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;
26 /* map is generic key/value storage optionally accesible by eBPF programs */
28 /* funcs callable from userspace (via syscall) */
29 int (*map_alloc_check)(union bpf_attr *attr);
30 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
31 void (*map_release)(struct bpf_map *map, struct file *map_file);
32 void (*map_free)(struct bpf_map *map);
33 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
35 /* funcs callable from userspace and from eBPF programs */
36 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
37 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
38 int (*map_delete_elem)(struct bpf_map *map, void *key);
40 /* funcs called by prog_array and perf_event_array map */
41 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
43 void (*map_fd_put_ptr)(void *ptr);
44 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
45 u32 (*map_fd_sys_lookup_elem)(void *ptr);
49 /* 1st cacheline with read-mostly members of which some
50 * are also accessed in fast-path (e.g. ops, max_entries).
52 const struct bpf_map_ops *ops ____cacheline_aligned;
53 struct bpf_map *inner_map_meta;
54 #ifdef CONFIG_SECURITY
57 enum bpf_map_type map_type;
68 /* 2nd cacheline with misc members to avoid false sharing
69 * particularly with refcounting.
71 struct user_struct *user ____cacheline_aligned;
74 struct work_struct work;
75 char name[BPF_OBJ_NAME_LEN];
78 struct bpf_offloaded_map;
80 struct bpf_map_dev_ops {
81 int (*map_get_next_key)(struct bpf_offloaded_map *map,
82 void *key, void *next_key);
83 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
84 void *key, void *value);
85 int (*map_update_elem)(struct bpf_offloaded_map *map,
86 void *key, void *value, u64 flags);
87 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
90 struct bpf_offloaded_map {
92 struct net_device *netdev;
93 const struct bpf_map_dev_ops *dev_ops;
95 struct list_head offloads;
98 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
100 return container_of(map, struct bpf_offloaded_map, map);
103 extern const struct bpf_map_ops bpf_map_offload_ops;
105 /* function argument constraints */
107 ARG_DONTCARE = 0, /* unused argument in helper function */
109 /* the following constraints used to prototype
110 * bpf_map_lookup/update/delete_elem() functions
112 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
113 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
114 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
116 /* the following constraints used to prototype bpf_memcmp() and other
117 * functions that access data on eBPF program stack
119 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
120 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
121 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
122 * helper function must fill all bytes or clear
123 * them in error case.
126 ARG_CONST_SIZE, /* number of bytes accessed from memory */
127 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
129 ARG_PTR_TO_CTX, /* pointer to context */
130 ARG_ANYTHING, /* any (initialized) argument is ok */
133 /* type of values returned from helper functions */
134 enum bpf_return_type {
135 RET_INTEGER, /* function returns integer */
136 RET_VOID, /* function doesn't return anything */
137 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
140 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
141 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
142 * instructions after verifying
144 struct bpf_func_proto {
145 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
148 enum bpf_return_type ret_type;
149 enum bpf_arg_type arg1_type;
150 enum bpf_arg_type arg2_type;
151 enum bpf_arg_type arg3_type;
152 enum bpf_arg_type arg4_type;
153 enum bpf_arg_type arg5_type;
156 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
157 * the first argument to eBPF programs.
158 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
162 enum bpf_access_type {
167 /* types of values stored in eBPF registers */
168 /* Pointer types represent:
171 * pointer + (u16) var
172 * pointer + (u16) var + imm
173 * if (range > 0) then [ptr, ptr + range - off) is safe to access
174 * if (id > 0) means that some 'var' was added
175 * if (off > 0) means that 'imm' was added
178 NOT_INIT = 0, /* nothing was written into register */
179 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
180 PTR_TO_CTX, /* reg points to bpf_context */
181 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
182 PTR_TO_MAP_VALUE, /* reg points to map element value */
183 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
184 PTR_TO_STACK, /* reg == frame_pointer + offset */
185 PTR_TO_PACKET_META, /* skb->data - meta_len */
186 PTR_TO_PACKET, /* reg points to skb->data */
187 PTR_TO_PACKET_END, /* skb->data + headlen */
190 /* The information passed from prog-specific *_is_valid_access
191 * back to the verifier.
193 struct bpf_insn_access_aux {
194 enum bpf_reg_type reg_type;
199 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
201 aux->ctx_field_size = size;
204 struct bpf_prog_ops {
205 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
206 union bpf_attr __user *uattr);
209 struct bpf_verifier_ops {
210 /* return eBPF function prototype for verification */
211 const struct bpf_func_proto *
212 (*get_func_proto)(enum bpf_func_id func_id,
213 const struct bpf_prog *prog);
215 /* return true if 'size' wide access at offset 'off' within bpf_context
216 * with 'type' (read or write) is allowed
218 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
219 const struct bpf_prog *prog,
220 struct bpf_insn_access_aux *info);
221 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
222 const struct bpf_prog *prog);
223 u32 (*convert_ctx_access)(enum bpf_access_type type,
224 const struct bpf_insn *src,
225 struct bpf_insn *dst,
226 struct bpf_prog *prog, u32 *target_size);
229 struct bpf_prog_offload_ops {
230 int (*insn_hook)(struct bpf_verifier_env *env,
231 int insn_idx, int prev_insn_idx);
234 struct bpf_prog_offload {
235 struct bpf_prog *prog;
236 struct net_device *netdev;
238 struct list_head offloads;
240 const struct bpf_prog_offload_ops *dev_ops;
245 struct bpf_prog_aux {
252 bool offload_requested;
253 struct bpf_prog **func;
254 void *jit_data; /* JIT specific data. arch dependent */
255 struct latch_tree_node ksym_tnode;
256 struct list_head ksym_lnode;
257 const struct bpf_prog_ops *ops;
258 struct bpf_map **used_maps;
259 struct bpf_prog *prog;
260 struct user_struct *user;
261 u64 load_time; /* ns since boottime */
262 char name[BPF_OBJ_NAME_LEN];
263 #ifdef CONFIG_SECURITY
266 struct bpf_prog_offload *offload;
268 struct work_struct work;
277 /* 'ownership' of prog_array is claimed by the first program that
278 * is going to use this map or by the first program which FD is stored
279 * in the map to make sure that all callers and callees have the same
280 * prog_type and JITed flag
282 enum bpf_prog_type owner_prog_type;
285 char value[0] __aligned(8);
286 void *ptrs[0] __aligned(8);
287 void __percpu *pptrs[0] __aligned(8);
291 #define MAX_TAIL_CALL_CNT 32
293 struct bpf_event_entry {
294 struct perf_event *event;
295 struct file *perf_file;
296 struct file *map_file;
300 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
301 int bpf_prog_calc_tag(struct bpf_prog *fp);
303 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
305 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
306 unsigned long off, unsigned long len);
308 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
309 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
311 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
312 union bpf_attr __user *uattr);
313 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
314 union bpf_attr __user *uattr);
316 /* an array of programs to be executed under rcu_lock.
319 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
321 * the structure returned by bpf_prog_array_alloc() should be populated
322 * with program pointers and the last pointer must be NULL.
323 * The user has to keep refcnt on the program and make sure the program
324 * is removed from the array before bpf_prog_put().
325 * The 'struct bpf_prog_array *' should only be replaced with xchg()
326 * since other cpus are walking the array of pointers in parallel.
328 struct bpf_prog_array {
330 struct bpf_prog *progs[0];
333 struct bpf_prog_array __rcu *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
334 void bpf_prog_array_free(struct bpf_prog_array __rcu *progs);
335 int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
336 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
337 __u32 __user *prog_ids, u32 cnt);
339 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
340 struct bpf_prog *old_prog);
341 int bpf_prog_array_copy_info(struct bpf_prog_array __rcu *array,
342 u32 *prog_ids, u32 request_cnt,
344 int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
345 struct bpf_prog *exclude_prog,
346 struct bpf_prog *include_prog,
347 struct bpf_prog_array **new_array);
349 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
351 struct bpf_prog **_prog, *__prog; \
352 struct bpf_prog_array *_array; \
355 _array = rcu_dereference(array); \
356 if (unlikely(check_non_null && !_array))\
358 _prog = _array->progs; \
359 while ((__prog = READ_ONCE(*_prog))) { \
360 _ret &= func(__prog, ctx); \
368 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
369 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
371 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
372 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
374 #ifdef CONFIG_BPF_SYSCALL
375 DECLARE_PER_CPU(int, bpf_prog_active);
377 extern const struct file_operations bpf_map_fops;
378 extern const struct file_operations bpf_prog_fops;
380 #define BPF_PROG_TYPE(_id, _name) \
381 extern const struct bpf_prog_ops _name ## _prog_ops; \
382 extern const struct bpf_verifier_ops _name ## _verifier_ops;
383 #define BPF_MAP_TYPE(_id, _ops) \
384 extern const struct bpf_map_ops _ops;
385 #include <linux/bpf_types.h>
389 extern const struct bpf_prog_ops bpf_offload_prog_ops;
390 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
391 extern const struct bpf_verifier_ops xdp_analyzer_ops;
393 struct bpf_prog *bpf_prog_get(u32 ufd);
394 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
396 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
397 void bpf_prog_sub(struct bpf_prog *prog, int i);
398 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
399 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
400 void bpf_prog_put(struct bpf_prog *prog);
401 int __bpf_prog_charge(struct user_struct *user, u32 pages);
402 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
404 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
405 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
407 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
408 struct bpf_map *__bpf_map_get(struct fd f);
409 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
410 void bpf_map_put_with_uref(struct bpf_map *map);
411 void bpf_map_put(struct bpf_map *map);
412 int bpf_map_precharge_memlock(u32 pages);
413 void *bpf_map_area_alloc(size_t size, int numa_node);
414 void bpf_map_area_free(void *base);
415 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
417 extern int sysctl_unprivileged_bpf_disabled;
419 int bpf_map_new_fd(struct bpf_map *map, int flags);
420 int bpf_prog_new_fd(struct bpf_prog *prog);
422 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
423 int bpf_obj_get_user(const char __user *pathname, int flags);
425 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
426 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
427 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
429 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
432 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
434 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
435 void *key, void *value, u64 map_flags);
436 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
437 void bpf_fd_array_map_clear(struct bpf_map *map);
438 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
439 void *key, void *value, u64 map_flags);
440 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
442 int bpf_get_file_flag(int flags);
444 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
445 * forced to use 'long' read/writes to try to atomically copy long counters.
446 * Best-effort only. No barriers here, since it _will_ race with concurrent
447 * updates from BPF programs. Called from bpf syscall and mostly used with
448 * size 8 or 16 bytes, so ask compiler to inline it.
450 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
452 const long *lsrc = src;
455 size /= sizeof(long);
460 /* verify correctness of eBPF program */
461 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
462 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
465 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
466 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
467 void __dev_map_flush(struct bpf_map *map);
469 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
470 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index);
471 void __cpu_map_flush(struct bpf_map *map);
473 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
474 struct net_device *dev_rx);
476 /* Return map's numa specified by userspace */
477 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
479 return (attr->map_flags & BPF_F_NUMA_NODE) ?
480 attr->numa_node : NUMA_NO_NODE;
483 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
485 #else /* !CONFIG_BPF_SYSCALL */
486 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
488 return ERR_PTR(-EOPNOTSUPP);
491 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
492 enum bpf_prog_type type,
495 return ERR_PTR(-EOPNOTSUPP);
498 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
501 return ERR_PTR(-EOPNOTSUPP);
504 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
508 static inline void bpf_prog_put(struct bpf_prog *prog)
512 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
514 return ERR_PTR(-EOPNOTSUPP);
517 static inline struct bpf_prog *__must_check
518 bpf_prog_inc_not_zero(struct bpf_prog *prog)
520 return ERR_PTR(-EOPNOTSUPP);
523 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
528 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
532 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
537 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
543 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
547 static inline void __dev_map_flush(struct bpf_map *map)
552 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
557 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index)
561 static inline void __cpu_map_flush(struct bpf_map *map)
566 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
567 struct xdp_buff *xdp,
568 struct net_device *dev_rx)
573 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
574 enum bpf_prog_type type)
576 return ERR_PTR(-EOPNOTSUPP);
578 #endif /* CONFIG_BPF_SYSCALL */
580 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
581 enum bpf_prog_type type)
583 return bpf_prog_get_type_dev(ufd, type, false);
586 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
588 int bpf_prog_offload_compile(struct bpf_prog *prog);
589 void bpf_prog_offload_destroy(struct bpf_prog *prog);
590 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
591 struct bpf_prog *prog);
593 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
595 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
596 int bpf_map_offload_update_elem(struct bpf_map *map,
597 void *key, void *value, u64 flags);
598 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
599 int bpf_map_offload_get_next_key(struct bpf_map *map,
600 void *key, void *next_key);
602 bool bpf_offload_dev_match(struct bpf_prog *prog, struct bpf_map *map);
604 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
605 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
607 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
609 return aux->offload_requested;
612 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
614 return unlikely(map->ops == &bpf_map_offload_ops);
617 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
618 void bpf_map_offload_map_free(struct bpf_map *map);
620 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
621 union bpf_attr *attr)
626 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
631 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
636 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
638 return ERR_PTR(-EOPNOTSUPP);
641 static inline void bpf_map_offload_map_free(struct bpf_map *map)
644 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
646 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_INET)
647 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
648 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
650 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
655 static inline int sock_map_prog(struct bpf_map *map,
656 struct bpf_prog *prog,
663 /* verifier prototypes for helper functions called from eBPF programs */
664 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
665 extern const struct bpf_func_proto bpf_map_update_elem_proto;
666 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
668 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
669 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
670 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
671 extern const struct bpf_func_proto bpf_tail_call_proto;
672 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
673 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
674 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
675 extern const struct bpf_func_proto bpf_get_current_comm_proto;
676 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
677 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
678 extern const struct bpf_func_proto bpf_get_stackid_proto;
679 extern const struct bpf_func_proto bpf_sock_map_update_proto;
681 /* Shared helpers among cBPF and eBPF. */
682 void bpf_user_rnd_init_once(void);
683 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
685 #endif /* _LINUX_BPF_H */