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>
21 /* map is generic key/value storage optionally accesible by eBPF programs */
23 /* funcs callable from userspace (via syscall) */
24 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
25 void (*map_release)(struct bpf_map *map, struct file *map_file);
26 void (*map_free)(struct bpf_map *map);
27 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
29 /* funcs callable from userspace and from eBPF programs */
30 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
31 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
32 int (*map_delete_elem)(struct bpf_map *map, void *key);
34 /* funcs called by prog_array and perf_event_array map */
35 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
37 void (*map_fd_put_ptr)(void *ptr);
38 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
39 u32 (*map_fd_sys_lookup_elem)(void *ptr);
44 enum bpf_map_type map_type;
51 struct user_struct *user;
52 const struct bpf_map_ops *ops;
53 struct work_struct work;
55 struct bpf_map *inner_map_meta;
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_MEM, /* pointer to valid memory (stack, packet, map value) */
73 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
74 * helper function must fill all bytes or clear
78 ARG_CONST_SIZE, /* number of bytes accessed from memory */
79 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory 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);
100 enum bpf_return_type ret_type;
101 enum bpf_arg_type arg1_type;
102 enum bpf_arg_type arg2_type;
103 enum bpf_arg_type arg3_type;
104 enum bpf_arg_type arg4_type;
105 enum bpf_arg_type arg5_type;
108 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
109 * the first argument to eBPF programs.
110 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
114 enum bpf_access_type {
119 /* types of values stored in eBPF registers */
121 NOT_INIT = 0, /* nothing was written into register */
122 UNKNOWN_VALUE, /* reg doesn't contain a valid pointer */
123 PTR_TO_CTX, /* reg points to bpf_context */
124 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
125 PTR_TO_MAP_VALUE, /* reg points to map element value */
126 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
127 FRAME_PTR, /* reg == frame_pointer */
128 PTR_TO_STACK, /* reg == frame_pointer + imm */
129 CONST_IMM, /* constant integer value */
131 /* PTR_TO_PACKET represents:
134 * skb->data + (u16) var
135 * skb->data + (u16) var + imm
136 * if (range > 0) then [ptr, ptr + range - off) is safe to access
137 * if (id > 0) means that some 'var' was added
138 * if (off > 0) menas that 'imm' was added
141 PTR_TO_PACKET_END, /* skb->data + headlen */
143 /* PTR_TO_MAP_VALUE_ADJ is used for doing pointer math inside of a map
144 * elem value. We only allow this if we can statically verify that
145 * access from this register are going to fall within the size of the
148 PTR_TO_MAP_VALUE_ADJ,
153 /* The information passed from prog-specific *_is_valid_access
154 * back to the verifier.
156 struct bpf_insn_access_aux {
157 enum bpf_reg_type reg_type;
159 int converted_op_size;
162 struct bpf_verifier_ops {
163 /* return eBPF function prototype for verification */
164 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
166 /* return true if 'size' wide access at offset 'off' within bpf_context
167 * with 'type' (read or write) is allowed
169 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
170 struct bpf_insn_access_aux *info);
171 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
172 const struct bpf_prog *prog);
173 u32 (*convert_ctx_access)(enum bpf_access_type type,
174 const struct bpf_insn *src,
175 struct bpf_insn *dst,
176 struct bpf_prog *prog);
177 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
178 union bpf_attr __user *uattr);
181 struct bpf_prog_aux {
187 struct latch_tree_node ksym_tnode;
188 struct list_head ksym_lnode;
189 const struct bpf_verifier_ops *ops;
190 struct bpf_map **used_maps;
191 struct bpf_prog *prog;
192 struct user_struct *user;
194 struct work_struct work;
202 /* 'ownership' of prog_array is claimed by the first program that
203 * is going to use this map or by the first program which FD is stored
204 * in the map to make sure that all callers and callees have the same
205 * prog_type and JITed flag
207 enum bpf_prog_type owner_prog_type;
210 char value[0] __aligned(8);
211 void *ptrs[0] __aligned(8);
212 void __percpu *pptrs[0] __aligned(8);
216 #define MAX_TAIL_CALL_CNT 32
218 struct bpf_event_entry {
219 struct perf_event *event;
220 struct file *perf_file;
221 struct file *map_file;
225 u64 bpf_tail_call(u64 ctx, u64 r2, u64 index, u64 r4, u64 r5);
226 u64 bpf_get_stackid(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
228 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
229 int bpf_prog_calc_tag(struct bpf_prog *fp);
231 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
233 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
234 unsigned long off, unsigned long len);
236 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
237 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
239 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
240 union bpf_attr __user *uattr);
241 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
242 union bpf_attr __user *uattr);
244 #ifdef CONFIG_BPF_SYSCALL
245 DECLARE_PER_CPU(int, bpf_prog_active);
247 #define BPF_PROG_TYPE(_id, _ops) \
248 extern const struct bpf_verifier_ops _ops;
249 #define BPF_MAP_TYPE(_id, _ops) \
250 extern const struct bpf_map_ops _ops;
251 #include <linux/bpf_types.h>
255 struct bpf_prog *bpf_prog_get(u32 ufd);
256 struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type);
257 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
258 void bpf_prog_sub(struct bpf_prog *prog, int i);
259 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
260 void bpf_prog_put(struct bpf_prog *prog);
261 int __bpf_prog_charge(struct user_struct *user, u32 pages);
262 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
264 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
265 struct bpf_map *__bpf_map_get(struct fd f);
266 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
267 void bpf_map_put_with_uref(struct bpf_map *map);
268 void bpf_map_put(struct bpf_map *map);
269 int bpf_map_precharge_memlock(u32 pages);
270 void *bpf_map_area_alloc(size_t size);
271 void bpf_map_area_free(void *base);
273 extern int sysctl_unprivileged_bpf_disabled;
275 int bpf_map_new_fd(struct bpf_map *map);
276 int bpf_prog_new_fd(struct bpf_prog *prog);
278 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
279 int bpf_obj_get_user(const char __user *pathname);
281 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
282 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
283 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
285 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
288 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
290 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
291 void *key, void *value, u64 map_flags);
292 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
293 void bpf_fd_array_map_clear(struct bpf_map *map);
294 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
295 void *key, void *value, u64 map_flags);
296 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
298 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
299 * forced to use 'long' read/writes to try to atomically copy long counters.
300 * Best-effort only. No barriers here, since it _will_ race with concurrent
301 * updates from BPF programs. Called from bpf syscall and mostly used with
302 * size 8 or 16 bytes, so ask compiler to inline it.
304 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
306 const long *lsrc = src;
309 size /= sizeof(long);
314 /* verify correctness of eBPF program */
315 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
317 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
319 return ERR_PTR(-EOPNOTSUPP);
322 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
323 enum bpf_prog_type type)
325 return ERR_PTR(-EOPNOTSUPP);
327 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
330 return ERR_PTR(-EOPNOTSUPP);
333 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
337 static inline void bpf_prog_put(struct bpf_prog *prog)
341 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
343 return ERR_PTR(-EOPNOTSUPP);
346 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
351 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
354 #endif /* CONFIG_BPF_SYSCALL */
356 /* verifier prototypes for helper functions called from eBPF programs */
357 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
358 extern const struct bpf_func_proto bpf_map_update_elem_proto;
359 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
361 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
362 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
363 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
364 extern const struct bpf_func_proto bpf_tail_call_proto;
365 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
366 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
367 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
368 extern const struct bpf_func_proto bpf_get_current_comm_proto;
369 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
370 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
371 extern const struct bpf_func_proto bpf_get_stackid_proto;
373 /* Shared helpers among cBPF and eBPF. */
374 void bpf_user_rnd_init_once(void);
375 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
377 #endif /* _LINUX_BPF_H */