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);
34 void (*map_release_uref)(struct bpf_map *map);
36 /* funcs callable from userspace and from eBPF programs */
37 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
38 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
39 int (*map_delete_elem)(struct bpf_map *map, void *key);
41 /* funcs called by prog_array and perf_event_array map */
42 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
44 void (*map_fd_put_ptr)(void *ptr);
45 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
46 u32 (*map_fd_sys_lookup_elem)(void *ptr);
50 /* 1st cacheline with read-mostly members of which some
51 * are also accessed in fast-path (e.g. ops, max_entries).
53 const struct bpf_map_ops *ops ____cacheline_aligned;
54 struct bpf_map *inner_map_meta;
55 #ifdef CONFIG_SECURITY
58 enum bpf_map_type map_type;
69 /* 2nd cacheline with misc members to avoid false sharing
70 * particularly with refcounting.
72 struct user_struct *user ____cacheline_aligned;
75 struct work_struct work;
76 char name[BPF_OBJ_NAME_LEN];
79 struct bpf_offloaded_map;
81 struct bpf_map_dev_ops {
82 int (*map_get_next_key)(struct bpf_offloaded_map *map,
83 void *key, void *next_key);
84 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
85 void *key, void *value);
86 int (*map_update_elem)(struct bpf_offloaded_map *map,
87 void *key, void *value, u64 flags);
88 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
91 struct bpf_offloaded_map {
93 struct net_device *netdev;
94 const struct bpf_map_dev_ops *dev_ops;
96 struct list_head offloads;
99 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
101 return container_of(map, struct bpf_offloaded_map, map);
104 extern const struct bpf_map_ops bpf_map_offload_ops;
106 /* function argument constraints */
108 ARG_DONTCARE = 0, /* unused argument in helper function */
110 /* the following constraints used to prototype
111 * bpf_map_lookup/update/delete_elem() functions
113 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
114 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
115 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
117 /* the following constraints used to prototype bpf_memcmp() and other
118 * functions that access data on eBPF program stack
120 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
121 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
122 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
123 * helper function must fill all bytes or clear
124 * them in error case.
127 ARG_CONST_SIZE, /* number of bytes accessed from memory */
128 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
130 ARG_PTR_TO_CTX, /* pointer to context */
131 ARG_ANYTHING, /* any (initialized) argument is ok */
134 /* type of values returned from helper functions */
135 enum bpf_return_type {
136 RET_INTEGER, /* function returns integer */
137 RET_VOID, /* function doesn't return anything */
138 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
141 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
142 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
143 * instructions after verifying
145 struct bpf_func_proto {
146 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
149 enum bpf_return_type ret_type;
150 enum bpf_arg_type arg1_type;
151 enum bpf_arg_type arg2_type;
152 enum bpf_arg_type arg3_type;
153 enum bpf_arg_type arg4_type;
154 enum bpf_arg_type arg5_type;
157 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
158 * the first argument to eBPF programs.
159 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
163 enum bpf_access_type {
168 /* types of values stored in eBPF registers */
169 /* Pointer types represent:
172 * pointer + (u16) var
173 * pointer + (u16) var + imm
174 * if (range > 0) then [ptr, ptr + range - off) is safe to access
175 * if (id > 0) means that some 'var' was added
176 * if (off > 0) means that 'imm' was added
179 NOT_INIT = 0, /* nothing was written into register */
180 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
181 PTR_TO_CTX, /* reg points to bpf_context */
182 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
183 PTR_TO_MAP_VALUE, /* reg points to map element value */
184 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
185 PTR_TO_STACK, /* reg == frame_pointer + offset */
186 PTR_TO_PACKET_META, /* skb->data - meta_len */
187 PTR_TO_PACKET, /* reg points to skb->data */
188 PTR_TO_PACKET_END, /* skb->data + headlen */
191 /* The information passed from prog-specific *_is_valid_access
192 * back to the verifier.
194 struct bpf_insn_access_aux {
195 enum bpf_reg_type reg_type;
200 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
202 aux->ctx_field_size = size;
205 struct bpf_prog_ops {
206 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
207 union bpf_attr __user *uattr);
210 struct bpf_verifier_ops {
211 /* return eBPF function prototype for verification */
212 const struct bpf_func_proto *
213 (*get_func_proto)(enum bpf_func_id func_id,
214 const struct bpf_prog *prog);
216 /* return true if 'size' wide access at offset 'off' within bpf_context
217 * with 'type' (read or write) is allowed
219 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
220 const struct bpf_prog *prog,
221 struct bpf_insn_access_aux *info);
222 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
223 const struct bpf_prog *prog);
224 u32 (*convert_ctx_access)(enum bpf_access_type type,
225 const struct bpf_insn *src,
226 struct bpf_insn *dst,
227 struct bpf_prog *prog, u32 *target_size);
230 struct bpf_prog_offload_ops {
231 int (*insn_hook)(struct bpf_verifier_env *env,
232 int insn_idx, int prev_insn_idx);
235 struct bpf_prog_offload {
236 struct bpf_prog *prog;
237 struct net_device *netdev;
239 struct list_head offloads;
241 const struct bpf_prog_offload_ops *dev_ops;
246 struct bpf_prog_aux {
253 bool offload_requested;
254 struct bpf_prog **func;
255 void *jit_data; /* JIT specific data. arch dependent */
256 struct latch_tree_node ksym_tnode;
257 struct list_head ksym_lnode;
258 const struct bpf_prog_ops *ops;
259 struct bpf_map **used_maps;
260 struct bpf_prog *prog;
261 struct user_struct *user;
262 u64 load_time; /* ns since boottime */
263 char name[BPF_OBJ_NAME_LEN];
264 #ifdef CONFIG_SECURITY
267 struct bpf_prog_offload *offload;
269 struct work_struct work;
278 /* 'ownership' of prog_array is claimed by the first program that
279 * is going to use this map or by the first program which FD is stored
280 * in the map to make sure that all callers and callees have the same
281 * prog_type and JITed flag
283 enum bpf_prog_type owner_prog_type;
286 char value[0] __aligned(8);
287 void *ptrs[0] __aligned(8);
288 void __percpu *pptrs[0] __aligned(8);
292 #define MAX_TAIL_CALL_CNT 32
294 struct bpf_event_entry {
295 struct perf_event *event;
296 struct file *perf_file;
297 struct file *map_file;
301 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
302 int bpf_prog_calc_tag(struct bpf_prog *fp);
304 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
306 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
307 unsigned long off, unsigned long len);
309 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
310 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
312 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
313 union bpf_attr __user *uattr);
314 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
315 union bpf_attr __user *uattr);
317 /* an array of programs to be executed under rcu_lock.
320 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
322 * the structure returned by bpf_prog_array_alloc() should be populated
323 * with program pointers and the last pointer must be NULL.
324 * The user has to keep refcnt on the program and make sure the program
325 * is removed from the array before bpf_prog_put().
326 * The 'struct bpf_prog_array *' should only be replaced with xchg()
327 * since other cpus are walking the array of pointers in parallel.
329 struct bpf_prog_array {
331 struct bpf_prog *progs[0];
334 struct bpf_prog_array __rcu *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
335 void bpf_prog_array_free(struct bpf_prog_array __rcu *progs);
336 int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
337 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
338 __u32 __user *prog_ids, u32 cnt);
340 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
341 struct bpf_prog *old_prog);
342 int bpf_prog_array_copy_info(struct bpf_prog_array __rcu *array,
343 u32 *prog_ids, u32 request_cnt,
345 int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
346 struct bpf_prog *exclude_prog,
347 struct bpf_prog *include_prog,
348 struct bpf_prog_array **new_array);
350 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
352 struct bpf_prog **_prog, *__prog; \
353 struct bpf_prog_array *_array; \
357 _array = rcu_dereference(array); \
358 if (unlikely(check_non_null && !_array))\
360 _prog = _array->progs; \
361 while ((__prog = READ_ONCE(*_prog))) { \
362 _ret &= func(__prog, ctx); \
367 preempt_enable_no_resched(); \
371 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
372 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
374 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
375 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
377 #ifdef CONFIG_BPF_SYSCALL
378 DECLARE_PER_CPU(int, bpf_prog_active);
380 extern const struct file_operations bpf_map_fops;
381 extern const struct file_operations bpf_prog_fops;
383 #define BPF_PROG_TYPE(_id, _name) \
384 extern const struct bpf_prog_ops _name ## _prog_ops; \
385 extern const struct bpf_verifier_ops _name ## _verifier_ops;
386 #define BPF_MAP_TYPE(_id, _ops) \
387 extern const struct bpf_map_ops _ops;
388 #include <linux/bpf_types.h>
392 extern const struct bpf_prog_ops bpf_offload_prog_ops;
393 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
394 extern const struct bpf_verifier_ops xdp_analyzer_ops;
396 struct bpf_prog *bpf_prog_get(u32 ufd);
397 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
399 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
400 void bpf_prog_sub(struct bpf_prog *prog, int i);
401 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
402 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
403 void bpf_prog_put(struct bpf_prog *prog);
404 int __bpf_prog_charge(struct user_struct *user, u32 pages);
405 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
407 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
408 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
410 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
411 struct bpf_map *__bpf_map_get(struct fd f);
412 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
413 void bpf_map_put_with_uref(struct bpf_map *map);
414 void bpf_map_put(struct bpf_map *map);
415 int bpf_map_precharge_memlock(u32 pages);
416 void *bpf_map_area_alloc(size_t size, int numa_node);
417 void bpf_map_area_free(void *base);
418 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
420 extern int sysctl_unprivileged_bpf_disabled;
422 int bpf_map_new_fd(struct bpf_map *map, int flags);
423 int bpf_prog_new_fd(struct bpf_prog *prog);
425 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
426 int bpf_obj_get_user(const char __user *pathname, int flags);
428 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
429 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
430 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
432 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
435 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
437 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
438 void *key, void *value, u64 map_flags);
439 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
440 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
441 void *key, void *value, u64 map_flags);
442 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
444 int bpf_get_file_flag(int flags);
446 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
447 * forced to use 'long' read/writes to try to atomically copy long counters.
448 * Best-effort only. No barriers here, since it _will_ race with concurrent
449 * updates from BPF programs. Called from bpf syscall and mostly used with
450 * size 8 or 16 bytes, so ask compiler to inline it.
452 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
454 const long *lsrc = src;
457 size /= sizeof(long);
462 /* verify correctness of eBPF program */
463 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
464 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
467 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
468 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
469 void __dev_map_flush(struct bpf_map *map);
471 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
472 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index);
473 void __cpu_map_flush(struct bpf_map *map);
475 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
476 struct net_device *dev_rx);
478 /* Return map's numa specified by userspace */
479 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
481 return (attr->map_flags & BPF_F_NUMA_NODE) ?
482 attr->numa_node : NUMA_NO_NODE;
485 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
487 #else /* !CONFIG_BPF_SYSCALL */
488 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
490 return ERR_PTR(-EOPNOTSUPP);
493 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
494 enum bpf_prog_type type,
497 return ERR_PTR(-EOPNOTSUPP);
500 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
503 return ERR_PTR(-EOPNOTSUPP);
506 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
510 static inline void bpf_prog_put(struct bpf_prog *prog)
514 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
516 return ERR_PTR(-EOPNOTSUPP);
519 static inline struct bpf_prog *__must_check
520 bpf_prog_inc_not_zero(struct bpf_prog *prog)
522 return ERR_PTR(-EOPNOTSUPP);
525 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
530 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
534 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
539 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
545 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
549 static inline void __dev_map_flush(struct bpf_map *map)
554 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
559 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index)
563 static inline void __cpu_map_flush(struct bpf_map *map)
568 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
569 struct xdp_buff *xdp,
570 struct net_device *dev_rx)
575 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
576 enum bpf_prog_type type)
578 return ERR_PTR(-EOPNOTSUPP);
580 #endif /* CONFIG_BPF_SYSCALL */
582 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
583 enum bpf_prog_type type)
585 return bpf_prog_get_type_dev(ufd, type, false);
588 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
590 int bpf_prog_offload_compile(struct bpf_prog *prog);
591 void bpf_prog_offload_destroy(struct bpf_prog *prog);
592 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
593 struct bpf_prog *prog);
595 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
597 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
598 int bpf_map_offload_update_elem(struct bpf_map *map,
599 void *key, void *value, u64 flags);
600 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
601 int bpf_map_offload_get_next_key(struct bpf_map *map,
602 void *key, void *next_key);
604 bool bpf_offload_dev_match(struct bpf_prog *prog, struct bpf_map *map);
606 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
607 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
609 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
611 return aux->offload_requested;
614 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
616 return unlikely(map->ops == &bpf_map_offload_ops);
619 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
620 void bpf_map_offload_map_free(struct bpf_map *map);
622 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
623 union bpf_attr *attr)
628 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
633 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
638 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
640 return ERR_PTR(-EOPNOTSUPP);
643 static inline void bpf_map_offload_map_free(struct bpf_map *map)
646 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
648 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_INET)
649 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
650 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
652 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
657 static inline int sock_map_prog(struct bpf_map *map,
658 struct bpf_prog *prog,
665 /* verifier prototypes for helper functions called from eBPF programs */
666 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
667 extern const struct bpf_func_proto bpf_map_update_elem_proto;
668 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
670 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
671 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
672 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
673 extern const struct bpf_func_proto bpf_tail_call_proto;
674 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
675 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
676 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
677 extern const struct bpf_func_proto bpf_get_current_comm_proto;
678 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
679 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
680 extern const struct bpf_func_proto bpf_get_stackid_proto;
681 extern const struct bpf_func_proto bpf_sock_map_update_proto;
683 /* Shared helpers among cBPF and eBPF. */
684 void bpf_user_rnd_init_once(void);
685 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
687 #endif /* _LINUX_BPF_H */