]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blob - include/linux/bpf.h
Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next
[mirror_ubuntu-jammy-kernel.git] / include / linux / bpf.h
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/refcount.h>
18 #include <linux/mutex.h>
19 #include <linux/module.h>
20 #include <linux/kallsyms.h>
21 #include <linux/capability.h>
22 #include <linux/sched/mm.h>
23 #include <linux/slab.h>
24 #include <linux/percpu-refcount.h>
25
26 struct bpf_verifier_env;
27 struct bpf_verifier_log;
28 struct perf_event;
29 struct bpf_prog;
30 struct bpf_prog_aux;
31 struct bpf_map;
32 struct sock;
33 struct seq_file;
34 struct btf;
35 struct btf_type;
36 struct exception_table_entry;
37 struct seq_operations;
38 struct bpf_iter_aux_info;
39 struct bpf_local_storage;
40 struct bpf_local_storage_map;
41 struct kobject;
42 struct mem_cgroup;
43 struct module;
44 struct bpf_func_state;
45
46 extern struct idr btf_idr;
47 extern spinlock_t btf_idr_lock;
48 extern struct kobject *btf_kobj;
49
50 typedef int (*bpf_iter_init_seq_priv_t)(void *private_data,
51 struct bpf_iter_aux_info *aux);
52 typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
53 struct bpf_iter_seq_info {
54 const struct seq_operations *seq_ops;
55 bpf_iter_init_seq_priv_t init_seq_private;
56 bpf_iter_fini_seq_priv_t fini_seq_private;
57 u32 seq_priv_size;
58 };
59
60 /* map is generic key/value storage optionally accessible by eBPF programs */
61 struct bpf_map_ops {
62 /* funcs callable from userspace (via syscall) */
63 int (*map_alloc_check)(union bpf_attr *attr);
64 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
65 void (*map_release)(struct bpf_map *map, struct file *map_file);
66 void (*map_free)(struct bpf_map *map);
67 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
68 void (*map_release_uref)(struct bpf_map *map);
69 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
70 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr,
71 union bpf_attr __user *uattr);
72 int (*map_lookup_and_delete_batch)(struct bpf_map *map,
73 const union bpf_attr *attr,
74 union bpf_attr __user *uattr);
75 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr,
76 union bpf_attr __user *uattr);
77 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr,
78 union bpf_attr __user *uattr);
79
80 /* funcs callable from userspace and from eBPF programs */
81 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
82 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
83 int (*map_delete_elem)(struct bpf_map *map, void *key);
84 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
85 int (*map_pop_elem)(struct bpf_map *map, void *value);
86 int (*map_peek_elem)(struct bpf_map *map, void *value);
87
88 /* funcs called by prog_array and perf_event_array map */
89 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
90 int fd);
91 void (*map_fd_put_ptr)(void *ptr);
92 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
93 u32 (*map_fd_sys_lookup_elem)(void *ptr);
94 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
95 struct seq_file *m);
96 int (*map_check_btf)(const struct bpf_map *map,
97 const struct btf *btf,
98 const struct btf_type *key_type,
99 const struct btf_type *value_type);
100
101 /* Prog poke tracking helpers. */
102 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
103 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
104 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
105 struct bpf_prog *new);
106
107 /* Direct value access helpers. */
108 int (*map_direct_value_addr)(const struct bpf_map *map,
109 u64 *imm, u32 off);
110 int (*map_direct_value_meta)(const struct bpf_map *map,
111 u64 imm, u32 *off);
112 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
113 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp,
114 struct poll_table_struct *pts);
115
116 /* Functions called by bpf_local_storage maps */
117 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap,
118 void *owner, u32 size);
119 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap,
120 void *owner, u32 size);
121 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner);
122
123 /* Misc helpers.*/
124 int (*map_redirect)(struct bpf_map *map, u32 ifindex, u64 flags);
125
126 /* map_meta_equal must be implemented for maps that can be
127 * used as an inner map. It is a runtime check to ensure
128 * an inner map can be inserted to an outer map.
129 *
130 * Some properties of the inner map has been used during the
131 * verification time. When inserting an inner map at the runtime,
132 * map_meta_equal has to ensure the inserting map has the same
133 * properties that the verifier has used earlier.
134 */
135 bool (*map_meta_equal)(const struct bpf_map *meta0,
136 const struct bpf_map *meta1);
137
138
139 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env,
140 struct bpf_func_state *caller,
141 struct bpf_func_state *callee);
142 int (*map_for_each_callback)(struct bpf_map *map, void *callback_fn,
143 void *callback_ctx, u64 flags);
144
145 /* BTF name and id of struct allocated by map_alloc */
146 const char * const map_btf_name;
147 int *map_btf_id;
148
149 /* bpf_iter info used to open a seq_file */
150 const struct bpf_iter_seq_info *iter_seq_info;
151 };
152
153 struct bpf_map {
154 /* The first two cachelines with read-mostly members of which some
155 * are also accessed in fast-path (e.g. ops, max_entries).
156 */
157 const struct bpf_map_ops *ops ____cacheline_aligned;
158 struct bpf_map *inner_map_meta;
159 #ifdef CONFIG_SECURITY
160 void *security;
161 #endif
162 enum bpf_map_type map_type;
163 u32 key_size;
164 u32 value_size;
165 u32 max_entries;
166 u32 map_flags;
167 int spin_lock_off; /* >=0 valid offset, <0 error */
168 u32 id;
169 int numa_node;
170 u32 btf_key_type_id;
171 u32 btf_value_type_id;
172 struct btf *btf;
173 #ifdef CONFIG_MEMCG_KMEM
174 struct mem_cgroup *memcg;
175 #endif
176 char name[BPF_OBJ_NAME_LEN];
177 u32 btf_vmlinux_value_type_id;
178 bool bypass_spec_v1;
179 bool frozen; /* write-once; write-protected by freeze_mutex */
180 /* 22 bytes hole */
181
182 /* The 3rd and 4th cacheline with misc members to avoid false sharing
183 * particularly with refcounting.
184 */
185 atomic64_t refcnt ____cacheline_aligned;
186 atomic64_t usercnt;
187 struct work_struct work;
188 struct mutex freeze_mutex;
189 u64 writecnt; /* writable mmap cnt; protected by freeze_mutex */
190 };
191
192 static inline bool map_value_has_spin_lock(const struct bpf_map *map)
193 {
194 return map->spin_lock_off >= 0;
195 }
196
197 static inline void check_and_init_map_lock(struct bpf_map *map, void *dst)
198 {
199 if (likely(!map_value_has_spin_lock(map)))
200 return;
201 *(struct bpf_spin_lock *)(dst + map->spin_lock_off) =
202 (struct bpf_spin_lock){};
203 }
204
205 /* copy everything but bpf_spin_lock */
206 static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
207 {
208 if (unlikely(map_value_has_spin_lock(map))) {
209 u32 off = map->spin_lock_off;
210
211 memcpy(dst, src, off);
212 memcpy(dst + off + sizeof(struct bpf_spin_lock),
213 src + off + sizeof(struct bpf_spin_lock),
214 map->value_size - off - sizeof(struct bpf_spin_lock));
215 } else {
216 memcpy(dst, src, map->value_size);
217 }
218 }
219 void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
220 bool lock_src);
221 int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
222
223 struct bpf_offload_dev;
224 struct bpf_offloaded_map;
225
226 struct bpf_map_dev_ops {
227 int (*map_get_next_key)(struct bpf_offloaded_map *map,
228 void *key, void *next_key);
229 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
230 void *key, void *value);
231 int (*map_update_elem)(struct bpf_offloaded_map *map,
232 void *key, void *value, u64 flags);
233 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
234 };
235
236 struct bpf_offloaded_map {
237 struct bpf_map map;
238 struct net_device *netdev;
239 const struct bpf_map_dev_ops *dev_ops;
240 void *dev_priv;
241 struct list_head offloads;
242 };
243
244 static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
245 {
246 return container_of(map, struct bpf_offloaded_map, map);
247 }
248
249 static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
250 {
251 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
252 }
253
254 static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
255 {
256 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
257 map->ops->map_seq_show_elem;
258 }
259
260 int map_check_no_btf(const struct bpf_map *map,
261 const struct btf *btf,
262 const struct btf_type *key_type,
263 const struct btf_type *value_type);
264
265 bool bpf_map_meta_equal(const struct bpf_map *meta0,
266 const struct bpf_map *meta1);
267
268 extern const struct bpf_map_ops bpf_map_offload_ops;
269
270 /* function argument constraints */
271 enum bpf_arg_type {
272 ARG_DONTCARE = 0, /* unused argument in helper function */
273
274 /* the following constraints used to prototype
275 * bpf_map_lookup/update/delete_elem() functions
276 */
277 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
278 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
279 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
280 ARG_PTR_TO_UNINIT_MAP_VALUE, /* pointer to valid memory used to store a map value */
281 ARG_PTR_TO_MAP_VALUE_OR_NULL, /* pointer to stack used as map value or NULL */
282
283 /* the following constraints used to prototype bpf_memcmp() and other
284 * functions that access data on eBPF program stack
285 */
286 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
287 ARG_PTR_TO_MEM_OR_NULL, /* pointer to valid memory or NULL */
288 ARG_PTR_TO_UNINIT_MEM, /* pointer to memory does not need to be initialized,
289 * helper function must fill all bytes or clear
290 * them in error case.
291 */
292
293 ARG_CONST_SIZE, /* number of bytes accessed from memory */
294 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
295
296 ARG_PTR_TO_CTX, /* pointer to context */
297 ARG_PTR_TO_CTX_OR_NULL, /* pointer to context or NULL */
298 ARG_ANYTHING, /* any (initialized) argument is ok */
299 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
300 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
301 ARG_PTR_TO_INT, /* pointer to int */
302 ARG_PTR_TO_LONG, /* pointer to long */
303 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
304 ARG_PTR_TO_SOCKET_OR_NULL, /* pointer to bpf_sock (fullsock) or NULL */
305 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
306 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */
307 ARG_PTR_TO_ALLOC_MEM_OR_NULL, /* pointer to dynamically allocated memory or NULL */
308 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
309 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
310 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */
311 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */
312 ARG_PTR_TO_STACK_OR_NULL, /* pointer to stack or NULL */
313 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */
314 __BPF_ARG_TYPE_MAX,
315 };
316
317 /* type of values returned from helper functions */
318 enum bpf_return_type {
319 RET_INTEGER, /* function returns integer */
320 RET_VOID, /* function doesn't return anything */
321 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
322 RET_PTR_TO_MAP_VALUE_OR_NULL, /* returns a pointer to map elem value or NULL */
323 RET_PTR_TO_SOCKET_OR_NULL, /* returns a pointer to a socket or NULL */
324 RET_PTR_TO_TCP_SOCK_OR_NULL, /* returns a pointer to a tcp_sock or NULL */
325 RET_PTR_TO_SOCK_COMMON_OR_NULL, /* returns a pointer to a sock_common or NULL */
326 RET_PTR_TO_ALLOC_MEM_OR_NULL, /* returns a pointer to dynamically allocated memory or NULL */
327 RET_PTR_TO_BTF_ID_OR_NULL, /* returns a pointer to a btf_id or NULL */
328 RET_PTR_TO_MEM_OR_BTF_ID_OR_NULL, /* returns a pointer to a valid memory or a btf_id or NULL */
329 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */
330 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */
331 };
332
333 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
334 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
335 * instructions after verifying
336 */
337 struct bpf_func_proto {
338 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
339 bool gpl_only;
340 bool pkt_access;
341 enum bpf_return_type ret_type;
342 union {
343 struct {
344 enum bpf_arg_type arg1_type;
345 enum bpf_arg_type arg2_type;
346 enum bpf_arg_type arg3_type;
347 enum bpf_arg_type arg4_type;
348 enum bpf_arg_type arg5_type;
349 };
350 enum bpf_arg_type arg_type[5];
351 };
352 union {
353 struct {
354 u32 *arg1_btf_id;
355 u32 *arg2_btf_id;
356 u32 *arg3_btf_id;
357 u32 *arg4_btf_id;
358 u32 *arg5_btf_id;
359 };
360 u32 *arg_btf_id[5];
361 };
362 int *ret_btf_id; /* return value btf_id */
363 bool (*allowed)(const struct bpf_prog *prog);
364 };
365
366 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
367 * the first argument to eBPF programs.
368 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
369 */
370 struct bpf_context;
371
372 enum bpf_access_type {
373 BPF_READ = 1,
374 BPF_WRITE = 2
375 };
376
377 /* types of values stored in eBPF registers */
378 /* Pointer types represent:
379 * pointer
380 * pointer + imm
381 * pointer + (u16) var
382 * pointer + (u16) var + imm
383 * if (range > 0) then [ptr, ptr + range - off) is safe to access
384 * if (id > 0) means that some 'var' was added
385 * if (off > 0) means that 'imm' was added
386 */
387 enum bpf_reg_type {
388 NOT_INIT = 0, /* nothing was written into register */
389 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
390 PTR_TO_CTX, /* reg points to bpf_context */
391 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
392 PTR_TO_MAP_VALUE, /* reg points to map element value */
393 PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
394 PTR_TO_STACK, /* reg == frame_pointer + offset */
395 PTR_TO_PACKET_META, /* skb->data - meta_len */
396 PTR_TO_PACKET, /* reg points to skb->data */
397 PTR_TO_PACKET_END, /* skb->data + headlen */
398 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
399 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
400 PTR_TO_SOCKET_OR_NULL, /* reg points to struct bpf_sock or NULL */
401 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
402 PTR_TO_SOCK_COMMON_OR_NULL, /* reg points to sock_common or NULL */
403 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
404 PTR_TO_TCP_SOCK_OR_NULL, /* reg points to struct tcp_sock or NULL */
405 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
406 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
407 /* PTR_TO_BTF_ID points to a kernel struct that does not need
408 * to be null checked by the BPF program. This does not imply the
409 * pointer is _not_ null and in practice this can easily be a null
410 * pointer when reading pointer chains. The assumption is program
411 * context will handle null pointer dereference typically via fault
412 * handling. The verifier must keep this in mind and can make no
413 * assumptions about null or non-null when doing branch analysis.
414 * Further, when passed into helpers the helpers can not, without
415 * additional context, assume the value is non-null.
416 */
417 PTR_TO_BTF_ID,
418 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
419 * been checked for null. Used primarily to inform the verifier
420 * an explicit null check is required for this struct.
421 */
422 PTR_TO_BTF_ID_OR_NULL,
423 PTR_TO_MEM, /* reg points to valid memory region */
424 PTR_TO_MEM_OR_NULL, /* reg points to valid memory region or NULL */
425 PTR_TO_RDONLY_BUF, /* reg points to a readonly buffer */
426 PTR_TO_RDONLY_BUF_OR_NULL, /* reg points to a readonly buffer or NULL */
427 PTR_TO_RDWR_BUF, /* reg points to a read/write buffer */
428 PTR_TO_RDWR_BUF_OR_NULL, /* reg points to a read/write buffer or NULL */
429 PTR_TO_PERCPU_BTF_ID, /* reg points to a percpu kernel variable */
430 PTR_TO_FUNC, /* reg points to a bpf program function */
431 PTR_TO_MAP_KEY, /* reg points to a map element key */
432 __BPF_REG_TYPE_MAX,
433 };
434
435 /* The information passed from prog-specific *_is_valid_access
436 * back to the verifier.
437 */
438 struct bpf_insn_access_aux {
439 enum bpf_reg_type reg_type;
440 union {
441 int ctx_field_size;
442 struct {
443 struct btf *btf;
444 u32 btf_id;
445 };
446 };
447 struct bpf_verifier_log *log; /* for verbose logs */
448 };
449
450 static inline void
451 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
452 {
453 aux->ctx_field_size = size;
454 }
455
456 struct bpf_prog_ops {
457 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
458 union bpf_attr __user *uattr);
459 };
460
461 struct bpf_verifier_ops {
462 /* return eBPF function prototype for verification */
463 const struct bpf_func_proto *
464 (*get_func_proto)(enum bpf_func_id func_id,
465 const struct bpf_prog *prog);
466
467 /* return true if 'size' wide access at offset 'off' within bpf_context
468 * with 'type' (read or write) is allowed
469 */
470 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
471 const struct bpf_prog *prog,
472 struct bpf_insn_access_aux *info);
473 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
474 const struct bpf_prog *prog);
475 int (*gen_ld_abs)(const struct bpf_insn *orig,
476 struct bpf_insn *insn_buf);
477 u32 (*convert_ctx_access)(enum bpf_access_type type,
478 const struct bpf_insn *src,
479 struct bpf_insn *dst,
480 struct bpf_prog *prog, u32 *target_size);
481 int (*btf_struct_access)(struct bpf_verifier_log *log,
482 const struct btf *btf,
483 const struct btf_type *t, int off, int size,
484 enum bpf_access_type atype,
485 u32 *next_btf_id);
486 bool (*check_kfunc_call)(u32 kfunc_btf_id);
487 };
488
489 struct bpf_prog_offload_ops {
490 /* verifier basic callbacks */
491 int (*insn_hook)(struct bpf_verifier_env *env,
492 int insn_idx, int prev_insn_idx);
493 int (*finalize)(struct bpf_verifier_env *env);
494 /* verifier optimization callbacks (called after .finalize) */
495 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
496 struct bpf_insn *insn);
497 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
498 /* program management callbacks */
499 int (*prepare)(struct bpf_prog *prog);
500 int (*translate)(struct bpf_prog *prog);
501 void (*destroy)(struct bpf_prog *prog);
502 };
503
504 struct bpf_prog_offload {
505 struct bpf_prog *prog;
506 struct net_device *netdev;
507 struct bpf_offload_dev *offdev;
508 void *dev_priv;
509 struct list_head offloads;
510 bool dev_state;
511 bool opt_failed;
512 void *jited_image;
513 u32 jited_len;
514 };
515
516 enum bpf_cgroup_storage_type {
517 BPF_CGROUP_STORAGE_SHARED,
518 BPF_CGROUP_STORAGE_PERCPU,
519 __BPF_CGROUP_STORAGE_MAX
520 };
521
522 #define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
523
524 /* The longest tracepoint has 12 args.
525 * See include/trace/bpf_probe.h
526 */
527 #define MAX_BPF_FUNC_ARGS 12
528
529 /* The maximum number of arguments passed through registers
530 * a single function may have.
531 */
532 #define MAX_BPF_FUNC_REG_ARGS 5
533
534 struct btf_func_model {
535 u8 ret_size;
536 u8 nr_args;
537 u8 arg_size[MAX_BPF_FUNC_ARGS];
538 };
539
540 /* Restore arguments before returning from trampoline to let original function
541 * continue executing. This flag is used for fentry progs when there are no
542 * fexit progs.
543 */
544 #define BPF_TRAMP_F_RESTORE_REGS BIT(0)
545 /* Call original function after fentry progs, but before fexit progs.
546 * Makes sense for fentry/fexit, normal calls and indirect calls.
547 */
548 #define BPF_TRAMP_F_CALL_ORIG BIT(1)
549 /* Skip current frame and return to parent. Makes sense for fentry/fexit
550 * programs only. Should not be used with normal calls and indirect calls.
551 */
552 #define BPF_TRAMP_F_SKIP_FRAME BIT(2)
553
554 /* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
555 * bytes on x86. Pick a number to fit into BPF_IMAGE_SIZE / 2
556 */
557 #define BPF_MAX_TRAMP_PROGS 38
558
559 struct bpf_tramp_progs {
560 struct bpf_prog *progs[BPF_MAX_TRAMP_PROGS];
561 int nr_progs;
562 };
563
564 /* Different use cases for BPF trampoline:
565 * 1. replace nop at the function entry (kprobe equivalent)
566 * flags = BPF_TRAMP_F_RESTORE_REGS
567 * fentry = a set of programs to run before returning from trampoline
568 *
569 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
570 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
571 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
572 * fentry = a set of program to run before calling original function
573 * fexit = a set of program to run after original function
574 *
575 * 3. replace direct call instruction anywhere in the function body
576 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
577 * With flags = 0
578 * fentry = a set of programs to run before returning from trampoline
579 * With flags = BPF_TRAMP_F_CALL_ORIG
580 * orig_call = original callback addr or direct function addr
581 * fentry = a set of program to run before calling original function
582 * fexit = a set of program to run after original function
583 */
584 struct bpf_tramp_image;
585 int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end,
586 const struct btf_func_model *m, u32 flags,
587 struct bpf_tramp_progs *tprogs,
588 void *orig_call);
589 /* these two functions are called from generated trampoline */
590 u64 notrace __bpf_prog_enter(struct bpf_prog *prog);
591 void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start);
592 u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog);
593 void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start);
594 void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
595 void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
596
597 struct bpf_ksym {
598 unsigned long start;
599 unsigned long end;
600 char name[KSYM_NAME_LEN];
601 struct list_head lnode;
602 struct latch_tree_node tnode;
603 bool prog;
604 };
605
606 enum bpf_tramp_prog_type {
607 BPF_TRAMP_FENTRY,
608 BPF_TRAMP_FEXIT,
609 BPF_TRAMP_MODIFY_RETURN,
610 BPF_TRAMP_MAX,
611 BPF_TRAMP_REPLACE, /* more than MAX */
612 };
613
614 struct bpf_tramp_image {
615 void *image;
616 struct bpf_ksym ksym;
617 struct percpu_ref pcref;
618 void *ip_after_call;
619 void *ip_epilogue;
620 union {
621 struct rcu_head rcu;
622 struct work_struct work;
623 };
624 };
625
626 struct bpf_trampoline {
627 /* hlist for trampoline_table */
628 struct hlist_node hlist;
629 /* serializes access to fields of this trampoline */
630 struct mutex mutex;
631 refcount_t refcnt;
632 u64 key;
633 struct {
634 struct btf_func_model model;
635 void *addr;
636 bool ftrace_managed;
637 } func;
638 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
639 * program by replacing one of its functions. func.addr is the address
640 * of the function it replaced.
641 */
642 struct bpf_prog *extension_prog;
643 /* list of BPF programs using this trampoline */
644 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
645 /* Number of attached programs. A counter per kind. */
646 int progs_cnt[BPF_TRAMP_MAX];
647 /* Executable image of trampoline */
648 struct bpf_tramp_image *cur_image;
649 u64 selector;
650 struct module *mod;
651 };
652
653 struct bpf_attach_target_info {
654 struct btf_func_model fmodel;
655 long tgt_addr;
656 const char *tgt_name;
657 const struct btf_type *tgt_type;
658 };
659
660 #define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
661
662 struct bpf_dispatcher_prog {
663 struct bpf_prog *prog;
664 refcount_t users;
665 };
666
667 struct bpf_dispatcher {
668 /* dispatcher mutex */
669 struct mutex mutex;
670 void *func;
671 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
672 int num_progs;
673 void *image;
674 u32 image_off;
675 struct bpf_ksym ksym;
676 };
677
678 static __always_inline unsigned int bpf_dispatcher_nop_func(
679 const void *ctx,
680 const struct bpf_insn *insnsi,
681 unsigned int (*bpf_func)(const void *,
682 const struct bpf_insn *))
683 {
684 return bpf_func(ctx, insnsi);
685 }
686 #ifdef CONFIG_BPF_JIT
687 int bpf_trampoline_link_prog(struct bpf_prog *prog, struct bpf_trampoline *tr);
688 int bpf_trampoline_unlink_prog(struct bpf_prog *prog, struct bpf_trampoline *tr);
689 struct bpf_trampoline *bpf_trampoline_get(u64 key,
690 struct bpf_attach_target_info *tgt_info);
691 void bpf_trampoline_put(struct bpf_trampoline *tr);
692 #define BPF_DISPATCHER_INIT(_name) { \
693 .mutex = __MUTEX_INITIALIZER(_name.mutex), \
694 .func = &_name##_func, \
695 .progs = {}, \
696 .num_progs = 0, \
697 .image = NULL, \
698 .image_off = 0, \
699 .ksym = { \
700 .name = #_name, \
701 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
702 }, \
703 }
704
705 #define DEFINE_BPF_DISPATCHER(name) \
706 noinline unsigned int bpf_dispatcher_##name##_func( \
707 const void *ctx, \
708 const struct bpf_insn *insnsi, \
709 unsigned int (*bpf_func)(const void *, \
710 const struct bpf_insn *)) \
711 { \
712 return bpf_func(ctx, insnsi); \
713 } \
714 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
715 struct bpf_dispatcher bpf_dispatcher_##name = \
716 BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
717 #define DECLARE_BPF_DISPATCHER(name) \
718 unsigned int bpf_dispatcher_##name##_func( \
719 const void *ctx, \
720 const struct bpf_insn *insnsi, \
721 unsigned int (*bpf_func)(const void *, \
722 const struct bpf_insn *)); \
723 extern struct bpf_dispatcher bpf_dispatcher_##name;
724 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
725 #define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
726 void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
727 struct bpf_prog *to);
728 /* Called only from JIT-enabled code, so there's no need for stubs. */
729 void *bpf_jit_alloc_exec_page(void);
730 void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym);
731 void bpf_image_ksym_del(struct bpf_ksym *ksym);
732 void bpf_ksym_add(struct bpf_ksym *ksym);
733 void bpf_ksym_del(struct bpf_ksym *ksym);
734 int bpf_jit_charge_modmem(u32 pages);
735 void bpf_jit_uncharge_modmem(u32 pages);
736 #else
737 static inline int bpf_trampoline_link_prog(struct bpf_prog *prog,
738 struct bpf_trampoline *tr)
739 {
740 return -ENOTSUPP;
741 }
742 static inline int bpf_trampoline_unlink_prog(struct bpf_prog *prog,
743 struct bpf_trampoline *tr)
744 {
745 return -ENOTSUPP;
746 }
747 static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
748 struct bpf_attach_target_info *tgt_info)
749 {
750 return ERR_PTR(-EOPNOTSUPP);
751 }
752 static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
753 #define DEFINE_BPF_DISPATCHER(name)
754 #define DECLARE_BPF_DISPATCHER(name)
755 #define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
756 #define BPF_DISPATCHER_PTR(name) NULL
757 static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
758 struct bpf_prog *from,
759 struct bpf_prog *to) {}
760 static inline bool is_bpf_image_address(unsigned long address)
761 {
762 return false;
763 }
764 #endif
765
766 struct bpf_func_info_aux {
767 u16 linkage;
768 bool unreliable;
769 };
770
771 enum bpf_jit_poke_reason {
772 BPF_POKE_REASON_TAIL_CALL,
773 };
774
775 /* Descriptor of pokes pointing /into/ the JITed image. */
776 struct bpf_jit_poke_descriptor {
777 void *tailcall_target;
778 void *tailcall_bypass;
779 void *bypass_addr;
780 union {
781 struct {
782 struct bpf_map *map;
783 u32 key;
784 } tail_call;
785 };
786 bool tailcall_target_stable;
787 u8 adj_off;
788 u16 reason;
789 u32 insn_idx;
790 };
791
792 /* reg_type info for ctx arguments */
793 struct bpf_ctx_arg_aux {
794 u32 offset;
795 enum bpf_reg_type reg_type;
796 u32 btf_id;
797 };
798
799 struct btf_mod_pair {
800 struct btf *btf;
801 struct module *module;
802 };
803
804 struct bpf_kfunc_desc_tab;
805
806 struct bpf_prog_aux {
807 atomic64_t refcnt;
808 u32 used_map_cnt;
809 u32 used_btf_cnt;
810 u32 max_ctx_offset;
811 u32 max_pkt_offset;
812 u32 max_tp_access;
813 u32 stack_depth;
814 u32 id;
815 u32 func_cnt; /* used by non-func prog as the number of func progs */
816 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
817 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
818 u32 ctx_arg_info_size;
819 u32 max_rdonly_access;
820 u32 max_rdwr_access;
821 struct btf *attach_btf;
822 const struct bpf_ctx_arg_aux *ctx_arg_info;
823 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
824 struct bpf_prog *dst_prog;
825 struct bpf_trampoline *dst_trampoline;
826 enum bpf_prog_type saved_dst_prog_type;
827 enum bpf_attach_type saved_dst_attach_type;
828 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
829 bool offload_requested;
830 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
831 bool func_proto_unreliable;
832 bool sleepable;
833 bool tail_call_reachable;
834 struct hlist_node tramp_hlist;
835 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
836 const struct btf_type *attach_func_proto;
837 /* function name for valid attach_btf_id */
838 const char *attach_func_name;
839 struct bpf_prog **func;
840 void *jit_data; /* JIT specific data. arch dependent */
841 struct bpf_jit_poke_descriptor *poke_tab;
842 struct bpf_kfunc_desc_tab *kfunc_tab;
843 u32 size_poke_tab;
844 struct bpf_ksym ksym;
845 const struct bpf_prog_ops *ops;
846 struct bpf_map **used_maps;
847 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
848 struct btf_mod_pair *used_btfs;
849 struct bpf_prog *prog;
850 struct user_struct *user;
851 u64 load_time; /* ns since boottime */
852 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
853 char name[BPF_OBJ_NAME_LEN];
854 #ifdef CONFIG_SECURITY
855 void *security;
856 #endif
857 struct bpf_prog_offload *offload;
858 struct btf *btf;
859 struct bpf_func_info *func_info;
860 struct bpf_func_info_aux *func_info_aux;
861 /* bpf_line_info loaded from userspace. linfo->insn_off
862 * has the xlated insn offset.
863 * Both the main and sub prog share the same linfo.
864 * The subprog can access its first linfo by
865 * using the linfo_idx.
866 */
867 struct bpf_line_info *linfo;
868 /* jited_linfo is the jited addr of the linfo. It has a
869 * one to one mapping to linfo:
870 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
871 * Both the main and sub prog share the same jited_linfo.
872 * The subprog can access its first jited_linfo by
873 * using the linfo_idx.
874 */
875 void **jited_linfo;
876 u32 func_info_cnt;
877 u32 nr_linfo;
878 /* subprog can use linfo_idx to access its first linfo and
879 * jited_linfo.
880 * main prog always has linfo_idx == 0
881 */
882 u32 linfo_idx;
883 u32 num_exentries;
884 struct exception_table_entry *extable;
885 union {
886 struct work_struct work;
887 struct rcu_head rcu;
888 };
889 };
890
891 struct bpf_array_aux {
892 /* 'Ownership' of prog array is claimed by the first program that
893 * is going to use this map or by the first program which FD is
894 * stored in the map to make sure that all callers and callees have
895 * the same prog type and JITed flag.
896 */
897 enum bpf_prog_type type;
898 bool jited;
899 /* Programs with direct jumps into programs part of this array. */
900 struct list_head poke_progs;
901 struct bpf_map *map;
902 struct mutex poke_mutex;
903 struct work_struct work;
904 };
905
906 struct bpf_link {
907 atomic64_t refcnt;
908 u32 id;
909 enum bpf_link_type type;
910 const struct bpf_link_ops *ops;
911 struct bpf_prog *prog;
912 struct work_struct work;
913 };
914
915 struct bpf_link_ops {
916 void (*release)(struct bpf_link *link);
917 void (*dealloc)(struct bpf_link *link);
918 int (*detach)(struct bpf_link *link);
919 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
920 struct bpf_prog *old_prog);
921 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
922 int (*fill_link_info)(const struct bpf_link *link,
923 struct bpf_link_info *info);
924 };
925
926 struct bpf_link_primer {
927 struct bpf_link *link;
928 struct file *file;
929 int fd;
930 u32 id;
931 };
932
933 struct bpf_struct_ops_value;
934 struct btf_member;
935
936 #define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
937 struct bpf_struct_ops {
938 const struct bpf_verifier_ops *verifier_ops;
939 int (*init)(struct btf *btf);
940 int (*check_member)(const struct btf_type *t,
941 const struct btf_member *member);
942 int (*init_member)(const struct btf_type *t,
943 const struct btf_member *member,
944 void *kdata, const void *udata);
945 int (*reg)(void *kdata);
946 void (*unreg)(void *kdata);
947 const struct btf_type *type;
948 const struct btf_type *value_type;
949 const char *name;
950 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
951 u32 type_id;
952 u32 value_id;
953 };
954
955 #if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
956 #define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
957 const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id);
958 void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log);
959 bool bpf_struct_ops_get(const void *kdata);
960 void bpf_struct_ops_put(const void *kdata);
961 int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
962 void *value);
963 static inline bool bpf_try_module_get(const void *data, struct module *owner)
964 {
965 if (owner == BPF_MODULE_OWNER)
966 return bpf_struct_ops_get(data);
967 else
968 return try_module_get(owner);
969 }
970 static inline void bpf_module_put(const void *data, struct module *owner)
971 {
972 if (owner == BPF_MODULE_OWNER)
973 bpf_struct_ops_put(data);
974 else
975 module_put(owner);
976 }
977 #else
978 static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id)
979 {
980 return NULL;
981 }
982 static inline void bpf_struct_ops_init(struct btf *btf,
983 struct bpf_verifier_log *log)
984 {
985 }
986 static inline bool bpf_try_module_get(const void *data, struct module *owner)
987 {
988 return try_module_get(owner);
989 }
990 static inline void bpf_module_put(const void *data, struct module *owner)
991 {
992 module_put(owner);
993 }
994 static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
995 void *key,
996 void *value)
997 {
998 return -EINVAL;
999 }
1000 #endif
1001
1002 struct bpf_array {
1003 struct bpf_map map;
1004 u32 elem_size;
1005 u32 index_mask;
1006 struct bpf_array_aux *aux;
1007 union {
1008 char value[0] __aligned(8);
1009 void *ptrs[0] __aligned(8);
1010 void __percpu *pptrs[0] __aligned(8);
1011 };
1012 };
1013
1014 #define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
1015 #define MAX_TAIL_CALL_CNT 32
1016
1017 #define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
1018 BPF_F_RDONLY_PROG | \
1019 BPF_F_WRONLY | \
1020 BPF_F_WRONLY_PROG)
1021
1022 #define BPF_MAP_CAN_READ BIT(0)
1023 #define BPF_MAP_CAN_WRITE BIT(1)
1024
1025 static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
1026 {
1027 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1028
1029 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
1030 * not possible.
1031 */
1032 if (access_flags & BPF_F_RDONLY_PROG)
1033 return BPF_MAP_CAN_READ;
1034 else if (access_flags & BPF_F_WRONLY_PROG)
1035 return BPF_MAP_CAN_WRITE;
1036 else
1037 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
1038 }
1039
1040 static inline bool bpf_map_flags_access_ok(u32 access_flags)
1041 {
1042 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
1043 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1044 }
1045
1046 struct bpf_event_entry {
1047 struct perf_event *event;
1048 struct file *perf_file;
1049 struct file *map_file;
1050 struct rcu_head rcu;
1051 };
1052
1053 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
1054 int bpf_prog_calc_tag(struct bpf_prog *fp);
1055
1056 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
1057
1058 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
1059 unsigned long off, unsigned long len);
1060 typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
1061 const struct bpf_insn *src,
1062 struct bpf_insn *dst,
1063 struct bpf_prog *prog,
1064 u32 *target_size);
1065
1066 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
1067 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
1068
1069 /* an array of programs to be executed under rcu_lock.
1070 *
1071 * Typical usage:
1072 * ret = BPF_PROG_RUN_ARRAY(&bpf_prog_array, ctx, BPF_PROG_RUN);
1073 *
1074 * the structure returned by bpf_prog_array_alloc() should be populated
1075 * with program pointers and the last pointer must be NULL.
1076 * The user has to keep refcnt on the program and make sure the program
1077 * is removed from the array before bpf_prog_put().
1078 * The 'struct bpf_prog_array *' should only be replaced with xchg()
1079 * since other cpus are walking the array of pointers in parallel.
1080 */
1081 struct bpf_prog_array_item {
1082 struct bpf_prog *prog;
1083 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1084 };
1085
1086 struct bpf_prog_array {
1087 struct rcu_head rcu;
1088 struct bpf_prog_array_item items[];
1089 };
1090
1091 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
1092 void bpf_prog_array_free(struct bpf_prog_array *progs);
1093 int bpf_prog_array_length(struct bpf_prog_array *progs);
1094 bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
1095 int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
1096 __u32 __user *prog_ids, u32 cnt);
1097
1098 void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
1099 struct bpf_prog *old_prog);
1100 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
1101 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
1102 struct bpf_prog *prog);
1103 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
1104 u32 *prog_ids, u32 request_cnt,
1105 u32 *prog_cnt);
1106 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
1107 struct bpf_prog *exclude_prog,
1108 struct bpf_prog *include_prog,
1109 struct bpf_prog_array **new_array);
1110
1111 /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
1112 #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0)
1113 /* BPF program asks to set CN on the packet. */
1114 #define BPF_RET_SET_CN (1 << 0)
1115
1116 /* For BPF_PROG_RUN_ARRAY_FLAGS and __BPF_PROG_RUN_ARRAY,
1117 * if bpf_cgroup_storage_set() failed, the rest of programs
1118 * will not execute. This should be a really rare scenario
1119 * as it requires BPF_CGROUP_STORAGE_NEST_MAX number of
1120 * preemptions all between bpf_cgroup_storage_set() and
1121 * bpf_cgroup_storage_unset() on the same cpu.
1122 */
1123 #define BPF_PROG_RUN_ARRAY_FLAGS(array, ctx, func, ret_flags) \
1124 ({ \
1125 struct bpf_prog_array_item *_item; \
1126 struct bpf_prog *_prog; \
1127 struct bpf_prog_array *_array; \
1128 u32 _ret = 1; \
1129 u32 func_ret; \
1130 migrate_disable(); \
1131 rcu_read_lock(); \
1132 _array = rcu_dereference(array); \
1133 _item = &_array->items[0]; \
1134 while ((_prog = READ_ONCE(_item->prog))) { \
1135 if (unlikely(bpf_cgroup_storage_set(_item->cgroup_storage))) \
1136 break; \
1137 func_ret = func(_prog, ctx); \
1138 _ret &= (func_ret & 1); \
1139 *(ret_flags) |= (func_ret >> 1); \
1140 bpf_cgroup_storage_unset(); \
1141 _item++; \
1142 } \
1143 rcu_read_unlock(); \
1144 migrate_enable(); \
1145 _ret; \
1146 })
1147
1148 #define __BPF_PROG_RUN_ARRAY(array, ctx, func, check_non_null, set_cg_storage) \
1149 ({ \
1150 struct bpf_prog_array_item *_item; \
1151 struct bpf_prog *_prog; \
1152 struct bpf_prog_array *_array; \
1153 u32 _ret = 1; \
1154 migrate_disable(); \
1155 rcu_read_lock(); \
1156 _array = rcu_dereference(array); \
1157 if (unlikely(check_non_null && !_array))\
1158 goto _out; \
1159 _item = &_array->items[0]; \
1160 while ((_prog = READ_ONCE(_item->prog))) { \
1161 if (!set_cg_storage) { \
1162 _ret &= func(_prog, ctx); \
1163 } else { \
1164 if (unlikely(bpf_cgroup_storage_set(_item->cgroup_storage))) \
1165 break; \
1166 _ret &= func(_prog, ctx); \
1167 bpf_cgroup_storage_unset(); \
1168 } \
1169 _item++; \
1170 } \
1171 _out: \
1172 rcu_read_unlock(); \
1173 migrate_enable(); \
1174 _ret; \
1175 })
1176
1177 /* To be used by __cgroup_bpf_run_filter_skb for EGRESS BPF progs
1178 * so BPF programs can request cwr for TCP packets.
1179 *
1180 * Current cgroup skb programs can only return 0 or 1 (0 to drop the
1181 * packet. This macro changes the behavior so the low order bit
1182 * indicates whether the packet should be dropped (0) or not (1)
1183 * and the next bit is a congestion notification bit. This could be
1184 * used by TCP to call tcp_enter_cwr()
1185 *
1186 * Hence, new allowed return values of CGROUP EGRESS BPF programs are:
1187 * 0: drop packet
1188 * 1: keep packet
1189 * 2: drop packet and cn
1190 * 3: keep packet and cn
1191 *
1192 * This macro then converts it to one of the NET_XMIT or an error
1193 * code that is then interpreted as drop packet (and no cn):
1194 * 0: NET_XMIT_SUCCESS skb should be transmitted
1195 * 1: NET_XMIT_DROP skb should be dropped and cn
1196 * 2: NET_XMIT_CN skb should be transmitted and cn
1197 * 3: -EPERM skb should be dropped
1198 */
1199 #define BPF_PROG_CGROUP_INET_EGRESS_RUN_ARRAY(array, ctx, func) \
1200 ({ \
1201 u32 _flags = 0; \
1202 bool _cn; \
1203 u32 _ret; \
1204 _ret = BPF_PROG_RUN_ARRAY_FLAGS(array, ctx, func, &_flags); \
1205 _cn = _flags & BPF_RET_SET_CN; \
1206 if (_ret) \
1207 _ret = (_cn ? NET_XMIT_CN : NET_XMIT_SUCCESS); \
1208 else \
1209 _ret = (_cn ? NET_XMIT_DROP : -EPERM); \
1210 _ret; \
1211 })
1212
1213 #define BPF_PROG_RUN_ARRAY(array, ctx, func) \
1214 __BPF_PROG_RUN_ARRAY(array, ctx, func, false, true)
1215
1216 #define BPF_PROG_RUN_ARRAY_CHECK(array, ctx, func) \
1217 __BPF_PROG_RUN_ARRAY(array, ctx, func, true, false)
1218
1219 #ifdef CONFIG_BPF_SYSCALL
1220 DECLARE_PER_CPU(int, bpf_prog_active);
1221 extern struct mutex bpf_stats_enabled_mutex;
1222
1223 /*
1224 * Block execution of BPF programs attached to instrumentation (perf,
1225 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
1226 * these events can happen inside a region which holds a map bucket lock
1227 * and can deadlock on it.
1228 *
1229 * Use the preemption safe inc/dec variants on RT because migrate disable
1230 * is preemptible on RT and preemption in the middle of the RMW operation
1231 * might lead to inconsistent state. Use the raw variants for non RT
1232 * kernels as migrate_disable() maps to preempt_disable() so the slightly
1233 * more expensive save operation can be avoided.
1234 */
1235 static inline void bpf_disable_instrumentation(void)
1236 {
1237 migrate_disable();
1238 if (IS_ENABLED(CONFIG_PREEMPT_RT))
1239 this_cpu_inc(bpf_prog_active);
1240 else
1241 __this_cpu_inc(bpf_prog_active);
1242 }
1243
1244 static inline void bpf_enable_instrumentation(void)
1245 {
1246 if (IS_ENABLED(CONFIG_PREEMPT_RT))
1247 this_cpu_dec(bpf_prog_active);
1248 else
1249 __this_cpu_dec(bpf_prog_active);
1250 migrate_enable();
1251 }
1252
1253 extern const struct file_operations bpf_map_fops;
1254 extern const struct file_operations bpf_prog_fops;
1255 extern const struct file_operations bpf_iter_fops;
1256
1257 #define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
1258 extern const struct bpf_prog_ops _name ## _prog_ops; \
1259 extern const struct bpf_verifier_ops _name ## _verifier_ops;
1260 #define BPF_MAP_TYPE(_id, _ops) \
1261 extern const struct bpf_map_ops _ops;
1262 #define BPF_LINK_TYPE(_id, _name)
1263 #include <linux/bpf_types.h>
1264 #undef BPF_PROG_TYPE
1265 #undef BPF_MAP_TYPE
1266 #undef BPF_LINK_TYPE
1267
1268 extern const struct bpf_prog_ops bpf_offload_prog_ops;
1269 extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
1270 extern const struct bpf_verifier_ops xdp_analyzer_ops;
1271
1272 struct bpf_prog *bpf_prog_get(u32 ufd);
1273 struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
1274 bool attach_drv);
1275 void bpf_prog_add(struct bpf_prog *prog, int i);
1276 void bpf_prog_sub(struct bpf_prog *prog, int i);
1277 void bpf_prog_inc(struct bpf_prog *prog);
1278 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
1279 void bpf_prog_put(struct bpf_prog *prog);
1280
1281 void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
1282 void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
1283
1284 struct bpf_map *bpf_map_get(u32 ufd);
1285 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
1286 struct bpf_map *__bpf_map_get(struct fd f);
1287 void bpf_map_inc(struct bpf_map *map);
1288 void bpf_map_inc_with_uref(struct bpf_map *map);
1289 struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
1290 void bpf_map_put_with_uref(struct bpf_map *map);
1291 void bpf_map_put(struct bpf_map *map);
1292 void *bpf_map_area_alloc(u64 size, int numa_node);
1293 void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
1294 void bpf_map_area_free(void *base);
1295 void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
1296 int generic_map_lookup_batch(struct bpf_map *map,
1297 const union bpf_attr *attr,
1298 union bpf_attr __user *uattr);
1299 int generic_map_update_batch(struct bpf_map *map,
1300 const union bpf_attr *attr,
1301 union bpf_attr __user *uattr);
1302 int generic_map_delete_batch(struct bpf_map *map,
1303 const union bpf_attr *attr,
1304 union bpf_attr __user *uattr);
1305 struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
1306 struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
1307
1308 #ifdef CONFIG_MEMCG_KMEM
1309 void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
1310 int node);
1311 void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags);
1312 void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
1313 size_t align, gfp_t flags);
1314 #else
1315 static inline void *
1316 bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
1317 int node)
1318 {
1319 return kmalloc_node(size, flags, node);
1320 }
1321
1322 static inline void *
1323 bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags)
1324 {
1325 return kzalloc(size, flags);
1326 }
1327
1328 static inline void __percpu *
1329 bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align,
1330 gfp_t flags)
1331 {
1332 return __alloc_percpu_gfp(size, align, flags);
1333 }
1334 #endif
1335
1336 extern int sysctl_unprivileged_bpf_disabled;
1337
1338 static inline bool bpf_allow_ptr_leaks(void)
1339 {
1340 return perfmon_capable();
1341 }
1342
1343 static inline bool bpf_allow_uninit_stack(void)
1344 {
1345 return perfmon_capable();
1346 }
1347
1348 static inline bool bpf_allow_ptr_to_map_access(void)
1349 {
1350 return perfmon_capable();
1351 }
1352
1353 static inline bool bpf_bypass_spec_v1(void)
1354 {
1355 return perfmon_capable();
1356 }
1357
1358 static inline bool bpf_bypass_spec_v4(void)
1359 {
1360 return perfmon_capable();
1361 }
1362
1363 int bpf_map_new_fd(struct bpf_map *map, int flags);
1364 int bpf_prog_new_fd(struct bpf_prog *prog);
1365
1366 void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1367 const struct bpf_link_ops *ops, struct bpf_prog *prog);
1368 int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
1369 int bpf_link_settle(struct bpf_link_primer *primer);
1370 void bpf_link_cleanup(struct bpf_link_primer *primer);
1371 void bpf_link_inc(struct bpf_link *link);
1372 void bpf_link_put(struct bpf_link *link);
1373 int bpf_link_new_fd(struct bpf_link *link);
1374 struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd);
1375 struct bpf_link *bpf_link_get_from_fd(u32 ufd);
1376
1377 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
1378 int bpf_obj_get_user(const char __user *pathname, int flags);
1379
1380 #define BPF_ITER_FUNC_PREFIX "bpf_iter_"
1381 #define DEFINE_BPF_ITER_FUNC(target, args...) \
1382 extern int bpf_iter_ ## target(args); \
1383 int __init bpf_iter_ ## target(args) { return 0; }
1384
1385 struct bpf_iter_aux_info {
1386 struct bpf_map *map;
1387 };
1388
1389 typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
1390 union bpf_iter_link_info *linfo,
1391 struct bpf_iter_aux_info *aux);
1392 typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
1393 typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
1394 struct seq_file *seq);
1395 typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
1396 struct bpf_link_info *info);
1397
1398 enum bpf_iter_feature {
1399 BPF_ITER_RESCHED = BIT(0),
1400 };
1401
1402 #define BPF_ITER_CTX_ARG_MAX 2
1403 struct bpf_iter_reg {
1404 const char *target;
1405 bpf_iter_attach_target_t attach_target;
1406 bpf_iter_detach_target_t detach_target;
1407 bpf_iter_show_fdinfo_t show_fdinfo;
1408 bpf_iter_fill_link_info_t fill_link_info;
1409 u32 ctx_arg_info_size;
1410 u32 feature;
1411 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
1412 const struct bpf_iter_seq_info *seq_info;
1413 };
1414
1415 struct bpf_iter_meta {
1416 __bpf_md_ptr(struct seq_file *, seq);
1417 u64 session_id;
1418 u64 seq_num;
1419 };
1420
1421 struct bpf_iter__bpf_map_elem {
1422 __bpf_md_ptr(struct bpf_iter_meta *, meta);
1423 __bpf_md_ptr(struct bpf_map *, map);
1424 __bpf_md_ptr(void *, key);
1425 __bpf_md_ptr(void *, value);
1426 };
1427
1428 int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
1429 void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
1430 bool bpf_iter_prog_supported(struct bpf_prog *prog);
1431 int bpf_iter_link_attach(const union bpf_attr *attr, struct bpf_prog *prog);
1432 int bpf_iter_new_fd(struct bpf_link *link);
1433 bool bpf_link_is_iter(struct bpf_link *link);
1434 struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
1435 int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
1436 void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
1437 struct seq_file *seq);
1438 int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
1439 struct bpf_link_info *info);
1440
1441 int map_set_for_each_callback_args(struct bpf_verifier_env *env,
1442 struct bpf_func_state *caller,
1443 struct bpf_func_state *callee);
1444
1445 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
1446 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
1447 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
1448 u64 flags);
1449 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
1450 u64 flags);
1451
1452 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
1453
1454 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
1455 void *key, void *value, u64 map_flags);
1456 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1457 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
1458 void *key, void *value, u64 map_flags);
1459 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1460
1461 int bpf_get_file_flag(int flags);
1462 int bpf_check_uarg_tail_zero(void __user *uaddr, size_t expected_size,
1463 size_t actual_size);
1464
1465 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
1466 * forced to use 'long' read/writes to try to atomically copy long counters.
1467 * Best-effort only. No barriers here, since it _will_ race with concurrent
1468 * updates from BPF programs. Called from bpf syscall and mostly used with
1469 * size 8 or 16 bytes, so ask compiler to inline it.
1470 */
1471 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
1472 {
1473 const long *lsrc = src;
1474 long *ldst = dst;
1475
1476 size /= sizeof(long);
1477 while (size--)
1478 *ldst++ = *lsrc++;
1479 }
1480
1481 /* verify correctness of eBPF program */
1482 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr,
1483 union bpf_attr __user *uattr);
1484
1485 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
1486 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
1487 #endif
1488
1489 struct btf *bpf_get_btf_vmlinux(void);
1490
1491 /* Map specifics */
1492 struct xdp_buff;
1493 struct sk_buff;
1494 struct bpf_dtab_netdev;
1495 struct bpf_cpu_map_entry;
1496
1497 void __dev_flush(void);
1498 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp,
1499 struct net_device *dev_rx);
1500 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
1501 struct net_device *dev_rx);
1502 int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
1503 struct bpf_prog *xdp_prog);
1504 bool dev_map_can_have_prog(struct bpf_map *map);
1505
1506 void __cpu_map_flush(void);
1507 int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_buff *xdp,
1508 struct net_device *dev_rx);
1509 bool cpu_map_prog_allowed(struct bpf_map *map);
1510
1511 /* Return map's numa specified by userspace */
1512 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
1513 {
1514 return (attr->map_flags & BPF_F_NUMA_NODE) ?
1515 attr->numa_node : NUMA_NO_NODE;
1516 }
1517
1518 struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
1519 int array_map_alloc_check(union bpf_attr *attr);
1520
1521 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
1522 union bpf_attr __user *uattr);
1523 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
1524 union bpf_attr __user *uattr);
1525 int bpf_prog_test_run_tracing(struct bpf_prog *prog,
1526 const union bpf_attr *kattr,
1527 union bpf_attr __user *uattr);
1528 int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1529 const union bpf_attr *kattr,
1530 union bpf_attr __user *uattr);
1531 int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
1532 const union bpf_attr *kattr,
1533 union bpf_attr __user *uattr);
1534 int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
1535 const union bpf_attr *kattr,
1536 union bpf_attr __user *uattr);
1537 bool bpf_prog_test_check_kfunc_call(u32 kfunc_id);
1538 bool btf_ctx_access(int off, int size, enum bpf_access_type type,
1539 const struct bpf_prog *prog,
1540 struct bpf_insn_access_aux *info);
1541 int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf,
1542 const struct btf_type *t, int off, int size,
1543 enum bpf_access_type atype,
1544 u32 *next_btf_id);
1545 bool btf_struct_ids_match(struct bpf_verifier_log *log,
1546 const struct btf *btf, u32 id, int off,
1547 const struct btf *need_btf, u32 need_type_id);
1548
1549 int btf_distill_func_proto(struct bpf_verifier_log *log,
1550 struct btf *btf,
1551 const struct btf_type *func_proto,
1552 const char *func_name,
1553 struct btf_func_model *m);
1554
1555 struct bpf_reg_state;
1556 int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog,
1557 struct bpf_reg_state *regs);
1558 int btf_check_kfunc_arg_match(struct bpf_verifier_env *env,
1559 const struct btf *btf, u32 func_id,
1560 struct bpf_reg_state *regs);
1561 int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog,
1562 struct bpf_reg_state *reg);
1563 int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
1564 struct btf *btf, const struct btf_type *t);
1565
1566 struct bpf_prog *bpf_prog_by_id(u32 id);
1567 struct bpf_link *bpf_link_by_id(u32 id);
1568
1569 const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
1570 void bpf_task_storage_free(struct task_struct *task);
1571 bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
1572 const struct btf_func_model *
1573 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
1574 const struct bpf_insn *insn);
1575 #else /* !CONFIG_BPF_SYSCALL */
1576 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
1577 {
1578 return ERR_PTR(-EOPNOTSUPP);
1579 }
1580
1581 static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
1582 enum bpf_prog_type type,
1583 bool attach_drv)
1584 {
1585 return ERR_PTR(-EOPNOTSUPP);
1586 }
1587
1588 static inline void bpf_prog_add(struct bpf_prog *prog, int i)
1589 {
1590 }
1591
1592 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
1593 {
1594 }
1595
1596 static inline void bpf_prog_put(struct bpf_prog *prog)
1597 {
1598 }
1599
1600 static inline void bpf_prog_inc(struct bpf_prog *prog)
1601 {
1602 }
1603
1604 static inline struct bpf_prog *__must_check
1605 bpf_prog_inc_not_zero(struct bpf_prog *prog)
1606 {
1607 return ERR_PTR(-EOPNOTSUPP);
1608 }
1609
1610 static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1611 const struct bpf_link_ops *ops,
1612 struct bpf_prog *prog)
1613 {
1614 }
1615
1616 static inline int bpf_link_prime(struct bpf_link *link,
1617 struct bpf_link_primer *primer)
1618 {
1619 return -EOPNOTSUPP;
1620 }
1621
1622 static inline int bpf_link_settle(struct bpf_link_primer *primer)
1623 {
1624 return -EOPNOTSUPP;
1625 }
1626
1627 static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
1628 {
1629 }
1630
1631 static inline void bpf_link_inc(struct bpf_link *link)
1632 {
1633 }
1634
1635 static inline void bpf_link_put(struct bpf_link *link)
1636 {
1637 }
1638
1639 static inline int bpf_obj_get_user(const char __user *pathname, int flags)
1640 {
1641 return -EOPNOTSUPP;
1642 }
1643
1644 static inline bool dev_map_can_have_prog(struct bpf_map *map)
1645 {
1646 return false;
1647 }
1648
1649 static inline void __dev_flush(void)
1650 {
1651 }
1652
1653 struct xdp_buff;
1654 struct bpf_dtab_netdev;
1655 struct bpf_cpu_map_entry;
1656
1657 static inline
1658 int dev_xdp_enqueue(struct net_device *dev, struct xdp_buff *xdp,
1659 struct net_device *dev_rx)
1660 {
1661 return 0;
1662 }
1663
1664 static inline
1665 int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp,
1666 struct net_device *dev_rx)
1667 {
1668 return 0;
1669 }
1670
1671 struct sk_buff;
1672
1673 static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
1674 struct sk_buff *skb,
1675 struct bpf_prog *xdp_prog)
1676 {
1677 return 0;
1678 }
1679
1680 static inline void __cpu_map_flush(void)
1681 {
1682 }
1683
1684 static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
1685 struct xdp_buff *xdp,
1686 struct net_device *dev_rx)
1687 {
1688 return 0;
1689 }
1690
1691 static inline bool cpu_map_prog_allowed(struct bpf_map *map)
1692 {
1693 return false;
1694 }
1695
1696 static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
1697 enum bpf_prog_type type)
1698 {
1699 return ERR_PTR(-EOPNOTSUPP);
1700 }
1701
1702 static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
1703 const union bpf_attr *kattr,
1704 union bpf_attr __user *uattr)
1705 {
1706 return -ENOTSUPP;
1707 }
1708
1709 static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
1710 const union bpf_attr *kattr,
1711 union bpf_attr __user *uattr)
1712 {
1713 return -ENOTSUPP;
1714 }
1715
1716 static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
1717 const union bpf_attr *kattr,
1718 union bpf_attr __user *uattr)
1719 {
1720 return -ENOTSUPP;
1721 }
1722
1723 static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1724 const union bpf_attr *kattr,
1725 union bpf_attr __user *uattr)
1726 {
1727 return -ENOTSUPP;
1728 }
1729
1730 static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
1731 const union bpf_attr *kattr,
1732 union bpf_attr __user *uattr)
1733 {
1734 return -ENOTSUPP;
1735 }
1736
1737 static inline bool bpf_prog_test_check_kfunc_call(u32 kfunc_id)
1738 {
1739 return false;
1740 }
1741
1742 static inline void bpf_map_put(struct bpf_map *map)
1743 {
1744 }
1745
1746 static inline struct bpf_prog *bpf_prog_by_id(u32 id)
1747 {
1748 return ERR_PTR(-ENOTSUPP);
1749 }
1750
1751 static inline const struct bpf_func_proto *
1752 bpf_base_func_proto(enum bpf_func_id func_id)
1753 {
1754 return NULL;
1755 }
1756
1757 static inline void bpf_task_storage_free(struct task_struct *task)
1758 {
1759 }
1760
1761 static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog)
1762 {
1763 return false;
1764 }
1765
1766 static inline const struct btf_func_model *
1767 bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
1768 const struct bpf_insn *insn)
1769 {
1770 return NULL;
1771 }
1772 #endif /* CONFIG_BPF_SYSCALL */
1773
1774 void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
1775 struct btf_mod_pair *used_btfs, u32 len);
1776
1777 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
1778 enum bpf_prog_type type)
1779 {
1780 return bpf_prog_get_type_dev(ufd, type, false);
1781 }
1782
1783 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
1784 struct bpf_map **used_maps, u32 len);
1785
1786 bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
1787
1788 int bpf_prog_offload_compile(struct bpf_prog *prog);
1789 void bpf_prog_offload_destroy(struct bpf_prog *prog);
1790 int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
1791 struct bpf_prog *prog);
1792
1793 int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
1794
1795 int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
1796 int bpf_map_offload_update_elem(struct bpf_map *map,
1797 void *key, void *value, u64 flags);
1798 int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
1799 int bpf_map_offload_get_next_key(struct bpf_map *map,
1800 void *key, void *next_key);
1801
1802 bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
1803
1804 struct bpf_offload_dev *
1805 bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
1806 void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
1807 void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
1808 int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
1809 struct net_device *netdev);
1810 void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
1811 struct net_device *netdev);
1812 bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
1813
1814 #if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
1815 int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
1816
1817 static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
1818 {
1819 return aux->offload_requested;
1820 }
1821
1822 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1823 {
1824 return unlikely(map->ops == &bpf_map_offload_ops);
1825 }
1826
1827 struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
1828 void bpf_map_offload_map_free(struct bpf_map *map);
1829 #else
1830 static inline int bpf_prog_offload_init(struct bpf_prog *prog,
1831 union bpf_attr *attr)
1832 {
1833 return -EOPNOTSUPP;
1834 }
1835
1836 static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
1837 {
1838 return false;
1839 }
1840
1841 static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
1842 {
1843 return false;
1844 }
1845
1846 static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
1847 {
1848 return ERR_PTR(-EOPNOTSUPP);
1849 }
1850
1851 static inline void bpf_map_offload_map_free(struct bpf_map *map)
1852 {
1853 }
1854 #endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
1855
1856 #if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
1857 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
1858 int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
1859 int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
1860 void sock_map_unhash(struct sock *sk);
1861 void sock_map_close(struct sock *sk, long timeout);
1862
1863 void bpf_sk_reuseport_detach(struct sock *sk);
1864 int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
1865 void *value);
1866 int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
1867 void *value, u64 map_flags);
1868 #else
1869 static inline void bpf_sk_reuseport_detach(struct sock *sk)
1870 {
1871 }
1872
1873 #ifdef CONFIG_BPF_SYSCALL
1874 static inline int sock_map_get_from_fd(const union bpf_attr *attr,
1875 struct bpf_prog *prog)
1876 {
1877 return -EINVAL;
1878 }
1879
1880 static inline int sock_map_prog_detach(const union bpf_attr *attr,
1881 enum bpf_prog_type ptype)
1882 {
1883 return -EOPNOTSUPP;
1884 }
1885
1886 static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
1887 u64 flags)
1888 {
1889 return -EOPNOTSUPP;
1890 }
1891
1892 static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
1893 void *key, void *value)
1894 {
1895 return -EOPNOTSUPP;
1896 }
1897
1898 static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
1899 void *key, void *value,
1900 u64 map_flags)
1901 {
1902 return -EOPNOTSUPP;
1903 }
1904 #endif /* CONFIG_BPF_SYSCALL */
1905 #endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
1906
1907 /* verifier prototypes for helper functions called from eBPF programs */
1908 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
1909 extern const struct bpf_func_proto bpf_map_update_elem_proto;
1910 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
1911 extern const struct bpf_func_proto bpf_map_push_elem_proto;
1912 extern const struct bpf_func_proto bpf_map_pop_elem_proto;
1913 extern const struct bpf_func_proto bpf_map_peek_elem_proto;
1914
1915 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
1916 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
1917 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
1918 extern const struct bpf_func_proto bpf_tail_call_proto;
1919 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
1920 extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
1921 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
1922 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
1923 extern const struct bpf_func_proto bpf_get_current_comm_proto;
1924 extern const struct bpf_func_proto bpf_get_stackid_proto;
1925 extern const struct bpf_func_proto bpf_get_stack_proto;
1926 extern const struct bpf_func_proto bpf_get_task_stack_proto;
1927 extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
1928 extern const struct bpf_func_proto bpf_get_stack_proto_pe;
1929 extern const struct bpf_func_proto bpf_sock_map_update_proto;
1930 extern const struct bpf_func_proto bpf_sock_hash_update_proto;
1931 extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
1932 extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
1933 extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
1934 extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
1935 extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
1936 extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
1937 extern const struct bpf_func_proto bpf_spin_lock_proto;
1938 extern const struct bpf_func_proto bpf_spin_unlock_proto;
1939 extern const struct bpf_func_proto bpf_get_local_storage_proto;
1940 extern const struct bpf_func_proto bpf_strtol_proto;
1941 extern const struct bpf_func_proto bpf_strtoul_proto;
1942 extern const struct bpf_func_proto bpf_tcp_sock_proto;
1943 extern const struct bpf_func_proto bpf_jiffies64_proto;
1944 extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
1945 extern const struct bpf_func_proto bpf_event_output_data_proto;
1946 extern const struct bpf_func_proto bpf_ringbuf_output_proto;
1947 extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
1948 extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
1949 extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
1950 extern const struct bpf_func_proto bpf_ringbuf_query_proto;
1951 extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
1952 extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
1953 extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
1954 extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
1955 extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
1956 extern const struct bpf_func_proto bpf_copy_from_user_proto;
1957 extern const struct bpf_func_proto bpf_snprintf_btf_proto;
1958 extern const struct bpf_func_proto bpf_snprintf_proto;
1959 extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
1960 extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
1961 extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto;
1962 extern const struct bpf_func_proto bpf_sock_from_file_proto;
1963 extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto;
1964 extern const struct bpf_func_proto bpf_task_storage_get_proto;
1965 extern const struct bpf_func_proto bpf_task_storage_delete_proto;
1966 extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
1967
1968 const struct bpf_func_proto *bpf_tracing_func_proto(
1969 enum bpf_func_id func_id, const struct bpf_prog *prog);
1970
1971 const struct bpf_func_proto *tracing_prog_func_proto(
1972 enum bpf_func_id func_id, const struct bpf_prog *prog);
1973
1974 /* Shared helpers among cBPF and eBPF. */
1975 void bpf_user_rnd_init_once(void);
1976 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1977 u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
1978
1979 #if defined(CONFIG_NET)
1980 bool bpf_sock_common_is_valid_access(int off, int size,
1981 enum bpf_access_type type,
1982 struct bpf_insn_access_aux *info);
1983 bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
1984 struct bpf_insn_access_aux *info);
1985 u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
1986 const struct bpf_insn *si,
1987 struct bpf_insn *insn_buf,
1988 struct bpf_prog *prog,
1989 u32 *target_size);
1990 #else
1991 static inline bool bpf_sock_common_is_valid_access(int off, int size,
1992 enum bpf_access_type type,
1993 struct bpf_insn_access_aux *info)
1994 {
1995 return false;
1996 }
1997 static inline bool bpf_sock_is_valid_access(int off, int size,
1998 enum bpf_access_type type,
1999 struct bpf_insn_access_aux *info)
2000 {
2001 return false;
2002 }
2003 static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
2004 const struct bpf_insn *si,
2005 struct bpf_insn *insn_buf,
2006 struct bpf_prog *prog,
2007 u32 *target_size)
2008 {
2009 return 0;
2010 }
2011 #endif
2012
2013 #ifdef CONFIG_INET
2014 struct sk_reuseport_kern {
2015 struct sk_buff *skb;
2016 struct sock *sk;
2017 struct sock *selected_sk;
2018 void *data_end;
2019 u32 hash;
2020 u32 reuseport_id;
2021 bool bind_inany;
2022 };
2023 bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
2024 struct bpf_insn_access_aux *info);
2025
2026 u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
2027 const struct bpf_insn *si,
2028 struct bpf_insn *insn_buf,
2029 struct bpf_prog *prog,
2030 u32 *target_size);
2031
2032 bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
2033 struct bpf_insn_access_aux *info);
2034
2035 u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
2036 const struct bpf_insn *si,
2037 struct bpf_insn *insn_buf,
2038 struct bpf_prog *prog,
2039 u32 *target_size);
2040 #else
2041 static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
2042 enum bpf_access_type type,
2043 struct bpf_insn_access_aux *info)
2044 {
2045 return false;
2046 }
2047
2048 static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
2049 const struct bpf_insn *si,
2050 struct bpf_insn *insn_buf,
2051 struct bpf_prog *prog,
2052 u32 *target_size)
2053 {
2054 return 0;
2055 }
2056 static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
2057 enum bpf_access_type type,
2058 struct bpf_insn_access_aux *info)
2059 {
2060 return false;
2061 }
2062
2063 static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
2064 const struct bpf_insn *si,
2065 struct bpf_insn *insn_buf,
2066 struct bpf_prog *prog,
2067 u32 *target_size)
2068 {
2069 return 0;
2070 }
2071 #endif /* CONFIG_INET */
2072
2073 enum bpf_text_poke_type {
2074 BPF_MOD_CALL,
2075 BPF_MOD_JUMP,
2076 };
2077
2078 int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
2079 void *addr1, void *addr2);
2080
2081 struct btf_id_set;
2082 bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
2083
2084 enum bpf_printf_mod_type {
2085 BPF_PRINTF_INT,
2086 BPF_PRINTF_LONG,
2087 BPF_PRINTF_LONG_LONG,
2088 };
2089
2090 /* Workaround for getting va_list handling working with different argument type
2091 * combinations generically for 32 and 64 bit archs.
2092 */
2093 #define BPF_CAST_FMT_ARG(arg_nb, args, mod) \
2094 (mod[arg_nb] == BPF_PRINTF_LONG_LONG || \
2095 (mod[arg_nb] == BPF_PRINTF_LONG && __BITS_PER_LONG == 64) \
2096 ? (u64)args[arg_nb] \
2097 : (u32)args[arg_nb])
2098
2099 int bpf_printf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
2100 u64 *final_args, enum bpf_printf_mod_type *mod,
2101 u32 num_args);
2102 void bpf_printf_cleanup(void);
2103
2104 #endif /* _LINUX_BPF_H */