Commit | Line | Data |
---|---|---|
99c55f7d AS |
1 | /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com |
2 | * | |
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. | |
6 | */ | |
7 | #ifndef _LINUX_BPF_H | |
8 | #define _LINUX_BPF_H 1 | |
9 | ||
10 | #include <uapi/linux/bpf.h> | |
74451e66 | 11 | |
99c55f7d | 12 | #include <linux/workqueue.h> |
db20fd2b | 13 | #include <linux/file.h> |
b121d1e7 | 14 | #include <linux/percpu.h> |
002245cc | 15 | #include <linux/err.h> |
74451e66 | 16 | #include <linux/rbtree_latch.h> |
99c55f7d | 17 | |
3b1efb19 | 18 | struct perf_event; |
99c55f7d AS |
19 | struct bpf_map; |
20 | ||
21 | /* map is generic key/value storage optionally accesible by eBPF programs */ | |
22 | struct bpf_map_ops { | |
23 | /* funcs callable from userspace (via syscall) */ | |
24 | struct bpf_map *(*map_alloc)(union bpf_attr *attr); | |
61d1b6a4 DB |
25 | void (*map_release)(struct bpf_map *map, struct file *map_file); |
26 | void (*map_free)(struct bpf_map *map); | |
db20fd2b AS |
27 | int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); |
28 | ||
29 | /* funcs callable from userspace and from eBPF programs */ | |
30 | void *(*map_lookup_elem)(struct bpf_map *map, void *key); | |
3274f520 | 31 | int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); |
db20fd2b | 32 | int (*map_delete_elem)(struct bpf_map *map, void *key); |
2a36f0b9 WN |
33 | |
34 | /* funcs called by prog_array and perf_event_array map */ | |
d056a788 DB |
35 | void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, |
36 | int fd); | |
37 | void (*map_fd_put_ptr)(void *ptr); | |
81ed18ab | 38 | u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); |
99c55f7d AS |
39 | }; |
40 | ||
41 | struct bpf_map { | |
42 | atomic_t refcnt; | |
43 | enum bpf_map_type map_type; | |
44 | u32 key_size; | |
45 | u32 value_size; | |
46 | u32 max_entries; | |
6c905981 | 47 | u32 map_flags; |
aaac3ba9 AS |
48 | u32 pages; |
49 | struct user_struct *user; | |
a2c83fff | 50 | const struct bpf_map_ops *ops; |
99c55f7d | 51 | struct work_struct work; |
c9da161c | 52 | atomic_t usercnt; |
56f668df | 53 | struct bpf_map *inner_map_meta; |
99c55f7d AS |
54 | }; |
55 | ||
56 | struct bpf_map_type_list { | |
57 | struct list_head list_node; | |
a2c83fff | 58 | const struct bpf_map_ops *ops; |
99c55f7d AS |
59 | enum bpf_map_type type; |
60 | }; | |
61 | ||
17a52670 AS |
62 | /* function argument constraints */ |
63 | enum bpf_arg_type { | |
80f1d68c | 64 | ARG_DONTCARE = 0, /* unused argument in helper function */ |
17a52670 AS |
65 | |
66 | /* the following constraints used to prototype | |
67 | * bpf_map_lookup/update/delete_elem() functions | |
68 | */ | |
69 | ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ | |
70 | ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ | |
71 | ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ | |
72 | ||
73 | /* the following constraints used to prototype bpf_memcmp() and other | |
74 | * functions that access data on eBPF program stack | |
75 | */ | |
39f19ebb AS |
76 | ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ |
77 | ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized, | |
78 | * helper function must fill all bytes or clear | |
79 | * them in error case. | |
435faee1 DB |
80 | */ |
81 | ||
39f19ebb AS |
82 | ARG_CONST_SIZE, /* number of bytes accessed from memory */ |
83 | ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ | |
80f1d68c | 84 | |
608cd71a | 85 | ARG_PTR_TO_CTX, /* pointer to context */ |
80f1d68c | 86 | ARG_ANYTHING, /* any (initialized) argument is ok */ |
17a52670 AS |
87 | }; |
88 | ||
89 | /* type of values returned from helper functions */ | |
90 | enum bpf_return_type { | |
91 | RET_INTEGER, /* function returns integer */ | |
92 | RET_VOID, /* function doesn't return anything */ | |
93 | RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */ | |
94 | }; | |
95 | ||
09756af4 AS |
96 | /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs |
97 | * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL | |
98 | * instructions after verifying | |
99 | */ | |
100 | struct bpf_func_proto { | |
101 | u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); | |
102 | bool gpl_only; | |
36bbef52 | 103 | bool pkt_access; |
17a52670 AS |
104 | enum bpf_return_type ret_type; |
105 | enum bpf_arg_type arg1_type; | |
106 | enum bpf_arg_type arg2_type; | |
107 | enum bpf_arg_type arg3_type; | |
108 | enum bpf_arg_type arg4_type; | |
109 | enum bpf_arg_type arg5_type; | |
110 | }; | |
111 | ||
112 | /* bpf_context is intentionally undefined structure. Pointer to bpf_context is | |
113 | * the first argument to eBPF programs. | |
114 | * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' | |
115 | */ | |
116 | struct bpf_context; | |
117 | ||
118 | enum bpf_access_type { | |
119 | BPF_READ = 1, | |
120 | BPF_WRITE = 2 | |
09756af4 AS |
121 | }; |
122 | ||
19de99f7 AS |
123 | /* types of values stored in eBPF registers */ |
124 | enum bpf_reg_type { | |
125 | NOT_INIT = 0, /* nothing was written into register */ | |
126 | UNKNOWN_VALUE, /* reg doesn't contain a valid pointer */ | |
127 | PTR_TO_CTX, /* reg points to bpf_context */ | |
128 | CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ | |
129 | PTR_TO_MAP_VALUE, /* reg points to map element value */ | |
130 | PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */ | |
131 | FRAME_PTR, /* reg == frame_pointer */ | |
132 | PTR_TO_STACK, /* reg == frame_pointer + imm */ | |
133 | CONST_IMM, /* constant integer value */ | |
134 | ||
135 | /* PTR_TO_PACKET represents: | |
136 | * skb->data | |
137 | * skb->data + imm | |
138 | * skb->data + (u16) var | |
139 | * skb->data + (u16) var + imm | |
140 | * if (range > 0) then [ptr, ptr + range - off) is safe to access | |
141 | * if (id > 0) means that some 'var' was added | |
142 | * if (off > 0) menas that 'imm' was added | |
143 | */ | |
144 | PTR_TO_PACKET, | |
145 | PTR_TO_PACKET_END, /* skb->data + headlen */ | |
48461135 JB |
146 | |
147 | /* PTR_TO_MAP_VALUE_ADJ is used for doing pointer math inside of a map | |
148 | * elem value. We only allow this if we can statically verify that | |
149 | * access from this register are going to fall within the size of the | |
150 | * map element. | |
151 | */ | |
152 | PTR_TO_MAP_VALUE_ADJ, | |
19de99f7 AS |
153 | }; |
154 | ||
ff936a04 AS |
155 | struct bpf_prog; |
156 | ||
09756af4 AS |
157 | struct bpf_verifier_ops { |
158 | /* return eBPF function prototype for verification */ | |
159 | const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id); | |
17a52670 AS |
160 | |
161 | /* return true if 'size' wide access at offset 'off' within bpf_context | |
162 | * with 'type' (read or write) is allowed | |
163 | */ | |
19de99f7 AS |
164 | bool (*is_valid_access)(int off, int size, enum bpf_access_type type, |
165 | enum bpf_reg_type *reg_type); | |
36bbef52 DB |
166 | int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, |
167 | const struct bpf_prog *prog); | |
6b8cc1d1 DB |
168 | u32 (*convert_ctx_access)(enum bpf_access_type type, |
169 | const struct bpf_insn *src, | |
170 | struct bpf_insn *dst, | |
171 | struct bpf_prog *prog); | |
09756af4 AS |
172 | }; |
173 | ||
174 | struct bpf_prog_type_list { | |
175 | struct list_head list_node; | |
a2c83fff | 176 | const struct bpf_verifier_ops *ops; |
09756af4 AS |
177 | enum bpf_prog_type type; |
178 | }; | |
179 | ||
09756af4 AS |
180 | struct bpf_prog_aux { |
181 | atomic_t refcnt; | |
24701ece | 182 | u32 used_map_cnt; |
32bbe007 | 183 | u32 max_ctx_offset; |
74451e66 DB |
184 | struct latch_tree_node ksym_tnode; |
185 | struct list_head ksym_lnode; | |
a2c83fff | 186 | const struct bpf_verifier_ops *ops; |
09756af4 | 187 | struct bpf_map **used_maps; |
09756af4 | 188 | struct bpf_prog *prog; |
aaac3ba9 | 189 | struct user_struct *user; |
abf2e7d6 AS |
190 | union { |
191 | struct work_struct work; | |
192 | struct rcu_head rcu; | |
193 | }; | |
09756af4 AS |
194 | }; |
195 | ||
04fd61ab AS |
196 | struct bpf_array { |
197 | struct bpf_map map; | |
198 | u32 elem_size; | |
199 | /* 'ownership' of prog_array is claimed by the first program that | |
200 | * is going to use this map or by the first program which FD is stored | |
201 | * in the map to make sure that all callers and callees have the same | |
202 | * prog_type and JITed flag | |
203 | */ | |
204 | enum bpf_prog_type owner_prog_type; | |
205 | bool owner_jited; | |
206 | union { | |
207 | char value[0] __aligned(8); | |
2a36f0b9 | 208 | void *ptrs[0] __aligned(8); |
a10423b8 | 209 | void __percpu *pptrs[0] __aligned(8); |
04fd61ab AS |
210 | }; |
211 | }; | |
3b1efb19 | 212 | |
04fd61ab AS |
213 | #define MAX_TAIL_CALL_CNT 32 |
214 | ||
3b1efb19 DB |
215 | struct bpf_event_entry { |
216 | struct perf_event *event; | |
217 | struct file *perf_file; | |
218 | struct file *map_file; | |
219 | struct rcu_head rcu; | |
220 | }; | |
221 | ||
04fd61ab | 222 | u64 bpf_tail_call(u64 ctx, u64 r2, u64 index, u64 r4, u64 r5); |
9940d67c | 223 | u64 bpf_get_stackid(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); |
d056a788 | 224 | |
04fd61ab | 225 | bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); |
f1f7714e | 226 | int bpf_prog_calc_tag(struct bpf_prog *fp); |
bd570ff9 | 227 | |
0756ea3e | 228 | const struct bpf_func_proto *bpf_get_trace_printk_proto(void); |
555c8a86 DB |
229 | |
230 | typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, | |
aa7145c1 | 231 | unsigned long off, unsigned long len); |
555c8a86 DB |
232 | |
233 | u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, | |
234 | void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); | |
04fd61ab | 235 | |
89aa0758 | 236 | #ifdef CONFIG_BPF_SYSCALL |
b121d1e7 AS |
237 | DECLARE_PER_CPU(int, bpf_prog_active); |
238 | ||
0fc174de | 239 | void bpf_register_prog_type(struct bpf_prog_type_list *tl); |
61e021f3 | 240 | void bpf_register_map_type(struct bpf_map_type_list *tl); |
0fc174de | 241 | |
0fc174de | 242 | struct bpf_prog *bpf_prog_get(u32 ufd); |
113214be | 243 | struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type); |
6d67942d | 244 | struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i); |
c540594f | 245 | void bpf_prog_sub(struct bpf_prog *prog, int i); |
6d67942d | 246 | struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog); |
61e021f3 | 247 | void bpf_prog_put(struct bpf_prog *prog); |
5ccb071e DB |
248 | int __bpf_prog_charge(struct user_struct *user, u32 pages); |
249 | void __bpf_prog_uncharge(struct user_struct *user, u32 pages); | |
61e021f3 | 250 | |
c9da161c | 251 | struct bpf_map *bpf_map_get_with_uref(u32 ufd); |
c2101297 | 252 | struct bpf_map *__bpf_map_get(struct fd f); |
6d67942d | 253 | struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref); |
c9da161c | 254 | void bpf_map_put_with_uref(struct bpf_map *map); |
61e021f3 | 255 | void bpf_map_put(struct bpf_map *map); |
6c905981 | 256 | int bpf_map_precharge_memlock(u32 pages); |
d407bd25 DB |
257 | void *bpf_map_area_alloc(size_t size); |
258 | void bpf_map_area_free(void *base); | |
61e021f3 | 259 | |
1be7f75d AS |
260 | extern int sysctl_unprivileged_bpf_disabled; |
261 | ||
b2197755 DB |
262 | int bpf_map_new_fd(struct bpf_map *map); |
263 | int bpf_prog_new_fd(struct bpf_prog *prog); | |
264 | ||
265 | int bpf_obj_pin_user(u32 ufd, const char __user *pathname); | |
266 | int bpf_obj_get_user(const char __user *pathname); | |
267 | ||
15a07b33 AS |
268 | int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); |
269 | int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); | |
270 | int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, | |
271 | u64 flags); | |
272 | int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, | |
273 | u64 flags); | |
d056a788 | 274 | |
557c0c6e | 275 | int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); |
15a07b33 | 276 | |
d056a788 DB |
277 | int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, |
278 | void *key, void *value, u64 map_flags); | |
279 | void bpf_fd_array_map_clear(struct bpf_map *map); | |
bcc6b1b7 MKL |
280 | int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, |
281 | void *key, void *value, u64 map_flags); | |
d056a788 | 282 | |
15a07b33 AS |
283 | /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and |
284 | * forced to use 'long' read/writes to try to atomically copy long counters. | |
285 | * Best-effort only. No barriers here, since it _will_ race with concurrent | |
286 | * updates from BPF programs. Called from bpf syscall and mostly used with | |
287 | * size 8 or 16 bytes, so ask compiler to inline it. | |
288 | */ | |
289 | static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) | |
290 | { | |
291 | const long *lsrc = src; | |
292 | long *ldst = dst; | |
293 | ||
294 | size /= sizeof(long); | |
295 | while (size--) | |
296 | *ldst++ = *lsrc++; | |
297 | } | |
298 | ||
61e021f3 | 299 | /* verify correctness of eBPF program */ |
9bac3d6d | 300 | int bpf_check(struct bpf_prog **fp, union bpf_attr *attr); |
89aa0758 | 301 | #else |
0fc174de DB |
302 | static inline void bpf_register_prog_type(struct bpf_prog_type_list *tl) |
303 | { | |
304 | } | |
305 | ||
306 | static inline struct bpf_prog *bpf_prog_get(u32 ufd) | |
307 | { | |
308 | return ERR_PTR(-EOPNOTSUPP); | |
309 | } | |
310 | ||
113214be DB |
311 | static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, |
312 | enum bpf_prog_type type) | |
313 | { | |
314 | return ERR_PTR(-EOPNOTSUPP); | |
315 | } | |
6d67942d DB |
316 | static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, |
317 | int i) | |
cc2e0b3f BB |
318 | { |
319 | return ERR_PTR(-EOPNOTSUPP); | |
320 | } | |
113214be | 321 | |
c540594f DB |
322 | static inline void bpf_prog_sub(struct bpf_prog *prog, int i) |
323 | { | |
324 | } | |
325 | ||
0fc174de DB |
326 | static inline void bpf_prog_put(struct bpf_prog *prog) |
327 | { | |
328 | } | |
6d67942d DB |
329 | |
330 | static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog) | |
aa6a5f3c AS |
331 | { |
332 | return ERR_PTR(-EOPNOTSUPP); | |
333 | } | |
5ccb071e DB |
334 | |
335 | static inline int __bpf_prog_charge(struct user_struct *user, u32 pages) | |
336 | { | |
337 | return 0; | |
338 | } | |
339 | ||
340 | static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages) | |
341 | { | |
342 | } | |
61e021f3 | 343 | #endif /* CONFIG_BPF_SYSCALL */ |
09756af4 | 344 | |
d0003ec0 | 345 | /* verifier prototypes for helper functions called from eBPF programs */ |
a2c83fff DB |
346 | extern const struct bpf_func_proto bpf_map_lookup_elem_proto; |
347 | extern const struct bpf_func_proto bpf_map_update_elem_proto; | |
348 | extern const struct bpf_func_proto bpf_map_delete_elem_proto; | |
d0003ec0 | 349 | |
03e69b50 | 350 | extern const struct bpf_func_proto bpf_get_prandom_u32_proto; |
c04167ce | 351 | extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; |
2d0e30c3 | 352 | extern const struct bpf_func_proto bpf_get_numa_node_id_proto; |
04fd61ab | 353 | extern const struct bpf_func_proto bpf_tail_call_proto; |
17ca8cbf | 354 | extern const struct bpf_func_proto bpf_ktime_get_ns_proto; |
ffeedafb AS |
355 | extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; |
356 | extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; | |
357 | extern const struct bpf_func_proto bpf_get_current_comm_proto; | |
4e10df9a AS |
358 | extern const struct bpf_func_proto bpf_skb_vlan_push_proto; |
359 | extern const struct bpf_func_proto bpf_skb_vlan_pop_proto; | |
d5a3b1f6 | 360 | extern const struct bpf_func_proto bpf_get_stackid_proto; |
03e69b50 | 361 | |
3ad00405 DB |
362 | /* Shared helpers among cBPF and eBPF. */ |
363 | void bpf_user_rnd_init_once(void); | |
364 | u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); | |
365 | ||
99c55f7d | 366 | #endif /* _LINUX_BPF_H */ |