]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/bpf.h
xdp: allow attaching programs loaded for specific device
[mirror_ubuntu-bionic-kernel.git] / include / linux / bpf.h
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
20 struct perf_event;
21 struct bpf_prog;
22 struct bpf_map;
23
24 /* map is generic key/value storage optionally accesible by eBPF programs */
25 struct bpf_map_ops {
26 /* funcs callable from userspace (via syscall) */
27 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
28 void (*map_release)(struct bpf_map *map, struct file *map_file);
29 void (*map_free)(struct bpf_map *map);
30 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
31
32 /* funcs callable from userspace and from eBPF programs */
33 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
34 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
35 int (*map_delete_elem)(struct bpf_map *map, void *key);
36
37 /* funcs called by prog_array and perf_event_array map */
38 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
39 int fd);
40 void (*map_fd_put_ptr)(void *ptr);
41 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
42 u32 (*map_fd_sys_lookup_elem)(void *ptr);
43 };
44
45 struct bpf_map {
46 atomic_t refcnt;
47 enum bpf_map_type map_type;
48 u32 key_size;
49 u32 value_size;
50 u32 max_entries;
51 u32 map_flags;
52 u32 pages;
53 u32 id;
54 int numa_node;
55 struct user_struct *user;
56 const struct bpf_map_ops *ops;
57 struct work_struct work;
58 atomic_t usercnt;
59 struct bpf_map *inner_map_meta;
60 char name[BPF_OBJ_NAME_LEN];
61 #ifdef CONFIG_SECURITY
62 void *security;
63 #endif
64 };
65
66 /* function argument constraints */
67 enum bpf_arg_type {
68 ARG_DONTCARE = 0, /* unused argument in helper function */
69
70 /* the following constraints used to prototype
71 * bpf_map_lookup/update/delete_elem() functions
72 */
73 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
74 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
75 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
76
77 /* the following constraints used to prototype bpf_memcmp() and other
78 * functions that access data on eBPF program stack
79 */
80 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
81 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
82 * helper function must fill all bytes or clear
83 * them in error case.
84 */
85
86 ARG_CONST_SIZE, /* number of bytes accessed from memory */
87 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
88
89 ARG_PTR_TO_CTX, /* pointer to context */
90 ARG_ANYTHING, /* any (initialized) argument is ok */
91 };
92
93 /* type of values returned from helper functions */
94 enum bpf_return_type {
95 RET_INTEGER, /* function returns integer */
96 RET_VOID, /* function doesn't return anything */
97 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
98 };
99
100 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
101 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
102 * instructions after verifying
103 */
104 struct bpf_func_proto {
105 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
106 bool gpl_only;
107 bool pkt_access;
108 enum bpf_return_type ret_type;
109 enum bpf_arg_type arg1_type;
110 enum bpf_arg_type arg2_type;
111 enum bpf_arg_type arg3_type;
112 enum bpf_arg_type arg4_type;
113 enum bpf_arg_type arg5_type;
114 };
115
116 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
117 * the first argument to eBPF programs.
118 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
119 */
120 struct bpf_context;
121
122 enum bpf_access_type {
123 BPF_READ = 1,
124 BPF_WRITE = 2
125 };
126
127 /* types of values stored in eBPF registers */
128 /* Pointer types represent:
129 * pointer
130 * pointer + imm
131 * pointer + (u16) var
132 * pointer + (u16) var + imm
133 * if (range > 0) then [ptr, ptr + range - off) is safe to access
134 * if (id > 0) means that some 'var' was added
135 * if (off > 0) means that 'imm' was added
136 */
137 enum bpf_reg_type {
138 NOT_INIT = 0, /* nothing was written into register */
139 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
140 PTR_TO_CTX, /* reg points to bpf_context */
141 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
142 PTR_TO_MAP_VALUE, /* reg points to map element value */
143 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
144 PTR_TO_STACK, /* reg == frame_pointer + offset */
145 PTR_TO_PACKET_META, /* skb->data - meta_len */
146 PTR_TO_PACKET, /* reg points to skb->data */
147 PTR_TO_PACKET_END, /* skb->data + headlen */
148 };
149
150 /* The information passed from prog-specific *_is_valid_access
151 * back to the verifier.
152 */
153 struct bpf_insn_access_aux {
154 enum bpf_reg_type reg_type;
155 int ctx_field_size;
156 };
157
158 static inline void
159 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
160 {
161 aux->ctx_field_size = size;
162 }
163
164 struct bpf_prog_ops {
165 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
166 union bpf_attr __user *uattr);
167 };
168
169 struct bpf_verifier_ops {
170 /* return eBPF function prototype for verification */
171 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
172
173 /* return true if 'size' wide access at offset 'off' within bpf_context
174 * with 'type' (read or write) is allowed
175 */
176 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
177 struct bpf_insn_access_aux *info);
178 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
179 const struct bpf_prog *prog);
180 u32 (*convert_ctx_access)(enum bpf_access_type type,
181 const struct bpf_insn *src,
182 struct bpf_insn *dst,
183 struct bpf_prog *prog, u32 *target_size);
184 };
185
186 struct bpf_dev_offload {
187 struct bpf_prog *prog;
188 struct net_device *netdev;
189 void *dev_priv;
190 struct list_head offloads;
191 bool dev_state;
192 bool verifier_running;
193 wait_queue_head_t verifier_done;
194 };
195
196 struct bpf_prog_aux {
197 atomic_t refcnt;
198 u32 used_map_cnt;
199 u32 max_ctx_offset;
200 u32 stack_depth;
201 u32 id;
202 struct latch_tree_node ksym_tnode;
203 struct list_head ksym_lnode;
204 const struct bpf_prog_ops *ops;
205 struct bpf_map **used_maps;
206 struct bpf_prog *prog;
207 struct user_struct *user;
208 u64 load_time; /* ns since boottime */
209 char name[BPF_OBJ_NAME_LEN];
210 #ifdef CONFIG_SECURITY
211 void *security;
212 #endif
213 struct bpf_dev_offload *offload;
214 union {
215 struct work_struct work;
216 struct rcu_head rcu;
217 };
218 };
219
220 struct bpf_array {
221 struct bpf_map map;
222 u32 elem_size;
223 /* 'ownership' of prog_array is claimed by the first program that
224 * is going to use this map or by the first program which FD is stored
225 * in the map to make sure that all callers and callees have the same
226 * prog_type and JITed flag
227 */
228 enum bpf_prog_type owner_prog_type;
229 bool owner_jited;
230 union {
231 char value[0] __aligned(8);
232 void *ptrs[0] __aligned(8);
233 void __percpu *pptrs[0] __aligned(8);
234 };
235 };
236
237 #define MAX_TAIL_CALL_CNT 32
238
239 struct bpf_event_entry {
240 struct perf_event *event;
241 struct file *perf_file;
242 struct file *map_file;
243 struct rcu_head rcu;
244 };
245
246 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
247 int bpf_prog_calc_tag(struct bpf_prog *fp);
248
249 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
250
251 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
252 unsigned long off, unsigned long len);
253
254 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
255 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
256
257 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
258 union bpf_attr __user *uattr);
259 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
260 union bpf_attr __user *uattr);
261
262 /* an array of programs to be executed under rcu_lock.
263 *
264 * Typical usage:
265 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
266 *
267 * the structure returned by bpf_prog_array_alloc() should be populated
268 * with program pointers and the last pointer must be NULL.
269 * The user has to keep refcnt on the program and make sure the program
270 * is removed from the array before bpf_prog_put().
271 * The 'struct bpf_prog_array *' should only be replaced with xchg()
272 * since other cpus are walking the array of pointers in parallel.
273 */
274 struct bpf_prog_array {
275 struct rcu_head rcu;
276 struct bpf_prog *progs[0];
277 };
278
279 struct bpf_prog_array __rcu *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
280 void bpf_prog_array_free(struct bpf_prog_array __rcu *progs);
281 int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
282 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
283 __u32 __user *prog_ids, u32 cnt);
284
285 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
286 struct bpf_prog *old_prog);
287 int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
288 struct bpf_prog *exclude_prog,
289 struct bpf_prog *include_prog,
290 struct bpf_prog_array **new_array);
291
292 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
293 ({ \
294 struct bpf_prog **_prog, *__prog; \
295 struct bpf_prog_array *_array; \
296 u32 _ret = 1; \
297 rcu_read_lock(); \
298 _array = rcu_dereference(array); \
299 if (unlikely(check_non_null && !_array))\
300 goto _out; \
301 _prog = _array->progs; \
302 while ((__prog = READ_ONCE(*_prog))) { \
303 _ret &= func(__prog, ctx); \
304 _prog++; \
305 } \
306 _out: \
307 rcu_read_unlock(); \
308 _ret; \
309 })
310
311 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
312 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
313
314 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
315 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
316
317 #ifdef CONFIG_BPF_SYSCALL
318 DECLARE_PER_CPU(int, bpf_prog_active);
319
320 extern const struct file_operations bpf_map_fops;
321 extern const struct file_operations bpf_prog_fops;
322
323 #define BPF_PROG_TYPE(_id, _name) \
324 extern const struct bpf_prog_ops _name ## _prog_ops; \
325 extern const struct bpf_verifier_ops _name ## _verifier_ops;
326 #define BPF_MAP_TYPE(_id, _ops) \
327 extern const struct bpf_map_ops _ops;
328 #include <linux/bpf_types.h>
329 #undef BPF_PROG_TYPE
330 #undef BPF_MAP_TYPE
331
332 extern const struct bpf_prog_ops bpf_offload_prog_ops;
333 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
334 extern const struct bpf_verifier_ops xdp_analyzer_ops;
335
336 struct bpf_prog *bpf_prog_get(u32 ufd);
337 struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type);
338 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
339 struct net_device *netdev);
340 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
341 void bpf_prog_sub(struct bpf_prog *prog, int i);
342 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
343 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
344 void bpf_prog_put(struct bpf_prog *prog);
345 int __bpf_prog_charge(struct user_struct *user, u32 pages);
346 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
347
348 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
349 struct bpf_map *__bpf_map_get(struct fd f);
350 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
351 void bpf_map_put_with_uref(struct bpf_map *map);
352 void bpf_map_put(struct bpf_map *map);
353 int bpf_map_precharge_memlock(u32 pages);
354 void *bpf_map_area_alloc(size_t size, int numa_node);
355 void bpf_map_area_free(void *base);
356
357 extern int sysctl_unprivileged_bpf_disabled;
358
359 int bpf_map_new_fd(struct bpf_map *map, int flags);
360 int bpf_prog_new_fd(struct bpf_prog *prog);
361
362 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
363 int bpf_obj_get_user(const char __user *pathname, int flags);
364
365 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
366 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
367 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
368 u64 flags);
369 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
370 u64 flags);
371
372 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
373
374 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
375 void *key, void *value, u64 map_flags);
376 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
377 void bpf_fd_array_map_clear(struct bpf_map *map);
378 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
379 void *key, void *value, u64 map_flags);
380 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
381
382 int bpf_get_file_flag(int flags);
383
384 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
385 * forced to use 'long' read/writes to try to atomically copy long counters.
386 * Best-effort only. No barriers here, since it _will_ race with concurrent
387 * updates from BPF programs. Called from bpf syscall and mostly used with
388 * size 8 or 16 bytes, so ask compiler to inline it.
389 */
390 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
391 {
392 const long *lsrc = src;
393 long *ldst = dst;
394
395 size /= sizeof(long);
396 while (size--)
397 *ldst++ = *lsrc++;
398 }
399
400 /* verify correctness of eBPF program */
401 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
402
403 /* Map specifics */
404 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
405 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
406 void __dev_map_flush(struct bpf_map *map);
407
408 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
409 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index);
410 void __cpu_map_flush(struct bpf_map *map);
411 struct xdp_buff;
412 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
413 struct net_device *dev_rx);
414
415 /* Return map's numa specified by userspace */
416 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
417 {
418 return (attr->map_flags & BPF_F_NUMA_NODE) ?
419 attr->numa_node : NUMA_NO_NODE;
420 }
421
422 #else /* !CONFIG_BPF_SYSCALL */
423 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
424 {
425 return ERR_PTR(-EOPNOTSUPP);
426 }
427
428 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
429 enum bpf_prog_type type)
430 {
431 return ERR_PTR(-EOPNOTSUPP);
432 }
433
434 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
435 enum bpf_prog_type type,
436 struct net_device *netdev)
437 {
438 return ERR_PTR(-EOPNOTSUPP);
439 }
440
441 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
442 int i)
443 {
444 return ERR_PTR(-EOPNOTSUPP);
445 }
446
447 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
448 {
449 }
450
451 static inline void bpf_prog_put(struct bpf_prog *prog)
452 {
453 }
454
455 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
456 {
457 return ERR_PTR(-EOPNOTSUPP);
458 }
459
460 static inline struct bpf_prog *__must_check
461 bpf_prog_inc_not_zero(struct bpf_prog *prog)
462 {
463 return ERR_PTR(-EOPNOTSUPP);
464 }
465
466 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
467 {
468 return 0;
469 }
470
471 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
472 {
473 }
474
475 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
476 {
477 return -EOPNOTSUPP;
478 }
479
480 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
481 u32 key)
482 {
483 return NULL;
484 }
485
486 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
487 {
488 }
489
490 static inline void __dev_map_flush(struct bpf_map *map)
491 {
492 }
493
494 static inline
495 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
496 {
497 return NULL;
498 }
499
500 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index)
501 {
502 }
503
504 static inline void __cpu_map_flush(struct bpf_map *map)
505 {
506 }
507
508 struct xdp_buff;
509 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
510 struct xdp_buff *xdp,
511 struct net_device *dev_rx)
512 {
513 return 0;
514 }
515 #endif /* CONFIG_BPF_SYSCALL */
516
517 int bpf_prog_offload_compile(struct bpf_prog *prog);
518 void bpf_prog_offload_destroy(struct bpf_prog *prog);
519 u32 bpf_prog_offload_ifindex(struct bpf_prog *prog);
520
521 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
522 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
523
524 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
525 {
526 return aux->offload;
527 }
528 #else
529 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
530 union bpf_attr *attr)
531 {
532 return -EOPNOTSUPP;
533 }
534
535 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
536 {
537 return false;
538 }
539 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
540
541 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
542 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
543 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
544 #else
545 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
546 {
547 return NULL;
548 }
549
550 static inline int sock_map_prog(struct bpf_map *map,
551 struct bpf_prog *prog,
552 u32 type)
553 {
554 return -EOPNOTSUPP;
555 }
556 #endif
557
558 /* verifier prototypes for helper functions called from eBPF programs */
559 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
560 extern const struct bpf_func_proto bpf_map_update_elem_proto;
561 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
562
563 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
564 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
565 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
566 extern const struct bpf_func_proto bpf_tail_call_proto;
567 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
568 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
569 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
570 extern const struct bpf_func_proto bpf_get_current_comm_proto;
571 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
572 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
573 extern const struct bpf_func_proto bpf_get_stackid_proto;
574 extern const struct bpf_func_proto bpf_sock_map_update_proto;
575
576 /* Shared helpers among cBPF and eBPF. */
577 void bpf_user_rnd_init_once(void);
578 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
579
580 #endif /* _LINUX_BPF_H */