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>
24 /* map is generic key/value storage optionally accesible by eBPF programs */
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);
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);
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,
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);
47 enum bpf_map_type map_type;
55 struct user_struct *user;
56 const struct bpf_map_ops *ops;
57 struct work_struct work;
59 struct bpf_map *inner_map_meta;
60 char name[BPF_OBJ_NAME_LEN];
61 #ifdef CONFIG_SECURITY
66 /* function argument constraints */
68 ARG_DONTCARE = 0, /* unused argument in helper function */
70 /* the following constraints used to prototype
71 * bpf_map_lookup/update/delete_elem() functions
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 */
77 /* the following constraints used to prototype bpf_memcmp() and other
78 * functions that access data on eBPF program stack
80 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
81 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
82 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
83 * helper function must fill all bytes or clear
87 ARG_CONST_SIZE, /* number of bytes accessed from memory */
88 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
90 ARG_PTR_TO_CTX, /* pointer to context */
91 ARG_ANYTHING, /* any (initialized) argument is ok */
94 /* type of values returned from helper functions */
95 enum bpf_return_type {
96 RET_INTEGER, /* function returns integer */
97 RET_VOID, /* function doesn't return anything */
98 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
101 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
102 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
103 * instructions after verifying
105 struct bpf_func_proto {
106 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
109 enum bpf_return_type ret_type;
110 enum bpf_arg_type arg1_type;
111 enum bpf_arg_type arg2_type;
112 enum bpf_arg_type arg3_type;
113 enum bpf_arg_type arg4_type;
114 enum bpf_arg_type arg5_type;
117 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
118 * the first argument to eBPF programs.
119 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
123 enum bpf_access_type {
128 /* types of values stored in eBPF registers */
129 /* Pointer types represent:
132 * pointer + (u16) var
133 * pointer + (u16) var + imm
134 * if (range > 0) then [ptr, ptr + range - off) is safe to access
135 * if (id > 0) means that some 'var' was added
136 * if (off > 0) means that 'imm' was added
139 NOT_INIT = 0, /* nothing was written into register */
140 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
141 PTR_TO_CTX, /* reg points to bpf_context */
142 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
143 PTR_TO_MAP_VALUE, /* reg points to map element value */
144 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
145 PTR_TO_STACK, /* reg == frame_pointer + offset */
146 PTR_TO_PACKET_META, /* skb->data - meta_len */
147 PTR_TO_PACKET, /* reg points to skb->data */
148 PTR_TO_PACKET_END, /* skb->data + headlen */
151 /* The information passed from prog-specific *_is_valid_access
152 * back to the verifier.
154 struct bpf_insn_access_aux {
155 enum bpf_reg_type reg_type;
160 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
162 aux->ctx_field_size = size;
165 struct bpf_prog_ops {
166 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
167 union bpf_attr __user *uattr);
170 struct bpf_verifier_ops {
171 /* return eBPF function prototype for verification */
172 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
174 /* return true if 'size' wide access at offset 'off' within bpf_context
175 * with 'type' (read or write) is allowed
177 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
178 struct bpf_insn_access_aux *info);
179 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
180 const struct bpf_prog *prog);
181 u32 (*convert_ctx_access)(enum bpf_access_type type,
182 const struct bpf_insn *src,
183 struct bpf_insn *dst,
184 struct bpf_prog *prog, u32 *target_size);
187 struct bpf_dev_offload {
188 struct bpf_prog *prog;
189 struct net_device *netdev;
191 struct list_head offloads;
193 bool verifier_running;
194 wait_queue_head_t verifier_done;
197 struct bpf_prog_aux {
203 struct latch_tree_node ksym_tnode;
204 struct list_head ksym_lnode;
205 const struct bpf_prog_ops *ops;
206 struct bpf_map **used_maps;
207 struct bpf_prog *prog;
208 struct user_struct *user;
209 u64 load_time; /* ns since boottime */
210 char name[BPF_OBJ_NAME_LEN];
211 #ifdef CONFIG_SECURITY
214 struct bpf_dev_offload *offload;
216 struct work_struct work;
224 /* 'ownership' of prog_array is claimed by the first program that
225 * is going to use this map or by the first program which FD is stored
226 * in the map to make sure that all callers and callees have the same
227 * prog_type and JITed flag
229 enum bpf_prog_type owner_prog_type;
232 char value[0] __aligned(8);
233 void *ptrs[0] __aligned(8);
234 void __percpu *pptrs[0] __aligned(8);
238 #define MAX_TAIL_CALL_CNT 32
240 struct bpf_event_entry {
241 struct perf_event *event;
242 struct file *perf_file;
243 struct file *map_file;
247 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
248 int bpf_prog_calc_tag(struct bpf_prog *fp);
250 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
252 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
253 unsigned long off, unsigned long len);
255 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
256 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
257 int bpf_event_query_prog_array(struct perf_event *event, void __user *info);
259 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
260 union bpf_attr __user *uattr);
261 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
262 union bpf_attr __user *uattr);
264 /* an array of programs to be executed under rcu_lock.
267 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
269 * the structure returned by bpf_prog_array_alloc() should be populated
270 * with program pointers and the last pointer must be NULL.
271 * The user has to keep refcnt on the program and make sure the program
272 * is removed from the array before bpf_prog_put().
273 * The 'struct bpf_prog_array *' should only be replaced with xchg()
274 * since other cpus are walking the array of pointers in parallel.
276 struct bpf_prog_array {
278 struct bpf_prog *progs[0];
281 struct bpf_prog_array __rcu *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
282 void bpf_prog_array_free(struct bpf_prog_array __rcu *progs);
283 int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
284 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
285 __u32 __user *prog_ids, u32 cnt);
287 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
288 struct bpf_prog *old_prog);
289 int bpf_prog_array_copy_info(struct bpf_prog_array __rcu *array,
290 __u32 __user *prog_ids, u32 request_cnt,
291 __u32 __user *prog_cnt);
292 int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
293 struct bpf_prog *exclude_prog,
294 struct bpf_prog *include_prog,
295 struct bpf_prog_array **new_array);
297 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
299 struct bpf_prog **_prog, *__prog; \
300 struct bpf_prog_array *_array; \
303 _array = rcu_dereference(array); \
304 if (unlikely(check_non_null && !_array))\
306 _prog = _array->progs; \
307 while ((__prog = READ_ONCE(*_prog))) { \
308 _ret &= func(__prog, ctx); \
316 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
317 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
319 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
320 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
322 #ifdef CONFIG_BPF_SYSCALL
323 DECLARE_PER_CPU(int, bpf_prog_active);
325 extern const struct file_operations bpf_map_fops;
326 extern const struct file_operations bpf_prog_fops;
328 #define BPF_PROG_TYPE(_id, _name) \
329 extern const struct bpf_prog_ops _name ## _prog_ops; \
330 extern const struct bpf_verifier_ops _name ## _verifier_ops;
331 #define BPF_MAP_TYPE(_id, _ops) \
332 extern const struct bpf_map_ops _ops;
333 #include <linux/bpf_types.h>
337 extern const struct bpf_prog_ops bpf_offload_prog_ops;
338 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
339 extern const struct bpf_verifier_ops xdp_analyzer_ops;
341 struct bpf_prog *bpf_prog_get(u32 ufd);
342 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
344 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
345 void bpf_prog_sub(struct bpf_prog *prog, int i);
346 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
347 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
348 void bpf_prog_put(struct bpf_prog *prog);
349 int __bpf_prog_charge(struct user_struct *user, u32 pages);
350 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
352 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
353 struct bpf_map *__bpf_map_get(struct fd f);
354 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
355 void bpf_map_put_with_uref(struct bpf_map *map);
356 void bpf_map_put(struct bpf_map *map);
357 int bpf_map_precharge_memlock(u32 pages);
358 void *bpf_map_area_alloc(size_t size, int numa_node);
359 void bpf_map_area_free(void *base);
361 extern int sysctl_unprivileged_bpf_disabled;
363 int bpf_map_new_fd(struct bpf_map *map, int flags);
364 int bpf_prog_new_fd(struct bpf_prog *prog);
366 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
367 int bpf_obj_get_user(const char __user *pathname, int flags);
369 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
370 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
371 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
373 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
376 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
378 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
379 void *key, void *value, u64 map_flags);
380 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
381 void bpf_fd_array_map_clear(struct bpf_map *map);
382 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
383 void *key, void *value, u64 map_flags);
384 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
386 int bpf_get_file_flag(int flags);
388 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
389 * forced to use 'long' read/writes to try to atomically copy long counters.
390 * Best-effort only. No barriers here, since it _will_ race with concurrent
391 * updates from BPF programs. Called from bpf syscall and mostly used with
392 * size 8 or 16 bytes, so ask compiler to inline it.
394 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
396 const long *lsrc = src;
399 size /= sizeof(long);
404 /* verify correctness of eBPF program */
405 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
408 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
409 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
410 void __dev_map_flush(struct bpf_map *map);
412 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
413 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index);
414 void __cpu_map_flush(struct bpf_map *map);
416 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
417 struct net_device *dev_rx);
419 /* Return map's numa specified by userspace */
420 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
422 return (attr->map_flags & BPF_F_NUMA_NODE) ?
423 attr->numa_node : NUMA_NO_NODE;
426 #else /* !CONFIG_BPF_SYSCALL */
427 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
429 return ERR_PTR(-EOPNOTSUPP);
432 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
433 enum bpf_prog_type type,
436 return ERR_PTR(-EOPNOTSUPP);
439 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
442 return ERR_PTR(-EOPNOTSUPP);
445 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
449 static inline void bpf_prog_put(struct bpf_prog *prog)
453 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
455 return ERR_PTR(-EOPNOTSUPP);
458 static inline struct bpf_prog *__must_check
459 bpf_prog_inc_not_zero(struct bpf_prog *prog)
461 return ERR_PTR(-EOPNOTSUPP);
464 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
469 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
473 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
478 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
484 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
488 static inline void __dev_map_flush(struct bpf_map *map)
493 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
498 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index)
502 static inline void __cpu_map_flush(struct bpf_map *map)
507 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
508 struct xdp_buff *xdp,
509 struct net_device *dev_rx)
513 #endif /* CONFIG_BPF_SYSCALL */
515 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
516 enum bpf_prog_type type)
518 return bpf_prog_get_type_dev(ufd, type, false);
521 int bpf_prog_offload_compile(struct bpf_prog *prog);
522 void bpf_prog_offload_destroy(struct bpf_prog *prog);
524 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
525 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
527 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
532 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
533 union bpf_attr *attr)
538 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
542 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
544 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
545 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
546 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
548 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
553 static inline int sock_map_prog(struct bpf_map *map,
554 struct bpf_prog *prog,
561 /* verifier prototypes for helper functions called from eBPF programs */
562 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
563 extern const struct bpf_func_proto bpf_map_update_elem_proto;
564 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
566 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
567 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
568 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
569 extern const struct bpf_func_proto bpf_tail_call_proto;
570 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
571 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
572 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
573 extern const struct bpf_func_proto bpf_get_current_comm_proto;
574 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
575 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
576 extern const struct bpf_func_proto bpf_get_stackid_proto;
577 extern const struct bpf_func_proto bpf_sock_map_update_proto;
579 /* Shared helpers among cBPF and eBPF. */
580 void bpf_user_rnd_init_once(void);
581 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
583 #if defined(__KERNEL__) && !defined(__ASSEMBLY__)
584 #ifdef CONFIG_BPF_KPROBE_OVERRIDE
585 #define BPF_ALLOW_ERROR_INJECTION(fname) \
586 static unsigned long __used \
587 __attribute__((__section__("_kprobe_error_inject_list"))) \
588 _eil_addr_##fname = (unsigned long)fname;
590 #define BPF_ALLOW_ERROR_INJECTION(fname)
594 #endif /* _LINUX_BPF_H */