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);
46 /* 1st cacheline with read-mostly members of which some
47 * are also accessed in fast-path (e.g. ops, max_entries).
49 const struct bpf_map_ops *ops ____cacheline_aligned;
50 struct bpf_map *inner_map_meta;
51 #ifdef CONFIG_SECURITY
54 enum bpf_map_type map_type;
65 /* 2nd cacheline with misc members to avoid false sharing
66 * particularly with refcounting.
68 struct user_struct *user ____cacheline_aligned;
71 struct work_struct work;
72 char name[BPF_OBJ_NAME_LEN];
75 /* function argument constraints */
77 ARG_DONTCARE = 0, /* unused argument in helper function */
79 /* the following constraints used to prototype
80 * bpf_map_lookup/update/delete_elem() functions
82 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
83 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
84 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
86 /* the following constraints used to prototype bpf_memcmp() and other
87 * functions that access data on eBPF program stack
89 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
90 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
91 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
92 * helper function must fill all bytes or clear
96 ARG_CONST_SIZE, /* number of bytes accessed from memory */
97 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
99 ARG_PTR_TO_CTX, /* pointer to context */
100 ARG_ANYTHING, /* any (initialized) argument is ok */
103 /* type of values returned from helper functions */
104 enum bpf_return_type {
105 RET_INTEGER, /* function returns integer */
106 RET_VOID, /* function doesn't return anything */
107 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
110 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
111 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
112 * instructions after verifying
114 struct bpf_func_proto {
115 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
118 enum bpf_return_type ret_type;
119 enum bpf_arg_type arg1_type;
120 enum bpf_arg_type arg2_type;
121 enum bpf_arg_type arg3_type;
122 enum bpf_arg_type arg4_type;
123 enum bpf_arg_type arg5_type;
126 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
127 * the first argument to eBPF programs.
128 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
132 enum bpf_access_type {
137 /* types of values stored in eBPF registers */
138 /* Pointer types represent:
141 * pointer + (u16) var
142 * pointer + (u16) var + imm
143 * if (range > 0) then [ptr, ptr + range - off) is safe to access
144 * if (id > 0) means that some 'var' was added
145 * if (off > 0) means that 'imm' was added
148 NOT_INIT = 0, /* nothing was written into register */
149 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
150 PTR_TO_CTX, /* reg points to bpf_context */
151 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
152 PTR_TO_MAP_VALUE, /* reg points to map element value */
153 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
154 PTR_TO_STACK, /* reg == frame_pointer + offset */
155 PTR_TO_PACKET_META, /* skb->data - meta_len */
156 PTR_TO_PACKET, /* reg points to skb->data */
157 PTR_TO_PACKET_END, /* skb->data + headlen */
160 /* The information passed from prog-specific *_is_valid_access
161 * back to the verifier.
163 struct bpf_insn_access_aux {
164 enum bpf_reg_type reg_type;
169 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
171 aux->ctx_field_size = size;
174 struct bpf_prog_ops {
175 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
176 union bpf_attr __user *uattr);
179 struct bpf_verifier_ops {
180 /* return eBPF function prototype for verification */
181 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
183 /* return true if 'size' wide access at offset 'off' within bpf_context
184 * with 'type' (read or write) is allowed
186 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
187 struct bpf_insn_access_aux *info);
188 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
189 const struct bpf_prog *prog);
190 u32 (*convert_ctx_access)(enum bpf_access_type type,
191 const struct bpf_insn *src,
192 struct bpf_insn *dst,
193 struct bpf_prog *prog, u32 *target_size);
196 struct bpf_dev_offload {
197 struct bpf_prog *prog;
198 struct net_device *netdev;
200 struct list_head offloads;
202 bool verifier_running;
203 wait_queue_head_t verifier_done;
206 struct bpf_prog_aux {
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;
234 /* 'ownership' of prog_array is claimed by the first program that
235 * is going to use this map or by the first program which FD is stored
236 * in the map to make sure that all callers and callees have the same
237 * prog_type and JITed flag
239 enum bpf_prog_type owner_prog_type;
242 char value[0] __aligned(8);
243 void *ptrs[0] __aligned(8);
244 void __percpu *pptrs[0] __aligned(8);
248 #define MAX_TAIL_CALL_CNT 32
250 struct bpf_event_entry {
251 struct perf_event *event;
252 struct file *perf_file;
253 struct file *map_file;
257 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
258 int bpf_prog_calc_tag(struct bpf_prog *fp);
260 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
262 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
263 unsigned long off, unsigned long len);
265 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
266 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
268 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
269 union bpf_attr __user *uattr);
270 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
271 union bpf_attr __user *uattr);
273 /* an array of programs to be executed under rcu_lock.
276 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
278 * the structure returned by bpf_prog_array_alloc() should be populated
279 * with program pointers and the last pointer must be NULL.
280 * The user has to keep refcnt on the program and make sure the program
281 * is removed from the array before bpf_prog_put().
282 * The 'struct bpf_prog_array *' should only be replaced with xchg()
283 * since other cpus are walking the array of pointers in parallel.
285 struct bpf_prog_array {
287 struct bpf_prog *progs[0];
290 struct bpf_prog_array __rcu *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
291 void bpf_prog_array_free(struct bpf_prog_array __rcu *progs);
292 int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
293 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
294 __u32 __user *prog_ids, u32 cnt);
296 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
297 struct bpf_prog *old_prog);
298 int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
299 struct bpf_prog *exclude_prog,
300 struct bpf_prog *include_prog,
301 struct bpf_prog_array **new_array);
303 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
305 struct bpf_prog **_prog, *__prog; \
306 struct bpf_prog_array *_array; \
309 _array = rcu_dereference(array); \
310 if (unlikely(check_non_null && !_array))\
312 _prog = _array->progs; \
313 while ((__prog = READ_ONCE(*_prog))) { \
314 _ret &= func(__prog, ctx); \
322 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
323 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
325 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
326 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
328 #ifdef CONFIG_BPF_SYSCALL
329 DECLARE_PER_CPU(int, bpf_prog_active);
331 extern const struct file_operations bpf_map_fops;
332 extern const struct file_operations bpf_prog_fops;
334 #define BPF_PROG_TYPE(_id, _name) \
335 extern const struct bpf_prog_ops _name ## _prog_ops; \
336 extern const struct bpf_verifier_ops _name ## _verifier_ops;
337 #define BPF_MAP_TYPE(_id, _ops) \
338 extern const struct bpf_map_ops _ops;
339 #include <linux/bpf_types.h>
343 extern const struct bpf_prog_ops bpf_offload_prog_ops;
344 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
345 extern const struct bpf_verifier_ops xdp_analyzer_ops;
347 struct bpf_prog *bpf_prog_get(u32 ufd);
348 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
350 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
351 void bpf_prog_sub(struct bpf_prog *prog, int i);
352 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
353 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
354 void bpf_prog_put(struct bpf_prog *prog);
355 int __bpf_prog_charge(struct user_struct *user, u32 pages);
356 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
358 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
359 struct bpf_map *__bpf_map_get(struct fd f);
360 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
361 void bpf_map_put_with_uref(struct bpf_map *map);
362 void bpf_map_put(struct bpf_map *map);
363 int bpf_map_precharge_memlock(u32 pages);
364 void *bpf_map_area_alloc(size_t size, int numa_node);
365 void bpf_map_area_free(void *base);
367 extern int sysctl_unprivileged_bpf_disabled;
369 int bpf_map_new_fd(struct bpf_map *map, int flags);
370 int bpf_prog_new_fd(struct bpf_prog *prog);
372 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
373 int bpf_obj_get_user(const char __user *pathname, int flags);
375 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
376 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
377 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
379 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
382 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
384 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
385 void *key, void *value, u64 map_flags);
386 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
387 void bpf_fd_array_map_clear(struct bpf_map *map);
388 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
389 void *key, void *value, u64 map_flags);
390 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
392 int bpf_get_file_flag(int flags);
394 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
395 * forced to use 'long' read/writes to try to atomically copy long counters.
396 * Best-effort only. No barriers here, since it _will_ race with concurrent
397 * updates from BPF programs. Called from bpf syscall and mostly used with
398 * size 8 or 16 bytes, so ask compiler to inline it.
400 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
402 const long *lsrc = src;
405 size /= sizeof(long);
410 /* verify correctness of eBPF program */
411 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
414 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
415 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
416 void __dev_map_flush(struct bpf_map *map);
418 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
419 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index);
420 void __cpu_map_flush(struct bpf_map *map);
422 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
423 struct net_device *dev_rx);
425 /* Return map's numa specified by userspace */
426 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
428 return (attr->map_flags & BPF_F_NUMA_NODE) ?
429 attr->numa_node : NUMA_NO_NODE;
432 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
434 #else /* !CONFIG_BPF_SYSCALL */
435 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
437 return ERR_PTR(-EOPNOTSUPP);
440 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
441 enum bpf_prog_type type,
444 return ERR_PTR(-EOPNOTSUPP);
447 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
450 return ERR_PTR(-EOPNOTSUPP);
453 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
457 static inline void bpf_prog_put(struct bpf_prog *prog)
461 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
463 return ERR_PTR(-EOPNOTSUPP);
466 static inline struct bpf_prog *__must_check
467 bpf_prog_inc_not_zero(struct bpf_prog *prog)
469 return ERR_PTR(-EOPNOTSUPP);
472 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
477 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
481 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
486 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
492 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
496 static inline void __dev_map_flush(struct bpf_map *map)
501 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
506 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index)
510 static inline void __cpu_map_flush(struct bpf_map *map)
515 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
516 struct xdp_buff *xdp,
517 struct net_device *dev_rx)
522 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
523 enum bpf_prog_type type)
525 return ERR_PTR(-EOPNOTSUPP);
527 #endif /* CONFIG_BPF_SYSCALL */
529 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
530 enum bpf_prog_type type)
532 return bpf_prog_get_type_dev(ufd, type, false);
535 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
537 int bpf_prog_offload_compile(struct bpf_prog *prog);
538 void bpf_prog_offload_destroy(struct bpf_prog *prog);
540 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
541 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
543 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
548 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
549 union bpf_attr *attr)
554 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
558 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
560 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
561 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
562 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
564 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
569 static inline int sock_map_prog(struct bpf_map *map,
570 struct bpf_prog *prog,
577 /* verifier prototypes for helper functions called from eBPF programs */
578 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
579 extern const struct bpf_func_proto bpf_map_update_elem_proto;
580 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
582 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
583 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
584 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
585 extern const struct bpf_func_proto bpf_tail_call_proto;
586 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
587 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
588 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
589 extern const struct bpf_func_proto bpf_get_current_comm_proto;
590 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
591 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
592 extern const struct bpf_func_proto bpf_get_stackid_proto;
593 extern const struct bpf_func_proto bpf_sock_map_update_proto;
595 /* Shared helpers among cBPF and eBPF. */
596 void bpf_user_rnd_init_once(void);
597 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
599 #endif /* _LINUX_BPF_H */