1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
7 #include <uapi/linux/bpf.h>
9 #include <linux/workqueue.h>
10 #include <linux/file.h>
11 #include <linux/percpu.h>
12 #include <linux/err.h>
13 #include <linux/rbtree_latch.h>
14 #include <linux/numa.h>
15 #include <linux/mm_types.h>
16 #include <linux/wait.h>
17 #include <linux/u64_stats_sync.h>
18 #include <linux/refcount.h>
19 #include <linux/mutex.h>
21 struct bpf_verifier_env;
22 struct bpf_verifier_log;
31 struct exception_table_entry;
33 extern struct idr btf_idr;
34 extern spinlock_t btf_idr_lock;
36 /* map is generic key/value storage optionally accesible by eBPF programs */
38 /* funcs callable from userspace (via syscall) */
39 int (*map_alloc_check)(union bpf_attr *attr);
40 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
41 void (*map_release)(struct bpf_map *map, struct file *map_file);
42 void (*map_free)(struct bpf_map *map);
43 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
44 void (*map_release_uref)(struct bpf_map *map);
45 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
47 /* funcs callable from userspace and from eBPF programs */
48 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
49 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
50 int (*map_delete_elem)(struct bpf_map *map, void *key);
51 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
52 int (*map_pop_elem)(struct bpf_map *map, void *value);
53 int (*map_peek_elem)(struct bpf_map *map, void *value);
55 /* funcs called by prog_array and perf_event_array map */
56 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
58 void (*map_fd_put_ptr)(void *ptr);
59 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
60 u32 (*map_fd_sys_lookup_elem)(void *ptr);
61 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
63 int (*map_check_btf)(const struct bpf_map *map,
64 const struct btf *btf,
65 const struct btf_type *key_type,
66 const struct btf_type *value_type);
68 /* Prog poke tracking helpers. */
69 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
70 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
71 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
72 struct bpf_prog *new);
74 /* Direct value access helpers. */
75 int (*map_direct_value_addr)(const struct bpf_map *map,
77 int (*map_direct_value_meta)(const struct bpf_map *map,
79 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
82 struct bpf_map_memory {
84 struct user_struct *user;
88 /* The first two cachelines with read-mostly members of which some
89 * are also accessed in fast-path (e.g. ops, max_entries).
91 const struct bpf_map_ops *ops ____cacheline_aligned;
92 struct bpf_map *inner_map_meta;
93 #ifdef CONFIG_SECURITY
96 enum bpf_map_type map_type;
101 int spin_lock_off; /* >=0 valid offset, <0 error */
105 u32 btf_value_type_id;
107 struct bpf_map_memory memory;
108 char name[BPF_OBJ_NAME_LEN];
110 bool frozen; /* write-once; write-protected by freeze_mutex */
113 /* The 3rd and 4th cacheline with misc members to avoid false sharing
114 * particularly with refcounting.
116 atomic64_t refcnt ____cacheline_aligned;
118 struct work_struct work;
119 struct mutex freeze_mutex;
120 u64 writecnt; /* writable mmap cnt; protected by freeze_mutex */
123 static inline bool map_value_has_spin_lock(const struct bpf_map *map)
125 return map->spin_lock_off >= 0;
128 static inline void check_and_init_map_lock(struct bpf_map *map, void *dst)
130 if (likely(!map_value_has_spin_lock(map)))
132 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) =
133 (struct bpf_spin_lock){};
136 /* copy everything but bpf_spin_lock */
137 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
139 if (unlikely(map_value_has_spin_lock(map))) {
140 u32 off = map->spin_lock_off;
142 memcpy(dst, src, off);
143 memcpy(dst + off + sizeof(struct bpf_spin_lock),
144 src + off + sizeof(struct bpf_spin_lock),
145 map->value_size - off - sizeof(struct bpf_spin_lock));
147 memcpy(dst, src, map->value_size);
150 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
153 struct bpf_offload_dev;
154 struct bpf_offloaded_map;
156 struct bpf_map_dev_ops {
157 int (*map_get_next_key)(struct bpf_offloaded_map *map,
158 void *key, void *next_key);
159 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
160 void *key, void *value);
161 int (*map_update_elem)(struct bpf_offloaded_map *map,
162 void *key, void *value, u64 flags);
163 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
166 struct bpf_offloaded_map {
168 struct net_device *netdev;
169 const struct bpf_map_dev_ops *dev_ops;
171 struct list_head offloads;
174 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
176 return container_of(map, struct bpf_offloaded_map, map);
179 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
181 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
184 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
186 return map->btf && map->ops->map_seq_show_elem;
189 int map_check_no_btf(const struct bpf_map *map,
190 const struct btf *btf,
191 const struct btf_type *key_type,
192 const struct btf_type *value_type);
194 extern const struct bpf_map_ops bpf_map_offload_ops;
196 /* function argument constraints */
198 ARG_DONTCARE = 0, /* unused argument in helper function */
200 /* the following constraints used to prototype
201 * bpf_map_lookup/update/delete_elem() functions
203 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
204 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
205 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
206 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */
207 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */
209 /* the following constraints used to prototype bpf_memcmp() and other
210 * functions that access data on eBPF program stack
212 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
213 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
214 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
215 * helper function must fill all bytes or clear
216 * them in error case.
219 ARG_CONST_SIZE, /* number of bytes accessed from memory */
220 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
222 ARG_PTR_TO_CTX, /* pointer to context */
223 ARG_ANYTHING, /* any (initialized) argument is ok */
224 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
225 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
226 ARG_PTR_TO_INT, /* pointer to int */
227 ARG_PTR_TO_LONG, /* pointer to long */
228 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
229 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
232 /* type of values returned from helper functions */
233 enum bpf_return_type {
234 RET_INTEGER, /* function returns integer */
235 RET_VOID, /* function doesn't return anything */
236 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
237 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
238 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */
239 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */
240 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */
243 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
244 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
245 * instructions after verifying
247 struct bpf_func_proto {
248 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
251 enum bpf_return_type ret_type;
254 enum bpf_arg_type arg1_type;
255 enum bpf_arg_type arg2_type;
256 enum bpf_arg_type arg3_type;
257 enum bpf_arg_type arg4_type;
258 enum bpf_arg_type arg5_type;
260 enum bpf_arg_type arg_type[5];
262 int *btf_id; /* BTF ids of arguments */
265 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
266 * the first argument to eBPF programs.
267 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
271 enum bpf_access_type {
276 /* types of values stored in eBPF registers */
277 /* Pointer types represent:
280 * pointer + (u16) var
281 * pointer + (u16) var + imm
282 * if (range > 0) then [ptr, ptr + range - off) is safe to access
283 * if (id > 0) means that some 'var' was added
284 * if (off > 0) means that 'imm' was added
287 NOT_INIT = 0, /* nothing was written into register */
288 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
289 PTR_TO_CTX, /* reg points to bpf_context */
290 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
291 PTR_TO_MAP_VALUE, /* reg points to map element value */
292 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
293 PTR_TO_STACK, /* reg == frame_pointer + offset */
294 PTR_TO_PACKET_META, /* skb->data - meta_len */
295 PTR_TO_PACKET, /* reg points to skb->data */
296 PTR_TO_PACKET_END, /* skb->data + headlen */
297 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
298 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
299 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */
300 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
301 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */
302 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
303 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */
304 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
305 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
306 PTR_TO_BTF_ID, /* reg points to kernel struct */
309 /* The information passed from prog-specific *_is_valid_access
310 * back to the verifier.
312 struct bpf_insn_access_aux {
313 enum bpf_reg_type reg_type;
318 struct bpf_verifier_log *log; /* for verbose logs */
322 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
324 aux->ctx_field_size = size;
327 struct bpf_prog_ops {
328 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
329 union bpf_attr __user *uattr);
332 struct bpf_verifier_ops {
333 /* return eBPF function prototype for verification */
334 const struct bpf_func_proto *
335 (*get_func_proto)(enum bpf_func_id func_id,
336 const struct bpf_prog *prog);
338 /* return true if 'size' wide access at offset 'off' within bpf_context
339 * with 'type' (read or write) is allowed
341 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
342 const struct bpf_prog *prog,
343 struct bpf_insn_access_aux *info);
344 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
345 const struct bpf_prog *prog);
346 int (*gen_ld_abs)(const struct bpf_insn *orig,
347 struct bpf_insn *insn_buf);
348 u32 (*convert_ctx_access)(enum bpf_access_type type,
349 const struct bpf_insn *src,
350 struct bpf_insn *dst,
351 struct bpf_prog *prog, u32 *target_size);
354 struct bpf_prog_offload_ops {
355 /* verifier basic callbacks */
356 int (*insn_hook)(struct bpf_verifier_env *env,
357 int insn_idx, int prev_insn_idx);
358 int (*finalize)(struct bpf_verifier_env *env);
359 /* verifier optimization callbacks (called after .finalize) */
360 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
361 struct bpf_insn *insn);
362 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
363 /* program management callbacks */
364 int (*prepare)(struct bpf_prog *prog);
365 int (*translate)(struct bpf_prog *prog);
366 void (*destroy)(struct bpf_prog *prog);
369 struct bpf_prog_offload {
370 struct bpf_prog *prog;
371 struct net_device *netdev;
372 struct bpf_offload_dev *offdev;
374 struct list_head offloads;
381 enum bpf_cgroup_storage_type {
382 BPF_CGROUP_STORAGE_SHARED,
383 BPF_CGROUP_STORAGE_PERCPU,
384 __BPF_CGROUP_STORAGE_MAX
387 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
389 /* The longest tracepoint has 12 args.
390 * See include/trace/bpf_probe.h
392 #define MAX_BPF_FUNC_ARGS 12
394 struct bpf_prog_stats {
397 struct u64_stats_sync syncp;
398 } __aligned(2 * sizeof(u64));
400 struct btf_func_model {
403 u8 arg_size[MAX_BPF_FUNC_ARGS];
406 /* Restore arguments before returning from trampoline to let original function
407 * continue executing. This flag is used for fentry progs when there are no
410 #define BPF_TRAMP_F_RESTORE_REGS BIT(0)
411 /* Call original function after fentry progs, but before fexit progs.
412 * Makes sense for fentry/fexit, normal calls and indirect calls.
414 #define BPF_TRAMP_F_CALL_ORIG BIT(1)
415 /* Skip current frame and return to parent. Makes sense for fentry/fexit
416 * programs only. Should not be used with normal calls and indirect calls.
418 #define BPF_TRAMP_F_SKIP_FRAME BIT(2)
420 /* Different use cases for BPF trampoline:
421 * 1. replace nop at the function entry (kprobe equivalent)
422 * flags = BPF_TRAMP_F_RESTORE_REGS
423 * fentry = a set of programs to run before returning from trampoline
425 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
426 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
427 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
428 * fentry = a set of program to run before calling original function
429 * fexit = a set of program to run after original function
431 * 3. replace direct call instruction anywhere in the function body
432 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
434 * fentry = a set of programs to run before returning from trampoline
435 * With flags = BPF_TRAMP_F_CALL_ORIG
436 * orig_call = original callback addr or direct function addr
437 * fentry = a set of program to run before calling original function
438 * fexit = a set of program to run after original function
440 int arch_prepare_bpf_trampoline(void *image, struct btf_func_model *m, u32 flags,
441 struct bpf_prog **fentry_progs, int fentry_cnt,
442 struct bpf_prog **fexit_progs, int fexit_cnt,
444 /* these two functions are called from generated trampoline */
445 u64 notrace __bpf_prog_enter(void);
446 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start);
448 enum bpf_tramp_prog_type {
454 struct bpf_trampoline {
455 /* hlist for trampoline_table */
456 struct hlist_node hlist;
457 /* serializes access to fields of this trampoline */
462 struct btf_func_model model;
466 /* list of BPF programs using this trampoline */
467 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
468 /* Number of attached programs. A counter per kind. */
469 int progs_cnt[BPF_TRAMP_MAX];
470 /* Executable image of trampoline */
474 #ifdef CONFIG_BPF_JIT
475 struct bpf_trampoline *bpf_trampoline_lookup(u64 key);
476 int bpf_trampoline_link_prog(struct bpf_prog *prog);
477 int bpf_trampoline_unlink_prog(struct bpf_prog *prog);
478 void bpf_trampoline_put(struct bpf_trampoline *tr);
480 static inline struct bpf_trampoline *bpf_trampoline_lookup(u64 key)
484 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog)
488 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog)
492 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
495 struct bpf_func_info_aux {
499 enum bpf_jit_poke_reason {
500 BPF_POKE_REASON_TAIL_CALL,
503 /* Descriptor of pokes pointing /into/ the JITed image. */
504 struct bpf_jit_poke_descriptor {
517 struct bpf_prog_aux {
525 u32 func_cnt; /* used by non-func prog as the number of func progs */
526 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
527 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
528 struct bpf_prog *linked_prog;
529 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
530 bool offload_requested;
531 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
532 bool func_proto_unreliable;
533 enum bpf_tramp_prog_type trampoline_prog_type;
534 struct bpf_trampoline *trampoline;
535 struct hlist_node tramp_hlist;
536 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
537 const struct btf_type *attach_func_proto;
538 /* function name for valid attach_btf_id */
539 const char *attach_func_name;
540 struct bpf_prog **func;
541 void *jit_data; /* JIT specific data. arch dependent */
542 struct bpf_jit_poke_descriptor *poke_tab;
544 struct latch_tree_node ksym_tnode;
545 struct list_head ksym_lnode;
546 const struct bpf_prog_ops *ops;
547 struct bpf_map **used_maps;
548 struct bpf_prog *prog;
549 struct user_struct *user;
550 u64 load_time; /* ns since boottime */
551 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
552 char name[BPF_OBJ_NAME_LEN];
553 #ifdef CONFIG_SECURITY
556 struct bpf_prog_offload *offload;
558 struct bpf_func_info *func_info;
559 struct bpf_func_info_aux *func_info_aux;
560 /* bpf_line_info loaded from userspace. linfo->insn_off
561 * has the xlated insn offset.
562 * Both the main and sub prog share the same linfo.
563 * The subprog can access its first linfo by
564 * using the linfo_idx.
566 struct bpf_line_info *linfo;
567 /* jited_linfo is the jited addr of the linfo. It has a
568 * one to one mapping to linfo:
569 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
570 * Both the main and sub prog share the same jited_linfo.
571 * The subprog can access its first jited_linfo by
572 * using the linfo_idx.
577 /* subprog can use linfo_idx to access its first linfo and
579 * main prog always has linfo_idx == 0
583 struct exception_table_entry *extable;
584 struct bpf_prog_stats __percpu *stats;
586 struct work_struct work;
591 struct bpf_array_aux {
592 /* 'Ownership' of prog array is claimed by the first program that
593 * is going to use this map or by the first program which FD is
594 * stored in the map to make sure that all callers and callees have
595 * the same prog type and JITed flag.
597 enum bpf_prog_type type;
599 /* Programs with direct jumps into programs part of this array. */
600 struct list_head poke_progs;
602 struct mutex poke_mutex;
603 struct work_struct work;
610 struct bpf_array_aux *aux;
612 char value[0] __aligned(8);
613 void *ptrs[0] __aligned(8);
614 void __percpu *pptrs[0] __aligned(8);
618 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
619 #define MAX_TAIL_CALL_CNT 32
621 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
622 BPF_F_RDONLY_PROG | \
626 #define BPF_MAP_CAN_READ BIT(0)
627 #define BPF_MAP_CAN_WRITE BIT(1)
629 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
631 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
633 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
636 if (access_flags & BPF_F_RDONLY_PROG)
637 return BPF_MAP_CAN_READ;
638 else if (access_flags & BPF_F_WRONLY_PROG)
639 return BPF_MAP_CAN_WRITE;
641 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
644 static inline bool bpf_map_flags_access_ok(u32 access_flags)
646 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
647 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
650 struct bpf_event_entry {
651 struct perf_event *event;
652 struct file *perf_file;
653 struct file *map_file;
657 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
658 int bpf_prog_calc_tag(struct bpf_prog *fp);
659 const char *kernel_type_name(u32 btf_type_id);
661 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
663 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
664 unsigned long off, unsigned long len);
665 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
666 const struct bpf_insn *src,
667 struct bpf_insn *dst,
668 struct bpf_prog *prog,
671 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
672 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
674 /* an array of programs to be executed under rcu_lock.
677 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
679 * the structure returned by bpf_prog_array_alloc() should be populated
680 * with program pointers and the last pointer must be NULL.
681 * The user has to keep refcnt on the program and make sure the program
682 * is removed from the array before bpf_prog_put().
683 * The 'struct bpf_prog_array *' should only be replaced with xchg()
684 * since other cpus are walking the array of pointers in parallel.
686 struct bpf_prog_array_item {
687 struct bpf_prog *prog;
688 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
691 struct bpf_prog_array {
693 struct bpf_prog_array_item items[0];
696 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
697 void bpf_prog_array_free(struct bpf_prog_array *progs);
698 int bpf_prog_array_length(struct bpf_prog_array *progs);
699 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
700 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
701 __u32 __user *prog_ids, u32 cnt);
703 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
704 struct bpf_prog *old_prog);
705 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
706 u32 *prog_ids, u32 request_cnt,
708 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
709 struct bpf_prog *exclude_prog,
710 struct bpf_prog *include_prog,
711 struct bpf_prog_array **new_array);
713 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
715 struct bpf_prog_array_item *_item; \
716 struct bpf_prog *_prog; \
717 struct bpf_prog_array *_array; \
721 _array = rcu_dereference(array); \
722 if (unlikely(check_non_null && !_array))\
724 _item = &_array->items[0]; \
725 while ((_prog = READ_ONCE(_item->prog))) { \
726 bpf_cgroup_storage_set(_item->cgroup_storage); \
727 _ret &= func(_prog, ctx); \
736 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs
737 * so BPF programs can request cwr for TCP packets.
739 * Current cgroup skb programs can only return 0 or 1 (0 to drop the
740 * packet. This macro changes the behavior so the low order bit
741 * indicates whether the packet should be dropped (0) or not (1)
742 * and the next bit is a congestion notification bit. This could be
743 * used by TCP to call tcp_enter_cwr()
745 * Hence, new allowed return values of CGROUP EGRESS BPF programs are:
748 * 2: drop packet and cn
749 * 3: keep packet and cn
751 * This macro then converts it to one of the NET_XMIT or an error
752 * code that is then interpreted as drop packet (and no cn):
753 * 0: NET_XMIT_SUCCESS skb should be transmitted
754 * 1: NET_XMIT_DROP skb should be dropped and cn
755 * 2: NET_XMIT_CN skb should be transmitted and cn
756 * 3: -EPERM skb should be dropped
758 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \
760 struct bpf_prog_array_item *_item; \
761 struct bpf_prog *_prog; \
762 struct bpf_prog_array *_array; \
768 _array = rcu_dereference(array); \
769 _item = &_array->items[0]; \
770 while ((_prog = READ_ONCE(_item->prog))) { \
771 bpf_cgroup_storage_set(_item->cgroup_storage); \
772 ret = func(_prog, ctx); \
780 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \
782 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \
786 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
787 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
789 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
790 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
792 #ifdef CONFIG_BPF_SYSCALL
793 DECLARE_PER_CPU(int, bpf_prog_active);
795 extern const struct file_operations bpf_map_fops;
796 extern const struct file_operations bpf_prog_fops;
798 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
799 extern const struct bpf_prog_ops _name ## _prog_ops; \
800 extern const struct bpf_verifier_ops _name ## _verifier_ops;
801 #define BPF_MAP_TYPE(_id, _ops) \
802 extern const struct bpf_map_ops _ops;
803 #include <linux/bpf_types.h>
807 extern const struct bpf_prog_ops bpf_offload_prog_ops;
808 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
809 extern const struct bpf_verifier_ops xdp_analyzer_ops;
811 struct bpf_prog *bpf_prog_get(u32 ufd);
812 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
814 void bpf_prog_add(struct bpf_prog *prog, int i);
815 void bpf_prog_sub(struct bpf_prog *prog, int i);
816 void bpf_prog_inc(struct bpf_prog *prog);
817 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
818 void bpf_prog_put(struct bpf_prog *prog);
819 int __bpf_prog_charge(struct user_struct *user, u32 pages);
820 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
821 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
822 struct bpf_map **used_maps, u32 len);
824 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
825 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
827 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
828 struct bpf_map *__bpf_map_get(struct fd f);
829 void bpf_map_inc(struct bpf_map *map);
830 void bpf_map_inc_with_uref(struct bpf_map *map);
831 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
832 void bpf_map_put_with_uref(struct bpf_map *map);
833 void bpf_map_put(struct bpf_map *map);
834 int bpf_map_charge_memlock(struct bpf_map *map, u32 pages);
835 void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages);
836 int bpf_map_charge_init(struct bpf_map_memory *mem, u64 size);
837 void bpf_map_charge_finish(struct bpf_map_memory *mem);
838 void bpf_map_charge_move(struct bpf_map_memory *dst,
839 struct bpf_map_memory *src);
840 void *bpf_map_area_alloc(u64 size, int numa_node);
841 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
842 void bpf_map_area_free(void *base);
843 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
845 extern int sysctl_unprivileged_bpf_disabled;
847 int bpf_map_new_fd(struct bpf_map *map, int flags);
848 int bpf_prog_new_fd(struct bpf_prog *prog);
850 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
851 int bpf_obj_get_user(const char __user *pathname, int flags);
853 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
854 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
855 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
857 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
860 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
862 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
863 void *key, void *value, u64 map_flags);
864 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
865 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
866 void *key, void *value, u64 map_flags);
867 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
869 int bpf_get_file_flag(int flags);
870 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size,
873 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
874 * forced to use 'long' read/writes to try to atomically copy long counters.
875 * Best-effort only. No barriers here, since it _will_ race with concurrent
876 * updates from BPF programs. Called from bpf syscall and mostly used with
877 * size 8 or 16 bytes, so ask compiler to inline it.
879 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
881 const long *lsrc = src;
884 size /= sizeof(long);
889 /* verify correctness of eBPF program */
890 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr,
891 union bpf_attr __user *uattr);
892 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
898 struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
899 struct bpf_dtab_netdev *__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key);
900 void __dev_map_flush(struct bpf_map *map);
901 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
902 struct net_device *dev_rx);
903 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
904 struct bpf_prog *xdp_prog);
906 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
907 void __cpu_map_flush(struct bpf_map *map);
908 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
909 struct net_device *dev_rx);
911 /* Return map's numa specified by userspace */
912 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
914 return (attr->map_flags & BPF_F_NUMA_NODE) ?
915 attr->numa_node : NUMA_NO_NODE;
918 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
919 int array_map_alloc_check(union bpf_attr *attr);
921 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
922 union bpf_attr __user *uattr);
923 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
924 union bpf_attr __user *uattr);
925 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
926 const union bpf_attr *kattr,
927 union bpf_attr __user *uattr);
928 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
929 const struct bpf_prog *prog,
930 struct bpf_insn_access_aux *info);
931 int btf_struct_access(struct bpf_verifier_log *log,
932 const struct btf_type *t, int off, int size,
933 enum bpf_access_type atype,
935 int btf_resolve_helper_id(struct bpf_verifier_log *log,
936 const struct bpf_func_proto *fn, int);
938 int btf_distill_func_proto(struct bpf_verifier_log *log,
940 const struct btf_type *func_proto,
941 const char *func_name,
942 struct btf_func_model *m);
944 int btf_check_func_arg_match(struct bpf_verifier_env *env, int subprog);
946 #else /* !CONFIG_BPF_SYSCALL */
947 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
949 return ERR_PTR(-EOPNOTSUPP);
952 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
953 enum bpf_prog_type type,
956 return ERR_PTR(-EOPNOTSUPP);
959 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
963 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
967 static inline void bpf_prog_put(struct bpf_prog *prog)
971 static inline void bpf_prog_inc(struct bpf_prog *prog)
975 static inline struct bpf_prog *__must_check
976 bpf_prog_inc_not_zero(struct bpf_prog *prog)
978 return ERR_PTR(-EOPNOTSUPP);
981 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
986 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
990 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
995 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
1001 static inline struct net_device *__dev_map_hash_lookup_elem(struct bpf_map *map,
1007 static inline void __dev_map_flush(struct bpf_map *map)
1012 struct bpf_dtab_netdev;
1015 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
1016 struct net_device *dev_rx)
1023 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
1024 struct sk_buff *skb,
1025 struct bpf_prog *xdp_prog)
1031 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
1036 static inline void __cpu_map_flush(struct bpf_map *map)
1040 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
1041 struct xdp_buff *xdp,
1042 struct net_device *dev_rx)
1047 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
1048 enum bpf_prog_type type)
1050 return ERR_PTR(-EOPNOTSUPP);
1053 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
1054 const union bpf_attr *kattr,
1055 union bpf_attr __user *uattr)
1060 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
1061 const union bpf_attr *kattr,
1062 union bpf_attr __user *uattr)
1067 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1068 const union bpf_attr *kattr,
1069 union bpf_attr __user *uattr)
1074 static inline void bpf_map_put(struct bpf_map *map)
1077 #endif /* CONFIG_BPF_SYSCALL */
1079 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
1080 enum bpf_prog_type type)
1082 return bpf_prog_get_type_dev(ufd, type, false);
1085 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
1087 int bpf_prog_offload_compile(struct bpf_prog *prog);
1088 void bpf_prog_offload_destroy(struct bpf_prog *prog);
1089 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
1090 struct bpf_prog *prog);
1092 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
1094 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
1095 int bpf_map_offload_update_elem(struct bpf_map *map,
1096 void *key, void *value, u64 flags);
1097 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
1098 int bpf_map_offload_get_next_key(struct bpf_map *map,
1099 void *key, void *next_key);
1101 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
1103 struct bpf_offload_dev *
1104 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
1105 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
1106 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
1107 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
1108 struct net_device *netdev);
1109 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
1110 struct net_device *netdev);
1111 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
1113 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
1114 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
1116 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
1118 return aux->offload_requested;
1121 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1123 return unlikely(map->ops == &bpf_map_offload_ops);
1126 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
1127 void bpf_map_offload_map_free(struct bpf_map *map);
1129 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
1130 union bpf_attr *attr)
1135 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
1140 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1145 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
1147 return ERR_PTR(-EOPNOTSUPP);
1150 static inline void bpf_map_offload_map_free(struct bpf_map *map)
1153 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
1155 #if defined(CONFIG_BPF_STREAM_PARSER)
1156 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, u32 which);
1157 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
1159 static inline int sock_map_prog_update(struct bpf_map *map,
1160 struct bpf_prog *prog, u32 which)
1165 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
1166 struct bpf_prog *prog)
1172 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
1173 void bpf_sk_reuseport_detach(struct sock *sk);
1174 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
1176 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
1177 void *value, u64 map_flags);
1179 static inline void bpf_sk_reuseport_detach(struct sock *sk)
1183 #ifdef CONFIG_BPF_SYSCALL
1184 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
1185 void *key, void *value)
1190 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
1191 void *key, void *value,
1196 #endif /* CONFIG_BPF_SYSCALL */
1197 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
1199 /* verifier prototypes for helper functions called from eBPF programs */
1200 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
1201 extern const struct bpf_func_proto bpf_map_update_elem_proto;
1202 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
1203 extern const struct bpf_func_proto bpf_map_push_elem_proto;
1204 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
1205 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
1207 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
1208 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
1209 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
1210 extern const struct bpf_func_proto bpf_tail_call_proto;
1211 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
1212 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
1213 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
1214 extern const struct bpf_func_proto bpf_get_current_comm_proto;
1215 extern const struct bpf_func_proto bpf_get_stackid_proto;
1216 extern const struct bpf_func_proto bpf_get_stack_proto;
1217 extern const struct bpf_func_proto bpf_sock_map_update_proto;
1218 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
1219 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
1220 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
1221 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
1222 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
1223 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
1224 extern const struct bpf_func_proto bpf_spin_lock_proto;
1225 extern const struct bpf_func_proto bpf_spin_unlock_proto;
1226 extern const struct bpf_func_proto bpf_get_local_storage_proto;
1227 extern const struct bpf_func_proto bpf_strtol_proto;
1228 extern const struct bpf_func_proto bpf_strtoul_proto;
1229 extern const struct bpf_func_proto bpf_tcp_sock_proto;
1231 /* Shared helpers among cBPF and eBPF. */
1232 void bpf_user_rnd_init_once(void);
1233 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1235 #if defined(CONFIG_NET)
1236 bool bpf_sock_common_is_valid_access(int off, int size,
1237 enum bpf_access_type type,
1238 struct bpf_insn_access_aux *info);
1239 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1240 struct bpf_insn_access_aux *info);
1241 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
1242 const struct bpf_insn *si,
1243 struct bpf_insn *insn_buf,
1244 struct bpf_prog *prog,
1247 static inline bool bpf_sock_common_is_valid_access(int off, int size,
1248 enum bpf_access_type type,
1249 struct bpf_insn_access_aux *info)
1253 static inline bool bpf_sock_is_valid_access(int off, int size,
1254 enum bpf_access_type type,
1255 struct bpf_insn_access_aux *info)
1259 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
1260 const struct bpf_insn *si,
1261 struct bpf_insn *insn_buf,
1262 struct bpf_prog *prog,
1270 struct sk_reuseport_kern {
1271 struct sk_buff *skb;
1273 struct sock *selected_sk;
1279 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1280 struct bpf_insn_access_aux *info);
1282 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
1283 const struct bpf_insn *si,
1284 struct bpf_insn *insn_buf,
1285 struct bpf_prog *prog,
1288 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1289 struct bpf_insn_access_aux *info);
1291 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
1292 const struct bpf_insn *si,
1293 struct bpf_insn *insn_buf,
1294 struct bpf_prog *prog,
1297 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
1298 enum bpf_access_type type,
1299 struct bpf_insn_access_aux *info)
1304 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
1305 const struct bpf_insn *si,
1306 struct bpf_insn *insn_buf,
1307 struct bpf_prog *prog,
1312 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
1313 enum bpf_access_type type,
1314 struct bpf_insn_access_aux *info)
1319 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
1320 const struct bpf_insn *si,
1321 struct bpf_insn *insn_buf,
1322 struct bpf_prog *prog,
1327 #endif /* CONFIG_INET */
1329 enum bpf_text_poke_type {
1334 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
1335 void *addr1, void *addr2);
1337 #endif /* _LINUX_BPF_H */