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;
28 /* map is generic key/value storage optionally accesible by eBPF programs */
30 /* funcs callable from userspace (via syscall) */
31 int (*map_alloc_check)(union bpf_attr *attr);
32 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
33 void (*map_release)(struct bpf_map *map, struct file *map_file);
34 void (*map_free)(struct bpf_map *map);
35 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
36 void (*map_release_uref)(struct bpf_map *map);
38 /* funcs callable from userspace and from eBPF programs */
39 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
40 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
41 int (*map_delete_elem)(struct bpf_map *map, void *key);
43 /* funcs called by prog_array and perf_event_array map */
44 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
46 void (*map_fd_put_ptr)(void *ptr);
47 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
48 u32 (*map_fd_sys_lookup_elem)(void *ptr);
49 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
51 int (*map_check_btf)(const struct bpf_map *map, const struct btf *btf,
52 u32 key_type_id, u32 value_type_id);
56 /* The first two cachelines with read-mostly members of which some
57 * are also accessed in fast-path (e.g. ops, max_entries).
59 const struct bpf_map_ops *ops ____cacheline_aligned;
60 struct bpf_map *inner_map_meta;
61 #ifdef CONFIG_SECURITY
64 enum bpf_map_type map_type;
78 /* The 3rd and 4th cacheline with misc members to avoid false sharing
79 * particularly with refcounting.
81 struct user_struct *user ____cacheline_aligned;
84 struct work_struct work;
85 char name[BPF_OBJ_NAME_LEN];
88 struct bpf_offloaded_map;
90 struct bpf_map_dev_ops {
91 int (*map_get_next_key)(struct bpf_offloaded_map *map,
92 void *key, void *next_key);
93 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
94 void *key, void *value);
95 int (*map_update_elem)(struct bpf_offloaded_map *map,
96 void *key, void *value, u64 flags);
97 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
100 struct bpf_offloaded_map {
102 struct net_device *netdev;
103 const struct bpf_map_dev_ops *dev_ops;
105 struct list_head offloads;
108 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
110 return container_of(map, struct bpf_offloaded_map, map);
113 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
115 return map->ops->map_seq_show_elem && map->ops->map_check_btf;
118 extern const struct bpf_map_ops bpf_map_offload_ops;
120 /* function argument constraints */
122 ARG_DONTCARE = 0, /* unused argument in helper function */
124 /* the following constraints used to prototype
125 * bpf_map_lookup/update/delete_elem() functions
127 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
128 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
129 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
131 /* the following constraints used to prototype bpf_memcmp() and other
132 * functions that access data on eBPF program stack
134 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
135 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
136 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
137 * helper function must fill all bytes or clear
138 * them in error case.
141 ARG_CONST_SIZE, /* number of bytes accessed from memory */
142 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
144 ARG_PTR_TO_CTX, /* pointer to context */
145 ARG_ANYTHING, /* any (initialized) argument is ok */
148 /* type of values returned from helper functions */
149 enum bpf_return_type {
150 RET_INTEGER, /* function returns integer */
151 RET_VOID, /* function doesn't return anything */
152 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
155 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
156 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
157 * instructions after verifying
159 struct bpf_func_proto {
160 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
163 enum bpf_return_type ret_type;
164 enum bpf_arg_type arg1_type;
165 enum bpf_arg_type arg2_type;
166 enum bpf_arg_type arg3_type;
167 enum bpf_arg_type arg4_type;
168 enum bpf_arg_type arg5_type;
171 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
172 * the first argument to eBPF programs.
173 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
177 enum bpf_access_type {
182 /* types of values stored in eBPF registers */
183 /* Pointer types represent:
186 * pointer + (u16) var
187 * pointer + (u16) var + imm
188 * if (range > 0) then [ptr, ptr + range - off) is safe to access
189 * if (id > 0) means that some 'var' was added
190 * if (off > 0) means that 'imm' was added
193 NOT_INIT = 0, /* nothing was written into register */
194 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
195 PTR_TO_CTX, /* reg points to bpf_context */
196 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
197 PTR_TO_MAP_VALUE, /* reg points to map element value */
198 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
199 PTR_TO_STACK, /* reg == frame_pointer + offset */
200 PTR_TO_PACKET_META, /* skb->data - meta_len */
201 PTR_TO_PACKET, /* reg points to skb->data */
202 PTR_TO_PACKET_END, /* skb->data + headlen */
205 /* The information passed from prog-specific *_is_valid_access
206 * back to the verifier.
208 struct bpf_insn_access_aux {
209 enum bpf_reg_type reg_type;
214 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
216 aux->ctx_field_size = size;
219 struct bpf_prog_ops {
220 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
221 union bpf_attr __user *uattr);
224 struct bpf_verifier_ops {
225 /* return eBPF function prototype for verification */
226 const struct bpf_func_proto *
227 (*get_func_proto)(enum bpf_func_id func_id,
228 const struct bpf_prog *prog);
230 /* return true if 'size' wide access at offset 'off' within bpf_context
231 * with 'type' (read or write) is allowed
233 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
234 const struct bpf_prog *prog,
235 struct bpf_insn_access_aux *info);
236 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
237 const struct bpf_prog *prog);
238 u32 (*convert_ctx_access)(enum bpf_access_type type,
239 const struct bpf_insn *src,
240 struct bpf_insn *dst,
241 struct bpf_prog *prog, u32 *target_size);
244 struct bpf_prog_offload_ops {
245 int (*insn_hook)(struct bpf_verifier_env *env,
246 int insn_idx, int prev_insn_idx);
249 struct bpf_prog_offload {
250 struct bpf_prog *prog;
251 struct net_device *netdev;
253 struct list_head offloads;
255 const struct bpf_prog_offload_ops *dev_ops;
260 struct bpf_prog_aux {
267 bool offload_requested;
268 struct bpf_prog **func;
269 void *jit_data; /* JIT specific data. arch dependent */
270 struct latch_tree_node ksym_tnode;
271 struct list_head ksym_lnode;
272 const struct bpf_prog_ops *ops;
273 struct bpf_map **used_maps;
274 struct bpf_prog *prog;
275 struct user_struct *user;
276 u64 load_time; /* ns since boottime */
277 char name[BPF_OBJ_NAME_LEN];
278 #ifdef CONFIG_SECURITY
281 struct bpf_prog_offload *offload;
283 struct work_struct work;
292 /* 'ownership' of prog_array is claimed by the first program that
293 * is going to use this map or by the first program which FD is stored
294 * in the map to make sure that all callers and callees have the same
295 * prog_type and JITed flag
297 enum bpf_prog_type owner_prog_type;
300 char value[0] __aligned(8);
301 void *ptrs[0] __aligned(8);
302 void __percpu *pptrs[0] __aligned(8);
306 #define MAX_TAIL_CALL_CNT 32
308 struct bpf_event_entry {
309 struct perf_event *event;
310 struct file *perf_file;
311 struct file *map_file;
315 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
316 int bpf_prog_calc_tag(struct bpf_prog *fp);
318 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
320 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
321 unsigned long off, unsigned long len);
323 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
324 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
326 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
327 union bpf_attr __user *uattr);
328 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
329 union bpf_attr __user *uattr);
331 /* an array of programs to be executed under rcu_lock.
334 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
336 * the structure returned by bpf_prog_array_alloc() should be populated
337 * with program pointers and the last pointer must be NULL.
338 * The user has to keep refcnt on the program and make sure the program
339 * is removed from the array before bpf_prog_put().
340 * The 'struct bpf_prog_array *' should only be replaced with xchg()
341 * since other cpus are walking the array of pointers in parallel.
343 struct bpf_prog_array {
345 struct bpf_prog *progs[0];
348 struct bpf_prog_array __rcu *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
349 void bpf_prog_array_free(struct bpf_prog_array __rcu *progs);
350 int bpf_prog_array_length(struct bpf_prog_array __rcu *progs);
351 int bpf_prog_array_copy_to_user(struct bpf_prog_array __rcu *progs,
352 __u32 __user *prog_ids, u32 cnt);
354 void bpf_prog_array_delete_safe(struct bpf_prog_array __rcu *progs,
355 struct bpf_prog *old_prog);
356 int bpf_prog_array_copy_info(struct bpf_prog_array __rcu *array,
357 u32 *prog_ids, u32 request_cnt,
359 int bpf_prog_array_copy(struct bpf_prog_array __rcu *old_array,
360 struct bpf_prog *exclude_prog,
361 struct bpf_prog *include_prog,
362 struct bpf_prog_array **new_array);
364 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null) \
366 struct bpf_prog **_prog, *__prog; \
367 struct bpf_prog_array *_array; \
371 _array = rcu_dereference(array); \
372 if (unlikely(check_non_null && !_array))\
374 _prog = _array->progs; \
375 while ((__prog = READ_ONCE(*_prog))) { \
376 _ret &= func(__prog, ctx); \
381 preempt_enable_no_resched(); \
385 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
386 __BPF_PROG_RUN_ARRAY(array, ctx, func, false)
388 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
389 __BPF_PROG_RUN_ARRAY(array, ctx, func, true)
391 #ifdef CONFIG_BPF_SYSCALL
392 DECLARE_PER_CPU(int, bpf_prog_active);
394 extern const struct file_operations bpf_map_fops;
395 extern const struct file_operations bpf_prog_fops;
397 #define BPF_PROG_TYPE(_id, _name) \
398 extern const struct bpf_prog_ops _name ## _prog_ops; \
399 extern const struct bpf_verifier_ops _name ## _verifier_ops;
400 #define BPF_MAP_TYPE(_id, _ops) \
401 extern const struct bpf_map_ops _ops;
402 #include <linux/bpf_types.h>
406 extern const struct bpf_prog_ops bpf_offload_prog_ops;
407 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
408 extern const struct bpf_verifier_ops xdp_analyzer_ops;
410 struct bpf_prog *bpf_prog_get(u32 ufd);
411 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
413 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
414 void bpf_prog_sub(struct bpf_prog *prog, int i);
415 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
416 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
417 void bpf_prog_put(struct bpf_prog *prog);
418 int __bpf_prog_charge(struct user_struct *user, u32 pages);
419 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
421 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
422 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
424 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
425 struct bpf_map *__bpf_map_get(struct fd f);
426 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
427 void bpf_map_put_with_uref(struct bpf_map *map);
428 void bpf_map_put(struct bpf_map *map);
429 int bpf_map_precharge_memlock(u32 pages);
430 void *bpf_map_area_alloc(size_t size, int numa_node);
431 void bpf_map_area_free(void *base);
432 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
434 extern int sysctl_unprivileged_bpf_disabled;
436 int bpf_map_new_fd(struct bpf_map *map, int flags);
437 int bpf_prog_new_fd(struct bpf_prog *prog);
439 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
440 int bpf_obj_get_user(const char __user *pathname, int flags);
442 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
443 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
444 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
446 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
449 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
451 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
452 void *key, void *value, u64 map_flags);
453 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
454 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
455 void *key, void *value, u64 map_flags);
456 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
458 int bpf_get_file_flag(int flags);
460 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
461 * forced to use 'long' read/writes to try to atomically copy long counters.
462 * Best-effort only. No barriers here, since it _will_ race with concurrent
463 * updates from BPF programs. Called from bpf syscall and mostly used with
464 * size 8 or 16 bytes, so ask compiler to inline it.
466 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
468 const long *lsrc = src;
471 size /= sizeof(long);
476 /* verify correctness of eBPF program */
477 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
478 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
481 struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
482 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
483 void __dev_map_flush(struct bpf_map *map);
485 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
486 void __cpu_map_insert_ctx(struct bpf_map *map, u32 index);
487 void __cpu_map_flush(struct bpf_map *map);
489 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
490 struct net_device *dev_rx);
492 /* Return map's numa specified by userspace */
493 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
495 return (attr->map_flags & BPF_F_NUMA_NODE) ?
496 attr->numa_node : NUMA_NO_NODE;
499 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
501 #else /* !CONFIG_BPF_SYSCALL */
502 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
504 return ERR_PTR(-EOPNOTSUPP);
507 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
508 enum bpf_prog_type type,
511 return ERR_PTR(-EOPNOTSUPP);
514 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
517 return ERR_PTR(-EOPNOTSUPP);
520 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
524 static inline void bpf_prog_put(struct bpf_prog *prog)
528 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
530 return ERR_PTR(-EOPNOTSUPP);
533 static inline struct bpf_prog *__must_check
534 bpf_prog_inc_not_zero(struct bpf_prog *prog)
536 return ERR_PTR(-EOPNOTSUPP);
539 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
544 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
548 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
553 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map,
559 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
563 static inline void __dev_map_flush(struct bpf_map *map)
568 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
573 static inline void __cpu_map_insert_ctx(struct bpf_map *map, u32 index)
577 static inline void __cpu_map_flush(struct bpf_map *map)
582 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
583 struct xdp_buff *xdp,
584 struct net_device *dev_rx)
589 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
590 enum bpf_prog_type type)
592 return ERR_PTR(-EOPNOTSUPP);
594 #endif /* CONFIG_BPF_SYSCALL */
596 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
597 enum bpf_prog_type type)
599 return bpf_prog_get_type_dev(ufd, type, false);
602 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
604 int bpf_prog_offload_compile(struct bpf_prog *prog);
605 void bpf_prog_offload_destroy(struct bpf_prog *prog);
606 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
607 struct bpf_prog *prog);
609 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
611 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
612 int bpf_map_offload_update_elem(struct bpf_map *map,
613 void *key, void *value, u64 flags);
614 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
615 int bpf_map_offload_get_next_key(struct bpf_map *map,
616 void *key, void *next_key);
618 bool bpf_offload_dev_match(struct bpf_prog *prog, struct bpf_map *map);
620 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
621 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
623 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
625 return aux->offload_requested;
628 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
630 return unlikely(map->ops == &bpf_map_offload_ops);
633 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
634 void bpf_map_offload_map_free(struct bpf_map *map);
636 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
637 union bpf_attr *attr)
642 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
647 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
652 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
654 return ERR_PTR(-EOPNOTSUPP);
657 static inline void bpf_map_offload_map_free(struct bpf_map *map)
660 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
662 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL) && defined(CONFIG_INET)
663 struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
664 int sock_map_prog(struct bpf_map *map, struct bpf_prog *prog, u32 type);
666 static inline struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
671 static inline int sock_map_prog(struct bpf_map *map,
672 struct bpf_prog *prog,
679 /* verifier prototypes for helper functions called from eBPF programs */
680 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
681 extern const struct bpf_func_proto bpf_map_update_elem_proto;
682 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
684 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
685 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
686 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
687 extern const struct bpf_func_proto bpf_tail_call_proto;
688 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
689 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
690 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
691 extern const struct bpf_func_proto bpf_get_current_comm_proto;
692 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
693 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
694 extern const struct bpf_func_proto bpf_get_stackid_proto;
695 extern const struct bpf_func_proto bpf_get_stack_proto;
696 extern const struct bpf_func_proto bpf_sock_map_update_proto;
698 /* Shared helpers among cBPF and eBPF. */
699 void bpf_user_rnd_init_once(void);
700 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
702 #endif /* _LINUX_BPF_H */