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>
17 #include <linux/numa.h>
18 #include <linux/wait.h>
20 struct bpf_verifier_env;
25 /* map is generic key/value storage optionally accesible by eBPF programs */
27 /* funcs callable from userspace (via syscall) */
28 int (*map_alloc_check)(union bpf_attr *attr);
29 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
30 void (*map_release)(struct bpf_map *map, struct file *map_file);
31 void (*map_free)(struct bpf_map *map);
32 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
34 /* funcs callable from userspace and from eBPF programs */
35 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
36 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
37 int (*map_delete_elem)(struct bpf_map *map, void *key);
39 /* funcs called by prog_array and perf_event_array map */
40 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
42 void (*map_fd_put_ptr)(void *ptr);
43 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
44 u32 (*map_fd_sys_lookup_elem)(void *ptr);
48 /* 1st cacheline with read-mostly members of which some
49 * are also accessed in fast-path (e.g. ops, max_entries).
51 const struct bpf_map_ops *ops ____cacheline_aligned;
52 struct bpf_map *inner_map_meta;
53 #ifdef CONFIG_SECURITY
56 enum bpf_map_type map_type;
67 /* 2nd cacheline with misc members to avoid false sharing
68 * particularly with refcounting.
70 struct user_struct *user ____cacheline_aligned;
73 struct work_struct work;
74 char name[BPF_OBJ_NAME_LEN];
77 struct bpf_offloaded_map;
79 struct bpf_map_dev_ops {
80 int (*map_get_next_key)(struct bpf_offloaded_map *map,
81 void *key, void *next_key);
82 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
83 void *key, void *value);
84 int (*map_update_elem)(struct bpf_offloaded_map *map,
85 void *key, void *value, u64 flags);
86 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
89 struct bpf_offloaded_map {
91 struct net_device *netdev;
92 const struct bpf_map_dev_ops *dev_ops;
94 struct list_head offloads;
97 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
99 return container_of(map, struct bpf_offloaded_map, map);
102 extern const struct bpf_map_ops bpf_map_offload_ops;
104 /* function argument constraints */
106 ARG_DONTCARE = 0, /* unused argument in helper function */
108 /* the following constraints used to prototype
109 * bpf_map_lookup/update/delete_elem() functions
111 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
112 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
113 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
115 /* the following constraints used to prototype bpf_memcmp() and other
116 * functions that access data on eBPF program stack
118 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
119 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
120 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
121 * helper function must fill all bytes or clear
122 * them in error case.
125 ARG_CONST_SIZE, /* number of bytes accessed from memory */
126 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
128 ARG_PTR_TO_CTX, /* pointer to context */
129 ARG_ANYTHING, /* any (initialized) argument is ok */
132 /* type of values returned from helper functions */
133 enum bpf_return_type {
134 RET_INTEGER, /* function returns integer */
135 RET_VOID, /* function doesn't return anything */
136 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
139 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
140 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
141 * instructions after verifying
143 struct bpf_func_proto {
144 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
147 enum bpf_return_type ret_type;
148 enum bpf_arg_type arg1_type;
149 enum bpf_arg_type arg2_type;
150 enum bpf_arg_type arg3_type;
151 enum bpf_arg_type arg4_type;
152 enum bpf_arg_type arg5_type;
155 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
156 * the first argument to eBPF programs.
157 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
161 enum bpf_access_type {
166 /* types of values stored in eBPF registers */
167 /* Pointer types represent:
170 * pointer + (u16) var
171 * pointer + (u16) var + imm
172 * if (range > 0) then [ptr, ptr + range - off) is safe to access
173 * if (id > 0) means that some 'var' was added
174 * if (off > 0) means that 'imm' was added
177 NOT_INIT = 0, /* nothing was written into register */
178 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
179 PTR_TO_CTX, /* reg points to bpf_context */
180 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
181 PTR_TO_MAP_VALUE, /* reg points to map element value */
182 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
183 PTR_TO_STACK, /* reg == frame_pointer + offset */
184 PTR_TO_PACKET_META, /* skb->data - meta_len */
185 PTR_TO_PACKET, /* reg points to skb->data */
186 PTR_TO_PACKET_END, /* skb->data + headlen */
189 /* The information passed from prog-specific *_is_valid_access
190 * back to the verifier.
192 struct bpf_insn_access_aux {
193 enum bpf_reg_type reg_type;
198 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
200 aux->ctx_field_size = size;
203 struct bpf_prog_ops {
204 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
205 union bpf_attr __user *uattr);
208 struct bpf_verifier_ops {
209 /* return eBPF function prototype for verification */
210 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
212 /* return true if 'size' wide access at offset 'off' within bpf_context
213 * with 'type' (read or write) is allowed
215 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
216 struct bpf_insn_access_aux *info);
217 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
218 const struct bpf_prog *prog);
219 u32 (*convert_ctx_access)(enum bpf_access_type type,
220 const struct bpf_insn *src,
221 struct bpf_insn *dst,
222 struct bpf_prog *prog, u32 *target_size);
225 struct bpf_prog_offload_ops {
226 int (*insn_hook)(struct bpf_verifier_env *env,
227 int insn_idx, int prev_insn_idx);
230 struct bpf_prog_offload {
231 struct bpf_prog *prog;
232 struct net_device *netdev;
234 struct list_head offloads;
236 const struct bpf_prog_offload_ops *dev_ops;
239 struct bpf_prog_aux {
246 bool offload_requested;
247 struct bpf_prog **func;
248 void *jit_data; /* JIT specific data. arch dependent */
249 struct latch_tree_node ksym_tnode;
250 struct list_head ksym_lnode;
251 const struct bpf_prog_ops *ops;
252 struct bpf_map **used_maps;
253 struct bpf_prog *prog;
254 struct user_struct *user;
255 u64 load_time; /* ns since boottime */
256 char name[BPF_OBJ_NAME_LEN];
257 #ifdef CONFIG_SECURITY
260 struct bpf_prog_offload *offload;
262 struct work_struct work;
271 /* 'ownership' of prog_array is claimed by the first program that
272 * is going to use this map or by the first program which FD is stored
273 * in the map to make sure that all callers and callees have the same
274 * prog_type and JITed flag
276 enum bpf_prog_type owner_prog_type;
279 char value[0] __aligned(8);
280 void *ptrs[0] __aligned(8);
281 void __percpu *pptrs[0] __aligned(8);
285 #define MAX_TAIL_CALL_CNT 32
287 struct bpf_event_entry {
288 struct perf_event *event;
289 struct file *perf_file;
290 struct file *map_file;
294 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
295 int bpf_prog_calc_tag(struct bpf_prog *fp);
297 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
299 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
300 unsigned long off, unsigned long len);
302 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
303 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
305 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
306 union bpf_attr __user *uattr);
307 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
308 union bpf_attr __user *uattr);
310 /* an array of programs to be executed under rcu_lock.
313 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
315 * the structure returned by bpf_prog_array_alloc() should be populated
316 * with program pointers and the last pointer must be NULL.
317 * The user has to keep refcnt on the program and make sure the program
318 * is removed from the array before bpf_prog_put().
319 * The 'struct bpf_prog_array *' should only be replaced with xchg()
320 * since other cpus are walking the array of pointers in parallel.
322 struct bpf_prog_array {
324 struct bpf_prog *progs[0];
327 struct bpf_prog_array __rcu *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
328 void bpf_prog_array_free(struct bpf_prog_array __rcu *progs);
329 int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
330 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
331 __u32 __user *prog_ids, u32 cnt);
333 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
334 struct bpf_prog *old_prog);
335 int bpf_prog_array_copy_info(struct bpf_prog_array __rcu *array,
336 __u32 __user *prog_ids, u32 request_cnt,
337 __u32 __user *prog_cnt);
338 int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
339 struct bpf_prog *exclude_prog,
340 struct bpf_prog *include_prog,
341 struct bpf_prog_array **new_array);
343 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
345 struct bpf_prog **_prog, *__prog; \
346 struct bpf_prog_array *_array; \
349 _array = rcu_dereference(array); \
350 if (unlikely(check_non_null && !_array))\
352 _prog = _array->progs; \
353 while ((__prog = READ_ONCE(*_prog))) { \
354 _ret &= func(__prog, ctx); \
362 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
363 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
365 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
366 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
368 #ifdef CONFIG_BPF_SYSCALL
369 DECLARE_PER_CPU(int, bpf_prog_active);
371 extern const struct file_operations bpf_map_fops;
372 extern const struct file_operations bpf_prog_fops;
374 #define BPF_PROG_TYPE(_id, _name) \
375 extern const struct bpf_prog_ops _name ## _prog_ops; \
376 extern const struct bpf_verifier_ops _name ## _verifier_ops;
377 #define BPF_MAP_TYPE(_id, _ops) \
378 extern const struct bpf_map_ops _ops;
379 #include <linux/bpf_types.h>
383 extern const struct bpf_prog_ops bpf_offload_prog_ops;
384 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
385 extern const struct bpf_verifier_ops xdp_analyzer_ops;
387 struct bpf_prog *bpf_prog_get(u32 ufd);
388 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
390 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
391 void bpf_prog_sub(struct bpf_prog *prog, int i);
392 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
393 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
394 void bpf_prog_put(struct bpf_prog *prog);
395 int __bpf_prog_charge(struct user_struct *user, u32 pages);
396 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
398 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
399 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
401 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
402 struct bpf_map *__bpf_map_get(struct fd f);
403 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
404 void bpf_map_put_with_uref(struct bpf_map *map);
405 void bpf_map_put(struct bpf_map *map);
406 int bpf_map_precharge_memlock(u32 pages);
407 void *bpf_map_area_alloc(size_t size, int numa_node);
408 void bpf_map_area_free(void *base);
409 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
411 extern int sysctl_unprivileged_bpf_disabled;
413 int bpf_map_new_fd(struct bpf_map *map, int flags);
414 int bpf_prog_new_fd(struct bpf_prog *prog);
416 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
417 int bpf_obj_get_user(const char __user *pathname, int flags);
419 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
420 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
421 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
423 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
426 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
428 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
429 void *key, void *value, u64 map_flags);
430 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
431 void bpf_fd_array_map_clear(struct bpf_map *map);
432 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
433 void *key, void *value, u64 map_flags);
434 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
436 int bpf_get_file_flag(int flags);
438 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
439 * forced to use 'long' read/writes to try to atomically copy long counters.
440 * Best-effort only. No barriers here, since it _will_ race with concurrent
441 * updates from BPF programs. Called from bpf syscall and mostly used with
442 * size 8 or 16 bytes, so ask compiler to inline it.
444 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
446 const long *lsrc = src;
449 size /= sizeof(long);
454 /* verify correctness of eBPF program */
455 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
456 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
459 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
460 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
461 void __dev_map_flush(struct bpf_map *map);
463 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
464 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index);
465 void __cpu_map_flush(struct bpf_map *map);
467 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
468 struct net_device *dev_rx);
470 /* Return map's numa specified by userspace */
471 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
473 return (attr->map_flags & BPF_F_NUMA_NODE) ?
474 attr->numa_node : NUMA_NO_NODE;
477 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
479 #else /* !CONFIG_BPF_SYSCALL */
480 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
482 return ERR_PTR(-EOPNOTSUPP);
485 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
486 enum bpf_prog_type type,
489 return ERR_PTR(-EOPNOTSUPP);
492 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
495 return ERR_PTR(-EOPNOTSUPP);
498 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
502 static inline void bpf_prog_put(struct bpf_prog *prog)
506 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
508 return ERR_PTR(-EOPNOTSUPP);
511 static inline struct bpf_prog *__must_check
512 bpf_prog_inc_not_zero(struct bpf_prog *prog)
514 return ERR_PTR(-EOPNOTSUPP);
517 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
522 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
526 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
531 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
537 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
541 static inline void __dev_map_flush(struct bpf_map *map)
546 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
551 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index)
555 static inline void __cpu_map_flush(struct bpf_map *map)
560 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
561 struct xdp_buff *xdp,
562 struct net_device *dev_rx)
567 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
568 enum bpf_prog_type type)
570 return ERR_PTR(-EOPNOTSUPP);
572 #endif /* CONFIG_BPF_SYSCALL */
574 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
575 enum bpf_prog_type type)
577 return bpf_prog_get_type_dev(ufd, type, false);
580 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
582 int bpf_prog_offload_compile(struct bpf_prog *prog);
583 void bpf_prog_offload_destroy(struct bpf_prog *prog);
584 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
585 struct bpf_prog *prog);
587 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
588 int bpf_map_offload_update_elem(struct bpf_map *map,
589 void *key, void *value, u64 flags);
590 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
591 int bpf_map_offload_get_next_key(struct bpf_map *map,
592 void *key, void *next_key);
594 bool bpf_offload_dev_match(struct bpf_prog *prog, struct bpf_map *map);
596 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
597 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
599 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
601 return aux->offload_requested;
604 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
606 return unlikely(map->ops == &bpf_map_offload_ops);
609 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
610 void bpf_map_offload_map_free(struct bpf_map *map);
612 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
613 union bpf_attr *attr)
618 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
623 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
628 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
630 return ERR_PTR(-EOPNOTSUPP);
633 static inline void bpf_map_offload_map_free(struct bpf_map *map)
636 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
638 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_INET)
639 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
640 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
642 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
647 static inline int sock_map_prog(struct bpf_map *map,
648 struct bpf_prog *prog,
655 /* verifier prototypes for helper functions called from eBPF programs */
656 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
657 extern const struct bpf_func_proto bpf_map_update_elem_proto;
658 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
660 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
661 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
662 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
663 extern const struct bpf_func_proto bpf_tail_call_proto;
664 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
665 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
666 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
667 extern const struct bpf_func_proto bpf_get_current_comm_proto;
668 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
669 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
670 extern const struct bpf_func_proto bpf_get_stackid_proto;
671 extern const struct bpf_func_proto bpf_sock_map_update_proto;
673 /* Shared helpers among cBPF and eBPF. */
674 void bpf_user_rnd_init_once(void);
675 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
677 #endif /* _LINUX_BPF_H */