]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - include/linux/bpf.h
bpf: add helper inlining infra and optimize map_array lookup
[mirror_ubuntu-jammy-kernel.git] / include / linux / bpf.h
CommitLineData
99c55f7d
AS
1/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
2 *
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.
6 */
7#ifndef _LINUX_BPF_H
8#define _LINUX_BPF_H 1
9
10#include <uapi/linux/bpf.h>
74451e66 11
99c55f7d 12#include <linux/workqueue.h>
db20fd2b 13#include <linux/file.h>
b121d1e7 14#include <linux/percpu.h>
002245cc 15#include <linux/err.h>
74451e66 16#include <linux/rbtree_latch.h>
99c55f7d 17
3b1efb19 18struct perf_event;
99c55f7d
AS
19struct bpf_map;
20
21/* map is generic key/value storage optionally accesible by eBPF programs */
22struct bpf_map_ops {
23 /* funcs callable from userspace (via syscall) */
24 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
61d1b6a4
DB
25 void (*map_release)(struct bpf_map *map, struct file *map_file);
26 void (*map_free)(struct bpf_map *map);
db20fd2b
AS
27 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
28
29 /* funcs callable from userspace and from eBPF programs */
30 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
3274f520 31 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
db20fd2b 32 int (*map_delete_elem)(struct bpf_map *map, void *key);
2a36f0b9
WN
33
34 /* funcs called by prog_array and perf_event_array map */
d056a788
DB
35 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
36 int fd);
37 void (*map_fd_put_ptr)(void *ptr);
81ed18ab 38 u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
99c55f7d
AS
39};
40
41struct bpf_map {
42 atomic_t refcnt;
43 enum bpf_map_type map_type;
44 u32 key_size;
45 u32 value_size;
46 u32 max_entries;
6c905981 47 u32 map_flags;
aaac3ba9
AS
48 u32 pages;
49 struct user_struct *user;
a2c83fff 50 const struct bpf_map_ops *ops;
99c55f7d 51 struct work_struct work;
c9da161c 52 atomic_t usercnt;
99c55f7d
AS
53};
54
55struct bpf_map_type_list {
56 struct list_head list_node;
a2c83fff 57 const struct bpf_map_ops *ops;
99c55f7d
AS
58 enum bpf_map_type type;
59};
60
17a52670
AS
61/* function argument constraints */
62enum bpf_arg_type {
80f1d68c 63 ARG_DONTCARE = 0, /* unused argument in helper function */
17a52670
AS
64
65 /* the following constraints used to prototype
66 * bpf_map_lookup/update/delete_elem() functions
67 */
68 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
69 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
70 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
71
72 /* the following constraints used to prototype bpf_memcmp() and other
73 * functions that access data on eBPF program stack
74 */
39f19ebb
AS
75 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
76 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
77 * helper function must fill all bytes or clear
78 * them in error case.
435faee1
DB
79 */
80
39f19ebb
AS
81 ARG_CONST_SIZE, /* number of bytes accessed from memory */
82 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
80f1d68c 83
608cd71a 84 ARG_PTR_TO_CTX, /* pointer to context */
80f1d68c 85 ARG_ANYTHING, /* any (initialized) argument is ok */
17a52670
AS
86};
87
88/* type of values returned from helper functions */
89enum bpf_return_type {
90 RET_INTEGER, /* function returns integer */
91 RET_VOID, /* function doesn't return anything */
92 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
93};
94
09756af4
AS
95/* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
96 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
97 * instructions after verifying
98 */
99struct bpf_func_proto {
100 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
101 bool gpl_only;
36bbef52 102 bool pkt_access;
17a52670
AS
103 enum bpf_return_type ret_type;
104 enum bpf_arg_type arg1_type;
105 enum bpf_arg_type arg2_type;
106 enum bpf_arg_type arg3_type;
107 enum bpf_arg_type arg4_type;
108 enum bpf_arg_type arg5_type;
109};
110
111/* bpf_context is intentionally undefined structure. Pointer to bpf_context is
112 * the first argument to eBPF programs.
113 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
114 */
115struct bpf_context;
116
117enum bpf_access_type {
118 BPF_READ = 1,
119 BPF_WRITE = 2
09756af4
AS
120};
121
19de99f7
AS
122/* types of values stored in eBPF registers */
123enum bpf_reg_type {
124 NOT_INIT = 0, /* nothing was written into register */
125 UNKNOWN_VALUE, /* reg doesn't contain a valid pointer */
126 PTR_TO_CTX, /* reg points to bpf_context */
127 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
128 PTR_TO_MAP_VALUE, /* reg points to map element value */
129 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
130 FRAME_PTR, /* reg == frame_pointer */
131 PTR_TO_STACK, /* reg == frame_pointer + imm */
132 CONST_IMM, /* constant integer value */
133
134 /* PTR_TO_PACKET represents:
135 * skb->data
136 * skb->data + imm
137 * skb->data + (u16) var
138 * skb->data + (u16) var + imm
139 * if (range > 0) then [ptr, ptr + range - off) is safe to access
140 * if (id > 0) means that some 'var' was added
141 * if (off > 0) menas that 'imm' was added
142 */
143 PTR_TO_PACKET,
144 PTR_TO_PACKET_END, /* skb->data + headlen */
48461135
JB
145
146 /* PTR_TO_MAP_VALUE_ADJ is used for doing pointer math inside of a map
147 * elem value. We only allow this if we can statically verify that
148 * access from this register are going to fall within the size of the
149 * map element.
150 */
151 PTR_TO_MAP_VALUE_ADJ,
19de99f7
AS
152};
153
ff936a04
AS
154struct bpf_prog;
155
09756af4
AS
156struct bpf_verifier_ops {
157 /* return eBPF function prototype for verification */
158 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
17a52670
AS
159
160 /* return true if 'size' wide access at offset 'off' within bpf_context
161 * with 'type' (read or write) is allowed
162 */
19de99f7
AS
163 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
164 enum bpf_reg_type *reg_type);
36bbef52
DB
165 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
166 const struct bpf_prog *prog);
6b8cc1d1
DB
167 u32 (*convert_ctx_access)(enum bpf_access_type type,
168 const struct bpf_insn *src,
169 struct bpf_insn *dst,
170 struct bpf_prog *prog);
09756af4
AS
171};
172
173struct bpf_prog_type_list {
174 struct list_head list_node;
a2c83fff 175 const struct bpf_verifier_ops *ops;
09756af4
AS
176 enum bpf_prog_type type;
177};
178
09756af4
AS
179struct bpf_prog_aux {
180 atomic_t refcnt;
24701ece 181 u32 used_map_cnt;
32bbe007 182 u32 max_ctx_offset;
74451e66
DB
183 struct latch_tree_node ksym_tnode;
184 struct list_head ksym_lnode;
a2c83fff 185 const struct bpf_verifier_ops *ops;
09756af4 186 struct bpf_map **used_maps;
09756af4 187 struct bpf_prog *prog;
aaac3ba9 188 struct user_struct *user;
abf2e7d6
AS
189 union {
190 struct work_struct work;
191 struct rcu_head rcu;
192 };
09756af4
AS
193};
194
04fd61ab
AS
195struct bpf_array {
196 struct bpf_map map;
197 u32 elem_size;
198 /* 'ownership' of prog_array is claimed by the first program that
199 * is going to use this map or by the first program which FD is stored
200 * in the map to make sure that all callers and callees have the same
201 * prog_type and JITed flag
202 */
203 enum bpf_prog_type owner_prog_type;
204 bool owner_jited;
205 union {
206 char value[0] __aligned(8);
2a36f0b9 207 void *ptrs[0] __aligned(8);
a10423b8 208 void __percpu *pptrs[0] __aligned(8);
04fd61ab
AS
209 };
210};
3b1efb19 211
04fd61ab
AS
212#define MAX_TAIL_CALL_CNT 32
213
3b1efb19
DB
214struct bpf_event_entry {
215 struct perf_event *event;
216 struct file *perf_file;
217 struct file *map_file;
218 struct rcu_head rcu;
219};
220
04fd61ab 221u64 bpf_tail_call(u64 ctx, u64 r2, u64 index, u64 r4, u64 r5);
9940d67c 222u64 bpf_get_stackid(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
d056a788 223
04fd61ab 224bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
f1f7714e 225int bpf_prog_calc_tag(struct bpf_prog *fp);
bd570ff9 226
0756ea3e 227const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
555c8a86
DB
228
229typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
aa7145c1 230 unsigned long off, unsigned long len);
555c8a86
DB
231
232u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
233 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
04fd61ab 234
89aa0758 235#ifdef CONFIG_BPF_SYSCALL
b121d1e7
AS
236DECLARE_PER_CPU(int, bpf_prog_active);
237
0fc174de 238void bpf_register_prog_type(struct bpf_prog_type_list *tl);
61e021f3 239void bpf_register_map_type(struct bpf_map_type_list *tl);
0fc174de 240
0fc174de 241struct bpf_prog *bpf_prog_get(u32 ufd);
113214be 242struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type);
6d67942d 243struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
c540594f 244void bpf_prog_sub(struct bpf_prog *prog, int i);
6d67942d 245struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
61e021f3 246void bpf_prog_put(struct bpf_prog *prog);
5ccb071e
DB
247int __bpf_prog_charge(struct user_struct *user, u32 pages);
248void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
61e021f3 249
c9da161c 250struct bpf_map *bpf_map_get_with_uref(u32 ufd);
c2101297 251struct bpf_map *__bpf_map_get(struct fd f);
6d67942d 252struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
c9da161c 253void bpf_map_put_with_uref(struct bpf_map *map);
61e021f3 254void bpf_map_put(struct bpf_map *map);
6c905981 255int bpf_map_precharge_memlock(u32 pages);
d407bd25
DB
256void *bpf_map_area_alloc(size_t size);
257void bpf_map_area_free(void *base);
61e021f3 258
1be7f75d
AS
259extern int sysctl_unprivileged_bpf_disabled;
260
b2197755
DB
261int bpf_map_new_fd(struct bpf_map *map);
262int bpf_prog_new_fd(struct bpf_prog *prog);
263
264int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
265int bpf_obj_get_user(const char __user *pathname);
266
15a07b33
AS
267int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
268int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
269int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
270 u64 flags);
271int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
272 u64 flags);
d056a788 273
557c0c6e 274int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
15a07b33 275
d056a788
DB
276int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
277 void *key, void *value, u64 map_flags);
278void bpf_fd_array_map_clear(struct bpf_map *map);
279
15a07b33
AS
280/* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
281 * forced to use 'long' read/writes to try to atomically copy long counters.
282 * Best-effort only. No barriers here, since it _will_ race with concurrent
283 * updates from BPF programs. Called from bpf syscall and mostly used with
284 * size 8 or 16 bytes, so ask compiler to inline it.
285 */
286static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
287{
288 const long *lsrc = src;
289 long *ldst = dst;
290
291 size /= sizeof(long);
292 while (size--)
293 *ldst++ = *lsrc++;
294}
295
61e021f3 296/* verify correctness of eBPF program */
9bac3d6d 297int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
89aa0758 298#else
0fc174de
DB
299static inline void bpf_register_prog_type(struct bpf_prog_type_list *tl)
300{
301}
302
303static inline struct bpf_prog *bpf_prog_get(u32 ufd)
304{
305 return ERR_PTR(-EOPNOTSUPP);
306}
307
113214be
DB
308static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
309 enum bpf_prog_type type)
310{
311 return ERR_PTR(-EOPNOTSUPP);
312}
6d67942d
DB
313static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
314 int i)
cc2e0b3f
BB
315{
316 return ERR_PTR(-EOPNOTSUPP);
317}
113214be 318
c540594f
DB
319static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
320{
321}
322
0fc174de
DB
323static inline void bpf_prog_put(struct bpf_prog *prog)
324{
325}
6d67942d
DB
326
327static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
aa6a5f3c
AS
328{
329 return ERR_PTR(-EOPNOTSUPP);
330}
5ccb071e
DB
331
332static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
333{
334 return 0;
335}
336
337static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
338{
339}
61e021f3 340#endif /* CONFIG_BPF_SYSCALL */
09756af4 341
d0003ec0 342/* verifier prototypes for helper functions called from eBPF programs */
a2c83fff
DB
343extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
344extern const struct bpf_func_proto bpf_map_update_elem_proto;
345extern const struct bpf_func_proto bpf_map_delete_elem_proto;
d0003ec0 346
03e69b50 347extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
c04167ce 348extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
2d0e30c3 349extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
04fd61ab 350extern const struct bpf_func_proto bpf_tail_call_proto;
17ca8cbf 351extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
ffeedafb
AS
352extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
353extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
354extern const struct bpf_func_proto bpf_get_current_comm_proto;
4e10df9a
AS
355extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
356extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
d5a3b1f6 357extern const struct bpf_func_proto bpf_get_stackid_proto;
03e69b50 358
3ad00405
DB
359/* Shared helpers among cBPF and eBPF. */
360void bpf_user_rnd_init_once(void);
361u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
362
99c55f7d 363#endif /* _LINUX_BPF_H */