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 struct bpf_map
*(*map_alloc
)(union bpf_attr
*attr
);
28 void (*map_release
)(struct bpf_map
*map
, struct file
*map_file
);
29 void (*map_free
)(struct bpf_map
*map
);
30 int (*map_get_next_key
)(struct bpf_map
*map
, void *key
, void *next_key
);
32 /* funcs callable from userspace and from eBPF programs */
33 void *(*map_lookup_elem
)(struct bpf_map
*map
, void *key
);
34 int (*map_update_elem
)(struct bpf_map
*map
, void *key
, void *value
, u64 flags
);
35 int (*map_delete_elem
)(struct bpf_map
*map
, void *key
);
37 /* funcs called by prog_array and perf_event_array map */
38 void *(*map_fd_get_ptr
)(struct bpf_map
*map
, struct file
*map_file
,
40 void (*map_fd_put_ptr
)(void *ptr
);
41 u32 (*map_gen_lookup
)(struct bpf_map
*map
, struct bpf_insn
*insn_buf
);
42 u32 (*map_fd_sys_lookup_elem
)(void *ptr
);
47 enum bpf_map_type map_type
;
55 struct user_struct
*user
;
56 const struct bpf_map_ops
*ops
;
57 struct work_struct work
;
59 struct bpf_map
*inner_map_meta
;
60 char name
[BPF_OBJ_NAME_LEN
];
61 #ifdef CONFIG_SECURITY
66 /* function argument constraints */
68 ARG_DONTCARE
= 0, /* unused argument in helper function */
70 /* the following constraints used to prototype
71 * bpf_map_lookup/update/delete_elem() functions
73 ARG_CONST_MAP_PTR
, /* const argument used as pointer to bpf_map */
74 ARG_PTR_TO_MAP_KEY
, /* pointer to stack used as map key */
75 ARG_PTR_TO_MAP_VALUE
, /* pointer to stack used as map value */
77 /* the following constraints used to prototype bpf_memcmp() and other
78 * functions that access data on eBPF program stack
80 ARG_PTR_TO_MEM
, /* pointer to valid memory (stack, packet, map value) */
81 ARG_PTR_TO_UNINIT_MEM
, /* pointer to memory does not need to be initialized,
82 * helper function must fill all bytes or clear
86 ARG_CONST_SIZE
, /* number of bytes accessed from memory */
87 ARG_CONST_SIZE_OR_ZERO
, /* number of bytes accessed from memory or 0 */
89 ARG_PTR_TO_CTX
, /* pointer to context */
90 ARG_ANYTHING
, /* any (initialized) argument is ok */
93 /* type of values returned from helper functions */
94 enum bpf_return_type
{
95 RET_INTEGER
, /* function returns integer */
96 RET_VOID
, /* function doesn't return anything */
97 RET_PTR_TO_MAP_VALUE_OR_NULL
, /* returns a pointer to map elem value or NULL */
100 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
101 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
102 * instructions after verifying
104 struct bpf_func_proto
{
105 u64 (*func
)(u64 r1
, u64 r2
, u64 r3
, u64 r4
, u64 r5
);
108 enum bpf_return_type ret_type
;
109 enum bpf_arg_type arg1_type
;
110 enum bpf_arg_type arg2_type
;
111 enum bpf_arg_type arg3_type
;
112 enum bpf_arg_type arg4_type
;
113 enum bpf_arg_type arg5_type
;
116 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
117 * the first argument to eBPF programs.
118 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
122 enum bpf_access_type
{
127 /* types of values stored in eBPF registers */
128 /* Pointer types represent:
131 * pointer + (u16) var
132 * pointer + (u16) var + imm
133 * if (range > 0) then [ptr, ptr + range - off) is safe to access
134 * if (id > 0) means that some 'var' was added
135 * if (off > 0) means that 'imm' was added
138 NOT_INIT
= 0, /* nothing was written into register */
139 SCALAR_VALUE
, /* reg doesn't contain a valid pointer */
140 PTR_TO_CTX
, /* reg points to bpf_context */
141 CONST_PTR_TO_MAP
, /* reg points to struct bpf_map */
142 PTR_TO_MAP_VALUE
, /* reg points to map element value */
143 PTR_TO_MAP_VALUE_OR_NULL
,/* points to map elem value or NULL */
144 PTR_TO_STACK
, /* reg == frame_pointer + offset */
145 PTR_TO_PACKET_META
, /* skb->data - meta_len */
146 PTR_TO_PACKET
, /* reg points to skb->data */
147 PTR_TO_PACKET_END
, /* skb->data + headlen */
150 /* The information passed from prog-specific *_is_valid_access
151 * back to the verifier.
153 struct bpf_insn_access_aux
{
154 enum bpf_reg_type reg_type
;
159 bpf_ctx_record_field_size(struct bpf_insn_access_aux
*aux
, u32 size
)
161 aux
->ctx_field_size
= size
;
164 struct bpf_prog_ops
{
165 int (*test_run
)(struct bpf_prog
*prog
, const union bpf_attr
*kattr
,
166 union bpf_attr __user
*uattr
);
169 struct bpf_verifier_ops
{
170 /* return eBPF function prototype for verification */
171 const struct bpf_func_proto
*(*get_func_proto
)(enum bpf_func_id func_id
);
173 /* return true if 'size' wide access at offset 'off' within bpf_context
174 * with 'type' (read or write) is allowed
176 bool (*is_valid_access
)(int off
, int size
, enum bpf_access_type type
,
177 struct bpf_insn_access_aux
*info
);
178 int (*gen_prologue
)(struct bpf_insn
*insn
, bool direct_write
,
179 const struct bpf_prog
*prog
);
180 u32 (*convert_ctx_access
)(enum bpf_access_type type
,
181 const struct bpf_insn
*src
,
182 struct bpf_insn
*dst
,
183 struct bpf_prog
*prog
, u32
*target_size
);
186 struct bpf_dev_offload
{
187 struct bpf_prog
*prog
;
188 struct net_device
*netdev
;
190 struct list_head offloads
;
192 bool verifier_running
;
193 wait_queue_head_t verifier_done
;
196 struct bpf_prog_aux
{
202 struct latch_tree_node ksym_tnode
;
203 struct list_head ksym_lnode
;
204 const struct bpf_prog_ops
*ops
;
205 struct bpf_map
**used_maps
;
206 struct bpf_prog
*prog
;
207 struct user_struct
*user
;
208 u64 load_time
; /* ns since boottime */
209 char name
[BPF_OBJ_NAME_LEN
];
210 #ifdef CONFIG_SECURITY
213 struct bpf_dev_offload
*offload
;
215 struct work_struct work
;
223 /* 'ownership' of prog_array is claimed by the first program that
224 * is going to use this map or by the first program which FD is stored
225 * in the map to make sure that all callers and callees have the same
226 * prog_type and JITed flag
228 enum bpf_prog_type owner_prog_type
;
231 char value
[0] __aligned(8);
232 void *ptrs
[0] __aligned(8);
233 void __percpu
*pptrs
[0] __aligned(8);
237 #define MAX_TAIL_CALL_CNT 32
239 struct bpf_event_entry
{
240 struct perf_event
*event
;
241 struct file
*perf_file
;
242 struct file
*map_file
;
246 bool bpf_prog_array_compatible(struct bpf_array
*array
, const struct bpf_prog
*fp
);
247 int bpf_prog_calc_tag(struct bpf_prog
*fp
);
249 const struct bpf_func_proto
*bpf_get_trace_printk_proto(void);
251 typedef unsigned long (*bpf_ctx_copy_t
)(void *dst
, const void *src
,
252 unsigned long off
, unsigned long len
);
254 u64
bpf_event_output(struct bpf_map
*map
, u64 flags
, void *meta
, u64 meta_size
,
255 void *ctx
, u64 ctx_size
, bpf_ctx_copy_t ctx_copy
);
257 int bpf_prog_test_run_xdp(struct bpf_prog
*prog
, const union bpf_attr
*kattr
,
258 union bpf_attr __user
*uattr
);
259 int bpf_prog_test_run_skb(struct bpf_prog
*prog
, const union bpf_attr
*kattr
,
260 union bpf_attr __user
*uattr
);
262 /* an array of programs to be executed under rcu_lock.
265 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
267 * the structure returned by bpf_prog_array_alloc() should be populated
268 * with program pointers and the last pointer must be NULL.
269 * The user has to keep refcnt on the program and make sure the program
270 * is removed from the array before bpf_prog_put().
271 * The 'struct bpf_prog_array *' should only be replaced with xchg()
272 * since other cpus are walking the array of pointers in parallel.
274 struct bpf_prog_array
{
276 struct bpf_prog
*progs
[0];
279 struct bpf_prog_array __rcu
*bpf_prog_array_alloc(u32 prog_cnt
, gfp_t flags
);
280 void bpf_prog_array_free(struct bpf_prog_array __rcu
*progs
);
281 int bpf_prog_array_length(struct bpf_prog_array __rcu
*progs
);
282 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu
*progs
,
283 __u32 __user
*prog_ids
, u32 cnt
);
285 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu
*progs
,
286 struct bpf_prog
*old_prog
);
287 int bpf_prog_array_copy(struct bpf_prog_array __rcu
*old_array
,
288 struct bpf_prog
*exclude_prog
,
289 struct bpf_prog
*include_prog
,
290 struct bpf_prog_array
**new_array
);
292 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
294 struct bpf_prog **_prog, *__prog; \
295 struct bpf_prog_array *_array; \
298 _array = rcu_dereference(array); \
299 if (unlikely(check_non_null && !_array))\
301 _prog = _array->progs; \
302 while ((__prog = READ_ONCE(*_prog))) { \
303 _ret &= func(__prog, ctx); \
311 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
312 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
314 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
315 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
317 #ifdef CONFIG_BPF_SYSCALL
318 DECLARE_PER_CPU(int, bpf_prog_active
);
320 extern const struct file_operations bpf_map_fops
;
321 extern const struct file_operations bpf_prog_fops
;
323 #define BPF_PROG_TYPE(_id, _name) \
324 extern const struct bpf_prog_ops _name ## _prog_ops; \
325 extern const struct bpf_verifier_ops _name ## _verifier_ops;
326 #define BPF_MAP_TYPE(_id, _ops) \
327 extern const struct bpf_map_ops _ops;
328 #include <linux/bpf_types.h>
332 extern const struct bpf_prog_ops bpf_offload_prog_ops
;
333 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops
;
334 extern const struct bpf_verifier_ops xdp_analyzer_ops
;
336 struct bpf_prog
*bpf_prog_get(u32 ufd
);
337 struct bpf_prog
*bpf_prog_get_type(u32 ufd
, enum bpf_prog_type type
);
338 struct bpf_prog
*bpf_prog_get_type_dev(u32 ufd
, enum bpf_prog_type type
,
339 struct net_device
*netdev
);
340 struct bpf_prog
* __must_check
bpf_prog_add(struct bpf_prog
*prog
, int i
);
341 void bpf_prog_sub(struct bpf_prog
*prog
, int i
);
342 struct bpf_prog
* __must_check
bpf_prog_inc(struct bpf_prog
*prog
);
343 struct bpf_prog
* __must_check
bpf_prog_inc_not_zero(struct bpf_prog
*prog
);
344 void bpf_prog_put(struct bpf_prog
*prog
);
345 int __bpf_prog_charge(struct user_struct
*user
, u32 pages
);
346 void __bpf_prog_uncharge(struct user_struct
*user
, u32 pages
);
348 struct bpf_map
*bpf_map_get_with_uref(u32 ufd
);
349 struct bpf_map
*__bpf_map_get(struct fd f
);
350 struct bpf_map
* __must_check
bpf_map_inc(struct bpf_map
*map
, bool uref
);
351 void bpf_map_put_with_uref(struct bpf_map
*map
);
352 void bpf_map_put(struct bpf_map
*map
);
353 int bpf_map_precharge_memlock(u32 pages
);
354 void *bpf_map_area_alloc(size_t size
, int numa_node
);
355 void bpf_map_area_free(void *base
);
357 extern int sysctl_unprivileged_bpf_disabled
;
359 int bpf_map_new_fd(struct bpf_map
*map
, int flags
);
360 int bpf_prog_new_fd(struct bpf_prog
*prog
);
362 int bpf_obj_pin_user(u32 ufd
, const char __user
*pathname
);
363 int bpf_obj_get_user(const char __user
*pathname
, int flags
);
365 int bpf_percpu_hash_copy(struct bpf_map
*map
, void *key
, void *value
);
366 int bpf_percpu_array_copy(struct bpf_map
*map
, void *key
, void *value
);
367 int bpf_percpu_hash_update(struct bpf_map
*map
, void *key
, void *value
,
369 int bpf_percpu_array_update(struct bpf_map
*map
, void *key
, void *value
,
372 int bpf_stackmap_copy(struct bpf_map
*map
, void *key
, void *value
);
374 int bpf_fd_array_map_update_elem(struct bpf_map
*map
, struct file
*map_file
,
375 void *key
, void *value
, u64 map_flags
);
376 int bpf_fd_array_map_lookup_elem(struct bpf_map
*map
, void *key
, u32
*value
);
377 void bpf_fd_array_map_clear(struct bpf_map
*map
);
378 int bpf_fd_htab_map_update_elem(struct bpf_map
*map
, struct file
*map_file
,
379 void *key
, void *value
, u64 map_flags
);
380 int bpf_fd_htab_map_lookup_elem(struct bpf_map
*map
, void *key
, u32
*value
);
382 int bpf_get_file_flag(int flags
);
384 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
385 * forced to use 'long' read/writes to try to atomically copy long counters.
386 * Best-effort only. No barriers here, since it _will_ race with concurrent
387 * updates from BPF programs. Called from bpf syscall and mostly used with
388 * size 8 or 16 bytes, so ask compiler to inline it.
390 static inline void bpf_long_memcpy(void *dst
, const void *src
, u32 size
)
392 const long *lsrc
= src
;
395 size
/= sizeof(long);
400 /* verify correctness of eBPF program */
401 int bpf_check(struct bpf_prog
**fp
, union bpf_attr
*attr
);
404 struct net_device
*__dev_map_lookup_elem(struct bpf_map
*map
, u32 key
);
405 void __dev_map_insert_ctx(struct bpf_map
*map
, u32 index
);
406 void __dev_map_flush(struct bpf_map
*map
);
408 struct bpf_cpu_map_entry
*__cpu_map_lookup_elem(struct bpf_map
*map
, u32 key
);
409 void __cpu_map_insert_ctx(struct bpf_map
*map
, u32 index
);
410 void __cpu_map_flush(struct bpf_map
*map
);
412 int cpu_map_enqueue(struct bpf_cpu_map_entry
*rcpu
, struct xdp_buff
*xdp
,
413 struct net_device
*dev_rx
);
415 /* Return map's numa specified by userspace */
416 static inline int bpf_map_attr_numa_node(const union bpf_attr
*attr
)
418 return (attr
->map_flags
& BPF_F_NUMA_NODE
) ?
419 attr
->numa_node
: NUMA_NO_NODE
;
422 #else /* !CONFIG_BPF_SYSCALL */
423 static inline struct bpf_prog
*bpf_prog_get(u32 ufd
)
425 return ERR_PTR(-EOPNOTSUPP
);
428 static inline struct bpf_prog
*bpf_prog_get_type(u32 ufd
,
429 enum bpf_prog_type type
)
431 return ERR_PTR(-EOPNOTSUPP
);
434 static inline struct bpf_prog
*bpf_prog_get_type_dev(u32 ufd
,
435 enum bpf_prog_type type
,
436 struct net_device
*netdev
)
438 return ERR_PTR(-EOPNOTSUPP
);
441 static inline struct bpf_prog
* __must_check
bpf_prog_add(struct bpf_prog
*prog
,
444 return ERR_PTR(-EOPNOTSUPP
);
447 static inline void bpf_prog_sub(struct bpf_prog
*prog
, int i
)
451 static inline void bpf_prog_put(struct bpf_prog
*prog
)
455 static inline struct bpf_prog
* __must_check
bpf_prog_inc(struct bpf_prog
*prog
)
457 return ERR_PTR(-EOPNOTSUPP
);
460 static inline struct bpf_prog
*__must_check
461 bpf_prog_inc_not_zero(struct bpf_prog
*prog
)
463 return ERR_PTR(-EOPNOTSUPP
);
466 static inline int __bpf_prog_charge(struct user_struct
*user
, u32 pages
)
471 static inline void __bpf_prog_uncharge(struct user_struct
*user
, u32 pages
)
475 static inline int bpf_obj_get_user(const char __user
*pathname
, int flags
)
480 static inline struct net_device
*__dev_map_lookup_elem(struct bpf_map
*map
,
486 static inline void __dev_map_insert_ctx(struct bpf_map
*map
, u32 index
)
490 static inline void __dev_map_flush(struct bpf_map
*map
)
495 struct bpf_cpu_map_entry
*__cpu_map_lookup_elem(struct bpf_map
*map
, u32 key
)
500 static inline void __cpu_map_insert_ctx(struct bpf_map
*map
, u32 index
)
504 static inline void __cpu_map_flush(struct bpf_map
*map
)
509 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry
*rcpu
,
510 struct xdp_buff
*xdp
,
511 struct net_device
*dev_rx
)
515 #endif /* CONFIG_BPF_SYSCALL */
517 int bpf_prog_offload_compile(struct bpf_prog
*prog
);
518 void bpf_prog_offload_destroy(struct bpf_prog
*prog
);
519 u32
bpf_prog_offload_ifindex(struct bpf_prog
*prog
);
521 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
522 int bpf_prog_offload_init(struct bpf_prog
*prog
, union bpf_attr
*attr
);
524 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux
*aux
)
529 static inline int bpf_prog_offload_init(struct bpf_prog
*prog
,
530 union bpf_attr
*attr
)
535 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux
*aux
)
539 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
541 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
542 struct sock
*__sock_map_lookup_elem(struct bpf_map
*map
, u32 key
);
543 int sock_map_prog(struct bpf_map
*map
, struct bpf_prog
*prog
, u32 type
);
545 static inline struct sock
*__sock_map_lookup_elem(struct bpf_map
*map
, u32 key
)
550 static inline int sock_map_prog(struct bpf_map
*map
,
551 struct bpf_prog
*prog
,
558 /* verifier prototypes for helper functions called from eBPF programs */
559 extern const struct bpf_func_proto bpf_map_lookup_elem_proto
;
560 extern const struct bpf_func_proto bpf_map_update_elem_proto
;
561 extern const struct bpf_func_proto bpf_map_delete_elem_proto
;
563 extern const struct bpf_func_proto bpf_get_prandom_u32_proto
;
564 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto
;
565 extern const struct bpf_func_proto bpf_get_numa_node_id_proto
;
566 extern const struct bpf_func_proto bpf_tail_call_proto
;
567 extern const struct bpf_func_proto bpf_ktime_get_ns_proto
;
568 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto
;
569 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto
;
570 extern const struct bpf_func_proto bpf_get_current_comm_proto
;
571 extern const struct bpf_func_proto bpf_skb_vlan_push_proto
;
572 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto
;
573 extern const struct bpf_func_proto bpf_get_stackid_proto
;
574 extern const struct bpf_func_proto bpf_sock_map_update_proto
;
576 /* Shared helpers among cBPF and eBPF. */
577 void bpf_user_rnd_init_once(void);
578 u64
bpf_user_rnd_u32(u64 r1
, u64 r2
, u64 r3
, u64 r4
, u64 r5
);
580 #endif /* _LINUX_BPF_H */