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>
11
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
19 struct perf_event;
20 struct bpf_prog;
21 struct bpf_map;
22
23 /* map is generic key/value storage optionally accesible by eBPF programs */
24 struct bpf_map_ops {
25 /* funcs callable from userspace (via syscall) */
26 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
27 void (*map_release)(struct bpf_map *map, struct file *map_file);
28 void (*map_free)(struct bpf_map *map);
29 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
30 void (*map_release_uref)(struct bpf_map *map);
31 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
32
33 /* funcs callable from userspace and from eBPF programs */
34 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
35 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
36 int (*map_delete_elem)(struct bpf_map *map, void *key);
37
38 /* funcs called by prog_array and perf_event_array map */
39 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
40 int fd);
41 void (*map_fd_put_ptr)(void *ptr);
42 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
43 u32 (*map_fd_sys_lookup_elem)(void *ptr);
44 };
45
46 struct bpf_map {
47 /* 1st cacheline with read-mostly members of which some
48 * are also accessed in fast-path (e.g. ops, max_entries).
49 */
50 const struct bpf_map_ops *ops ____cacheline_aligned;
51 struct bpf_map *inner_map_meta;
52 #ifdef CONFIG_SECURITY
53 void *security;
54 #endif
55 enum bpf_map_type map_type;
56 u32 key_size;
57 u32 value_size;
58 u32 max_entries;
59 u32 map_flags;
60 u32 pages;
61 u32 id;
62 int numa_node;
63 bool unpriv_array;
64 /* 7 bytes hole */
65
66 /* 2nd cacheline with misc members to avoid false sharing
67 * particularly with refcounting.
68 */
69 struct user_struct *user ____cacheline_aligned;
70 atomic_t refcnt;
71 atomic_t usercnt;
72 struct work_struct work;
73 };
74
75 /* function argument constraints */
76 enum bpf_arg_type {
77 ARG_DONTCARE = 0, /* unused argument in helper function */
78
79 /* the following constraints used to prototype
80 * bpf_map_lookup/update/delete_elem() functions
81 */
82 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
83 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
84 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
85
86 /* the following constraints used to prototype bpf_memcmp() and other
87 * functions that access data on eBPF program stack
88 */
89 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
90 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
91 * helper function must fill all bytes or clear
92 * them in error case.
93 */
94
95 ARG_CONST_SIZE, /* number of bytes accessed from memory */
96 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
97
98 ARG_PTR_TO_CTX, /* pointer to context */
99 ARG_ANYTHING, /* any (initialized) argument is ok */
100 };
101
102 /* type of values returned from helper functions */
103 enum bpf_return_type {
104 RET_INTEGER, /* function returns integer */
105 RET_VOID, /* function doesn't return anything */
106 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
107 };
108
109 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
110 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
111 * instructions after verifying
112 */
113 struct bpf_func_proto {
114 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
115 bool gpl_only;
116 bool pkt_access;
117 enum bpf_return_type ret_type;
118 enum bpf_arg_type arg1_type;
119 enum bpf_arg_type arg2_type;
120 enum bpf_arg_type arg3_type;
121 enum bpf_arg_type arg4_type;
122 enum bpf_arg_type arg5_type;
123 };
124
125 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
126 * the first argument to eBPF programs.
127 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
128 */
129 struct bpf_context;
130
131 enum bpf_access_type {
132 BPF_READ = 1,
133 BPF_WRITE = 2
134 };
135
136 /* types of values stored in eBPF registers */
137 /* Pointer types represent:
138 * pointer
139 * pointer + imm
140 * pointer + (u16) var
141 * pointer + (u16) var + imm
142 * if (range > 0) then [ptr, ptr + range - off) is safe to access
143 * if (id > 0) means that some 'var' was added
144 * if (off > 0) means that 'imm' was added
145 */
146 enum bpf_reg_type {
147 NOT_INIT = 0, /* nothing was written into register */
148 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
149 PTR_TO_CTX, /* reg points to bpf_context */
150 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
151 PTR_TO_MAP_VALUE, /* reg points to map element value */
152 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
153 PTR_TO_STACK, /* reg == frame_pointer + offset */
154 PTR_TO_PACKET, /* reg points to skb->data */
155 PTR_TO_PACKET_END, /* skb->data + headlen */
156 };
157
158 /* The information passed from prog-specific *_is_valid_access
159 * back to the verifier.
160 */
161 struct bpf_insn_access_aux {
162 enum bpf_reg_type reg_type;
163 int ctx_field_size;
164 };
165
166 static inline void
bpf_ctx_record_field_size(struct bpf_insn_access_aux * aux,u32 size)167 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
168 {
169 aux->ctx_field_size = size;
170 }
171
172 struct bpf_verifier_ops {
173 /* return eBPF function prototype for verification */
174 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
175
176 /* return true if 'size' wide access at offset 'off' within bpf_context
177 * with 'type' (read or write) is allowed
178 */
179 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
180 struct bpf_insn_access_aux *info);
181 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
182 const struct bpf_prog *prog);
183 u32 (*convert_ctx_access)(enum bpf_access_type type,
184 const struct bpf_insn *src,
185 struct bpf_insn *dst,
186 struct bpf_prog *prog, u32 *target_size);
187 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
188 union bpf_attr __user *uattr);
189 };
190
191 struct bpf_prog_aux {
192 atomic_t refcnt;
193 u32 used_map_cnt;
194 u32 max_ctx_offset;
195 u32 stack_depth;
196 u32 id;
197 struct latch_tree_node ksym_tnode;
198 struct list_head ksym_lnode;
199 const struct bpf_verifier_ops *ops;
200 struct bpf_map **used_maps;
201 struct bpf_prog *prog;
202 struct user_struct *user;
203 #ifdef CONFIG_SECURITY
204 void *security;
205 #endif
206 union {
207 struct work_struct work;
208 struct rcu_head rcu;
209 };
210 };
211
212 struct bpf_array {
213 struct bpf_map map;
214 u32 elem_size;
215 u32 index_mask;
216 /* 'ownership' of prog_array is claimed by the first program that
217 * is going to use this map or by the first program which FD is stored
218 * in the map to make sure that all callers and callees have the same
219 * prog_type and JITed flag
220 */
221 enum bpf_prog_type owner_prog_type;
222 bool owner_jited;
223 union {
224 char value[0] __aligned(8);
225 void *ptrs[0] __aligned(8);
226 void __percpu *pptrs[0] __aligned(8);
227 };
228 };
229
230 #define MAX_TAIL_CALL_CNT 32
231
232 struct bpf_event_entry {
233 struct perf_event *event;
234 struct file *perf_file;
235 struct file *map_file;
236 struct rcu_head rcu;
237 };
238
239 u64 bpf_tail_call(u64 ctx, u64 r2, u64 index, u64 r4, u64 r5);
240 u64 bpf_get_stackid(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
241
242 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
243 int bpf_prog_calc_tag(struct bpf_prog *fp);
244
245 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
246
247 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
248 unsigned long off, unsigned long len);
249
250 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
251 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
252
253 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
254 union bpf_attr __user *uattr);
255 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
256 union bpf_attr __user *uattr);
257
258 #ifdef CONFIG_BPF_SYSCALL
259 DECLARE_PER_CPU(int, bpf_prog_active);
260
261 extern const struct file_operations bpf_map_fops;
262 extern const struct file_operations bpf_prog_fops;
263
264 #define BPF_PROG_TYPE(_id, _ops) \
265 extern const struct bpf_verifier_ops _ops;
266 #define BPF_MAP_TYPE(_id, _ops) \
267 extern const struct bpf_map_ops _ops;
268 #include <linux/bpf_types.h>
269 #undef BPF_PROG_TYPE
270 #undef BPF_MAP_TYPE
271
272 struct bpf_prog *bpf_prog_get(u32 ufd);
273 struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type);
274 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
275 void bpf_prog_sub(struct bpf_prog *prog, int i);
276 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
277 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
278 void bpf_prog_put(struct bpf_prog *prog);
279 int __bpf_prog_charge(struct user_struct *user, u32 pages);
280 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
281
282 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
283 struct bpf_map *__bpf_map_get(struct fd f);
284 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
285 void bpf_map_put_with_uref(struct bpf_map *map);
286 void bpf_map_put(struct bpf_map *map);
287 int bpf_map_precharge_memlock(u32 pages);
288 void *bpf_map_area_alloc(size_t size, int numa_node);
289 void bpf_map_area_free(void *base);
290
291 extern int sysctl_unprivileged_bpf_disabled;
292
293 int bpf_map_new_fd(struct bpf_map *map, int flags);
294 int bpf_prog_new_fd(struct bpf_prog *prog);
295
296 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
297 int bpf_obj_get_user(const char __user *pathname, int flags);
298
299 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
300 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
301 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
302 u64 flags);
303 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
304 u64 flags);
305
306 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
307
308 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
309 void *key, void *value, u64 map_flags);
310 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
311 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
312 void *key, void *value, u64 map_flags);
313 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
314
315 int bpf_get_file_flag(int flags);
316
317 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
318 * forced to use 'long' read/writes to try to atomically copy long counters.
319 * Best-effort only. No barriers here, since it _will_ race with concurrent
320 * updates from BPF programs. Called from bpf syscall and mostly used with
321 * size 8 or 16 bytes, so ask compiler to inline it.
322 */
bpf_long_memcpy(void * dst,const void * src,u32 size)323 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
324 {
325 const long *lsrc = src;
326 long *ldst = dst;
327
328 size /= sizeof(long);
329 while (size--)
330 *ldst++ = *lsrc++;
331 }
332
333 /* verify correctness of eBPF program */
334 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
335
336 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
337
338 /* Map specifics */
339 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
340 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
341 void __dev_map_flush(struct bpf_map *map);
342
343 /* Return map's numa specified by userspace */
bpf_map_attr_numa_node(const union bpf_attr * attr)344 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
345 {
346 return (attr->map_flags & BPF_F_NUMA_NODE) ?
347 attr->numa_node : NUMA_NO_NODE;
348 }
349
350 #else
bpf_prog_get(u32 ufd)351 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
352 {
353 return ERR_PTR(-EOPNOTSUPP);
354 }
355
bpf_prog_get_type(u32 ufd,enum bpf_prog_type type)356 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
357 enum bpf_prog_type type)
358 {
359 return ERR_PTR(-EOPNOTSUPP);
360 }
bpf_prog_add(struct bpf_prog * prog,int i)361 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
362 int i)
363 {
364 return ERR_PTR(-EOPNOTSUPP);
365 }
366
bpf_prog_sub(struct bpf_prog * prog,int i)367 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
368 {
369 }
370
bpf_prog_put(struct bpf_prog * prog)371 static inline void bpf_prog_put(struct bpf_prog *prog)
372 {
373 }
374
bpf_prog_inc(struct bpf_prog * prog)375 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
376 {
377 return ERR_PTR(-EOPNOTSUPP);
378 }
379
380 static inline struct bpf_prog *__must_check
bpf_prog_inc_not_zero(struct bpf_prog * prog)381 bpf_prog_inc_not_zero(struct bpf_prog *prog)
382 {
383 return ERR_PTR(-EOPNOTSUPP);
384 }
385
__bpf_prog_charge(struct user_struct * user,u32 pages)386 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
387 {
388 return 0;
389 }
390
__bpf_prog_uncharge(struct user_struct * user,u32 pages)391 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
392 {
393 }
394
bpf_obj_get_user(const char __user * pathname,int flags)395 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
396 {
397 return -EOPNOTSUPP;
398 }
399
bpf_prog_get_type_path(const char * name,enum bpf_prog_type type)400 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
401 enum bpf_prog_type type)
402 {
403 return ERR_PTR(-EOPNOTSUPP);
404 }
405
__dev_map_lookup_elem(struct bpf_map * map,u32 key)406 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
407 u32 key)
408 {
409 return NULL;
410 }
411
__dev_map_insert_ctx(struct bpf_map * map,u32 index)412 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
413 {
414 }
415
__dev_map_flush(struct bpf_map * map)416 static inline void __dev_map_flush(struct bpf_map *map)
417 {
418 }
419 #endif /* CONFIG_BPF_SYSCALL */
420
421 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
422 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
423 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
424 #else
__sock_map_lookup_elem(struct bpf_map * map,u32 key)425 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
426 {
427 return NULL;
428 }
429
sock_map_prog(struct bpf_map * map,struct bpf_prog * prog,u32 type)430 static inline int sock_map_prog(struct bpf_map *map,
431 struct bpf_prog *prog,
432 u32 type)
433 {
434 return -EOPNOTSUPP;
435 }
436 #endif
437
438 /* verifier prototypes for helper functions called from eBPF programs */
439 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
440 extern const struct bpf_func_proto bpf_map_update_elem_proto;
441 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
442
443 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
444 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
445 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
446 extern const struct bpf_func_proto bpf_tail_call_proto;
447 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
448 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
449 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
450 extern const struct bpf_func_proto bpf_get_current_comm_proto;
451 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
452 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
453 extern const struct bpf_func_proto bpf_get_stackid_proto;
454 extern const struct bpf_func_proto bpf_sock_map_update_proto;
455
456 /* Shared helpers among cBPF and eBPF. */
457 void bpf_user_rnd_init_once(void);
458 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
459
460 #endif /* _LINUX_BPF_H */
461