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 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
29 void (*map_release)(struct bpf_map *map, struct file *map_file);
30 void (*map_free)(struct bpf_map *map);
31 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
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);
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,
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);
48 enum bpf_map_type map_type;
56 struct user_struct *user;
57 const struct bpf_map_ops *ops;
58 struct work_struct work;
60 struct bpf_map *inner_map_meta;
61 char name[BPF_OBJ_NAME_LEN];
62 #ifdef CONFIG_SECURITY
67 /* function argument constraints */
69 ARG_DONTCARE = 0, /* unused argument in helper function */
71 /* the following constraints used to prototype
72 * bpf_map_lookup/update/delete_elem() functions
74 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
75 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
76 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
78 /* the following constraints used to prototype bpf_memcmp() and other
79 * functions that access data on eBPF program stack
81 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
82 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
83 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
84 * helper function must fill all bytes or clear
88 ARG_CONST_SIZE, /* number of bytes accessed from memory */
89 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
91 ARG_PTR_TO_CTX, /* pointer to context */
92 ARG_ANYTHING, /* any (initialized) argument is ok */
95 /* type of values returned from helper functions */
96 enum bpf_return_type {
97 RET_INTEGER, /* function returns integer */
98 RET_VOID, /* function doesn't return anything */
99 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
102 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
103 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
104 * instructions after verifying
106 struct bpf_func_proto {
107 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
110 enum bpf_return_type ret_type;
111 enum bpf_arg_type arg1_type;
112 enum bpf_arg_type arg2_type;
113 enum bpf_arg_type arg3_type;
114 enum bpf_arg_type arg4_type;
115 enum bpf_arg_type arg5_type;
118 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
119 * the first argument to eBPF programs.
120 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
124 enum bpf_access_type {
129 /* types of values stored in eBPF registers */
130 /* Pointer types represent:
133 * pointer + (u16) var
134 * pointer + (u16) var + imm
135 * if (range > 0) then [ptr, ptr + range - off) is safe to access
136 * if (id > 0) means that some 'var' was added
137 * if (off > 0) means that 'imm' was added
140 NOT_INIT = 0, /* nothing was written into register */
141 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
142 PTR_TO_CTX, /* reg points to bpf_context */
143 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
144 PTR_TO_MAP_VALUE, /* reg points to map element value */
145 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
146 PTR_TO_STACK, /* reg == frame_pointer + offset */
147 PTR_TO_PACKET_META, /* skb->data - meta_len */
148 PTR_TO_PACKET, /* reg points to skb->data */
149 PTR_TO_PACKET_END, /* skb->data + headlen */
152 /* The information passed from prog-specific *_is_valid_access
153 * back to the verifier.
155 struct bpf_insn_access_aux {
156 enum bpf_reg_type reg_type;
161 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
163 aux->ctx_field_size = size;
166 struct bpf_prog_ops {
167 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
168 union bpf_attr __user *uattr);
171 struct bpf_verifier_ops {
172 /* return eBPF function prototype for verification */
173 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
175 /* return true if 'size' wide access at offset 'off' within bpf_context
176 * with 'type' (read or write) is allowed
178 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
179 struct bpf_insn_access_aux *info);
180 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
181 const struct bpf_prog *prog);
182 u32 (*convert_ctx_access)(enum bpf_access_type type,
183 const struct bpf_insn *src,
184 struct bpf_insn *dst,
185 struct bpf_prog *prog, u32 *target_size);
188 struct bpf_prog_offload_ops {
189 int (*insn_hook)(struct bpf_verifier_env *env,
190 int insn_idx, int prev_insn_idx);
193 struct bpf_dev_offload {
194 struct bpf_prog *prog;
195 struct net_device *netdev;
197 struct list_head offloads;
199 const struct bpf_prog_offload_ops *dev_ops;
202 struct bpf_prog_aux {
209 bool offload_requested;
210 struct bpf_prog **func;
211 void *jit_data; /* JIT specific data. arch dependent */
212 struct latch_tree_node ksym_tnode;
213 struct list_head ksym_lnode;
214 const struct bpf_prog_ops *ops;
215 struct bpf_map **used_maps;
216 struct bpf_prog *prog;
217 struct user_struct *user;
218 u64 load_time; /* ns since boottime */
219 char name[BPF_OBJ_NAME_LEN];
220 #ifdef CONFIG_SECURITY
223 struct bpf_dev_offload *offload;
225 struct work_struct work;
233 /* 'ownership' of prog_array is claimed by the first program that
234 * is going to use this map or by the first program which FD is stored
235 * in the map to make sure that all callers and callees have the same
236 * prog_type and JITed flag
238 enum bpf_prog_type owner_prog_type;
241 char value[0] __aligned(8);
242 void *ptrs[0] __aligned(8);
243 void __percpu *pptrs[0] __aligned(8);
247 #define MAX_TAIL_CALL_CNT 32
249 struct bpf_event_entry {
250 struct perf_event *event;
251 struct file *perf_file;
252 struct file *map_file;
256 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
257 int bpf_prog_calc_tag(struct bpf_prog *fp);
259 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
261 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
262 unsigned long off, unsigned long len);
264 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
265 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
267 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
268 union bpf_attr __user *uattr);
269 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
270 union bpf_attr __user *uattr);
272 /* an array of programs to be executed under rcu_lock.
275 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
277 * the structure returned by bpf_prog_array_alloc() should be populated
278 * with program pointers and the last pointer must be NULL.
279 * The user has to keep refcnt on the program and make sure the program
280 * is removed from the array before bpf_prog_put().
281 * The 'struct bpf_prog_array *' should only be replaced with xchg()
282 * since other cpus are walking the array of pointers in parallel.
284 struct bpf_prog_array {
286 struct bpf_prog *progs[0];
289 struct bpf_prog_array __rcu *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
290 void bpf_prog_array_free(struct bpf_prog_array __rcu *progs);
291 int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
292 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
293 __u32 __user *prog_ids, u32 cnt);
295 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
296 struct bpf_prog *old_prog);
297 int bpf_prog_array_copy_info(struct bpf_prog_array __rcu *array,
298 __u32 __user *prog_ids, u32 request_cnt,
299 __u32 __user *prog_cnt);
300 int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
301 struct bpf_prog *exclude_prog,
302 struct bpf_prog *include_prog,
303 struct bpf_prog_array **new_array);
305 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
307 struct bpf_prog **_prog, *__prog; \
308 struct bpf_prog_array *_array; \
311 _array = rcu_dereference(array); \
312 if (unlikely(check_non_null && !_array))\
314 _prog = _array->progs; \
315 while ((__prog = READ_ONCE(*_prog))) { \
316 _ret &= func(__prog, ctx); \
324 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
325 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
327 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
328 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
330 #ifdef CONFIG_BPF_SYSCALL
331 DECLARE_PER_CPU(int, bpf_prog_active);
333 extern const struct file_operations bpf_map_fops;
334 extern const struct file_operations bpf_prog_fops;
336 #define BPF_PROG_TYPE(_id, _name) \
337 extern const struct bpf_prog_ops _name ## _prog_ops; \
338 extern const struct bpf_verifier_ops _name ## _verifier_ops;
339 #define BPF_MAP_TYPE(_id, _ops) \
340 extern const struct bpf_map_ops _ops;
341 #include <linux/bpf_types.h>
345 extern const struct bpf_prog_ops bpf_offload_prog_ops;
346 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
347 extern const struct bpf_verifier_ops xdp_analyzer_ops;
349 struct bpf_prog *bpf_prog_get(u32 ufd);
350 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
352 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
353 void bpf_prog_sub(struct bpf_prog *prog, int i);
354 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
355 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
356 void bpf_prog_put(struct bpf_prog *prog);
357 int __bpf_prog_charge(struct user_struct *user, u32 pages);
358 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
360 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
361 struct bpf_map *__bpf_map_get(struct fd f);
362 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
363 void bpf_map_put_with_uref(struct bpf_map *map);
364 void bpf_map_put(struct bpf_map *map);
365 int bpf_map_precharge_memlock(u32 pages);
366 void *bpf_map_area_alloc(size_t size, int numa_node);
367 void bpf_map_area_free(void *base);
369 extern int sysctl_unprivileged_bpf_disabled;
371 int bpf_map_new_fd(struct bpf_map *map, int flags);
372 int bpf_prog_new_fd(struct bpf_prog *prog);
374 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
375 int bpf_obj_get_user(const char __user *pathname, int flags);
377 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
378 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
379 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
381 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
384 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
386 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
387 void *key, void *value, u64 map_flags);
388 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
389 void bpf_fd_array_map_clear(struct bpf_map *map);
390 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
391 void *key, void *value, u64 map_flags);
392 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
394 int bpf_get_file_flag(int flags);
396 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
397 * forced to use 'long' read/writes to try to atomically copy long counters.
398 * Best-effort only. No barriers here, since it _will_ race with concurrent
399 * updates from BPF programs. Called from bpf syscall and mostly used with
400 * size 8 or 16 bytes, so ask compiler to inline it.
402 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
404 const long *lsrc = src;
407 size /= sizeof(long);
412 /* verify correctness of eBPF program */
413 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
414 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
417 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
418 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
419 void __dev_map_flush(struct bpf_map *map);
421 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
422 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index);
423 void __cpu_map_flush(struct bpf_map *map);
425 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
426 struct net_device *dev_rx);
428 /* Return map's numa specified by userspace */
429 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
431 return (attr->map_flags & BPF_F_NUMA_NODE) ?
432 attr->numa_node : NUMA_NO_NODE;
435 #else /* !CONFIG_BPF_SYSCALL */
436 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
438 return ERR_PTR(-EOPNOTSUPP);
441 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
442 enum bpf_prog_type type,
445 return ERR_PTR(-EOPNOTSUPP);
448 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
451 return ERR_PTR(-EOPNOTSUPP);
454 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
458 static inline void bpf_prog_put(struct bpf_prog *prog)
462 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
464 return ERR_PTR(-EOPNOTSUPP);
467 static inline struct bpf_prog *__must_check
468 bpf_prog_inc_not_zero(struct bpf_prog *prog)
470 return ERR_PTR(-EOPNOTSUPP);
473 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
478 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
482 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
487 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
493 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
497 static inline void __dev_map_flush(struct bpf_map *map)
502 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
507 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index)
511 static inline void __cpu_map_flush(struct bpf_map *map)
516 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
517 struct xdp_buff *xdp,
518 struct net_device *dev_rx)
522 #endif /* CONFIG_BPF_SYSCALL */
524 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
525 enum bpf_prog_type type)
527 return bpf_prog_get_type_dev(ufd, type, false);
530 int bpf_prog_offload_compile(struct bpf_prog *prog);
531 void bpf_prog_offload_destroy(struct bpf_prog *prog);
533 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
534 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
536 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
538 return aux->offload_requested;
541 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
542 union bpf_attr *attr)
547 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
551 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
553 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
554 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
555 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
557 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
562 static inline int sock_map_prog(struct bpf_map *map,
563 struct bpf_prog *prog,
570 /* verifier prototypes for helper functions called from eBPF programs */
571 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
572 extern const struct bpf_func_proto bpf_map_update_elem_proto;
573 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
575 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
576 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
577 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
578 extern const struct bpf_func_proto bpf_tail_call_proto;
579 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
580 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
581 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
582 extern const struct bpf_func_proto bpf_get_current_comm_proto;
583 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
584 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
585 extern const struct bpf_func_proto bpf_get_stackid_proto;
586 extern const struct bpf_func_proto bpf_sock_map_update_proto;
588 /* Shared helpers among cBPF and eBPF. */
589 void bpf_user_rnd_init_once(void);
590 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
592 #if defined(__KERNEL__) && !defined(__ASSEMBLY__)
593 #ifdef CONFIG_BPF_KPROBE_OVERRIDE
594 #define BPF_ALLOW_ERROR_INJECTION(fname) \
595 static unsigned long __used \
596 __attribute__((__section__("_kprobe_error_inject_list"))) \
597 _eil_addr_##fname = (unsigned long)fname;
599 #define BPF_ALLOW_ERROR_INJECTION(fname)
603 #endif /* _LINUX_BPF_H */