bpf_map_kptr_off_contains(struct bpf_map *map, u32 offset);
1608 void bpf_map_free_kptr_off_tab(struct bpf_map *map);
1609 struct bpf_map_value_off *bpf_map_copy_kptr_off_tab(const struct bpf_map *map);
1610 bool bpf_map_equal_kptr_off_tab(const struct bpf_map *map_a, const struct bpf_map *map_b);
1611 void bpf_map_free_kptrs(struct bpf_map *map, void *map_value);
1613 struct bpf_map *bpf_map_get(u32 ufd);
1614 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
1615 struct bpf_map *__bpf_map_get(struct fd f);
1616 void bpf_map_inc(struct bpf_map *map);
1617 void bpf_map_inc_with_uref(struct bpf_map *map);
1618 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
1619 void bpf_map_put_with_uref(struct bpf_map *map);
1620 void bpf_map_put(struct bpf_map *map);
1621 void *bpf_map_area_alloc(u64 size, int numa_node);
1622 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
1623 void bpf_map_area_free(void *base);
1624 bool bpf_map_write_active(const struct bpf_map *map);
1625 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
1626 int generic_map_lookup_batch(struct bpf_map *map,
1627 const union bpf_attr *attr,
1628 union bpf_attr __user *uattr);
1629 int generic_map_update_batch(struct bpf_map *map,
1630 const union bpf_attr *attr,
1631 union bpf_attr __user *uattr);
1632 int generic_map_delete_batch(struct bpf_map *map,
1633 const union bpf_attr *attr,
1634 union bpf_attr __user *uattr);
1635 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
1636 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
1638 #ifdef CONFIG_MEMCG_KMEM
1639 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
1641 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags);
1642 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
1643 size_t align, gfp_t flags);
1645 static inline void *
1646 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
1649 return kmalloc_node(size, flags, node);
1652 static inline void *
1653 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags)
1655 return kzalloc(size, flags);
1658 static inline void __percpu *
1659 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align,
1662 return __alloc_percpu_gfp(size, align, flags);
1666 extern int sysctl_unprivileged_bpf_disabled;
1668 static inline bool bpf_allow_ptr_leaks(void)
1670 return perfmon_capable();
1673 static inline bool bpf_allow_uninit_stack(void)
1675 return perfmon_capable();
1678 static inline bool bpf_allow_ptr_to_map_access(void)
1680 return perfmon_capable();
1683 static inline bool bpf_bypass_spec_v1(void)
1685 return perfmon_capable();
1688 static inline bool bpf_bypass_spec_v4(void)
1690 return perfmon_capable();
1693 int bpf_map_new_fd(struct bpf_map *map, int flags);
1694 int bpf_prog_new_fd(struct bpf_prog *prog);
1696 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1697 const struct bpf_link_ops *ops, struct bpf_prog *prog);
1698 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
1699 int bpf_link_settle(struct bpf_link_primer *primer);
1700 void bpf_link_cleanup(struct bpf_link_primer *primer);
1701 void bpf_link_inc(struct bpf_link *link);
1702 void bpf_link_put(struct bpf_link *link);
1703 int bpf_link_new_fd(struct bpf_link *link);
1704 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd);
1705 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
1706 struct bpf_link *bpf_link_get_curr_or_next(u32 *id);
1708 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
1709 int bpf_obj_get_user(const char __user *pathname, int flags);
1711 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
1712 #define DEFINE_BPF_ITER_FUNC(target, args...) \
1713 extern int bpf_iter_ ## target(args); \
1714 int __init bpf_iter_ ## target(args) { return 0; }
1716 struct bpf_iter_aux_info {
1717 struct bpf_map *map;
1720 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
1721 union bpf_iter_link_info *linfo,
1722 struct bpf_iter_aux_info *aux);
1723 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
1724 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
1725 struct seq_file *seq);
1726 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
1727 struct bpf_link_info *info);
1728 typedef const struct bpf_func_proto *
1729 (*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id,
1730 const struct bpf_prog *prog);
1732 enum bpf_iter_feature {
1733 BPF_ITER_RESCHED = BIT(0),
1736 #define BPF_ITER_CTX_ARG_MAX 2
1737 struct bpf_iter_reg {
1739 bpf_iter_attach_target_t attach_target;
1740 bpf_iter_detach_target_t detach_target;
1741 bpf_iter_show_fdinfo_t show_fdinfo;
1742 bpf_iter_fill_link_info_t fill_link_info;
1743 bpf_iter_get_func_proto_t get_func_proto;
1744 u32 ctx_arg_info_size;
1746 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
1747 const struct bpf_iter_seq_info *seq_info;
1750 struct bpf_iter_meta {
1751 __bpf_md_ptr(struct seq_file *, seq);
1756 struct bpf_iter__bpf_map_elem {
1757 __bpf_md_ptr(struct bpf_iter_meta *, meta);
1758 __bpf_md_ptr(struct bpf_map *, map);
1759 __bpf_md_ptr(void *, key);
1760 __bpf_md_ptr(void *, value);
1763 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
1764 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
1765 bool bpf_iter_prog_supported(struct bpf_prog *prog);
1766 const struct bpf_func_proto *
1767 bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
1768 int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog);
1769 int bpf_iter_new_fd(struct bpf_link *link);
1770 bool bpf_link_is_iter(struct bpf_link *link);
1771 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
1772 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
1773 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
1774 struct seq_file *seq);
1775 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
1776 struct bpf_link_info *info);
1778 int map_set_for_each_callback_args(struct bpf_verifier_env *env,
1779 struct bpf_func_state *caller,
1780 struct bpf_func_state *callee);
1782 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
1783 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
1784 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
1786 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
1789 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
1791 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
1792 void *key, void *value, u64 map_flags);
1793 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1794 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
1795 void *key, void *value, u64 map_flags);
1796 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1798 int bpf_get_file_flag(int flags);
1799 int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size,
1800 size_t actual_size);
1802 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
1803 * forced to use 'long' read/writes to try to atomically copy long counters.
1804 * Best-effort only. No barriers here, since it _will_ race with concurrent
1805 * updates from BPF programs. Called from bpf syscall and mostly used with
1806 * size 8 or 16 bytes, so ask compiler to inline it.
1808 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
1810 const long *lsrc = src;
1813 size /= sizeof(long);
1818 /* verify correctness of eBPF program */
1819 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr);
1821 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
1822 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
1825 struct btf *bpf_get_btf_vmlinux(void);
1830 struct bpf_dtab_netdev;
1831 struct bpf_cpu_map_entry;
1833 void __dev_flush(void);
1834 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
1835 struct net_device *dev_rx);
1836 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
1837 struct net_device *dev_rx);
1838 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
1839 struct bpf_map *map, bool exclude_ingress);
1840 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
1841 struct bpf_prog *xdp_prog);
1842 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
1843 struct bpf_prog *xdp_prog, struct bpf_map *map,
1844 bool exclude_ingress);
1846 void __cpu_map_flush(void);
1847 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf,
1848 struct net_device *dev_rx);
1849 int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
1850 struct sk_buff *skb);
1852 /* Return map's numa specified by userspace */
1853 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
1855 return (attr->map_flags & BPF_F_NUMA_NODE) ?
1856 attr->numa_node : NUMA_NO_NODE;
1859 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
1860 int array_map_alloc_check(union bpf_attr *attr);
1862 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
1863 union bpf_attr __user *uattr);
1864 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
1865 union bpf_attr __user *uattr);
1866 int bpf_prog_test_run_tracing(struct bpf_prog *prog,
1867 const union bpf_attr *kattr,
1868 union bpf_attr __user *uattr);
1869 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1870 const union bpf_attr *kattr,
1871 union bpf_attr __user *uattr);
1872 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
1873 const union bpf_attr *kattr,
1874 union bpf_attr __user *uattr);
1875 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
1876 const union bpf_attr *kattr,
1877 union bpf_attr __user *uattr);
1878 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
1879 const struct bpf_prog *prog,
1880 struct bpf_insn_access_aux *info);
1882 static inline bool bpf_tracing_ctx_access(int off, int size,
1883 enum bpf_access_type type)
1885 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
1887 if (type != BPF_READ)
1889 if (off % size != 0)
1894 static inline bool bpf_tracing_btf_ctx_access(int off, int size,
1895 enum bpf_access_type type,
1896 const struct bpf_prog *prog,
1897 struct bpf_insn_access_aux *info)
1899 if (!bpf_tracing_ctx_access(off, size, type))
1901 return btf_ctx_access(off, size, type, prog, info);
1904 int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf,
1905 const struct btf_type *t, int off, int size,
1906 enum bpf_access_type atype,
1907 u32 *next_btf_id, enum bpf_type_flag *flag);
1908 bool btf_struct_ids_match(struct bpf_verifier_log *log,
1909 const struct btf *btf, u32 id, int off,
1910 const struct btf *need_btf, u32 need_type_id,
1913 int btf_distill_func_proto(struct bpf_verifier_log *log,
1915 const struct btf_type *func_proto,
1916 const char *func_name,
1917 struct btf_func_model *m);
1919 struct bpf_reg_state;
1920 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog,
1921 struct bpf_reg_state *regs);
1922 int btf_check_kfunc_arg_match(struct bpf_verifier_env *env,
1923 const struct btf *btf, u32 func_id,
1924 struct bpf_reg_state *regs);
1925 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog,
1926 struct bpf_reg_state *reg);
1927 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
1928 struct btf *btf, const struct btf_type *t);
1930 struct bpf_prog *bpf_prog_by_id(u32 id);
1931 struct bpf_link *bpf_link_by_id(u32 id);
1933 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
1934 void bpf_task_storage_free(struct task_struct *task);
1935 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
1936 const struct btf_func_model *
1937 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
1938 const struct bpf_insn *insn);
1939 struct bpf_core_ctx {
1940 struct bpf_verifier_log *log;
1941 const struct btf *btf;
1944 int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo,
1945 int relo_idx, void *insn);
1947 static inline bool unprivileged_ebpf_enabled(void)
1949 return !sysctl_unprivileged_bpf_disabled;
1952 #else /* !CONFIG_BPF_SYSCALL */
1953 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
1955 return ERR_PTR(-EOPNOTSUPP);
1958 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
1959 enum bpf_prog_type type,
1962 return ERR_PTR(-EOPNOTSUPP);
1965 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
1969 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
1973 static inline void bpf_prog_put(struct bpf_prog *prog)
1977 static inline void bpf_prog_inc(struct bpf_prog *prog)
1981 static inline struct bpf_prog *__must_check
1982 bpf_prog_inc_not_zero(struct bpf_prog *prog)
1984 return ERR_PTR(-EOPNOTSUPP);
1987 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1988 const struct bpf_link_ops *ops,
1989 struct bpf_prog *prog)
1993 static inline int bpf_link_prime(struct bpf_link *link,
1994 struct bpf_link_primer *primer)
1999 static inline int bpf_link_settle(struct bpf_link_primer *primer)
2004 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
2008 static inline void bpf_link_inc(struct bpf_link *link)
2012 static inline void bpf_link_put(struct bpf_link *link)
2016 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
2021 static inline void __dev_flush(void)
2026 struct bpf_dtab_netdev;
2027 struct bpf_cpu_map_entry;
2030 int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2031 struct net_device *dev_rx)
2037 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2038 struct net_device *dev_rx)
2044 int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2045 struct bpf_map *map, bool exclude_ingress)
2052 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
2053 struct sk_buff *skb,
2054 struct bpf_prog *xdp_prog)
2060 int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2061 struct bpf_prog *xdp_prog, struct bpf_map *map,
2062 bool exclude_ingress)
2067 static inline void __cpu_map_flush(void)
2071 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
2072 struct xdp_frame *xdpf,
2073 struct net_device *dev_rx)
2078 static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2079 struct sk_buff *skb)
2084 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
2085 enum bpf_prog_type type)
2087 return ERR_PTR(-EOPNOTSUPP);
2090 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
2091 const union bpf_attr *kattr,
2092 union bpf_attr __user *uattr)
2097 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
2098 const union bpf_attr *kattr,
2099 union bpf_attr __user *uattr)
2104 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2105 const union bpf_attr *kattr,
2106 union bpf_attr __user *uattr)
2111 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2112 const union bpf_attr *kattr,
2113 union bpf_attr __user *uattr)
2118 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2119 const union bpf_attr *kattr,
2120 union bpf_attr __user *uattr)
2125 static inline void bpf_map_put(struct bpf_map *map)
2129 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
2131 return ERR_PTR(-ENOTSUPP);
2134 static inline const struct bpf_func_proto *
2135 bpf_base_func_proto(enum bpf_func_id func_id)
2140 static inline void bpf_task_storage_free(struct task_struct *task)
2144 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog)
2149 static inline const struct btf_func_model *
2150 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2151 const struct bpf_insn *insn)
2156 static inline bool unprivileged_ebpf_enabled(void)
2161 #endif /* CONFIG_BPF_SYSCALL */
2163 void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
2164 struct btf_mod_pair *used_btfs, u32 len);
2166 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
2167 enum bpf_prog_type type)
2169 return bpf_prog_get_type_dev(ufd, type, false);
2172 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
2173 struct bpf_map **used_maps, u32 len);
2175 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
2177 int bpf_prog_offload_compile(struct bpf_prog *prog);
2178 void bpf_prog_offload_destroy(struct bpf_prog *prog);
2179 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
2180 struct bpf_prog *prog);
2182 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
2184 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
2185 int bpf_map_offload_update_elem(struct bpf_map *map,
2186 void *key, void *value, u64 flags);
2187 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
2188 int bpf_map_offload_get_next_key(struct bpf_map *map,
2189 void *key, void *next_key);
2191 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
2193 struct bpf_offload_dev *
2194 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
2195 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
2196 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
2197 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
2198 struct net_device *netdev);
2199 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
2200 struct net_device *netdev);
2201 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
2203 void unpriv_ebpf_notify(int new_state);
2205 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
2206 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
2208 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
2210 return aux->offload_requested;
2213 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
2215 return unlikely(map->ops == &bpf_map_offload_ops);
2218 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
2219 void bpf_map_offload_map_free(struct bpf_map *map);
2220 int bpf_prog_test_run_syscall(struct bpf_prog *prog,
2221 const union bpf_attr *kattr,
2222 union bpf_attr __user *uattr);
2224 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
2225 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
2226 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
2227 int sock_map_bpf_prog_query(const union bpf_attr *attr,
2228 union bpf_attr __user *uattr);
2230 void sock_map_unhash(struct sock *sk);
2231 void sock_map_destroy(struct sock *sk);
2232 void sock_map_close(struct sock *sk, long timeout);
2234 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
2235 union bpf_attr *attr)
2240 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
2245 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
2250 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
2252 return ERR_PTR(-EOPNOTSUPP);
2255 static inline void bpf_map_offload_map_free(struct bpf_map *map)
2259 static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog,
2260 const union bpf_attr *kattr,
2261 union bpf_attr __user *uattr)
2266 #ifdef CONFIG_BPF_SYSCALL
2267 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
2268 struct bpf_prog *prog)
2273 static inline int sock_map_prog_detach(const union bpf_attr *attr,
2274 enum bpf_prog_type ptype)
2279 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
2285 static inline int sock_map_bpf_prog_query(const union bpf_attr *attr,
2286 union bpf_attr __user *uattr)
2290 #endif /* CONFIG_BPF_SYSCALL */
2291 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
2293 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
2294 void bpf_sk_reuseport_detach(struct sock *sk);
2295 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
2297 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
2298 void *value, u64 map_flags);
2300 static inline void bpf_sk_reuseport_detach(struct sock *sk)
2304 #ifdef CONFIG_BPF_SYSCALL
2305 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
2306 void *key, void *value)
2311 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
2312 void *key, void *value,
2317 #endif /* CONFIG_BPF_SYSCALL */
2318 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
2320 /* verifier prototypes for helper functions called from eBPF programs */
2321 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
2322 extern const struct bpf_func_proto bpf_map_update_elem_proto;
2323 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
2324 extern const struct bpf_func_proto bpf_map_push_elem_proto;
2325 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
2326 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
2327 extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto;
2329 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
2330 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
2331 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
2332 extern const struct bpf_func_proto bpf_tail_call_proto;
2333 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
2334 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
2335 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
2336 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
2337 extern const struct bpf_func_proto bpf_get_current_comm_proto;
2338 extern const struct bpf_func_proto bpf_get_stackid_proto;
2339 extern const struct bpf_func_proto bpf_get_stack_proto;
2340 extern const struct bpf_func_proto bpf_get_task_stack_proto;
2341 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
2342 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
2343 extern const struct bpf_func_proto bpf_sock_map_update_proto;
2344 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
2345 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
2346 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
2347 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
2348 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
2349 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
2350 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
2351 extern const struct bpf_func_proto bpf_spin_lock_proto;
2352 extern const struct bpf_func_proto bpf_spin_unlock_proto;
2353 extern const struct bpf_func_proto bpf_get_local_storage_proto;
2354 extern const struct bpf_func_proto bpf_strtol_proto;
2355 extern const struct bpf_func_proto bpf_strtoul_proto;
2356 extern const struct bpf_func_proto bpf_tcp_sock_proto;
2357 extern const struct bpf_func_proto bpf_jiffies64_proto;
2358 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
2359 extern const struct bpf_func_proto bpf_event_output_data_proto;
2360 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
2361 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
2362 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
2363 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
2364 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
2365 extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto;
2366 extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto;
2367 extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto;
2368 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
2369 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
2370 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
2371 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
2372 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
2373 extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto;
2374 extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto;
2375 extern const struct bpf_func_proto bpf_copy_from_user_proto;
2376 extern const struct bpf_func_proto bpf_snprintf_btf_proto;
2377 extern const struct bpf_func_proto bpf_snprintf_proto;
2378 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
2379 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
2380 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto;
2381 extern const struct bpf_func_proto bpf_sock_from_file_proto;
2382 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto;
2383 extern const struct bpf_func_proto bpf_task_storage_get_proto;
2384 extern const struct bpf_func_proto bpf_task_storage_delete_proto;
2385 extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
2386 extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto;
2387 extern const struct bpf_func_proto bpf_sk_setsockopt_proto;
2388 extern const struct bpf_func_proto bpf_sk_getsockopt_proto;
2389 extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto;
2390 extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto;
2391 extern const struct bpf_func_proto bpf_find_vma_proto;
2392 extern const struct bpf_func_proto bpf_loop_proto;
2393 extern const struct bpf_func_proto bpf_copy_from_user_task_proto;
2394 extern const struct bpf_func_proto bpf_set_retval_proto;
2395 extern const struct bpf_func_proto bpf_get_retval_proto;
2397 const struct bpf_func_proto *tracing_prog_func_proto(
2398 enum bpf_func_id func_id, const struct bpf_prog *prog);
2400 /* Shared helpers among cBPF and eBPF. */
2401 void bpf_user_rnd_init_once(void);
2402 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
2403 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
2405 #if defined(CONFIG_NET)
2406 bool bpf_sock_common_is_valid_access(int off, int size,
2407 enum bpf_access_type type,
2408 struct bpf_insn_access_aux *info);
2409 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
2410 struct bpf_insn_access_aux *info);
2411 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
2412 const struct bpf_insn *si,
2413 struct bpf_insn *insn_buf,
2414 struct bpf_prog *prog,
2417 static inline bool bpf_sock_common_is_valid_access(int off, int size,
2418 enum bpf_access_type type,
2419 struct bpf_insn_access_aux *info)
2423 static inline bool bpf_sock_is_valid_access(int off, int size,
2424 enum bpf_access_type type,
2425 struct bpf_insn_access_aux *info)
2429 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
2430 const struct bpf_insn *si,
2431 struct bpf_insn *insn_buf,
2432 struct bpf_prog *prog,
2440 struct sk_reuseport_kern {
2441 struct sk_buff *skb;
2443 struct sock *selected_sk;
2444 struct sock *migrating_sk;
2450 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
2451 struct bpf_insn_access_aux *info);
2453 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
2454 const struct bpf_insn *si,
2455 struct bpf_insn *insn_buf,
2456 struct bpf_prog *prog,
2459 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
2460 struct bpf_insn_access_aux *info);
2462 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
2463 const struct bpf_insn *si,
2464 struct bpf_insn *insn_buf,
2465 struct bpf_prog *prog,
2468 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
2469 enum bpf_access_type type,
2470 struct bpf_insn_access_aux *info)
2475 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
2476 const struct bpf_insn *si,
2477 struct bpf_insn *insn_buf,
2478 struct bpf_prog *prog,
2483 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
2484 enum bpf_access_type type,
2485 struct bpf_insn_access_aux *info)
2490 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
2491 const struct bpf_insn *si,
2492 struct bpf_insn *insn_buf,
2493 struct bpf_prog *prog,
2498 #endif /* CONFIG_INET */
2500 enum bpf_text_poke_type {
2505 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
2506 void *addr1, void *addr2);
2508 void *bpf_arch_text_copy(void *dst, void *src, size_t len);
2509 int bpf_arch_text_invalidate(void *dst, size_t len);
2512 bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
2514 #define MAX_BPRINTF_VARARGS 12
2516 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
2517 u32 **bin_buf, u32 num_args);
2518 void bpf_bprintf_cleanup(void);
2520 /* the implementation of the opaque uapi struct bpf_dynptr */
2521 struct bpf_dynptr_kern {
2523 /* Size represents the number of usable bytes of dynptr data.
2524 * If for example the offset is at 4 for a local dynptr whose data is
2525 * of type u64, the number of usable bytes is 4.
2527 * The upper 8 bits are reserved. It is as follows:
2528 * Bits 0 - 23 = size
2529 * Bits 24 - 30 = dynptr type
2530 * Bit 31 = whether dynptr is read-only
2536 enum bpf_dynptr_type {
2537 BPF_DYNPTR_TYPE_INVALID,
2538 /* Points to memory that is local to the bpf program */
2539 BPF_DYNPTR_TYPE_LOCAL,
2540 /* Underlying data is a ringbuf record */
2541 BPF_DYNPTR_TYPE_RINGBUF,
2544 void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2545 enum bpf_dynptr_type type, u32 offset, u32 size);
2546 void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr);
2547 int bpf_dynptr_check_size(u32 size);
2549 #ifdef CONFIG_BPF_LSM
2550 void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype);
2551 void bpf_cgroup_atype_put(int cgroup_atype);
2553 static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {}
2554 static inline void bpf_cgroup_atype_put(int cgroup_atype) {}
2555 #endif /* CONFIG_BPF_LSM */
2557 #endif /* _LINUX_BPF_H */