1 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 * This program is free software; you can redistribute it and/or
4 * modify it under the terms of version 2 of the GNU General Public
5 * License as published by the Free Software Foundation.
10 #include <uapi/linux/bpf.h>
12 #include <linux/workqueue.h>
13 #include <linux/file.h>
14 #include <linux/percpu.h>
15 #include <linux/err.h>
16 #include <linux/rbtree_latch.h>
17 #include <linux/numa.h>
18 #include <linux/wait.h>
24 /* map is generic key/value storage optionally accesible by eBPF programs */
26 /* funcs callable from userspace (via syscall) */
27 int (*map_alloc_check
)(union bpf_attr
*attr
);
28 struct bpf_map
*(*map_alloc
)(union bpf_attr
*attr
);
29 void (*map_release
)(struct bpf_map
*map
, struct file
*map_file
);
30 void (*map_free
)(struct bpf_map
*map
);
31 int (*map_get_next_key
)(struct bpf_map
*map
, void *key
, void *next_key
);
32 void (*map_release_uref
)(struct bpf_map
*map
);
33 void *(*map_lookup_elem_sys_only
)(struct bpf_map
*map
, void *key
);
35 /* funcs callable from userspace and from eBPF programs */
36 void *(*map_lookup_elem
)(struct bpf_map
*map
, void *key
);
37 int (*map_update_elem
)(struct bpf_map
*map
, void *key
, void *value
, u64 flags
);
38 int (*map_delete_elem
)(struct bpf_map
*map
, void *key
);
40 /* funcs called by prog_array and perf_event_array map */
41 void *(*map_fd_get_ptr
)(struct bpf_map
*map
, struct file
*map_file
,
43 void (*map_fd_put_ptr
)(void *ptr
);
44 u32 (*map_gen_lookup
)(struct bpf_map
*map
, struct bpf_insn
*insn_buf
);
45 u32 (*map_fd_sys_lookup_elem
)(void *ptr
);
49 /* 1st cacheline with read-mostly members of which some
50 * are also accessed in fast-path (e.g. ops, max_entries).
52 const struct bpf_map_ops
*ops ____cacheline_aligned
;
53 struct bpf_map
*inner_map_meta
;
54 #ifdef CONFIG_SECURITY
57 enum bpf_map_type map_type
;
68 /* 2nd cacheline with misc members to avoid false sharing
69 * particularly with refcounting.
71 struct user_struct
*user ____cacheline_aligned
;
74 struct work_struct work
;
75 char name
[BPF_OBJ_NAME_LEN
];
78 /* function argument constraints */
80 ARG_DONTCARE
= 0, /* unused argument in helper function */
82 /* the following constraints used to prototype
83 * bpf_map_lookup/update/delete_elem() functions
85 ARG_CONST_MAP_PTR
, /* const argument used as pointer to bpf_map */
86 ARG_PTR_TO_MAP_KEY
, /* pointer to stack used as map key */
87 ARG_PTR_TO_MAP_VALUE
, /* pointer to stack used as map value */
89 /* the following constraints used to prototype bpf_memcmp() and other
90 * functions that access data on eBPF program stack
92 ARG_PTR_TO_MEM
, /* pointer to valid memory (stack, packet, map value) */
93 ARG_PTR_TO_MEM_OR_NULL
, /* pointer to valid memory or NULL */
94 ARG_PTR_TO_UNINIT_MEM
, /* pointer to memory does not need to be initialized,
95 * helper function must fill all bytes or clear
99 ARG_CONST_SIZE
, /* number of bytes accessed from memory */
100 ARG_CONST_SIZE_OR_ZERO
, /* number of bytes accessed from memory or 0 */
102 ARG_PTR_TO_CTX
, /* pointer to context */
103 ARG_ANYTHING
, /* any (initialized) argument is ok */
106 /* type of values returned from helper functions */
107 enum bpf_return_type
{
108 RET_INTEGER
, /* function returns integer */
109 RET_VOID
, /* function doesn't return anything */
110 RET_PTR_TO_MAP_VALUE_OR_NULL
, /* returns a pointer to map elem value or NULL */
113 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
114 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
115 * instructions after verifying
117 struct bpf_func_proto
{
118 u64 (*func
)(u64 r1
, u64 r2
, u64 r3
, u64 r4
, u64 r5
);
121 enum bpf_return_type ret_type
;
122 enum bpf_arg_type arg1_type
;
123 enum bpf_arg_type arg2_type
;
124 enum bpf_arg_type arg3_type
;
125 enum bpf_arg_type arg4_type
;
126 enum bpf_arg_type arg5_type
;
129 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
130 * the first argument to eBPF programs.
131 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
135 enum bpf_access_type
{
140 /* types of values stored in eBPF registers */
141 /* Pointer types represent:
144 * pointer + (u16) var
145 * pointer + (u16) var + imm
146 * if (range > 0) then [ptr, ptr + range - off) is safe to access
147 * if (id > 0) means that some 'var' was added
148 * if (off > 0) means that 'imm' was added
151 NOT_INIT
= 0, /* nothing was written into register */
152 SCALAR_VALUE
, /* reg doesn't contain a valid pointer */
153 PTR_TO_CTX
, /* reg points to bpf_context */
154 CONST_PTR_TO_MAP
, /* reg points to struct bpf_map */
155 PTR_TO_MAP_VALUE
, /* reg points to map element value */
156 PTR_TO_MAP_VALUE_OR_NULL
,/* points to map elem value or NULL */
157 PTR_TO_STACK
, /* reg == frame_pointer + offset */
158 PTR_TO_PACKET_META
, /* skb->data - meta_len */
159 PTR_TO_PACKET
, /* reg points to skb->data */
160 PTR_TO_PACKET_END
, /* skb->data + headlen */
163 /* The information passed from prog-specific *_is_valid_access
164 * back to the verifier.
166 struct bpf_insn_access_aux
{
167 enum bpf_reg_type reg_type
;
172 bpf_ctx_record_field_size(struct bpf_insn_access_aux
*aux
, u32 size
)
174 aux
->ctx_field_size
= size
;
177 struct bpf_prog_ops
{
178 int (*test_run
)(struct bpf_prog
*prog
, const union bpf_attr
*kattr
,
179 union bpf_attr __user
*uattr
);
182 struct bpf_verifier_ops
{
183 /* return eBPF function prototype for verification */
184 const struct bpf_func_proto
*(*get_func_proto
)(enum bpf_func_id func_id
);
186 /* return true if 'size' wide access at offset 'off' within bpf_context
187 * with 'type' (read or write) is allowed
189 bool (*is_valid_access
)(int off
, int size
, enum bpf_access_type type
,
190 struct bpf_insn_access_aux
*info
);
191 int (*gen_prologue
)(struct bpf_insn
*insn
, bool direct_write
,
192 const struct bpf_prog
*prog
);
193 u32 (*convert_ctx_access
)(enum bpf_access_type type
,
194 const struct bpf_insn
*src
,
195 struct bpf_insn
*dst
,
196 struct bpf_prog
*prog
, u32
*target_size
);
199 struct bpf_dev_offload
{
200 struct bpf_prog
*prog
;
201 struct net_device
*netdev
;
203 struct list_head offloads
;
205 bool verifier_running
;
206 wait_queue_head_t verifier_done
;
209 struct bpf_prog_aux
{
215 struct latch_tree_node ksym_tnode
;
216 struct list_head ksym_lnode
;
217 const struct bpf_prog_ops
*ops
;
218 struct bpf_map
**used_maps
;
219 struct bpf_prog
*prog
;
220 struct user_struct
*user
;
221 u64 load_time
; /* ns since boottime */
222 char name
[BPF_OBJ_NAME_LEN
];
223 #ifdef CONFIG_SECURITY
226 struct bpf_dev_offload
*offload
;
228 struct work_struct work
;
237 /* 'ownership' of prog_array is claimed by the first program that
238 * is going to use this map or by the first program which FD is stored
239 * in the map to make sure that all callers and callees have the same
240 * prog_type and JITed flag
242 enum bpf_prog_type owner_prog_type
;
245 char value
[0] __aligned(8);
246 void *ptrs
[0] __aligned(8);
247 void __percpu
*pptrs
[0] __aligned(8);
251 #define MAX_TAIL_CALL_CNT 32
253 struct bpf_event_entry
{
254 struct perf_event
*event
;
255 struct file
*perf_file
;
256 struct file
*map_file
;
260 bool bpf_prog_array_compatible(struct bpf_array
*array
, const struct bpf_prog
*fp
);
261 int bpf_prog_calc_tag(struct bpf_prog
*fp
);
263 const struct bpf_func_proto
*bpf_get_trace_printk_proto(void);
265 typedef unsigned long (*bpf_ctx_copy_t
)(void *dst
, const void *src
,
266 unsigned long off
, unsigned long len
);
268 u64
bpf_event_output(struct bpf_map
*map
, u64 flags
, void *meta
, u64 meta_size
,
269 void *ctx
, u64 ctx_size
, bpf_ctx_copy_t ctx_copy
);
271 int bpf_prog_test_run_xdp(struct bpf_prog
*prog
, const union bpf_attr
*kattr
,
272 union bpf_attr __user
*uattr
);
273 int bpf_prog_test_run_skb(struct bpf_prog
*prog
, const union bpf_attr
*kattr
,
274 union bpf_attr __user
*uattr
);
276 /* an array of programs to be executed under rcu_lock.
279 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
281 * the structure returned by bpf_prog_array_alloc() should be populated
282 * with program pointers and the last pointer must be NULL.
283 * The user has to keep refcnt on the program and make sure the program
284 * is removed from the array before bpf_prog_put().
285 * The 'struct bpf_prog_array *' should only be replaced with xchg()
286 * since other cpus are walking the array of pointers in parallel.
288 struct bpf_prog_array
{
290 struct bpf_prog
*progs
[0];
293 struct bpf_prog_array __rcu
*bpf_prog_array_alloc(u32 prog_cnt
, gfp_t flags
);
294 void bpf_prog_array_free(struct bpf_prog_array __rcu
*progs
);
295 int bpf_prog_array_length(struct bpf_prog_array __rcu
*progs
);
296 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu
*progs
,
297 __u32 __user
*prog_ids
, u32 cnt
);
299 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu
*progs
,
300 struct bpf_prog
*old_prog
);
301 int bpf_prog_array_copy(struct bpf_prog_array __rcu
*old_array
,
302 struct bpf_prog
*exclude_prog
,
303 struct bpf_prog
*include_prog
,
304 struct bpf_prog_array
**new_array
);
306 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
308 struct bpf_prog **_prog, *__prog; \
309 struct bpf_prog_array *_array; \
312 _array = rcu_dereference(array); \
313 if (unlikely(check_non_null && !_array))\
315 _prog = _array->progs; \
316 while ((__prog = READ_ONCE(*_prog))) { \
317 _ret &= func(__prog, ctx); \
325 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
326 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
328 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
329 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
331 #ifdef CONFIG_BPF_SYSCALL
332 DECLARE_PER_CPU(int, bpf_prog_active
);
334 extern const struct file_operations bpf_map_fops
;
335 extern const struct file_operations bpf_prog_fops
;
337 #define BPF_PROG_TYPE(_id, _name) \
338 extern const struct bpf_prog_ops _name ## _prog_ops; \
339 extern const struct bpf_verifier_ops _name ## _verifier_ops;
340 #define BPF_MAP_TYPE(_id, _ops) \
341 extern const struct bpf_map_ops _ops;
342 #include <linux/bpf_types.h>
346 extern const struct bpf_prog_ops bpf_offload_prog_ops
;
347 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops
;
348 extern const struct bpf_verifier_ops xdp_analyzer_ops
;
350 struct bpf_prog
*bpf_prog_get(u32 ufd
);
351 struct bpf_prog
*bpf_prog_get_type_dev(u32 ufd
, enum bpf_prog_type type
,
353 struct bpf_prog
* __must_check
bpf_prog_add(struct bpf_prog
*prog
, int i
);
354 void bpf_prog_sub(struct bpf_prog
*prog
, int i
);
355 struct bpf_prog
* __must_check
bpf_prog_inc(struct bpf_prog
*prog
);
356 struct bpf_prog
* __must_check
bpf_prog_inc_not_zero(struct bpf_prog
*prog
);
357 void bpf_prog_put(struct bpf_prog
*prog
);
358 int __bpf_prog_charge(struct user_struct
*user
, u32 pages
);
359 void __bpf_prog_uncharge(struct user_struct
*user
, u32 pages
);
361 struct bpf_map
*bpf_map_get_with_uref(u32 ufd
);
362 struct bpf_map
*__bpf_map_get(struct fd f
);
363 struct bpf_map
* __must_check
bpf_map_inc(struct bpf_map
*map
, bool uref
);
364 void bpf_map_put_with_uref(struct bpf_map
*map
);
365 void bpf_map_put(struct bpf_map
*map
);
366 int bpf_map_precharge_memlock(u32 pages
);
367 void *bpf_map_area_alloc(size_t size
, int numa_node
);
368 void bpf_map_area_free(void *base
);
370 extern int sysctl_unprivileged_bpf_disabled
;
372 int bpf_map_new_fd(struct bpf_map
*map
, int flags
);
373 int bpf_prog_new_fd(struct bpf_prog
*prog
);
375 int bpf_obj_pin_user(u32 ufd
, const char __user
*pathname
);
376 int bpf_obj_get_user(const char __user
*pathname
, int flags
);
378 int bpf_percpu_hash_copy(struct bpf_map
*map
, void *key
, void *value
);
379 int bpf_percpu_array_copy(struct bpf_map
*map
, void *key
, void *value
);
380 int bpf_percpu_hash_update(struct bpf_map
*map
, void *key
, void *value
,
382 int bpf_percpu_array_update(struct bpf_map
*map
, void *key
, void *value
,
385 int bpf_stackmap_copy(struct bpf_map
*map
, void *key
, void *value
);
387 int bpf_fd_array_map_update_elem(struct bpf_map
*map
, struct file
*map_file
,
388 void *key
, void *value
, u64 map_flags
);
389 int bpf_fd_array_map_lookup_elem(struct bpf_map
*map
, void *key
, u32
*value
);
390 int bpf_fd_htab_map_update_elem(struct bpf_map
*map
, struct file
*map_file
,
391 void *key
, void *value
, u64 map_flags
);
392 int bpf_fd_htab_map_lookup_elem(struct bpf_map
*map
, void *key
, u32
*value
);
394 int bpf_get_file_flag(int flags
);
396 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
397 * forced to use 'long' read/writes to try to atomically copy long counters.
398 * Best-effort only. No barriers here, since it _will_ race with concurrent
399 * updates from BPF programs. Called from bpf syscall and mostly used with
400 * size 8 or 16 bytes, so ask compiler to inline it.
402 static inline void bpf_long_memcpy(void *dst
, const void *src
, u32 size
)
404 const long *lsrc
= src
;
407 size
/= sizeof(long);
412 /* verify correctness of eBPF program */
413 int bpf_check(struct bpf_prog
**fp
, union bpf_attr
*attr
);
416 struct net_device
*__dev_map_lookup_elem(struct bpf_map
*map
, u32 key
);
417 void __dev_map_insert_ctx(struct bpf_map
*map
, u32 index
);
418 void __dev_map_flush(struct bpf_map
*map
);
420 struct bpf_cpu_map_entry
*__cpu_map_lookup_elem(struct bpf_map
*map
, u32 key
);
421 void __cpu_map_insert_ctx(struct bpf_map
*map
, u32 index
);
422 void __cpu_map_flush(struct bpf_map
*map
);
424 int cpu_map_enqueue(struct bpf_cpu_map_entry
*rcpu
, struct xdp_buff
*xdp
,
425 struct net_device
*dev_rx
);
427 /* Return map's numa specified by userspace */
428 static inline int bpf_map_attr_numa_node(const union bpf_attr
*attr
)
430 return (attr
->map_flags
& BPF_F_NUMA_NODE
) ?
431 attr
->numa_node
: NUMA_NO_NODE
;
434 struct bpf_prog
*bpf_prog_get_type_path(const char *name
, enum bpf_prog_type type
);
436 #else /* !CONFIG_BPF_SYSCALL */
437 static inline struct bpf_prog
*bpf_prog_get(u32 ufd
)
439 return ERR_PTR(-EOPNOTSUPP
);
442 static inline struct bpf_prog
*bpf_prog_get_type_dev(u32 ufd
,
443 enum bpf_prog_type type
,
446 return ERR_PTR(-EOPNOTSUPP
);
449 static inline struct bpf_prog
* __must_check
bpf_prog_add(struct bpf_prog
*prog
,
452 return ERR_PTR(-EOPNOTSUPP
);
455 static inline void bpf_prog_sub(struct bpf_prog
*prog
, int i
)
459 static inline void bpf_prog_put(struct bpf_prog
*prog
)
463 static inline struct bpf_prog
* __must_check
bpf_prog_inc(struct bpf_prog
*prog
)
465 return ERR_PTR(-EOPNOTSUPP
);
468 static inline struct bpf_prog
*__must_check
469 bpf_prog_inc_not_zero(struct bpf_prog
*prog
)
471 return ERR_PTR(-EOPNOTSUPP
);
474 static inline int __bpf_prog_charge(struct user_struct
*user
, u32 pages
)
479 static inline void __bpf_prog_uncharge(struct user_struct
*user
, u32 pages
)
483 static inline int bpf_obj_get_user(const char __user
*pathname
, int flags
)
488 static inline struct net_device
*__dev_map_lookup_elem(struct bpf_map
*map
,
494 static inline void __dev_map_insert_ctx(struct bpf_map
*map
, u32 index
)
498 static inline void __dev_map_flush(struct bpf_map
*map
)
503 struct bpf_cpu_map_entry
*__cpu_map_lookup_elem(struct bpf_map
*map
, u32 key
)
508 static inline void __cpu_map_insert_ctx(struct bpf_map
*map
, u32 index
)
512 static inline void __cpu_map_flush(struct bpf_map
*map
)
517 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry
*rcpu
,
518 struct xdp_buff
*xdp
,
519 struct net_device
*dev_rx
)
524 static inline struct bpf_prog
*bpf_prog_get_type_path(const char *name
,
525 enum bpf_prog_type type
)
527 return ERR_PTR(-EOPNOTSUPP
);
529 #endif /* CONFIG_BPF_SYSCALL */
531 static inline struct bpf_prog
*bpf_prog_get_type(u32 ufd
,
532 enum bpf_prog_type type
)
534 return bpf_prog_get_type_dev(ufd
, type
, false);
537 bool bpf_prog_get_ok(struct bpf_prog
*, enum bpf_prog_type
*, bool);
539 int bpf_prog_offload_compile(struct bpf_prog
*prog
);
540 void bpf_prog_offload_destroy(struct bpf_prog
*prog
);
542 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
543 int bpf_prog_offload_init(struct bpf_prog
*prog
, union bpf_attr
*attr
);
545 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux
*aux
)
550 static inline int bpf_prog_offload_init(struct bpf_prog
*prog
,
551 union bpf_attr
*attr
)
556 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux
*aux
)
560 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
562 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
563 struct sock
*__sock_map_lookup_elem(struct bpf_map
*map
, u32 key
);
564 int sock_map_prog(struct bpf_map
*map
, struct bpf_prog
*prog
, u32 type
);
566 static inline struct sock
*__sock_map_lookup_elem(struct bpf_map
*map
, u32 key
)
571 static inline int sock_map_prog(struct bpf_map
*map
,
572 struct bpf_prog
*prog
,
579 /* verifier prototypes for helper functions called from eBPF programs */
580 extern const struct bpf_func_proto bpf_map_lookup_elem_proto
;
581 extern const struct bpf_func_proto bpf_map_update_elem_proto
;
582 extern const struct bpf_func_proto bpf_map_delete_elem_proto
;
584 extern const struct bpf_func_proto bpf_get_prandom_u32_proto
;
585 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto
;
586 extern const struct bpf_func_proto bpf_get_numa_node_id_proto
;
587 extern const struct bpf_func_proto bpf_tail_call_proto
;
588 extern const struct bpf_func_proto bpf_ktime_get_ns_proto
;
589 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto
;
590 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto
;
591 extern const struct bpf_func_proto bpf_get_current_comm_proto
;
592 extern const struct bpf_func_proto bpf_skb_vlan_push_proto
;
593 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto
;
594 extern const struct bpf_func_proto bpf_get_stackid_proto
;
595 extern const struct bpf_func_proto bpf_sock_map_update_proto
;
597 /* Shared helpers among cBPF and eBPF. */
598 void bpf_user_rnd_init_once(void);
599 u64
bpf_user_rnd_u32(u64 r1
, u64 r2
, u64 r3
, u64 r4
, u64 r5
);
601 #endif /* _LINUX_BPF_H */