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>
23 /* map is generic key/value storage optionally accesible by eBPF programs */
25 /* funcs callable from userspace (via syscall) */
26 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
27 void (*map_release)(struct bpf_map *map, struct file *map_file);
28 void (*map_free)(struct bpf_map *map);
29 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
30 void (*map_release_uref)(struct bpf_map *map);
31 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *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);
47 /* 1st cacheline with read-mostly members of which some
48 * are also accessed in fast-path (e.g. ops, max_entries).
50 const struct bpf_map_ops *ops ____cacheline_aligned;
51 struct bpf_map *inner_map_meta;
52 #ifdef CONFIG_SECURITY
55 enum bpf_map_type map_type;
66 /* 2nd cacheline with misc members to avoid false sharing
67 * particularly with refcounting.
69 struct user_struct *user ____cacheline_aligned;
72 struct work_struct work;
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_UNINIT_MEM, /* pointer to memory does not need to be initialized,
91 * helper function must fill all bytes or clear
95 ARG_CONST_SIZE, /* number of bytes accessed from memory */
96 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
98 ARG_PTR_TO_CTX, /* pointer to context */
99 ARG_ANYTHING, /* any (initialized) argument is ok */
102 /* type of values returned from helper functions */
103 enum bpf_return_type {
104 RET_INTEGER, /* function returns integer */
105 RET_VOID, /* function doesn't return anything */
106 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
109 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
110 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
111 * instructions after verifying
113 struct bpf_func_proto {
114 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
117 enum bpf_return_type ret_type;
118 enum bpf_arg_type arg1_type;
119 enum bpf_arg_type arg2_type;
120 enum bpf_arg_type arg3_type;
121 enum bpf_arg_type arg4_type;
122 enum bpf_arg_type arg5_type;
125 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
126 * the first argument to eBPF programs.
127 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
131 enum bpf_access_type {
136 /* types of values stored in eBPF registers */
137 /* Pointer types represent:
140 * pointer + (u16) var
141 * pointer + (u16) var + imm
142 * if (range > 0) then [ptr, ptr + range - off) is safe to access
143 * if (id > 0) means that some 'var' was added
144 * if (off > 0) means that 'imm' was added
147 NOT_INIT = 0, /* nothing was written into register */
148 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
149 PTR_TO_CTX, /* reg points to bpf_context */
150 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
151 PTR_TO_MAP_VALUE, /* reg points to map element value */
152 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
153 PTR_TO_STACK, /* reg == frame_pointer + offset */
154 PTR_TO_PACKET, /* reg points to skb->data */
155 PTR_TO_PACKET_END, /* skb->data + headlen */
158 /* The information passed from prog-specific *_is_valid_access
159 * back to the verifier.
161 struct bpf_insn_access_aux {
162 enum bpf_reg_type reg_type;
167 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
169 aux->ctx_field_size = size;
172 struct bpf_verifier_ops {
173 /* return eBPF function prototype for verification */
174 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
176 /* return true if 'size' wide access at offset 'off' within bpf_context
177 * with 'type' (read or write) is allowed
179 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
180 struct bpf_insn_access_aux *info);
181 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
182 const struct bpf_prog *prog);
183 u32 (*convert_ctx_access)(enum bpf_access_type type,
184 const struct bpf_insn *src,
185 struct bpf_insn *dst,
186 struct bpf_prog *prog, u32 *target_size);
187 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
188 union bpf_attr __user *uattr);
191 struct bpf_prog_aux {
197 struct latch_tree_node ksym_tnode;
198 struct list_head ksym_lnode;
199 const struct bpf_verifier_ops *ops;
200 struct bpf_map **used_maps;
201 struct bpf_prog *prog;
202 struct user_struct *user;
204 struct work_struct work;
213 /* 'ownership' of prog_array is claimed by the first program that
214 * is going to use this map or by the first program which FD is stored
215 * in the map to make sure that all callers and callees have the same
216 * prog_type and JITed flag
218 enum bpf_prog_type owner_prog_type;
221 char value[0] __aligned(8);
222 void *ptrs[0] __aligned(8);
223 void __percpu *pptrs[0] __aligned(8);
227 #define MAX_TAIL_CALL_CNT 32
229 struct bpf_event_entry {
230 struct perf_event *event;
231 struct file *perf_file;
232 struct file *map_file;
236 u64 bpf_tail_call(u64 ctx, u64 r2, u64 index, u64 r4, u64 r5);
237 u64 bpf_get_stackid(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
239 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
240 int bpf_prog_calc_tag(struct bpf_prog *fp);
242 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
244 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
245 unsigned long off, unsigned long len);
247 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
248 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
250 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
251 union bpf_attr __user *uattr);
252 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
253 union bpf_attr __user *uattr);
255 #ifdef CONFIG_BPF_SYSCALL
256 DECLARE_PER_CPU(int, bpf_prog_active);
258 #define BPF_PROG_TYPE(_id, _ops) \
259 extern const struct bpf_verifier_ops _ops;
260 #define BPF_MAP_TYPE(_id, _ops) \
261 extern const struct bpf_map_ops _ops;
262 #include <linux/bpf_types.h>
266 struct bpf_prog *bpf_prog_get(u32 ufd);
267 struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type);
268 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
269 void bpf_prog_sub(struct bpf_prog *prog, int i);
270 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
271 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
272 void bpf_prog_put(struct bpf_prog *prog);
273 int __bpf_prog_charge(struct user_struct *user, u32 pages);
274 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
276 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
277 struct bpf_map *__bpf_map_get(struct fd f);
278 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
279 void bpf_map_put_with_uref(struct bpf_map *map);
280 void bpf_map_put(struct bpf_map *map);
281 int bpf_map_precharge_memlock(u32 pages);
282 void *bpf_map_area_alloc(size_t size, int numa_node);
283 void bpf_map_area_free(void *base);
285 extern int sysctl_unprivileged_bpf_disabled;
287 int bpf_map_new_fd(struct bpf_map *map);
288 int bpf_prog_new_fd(struct bpf_prog *prog);
290 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
291 int bpf_obj_get_user(const char __user *pathname);
293 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
294 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
295 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
297 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
300 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
302 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
303 void *key, void *value, u64 map_flags);
304 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
305 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
306 void *key, void *value, u64 map_flags);
307 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
309 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
310 * forced to use 'long' read/writes to try to atomically copy long counters.
311 * Best-effort only. No barriers here, since it _will_ race with concurrent
312 * updates from BPF programs. Called from bpf syscall and mostly used with
313 * size 8 or 16 bytes, so ask compiler to inline it.
315 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
317 const long *lsrc = src;
320 size /= sizeof(long);
325 /* verify correctness of eBPF program */
326 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
329 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
330 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
331 void __dev_map_flush(struct bpf_map *map);
333 /* Return map's numa specified by userspace */
334 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
336 return (attr->map_flags & BPF_F_NUMA_NODE) ?
337 attr->numa_node : NUMA_NO_NODE;
340 static inline bool unprivileged_ebpf_enabled(void)
342 return !sysctl_unprivileged_bpf_disabled;
346 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
348 return ERR_PTR(-EOPNOTSUPP);
351 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
352 enum bpf_prog_type type)
354 return ERR_PTR(-EOPNOTSUPP);
356 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
359 return ERR_PTR(-EOPNOTSUPP);
362 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
366 static inline void bpf_prog_put(struct bpf_prog *prog)
370 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
372 return ERR_PTR(-EOPNOTSUPP);
375 static inline struct bpf_prog *__must_check
376 bpf_prog_inc_not_zero(struct bpf_prog *prog)
378 return ERR_PTR(-EOPNOTSUPP);
381 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
386 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
390 static inline int bpf_obj_get_user(const char __user *pathname)
395 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
401 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
405 static inline void __dev_map_flush(struct bpf_map *map)
409 static inline bool unprivileged_ebpf_enabled(void)
414 #endif /* CONFIG_BPF_SYSCALL */
416 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
417 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
418 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
420 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
425 static inline int sock_map_prog(struct bpf_map *map,
426 struct bpf_prog *prog,
433 /* verifier prototypes for helper functions called from eBPF programs */
434 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
435 extern const struct bpf_func_proto bpf_map_update_elem_proto;
436 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
438 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
439 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
440 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
441 extern const struct bpf_func_proto bpf_tail_call_proto;
442 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
443 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
444 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
445 extern const struct bpf_func_proto bpf_get_current_comm_proto;
446 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
447 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
448 extern const struct bpf_func_proto bpf_get_stackid_proto;
449 extern const struct bpf_func_proto bpf_sock_map_update_proto;
451 /* Shared helpers among cBPF and eBPF. */
452 void bpf_user_rnd_init_once(void);
453 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
455 #endif /* _LINUX_BPF_H */