1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
3 */
4 #ifndef _LINUX_BPF_H
5 #define _LINUX_BPF_H 1
6
7 #include <uapi/linux/bpf.h>
8
9 #include <linux/workqueue.h>
10 #include <linux/file.h>
11 #include <linux/percpu.h>
12 #include <linux/err.h>
13 #include <linux/rbtree_latch.h>
14 #include <linux/numa.h>
15 #include <linux/mm_types.h>
16 #include <linux/wait.h>
17 #include <linux/u64_stats_sync.h>
18 #include <linux/refcount.h>
19 #include <linux/mutex.h>
20 #include <linux/module.h>
21 #include <linux/kallsyms.h>
22 #include <linux/capability.h>
23 #include <linux/percpu-refcount.h>
24
25 #define BPF_TWO 2
26
27 struct bpf_verifier_env;
28 struct bpf_verifier_log;
29 struct perf_event;
30 struct bpf_prog;
31 struct bpf_prog_aux;
32 struct bpf_map;
33 struct sock;
34 struct seq_file;
35 struct btf;
36 struct btf_type;
37 struct exception_table_entry;
38 struct seq_operations;
39 struct bpf_iter_aux_info;
40 struct bpf_local_storage;
41 struct bpf_local_storage_map;
42
43 extern struct idr btf_idr;
44 extern spinlock_t btf_idr_lock;
45
46 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data, struct bpf_iter_aux_info *aux);
47 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
48 struct bpf_iter_seq_info {
49 const struct seq_operations *seq_ops;
50 bpf_iter_init_seq_priv_t init_seq_private;
51 bpf_iter_fini_seq_priv_t fini_seq_private;
52 u32 seq_priv_size;
53 };
54
55 /* map is generic key/value storage optionally accesible by eBPF programs */
56 struct bpf_map_ops {
57 /* funcs callable from userspace (via syscall) */
58 int (*map_alloc_check)(union bpf_attr *attr);
59 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
60 void (*map_release)(struct bpf_map *map, struct file *map_file);
61 void (*map_free)(struct bpf_map *map);
62 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
63 void (*map_release_uref)(struct bpf_map *map);
64 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
65 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr, union bpf_attr __user *uattr);
66 int (*map_lookup_and_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, union bpf_attr __user *uattr);
67 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr, union bpf_attr __user *uattr);
68 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr, union bpf_attr __user *uattr);
69
70 /* funcs callable from userspace and from eBPF programs */
71 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
72 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
73 int (*map_delete_elem)(struct bpf_map *map, void *key);
74 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
75 int (*map_pop_elem)(struct bpf_map *map, void *value);
76 int (*map_peek_elem)(struct bpf_map *map, void *value);
77
78 /* funcs called by prog_array and perf_event_array map */
79 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file, int fd);
80 void (*map_fd_put_ptr)(void *ptr);
81 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
82 u32 (*map_fd_sys_lookup_elem)(void *ptr);
83 void (*map_seq_show_elem)(struct bpf_map *map, void *key, struct seq_file *m);
84 int (*map_check_btf)(const struct bpf_map *map, const struct btf *btf, const struct btf_type *key_type,
85 const struct btf_type *value_type);
86
87 /* Prog poke tracking helpers. */
88 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
89 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
90 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old, struct bpf_prog *new);
91
92 /* Direct value access helpers. */
93 int (*map_direct_value_addr)(const struct bpf_map *map, u64 *imm, u32 off);
94 int (*map_direct_value_meta)(const struct bpf_map *map, u64 imm, u32 *off);
95 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
96 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp, struct poll_table_struct *pts);
97
98 /* Functions called by bpf_local_storage maps */
99 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap, void *owner, u32 size);
100 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap, void *owner, u32 size);
101 struct bpf_local_storage __rcu **(*map_owner_storage_ptr)(void *owner);
102
103 /* map_meta_equal must be implemented for maps that can be
104 * used as an inner map. It is a runtime check to ensure
105 * an inner map can be inserted to an outer map.
106 *
107 * Some properties of the inner map has been used during the
108 * verification time. When inserting an inner map at the runtime,
109 * map_meta_equal has to ensure the inserting map has the same
110 * properties that the verifier has used earlier.
111 */
112 bool (*map_meta_equal)(const struct bpf_map *meta0, const struct bpf_map *meta1);
113
114 /* BTF name and id of struct allocated by map_alloc */
115 const char *const map_btf_name;
116 int *map_btf_id;
117
118 /* bpf_iter info used to open a seq_file */
119 const struct bpf_iter_seq_info *iter_seq_info;
120 };
121
122 struct bpf_map_memory {
123 u32 pages;
124 struct user_struct *user;
125 };
126
127 struct bpf_map {
128 /* The first two cachelines with read-mostly members of which some
129 * are also accessed in fast-path (e.g. ops, max_entries).
130 */
131 const struct bpf_map_ops *ops ____cacheline_aligned;
132 struct bpf_map *inner_map_meta;
133 #ifdef CONFIG_SECURITY
134 void *security;
135 #endif
136 enum bpf_map_type map_type;
137 u32 key_size;
138 u32 value_size;
139 u32 max_entries;
140 u32 map_flags;
141 int spin_lock_off; /* >=0 valid offset, <0 error */
142 u32 id;
143 int numa_node;
144 u32 btf_key_type_id;
145 u32 btf_value_type_id;
146 struct btf *btf;
147 struct bpf_map_memory memory;
148 char name[BPF_OBJ_NAME_LEN];
149 u32 btf_vmlinux_value_type_id;
150 bool bypass_spec_v1;
151 bool frozen; /* write-once; write-protected by freeze_mutex */
152 /* 22 bytes hole */
153
154 /* The 3rd and 4th cacheline with misc members to avoid false sharing
155 * particularly with refcounting.
156 */
157 atomic64_t refcnt ____cacheline_aligned;
158 atomic64_t usercnt;
159 struct work_struct work;
160 struct mutex freeze_mutex;
161 u64 writecnt; /* writable mmap cnt; protected by freeze_mutex */
162 };
163
map_value_has_spin_lock(const struct bpf_map *map)164 static inline bool map_value_has_spin_lock(const struct bpf_map *map)
165 {
166 return map->spin_lock_off >= 0;
167 }
168
check_and_init_map_lock(struct bpf_map *map, void *dst)169 static inline void check_and_init_map_lock(struct bpf_map *map, void *dst)
170 {
171 if (likely(!map_value_has_spin_lock(map))) {
172 return;
173 }
174 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) = (struct bpf_spin_lock) {};
175 }
176
177 /* copy everything but bpf_spin_lock */
copy_map_value(struct bpf_map *map, void *dst, void *src)178 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
179 {
180 if (unlikely(map_value_has_spin_lock(map))) {
181 u32 off = map->spin_lock_off;
182
183 memcpy(dst, src, off);
184 memcpy(dst + off + sizeof(struct bpf_spin_lock), src + off + sizeof(struct bpf_spin_lock),
185 map->value_size - off - sizeof(struct bpf_spin_lock));
186 } else {
187 memcpy(dst, src, map->value_size);
188 }
189 }
190 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src, bool lock_src);
191 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
192
193 struct bpf_offload_dev;
194 struct bpf_offloaded_map;
195
196 struct bpf_map_dev_ops {
197 int (*map_get_next_key)(struct bpf_offloaded_map *map, void *key, void *next_key);
198 int (*map_lookup_elem)(struct bpf_offloaded_map *map, void *key, void *value);
199 int (*map_update_elem)(struct bpf_offloaded_map *map, void *key, void *value, u64 flags);
200 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
201 };
202
203 struct bpf_offloaded_map {
204 struct bpf_map map;
205 struct net_device *netdev;
206 const struct bpf_map_dev_ops *dev_ops;
207 void *dev_priv;
208 struct list_head offloads;
209 };
210
map_to_offmap(struct bpf_map *map)211 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
212 {
213 return container_of(map, struct bpf_offloaded_map, map);
214 }
215
bpf_map_offload_neutral(const struct bpf_map *map)216 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
217 {
218 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
219 }
220
bpf_map_support_seq_show(const struct bpf_map *map)221 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
222 {
223 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) && map->ops->map_seq_show_elem;
224 }
225
226 int map_check_no_btf(const struct bpf_map *map, const struct btf *btf, const struct btf_type *key_type,
227 const struct btf_type *value_type);
228
229 bool bpf_map_meta_equal(const struct bpf_map *meta0, const struct bpf_map *meta1);
230
231 extern const struct bpf_map_ops bpf_map_offload_ops;
232
233 /* bpf_type_flag contains a set of flags that are applicable to the values of
234 * arg_type, ret_type and reg_type. For example, a pointer value may be null,
235 * or a memory is read-only. We classify types into two categories: base types
236 * and extended types. Extended types are base types combined with a type flag.
237 *
238 * Currently there are no more than 32 base types in arg_type, ret_type and
239 * reg_types.
240 */
241 #define BPF_BASE_TYPE_BITS 8
242
243 enum bpf_type_flag {
244 /* PTR may be NULL. */
245 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS),
246
247 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is
248 * compatible with both mutable and immutable memory.
249 */
250 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS),
251
252 /* MEM was "allocated" from a different helper, and cannot be mixed
253 * with regular non-MEM_ALLOC'ed MEM types.
254 */
255 MEM_ALLOC = BIT(2 + BPF_BASE_TYPE_BITS),
256
257 __BPF_TYPE_LAST_FLAG = MEM_ALLOC,
258 };
259
260 /* Max number of base types. */
261 #define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS)
262
263 /* Max number of all types. */
264 #define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1))
265
266 /* function argument constraints */
267 enum bpf_arg_type {
268 ARG_DONTCARE = 0, /* unused argument in helper function */
269
270 /* the following constraints used to prototype
271 * bpf_map_lookup/update/delete_elem() functions
272 */
273 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
274 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
275 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
276 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */
277
278 /* the following constraints used to prototype bpf_memcmp() and other
279 * functions that access data on eBPF program stack
280 */
281 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
282 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
283 * helper function must fill all bytes or clear
284 * them in error case.
285 */
286
287 ARG_CONST_SIZE, /* number of bytes accessed from memory */
288 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
289
290 ARG_PTR_TO_CTX, /* pointer to context */
291 ARG_ANYTHING, /* any (initialized) argument is ok */
292 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
293 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
294 ARG_PTR_TO_INT, /* pointer to int */
295 ARG_PTR_TO_LONG, /* pointer to long */
296 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
297 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
298 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */
299 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
300 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
301 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */
302 __BPF_ARG_TYPE_MAX,
303
304 /* Extended arg_types. */
305 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE,
306 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM,
307 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX,
308 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET,
309 ARG_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_ALLOC_MEM,
310
311 /* This must be the last entry. Its purpose is to ensure the enum is
312 * wide enough to hold the higher bits reserved for bpf_type_flag.
313 */
314 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT,
315 };
316 static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
317
318 /* type of values returned from helper functions */
319 enum bpf_return_type {
320 RET_INTEGER, /* function returns integer */
321 RET_VOID, /* function doesn't return anything */
322 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
323 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */
324 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */
325 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */
326 RET_PTR_TO_ALLOC_MEM, /* returns a pointer to dynamically allocated memory */
327 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */
328 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */
329 __BPF_RET_TYPE_MAX,
330
331 /* Extended ret_types. */
332 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE,
333 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET,
334 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK,
335 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON,
336 RET_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_ALLOC | RET_PTR_TO_ALLOC_MEM,
337 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID,
338
339 /* This must be the last entry. Its purpose is to ensure the enum is
340 * wide enough to hold the higher bits reserved for bpf_type_flag.
341 */
342 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT,
343 };
344 static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
345
346 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
347 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
348 * instructions after verifying
349 */
350 struct bpf_func_proto {
351 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
352 bool gpl_only;
353 bool pkt_access;
354 enum bpf_return_type ret_type;
355 union {
356 struct {
357 enum bpf_arg_type arg1_type;
358 enum bpf_arg_type arg2_type;
359 enum bpf_arg_type arg3_type;
360 enum bpf_arg_type arg4_type;
361 enum bpf_arg_type arg5_type;
362 };
363 enum bpf_arg_type arg_type[5];
364 };
365 union {
366 struct {
367 u32 *arg1_btf_id;
368 u32 *arg2_btf_id;
369 u32 *arg3_btf_id;
370 u32 *arg4_btf_id;
371 u32 *arg5_btf_id;
372 };
373 u32 *arg_btf_id[5];
374 };
375 int *ret_btf_id; /* return value btf_id */
376 bool (*allowed)(const struct bpf_prog *prog);
377 };
378
379 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
380 * the first argument to eBPF programs.
381 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
382 */
383 struct bpf_context;
384
385 enum bpf_access_type { BPF_READ = 1, BPF_WRITE = 2 };
386
387 /* types of values stored in eBPF registers */
388 /* Pointer types represent:
389 * pointer
390 * pointer + imm
391 * pointer + (u16) var
392 * pointer + (u16) var + imm
393 * if (range > 0) then [ptr, ptr + range - off) is safe to access
394 * if (id > 0) means that some 'var' was added
395 * if (off > 0) means that 'imm' was added
396 */
397 enum bpf_reg_type {
398 NOT_INIT = 0, /* nothing was written into register */
399 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
400 PTR_TO_CTX, /* reg points to bpf_context */
401 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
402 PTR_TO_MAP_VALUE, /* reg points to map element value */
403 PTR_TO_STACK, /* reg == frame_pointer + offset */
404 PTR_TO_PACKET_META, /* skb->data - meta_len */
405 PTR_TO_PACKET, /* reg points to skb->data */
406 PTR_TO_PACKET_END, /* skb->data + headlen */
407 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
408 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
409 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
410 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
411 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
412 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
413 /* PTR_TO_BTF_ID points to a kernel struct that does not need
414 * to be null checked by the BPF program. This does not imply the
415 * pointer is _not_ null and in practice this can easily be a null
416 * pointer when reading pointer chains. The assumption is program
417 * context will handle null pointer dereference typically via fault
418 * handling. The verifier must keep this in mind and can make no
419 * assumptions about null or non-null when doing branch analysis.
420 * Further, when passed into helpers the helpers can not, without
421 * additional context, assume the value is non-null.
422 */
423 PTR_TO_BTF_ID,
424 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
425 * been checked for null. Used primarily to inform the verifier
426 * an explicit null check is required for this struct.
427 */
428 PTR_TO_MEM, /* reg points to valid memory region */
429 PTR_TO_BUF, /* reg points to a read/write buffer */
430 PTR_TO_PERCPU_BTF_ID, /* reg points to a percpu kernel variable */
431 __BPF_REG_TYPE_MAX,
432
433 /* Extended reg_types. */
434 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE,
435 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET,
436 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON,
437 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK,
438 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID,
439
440 /* This must be the last entry. Its purpose is to ensure the enum is
441 * wide enough to hold the higher bits reserved for bpf_type_flag.
442 */
443 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT,
444 };
445 static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
446
447 /* The information passed from prog-specific *_is_valid_access
448 * back to the verifier.
449 */
450 struct bpf_insn_access_aux {
451 enum bpf_reg_type reg_type;
452 union {
453 int ctx_field_size;
454 u32 btf_id;
455 };
456 struct bpf_verifier_log *log; /* for verbose logs */
457 };
458
bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)459 static inline void bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
460 {
461 aux->ctx_field_size = size;
462 }
463
464 struct bpf_prog_ops {
465 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr);
466 };
467
468 struct bpf_verifier_ops {
469 /* return eBPF function prototype for verification */
470 const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id, const struct bpf_prog *prog);
471
472 /* return true if 'size' wide access at offset 'off' within bpf_context
473 * with 'type' (read or write) is allowed
474 */
475 bool (*is_valid_access)(int off, int size, enum bpf_access_type type, const struct bpf_prog *prog,
476 struct bpf_insn_access_aux *info);
477 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write, const struct bpf_prog *prog);
478 int (*gen_ld_abs)(const struct bpf_insn *orig, struct bpf_insn *insn_buf);
479 u32 (*convert_ctx_access)(enum bpf_access_type type, const struct bpf_insn *src, struct bpf_insn *dst,
480 struct bpf_prog *prog, u32 *target_size);
481 int (*btf_struct_access)(struct bpf_verifier_log *log, const struct btf_type *t, int off, int size,
482 enum bpf_access_type atype, u32 *next_btf_id);
483 };
484
485 struct bpf_prog_offload_ops {
486 /* verifier basic callbacks */
487 int (*insn_hook)(struct bpf_verifier_env *env, int insn_idx, int prev_insn_idx);
488 int (*finalize)(struct bpf_verifier_env *env);
489 /* verifier optimization callbacks (called after .finalize) */
490 int (*replace_insn)(struct bpf_verifier_env *env, u32 off, struct bpf_insn *insn);
491 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
492 /* program management callbacks */
493 int (*prepare)(struct bpf_prog *prog);
494 int (*translate)(struct bpf_prog *prog);
495 void (*destroy)(struct bpf_prog *prog);
496 };
497
498 struct bpf_prog_offload {
499 struct bpf_prog *prog;
500 struct net_device *netdev;
501 struct bpf_offload_dev *offdev;
502 void *dev_priv;
503 struct list_head offloads;
504 bool dev_state;
505 bool opt_failed;
506 void *jited_image;
507 u32 jited_len;
508 };
509
510 enum bpf_cgroup_storage_type { BPF_CGROUP_STORAGE_SHARED, BPF_CGROUP_STORAGE_PERCPU, __BPF_CGROUP_STORAGE_MAX };
511
512 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
513
514 /* The longest tracepoint has 12 args.
515 * See include/trace/bpf_probe.h
516 */
517 #define MAX_BPF_FUNC_ARGS 12
518
519 struct bpf_prog_stats {
520 u64 cnt;
521 u64 nsecs;
522 struct u64_stats_sync syncp;
523 } __aligned(BPF_TWO * sizeof(u64));
524
525 struct btf_func_model {
526 u8 ret_size;
527 u8 nr_args;
528 u8 arg_size[MAX_BPF_FUNC_ARGS];
529 };
530
531 /* Restore arguments before returning from trampoline to let original function
532 * continue executing. This flag is used for fentry progs when there are no
533 * fexit progs.
534 */
535 #define BPF_TRAMP_F_RESTORE_REGS BIT(0)
536 /* Call original function after fentry progs, but before fexit progs.
537 * Makes sense for fentry/fexit, normal calls and indirect calls.
538 */
539 #define BPF_TRAMP_F_CALL_ORIG BIT(1)
540 /* Skip current frame and return to parent. Makes sense for fentry/fexit
541 * programs only. Should not be used with normal calls and indirect calls.
542 */
543 #define BPF_TRAMP_F_SKIP_FRAME BIT(2)
544 /* Return the return value of fentry prog. Only used by bpf_struct_ops. */
545 #define BPF_TRAMP_F_RET_FENTRY_RET BIT(4)
546
547 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
548 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2
549 */
550 #define BPF_MAX_TRAMP_PROGS 40
551
552 struct bpf_tramp_progs {
553 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS];
554 int nr_progs;
555 };
556
557 /* Different use cases for BPF trampoline:
558 * 1. replace nop at the function entry (kprobe equivalent)
559 * flags = BPF_TRAMP_F_RESTORE_REGS
560 * fentry = a set of programs to run before returning from trampoline
561 *
562 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
563 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
564 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
565 * fentry = a set of program to run before calling original function
566 * fexit = a set of program to run after original function
567 *
568 * 3. replace direct call instruction anywhere in the function body
569 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
570 * With flags = 0
571 * fentry = a set of programs to run before returning from trampoline
572 * With flags = BPF_TRAMP_F_CALL_ORIG
573 * orig_call = original callback addr or direct function addr
574 * fentry = a set of program to run before calling original function
575 * fexit = a set of program to run after original function
576 */
577 struct bpf_tramp_image;
578 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end,
579 const struct btf_func_model *m, u32 flags, struct bpf_tramp_progs *tprogs,
580 void *orig_call);
581 /* these two functions are called from generated trampoline */
582 u64 notrace __bpf_prog_enter(void);
583 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start);
584 void notrace __bpf_prog_enter_sleepable(void);
585 void notrace __bpf_prog_exit_sleepable(void);
586 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
587 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
588
589 struct bpf_ksym {
590 unsigned long start;
591 unsigned long end;
592 char name[KSYM_NAME_LEN];
593 struct list_head lnode;
594 struct latch_tree_node tnode;
595 bool prog;
596 };
597
598 enum bpf_tramp_prog_type {
599 BPF_TRAMP_FENTRY,
600 BPF_TRAMP_FEXIT,
601 BPF_TRAMP_MODIFY_RETURN,
602 BPF_TRAMP_MAX,
603 BPF_TRAMP_REPLACE, /* more than MAX */
604 };
605
606 struct bpf_tramp_image {
607 void *image;
608 struct bpf_ksym ksym;
609 struct percpu_ref pcref;
610 void *ip_after_call;
611 void *ip_epilogue;
612 union {
613 struct rcu_head rcu;
614 struct work_struct work;
615 };
616 };
617
618 struct bpf_trampoline {
619 /* hlist for trampoline_table */
620 struct hlist_node hlist;
621 /* serializes access to fields of this trampoline */
622 struct mutex mutex;
623 refcount_t refcnt;
624 u64 key;
625 struct {
626 struct btf_func_model model;
627 void *addr;
628 bool ftrace_managed;
629 } func;
630 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
631 * program by replacing one of its functions. func.addr is the address
632 * of the function it replaced.
633 */
634 struct bpf_prog *extension_prog;
635 /* list of BPF programs using this trampoline */
636 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
637 /* Number of attached programs. A counter per kind. */
638 int progs_cnt[BPF_TRAMP_MAX];
639 /* Executable image of trampoline */
640 struct bpf_tramp_image *cur_image;
641 u64 selector;
642 };
643
644 struct bpf_attach_target_info {
645 struct btf_func_model fmodel;
646 long tgt_addr;
647 const char *tgt_name;
648 const struct btf_type *tgt_type;
649 };
650
651 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
652
653 struct bpf_dispatcher_prog {
654 struct bpf_prog *prog;
655 refcount_t users;
656 };
657
658 struct bpf_dispatcher {
659 /* dispatcher mutex */
660 struct mutex mutex;
661 void *func;
662 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
663 int num_progs;
664 void *image;
665 u32 image_off;
666 struct bpf_ksym ksym;
667 };
668
bpf_dispatcher_nop_func(const void *ctx, const struct bpf_insn *insnsi, unsigned int (*bpf_func)(const void *, const struct bpf_insn *))669 static __always_inline unsigned int bpf_dispatcher_nop_func(const void *ctx, const struct bpf_insn *insnsi,
670 unsigned int (*bpf_func)(const void *,
671 const struct bpf_insn *))
672 {
673 return bpf_func(ctx, insnsi);
674 }
675 #ifdef CONFIG_BPF_JIT
676 int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr);
677 int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr);
678 struct bpf_trampoline *bpf_trampoline_get(u64 key, struct bpf_attach_target_info *tgt_info);
679 void bpf_trampoline_put(struct bpf_trampoline *tr);
680 #define BPF_DISPATCHER_INIT(_name) \
681 { \
682 .mutex = __MUTEX_INITIALIZER(_name.mutex), .func = &_name##_func, .progs = {}, .num_progs = 0, .image = NULL, \
683 .image_off = 0, \
684 .ksym = { \
685 .name = #_name, \
686 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
687 }, \
688 }
689
690 #define DEFINE_BPF_DISPATCHER(name) \
691 noinline unsigned int bpf_dispatcher_##name##_func( \
692 const void *ctx, const struct bpf_insn *insnsi, \
693 unsigned int (*bpf_func)(const void *, const struct bpf_insn *)) \
694 { \
695 return bpf_func(ctx, insnsi); \
696 } \
697 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
698 struct bpf_dispatcher bpf_dispatcher_##name = BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
699 #define DECLARE_BPF_DISPATCHER(name) \
700 unsigned int bpf_dispatcher_##name##_func(const void *ctx, const struct bpf_insn *insnsi, \
701 unsigned int (*bpf_func)(const void *, const struct bpf_insn *)); \
702 extern struct bpf_dispatcher bpf_dispatcher_##name;
703 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
704 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
705 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, struct bpf_prog *to);
706 /* Called only from JIT-enabled code, so there's no need for stubs. */
707 void *bpf_jit_alloc_exec_page(void);
708 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym);
709 void bpf_image_ksym_del(struct bpf_ksym *ksym);
710 void bpf_ksym_add(struct bpf_ksym *ksym);
711 void bpf_ksym_del(struct bpf_ksym *ksym);
712 int bpf_jit_charge_modmem(u32 pages);
713 void bpf_jit_uncharge_modmem(u32 pages);
714 #else
bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr)715 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr)
716 {
717 return -ENOTSUPP;
718 }
bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr)719 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr)
720 {
721 return -ENOTSUPP;
722 }
bpf_trampoline_get(u64 key, struct bpf_attach_target_info *tgt_info)723 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key, struct bpf_attach_target_info *tgt_info)
724 {
725 return ERR_PTR(-EOPNOTSUPP);
726 }
bpf_trampoline_put(struct bpf_trampoline *tr)727 static inline void bpf_trampoline_put(struct bpf_trampoline *tr)
728 {
729 }
730 #define DEFINE_BPF_DISPATCHER(name)
731 #define DECLARE_BPF_DISPATCHER(name)
732 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
733 #define BPF_DISPATCHER_PTR(name) NULL
bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, struct bpf_prog *to)734 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from, struct bpf_prog *to)
735 {
736 }
is_bpf_image_address(unsigned long address)737 static inline bool is_bpf_image_address(unsigned long address)
738 {
739 return false;
740 }
741 #endif
742
743 struct bpf_func_info_aux {
744 u16 linkage;
745 bool unreliable;
746 };
747
748 enum bpf_jit_poke_reason {
749 BPF_POKE_REASON_TAIL_CALL,
750 };
751
752 /* Descriptor of pokes pointing /into/ the JITed image. */
753 struct bpf_jit_poke_descriptor {
754 void *tailcall_target;
755 void *tailcall_bypass;
756 void *bypass_addr;
757 void *aux;
758 union {
759 struct {
760 struct bpf_map *map;
761 u32 key;
762 } tail_call;
763 };
764 bool tailcall_target_stable;
765 u8 adj_off;
766 u16 reason;
767 u32 insn_idx;
768 };
769
770 /* reg_type info for ctx arguments */
771 struct bpf_ctx_arg_aux {
772 u32 offset;
773 enum bpf_reg_type reg_type;
774 u32 btf_id;
775 };
776
777 struct bpf_prog_aux {
778 atomic64_t refcnt;
779 u32 used_map_cnt;
780 u32 max_ctx_offset;
781 u32 max_pkt_offset;
782 u32 max_tp_access;
783 u32 stack_depth;
784 u32 id;
785 u32 func_cnt; /* used by non-func prog as the number of func progs */
786 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
787 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
788 u32 ctx_arg_info_size;
789 u32 max_rdonly_access;
790 u32 max_rdwr_access;
791 const struct bpf_ctx_arg_aux *ctx_arg_info;
792 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
793 struct bpf_prog *dst_prog;
794 struct bpf_trampoline *dst_trampoline;
795 enum bpf_prog_type saved_dst_prog_type;
796 enum bpf_attach_type saved_dst_attach_type;
797 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
798 bool offload_requested;
799 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
800 bool func_proto_unreliable;
801 bool sleepable;
802 bool tail_call_reachable;
803 struct hlist_node tramp_hlist;
804 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
805 const struct btf_type *attach_func_proto;
806 /* function name for valid attach_btf_id */
807 const char *attach_func_name;
808 struct bpf_prog **func;
809 void *jit_data; /* JIT specific data. arch dependent */
810 struct bpf_jit_poke_descriptor *poke_tab;
811 u32 size_poke_tab;
812 struct bpf_ksym ksym;
813 const struct bpf_prog_ops *ops;
814 struct bpf_map **used_maps;
815 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
816 struct bpf_prog *prog;
817 struct user_struct *user;
818 u64 load_time; /* ns since boottime */
819 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
820 char name[BPF_OBJ_NAME_LEN];
821 #ifdef CONFIG_SECURITY
822 void *security;
823 #endif
824 struct bpf_prog_offload *offload;
825 struct btf *btf;
826 struct bpf_func_info *func_info;
827 struct bpf_func_info_aux *func_info_aux;
828 /* bpf_line_info loaded from userspace. linfo->insn_off
829 * has the xlated insn offset.
830 * Both the main and sub prog share the same linfo.
831 * The subprog can access its first linfo by
832 * using the linfo_idx.
833 */
834 struct bpf_line_info *linfo;
835 /* jited_linfo is the jited addr of the linfo. It has a
836 * one to one mapping to linfo:
837 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
838 * Both the main and sub prog share the same jited_linfo.
839 * The subprog can access its first jited_linfo by
840 * using the linfo_idx.
841 */
842 void **jited_linfo;
843 u32 func_info_cnt;
844 u32 nr_linfo;
845 /* subprog can use linfo_idx to access its first linfo and
846 * jited_linfo.
847 * main prog always has linfo_idx == 0
848 */
849 u32 linfo_idx;
850 u32 num_exentries;
851 struct exception_table_entry *extable;
852 struct bpf_prog_stats __percpu *stats;
853 union {
854 struct work_struct work;
855 struct rcu_head rcu;
856 };
857 };
858
859 struct bpf_array_aux {
860 /* 'Ownership' of prog array is claimed by the first program that
861 * is going to use this map or by the first program which FD is
862 * stored in the map to make sure that all callers and callees have
863 * the same prog type and JITed flag.
864 */
865 struct {
866 spinlock_t lock;
867 enum bpf_prog_type type;
868 bool jited;
869 } owner;
870 /* Programs with direct jumps into programs part of this array. */
871 struct list_head poke_progs;
872 struct bpf_map *map;
873 struct mutex poke_mutex;
874 struct work_struct work;
875 };
876
877 struct bpf_link {
878 atomic64_t refcnt;
879 u32 id;
880 enum bpf_link_type type;
881 const struct bpf_link_ops *ops;
882 struct bpf_prog *prog;
883 struct work_struct work;
884 };
885
886 struct bpf_link_ops {
887 void (*release)(struct bpf_link *link);
888 void (*dealloc)(struct bpf_link *link);
889 int (*detach)(struct bpf_link *link);
890 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog, struct bpf_prog *old_prog);
891 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
892 int (*fill_link_info)(const struct bpf_link *link, struct bpf_link_info *info);
893 };
894
895 struct bpf_link_primer {
896 struct bpf_link *link;
897 struct file *file;
898 int fd;
899 u32 id;
900 };
901
902 struct bpf_struct_ops_value;
903 struct btf_type;
904 struct btf_member;
905
906 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
907 struct bpf_struct_ops {
908 const struct bpf_verifier_ops *verifier_ops;
909 int (*init)(struct btf *btf);
910 int (*check_member)(const struct btf_type *t, const struct btf_member *member);
911 int (*init_member)(const struct btf_type *t, const struct btf_member *member, void *kdata, const void *udata);
912 int (*reg)(void *kdata);
913 void (*unreg)(void *kdata);
914 const struct btf_type *type;
915 const struct btf_type *value_type;
916 const char *name;
917 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
918 u32 type_id;
919 u32 value_id;
920 };
921
922 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
923 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
924 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id);
925 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log);
926 bool bpf_struct_ops_get(const void *kdata);
927 void bpf_struct_ops_put(const void *kdata);
928 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, void *value);
bpf_try_module_get(const void *data, struct module *owner)929 static inline bool bpf_try_module_get(const void *data, struct module *owner)
930 {
931 if (owner == BPF_MODULE_OWNER) {
932 return bpf_struct_ops_get(data);
933 } else {
934 return try_module_get(owner);
935 }
936 }
bpf_module_put(const void *data, struct module *owner)937 static inline void bpf_module_put(const void *data, struct module *owner)
938 {
939 if (owner == BPF_MODULE_OWNER) {
940 bpf_struct_ops_put(data);
941 } else {
942 module_put(owner);
943 }
944 }
945 #else
bpf_struct_ops_find(u32 type_id)946 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id)
947 {
948 return NULL;
949 }
bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log)950 static inline void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log)
951 {
952 }
bpf_try_module_get(const void *data, struct module *owner)953 static inline bool bpf_try_module_get(const void *data, struct module *owner)
954 {
955 return try_module_get(owner);
956 }
bpf_module_put(const void *data, struct module *owner)957 static inline void bpf_module_put(const void *data, struct module *owner)
958 {
959 module_put(owner);
960 }
bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, void *value)961 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key, void *value)
962 {
963 return -EINVAL;
964 }
965 #endif
966
967 struct bpf_array {
968 struct bpf_map map;
969 u32 elem_size;
970 u32 index_mask;
971 struct bpf_array_aux *aux;
972 union {
973 char value[0] __aligned(8);
974 void *ptrs[0] __aligned(8);
975 void __percpu *pptrs[0] __aligned(8);
976 };
977 };
978
979 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
980 #define MAX_TAIL_CALL_CNT 32
981
982 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | BPF_F_RDONLY_PROG | BPF_F_WRONLY | BPF_F_WRONLY_PROG)
983
984 #define BPF_MAP_CAN_READ BIT(0)
985 #define BPF_MAP_CAN_WRITE BIT(1)
986
bpf_map_flags_to_cap(struct bpf_map *map)987 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
988 {
989 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
990
991 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
992 * not possible.
993 */
994 if (access_flags & BPF_F_RDONLY_PROG) {
995 return BPF_MAP_CAN_READ;
996 } else if (access_flags & BPF_F_WRONLY_PROG) {
997 return BPF_MAP_CAN_WRITE;
998 } else {
999 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
1000 }
1001 }
1002
bpf_map_flags_access_ok(u32 access_flags)1003 static inline bool bpf_map_flags_access_ok(u32 access_flags)
1004 {
1005 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) != (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1006 }
1007
1008 struct bpf_event_entry {
1009 struct perf_event *event;
1010 struct file *perf_file;
1011 struct file *map_file;
1012 struct rcu_head rcu;
1013 };
1014
1015 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
1016 int bpf_prog_calc_tag(struct bpf_prog *fp);
1017 const char *kernel_type_name(u32 btf_type_id);
1018
1019 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
1020
1021 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src, unsigned long off, unsigned long len);
1022 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type, const struct bpf_insn *src, struct bpf_insn *dst,
1023 struct bpf_prog *prog, u32 *target_size);
1024
1025 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size, void *ctx, u64 ctx_size,
1026 bpf_ctx_copy_t ctx_copy);
1027
1028 /* an array of programs to be executed under rcu_lock.
1029 *
1030 * Typical usage:
1031 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
1032 *
1033 * the structure returned by bpf_prog_array_alloc() should be populated
1034 * with program pointers and the last pointer must be NULL.
1035 * The user has to keep refcnt on the program and make sure the program
1036 * is removed from the array before bpf_prog_put().
1037 * The 'struct bpf_prog_array *' should only be replaced with xchg()
1038 * since other cpus are walking the array of pointers in parallel.
1039 */
1040 struct bpf_prog_array_item {
1041 struct bpf_prog *prog;
1042 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1043 };
1044
1045 struct bpf_prog_array {
1046 struct rcu_head rcu;
1047 struct bpf_prog_array_item items[];
1048 };
1049
1050 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
1051 void bpf_prog_array_free(struct bpf_prog_array *progs);
1052 int bpf_prog_array_length(struct bpf_prog_array *progs);
1053 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
1054 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs, __u32 __user *prog_ids, u32 cnt);
1055
1056 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs, struct bpf_prog *old_prog);
1057 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
1058 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index, struct bpf_prog *prog);
1059 int bpf_prog_array_copy_info(struct bpf_prog_array *array, u32 *prog_ids, u32 request_cnt, u32 *prog_cnt);
1060 int bpf_prog_array_copy(struct bpf_prog_array *old_array, struct bpf_prog *exclude_prog, struct bpf_prog *include_prog,
1061 struct bpf_prog_array **new_array);
1062
1063 struct bpf_run_ctx {
1064 };
1065
1066 struct bpf_cg_run_ctx {
1067 struct bpf_run_ctx run_ctx;
1068 struct bpf_prog_array_item *prog_item;
1069 };
1070
1071 #define I_BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null, set_cg_storage) \
1072 ( { \
1073 struct bpf_prog_array_item *_item; \
1074 struct bpf_prog *_prog; \
1075 struct bpf_prog_array *_array; \
1076 struct bpf_run_ctx *old_run_ctx; \
1077 struct bpf_cg_run_ctx run_ctx; \
1078 u32 _ret = 1; \
1079 migrate_disable(); \
1080 rcu_read_lock(); \
1081 _array = rcu_dereference(array); \
1082 if (unlikely((check_non_null) && !_array)) \
1083 goto _out; \
1084 _item = &_array->items[0]; \
1085 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); \
1086 while ((_prog = READ_ONCE(_item->prog))) { \
1087 run_ctx.prog_item = _item; \
1088 _ret &= func(_prog, ctx); \
1089 _item++; \
1090 } \
1091 bpf_reset_run_ctx(old_run_ctx); \
1092 _out: \
1093 rcu_read_unlock(); \
1094 migrate_enable(); \
1095 _ret; \
1096 })
1097
1098 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs
1099 * so BPF programs can request cwr for TCP packets.
1100 *
1101 * Current cgroup skb programs can only return 0 or 1 (0 to drop the
1102 * packet. This macro changes the behavior so the low order bit
1103 * indicates whether the packet should be dropped (0) or not (1)
1104 * and the next bit is a congestion notification bit. This could be
1105 * used by TCP to call tcp_enter_cwr()
1106 *
1107 * Hence, new allowed return values of CGROUP EGRESS BPF programs are:
1108 * 0: drop packet
1109 * 1: keep packet
1110 * 2: drop packet and cn
1111 * 3: keep packet and cn
1112 *
1113 * This macro then converts it to one of the NET_XMIT or an error
1114 * code that is then interpreted as drop packet (and no cn):
1115 * 0: NET_XMIT_SUCCESS skb should be transmitted
1116 * 1: NET_XMIT_DROP skb should be dropped and cn
1117 * 2: NET_XMIT_CN skb should be transmitted and cn
1118 * 3: -EPERM skb should be dropped
1119 */
1120 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \
1121 ( { \
1122 struct bpf_prog_array_item *_item; \
1123 struct bpf_prog *_prog; \
1124 struct bpf_prog_array *_array; \
1125 struct bpf_run_ctx *old_run_ctx; \
1126 struct bpf_cg_run_ctx run_ctx; \
1127 u32 ret; \
1128 u32 _ret = 1; \
1129 u32 _cn = 0; \
1130 migrate_disable(); \
1131 rcu_read_lock(); \
1132 _array = rcu_dereference(array); \
1133 _item = &_array->items[0]; \
1134 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx); \
1135 while ((_prog = READ_ONCE(_item->prog))) { \
1136 run_ctx.prog_item = _item; \
1137 ret = func(_prog, ctx); \
1138 _ret &= (ret & 1); \
1139 _cn |= (ret & 2); \
1140 _item++; \
1141 } \
1142 bpf_reset_run_ctx(old_run_ctx); \
1143 rcu_read_unlock(); \
1144 migrate_enable(); \
1145 if (_ret) \
1146 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \
1147 else \
1148 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \
1149 _ret; \
1150 })
1151
1152 #define BPF_PROG_RUN_ARRAY(array, ctx, func) I_BPF_PROG_RUN_ARRAY(array, ctx, func, false, true)
1153
1154 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) I_BPF_PROG_RUN_ARRAY(array, ctx, func, true, false)
1155
1156 #ifdef CONFIG_BPF_SYSCALL
1157 DECLARE_PER_CPU(int, bpf_prog_active);
1158 extern struct mutex bpf_stats_enabled_mutex;
1159
1160 /*
1161 * Block execution of BPF programs attached to instrumentation (perf,
1162 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
1163 * these events can happen inside a region which holds a map bucket lock
1164 * and can deadlock on it.
1165 *
1166 * Use the preemption safe inc/dec variants on RT because migrate disable
1167 * is preemptible on RT and preemption in the middle of the RMW operation
1168 * might lead to inconsistent state. Use the raw variants for non RT
1169 * kernels as migrate_disable() maps to preempt_disable() so the slightly
1170 * more expensive save operation can be avoided.
1171 */
bpf_disable_instrumentation(void)1172 static inline void bpf_disable_instrumentation(void)
1173 {
1174 migrate_disable();
1175 if (IS_ENABLED(CONFIG_PREEMPT_RT)) {
1176 this_cpu_inc(bpf_prog_active);
1177 } else {
1178 __this_cpu_inc(bpf_prog_active);
1179 }
1180 }
1181
bpf_enable_instrumentation(void)1182 static inline void bpf_enable_instrumentation(void)
1183 {
1184 if (IS_ENABLED(CONFIG_PREEMPT_RT)) {
1185 this_cpu_dec(bpf_prog_active);
1186 } else {
1187 __this_cpu_dec(bpf_prog_active);
1188 }
1189 migrate_enable();
1190 }
1191
bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)1192 static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)
1193 {
1194 struct bpf_run_ctx *old_ctx;
1195
1196 old_ctx = current->bpf_ctx;
1197 current->bpf_ctx = new_ctx;
1198 return old_ctx;
1199 }
1200
bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)1201 static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
1202 {
1203 current->bpf_ctx = old_ctx;
1204 }
1205
1206 extern const struct file_operations bpf_map_fops;
1207 extern const struct file_operations bpf_prog_fops;
1208 extern const struct file_operations bpf_iter_fops;
1209
1210 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
1211 extern const struct bpf_prog_ops _name##_prog_ops; \
1212 extern const struct bpf_verifier_ops _name##_verifier_ops;
1213 #define BPF_MAP_TYPE(_id, _ops) extern const struct bpf_map_ops _ops;
1214 #define BPF_LINK_TYPE(_id, _name)
1215 #include <linux/bpf_types.h>
1216 #undef BPF_PROG_TYPE
1217 #undef BPF_MAP_TYPE
1218 #undef BPF_LINK_TYPE
1219
1220 extern const struct bpf_prog_ops bpf_offload_prog_ops;
1221 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
1222 extern const struct bpf_verifier_ops xdp_analyzer_ops;
1223
1224 struct bpf_prog *bpf_prog_get(u32 ufd);
1225 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, bool attach_drv);
1226 void bpf_prog_add(struct bpf_prog *prog, int i);
1227 void bpf_prog_sub(struct bpf_prog *prog, int i);
1228 void bpf_prog_inc(struct bpf_prog *prog);
1229 struct bpf_prog *__must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
1230 void bpf_prog_put(struct bpf_prog *prog);
1231 int __bpf_prog_charge(struct user_struct *user, u32 pages);
1232 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
1233
1234 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
1235 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
1236
1237 struct bpf_map *bpf_map_get(u32 ufd);
1238 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
1239 struct bpf_map *__bpf_map_get(struct fd f);
1240 void bpf_map_inc(struct bpf_map *map);
1241 void bpf_map_inc_with_uref(struct bpf_map *map);
1242 struct bpf_map *__must_check bpf_map_inc_not_zero(struct bpf_map *map);
1243 void bpf_map_put_with_uref(struct bpf_map *map);
1244 void bpf_map_put(struct bpf_map *map);
1245 int bpf_map_charge_memlock(struct bpf_map *map, u32 pages);
1246 void bpf_map_uncharge_memlock(struct bpf_map *map, u32 pages);
1247 int bpf_map_charge_init(struct bpf_map_memory *mem, u64 size);
1248 void bpf_map_charge_finish(struct bpf_map_memory *mem);
1249 void bpf_map_charge_move(struct bpf_map_memory *dst, struct bpf_map_memory *src);
1250 void *bpf_map_area_alloc(u64 size, int numa_node);
1251 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
1252 void bpf_map_area_free(void *base);
1253 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
1254 int generic_map_lookup_batch(struct bpf_map *map, const union bpf_attr *attr, union bpf_attr __user *uattr);
1255 int generic_map_update_batch(struct bpf_map *map, const union bpf_attr *attr, union bpf_attr __user *uattr);
1256 int generic_map_delete_batch(struct bpf_map *map, const union bpf_attr *attr, union bpf_attr __user *uattr);
1257 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
1258 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
1259
1260 extern int sysctl_unprivileged_bpf_disabled;
1261
bpf_allow_ptr_leaks(void)1262 static inline bool bpf_allow_ptr_leaks(void)
1263 {
1264 return perfmon_capable();
1265 }
1266
bpf_allow_uninit_stack(void)1267 static inline bool bpf_allow_uninit_stack(void)
1268 {
1269 return perfmon_capable();
1270 }
1271
bpf_allow_ptr_to_map_access(void)1272 static inline bool bpf_allow_ptr_to_map_access(void)
1273 {
1274 return perfmon_capable();
1275 }
1276
bpf_bypass_spec_v1(void)1277 static inline bool bpf_bypass_spec_v1(void)
1278 {
1279 return perfmon_capable();
1280 }
1281
bpf_bypass_spec_v4(void)1282 static inline bool bpf_bypass_spec_v4(void)
1283 {
1284 return perfmon_capable();
1285 }
1286
1287 int bpf_map_new_fd(struct bpf_map *map, int flags);
1288 int bpf_prog_new_fd(struct bpf_prog *prog);
1289
1290 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, const struct bpf_link_ops *ops,
1291 struct bpf_prog *prog);
1292 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
1293 int bpf_link_settle(struct bpf_link_primer *primer);
1294 void bpf_link_cleanup(struct bpf_link_primer *primer);
1295 void bpf_link_inc(struct bpf_link *link);
1296 void bpf_link_put(struct bpf_link *link);
1297 int bpf_link_new_fd(struct bpf_link *link);
1298 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd);
1299 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
1300
1301 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
1302 int bpf_obj_get_user(const char __user *pathname, int flags);
1303
1304 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
1305 #define DEFINE_BPF_ITER_FUNC(target, args...) \
1306 extern int bpf_iter_##target(args); \
1307 int __init bpf_iter_##target(args) \
1308 { \
1309 return 0; \
1310 }
1311
1312 struct bpf_iter_aux_info {
1313 struct bpf_map *map;
1314 };
1315
1316 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog, union bpf_iter_link_info *linfo,
1317 struct bpf_iter_aux_info *aux);
1318 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
1319 typedef void (*bpf_iter_show_fdinfo_t)(const struct bpf_iter_aux_info *aux, struct seq_file *seq);
1320 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux, struct bpf_link_info *info);
1321
1322 #define BPF_ITER_CTX_ARG_MAX 2
1323 struct bpf_iter_reg {
1324 const char *target;
1325 bpf_iter_attach_target_t attach_target;
1326 bpf_iter_detach_target_t detach_target;
1327 bpf_iter_show_fdinfo_t show_fdinfo;
1328 bpf_iter_fill_link_info_t fill_link_info;
1329 u32 ctx_arg_info_size;
1330 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
1331 const struct bpf_iter_seq_info *seq_info;
1332 };
1333
1334 struct bpf_iter_meta {
1335 __bpf_md_ptr(struct seq_file *, seq);
1336 u64 session_id;
1337 u64 seq_num;
1338 };
1339
1340 struct bpf_iter__bpf_map_elem {
1341 __bpf_md_ptr(struct bpf_iter_meta *, meta);
1342 __bpf_md_ptr(struct bpf_map *, map);
1343 __bpf_md_ptr(void *, key);
1344 __bpf_md_ptr(void *, value);
1345 };
1346
1347 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
1348 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
1349 bool bpf_iter_prog_supported(struct bpf_prog *prog);
1350 int bpf_iter_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
1351 int bpf_iter_new_fd(struct bpf_link *link);
1352 bool bpf_link_is_iter(struct bpf_link *link);
1353 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
1354 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
1355 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux, struct seq_file *seq);
1356 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux, struct bpf_link_info *info);
1357
1358 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
1359 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
1360 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value, u64 flags);
1361 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value, u64 flags);
1362
1363 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
1364
1365 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file, void *key, void *value, u64 map_flags);
1366 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1367 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file, void *key, void *value, u64 map_flags);
1368 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1369
1370 int bpf_get_file_flag(int flags);
1371 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size, size_t actual_size);
1372
1373 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
1374 * forced to use 'long' read/writes to try to atomically copy long counters.
1375 * Best-effort only. No barriers here, since it _will_ race with concurrent
1376 * updates from BPF programs. Called from bpf syscall and mostly used with
1377 * size 8 or 16 bytes, so ask compiler to inline it.
1378 */
bpf_long_memcpy(void *dst, const void *src, u32 size)1379 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
1380 {
1381 const long *lsrc = src;
1382 long *ldst = dst;
1383
1384 size /= sizeof(long);
1385 while (size--) {
1386 *ldst++ = *lsrc++;
1387 }
1388 }
1389
1390 /* verify correctness of eBPF program */
1391 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, union bpf_attr __user *uattr);
1392
1393 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
1394 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
1395 #endif
1396
1397 struct btf *bpf_get_btf_vmlinux(void);
1398
1399 /* Map specifics */
1400 struct xdp_buff;
1401 struct sk_buff;
1402
1403 struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
1404 struct bpf_dtab_netdev *__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key);
1405 void __dev_flush(void);
1406 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, struct net_device *dev_rx);
1407 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, struct net_device *dev_rx);
1408 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, struct bpf_prog *xdp_prog);
1409 bool dev_map_can_have_prog(struct bpf_map *map);
1410
1411 struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key);
1412 void __cpu_map_flush(void);
1413 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, struct net_device *dev_rx);
1414 bool cpu_map_prog_allowed(struct bpf_map *map);
1415
1416 /* Return map's numa specified by userspace */
bpf_map_attr_numa_node(const union bpf_attr *attr)1417 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
1418 {
1419 return (attr->map_flags & BPF_F_NUMA_NODE) ? attr->numa_node : NUMA_NO_NODE;
1420 }
1421
1422 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
1423 int array_map_alloc_check(union bpf_attr *attr);
1424
1425 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr);
1426 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr);
1427 int bpf_prog_test_run_tracing(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr);
1428 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr);
1429 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr);
1430 bool btf_ctx_access(int off, int size, enum bpf_access_type type, const struct bpf_prog *prog,
1431 struct bpf_insn_access_aux *info);
1432 int btf_struct_access(struct bpf_verifier_log *log, const struct btf_type *t, int off, int size,
1433 enum bpf_access_type atype, u32 *next_btf_id);
1434 bool btf_struct_ids_match(struct bpf_verifier_log *log, int off, u32 id, u32 need_type_id);
1435
1436 int btf_distill_func_proto(struct bpf_verifier_log *log, struct btf *btf, const struct btf_type *func_proto,
1437 const char *func_name, struct btf_func_model *m);
1438
1439 struct bpf_reg_state;
1440 int btf_check_func_arg_match(struct bpf_verifier_env *env, int subprog, struct bpf_reg_state *regs);
1441 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog, struct bpf_reg_state *reg);
1442 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog, struct btf *btf,
1443 const struct btf_type *t);
1444
1445 struct bpf_prog *bpf_prog_by_id(u32 id);
1446 struct bpf_link *bpf_link_by_id(u32 id);
1447
1448 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
1449
unprivileged_ebpf_enabled(void)1450 static inline bool unprivileged_ebpf_enabled(void)
1451 {
1452 return !sysctl_unprivileged_bpf_disabled;
1453 }
1454
1455 #else /* !CONFIG_BPF_SYSCALL */
bpf_prog_get(u32 ufd)1456 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
1457 {
1458 return ERR_PTR(-EOPNOTSUPP);
1459 }
1460
bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, bool attach_drv)1461 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type, bool attach_drv)
1462 {
1463 return ERR_PTR(-EOPNOTSUPP);
1464 }
1465
bpf_prog_add(struct bpf_prog *prog, int i)1466 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
1467 {
1468 }
1469
bpf_prog_sub(struct bpf_prog *prog, int i)1470 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
1471 {
1472 }
1473
bpf_prog_put(struct bpf_prog *prog)1474 static inline void bpf_prog_put(struct bpf_prog *prog)
1475 {
1476 }
1477
bpf_prog_inc(struct bpf_prog *prog)1478 static inline void bpf_prog_inc(struct bpf_prog *prog)
1479 {
1480 }
1481
bpf_prog_inc_not_zero(struct bpf_prog *prog)1482 static inline struct bpf_prog *__must_check bpf_prog_inc_not_zero(struct bpf_prog *prog)
1483 {
1484 return ERR_PTR(-EOPNOTSUPP);
1485 }
1486
__bpf_prog_charge(struct user_struct *user, u32 pages)1487 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
1488 {
1489 return 0;
1490 }
1491
__bpf_prog_uncharge(struct user_struct *user, u32 pages)1492 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
1493 {
1494 }
1495
bpf_link_init(struct bpf_link *link, enum bpf_link_type type, const struct bpf_link_ops *ops, struct bpf_prog *prog)1496 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type, const struct bpf_link_ops *ops,
1497 struct bpf_prog *prog)
1498 {
1499 }
1500
bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer)1501 static inline int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer)
1502 {
1503 return -EOPNOTSUPP;
1504 }
1505
bpf_link_settle(struct bpf_link_primer *primer)1506 static inline int bpf_link_settle(struct bpf_link_primer *primer)
1507 {
1508 return -EOPNOTSUPP;
1509 }
1510
bpf_link_cleanup(struct bpf_link_primer *primer)1511 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
1512 {
1513 }
1514
bpf_link_inc(struct bpf_link *link)1515 static inline void bpf_link_inc(struct bpf_link *link)
1516 {
1517 }
1518
bpf_link_put(struct bpf_link *link)1519 static inline void bpf_link_put(struct bpf_link *link)
1520 {
1521 }
1522
bpf_obj_get_user(const char __user *pathname, int flags)1523 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
1524 {
1525 return -EOPNOTSUPP;
1526 }
1527
__dev_map_lookup_elem(struct bpf_map *map, u32 key)1528 static inline struct net_device *__dev_map_lookup_elem(struct bpf_map *map, u32 key)
1529 {
1530 return NULL;
1531 }
1532
__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key)1533 static inline struct net_device *__dev_map_hash_lookup_elem(struct bpf_map *map, u32 key)
1534 {
1535 return NULL;
1536 }
dev_map_can_have_prog(struct bpf_map *map)1537 static inline bool dev_map_can_have_prog(struct bpf_map *map)
1538 {
1539 return false;
1540 }
1541
__dev_flush(void)1542 static inline void __dev_flush(void)
1543 {
1544 }
1545
1546 struct xdp_buff;
1547 struct bpf_dtab_netdev;
1548
dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, struct net_device *dev_rx)1549 static inline int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp, struct net_device *dev_rx)
1550 {
1551 return 0;
1552 }
1553
dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, struct net_device *dev_rx)1554 static inline int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, struct net_device *dev_rx)
1555 {
1556 return 0;
1557 }
1558
1559 struct sk_buff;
1560
dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, struct bpf_prog *xdp_prog)1561 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb, struct bpf_prog *xdp_prog)
1562 {
1563 return 0;
1564 }
1565
__cpu_map_lookup_elem(struct bpf_map *map, u32 key)1566 static inline struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key)
1567 {
1568 return NULL;
1569 }
1570
__cpu_map_flush(void)1571 static inline void __cpu_map_flush(void)
1572 {
1573 }
1574
cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, struct net_device *dev_rx)1575 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp, struct net_device *dev_rx)
1576 {
1577 return 0;
1578 }
1579
cpu_map_prog_allowed(struct bpf_map *map)1580 static inline bool cpu_map_prog_allowed(struct bpf_map *map)
1581 {
1582 return false;
1583 }
1584
bpf_prog_get_type_path(const char *name, enum bpf_prog_type type)1585 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type)
1586 {
1587 return ERR_PTR(-EOPNOTSUPP);
1588 }
1589
bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr)1590 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
1591 union bpf_attr __user *uattr)
1592 {
1593 return -ENOTSUPP;
1594 }
1595
bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr)1596 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
1597 union bpf_attr __user *uattr)
1598 {
1599 return -ENOTSUPP;
1600 }
1601
bpf_prog_test_run_tracing(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr)1602 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog, const union bpf_attr *kattr,
1603 union bpf_attr __user *uattr)
1604 {
1605 return -ENOTSUPP;
1606 }
1607
bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, const union bpf_attr *kattr, union bpf_attr __user *uattr)1608 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog, const union bpf_attr *kattr,
1609 union bpf_attr __user *uattr)
1610 {
1611 return -ENOTSUPP;
1612 }
1613
bpf_map_put(struct bpf_map *map)1614 static inline void bpf_map_put(struct bpf_map *map)
1615 {
1616 }
1617
bpf_prog_by_id(u32 id)1618 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
1619 {
1620 return ERR_PTR(-ENOTSUPP);
1621 }
1622
bpf_base_func_proto(enum bpf_func_id func_id)1623 static inline const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id)
1624 {
1625 return NULL;
1626 }
1627
unprivileged_ebpf_enabled(void)1628 static inline bool unprivileged_ebpf_enabled(void)
1629 {
1630 return false;
1631 }
1632
1633 #endif /* CONFIG_BPF_SYSCALL */
1634
bpf_prog_get_type(u32 ufd, enum bpf_prog_type type)1635 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type)
1636 {
1637 return bpf_prog_get_type_dev(ufd, type, false);
1638 }
1639
1640 void __bpf_free_used_maps(struct bpf_prog_aux *aux, struct bpf_map **used_maps, u32 len);
1641
1642 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
1643
1644 int bpf_prog_offload_compile(struct bpf_prog *prog);
1645 void bpf_prog_offload_destroy(struct bpf_prog *prog);
1646 int bpf_prog_offload_info_fill(struct bpf_prog_info *info, struct bpf_prog *prog);
1647
1648 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
1649
1650 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
1651 int bpf_map_offload_update_elem(struct bpf_map *map, void *key, void *value, u64 flags);
1652 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
1653 int bpf_map_offload_get_next_key(struct bpf_map *map, void *key, void *next_key);
1654
1655 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
1656
1657 struct bpf_offload_dev *bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
1658 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
1659 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
1660 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev, struct net_device *netdev);
1661 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev, struct net_device *netdev);
1662 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
1663
1664 void unpriv_ebpf_notify(int new_state);
1665
1666 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
1667 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
1668
bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)1669 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
1670 {
1671 return aux->offload_requested;
1672 }
1673
bpf_map_is_dev_bound(struct bpf_map *map)1674 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1675 {
1676 return unlikely(map->ops == &bpf_map_offload_ops);
1677 }
1678
1679 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
1680 void bpf_map_offload_map_free(struct bpf_map *map);
1681 #else
bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr)1682 static inline int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr)
1683 {
1684 return -EOPNOTSUPP;
1685 }
1686
bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)1687 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
1688 {
1689 return false;
1690 }
1691
bpf_map_is_dev_bound(struct bpf_map *map)1692 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1693 {
1694 return false;
1695 }
1696
bpf_map_offload_map_alloc(union bpf_attr *attr)1697 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
1698 {
1699 return ERR_PTR(-EOPNOTSUPP);
1700 }
1701
bpf_map_offload_map_free(struct bpf_map *map)1702 static inline void bpf_map_offload_map_free(struct bpf_map *map)
1703 {
1704 }
1705 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
1706
1707 #if defined(CONFIG_BPF_STREAM_PARSER)
1708 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, struct bpf_prog *old, u32 which);
1709 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
1710 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
1711 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
1712 void sock_map_unhash(struct sock *sk);
1713 void sock_map_close(struct sock *sk, long timeout);
1714 #else
sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, struct bpf_prog *old, u32 which)1715 static inline int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog, struct bpf_prog *old, u32 which)
1716 {
1717 return -EOPNOTSUPP;
1718 }
1719
sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog)1720 static inline int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog)
1721 {
1722 return -EINVAL;
1723 }
1724
sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype)1725 static inline int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype)
1726 {
1727 return -EOPNOTSUPP;
1728 }
1729
sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags)1730 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags)
1731 {
1732 return -EOPNOTSUPP;
1733 }
1734 #endif /* CONFIG_BPF_STREAM_PARSER */
1735
1736 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
1737 void bpf_sk_reuseport_detach(struct sock *sk);
1738 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, void *value);
1739 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, void *value, u64 map_flags);
1740 #else
bpf_sk_reuseport_detach(struct sock *sk)1741 static inline void bpf_sk_reuseport_detach(struct sock *sk)
1742 {
1743 }
1744
1745 #ifdef CONFIG_BPF_SYSCALL
bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, void *value)1746 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key, void *value)
1747 {
1748 return -EOPNOTSUPP;
1749 }
1750
bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, void *value, u64 map_flags)1751 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key, void *value, u64 map_flags)
1752 {
1753 return -EOPNOTSUPP;
1754 }
1755 #endif /* CONFIG_BPF_SYSCALL */
1756 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
1757
1758 /* verifier prototypes for helper functions called from eBPF programs */
1759 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
1760 extern const struct bpf_func_proto bpf_map_update_elem_proto;
1761 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
1762 extern const struct bpf_func_proto bpf_map_push_elem_proto;
1763 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
1764 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
1765
1766 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
1767 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
1768 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
1769 extern const struct bpf_func_proto bpf_tail_call_proto;
1770 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
1771 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
1772 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
1773 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
1774 extern const struct bpf_func_proto bpf_get_current_comm_proto;
1775 extern const struct bpf_func_proto bpf_get_stackid_proto;
1776 extern const struct bpf_func_proto bpf_get_stack_proto;
1777 extern const struct bpf_func_proto bpf_get_task_stack_proto;
1778 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
1779 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
1780 extern const struct bpf_func_proto bpf_sock_map_update_proto;
1781 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
1782 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
1783 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
1784 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
1785 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
1786 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
1787 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
1788 extern const struct bpf_func_proto bpf_spin_lock_proto;
1789 extern const struct bpf_func_proto bpf_spin_unlock_proto;
1790 extern const struct bpf_func_proto bpf_get_local_storage_proto;
1791 extern const struct bpf_func_proto bpf_strtol_proto;
1792 extern const struct bpf_func_proto bpf_strtoul_proto;
1793 extern const struct bpf_func_proto bpf_tcp_sock_proto;
1794 extern const struct bpf_func_proto bpf_jiffies64_proto;
1795 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
1796 extern const struct bpf_func_proto bpf_event_output_data_proto;
1797 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
1798 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
1799 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
1800 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
1801 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
1802 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
1803 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
1804 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
1805 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
1806 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
1807 extern const struct bpf_func_proto bpf_copy_from_user_proto;
1808 extern const struct bpf_func_proto bpf_snprintf_btf_proto;
1809 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
1810 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
1811
1812 const struct bpf_func_proto *bpf_tracing_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
1813
1814 const struct bpf_func_proto *tracing_prog_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
1815
1816 /* Shared helpers among cBPF and eBPF. */
1817 void bpf_user_rnd_init_once(void);
1818 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1819 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1820
1821 #if defined(CONFIG_NET)
1822 bool bpf_sock_common_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info);
1823 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info);
1824 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si, struct bpf_insn *insn_buf,
1825 struct bpf_prog *prog, u32 *target_size);
1826 #else
bpf_sock_common_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info)1827 static inline bool bpf_sock_common_is_valid_access(int off, int size, enum bpf_access_type type,
1828 struct bpf_insn_access_aux *info)
1829 {
1830 return false;
1831 }
bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info)1832 static inline bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1833 struct bpf_insn_access_aux *info)
1834 {
1835 return false;
1836 }
bpf_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si, struct bpf_insn *insn_buf, struct bpf_prog *prog, u32 *target_size)1837 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si,
1838 struct bpf_insn *insn_buf, struct bpf_prog *prog, u32 *target_size)
1839 {
1840 return 0;
1841 }
1842 #endif
1843
1844 #ifdef CONFIG_INET
1845 struct sk_reuseport_kern {
1846 struct sk_buff *skb;
1847 struct sock *sk;
1848 struct sock *selected_sk;
1849 void *data_end;
1850 u32 hash;
1851 u32 reuseport_id;
1852 bool bind_inany;
1853 };
1854 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info);
1855
1856 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si, struct bpf_insn *insn_buf,
1857 struct bpf_prog *prog, u32 *target_size);
1858
1859 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info);
1860
1861 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si, struct bpf_insn *insn_buf,
1862 struct bpf_prog *prog, u32 *target_size);
1863 #else
bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info)1864 static inline bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1865 struct bpf_insn_access_aux *info)
1866 {
1867 return false;
1868 }
1869
bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si, struct bpf_insn *insn_buf, struct bpf_prog *prog, u32 *target_size)1870 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si,
1871 struct bpf_insn *insn_buf, struct bpf_prog *prog, u32 *target_size)
1872 {
1873 return 0;
1874 }
bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info)1875 static inline bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1876 struct bpf_insn_access_aux *info)
1877 {
1878 return false;
1879 }
1880
bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si, struct bpf_insn *insn_buf, struct bpf_prog *prog, u32 *target_size)1881 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type, const struct bpf_insn *si,
1882 struct bpf_insn *insn_buf, struct bpf_prog *prog, u32 *target_size)
1883 {
1884 return 0;
1885 }
1886 #endif /* CONFIG_INET */
1887
1888 enum bpf_text_poke_type {
1889 BPF_MOD_CALL,
1890 BPF_MOD_JUMP,
1891 };
1892
1893 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t, void *addr1, void *addr2);
1894
1895 struct btf_id_set;
1896 bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
1897
1898 #endif /* _LINUX_BPF_H */
1899