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>
11 #include <linux/workqueue.h>
12 #include <linux/file.h>
13 #include <linux/percpu.h>
14 #include <linux/err.h>
19 /* map is generic key/value storage optionally accesible by eBPF programs */
21 /* funcs callable from userspace (via syscall) */
22 struct bpf_map
*(*map_alloc
)(union bpf_attr
*attr
);
23 void (*map_release
)(struct bpf_map
*map
, struct file
*map_file
);
24 void (*map_free
)(struct bpf_map
*map
);
25 int (*map_get_next_key
)(struct bpf_map
*map
, void *key
, void *next_key
);
27 /* funcs callable from userspace and from eBPF programs */
28 void *(*map_lookup_elem
)(struct bpf_map
*map
, void *key
);
29 int (*map_update_elem
)(struct bpf_map
*map
, void *key
, void *value
, u64 flags
);
30 int (*map_delete_elem
)(struct bpf_map
*map
, void *key
);
32 /* funcs called by prog_array and perf_event_array map */
33 void *(*map_fd_get_ptr
)(struct bpf_map
*map
, struct file
*map_file
,
35 void (*map_fd_put_ptr
)(void *ptr
);
40 enum bpf_map_type map_type
;
46 struct user_struct
*user
;
47 const struct bpf_map_ops
*ops
;
48 struct work_struct work
;
52 struct bpf_map_type_list
{
53 struct list_head list_node
;
54 const struct bpf_map_ops
*ops
;
55 enum bpf_map_type type
;
58 /* function argument constraints */
60 ARG_DONTCARE
= 0, /* unused argument in helper function */
62 /* the following constraints used to prototype
63 * bpf_map_lookup/update/delete_elem() functions
65 ARG_CONST_MAP_PTR
, /* const argument used as pointer to bpf_map */
66 ARG_PTR_TO_MAP_KEY
, /* pointer to stack used as map key */
67 ARG_PTR_TO_MAP_VALUE
, /* pointer to stack used as map value */
69 /* the following constraints used to prototype bpf_memcmp() and other
70 * functions that access data on eBPF program stack
72 ARG_PTR_TO_STACK
, /* any pointer to eBPF program stack */
73 ARG_PTR_TO_RAW_STACK
, /* any pointer to eBPF program stack, area does not
74 * need to be initialized, helper function must fill
75 * all bytes or clear them in error case.
78 ARG_CONST_STACK_SIZE
, /* number of bytes accessed from stack */
79 ARG_CONST_STACK_SIZE_OR_ZERO
, /* number of bytes accessed from stack or 0 */
81 ARG_PTR_TO_CTX
, /* pointer to context */
82 ARG_ANYTHING
, /* any (initialized) argument is ok */
85 /* type of values returned from helper functions */
86 enum bpf_return_type
{
87 RET_INTEGER
, /* function returns integer */
88 RET_VOID
, /* function doesn't return anything */
89 RET_PTR_TO_MAP_VALUE_OR_NULL
, /* returns a pointer to map elem value or NULL */
92 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
93 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
94 * instructions after verifying
96 struct bpf_func_proto
{
97 u64 (*func
)(u64 r1
, u64 r2
, u64 r3
, u64 r4
, u64 r5
);
99 enum bpf_return_type ret_type
;
100 enum bpf_arg_type arg1_type
;
101 enum bpf_arg_type arg2_type
;
102 enum bpf_arg_type arg3_type
;
103 enum bpf_arg_type arg4_type
;
104 enum bpf_arg_type arg5_type
;
107 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
108 * the first argument to eBPF programs.
109 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
113 enum bpf_access_type
{
118 /* types of values stored in eBPF registers */
120 NOT_INIT
= 0, /* nothing was written into register */
121 UNKNOWN_VALUE
, /* reg doesn't contain a valid pointer */
122 PTR_TO_CTX
, /* reg points to bpf_context */
123 CONST_PTR_TO_MAP
, /* reg points to struct bpf_map */
124 PTR_TO_MAP_VALUE
, /* reg points to map element value */
125 PTR_TO_MAP_VALUE_OR_NULL
,/* points to map elem value or NULL */
126 FRAME_PTR
, /* reg == frame_pointer */
127 PTR_TO_STACK
, /* reg == frame_pointer + imm */
128 CONST_IMM
, /* constant integer value */
130 /* PTR_TO_PACKET represents:
133 * skb->data + (u16) var
134 * skb->data + (u16) var + imm
135 * if (range > 0) then [ptr, ptr + range - off) is safe to access
136 * if (id > 0) means that some 'var' was added
137 * if (off > 0) menas that 'imm' was added
140 PTR_TO_PACKET_END
, /* skb->data + headlen */
145 struct bpf_verifier_ops
{
146 /* return eBPF function prototype for verification */
147 const struct bpf_func_proto
*(*get_func_proto
)(enum bpf_func_id func_id
);
149 /* return true if 'size' wide access at offset 'off' within bpf_context
150 * with 'type' (read or write) is allowed
152 bool (*is_valid_access
)(int off
, int size
, enum bpf_access_type type
,
153 enum bpf_reg_type
*reg_type
);
155 u32 (*convert_ctx_access
)(enum bpf_access_type type
, int dst_reg
,
156 int src_reg
, int ctx_off
,
157 struct bpf_insn
*insn
, struct bpf_prog
*prog
);
160 struct bpf_prog_type_list
{
161 struct list_head list_node
;
162 const struct bpf_verifier_ops
*ops
;
163 enum bpf_prog_type type
;
166 struct bpf_prog_aux
{
170 const struct bpf_verifier_ops
*ops
;
171 struct bpf_map
**used_maps
;
172 struct bpf_prog
*prog
;
173 struct user_struct
*user
;
175 struct work_struct work
;
183 /* 'ownership' of prog_array is claimed by the first program that
184 * is going to use this map or by the first program which FD is stored
185 * in the map to make sure that all callers and callees have the same
186 * prog_type and JITed flag
188 enum bpf_prog_type owner_prog_type
;
191 char value
[0] __aligned(8);
192 void *ptrs
[0] __aligned(8);
193 void __percpu
*pptrs
[0] __aligned(8);
197 #define MAX_TAIL_CALL_CNT 32
199 struct bpf_event_entry
{
200 struct perf_event
*event
;
201 struct file
*perf_file
;
202 struct file
*map_file
;
206 u64
bpf_tail_call(u64 ctx
, u64 r2
, u64 index
, u64 r4
, u64 r5
);
207 u64
bpf_get_stackid(u64 r1
, u64 r2
, u64 r3
, u64 r4
, u64 r5
);
209 bool bpf_prog_array_compatible(struct bpf_array
*array
, const struct bpf_prog
*fp
);
211 const struct bpf_func_proto
*bpf_get_trace_printk_proto(void);
213 typedef unsigned long (*bpf_ctx_copy_t
)(void *dst
, const void *src
,
214 unsigned long off
, unsigned long len
);
216 u64
bpf_event_output(struct bpf_map
*map
, u64 flags
, void *meta
, u64 meta_size
,
217 void *ctx
, u64 ctx_size
, bpf_ctx_copy_t ctx_copy
);
219 #ifdef CONFIG_BPF_SYSCALL
220 DECLARE_PER_CPU(int, bpf_prog_active
);
222 void bpf_register_prog_type(struct bpf_prog_type_list
*tl
);
223 void bpf_register_map_type(struct bpf_map_type_list
*tl
);
225 struct bpf_prog
*bpf_prog_get(u32 ufd
);
226 struct bpf_prog
*bpf_prog_get_type(u32 ufd
, enum bpf_prog_type type
);
227 struct bpf_prog
*bpf_prog_add(struct bpf_prog
*prog
, int i
);
228 struct bpf_prog
*bpf_prog_inc(struct bpf_prog
*prog
);
229 void bpf_prog_put(struct bpf_prog
*prog
);
231 struct bpf_map
*bpf_map_get_with_uref(u32 ufd
);
232 struct bpf_map
*__bpf_map_get(struct fd f
);
233 struct bpf_map
*bpf_map_inc(struct bpf_map
*map
, bool uref
);
234 void bpf_map_put_with_uref(struct bpf_map
*map
);
235 void bpf_map_put(struct bpf_map
*map
);
236 int bpf_map_precharge_memlock(u32 pages
);
238 extern int sysctl_unprivileged_bpf_disabled
;
240 int bpf_map_new_fd(struct bpf_map
*map
);
241 int bpf_prog_new_fd(struct bpf_prog
*prog
);
243 int bpf_obj_pin_user(u32 ufd
, const char __user
*pathname
);
244 int bpf_obj_get_user(const char __user
*pathname
);
246 int bpf_percpu_hash_copy(struct bpf_map
*map
, void *key
, void *value
);
247 int bpf_percpu_array_copy(struct bpf_map
*map
, void *key
, void *value
);
248 int bpf_percpu_hash_update(struct bpf_map
*map
, void *key
, void *value
,
250 int bpf_percpu_array_update(struct bpf_map
*map
, void *key
, void *value
,
253 int bpf_stackmap_copy(struct bpf_map
*map
, void *key
, void *value
);
255 int bpf_fd_array_map_update_elem(struct bpf_map
*map
, struct file
*map_file
,
256 void *key
, void *value
, u64 map_flags
);
257 void bpf_fd_array_map_clear(struct bpf_map
*map
);
259 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
260 * forced to use 'long' read/writes to try to atomically copy long counters.
261 * Best-effort only. No barriers here, since it _will_ race with concurrent
262 * updates from BPF programs. Called from bpf syscall and mostly used with
263 * size 8 or 16 bytes, so ask compiler to inline it.
265 static inline void bpf_long_memcpy(void *dst
, const void *src
, u32 size
)
267 const long *lsrc
= src
;
270 size
/= sizeof(long);
275 /* verify correctness of eBPF program */
276 int bpf_check(struct bpf_prog
**fp
, union bpf_attr
*attr
);
278 static inline void bpf_register_prog_type(struct bpf_prog_type_list
*tl
)
282 static inline struct bpf_prog
*bpf_prog_get(u32 ufd
)
284 return ERR_PTR(-EOPNOTSUPP
);
287 static inline struct bpf_prog
*bpf_prog_get_type(u32 ufd
,
288 enum bpf_prog_type type
)
290 return ERR_PTR(-EOPNOTSUPP
);
292 static inline struct bpf_prog
*bpf_prog_add(struct bpf_prog
*prog
, int i
)
294 return ERR_PTR(-EOPNOTSUPP
);
297 static inline void bpf_prog_put(struct bpf_prog
*prog
)
300 #endif /* CONFIG_BPF_SYSCALL */
302 /* verifier prototypes for helper functions called from eBPF programs */
303 extern const struct bpf_func_proto bpf_map_lookup_elem_proto
;
304 extern const struct bpf_func_proto bpf_map_update_elem_proto
;
305 extern const struct bpf_func_proto bpf_map_delete_elem_proto
;
307 extern const struct bpf_func_proto bpf_get_prandom_u32_proto
;
308 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto
;
309 extern const struct bpf_func_proto bpf_tail_call_proto
;
310 extern const struct bpf_func_proto bpf_ktime_get_ns_proto
;
311 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto
;
312 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto
;
313 extern const struct bpf_func_proto bpf_get_current_comm_proto
;
314 extern const struct bpf_func_proto bpf_skb_vlan_push_proto
;
315 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto
;
316 extern const struct bpf_func_proto bpf_get_stackid_proto
;
318 /* Shared helpers among cBPF and eBPF. */
319 void bpf_user_rnd_init_once(void);
320 u64
bpf_user_rnd_u32(u64 r1
, u64 r2
, u64 r3
, u64 r4
, u64 r5
);
322 #endif /* _LINUX_BPF_H */