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#include <uapi/linux/filter.h>
9
10#include <linux/workqueue.h>
11#include <linux/file.h>
12#include <linux/percpu.h>
13#include <linux/err.h>
14#include <linux/rbtree_latch.h>
15#include <linux/numa.h>
16#include <linux/mm_types.h>
17#include <linux/wait.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/sched/mm.h>
24#include <linux/slab.h>
25#include <linux/percpu-refcount.h>
26#include <linux/stddef.h>
27#include <linux/bpfptr.h>
28#include <linux/btf.h>
29#include <linux/rcupdate_trace.h>
30
31struct bpf_verifier_env;
32struct bpf_verifier_log;
33struct perf_event;
34struct bpf_prog;
35struct bpf_prog_aux;
36struct bpf_map;
37struct sock;
38struct seq_file;
39struct btf;
40struct btf_type;
41struct exception_table_entry;
42struct seq_operations;
43struct bpf_iter_aux_info;
44struct bpf_local_storage;
45struct bpf_local_storage_map;
46struct kobject;
47struct mem_cgroup;
48struct module;
49struct bpf_func_state;
50struct ftrace_ops;
51
52extern struct idr btf_idr;
53extern spinlock_t btf_idr_lock;
54extern struct kobject *btf_kobj;
55
56typedef u64 (*bpf_callback_t)(u64, u64, u64, u64, u64);
57typedef int (*bpf_iter_init_seq_priv_t)(void *private_data,
58 struct bpf_iter_aux_info *aux);
59typedef void (*bpf_iter_fini_seq_priv_t)(void *private_data);
60typedef unsigned int (*bpf_func_t)(const void *,
61 const struct bpf_insn *);
62struct bpf_iter_seq_info {
63 const struct seq_operations *seq_ops;
64 bpf_iter_init_seq_priv_t init_seq_private;
65 bpf_iter_fini_seq_priv_t fini_seq_private;
66 u32 seq_priv_size;
67};
68
69/* map is generic key/value storage optionally accessible by eBPF programs */
70struct bpf_map_ops {
71 /* funcs callable from userspace (via syscall) */
72 int (*map_alloc_check)(union bpf_attr *attr);
73 struct bpf_map *(*map_alloc)(union bpf_attr *attr);
74 void (*map_release)(struct bpf_map *map, struct file *map_file);
75 void (*map_free)(struct bpf_map *map);
76 int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
77 void (*map_release_uref)(struct bpf_map *map);
78 void *(*map_lookup_elem_sys_only)(struct bpf_map *map, void *key);
79 int (*map_lookup_batch)(struct bpf_map *map, const union bpf_attr *attr,
80 union bpf_attr __user *uattr);
81 int (*map_lookup_and_delete_elem)(struct bpf_map *map, void *key,
82 void *value, u64 flags);
83 int (*map_lookup_and_delete_batch)(struct bpf_map *map,
84 const union bpf_attr *attr,
85 union bpf_attr __user *uattr);
86 int (*map_update_batch)(struct bpf_map *map, const union bpf_attr *attr,
87 union bpf_attr __user *uattr);
88 int (*map_delete_batch)(struct bpf_map *map, const union bpf_attr *attr,
89 union bpf_attr __user *uattr);
90
91 /* funcs callable from userspace and from eBPF programs */
92 void *(*map_lookup_elem)(struct bpf_map *map, void *key);
93 int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
94 int (*map_delete_elem)(struct bpf_map *map, void *key);
95 int (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
96 int (*map_pop_elem)(struct bpf_map *map, void *value);
97 int (*map_peek_elem)(struct bpf_map *map, void *value);
98 void *(*map_lookup_percpu_elem)(struct bpf_map *map, void *key, u32 cpu);
99
100 /* funcs called by prog_array and perf_event_array map */
101 void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
102 int fd);
103 void (*map_fd_put_ptr)(void *ptr);
104 int (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
105 u32 (*map_fd_sys_lookup_elem)(void *ptr);
106 void (*map_seq_show_elem)(struct bpf_map *map, void *key,
107 struct seq_file *m);
108 int (*map_check_btf)(const struct bpf_map *map,
109 const struct btf *btf,
110 const struct btf_type *key_type,
111 const struct btf_type *value_type);
112
113 /* Prog poke tracking helpers. */
114 int (*map_poke_track)(struct bpf_map *map, struct bpf_prog_aux *aux);
115 void (*map_poke_untrack)(struct bpf_map *map, struct bpf_prog_aux *aux);
116 void (*map_poke_run)(struct bpf_map *map, u32 key, struct bpf_prog *old,
117 struct bpf_prog *new);
118
119 /* Direct value access helpers. */
120 int (*map_direct_value_addr)(const struct bpf_map *map,
121 u64 *imm, u32 off);
122 int (*map_direct_value_meta)(const struct bpf_map *map,
123 u64 imm, u32 *off);
124 int (*map_mmap)(struct bpf_map *map, struct vm_area_struct *vma);
125 __poll_t (*map_poll)(struct bpf_map *map, struct file *filp,
126 struct poll_table_struct *pts);
127
128 /* Functions called by bpf_local_storage maps */
129 int (*map_local_storage_charge)(struct bpf_local_storage_map *smap,
130 void *owner, u32 size);
131 void (*map_local_storage_uncharge)(struct bpf_local_storage_map *smap,
132 void *owner, u32 size);
133 struct bpf_local_storage __rcu ** (*map_owner_storage_ptr)(void *owner);
134
135 /* Misc helpers.*/
136 int (*map_redirect)(struct bpf_map *map, u32 ifindex, u64 flags);
137
138 /* map_meta_equal must be implemented for maps that can be
139 * used as an inner map. It is a runtime check to ensure
140 * an inner map can be inserted to an outer map.
141 *
142 * Some properties of the inner map has been used during the
143 * verification time. When inserting an inner map at the runtime,
144 * map_meta_equal has to ensure the inserting map has the same
145 * properties that the verifier has used earlier.
146 */
147 bool (*map_meta_equal)(const struct bpf_map *meta0,
148 const struct bpf_map *meta1);
149
150
151 int (*map_set_for_each_callback_args)(struct bpf_verifier_env *env,
152 struct bpf_func_state *caller,
153 struct bpf_func_state *callee);
154 int (*map_for_each_callback)(struct bpf_map *map,
155 bpf_callback_t callback_fn,
156 void *callback_ctx, u64 flags);
157
158 /* BTF id of struct allocated by map_alloc */
159 int *map_btf_id;
160
161 /* bpf_iter info used to open a seq_file */
162 const struct bpf_iter_seq_info *iter_seq_info;
163};
164
165enum {
166 /* Support at most 8 pointers in a BPF map value */
167 BPF_MAP_VALUE_OFF_MAX = 8,
168 BPF_MAP_OFF_ARR_MAX = BPF_MAP_VALUE_OFF_MAX +
169 1 + /* for bpf_spin_lock */
170 1, /* for bpf_timer */
171};
172
173enum bpf_kptr_type {
174 BPF_KPTR_UNREF,
175 BPF_KPTR_REF,
176};
177
178struct bpf_map_value_off_desc {
179 u32 offset;
180 enum bpf_kptr_type type;
181 struct {
182 struct btf *btf;
183 struct module *module;
184 btf_dtor_kfunc_t dtor;
185 u32 btf_id;
186 } kptr;
187};
188
189struct bpf_map_value_off {
190 u32 nr_off;
191 struct bpf_map_value_off_desc off[];
192};
193
194struct bpf_map_off_arr {
195 u32 cnt;
196 u32 field_off[BPF_MAP_OFF_ARR_MAX];
197 u8 field_sz[BPF_MAP_OFF_ARR_MAX];
198};
199
200struct bpf_map {
201 /* The first two cachelines with read-mostly members of which some
202 * are also accessed in fast-path (e.g. ops, max_entries).
203 */
204 const struct bpf_map_ops *ops ____cacheline_aligned;
205 struct bpf_map *inner_map_meta;
206#ifdef CONFIG_SECURITY
207 void *security;
208#endif
209 enum bpf_map_type map_type;
210 u32 key_size;
211 u32 value_size;
212 u32 max_entries;
213 u64 map_extra; /* any per-map-type extra fields */
214 u32 map_flags;
215 int spin_lock_off; /* >=0 valid offset, <0 error */
216 struct bpf_map_value_off *kptr_off_tab;
217 int timer_off; /* >=0 valid offset, <0 error */
218 u32 id;
219 int numa_node;
220 u32 btf_key_type_id;
221 u32 btf_value_type_id;
222 u32 btf_vmlinux_value_type_id;
223 struct btf *btf;
224#ifdef CONFIG_MEMCG_KMEM
225 struct obj_cgroup *objcg;
226#endif
227 char name[BPF_OBJ_NAME_LEN];
228 struct bpf_map_off_arr *off_arr;
229 /* The 3rd and 4th cacheline with misc members to avoid false sharing
230 * particularly with refcounting.
231 */
232 atomic64_t refcnt ____cacheline_aligned;
233 atomic64_t usercnt;
234 struct work_struct work;
235 struct mutex freeze_mutex;
236 atomic64_t writecnt;
237 /* 'Ownership' of program-containing map is claimed by the first program
238 * that is going to use this map or by the first program which FD is
239 * stored in the map to make sure that all callers and callees have the
240 * same prog type, JITed flag and xdp_has_frags flag.
241 */
242 struct {
243 spinlock_t lock;
244 enum bpf_prog_type type;
245 bool jited;
246 bool xdp_has_frags;
247 } owner;
248 bool bypass_spec_v1;
249 bool frozen; /* write-once; write-protected by freeze_mutex */
250};
251
252static inline bool map_value_has_spin_lock(const struct bpf_map *map)
253{
254 return map->spin_lock_off >= 0;
255}
256
257static inline bool map_value_has_timer(const struct bpf_map *map)
258{
259 return map->timer_off >= 0;
260}
261
262static inline bool map_value_has_kptrs(const struct bpf_map *map)
263{
264 return !IS_ERR_OR_NULL(map->kptr_off_tab);
265}
266
267static inline void check_and_init_map_value(struct bpf_map *map, void *dst)
268{
269 if (unlikely(map_value_has_spin_lock(map)))
270 memset(dst + map->spin_lock_off, 0, sizeof(struct bpf_spin_lock));
271 if (unlikely(map_value_has_timer(map)))
272 memset(dst + map->timer_off, 0, sizeof(struct bpf_timer));
273 if (unlikely(map_value_has_kptrs(map))) {
274 struct bpf_map_value_off *tab = map->kptr_off_tab;
275 int i;
276
277 for (i = 0; i < tab->nr_off; i++)
278 *(u64 *)(dst + tab->off[i].offset) = 0;
279 }
280}
281
282/* copy everything but bpf_spin_lock and bpf_timer. There could be one of each. */
283static inline void copy_map_value(struct bpf_map *map, void *dst, void *src)
284{
285 u32 curr_off = 0;
286 int i;
287
288 if (likely(!map->off_arr)) {
289 memcpy(dst, src, map->value_size);
290 return;
291 }
292
293 for (i = 0; i < map->off_arr->cnt; i++) {
294 u32 next_off = map->off_arr->field_off[i];
295
296 memcpy(dst + curr_off, src + curr_off, next_off - curr_off);
297 curr_off += map->off_arr->field_sz[i];
298 }
299 memcpy(dst + curr_off, src + curr_off, map->value_size - curr_off);
300}
301void copy_map_value_locked(struct bpf_map *map, void *dst, void *src,
302 bool lock_src);
303void bpf_timer_cancel_and_free(void *timer);
304int bpf_obj_name_cpy(char *dst, const char *src, unsigned int size);
305
306struct bpf_offload_dev;
307struct bpf_offloaded_map;
308
309struct bpf_map_dev_ops {
310 int (*map_get_next_key)(struct bpf_offloaded_map *map,
311 void *key, void *next_key);
312 int (*map_lookup_elem)(struct bpf_offloaded_map *map,
313 void *key, void *value);
314 int (*map_update_elem)(struct bpf_offloaded_map *map,
315 void *key, void *value, u64 flags);
316 int (*map_delete_elem)(struct bpf_offloaded_map *map, void *key);
317};
318
319struct bpf_offloaded_map {
320 struct bpf_map map;
321 struct net_device *netdev;
322 const struct bpf_map_dev_ops *dev_ops;
323 void *dev_priv;
324 struct list_head offloads;
325};
326
327static inline struct bpf_offloaded_map *map_to_offmap(struct bpf_map *map)
328{
329 return container_of(map, struct bpf_offloaded_map, map);
330}
331
332static inline bool bpf_map_offload_neutral(const struct bpf_map *map)
333{
334 return map->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
335}
336
337static inline bool bpf_map_support_seq_show(const struct bpf_map *map)
338{
339 return (map->btf_value_type_id || map->btf_vmlinux_value_type_id) &&
340 map->ops->map_seq_show_elem;
341}
342
343int map_check_no_btf(const struct bpf_map *map,
344 const struct btf *btf,
345 const struct btf_type *key_type,
346 const struct btf_type *value_type);
347
348bool bpf_map_meta_equal(const struct bpf_map *meta0,
349 const struct bpf_map *meta1);
350
351extern const struct bpf_map_ops bpf_map_offload_ops;
352
353/* bpf_type_flag contains a set of flags that are applicable to the values of
354 * arg_type, ret_type and reg_type. For example, a pointer value may be null,
355 * or a memory is read-only. We classify types into two categories: base types
356 * and extended types. Extended types are base types combined with a type flag.
357 *
358 * Currently there are no more than 32 base types in arg_type, ret_type and
359 * reg_types.
360 */
361#define BPF_BASE_TYPE_BITS 8
362
363enum bpf_type_flag {
364 /* PTR may be NULL. */
365 PTR_MAYBE_NULL = BIT(0 + BPF_BASE_TYPE_BITS),
366
367 /* MEM is read-only. When applied on bpf_arg, it indicates the arg is
368 * compatible with both mutable and immutable memory.
369 */
370 MEM_RDONLY = BIT(1 + BPF_BASE_TYPE_BITS),
371
372 /* MEM was "allocated" from a different helper, and cannot be mixed
373 * with regular non-MEM_ALLOC'ed MEM types.
374 */
375 MEM_ALLOC = BIT(2 + BPF_BASE_TYPE_BITS),
376
377 /* MEM is in user address space. */
378 MEM_USER = BIT(3 + BPF_BASE_TYPE_BITS),
379
380 /* MEM is a percpu memory. MEM_PERCPU tags PTR_TO_BTF_ID. When tagged
381 * with MEM_PERCPU, PTR_TO_BTF_ID _cannot_ be directly accessed. In
382 * order to drop this tag, it must be passed into bpf_per_cpu_ptr()
383 * or bpf_this_cpu_ptr(), which will return the pointer corresponding
384 * to the specified cpu.
385 */
386 MEM_PERCPU = BIT(4 + BPF_BASE_TYPE_BITS),
387
388 /* Indicates that the argument will be released. */
389 OBJ_RELEASE = BIT(5 + BPF_BASE_TYPE_BITS),
390
391 /* PTR is not trusted. This is only used with PTR_TO_BTF_ID, to mark
392 * unreferenced and referenced kptr loaded from map value using a load
393 * instruction, so that they can only be dereferenced but not escape the
394 * BPF program into the kernel (i.e. cannot be passed as arguments to
395 * kfunc or bpf helpers).
396 */
397 PTR_UNTRUSTED = BIT(6 + BPF_BASE_TYPE_BITS),
398
399 MEM_UNINIT = BIT(7 + BPF_BASE_TYPE_BITS),
400
401 /* DYNPTR points to memory local to the bpf program. */
402 DYNPTR_TYPE_LOCAL = BIT(8 + BPF_BASE_TYPE_BITS),
403
404 /* DYNPTR points to a ringbuf record. */
405 DYNPTR_TYPE_RINGBUF = BIT(9 + BPF_BASE_TYPE_BITS),
406
407 /* Size is known at compile time. */
408 MEM_FIXED_SIZE = BIT(10 + BPF_BASE_TYPE_BITS),
409
410 __BPF_TYPE_FLAG_MAX,
411 __BPF_TYPE_LAST_FLAG = __BPF_TYPE_FLAG_MAX - 1,
412};
413
414#define DYNPTR_TYPE_FLAG_MASK (DYNPTR_TYPE_LOCAL | DYNPTR_TYPE_RINGBUF)
415
416/* Max number of base types. */
417#define BPF_BASE_TYPE_LIMIT (1UL << BPF_BASE_TYPE_BITS)
418
419/* Max number of all types. */
420#define BPF_TYPE_LIMIT (__BPF_TYPE_LAST_FLAG | (__BPF_TYPE_LAST_FLAG - 1))
421
422/* function argument constraints */
423enum bpf_arg_type {
424 ARG_DONTCARE = 0, /* unused argument in helper function */
425
426 /* the following constraints used to prototype
427 * bpf_map_lookup/update/delete_elem() functions
428 */
429 ARG_CONST_MAP_PTR, /* const argument used as pointer to bpf_map */
430 ARG_PTR_TO_MAP_KEY, /* pointer to stack used as map key */
431 ARG_PTR_TO_MAP_VALUE, /* pointer to stack used as map value */
432
433 /* Used to prototype bpf_memcmp() and other functions that access data
434 * on eBPF program stack
435 */
436 ARG_PTR_TO_MEM, /* pointer to valid memory (stack, packet, map value) */
437
438 ARG_CONST_SIZE, /* number of bytes accessed from memory */
439 ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
440
441 ARG_PTR_TO_CTX, /* pointer to context */
442 ARG_ANYTHING, /* any (initialized) argument is ok */
443 ARG_PTR_TO_SPIN_LOCK, /* pointer to bpf_spin_lock */
444 ARG_PTR_TO_SOCK_COMMON, /* pointer to sock_common */
445 ARG_PTR_TO_INT, /* pointer to int */
446 ARG_PTR_TO_LONG, /* pointer to long */
447 ARG_PTR_TO_SOCKET, /* pointer to bpf_sock (fullsock) */
448 ARG_PTR_TO_BTF_ID, /* pointer to in-kernel struct */
449 ARG_PTR_TO_ALLOC_MEM, /* pointer to dynamically allocated memory */
450 ARG_CONST_ALLOC_SIZE_OR_ZERO, /* number of allocated bytes requested */
451 ARG_PTR_TO_BTF_ID_SOCK_COMMON, /* pointer to in-kernel sock_common or bpf-mirrored bpf_sock */
452 ARG_PTR_TO_PERCPU_BTF_ID, /* pointer to in-kernel percpu type */
453 ARG_PTR_TO_FUNC, /* pointer to a bpf program function */
454 ARG_PTR_TO_STACK, /* pointer to stack */
455 ARG_PTR_TO_CONST_STR, /* pointer to a null terminated read-only string */
456 ARG_PTR_TO_TIMER, /* pointer to bpf_timer */
457 ARG_PTR_TO_KPTR, /* pointer to referenced kptr */
458 ARG_PTR_TO_DYNPTR, /* pointer to bpf_dynptr. See bpf_type_flag for dynptr type */
459 __BPF_ARG_TYPE_MAX,
460
461 /* Extended arg_types. */
462 ARG_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MAP_VALUE,
463 ARG_PTR_TO_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_MEM,
464 ARG_PTR_TO_CTX_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_CTX,
465 ARG_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_SOCKET,
466 ARG_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_ALLOC_MEM,
467 ARG_PTR_TO_STACK_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_STACK,
468 ARG_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | ARG_PTR_TO_BTF_ID,
469 /* pointer to memory does not need to be initialized, helper function must fill
470 * all bytes or clear them in error case.
471 */
472 ARG_PTR_TO_UNINIT_MEM = MEM_UNINIT | ARG_PTR_TO_MEM,
473 /* Pointer to valid memory of size known at compile time. */
474 ARG_PTR_TO_FIXED_SIZE_MEM = MEM_FIXED_SIZE | ARG_PTR_TO_MEM,
475
476 /* This must be the last entry. Its purpose is to ensure the enum is
477 * wide enough to hold the higher bits reserved for bpf_type_flag.
478 */
479 __BPF_ARG_TYPE_LIMIT = BPF_TYPE_LIMIT,
480};
481static_assert(__BPF_ARG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
482
483/* type of values returned from helper functions */
484enum bpf_return_type {
485 RET_INTEGER, /* function returns integer */
486 RET_VOID, /* function doesn't return anything */
487 RET_PTR_TO_MAP_VALUE, /* returns a pointer to map elem value */
488 RET_PTR_TO_SOCKET, /* returns a pointer to a socket */
489 RET_PTR_TO_TCP_SOCK, /* returns a pointer to a tcp_sock */
490 RET_PTR_TO_SOCK_COMMON, /* returns a pointer to a sock_common */
491 RET_PTR_TO_ALLOC_MEM, /* returns a pointer to dynamically allocated memory */
492 RET_PTR_TO_MEM_OR_BTF_ID, /* returns a pointer to a valid memory or a btf_id */
493 RET_PTR_TO_BTF_ID, /* returns a pointer to a btf_id */
494 __BPF_RET_TYPE_MAX,
495
496 /* Extended ret_types. */
497 RET_PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_MAP_VALUE,
498 RET_PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCKET,
499 RET_PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_TCP_SOCK,
500 RET_PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_SOCK_COMMON,
501 RET_PTR_TO_ALLOC_MEM_OR_NULL = PTR_MAYBE_NULL | MEM_ALLOC | RET_PTR_TO_ALLOC_MEM,
502 RET_PTR_TO_DYNPTR_MEM_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_ALLOC_MEM,
503 RET_PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | RET_PTR_TO_BTF_ID,
504
505 /* This must be the last entry. Its purpose is to ensure the enum is
506 * wide enough to hold the higher bits reserved for bpf_type_flag.
507 */
508 __BPF_RET_TYPE_LIMIT = BPF_TYPE_LIMIT,
509};
510static_assert(__BPF_RET_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
511
512/* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
513 * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
514 * instructions after verifying
515 */
516struct bpf_func_proto {
517 u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
518 bool gpl_only;
519 bool pkt_access;
520 enum bpf_return_type ret_type;
521 union {
522 struct {
523 enum bpf_arg_type arg1_type;
524 enum bpf_arg_type arg2_type;
525 enum bpf_arg_type arg3_type;
526 enum bpf_arg_type arg4_type;
527 enum bpf_arg_type arg5_type;
528 };
529 enum bpf_arg_type arg_type[5];
530 };
531 union {
532 struct {
533 u32 *arg1_btf_id;
534 u32 *arg2_btf_id;
535 u32 *arg3_btf_id;
536 u32 *arg4_btf_id;
537 u32 *arg5_btf_id;
538 };
539 u32 *arg_btf_id[5];
540 struct {
541 size_t arg1_size;
542 size_t arg2_size;
543 size_t arg3_size;
544 size_t arg4_size;
545 size_t arg5_size;
546 };
547 size_t arg_size[5];
548 };
549 int *ret_btf_id; /* return value btf_id */
550 bool (*allowed)(const struct bpf_prog *prog);
551};
552
553/* bpf_context is intentionally undefined structure. Pointer to bpf_context is
554 * the first argument to eBPF programs.
555 * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
556 */
557struct bpf_context;
558
559enum bpf_access_type {
560 BPF_READ = 1,
561 BPF_WRITE = 2
562};
563
564/* types of values stored in eBPF registers */
565/* Pointer types represent:
566 * pointer
567 * pointer + imm
568 * pointer + (u16) var
569 * pointer + (u16) var + imm
570 * if (range > 0) then [ptr, ptr + range - off) is safe to access
571 * if (id > 0) means that some 'var' was added
572 * if (off > 0) means that 'imm' was added
573 */
574enum bpf_reg_type {
575 NOT_INIT = 0, /* nothing was written into register */
576 SCALAR_VALUE, /* reg doesn't contain a valid pointer */
577 PTR_TO_CTX, /* reg points to bpf_context */
578 CONST_PTR_TO_MAP, /* reg points to struct bpf_map */
579 PTR_TO_MAP_VALUE, /* reg points to map element value */
580 PTR_TO_MAP_KEY, /* reg points to a map element key */
581 PTR_TO_STACK, /* reg == frame_pointer + offset */
582 PTR_TO_PACKET_META, /* skb->data - meta_len */
583 PTR_TO_PACKET, /* reg points to skb->data */
584 PTR_TO_PACKET_END, /* skb->data + headlen */
585 PTR_TO_FLOW_KEYS, /* reg points to bpf_flow_keys */
586 PTR_TO_SOCKET, /* reg points to struct bpf_sock */
587 PTR_TO_SOCK_COMMON, /* reg points to sock_common */
588 PTR_TO_TCP_SOCK, /* reg points to struct tcp_sock */
589 PTR_TO_TP_BUFFER, /* reg points to a writable raw tp's buffer */
590 PTR_TO_XDP_SOCK, /* reg points to struct xdp_sock */
591 /* PTR_TO_BTF_ID points to a kernel struct that does not need
592 * to be null checked by the BPF program. This does not imply the
593 * pointer is _not_ null and in practice this can easily be a null
594 * pointer when reading pointer chains. The assumption is program
595 * context will handle null pointer dereference typically via fault
596 * handling. The verifier must keep this in mind and can make no
597 * assumptions about null or non-null when doing branch analysis.
598 * Further, when passed into helpers the helpers can not, without
599 * additional context, assume the value is non-null.
600 */
601 PTR_TO_BTF_ID,
602 /* PTR_TO_BTF_ID_OR_NULL points to a kernel struct that has not
603 * been checked for null. Used primarily to inform the verifier
604 * an explicit null check is required for this struct.
605 */
606 PTR_TO_MEM, /* reg points to valid memory region */
607 PTR_TO_BUF, /* reg points to a read/write buffer */
608 PTR_TO_FUNC, /* reg points to a bpf program function */
609 __BPF_REG_TYPE_MAX,
610
611 /* Extended reg_types. */
612 PTR_TO_MAP_VALUE_OR_NULL = PTR_MAYBE_NULL | PTR_TO_MAP_VALUE,
613 PTR_TO_SOCKET_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCKET,
614 PTR_TO_SOCK_COMMON_OR_NULL = PTR_MAYBE_NULL | PTR_TO_SOCK_COMMON,
615 PTR_TO_TCP_SOCK_OR_NULL = PTR_MAYBE_NULL | PTR_TO_TCP_SOCK,
616 PTR_TO_BTF_ID_OR_NULL = PTR_MAYBE_NULL | PTR_TO_BTF_ID,
617
618 /* This must be the last entry. Its purpose is to ensure the enum is
619 * wide enough to hold the higher bits reserved for bpf_type_flag.
620 */
621 __BPF_REG_TYPE_LIMIT = BPF_TYPE_LIMIT,
622};
623static_assert(__BPF_REG_TYPE_MAX <= BPF_BASE_TYPE_LIMIT);
624
625/* The information passed from prog-specific *_is_valid_access
626 * back to the verifier.
627 */
628struct bpf_insn_access_aux {
629 enum bpf_reg_type reg_type;
630 union {
631 int ctx_field_size;
632 struct {
633 struct btf *btf;
634 u32 btf_id;
635 };
636 };
637 struct bpf_verifier_log *log; /* for verbose logs */
638};
639
640static inline void
641bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
642{
643 aux->ctx_field_size = size;
644}
645
646static inline bool bpf_pseudo_func(const struct bpf_insn *insn)
647{
648 return insn->code == (BPF_LD | BPF_IMM | BPF_DW) &&
649 insn->src_reg == BPF_PSEUDO_FUNC;
650}
651
652struct bpf_prog_ops {
653 int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
654 union bpf_attr __user *uattr);
655};
656
657struct bpf_verifier_ops {
658 /* return eBPF function prototype for verification */
659 const struct bpf_func_proto *
660 (*get_func_proto)(enum bpf_func_id func_id,
661 const struct bpf_prog *prog);
662
663 /* return true if 'size' wide access at offset 'off' within bpf_context
664 * with 'type' (read or write) is allowed
665 */
666 bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
667 const struct bpf_prog *prog,
668 struct bpf_insn_access_aux *info);
669 int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
670 const struct bpf_prog *prog);
671 int (*gen_ld_abs)(const struct bpf_insn *orig,
672 struct bpf_insn *insn_buf);
673 u32 (*convert_ctx_access)(enum bpf_access_type type,
674 const struct bpf_insn *src,
675 struct bpf_insn *dst,
676 struct bpf_prog *prog, u32 *target_size);
677 int (*btf_struct_access)(struct bpf_verifier_log *log,
678 const struct btf *btf,
679 const struct btf_type *t, int off, int size,
680 enum bpf_access_type atype,
681 u32 *next_btf_id, enum bpf_type_flag *flag);
682};
683
684struct bpf_prog_offload_ops {
685 /* verifier basic callbacks */
686 int (*insn_hook)(struct bpf_verifier_env *env,
687 int insn_idx, int prev_insn_idx);
688 int (*finalize)(struct bpf_verifier_env *env);
689 /* verifier optimization callbacks (called after .finalize) */
690 int (*replace_insn)(struct bpf_verifier_env *env, u32 off,
691 struct bpf_insn *insn);
692 int (*remove_insns)(struct bpf_verifier_env *env, u32 off, u32 cnt);
693 /* program management callbacks */
694 int (*prepare)(struct bpf_prog *prog);
695 int (*translate)(struct bpf_prog *prog);
696 void (*destroy)(struct bpf_prog *prog);
697};
698
699struct bpf_prog_offload {
700 struct bpf_prog *prog;
701 struct net_device *netdev;
702 struct bpf_offload_dev *offdev;
703 void *dev_priv;
704 struct list_head offloads;
705 bool dev_state;
706 bool opt_failed;
707 void *jited_image;
708 u32 jited_len;
709};
710
711enum bpf_cgroup_storage_type {
712 BPF_CGROUP_STORAGE_SHARED,
713 BPF_CGROUP_STORAGE_PERCPU,
714 __BPF_CGROUP_STORAGE_MAX
715};
716
717#define MAX_BPF_CGROUP_STORAGE_TYPE __BPF_CGROUP_STORAGE_MAX
718
719/* The longest tracepoint has 12 args.
720 * See include/trace/bpf_probe.h
721 */
722#define MAX_BPF_FUNC_ARGS 12
723
724/* The maximum number of arguments passed through registers
725 * a single function may have.
726 */
727#define MAX_BPF_FUNC_REG_ARGS 5
728
729struct btf_func_model {
730 u8 ret_size;
731 u8 nr_args;
732 u8 arg_size[MAX_BPF_FUNC_ARGS];
733};
734
735/* Restore arguments before returning from trampoline to let original function
736 * continue executing. This flag is used for fentry progs when there are no
737 * fexit progs.
738 */
739#define BPF_TRAMP_F_RESTORE_REGS BIT(0)
740/* Call original function after fentry progs, but before fexit progs.
741 * Makes sense for fentry/fexit, normal calls and indirect calls.
742 */
743#define BPF_TRAMP_F_CALL_ORIG BIT(1)
744/* Skip current frame and return to parent. Makes sense for fentry/fexit
745 * programs only. Should not be used with normal calls and indirect calls.
746 */
747#define BPF_TRAMP_F_SKIP_FRAME BIT(2)
748/* Store IP address of the caller on the trampoline stack,
749 * so it's available for trampoline's programs.
750 */
751#define BPF_TRAMP_F_IP_ARG BIT(3)
752/* Return the return value of fentry prog. Only used by bpf_struct_ops. */
753#define BPF_TRAMP_F_RET_FENTRY_RET BIT(4)
754
755/* Get original function from stack instead of from provided direct address.
756 * Makes sense for trampolines with fexit or fmod_ret programs.
757 */
758#define BPF_TRAMP_F_ORIG_STACK BIT(5)
759
760/* This trampoline is on a function with another ftrace_ops with IPMODIFY,
761 * e.g., a live patch. This flag is set and cleared by ftrace call backs,
762 */
763#define BPF_TRAMP_F_SHARE_IPMODIFY BIT(6)
764
765/* Each call __bpf_prog_enter + call bpf_func + call __bpf_prog_exit is ~50
766 * bytes on x86.
767 */
768#define BPF_MAX_TRAMP_LINKS 38
769
770struct bpf_tramp_links {
771 struct bpf_tramp_link *links[BPF_MAX_TRAMP_LINKS];
772 int nr_links;
773};
774
775struct bpf_tramp_run_ctx;
776
777/* Different use cases for BPF trampoline:
778 * 1. replace nop at the function entry (kprobe equivalent)
779 * flags = BPF_TRAMP_F_RESTORE_REGS
780 * fentry = a set of programs to run before returning from trampoline
781 *
782 * 2. replace nop at the function entry (kprobe + kretprobe equivalent)
783 * flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME
784 * orig_call = fentry_ip + MCOUNT_INSN_SIZE
785 * fentry = a set of program to run before calling original function
786 * fexit = a set of program to run after original function
787 *
788 * 3. replace direct call instruction anywhere in the function body
789 * or assign a function pointer for indirect call (like tcp_congestion_ops->cong_avoid)
790 * With flags = 0
791 * fentry = a set of programs to run before returning from trampoline
792 * With flags = BPF_TRAMP_F_CALL_ORIG
793 * orig_call = original callback addr or direct function addr
794 * fentry = a set of program to run before calling original function
795 * fexit = a set of program to run after original function
796 */
797struct bpf_tramp_image;
798int arch_prepare_bpf_trampoline(struct bpf_tramp_image *tr, void *image, void *image_end,
799 const struct btf_func_model *m, u32 flags,
800 struct bpf_tramp_links *tlinks,
801 void *orig_call);
802/* these two functions are called from generated trampoline */
803u64 notrace __bpf_prog_enter(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx);
804void notrace __bpf_prog_exit(struct bpf_prog *prog, u64 start, struct bpf_tramp_run_ctx *run_ctx);
805u64 notrace __bpf_prog_enter_sleepable(struct bpf_prog *prog, struct bpf_tramp_run_ctx *run_ctx);
806void notrace __bpf_prog_exit_sleepable(struct bpf_prog *prog, u64 start,
807 struct bpf_tramp_run_ctx *run_ctx);
808u64 notrace __bpf_prog_enter_lsm_cgroup(struct bpf_prog *prog,
809 struct bpf_tramp_run_ctx *run_ctx);
810void notrace __bpf_prog_exit_lsm_cgroup(struct bpf_prog *prog, u64 start,
811 struct bpf_tramp_run_ctx *run_ctx);
812void notrace __bpf_tramp_enter(struct bpf_tramp_image *tr);
813void notrace __bpf_tramp_exit(struct bpf_tramp_image *tr);
814
815struct bpf_ksym {
816 unsigned long start;
817 unsigned long end;
818 char name[KSYM_NAME_LEN];
819 struct list_head lnode;
820 struct latch_tree_node tnode;
821 bool prog;
822};
823
824enum bpf_tramp_prog_type {
825 BPF_TRAMP_FENTRY,
826 BPF_TRAMP_FEXIT,
827 BPF_TRAMP_MODIFY_RETURN,
828 BPF_TRAMP_MAX,
829 BPF_TRAMP_REPLACE, /* more than MAX */
830};
831
832struct bpf_tramp_image {
833 void *image;
834 struct bpf_ksym ksym;
835 struct percpu_ref pcref;
836 void *ip_after_call;
837 void *ip_epilogue;
838 union {
839 struct rcu_head rcu;
840 struct work_struct work;
841 };
842};
843
844struct bpf_trampoline {
845 /* hlist for trampoline_table */
846 struct hlist_node hlist;
847 struct ftrace_ops *fops;
848 /* serializes access to fields of this trampoline */
849 struct mutex mutex;
850 refcount_t refcnt;
851 u32 flags;
852 u64 key;
853 struct {
854 struct btf_func_model model;
855 void *addr;
856 bool ftrace_managed;
857 } func;
858 /* if !NULL this is BPF_PROG_TYPE_EXT program that extends another BPF
859 * program by replacing one of its functions. func.addr is the address
860 * of the function it replaced.
861 */
862 struct bpf_prog *extension_prog;
863 /* list of BPF programs using this trampoline */
864 struct hlist_head progs_hlist[BPF_TRAMP_MAX];
865 /* Number of attached programs. A counter per kind. */
866 int progs_cnt[BPF_TRAMP_MAX];
867 /* Executable image of trampoline */
868 struct bpf_tramp_image *cur_image;
869 u64 selector;
870 struct module *mod;
871};
872
873struct bpf_attach_target_info {
874 struct btf_func_model fmodel;
875 long tgt_addr;
876 const char *tgt_name;
877 const struct btf_type *tgt_type;
878};
879
880#define BPF_DISPATCHER_MAX 48 /* Fits in 2048B */
881
882struct bpf_dispatcher_prog {
883 struct bpf_prog *prog;
884 refcount_t users;
885};
886
887struct bpf_dispatcher {
888 /* dispatcher mutex */
889 struct mutex mutex;
890 void *func;
891 struct bpf_dispatcher_prog progs[BPF_DISPATCHER_MAX];
892 int num_progs;
893 void *image;
894 u32 image_off;
895 struct bpf_ksym ksym;
896};
897
898static __always_inline __nocfi unsigned int bpf_dispatcher_nop_func(
899 const void *ctx,
900 const struct bpf_insn *insnsi,
901 bpf_func_t bpf_func)
902{
903 return bpf_func(ctx, insnsi);
904}
905
906#ifdef CONFIG_BPF_JIT
907int bpf_trampoline_link_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
908int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link, struct bpf_trampoline *tr);
909struct bpf_trampoline *bpf_trampoline_get(u64 key,
910 struct bpf_attach_target_info *tgt_info);
911void bpf_trampoline_put(struct bpf_trampoline *tr);
912int arch_prepare_bpf_dispatcher(void *image, s64 *funcs, int num_funcs);
913#define BPF_DISPATCHER_INIT(_name) { \
914 .mutex = __MUTEX_INITIALIZER(_name.mutex), \
915 .func = &_name##_func, \
916 .progs = {}, \
917 .num_progs = 0, \
918 .image = NULL, \
919 .image_off = 0, \
920 .ksym = { \
921 .name = #_name, \
922 .lnode = LIST_HEAD_INIT(_name.ksym.lnode), \
923 }, \
924}
925
926#define DEFINE_BPF_DISPATCHER(name) \
927 noinline __nocfi unsigned int bpf_dispatcher_##name##_func( \
928 const void *ctx, \
929 const struct bpf_insn *insnsi, \
930 bpf_func_t bpf_func) \
931 { \
932 return bpf_func(ctx, insnsi); \
933 } \
934 EXPORT_SYMBOL(bpf_dispatcher_##name##_func); \
935 struct bpf_dispatcher bpf_dispatcher_##name = \
936 BPF_DISPATCHER_INIT(bpf_dispatcher_##name);
937#define DECLARE_BPF_DISPATCHER(name) \
938 unsigned int bpf_dispatcher_##name##_func( \
939 const void *ctx, \
940 const struct bpf_insn *insnsi, \
941 bpf_func_t bpf_func); \
942 extern struct bpf_dispatcher bpf_dispatcher_##name;
943#define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_##name##_func
944#define BPF_DISPATCHER_PTR(name) (&bpf_dispatcher_##name)
945void bpf_dispatcher_change_prog(struct bpf_dispatcher *d, struct bpf_prog *from,
946 struct bpf_prog *to);
947/* Called only from JIT-enabled code, so there's no need for stubs. */
948void *bpf_jit_alloc_exec_page(void);
949void bpf_image_ksym_add(void *data, struct bpf_ksym *ksym);
950void bpf_image_ksym_del(struct bpf_ksym *ksym);
951void bpf_ksym_add(struct bpf_ksym *ksym);
952void bpf_ksym_del(struct bpf_ksym *ksym);
953int bpf_jit_charge_modmem(u32 size);
954void bpf_jit_uncharge_modmem(u32 size);
955bool bpf_prog_has_trampoline(const struct bpf_prog *prog);
956#else
957static inline int bpf_trampoline_link_prog(struct bpf_tramp_link *link,
958 struct bpf_trampoline *tr)
959{
960 return -ENOTSUPP;
961}
962static inline int bpf_trampoline_unlink_prog(struct bpf_tramp_link *link,
963 struct bpf_trampoline *tr)
964{
965 return -ENOTSUPP;
966}
967static inline struct bpf_trampoline *bpf_trampoline_get(u64 key,
968 struct bpf_attach_target_info *tgt_info)
969{
970 return ERR_PTR(-EOPNOTSUPP);
971}
972static inline void bpf_trampoline_put(struct bpf_trampoline *tr) {}
973#define DEFINE_BPF_DISPATCHER(name)
974#define DECLARE_BPF_DISPATCHER(name)
975#define BPF_DISPATCHER_FUNC(name) bpf_dispatcher_nop_func
976#define BPF_DISPATCHER_PTR(name) NULL
977static inline void bpf_dispatcher_change_prog(struct bpf_dispatcher *d,
978 struct bpf_prog *from,
979 struct bpf_prog *to) {}
980static inline bool is_bpf_image_address(unsigned long address)
981{
982 return false;
983}
984static inline bool bpf_prog_has_trampoline(const struct bpf_prog *prog)
985{
986 return false;
987}
988#endif
989
990struct bpf_func_info_aux {
991 u16 linkage;
992 bool unreliable;
993};
994
995enum bpf_jit_poke_reason {
996 BPF_POKE_REASON_TAIL_CALL,
997};
998
999/* Descriptor of pokes pointing /into/ the JITed image. */
1000struct bpf_jit_poke_descriptor {
1001 void *tailcall_target;
1002 void *tailcall_bypass;
1003 void *bypass_addr;
1004 void *aux;
1005 union {
1006 struct {
1007 struct bpf_map *map;
1008 u32 key;
1009 } tail_call;
1010 };
1011 bool tailcall_target_stable;
1012 u8 adj_off;
1013 u16 reason;
1014 u32 insn_idx;
1015};
1016
1017/* reg_type info for ctx arguments */
1018struct bpf_ctx_arg_aux {
1019 u32 offset;
1020 enum bpf_reg_type reg_type;
1021 u32 btf_id;
1022};
1023
1024struct btf_mod_pair {
1025 struct btf *btf;
1026 struct module *module;
1027};
1028
1029struct bpf_kfunc_desc_tab;
1030
1031struct bpf_prog_aux {
1032 atomic64_t refcnt;
1033 u32 used_map_cnt;
1034 u32 used_btf_cnt;
1035 u32 max_ctx_offset;
1036 u32 max_pkt_offset;
1037 u32 max_tp_access;
1038 u32 stack_depth;
1039 u32 id;
1040 u32 func_cnt; /* used by non-func prog as the number of func progs */
1041 u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
1042 u32 attach_btf_id; /* in-kernel BTF type id to attach to */
1043 u32 ctx_arg_info_size;
1044 u32 max_rdonly_access;
1045 u32 max_rdwr_access;
1046 struct btf *attach_btf;
1047 const struct bpf_ctx_arg_aux *ctx_arg_info;
1048 struct mutex dst_mutex; /* protects dst_* pointers below, *after* prog becomes visible */
1049 struct bpf_prog *dst_prog;
1050 struct bpf_trampoline *dst_trampoline;
1051 enum bpf_prog_type saved_dst_prog_type;
1052 enum bpf_attach_type saved_dst_attach_type;
1053 bool verifier_zext; /* Zero extensions has been inserted by verifier. */
1054 bool offload_requested;
1055 bool attach_btf_trace; /* true if attaching to BTF-enabled raw tp */
1056 bool func_proto_unreliable;
1057 bool sleepable;
1058 bool tail_call_reachable;
1059 bool xdp_has_frags;
1060 /* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
1061 const struct btf_type *attach_func_proto;
1062 /* function name for valid attach_btf_id */
1063 const char *attach_func_name;
1064 struct bpf_prog **func;
1065 void *jit_data; /* JIT specific data. arch dependent */
1066 struct bpf_jit_poke_descriptor *poke_tab;
1067 struct bpf_kfunc_desc_tab *kfunc_tab;
1068 struct bpf_kfunc_btf_tab *kfunc_btf_tab;
1069 u32 size_poke_tab;
1070 struct bpf_ksym ksym;
1071 const struct bpf_prog_ops *ops;
1072 struct bpf_map **used_maps;
1073 struct mutex used_maps_mutex; /* mutex for used_maps and used_map_cnt */
1074 struct btf_mod_pair *used_btfs;
1075 struct bpf_prog *prog;
1076 struct user_struct *user;
1077 u64 load_time; /* ns since boottime */
1078 u32 verified_insns;
1079 int cgroup_atype; /* enum cgroup_bpf_attach_type */
1080 struct bpf_map *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1081 char name[BPF_OBJ_NAME_LEN];
1082#ifdef CONFIG_SECURITY
1083 void *security;
1084#endif
1085 struct bpf_prog_offload *offload;
1086 struct btf *btf;
1087 struct bpf_func_info *func_info;
1088 struct bpf_func_info_aux *func_info_aux;
1089 /* bpf_line_info loaded from userspace. linfo->insn_off
1090 * has the xlated insn offset.
1091 * Both the main and sub prog share the same linfo.
1092 * The subprog can access its first linfo by
1093 * using the linfo_idx.
1094 */
1095 struct bpf_line_info *linfo;
1096 /* jited_linfo is the jited addr of the linfo. It has a
1097 * one to one mapping to linfo:
1098 * jited_linfo[i] is the jited addr for the linfo[i]->insn_off.
1099 * Both the main and sub prog share the same jited_linfo.
1100 * The subprog can access its first jited_linfo by
1101 * using the linfo_idx.
1102 */
1103 void **jited_linfo;
1104 u32 func_info_cnt;
1105 u32 nr_linfo;
1106 /* subprog can use linfo_idx to access its first linfo and
1107 * jited_linfo.
1108 * main prog always has linfo_idx == 0
1109 */
1110 u32 linfo_idx;
1111 u32 num_exentries;
1112 struct exception_table_entry *extable;
1113 union {
1114 struct work_struct work;
1115 struct rcu_head rcu;
1116 };
1117};
1118
1119struct bpf_prog {
1120 u16 pages; /* Number of allocated pages */
1121 u16 jited:1, /* Is our filter JIT'ed? */
1122 jit_requested:1,/* archs need to JIT the prog */
1123 gpl_compatible:1, /* Is filter GPL compatible? */
1124 cb_access:1, /* Is control block accessed? */
1125 dst_needed:1, /* Do we need dst entry? */
1126 blinding_requested:1, /* needs constant blinding */
1127 blinded:1, /* Was blinded */
1128 is_func:1, /* program is a bpf function */
1129 kprobe_override:1, /* Do we override a kprobe? */
1130 has_callchain_buf:1, /* callchain buffer allocated? */
1131 enforce_expected_attach_type:1, /* Enforce expected_attach_type checking at attach time */
1132 call_get_stack:1, /* Do we call bpf_get_stack() or bpf_get_stackid() */
1133 call_get_func_ip:1, /* Do we call get_func_ip() */
1134 tstamp_type_access:1; /* Accessed __sk_buff->tstamp_type */
1135 enum bpf_prog_type type; /* Type of BPF program */
1136 enum bpf_attach_type expected_attach_type; /* For some prog types */
1137 u32 len; /* Number of filter blocks */
1138 u32 jited_len; /* Size of jited insns in bytes */
1139 u8 tag[BPF_TAG_SIZE];
1140 struct bpf_prog_stats __percpu *stats;
1141 int __percpu *active;
1142 unsigned int (*bpf_func)(const void *ctx,
1143 const struct bpf_insn *insn);
1144 struct bpf_prog_aux *aux; /* Auxiliary fields */
1145 struct sock_fprog_kern *orig_prog; /* Original BPF program */
1146 /* Instructions for interpreter */
1147 union {
1148 DECLARE_FLEX_ARRAY(struct sock_filter, insns);
1149 DECLARE_FLEX_ARRAY(struct bpf_insn, insnsi);
1150 };
1151};
1152
1153struct bpf_array_aux {
1154 /* Programs with direct jumps into programs part of this array. */
1155 struct list_head poke_progs;
1156 struct bpf_map *map;
1157 struct mutex poke_mutex;
1158 struct work_struct work;
1159};
1160
1161struct bpf_link {
1162 atomic64_t refcnt;
1163 u32 id;
1164 enum bpf_link_type type;
1165 const struct bpf_link_ops *ops;
1166 struct bpf_prog *prog;
1167 struct work_struct work;
1168};
1169
1170struct bpf_link_ops {
1171 void (*release)(struct bpf_link *link);
1172 void (*dealloc)(struct bpf_link *link);
1173 int (*detach)(struct bpf_link *link);
1174 int (*update_prog)(struct bpf_link *link, struct bpf_prog *new_prog,
1175 struct bpf_prog *old_prog);
1176 void (*show_fdinfo)(const struct bpf_link *link, struct seq_file *seq);
1177 int (*fill_link_info)(const struct bpf_link *link,
1178 struct bpf_link_info *info);
1179};
1180
1181struct bpf_tramp_link {
1182 struct bpf_link link;
1183 struct hlist_node tramp_hlist;
1184 u64 cookie;
1185};
1186
1187struct bpf_shim_tramp_link {
1188 struct bpf_tramp_link link;
1189 struct bpf_trampoline *trampoline;
1190};
1191
1192struct bpf_tracing_link {
1193 struct bpf_tramp_link link;
1194 enum bpf_attach_type attach_type;
1195 struct bpf_trampoline *trampoline;
1196 struct bpf_prog *tgt_prog;
1197};
1198
1199struct bpf_link_primer {
1200 struct bpf_link *link;
1201 struct file *file;
1202 int fd;
1203 u32 id;
1204};
1205
1206struct bpf_struct_ops_value;
1207struct btf_member;
1208
1209#define BPF_STRUCT_OPS_MAX_NR_MEMBERS 64
1210struct bpf_struct_ops {
1211 const struct bpf_verifier_ops *verifier_ops;
1212 int (*init)(struct btf *btf);
1213 int (*check_member)(const struct btf_type *t,
1214 const struct btf_member *member);
1215 int (*init_member)(const struct btf_type *t,
1216 const struct btf_member *member,
1217 void *kdata, const void *udata);
1218 int (*reg)(void *kdata);
1219 void (*unreg)(void *kdata);
1220 const struct btf_type *type;
1221 const struct btf_type *value_type;
1222 const char *name;
1223 struct btf_func_model func_models[BPF_STRUCT_OPS_MAX_NR_MEMBERS];
1224 u32 type_id;
1225 u32 value_id;
1226};
1227
1228#if defined(CONFIG_BPF_JIT) && defined(CONFIG_BPF_SYSCALL)
1229#define BPF_MODULE_OWNER ((void *)((0xeB9FUL << 2) + POISON_POINTER_DELTA))
1230const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id);
1231void bpf_struct_ops_init(struct btf *btf, struct bpf_verifier_log *log);
1232bool bpf_struct_ops_get(const void *kdata);
1233void bpf_struct_ops_put(const void *kdata);
1234int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map, void *key,
1235 void *value);
1236int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks,
1237 struct bpf_tramp_link *link,
1238 const struct btf_func_model *model,
1239 void *image, void *image_end);
1240static inline bool bpf_try_module_get(const void *data, struct module *owner)
1241{
1242 if (owner == BPF_MODULE_OWNER)
1243 return bpf_struct_ops_get(data);
1244 else
1245 return try_module_get(owner);
1246}
1247static inline void bpf_module_put(const void *data, struct module *owner)
1248{
1249 if (owner == BPF_MODULE_OWNER)
1250 bpf_struct_ops_put(data);
1251 else
1252 module_put(owner);
1253}
1254
1255#ifdef CONFIG_NET
1256/* Define it here to avoid the use of forward declaration */
1257struct bpf_dummy_ops_state {
1258 int val;
1259};
1260
1261struct bpf_dummy_ops {
1262 int (*test_1)(struct bpf_dummy_ops_state *cb);
1263 int (*test_2)(struct bpf_dummy_ops_state *cb, int a1, unsigned short a2,
1264 char a3, unsigned long a4);
1265};
1266
1267int bpf_struct_ops_test_run(struct bpf_prog *prog, const union bpf_attr *kattr,
1268 union bpf_attr __user *uattr);
1269#endif
1270#else
1271static inline const struct bpf_struct_ops *bpf_struct_ops_find(u32 type_id)
1272{
1273 return NULL;
1274}
1275static inline void bpf_struct_ops_init(struct btf *btf,
1276 struct bpf_verifier_log *log)
1277{
1278}
1279static inline bool bpf_try_module_get(const void *data, struct module *owner)
1280{
1281 return try_module_get(owner);
1282}
1283static inline void bpf_module_put(const void *data, struct module *owner)
1284{
1285 module_put(owner);
1286}
1287static inline int bpf_struct_ops_map_sys_lookup_elem(struct bpf_map *map,
1288 void *key,
1289 void *value)
1290{
1291 return -EINVAL;
1292}
1293#endif
1294
1295#if defined(CONFIG_CGROUP_BPF) && defined(CONFIG_BPF_LSM)
1296int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1297 int cgroup_atype);
1298void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog);
1299#else
1300static inline int bpf_trampoline_link_cgroup_shim(struct bpf_prog *prog,
1301 int cgroup_atype)
1302{
1303 return -EOPNOTSUPP;
1304}
1305static inline void bpf_trampoline_unlink_cgroup_shim(struct bpf_prog *prog)
1306{
1307}
1308#endif
1309
1310struct bpf_array {
1311 struct bpf_map map;
1312 u32 elem_size;
1313 u32 index_mask;
1314 struct bpf_array_aux *aux;
1315 union {
1316 char value[0] __aligned(8);
1317 void *ptrs[0] __aligned(8);
1318 void __percpu *pptrs[0] __aligned(8);
1319 };
1320};
1321
1322#define BPF_COMPLEXITY_LIMIT_INSNS 1000000 /* yes. 1M insns */
1323#define MAX_TAIL_CALL_CNT 33
1324
1325/* Maximum number of loops for bpf_loop */
1326#define BPF_MAX_LOOPS BIT(23)
1327
1328#define BPF_F_ACCESS_MASK (BPF_F_RDONLY | \
1329 BPF_F_RDONLY_PROG | \
1330 BPF_F_WRONLY | \
1331 BPF_F_WRONLY_PROG)
1332
1333#define BPF_MAP_CAN_READ BIT(0)
1334#define BPF_MAP_CAN_WRITE BIT(1)
1335
1336static inline u32 bpf_map_flags_to_cap(struct bpf_map *map)
1337{
1338 u32 access_flags = map->map_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1339
1340 /* Combination of BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG is
1341 * not possible.
1342 */
1343 if (access_flags & BPF_F_RDONLY_PROG)
1344 return BPF_MAP_CAN_READ;
1345 else if (access_flags & BPF_F_WRONLY_PROG)
1346 return BPF_MAP_CAN_WRITE;
1347 else
1348 return BPF_MAP_CAN_READ | BPF_MAP_CAN_WRITE;
1349}
1350
1351static inline bool bpf_map_flags_access_ok(u32 access_flags)
1352{
1353 return (access_flags & (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG)) !=
1354 (BPF_F_RDONLY_PROG | BPF_F_WRONLY_PROG);
1355}
1356
1357struct bpf_event_entry {
1358 struct perf_event *event;
1359 struct file *perf_file;
1360 struct file *map_file;
1361 struct rcu_head rcu;
1362};
1363
1364static inline bool map_type_contains_progs(struct bpf_map *map)
1365{
1366 return map->map_type == BPF_MAP_TYPE_PROG_ARRAY ||
1367 map->map_type == BPF_MAP_TYPE_DEVMAP ||
1368 map->map_type == BPF_MAP_TYPE_CPUMAP;
1369}
1370
1371bool bpf_prog_map_compatible(struct bpf_map *map, const struct bpf_prog *fp);
1372int bpf_prog_calc_tag(struct bpf_prog *fp);
1373
1374const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
1375const struct bpf_func_proto *bpf_get_trace_vprintk_proto(void);
1376
1377typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
1378 unsigned long off, unsigned long len);
1379typedef u32 (*bpf_convert_ctx_access_t)(enum bpf_access_type type,
1380 const struct bpf_insn *src,
1381 struct bpf_insn *dst,
1382 struct bpf_prog *prog,
1383 u32 *target_size);
1384
1385u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
1386 void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
1387
1388/* an array of programs to be executed under rcu_lock.
1389 *
1390 * Typical usage:
1391 * ret = bpf_prog_run_array(rcu_dereference(&bpf_prog_array), ctx, bpf_prog_run);
1392 *
1393 * the structure returned by bpf_prog_array_alloc() should be populated
1394 * with program pointers and the last pointer must be NULL.
1395 * The user has to keep refcnt on the program and make sure the program
1396 * is removed from the array before bpf_prog_put().
1397 * The 'struct bpf_prog_array *' should only be replaced with xchg()
1398 * since other cpus are walking the array of pointers in parallel.
1399 */
1400struct bpf_prog_array_item {
1401 struct bpf_prog *prog;
1402 union {
1403 struct bpf_cgroup_storage *cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE];
1404 u64 bpf_cookie;
1405 };
1406};
1407
1408struct bpf_prog_array {
1409 struct rcu_head rcu;
1410 struct bpf_prog_array_item items[];
1411};
1412
1413struct bpf_empty_prog_array {
1414 struct bpf_prog_array hdr;
1415 struct bpf_prog *null_prog;
1416};
1417
1418/* to avoid allocating empty bpf_prog_array for cgroups that
1419 * don't have bpf program attached use one global 'bpf_empty_prog_array'
1420 * It will not be modified the caller of bpf_prog_array_alloc()
1421 * (since caller requested prog_cnt == 0)
1422 * that pointer should be 'freed' by bpf_prog_array_free()
1423 */
1424extern struct bpf_empty_prog_array bpf_empty_prog_array;
1425
1426struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags);
1427void bpf_prog_array_free(struct bpf_prog_array *progs);
1428/* Use when traversal over the bpf_prog_array uses tasks_trace rcu */
1429void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs);
1430int bpf_prog_array_length(struct bpf_prog_array *progs);
1431bool bpf_prog_array_is_empty(struct bpf_prog_array *array);
1432int bpf_prog_array_copy_to_user(struct bpf_prog_array *progs,
1433 __u32 __user *prog_ids, u32 cnt);
1434
1435void bpf_prog_array_delete_safe(struct bpf_prog_array *progs,
1436 struct bpf_prog *old_prog);
1437int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index);
1438int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
1439 struct bpf_prog *prog);
1440int bpf_prog_array_copy_info(struct bpf_prog_array *array,
1441 u32 *prog_ids, u32 request_cnt,
1442 u32 *prog_cnt);
1443int bpf_prog_array_copy(struct bpf_prog_array *old_array,
1444 struct bpf_prog *exclude_prog,
1445 struct bpf_prog *include_prog,
1446 u64 bpf_cookie,
1447 struct bpf_prog_array **new_array);
1448
1449struct bpf_run_ctx {};
1450
1451struct bpf_cg_run_ctx {
1452 struct bpf_run_ctx run_ctx;
1453 const struct bpf_prog_array_item *prog_item;
1454 int retval;
1455};
1456
1457struct bpf_trace_run_ctx {
1458 struct bpf_run_ctx run_ctx;
1459 u64 bpf_cookie;
1460};
1461
1462struct bpf_tramp_run_ctx {
1463 struct bpf_run_ctx run_ctx;
1464 u64 bpf_cookie;
1465 struct bpf_run_ctx *saved_run_ctx;
1466};
1467
1468static inline struct bpf_run_ctx *bpf_set_run_ctx(struct bpf_run_ctx *new_ctx)
1469{
1470 struct bpf_run_ctx *old_ctx = NULL;
1471
1472#ifdef CONFIG_BPF_SYSCALL
1473 old_ctx = current->bpf_ctx;
1474 current->bpf_ctx = new_ctx;
1475#endif
1476 return old_ctx;
1477}
1478
1479static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
1480{
1481#ifdef CONFIG_BPF_SYSCALL
1482 current->bpf_ctx = old_ctx;
1483#endif
1484}
1485
1486/* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
1487#define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0)
1488/* BPF program asks to set CN on the packet. */
1489#define BPF_RET_SET_CN (1 << 0)
1490
1491typedef u32 (*bpf_prog_run_fn)(const struct bpf_prog *prog, const void *ctx);
1492
1493static __always_inline u32
1494bpf_prog_run_array(const struct bpf_prog_array *array,
1495 const void *ctx, bpf_prog_run_fn run_prog)
1496{
1497 const struct bpf_prog_array_item *item;
1498 const struct bpf_prog *prog;
1499 struct bpf_run_ctx *old_run_ctx;
1500 struct bpf_trace_run_ctx run_ctx;
1501 u32 ret = 1;
1502
1503 RCU_LOCKDEP_WARN(!rcu_read_lock_held(), "no rcu lock held");
1504
1505 if (unlikely(!array))
1506 return ret;
1507
1508 migrate_disable();
1509 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
1510 item = &array->items[0];
1511 while ((prog = READ_ONCE(item->prog))) {
1512 run_ctx.bpf_cookie = item->bpf_cookie;
1513 ret &= run_prog(prog, ctx);
1514 item++;
1515 }
1516 bpf_reset_run_ctx(old_run_ctx);
1517 migrate_enable();
1518 return ret;
1519}
1520
1521/* Notes on RCU design for bpf_prog_arrays containing sleepable programs:
1522 *
1523 * We use the tasks_trace rcu flavor read section to protect the bpf_prog_array
1524 * overall. As a result, we must use the bpf_prog_array_free_sleepable
1525 * in order to use the tasks_trace rcu grace period.
1526 *
1527 * When a non-sleepable program is inside the array, we take the rcu read
1528 * section and disable preemption for that program alone, so it can access
1529 * rcu-protected dynamically sized maps.
1530 */
1531static __always_inline u32
1532bpf_prog_run_array_sleepable(const struct bpf_prog_array __rcu *array_rcu,
1533 const void *ctx, bpf_prog_run_fn run_prog)
1534{
1535 const struct bpf_prog_array_item *item;
1536 const struct bpf_prog *prog;
1537 const struct bpf_prog_array *array;
1538 struct bpf_run_ctx *old_run_ctx;
1539 struct bpf_trace_run_ctx run_ctx;
1540 u32 ret = 1;
1541
1542 might_fault();
1543
1544 rcu_read_lock_trace();
1545 migrate_disable();
1546
1547 array = rcu_dereference_check(array_rcu, rcu_read_lock_trace_held());
1548 if (unlikely(!array))
1549 goto out;
1550 old_run_ctx = bpf_set_run_ctx(&run_ctx.run_ctx);
1551 item = &array->items[0];
1552 while ((prog = READ_ONCE(item->prog))) {
1553 if (!prog->aux->sleepable)
1554 rcu_read_lock();
1555
1556 run_ctx.bpf_cookie = item->bpf_cookie;
1557 ret &= run_prog(prog, ctx);
1558 item++;
1559
1560 if (!prog->aux->sleepable)
1561 rcu_read_unlock();
1562 }
1563 bpf_reset_run_ctx(old_run_ctx);
1564out:
1565 migrate_enable();
1566 rcu_read_unlock_trace();
1567 return ret;
1568}
1569
1570#ifdef CONFIG_BPF_SYSCALL
1571DECLARE_PER_CPU(int, bpf_prog_active);
1572extern struct mutex bpf_stats_enabled_mutex;
1573
1574/*
1575 * Block execution of BPF programs attached to instrumentation (perf,
1576 * kprobes, tracepoints) to prevent deadlocks on map operations as any of
1577 * these events can happen inside a region which holds a map bucket lock
1578 * and can deadlock on it.
1579 */
1580static inline void bpf_disable_instrumentation(void)
1581{
1582 migrate_disable();
1583 this_cpu_inc(bpf_prog_active);
1584}
1585
1586static inline void bpf_enable_instrumentation(void)
1587{
1588 this_cpu_dec(bpf_prog_active);
1589 migrate_enable();
1590}
1591
1592extern const struct file_operations bpf_map_fops;
1593extern const struct file_operations bpf_prog_fops;
1594extern const struct file_operations bpf_iter_fops;
1595
1596#define BPF_PROG_TYPE(_id, _name, prog_ctx_type, kern_ctx_type) \
1597 extern const struct bpf_prog_ops _name ## _prog_ops; \
1598 extern const struct bpf_verifier_ops _name ## _verifier_ops;
1599#define BPF_MAP_TYPE(_id, _ops) \
1600 extern const struct bpf_map_ops _ops;
1601#define BPF_LINK_TYPE(_id, _name)
1602#include <linux/bpf_types.h>
1603#undef BPF_PROG_TYPE
1604#undef BPF_MAP_TYPE
1605#undef BPF_LINK_TYPE
1606
1607extern const struct bpf_prog_ops bpf_offload_prog_ops;
1608extern const struct bpf_verifier_ops tc_cls_act_analyzer_ops;
1609extern const struct bpf_verifier_ops xdp_analyzer_ops;
1610
1611struct bpf_prog *bpf_prog_get(u32 ufd);
1612struct bpf_prog *bpf_prog_get_type_dev(u32 ufd, enum bpf_prog_type type,
1613 bool attach_drv);
1614void bpf_prog_add(struct bpf_prog *prog, int i);
1615void bpf_prog_sub(struct bpf_prog *prog, int i);
1616void bpf_prog_inc(struct bpf_prog *prog);
1617struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
1618void bpf_prog_put(struct bpf_prog *prog);
1619
1620void bpf_prog_free_id(struct bpf_prog *prog, bool do_idr_lock);
1621void bpf_map_free_id(struct bpf_map *map, bool do_idr_lock);
1622
1623struct bpf_map_value_off_desc *bpf_map_kptr_off_contains(struct bpf_map *map, u32 offset);
1624void bpf_map_free_kptr_off_tab(struct bpf_map *map);
1625struct bpf_map_value_off *bpf_map_copy_kptr_off_tab(const struct bpf_map *map);
1626bool bpf_map_equal_kptr_off_tab(const struct bpf_map *map_a, const struct bpf_map *map_b);
1627void bpf_map_free_kptrs(struct bpf_map *map, void *map_value);
1628
1629struct bpf_map *bpf_map_get(u32 ufd);
1630struct bpf_map *bpf_map_get_with_uref(u32 ufd);
1631struct bpf_map *__bpf_map_get(struct fd f);
1632void bpf_map_inc(struct bpf_map *map);
1633void bpf_map_inc_with_uref(struct bpf_map *map);
1634struct bpf_map * __must_check bpf_map_inc_not_zero(struct bpf_map *map);
1635void bpf_map_put_with_uref(struct bpf_map *map);
1636void bpf_map_put(struct bpf_map *map);
1637void *bpf_map_area_alloc(u64 size, int numa_node);
1638void *bpf_map_area_mmapable_alloc(u64 size, int numa_node);
1639void bpf_map_area_free(void *base);
1640bool bpf_map_write_active(const struct bpf_map *map);
1641void bpf_map_init_from_attr(struct bpf_map *map, union bpf_attr *attr);
1642int generic_map_lookup_batch(struct bpf_map *map,
1643 const union bpf_attr *attr,
1644 union bpf_attr __user *uattr);
1645int generic_map_update_batch(struct bpf_map *map,
1646 const union bpf_attr *attr,
1647 union bpf_attr __user *uattr);
1648int generic_map_delete_batch(struct bpf_map *map,
1649 const union bpf_attr *attr,
1650 union bpf_attr __user *uattr);
1651struct bpf_map *bpf_map_get_curr_or_next(u32 *id);
1652struct bpf_prog *bpf_prog_get_curr_or_next(u32 *id);
1653
1654#ifdef CONFIG_MEMCG_KMEM
1655void *bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
1656 int node);
1657void *bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags);
1658void __percpu *bpf_map_alloc_percpu(const struct bpf_map *map, size_t size,
1659 size_t align, gfp_t flags);
1660#else
1661static inline void *
1662bpf_map_kmalloc_node(const struct bpf_map *map, size_t size, gfp_t flags,
1663 int node)
1664{
1665 return kmalloc_node(size, flags, node);
1666}
1667
1668static inline void *
1669bpf_map_kzalloc(const struct bpf_map *map, size_t size, gfp_t flags)
1670{
1671 return kzalloc(size, flags);
1672}
1673
1674static inline void __percpu *
1675bpf_map_alloc_percpu(const struct bpf_map *map, size_t size, size_t align,
1676 gfp_t flags)
1677{
1678 return __alloc_percpu_gfp(size, align, flags);
1679}
1680#endif
1681
1682extern int sysctl_unprivileged_bpf_disabled;
1683
1684static inline bool bpf_allow_ptr_leaks(void)
1685{
1686 return perfmon_capable();
1687}
1688
1689static inline bool bpf_allow_uninit_stack(void)
1690{
1691 return perfmon_capable();
1692}
1693
1694static inline bool bpf_allow_ptr_to_map_access(void)
1695{
1696 return perfmon_capable();
1697}
1698
1699static inline bool bpf_bypass_spec_v1(void)
1700{
1701 return perfmon_capable();
1702}
1703
1704static inline bool bpf_bypass_spec_v4(void)
1705{
1706 return perfmon_capable();
1707}
1708
1709int bpf_map_new_fd(struct bpf_map *map, int flags);
1710int bpf_prog_new_fd(struct bpf_prog *prog);
1711
1712void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
1713 const struct bpf_link_ops *ops, struct bpf_prog *prog);
1714int bpf_link_prime(struct bpf_link *link, struct bpf_link_primer *primer);
1715int bpf_link_settle(struct bpf_link_primer *primer);
1716void bpf_link_cleanup(struct bpf_link_primer *primer);
1717void bpf_link_inc(struct bpf_link *link);
1718void bpf_link_put(struct bpf_link *link);
1719int bpf_link_new_fd(struct bpf_link *link);
1720struct file *bpf_link_new_file(struct bpf_link *link, int *reserved_fd);
1721struct bpf_link *bpf_link_get_from_fd(u32 ufd);
1722struct bpf_link *bpf_link_get_curr_or_next(u32 *id);
1723
1724int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
1725int bpf_obj_get_user(const char __user *pathname, int flags);
1726
1727#define BPF_ITER_FUNC_PREFIX "bpf_iter_"
1728#define DEFINE_BPF_ITER_FUNC(target, args...) \
1729 extern int bpf_iter_ ## target(args); \
1730 int __init bpf_iter_ ## target(args) { return 0; }
1731
1732struct bpf_iter_aux_info {
1733 struct bpf_map *map;
1734};
1735
1736typedef int (*bpf_iter_attach_target_t)(struct bpf_prog *prog,
1737 union bpf_iter_link_info *linfo,
1738 struct bpf_iter_aux_info *aux);
1739typedef void (*bpf_iter_detach_target_t)(struct bpf_iter_aux_info *aux);
1740typedef void (*bpf_iter_show_fdinfo_t) (const struct bpf_iter_aux_info *aux,
1741 struct seq_file *seq);
1742typedef int (*bpf_iter_fill_link_info_t)(const struct bpf_iter_aux_info *aux,
1743 struct bpf_link_info *info);
1744typedef const struct bpf_func_proto *
1745(*bpf_iter_get_func_proto_t)(enum bpf_func_id func_id,
1746 const struct bpf_prog *prog);
1747
1748enum bpf_iter_feature {
1749 BPF_ITER_RESCHED = BIT(0),
1750};
1751
1752#define BPF_ITER_CTX_ARG_MAX 2
1753struct bpf_iter_reg {
1754 const char *target;
1755 bpf_iter_attach_target_t attach_target;
1756 bpf_iter_detach_target_t detach_target;
1757 bpf_iter_show_fdinfo_t show_fdinfo;
1758 bpf_iter_fill_link_info_t fill_link_info;
1759 bpf_iter_get_func_proto_t get_func_proto;
1760 u32 ctx_arg_info_size;
1761 u32 feature;
1762 struct bpf_ctx_arg_aux ctx_arg_info[BPF_ITER_CTX_ARG_MAX];
1763 const struct bpf_iter_seq_info *seq_info;
1764};
1765
1766struct bpf_iter_meta {
1767 __bpf_md_ptr(struct seq_file *, seq);
1768 u64 session_id;
1769 u64 seq_num;
1770};
1771
1772struct bpf_iter__bpf_map_elem {
1773 __bpf_md_ptr(struct bpf_iter_meta *, meta);
1774 __bpf_md_ptr(struct bpf_map *, map);
1775 __bpf_md_ptr(void *, key);
1776 __bpf_md_ptr(void *, value);
1777};
1778
1779int bpf_iter_reg_target(const struct bpf_iter_reg *reg_info);
1780void bpf_iter_unreg_target(const struct bpf_iter_reg *reg_info);
1781bool bpf_iter_prog_supported(struct bpf_prog *prog);
1782const struct bpf_func_proto *
1783bpf_iter_get_func_proto(enum bpf_func_id func_id, const struct bpf_prog *prog);
1784int bpf_iter_link_attach(const union bpf_attr *attr, bpfptr_t uattr, struct bpf_prog *prog);
1785int bpf_iter_new_fd(struct bpf_link *link);
1786bool bpf_link_is_iter(struct bpf_link *link);
1787struct bpf_prog *bpf_iter_get_info(struct bpf_iter_meta *meta, bool in_stop);
1788int bpf_iter_run_prog(struct bpf_prog *prog, void *ctx);
1789void bpf_iter_map_show_fdinfo(const struct bpf_iter_aux_info *aux,
1790 struct seq_file *seq);
1791int bpf_iter_map_fill_link_info(const struct bpf_iter_aux_info *aux,
1792 struct bpf_link_info *info);
1793
1794int map_set_for_each_callback_args(struct bpf_verifier_env *env,
1795 struct bpf_func_state *caller,
1796 struct bpf_func_state *callee);
1797
1798int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
1799int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
1800int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
1801 u64 flags);
1802int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
1803 u64 flags);
1804
1805int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
1806
1807int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
1808 void *key, void *value, u64 map_flags);
1809int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1810int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
1811 void *key, void *value, u64 map_flags);
1812int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
1813
1814int bpf_get_file_flag(int flags);
1815int bpf_check_uarg_tail_zero(bpfptr_t uaddr, size_t expected_size,
1816 size_t actual_size);
1817
1818/* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
1819 * forced to use 'long' read/writes to try to atomically copy long counters.
1820 * Best-effort only. No barriers here, since it _will_ race with concurrent
1821 * updates from BPF programs. Called from bpf syscall and mostly used with
1822 * size 8 or 16 bytes, so ask compiler to inline it.
1823 */
1824static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
1825{
1826 const long *lsrc = src;
1827 long *ldst = dst;
1828
1829 size /= sizeof(long);
1830 while (size--)
1831 *ldst++ = *lsrc++;
1832}
1833
1834/* verify correctness of eBPF program */
1835int bpf_check(struct bpf_prog **fp, union bpf_attr *attr, bpfptr_t uattr);
1836
1837#ifndef CONFIG_BPF_JIT_ALWAYS_ON
1838void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth);
1839#endif
1840
1841struct btf *bpf_get_btf_vmlinux(void);
1842
1843/* Map specifics */
1844struct xdp_frame;
1845struct sk_buff;
1846struct bpf_dtab_netdev;
1847struct bpf_cpu_map_entry;
1848
1849void __dev_flush(void);
1850int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
1851 struct net_device *dev_rx);
1852int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
1853 struct net_device *dev_rx);
1854int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
1855 struct bpf_map *map, bool exclude_ingress);
1856int dev_map_generic_redirect(struct bpf_dtab_netdev *dst, struct sk_buff *skb,
1857 struct bpf_prog *xdp_prog);
1858int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
1859 struct bpf_prog *xdp_prog, struct bpf_map *map,
1860 bool exclude_ingress);
1861
1862void __cpu_map_flush(void);
1863int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu, struct xdp_frame *xdpf,
1864 struct net_device *dev_rx);
1865int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
1866 struct sk_buff *skb);
1867
1868/* Return map's numa specified by userspace */
1869static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
1870{
1871 return (attr->map_flags & BPF_F_NUMA_NODE) ?
1872 attr->numa_node : NUMA_NO_NODE;
1873}
1874
1875struct bpf_prog *bpf_prog_get_type_path(const char *name, enum bpf_prog_type type);
1876int array_map_alloc_check(union bpf_attr *attr);
1877
1878int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
1879 union bpf_attr __user *uattr);
1880int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
1881 union bpf_attr __user *uattr);
1882int bpf_prog_test_run_tracing(struct bpf_prog *prog,
1883 const union bpf_attr *kattr,
1884 union bpf_attr __user *uattr);
1885int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
1886 const union bpf_attr *kattr,
1887 union bpf_attr __user *uattr);
1888int bpf_prog_test_run_raw_tp(struct bpf_prog *prog,
1889 const union bpf_attr *kattr,
1890 union bpf_attr __user *uattr);
1891int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
1892 const union bpf_attr *kattr,
1893 union bpf_attr __user *uattr);
1894bool btf_ctx_access(int off, int size, enum bpf_access_type type,
1895 const struct bpf_prog *prog,
1896 struct bpf_insn_access_aux *info);
1897
1898static inline bool bpf_tracing_ctx_access(int off, int size,
1899 enum bpf_access_type type)
1900{
1901 if (off < 0 || off >= sizeof(__u64) * MAX_BPF_FUNC_ARGS)
1902 return false;
1903 if (type != BPF_READ)
1904 return false;
1905 if (off % size != 0)
1906 return false;
1907 return true;
1908}
1909
1910static inline bool bpf_tracing_btf_ctx_access(int off, int size,
1911 enum bpf_access_type type,
1912 const struct bpf_prog *prog,
1913 struct bpf_insn_access_aux *info)
1914{
1915 if (!bpf_tracing_ctx_access(off, size, type))
1916 return false;
1917 return btf_ctx_access(off, size, type, prog, info);
1918}
1919
1920int btf_struct_access(struct bpf_verifier_log *log, const struct btf *btf,
1921 const struct btf_type *t, int off, int size,
1922 enum bpf_access_type atype,
1923 u32 *next_btf_id, enum bpf_type_flag *flag);
1924bool btf_struct_ids_match(struct bpf_verifier_log *log,
1925 const struct btf *btf, u32 id, int off,
1926 const struct btf *need_btf, u32 need_type_id,
1927 bool strict);
1928
1929int btf_distill_func_proto(struct bpf_verifier_log *log,
1930 struct btf *btf,
1931 const struct btf_type *func_proto,
1932 const char *func_name,
1933 struct btf_func_model *m);
1934
1935struct bpf_reg_state;
1936int btf_check_subprog_arg_match(struct bpf_verifier_env *env, int subprog,
1937 struct bpf_reg_state *regs);
1938int btf_check_kfunc_arg_match(struct bpf_verifier_env *env,
1939 const struct btf *btf, u32 func_id,
1940 struct bpf_reg_state *regs,
1941 u32 kfunc_flags);
1942int btf_prepare_func_args(struct bpf_verifier_env *env, int subprog,
1943 struct bpf_reg_state *reg);
1944int btf_check_type_match(struct bpf_verifier_log *log, const struct bpf_prog *prog,
1945 struct btf *btf, const struct btf_type *t);
1946
1947struct bpf_prog *bpf_prog_by_id(u32 id);
1948struct bpf_link *bpf_link_by_id(u32 id);
1949
1950const struct bpf_func_proto *bpf_base_func_proto(enum bpf_func_id func_id);
1951void bpf_task_storage_free(struct task_struct *task);
1952bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog);
1953const struct btf_func_model *
1954bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
1955 const struct bpf_insn *insn);
1956struct bpf_core_ctx {
1957 struct bpf_verifier_log *log;
1958 const struct btf *btf;
1959};
1960
1961int bpf_core_apply(struct bpf_core_ctx *ctx, const struct bpf_core_relo *relo,
1962 int relo_idx, void *insn);
1963
1964static inline bool unprivileged_ebpf_enabled(void)
1965{
1966 return !sysctl_unprivileged_bpf_disabled;
1967}
1968
1969#else /* !CONFIG_BPF_SYSCALL */
1970static inline struct bpf_prog *bpf_prog_get(u32 ufd)
1971{
1972 return ERR_PTR(-EOPNOTSUPP);
1973}
1974
1975static inline struct bpf_prog *bpf_prog_get_type_dev(u32 ufd,
1976 enum bpf_prog_type type,
1977 bool attach_drv)
1978{
1979 return ERR_PTR(-EOPNOTSUPP);
1980}
1981
1982static inline void bpf_prog_add(struct bpf_prog *prog, int i)
1983{
1984}
1985
1986static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
1987{
1988}
1989
1990static inline void bpf_prog_put(struct bpf_prog *prog)
1991{
1992}
1993
1994static inline void bpf_prog_inc(struct bpf_prog *prog)
1995{
1996}
1997
1998static inline struct bpf_prog *__must_check
1999bpf_prog_inc_not_zero(struct bpf_prog *prog)
2000{
2001 return ERR_PTR(-EOPNOTSUPP);
2002}
2003
2004static inline void bpf_link_init(struct bpf_link *link, enum bpf_link_type type,
2005 const struct bpf_link_ops *ops,
2006 struct bpf_prog *prog)
2007{
2008}
2009
2010static inline int bpf_link_prime(struct bpf_link *link,
2011 struct bpf_link_primer *primer)
2012{
2013 return -EOPNOTSUPP;
2014}
2015
2016static inline int bpf_link_settle(struct bpf_link_primer *primer)
2017{
2018 return -EOPNOTSUPP;
2019}
2020
2021static inline void bpf_link_cleanup(struct bpf_link_primer *primer)
2022{
2023}
2024
2025static inline void bpf_link_inc(struct bpf_link *link)
2026{
2027}
2028
2029static inline void bpf_link_put(struct bpf_link *link)
2030{
2031}
2032
2033static inline int bpf_obj_get_user(const char __user *pathname, int flags)
2034{
2035 return -EOPNOTSUPP;
2036}
2037
2038static inline void __dev_flush(void)
2039{
2040}
2041
2042struct xdp_frame;
2043struct bpf_dtab_netdev;
2044struct bpf_cpu_map_entry;
2045
2046static inline
2047int dev_xdp_enqueue(struct net_device *dev, struct xdp_frame *xdpf,
2048 struct net_device *dev_rx)
2049{
2050 return 0;
2051}
2052
2053static inline
2054int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_frame *xdpf,
2055 struct net_device *dev_rx)
2056{
2057 return 0;
2058}
2059
2060static inline
2061int dev_map_enqueue_multi(struct xdp_frame *xdpf, struct net_device *dev_rx,
2062 struct bpf_map *map, bool exclude_ingress)
2063{
2064 return 0;
2065}
2066
2067struct sk_buff;
2068
2069static inline int dev_map_generic_redirect(struct bpf_dtab_netdev *dst,
2070 struct sk_buff *skb,
2071 struct bpf_prog *xdp_prog)
2072{
2073 return 0;
2074}
2075
2076static inline
2077int dev_map_redirect_multi(struct net_device *dev, struct sk_buff *skb,
2078 struct bpf_prog *xdp_prog, struct bpf_map *map,
2079 bool exclude_ingress)
2080{
2081 return 0;
2082}
2083
2084static inline void __cpu_map_flush(void)
2085{
2086}
2087
2088static inline int cpu_map_enqueue(struct bpf_cpu_map_entry *rcpu,
2089 struct xdp_frame *xdpf,
2090 struct net_device *dev_rx)
2091{
2092 return 0;
2093}
2094
2095static inline int cpu_map_generic_redirect(struct bpf_cpu_map_entry *rcpu,
2096 struct sk_buff *skb)
2097{
2098 return -EOPNOTSUPP;
2099}
2100
2101static inline struct bpf_prog *bpf_prog_get_type_path(const char *name,
2102 enum bpf_prog_type type)
2103{
2104 return ERR_PTR(-EOPNOTSUPP);
2105}
2106
2107static inline int bpf_prog_test_run_xdp(struct bpf_prog *prog,
2108 const union bpf_attr *kattr,
2109 union bpf_attr __user *uattr)
2110{
2111 return -ENOTSUPP;
2112}
2113
2114static inline int bpf_prog_test_run_skb(struct bpf_prog *prog,
2115 const union bpf_attr *kattr,
2116 union bpf_attr __user *uattr)
2117{
2118 return -ENOTSUPP;
2119}
2120
2121static inline int bpf_prog_test_run_tracing(struct bpf_prog *prog,
2122 const union bpf_attr *kattr,
2123 union bpf_attr __user *uattr)
2124{
2125 return -ENOTSUPP;
2126}
2127
2128static inline int bpf_prog_test_run_flow_dissector(struct bpf_prog *prog,
2129 const union bpf_attr *kattr,
2130 union bpf_attr __user *uattr)
2131{
2132 return -ENOTSUPP;
2133}
2134
2135static inline int bpf_prog_test_run_sk_lookup(struct bpf_prog *prog,
2136 const union bpf_attr *kattr,
2137 union bpf_attr __user *uattr)
2138{
2139 return -ENOTSUPP;
2140}
2141
2142static inline void bpf_map_put(struct bpf_map *map)
2143{
2144}
2145
2146static inline struct bpf_prog *bpf_prog_by_id(u32 id)
2147{
2148 return ERR_PTR(-ENOTSUPP);
2149}
2150
2151static inline const struct bpf_func_proto *
2152bpf_base_func_proto(enum bpf_func_id func_id)
2153{
2154 return NULL;
2155}
2156
2157static inline void bpf_task_storage_free(struct task_struct *task)
2158{
2159}
2160
2161static inline bool bpf_prog_has_kfunc_call(const struct bpf_prog *prog)
2162{
2163 return false;
2164}
2165
2166static inline const struct btf_func_model *
2167bpf_jit_find_kfunc_model(const struct bpf_prog *prog,
2168 const struct bpf_insn *insn)
2169{
2170 return NULL;
2171}
2172
2173static inline bool unprivileged_ebpf_enabled(void)
2174{
2175 return false;
2176}
2177
2178#endif /* CONFIG_BPF_SYSCALL */
2179
2180void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
2181 struct btf_mod_pair *used_btfs, u32 len);
2182
2183static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
2184 enum bpf_prog_type type)
2185{
2186 return bpf_prog_get_type_dev(ufd, type, false);
2187}
2188
2189void __bpf_free_used_maps(struct bpf_prog_aux *aux,
2190 struct bpf_map **used_maps, u32 len);
2191
2192bool bpf_prog_get_ok(struct bpf_prog *, enum bpf_prog_type *, bool);
2193
2194int bpf_prog_offload_compile(struct bpf_prog *prog);
2195void bpf_prog_offload_destroy(struct bpf_prog *prog);
2196int bpf_prog_offload_info_fill(struct bpf_prog_info *info,
2197 struct bpf_prog *prog);
2198
2199int bpf_map_offload_info_fill(struct bpf_map_info *info, struct bpf_map *map);
2200
2201int bpf_map_offload_lookup_elem(struct bpf_map *map, void *key, void *value);
2202int bpf_map_offload_update_elem(struct bpf_map *map,
2203 void *key, void *value, u64 flags);
2204int bpf_map_offload_delete_elem(struct bpf_map *map, void *key);
2205int bpf_map_offload_get_next_key(struct bpf_map *map,
2206 void *key, void *next_key);
2207
2208bool bpf_offload_prog_map_match(struct bpf_prog *prog, struct bpf_map *map);
2209
2210struct bpf_offload_dev *
2211bpf_offload_dev_create(const struct bpf_prog_offload_ops *ops, void *priv);
2212void bpf_offload_dev_destroy(struct bpf_offload_dev *offdev);
2213void *bpf_offload_dev_priv(struct bpf_offload_dev *offdev);
2214int bpf_offload_dev_netdev_register(struct bpf_offload_dev *offdev,
2215 struct net_device *netdev);
2216void bpf_offload_dev_netdev_unregister(struct bpf_offload_dev *offdev,
2217 struct net_device *netdev);
2218bool bpf_offload_dev_match(struct bpf_prog *prog, struct net_device *netdev);
2219
2220void unpriv_ebpf_notify(int new_state);
2221
2222#if defined(CONFIG_NET) && defined(CONFIG_BPF_SYSCALL)
2223int bpf_prog_offload_init(struct bpf_prog *prog, union bpf_attr *attr);
2224
2225static inline bool bpf_prog_is_dev_bound(const struct bpf_prog_aux *aux)
2226{
2227 return aux->offload_requested;
2228}
2229
2230static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
2231{
2232 return unlikely(map->ops == &bpf_map_offload_ops);
2233}
2234
2235struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr);
2236void bpf_map_offload_map_free(struct bpf_map *map);
2237int bpf_prog_test_run_syscall(struct bpf_prog *prog,
2238 const union bpf_attr *kattr,
2239 union bpf_attr __user *uattr);
2240
2241int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog);
2242int sock_map_prog_detach(const union bpf_attr *attr, enum bpf_prog_type ptype);
2243int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value, u64 flags);
2244int sock_map_bpf_prog_query(const union bpf_attr *attr,
2245 union bpf_attr __user *uattr);
2246
2247void sock_map_unhash(struct sock *sk);
2248void sock_map_destroy(struct sock *sk);
2249void sock_map_close(struct sock *sk, long timeout);
2250#else
2251static inline int bpf_prog_offload_init(struct bpf_prog *prog,
2252 union bpf_attr *attr)
2253{
2254 return -EOPNOTSUPP;
2255}
2256
2257static inline bool bpf_prog_is_dev_bound(struct bpf_prog_aux *aux)
2258{
2259 return false;
2260}
2261
2262static inline bool bpf_map_is_dev_bound(struct bpf_map *map)
2263{
2264 return false;
2265}
2266
2267static inline struct bpf_map *bpf_map_offload_map_alloc(union bpf_attr *attr)
2268{
2269 return ERR_PTR(-EOPNOTSUPP);
2270}
2271
2272static inline void bpf_map_offload_map_free(struct bpf_map *map)
2273{
2274}
2275
2276static inline int bpf_prog_test_run_syscall(struct bpf_prog *prog,
2277 const union bpf_attr *kattr,
2278 union bpf_attr __user *uattr)
2279{
2280 return -ENOTSUPP;
2281}
2282
2283#ifdef CONFIG_BPF_SYSCALL
2284static inline int sock_map_get_from_fd(const union bpf_attr *attr,
2285 struct bpf_prog *prog)
2286{
2287 return -EINVAL;
2288}
2289
2290static inline int sock_map_prog_detach(const union bpf_attr *attr,
2291 enum bpf_prog_type ptype)
2292{
2293 return -EOPNOTSUPP;
2294}
2295
2296static inline int sock_map_update_elem_sys(struct bpf_map *map, void *key, void *value,
2297 u64 flags)
2298{
2299 return -EOPNOTSUPP;
2300}
2301
2302static inline int sock_map_bpf_prog_query(const union bpf_attr *attr,
2303 union bpf_attr __user *uattr)
2304{
2305 return -EINVAL;
2306}
2307#endif /* CONFIG_BPF_SYSCALL */
2308#endif /* CONFIG_NET && CONFIG_BPF_SYSCALL */
2309
2310#if defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL)
2311void bpf_sk_reuseport_detach(struct sock *sk);
2312int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map, void *key,
2313 void *value);
2314int bpf_fd_reuseport_array_update_elem(struct bpf_map *map, void *key,
2315 void *value, u64 map_flags);
2316#else
2317static inline void bpf_sk_reuseport_detach(struct sock *sk)
2318{
2319}
2320
2321#ifdef CONFIG_BPF_SYSCALL
2322static inline int bpf_fd_reuseport_array_lookup_elem(struct bpf_map *map,
2323 void *key, void *value)
2324{
2325 return -EOPNOTSUPP;
2326}
2327
2328static inline int bpf_fd_reuseport_array_update_elem(struct bpf_map *map,
2329 void *key, void *value,
2330 u64 map_flags)
2331{
2332 return -EOPNOTSUPP;
2333}
2334#endif /* CONFIG_BPF_SYSCALL */
2335#endif /* defined(CONFIG_INET) && defined(CONFIG_BPF_SYSCALL) */
2336
2337/* verifier prototypes for helper functions called from eBPF programs */
2338extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
2339extern const struct bpf_func_proto bpf_map_update_elem_proto;
2340extern const struct bpf_func_proto bpf_map_delete_elem_proto;
2341extern const struct bpf_func_proto bpf_map_push_elem_proto;
2342extern const struct bpf_func_proto bpf_map_pop_elem_proto;
2343extern const struct bpf_func_proto bpf_map_peek_elem_proto;
2344extern const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto;
2345
2346extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
2347extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
2348extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
2349extern const struct bpf_func_proto bpf_tail_call_proto;
2350extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
2351extern const struct bpf_func_proto bpf_ktime_get_boot_ns_proto;
2352extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
2353extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
2354extern const struct bpf_func_proto bpf_get_current_comm_proto;
2355extern const struct bpf_func_proto bpf_get_stackid_proto;
2356extern const struct bpf_func_proto bpf_get_stack_proto;
2357extern const struct bpf_func_proto bpf_get_task_stack_proto;
2358extern const struct bpf_func_proto bpf_get_stackid_proto_pe;
2359extern const struct bpf_func_proto bpf_get_stack_proto_pe;
2360extern const struct bpf_func_proto bpf_sock_map_update_proto;
2361extern const struct bpf_func_proto bpf_sock_hash_update_proto;
2362extern const struct bpf_func_proto bpf_get_current_cgroup_id_proto;
2363extern const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto;
2364extern const struct bpf_func_proto bpf_msg_redirect_hash_proto;
2365extern const struct bpf_func_proto bpf_msg_redirect_map_proto;
2366extern const struct bpf_func_proto bpf_sk_redirect_hash_proto;
2367extern const struct bpf_func_proto bpf_sk_redirect_map_proto;
2368extern const struct bpf_func_proto bpf_spin_lock_proto;
2369extern const struct bpf_func_proto bpf_spin_unlock_proto;
2370extern const struct bpf_func_proto bpf_get_local_storage_proto;
2371extern const struct bpf_func_proto bpf_strtol_proto;
2372extern const struct bpf_func_proto bpf_strtoul_proto;
2373extern const struct bpf_func_proto bpf_tcp_sock_proto;
2374extern const struct bpf_func_proto bpf_jiffies64_proto;
2375extern const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto;
2376extern const struct bpf_func_proto bpf_event_output_data_proto;
2377extern const struct bpf_func_proto bpf_ringbuf_output_proto;
2378extern const struct bpf_func_proto bpf_ringbuf_reserve_proto;
2379extern const struct bpf_func_proto bpf_ringbuf_submit_proto;
2380extern const struct bpf_func_proto bpf_ringbuf_discard_proto;
2381extern const struct bpf_func_proto bpf_ringbuf_query_proto;
2382extern const struct bpf_func_proto bpf_ringbuf_reserve_dynptr_proto;
2383extern const struct bpf_func_proto bpf_ringbuf_submit_dynptr_proto;
2384extern const struct bpf_func_proto bpf_ringbuf_discard_dynptr_proto;
2385extern const struct bpf_func_proto bpf_skc_to_tcp6_sock_proto;
2386extern const struct bpf_func_proto bpf_skc_to_tcp_sock_proto;
2387extern const struct bpf_func_proto bpf_skc_to_tcp_timewait_sock_proto;
2388extern const struct bpf_func_proto bpf_skc_to_tcp_request_sock_proto;
2389extern const struct bpf_func_proto bpf_skc_to_udp6_sock_proto;
2390extern const struct bpf_func_proto bpf_skc_to_unix_sock_proto;
2391extern const struct bpf_func_proto bpf_skc_to_mptcp_sock_proto;
2392extern const struct bpf_func_proto bpf_copy_from_user_proto;
2393extern const struct bpf_func_proto bpf_snprintf_btf_proto;
2394extern const struct bpf_func_proto bpf_snprintf_proto;
2395extern const struct bpf_func_proto bpf_per_cpu_ptr_proto;
2396extern const struct bpf_func_proto bpf_this_cpu_ptr_proto;
2397extern const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto;
2398extern const struct bpf_func_proto bpf_sock_from_file_proto;
2399extern const struct bpf_func_proto bpf_get_socket_ptr_cookie_proto;
2400extern const struct bpf_func_proto bpf_task_storage_get_proto;
2401extern const struct bpf_func_proto bpf_task_storage_delete_proto;
2402extern const struct bpf_func_proto bpf_for_each_map_elem_proto;
2403extern const struct bpf_func_proto bpf_btf_find_by_name_kind_proto;
2404extern const struct bpf_func_proto bpf_sk_setsockopt_proto;
2405extern const struct bpf_func_proto bpf_sk_getsockopt_proto;
2406extern const struct bpf_func_proto bpf_unlocked_sk_setsockopt_proto;
2407extern const struct bpf_func_proto bpf_unlocked_sk_getsockopt_proto;
2408extern const struct bpf_func_proto bpf_find_vma_proto;
2409extern const struct bpf_func_proto bpf_loop_proto;
2410extern const struct bpf_func_proto bpf_copy_from_user_task_proto;
2411extern const struct bpf_func_proto bpf_set_retval_proto;
2412extern const struct bpf_func_proto bpf_get_retval_proto;
2413
2414const struct bpf_func_proto *tracing_prog_func_proto(
2415 enum bpf_func_id func_id, const struct bpf_prog *prog);
2416
2417/* Shared helpers among cBPF and eBPF. */
2418void bpf_user_rnd_init_once(void);
2419u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
2420u64 bpf_get_raw_cpu_id(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
2421
2422#if defined(CONFIG_NET)
2423bool bpf_sock_common_is_valid_access(int off, int size,
2424 enum bpf_access_type type,
2425 struct bpf_insn_access_aux *info);
2426bool bpf_sock_is_valid_access(int off, int size, enum bpf_access_type type,
2427 struct bpf_insn_access_aux *info);
2428u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
2429 const struct bpf_insn *si,
2430 struct bpf_insn *insn_buf,
2431 struct bpf_prog *prog,
2432 u32 *target_size);
2433#else
2434static inline bool bpf_sock_common_is_valid_access(int off, int size,
2435 enum bpf_access_type type,
2436 struct bpf_insn_access_aux *info)
2437{
2438 return false;
2439}
2440static inline bool bpf_sock_is_valid_access(int off, int size,
2441 enum bpf_access_type type,
2442 struct bpf_insn_access_aux *info)
2443{
2444 return false;
2445}
2446static inline u32 bpf_sock_convert_ctx_access(enum bpf_access_type type,
2447 const struct bpf_insn *si,
2448 struct bpf_insn *insn_buf,
2449 struct bpf_prog *prog,
2450 u32 *target_size)
2451{
2452 return 0;
2453}
2454#endif
2455
2456#ifdef CONFIG_INET
2457struct sk_reuseport_kern {
2458 struct sk_buff *skb;
2459 struct sock *sk;
2460 struct sock *selected_sk;
2461 struct sock *migrating_sk;
2462 void *data_end;
2463 u32 hash;
2464 u32 reuseport_id;
2465 bool bind_inany;
2466};
2467bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
2468 struct bpf_insn_access_aux *info);
2469
2470u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
2471 const struct bpf_insn *si,
2472 struct bpf_insn *insn_buf,
2473 struct bpf_prog *prog,
2474 u32 *target_size);
2475
2476bool bpf_xdp_sock_is_valid_access(int off, int size, enum bpf_access_type type,
2477 struct bpf_insn_access_aux *info);
2478
2479u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
2480 const struct bpf_insn *si,
2481 struct bpf_insn *insn_buf,
2482 struct bpf_prog *prog,
2483 u32 *target_size);
2484#else
2485static inline bool bpf_tcp_sock_is_valid_access(int off, int size,
2486 enum bpf_access_type type,
2487 struct bpf_insn_access_aux *info)
2488{
2489 return false;
2490}
2491
2492static inline u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type,
2493 const struct bpf_insn *si,
2494 struct bpf_insn *insn_buf,
2495 struct bpf_prog *prog,
2496 u32 *target_size)
2497{
2498 return 0;
2499}
2500static inline bool bpf_xdp_sock_is_valid_access(int off, int size,
2501 enum bpf_access_type type,
2502 struct bpf_insn_access_aux *info)
2503{
2504 return false;
2505}
2506
2507static inline u32 bpf_xdp_sock_convert_ctx_access(enum bpf_access_type type,
2508 const struct bpf_insn *si,
2509 struct bpf_insn *insn_buf,
2510 struct bpf_prog *prog,
2511 u32 *target_size)
2512{
2513 return 0;
2514}
2515#endif /* CONFIG_INET */
2516
2517enum bpf_text_poke_type {
2518 BPF_MOD_CALL,
2519 BPF_MOD_JUMP,
2520};
2521
2522int bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
2523 void *addr1, void *addr2);
2524
2525void *bpf_arch_text_copy(void *dst, void *src, size_t len);
2526int bpf_arch_text_invalidate(void *dst, size_t len);
2527
2528struct btf_id_set;
2529bool btf_id_set_contains(const struct btf_id_set *set, u32 id);
2530
2531#define MAX_BPRINTF_VARARGS 12
2532
2533int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
2534 u32 **bin_buf, u32 num_args);
2535void bpf_bprintf_cleanup(void);
2536
2537/* the implementation of the opaque uapi struct bpf_dynptr */
2538struct bpf_dynptr_kern {
2539 void *data;
2540 /* Size represents the number of usable bytes of dynptr data.
2541 * If for example the offset is at 4 for a local dynptr whose data is
2542 * of type u64, the number of usable bytes is 4.
2543 *
2544 * The upper 8 bits are reserved. It is as follows:
2545 * Bits 0 - 23 = size
2546 * Bits 24 - 30 = dynptr type
2547 * Bit 31 = whether dynptr is read-only
2548 */
2549 u32 size;
2550 u32 offset;
2551} __aligned(8);
2552
2553enum bpf_dynptr_type {
2554 BPF_DYNPTR_TYPE_INVALID,
2555 /* Points to memory that is local to the bpf program */
2556 BPF_DYNPTR_TYPE_LOCAL,
2557 /* Underlying data is a ringbuf record */
2558 BPF_DYNPTR_TYPE_RINGBUF,
2559};
2560
2561void bpf_dynptr_init(struct bpf_dynptr_kern *ptr, void *data,
2562 enum bpf_dynptr_type type, u32 offset, u32 size);
2563void bpf_dynptr_set_null(struct bpf_dynptr_kern *ptr);
2564int bpf_dynptr_check_size(u32 size);
2565
2566#ifdef CONFIG_BPF_LSM
2567void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype);
2568void bpf_cgroup_atype_put(int cgroup_atype);
2569#else
2570static inline void bpf_cgroup_atype_get(u32 attach_btf_id, int cgroup_atype) {}
2571static inline void bpf_cgroup_atype_put(int cgroup_atype) {}
2572#endif /* CONFIG_BPF_LSM */
2573
2574#endif /* _LINUX_BPF_H */
2575

source code of linux/include/linux/bpf.h