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 | #include <linux/wait.h> |
19 | #include <linux/u64_stats_sync.h> |
20 | |
21 | struct bpf_verifier_env; |
22 | struct perf_event; |
23 | struct bpf_prog; |
24 | struct bpf_map; |
25 | struct sock; |
26 | struct seq_file; |
27 | struct btf; |
28 | struct btf_type; |
29 | |
30 | /* map is generic key/value storage optionally accesible by eBPF programs */ |
31 | struct bpf_map_ops { |
32 | /* funcs callable from userspace (via syscall) */ |
33 | int (*map_alloc_check)(union bpf_attr *attr); |
34 | struct bpf_map *(*map_alloc)(union bpf_attr *attr); |
35 | void (*map_release)(struct bpf_map *map, struct file *map_file); |
36 | void (*map_free)(struct bpf_map *map); |
37 | int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key); |
38 | void (*map_release_uref)(struct bpf_map *map); |
39 | |
40 | /* funcs callable from userspace and from eBPF programs */ |
41 | void *(*map_lookup_elem)(struct bpf_map *map, void *key); |
42 | int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags); |
43 | int (*map_delete_elem)(struct bpf_map *map, void *key); |
44 | int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags); |
45 | int (*map_pop_elem)(struct bpf_map *map, void *value); |
46 | int (*map_peek_elem)(struct bpf_map *map, void *value); |
47 | |
48 | /* funcs called by prog_array and perf_event_array map */ |
49 | void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, |
50 | int fd); |
51 | void (*map_fd_put_ptr)(void *ptr); |
52 | u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf); |
53 | u32 (*map_fd_sys_lookup_elem)(void *ptr); |
54 | void (*map_seq_show_elem)(struct bpf_map *map, void *key, |
55 | struct seq_file *m); |
56 | int (*map_check_btf)(const struct bpf_map *map, |
57 | const struct btf *btf, |
58 | const struct btf_type *key_type, |
59 | const struct btf_type *value_type); |
60 | }; |
61 | |
62 | struct bpf_map { |
63 | /* The first two cachelines with read-mostly members of which some |
64 | * are also accessed in fast-path (e.g. ops, max_entries). |
65 | */ |
66 | const struct bpf_map_ops *ops ____cacheline_aligned; |
67 | struct bpf_map *inner_map_meta; |
68 | #ifdef CONFIG_SECURITY |
69 | void *security; |
70 | #endif |
71 | enum bpf_map_type map_type; |
72 | u32 key_size; |
73 | u32 value_size; |
74 | u32 max_entries; |
75 | u32 map_flags; |
76 | int spin_lock_off; /* >=0 valid offset, <0 error */ |
77 | u32 id; |
78 | int numa_node; |
79 | u32 btf_key_type_id; |
80 | u32 btf_value_type_id; |
81 | struct btf *btf; |
82 | u32 pages; |
83 | bool unpriv_array; |
84 | /* 51 bytes hole */ |
85 | |
86 | /* The 3rd and 4th cacheline with misc members to avoid false sharing |
87 | * particularly with refcounting. |
88 | */ |
89 | struct user_struct *user ____cacheline_aligned; |
90 | atomic_t refcnt; |
91 | atomic_t usercnt; |
92 | struct work_struct work; |
93 | char name[BPF_OBJ_NAME_LEN]; |
94 | }; |
95 | |
96 | static inline bool map_value_has_spin_lock(const struct bpf_map *map) |
97 | { |
98 | return map->spin_lock_off >= 0; |
99 | } |
100 | |
101 | static inline void check_and_init_map_lock(struct bpf_map *map, void *dst) |
102 | { |
103 | if (likely(!map_value_has_spin_lock(map))) |
104 | return; |
105 | *(struct bpf_spin_lock *)(dst + map->spin_lock_off) = |
106 | (struct bpf_spin_lock){}; |
107 | } |
108 | |
109 | /* copy everything but bpf_spin_lock */ |
110 | static inline void copy_map_value(struct bpf_map *map, void *dst, void *src) |
111 | { |
112 | if (unlikely(map_value_has_spin_lock(map))) { |
113 | u32 off = map->spin_lock_off; |
114 | |
115 | memcpy(dst, src, off); |
116 | memcpy(dst + off + sizeof(struct bpf_spin_lock), |
117 | src + off + sizeof(struct bpf_spin_lock), |
118 | map->value_size - off - sizeof(struct bpf_spin_lock)); |
119 | } else { |
120 | memcpy(dst, src, map->value_size); |
121 | } |
122 | } |
123 | void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, |
124 | bool lock_src); |
125 | |
126 | struct bpf_offload_dev; |
127 | struct bpf_offloaded_map; |
128 | |
129 | struct bpf_map_dev_ops { |
130 | int (*map_get_next_key)(struct bpf_offloaded_map *map, |
131 | void *key, void *next_key); |
132 | int (*map_lookup_elem)(struct bpf_offloaded_map *map, |
133 | void *key, void *value); |
134 | int (*map_update_elem)(struct bpf_offloaded_map *map, |
135 | void *key, void *value, u64 flags); |
136 | int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key); |
137 | }; |
138 | |
139 | struct bpf_offloaded_map { |
140 | struct bpf_map map; |
141 | struct net_device *netdev; |
142 | const struct bpf_map_dev_ops *dev_ops; |
143 | void *dev_priv; |
144 | struct list_head offloads; |
145 | }; |
146 | |
147 | static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map) |
148 | { |
149 | return container_of(map, struct bpf_offloaded_map, map); |
150 | } |
151 | |
152 | static inline bool bpf_map_offload_neutral(const struct bpf_map *map) |
153 | { |
154 | return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY; |
155 | } |
156 | |
157 | static inline bool bpf_map_support_seq_show(const struct bpf_map *map) |
158 | { |
159 | return map->btf && map->ops->map_seq_show_elem; |
160 | } |
161 | |
162 | int map_check_no_btf(const struct bpf_map *map, |
163 | const struct btf *btf, |
164 | const struct btf_type *key_type, |
165 | const struct btf_type *value_type); |
166 | |
167 | extern const struct bpf_map_ops bpf_map_offload_ops; |
168 | |
169 | /* function argument constraints */ |
170 | enum bpf_arg_type { |
171 | ARG_DONTCARE = 0, /* unused argument in helper function */ |
172 | |
173 | /* the following constraints used to prototype |
174 | * bpf_map_lookup/update/delete_elem() functions |
175 | */ |
176 | ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */ |
177 | ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */ |
178 | ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */ |
179 | ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */ |
180 | |
181 | /* the following constraints used to prototype bpf_memcmp() and other |
182 | * functions that access data on eBPF program stack |
183 | */ |
184 | ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */ |
185 | ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */ |
186 | ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized, |
187 | * helper function must fill all bytes or clear |
188 | * them in error case. |
189 | */ |
190 | |
191 | ARG_CONST_SIZE, /* number of bytes accessed from memory */ |
192 | ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */ |
193 | |
194 | ARG_PTR_TO_CTX, /* pointer to context */ |
195 | ARG_ANYTHING, /* any (initialized) argument is ok */ |
196 | ARG_PTR_TO_SOCKET, /* pointer to bpf_sock */ |
197 | ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */ |
198 | ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */ |
199 | }; |
200 | |
201 | /* type of values returned from helper functions */ |
202 | enum bpf_return_type { |
203 | RET_INTEGER, /* function returns integer */ |
204 | RET_VOID, /* function doesn't return anything */ |
205 | RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */ |
206 | RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */ |
207 | RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */ |
208 | RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */ |
209 | }; |
210 | |
211 | /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs |
212 | * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL |
213 | * instructions after verifying |
214 | */ |
215 | struct bpf_func_proto { |
216 | u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); |
217 | bool gpl_only; |
218 | bool pkt_access; |
219 | enum bpf_return_type ret_type; |
220 | enum bpf_arg_type arg1_type; |
221 | enum bpf_arg_type arg2_type; |
222 | enum bpf_arg_type arg3_type; |
223 | enum bpf_arg_type arg4_type; |
224 | enum bpf_arg_type arg5_type; |
225 | }; |
226 | |
227 | /* bpf_context is intentionally undefined structure. Pointer to bpf_context is |
228 | * the first argument to eBPF programs. |
229 | * For socket filters: 'struct bpf_context *' == 'struct sk_buff *' |
230 | */ |
231 | struct bpf_context; |
232 | |
233 | enum bpf_access_type { |
234 | BPF_READ = 1, |
235 | BPF_WRITE = 2 |
236 | }; |
237 | |
238 | /* types of values stored in eBPF registers */ |
239 | /* Pointer types represent: |
240 | * pointer |
241 | * pointer + imm |
242 | * pointer + (u16) var |
243 | * pointer + (u16) var + imm |
244 | * if (range > 0) then [ptr, ptr + range - off) is safe to access |
245 | * if (id > 0) means that some 'var' was added |
246 | * if (off > 0) means that 'imm' was added |
247 | */ |
248 | enum bpf_reg_type { |
249 | NOT_INIT = 0, /* nothing was written into register */ |
250 | SCALAR_VALUE, /* reg doesn't contain a valid pointer */ |
251 | PTR_TO_CTX, /* reg points to bpf_context */ |
252 | CONST_PTR_TO_MAP, /* reg points to struct bpf_map */ |
253 | PTR_TO_MAP_VALUE, /* reg points to map element value */ |
254 | PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */ |
255 | PTR_TO_STACK, /* reg == frame_pointer + offset */ |
256 | PTR_TO_PACKET_META, /* skb->data - meta_len */ |
257 | PTR_TO_PACKET, /* reg points to skb->data */ |
258 | PTR_TO_PACKET_END, /* skb->data + headlen */ |
259 | PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */ |
260 | PTR_TO_SOCKET, /* reg points to struct bpf_sock */ |
261 | PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */ |
262 | PTR_TO_SOCK_COMMON, /* reg points to sock_common */ |
263 | PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */ |
264 | PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */ |
265 | PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */ |
266 | }; |
267 | |
268 | /* The information passed from prog-specific *_is_valid_access |
269 | * back to the verifier. |
270 | */ |
271 | struct bpf_insn_access_aux { |
272 | enum bpf_reg_type reg_type; |
273 | int ctx_field_size; |
274 | }; |
275 | |
276 | static inline void |
277 | bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size) |
278 | { |
279 | aux->ctx_field_size = size; |
280 | } |
281 | |
282 | struct bpf_prog_ops { |
283 | int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, |
284 | union bpf_attr __user *uattr); |
285 | }; |
286 | |
287 | struct bpf_verifier_ops { |
288 | /* return eBPF function prototype for verification */ |
289 | const struct bpf_func_proto * |
290 | (*get_func_proto)(enum bpf_func_id func_id, |
291 | const struct bpf_prog *prog); |
292 | |
293 | /* return true if 'size' wide access at offset 'off' within bpf_context |
294 | * with 'type' (read or write) is allowed |
295 | */ |
296 | bool (*is_valid_access)(int off, int size, enum bpf_access_type type, |
297 | const struct bpf_prog *prog, |
298 | struct bpf_insn_access_aux *info); |
299 | int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, |
300 | const struct bpf_prog *prog); |
301 | int (*gen_ld_abs)(const struct bpf_insn *orig, |
302 | struct bpf_insn *insn_buf); |
303 | u32 (*convert_ctx_access)(enum bpf_access_type type, |
304 | const struct bpf_insn *src, |
305 | struct bpf_insn *dst, |
306 | struct bpf_prog *prog, u32 *target_size); |
307 | }; |
308 | |
309 | struct bpf_prog_offload_ops { |
310 | /* verifier basic callbacks */ |
311 | int (*insn_hook)(struct bpf_verifier_env *env, |
312 | int insn_idx, int prev_insn_idx); |
313 | int (*finalize)(struct bpf_verifier_env *env); |
314 | /* verifier optimization callbacks (called after .finalize) */ |
315 | int (*replace_insn)(struct bpf_verifier_env *env, u32 off, |
316 | struct bpf_insn *insn); |
317 | int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt); |
318 | /* program management callbacks */ |
319 | int (*prepare)(struct bpf_prog *prog); |
320 | int (*translate)(struct bpf_prog *prog); |
321 | void (*destroy)(struct bpf_prog *prog); |
322 | }; |
323 | |
324 | struct bpf_prog_offload { |
325 | struct bpf_prog *prog; |
326 | struct net_device *netdev; |
327 | struct bpf_offload_dev *offdev; |
328 | void *dev_priv; |
329 | struct list_head offloads; |
330 | bool dev_state; |
331 | bool opt_failed; |
332 | void *jited_image; |
333 | u32 jited_len; |
334 | }; |
335 | |
336 | enum bpf_cgroup_storage_type { |
337 | BPF_CGROUP_STORAGE_SHARED, |
338 | BPF_CGROUP_STORAGE_PERCPU, |
339 | __BPF_CGROUP_STORAGE_MAX |
340 | }; |
341 | |
342 | #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX |
343 | |
344 | struct bpf_prog_stats { |
345 | u64 cnt; |
346 | u64 nsecs; |
347 | struct u64_stats_sync syncp; |
348 | }; |
349 | |
350 | struct bpf_prog_aux { |
351 | atomic_t refcnt; |
352 | u32 used_map_cnt; |
353 | u32 max_ctx_offset; |
354 | u32 max_pkt_offset; |
355 | u32 stack_depth; |
356 | u32 id; |
357 | u32 func_cnt; /* used by non-func prog as the number of func progs */ |
358 | u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */ |
359 | bool offload_requested; |
360 | struct bpf_prog **func; |
361 | void *jit_data; /* JIT specific data. arch dependent */ |
362 | struct latch_tree_node ksym_tnode; |
363 | struct list_head ksym_lnode; |
364 | const struct bpf_prog_ops *ops; |
365 | struct bpf_map **used_maps; |
366 | struct bpf_prog *prog; |
367 | struct user_struct *user; |
368 | u64 load_time; /* ns since boottime */ |
369 | struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; |
370 | char name[BPF_OBJ_NAME_LEN]; |
371 | #ifdef CONFIG_SECURITY |
372 | void *security; |
373 | #endif |
374 | struct bpf_prog_offload *offload; |
375 | struct btf *btf; |
376 | struct bpf_func_info *func_info; |
377 | /* bpf_line_info loaded from userspace. linfo->insn_off |
378 | * has the xlated insn offset. |
379 | * Both the main and sub prog share the same linfo. |
380 | * The subprog can access its first linfo by |
381 | * using the linfo_idx. |
382 | */ |
383 | struct bpf_line_info *linfo; |
384 | /* jited_linfo is the jited addr of the linfo. It has a |
385 | * one to one mapping to linfo: |
386 | * jited_linfo[i] is the jited addr for the linfo[i]->insn_off. |
387 | * Both the main and sub prog share the same jited_linfo. |
388 | * The subprog can access its first jited_linfo by |
389 | * using the linfo_idx. |
390 | */ |
391 | void **jited_linfo; |
392 | u32 func_info_cnt; |
393 | u32 nr_linfo; |
394 | /* subprog can use linfo_idx to access its first linfo and |
395 | * jited_linfo. |
396 | * main prog always has linfo_idx == 0 |
397 | */ |
398 | u32 linfo_idx; |
399 | struct bpf_prog_stats __percpu *stats; |
400 | union { |
401 | struct work_struct work; |
402 | struct rcu_head rcu; |
403 | }; |
404 | }; |
405 | |
406 | struct bpf_array { |
407 | struct bpf_map map; |
408 | u32 elem_size; |
409 | u32 index_mask; |
410 | /* 'ownership' of prog_array is claimed by the first program that |
411 | * is going to use this map or by the first program which FD is stored |
412 | * in the map to make sure that all callers and callees have the same |
413 | * prog_type and JITed flag |
414 | */ |
415 | enum bpf_prog_type owner_prog_type; |
416 | bool owner_jited; |
417 | union { |
418 | char value[0] __aligned(8); |
419 | void *ptrs[0] __aligned(8); |
420 | void __percpu *pptrs[0] __aligned(8); |
421 | }; |
422 | }; |
423 | |
424 | #define MAX_TAIL_CALL_CNT 32 |
425 | |
426 | struct bpf_event_entry { |
427 | struct perf_event *event; |
428 | struct file *perf_file; |
429 | struct file *map_file; |
430 | struct rcu_head rcu; |
431 | }; |
432 | |
433 | bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp); |
434 | int bpf_prog_calc_tag(struct bpf_prog *fp); |
435 | |
436 | const struct bpf_func_proto *bpf_get_trace_printk_proto(void); |
437 | |
438 | typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, |
439 | unsigned long off, unsigned long len); |
440 | typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, |
441 | const struct bpf_insn *src, |
442 | struct bpf_insn *dst, |
443 | struct bpf_prog *prog, |
444 | u32 *target_size); |
445 | |
446 | u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, |
447 | void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy); |
448 | |
449 | int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, |
450 | union bpf_attr __user *uattr); |
451 | int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, |
452 | union bpf_attr __user *uattr); |
453 | int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, |
454 | const union bpf_attr *kattr, |
455 | union bpf_attr __user *uattr); |
456 | |
457 | /* an array of programs to be executed under rcu_lock. |
458 | * |
459 | * Typical usage: |
460 | * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN); |
461 | * |
462 | * the structure returned by bpf_prog_array_alloc() should be populated |
463 | * with program pointers and the last pointer must be NULL. |
464 | * The user has to keep refcnt on the program and make sure the program |
465 | * is removed from the array before bpf_prog_put(). |
466 | * The 'struct bpf_prog_array *' should only be replaced with xchg() |
467 | * since other cpus are walking the array of pointers in parallel. |
468 | */ |
469 | struct bpf_prog_array_item { |
470 | struct bpf_prog *prog; |
471 | struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]; |
472 | }; |
473 | |
474 | struct bpf_prog_array { |
475 | struct rcu_head rcu; |
476 | struct bpf_prog_array_item items[0]; |
477 | }; |
478 | |
479 | struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags); |
480 | void bpf_prog_array_free(struct bpf_prog_array __rcu *progs); |
481 | int bpf_prog_array_length(struct bpf_prog_array __rcu *progs); |
482 | int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs, |
483 | __u32 __user *prog_ids, u32 cnt); |
484 | |
485 | void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs, |
486 | struct bpf_prog *old_prog); |
487 | int bpf_prog_array_copy_info(struct bpf_prog_array __rcu *array, |
488 | u32 *prog_ids, u32 request_cnt, |
489 | u32 *prog_cnt); |
490 | int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array, |
491 | struct bpf_prog *exclude_prog, |
492 | struct bpf_prog *include_prog, |
493 | struct bpf_prog_array **new_array); |
494 | |
495 | #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \ |
496 | ({ \ |
497 | struct bpf_prog_array_item *_item; \ |
498 | struct bpf_prog *_prog; \ |
499 | struct bpf_prog_array *_array; \ |
500 | u32 _ret = 1; \ |
501 | preempt_disable(); \ |
502 | rcu_read_lock(); \ |
503 | _array = rcu_dereference(array); \ |
504 | if (unlikely(check_non_null && !_array))\ |
505 | goto _out; \ |
506 | _item = &_array->items[0]; \ |
507 | while ((_prog = READ_ONCE(_item->prog))) { \ |
508 | bpf_cgroup_storage_set(_item->cgroup_storage); \ |
509 | _ret &= func(_prog, ctx); \ |
510 | _item++; \ |
511 | } \ |
512 | _out: \ |
513 | rcu_read_unlock(); \ |
514 | preempt_enable_no_resched(); \ |
515 | _ret; \ |
516 | }) |
517 | |
518 | #define BPF_PROG_RUN_ARRAY(array, ctx, func) \ |
519 | __BPF_PROG_RUN_ARRAY(array, ctx, func, false) |
520 | |
521 | #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \ |
522 | __BPF_PROG_RUN_ARRAY(array, ctx, func, true) |
523 | |
524 | #ifdef CONFIG_BPF_SYSCALL |
525 | DECLARE_PER_CPU(int, bpf_prog_active); |
526 | |
527 | extern const struct file_operations bpf_map_fops; |
528 | extern const struct file_operations bpf_prog_fops; |
529 | |
530 | #define BPF_PROG_TYPE(_id, _name) \ |
531 | extern const struct bpf_prog_ops _name ## _prog_ops; \ |
532 | extern const struct bpf_verifier_ops _name ## _verifier_ops; |
533 | #define BPF_MAP_TYPE(_id, _ops) \ |
534 | extern const struct bpf_map_ops _ops; |
535 | #include <linux/bpf_types.h> |
536 | #undef BPF_PROG_TYPE |
537 | #undef BPF_MAP_TYPE |
538 | |
539 | extern const struct bpf_prog_ops bpf_offload_prog_ops; |
540 | extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops; |
541 | extern const struct bpf_verifier_ops xdp_analyzer_ops; |
542 | |
543 | struct bpf_prog *bpf_prog_get(u32 ufd); |
544 | struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, |
545 | bool attach_drv); |
546 | struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i); |
547 | void bpf_prog_sub(struct bpf_prog *prog, int i); |
548 | struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog); |
549 | struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog); |
550 | void bpf_prog_put(struct bpf_prog *prog); |
551 | int __bpf_prog_charge(struct user_struct *user, u32 pages); |
552 | void __bpf_prog_uncharge(struct user_struct *user, u32 pages); |
553 | |
554 | void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock); |
555 | void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock); |
556 | |
557 | struct bpf_map *bpf_map_get_with_uref(u32 ufd); |
558 | struct bpf_map *__bpf_map_get(struct fd f); |
559 | struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref); |
560 | void bpf_map_put_with_uref(struct bpf_map *map); |
561 | void bpf_map_put(struct bpf_map *map); |
562 | int bpf_map_precharge_memlock(u32 pages); |
563 | int bpf_map_charge_memlock(struct bpf_map *map, u32 pages); |
564 | void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages); |
565 | void *bpf_map_area_alloc(size_t size, int numa_node); |
566 | void bpf_map_area_free(void *base); |
567 | void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr); |
568 | |
569 | extern int sysctl_unprivileged_bpf_disabled; |
570 | extern int sysctl_bpf_stats_enabled; |
571 | |
572 | int bpf_map_new_fd(struct bpf_map *map, int flags); |
573 | int bpf_prog_new_fd(struct bpf_prog *prog); |
574 | |
575 | int bpf_obj_pin_user(u32 ufd, const char __user *pathname); |
576 | int bpf_obj_get_user(const char __user *pathname, int flags); |
577 | |
578 | int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value); |
579 | int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value); |
580 | int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, |
581 | u64 flags); |
582 | int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, |
583 | u64 flags); |
584 | |
585 | int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value); |
586 | |
587 | int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, |
588 | void *key, void *value, u64 map_flags); |
589 | int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); |
590 | int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, |
591 | void *key, void *value, u64 map_flags); |
592 | int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value); |
593 | |
594 | int bpf_get_file_flag(int flags); |
595 | int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size, |
596 | size_t actual_size); |
597 | |
598 | /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and |
599 | * forced to use 'long' read/writes to try to atomically copy long counters. |
600 | * Best-effort only. No barriers here, since it _will_ race with concurrent |
601 | * updates from BPF programs. Called from bpf syscall and mostly used with |
602 | * size 8 or 16 bytes, so ask compiler to inline it. |
603 | */ |
604 | static inline void bpf_long_memcpy(void *dst, const void *src, u32 size) |
605 | { |
606 | const long *lsrc = src; |
607 | long *ldst = dst; |
608 | |
609 | size /= sizeof(long); |
610 | while (size--) |
611 | *ldst++ = *lsrc++; |
612 | } |
613 | |
614 | /* verify correctness of eBPF program */ |
615 | int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, |
616 | union bpf_attr __user *uattr); |
617 | void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth); |
618 | |
619 | /* Map specifics */ |
620 | struct xdp_buff; |
621 | struct sk_buff; |
622 | |
623 | struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key); |
624 | void __dev_map_insert_ctx(struct bpf_map *map, u32 index); |
625 | void __dev_map_flush(struct bpf_map *map); |
626 | int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, |
627 | struct net_device *dev_rx); |
628 | int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, |
629 | struct bpf_prog *xdp_prog); |
630 | |
631 | struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key); |
632 | void __cpu_map_insert_ctx(struct bpf_map *map, u32 index); |
633 | void __cpu_map_flush(struct bpf_map *map); |
634 | int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, |
635 | struct net_device *dev_rx); |
636 | |
637 | /* Return map's numa specified by userspace */ |
638 | static inline int bpf_map_attr_numa_node(const union bpf_attr *attr) |
639 | { |
640 | return (attr->map_flags & BPF_F_NUMA_NODE) ? |
641 | attr->numa_node : NUMA_NO_NODE; |
642 | } |
643 | |
644 | struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type); |
645 | int array_map_alloc_check(union bpf_attr *attr); |
646 | |
647 | #else /* !CONFIG_BPF_SYSCALL */ |
648 | static inline struct bpf_prog *bpf_prog_get(u32 ufd) |
649 | { |
650 | return ERR_PTR(-EOPNOTSUPP); |
651 | } |
652 | |
653 | static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, |
654 | enum bpf_prog_type type, |
655 | bool attach_drv) |
656 | { |
657 | return ERR_PTR(-EOPNOTSUPP); |
658 | } |
659 | |
660 | static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, |
661 | int i) |
662 | { |
663 | return ERR_PTR(-EOPNOTSUPP); |
664 | } |
665 | |
666 | static inline void bpf_prog_sub(struct bpf_prog *prog, int i) |
667 | { |
668 | } |
669 | |
670 | static inline void bpf_prog_put(struct bpf_prog *prog) |
671 | { |
672 | } |
673 | |
674 | static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog) |
675 | { |
676 | return ERR_PTR(-EOPNOTSUPP); |
677 | } |
678 | |
679 | static inline struct bpf_prog *__must_check |
680 | bpf_prog_inc_not_zero(struct bpf_prog *prog) |
681 | { |
682 | return ERR_PTR(-EOPNOTSUPP); |
683 | } |
684 | |
685 | static inline int __bpf_prog_charge(struct user_struct *user, u32 pages) |
686 | { |
687 | return 0; |
688 | } |
689 | |
690 | static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages) |
691 | { |
692 | } |
693 | |
694 | static inline int bpf_obj_get_user(const char __user *pathname, int flags) |
695 | { |
696 | return -EOPNOTSUPP; |
697 | } |
698 | |
699 | static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map, |
700 | u32 key) |
701 | { |
702 | return NULL; |
703 | } |
704 | |
705 | static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index) |
706 | { |
707 | } |
708 | |
709 | static inline void __dev_map_flush(struct bpf_map *map) |
710 | { |
711 | } |
712 | |
713 | struct xdp_buff; |
714 | struct bpf_dtab_netdev; |
715 | |
716 | static inline |
717 | int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, |
718 | struct net_device *dev_rx) |
719 | { |
720 | return 0; |
721 | } |
722 | |
723 | struct sk_buff; |
724 | |
725 | static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, |
726 | struct sk_buff *skb, |
727 | struct bpf_prog *xdp_prog) |
728 | { |
729 | return 0; |
730 | } |
731 | |
732 | static inline |
733 | struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key) |
734 | { |
735 | return NULL; |
736 | } |
737 | |
738 | static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index) |
739 | { |
740 | } |
741 | |
742 | static inline void __cpu_map_flush(struct bpf_map *map) |
743 | { |
744 | } |
745 | |
746 | static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, |
747 | struct xdp_buff *xdp, |
748 | struct net_device *dev_rx) |
749 | { |
750 | return 0; |
751 | } |
752 | |
753 | static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, |
754 | enum bpf_prog_type type) |
755 | { |
756 | return ERR_PTR(-EOPNOTSUPP); |
757 | } |
758 | #endif /* CONFIG_BPF_SYSCALL */ |
759 | |
760 | static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, |
761 | enum bpf_prog_type type) |
762 | { |
763 | return bpf_prog_get_type_dev(ufd, type, false); |
764 | } |
765 | |
766 | bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool); |
767 | |
768 | int bpf_prog_offload_compile(struct bpf_prog *prog); |
769 | void bpf_prog_offload_destroy(struct bpf_prog *prog); |
770 | int bpf_prog_offload_info_fill(struct bpf_prog_info *info, |
771 | struct bpf_prog *prog); |
772 | |
773 | int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map); |
774 | |
775 | int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value); |
776 | int bpf_map_offload_update_elem(struct bpf_map *map, |
777 | void *key, void *value, u64 flags); |
778 | int bpf_map_offload_delete_elem(struct bpf_map *map, void *key); |
779 | int bpf_map_offload_get_next_key(struct bpf_map *map, |
780 | void *key, void *next_key); |
781 | |
782 | bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map); |
783 | |
784 | struct bpf_offload_dev * |
785 | bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv); |
786 | void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev); |
787 | void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev); |
788 | int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, |
789 | struct net_device *netdev); |
790 | void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, |
791 | struct net_device *netdev); |
792 | bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev); |
793 | |
794 | #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL) |
795 | int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr); |
796 | |
797 | static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux) |
798 | { |
799 | return aux->offload_requested; |
800 | } |
801 | |
802 | static inline bool bpf_map_is_dev_bound(struct bpf_map *map) |
803 | { |
804 | return unlikely(map->ops == &bpf_map_offload_ops); |
805 | } |
806 | |
807 | struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr); |
808 | void bpf_map_offload_map_free(struct bpf_map *map); |
809 | #else |
810 | static inline int bpf_prog_offload_init(struct bpf_prog *prog, |
811 | union bpf_attr *attr) |
812 | { |
813 | return -EOPNOTSUPP; |
814 | } |
815 | |
816 | static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux) |
817 | { |
818 | return false; |
819 | } |
820 | |
821 | static inline bool bpf_map_is_dev_bound(struct bpf_map *map) |
822 | { |
823 | return false; |
824 | } |
825 | |
826 | static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr) |
827 | { |
828 | return ERR_PTR(-EOPNOTSUPP); |
829 | } |
830 | |
831 | static inline void bpf_map_offload_map_free(struct bpf_map *map) |
832 | { |
833 | } |
834 | #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */ |
835 | |
836 | #if defined(CONFIG_BPF_STREAM_PARSER) |
837 | int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, u32 which); |
838 | int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog); |
839 | #else |
840 | static inline int sock_map_prog_update(struct bpf_map *map, |
841 | struct bpf_prog *prog, u32 which) |
842 | { |
843 | return -EOPNOTSUPP; |
844 | } |
845 | |
846 | static inline int sock_map_get_from_fd(const union bpf_attr *attr, |
847 | struct bpf_prog *prog) |
848 | { |
849 | return -EINVAL; |
850 | } |
851 | #endif |
852 | |
853 | #if defined(CONFIG_XDP_SOCKETS) |
854 | struct xdp_sock; |
855 | struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map, u32 key); |
856 | int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp, |
857 | struct xdp_sock *xs); |
858 | void __xsk_map_flush(struct bpf_map *map); |
859 | #else |
860 | struct xdp_sock; |
861 | static inline struct xdp_sock *__xsk_map_lookup_elem(struct bpf_map *map, |
862 | u32 key) |
863 | { |
864 | return NULL; |
865 | } |
866 | |
867 | static inline int __xsk_map_redirect(struct bpf_map *map, struct xdp_buff *xdp, |
868 | struct xdp_sock *xs) |
869 | { |
870 | return -EOPNOTSUPP; |
871 | } |
872 | |
873 | static inline void __xsk_map_flush(struct bpf_map *map) |
874 | { |
875 | } |
876 | #endif |
877 | |
878 | #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) |
879 | void bpf_sk_reuseport_detach(struct sock *sk); |
880 | int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, |
881 | void *value); |
882 | int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, |
883 | void *value, u64 map_flags); |
884 | #else |
885 | static inline void bpf_sk_reuseport_detach(struct sock *sk) |
886 | { |
887 | } |
888 | |
889 | #ifdef CONFIG_BPF_SYSCALL |
890 | static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, |
891 | void *key, void *value) |
892 | { |
893 | return -EOPNOTSUPP; |
894 | } |
895 | |
896 | static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, |
897 | void *key, void *value, |
898 | u64 map_flags) |
899 | { |
900 | return -EOPNOTSUPP; |
901 | } |
902 | #endif /* CONFIG_BPF_SYSCALL */ |
903 | #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */ |
904 | |
905 | /* verifier prototypes for helper functions called from eBPF programs */ |
906 | extern const struct bpf_func_proto bpf_map_lookup_elem_proto; |
907 | extern const struct bpf_func_proto bpf_map_update_elem_proto; |
908 | extern const struct bpf_func_proto bpf_map_delete_elem_proto; |
909 | extern const struct bpf_func_proto bpf_map_push_elem_proto; |
910 | extern const struct bpf_func_proto bpf_map_pop_elem_proto; |
911 | extern const struct bpf_func_proto bpf_map_peek_elem_proto; |
912 | |
913 | extern const struct bpf_func_proto bpf_get_prandom_u32_proto; |
914 | extern const struct bpf_func_proto bpf_get_smp_processor_id_proto; |
915 | extern const struct bpf_func_proto bpf_get_numa_node_id_proto; |
916 | extern const struct bpf_func_proto bpf_tail_call_proto; |
917 | extern const struct bpf_func_proto bpf_ktime_get_ns_proto; |
918 | extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto; |
919 | extern const struct bpf_func_proto bpf_get_current_uid_gid_proto; |
920 | extern const struct bpf_func_proto bpf_get_current_comm_proto; |
921 | extern const struct bpf_func_proto bpf_get_stackid_proto; |
922 | extern const struct bpf_func_proto bpf_get_stack_proto; |
923 | extern const struct bpf_func_proto bpf_sock_map_update_proto; |
924 | extern const struct bpf_func_proto bpf_sock_hash_update_proto; |
925 | extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto; |
926 | extern const struct bpf_func_proto bpf_msg_redirect_hash_proto; |
927 | extern const struct bpf_func_proto bpf_msg_redirect_map_proto; |
928 | extern const struct bpf_func_proto bpf_sk_redirect_hash_proto; |
929 | extern const struct bpf_func_proto bpf_sk_redirect_map_proto; |
930 | extern const struct bpf_func_proto bpf_spin_lock_proto; |
931 | extern const struct bpf_func_proto bpf_spin_unlock_proto; |
932 | extern const struct bpf_func_proto bpf_get_local_storage_proto; |
933 | |
934 | /* Shared helpers among cBPF and eBPF. */ |
935 | void bpf_user_rnd_init_once(void); |
936 | u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5); |
937 | |
938 | #if defined(CONFIG_NET) |
939 | bool bpf_sock_common_is_valid_access(int off, int size, |
940 | enum bpf_access_type type, |
941 | struct bpf_insn_access_aux *info); |
942 | bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, |
943 | struct bpf_insn_access_aux *info); |
944 | u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, |
945 | const struct bpf_insn *si, |
946 | struct bpf_insn *insn_buf, |
947 | struct bpf_prog *prog, |
948 | u32 *target_size); |
949 | #else |
950 | static inline bool bpf_sock_common_is_valid_access(int off, int size, |
951 | enum bpf_access_type type, |
952 | struct bpf_insn_access_aux *info) |
953 | { |
954 | return false; |
955 | } |
956 | static inline bool bpf_sock_is_valid_access(int off, int size, |
957 | enum bpf_access_type type, |
958 | struct bpf_insn_access_aux *info) |
959 | { |
960 | return false; |
961 | } |
962 | static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, |
963 | const struct bpf_insn *si, |
964 | struct bpf_insn *insn_buf, |
965 | struct bpf_prog *prog, |
966 | u32 *target_size) |
967 | { |
968 | return 0; |
969 | } |
970 | #endif |
971 | |
972 | #ifdef CONFIG_INET |
973 | bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, |
974 | struct bpf_insn_access_aux *info); |
975 | |
976 | u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, |
977 | const struct bpf_insn *si, |
978 | struct bpf_insn *insn_buf, |
979 | struct bpf_prog *prog, |
980 | u32 *target_size); |
981 | #else |
982 | static inline bool bpf_tcp_sock_is_valid_access(int off, int size, |
983 | enum bpf_access_type type, |
984 | struct bpf_insn_access_aux *info) |
985 | { |
986 | return false; |
987 | } |
988 | |
989 | static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, |
990 | const struct bpf_insn *si, |
991 | struct bpf_insn *insn_buf, |
992 | struct bpf_prog *prog, |
993 | u32 *target_size) |
994 | { |
995 | return 0; |
996 | } |
997 | #endif /* CONFIG_INET */ |
998 | |
999 | #endif /* _LINUX_BPF_H */ |
1000 | |